diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,187639 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "global_step": 31269, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 2.129925452609159e-08, + "loss": 1.8506, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 4.259850905218318e-08, + "loss": 2.1582, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 6.389776357827476e-08, + "loss": 1.9258, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 8.519701810436636e-08, + "loss": 2.0449, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 1.0649627263045795e-07, + "loss": 2.002, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 1.2779552715654952e-07, + "loss": 1.8867, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 1.4909478168264114e-07, + "loss": 1.8145, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 1.7039403620873272e-07, + "loss": 2.0293, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 1.9169329073482428e-07, + "loss": 1.875, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 2.129925452609159e-07, + "loss": 2.1377, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 2.3429179978700748e-07, + "loss": 1.8057, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 2.5559105431309904e-07, + "loss": 2.0225, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 2.7689030883919063e-07, + "loss": 2.1406, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 2.9818956336528227e-07, + "loss": 2.167, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 3.194888178913738e-07, + "loss": 2.1221, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 3.4078807241746545e-07, + "loss": 2.1455, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 3.6208732694355703e-07, + "loss": 2.0371, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 3.8338658146964857e-07, + "loss": 2.1172, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 4.046858359957402e-07, + "loss": 1.9307, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 4.259850905218318e-07, + "loss": 1.7998, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 4.4728434504792333e-07, + "loss": 2.0293, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 4.6858359957401497e-07, + "loss": 1.8506, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 4.898828541001065e-07, + "loss": 1.8555, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 5.111821086261981e-07, + "loss": 2.0088, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 5.324813631522898e-07, + "loss": 2.0508, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 5.537806176783813e-07, + "loss": 1.7383, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 5.750798722044729e-07, + "loss": 1.8584, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 5.963791267305645e-07, + "loss": 1.6201, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 6.17678381256656e-07, + "loss": 1.7432, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 6.389776357827476e-07, + "loss": 1.5371, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 6.602768903088393e-07, + "loss": 1.5615, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 6.815761448349309e-07, + "loss": 1.4678, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 7.028753993610224e-07, + "loss": 1.5088, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 7.241746538871141e-07, + "loss": 0.3462, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 7.454739084132057e-07, + "loss": 1.5801, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 7.667731629392971e-07, + "loss": 1.4033, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 7.880724174653888e-07, + "loss": 1.6211, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 8.093716719914804e-07, + "loss": 1.3604, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 8.306709265175719e-07, + "loss": 1.5469, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 8.519701810436636e-07, + "loss": 1.4883, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 8.732694355697552e-07, + "loss": 1.5605, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 8.945686900958467e-07, + "loss": 1.4189, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 9.158679446219383e-07, + "loss": 1.5889, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 9.371671991480299e-07, + "loss": 1.416, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 9.584664536741215e-07, + "loss": 1.4531, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 9.79765708200213e-07, + "loss": 1.4268, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 1.0010649627263047e-06, + "loss": 1.3076, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 1.0223642172523962e-06, + "loss": 1.3672, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 1.0436634717784879e-06, + "loss": 1.5527, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 1.0649627263045796e-06, + "loss": 1.4253, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 1.086261980830671e-06, + "loss": 1.2837, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 1.1075612353567625e-06, + "loss": 1.3633, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 1.1288604898828542e-06, + "loss": 1.46, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 1.1501597444089457e-06, + "loss": 1.374, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 1.1714589989350374e-06, + "loss": 1.3555, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 1.192758253461129e-06, + "loss": 1.1641, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 1.2140575079872206e-06, + "loss": 1.3105, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 1.235356762513312e-06, + "loss": 1.2017, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 1.2566560170394037e-06, + "loss": 1.2778, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 1.2779552715654952e-06, + "loss": 1.292, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 1.2992545260915871e-06, + "loss": 1.2954, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 1.3205537806176786e-06, + "loss": 1.3369, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 1.34185303514377e-06, + "loss": 1.3164, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 1.3631522896698618e-06, + "loss": 1.3838, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 1.3844515441959533e-06, + "loss": 1.375, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 1.4057507987220447e-06, + "loss": 1.2607, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 1.4270500532481367e-06, + "loss": 1.2412, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 1.4483493077742281e-06, + "loss": 1.2637, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 1.4696485623003196e-06, + "loss": 1.2671, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 1.4909478168264113e-06, + "loss": 1.1387, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 1.5122470713525028e-06, + "loss": 1.3315, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 1.5335463258785943e-06, + "loss": 1.4385, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 1.5548455804046862e-06, + "loss": 1.2891, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 1.5761448349307777e-06, + "loss": 1.2822, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 1.5974440894568691e-06, + "loss": 1.2451, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 1.6187433439829608e-06, + "loss": 1.1919, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 1.6400425985090523e-06, + "loss": 1.2383, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 1.6613418530351438e-06, + "loss": 1.1431, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 1.6826411075612357e-06, + "loss": 1.2842, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 1.7039403620873272e-06, + "loss": 1.3359, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 1.7252396166134187e-06, + "loss": 1.313, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 1.7465388711395104e-06, + "loss": 1.2227, + "step": 82 + }, + { + "epoch": 0.0, + "learning_rate": 1.7678381256656018e-06, + "loss": 1.3389, + "step": 83 + }, + { + "epoch": 0.0, + "learning_rate": 1.7891373801916933e-06, + "loss": 1.2856, + "step": 84 + }, + { + "epoch": 0.0, + "learning_rate": 1.8104366347177852e-06, + "loss": 1.1699, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 1.8317358892438767e-06, + "loss": 1.2354, + "step": 86 + }, + { + "epoch": 0.0, + "learning_rate": 1.8530351437699682e-06, + "loss": 1.187, + "step": 87 + }, + { + "epoch": 0.0, + "learning_rate": 1.8743343982960599e-06, + "loss": 1.1973, + "step": 88 + }, + { + "epoch": 0.0, + "learning_rate": 1.8956336528221514e-06, + "loss": 1.1904, + "step": 89 + }, + { + "epoch": 0.0, + "learning_rate": 1.916932907348243e-06, + "loss": 1.2295, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 1.9382321618743345e-06, + "loss": 1.2598, + "step": 91 + }, + { + "epoch": 0.0, + "learning_rate": 1.959531416400426e-06, + "loss": 1.2676, + "step": 92 + }, + { + "epoch": 0.0, + "learning_rate": 1.9808306709265175e-06, + "loss": 1.2637, + "step": 93 + }, + { + "epoch": 0.0, + "learning_rate": 2.0021299254526094e-06, + "loss": 1.252, + "step": 94 + }, + { + "epoch": 0.0, + "learning_rate": 2.023429179978701e-06, + "loss": 1.1494, + "step": 95 + }, + { + "epoch": 0.0, + "learning_rate": 2.0447284345047924e-06, + "loss": 1.3042, + "step": 96 + }, + { + "epoch": 0.0, + "learning_rate": 2.0660276890308843e-06, + "loss": 1.1626, + "step": 97 + }, + { + "epoch": 0.0, + "learning_rate": 2.0873269435569757e-06, + "loss": 1.1538, + "step": 98 + }, + { + "epoch": 0.0, + "learning_rate": 2.1086261980830672e-06, + "loss": 1.2178, + "step": 99 + }, + { + "epoch": 0.0, + "learning_rate": 2.129925452609159e-06, + "loss": 1.2183, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 2.1512247071352506e-06, + "loss": 1.2158, + "step": 101 + }, + { + "epoch": 0.0, + "learning_rate": 2.172523961661342e-06, + "loss": 1.2783, + "step": 102 + }, + { + "epoch": 0.0, + "learning_rate": 2.1938232161874336e-06, + "loss": 1.2461, + "step": 103 + }, + { + "epoch": 0.0, + "learning_rate": 2.215122470713525e-06, + "loss": 1.2329, + "step": 104 + }, + { + "epoch": 0.0, + "learning_rate": 2.2364217252396165e-06, + "loss": 1.2544, + "step": 105 + }, + { + "epoch": 0.0, + "learning_rate": 2.2577209797657084e-06, + "loss": 1.2998, + "step": 106 + }, + { + "epoch": 0.0, + "learning_rate": 2.2790202342918e-06, + "loss": 1.2363, + "step": 107 + }, + { + "epoch": 0.0, + "learning_rate": 2.3003194888178914e-06, + "loss": 1.1348, + "step": 108 + }, + { + "epoch": 0.0, + "learning_rate": 2.3216187433439833e-06, + "loss": 1.0625, + "step": 109 + }, + { + "epoch": 0.0, + "learning_rate": 2.342917997870075e-06, + "loss": 1.3154, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 2.3642172523961663e-06, + "loss": 1.2383, + "step": 111 + }, + { + "epoch": 0.0, + "learning_rate": 2.385516506922258e-06, + "loss": 1.228, + "step": 112 + }, + { + "epoch": 0.0, + "learning_rate": 2.4068157614483497e-06, + "loss": 1.0532, + "step": 113 + }, + { + "epoch": 0.0, + "learning_rate": 2.428115015974441e-06, + "loss": 1.1602, + "step": 114 + }, + { + "epoch": 0.0, + "learning_rate": 2.4494142705005326e-06, + "loss": 1.2217, + "step": 115 + }, + { + "epoch": 0.0, + "learning_rate": 2.470713525026624e-06, + "loss": 0.3672, + "step": 116 + }, + { + "epoch": 0.0, + "learning_rate": 2.4920127795527156e-06, + "loss": 1.1709, + "step": 117 + }, + { + "epoch": 0.0, + "learning_rate": 2.5133120340788075e-06, + "loss": 0.3816, + "step": 118 + }, + { + "epoch": 0.0, + "learning_rate": 2.534611288604899e-06, + "loss": 1.2549, + "step": 119 + }, + { + "epoch": 0.0, + "learning_rate": 2.5559105431309904e-06, + "loss": 1.269, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 2.577209797657082e-06, + "loss": 1.3438, + "step": 121 + }, + { + "epoch": 0.0, + "learning_rate": 2.5985090521831743e-06, + "loss": 1.1426, + "step": 122 + }, + { + "epoch": 0.0, + "learning_rate": 2.6198083067092657e-06, + "loss": 1.1885, + "step": 123 + }, + { + "epoch": 0.0, + "learning_rate": 2.6411075612353572e-06, + "loss": 1.1499, + "step": 124 + }, + { + "epoch": 0.0, + "learning_rate": 2.6624068157614487e-06, + "loss": 1.043, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 2.68370607028754e-06, + "loss": 1.2451, + "step": 126 + }, + { + "epoch": 0.0, + "learning_rate": 2.7050053248136317e-06, + "loss": 1.2578, + "step": 127 + }, + { + "epoch": 0.0, + "learning_rate": 2.7263045793397236e-06, + "loss": 1.2803, + "step": 128 + }, + { + "epoch": 0.0, + "learning_rate": 2.747603833865815e-06, + "loss": 1.2461, + "step": 129 + }, + { + "epoch": 0.0, + "learning_rate": 2.7689030883919065e-06, + "loss": 1.2842, + "step": 130 + }, + { + "epoch": 0.0, + "learning_rate": 2.790202342917998e-06, + "loss": 1.2622, + "step": 131 + }, + { + "epoch": 0.0, + "learning_rate": 2.8115015974440895e-06, + "loss": 1.145, + "step": 132 + }, + { + "epoch": 0.0, + "learning_rate": 2.832800851970181e-06, + "loss": 1.229, + "step": 133 + }, + { + "epoch": 0.0, + "learning_rate": 2.8541001064962733e-06, + "loss": 1.2866, + "step": 134 + }, + { + "epoch": 0.0, + "learning_rate": 2.8753993610223648e-06, + "loss": 1.1948, + "step": 135 + }, + { + "epoch": 0.0, + "learning_rate": 2.8966986155484563e-06, + "loss": 1.2021, + "step": 136 + }, + { + "epoch": 0.0, + "learning_rate": 2.9179978700745477e-06, + "loss": 1.2715, + "step": 137 + }, + { + "epoch": 0.0, + "learning_rate": 2.9392971246006392e-06, + "loss": 1.2104, + "step": 138 + }, + { + "epoch": 0.0, + "learning_rate": 2.9605963791267307e-06, + "loss": 1.2339, + "step": 139 + }, + { + "epoch": 0.0, + "learning_rate": 2.9818956336528226e-06, + "loss": 1.1562, + "step": 140 + }, + { + "epoch": 0.0, + "learning_rate": 3.003194888178914e-06, + "loss": 1.2031, + "step": 141 + }, + { + "epoch": 0.0, + "learning_rate": 3.0244941427050056e-06, + "loss": 1.3145, + "step": 142 + }, + { + "epoch": 0.0, + "learning_rate": 3.045793397231097e-06, + "loss": 1.2939, + "step": 143 + }, + { + "epoch": 0.0, + "learning_rate": 3.0670926517571885e-06, + "loss": 1.1719, + "step": 144 + }, + { + "epoch": 0.0, + "learning_rate": 3.08839190628328e-06, + "loss": 1.1641, + "step": 145 + }, + { + "epoch": 0.0, + "learning_rate": 3.1096911608093723e-06, + "loss": 1.3574, + "step": 146 + }, + { + "epoch": 0.0, + "learning_rate": 3.130990415335464e-06, + "loss": 1.1235, + "step": 147 + }, + { + "epoch": 0.0, + "learning_rate": 3.1522896698615553e-06, + "loss": 1.1519, + "step": 148 + }, + { + "epoch": 0.0, + "learning_rate": 3.173588924387647e-06, + "loss": 1.1538, + "step": 149 + }, + { + "epoch": 0.0, + "learning_rate": 3.1948881789137383e-06, + "loss": 1.103, + "step": 150 + }, + { + "epoch": 0.0, + "learning_rate": 3.2161874334398298e-06, + "loss": 1.1484, + "step": 151 + }, + { + "epoch": 0.0, + "learning_rate": 3.2374866879659217e-06, + "loss": 1.1978, + "step": 152 + }, + { + "epoch": 0.0, + "learning_rate": 3.258785942492013e-06, + "loss": 1.0122, + "step": 153 + }, + { + "epoch": 0.0, + "learning_rate": 3.2800851970181046e-06, + "loss": 1.2598, + "step": 154 + }, + { + "epoch": 0.0, + "learning_rate": 3.301384451544196e-06, + "loss": 1.1597, + "step": 155 + }, + { + "epoch": 0.0, + "learning_rate": 3.3226837060702876e-06, + "loss": 1.1099, + "step": 156 + }, + { + "epoch": 0.01, + "learning_rate": 3.343982960596379e-06, + "loss": 1.189, + "step": 157 + }, + { + "epoch": 0.01, + "learning_rate": 3.3652822151224714e-06, + "loss": 1.1899, + "step": 158 + }, + { + "epoch": 0.01, + "learning_rate": 3.386581469648563e-06, + "loss": 1.1357, + "step": 159 + }, + { + "epoch": 0.01, + "learning_rate": 3.4078807241746544e-06, + "loss": 1.2461, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 3.429179978700746e-06, + "loss": 1.1138, + "step": 161 + }, + { + "epoch": 0.01, + "learning_rate": 3.4504792332268373e-06, + "loss": 1.2495, + "step": 162 + }, + { + "epoch": 0.01, + "learning_rate": 3.471778487752929e-06, + "loss": 1.1035, + "step": 163 + }, + { + "epoch": 0.01, + "learning_rate": 3.4930777422790207e-06, + "loss": 1.1367, + "step": 164 + }, + { + "epoch": 0.01, + "learning_rate": 3.514376996805112e-06, + "loss": 1.1973, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 3.5356762513312037e-06, + "loss": 1.1323, + "step": 166 + }, + { + "epoch": 0.01, + "learning_rate": 3.556975505857295e-06, + "loss": 1.084, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 3.5782747603833866e-06, + "loss": 1.1328, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 3.599574014909478e-06, + "loss": 1.4463, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 3.6208732694355704e-06, + "loss": 1.1606, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 3.642172523961662e-06, + "loss": 1.2979, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 3.6634717784877534e-06, + "loss": 1.1875, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 3.684771033013845e-06, + "loss": 1.1655, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 3.7060702875399364e-06, + "loss": 1.293, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 3.727369542066028e-06, + "loss": 1.1865, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 3.7486687965921197e-06, + "loss": 1.2251, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 3.7699680511182112e-06, + "loss": 1.2412, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 3.7912673056443027e-06, + "loss": 1.1309, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 3.812566560170394e-06, + "loss": 1.0547, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 3.833865814696486e-06, + "loss": 0.408, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 3.855165069222578e-06, + "loss": 1.228, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 3.876464323748669e-06, + "loss": 0.4412, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 3.8977635782747605e-06, + "loss": 0.3838, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 3.919062832800852e-06, + "loss": 1.0981, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 3.9403620873269435e-06, + "loss": 1.1572, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 3.961661341853035e-06, + "loss": 1.2041, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 3.9829605963791265e-06, + "loss": 1.3218, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 4.004259850905219e-06, + "loss": 1.2002, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 4.02555910543131e-06, + "loss": 1.1816, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 4.046858359957402e-06, + "loss": 1.1802, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 4.068157614483493e-06, + "loss": 1.3027, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 4.089456869009585e-06, + "loss": 1.1265, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 4.110756123535676e-06, + "loss": 1.1309, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 4.1320553780617685e-06, + "loss": 1.1641, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 4.15335463258786e-06, + "loss": 1.2017, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 4.1746538871139515e-06, + "loss": 1.0991, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 4.195953141640043e-06, + "loss": 1.1597, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 4.2172523961661345e-06, + "loss": 1.1973, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 4.238551650692226e-06, + "loss": 1.104, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 4.259850905218318e-06, + "loss": 1.2549, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 4.28115015974441e-06, + "loss": 1.1821, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 4.302449414270501e-06, + "loss": 1.1655, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 4.323748668796593e-06, + "loss": 1.2056, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 4.345047923322684e-06, + "loss": 1.1343, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 4.366347177848776e-06, + "loss": 1.1436, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 4.387646432374867e-06, + "loss": 1.1997, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 4.408945686900959e-06, + "loss": 1.2168, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 4.43024494142705e-06, + "loss": 1.0649, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 4.451544195953142e-06, + "loss": 1.1802, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 4.472843450479233e-06, + "loss": 1.0908, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 4.494142705005325e-06, + "loss": 1.125, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 4.515441959531417e-06, + "loss": 1.0225, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 4.536741214057508e-06, + "loss": 1.2266, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 4.5580404685836e-06, + "loss": 1.248, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 4.579339723109691e-06, + "loss": 1.105, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 4.600638977635783e-06, + "loss": 1.0, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 4.621938232161875e-06, + "loss": 1.0986, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 4.643237486687967e-06, + "loss": 1.1372, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 4.664536741214058e-06, + "loss": 1.0713, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 4.68583599574015e-06, + "loss": 1.1284, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 4.707135250266241e-06, + "loss": 1.2773, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 4.7284345047923325e-06, + "loss": 1.2871, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 4.749733759318425e-06, + "loss": 1.0811, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 4.771033013844516e-06, + "loss": 0.306, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 4.792332268370608e-06, + "loss": 1.103, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 4.813631522896699e-06, + "loss": 1.1421, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 4.834930777422791e-06, + "loss": 1.0913, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 4.856230031948882e-06, + "loss": 1.313, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 4.877529286474974e-06, + "loss": 1.1499, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 4.898828541001065e-06, + "loss": 1.2305, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 4.920127795527157e-06, + "loss": 1.0762, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 4.941427050053248e-06, + "loss": 1.1206, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 4.96272630457934e-06, + "loss": 1.127, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 4.984025559105431e-06, + "loss": 1.0693, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 5.0053248136315235e-06, + "loss": 1.1997, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 5.026624068157615e-06, + "loss": 1.1245, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 5.0479233226837065e-06, + "loss": 1.1338, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 5.069222577209798e-06, + "loss": 1.167, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 5.090521831735889e-06, + "loss": 1.2041, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 5.111821086261981e-06, + "loss": 0.3452, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 5.133120340788072e-06, + "loss": 1.126, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 5.154419595314164e-06, + "loss": 1.0854, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 5.175718849840255e-06, + "loss": 1.168, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 5.1970181043663485e-06, + "loss": 1.2881, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 5.21831735889244e-06, + "loss": 1.1245, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 5.2396166134185315e-06, + "loss": 1.1514, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 5.260915867944623e-06, + "loss": 1.1396, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 5.2822151224707144e-06, + "loss": 1.064, + "step": 248 + }, + { + "epoch": 0.01, + "learning_rate": 5.303514376996806e-06, + "loss": 1.1191, + "step": 249 + }, + { + "epoch": 0.01, + "learning_rate": 5.324813631522897e-06, + "loss": 1.1826, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 5.346112886048989e-06, + "loss": 1.1733, + "step": 251 + }, + { + "epoch": 0.01, + "learning_rate": 5.36741214057508e-06, + "loss": 1.0869, + "step": 252 + }, + { + "epoch": 0.01, + "learning_rate": 5.388711395101172e-06, + "loss": 1.1074, + "step": 253 + }, + { + "epoch": 0.01, + "learning_rate": 5.410010649627263e-06, + "loss": 1.2012, + "step": 254 + }, + { + "epoch": 0.01, + "learning_rate": 5.431309904153355e-06, + "loss": 1.1362, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 5.452609158679447e-06, + "loss": 1.1704, + "step": 256 + }, + { + "epoch": 0.01, + "learning_rate": 5.473908413205539e-06, + "loss": 1.104, + "step": 257 + }, + { + "epoch": 0.01, + "learning_rate": 5.49520766773163e-06, + "loss": 1.1914, + "step": 258 + }, + { + "epoch": 0.01, + "learning_rate": 5.516506922257722e-06, + "loss": 1.1426, + "step": 259 + }, + { + "epoch": 0.01, + "learning_rate": 5.537806176783813e-06, + "loss": 1.0718, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 5.5591054313099045e-06, + "loss": 1.1548, + "step": 261 + }, + { + "epoch": 0.01, + "learning_rate": 5.580404685835996e-06, + "loss": 1.1494, + "step": 262 + }, + { + "epoch": 0.01, + "learning_rate": 5.6017039403620875e-06, + "loss": 1.1992, + "step": 263 + }, + { + "epoch": 0.01, + "learning_rate": 5.623003194888179e-06, + "loss": 1.1597, + "step": 264 + }, + { + "epoch": 0.01, + "learning_rate": 5.6443024494142705e-06, + "loss": 1.2285, + "step": 265 + }, + { + "epoch": 0.01, + "learning_rate": 5.665601703940362e-06, + "loss": 1.0786, + "step": 266 + }, + { + "epoch": 0.01, + "learning_rate": 5.6869009584664534e-06, + "loss": 1.1255, + "step": 267 + }, + { + "epoch": 0.01, + "learning_rate": 5.708200212992547e-06, + "loss": 1.0444, + "step": 268 + }, + { + "epoch": 0.01, + "learning_rate": 5.729499467518638e-06, + "loss": 1.1875, + "step": 269 + }, + { + "epoch": 0.01, + "learning_rate": 5.7507987220447296e-06, + "loss": 1.1577, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 5.772097976570821e-06, + "loss": 1.084, + "step": 271 + }, + { + "epoch": 0.01, + "learning_rate": 5.7933972310969125e-06, + "loss": 1.0063, + "step": 272 + }, + { + "epoch": 0.01, + "learning_rate": 5.814696485623004e-06, + "loss": 1.1655, + "step": 273 + }, + { + "epoch": 0.01, + "learning_rate": 5.8359957401490955e-06, + "loss": 0.3628, + "step": 274 + }, + { + "epoch": 0.01, + "learning_rate": 5.857294994675187e-06, + "loss": 1.1118, + "step": 275 + }, + { + "epoch": 0.01, + "learning_rate": 5.8785942492012785e-06, + "loss": 1.1016, + "step": 276 + }, + { + "epoch": 0.01, + "learning_rate": 5.89989350372737e-06, + "loss": 1.1245, + "step": 277 + }, + { + "epoch": 0.01, + "learning_rate": 5.921192758253461e-06, + "loss": 1.2832, + "step": 278 + }, + { + "epoch": 0.01, + "learning_rate": 5.942492012779553e-06, + "loss": 1.0562, + "step": 279 + }, + { + "epoch": 0.01, + "learning_rate": 5.963791267305645e-06, + "loss": 1.1558, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 5.985090521831737e-06, + "loss": 1.0884, + "step": 281 + }, + { + "epoch": 0.01, + "learning_rate": 6.006389776357828e-06, + "loss": 1.2471, + "step": 282 + }, + { + "epoch": 0.01, + "learning_rate": 6.02768903088392e-06, + "loss": 1.252, + "step": 283 + }, + { + "epoch": 0.01, + "learning_rate": 6.048988285410011e-06, + "loss": 1.1348, + "step": 284 + }, + { + "epoch": 0.01, + "learning_rate": 6.070287539936103e-06, + "loss": 1.2388, + "step": 285 + }, + { + "epoch": 0.01, + "learning_rate": 6.091586794462194e-06, + "loss": 1.188, + "step": 286 + }, + { + "epoch": 0.01, + "learning_rate": 6.112886048988286e-06, + "loss": 1.2183, + "step": 287 + }, + { + "epoch": 0.01, + "learning_rate": 6.134185303514377e-06, + "loss": 1.0806, + "step": 288 + }, + { + "epoch": 0.01, + "learning_rate": 6.1554845580404686e-06, + "loss": 1.3164, + "step": 289 + }, + { + "epoch": 0.01, + "learning_rate": 6.17678381256656e-06, + "loss": 1.0991, + "step": 290 + }, + { + "epoch": 0.01, + "learning_rate": 6.1980830670926515e-06, + "loss": 1.1196, + "step": 291 + }, + { + "epoch": 0.01, + "learning_rate": 6.219382321618745e-06, + "loss": 1.0947, + "step": 292 + }, + { + "epoch": 0.01, + "learning_rate": 6.240681576144836e-06, + "loss": 1.0771, + "step": 293 + }, + { + "epoch": 0.01, + "learning_rate": 6.261980830670928e-06, + "loss": 1.2046, + "step": 294 + }, + { + "epoch": 0.01, + "learning_rate": 6.283280085197019e-06, + "loss": 1.1138, + "step": 295 + }, + { + "epoch": 0.01, + "learning_rate": 6.304579339723111e-06, + "loss": 1.1392, + "step": 296 + }, + { + "epoch": 0.01, + "learning_rate": 6.325878594249202e-06, + "loss": 0.3701, + "step": 297 + }, + { + "epoch": 0.01, + "learning_rate": 6.347177848775294e-06, + "loss": 1.1948, + "step": 298 + }, + { + "epoch": 0.01, + "learning_rate": 6.368477103301385e-06, + "loss": 1.1919, + "step": 299 + }, + { + "epoch": 0.01, + "learning_rate": 6.3897763578274765e-06, + "loss": 1.1602, + "step": 300 + }, + { + "epoch": 0.01, + "learning_rate": 6.411075612353568e-06, + "loss": 1.2236, + "step": 301 + }, + { + "epoch": 0.01, + "learning_rate": 6.4323748668796595e-06, + "loss": 1.2383, + "step": 302 + }, + { + "epoch": 0.01, + "learning_rate": 6.453674121405751e-06, + "loss": 1.2466, + "step": 303 + }, + { + "epoch": 0.01, + "learning_rate": 6.474973375931843e-06, + "loss": 1.168, + "step": 304 + }, + { + "epoch": 0.01, + "learning_rate": 6.496272630457935e-06, + "loss": 1.144, + "step": 305 + }, + { + "epoch": 0.01, + "learning_rate": 6.517571884984026e-06, + "loss": 1.2002, + "step": 306 + }, + { + "epoch": 0.01, + "learning_rate": 6.538871139510118e-06, + "loss": 1.0981, + "step": 307 + }, + { + "epoch": 0.01, + "learning_rate": 6.560170394036209e-06, + "loss": 1.2002, + "step": 308 + }, + { + "epoch": 0.01, + "learning_rate": 6.581469648562301e-06, + "loss": 1.1094, + "step": 309 + }, + { + "epoch": 0.01, + "learning_rate": 6.602768903088392e-06, + "loss": 1.1182, + "step": 310 + }, + { + "epoch": 0.01, + "learning_rate": 6.624068157614484e-06, + "loss": 1.106, + "step": 311 + }, + { + "epoch": 0.01, + "learning_rate": 6.645367412140575e-06, + "loss": 1.248, + "step": 312 + }, + { + "epoch": 0.01, + "learning_rate": 6.666666666666667e-06, + "loss": 1.1982, + "step": 313 + }, + { + "epoch": 0.01, + "learning_rate": 6.687965921192758e-06, + "loss": 1.2476, + "step": 314 + }, + { + "epoch": 0.01, + "learning_rate": 6.709265175718851e-06, + "loss": 1.1802, + "step": 315 + }, + { + "epoch": 0.01, + "learning_rate": 6.730564430244943e-06, + "loss": 1.063, + "step": 316 + }, + { + "epoch": 0.01, + "learning_rate": 6.751863684771034e-06, + "loss": 1.2773, + "step": 317 + }, + { + "epoch": 0.01, + "learning_rate": 6.773162939297126e-06, + "loss": 1.1221, + "step": 318 + }, + { + "epoch": 0.01, + "learning_rate": 6.794462193823217e-06, + "loss": 1.2275, + "step": 319 + }, + { + "epoch": 0.01, + "learning_rate": 6.815761448349309e-06, + "loss": 1.1758, + "step": 320 + }, + { + "epoch": 0.01, + "learning_rate": 6.8370607028754e-06, + "loss": 1.1558, + "step": 321 + }, + { + "epoch": 0.01, + "learning_rate": 6.858359957401492e-06, + "loss": 1.1235, + "step": 322 + }, + { + "epoch": 0.01, + "learning_rate": 6.879659211927583e-06, + "loss": 1.2212, + "step": 323 + }, + { + "epoch": 0.01, + "learning_rate": 6.900958466453675e-06, + "loss": 1.1064, + "step": 324 + }, + { + "epoch": 0.01, + "learning_rate": 6.922257720979766e-06, + "loss": 1.0576, + "step": 325 + }, + { + "epoch": 0.01, + "learning_rate": 6.943556975505858e-06, + "loss": 1.0239, + "step": 326 + }, + { + "epoch": 0.01, + "learning_rate": 6.96485623003195e-06, + "loss": 1.1548, + "step": 327 + }, + { + "epoch": 0.01, + "learning_rate": 6.986155484558041e-06, + "loss": 1.1465, + "step": 328 + }, + { + "epoch": 0.01, + "learning_rate": 7.007454739084133e-06, + "loss": 1.1138, + "step": 329 + }, + { + "epoch": 0.01, + "learning_rate": 7.028753993610224e-06, + "loss": 1.1035, + "step": 330 + }, + { + "epoch": 0.01, + "learning_rate": 7.050053248136316e-06, + "loss": 1.0957, + "step": 331 + }, + { + "epoch": 0.01, + "learning_rate": 7.071352502662407e-06, + "loss": 1.1772, + "step": 332 + }, + { + "epoch": 0.01, + "learning_rate": 7.092651757188499e-06, + "loss": 1.1655, + "step": 333 + }, + { + "epoch": 0.01, + "learning_rate": 7.11395101171459e-06, + "loss": 1.0825, + "step": 334 + }, + { + "epoch": 0.01, + "learning_rate": 7.135250266240682e-06, + "loss": 1.1499, + "step": 335 + }, + { + "epoch": 0.01, + "learning_rate": 7.156549520766773e-06, + "loss": 1.1343, + "step": 336 + }, + { + "epoch": 0.01, + "learning_rate": 7.177848775292865e-06, + "loss": 0.3757, + "step": 337 + }, + { + "epoch": 0.01, + "learning_rate": 7.199148029818956e-06, + "loss": 1.0669, + "step": 338 + }, + { + "epoch": 0.01, + "learning_rate": 7.220447284345049e-06, + "loss": 1.1704, + "step": 339 + }, + { + "epoch": 0.01, + "learning_rate": 7.241746538871141e-06, + "loss": 1.1221, + "step": 340 + }, + { + "epoch": 0.01, + "learning_rate": 7.263045793397232e-06, + "loss": 1.2109, + "step": 341 + }, + { + "epoch": 0.01, + "learning_rate": 7.284345047923324e-06, + "loss": 1.0654, + "step": 342 + }, + { + "epoch": 0.01, + "learning_rate": 7.305644302449415e-06, + "loss": 1.0635, + "step": 343 + }, + { + "epoch": 0.01, + "learning_rate": 7.326943556975507e-06, + "loss": 1.2349, + "step": 344 + }, + { + "epoch": 0.01, + "learning_rate": 7.348242811501598e-06, + "loss": 1.0776, + "step": 345 + }, + { + "epoch": 0.01, + "learning_rate": 7.36954206602769e-06, + "loss": 1.1396, + "step": 346 + }, + { + "epoch": 0.01, + "learning_rate": 7.390841320553781e-06, + "loss": 1.1152, + "step": 347 + }, + { + "epoch": 0.01, + "learning_rate": 7.412140575079873e-06, + "loss": 1.2002, + "step": 348 + }, + { + "epoch": 0.01, + "learning_rate": 7.433439829605964e-06, + "loss": 1.0986, + "step": 349 + }, + { + "epoch": 0.01, + "learning_rate": 7.454739084132056e-06, + "loss": 1.1401, + "step": 350 + }, + { + "epoch": 0.01, + "learning_rate": 7.476038338658148e-06, + "loss": 1.1421, + "step": 351 + }, + { + "epoch": 0.01, + "learning_rate": 7.4973375931842395e-06, + "loss": 1.1118, + "step": 352 + }, + { + "epoch": 0.01, + "learning_rate": 7.518636847710331e-06, + "loss": 1.1201, + "step": 353 + }, + { + "epoch": 0.01, + "learning_rate": 7.5399361022364225e-06, + "loss": 1.0371, + "step": 354 + }, + { + "epoch": 0.01, + "learning_rate": 7.561235356762514e-06, + "loss": 1.1475, + "step": 355 + }, + { + "epoch": 0.01, + "learning_rate": 7.582534611288605e-06, + "loss": 1.0771, + "step": 356 + }, + { + "epoch": 0.01, + "learning_rate": 7.603833865814697e-06, + "loss": 1.1089, + "step": 357 + }, + { + "epoch": 0.01, + "learning_rate": 7.625133120340788e-06, + "loss": 1.1489, + "step": 358 + }, + { + "epoch": 0.01, + "learning_rate": 7.64643237486688e-06, + "loss": 1.1494, + "step": 359 + }, + { + "epoch": 0.01, + "learning_rate": 7.667731629392972e-06, + "loss": 1.105, + "step": 360 + }, + { + "epoch": 0.01, + "learning_rate": 7.689030883919063e-06, + "loss": 1.1187, + "step": 361 + }, + { + "epoch": 0.01, + "learning_rate": 7.710330138445155e-06, + "loss": 1.0879, + "step": 362 + }, + { + "epoch": 0.01, + "learning_rate": 7.731629392971247e-06, + "loss": 1.2056, + "step": 363 + }, + { + "epoch": 0.01, + "learning_rate": 7.752928647497338e-06, + "loss": 1.0713, + "step": 364 + }, + { + "epoch": 0.01, + "learning_rate": 7.77422790202343e-06, + "loss": 1.1821, + "step": 365 + }, + { + "epoch": 0.01, + "learning_rate": 7.795527156549521e-06, + "loss": 1.2339, + "step": 366 + }, + { + "epoch": 0.01, + "learning_rate": 7.816826411075613e-06, + "loss": 1.1909, + "step": 367 + }, + { + "epoch": 0.01, + "learning_rate": 7.838125665601704e-06, + "loss": 1.2773, + "step": 368 + }, + { + "epoch": 0.01, + "learning_rate": 7.859424920127796e-06, + "loss": 1.061, + "step": 369 + }, + { + "epoch": 0.01, + "learning_rate": 7.880724174653887e-06, + "loss": 1.0566, + "step": 370 + }, + { + "epoch": 0.01, + "learning_rate": 7.90202342917998e-06, + "loss": 1.124, + "step": 371 + }, + { + "epoch": 0.01, + "learning_rate": 7.92332268370607e-06, + "loss": 1.2969, + "step": 372 + }, + { + "epoch": 0.01, + "learning_rate": 7.944621938232162e-06, + "loss": 1.0913, + "step": 373 + }, + { + "epoch": 0.01, + "learning_rate": 7.965921192758253e-06, + "loss": 1.1577, + "step": 374 + }, + { + "epoch": 0.01, + "learning_rate": 7.987220447284347e-06, + "loss": 1.1914, + "step": 375 + }, + { + "epoch": 0.01, + "learning_rate": 8.008519701810438e-06, + "loss": 1.2217, + "step": 376 + }, + { + "epoch": 0.01, + "learning_rate": 8.02981895633653e-06, + "loss": 1.1309, + "step": 377 + }, + { + "epoch": 0.01, + "learning_rate": 8.05111821086262e-06, + "loss": 1.0923, + "step": 378 + }, + { + "epoch": 0.01, + "learning_rate": 8.072417465388713e-06, + "loss": 1.1035, + "step": 379 + }, + { + "epoch": 0.01, + "learning_rate": 8.093716719914804e-06, + "loss": 1.1157, + "step": 380 + }, + { + "epoch": 0.01, + "learning_rate": 8.115015974440896e-06, + "loss": 1.1055, + "step": 381 + }, + { + "epoch": 0.01, + "learning_rate": 8.136315228966986e-06, + "loss": 1.2485, + "step": 382 + }, + { + "epoch": 0.01, + "learning_rate": 8.157614483493079e-06, + "loss": 1.1704, + "step": 383 + }, + { + "epoch": 0.01, + "learning_rate": 8.17891373801917e-06, + "loss": 1.0605, + "step": 384 + }, + { + "epoch": 0.01, + "learning_rate": 8.200212992545262e-06, + "loss": 1.1157, + "step": 385 + }, + { + "epoch": 0.01, + "learning_rate": 8.221512247071352e-06, + "loss": 1.1235, + "step": 386 + }, + { + "epoch": 0.01, + "learning_rate": 8.242811501597445e-06, + "loss": 1.1597, + "step": 387 + }, + { + "epoch": 0.01, + "learning_rate": 8.264110756123537e-06, + "loss": 1.1953, + "step": 388 + }, + { + "epoch": 0.01, + "learning_rate": 8.285410010649628e-06, + "loss": 1.0571, + "step": 389 + }, + { + "epoch": 0.01, + "learning_rate": 8.30670926517572e-06, + "loss": 1.1084, + "step": 390 + }, + { + "epoch": 0.01, + "learning_rate": 8.32800851970181e-06, + "loss": 1.1289, + "step": 391 + }, + { + "epoch": 0.01, + "learning_rate": 8.349307774227903e-06, + "loss": 1.2109, + "step": 392 + }, + { + "epoch": 0.01, + "learning_rate": 8.370607028753994e-06, + "loss": 1.042, + "step": 393 + }, + { + "epoch": 0.01, + "learning_rate": 8.391906283280086e-06, + "loss": 1.1704, + "step": 394 + }, + { + "epoch": 0.01, + "learning_rate": 8.413205537806177e-06, + "loss": 1.1201, + "step": 395 + }, + { + "epoch": 0.01, + "learning_rate": 8.434504792332269e-06, + "loss": 1.1226, + "step": 396 + }, + { + "epoch": 0.01, + "learning_rate": 8.45580404685836e-06, + "loss": 1.1216, + "step": 397 + }, + { + "epoch": 0.01, + "learning_rate": 8.477103301384452e-06, + "loss": 1.1328, + "step": 398 + }, + { + "epoch": 0.01, + "learning_rate": 8.498402555910544e-06, + "loss": 1.0884, + "step": 399 + }, + { + "epoch": 0.01, + "learning_rate": 8.519701810436637e-06, + "loss": 1.2207, + "step": 400 + }, + { + "epoch": 0.01, + "learning_rate": 8.541001064962727e-06, + "loss": 1.1055, + "step": 401 + }, + { + "epoch": 0.01, + "learning_rate": 8.56230031948882e-06, + "loss": 1.2378, + "step": 402 + }, + { + "epoch": 0.01, + "learning_rate": 8.58359957401491e-06, + "loss": 1.2207, + "step": 403 + }, + { + "epoch": 0.01, + "learning_rate": 8.604898828541002e-06, + "loss": 1.0356, + "step": 404 + }, + { + "epoch": 0.01, + "learning_rate": 8.626198083067093e-06, + "loss": 1.1636, + "step": 405 + }, + { + "epoch": 0.01, + "learning_rate": 8.647497337593185e-06, + "loss": 1.1577, + "step": 406 + }, + { + "epoch": 0.01, + "learning_rate": 8.668796592119276e-06, + "loss": 1.1592, + "step": 407 + }, + { + "epoch": 0.01, + "learning_rate": 8.690095846645368e-06, + "loss": 1.1392, + "step": 408 + }, + { + "epoch": 0.01, + "learning_rate": 8.711395101171459e-06, + "loss": 1.0449, + "step": 409 + }, + { + "epoch": 0.01, + "learning_rate": 8.732694355697551e-06, + "loss": 1.0962, + "step": 410 + }, + { + "epoch": 0.01, + "learning_rate": 8.753993610223644e-06, + "loss": 1.3008, + "step": 411 + }, + { + "epoch": 0.01, + "learning_rate": 8.775292864749734e-06, + "loss": 1.0015, + "step": 412 + }, + { + "epoch": 0.01, + "learning_rate": 8.796592119275827e-06, + "loss": 1.1484, + "step": 413 + }, + { + "epoch": 0.01, + "learning_rate": 8.817891373801917e-06, + "loss": 1.1069, + "step": 414 + }, + { + "epoch": 0.01, + "learning_rate": 8.83919062832801e-06, + "loss": 1.1294, + "step": 415 + }, + { + "epoch": 0.01, + "learning_rate": 8.8604898828541e-06, + "loss": 1.0767, + "step": 416 + }, + { + "epoch": 0.01, + "learning_rate": 8.881789137380193e-06, + "loss": 1.1025, + "step": 417 + }, + { + "epoch": 0.01, + "learning_rate": 8.903088391906283e-06, + "loss": 1.1997, + "step": 418 + }, + { + "epoch": 0.01, + "learning_rate": 8.924387646432376e-06, + "loss": 1.1802, + "step": 419 + }, + { + "epoch": 0.01, + "learning_rate": 8.945686900958466e-06, + "loss": 1.1885, + "step": 420 + }, + { + "epoch": 0.01, + "learning_rate": 8.966986155484558e-06, + "loss": 1.1577, + "step": 421 + }, + { + "epoch": 0.01, + "learning_rate": 8.98828541001065e-06, + "loss": 1.0391, + "step": 422 + }, + { + "epoch": 0.01, + "learning_rate": 9.009584664536743e-06, + "loss": 1.1919, + "step": 423 + }, + { + "epoch": 0.01, + "learning_rate": 9.030883919062834e-06, + "loss": 1.0811, + "step": 424 + }, + { + "epoch": 0.01, + "learning_rate": 9.052183173588926e-06, + "loss": 1.1069, + "step": 425 + }, + { + "epoch": 0.01, + "learning_rate": 9.073482428115017e-06, + "loss": 1.1636, + "step": 426 + }, + { + "epoch": 0.01, + "learning_rate": 9.094781682641109e-06, + "loss": 1.1099, + "step": 427 + }, + { + "epoch": 0.01, + "learning_rate": 9.1160809371672e-06, + "loss": 1.0723, + "step": 428 + }, + { + "epoch": 0.01, + "learning_rate": 9.137380191693292e-06, + "loss": 1.0166, + "step": 429 + }, + { + "epoch": 0.01, + "learning_rate": 9.158679446219383e-06, + "loss": 1.2012, + "step": 430 + }, + { + "epoch": 0.01, + "learning_rate": 9.179978700745475e-06, + "loss": 1.125, + "step": 431 + }, + { + "epoch": 0.01, + "learning_rate": 9.201277955271566e-06, + "loss": 1.0278, + "step": 432 + }, + { + "epoch": 0.01, + "learning_rate": 9.222577209797658e-06, + "loss": 1.0898, + "step": 433 + }, + { + "epoch": 0.01, + "learning_rate": 9.24387646432375e-06, + "loss": 1.1255, + "step": 434 + }, + { + "epoch": 0.01, + "learning_rate": 9.265175718849841e-06, + "loss": 1.1113, + "step": 435 + }, + { + "epoch": 0.01, + "learning_rate": 9.286474973375933e-06, + "loss": 1.2148, + "step": 436 + }, + { + "epoch": 0.01, + "learning_rate": 9.307774227902024e-06, + "loss": 1.1587, + "step": 437 + }, + { + "epoch": 0.01, + "learning_rate": 9.329073482428116e-06, + "loss": 0.999, + "step": 438 + }, + { + "epoch": 0.01, + "learning_rate": 9.350372736954207e-06, + "loss": 1.0718, + "step": 439 + }, + { + "epoch": 0.01, + "learning_rate": 9.3716719914803e-06, + "loss": 1.1025, + "step": 440 + }, + { + "epoch": 0.01, + "learning_rate": 9.39297124600639e-06, + "loss": 1.106, + "step": 441 + }, + { + "epoch": 0.01, + "learning_rate": 9.414270500532482e-06, + "loss": 1.189, + "step": 442 + }, + { + "epoch": 0.01, + "learning_rate": 9.435569755058573e-06, + "loss": 1.1875, + "step": 443 + }, + { + "epoch": 0.01, + "learning_rate": 9.456869009584665e-06, + "loss": 1.0679, + "step": 444 + }, + { + "epoch": 0.01, + "learning_rate": 9.478168264110756e-06, + "loss": 1.1582, + "step": 445 + }, + { + "epoch": 0.01, + "learning_rate": 9.49946751863685e-06, + "loss": 1.3345, + "step": 446 + }, + { + "epoch": 0.01, + "learning_rate": 9.52076677316294e-06, + "loss": 1.041, + "step": 447 + }, + { + "epoch": 0.01, + "learning_rate": 9.542066027689033e-06, + "loss": 1.1641, + "step": 448 + }, + { + "epoch": 0.01, + "learning_rate": 9.563365282215123e-06, + "loss": 1.0981, + "step": 449 + }, + { + "epoch": 0.01, + "learning_rate": 9.584664536741216e-06, + "loss": 1.0674, + "step": 450 + }, + { + "epoch": 0.01, + "learning_rate": 9.605963791267306e-06, + "loss": 1.1689, + "step": 451 + }, + { + "epoch": 0.01, + "learning_rate": 9.627263045793399e-06, + "loss": 1.1689, + "step": 452 + }, + { + "epoch": 0.01, + "learning_rate": 9.64856230031949e-06, + "loss": 1.145, + "step": 453 + }, + { + "epoch": 0.01, + "learning_rate": 9.669861554845582e-06, + "loss": 1.1772, + "step": 454 + }, + { + "epoch": 0.01, + "learning_rate": 9.691160809371672e-06, + "loss": 1.082, + "step": 455 + }, + { + "epoch": 0.01, + "learning_rate": 9.712460063897765e-06, + "loss": 1.2104, + "step": 456 + }, + { + "epoch": 0.01, + "learning_rate": 9.733759318423855e-06, + "loss": 1.207, + "step": 457 + }, + { + "epoch": 0.01, + "learning_rate": 9.755058572949948e-06, + "loss": 1.1445, + "step": 458 + }, + { + "epoch": 0.01, + "learning_rate": 9.77635782747604e-06, + "loss": 1.1357, + "step": 459 + }, + { + "epoch": 0.01, + "learning_rate": 9.79765708200213e-06, + "loss": 1.126, + "step": 460 + }, + { + "epoch": 0.01, + "learning_rate": 9.818956336528223e-06, + "loss": 1.0869, + "step": 461 + }, + { + "epoch": 0.01, + "learning_rate": 9.840255591054313e-06, + "loss": 1.2637, + "step": 462 + }, + { + "epoch": 0.01, + "learning_rate": 9.861554845580406e-06, + "loss": 1.1846, + "step": 463 + }, + { + "epoch": 0.01, + "learning_rate": 9.882854100106496e-06, + "loss": 0.3406, + "step": 464 + }, + { + "epoch": 0.01, + "learning_rate": 9.904153354632589e-06, + "loss": 1.1035, + "step": 465 + }, + { + "epoch": 0.01, + "learning_rate": 9.92545260915868e-06, + "loss": 1.1177, + "step": 466 + }, + { + "epoch": 0.01, + "learning_rate": 9.946751863684772e-06, + "loss": 1.1416, + "step": 467 + }, + { + "epoch": 0.01, + "learning_rate": 9.968051118210862e-06, + "loss": 1.1206, + "step": 468 + }, + { + "epoch": 0.01, + "learning_rate": 9.989350372736955e-06, + "loss": 1.0825, + "step": 469 + }, + { + "epoch": 0.02, + "learning_rate": 1.0010649627263047e-05, + "loss": 1.0532, + "step": 470 + }, + { + "epoch": 0.02, + "learning_rate": 1.0031948881789138e-05, + "loss": 1.0625, + "step": 471 + }, + { + "epoch": 0.02, + "learning_rate": 1.005324813631523e-05, + "loss": 1.1895, + "step": 472 + }, + { + "epoch": 0.02, + "learning_rate": 1.007454739084132e-05, + "loss": 0.3713, + "step": 473 + }, + { + "epoch": 0.02, + "learning_rate": 1.0095846645367413e-05, + "loss": 1.1392, + "step": 474 + }, + { + "epoch": 0.02, + "learning_rate": 1.0117145899893504e-05, + "loss": 1.1602, + "step": 475 + }, + { + "epoch": 0.02, + "learning_rate": 1.0138445154419596e-05, + "loss": 1.123, + "step": 476 + }, + { + "epoch": 0.02, + "learning_rate": 1.0159744408945688e-05, + "loss": 0.3911, + "step": 477 + }, + { + "epoch": 0.02, + "learning_rate": 1.0181043663471779e-05, + "loss": 1.0908, + "step": 478 + }, + { + "epoch": 0.02, + "learning_rate": 1.0202342917997871e-05, + "loss": 1.1274, + "step": 479 + }, + { + "epoch": 0.02, + "learning_rate": 1.0223642172523962e-05, + "loss": 1.1055, + "step": 480 + }, + { + "epoch": 0.02, + "learning_rate": 1.0244941427050054e-05, + "loss": 1.124, + "step": 481 + }, + { + "epoch": 0.02, + "learning_rate": 1.0266240681576145e-05, + "loss": 1.085, + "step": 482 + }, + { + "epoch": 0.02, + "learning_rate": 1.0287539936102237e-05, + "loss": 1.2202, + "step": 483 + }, + { + "epoch": 0.02, + "learning_rate": 1.0308839190628328e-05, + "loss": 1.1895, + "step": 484 + }, + { + "epoch": 0.02, + "learning_rate": 1.033013844515442e-05, + "loss": 1.0601, + "step": 485 + }, + { + "epoch": 0.02, + "learning_rate": 1.035143769968051e-05, + "loss": 1.1772, + "step": 486 + }, + { + "epoch": 0.02, + "learning_rate": 1.0372736954206603e-05, + "loss": 1.0947, + "step": 487 + }, + { + "epoch": 0.02, + "learning_rate": 1.0394036208732697e-05, + "loss": 1.2036, + "step": 488 + }, + { + "epoch": 0.02, + "learning_rate": 1.0415335463258786e-05, + "loss": 1.2832, + "step": 489 + }, + { + "epoch": 0.02, + "learning_rate": 1.043663471778488e-05, + "loss": 1.1421, + "step": 490 + }, + { + "epoch": 0.02, + "learning_rate": 1.0457933972310969e-05, + "loss": 1.0591, + "step": 491 + }, + { + "epoch": 0.02, + "learning_rate": 1.0479233226837063e-05, + "loss": 1.0264, + "step": 492 + }, + { + "epoch": 0.02, + "learning_rate": 1.0500532481363152e-05, + "loss": 1.0908, + "step": 493 + }, + { + "epoch": 0.02, + "learning_rate": 1.0521831735889246e-05, + "loss": 1.2617, + "step": 494 + }, + { + "epoch": 0.02, + "learning_rate": 1.0543130990415335e-05, + "loss": 1.1729, + "step": 495 + }, + { + "epoch": 0.02, + "learning_rate": 1.0564430244941429e-05, + "loss": 0.4077, + "step": 496 + }, + { + "epoch": 0.02, + "learning_rate": 1.0585729499467518e-05, + "loss": 1.0, + "step": 497 + }, + { + "epoch": 0.02, + "learning_rate": 1.0607028753993612e-05, + "loss": 1.165, + "step": 498 + }, + { + "epoch": 0.02, + "learning_rate": 1.06283280085197e-05, + "loss": 1.0044, + "step": 499 + }, + { + "epoch": 0.02, + "learning_rate": 1.0649627263045795e-05, + "loss": 1.061, + "step": 500 + }, + { + "epoch": 0.02, + "learning_rate": 1.0670926517571887e-05, + "loss": 1.0386, + "step": 501 + }, + { + "epoch": 0.02, + "learning_rate": 1.0692225772097978e-05, + "loss": 1.0391, + "step": 502 + }, + { + "epoch": 0.02, + "learning_rate": 1.071352502662407e-05, + "loss": 1.1948, + "step": 503 + }, + { + "epoch": 0.02, + "learning_rate": 1.073482428115016e-05, + "loss": 1.1387, + "step": 504 + }, + { + "epoch": 0.02, + "learning_rate": 1.0756123535676253e-05, + "loss": 1.1504, + "step": 505 + }, + { + "epoch": 0.02, + "learning_rate": 1.0777422790202344e-05, + "loss": 1.0972, + "step": 506 + }, + { + "epoch": 0.02, + "learning_rate": 1.0798722044728436e-05, + "loss": 1.2529, + "step": 507 + }, + { + "epoch": 0.02, + "learning_rate": 1.0820021299254527e-05, + "loss": 1.2944, + "step": 508 + }, + { + "epoch": 0.02, + "learning_rate": 1.0841320553780619e-05, + "loss": 1.2085, + "step": 509 + }, + { + "epoch": 0.02, + "learning_rate": 1.086261980830671e-05, + "loss": 1.1172, + "step": 510 + }, + { + "epoch": 0.02, + "learning_rate": 1.0883919062832802e-05, + "loss": 1.1631, + "step": 511 + }, + { + "epoch": 0.02, + "learning_rate": 1.0905218317358894e-05, + "loss": 1.1167, + "step": 512 + }, + { + "epoch": 0.02, + "learning_rate": 1.0926517571884985e-05, + "loss": 1.2075, + "step": 513 + }, + { + "epoch": 0.02, + "learning_rate": 1.0947816826411077e-05, + "loss": 1.1611, + "step": 514 + }, + { + "epoch": 0.02, + "learning_rate": 1.0969116080937168e-05, + "loss": 1.1631, + "step": 515 + }, + { + "epoch": 0.02, + "learning_rate": 1.099041533546326e-05, + "loss": 1.0537, + "step": 516 + }, + { + "epoch": 0.02, + "learning_rate": 1.101171458998935e-05, + "loss": 1.146, + "step": 517 + }, + { + "epoch": 0.02, + "learning_rate": 1.1033013844515443e-05, + "loss": 1.1074, + "step": 518 + }, + { + "epoch": 0.02, + "learning_rate": 1.1054313099041534e-05, + "loss": 1.3052, + "step": 519 + }, + { + "epoch": 0.02, + "learning_rate": 1.1075612353567626e-05, + "loss": 1.0845, + "step": 520 + }, + { + "epoch": 0.02, + "learning_rate": 1.1096911608093717e-05, + "loss": 1.0532, + "step": 521 + }, + { + "epoch": 0.02, + "learning_rate": 1.1118210862619809e-05, + "loss": 1.2188, + "step": 522 + }, + { + "epoch": 0.02, + "learning_rate": 1.1139510117145901e-05, + "loss": 0.3821, + "step": 523 + }, + { + "epoch": 0.02, + "learning_rate": 1.1160809371671992e-05, + "loss": 1.0601, + "step": 524 + }, + { + "epoch": 0.02, + "learning_rate": 1.1182108626198084e-05, + "loss": 1.1846, + "step": 525 + }, + { + "epoch": 0.02, + "learning_rate": 1.1203407880724175e-05, + "loss": 1.1704, + "step": 526 + }, + { + "epoch": 0.02, + "learning_rate": 1.1224707135250267e-05, + "loss": 1.1084, + "step": 527 + }, + { + "epoch": 0.02, + "learning_rate": 1.1246006389776358e-05, + "loss": 1.1406, + "step": 528 + }, + { + "epoch": 0.02, + "learning_rate": 1.126730564430245e-05, + "loss": 1.082, + "step": 529 + }, + { + "epoch": 0.02, + "learning_rate": 1.1288604898828541e-05, + "loss": 1.2324, + "step": 530 + }, + { + "epoch": 0.02, + "learning_rate": 1.1309904153354633e-05, + "loss": 1.1343, + "step": 531 + }, + { + "epoch": 0.02, + "learning_rate": 1.1331203407880724e-05, + "loss": 1.1182, + "step": 532 + }, + { + "epoch": 0.02, + "learning_rate": 1.1352502662406816e-05, + "loss": 1.2393, + "step": 533 + }, + { + "epoch": 0.02, + "learning_rate": 1.1373801916932907e-05, + "loss": 1.0923, + "step": 534 + }, + { + "epoch": 0.02, + "learning_rate": 1.1395101171459e-05, + "loss": 1.2695, + "step": 535 + }, + { + "epoch": 0.02, + "learning_rate": 1.1416400425985093e-05, + "loss": 1.209, + "step": 536 + }, + { + "epoch": 0.02, + "learning_rate": 1.1437699680511182e-05, + "loss": 1.0938, + "step": 537 + }, + { + "epoch": 0.02, + "learning_rate": 1.1458998935037276e-05, + "loss": 1.1074, + "step": 538 + }, + { + "epoch": 0.02, + "learning_rate": 1.1480298189563365e-05, + "loss": 0.3638, + "step": 539 + }, + { + "epoch": 0.02, + "learning_rate": 1.1501597444089459e-05, + "loss": 1.2148, + "step": 540 + }, + { + "epoch": 0.02, + "learning_rate": 1.1522896698615548e-05, + "loss": 1.1216, + "step": 541 + }, + { + "epoch": 0.02, + "learning_rate": 1.1544195953141642e-05, + "loss": 1.061, + "step": 542 + }, + { + "epoch": 0.02, + "learning_rate": 1.1565495207667731e-05, + "loss": 1.125, + "step": 543 + }, + { + "epoch": 0.02, + "learning_rate": 1.1586794462193825e-05, + "loss": 1.1611, + "step": 544 + }, + { + "epoch": 0.02, + "learning_rate": 1.1608093716719914e-05, + "loss": 1.1016, + "step": 545 + }, + { + "epoch": 0.02, + "learning_rate": 1.1629392971246008e-05, + "loss": 1.2451, + "step": 546 + }, + { + "epoch": 0.02, + "learning_rate": 1.16506922257721e-05, + "loss": 1.1567, + "step": 547 + }, + { + "epoch": 0.02, + "learning_rate": 1.1671991480298191e-05, + "loss": 1.2358, + "step": 548 + }, + { + "epoch": 0.02, + "learning_rate": 1.1693290734824283e-05, + "loss": 1.0459, + "step": 549 + }, + { + "epoch": 0.02, + "learning_rate": 1.1714589989350374e-05, + "loss": 1.0703, + "step": 550 + }, + { + "epoch": 0.02, + "learning_rate": 1.1735889243876466e-05, + "loss": 1.0479, + "step": 551 + }, + { + "epoch": 0.02, + "learning_rate": 1.1757188498402557e-05, + "loss": 1.106, + "step": 552 + }, + { + "epoch": 0.02, + "learning_rate": 1.177848775292865e-05, + "loss": 1.1953, + "step": 553 + }, + { + "epoch": 0.02, + "learning_rate": 1.179978700745474e-05, + "loss": 1.208, + "step": 554 + }, + { + "epoch": 0.02, + "learning_rate": 1.1821086261980832e-05, + "loss": 1.1025, + "step": 555 + }, + { + "epoch": 0.02, + "learning_rate": 1.1842385516506923e-05, + "loss": 1.1211, + "step": 556 + }, + { + "epoch": 0.02, + "learning_rate": 1.1863684771033015e-05, + "loss": 1.2217, + "step": 557 + }, + { + "epoch": 0.02, + "learning_rate": 1.1884984025559106e-05, + "loss": 0.9893, + "step": 558 + }, + { + "epoch": 0.02, + "learning_rate": 1.1906283280085198e-05, + "loss": 0.9829, + "step": 559 + }, + { + "epoch": 0.02, + "learning_rate": 1.192758253461129e-05, + "loss": 1.2305, + "step": 560 + }, + { + "epoch": 0.02, + "learning_rate": 1.1948881789137381e-05, + "loss": 1.248, + "step": 561 + }, + { + "epoch": 0.02, + "learning_rate": 1.1970181043663473e-05, + "loss": 1.0825, + "step": 562 + }, + { + "epoch": 0.02, + "learning_rate": 1.1991480298189564e-05, + "loss": 1.0903, + "step": 563 + }, + { + "epoch": 0.02, + "learning_rate": 1.2012779552715656e-05, + "loss": 1.2026, + "step": 564 + }, + { + "epoch": 0.02, + "learning_rate": 1.2034078807241747e-05, + "loss": 1.165, + "step": 565 + }, + { + "epoch": 0.02, + "learning_rate": 1.205537806176784e-05, + "loss": 1.1592, + "step": 566 + }, + { + "epoch": 0.02, + "learning_rate": 1.207667731629393e-05, + "loss": 1.0581, + "step": 567 + }, + { + "epoch": 0.02, + "learning_rate": 1.2097976570820022e-05, + "loss": 1.0869, + "step": 568 + }, + { + "epoch": 0.02, + "learning_rate": 1.2119275825346113e-05, + "loss": 1.0781, + "step": 569 + }, + { + "epoch": 0.02, + "learning_rate": 1.2140575079872205e-05, + "loss": 1.1084, + "step": 570 + }, + { + "epoch": 0.02, + "learning_rate": 1.2161874334398298e-05, + "loss": 1.0752, + "step": 571 + }, + { + "epoch": 0.02, + "learning_rate": 1.2183173588924388e-05, + "loss": 1.2388, + "step": 572 + }, + { + "epoch": 0.02, + "learning_rate": 1.220447284345048e-05, + "loss": 1.1206, + "step": 573 + }, + { + "epoch": 0.02, + "learning_rate": 1.2225772097976571e-05, + "loss": 1.2505, + "step": 574 + }, + { + "epoch": 0.02, + "learning_rate": 1.2247071352502664e-05, + "loss": 1.0693, + "step": 575 + }, + { + "epoch": 0.02, + "learning_rate": 1.2268370607028754e-05, + "loss": 1.1309, + "step": 576 + }, + { + "epoch": 0.02, + "learning_rate": 1.2289669861554846e-05, + "loss": 1.1357, + "step": 577 + }, + { + "epoch": 0.02, + "learning_rate": 1.2310969116080937e-05, + "loss": 1.1143, + "step": 578 + }, + { + "epoch": 0.02, + "learning_rate": 1.233226837060703e-05, + "loss": 1.1318, + "step": 579 + }, + { + "epoch": 0.02, + "learning_rate": 1.235356762513312e-05, + "loss": 1.0835, + "step": 580 + }, + { + "epoch": 0.02, + "learning_rate": 1.2374866879659212e-05, + "loss": 1.0601, + "step": 581 + }, + { + "epoch": 0.02, + "learning_rate": 1.2396166134185303e-05, + "loss": 1.1392, + "step": 582 + }, + { + "epoch": 0.02, + "learning_rate": 1.2417465388711395e-05, + "loss": 1.0752, + "step": 583 + }, + { + "epoch": 0.02, + "learning_rate": 1.243876464323749e-05, + "loss": 1.0986, + "step": 584 + }, + { + "epoch": 0.02, + "learning_rate": 1.2460063897763578e-05, + "loss": 1.0918, + "step": 585 + }, + { + "epoch": 0.02, + "learning_rate": 1.2481363152289672e-05, + "loss": 1.0601, + "step": 586 + }, + { + "epoch": 0.02, + "learning_rate": 1.2502662406815761e-05, + "loss": 1.1807, + "step": 587 + }, + { + "epoch": 0.02, + "learning_rate": 1.2523961661341855e-05, + "loss": 1.1323, + "step": 588 + }, + { + "epoch": 0.02, + "learning_rate": 1.2545260915867944e-05, + "loss": 1.207, + "step": 589 + }, + { + "epoch": 0.02, + "learning_rate": 1.2566560170394038e-05, + "loss": 1.1387, + "step": 590 + }, + { + "epoch": 0.02, + "learning_rate": 1.2587859424920127e-05, + "loss": 1.1162, + "step": 591 + }, + { + "epoch": 0.02, + "learning_rate": 1.2609158679446221e-05, + "loss": 1.0522, + "step": 592 + }, + { + "epoch": 0.02, + "learning_rate": 1.263045793397231e-05, + "loss": 1.1069, + "step": 593 + }, + { + "epoch": 0.02, + "learning_rate": 1.2651757188498404e-05, + "loss": 1.1406, + "step": 594 + }, + { + "epoch": 0.02, + "learning_rate": 1.2673056443024497e-05, + "loss": 1.1655, + "step": 595 + }, + { + "epoch": 0.02, + "learning_rate": 1.2694355697550587e-05, + "loss": 0.9995, + "step": 596 + }, + { + "epoch": 0.02, + "learning_rate": 1.271565495207668e-05, + "loss": 1.0508, + "step": 597 + }, + { + "epoch": 0.02, + "learning_rate": 1.273695420660277e-05, + "loss": 1.2109, + "step": 598 + }, + { + "epoch": 0.02, + "learning_rate": 1.2758253461128862e-05, + "loss": 1.1226, + "step": 599 + }, + { + "epoch": 0.02, + "learning_rate": 1.2779552715654953e-05, + "loss": 1.2183, + "step": 600 + }, + { + "epoch": 0.02, + "learning_rate": 1.2800851970181045e-05, + "loss": 1.167, + "step": 601 + }, + { + "epoch": 0.02, + "learning_rate": 1.2822151224707136e-05, + "loss": 1.1924, + "step": 602 + }, + { + "epoch": 0.02, + "learning_rate": 1.2843450479233228e-05, + "loss": 1.1089, + "step": 603 + }, + { + "epoch": 0.02, + "learning_rate": 1.2864749733759319e-05, + "loss": 1.0781, + "step": 604 + }, + { + "epoch": 0.02, + "learning_rate": 1.2886048988285411e-05, + "loss": 1.1055, + "step": 605 + }, + { + "epoch": 0.02, + "learning_rate": 1.2907348242811502e-05, + "loss": 1.1738, + "step": 606 + }, + { + "epoch": 0.02, + "learning_rate": 1.2928647497337594e-05, + "loss": 1.1274, + "step": 607 + }, + { + "epoch": 0.02, + "learning_rate": 1.2949946751863687e-05, + "loss": 1.2021, + "step": 608 + }, + { + "epoch": 0.02, + "learning_rate": 1.2971246006389777e-05, + "loss": 1.1904, + "step": 609 + }, + { + "epoch": 0.02, + "learning_rate": 1.299254526091587e-05, + "loss": 1.0205, + "step": 610 + }, + { + "epoch": 0.02, + "learning_rate": 1.301384451544196e-05, + "loss": 1.0659, + "step": 611 + }, + { + "epoch": 0.02, + "learning_rate": 1.3035143769968053e-05, + "loss": 1.1021, + "step": 612 + }, + { + "epoch": 0.02, + "learning_rate": 1.3056443024494143e-05, + "loss": 1.2061, + "step": 613 + }, + { + "epoch": 0.02, + "learning_rate": 1.3077742279020236e-05, + "loss": 1.1309, + "step": 614 + }, + { + "epoch": 0.02, + "learning_rate": 1.3099041533546326e-05, + "loss": 1.0483, + "step": 615 + }, + { + "epoch": 0.02, + "learning_rate": 1.3120340788072418e-05, + "loss": 1.147, + "step": 616 + }, + { + "epoch": 0.02, + "learning_rate": 1.3141640042598509e-05, + "loss": 1.2397, + "step": 617 + }, + { + "epoch": 0.02, + "learning_rate": 1.3162939297124601e-05, + "loss": 1.2324, + "step": 618 + }, + { + "epoch": 0.02, + "learning_rate": 1.3184238551650694e-05, + "loss": 1.1465, + "step": 619 + }, + { + "epoch": 0.02, + "learning_rate": 1.3205537806176784e-05, + "loss": 1.1445, + "step": 620 + }, + { + "epoch": 0.02, + "learning_rate": 1.3226837060702877e-05, + "loss": 1.0801, + "step": 621 + }, + { + "epoch": 0.02, + "learning_rate": 1.3248136315228967e-05, + "loss": 1.1279, + "step": 622 + }, + { + "epoch": 0.02, + "learning_rate": 1.326943556975506e-05, + "loss": 1.1099, + "step": 623 + }, + { + "epoch": 0.02, + "learning_rate": 1.329073482428115e-05, + "loss": 1.1768, + "step": 624 + }, + { + "epoch": 0.02, + "learning_rate": 1.3312034078807243e-05, + "loss": 1.1646, + "step": 625 + }, + { + "epoch": 0.02, + "learning_rate": 1.3333333333333333e-05, + "loss": 1.1187, + "step": 626 + }, + { + "epoch": 0.02, + "learning_rate": 1.3354632587859426e-05, + "loss": 1.1758, + "step": 627 + }, + { + "epoch": 0.02, + "learning_rate": 1.3375931842385516e-05, + "loss": 1.1274, + "step": 628 + }, + { + "epoch": 0.02, + "learning_rate": 1.3397231096911609e-05, + "loss": 1.167, + "step": 629 + }, + { + "epoch": 0.02, + "learning_rate": 1.3418530351437703e-05, + "loss": 1.1572, + "step": 630 + }, + { + "epoch": 0.02, + "learning_rate": 1.3439829605963792e-05, + "loss": 1.1943, + "step": 631 + }, + { + "epoch": 0.02, + "learning_rate": 1.3461128860489886e-05, + "loss": 1.1113, + "step": 632 + }, + { + "epoch": 0.02, + "learning_rate": 1.3482428115015975e-05, + "loss": 1.2402, + "step": 633 + }, + { + "epoch": 0.02, + "learning_rate": 1.3503727369542069e-05, + "loss": 1.1812, + "step": 634 + }, + { + "epoch": 0.02, + "learning_rate": 1.3525026624068157e-05, + "loss": 1.1396, + "step": 635 + }, + { + "epoch": 0.02, + "learning_rate": 1.3546325878594251e-05, + "loss": 1.145, + "step": 636 + }, + { + "epoch": 0.02, + "learning_rate": 1.356762513312034e-05, + "loss": 1.1548, + "step": 637 + }, + { + "epoch": 0.02, + "learning_rate": 1.3588924387646434e-05, + "loss": 1.1323, + "step": 638 + }, + { + "epoch": 0.02, + "learning_rate": 1.3610223642172523e-05, + "loss": 1.1348, + "step": 639 + }, + { + "epoch": 0.02, + "learning_rate": 1.3631522896698617e-05, + "loss": 1.2559, + "step": 640 + }, + { + "epoch": 0.02, + "learning_rate": 1.3652822151224708e-05, + "loss": 1.1714, + "step": 641 + }, + { + "epoch": 0.02, + "learning_rate": 1.36741214057508e-05, + "loss": 1.1401, + "step": 642 + }, + { + "epoch": 0.02, + "learning_rate": 1.3695420660276893e-05, + "loss": 1.1758, + "step": 643 + }, + { + "epoch": 0.02, + "learning_rate": 1.3716719914802983e-05, + "loss": 1.2002, + "step": 644 + }, + { + "epoch": 0.02, + "learning_rate": 1.3738019169329076e-05, + "loss": 1.1997, + "step": 645 + }, + { + "epoch": 0.02, + "learning_rate": 1.3759318423855166e-05, + "loss": 1.2373, + "step": 646 + }, + { + "epoch": 0.02, + "learning_rate": 1.3780617678381259e-05, + "loss": 1.0605, + "step": 647 + }, + { + "epoch": 0.02, + "learning_rate": 1.380191693290735e-05, + "loss": 1.1499, + "step": 648 + }, + { + "epoch": 0.02, + "learning_rate": 1.3823216187433442e-05, + "loss": 1.1309, + "step": 649 + }, + { + "epoch": 0.02, + "learning_rate": 1.3844515441959532e-05, + "loss": 1.124, + "step": 650 + }, + { + "epoch": 0.02, + "learning_rate": 1.3865814696485625e-05, + "loss": 1.1318, + "step": 651 + }, + { + "epoch": 0.02, + "learning_rate": 1.3887113951011715e-05, + "loss": 1.103, + "step": 652 + }, + { + "epoch": 0.02, + "learning_rate": 1.3908413205537808e-05, + "loss": 1.1836, + "step": 653 + }, + { + "epoch": 0.02, + "learning_rate": 1.39297124600639e-05, + "loss": 1.0068, + "step": 654 + }, + { + "epoch": 0.02, + "learning_rate": 1.395101171458999e-05, + "loss": 1.1655, + "step": 655 + }, + { + "epoch": 0.02, + "learning_rate": 1.3972310969116083e-05, + "loss": 1.1729, + "step": 656 + }, + { + "epoch": 0.02, + "learning_rate": 1.3993610223642173e-05, + "loss": 1.0913, + "step": 657 + }, + { + "epoch": 0.02, + "learning_rate": 1.4014909478168266e-05, + "loss": 1.1538, + "step": 658 + }, + { + "epoch": 0.02, + "learning_rate": 1.4036208732694356e-05, + "loss": 1.0571, + "step": 659 + }, + { + "epoch": 0.02, + "learning_rate": 1.4057507987220449e-05, + "loss": 1.0435, + "step": 660 + }, + { + "epoch": 0.02, + "learning_rate": 1.407880724174654e-05, + "loss": 1.186, + "step": 661 + }, + { + "epoch": 0.02, + "learning_rate": 1.4100106496272632e-05, + "loss": 1.1421, + "step": 662 + }, + { + "epoch": 0.02, + "learning_rate": 1.4121405750798722e-05, + "loss": 1.2207, + "step": 663 + }, + { + "epoch": 0.02, + "learning_rate": 1.4142705005324815e-05, + "loss": 1.1968, + "step": 664 + }, + { + "epoch": 0.02, + "learning_rate": 1.4164004259850905e-05, + "loss": 1.187, + "step": 665 + }, + { + "epoch": 0.02, + "learning_rate": 1.4185303514376998e-05, + "loss": 1.2402, + "step": 666 + }, + { + "epoch": 0.02, + "learning_rate": 1.420660276890309e-05, + "loss": 1.105, + "step": 667 + }, + { + "epoch": 0.02, + "learning_rate": 1.422790202342918e-05, + "loss": 1.167, + "step": 668 + }, + { + "epoch": 0.02, + "learning_rate": 1.4249201277955273e-05, + "loss": 1.1387, + "step": 669 + }, + { + "epoch": 0.02, + "learning_rate": 1.4270500532481364e-05, + "loss": 1.0898, + "step": 670 + }, + { + "epoch": 0.02, + "learning_rate": 1.4291799787007456e-05, + "loss": 1.1963, + "step": 671 + }, + { + "epoch": 0.02, + "learning_rate": 1.4313099041533547e-05, + "loss": 1.0913, + "step": 672 + }, + { + "epoch": 0.02, + "learning_rate": 1.4334398296059639e-05, + "loss": 1.145, + "step": 673 + }, + { + "epoch": 0.02, + "learning_rate": 1.435569755058573e-05, + "loss": 1.1797, + "step": 674 + }, + { + "epoch": 0.02, + "learning_rate": 1.4376996805111822e-05, + "loss": 1.1606, + "step": 675 + }, + { + "epoch": 0.02, + "learning_rate": 1.4398296059637912e-05, + "loss": 1.1118, + "step": 676 + }, + { + "epoch": 0.02, + "learning_rate": 1.4419595314164005e-05, + "loss": 1.2354, + "step": 677 + }, + { + "epoch": 0.02, + "learning_rate": 1.4440894568690099e-05, + "loss": 1.1899, + "step": 678 + }, + { + "epoch": 0.02, + "learning_rate": 1.4462193823216188e-05, + "loss": 1.1436, + "step": 679 + }, + { + "epoch": 0.02, + "learning_rate": 1.4483493077742282e-05, + "loss": 1.1348, + "step": 680 + }, + { + "epoch": 0.02, + "learning_rate": 1.450479233226837e-05, + "loss": 1.062, + "step": 681 + }, + { + "epoch": 0.02, + "learning_rate": 1.4526091586794465e-05, + "loss": 1.1021, + "step": 682 + }, + { + "epoch": 0.02, + "learning_rate": 1.4547390841320554e-05, + "loss": 1.2842, + "step": 683 + }, + { + "epoch": 0.02, + "learning_rate": 1.4568690095846648e-05, + "loss": 1.1587, + "step": 684 + }, + { + "epoch": 0.02, + "learning_rate": 1.4589989350372737e-05, + "loss": 1.1494, + "step": 685 + }, + { + "epoch": 0.02, + "learning_rate": 1.461128860489883e-05, + "loss": 1.0967, + "step": 686 + }, + { + "epoch": 0.02, + "learning_rate": 1.4632587859424921e-05, + "loss": 1.0576, + "step": 687 + }, + { + "epoch": 0.02, + "learning_rate": 1.4653887113951014e-05, + "loss": 1.0371, + "step": 688 + }, + { + "epoch": 0.02, + "learning_rate": 1.4675186368477104e-05, + "loss": 0.9722, + "step": 689 + }, + { + "epoch": 0.02, + "learning_rate": 1.4696485623003197e-05, + "loss": 1.3174, + "step": 690 + }, + { + "epoch": 0.02, + "learning_rate": 1.4717784877529289e-05, + "loss": 1.0796, + "step": 691 + }, + { + "epoch": 0.02, + "learning_rate": 1.473908413205538e-05, + "loss": 1.2119, + "step": 692 + }, + { + "epoch": 0.02, + "learning_rate": 1.4760383386581472e-05, + "loss": 1.0771, + "step": 693 + }, + { + "epoch": 0.02, + "learning_rate": 1.4781682641107562e-05, + "loss": 1.1768, + "step": 694 + }, + { + "epoch": 0.02, + "learning_rate": 1.4802981895633655e-05, + "loss": 1.23, + "step": 695 + }, + { + "epoch": 0.02, + "learning_rate": 1.4824281150159745e-05, + "loss": 1.1636, + "step": 696 + }, + { + "epoch": 0.02, + "learning_rate": 1.4845580404685838e-05, + "loss": 1.2607, + "step": 697 + }, + { + "epoch": 0.02, + "learning_rate": 1.4866879659211928e-05, + "loss": 1.0884, + "step": 698 + }, + { + "epoch": 0.02, + "learning_rate": 1.488817891373802e-05, + "loss": 1.1196, + "step": 699 + }, + { + "epoch": 0.02, + "learning_rate": 1.4909478168264111e-05, + "loss": 1.103, + "step": 700 + }, + { + "epoch": 0.02, + "learning_rate": 1.4930777422790204e-05, + "loss": 1.1436, + "step": 701 + }, + { + "epoch": 0.02, + "learning_rate": 1.4952076677316296e-05, + "loss": 1.1382, + "step": 702 + }, + { + "epoch": 0.02, + "learning_rate": 1.4973375931842387e-05, + "loss": 1.2075, + "step": 703 + }, + { + "epoch": 0.02, + "learning_rate": 1.4994675186368479e-05, + "loss": 1.0308, + "step": 704 + }, + { + "epoch": 0.02, + "learning_rate": 1.501597444089457e-05, + "loss": 1.2412, + "step": 705 + }, + { + "epoch": 0.02, + "learning_rate": 1.5037273695420662e-05, + "loss": 1.1616, + "step": 706 + }, + { + "epoch": 0.02, + "learning_rate": 1.5058572949946753e-05, + "loss": 1.2017, + "step": 707 + }, + { + "epoch": 0.02, + "learning_rate": 1.5079872204472845e-05, + "loss": 1.1367, + "step": 708 + }, + { + "epoch": 0.02, + "learning_rate": 1.5101171458998936e-05, + "loss": 1.1279, + "step": 709 + }, + { + "epoch": 0.02, + "learning_rate": 1.5122470713525028e-05, + "loss": 1.2153, + "step": 710 + }, + { + "epoch": 0.02, + "learning_rate": 1.5143769968051119e-05, + "loss": 1.1719, + "step": 711 + }, + { + "epoch": 0.02, + "learning_rate": 1.516506922257721e-05, + "loss": 1.2441, + "step": 712 + }, + { + "epoch": 0.02, + "learning_rate": 1.5186368477103301e-05, + "loss": 1.0479, + "step": 713 + }, + { + "epoch": 0.02, + "learning_rate": 1.5207667731629394e-05, + "loss": 1.1094, + "step": 714 + }, + { + "epoch": 0.02, + "learning_rate": 1.5228966986155486e-05, + "loss": 1.0811, + "step": 715 + }, + { + "epoch": 0.02, + "learning_rate": 1.5250266240681577e-05, + "loss": 1.0513, + "step": 716 + }, + { + "epoch": 0.02, + "learning_rate": 1.527156549520767e-05, + "loss": 1.1416, + "step": 717 + }, + { + "epoch": 0.02, + "learning_rate": 1.529286474973376e-05, + "loss": 1.1201, + "step": 718 + }, + { + "epoch": 0.02, + "learning_rate": 1.5314164004259852e-05, + "loss": 1.2847, + "step": 719 + }, + { + "epoch": 0.02, + "learning_rate": 1.5335463258785944e-05, + "loss": 1.1118, + "step": 720 + }, + { + "epoch": 0.02, + "learning_rate": 1.5356762513312037e-05, + "loss": 1.271, + "step": 721 + }, + { + "epoch": 0.02, + "learning_rate": 1.5378061767838126e-05, + "loss": 1.1108, + "step": 722 + }, + { + "epoch": 0.02, + "learning_rate": 1.5399361022364218e-05, + "loss": 1.1904, + "step": 723 + }, + { + "epoch": 0.02, + "learning_rate": 1.542066027689031e-05, + "loss": 1.0732, + "step": 724 + }, + { + "epoch": 0.02, + "learning_rate": 1.5441959531416403e-05, + "loss": 1.1802, + "step": 725 + }, + { + "epoch": 0.02, + "learning_rate": 1.5463258785942495e-05, + "loss": 1.124, + "step": 726 + }, + { + "epoch": 0.02, + "learning_rate": 1.5484558040468584e-05, + "loss": 1.0112, + "step": 727 + }, + { + "epoch": 0.02, + "learning_rate": 1.5505857294994676e-05, + "loss": 1.1973, + "step": 728 + }, + { + "epoch": 0.02, + "learning_rate": 1.552715654952077e-05, + "loss": 1.2012, + "step": 729 + }, + { + "epoch": 0.02, + "learning_rate": 1.554845580404686e-05, + "loss": 1.2271, + "step": 730 + }, + { + "epoch": 0.02, + "learning_rate": 1.556975505857295e-05, + "loss": 1.0713, + "step": 731 + }, + { + "epoch": 0.02, + "learning_rate": 1.5591054313099042e-05, + "loss": 1.2061, + "step": 732 + }, + { + "epoch": 0.02, + "learning_rate": 1.5612353567625134e-05, + "loss": 0.4187, + "step": 733 + }, + { + "epoch": 0.02, + "learning_rate": 1.5633652822151227e-05, + "loss": 1.1177, + "step": 734 + }, + { + "epoch": 0.02, + "learning_rate": 1.5654952076677316e-05, + "loss": 1.1372, + "step": 735 + }, + { + "epoch": 0.02, + "learning_rate": 1.5676251331203408e-05, + "loss": 1.1377, + "step": 736 + }, + { + "epoch": 0.02, + "learning_rate": 1.56975505857295e-05, + "loss": 1.1157, + "step": 737 + }, + { + "epoch": 0.02, + "learning_rate": 1.5718849840255593e-05, + "loss": 1.1572, + "step": 738 + }, + { + "epoch": 0.02, + "learning_rate": 1.5740149094781685e-05, + "loss": 1.1519, + "step": 739 + }, + { + "epoch": 0.02, + "learning_rate": 1.5761448349307774e-05, + "loss": 1.1934, + "step": 740 + }, + { + "epoch": 0.02, + "learning_rate": 1.5782747603833866e-05, + "loss": 1.2842, + "step": 741 + }, + { + "epoch": 0.02, + "learning_rate": 1.580404685835996e-05, + "loss": 1.1787, + "step": 742 + }, + { + "epoch": 0.02, + "learning_rate": 1.582534611288605e-05, + "loss": 1.0386, + "step": 743 + }, + { + "epoch": 0.02, + "learning_rate": 1.584664536741214e-05, + "loss": 1.0845, + "step": 744 + }, + { + "epoch": 0.02, + "learning_rate": 1.5867944621938232e-05, + "loss": 1.1572, + "step": 745 + }, + { + "epoch": 0.02, + "learning_rate": 1.5889243876464325e-05, + "loss": 1.292, + "step": 746 + }, + { + "epoch": 0.02, + "learning_rate": 1.5910543130990417e-05, + "loss": 1.1602, + "step": 747 + }, + { + "epoch": 0.02, + "learning_rate": 1.5931842385516506e-05, + "loss": 1.1846, + "step": 748 + }, + { + "epoch": 0.02, + "learning_rate": 1.5953141640042598e-05, + "loss": 1.0522, + "step": 749 + }, + { + "epoch": 0.02, + "learning_rate": 1.5974440894568694e-05, + "loss": 1.1855, + "step": 750 + }, + { + "epoch": 0.02, + "learning_rate": 1.5995740149094783e-05, + "loss": 1.2334, + "step": 751 + }, + { + "epoch": 0.02, + "learning_rate": 1.6017039403620875e-05, + "loss": 1.1558, + "step": 752 + }, + { + "epoch": 0.02, + "learning_rate": 1.6038338658146964e-05, + "loss": 1.0728, + "step": 753 + }, + { + "epoch": 0.02, + "learning_rate": 1.605963791267306e-05, + "loss": 1.0967, + "step": 754 + }, + { + "epoch": 0.02, + "learning_rate": 1.608093716719915e-05, + "loss": 1.2422, + "step": 755 + }, + { + "epoch": 0.02, + "learning_rate": 1.610223642172524e-05, + "loss": 1.0005, + "step": 756 + }, + { + "epoch": 0.02, + "learning_rate": 1.6123535676251333e-05, + "loss": 1.1855, + "step": 757 + }, + { + "epoch": 0.02, + "learning_rate": 1.6144834930777426e-05, + "loss": 1.0913, + "step": 758 + }, + { + "epoch": 0.02, + "learning_rate": 1.6166134185303515e-05, + "loss": 1.1021, + "step": 759 + }, + { + "epoch": 0.02, + "learning_rate": 1.6187433439829607e-05, + "loss": 1.2051, + "step": 760 + }, + { + "epoch": 0.02, + "learning_rate": 1.62087326943557e-05, + "loss": 1.2109, + "step": 761 + }, + { + "epoch": 0.02, + "learning_rate": 1.623003194888179e-05, + "loss": 1.1245, + "step": 762 + }, + { + "epoch": 0.02, + "learning_rate": 1.6251331203407884e-05, + "loss": 1.0493, + "step": 763 + }, + { + "epoch": 0.02, + "learning_rate": 1.6272630457933973e-05, + "loss": 1.0234, + "step": 764 + }, + { + "epoch": 0.02, + "learning_rate": 1.6293929712460065e-05, + "loss": 1.0952, + "step": 765 + }, + { + "epoch": 0.02, + "learning_rate": 1.6315228966986158e-05, + "loss": 1.0952, + "step": 766 + }, + { + "epoch": 0.02, + "learning_rate": 1.633652822151225e-05, + "loss": 1.2036, + "step": 767 + }, + { + "epoch": 0.02, + "learning_rate": 1.635782747603834e-05, + "loss": 1.1997, + "step": 768 + }, + { + "epoch": 0.02, + "learning_rate": 1.637912673056443e-05, + "loss": 1.146, + "step": 769 + }, + { + "epoch": 0.02, + "learning_rate": 1.6400425985090524e-05, + "loss": 1.0552, + "step": 770 + }, + { + "epoch": 0.02, + "learning_rate": 1.6421725239616616e-05, + "loss": 1.0488, + "step": 771 + }, + { + "epoch": 0.02, + "learning_rate": 1.6443024494142705e-05, + "loss": 1.1797, + "step": 772 + }, + { + "epoch": 0.02, + "learning_rate": 1.6464323748668797e-05, + "loss": 1.1885, + "step": 773 + }, + { + "epoch": 0.02, + "learning_rate": 1.648562300319489e-05, + "loss": 1.1611, + "step": 774 + }, + { + "epoch": 0.02, + "learning_rate": 1.6506922257720982e-05, + "loss": 1.084, + "step": 775 + }, + { + "epoch": 0.02, + "learning_rate": 1.6528221512247074e-05, + "loss": 1.2441, + "step": 776 + }, + { + "epoch": 0.02, + "learning_rate": 1.6549520766773163e-05, + "loss": 1.126, + "step": 777 + }, + { + "epoch": 0.02, + "learning_rate": 1.6570820021299255e-05, + "loss": 1.0605, + "step": 778 + }, + { + "epoch": 0.02, + "learning_rate": 1.6592119275825348e-05, + "loss": 1.1138, + "step": 779 + }, + { + "epoch": 0.02, + "learning_rate": 1.661341853035144e-05, + "loss": 1.2603, + "step": 780 + }, + { + "epoch": 0.02, + "learning_rate": 1.663471778487753e-05, + "loss": 1.126, + "step": 781 + }, + { + "epoch": 0.03, + "learning_rate": 1.665601703940362e-05, + "loss": 1.1084, + "step": 782 + }, + { + "epoch": 0.03, + "learning_rate": 1.6677316293929714e-05, + "loss": 1.1426, + "step": 783 + }, + { + "epoch": 0.03, + "learning_rate": 1.6698615548455806e-05, + "loss": 1.0942, + "step": 784 + }, + { + "epoch": 0.03, + "learning_rate": 1.6719914802981898e-05, + "loss": 1.1729, + "step": 785 + }, + { + "epoch": 0.03, + "learning_rate": 1.6741214057507987e-05, + "loss": 1.23, + "step": 786 + }, + { + "epoch": 0.03, + "learning_rate": 1.676251331203408e-05, + "loss": 1.1494, + "step": 787 + }, + { + "epoch": 0.03, + "learning_rate": 1.6783812566560172e-05, + "loss": 1.1724, + "step": 788 + }, + { + "epoch": 0.03, + "learning_rate": 1.6805111821086264e-05, + "loss": 1.2812, + "step": 789 + }, + { + "epoch": 0.03, + "learning_rate": 1.6826411075612353e-05, + "loss": 1.2988, + "step": 790 + }, + { + "epoch": 0.03, + "learning_rate": 1.6847710330138445e-05, + "loss": 1.1572, + "step": 791 + }, + { + "epoch": 0.03, + "learning_rate": 1.6869009584664538e-05, + "loss": 1.2134, + "step": 792 + }, + { + "epoch": 0.03, + "learning_rate": 1.689030883919063e-05, + "loss": 1.1509, + "step": 793 + }, + { + "epoch": 0.03, + "learning_rate": 1.691160809371672e-05, + "loss": 1.103, + "step": 794 + }, + { + "epoch": 0.03, + "learning_rate": 1.693290734824281e-05, + "loss": 1.1621, + "step": 795 + }, + { + "epoch": 0.03, + "learning_rate": 1.6954206602768904e-05, + "loss": 0.3845, + "step": 796 + }, + { + "epoch": 0.03, + "learning_rate": 1.6975505857294996e-05, + "loss": 1.1582, + "step": 797 + }, + { + "epoch": 0.03, + "learning_rate": 1.699680511182109e-05, + "loss": 1.2295, + "step": 798 + }, + { + "epoch": 0.03, + "learning_rate": 1.7018104366347177e-05, + "loss": 1.3047, + "step": 799 + }, + { + "epoch": 0.03, + "learning_rate": 1.7039403620873273e-05, + "loss": 1.1602, + "step": 800 + }, + { + "epoch": 0.03, + "learning_rate": 1.7060702875399362e-05, + "loss": 1.2095, + "step": 801 + }, + { + "epoch": 0.03, + "learning_rate": 1.7082002129925454e-05, + "loss": 1.001, + "step": 802 + }, + { + "epoch": 0.03, + "learning_rate": 1.7103301384451547e-05, + "loss": 1.0728, + "step": 803 + }, + { + "epoch": 0.03, + "learning_rate": 1.712460063897764e-05, + "loss": 1.0444, + "step": 804 + }, + { + "epoch": 0.03, + "learning_rate": 1.7145899893503728e-05, + "loss": 1.1177, + "step": 805 + }, + { + "epoch": 0.03, + "learning_rate": 1.716719914802982e-05, + "loss": 1.0659, + "step": 806 + }, + { + "epoch": 0.03, + "learning_rate": 1.7188498402555913e-05, + "loss": 1.1646, + "step": 807 + }, + { + "epoch": 0.03, + "learning_rate": 1.7209797657082005e-05, + "loss": 1.1401, + "step": 808 + }, + { + "epoch": 0.03, + "learning_rate": 1.7231096911608097e-05, + "loss": 1.1353, + "step": 809 + }, + { + "epoch": 0.03, + "learning_rate": 1.7252396166134186e-05, + "loss": 1.0923, + "step": 810 + }, + { + "epoch": 0.03, + "learning_rate": 1.727369542066028e-05, + "loss": 1.0498, + "step": 811 + }, + { + "epoch": 0.03, + "learning_rate": 1.729499467518637e-05, + "loss": 0.4221, + "step": 812 + }, + { + "epoch": 0.03, + "learning_rate": 1.7316293929712463e-05, + "loss": 1.0854, + "step": 813 + }, + { + "epoch": 0.03, + "learning_rate": 1.7337593184238552e-05, + "loss": 1.2134, + "step": 814 + }, + { + "epoch": 0.03, + "learning_rate": 1.7358892438764644e-05, + "loss": 1.2231, + "step": 815 + }, + { + "epoch": 0.03, + "learning_rate": 1.7380191693290737e-05, + "loss": 1.1533, + "step": 816 + }, + { + "epoch": 0.03, + "learning_rate": 1.740149094781683e-05, + "loss": 1.2236, + "step": 817 + }, + { + "epoch": 0.03, + "learning_rate": 1.7422790202342918e-05, + "loss": 1.061, + "step": 818 + }, + { + "epoch": 0.03, + "learning_rate": 1.744408945686901e-05, + "loss": 1.1987, + "step": 819 + }, + { + "epoch": 0.03, + "learning_rate": 1.7465388711395103e-05, + "loss": 1.2139, + "step": 820 + }, + { + "epoch": 0.03, + "learning_rate": 1.7486687965921195e-05, + "loss": 1.2124, + "step": 821 + }, + { + "epoch": 0.03, + "learning_rate": 1.7507987220447287e-05, + "loss": 1.1665, + "step": 822 + }, + { + "epoch": 0.03, + "learning_rate": 1.7529286474973376e-05, + "loss": 1.2119, + "step": 823 + }, + { + "epoch": 0.03, + "learning_rate": 1.755058572949947e-05, + "loss": 1.1797, + "step": 824 + }, + { + "epoch": 0.03, + "learning_rate": 1.757188498402556e-05, + "loss": 1.0088, + "step": 825 + }, + { + "epoch": 0.03, + "learning_rate": 1.7593184238551653e-05, + "loss": 1.1235, + "step": 826 + }, + { + "epoch": 0.03, + "learning_rate": 1.7614483493077742e-05, + "loss": 1.0918, + "step": 827 + }, + { + "epoch": 0.03, + "learning_rate": 1.7635782747603835e-05, + "loss": 1.2188, + "step": 828 + }, + { + "epoch": 0.03, + "learning_rate": 1.7657082002129927e-05, + "loss": 1.2202, + "step": 829 + }, + { + "epoch": 0.03, + "learning_rate": 1.767838125665602e-05, + "loss": 1.2842, + "step": 830 + }, + { + "epoch": 0.03, + "learning_rate": 1.7699680511182108e-05, + "loss": 1.2852, + "step": 831 + }, + { + "epoch": 0.03, + "learning_rate": 1.77209797657082e-05, + "loss": 1.252, + "step": 832 + }, + { + "epoch": 0.03, + "learning_rate": 1.7742279020234293e-05, + "loss": 1.1572, + "step": 833 + }, + { + "epoch": 0.03, + "learning_rate": 1.7763578274760385e-05, + "loss": 1.1699, + "step": 834 + }, + { + "epoch": 0.03, + "learning_rate": 1.7784877529286477e-05, + "loss": 1.0469, + "step": 835 + }, + { + "epoch": 0.03, + "learning_rate": 1.7806176783812566e-05, + "loss": 1.1787, + "step": 836 + }, + { + "epoch": 0.03, + "learning_rate": 1.782747603833866e-05, + "loss": 1.0762, + "step": 837 + }, + { + "epoch": 0.03, + "learning_rate": 1.784877529286475e-05, + "loss": 1.1709, + "step": 838 + }, + { + "epoch": 0.03, + "learning_rate": 1.7870074547390843e-05, + "loss": 0.4316, + "step": 839 + }, + { + "epoch": 0.03, + "learning_rate": 1.7891373801916932e-05, + "loss": 1.1084, + "step": 840 + }, + { + "epoch": 0.03, + "learning_rate": 1.7912673056443025e-05, + "loss": 1.2153, + "step": 841 + }, + { + "epoch": 0.03, + "learning_rate": 1.7933972310969117e-05, + "loss": 1.1191, + "step": 842 + }, + { + "epoch": 0.03, + "learning_rate": 1.795527156549521e-05, + "loss": 1.1709, + "step": 843 + }, + { + "epoch": 0.03, + "learning_rate": 1.79765708200213e-05, + "loss": 1.2085, + "step": 844 + }, + { + "epoch": 0.03, + "learning_rate": 1.799787007454739e-05, + "loss": 1.2188, + "step": 845 + }, + { + "epoch": 0.03, + "learning_rate": 1.8019169329073486e-05, + "loss": 1.147, + "step": 846 + }, + { + "epoch": 0.03, + "learning_rate": 1.8040468583599575e-05, + "loss": 1.0771, + "step": 847 + }, + { + "epoch": 0.03, + "learning_rate": 1.8061767838125668e-05, + "loss": 1.2085, + "step": 848 + }, + { + "epoch": 0.03, + "learning_rate": 1.808306709265176e-05, + "loss": 1.2471, + "step": 849 + }, + { + "epoch": 0.03, + "learning_rate": 1.8104366347177852e-05, + "loss": 1.2446, + "step": 850 + }, + { + "epoch": 0.03, + "learning_rate": 1.812566560170394e-05, + "loss": 1.1357, + "step": 851 + }, + { + "epoch": 0.03, + "learning_rate": 1.8146964856230033e-05, + "loss": 1.2773, + "step": 852 + }, + { + "epoch": 0.03, + "learning_rate": 1.8168264110756126e-05, + "loss": 1.1792, + "step": 853 + }, + { + "epoch": 0.03, + "learning_rate": 1.8189563365282218e-05, + "loss": 1.1514, + "step": 854 + }, + { + "epoch": 0.03, + "learning_rate": 1.8210862619808307e-05, + "loss": 1.2432, + "step": 855 + }, + { + "epoch": 0.03, + "learning_rate": 1.82321618743344e-05, + "loss": 1.1797, + "step": 856 + }, + { + "epoch": 0.03, + "learning_rate": 1.8253461128860492e-05, + "loss": 1.187, + "step": 857 + }, + { + "epoch": 0.03, + "learning_rate": 1.8274760383386584e-05, + "loss": 0.4062, + "step": 858 + }, + { + "epoch": 0.03, + "learning_rate": 1.8296059637912676e-05, + "loss": 1.1426, + "step": 859 + }, + { + "epoch": 0.03, + "learning_rate": 1.8317358892438765e-05, + "loss": 0.9736, + "step": 860 + }, + { + "epoch": 0.03, + "learning_rate": 1.8338658146964858e-05, + "loss": 1.2368, + "step": 861 + }, + { + "epoch": 0.03, + "learning_rate": 1.835995740149095e-05, + "loss": 1.2324, + "step": 862 + }, + { + "epoch": 0.03, + "learning_rate": 1.8381256656017042e-05, + "loss": 1.1689, + "step": 863 + }, + { + "epoch": 0.03, + "learning_rate": 1.840255591054313e-05, + "loss": 1.1802, + "step": 864 + }, + { + "epoch": 0.03, + "learning_rate": 1.8423855165069224e-05, + "loss": 1.103, + "step": 865 + }, + { + "epoch": 0.03, + "learning_rate": 1.8445154419595316e-05, + "loss": 1.1836, + "step": 866 + }, + { + "epoch": 0.03, + "learning_rate": 1.8466453674121408e-05, + "loss": 1.2852, + "step": 867 + }, + { + "epoch": 0.03, + "learning_rate": 1.84877529286475e-05, + "loss": 1.042, + "step": 868 + }, + { + "epoch": 0.03, + "learning_rate": 1.850905218317359e-05, + "loss": 1.0688, + "step": 869 + }, + { + "epoch": 0.03, + "learning_rate": 1.8530351437699682e-05, + "loss": 1.186, + "step": 870 + }, + { + "epoch": 0.03, + "learning_rate": 1.8551650692225774e-05, + "loss": 1.1426, + "step": 871 + }, + { + "epoch": 0.03, + "learning_rate": 1.8572949946751866e-05, + "loss": 1.2524, + "step": 872 + }, + { + "epoch": 0.03, + "learning_rate": 1.8594249201277955e-05, + "loss": 1.1489, + "step": 873 + }, + { + "epoch": 0.03, + "learning_rate": 1.8615548455804048e-05, + "loss": 1.2046, + "step": 874 + }, + { + "epoch": 0.03, + "learning_rate": 1.863684771033014e-05, + "loss": 1.1729, + "step": 875 + }, + { + "epoch": 0.03, + "learning_rate": 1.8658146964856232e-05, + "loss": 1.1416, + "step": 876 + }, + { + "epoch": 0.03, + "learning_rate": 1.867944621938232e-05, + "loss": 1.0942, + "step": 877 + }, + { + "epoch": 0.03, + "learning_rate": 1.8700745473908414e-05, + "loss": 1.1719, + "step": 878 + }, + { + "epoch": 0.03, + "learning_rate": 1.8722044728434506e-05, + "loss": 1.2217, + "step": 879 + }, + { + "epoch": 0.03, + "learning_rate": 1.87433439829606e-05, + "loss": 1.0874, + "step": 880 + }, + { + "epoch": 0.03, + "learning_rate": 1.876464323748669e-05, + "loss": 1.0884, + "step": 881 + }, + { + "epoch": 0.03, + "learning_rate": 1.878594249201278e-05, + "loss": 1.1938, + "step": 882 + }, + { + "epoch": 0.03, + "learning_rate": 1.8807241746538872e-05, + "loss": 1.1455, + "step": 883 + }, + { + "epoch": 0.03, + "learning_rate": 1.8828541001064964e-05, + "loss": 1.1606, + "step": 884 + }, + { + "epoch": 0.03, + "learning_rate": 1.8849840255591057e-05, + "loss": 1.0723, + "step": 885 + }, + { + "epoch": 0.03, + "learning_rate": 1.8871139510117146e-05, + "loss": 1.2393, + "step": 886 + }, + { + "epoch": 0.03, + "learning_rate": 1.8892438764643238e-05, + "loss": 1.104, + "step": 887 + }, + { + "epoch": 0.03, + "learning_rate": 1.891373801916933e-05, + "loss": 1.1885, + "step": 888 + }, + { + "epoch": 0.03, + "learning_rate": 1.8935037273695422e-05, + "loss": 1.0996, + "step": 889 + }, + { + "epoch": 0.03, + "learning_rate": 1.895633652822151e-05, + "loss": 1.1421, + "step": 890 + }, + { + "epoch": 0.03, + "learning_rate": 1.8977635782747604e-05, + "loss": 1.2422, + "step": 891 + }, + { + "epoch": 0.03, + "learning_rate": 1.89989350372737e-05, + "loss": 1.1763, + "step": 892 + }, + { + "epoch": 0.03, + "learning_rate": 1.902023429179979e-05, + "loss": 1.103, + "step": 893 + }, + { + "epoch": 0.03, + "learning_rate": 1.904153354632588e-05, + "loss": 1.1094, + "step": 894 + }, + { + "epoch": 0.03, + "learning_rate": 1.9062832800851973e-05, + "loss": 1.1421, + "step": 895 + }, + { + "epoch": 0.03, + "learning_rate": 1.9084132055378065e-05, + "loss": 1.1372, + "step": 896 + }, + { + "epoch": 0.03, + "learning_rate": 1.9105431309904154e-05, + "loss": 1.2188, + "step": 897 + }, + { + "epoch": 0.03, + "learning_rate": 1.9126730564430247e-05, + "loss": 1.1245, + "step": 898 + }, + { + "epoch": 0.03, + "learning_rate": 1.914802981895634e-05, + "loss": 1.1313, + "step": 899 + }, + { + "epoch": 0.03, + "learning_rate": 1.916932907348243e-05, + "loss": 1.123, + "step": 900 + }, + { + "epoch": 0.03, + "learning_rate": 1.919062832800852e-05, + "loss": 1.1309, + "step": 901 + }, + { + "epoch": 0.03, + "learning_rate": 1.9211927582534613e-05, + "loss": 1.2002, + "step": 902 + }, + { + "epoch": 0.03, + "learning_rate": 1.9233226837060705e-05, + "loss": 1.1221, + "step": 903 + }, + { + "epoch": 0.03, + "learning_rate": 1.9254526091586797e-05, + "loss": 1.1543, + "step": 904 + }, + { + "epoch": 0.03, + "learning_rate": 1.927582534611289e-05, + "loss": 1.2129, + "step": 905 + }, + { + "epoch": 0.03, + "learning_rate": 1.929712460063898e-05, + "loss": 1.21, + "step": 906 + }, + { + "epoch": 0.03, + "learning_rate": 1.931842385516507e-05, + "loss": 1.3257, + "step": 907 + }, + { + "epoch": 0.03, + "learning_rate": 1.9339723109691163e-05, + "loss": 1.0508, + "step": 908 + }, + { + "epoch": 0.03, + "learning_rate": 1.9361022364217256e-05, + "loss": 1.0928, + "step": 909 + }, + { + "epoch": 0.03, + "learning_rate": 1.9382321618743344e-05, + "loss": 1.0854, + "step": 910 + }, + { + "epoch": 0.03, + "learning_rate": 1.9403620873269437e-05, + "loss": 1.1768, + "step": 911 + }, + { + "epoch": 0.03, + "learning_rate": 1.942492012779553e-05, + "loss": 0.3784, + "step": 912 + }, + { + "epoch": 0.03, + "learning_rate": 1.944621938232162e-05, + "loss": 1.1157, + "step": 913 + }, + { + "epoch": 0.03, + "learning_rate": 1.946751863684771e-05, + "loss": 1.0894, + "step": 914 + }, + { + "epoch": 0.03, + "learning_rate": 1.9488817891373803e-05, + "loss": 1.2695, + "step": 915 + }, + { + "epoch": 0.03, + "learning_rate": 1.9510117145899895e-05, + "loss": 1.1973, + "step": 916 + }, + { + "epoch": 0.03, + "learning_rate": 1.9531416400425987e-05, + "loss": 0.3718, + "step": 917 + }, + { + "epoch": 0.03, + "learning_rate": 1.955271565495208e-05, + "loss": 1.2744, + "step": 918 + }, + { + "epoch": 0.03, + "learning_rate": 1.957401490947817e-05, + "loss": 1.1035, + "step": 919 + }, + { + "epoch": 0.03, + "learning_rate": 1.959531416400426e-05, + "loss": 1.1772, + "step": 920 + }, + { + "epoch": 0.03, + "learning_rate": 1.9616613418530353e-05, + "loss": 1.125, + "step": 921 + }, + { + "epoch": 0.03, + "learning_rate": 1.9637912673056446e-05, + "loss": 1.1064, + "step": 922 + }, + { + "epoch": 0.03, + "learning_rate": 1.9659211927582535e-05, + "loss": 1.1118, + "step": 923 + }, + { + "epoch": 0.03, + "learning_rate": 1.9680511182108627e-05, + "loss": 1.1636, + "step": 924 + }, + { + "epoch": 0.03, + "learning_rate": 1.970181043663472e-05, + "loss": 1.146, + "step": 925 + }, + { + "epoch": 0.03, + "learning_rate": 1.972310969116081e-05, + "loss": 1.1147, + "step": 926 + }, + { + "epoch": 0.03, + "learning_rate": 1.97444089456869e-05, + "loss": 1.0747, + "step": 927 + }, + { + "epoch": 0.03, + "learning_rate": 1.9765708200212993e-05, + "loss": 1.2109, + "step": 928 + }, + { + "epoch": 0.03, + "learning_rate": 1.9787007454739085e-05, + "loss": 1.189, + "step": 929 + }, + { + "epoch": 0.03, + "learning_rate": 1.9808306709265177e-05, + "loss": 1.2354, + "step": 930 + }, + { + "epoch": 0.03, + "learning_rate": 1.982960596379127e-05, + "loss": 1.0215, + "step": 931 + }, + { + "epoch": 0.03, + "learning_rate": 1.985090521831736e-05, + "loss": 1.1147, + "step": 932 + }, + { + "epoch": 0.03, + "learning_rate": 1.987220447284345e-05, + "loss": 1.2144, + "step": 933 + }, + { + "epoch": 0.03, + "learning_rate": 1.9893503727369543e-05, + "loss": 1.1133, + "step": 934 + }, + { + "epoch": 0.03, + "learning_rate": 1.9914802981895636e-05, + "loss": 1.1406, + "step": 935 + }, + { + "epoch": 0.03, + "learning_rate": 1.9936102236421725e-05, + "loss": 1.1611, + "step": 936 + }, + { + "epoch": 0.03, + "learning_rate": 1.9957401490947817e-05, + "loss": 1.2275, + "step": 937 + }, + { + "epoch": 0.03, + "learning_rate": 1.997870074547391e-05, + "loss": 1.1167, + "step": 938 + }, + { + "epoch": 0.03, + "learning_rate": 2e-05, + "loss": 1.1587, + "step": 939 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999946355535e-05, + "loss": 1.0967, + "step": 940 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999785422138e-05, + "loss": 1.2061, + "step": 941 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999517199807e-05, + "loss": 1.1538, + "step": 942 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999141688553e-05, + "loss": 1.147, + "step": 943 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999998658888372e-05, + "loss": 1.1929, + "step": 944 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999998068799278e-05, + "loss": 1.063, + "step": 945 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999997371421267e-05, + "loss": 0.3838, + "step": 946 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999996566754354e-05, + "loss": 1.0654, + "step": 947 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999995654798547e-05, + "loss": 1.0576, + "step": 948 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999994635553847e-05, + "loss": 1.2104, + "step": 949 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999993509020278e-05, + "loss": 1.1489, + "step": 950 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999992275197847e-05, + "loss": 1.2051, + "step": 951 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999990934086563e-05, + "loss": 1.2451, + "step": 952 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999989485686444e-05, + "loss": 1.1646, + "step": 953 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999987929997506e-05, + "loss": 1.1621, + "step": 954 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999986267019767e-05, + "loss": 1.1274, + "step": 955 + }, + { + "epoch": 0.03, + "learning_rate": 1.999998449675324e-05, + "loss": 1.23, + "step": 956 + }, + { + "epoch": 0.03, + "learning_rate": 1.999998261919795e-05, + "loss": 1.2842, + "step": 957 + }, + { + "epoch": 0.03, + "learning_rate": 1.999998063435391e-05, + "loss": 1.2363, + "step": 958 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999978542221146e-05, + "loss": 1.2539, + "step": 959 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999976342799685e-05, + "loss": 1.1694, + "step": 960 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999974036089537e-05, + "loss": 1.1465, + "step": 961 + }, + { + "epoch": 0.03, + "learning_rate": 1.999997162209074e-05, + "loss": 1.1719, + "step": 962 + }, + { + "epoch": 0.03, + "learning_rate": 1.999996910080331e-05, + "loss": 1.0029, + "step": 963 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999966472227287e-05, + "loss": 1.1431, + "step": 964 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999963736362686e-05, + "loss": 1.2627, + "step": 965 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999960893209544e-05, + "loss": 1.1606, + "step": 966 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999957942767888e-05, + "loss": 1.124, + "step": 967 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999954885037746e-05, + "loss": 1.1821, + "step": 968 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999951720019162e-05, + "loss": 1.3389, + "step": 969 + }, + { + "epoch": 0.03, + "learning_rate": 1.999994844771216e-05, + "loss": 1.1265, + "step": 970 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999945068116778e-05, + "loss": 1.0767, + "step": 971 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999941581233057e-05, + "loss": 1.1455, + "step": 972 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999937987061028e-05, + "loss": 1.1411, + "step": 973 + }, + { + "epoch": 0.03, + "learning_rate": 1.999993428560073e-05, + "loss": 1.0596, + "step": 974 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999930476852207e-05, + "loss": 1.0923, + "step": 975 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999926560815496e-05, + "loss": 1.1187, + "step": 976 + }, + { + "epoch": 0.03, + "learning_rate": 1.999992253749064e-05, + "loss": 1.2363, + "step": 977 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999918406877684e-05, + "loss": 1.1992, + "step": 978 + }, + { + "epoch": 0.03, + "learning_rate": 1.999991416897667e-05, + "loss": 1.2354, + "step": 979 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999909823787642e-05, + "loss": 1.1875, + "step": 980 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999905371310653e-05, + "loss": 1.1289, + "step": 981 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999900811545744e-05, + "loss": 1.1108, + "step": 982 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999896144492966e-05, + "loss": 1.2441, + "step": 983 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999891370152375e-05, + "loss": 1.2041, + "step": 984 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999886488524013e-05, + "loss": 1.1191, + "step": 985 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999881499607936e-05, + "loss": 1.1592, + "step": 986 + }, + { + "epoch": 0.03, + "learning_rate": 1.99998764034042e-05, + "loss": 1.1577, + "step": 987 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999871199912856e-05, + "loss": 1.1123, + "step": 988 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999865889133957e-05, + "loss": 1.1143, + "step": 989 + }, + { + "epoch": 0.03, + "learning_rate": 1.999986047106757e-05, + "loss": 1.1177, + "step": 990 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999854945713748e-05, + "loss": 1.1597, + "step": 991 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999849313072544e-05, + "loss": 1.2256, + "step": 992 + }, + { + "epoch": 0.03, + "learning_rate": 1.999984357314403e-05, + "loss": 1.2368, + "step": 993 + }, + { + "epoch": 0.03, + "learning_rate": 1.999983772592826e-05, + "loss": 1.1797, + "step": 994 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999831771425298e-05, + "loss": 0.4487, + "step": 995 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999825709635207e-05, + "loss": 1.0371, + "step": 996 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999819540558056e-05, + "loss": 1.1694, + "step": 997 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999813264193906e-05, + "loss": 1.1406, + "step": 998 + }, + { + "epoch": 0.03, + "learning_rate": 1.999980688054283e-05, + "loss": 1.1509, + "step": 999 + }, + { + "epoch": 0.03, + "learning_rate": 1.999980038960489e-05, + "loss": 1.1021, + "step": 1000 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999793791380164e-05, + "loss": 1.0996, + "step": 1001 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999787085868713e-05, + "loss": 1.2705, + "step": 1002 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999780273070618e-05, + "loss": 1.062, + "step": 1003 + }, + { + "epoch": 0.03, + "learning_rate": 1.999977335298594e-05, + "loss": 1.2256, + "step": 1004 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999766325614767e-05, + "loss": 0.9707, + "step": 1005 + }, + { + "epoch": 0.03, + "learning_rate": 1.999975919095717e-05, + "loss": 1.1841, + "step": 1006 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999751949013224e-05, + "loss": 1.2949, + "step": 1007 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999744599783005e-05, + "loss": 1.1748, + "step": 1008 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999737143266592e-05, + "loss": 1.1738, + "step": 1009 + }, + { + "epoch": 0.03, + "learning_rate": 1.999972957946407e-05, + "loss": 1.0757, + "step": 1010 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999721908375512e-05, + "loss": 1.2402, + "step": 1011 + }, + { + "epoch": 0.03, + "learning_rate": 1.999971413000101e-05, + "loss": 1.1172, + "step": 1012 + }, + { + "epoch": 0.03, + "learning_rate": 1.999970624434064e-05, + "loss": 1.145, + "step": 1013 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999698251394492e-05, + "loss": 1.2539, + "step": 1014 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999690151162645e-05, + "loss": 1.125, + "step": 1015 + }, + { + "epoch": 0.03, + "learning_rate": 1.999968194364519e-05, + "loss": 1.1436, + "step": 1016 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999673628842214e-05, + "loss": 1.1353, + "step": 1017 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999665206753812e-05, + "loss": 1.2061, + "step": 1018 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999656677380065e-05, + "loss": 1.0405, + "step": 1019 + }, + { + "epoch": 0.03, + "learning_rate": 1.999964804072107e-05, + "loss": 1.1616, + "step": 1020 + }, + { + "epoch": 0.03, + "learning_rate": 1.999963929677692e-05, + "loss": 1.1162, + "step": 1021 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999630445547705e-05, + "loss": 1.1953, + "step": 1022 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999621487033524e-05, + "loss": 1.1689, + "step": 1023 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999612421234472e-05, + "loss": 1.1167, + "step": 1024 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999603248150644e-05, + "loss": 1.2227, + "step": 1025 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999593967782144e-05, + "loss": 1.1582, + "step": 1026 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999584580129064e-05, + "loss": 1.2109, + "step": 1027 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999575085191507e-05, + "loss": 1.1274, + "step": 1028 + }, + { + "epoch": 0.03, + "learning_rate": 1.999956548296958e-05, + "loss": 1.1152, + "step": 1029 + }, + { + "epoch": 0.03, + "learning_rate": 1.999955577346338e-05, + "loss": 1.1821, + "step": 1030 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999545956673018e-05, + "loss": 0.3579, + "step": 1031 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999536032598587e-05, + "loss": 1.1294, + "step": 1032 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999526001240207e-05, + "loss": 1.0415, + "step": 1033 + }, + { + "epoch": 0.03, + "learning_rate": 1.999951586259798e-05, + "loss": 1.2798, + "step": 1034 + }, + { + "epoch": 0.03, + "learning_rate": 1.999950561667201e-05, + "loss": 1.0146, + "step": 1035 + }, + { + "epoch": 0.03, + "learning_rate": 1.999949526346242e-05, + "loss": 1.1953, + "step": 1036 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999484802969304e-05, + "loss": 0.9624, + "step": 1037 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999474235192785e-05, + "loss": 1.2026, + "step": 1038 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999463560132977e-05, + "loss": 1.0972, + "step": 1039 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999452777789992e-05, + "loss": 1.0269, + "step": 1040 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999441888163945e-05, + "loss": 1.2139, + "step": 1041 + }, + { + "epoch": 0.03, + "learning_rate": 1.999943089125495e-05, + "loss": 1.0464, + "step": 1042 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999419787063132e-05, + "loss": 1.3208, + "step": 1043 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999408575588607e-05, + "loss": 1.2847, + "step": 1044 + }, + { + "epoch": 0.03, + "learning_rate": 1.999939725683149e-05, + "loss": 1.1567, + "step": 1045 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999385830791913e-05, + "loss": 1.2295, + "step": 1046 + }, + { + "epoch": 0.03, + "learning_rate": 1.999937429746999e-05, + "loss": 1.3857, + "step": 1047 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999362656865848e-05, + "loss": 1.1074, + "step": 1048 + }, + { + "epoch": 0.03, + "learning_rate": 1.999935090897961e-05, + "loss": 1.0474, + "step": 1049 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999339053811407e-05, + "loss": 1.2373, + "step": 1050 + }, + { + "epoch": 0.03, + "learning_rate": 1.999932709136136e-05, + "loss": 1.1948, + "step": 1051 + }, + { + "epoch": 0.03, + "learning_rate": 1.99993150216296e-05, + "loss": 1.106, + "step": 1052 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999302844616256e-05, + "loss": 1.333, + "step": 1053 + }, + { + "epoch": 0.03, + "learning_rate": 1.999929056032146e-05, + "loss": 1.1831, + "step": 1054 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999278168745345e-05, + "loss": 1.2476, + "step": 1055 + }, + { + "epoch": 0.03, + "learning_rate": 1.999926566988804e-05, + "loss": 1.168, + "step": 1056 + }, + { + "epoch": 0.03, + "learning_rate": 1.999925306374968e-05, + "loss": 1.2056, + "step": 1057 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999240350330404e-05, + "loss": 1.1865, + "step": 1058 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999227529630343e-05, + "loss": 1.1538, + "step": 1059 + }, + { + "epoch": 0.03, + "learning_rate": 1.999921460164964e-05, + "loss": 1.2085, + "step": 1060 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999201566388427e-05, + "loss": 1.0625, + "step": 1061 + }, + { + "epoch": 0.03, + "learning_rate": 1.999918842384685e-05, + "loss": 1.1689, + "step": 1062 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999175174025048e-05, + "loss": 1.1309, + "step": 1063 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999161816923162e-05, + "loss": 1.1323, + "step": 1064 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999148352541334e-05, + "loss": 1.1123, + "step": 1065 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999134780879714e-05, + "loss": 1.1748, + "step": 1066 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999121101938447e-05, + "loss": 1.2012, + "step": 1067 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999107315717672e-05, + "loss": 1.2354, + "step": 1068 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999093422217542e-05, + "loss": 1.0254, + "step": 1069 + }, + { + "epoch": 0.03, + "learning_rate": 1.999907942143821e-05, + "loss": 0.4512, + "step": 1070 + }, + { + "epoch": 0.03, + "learning_rate": 1.999906531337982e-05, + "loss": 1.0737, + "step": 1071 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999051098042524e-05, + "loss": 1.1108, + "step": 1072 + }, + { + "epoch": 0.03, + "learning_rate": 1.999903677542648e-05, + "loss": 1.1621, + "step": 1073 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999022345531834e-05, + "loss": 1.1221, + "step": 1074 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999007808358746e-05, + "loss": 1.2231, + "step": 1075 + }, + { + "epoch": 0.03, + "learning_rate": 1.999899316390737e-05, + "loss": 1.0767, + "step": 1076 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998978412177866e-05, + "loss": 1.106, + "step": 1077 + }, + { + "epoch": 0.03, + "learning_rate": 1.999896355317039e-05, + "loss": 1.1929, + "step": 1078 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998948586885098e-05, + "loss": 1.1543, + "step": 1079 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998933513322155e-05, + "loss": 1.106, + "step": 1080 + }, + { + "epoch": 0.03, + "learning_rate": 1.999891833248172e-05, + "loss": 0.9878, + "step": 1081 + }, + { + "epoch": 0.03, + "learning_rate": 1.999890304436396e-05, + "loss": 1.1953, + "step": 1082 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998887648969038e-05, + "loss": 1.124, + "step": 1083 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998872146297118e-05, + "loss": 1.2285, + "step": 1084 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998856536348363e-05, + "loss": 1.4229, + "step": 1085 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998840819122946e-05, + "loss": 1.1426, + "step": 1086 + }, + { + "epoch": 0.03, + "learning_rate": 1.999882499462103e-05, + "loss": 1.1094, + "step": 1087 + }, + { + "epoch": 0.03, + "learning_rate": 1.999880906284279e-05, + "loss": 1.1562, + "step": 1088 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998793023788395e-05, + "loss": 1.1992, + "step": 1089 + }, + { + "epoch": 0.03, + "learning_rate": 1.999877687745802e-05, + "loss": 1.2188, + "step": 1090 + }, + { + "epoch": 0.03, + "learning_rate": 1.999876062385183e-05, + "loss": 1.2637, + "step": 1091 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998744262970007e-05, + "loss": 1.1309, + "step": 1092 + }, + { + "epoch": 0.03, + "learning_rate": 1.999872779481272e-05, + "loss": 1.1436, + "step": 1093 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998711219380157e-05, + "loss": 1.127, + "step": 1094 + }, + { + "epoch": 0.04, + "learning_rate": 1.999869453667248e-05, + "loss": 0.9995, + "step": 1095 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998677746689886e-05, + "loss": 1.1035, + "step": 1096 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998660849432544e-05, + "loss": 1.1211, + "step": 1097 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998643844900633e-05, + "loss": 1.1187, + "step": 1098 + }, + { + "epoch": 0.04, + "learning_rate": 1.999862673309434e-05, + "loss": 1.1377, + "step": 1099 + }, + { + "epoch": 0.04, + "learning_rate": 1.999860951401385e-05, + "loss": 1.1274, + "step": 1100 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998592187659343e-05, + "loss": 1.1846, + "step": 1101 + }, + { + "epoch": 0.04, + "learning_rate": 1.999857475403101e-05, + "loss": 1.21, + "step": 1102 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998557213129035e-05, + "loss": 1.1797, + "step": 1103 + }, + { + "epoch": 0.04, + "learning_rate": 1.999853956495361e-05, + "loss": 1.2363, + "step": 1104 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998521809504918e-05, + "loss": 1.105, + "step": 1105 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998503946783155e-05, + "loss": 1.21, + "step": 1106 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998485976788506e-05, + "loss": 0.3887, + "step": 1107 + }, + { + "epoch": 0.04, + "learning_rate": 1.999846789952117e-05, + "loss": 1.1582, + "step": 1108 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998449714981343e-05, + "loss": 1.3188, + "step": 1109 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998431423169213e-05, + "loss": 1.2153, + "step": 1110 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998413024084982e-05, + "loss": 1.2393, + "step": 1111 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998394517728843e-05, + "loss": 1.0908, + "step": 1112 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998375904100996e-05, + "loss": 1.1357, + "step": 1113 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998357183201643e-05, + "loss": 1.2236, + "step": 1114 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998338355030982e-05, + "loss": 1.1987, + "step": 1115 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998319419589215e-05, + "loss": 1.207, + "step": 1116 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998300376876547e-05, + "loss": 1.3408, + "step": 1117 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998281226893183e-05, + "loss": 1.1987, + "step": 1118 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998261969639324e-05, + "loss": 1.2451, + "step": 1119 + }, + { + "epoch": 0.04, + "learning_rate": 1.999824260511518e-05, + "loss": 1.1567, + "step": 1120 + }, + { + "epoch": 0.04, + "learning_rate": 1.999822313332096e-05, + "loss": 1.144, + "step": 1121 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998203554256876e-05, + "loss": 1.1318, + "step": 1122 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998183867923126e-05, + "loss": 1.1704, + "step": 1123 + }, + { + "epoch": 0.04, + "learning_rate": 1.999816407431993e-05, + "loss": 1.1167, + "step": 1124 + }, + { + "epoch": 0.04, + "learning_rate": 1.99981441734475e-05, + "loss": 1.2046, + "step": 1125 + }, + { + "epoch": 0.04, + "learning_rate": 1.999812416530605e-05, + "loss": 1.1973, + "step": 1126 + }, + { + "epoch": 0.04, + "learning_rate": 1.999810404989579e-05, + "loss": 1.0903, + "step": 1127 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998083827216938e-05, + "loss": 1.207, + "step": 1128 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998063497269716e-05, + "loss": 1.2202, + "step": 1129 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998043060054336e-05, + "loss": 1.1367, + "step": 1130 + }, + { + "epoch": 0.04, + "learning_rate": 1.999802251557102e-05, + "loss": 1.1611, + "step": 1131 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998001863819985e-05, + "loss": 1.1709, + "step": 1132 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997981104801457e-05, + "loss": 1.1221, + "step": 1133 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997960238515656e-05, + "loss": 1.0918, + "step": 1134 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997939264962813e-05, + "loss": 1.125, + "step": 1135 + }, + { + "epoch": 0.04, + "learning_rate": 1.999791818414314e-05, + "loss": 1.2617, + "step": 1136 + }, + { + "epoch": 0.04, + "learning_rate": 1.999789699605687e-05, + "loss": 1.144, + "step": 1137 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997875700704233e-05, + "loss": 1.0923, + "step": 1138 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997854298085454e-05, + "loss": 1.1006, + "step": 1139 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997832788200764e-05, + "loss": 1.2124, + "step": 1140 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997811171050393e-05, + "loss": 1.146, + "step": 1141 + }, + { + "epoch": 0.04, + "learning_rate": 1.999778944663457e-05, + "loss": 1.2305, + "step": 1142 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997767614953536e-05, + "loss": 1.1338, + "step": 1143 + }, + { + "epoch": 0.04, + "learning_rate": 1.999774567600752e-05, + "loss": 1.3286, + "step": 1144 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997723629796755e-05, + "loss": 1.2051, + "step": 1145 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997701476321483e-05, + "loss": 1.2319, + "step": 1146 + }, + { + "epoch": 0.04, + "learning_rate": 1.999767921558194e-05, + "loss": 1.0801, + "step": 1147 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997656847578358e-05, + "loss": 1.1577, + "step": 1148 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997634372310987e-05, + "loss": 1.1558, + "step": 1149 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997611789780062e-05, + "loss": 1.0713, + "step": 1150 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997589099985832e-05, + "loss": 1.1123, + "step": 1151 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997566302928533e-05, + "loss": 1.2207, + "step": 1152 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997543398608413e-05, + "loss": 1.1548, + "step": 1153 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997520387025716e-05, + "loss": 1.1152, + "step": 1154 + }, + { + "epoch": 0.04, + "learning_rate": 1.999749726818069e-05, + "loss": 1.1406, + "step": 1155 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997474042073582e-05, + "loss": 1.1899, + "step": 1156 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997450708704643e-05, + "loss": 1.2119, + "step": 1157 + }, + { + "epoch": 0.04, + "learning_rate": 1.999742726807412e-05, + "loss": 1.1016, + "step": 1158 + }, + { + "epoch": 0.04, + "learning_rate": 1.999740372018227e-05, + "loss": 1.0874, + "step": 1159 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997380065029346e-05, + "loss": 1.1865, + "step": 1160 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997356302615594e-05, + "loss": 1.1465, + "step": 1161 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997332432941274e-05, + "loss": 1.1353, + "step": 1162 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997308456006643e-05, + "loss": 1.0684, + "step": 1163 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997284371811955e-05, + "loss": 1.0625, + "step": 1164 + }, + { + "epoch": 0.04, + "learning_rate": 1.999726018035747e-05, + "loss": 1.126, + "step": 1165 + }, + { + "epoch": 0.04, + "learning_rate": 1.999723588164345e-05, + "loss": 1.2251, + "step": 1166 + }, + { + "epoch": 0.04, + "learning_rate": 1.999721147567015e-05, + "loss": 1.0806, + "step": 1167 + }, + { + "epoch": 0.04, + "learning_rate": 1.999718696243784e-05, + "loss": 1.0713, + "step": 1168 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997162341946776e-05, + "loss": 1.0825, + "step": 1169 + }, + { + "epoch": 0.04, + "learning_rate": 1.999713761419722e-05, + "loss": 1.1558, + "step": 1170 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997112779189445e-05, + "loss": 1.1426, + "step": 1171 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997087836923717e-05, + "loss": 0.9888, + "step": 1172 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997062787400298e-05, + "loss": 1.21, + "step": 1173 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997037630619458e-05, + "loss": 1.105, + "step": 1174 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997012366581468e-05, + "loss": 1.1167, + "step": 1175 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996986995286603e-05, + "loss": 1.1455, + "step": 1176 + }, + { + "epoch": 0.04, + "learning_rate": 1.999696151673513e-05, + "loss": 1.2397, + "step": 1177 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996935930927326e-05, + "loss": 1.1514, + "step": 1178 + }, + { + "epoch": 0.04, + "learning_rate": 1.999691023786346e-05, + "loss": 1.1104, + "step": 1179 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996884437543815e-05, + "loss": 1.1626, + "step": 1180 + }, + { + "epoch": 0.04, + "learning_rate": 1.999685852996866e-05, + "loss": 1.0889, + "step": 1181 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996832515138278e-05, + "loss": 1.123, + "step": 1182 + }, + { + "epoch": 0.04, + "learning_rate": 1.999680639305295e-05, + "loss": 1.1797, + "step": 1183 + }, + { + "epoch": 0.04, + "learning_rate": 1.999678016371295e-05, + "loss": 1.1221, + "step": 1184 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996753827118565e-05, + "loss": 1.0898, + "step": 1185 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996727383270074e-05, + "loss": 1.1387, + "step": 1186 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996700832167764e-05, + "loss": 1.1841, + "step": 1187 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996674173811913e-05, + "loss": 1.1914, + "step": 1188 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996647408202815e-05, + "loss": 1.1548, + "step": 1189 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996620535340753e-05, + "loss": 1.2109, + "step": 1190 + }, + { + "epoch": 0.04, + "learning_rate": 1.999659355522602e-05, + "loss": 1.2246, + "step": 1191 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996566467858897e-05, + "loss": 1.1416, + "step": 1192 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996539273239684e-05, + "loss": 1.3193, + "step": 1193 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996511971368667e-05, + "loss": 1.2188, + "step": 1194 + }, + { + "epoch": 0.04, + "learning_rate": 1.999648456224614e-05, + "loss": 1.1274, + "step": 1195 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996457045872394e-05, + "loss": 1.189, + "step": 1196 + }, + { + "epoch": 0.04, + "learning_rate": 1.999642942224773e-05, + "loss": 1.2373, + "step": 1197 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996401691372442e-05, + "loss": 1.2524, + "step": 1198 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996373853246825e-05, + "loss": 1.1436, + "step": 1199 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996345907871184e-05, + "loss": 1.082, + "step": 1200 + }, + { + "epoch": 0.04, + "learning_rate": 1.999631785524581e-05, + "loss": 1.1807, + "step": 1201 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996289695371014e-05, + "loss": 1.1353, + "step": 1202 + }, + { + "epoch": 0.04, + "learning_rate": 1.999626142824709e-05, + "loss": 1.228, + "step": 1203 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996233053874342e-05, + "loss": 1.1792, + "step": 1204 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996204572253082e-05, + "loss": 0.9819, + "step": 1205 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996175983383607e-05, + "loss": 1.1392, + "step": 1206 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996147287266225e-05, + "loss": 1.2471, + "step": 1207 + }, + { + "epoch": 0.04, + "learning_rate": 1.999611848390125e-05, + "loss": 1.2065, + "step": 1208 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996089573288985e-05, + "loss": 1.0918, + "step": 1209 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996060555429743e-05, + "loss": 1.2124, + "step": 1210 + }, + { + "epoch": 0.04, + "learning_rate": 1.999603143032383e-05, + "loss": 1.1089, + "step": 1211 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996002197971568e-05, + "loss": 1.1094, + "step": 1212 + }, + { + "epoch": 0.04, + "learning_rate": 1.999597285837326e-05, + "loss": 1.1689, + "step": 1213 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995943411529233e-05, + "loss": 1.208, + "step": 1214 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995913857439792e-05, + "loss": 1.1338, + "step": 1215 + }, + { + "epoch": 0.04, + "learning_rate": 1.999588419610526e-05, + "loss": 1.3179, + "step": 1216 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995854427525955e-05, + "loss": 1.2617, + "step": 1217 + }, + { + "epoch": 0.04, + "learning_rate": 1.999582455170219e-05, + "loss": 1.1895, + "step": 1218 + }, + { + "epoch": 0.04, + "learning_rate": 1.999579456863429e-05, + "loss": 1.1543, + "step": 1219 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995764478322582e-05, + "loss": 1.166, + "step": 1220 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995734280767382e-05, + "loss": 1.1582, + "step": 1221 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995703975969018e-05, + "loss": 1.2095, + "step": 1222 + }, + { + "epoch": 0.04, + "learning_rate": 1.999567356392781e-05, + "loss": 0.4409, + "step": 1223 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995643044644088e-05, + "loss": 1.1528, + "step": 1224 + }, + { + "epoch": 0.04, + "learning_rate": 1.999561241811818e-05, + "loss": 1.1118, + "step": 1225 + }, + { + "epoch": 0.04, + "learning_rate": 1.999558168435041e-05, + "loss": 1.0718, + "step": 1226 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995550843341116e-05, + "loss": 1.0981, + "step": 1227 + }, + { + "epoch": 0.04, + "learning_rate": 1.999551989509062e-05, + "loss": 1.1035, + "step": 1228 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995488839599256e-05, + "loss": 1.1914, + "step": 1229 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995457676867363e-05, + "loss": 1.2593, + "step": 1230 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995426406895274e-05, + "loss": 1.0459, + "step": 1231 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995395029683316e-05, + "loss": 1.209, + "step": 1232 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995363545231832e-05, + "loss": 1.2432, + "step": 1233 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995331953541167e-05, + "loss": 1.2134, + "step": 1234 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995300254611644e-05, + "loss": 1.0972, + "step": 1235 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995268448443617e-05, + "loss": 1.21, + "step": 1236 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995236535037417e-05, + "loss": 1.1816, + "step": 1237 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995204514393397e-05, + "loss": 1.0903, + "step": 1238 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995172386511893e-05, + "loss": 1.2275, + "step": 1239 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995140151393252e-05, + "loss": 1.2705, + "step": 1240 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995107809037814e-05, + "loss": 1.2422, + "step": 1241 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995075359445935e-05, + "loss": 1.1548, + "step": 1242 + }, + { + "epoch": 0.04, + "learning_rate": 1.999504280261796e-05, + "loss": 1.0864, + "step": 1243 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995010138554236e-05, + "loss": 1.2578, + "step": 1244 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994977367255117e-05, + "loss": 1.2534, + "step": 1245 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994944488720952e-05, + "loss": 1.126, + "step": 1246 + }, + { + "epoch": 0.04, + "learning_rate": 1.999491150295209e-05, + "loss": 1.0625, + "step": 1247 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994878409948897e-05, + "loss": 1.188, + "step": 1248 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994845209711714e-05, + "loss": 1.1191, + "step": 1249 + }, + { + "epoch": 0.04, + "learning_rate": 1.999481190224091e-05, + "loss": 1.0576, + "step": 1250 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994778487536833e-05, + "loss": 1.2383, + "step": 1251 + }, + { + "epoch": 0.04, + "learning_rate": 1.999474496559984e-05, + "loss": 1.1172, + "step": 1252 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994711336430303e-05, + "loss": 1.2065, + "step": 1253 + }, + { + "epoch": 0.04, + "learning_rate": 1.999467760002857e-05, + "loss": 1.1313, + "step": 1254 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994643756395012e-05, + "loss": 1.0781, + "step": 1255 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994609805529984e-05, + "loss": 1.1206, + "step": 1256 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994575747433855e-05, + "loss": 1.1582, + "step": 1257 + }, + { + "epoch": 0.04, + "learning_rate": 1.999454158210699e-05, + "loss": 1.1265, + "step": 1258 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994507309549755e-05, + "loss": 1.2266, + "step": 1259 + }, + { + "epoch": 0.04, + "learning_rate": 1.999447292976252e-05, + "loss": 1.1509, + "step": 1260 + }, + { + "epoch": 0.04, + "learning_rate": 1.999443844274565e-05, + "loss": 1.1826, + "step": 1261 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994403848499515e-05, + "loss": 1.1147, + "step": 1262 + }, + { + "epoch": 0.04, + "learning_rate": 1.999436914702449e-05, + "loss": 1.1416, + "step": 1263 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994334338320947e-05, + "loss": 1.1411, + "step": 1264 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994299422389252e-05, + "loss": 1.0723, + "step": 1265 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994264399229794e-05, + "loss": 1.2114, + "step": 1266 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994229268842933e-05, + "loss": 0.4353, + "step": 1267 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994194031229054e-05, + "loss": 1.2651, + "step": 1268 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994158686388535e-05, + "loss": 1.2061, + "step": 1269 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994123234321756e-05, + "loss": 1.2207, + "step": 1270 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994087675029094e-05, + "loss": 1.1772, + "step": 1271 + }, + { + "epoch": 0.04, + "learning_rate": 1.999405200851093e-05, + "loss": 1.1396, + "step": 1272 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994016234767656e-05, + "loss": 1.3174, + "step": 1273 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993980353799642e-05, + "loss": 1.0815, + "step": 1274 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993944365607283e-05, + "loss": 1.084, + "step": 1275 + }, + { + "epoch": 0.04, + "learning_rate": 1.999390827019096e-05, + "loss": 0.9395, + "step": 1276 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993872067551062e-05, + "loss": 1.0947, + "step": 1277 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993835757687977e-05, + "loss": 1.2007, + "step": 1278 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993799340602094e-05, + "loss": 1.2017, + "step": 1279 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993762816293807e-05, + "loss": 1.1299, + "step": 1280 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993726184763506e-05, + "loss": 1.1523, + "step": 1281 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993689446011583e-05, + "loss": 1.1401, + "step": 1282 + }, + { + "epoch": 0.04, + "learning_rate": 1.999365260003843e-05, + "loss": 1.1689, + "step": 1283 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993615646844447e-05, + "loss": 1.1299, + "step": 1284 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993578586430028e-05, + "loss": 1.2241, + "step": 1285 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993541418795573e-05, + "loss": 1.1001, + "step": 1286 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993504143941478e-05, + "loss": 1.1323, + "step": 1287 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993466761868143e-05, + "loss": 1.186, + "step": 1288 + }, + { + "epoch": 0.04, + "learning_rate": 1.999342927257597e-05, + "loss": 1.0962, + "step": 1289 + }, + { + "epoch": 0.04, + "learning_rate": 1.999339167606536e-05, + "loss": 1.0864, + "step": 1290 + }, + { + "epoch": 0.04, + "learning_rate": 1.999335397233672e-05, + "loss": 1.2471, + "step": 1291 + }, + { + "epoch": 0.04, + "learning_rate": 1.999331616139045e-05, + "loss": 1.0024, + "step": 1292 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993278243226958e-05, + "loss": 1.2139, + "step": 1293 + }, + { + "epoch": 0.04, + "learning_rate": 1.999324021784665e-05, + "loss": 1.1392, + "step": 1294 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993202085249935e-05, + "loss": 0.425, + "step": 1295 + }, + { + "epoch": 0.04, + "learning_rate": 1.999316384543722e-05, + "loss": 1.1924, + "step": 1296 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993125498408923e-05, + "loss": 1.1294, + "step": 1297 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993087044165444e-05, + "loss": 1.1279, + "step": 1298 + }, + { + "epoch": 0.04, + "learning_rate": 1.99930484827072e-05, + "loss": 1.2231, + "step": 1299 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993009814034608e-05, + "loss": 1.0957, + "step": 1300 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992971038148077e-05, + "loss": 1.1128, + "step": 1301 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992932155048033e-05, + "loss": 1.2285, + "step": 1302 + }, + { + "epoch": 0.04, + "learning_rate": 1.999289316473488e-05, + "loss": 1.1567, + "step": 1303 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992854067209046e-05, + "loss": 1.3154, + "step": 1304 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992814862470947e-05, + "loss": 1.2744, + "step": 1305 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992775550521005e-05, + "loss": 1.1733, + "step": 1306 + }, + { + "epoch": 0.04, + "learning_rate": 1.999273613135964e-05, + "loss": 1.0142, + "step": 1307 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992696604987277e-05, + "loss": 1.0605, + "step": 1308 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992656971404335e-05, + "loss": 1.1948, + "step": 1309 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992617230611248e-05, + "loss": 1.207, + "step": 1310 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992577382608434e-05, + "loss": 1.2021, + "step": 1311 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992537427396325e-05, + "loss": 1.251, + "step": 1312 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992497364975347e-05, + "loss": 1.1064, + "step": 1313 + }, + { + "epoch": 0.04, + "learning_rate": 1.999245719534593e-05, + "loss": 1.252, + "step": 1314 + }, + { + "epoch": 0.04, + "learning_rate": 1.999241691850851e-05, + "loss": 1.1958, + "step": 1315 + }, + { + "epoch": 0.04, + "learning_rate": 1.999237653446352e-05, + "loss": 1.208, + "step": 1316 + }, + { + "epoch": 0.04, + "learning_rate": 1.999233604321138e-05, + "loss": 1.2031, + "step": 1317 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992295444752536e-05, + "loss": 1.1611, + "step": 1318 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992254739087423e-05, + "loss": 1.1777, + "step": 1319 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992213926216474e-05, + "loss": 1.0376, + "step": 1320 + }, + { + "epoch": 0.04, + "learning_rate": 1.999217300614013e-05, + "loss": 1.2158, + "step": 1321 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992131978858825e-05, + "loss": 1.1997, + "step": 1322 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992090844373003e-05, + "loss": 1.0767, + "step": 1323 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992049602683107e-05, + "loss": 1.1572, + "step": 1324 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992008253789577e-05, + "loss": 1.1318, + "step": 1325 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991966797692856e-05, + "loss": 1.1538, + "step": 1326 + }, + { + "epoch": 0.04, + "learning_rate": 1.999192523439339e-05, + "loss": 1.082, + "step": 1327 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991883563891626e-05, + "loss": 1.0552, + "step": 1328 + }, + { + "epoch": 0.04, + "learning_rate": 1.999184178618801e-05, + "loss": 1.1304, + "step": 1329 + }, + { + "epoch": 0.04, + "learning_rate": 1.999179990128299e-05, + "loss": 0.9785, + "step": 1330 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991757909177013e-05, + "loss": 1.0967, + "step": 1331 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991715809870533e-05, + "loss": 1.2217, + "step": 1332 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991673603364004e-05, + "loss": 0.4438, + "step": 1333 + }, + { + "epoch": 0.04, + "learning_rate": 1.999163128965787e-05, + "loss": 1.2637, + "step": 1334 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991588868752594e-05, + "loss": 1.1382, + "step": 1335 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991546340648627e-05, + "loss": 1.1289, + "step": 1336 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991503705346424e-05, + "loss": 1.1421, + "step": 1337 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991460962846448e-05, + "loss": 1.1714, + "step": 1338 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991418113149154e-05, + "loss": 1.1133, + "step": 1339 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991375156254998e-05, + "loss": 1.1489, + "step": 1340 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991332092164442e-05, + "loss": 1.1743, + "step": 1341 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991288920877957e-05, + "loss": 1.2246, + "step": 1342 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991245642395997e-05, + "loss": 1.2861, + "step": 1343 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991202256719032e-05, + "loss": 1.1182, + "step": 1344 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991158763847516e-05, + "loss": 1.1914, + "step": 1345 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991115163781935e-05, + "loss": 1.1069, + "step": 1346 + }, + { + "epoch": 0.04, + "learning_rate": 1.999107145652274e-05, + "loss": 1.2251, + "step": 1347 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991027642070408e-05, + "loss": 1.1245, + "step": 1348 + }, + { + "epoch": 0.04, + "learning_rate": 1.99909837204254e-05, + "loss": 1.1885, + "step": 1349 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990939691588204e-05, + "loss": 1.145, + "step": 1350 + }, + { + "epoch": 0.04, + "learning_rate": 1.999089555555928e-05, + "loss": 1.0889, + "step": 1351 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990851312339098e-05, + "loss": 1.1787, + "step": 1352 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990806961928143e-05, + "loss": 1.1963, + "step": 1353 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990762504326887e-05, + "loss": 1.1025, + "step": 1354 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990717939535807e-05, + "loss": 1.1265, + "step": 1355 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990673267555378e-05, + "loss": 1.3022, + "step": 1356 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990628488386083e-05, + "loss": 1.0913, + "step": 1357 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990583602028405e-05, + "loss": 1.1104, + "step": 1358 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990538608482816e-05, + "loss": 1.1636, + "step": 1359 + }, + { + "epoch": 0.04, + "learning_rate": 1.999049350774981e-05, + "loss": 1.1377, + "step": 1360 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990448299829862e-05, + "loss": 1.1313, + "step": 1361 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990402984723462e-05, + "loss": 1.167, + "step": 1362 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990357562431097e-05, + "loss": 1.1162, + "step": 1363 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990312032953248e-05, + "loss": 1.0962, + "step": 1364 + }, + { + "epoch": 0.04, + "learning_rate": 1.999026639629041e-05, + "loss": 1.0801, + "step": 1365 + }, + { + "epoch": 0.04, + "learning_rate": 1.999022065244307e-05, + "loss": 1.1333, + "step": 1366 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990174801411718e-05, + "loss": 1.2607, + "step": 1367 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990128843196847e-05, + "loss": 1.022, + "step": 1368 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990082777798952e-05, + "loss": 1.1182, + "step": 1369 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990036605218524e-05, + "loss": 1.168, + "step": 1370 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989990325456058e-05, + "loss": 1.0786, + "step": 1371 + }, + { + "epoch": 0.04, + "learning_rate": 1.998994393851206e-05, + "loss": 1.2021, + "step": 1372 + }, + { + "epoch": 0.04, + "learning_rate": 1.998989744438701e-05, + "loss": 1.0337, + "step": 1373 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989850843081423e-05, + "loss": 1.1006, + "step": 1374 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989804134595786e-05, + "loss": 1.1855, + "step": 1375 + }, + { + "epoch": 0.04, + "learning_rate": 1.998975731893061e-05, + "loss": 1.0684, + "step": 1376 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989710396086396e-05, + "loss": 1.1733, + "step": 1377 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989663366063647e-05, + "loss": 1.1533, + "step": 1378 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989616228862864e-05, + "loss": 1.2988, + "step": 1379 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989568984484556e-05, + "loss": 1.1953, + "step": 1380 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989521632929225e-05, + "loss": 1.1562, + "step": 1381 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989474174197385e-05, + "loss": 1.0977, + "step": 1382 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989426608289545e-05, + "loss": 1.1001, + "step": 1383 + }, + { + "epoch": 0.04, + "learning_rate": 1.998937893520621e-05, + "loss": 1.1479, + "step": 1384 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989331154947896e-05, + "loss": 1.2617, + "step": 1385 + }, + { + "epoch": 0.04, + "learning_rate": 1.998928326751511e-05, + "loss": 1.1719, + "step": 1386 + }, + { + "epoch": 0.04, + "learning_rate": 1.998923527290838e-05, + "loss": 1.1924, + "step": 1387 + }, + { + "epoch": 0.04, + "learning_rate": 1.99891871711282e-05, + "loss": 1.1079, + "step": 1388 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989138962175105e-05, + "loss": 1.2803, + "step": 1389 + }, + { + "epoch": 0.04, + "learning_rate": 1.99890906460496e-05, + "loss": 1.2173, + "step": 1390 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989042222752212e-05, + "loss": 1.0918, + "step": 1391 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988993692283455e-05, + "loss": 1.1104, + "step": 1392 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988945054643848e-05, + "loss": 1.2627, + "step": 1393 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988896309833915e-05, + "loss": 1.3018, + "step": 1394 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988847457854182e-05, + "loss": 1.1899, + "step": 1395 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988798498705175e-05, + "loss": 1.2749, + "step": 1396 + }, + { + "epoch": 0.04, + "learning_rate": 1.998874943238741e-05, + "loss": 1.3086, + "step": 1397 + }, + { + "epoch": 0.04, + "learning_rate": 1.998870025890142e-05, + "loss": 1.2036, + "step": 1398 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988650978247733e-05, + "loss": 1.2627, + "step": 1399 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988601590426874e-05, + "loss": 1.1543, + "step": 1400 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988552095439372e-05, + "loss": 1.1792, + "step": 1401 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988502493285767e-05, + "loss": 1.1587, + "step": 1402 + }, + { + "epoch": 0.04, + "learning_rate": 1.998845278396658e-05, + "loss": 1.1157, + "step": 1403 + }, + { + "epoch": 0.04, + "learning_rate": 1.998840296748235e-05, + "loss": 1.1289, + "step": 1404 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988353043833614e-05, + "loss": 1.2373, + "step": 1405 + }, + { + "epoch": 0.04, + "learning_rate": 1.99883030130209e-05, + "loss": 1.0664, + "step": 1406 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988252875044752e-05, + "loss": 1.1089, + "step": 1407 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988202629905706e-05, + "loss": 1.1685, + "step": 1408 + }, + { + "epoch": 0.05, + "learning_rate": 1.99881522776043e-05, + "loss": 1.1694, + "step": 1409 + }, + { + "epoch": 0.05, + "learning_rate": 1.998810181814107e-05, + "loss": 1.1016, + "step": 1410 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988051251516566e-05, + "loss": 1.2197, + "step": 1411 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988000577731323e-05, + "loss": 1.166, + "step": 1412 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987949796785892e-05, + "loss": 1.0811, + "step": 1413 + }, + { + "epoch": 0.05, + "learning_rate": 1.998789890868081e-05, + "loss": 0.407, + "step": 1414 + }, + { + "epoch": 0.05, + "learning_rate": 1.998784791341663e-05, + "loss": 1.1401, + "step": 1415 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987796810993892e-05, + "loss": 1.0552, + "step": 1416 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987745601413154e-05, + "loss": 1.2471, + "step": 1417 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987694284674955e-05, + "loss": 1.2217, + "step": 1418 + }, + { + "epoch": 0.05, + "learning_rate": 1.998764286077985e-05, + "loss": 1.0547, + "step": 1419 + }, + { + "epoch": 0.05, + "learning_rate": 1.998759132972839e-05, + "loss": 1.1675, + "step": 1420 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987539691521135e-05, + "loss": 1.062, + "step": 1421 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987487946158627e-05, + "loss": 1.1553, + "step": 1422 + }, + { + "epoch": 0.05, + "learning_rate": 1.998743609364143e-05, + "loss": 1.1582, + "step": 1423 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987384133970096e-05, + "loss": 1.2109, + "step": 1424 + }, + { + "epoch": 0.05, + "learning_rate": 1.998733206714518e-05, + "loss": 1.1724, + "step": 1425 + }, + { + "epoch": 0.05, + "learning_rate": 1.998727989316725e-05, + "loss": 1.0605, + "step": 1426 + }, + { + "epoch": 0.05, + "learning_rate": 1.998722761203686e-05, + "loss": 1.146, + "step": 1427 + }, + { + "epoch": 0.05, + "learning_rate": 1.998717522375457e-05, + "loss": 1.0908, + "step": 1428 + }, + { + "epoch": 0.05, + "learning_rate": 1.998712272832094e-05, + "loss": 1.2148, + "step": 1429 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987070125736537e-05, + "loss": 1.1304, + "step": 1430 + }, + { + "epoch": 0.05, + "learning_rate": 1.998701741600193e-05, + "loss": 1.2104, + "step": 1431 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986964599117674e-05, + "loss": 1.1772, + "step": 1432 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986911675084344e-05, + "loss": 1.0801, + "step": 1433 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986858643902502e-05, + "loss": 1.23, + "step": 1434 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986805505572723e-05, + "loss": 1.1089, + "step": 1435 + }, + { + "epoch": 0.05, + "learning_rate": 1.998675226009557e-05, + "loss": 1.1221, + "step": 1436 + }, + { + "epoch": 0.05, + "learning_rate": 1.998669890747162e-05, + "loss": 1.1426, + "step": 1437 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986645447701442e-05, + "loss": 1.2188, + "step": 1438 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986591880785612e-05, + "loss": 1.0679, + "step": 1439 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986538206724703e-05, + "loss": 1.1382, + "step": 1440 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986484425519292e-05, + "loss": 1.1069, + "step": 1441 + }, + { + "epoch": 0.05, + "learning_rate": 1.998643053716996e-05, + "loss": 1.1572, + "step": 1442 + }, + { + "epoch": 0.05, + "learning_rate": 1.998637654167727e-05, + "loss": 1.1699, + "step": 1443 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986322439041824e-05, + "loss": 1.1221, + "step": 1444 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986268229264186e-05, + "loss": 1.2129, + "step": 1445 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986213912344938e-05, + "loss": 1.1724, + "step": 1446 + }, + { + "epoch": 0.05, + "learning_rate": 1.998615948828467e-05, + "loss": 1.1455, + "step": 1447 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986104957083966e-05, + "loss": 1.1035, + "step": 1448 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986050318743406e-05, + "loss": 1.04, + "step": 1449 + }, + { + "epoch": 0.05, + "learning_rate": 1.998599557326358e-05, + "loss": 1.1768, + "step": 1450 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985940720645068e-05, + "loss": 1.0371, + "step": 1451 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985885760888463e-05, + "loss": 1.084, + "step": 1452 + }, + { + "epoch": 0.05, + "learning_rate": 1.998583069399436e-05, + "loss": 1.1562, + "step": 1453 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985775519963347e-05, + "loss": 1.1836, + "step": 1454 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985720238796012e-05, + "loss": 1.0483, + "step": 1455 + }, + { + "epoch": 0.05, + "learning_rate": 1.998566485049295e-05, + "loss": 1.0762, + "step": 1456 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985609355054756e-05, + "loss": 1.2046, + "step": 1457 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985553752482026e-05, + "loss": 1.1562, + "step": 1458 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985498042775355e-05, + "loss": 1.0171, + "step": 1459 + }, + { + "epoch": 0.05, + "learning_rate": 1.998544222593534e-05, + "loss": 1.1953, + "step": 1460 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985386301962585e-05, + "loss": 1.02, + "step": 1461 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985330270857682e-05, + "loss": 1.2178, + "step": 1462 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985274132621238e-05, + "loss": 1.1758, + "step": 1463 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985217887253856e-05, + "loss": 1.209, + "step": 1464 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985161534756135e-05, + "loss": 1.1333, + "step": 1465 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985105075128685e-05, + "loss": 1.0771, + "step": 1466 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985048508372103e-05, + "loss": 1.1416, + "step": 1467 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984991834487005e-05, + "loss": 1.1416, + "step": 1468 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984935053473994e-05, + "loss": 1.0635, + "step": 1469 + }, + { + "epoch": 0.05, + "learning_rate": 1.998487816533368e-05, + "loss": 1.1206, + "step": 1470 + }, + { + "epoch": 0.05, + "learning_rate": 1.998482117006668e-05, + "loss": 1.1548, + "step": 1471 + }, + { + "epoch": 0.05, + "learning_rate": 1.998476406767359e-05, + "loss": 1.0732, + "step": 1472 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984706858155037e-05, + "loss": 1.1172, + "step": 1473 + }, + { + "epoch": 0.05, + "learning_rate": 1.998464954151163e-05, + "loss": 1.1221, + "step": 1474 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984592117743983e-05, + "loss": 1.1094, + "step": 1475 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984534586852715e-05, + "loss": 1.2095, + "step": 1476 + }, + { + "epoch": 0.05, + "learning_rate": 1.998447694883844e-05, + "loss": 1.2793, + "step": 1477 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984419203701778e-05, + "loss": 1.1978, + "step": 1478 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984361351443343e-05, + "loss": 1.1519, + "step": 1479 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984303392063768e-05, + "loss": 1.2988, + "step": 1480 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984245325563663e-05, + "loss": 1.1753, + "step": 1481 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984187151943662e-05, + "loss": 1.1294, + "step": 1482 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984128871204378e-05, + "loss": 1.2158, + "step": 1483 + }, + { + "epoch": 0.05, + "learning_rate": 1.998407048334644e-05, + "loss": 1.2412, + "step": 1484 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984011988370478e-05, + "loss": 1.1055, + "step": 1485 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983953386277116e-05, + "loss": 1.2666, + "step": 1486 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983894677066984e-05, + "loss": 1.1357, + "step": 1487 + }, + { + "epoch": 0.05, + "learning_rate": 1.998383586074071e-05, + "loss": 1.1821, + "step": 1488 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983776937298932e-05, + "loss": 1.1416, + "step": 1489 + }, + { + "epoch": 0.05, + "learning_rate": 1.998371790674227e-05, + "loss": 1.1782, + "step": 1490 + }, + { + "epoch": 0.05, + "learning_rate": 1.998365876907137e-05, + "loss": 1.1265, + "step": 1491 + }, + { + "epoch": 0.05, + "learning_rate": 1.998359952428686e-05, + "loss": 1.0859, + "step": 1492 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983540172389374e-05, + "loss": 1.1895, + "step": 1493 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983480713379552e-05, + "loss": 1.1357, + "step": 1494 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983421147258033e-05, + "loss": 1.1646, + "step": 1495 + }, + { + "epoch": 0.05, + "learning_rate": 1.998336147402545e-05, + "loss": 1.1582, + "step": 1496 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983301693682452e-05, + "loss": 1.1543, + "step": 1497 + }, + { + "epoch": 0.05, + "learning_rate": 1.998324180622967e-05, + "loss": 1.1211, + "step": 1498 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983181811667756e-05, + "loss": 1.1035, + "step": 1499 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983121709997354e-05, + "loss": 1.189, + "step": 1500 + }, + { + "epoch": 0.05, + "learning_rate": 1.99830615012191e-05, + "loss": 1.0083, + "step": 1501 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983001185333646e-05, + "loss": 1.1572, + "step": 1502 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982940762341637e-05, + "loss": 1.1265, + "step": 1503 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982880232243725e-05, + "loss": 1.0518, + "step": 1504 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982819595040554e-05, + "loss": 1.2461, + "step": 1505 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982758850732777e-05, + "loss": 1.0732, + "step": 1506 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982697999321048e-05, + "loss": 1.2271, + "step": 1507 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982637040806018e-05, + "loss": 1.1821, + "step": 1508 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982575975188337e-05, + "loss": 1.1274, + "step": 1509 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982514802468666e-05, + "loss": 1.2979, + "step": 1510 + }, + { + "epoch": 0.05, + "learning_rate": 1.998245352264766e-05, + "loss": 1.103, + "step": 1511 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982392135725973e-05, + "loss": 1.1191, + "step": 1512 + }, + { + "epoch": 0.05, + "learning_rate": 1.998233064170427e-05, + "loss": 1.1602, + "step": 1513 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982269040583207e-05, + "loss": 1.1479, + "step": 1514 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982207332363442e-05, + "loss": 1.0723, + "step": 1515 + }, + { + "epoch": 0.05, + "learning_rate": 1.998214551704564e-05, + "loss": 1.0747, + "step": 1516 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982083594630464e-05, + "loss": 1.1489, + "step": 1517 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982021565118583e-05, + "loss": 1.1611, + "step": 1518 + }, + { + "epoch": 0.05, + "learning_rate": 1.998195942851066e-05, + "loss": 1.167, + "step": 1519 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981897184807354e-05, + "loss": 1.2568, + "step": 1520 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981834834009337e-05, + "loss": 1.1406, + "step": 1521 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981772376117285e-05, + "loss": 1.126, + "step": 1522 + }, + { + "epoch": 0.05, + "learning_rate": 1.998170981113186e-05, + "loss": 1.0635, + "step": 1523 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981647139053737e-05, + "loss": 1.1846, + "step": 1524 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981584359883586e-05, + "loss": 1.2012, + "step": 1525 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981521473622083e-05, + "loss": 1.0747, + "step": 1526 + }, + { + "epoch": 0.05, + "learning_rate": 1.99814584802699e-05, + "loss": 1.1216, + "step": 1527 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981395379827715e-05, + "loss": 1.0747, + "step": 1528 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981332172296207e-05, + "loss": 1.103, + "step": 1529 + }, + { + "epoch": 0.05, + "learning_rate": 1.998126885767605e-05, + "loss": 0.4695, + "step": 1530 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981205435967924e-05, + "loss": 1.2549, + "step": 1531 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981141907172513e-05, + "loss": 1.0947, + "step": 1532 + }, + { + "epoch": 0.05, + "learning_rate": 1.998107827129049e-05, + "loss": 1.1689, + "step": 1533 + }, + { + "epoch": 0.05, + "learning_rate": 1.998101452832255e-05, + "loss": 1.21, + "step": 1534 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980950678269368e-05, + "loss": 1.1348, + "step": 1535 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980886721131632e-05, + "loss": 1.0625, + "step": 1536 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980822656910026e-05, + "loss": 1.062, + "step": 1537 + }, + { + "epoch": 0.05, + "learning_rate": 1.998075848560524e-05, + "loss": 1.2056, + "step": 1538 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980694207217964e-05, + "loss": 1.2158, + "step": 1539 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980629821748884e-05, + "loss": 1.1704, + "step": 1540 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980565329198688e-05, + "loss": 1.1655, + "step": 1541 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980500729568074e-05, + "loss": 1.1924, + "step": 1542 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980436022857738e-05, + "loss": 1.2739, + "step": 1543 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980371209068365e-05, + "loss": 1.0503, + "step": 1544 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980306288200655e-05, + "loss": 1.1045, + "step": 1545 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980241260255306e-05, + "loss": 1.0801, + "step": 1546 + }, + { + "epoch": 0.05, + "learning_rate": 1.998017612523301e-05, + "loss": 1.4316, + "step": 1547 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980110883134473e-05, + "loss": 1.1587, + "step": 1548 + }, + { + "epoch": 0.05, + "learning_rate": 1.998004553396039e-05, + "loss": 1.0249, + "step": 1549 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979980077711465e-05, + "loss": 1.1763, + "step": 1550 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979914514388397e-05, + "loss": 1.0386, + "step": 1551 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979848843991895e-05, + "loss": 1.0737, + "step": 1552 + }, + { + "epoch": 0.05, + "learning_rate": 1.997978306652266e-05, + "loss": 1.1831, + "step": 1553 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979717181981394e-05, + "loss": 1.2192, + "step": 1554 + }, + { + "epoch": 0.05, + "learning_rate": 1.997965119036881e-05, + "loss": 1.0586, + "step": 1555 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979585091685614e-05, + "loss": 0.9717, + "step": 1556 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979518885932512e-05, + "loss": 1.0156, + "step": 1557 + }, + { + "epoch": 0.05, + "learning_rate": 1.997945257311022e-05, + "loss": 1.0854, + "step": 1558 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979386153219446e-05, + "loss": 1.2046, + "step": 1559 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979319626260903e-05, + "loss": 1.0962, + "step": 1560 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979252992235305e-05, + "loss": 1.1475, + "step": 1561 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979186251143367e-05, + "loss": 1.1177, + "step": 1562 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979119402985803e-05, + "loss": 1.1011, + "step": 1563 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979052447763336e-05, + "loss": 1.3447, + "step": 1564 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978985385476677e-05, + "loss": 1.1353, + "step": 1565 + }, + { + "epoch": 0.05, + "learning_rate": 1.997891821612655e-05, + "loss": 1.1685, + "step": 1566 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978850939713676e-05, + "loss": 1.1997, + "step": 1567 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978783556238774e-05, + "loss": 1.1484, + "step": 1568 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978716065702566e-05, + "loss": 1.209, + "step": 1569 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978648468105786e-05, + "loss": 1.1172, + "step": 1570 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978580763449146e-05, + "loss": 1.2266, + "step": 1571 + }, + { + "epoch": 0.05, + "learning_rate": 1.997851295173338e-05, + "loss": 1.2261, + "step": 1572 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978445032959212e-05, + "loss": 1.1372, + "step": 1573 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978377007127374e-05, + "loss": 1.0317, + "step": 1574 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978308874238595e-05, + "loss": 1.0952, + "step": 1575 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978240634293603e-05, + "loss": 1.1597, + "step": 1576 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978172287293133e-05, + "loss": 1.2178, + "step": 1577 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978103833237917e-05, + "loss": 1.1143, + "step": 1578 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978035272128694e-05, + "loss": 1.0391, + "step": 1579 + }, + { + "epoch": 0.05, + "learning_rate": 1.997796660396619e-05, + "loss": 1.2139, + "step": 1580 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977897828751153e-05, + "loss": 1.0396, + "step": 1581 + }, + { + "epoch": 0.05, + "learning_rate": 1.997782894648431e-05, + "loss": 1.0806, + "step": 1582 + }, + { + "epoch": 0.05, + "learning_rate": 1.997775995716641e-05, + "loss": 1.0776, + "step": 1583 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977690860798184e-05, + "loss": 1.187, + "step": 1584 + }, + { + "epoch": 0.05, + "learning_rate": 1.997762165738038e-05, + "loss": 1.1089, + "step": 1585 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977552346913736e-05, + "loss": 1.0586, + "step": 1586 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977482929399e-05, + "loss": 1.0688, + "step": 1587 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977413404836915e-05, + "loss": 1.1543, + "step": 1588 + }, + { + "epoch": 0.05, + "learning_rate": 1.997734377322823e-05, + "loss": 1.147, + "step": 1589 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977274034573684e-05, + "loss": 1.1021, + "step": 1590 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977204188874032e-05, + "loss": 1.2715, + "step": 1591 + }, + { + "epoch": 0.05, + "learning_rate": 1.997713423613002e-05, + "loss": 1.1865, + "step": 1592 + }, + { + "epoch": 0.05, + "learning_rate": 1.99770641763424e-05, + "loss": 1.1343, + "step": 1593 + }, + { + "epoch": 0.05, + "learning_rate": 1.997699400951192e-05, + "loss": 1.1504, + "step": 1594 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976923735639344e-05, + "loss": 1.1587, + "step": 1595 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976853354725414e-05, + "loss": 1.2139, + "step": 1596 + }, + { + "epoch": 0.05, + "learning_rate": 1.997678286677089e-05, + "loss": 1.0366, + "step": 1597 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976712271776526e-05, + "loss": 1.1875, + "step": 1598 + }, + { + "epoch": 0.05, + "learning_rate": 1.997664156974308e-05, + "loss": 1.1201, + "step": 1599 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976570760671317e-05, + "loss": 1.0557, + "step": 1600 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976499844561988e-05, + "loss": 1.0459, + "step": 1601 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976428821415856e-05, + "loss": 1.2134, + "step": 1602 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976357691233684e-05, + "loss": 1.0659, + "step": 1603 + }, + { + "epoch": 0.05, + "learning_rate": 1.997628645401624e-05, + "loss": 1.1348, + "step": 1604 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976215109764277e-05, + "loss": 1.1807, + "step": 1605 + }, + { + "epoch": 0.05, + "learning_rate": 1.997614365847857e-05, + "loss": 1.105, + "step": 1606 + }, + { + "epoch": 0.05, + "learning_rate": 1.997607210015988e-05, + "loss": 1.2007, + "step": 1607 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976000434808985e-05, + "loss": 1.1294, + "step": 1608 + }, + { + "epoch": 0.05, + "learning_rate": 1.997592866242664e-05, + "loss": 1.0376, + "step": 1609 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975856783013624e-05, + "loss": 1.2393, + "step": 1610 + }, + { + "epoch": 0.05, + "learning_rate": 1.99757847965707e-05, + "loss": 1.0654, + "step": 1611 + }, + { + "epoch": 0.05, + "learning_rate": 1.997571270309865e-05, + "loss": 1.0488, + "step": 1612 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975640502598243e-05, + "loss": 1.2017, + "step": 1613 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975568195070253e-05, + "loss": 0.9634, + "step": 1614 + }, + { + "epoch": 0.05, + "learning_rate": 1.997549578051546e-05, + "loss": 1.2373, + "step": 1615 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975423258934633e-05, + "loss": 1.0356, + "step": 1616 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975350630328558e-05, + "loss": 1.0933, + "step": 1617 + }, + { + "epoch": 0.05, + "learning_rate": 1.997527789469801e-05, + "loss": 1.1084, + "step": 1618 + }, + { + "epoch": 0.05, + "learning_rate": 1.997520505204377e-05, + "loss": 1.25, + "step": 1619 + }, + { + "epoch": 0.05, + "learning_rate": 1.997513210236662e-05, + "loss": 0.4346, + "step": 1620 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975059045667343e-05, + "loss": 1.1904, + "step": 1621 + }, + { + "epoch": 0.05, + "learning_rate": 1.997498588194672e-05, + "loss": 1.1011, + "step": 1622 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974912611205538e-05, + "loss": 1.1245, + "step": 1623 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974839233444588e-05, + "loss": 1.1167, + "step": 1624 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974765748664648e-05, + "loss": 1.1323, + "step": 1625 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974692156866516e-05, + "loss": 1.1772, + "step": 1626 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974618458050972e-05, + "loss": 1.0986, + "step": 1627 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974544652218812e-05, + "loss": 1.0586, + "step": 1628 + }, + { + "epoch": 0.05, + "learning_rate": 1.997447073937083e-05, + "loss": 1.0581, + "step": 1629 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974396719507813e-05, + "loss": 1.0977, + "step": 1630 + }, + { + "epoch": 0.05, + "learning_rate": 1.997432259263056e-05, + "loss": 1.1479, + "step": 1631 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974248358739863e-05, + "loss": 1.1509, + "step": 1632 + }, + { + "epoch": 0.05, + "learning_rate": 1.997417401783652e-05, + "loss": 0.4248, + "step": 1633 + }, + { + "epoch": 0.05, + "learning_rate": 1.997409956992133e-05, + "loss": 1.0957, + "step": 1634 + }, + { + "epoch": 0.05, + "learning_rate": 1.997402501499509e-05, + "loss": 1.0918, + "step": 1635 + }, + { + "epoch": 0.05, + "learning_rate": 1.99739503530586e-05, + "loss": 1.1885, + "step": 1636 + }, + { + "epoch": 0.05, + "learning_rate": 1.997387558411266e-05, + "loss": 1.1426, + "step": 1637 + }, + { + "epoch": 0.05, + "learning_rate": 1.997380070815808e-05, + "loss": 1.123, + "step": 1638 + }, + { + "epoch": 0.05, + "learning_rate": 1.997372572519565e-05, + "loss": 1.1431, + "step": 1639 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973650635226184e-05, + "loss": 1.1211, + "step": 1640 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973575438250483e-05, + "loss": 1.0664, + "step": 1641 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973500134269357e-05, + "loss": 1.0933, + "step": 1642 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973424723283612e-05, + "loss": 1.0635, + "step": 1643 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973349205294062e-05, + "loss": 1.0898, + "step": 1644 + }, + { + "epoch": 0.05, + "learning_rate": 1.997327358030151e-05, + "loss": 1.0562, + "step": 1645 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973197848306765e-05, + "loss": 1.166, + "step": 1646 + }, + { + "epoch": 0.05, + "learning_rate": 1.997312200931065e-05, + "loss": 1.1353, + "step": 1647 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973046063313974e-05, + "loss": 1.0688, + "step": 1648 + }, + { + "epoch": 0.05, + "learning_rate": 1.997297001031755e-05, + "loss": 1.1694, + "step": 1649 + }, + { + "epoch": 0.05, + "learning_rate": 1.99728938503222e-05, + "loss": 1.0264, + "step": 1650 + }, + { + "epoch": 0.05, + "learning_rate": 1.997281758332873e-05, + "loss": 1.1484, + "step": 1651 + }, + { + "epoch": 0.05, + "learning_rate": 1.997274120933797e-05, + "loss": 1.1509, + "step": 1652 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972664728350727e-05, + "loss": 1.0928, + "step": 1653 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972588140367833e-05, + "loss": 1.0405, + "step": 1654 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972511445390106e-05, + "loss": 1.0967, + "step": 1655 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972434643418368e-05, + "loss": 1.2344, + "step": 1656 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972357734453444e-05, + "loss": 0.959, + "step": 1657 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972280718496155e-05, + "loss": 1.1631, + "step": 1658 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972203595547334e-05, + "loss": 1.1587, + "step": 1659 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972126365607805e-05, + "loss": 0.9478, + "step": 1660 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972049028678396e-05, + "loss": 1.1567, + "step": 1661 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971971584759937e-05, + "loss": 1.0264, + "step": 1662 + }, + { + "epoch": 0.05, + "learning_rate": 1.997189403385326e-05, + "loss": 1.0947, + "step": 1663 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971816375959193e-05, + "loss": 1.1597, + "step": 1664 + }, + { + "epoch": 0.05, + "learning_rate": 1.997173861107858e-05, + "loss": 1.1807, + "step": 1665 + }, + { + "epoch": 0.05, + "learning_rate": 1.997166073921224e-05, + "loss": 1.1157, + "step": 1666 + }, + { + "epoch": 0.05, + "learning_rate": 1.997158276036102e-05, + "loss": 0.9897, + "step": 1667 + }, + { + "epoch": 0.05, + "learning_rate": 1.997150467452575e-05, + "loss": 1.1963, + "step": 1668 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971426481707276e-05, + "loss": 1.2617, + "step": 1669 + }, + { + "epoch": 0.05, + "learning_rate": 1.997134818190643e-05, + "loss": 1.0464, + "step": 1670 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971269775124053e-05, + "loss": 1.0488, + "step": 1671 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971191261360982e-05, + "loss": 1.2549, + "step": 1672 + }, + { + "epoch": 0.05, + "learning_rate": 1.997111264061807e-05, + "loss": 1.1382, + "step": 1673 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971033912896155e-05, + "loss": 1.2051, + "step": 1674 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970955078196082e-05, + "loss": 1.2275, + "step": 1675 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970876136518692e-05, + "loss": 1.1245, + "step": 1676 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970797087864842e-05, + "loss": 0.4214, + "step": 1677 + }, + { + "epoch": 0.05, + "learning_rate": 1.997071793223537e-05, + "loss": 1.019, + "step": 1678 + }, + { + "epoch": 0.05, + "learning_rate": 1.997063866963113e-05, + "loss": 1.2104, + "step": 1679 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970559300052974e-05, + "loss": 1.1919, + "step": 1680 + }, + { + "epoch": 0.05, + "learning_rate": 1.997047982350175e-05, + "loss": 1.0669, + "step": 1681 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970400239978313e-05, + "loss": 1.144, + "step": 1682 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970320549483516e-05, + "loss": 1.1968, + "step": 1683 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970240752018216e-05, + "loss": 1.2168, + "step": 1684 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970160847583262e-05, + "loss": 1.1235, + "step": 1685 + }, + { + "epoch": 0.05, + "learning_rate": 1.997008083617952e-05, + "loss": 1.2446, + "step": 1686 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970000717807846e-05, + "loss": 1.0605, + "step": 1687 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969920492469097e-05, + "loss": 1.1465, + "step": 1688 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969840160164134e-05, + "loss": 1.1318, + "step": 1689 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969759720893825e-05, + "loss": 1.124, + "step": 1690 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969679174659023e-05, + "loss": 1.1626, + "step": 1691 + }, + { + "epoch": 0.05, + "learning_rate": 1.99695985214606e-05, + "loss": 1.1455, + "step": 1692 + }, + { + "epoch": 0.05, + "learning_rate": 1.996951776129942e-05, + "loss": 1.0029, + "step": 1693 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969436894176345e-05, + "loss": 1.2344, + "step": 1694 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969355920092245e-05, + "loss": 1.1689, + "step": 1695 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969274839047992e-05, + "loss": 1.1592, + "step": 1696 + }, + { + "epoch": 0.05, + "learning_rate": 1.996919365104445e-05, + "loss": 1.228, + "step": 1697 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969112356082498e-05, + "loss": 1.1646, + "step": 1698 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969030954163e-05, + "loss": 1.1118, + "step": 1699 + }, + { + "epoch": 0.05, + "learning_rate": 1.996894944528684e-05, + "loss": 1.1592, + "step": 1700 + }, + { + "epoch": 0.05, + "learning_rate": 1.996886782945488e-05, + "loss": 1.1978, + "step": 1701 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968786106668e-05, + "loss": 1.2773, + "step": 1702 + }, + { + "epoch": 0.05, + "learning_rate": 1.996870427692708e-05, + "loss": 1.0923, + "step": 1703 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968622340232993e-05, + "loss": 1.0933, + "step": 1704 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968540296586628e-05, + "loss": 1.2383, + "step": 1705 + }, + { + "epoch": 0.05, + "learning_rate": 1.996845814598885e-05, + "loss": 1.2026, + "step": 1706 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968375888440554e-05, + "loss": 1.2402, + "step": 1707 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968293523942615e-05, + "loss": 1.1006, + "step": 1708 + }, + { + "epoch": 0.05, + "learning_rate": 1.996821105249592e-05, + "loss": 1.3115, + "step": 1709 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968128474101355e-05, + "loss": 1.1245, + "step": 1710 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968045788759797e-05, + "loss": 0.9673, + "step": 1711 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967962996472146e-05, + "loss": 1.1484, + "step": 1712 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967880097239282e-05, + "loss": 1.1235, + "step": 1713 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967797091062097e-05, + "loss": 1.1396, + "step": 1714 + }, + { + "epoch": 0.05, + "learning_rate": 1.996771397794148e-05, + "loss": 1.1006, + "step": 1715 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967630757878322e-05, + "loss": 1.1602, + "step": 1716 + }, + { + "epoch": 0.05, + "learning_rate": 1.996754743087352e-05, + "loss": 1.166, + "step": 1717 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967463996927964e-05, + "loss": 1.1772, + "step": 1718 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967380456042552e-05, + "loss": 1.2373, + "step": 1719 + }, + { + "epoch": 0.06, + "learning_rate": 1.996729680821818e-05, + "loss": 1.1401, + "step": 1720 + }, + { + "epoch": 0.06, + "learning_rate": 1.9967213053455744e-05, + "loss": 1.168, + "step": 1721 + }, + { + "epoch": 0.06, + "learning_rate": 1.996712919175614e-05, + "loss": 1.1445, + "step": 1722 + }, + { + "epoch": 0.06, + "learning_rate": 1.9967045223120273e-05, + "loss": 1.146, + "step": 1723 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966961147549044e-05, + "loss": 1.2036, + "step": 1724 + }, + { + "epoch": 0.06, + "learning_rate": 1.996687696504335e-05, + "loss": 1.1528, + "step": 1725 + }, + { + "epoch": 0.06, + "learning_rate": 1.99667926756041e-05, + "loss": 1.127, + "step": 1726 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966708279232192e-05, + "loss": 1.0615, + "step": 1727 + }, + { + "epoch": 0.06, + "learning_rate": 1.996662377592854e-05, + "loss": 1.103, + "step": 1728 + }, + { + "epoch": 0.06, + "learning_rate": 1.996653916569404e-05, + "loss": 1.1123, + "step": 1729 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966454448529604e-05, + "loss": 1.0635, + "step": 1730 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966369624436142e-05, + "loss": 1.189, + "step": 1731 + }, + { + "epoch": 0.06, + "learning_rate": 1.996628469341457e-05, + "loss": 1.042, + "step": 1732 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966199655465786e-05, + "loss": 1.1216, + "step": 1733 + }, + { + "epoch": 0.06, + "learning_rate": 1.996611451059072e-05, + "loss": 1.1641, + "step": 1734 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966029258790267e-05, + "loss": 1.062, + "step": 1735 + }, + { + "epoch": 0.06, + "learning_rate": 1.996594390006535e-05, + "loss": 1.1392, + "step": 1736 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965858434416886e-05, + "loss": 1.0962, + "step": 1737 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965772861845793e-05, + "loss": 1.1655, + "step": 1738 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965687182352985e-05, + "loss": 1.1968, + "step": 1739 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965601395939385e-05, + "loss": 1.1074, + "step": 1740 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965515502605907e-05, + "loss": 1.1299, + "step": 1741 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965429502353482e-05, + "loss": 1.1509, + "step": 1742 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965343395183023e-05, + "loss": 1.1094, + "step": 1743 + }, + { + "epoch": 0.06, + "learning_rate": 1.996525718109546e-05, + "loss": 1.0898, + "step": 1744 + }, + { + "epoch": 0.06, + "learning_rate": 1.996517086009172e-05, + "loss": 1.0342, + "step": 1745 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965084432172723e-05, + "loss": 1.0771, + "step": 1746 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964997897339397e-05, + "loss": 1.1562, + "step": 1747 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964911255592672e-05, + "loss": 1.0825, + "step": 1748 + }, + { + "epoch": 0.06, + "learning_rate": 1.996482450693348e-05, + "loss": 1.1426, + "step": 1749 + }, + { + "epoch": 0.06, + "learning_rate": 1.996473765136275e-05, + "loss": 1.2373, + "step": 1750 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964650688881412e-05, + "loss": 1.251, + "step": 1751 + }, + { + "epoch": 0.06, + "learning_rate": 1.99645636194904e-05, + "loss": 1.1367, + "step": 1752 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964476443190653e-05, + "loss": 1.0669, + "step": 1753 + }, + { + "epoch": 0.06, + "learning_rate": 1.99643891599831e-05, + "loss": 1.1885, + "step": 1754 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964301769868673e-05, + "loss": 1.1167, + "step": 1755 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964214272848322e-05, + "loss": 1.1206, + "step": 1756 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964126668922978e-05, + "loss": 1.1182, + "step": 1757 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964038958093584e-05, + "loss": 1.0972, + "step": 1758 + }, + { + "epoch": 0.06, + "learning_rate": 1.996395114036108e-05, + "loss": 1.3311, + "step": 1759 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963863215726406e-05, + "loss": 1.2051, + "step": 1760 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963775184190508e-05, + "loss": 1.2261, + "step": 1761 + }, + { + "epoch": 0.06, + "learning_rate": 1.996368704575433e-05, + "loss": 1.2349, + "step": 1762 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963598800418814e-05, + "loss": 1.124, + "step": 1763 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963510448184913e-05, + "loss": 1.0649, + "step": 1764 + }, + { + "epoch": 0.06, + "learning_rate": 1.996342198905357e-05, + "loss": 1.1128, + "step": 1765 + }, + { + "epoch": 0.06, + "learning_rate": 1.996333342302574e-05, + "loss": 0.4128, + "step": 1766 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963244750102365e-05, + "loss": 1.1353, + "step": 1767 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963155970284402e-05, + "loss": 1.1416, + "step": 1768 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963067083572806e-05, + "loss": 1.0342, + "step": 1769 + }, + { + "epoch": 0.06, + "learning_rate": 1.996297808996852e-05, + "loss": 1.186, + "step": 1770 + }, + { + "epoch": 0.06, + "learning_rate": 1.996288898947251e-05, + "loss": 1.2578, + "step": 1771 + }, + { + "epoch": 0.06, + "learning_rate": 1.996279978208573e-05, + "loss": 1.2764, + "step": 1772 + }, + { + "epoch": 0.06, + "learning_rate": 1.996271046780913e-05, + "loss": 1.1489, + "step": 1773 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962621046643673e-05, + "loss": 1.1846, + "step": 1774 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962531518590317e-05, + "loss": 1.2837, + "step": 1775 + }, + { + "epoch": 0.06, + "learning_rate": 1.996244188365003e-05, + "loss": 1.0508, + "step": 1776 + }, + { + "epoch": 0.06, + "learning_rate": 1.996235214182376e-05, + "loss": 1.1611, + "step": 1777 + }, + { + "epoch": 0.06, + "learning_rate": 1.996226229311248e-05, + "loss": 1.1768, + "step": 1778 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962172337517154e-05, + "loss": 1.1299, + "step": 1779 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962082275038744e-05, + "loss": 1.1895, + "step": 1780 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961992105678213e-05, + "loss": 1.1836, + "step": 1781 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961901829436535e-05, + "loss": 1.146, + "step": 1782 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961811446314673e-05, + "loss": 1.1968, + "step": 1783 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961720956313605e-05, + "loss": 1.272, + "step": 1784 + }, + { + "epoch": 0.06, + "learning_rate": 1.996163035943429e-05, + "loss": 1.166, + "step": 1785 + }, + { + "epoch": 0.06, + "learning_rate": 1.996153965567771e-05, + "loss": 1.1001, + "step": 1786 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961448845044832e-05, + "loss": 1.04, + "step": 1787 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961357927536633e-05, + "loss": 1.1562, + "step": 1788 + }, + { + "epoch": 0.06, + "learning_rate": 1.996126690315409e-05, + "loss": 1.1738, + "step": 1789 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961175771898177e-05, + "loss": 1.1636, + "step": 1790 + }, + { + "epoch": 0.06, + "learning_rate": 1.996108453376987e-05, + "loss": 1.0415, + "step": 1791 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960993188770154e-05, + "loss": 1.0542, + "step": 1792 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960901736900002e-05, + "loss": 1.1367, + "step": 1793 + }, + { + "epoch": 0.06, + "learning_rate": 1.99608101781604e-05, + "loss": 1.2451, + "step": 1794 + }, + { + "epoch": 0.06, + "learning_rate": 1.996071851255233e-05, + "loss": 1.1309, + "step": 1795 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960626740076773e-05, + "loss": 1.0815, + "step": 1796 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960534860734718e-05, + "loss": 1.1001, + "step": 1797 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960442874527143e-05, + "loss": 1.144, + "step": 1798 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960350781455043e-05, + "loss": 1.1816, + "step": 1799 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960258581519402e-05, + "loss": 1.1445, + "step": 1800 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960166274721207e-05, + "loss": 1.3115, + "step": 1801 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960073861061454e-05, + "loss": 1.0723, + "step": 1802 + }, + { + "epoch": 0.06, + "learning_rate": 1.995998134054113e-05, + "loss": 1.0688, + "step": 1803 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959888713161232e-05, + "loss": 1.1016, + "step": 1804 + }, + { + "epoch": 0.06, + "learning_rate": 1.995979597892275e-05, + "loss": 1.1079, + "step": 1805 + }, + { + "epoch": 0.06, + "learning_rate": 1.995970313782668e-05, + "loss": 1.1099, + "step": 1806 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959610189874018e-05, + "loss": 1.1694, + "step": 1807 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959517135065764e-05, + "loss": 1.2021, + "step": 1808 + }, + { + "epoch": 0.06, + "learning_rate": 1.995942397340291e-05, + "loss": 1.1094, + "step": 1809 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959330704886462e-05, + "loss": 1.1719, + "step": 1810 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959237329517417e-05, + "loss": 1.0728, + "step": 1811 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959143847296775e-05, + "loss": 1.1284, + "step": 1812 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959050258225546e-05, + "loss": 1.2666, + "step": 1813 + }, + { + "epoch": 0.06, + "learning_rate": 1.995895656230473e-05, + "loss": 1.1172, + "step": 1814 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958862759535327e-05, + "loss": 1.23, + "step": 1815 + }, + { + "epoch": 0.06, + "learning_rate": 1.995876884991835e-05, + "loss": 0.4551, + "step": 1816 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958674833454807e-05, + "loss": 1.0684, + "step": 1817 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958580710145705e-05, + "loss": 1.1338, + "step": 1818 + }, + { + "epoch": 0.06, + "learning_rate": 1.995848647999205e-05, + "loss": 1.0415, + "step": 1819 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958392142994865e-05, + "loss": 1.0913, + "step": 1820 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958297699155147e-05, + "loss": 1.0728, + "step": 1821 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958203148473915e-05, + "loss": 0.9722, + "step": 1822 + }, + { + "epoch": 0.06, + "learning_rate": 1.995810849095219e-05, + "loss": 1.0479, + "step": 1823 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958013726590978e-05, + "loss": 1.1553, + "step": 1824 + }, + { + "epoch": 0.06, + "learning_rate": 1.99579188553913e-05, + "loss": 1.0918, + "step": 1825 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957823877354176e-05, + "loss": 0.469, + "step": 1826 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957728792480617e-05, + "loss": 1.3008, + "step": 1827 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957633600771654e-05, + "loss": 1.0947, + "step": 1828 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957538302228302e-05, + "loss": 1.1455, + "step": 1829 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957442896851584e-05, + "loss": 1.1685, + "step": 1830 + }, + { + "epoch": 0.06, + "learning_rate": 1.995734738464252e-05, + "loss": 1.0845, + "step": 1831 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957251765602143e-05, + "loss": 1.1533, + "step": 1832 + }, + { + "epoch": 0.06, + "learning_rate": 1.995715603973148e-05, + "loss": 1.1655, + "step": 1833 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957060207031542e-05, + "loss": 1.1123, + "step": 1834 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956964267503376e-05, + "loss": 1.0342, + "step": 1835 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956868221148e-05, + "loss": 1.1606, + "step": 1836 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956772067966444e-05, + "loss": 1.1636, + "step": 1837 + }, + { + "epoch": 0.06, + "learning_rate": 1.995667580795975e-05, + "loss": 1.2192, + "step": 1838 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956579441128942e-05, + "loss": 1.1533, + "step": 1839 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956482967475054e-05, + "loss": 1.3232, + "step": 1840 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956386386999124e-05, + "loss": 1.2051, + "step": 1841 + }, + { + "epoch": 0.06, + "learning_rate": 1.995628969970219e-05, + "loss": 1.043, + "step": 1842 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956192905585284e-05, + "loss": 1.1582, + "step": 1843 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956096004649444e-05, + "loss": 1.272, + "step": 1844 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955998996895716e-05, + "loss": 1.1162, + "step": 1845 + }, + { + "epoch": 0.06, + "learning_rate": 1.995590188232514e-05, + "loss": 1.125, + "step": 1846 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955804660938752e-05, + "loss": 1.0474, + "step": 1847 + }, + { + "epoch": 0.06, + "learning_rate": 1.99557073327376e-05, + "loss": 1.2114, + "step": 1848 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955609897722726e-05, + "loss": 1.2007, + "step": 1849 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955512355895176e-05, + "loss": 1.106, + "step": 1850 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955414707255998e-05, + "loss": 1.0591, + "step": 1851 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955316951806236e-05, + "loss": 1.1709, + "step": 1852 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955219089546944e-05, + "loss": 1.2168, + "step": 1853 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955121120479167e-05, + "loss": 1.1509, + "step": 1854 + }, + { + "epoch": 0.06, + "learning_rate": 1.995502304460396e-05, + "loss": 1.2549, + "step": 1855 + }, + { + "epoch": 0.06, + "learning_rate": 1.995492486192237e-05, + "loss": 1.083, + "step": 1856 + }, + { + "epoch": 0.06, + "learning_rate": 1.995482657243546e-05, + "loss": 1.1416, + "step": 1857 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954728176144273e-05, + "loss": 1.1167, + "step": 1858 + }, + { + "epoch": 0.06, + "learning_rate": 1.995462967304987e-05, + "loss": 1.1016, + "step": 1859 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954531063153308e-05, + "loss": 1.0713, + "step": 1860 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954432346455646e-05, + "loss": 1.2051, + "step": 1861 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954333522957942e-05, + "loss": 1.1436, + "step": 1862 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954234592661258e-05, + "loss": 0.9873, + "step": 1863 + }, + { + "epoch": 0.06, + "learning_rate": 1.995413555556665e-05, + "loss": 1.0972, + "step": 1864 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954036411675188e-05, + "loss": 1.1797, + "step": 1865 + }, + { + "epoch": 0.06, + "learning_rate": 1.995393716098793e-05, + "loss": 1.1318, + "step": 1866 + }, + { + "epoch": 0.06, + "learning_rate": 1.995383780350594e-05, + "loss": 1.165, + "step": 1867 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953738339230287e-05, + "loss": 1.1509, + "step": 1868 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953638768162042e-05, + "loss": 1.1523, + "step": 1869 + }, + { + "epoch": 0.06, + "learning_rate": 1.995353909030227e-05, + "loss": 1.1255, + "step": 1870 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953439305652033e-05, + "loss": 1.0498, + "step": 1871 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953339414212414e-05, + "loss": 1.0874, + "step": 1872 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953239415984478e-05, + "loss": 1.1421, + "step": 1873 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953139310969295e-05, + "loss": 1.0981, + "step": 1874 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953039099167948e-05, + "loss": 1.2227, + "step": 1875 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952938780581502e-05, + "loss": 1.0542, + "step": 1876 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952838355211043e-05, + "loss": 1.186, + "step": 1877 + }, + { + "epoch": 0.06, + "learning_rate": 1.995273782305764e-05, + "loss": 1.1572, + "step": 1878 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952637184122378e-05, + "loss": 1.1035, + "step": 1879 + }, + { + "epoch": 0.06, + "learning_rate": 1.995253643840633e-05, + "loss": 1.1714, + "step": 1880 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952435585910584e-05, + "loss": 0.4692, + "step": 1881 + }, + { + "epoch": 0.06, + "learning_rate": 1.995233462663622e-05, + "loss": 1.0103, + "step": 1882 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952233560584316e-05, + "loss": 1.1602, + "step": 1883 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952132387755965e-05, + "loss": 1.1973, + "step": 1884 + }, + { + "epoch": 0.06, + "learning_rate": 1.995203110815225e-05, + "loss": 1.1631, + "step": 1885 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951929721774247e-05, + "loss": 1.1899, + "step": 1886 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951828228623057e-05, + "loss": 1.0474, + "step": 1887 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951726628699765e-05, + "loss": 1.0586, + "step": 1888 + }, + { + "epoch": 0.06, + "learning_rate": 1.995162492200546e-05, + "loss": 1.0781, + "step": 1889 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951523108541234e-05, + "loss": 1.208, + "step": 1890 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951421188308178e-05, + "loss": 1.0254, + "step": 1891 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951319161307384e-05, + "loss": 1.144, + "step": 1892 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951217027539954e-05, + "loss": 1.1523, + "step": 1893 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951114787006976e-05, + "loss": 1.2324, + "step": 1894 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951012439709547e-05, + "loss": 1.0859, + "step": 1895 + }, + { + "epoch": 0.06, + "learning_rate": 1.995090998564877e-05, + "loss": 1.2817, + "step": 1896 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950807424825744e-05, + "loss": 1.0913, + "step": 1897 + }, + { + "epoch": 0.06, + "learning_rate": 1.995070475724157e-05, + "loss": 1.1636, + "step": 1898 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950601982897338e-05, + "loss": 1.208, + "step": 1899 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950499101794165e-05, + "loss": 1.0693, + "step": 1900 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950396113933152e-05, + "loss": 1.2119, + "step": 1901 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950293019315396e-05, + "loss": 1.1768, + "step": 1902 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950189817942008e-05, + "loss": 1.1758, + "step": 1903 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950086509814097e-05, + "loss": 1.1279, + "step": 1904 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949983094932773e-05, + "loss": 1.0991, + "step": 1905 + }, + { + "epoch": 0.06, + "learning_rate": 1.994987957329914e-05, + "loss": 1.0845, + "step": 1906 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949775944914308e-05, + "loss": 1.1226, + "step": 1907 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949672209779395e-05, + "loss": 1.1157, + "step": 1908 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949568367895512e-05, + "loss": 1.062, + "step": 1909 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949464419263773e-05, + "loss": 1.1719, + "step": 1910 + }, + { + "epoch": 0.06, + "learning_rate": 1.994936036388529e-05, + "loss": 1.1919, + "step": 1911 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949256201761183e-05, + "loss": 1.1523, + "step": 1912 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949151932892565e-05, + "loss": 1.0161, + "step": 1913 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949047557280564e-05, + "loss": 1.0562, + "step": 1914 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948943074926287e-05, + "loss": 1.2329, + "step": 1915 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948838485830867e-05, + "loss": 1.1416, + "step": 1916 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948733789995417e-05, + "loss": 1.1411, + "step": 1917 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948628987421066e-05, + "loss": 1.1963, + "step": 1918 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948524078108936e-05, + "loss": 1.1143, + "step": 1919 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948419062060155e-05, + "loss": 1.1709, + "step": 1920 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948313939275844e-05, + "loss": 1.0225, + "step": 1921 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948208709757138e-05, + "loss": 1.0542, + "step": 1922 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948103373505163e-05, + "loss": 1.0444, + "step": 1923 + }, + { + "epoch": 0.06, + "learning_rate": 1.994799793052105e-05, + "loss": 1.249, + "step": 1924 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947892380805926e-05, + "loss": 1.1689, + "step": 1925 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947786724360924e-05, + "loss": 1.1118, + "step": 1926 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947680961187185e-05, + "loss": 1.0859, + "step": 1927 + }, + { + "epoch": 0.06, + "learning_rate": 1.994757509128584e-05, + "loss": 1.0391, + "step": 1928 + }, + { + "epoch": 0.06, + "learning_rate": 1.994746911465802e-05, + "loss": 1.1943, + "step": 1929 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947363031304864e-05, + "loss": 1.0967, + "step": 1930 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947256841227516e-05, + "loss": 1.1748, + "step": 1931 + }, + { + "epoch": 0.06, + "learning_rate": 1.994715054442711e-05, + "loss": 1.1045, + "step": 1932 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947044140904786e-05, + "loss": 1.0347, + "step": 1933 + }, + { + "epoch": 0.06, + "learning_rate": 1.994693763066169e-05, + "loss": 1.061, + "step": 1934 + }, + { + "epoch": 0.06, + "learning_rate": 1.994683101369896e-05, + "loss": 1.0371, + "step": 1935 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946724290017744e-05, + "loss": 0.9771, + "step": 1936 + }, + { + "epoch": 0.06, + "learning_rate": 1.994661745961918e-05, + "loss": 1.1733, + "step": 1937 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946510522504424e-05, + "loss": 1.0786, + "step": 1938 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946403478674616e-05, + "loss": 1.1787, + "step": 1939 + }, + { + "epoch": 0.06, + "learning_rate": 1.994629632813091e-05, + "loss": 1.1191, + "step": 1940 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946189070874447e-05, + "loss": 1.188, + "step": 1941 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946081706906386e-05, + "loss": 1.1099, + "step": 1942 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945974236227878e-05, + "loss": 1.1553, + "step": 1943 + }, + { + "epoch": 0.06, + "learning_rate": 1.994586665884007e-05, + "loss": 1.1777, + "step": 1944 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945758974744123e-05, + "loss": 1.0264, + "step": 1945 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945651183941187e-05, + "loss": 1.0786, + "step": 1946 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945543286432426e-05, + "loss": 1.0933, + "step": 1947 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945435282218987e-05, + "loss": 1.0303, + "step": 1948 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945327171302037e-05, + "loss": 1.1504, + "step": 1949 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945218953682736e-05, + "loss": 1.022, + "step": 1950 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945110629362238e-05, + "loss": 0.4817, + "step": 1951 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945002198341714e-05, + "loss": 1.2207, + "step": 1952 + }, + { + "epoch": 0.06, + "learning_rate": 1.994489366062232e-05, + "loss": 1.1973, + "step": 1953 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944785016205222e-05, + "loss": 1.0449, + "step": 1954 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944676265091592e-05, + "loss": 1.1694, + "step": 1955 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944567407282588e-05, + "loss": 1.0117, + "step": 1956 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944458442779386e-05, + "loss": 1.1172, + "step": 1957 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944349371583148e-05, + "loss": 1.1636, + "step": 1958 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944240193695043e-05, + "loss": 1.2344, + "step": 1959 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944130909116254e-05, + "loss": 1.0737, + "step": 1960 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944021517847943e-05, + "loss": 1.165, + "step": 1961 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943912019891285e-05, + "loss": 1.1074, + "step": 1962 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943802415247456e-05, + "loss": 1.1528, + "step": 1963 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943692703917635e-05, + "loss": 1.105, + "step": 1964 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943582885902995e-05, + "loss": 1.1133, + "step": 1965 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943472961204715e-05, + "loss": 1.1167, + "step": 1966 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943362929823976e-05, + "loss": 1.2041, + "step": 1967 + }, + { + "epoch": 0.06, + "learning_rate": 1.994325279176196e-05, + "loss": 1.0786, + "step": 1968 + }, + { + "epoch": 0.06, + "learning_rate": 1.994314254701984e-05, + "loss": 1.0508, + "step": 1969 + }, + { + "epoch": 0.06, + "learning_rate": 1.994303219559881e-05, + "loss": 1.1191, + "step": 1970 + }, + { + "epoch": 0.06, + "learning_rate": 1.994292173750005e-05, + "loss": 1.1353, + "step": 1971 + }, + { + "epoch": 0.06, + "learning_rate": 1.994281117272474e-05, + "loss": 1.1104, + "step": 1972 + }, + { + "epoch": 0.06, + "learning_rate": 1.994270050127407e-05, + "loss": 1.2627, + "step": 1973 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942589723149233e-05, + "loss": 1.1807, + "step": 1974 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942478838351406e-05, + "loss": 1.1475, + "step": 1975 + }, + { + "epoch": 0.06, + "learning_rate": 1.994236784688179e-05, + "loss": 1.1167, + "step": 1976 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942256748741567e-05, + "loss": 1.04, + "step": 1977 + }, + { + "epoch": 0.06, + "learning_rate": 1.994214554393193e-05, + "loss": 1.0659, + "step": 1978 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942034232454084e-05, + "loss": 1.1528, + "step": 1979 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941922814309208e-05, + "loss": 1.1421, + "step": 1980 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941811289498503e-05, + "loss": 1.0732, + "step": 1981 + }, + { + "epoch": 0.06, + "learning_rate": 1.994169965802317e-05, + "loss": 1.0825, + "step": 1982 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941587919884394e-05, + "loss": 1.2861, + "step": 1983 + }, + { + "epoch": 0.06, + "learning_rate": 1.994147607508339e-05, + "loss": 1.1118, + "step": 1984 + }, + { + "epoch": 0.06, + "learning_rate": 1.994136412362135e-05, + "loss": 1.2646, + "step": 1985 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941252065499473e-05, + "loss": 1.2012, + "step": 1986 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941139900718964e-05, + "loss": 1.1416, + "step": 1987 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941027629281026e-05, + "loss": 1.1753, + "step": 1988 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940915251186867e-05, + "loss": 1.1421, + "step": 1989 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940802766437687e-05, + "loss": 1.1323, + "step": 1990 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940690175034693e-05, + "loss": 1.0894, + "step": 1991 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940577476979097e-05, + "loss": 1.1753, + "step": 1992 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940464672272108e-05, + "loss": 1.1147, + "step": 1993 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940351760914934e-05, + "loss": 1.126, + "step": 1994 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940238742908786e-05, + "loss": 1.2461, + "step": 1995 + }, + { + "epoch": 0.06, + "learning_rate": 1.994012561825488e-05, + "loss": 1.1646, + "step": 1996 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940012386954424e-05, + "loss": 1.2432, + "step": 1997 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939899049008638e-05, + "loss": 1.1328, + "step": 1998 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939785604418734e-05, + "loss": 1.0527, + "step": 1999 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939672053185936e-05, + "loss": 1.252, + "step": 2000 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939558395311452e-05, + "loss": 1.1128, + "step": 2001 + }, + { + "epoch": 0.06, + "learning_rate": 1.993944463079651e-05, + "loss": 1.1123, + "step": 2002 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939330759642325e-05, + "loss": 1.0117, + "step": 2003 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939216781850127e-05, + "loss": 1.1514, + "step": 2004 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939102697421127e-05, + "loss": 1.1924, + "step": 2005 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938988506356557e-05, + "loss": 1.084, + "step": 2006 + }, + { + "epoch": 0.06, + "learning_rate": 1.993887420865764e-05, + "loss": 1.1367, + "step": 2007 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938759804325604e-05, + "loss": 1.1235, + "step": 2008 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938645293361674e-05, + "loss": 1.3081, + "step": 2009 + }, + { + "epoch": 0.06, + "learning_rate": 1.993853067576708e-05, + "loss": 1.1851, + "step": 2010 + }, + { + "epoch": 0.06, + "learning_rate": 1.993841595154305e-05, + "loss": 1.124, + "step": 2011 + }, + { + "epoch": 0.06, + "learning_rate": 1.993830112069082e-05, + "loss": 0.9736, + "step": 2012 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938186183211614e-05, + "loss": 1.1748, + "step": 2013 + }, + { + "epoch": 0.06, + "learning_rate": 1.993807113910667e-05, + "loss": 1.0435, + "step": 2014 + }, + { + "epoch": 0.06, + "learning_rate": 1.993795598837722e-05, + "loss": 1.1787, + "step": 2015 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937840731024508e-05, + "loss": 1.2168, + "step": 2016 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937725367049756e-05, + "loss": 1.0815, + "step": 2017 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937609896454214e-05, + "loss": 1.0669, + "step": 2018 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937494319239112e-05, + "loss": 1.0625, + "step": 2019 + }, + { + "epoch": 0.06, + "learning_rate": 1.99373786354057e-05, + "loss": 1.0884, + "step": 2020 + }, + { + "epoch": 0.06, + "learning_rate": 1.993726284495521e-05, + "loss": 1.2461, + "step": 2021 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937146947888884e-05, + "loss": 1.167, + "step": 2022 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937030944207977e-05, + "loss": 1.0396, + "step": 2023 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936914833913722e-05, + "loss": 1.1924, + "step": 2024 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936798617007364e-05, + "loss": 1.0405, + "step": 2025 + }, + { + "epoch": 0.06, + "learning_rate": 1.993668229349016e-05, + "loss": 1.1772, + "step": 2026 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936565863363352e-05, + "loss": 1.1797, + "step": 2027 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936449326628188e-05, + "loss": 1.1455, + "step": 2028 + }, + { + "epoch": 0.06, + "learning_rate": 1.993633268328592e-05, + "loss": 1.1201, + "step": 2029 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936215933337797e-05, + "loss": 1.1631, + "step": 2030 + }, + { + "epoch": 0.06, + "learning_rate": 1.993609907678508e-05, + "loss": 1.1338, + "step": 2031 + }, + { + "epoch": 0.06, + "learning_rate": 1.993598211362901e-05, + "loss": 1.1753, + "step": 2032 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935865043870853e-05, + "loss": 1.1182, + "step": 2033 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935747867511857e-05, + "loss": 1.0869, + "step": 2034 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935630584553288e-05, + "loss": 1.2686, + "step": 2035 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935513194996395e-05, + "loss": 1.167, + "step": 2036 + }, + { + "epoch": 0.07, + "learning_rate": 1.993539569884244e-05, + "loss": 1.0698, + "step": 2037 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935278096092688e-05, + "loss": 1.1538, + "step": 2038 + }, + { + "epoch": 0.07, + "learning_rate": 1.99351603867484e-05, + "loss": 1.0981, + "step": 2039 + }, + { + "epoch": 0.07, + "learning_rate": 1.993504257081083e-05, + "loss": 1.0356, + "step": 2040 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934924648281254e-05, + "loss": 1.0747, + "step": 2041 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934806619160933e-05, + "loss": 1.0996, + "step": 2042 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934688483451127e-05, + "loss": 1.144, + "step": 2043 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934570241153114e-05, + "loss": 1.1816, + "step": 2044 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934451892268155e-05, + "loss": 1.0254, + "step": 2045 + }, + { + "epoch": 0.07, + "learning_rate": 1.993433343679752e-05, + "loss": 1.1646, + "step": 2046 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934214874742486e-05, + "loss": 1.0205, + "step": 2047 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934096206104317e-05, + "loss": 1.0469, + "step": 2048 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933977430884294e-05, + "loss": 1.1045, + "step": 2049 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933858549083688e-05, + "loss": 1.0752, + "step": 2050 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933739560703768e-05, + "loss": 1.2217, + "step": 2051 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933620465745823e-05, + "loss": 1.1982, + "step": 2052 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933501264211122e-05, + "loss": 1.2637, + "step": 2053 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933381956100944e-05, + "loss": 1.0615, + "step": 2054 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933262541416576e-05, + "loss": 1.0039, + "step": 2055 + }, + { + "epoch": 0.07, + "learning_rate": 1.993314302015929e-05, + "loss": 1.0986, + "step": 2056 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933023392330377e-05, + "loss": 1.2153, + "step": 2057 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932903657931116e-05, + "loss": 1.1265, + "step": 2058 + }, + { + "epoch": 0.07, + "learning_rate": 1.993278381696279e-05, + "loss": 1.1953, + "step": 2059 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932663869426684e-05, + "loss": 0.4956, + "step": 2060 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932543815324093e-05, + "loss": 0.9619, + "step": 2061 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932423654656297e-05, + "loss": 1.1182, + "step": 2062 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932303387424587e-05, + "loss": 1.1812, + "step": 2063 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932183013630257e-05, + "loss": 1.1138, + "step": 2064 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932062533274594e-05, + "loss": 1.2344, + "step": 2065 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931941946358894e-05, + "loss": 1.084, + "step": 2066 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931821252884443e-05, + "loss": 1.2261, + "step": 2067 + }, + { + "epoch": 0.07, + "learning_rate": 1.993170045285255e-05, + "loss": 1.2178, + "step": 2068 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931579546264497e-05, + "loss": 0.447, + "step": 2069 + }, + { + "epoch": 0.07, + "learning_rate": 1.993145853312159e-05, + "loss": 1.1802, + "step": 2070 + }, + { + "epoch": 0.07, + "learning_rate": 1.993133741342512e-05, + "loss": 0.9644, + "step": 2071 + }, + { + "epoch": 0.07, + "learning_rate": 1.99312161871764e-05, + "loss": 1.0679, + "step": 2072 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931094854376716e-05, + "loss": 1.1753, + "step": 2073 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930973415027376e-05, + "loss": 1.1152, + "step": 2074 + }, + { + "epoch": 0.07, + "learning_rate": 1.993085186912968e-05, + "loss": 1.1523, + "step": 2075 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930730216684937e-05, + "loss": 1.0825, + "step": 2076 + }, + { + "epoch": 0.07, + "learning_rate": 1.993060845769445e-05, + "loss": 1.1401, + "step": 2077 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930486592159527e-05, + "loss": 1.2412, + "step": 2078 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930364620081467e-05, + "loss": 1.2129, + "step": 2079 + }, + { + "epoch": 0.07, + "learning_rate": 1.993024254146159e-05, + "loss": 0.4399, + "step": 2080 + }, + { + "epoch": 0.07, + "learning_rate": 1.99301203563012e-05, + "loss": 1.1187, + "step": 2081 + }, + { + "epoch": 0.07, + "learning_rate": 1.992999806460161e-05, + "loss": 1.0933, + "step": 2082 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929875666364125e-05, + "loss": 1.126, + "step": 2083 + }, + { + "epoch": 0.07, + "learning_rate": 1.992975316159007e-05, + "loss": 0.4326, + "step": 2084 + }, + { + "epoch": 0.07, + "learning_rate": 1.992963055028075e-05, + "loss": 0.447, + "step": 2085 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929507832437484e-05, + "loss": 1.1074, + "step": 2086 + }, + { + "epoch": 0.07, + "learning_rate": 1.992938500806159e-05, + "loss": 1.2471, + "step": 2087 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929262077154383e-05, + "loss": 1.1309, + "step": 2088 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929139039717184e-05, + "loss": 1.1416, + "step": 2089 + }, + { + "epoch": 0.07, + "learning_rate": 1.992901589575131e-05, + "loss": 1.1636, + "step": 2090 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928892645258088e-05, + "loss": 1.1782, + "step": 2091 + }, + { + "epoch": 0.07, + "learning_rate": 1.992876928823883e-05, + "loss": 1.0825, + "step": 2092 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928645824694873e-05, + "loss": 1.1987, + "step": 2093 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928522254627534e-05, + "loss": 1.168, + "step": 2094 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928398578038138e-05, + "loss": 1.2559, + "step": 2095 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928274794928013e-05, + "loss": 1.0625, + "step": 2096 + }, + { + "epoch": 0.07, + "learning_rate": 1.992815090529849e-05, + "loss": 1.0703, + "step": 2097 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928026909150892e-05, + "loss": 1.1416, + "step": 2098 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927902806486556e-05, + "loss": 0.9531, + "step": 2099 + }, + { + "epoch": 0.07, + "learning_rate": 1.992777859730681e-05, + "loss": 1.0918, + "step": 2100 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927654281612986e-05, + "loss": 1.146, + "step": 2101 + }, + { + "epoch": 0.07, + "learning_rate": 1.992752985940642e-05, + "loss": 1.2056, + "step": 2102 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927405330688444e-05, + "loss": 1.1167, + "step": 2103 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927280695460396e-05, + "loss": 1.1211, + "step": 2104 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927155953723613e-05, + "loss": 1.1182, + "step": 2105 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927031105479436e-05, + "loss": 1.2319, + "step": 2106 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926906150729198e-05, + "loss": 1.0522, + "step": 2107 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926781089474247e-05, + "loss": 1.1235, + "step": 2108 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926655921715924e-05, + "loss": 1.1753, + "step": 2109 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926530647455564e-05, + "loss": 1.0996, + "step": 2110 + }, + { + "epoch": 0.07, + "learning_rate": 1.992640526669452e-05, + "loss": 1.0596, + "step": 2111 + }, + { + "epoch": 0.07, + "learning_rate": 1.992627977943413e-05, + "loss": 0.9922, + "step": 2112 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926154185675744e-05, + "loss": 1.1055, + "step": 2113 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926028485420713e-05, + "loss": 1.1636, + "step": 2114 + }, + { + "epoch": 0.07, + "learning_rate": 1.992590267867038e-05, + "loss": 1.0474, + "step": 2115 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925776765426096e-05, + "loss": 1.0742, + "step": 2116 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925650745689214e-05, + "loss": 1.0747, + "step": 2117 + }, + { + "epoch": 0.07, + "learning_rate": 1.992552461946108e-05, + "loss": 1.2197, + "step": 2118 + }, + { + "epoch": 0.07, + "learning_rate": 1.992539838674306e-05, + "loss": 1.2305, + "step": 2119 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925272047536492e-05, + "loss": 0.4211, + "step": 2120 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925145601842747e-05, + "loss": 1.2256, + "step": 2121 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925019049663166e-05, + "loss": 1.2109, + "step": 2122 + }, + { + "epoch": 0.07, + "learning_rate": 1.992489239099912e-05, + "loss": 1.1709, + "step": 2123 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924765625851964e-05, + "loss": 1.1348, + "step": 2124 + }, + { + "epoch": 0.07, + "learning_rate": 1.992463875422306e-05, + "loss": 1.1875, + "step": 2125 + }, + { + "epoch": 0.07, + "learning_rate": 1.992451177611376e-05, + "loss": 1.2373, + "step": 2126 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924384691525435e-05, + "loss": 1.1182, + "step": 2127 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924257500459446e-05, + "loss": 1.0625, + "step": 2128 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924130202917154e-05, + "loss": 1.0332, + "step": 2129 + }, + { + "epoch": 0.07, + "learning_rate": 1.992400279889993e-05, + "loss": 1.1235, + "step": 2130 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923875288409144e-05, + "loss": 1.2349, + "step": 2131 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923747671446154e-05, + "loss": 1.1226, + "step": 2132 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923619948012338e-05, + "loss": 1.0503, + "step": 2133 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923492118109063e-05, + "loss": 1.1138, + "step": 2134 + }, + { + "epoch": 0.07, + "learning_rate": 1.99233641817377e-05, + "loss": 1.1309, + "step": 2135 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923236138899618e-05, + "loss": 1.0415, + "step": 2136 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923107989596197e-05, + "loss": 1.1016, + "step": 2137 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922979733828815e-05, + "loss": 1.0972, + "step": 2138 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922851371598834e-05, + "loss": 1.0674, + "step": 2139 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922722902907647e-05, + "loss": 1.0723, + "step": 2140 + }, + { + "epoch": 0.07, + "learning_rate": 1.992259432775662e-05, + "loss": 1.2134, + "step": 2141 + }, + { + "epoch": 0.07, + "learning_rate": 1.992246564614714e-05, + "loss": 1.2627, + "step": 2142 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922336858080584e-05, + "loss": 1.0347, + "step": 2143 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922207963558334e-05, + "loss": 1.2329, + "step": 2144 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922078962581777e-05, + "loss": 1.1055, + "step": 2145 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921949855152294e-05, + "loss": 1.2168, + "step": 2146 + }, + { + "epoch": 0.07, + "learning_rate": 1.992182064127127e-05, + "loss": 1.1787, + "step": 2147 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921691320940087e-05, + "loss": 1.1196, + "step": 2148 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921561894160142e-05, + "loss": 1.0054, + "step": 2149 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921432360932817e-05, + "loss": 1.2705, + "step": 2150 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921302721259502e-05, + "loss": 1.0215, + "step": 2151 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921172975141593e-05, + "loss": 1.0537, + "step": 2152 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921043122580476e-05, + "loss": 1.064, + "step": 2153 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920913163577542e-05, + "loss": 1.1729, + "step": 2154 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920783098134196e-05, + "loss": 1.0347, + "step": 2155 + }, + { + "epoch": 0.07, + "learning_rate": 1.992065292625182e-05, + "loss": 1.2119, + "step": 2156 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920522647931826e-05, + "loss": 1.1836, + "step": 2157 + }, + { + "epoch": 0.07, + "learning_rate": 1.99203922631756e-05, + "loss": 1.0723, + "step": 2158 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920261771984544e-05, + "loss": 1.0488, + "step": 2159 + }, + { + "epoch": 0.07, + "learning_rate": 1.992013117436006e-05, + "loss": 1.1729, + "step": 2160 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920000470303544e-05, + "loss": 1.1504, + "step": 2161 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919869659816403e-05, + "loss": 1.1802, + "step": 2162 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919738742900044e-05, + "loss": 1.1255, + "step": 2163 + }, + { + "epoch": 0.07, + "learning_rate": 1.991960771955586e-05, + "loss": 1.1396, + "step": 2164 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919476589785266e-05, + "loss": 1.0327, + "step": 2165 + }, + { + "epoch": 0.07, + "learning_rate": 1.991934535358967e-05, + "loss": 1.0874, + "step": 2166 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919214010970475e-05, + "loss": 1.2559, + "step": 2167 + }, + { + "epoch": 0.07, + "learning_rate": 1.991908256192909e-05, + "loss": 1.1074, + "step": 2168 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918951006466927e-05, + "loss": 1.0884, + "step": 2169 + }, + { + "epoch": 0.07, + "learning_rate": 1.99188193445854e-05, + "loss": 1.1509, + "step": 2170 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918687576285914e-05, + "loss": 1.1377, + "step": 2171 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918555701569892e-05, + "loss": 1.1826, + "step": 2172 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918423720438742e-05, + "loss": 1.1812, + "step": 2173 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918291632893885e-05, + "loss": 1.0879, + "step": 2174 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918159438936735e-05, + "loss": 1.0322, + "step": 2175 + }, + { + "epoch": 0.07, + "learning_rate": 1.991802713856871e-05, + "loss": 1.0439, + "step": 2176 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917894731791233e-05, + "loss": 1.0986, + "step": 2177 + }, + { + "epoch": 0.07, + "learning_rate": 1.991776221860572e-05, + "loss": 1.0752, + "step": 2178 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917629599013596e-05, + "loss": 1.1279, + "step": 2179 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917496873016282e-05, + "loss": 1.0137, + "step": 2180 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917364040615204e-05, + "loss": 1.0747, + "step": 2181 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917231101811786e-05, + "loss": 1.2378, + "step": 2182 + }, + { + "epoch": 0.07, + "learning_rate": 1.991709805660745e-05, + "loss": 1.1909, + "step": 2183 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916964905003632e-05, + "loss": 1.1494, + "step": 2184 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916831647001753e-05, + "loss": 1.127, + "step": 2185 + }, + { + "epoch": 0.07, + "learning_rate": 1.991669828260325e-05, + "loss": 1.125, + "step": 2186 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916564811809543e-05, + "loss": 1.188, + "step": 2187 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916431234622075e-05, + "loss": 1.1016, + "step": 2188 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916297551042277e-05, + "loss": 1.1162, + "step": 2189 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916163761071578e-05, + "loss": 1.1279, + "step": 2190 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916029864711416e-05, + "loss": 1.1528, + "step": 2191 + }, + { + "epoch": 0.07, + "learning_rate": 1.991589586196323e-05, + "loss": 1.0723, + "step": 2192 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915761752828457e-05, + "loss": 1.1792, + "step": 2193 + }, + { + "epoch": 0.07, + "learning_rate": 1.991562753730853e-05, + "loss": 1.1572, + "step": 2194 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915493215404897e-05, + "loss": 1.0884, + "step": 2195 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915358787118994e-05, + "loss": 1.0703, + "step": 2196 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915224252452265e-05, + "loss": 1.1284, + "step": 2197 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915089611406157e-05, + "loss": 1.0386, + "step": 2198 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914954863982106e-05, + "loss": 1.1294, + "step": 2199 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914820010181566e-05, + "loss": 1.1406, + "step": 2200 + }, + { + "epoch": 0.07, + "learning_rate": 1.991468505000598e-05, + "loss": 1.0884, + "step": 2201 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914549983456797e-05, + "loss": 1.1816, + "step": 2202 + }, + { + "epoch": 0.07, + "learning_rate": 1.991441481053546e-05, + "loss": 1.1655, + "step": 2203 + }, + { + "epoch": 0.07, + "learning_rate": 1.991427953124343e-05, + "loss": 1.1094, + "step": 2204 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914144145582155e-05, + "loss": 1.1831, + "step": 2205 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914008653553083e-05, + "loss": 1.1733, + "step": 2206 + }, + { + "epoch": 0.07, + "learning_rate": 1.991387305515767e-05, + "loss": 1.1475, + "step": 2207 + }, + { + "epoch": 0.07, + "learning_rate": 1.991373735039737e-05, + "loss": 1.0259, + "step": 2208 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913601539273644e-05, + "loss": 1.1865, + "step": 2209 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913465621787943e-05, + "loss": 1.0918, + "step": 2210 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913329597941727e-05, + "loss": 1.1484, + "step": 2211 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913193467736457e-05, + "loss": 1.1489, + "step": 2212 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913057231173594e-05, + "loss": 1.1152, + "step": 2213 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912920888254597e-05, + "loss": 1.2812, + "step": 2214 + }, + { + "epoch": 0.07, + "learning_rate": 1.991278443898093e-05, + "loss": 1.0811, + "step": 2215 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912647883354054e-05, + "loss": 1.1562, + "step": 2216 + }, + { + "epoch": 0.07, + "learning_rate": 1.991251122137544e-05, + "loss": 1.1191, + "step": 2217 + }, + { + "epoch": 0.07, + "learning_rate": 1.991237445304655e-05, + "loss": 1.2256, + "step": 2218 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912237578368855e-05, + "loss": 1.1396, + "step": 2219 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912100597343818e-05, + "loss": 1.1582, + "step": 2220 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911963509972914e-05, + "loss": 1.2207, + "step": 2221 + }, + { + "epoch": 0.07, + "learning_rate": 1.991182631625761e-05, + "loss": 1.1289, + "step": 2222 + }, + { + "epoch": 0.07, + "learning_rate": 1.991168901619938e-05, + "loss": 1.1523, + "step": 2223 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911551609799697e-05, + "loss": 1.1133, + "step": 2224 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911414097060034e-05, + "loss": 1.1919, + "step": 2225 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911276477981866e-05, + "loss": 1.0923, + "step": 2226 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911138752566673e-05, + "loss": 1.0581, + "step": 2227 + }, + { + "epoch": 0.07, + "learning_rate": 1.991100092081593e-05, + "loss": 1.1235, + "step": 2228 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910862982731114e-05, + "loss": 0.4734, + "step": 2229 + }, + { + "epoch": 0.07, + "learning_rate": 1.991072493831371e-05, + "loss": 1.1934, + "step": 2230 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910586787565194e-05, + "loss": 1.022, + "step": 2231 + }, + { + "epoch": 0.07, + "learning_rate": 1.991044853048705e-05, + "loss": 1.248, + "step": 2232 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910310167080764e-05, + "loss": 1.2139, + "step": 2233 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910171697347818e-05, + "loss": 1.1431, + "step": 2234 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910033121289695e-05, + "loss": 1.2432, + "step": 2235 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909894438907886e-05, + "loss": 1.1562, + "step": 2236 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909755650203877e-05, + "loss": 1.0918, + "step": 2237 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909616755179157e-05, + "loss": 0.9707, + "step": 2238 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909477753835217e-05, + "loss": 1.1182, + "step": 2239 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909338646173548e-05, + "loss": 1.1479, + "step": 2240 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909199432195644e-05, + "loss": 1.1982, + "step": 2241 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909060111902993e-05, + "loss": 1.188, + "step": 2242 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908920685297098e-05, + "loss": 1.1304, + "step": 2243 + }, + { + "epoch": 0.07, + "learning_rate": 1.990878115237945e-05, + "loss": 1.0693, + "step": 2244 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908641513151546e-05, + "loss": 1.0913, + "step": 2245 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908501767614884e-05, + "loss": 1.0854, + "step": 2246 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908361915770962e-05, + "loss": 1.2139, + "step": 2247 + }, + { + "epoch": 0.07, + "learning_rate": 1.990822195762129e-05, + "loss": 1.062, + "step": 2248 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908081893167357e-05, + "loss": 1.0425, + "step": 2249 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907941722410672e-05, + "loss": 1.1431, + "step": 2250 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907801445352738e-05, + "loss": 1.1025, + "step": 2251 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907661061995062e-05, + "loss": 1.0195, + "step": 2252 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907520572339143e-05, + "loss": 1.0786, + "step": 2253 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907379976386497e-05, + "loss": 1.022, + "step": 2254 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907239274138627e-05, + "loss": 1.2085, + "step": 2255 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907098465597046e-05, + "loss": 1.1323, + "step": 2256 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906957550763262e-05, + "loss": 1.0493, + "step": 2257 + }, + { + "epoch": 0.07, + "learning_rate": 1.990681652963879e-05, + "loss": 1.2031, + "step": 2258 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906675402225137e-05, + "loss": 1.2061, + "step": 2259 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906534168523825e-05, + "loss": 1.1338, + "step": 2260 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906392828536364e-05, + "loss": 1.1748, + "step": 2261 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906251382264274e-05, + "loss": 1.1147, + "step": 2262 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906109829709067e-05, + "loss": 1.2119, + "step": 2263 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905968170872266e-05, + "loss": 1.1123, + "step": 2264 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905826405755388e-05, + "loss": 0.5374, + "step": 2265 + }, + { + "epoch": 0.07, + "learning_rate": 1.990568453435996e-05, + "loss": 1.0605, + "step": 2266 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905542556687498e-05, + "loss": 1.1553, + "step": 2267 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905400472739528e-05, + "loss": 0.469, + "step": 2268 + }, + { + "epoch": 0.07, + "learning_rate": 1.990525828251757e-05, + "loss": 0.5073, + "step": 2269 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905115986023157e-05, + "loss": 1.2358, + "step": 2270 + }, + { + "epoch": 0.07, + "learning_rate": 1.990497358325781e-05, + "loss": 0.9985, + "step": 2271 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904831074223062e-05, + "loss": 1.1816, + "step": 2272 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904688458920435e-05, + "loss": 1.1167, + "step": 2273 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904545737351463e-05, + "loss": 1.1846, + "step": 2274 + }, + { + "epoch": 0.07, + "learning_rate": 1.990440290951768e-05, + "loss": 1.0845, + "step": 2275 + }, + { + "epoch": 0.07, + "learning_rate": 1.990425997542061e-05, + "loss": 1.1221, + "step": 2276 + }, + { + "epoch": 0.07, + "learning_rate": 1.99041169350618e-05, + "loss": 1.166, + "step": 2277 + }, + { + "epoch": 0.07, + "learning_rate": 1.990397378844277e-05, + "loss": 1.0381, + "step": 2278 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903830535565064e-05, + "loss": 1.1323, + "step": 2279 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903687176430222e-05, + "loss": 1.1953, + "step": 2280 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903543711039772e-05, + "loss": 1.0894, + "step": 2281 + }, + { + "epoch": 0.07, + "learning_rate": 1.990340013939526e-05, + "loss": 1.1704, + "step": 2282 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903256461498226e-05, + "loss": 1.0796, + "step": 2283 + }, + { + "epoch": 0.07, + "learning_rate": 1.990311267735021e-05, + "loss": 0.9849, + "step": 2284 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902968786952753e-05, + "loss": 1.126, + "step": 2285 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902824790307406e-05, + "loss": 1.0601, + "step": 2286 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902680687415704e-05, + "loss": 1.1465, + "step": 2287 + }, + { + "epoch": 0.07, + "learning_rate": 1.99025364782792e-05, + "loss": 1.0938, + "step": 2288 + }, + { + "epoch": 0.07, + "learning_rate": 1.990239216289944e-05, + "loss": 1.1279, + "step": 2289 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902247741277973e-05, + "loss": 1.2417, + "step": 2290 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902103213416343e-05, + "loss": 1.1313, + "step": 2291 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901958579316104e-05, + "loss": 1.2544, + "step": 2292 + }, + { + "epoch": 0.07, + "learning_rate": 1.990181383897881e-05, + "loss": 1.0527, + "step": 2293 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901668992406016e-05, + "loss": 1.1016, + "step": 2294 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901524039599266e-05, + "loss": 1.0698, + "step": 2295 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901378980560128e-05, + "loss": 1.0757, + "step": 2296 + }, + { + "epoch": 0.07, + "learning_rate": 1.990123381529015e-05, + "loss": 1.2686, + "step": 2297 + }, + { + "epoch": 0.07, + "learning_rate": 1.990108854379089e-05, + "loss": 1.1763, + "step": 2298 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900943166063905e-05, + "loss": 1.0288, + "step": 2299 + }, + { + "epoch": 0.07, + "learning_rate": 1.990079768211076e-05, + "loss": 1.167, + "step": 2300 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900652091933015e-05, + "loss": 1.1924, + "step": 2301 + }, + { + "epoch": 0.07, + "learning_rate": 1.990050639553223e-05, + "loss": 1.0967, + "step": 2302 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900360592909965e-05, + "loss": 1.1084, + "step": 2303 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900214684067794e-05, + "loss": 1.1597, + "step": 2304 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900068669007274e-05, + "loss": 1.1182, + "step": 2305 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899922547729972e-05, + "loss": 1.126, + "step": 2306 + }, + { + "epoch": 0.07, + "learning_rate": 1.989977632023746e-05, + "loss": 1.1284, + "step": 2307 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899629986531302e-05, + "loss": 1.103, + "step": 2308 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899483546613077e-05, + "loss": 1.1172, + "step": 2309 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899337000484343e-05, + "loss": 1.1426, + "step": 2310 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899190348146683e-05, + "loss": 1.0615, + "step": 2311 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899043589601662e-05, + "loss": 1.144, + "step": 2312 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898896724850863e-05, + "loss": 1.0366, + "step": 2313 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898749753895858e-05, + "loss": 1.1084, + "step": 2314 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898602676738222e-05, + "loss": 1.0698, + "step": 2315 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898455493379534e-05, + "loss": 1.064, + "step": 2316 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898308203821376e-05, + "loss": 1.1621, + "step": 2317 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898160808065325e-05, + "loss": 1.1836, + "step": 2318 + }, + { + "epoch": 0.07, + "learning_rate": 1.989801330611296e-05, + "loss": 1.147, + "step": 2319 + }, + { + "epoch": 0.07, + "learning_rate": 1.989786569796587e-05, + "loss": 1.1011, + "step": 2320 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897717983625634e-05, + "loss": 1.1533, + "step": 2321 + }, + { + "epoch": 0.07, + "learning_rate": 1.989757016309384e-05, + "loss": 1.2349, + "step": 2322 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897422236372073e-05, + "loss": 1.0679, + "step": 2323 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897274203461917e-05, + "loss": 1.0767, + "step": 2324 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897126064364964e-05, + "loss": 1.0005, + "step": 2325 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896977819082798e-05, + "loss": 1.0645, + "step": 2326 + }, + { + "epoch": 0.07, + "learning_rate": 1.989682946761702e-05, + "loss": 1.2095, + "step": 2327 + }, + { + "epoch": 0.07, + "learning_rate": 1.989668100996921e-05, + "loss": 1.1387, + "step": 2328 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896532446140967e-05, + "loss": 1.0718, + "step": 2329 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896383776133886e-05, + "loss": 1.2368, + "step": 2330 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896234999949558e-05, + "loss": 1.1465, + "step": 2331 + }, + { + "epoch": 0.07, + "learning_rate": 1.989608611758958e-05, + "loss": 1.1597, + "step": 2332 + }, + { + "epoch": 0.07, + "learning_rate": 1.989593712905555e-05, + "loss": 1.1357, + "step": 2333 + }, + { + "epoch": 0.07, + "learning_rate": 1.989578803434907e-05, + "loss": 1.0908, + "step": 2334 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895638833471733e-05, + "loss": 1.1396, + "step": 2335 + }, + { + "epoch": 0.07, + "learning_rate": 1.989548952642515e-05, + "loss": 1.0278, + "step": 2336 + }, + { + "epoch": 0.07, + "learning_rate": 1.989534011321091e-05, + "loss": 1.0254, + "step": 2337 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895190593830623e-05, + "loss": 1.1719, + "step": 2338 + }, + { + "epoch": 0.07, + "learning_rate": 1.989504096828589e-05, + "loss": 1.1265, + "step": 2339 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894891236578325e-05, + "loss": 1.1812, + "step": 2340 + }, + { + "epoch": 0.07, + "learning_rate": 1.989474139870952e-05, + "loss": 1.0625, + "step": 2341 + }, + { + "epoch": 0.07, + "learning_rate": 1.98945914546811e-05, + "loss": 0.9731, + "step": 2342 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894441404494657e-05, + "loss": 1.0776, + "step": 2343 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894291248151813e-05, + "loss": 1.1602, + "step": 2344 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894140985654174e-05, + "loss": 1.1226, + "step": 2345 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893990617003348e-05, + "loss": 1.1616, + "step": 2346 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893840142200957e-05, + "loss": 1.1191, + "step": 2347 + }, + { + "epoch": 0.08, + "learning_rate": 1.989368956124861e-05, + "loss": 1.2554, + "step": 2348 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893538874147928e-05, + "loss": 1.0562, + "step": 2349 + }, + { + "epoch": 0.08, + "learning_rate": 1.989338808090052e-05, + "loss": 1.2188, + "step": 2350 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893237181508006e-05, + "loss": 1.165, + "step": 2351 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893086175972006e-05, + "loss": 0.9966, + "step": 2352 + }, + { + "epoch": 0.08, + "learning_rate": 1.989293506429414e-05, + "loss": 1.1177, + "step": 2353 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892783846476034e-05, + "loss": 1.1992, + "step": 2354 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892632522519305e-05, + "loss": 1.2007, + "step": 2355 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892481092425573e-05, + "loss": 0.96, + "step": 2356 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892329556196472e-05, + "loss": 1.1279, + "step": 2357 + }, + { + "epoch": 0.08, + "learning_rate": 1.989217791383362e-05, + "loss": 0.9956, + "step": 2358 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892026165338652e-05, + "loss": 1.1636, + "step": 2359 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891874310713187e-05, + "loss": 1.1021, + "step": 2360 + }, + { + "epoch": 0.08, + "learning_rate": 1.989172234995886e-05, + "loss": 1.0996, + "step": 2361 + }, + { + "epoch": 0.08, + "learning_rate": 1.98915702830773e-05, + "loss": 1.0229, + "step": 2362 + }, + { + "epoch": 0.08, + "learning_rate": 1.989141811007014e-05, + "loss": 1.1069, + "step": 2363 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891265830939008e-05, + "loss": 1.1953, + "step": 2364 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891113445685537e-05, + "loss": 1.167, + "step": 2365 + }, + { + "epoch": 0.08, + "learning_rate": 1.989096095431137e-05, + "loss": 1.1504, + "step": 2366 + }, + { + "epoch": 0.08, + "learning_rate": 1.989080835681814e-05, + "loss": 1.2295, + "step": 2367 + }, + { + "epoch": 0.08, + "learning_rate": 1.989065565320748e-05, + "loss": 0.9863, + "step": 2368 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890502843481034e-05, + "loss": 1.1582, + "step": 2369 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890349927640437e-05, + "loss": 1.1982, + "step": 2370 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890196905687333e-05, + "loss": 1.1636, + "step": 2371 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890043777623363e-05, + "loss": 1.2231, + "step": 2372 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889890543450166e-05, + "loss": 1.0889, + "step": 2373 + }, + { + "epoch": 0.08, + "learning_rate": 1.988973720316939e-05, + "loss": 1.0244, + "step": 2374 + }, + { + "epoch": 0.08, + "learning_rate": 1.988958375678268e-05, + "loss": 1.0601, + "step": 2375 + }, + { + "epoch": 0.08, + "learning_rate": 1.988943020429168e-05, + "loss": 1.1899, + "step": 2376 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889276545698043e-05, + "loss": 1.0869, + "step": 2377 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889122781003408e-05, + "loss": 1.1572, + "step": 2378 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888968910209433e-05, + "loss": 1.1836, + "step": 2379 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888814933317768e-05, + "loss": 1.1006, + "step": 2380 + }, + { + "epoch": 0.08, + "learning_rate": 1.988866085033006e-05, + "loss": 1.0801, + "step": 2381 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888506661247967e-05, + "loss": 1.1436, + "step": 2382 + }, + { + "epoch": 0.08, + "learning_rate": 1.988835236607314e-05, + "loss": 1.0742, + "step": 2383 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888197964807236e-05, + "loss": 1.0869, + "step": 2384 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888043457451915e-05, + "loss": 1.1128, + "step": 2385 + }, + { + "epoch": 0.08, + "learning_rate": 1.988788884400883e-05, + "loss": 1.1362, + "step": 2386 + }, + { + "epoch": 0.08, + "learning_rate": 1.988773412447964e-05, + "loss": 1.0938, + "step": 2387 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887579298866006e-05, + "loss": 1.1841, + "step": 2388 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887424367169587e-05, + "loss": 1.1914, + "step": 2389 + }, + { + "epoch": 0.08, + "learning_rate": 1.988726932939205e-05, + "loss": 0.4304, + "step": 2390 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887114185535055e-05, + "loss": 1.2808, + "step": 2391 + }, + { + "epoch": 0.08, + "learning_rate": 1.988695893560027e-05, + "loss": 0.9458, + "step": 2392 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886803579589354e-05, + "loss": 1.1943, + "step": 2393 + }, + { + "epoch": 0.08, + "learning_rate": 1.988664811750398e-05, + "loss": 1.1182, + "step": 2394 + }, + { + "epoch": 0.08, + "learning_rate": 1.988649254934581e-05, + "loss": 1.1924, + "step": 2395 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886336875116523e-05, + "loss": 1.1289, + "step": 2396 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886181094817777e-05, + "loss": 1.1289, + "step": 2397 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886025208451255e-05, + "loss": 1.0454, + "step": 2398 + }, + { + "epoch": 0.08, + "learning_rate": 1.988586921601862e-05, + "loss": 1.1548, + "step": 2399 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885713117521553e-05, + "loss": 1.0859, + "step": 2400 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885556912961717e-05, + "loss": 1.2266, + "step": 2401 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885400602340807e-05, + "loss": 1.0723, + "step": 2402 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885244185660482e-05, + "loss": 1.1553, + "step": 2403 + }, + { + "epoch": 0.08, + "learning_rate": 1.988508766292243e-05, + "loss": 1.165, + "step": 2404 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884931034128328e-05, + "loss": 1.1816, + "step": 2405 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884774299279853e-05, + "loss": 1.2568, + "step": 2406 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884617458378692e-05, + "loss": 1.083, + "step": 2407 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884460511426526e-05, + "loss": 1.04, + "step": 2408 + }, + { + "epoch": 0.08, + "learning_rate": 1.988430345842504e-05, + "loss": 1.0464, + "step": 2409 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884146299375912e-05, + "loss": 1.229, + "step": 2410 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883989034280838e-05, + "loss": 1.084, + "step": 2411 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883831663141497e-05, + "loss": 1.1367, + "step": 2412 + }, + { + "epoch": 0.08, + "learning_rate": 1.988367418595958e-05, + "loss": 1.1255, + "step": 2413 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883516602736784e-05, + "loss": 1.1025, + "step": 2414 + }, + { + "epoch": 0.08, + "learning_rate": 1.988335891347479e-05, + "loss": 1.1108, + "step": 2415 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883201118175295e-05, + "loss": 0.9932, + "step": 2416 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883043216839984e-05, + "loss": 1.1167, + "step": 2417 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882885209470565e-05, + "loss": 1.0776, + "step": 2418 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882727096068724e-05, + "loss": 1.2061, + "step": 2419 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882568876636153e-05, + "loss": 1.1631, + "step": 2420 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882410551174563e-05, + "loss": 1.0713, + "step": 2421 + }, + { + "epoch": 0.08, + "learning_rate": 1.988225211968564e-05, + "loss": 1.0913, + "step": 2422 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882093582171093e-05, + "loss": 1.0659, + "step": 2423 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881934938632615e-05, + "loss": 1.1421, + "step": 2424 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881776189071914e-05, + "loss": 1.1709, + "step": 2425 + }, + { + "epoch": 0.08, + "learning_rate": 1.988161733349069e-05, + "loss": 1.064, + "step": 2426 + }, + { + "epoch": 0.08, + "learning_rate": 1.988145837189065e-05, + "loss": 1.2222, + "step": 2427 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881299304273496e-05, + "loss": 1.1201, + "step": 2428 + }, + { + "epoch": 0.08, + "learning_rate": 1.988114013064094e-05, + "loss": 1.2231, + "step": 2429 + }, + { + "epoch": 0.08, + "learning_rate": 1.988098085099468e-05, + "loss": 1.0449, + "step": 2430 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880821465336434e-05, + "loss": 1.1016, + "step": 2431 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880661973667914e-05, + "loss": 0.9878, + "step": 2432 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880502375990823e-05, + "loss": 1.3379, + "step": 2433 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880342672306874e-05, + "loss": 1.0254, + "step": 2434 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880182862617788e-05, + "loss": 1.1099, + "step": 2435 + }, + { + "epoch": 0.08, + "learning_rate": 1.988002294692527e-05, + "loss": 1.1113, + "step": 2436 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879862925231044e-05, + "loss": 1.0864, + "step": 2437 + }, + { + "epoch": 0.08, + "learning_rate": 1.987970279753682e-05, + "loss": 0.427, + "step": 2438 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879542563844324e-05, + "loss": 1.0679, + "step": 2439 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879382224155264e-05, + "loss": 1.1235, + "step": 2440 + }, + { + "epoch": 0.08, + "learning_rate": 1.987922177847137e-05, + "loss": 1.0913, + "step": 2441 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879061226794363e-05, + "loss": 1.0718, + "step": 2442 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878900569125957e-05, + "loss": 1.2314, + "step": 2443 + }, + { + "epoch": 0.08, + "learning_rate": 1.987873980546789e-05, + "loss": 1.0713, + "step": 2444 + }, + { + "epoch": 0.08, + "learning_rate": 1.987857893582187e-05, + "loss": 1.1343, + "step": 2445 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878417960189633e-05, + "loss": 1.1587, + "step": 2446 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878256878572902e-05, + "loss": 1.0669, + "step": 2447 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878095690973413e-05, + "loss": 1.1748, + "step": 2448 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877934397392888e-05, + "loss": 1.1382, + "step": 2449 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877772997833057e-05, + "loss": 1.1436, + "step": 2450 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877611492295655e-05, + "loss": 1.2314, + "step": 2451 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877449880782414e-05, + "loss": 1.1055, + "step": 2452 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877288163295068e-05, + "loss": 1.1167, + "step": 2453 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877126339835352e-05, + "loss": 1.1689, + "step": 2454 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876964410405e-05, + "loss": 0.9092, + "step": 2455 + }, + { + "epoch": 0.08, + "learning_rate": 1.987680237500575e-05, + "loss": 1.1338, + "step": 2456 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876640233639346e-05, + "loss": 1.1318, + "step": 2457 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876477986307518e-05, + "loss": 1.165, + "step": 2458 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876315633012017e-05, + "loss": 1.1172, + "step": 2459 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876153173754578e-05, + "loss": 1.085, + "step": 2460 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875990608536947e-05, + "loss": 1.2227, + "step": 2461 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875827937360864e-05, + "loss": 1.1685, + "step": 2462 + }, + { + "epoch": 0.08, + "learning_rate": 1.987566516022808e-05, + "loss": 1.0664, + "step": 2463 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875502277140336e-05, + "loss": 1.0708, + "step": 2464 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875339288099384e-05, + "loss": 0.9863, + "step": 2465 + }, + { + "epoch": 0.08, + "learning_rate": 1.987517619310697e-05, + "loss": 1.0869, + "step": 2466 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875012992164844e-05, + "loss": 1.1562, + "step": 2467 + }, + { + "epoch": 0.08, + "learning_rate": 1.987484968527476e-05, + "loss": 1.1665, + "step": 2468 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874686272438467e-05, + "loss": 1.0923, + "step": 2469 + }, + { + "epoch": 0.08, + "learning_rate": 1.987452275365772e-05, + "loss": 1.1147, + "step": 2470 + }, + { + "epoch": 0.08, + "learning_rate": 1.987435912893427e-05, + "loss": 1.1006, + "step": 2471 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874195398269877e-05, + "loss": 0.9961, + "step": 2472 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874031561666295e-05, + "loss": 1.1104, + "step": 2473 + }, + { + "epoch": 0.08, + "learning_rate": 1.987386761912528e-05, + "loss": 1.2417, + "step": 2474 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873703570648598e-05, + "loss": 1.1953, + "step": 2475 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873539416238003e-05, + "loss": 1.2046, + "step": 2476 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873375155895255e-05, + "loss": 1.1797, + "step": 2477 + }, + { + "epoch": 0.08, + "learning_rate": 1.987321078962212e-05, + "loss": 1.0713, + "step": 2478 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873046317420362e-05, + "loss": 0.8433, + "step": 2479 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872881739291744e-05, + "loss": 1.0308, + "step": 2480 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872717055238027e-05, + "loss": 1.1113, + "step": 2481 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872552265260987e-05, + "loss": 1.1206, + "step": 2482 + }, + { + "epoch": 0.08, + "learning_rate": 1.987238736936239e-05, + "loss": 0.9575, + "step": 2483 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872222367543997e-05, + "loss": 1.2119, + "step": 2484 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872057259807588e-05, + "loss": 1.0938, + "step": 2485 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871892046154927e-05, + "loss": 1.1006, + "step": 2486 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871726726587792e-05, + "loss": 1.064, + "step": 2487 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871561301107952e-05, + "loss": 1.1685, + "step": 2488 + }, + { + "epoch": 0.08, + "learning_rate": 1.987139576971719e-05, + "loss": 1.1401, + "step": 2489 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871230132417273e-05, + "loss": 1.1172, + "step": 2490 + }, + { + "epoch": 0.08, + "learning_rate": 1.987106438920998e-05, + "loss": 1.0894, + "step": 2491 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870898540097095e-05, + "loss": 1.103, + "step": 2492 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870732585080392e-05, + "loss": 1.1616, + "step": 2493 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870566524161653e-05, + "loss": 1.1064, + "step": 2494 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870400357342657e-05, + "loss": 1.1431, + "step": 2495 + }, + { + "epoch": 0.08, + "learning_rate": 1.987023408462519e-05, + "loss": 1.0864, + "step": 2496 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870067706011036e-05, + "loss": 1.1782, + "step": 2497 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869901221501976e-05, + "loss": 1.1572, + "step": 2498 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869734631099807e-05, + "loss": 1.1089, + "step": 2499 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869567934806303e-05, + "loss": 1.1729, + "step": 2500 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869401132623258e-05, + "loss": 1.2388, + "step": 2501 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869234224552462e-05, + "loss": 1.0933, + "step": 2502 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869067210595707e-05, + "loss": 1.1421, + "step": 2503 + }, + { + "epoch": 0.08, + "learning_rate": 1.986890009075478e-05, + "loss": 0.9419, + "step": 2504 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868732865031482e-05, + "loss": 1.2705, + "step": 2505 + }, + { + "epoch": 0.08, + "learning_rate": 1.98685655334276e-05, + "loss": 1.0049, + "step": 2506 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868398095944935e-05, + "loss": 1.1548, + "step": 2507 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868230552585277e-05, + "loss": 1.2031, + "step": 2508 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868062903350426e-05, + "loss": 0.9961, + "step": 2509 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867895148242183e-05, + "loss": 1.1992, + "step": 2510 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867727287262347e-05, + "loss": 1.0532, + "step": 2511 + }, + { + "epoch": 0.08, + "learning_rate": 1.986755932041272e-05, + "loss": 1.0298, + "step": 2512 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867391247695098e-05, + "loss": 1.0356, + "step": 2513 + }, + { + "epoch": 0.08, + "learning_rate": 1.986722306911129e-05, + "loss": 1.0776, + "step": 2514 + }, + { + "epoch": 0.08, + "learning_rate": 1.98670547846631e-05, + "loss": 1.1826, + "step": 2515 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866886394352333e-05, + "loss": 1.1494, + "step": 2516 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866717898180795e-05, + "loss": 1.1265, + "step": 2517 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866549296150292e-05, + "loss": 1.1304, + "step": 2518 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866380588262636e-05, + "loss": 1.2515, + "step": 2519 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866211774519635e-05, + "loss": 1.1558, + "step": 2520 + }, + { + "epoch": 0.08, + "learning_rate": 1.98660428549231e-05, + "loss": 0.9946, + "step": 2521 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865873829474847e-05, + "loss": 1.2236, + "step": 2522 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865704698176684e-05, + "loss": 1.1953, + "step": 2523 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865535461030432e-05, + "loss": 1.1572, + "step": 2524 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865366118037902e-05, + "loss": 1.1719, + "step": 2525 + }, + { + "epoch": 0.08, + "learning_rate": 1.986519666920091e-05, + "loss": 1.0898, + "step": 2526 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865027114521278e-05, + "loss": 1.2217, + "step": 2527 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864857454000823e-05, + "loss": 1.0942, + "step": 2528 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864687687641365e-05, + "loss": 1.0059, + "step": 2529 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864517815444726e-05, + "loss": 1.2891, + "step": 2530 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864347837412727e-05, + "loss": 1.1597, + "step": 2531 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864177753547197e-05, + "loss": 1.0415, + "step": 2532 + }, + { + "epoch": 0.08, + "learning_rate": 1.986400756384995e-05, + "loss": 1.209, + "step": 2533 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863837268322825e-05, + "loss": 1.0713, + "step": 2534 + }, + { + "epoch": 0.08, + "learning_rate": 1.986366686696764e-05, + "loss": 1.0894, + "step": 2535 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863496359786226e-05, + "loss": 0.4614, + "step": 2536 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863325746780413e-05, + "loss": 1.1187, + "step": 2537 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863155027952033e-05, + "loss": 1.2031, + "step": 2538 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862984203302912e-05, + "loss": 1.0127, + "step": 2539 + }, + { + "epoch": 0.08, + "learning_rate": 1.986281327283489e-05, + "loss": 1.2271, + "step": 2540 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862642236549794e-05, + "loss": 1.0957, + "step": 2541 + }, + { + "epoch": 0.08, + "learning_rate": 1.986247109444946e-05, + "loss": 1.1191, + "step": 2542 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862299846535734e-05, + "loss": 1.2217, + "step": 2543 + }, + { + "epoch": 0.08, + "learning_rate": 1.986212849281044e-05, + "loss": 1.1392, + "step": 2544 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861957033275425e-05, + "loss": 1.0737, + "step": 2545 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861785467932522e-05, + "loss": 1.2388, + "step": 2546 + }, + { + "epoch": 0.08, + "learning_rate": 1.986161379678358e-05, + "loss": 1.0688, + "step": 2547 + }, + { + "epoch": 0.08, + "learning_rate": 1.986144201983043e-05, + "loss": 1.167, + "step": 2548 + }, + { + "epoch": 0.08, + "learning_rate": 1.986127013707493e-05, + "loss": 1.0444, + "step": 2549 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861098148518906e-05, + "loss": 1.084, + "step": 2550 + }, + { + "epoch": 0.08, + "learning_rate": 1.986092605416422e-05, + "loss": 1.332, + "step": 2551 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860753854012705e-05, + "loss": 1.1333, + "step": 2552 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860581548066216e-05, + "loss": 1.0454, + "step": 2553 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860409136326604e-05, + "loss": 1.1113, + "step": 2554 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860236618795713e-05, + "loss": 1.1074, + "step": 2555 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860063995475395e-05, + "loss": 1.1187, + "step": 2556 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859891266367503e-05, + "loss": 1.1201, + "step": 2557 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859718431473892e-05, + "loss": 1.0801, + "step": 2558 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859545490796414e-05, + "loss": 1.1118, + "step": 2559 + }, + { + "epoch": 0.08, + "learning_rate": 1.985937244433692e-05, + "loss": 1.063, + "step": 2560 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859199292097277e-05, + "loss": 1.0908, + "step": 2561 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859026034079336e-05, + "loss": 1.021, + "step": 2562 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858852670284956e-05, + "loss": 1.0688, + "step": 2563 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858679200716e-05, + "loss": 1.0806, + "step": 2564 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858505625374325e-05, + "loss": 1.2148, + "step": 2565 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858331944261797e-05, + "loss": 1.1689, + "step": 2566 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858158157380278e-05, + "loss": 1.0791, + "step": 2567 + }, + { + "epoch": 0.08, + "learning_rate": 1.985798426473163e-05, + "loss": 1.0913, + "step": 2568 + }, + { + "epoch": 0.08, + "learning_rate": 1.985781026631772e-05, + "loss": 1.0742, + "step": 2569 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857636162140422e-05, + "loss": 1.1162, + "step": 2570 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857461952201596e-05, + "loss": 1.1426, + "step": 2571 + }, + { + "epoch": 0.08, + "learning_rate": 1.985728763650311e-05, + "loss": 1.0972, + "step": 2572 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857113215046838e-05, + "loss": 1.2148, + "step": 2573 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856938687834653e-05, + "loss": 0.9771, + "step": 2574 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856764054868424e-05, + "loss": 1.1792, + "step": 2575 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856589316150022e-05, + "loss": 1.2168, + "step": 2576 + }, + { + "epoch": 0.08, + "learning_rate": 1.985641447168133e-05, + "loss": 1.2026, + "step": 2577 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856239521464218e-05, + "loss": 1.2451, + "step": 2578 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856064465500566e-05, + "loss": 1.1431, + "step": 2579 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855889303792243e-05, + "loss": 0.9536, + "step": 2580 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855714036341144e-05, + "loss": 1.2202, + "step": 2581 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855538663149135e-05, + "loss": 1.1924, + "step": 2582 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855363184218108e-05, + "loss": 1.0112, + "step": 2583 + }, + { + "epoch": 0.08, + "learning_rate": 1.985518759954994e-05, + "loss": 1.1719, + "step": 2584 + }, + { + "epoch": 0.08, + "learning_rate": 1.985501190914652e-05, + "loss": 1.1953, + "step": 2585 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854836113009723e-05, + "loss": 1.127, + "step": 2586 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854660211141446e-05, + "loss": 1.2158, + "step": 2587 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854484203543566e-05, + "loss": 1.1611, + "step": 2588 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854308090217983e-05, + "loss": 1.0581, + "step": 2589 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854131871166578e-05, + "loss": 1.125, + "step": 2590 + }, + { + "epoch": 0.08, + "learning_rate": 1.985395554639124e-05, + "loss": 1.1782, + "step": 2591 + }, + { + "epoch": 0.08, + "learning_rate": 1.985377911589387e-05, + "loss": 1.1885, + "step": 2592 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853602579676353e-05, + "loss": 1.1558, + "step": 2593 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853425937740588e-05, + "loss": 0.9956, + "step": 2594 + }, + { + "epoch": 0.08, + "learning_rate": 1.985324919008847e-05, + "loss": 1.064, + "step": 2595 + }, + { + "epoch": 0.08, + "learning_rate": 1.985307233672189e-05, + "loss": 1.0869, + "step": 2596 + }, + { + "epoch": 0.08, + "learning_rate": 1.985289537764275e-05, + "loss": 1.0991, + "step": 2597 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852718312852944e-05, + "loss": 0.9702, + "step": 2598 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852541142354375e-05, + "loss": 1.1748, + "step": 2599 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852363866148948e-05, + "loss": 1.1445, + "step": 2600 + }, + { + "epoch": 0.08, + "learning_rate": 1.985218648423856e-05, + "loss": 1.1909, + "step": 2601 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852008996625113e-05, + "loss": 1.0269, + "step": 2602 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851831403310513e-05, + "loss": 0.9985, + "step": 2603 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851653704296664e-05, + "loss": 1.2183, + "step": 2604 + }, + { + "epoch": 0.08, + "learning_rate": 1.985147589958548e-05, + "loss": 1.0884, + "step": 2605 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851297989178858e-05, + "loss": 1.1094, + "step": 2606 + }, + { + "epoch": 0.08, + "learning_rate": 1.985111997307871e-05, + "loss": 1.0146, + "step": 2607 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850941851286952e-05, + "loss": 0.4949, + "step": 2608 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850763623805486e-05, + "loss": 0.9946, + "step": 2609 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850585290636233e-05, + "loss": 1.1514, + "step": 2610 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850406851781097e-05, + "loss": 1.1299, + "step": 2611 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850228307242e-05, + "loss": 1.1323, + "step": 2612 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850049657020857e-05, + "loss": 1.0068, + "step": 2613 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849870901119578e-05, + "loss": 1.0112, + "step": 2614 + }, + { + "epoch": 0.08, + "learning_rate": 1.984969203954009e-05, + "loss": 1.0645, + "step": 2615 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849513072284307e-05, + "loss": 1.0806, + "step": 2616 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849333999354147e-05, + "loss": 1.1543, + "step": 2617 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849154820751535e-05, + "loss": 1.0986, + "step": 2618 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848975536478394e-05, + "loss": 1.1133, + "step": 2619 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848796146536643e-05, + "loss": 1.1089, + "step": 2620 + }, + { + "epoch": 0.08, + "learning_rate": 1.984861665092821e-05, + "loss": 1.1294, + "step": 2621 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848437049655022e-05, + "loss": 1.1699, + "step": 2622 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848257342719e-05, + "loss": 1.2129, + "step": 2623 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848077530122083e-05, + "loss": 1.0625, + "step": 2624 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847897611866188e-05, + "loss": 1.0166, + "step": 2625 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847717587953253e-05, + "loss": 1.0913, + "step": 2626 + }, + { + "epoch": 0.08, + "learning_rate": 1.984753745838521e-05, + "loss": 0.478, + "step": 2627 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847357223163984e-05, + "loss": 1.0747, + "step": 2628 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847176882291513e-05, + "loss": 1.0557, + "step": 2629 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846996435769738e-05, + "loss": 1.0464, + "step": 2630 + }, + { + "epoch": 0.08, + "learning_rate": 1.984681588360058e-05, + "loss": 1.1182, + "step": 2631 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846635225785997e-05, + "loss": 1.0737, + "step": 2632 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846454462327907e-05, + "loss": 1.0391, + "step": 2633 + }, + { + "epoch": 0.08, + "learning_rate": 1.984627359322826e-05, + "loss": 1.1528, + "step": 2634 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846092618488995e-05, + "loss": 1.1255, + "step": 2635 + }, + { + "epoch": 0.08, + "learning_rate": 1.984591153811206e-05, + "loss": 1.0918, + "step": 2636 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845730352099383e-05, + "loss": 1.2041, + "step": 2637 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845549060452917e-05, + "loss": 1.126, + "step": 2638 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845367663174605e-05, + "loss": 0.8955, + "step": 2639 + }, + { + "epoch": 0.08, + "learning_rate": 1.98451861602664e-05, + "loss": 1.0718, + "step": 2640 + }, + { + "epoch": 0.08, + "learning_rate": 1.984500455173024e-05, + "loss": 1.1333, + "step": 2641 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844822837568076e-05, + "loss": 1.2954, + "step": 2642 + }, + { + "epoch": 0.08, + "learning_rate": 1.984464101778186e-05, + "loss": 1.1787, + "step": 2643 + }, + { + "epoch": 0.08, + "learning_rate": 1.984445909237354e-05, + "loss": 1.0298, + "step": 2644 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844277061345073e-05, + "loss": 1.0459, + "step": 2645 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844094924698403e-05, + "loss": 1.0864, + "step": 2646 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843912682435493e-05, + "loss": 1.1792, + "step": 2647 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843730334558295e-05, + "loss": 1.1118, + "step": 2648 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843547881068763e-05, + "loss": 1.0703, + "step": 2649 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843365321968856e-05, + "loss": 1.1064, + "step": 2650 + }, + { + "epoch": 0.08, + "learning_rate": 1.984318265726054e-05, + "loss": 1.1499, + "step": 2651 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842999886945757e-05, + "loss": 1.2002, + "step": 2652 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842817011026488e-05, + "loss": 1.0645, + "step": 2653 + }, + { + "epoch": 0.08, + "learning_rate": 1.984263402950468e-05, + "loss": 1.2261, + "step": 2654 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842450942382306e-05, + "loss": 1.188, + "step": 2655 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842267749661324e-05, + "loss": 1.166, + "step": 2656 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842084451343705e-05, + "loss": 1.0186, + "step": 2657 + }, + { + "epoch": 0.09, + "learning_rate": 1.984190104743141e-05, + "loss": 1.1177, + "step": 2658 + }, + { + "epoch": 0.09, + "learning_rate": 1.984171753792641e-05, + "loss": 1.1631, + "step": 2659 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841533922830673e-05, + "loss": 1.0923, + "step": 2660 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841350202146167e-05, + "loss": 1.0415, + "step": 2661 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841166375874868e-05, + "loss": 1.1221, + "step": 2662 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840982444018743e-05, + "loss": 1.0586, + "step": 2663 + }, + { + "epoch": 0.09, + "learning_rate": 1.984079840657977e-05, + "loss": 1.1367, + "step": 2664 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840614263559922e-05, + "loss": 1.0913, + "step": 2665 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840430014961173e-05, + "loss": 1.1929, + "step": 2666 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840245660785496e-05, + "loss": 1.3628, + "step": 2667 + }, + { + "epoch": 0.09, + "learning_rate": 1.984006120103488e-05, + "loss": 1.1021, + "step": 2668 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839876635711296e-05, + "loss": 0.9424, + "step": 2669 + }, + { + "epoch": 0.09, + "learning_rate": 1.983969196481673e-05, + "loss": 1.2046, + "step": 2670 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839507188353152e-05, + "loss": 1.2285, + "step": 2671 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839322306322555e-05, + "loss": 1.1929, + "step": 2672 + }, + { + "epoch": 0.09, + "learning_rate": 1.983913731872692e-05, + "loss": 1.1406, + "step": 2673 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838952225568228e-05, + "loss": 1.1826, + "step": 2674 + }, + { + "epoch": 0.09, + "learning_rate": 1.983876702684847e-05, + "loss": 1.1655, + "step": 2675 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838581722569633e-05, + "loss": 1.0928, + "step": 2676 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838396312733697e-05, + "loss": 1.1704, + "step": 2677 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838210797342664e-05, + "loss": 1.104, + "step": 2678 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838025176398515e-05, + "loss": 1.0957, + "step": 2679 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837839449903242e-05, + "loss": 1.0283, + "step": 2680 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837653617858842e-05, + "loss": 1.1704, + "step": 2681 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837467680267302e-05, + "loss": 1.0684, + "step": 2682 + }, + { + "epoch": 0.09, + "learning_rate": 1.983728163713063e-05, + "loss": 0.4453, + "step": 2683 + }, + { + "epoch": 0.09, + "learning_rate": 1.983709548845081e-05, + "loss": 1.168, + "step": 2684 + }, + { + "epoch": 0.09, + "learning_rate": 1.983690923422984e-05, + "loss": 1.0317, + "step": 2685 + }, + { + "epoch": 0.09, + "learning_rate": 1.983672287446972e-05, + "loss": 1.0742, + "step": 2686 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836536409172453e-05, + "loss": 1.1714, + "step": 2687 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836349838340034e-05, + "loss": 1.1279, + "step": 2688 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836163161974472e-05, + "loss": 1.1851, + "step": 2689 + }, + { + "epoch": 0.09, + "learning_rate": 1.983597638007776e-05, + "loss": 1.1562, + "step": 2690 + }, + { + "epoch": 0.09, + "learning_rate": 1.983578949265191e-05, + "loss": 1.1104, + "step": 2691 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835602499698928e-05, + "loss": 1.021, + "step": 2692 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835415401220812e-05, + "loss": 1.1587, + "step": 2693 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835228197219573e-05, + "loss": 1.123, + "step": 2694 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835040887697222e-05, + "loss": 1.1328, + "step": 2695 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834853472655772e-05, + "loss": 1.0654, + "step": 2696 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834665952097223e-05, + "loss": 1.1885, + "step": 2697 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834478326023596e-05, + "loss": 1.2021, + "step": 2698 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834290594436897e-05, + "loss": 1.146, + "step": 2699 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834102757339147e-05, + "loss": 1.0059, + "step": 2700 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833914814732357e-05, + "loss": 1.0425, + "step": 2701 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833726766618547e-05, + "loss": 0.5068, + "step": 2702 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833538612999732e-05, + "loss": 1.1377, + "step": 2703 + }, + { + "epoch": 0.09, + "learning_rate": 1.983335035387793e-05, + "loss": 1.1255, + "step": 2704 + }, + { + "epoch": 0.09, + "learning_rate": 1.983316198925516e-05, + "loss": 1.0483, + "step": 2705 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832973519133447e-05, + "loss": 1.0518, + "step": 2706 + }, + { + "epoch": 0.09, + "learning_rate": 1.983278494351481e-05, + "loss": 1.0737, + "step": 2707 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832596262401272e-05, + "loss": 1.0747, + "step": 2708 + }, + { + "epoch": 0.09, + "learning_rate": 1.983240747579486e-05, + "loss": 1.1396, + "step": 2709 + }, + { + "epoch": 0.09, + "learning_rate": 1.98322185836976e-05, + "loss": 1.0352, + "step": 2710 + }, + { + "epoch": 0.09, + "learning_rate": 1.983202958611151e-05, + "loss": 1.2129, + "step": 2711 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831840483038626e-05, + "loss": 1.0981, + "step": 2712 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831651274480977e-05, + "loss": 1.21, + "step": 2713 + }, + { + "epoch": 0.09, + "learning_rate": 1.983146196044059e-05, + "loss": 1.0571, + "step": 2714 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831272540919497e-05, + "loss": 1.0879, + "step": 2715 + }, + { + "epoch": 0.09, + "learning_rate": 1.983108301591973e-05, + "loss": 1.1807, + "step": 2716 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830893385443323e-05, + "loss": 1.1826, + "step": 2717 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830703649492312e-05, + "loss": 0.999, + "step": 2718 + }, + { + "epoch": 0.09, + "learning_rate": 1.983051380806873e-05, + "loss": 1.1084, + "step": 2719 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830323861174615e-05, + "loss": 1.1924, + "step": 2720 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830133808812e-05, + "loss": 1.002, + "step": 2721 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829943650982935e-05, + "loss": 1.2075, + "step": 2722 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829753387689452e-05, + "loss": 1.1602, + "step": 2723 + }, + { + "epoch": 0.09, + "learning_rate": 1.982956301893359e-05, + "loss": 1.0737, + "step": 2724 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829372544717402e-05, + "loss": 1.1143, + "step": 2725 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829181965042924e-05, + "loss": 1.0425, + "step": 2726 + }, + { + "epoch": 0.09, + "learning_rate": 1.98289912799122e-05, + "loss": 1.146, + "step": 2727 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828800489327277e-05, + "loss": 1.1875, + "step": 2728 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828609593290204e-05, + "loss": 1.0601, + "step": 2729 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828418591803025e-05, + "loss": 1.1167, + "step": 2730 + }, + { + "epoch": 0.09, + "learning_rate": 1.98282274848678e-05, + "loss": 1.1196, + "step": 2731 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828036272486565e-05, + "loss": 1.084, + "step": 2732 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827844954661376e-05, + "loss": 1.1914, + "step": 2733 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827653531394293e-05, + "loss": 1.1616, + "step": 2734 + }, + { + "epoch": 0.09, + "learning_rate": 1.982746200268736e-05, + "loss": 1.0244, + "step": 2735 + }, + { + "epoch": 0.09, + "learning_rate": 1.982727036854264e-05, + "loss": 1.1597, + "step": 2736 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827078628962182e-05, + "loss": 1.1152, + "step": 2737 + }, + { + "epoch": 0.09, + "learning_rate": 1.982688678394805e-05, + "loss": 1.0991, + "step": 2738 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826694833502295e-05, + "loss": 1.1572, + "step": 2739 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826502777626982e-05, + "loss": 1.1953, + "step": 2740 + }, + { + "epoch": 0.09, + "learning_rate": 1.982631061632417e-05, + "loss": 1.209, + "step": 2741 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826118349595918e-05, + "loss": 1.0986, + "step": 2742 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825925977444292e-05, + "loss": 1.1289, + "step": 2743 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825733499871357e-05, + "loss": 1.1216, + "step": 2744 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825540916879174e-05, + "loss": 1.1113, + "step": 2745 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825348228469813e-05, + "loss": 1.085, + "step": 2746 + }, + { + "epoch": 0.09, + "learning_rate": 1.982515543464534e-05, + "loss": 1.1987, + "step": 2747 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824962535407822e-05, + "loss": 0.9346, + "step": 2748 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824769530759327e-05, + "loss": 0.4775, + "step": 2749 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824576420701933e-05, + "loss": 1.083, + "step": 2750 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824383205237706e-05, + "loss": 1.1157, + "step": 2751 + }, + { + "epoch": 0.09, + "learning_rate": 1.982418988436872e-05, + "loss": 1.1797, + "step": 2752 + }, + { + "epoch": 0.09, + "learning_rate": 1.982399645809705e-05, + "loss": 1.0376, + "step": 2753 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823802926424766e-05, + "loss": 1.1484, + "step": 2754 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823609289353952e-05, + "loss": 1.0649, + "step": 2755 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823415546886685e-05, + "loss": 1.0713, + "step": 2756 + }, + { + "epoch": 0.09, + "learning_rate": 1.982322169902504e-05, + "loss": 1.1167, + "step": 2757 + }, + { + "epoch": 0.09, + "learning_rate": 1.98230277457711e-05, + "loss": 1.1489, + "step": 2758 + }, + { + "epoch": 0.09, + "learning_rate": 1.982283368712694e-05, + "loss": 1.1245, + "step": 2759 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822639523094645e-05, + "loss": 1.0923, + "step": 2760 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822445253676303e-05, + "loss": 1.1538, + "step": 2761 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822250878873995e-05, + "loss": 1.2075, + "step": 2762 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822056398689803e-05, + "loss": 1.1421, + "step": 2763 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821861813125816e-05, + "loss": 1.0127, + "step": 2764 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821667122184123e-05, + "loss": 1.0269, + "step": 2765 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821472325866812e-05, + "loss": 1.1455, + "step": 2766 + }, + { + "epoch": 0.09, + "learning_rate": 1.982127742417597e-05, + "loss": 1.165, + "step": 2767 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821082417113694e-05, + "loss": 1.0576, + "step": 2768 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820887304682074e-05, + "loss": 1.1958, + "step": 2769 + }, + { + "epoch": 0.09, + "learning_rate": 1.98206920868832e-05, + "loss": 1.1875, + "step": 2770 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820496763719173e-05, + "loss": 1.0562, + "step": 2771 + }, + { + "epoch": 0.09, + "learning_rate": 1.982030133519208e-05, + "loss": 1.0977, + "step": 2772 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820105801304024e-05, + "loss": 1.1846, + "step": 2773 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819910162057102e-05, + "loss": 0.897, + "step": 2774 + }, + { + "epoch": 0.09, + "learning_rate": 1.981971441745341e-05, + "loss": 1.1704, + "step": 2775 + }, + { + "epoch": 0.09, + "learning_rate": 1.981951856749505e-05, + "loss": 1.1084, + "step": 2776 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819322612184123e-05, + "loss": 1.1191, + "step": 2777 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819126551522734e-05, + "loss": 0.96, + "step": 2778 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818930385512988e-05, + "loss": 1.0518, + "step": 2779 + }, + { + "epoch": 0.09, + "learning_rate": 1.981873411415698e-05, + "loss": 1.1304, + "step": 2780 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818537737456826e-05, + "loss": 1.021, + "step": 2781 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818341255414625e-05, + "loss": 1.1187, + "step": 2782 + }, + { + "epoch": 0.09, + "learning_rate": 1.981814466803249e-05, + "loss": 1.1523, + "step": 2783 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817947975312527e-05, + "loss": 1.1948, + "step": 2784 + }, + { + "epoch": 0.09, + "learning_rate": 1.981775117725685e-05, + "loss": 1.2109, + "step": 2785 + }, + { + "epoch": 0.09, + "learning_rate": 1.981755427386757e-05, + "loss": 1.1831, + "step": 2786 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817357265146798e-05, + "loss": 1.0044, + "step": 2787 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817160151096645e-05, + "loss": 1.1426, + "step": 2788 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816962931719233e-05, + "loss": 1.0225, + "step": 2789 + }, + { + "epoch": 0.09, + "learning_rate": 1.981676560701667e-05, + "loss": 1.2256, + "step": 2790 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816568176991077e-05, + "loss": 0.9912, + "step": 2791 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816370641644575e-05, + "loss": 1.144, + "step": 2792 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816173000979276e-05, + "loss": 1.1875, + "step": 2793 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815975254997307e-05, + "loss": 1.2285, + "step": 2794 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815777403700786e-05, + "loss": 1.0273, + "step": 2795 + }, + { + "epoch": 0.09, + "learning_rate": 1.981557944709184e-05, + "loss": 1.0767, + "step": 2796 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815381385172588e-05, + "loss": 1.1587, + "step": 2797 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815183217945155e-05, + "loss": 1.0688, + "step": 2798 + }, + { + "epoch": 0.09, + "learning_rate": 1.981498494541167e-05, + "loss": 1.1353, + "step": 2799 + }, + { + "epoch": 0.09, + "learning_rate": 1.981478656757426e-05, + "loss": 1.1045, + "step": 2800 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814588084435055e-05, + "loss": 0.4585, + "step": 2801 + }, + { + "epoch": 0.09, + "learning_rate": 1.981438949599618e-05, + "loss": 1.1206, + "step": 2802 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814190802259765e-05, + "loss": 1.2021, + "step": 2803 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813992003227946e-05, + "loss": 1.0786, + "step": 2804 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813793098902858e-05, + "loss": 1.1313, + "step": 2805 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813594089286627e-05, + "loss": 1.0835, + "step": 2806 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813394974381396e-05, + "loss": 1.0771, + "step": 2807 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813195754189295e-05, + "loss": 1.0684, + "step": 2808 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812996428712466e-05, + "loss": 1.2544, + "step": 2809 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812796997953043e-05, + "loss": 1.0811, + "step": 2810 + }, + { + "epoch": 0.09, + "learning_rate": 1.981259746191317e-05, + "loss": 1.1992, + "step": 2811 + }, + { + "epoch": 0.09, + "learning_rate": 1.981239782059499e-05, + "loss": 1.1494, + "step": 2812 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812198074000638e-05, + "loss": 1.1191, + "step": 2813 + }, + { + "epoch": 0.09, + "learning_rate": 1.981199822213226e-05, + "loss": 1.2075, + "step": 2814 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811798264991998e-05, + "loss": 1.0854, + "step": 2815 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811598202582007e-05, + "loss": 1.1279, + "step": 2816 + }, + { + "epoch": 0.09, + "learning_rate": 1.981139803490442e-05, + "loss": 1.0742, + "step": 2817 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811197761961393e-05, + "loss": 1.1782, + "step": 2818 + }, + { + "epoch": 0.09, + "learning_rate": 1.981099738375507e-05, + "loss": 1.0449, + "step": 2819 + }, + { + "epoch": 0.09, + "learning_rate": 1.981079690028761e-05, + "loss": 1.1943, + "step": 2820 + }, + { + "epoch": 0.09, + "learning_rate": 1.981059631156115e-05, + "loss": 1.2637, + "step": 2821 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810395617577853e-05, + "loss": 1.1616, + "step": 2822 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810194818339868e-05, + "loss": 1.0571, + "step": 2823 + }, + { + "epoch": 0.09, + "learning_rate": 1.980999391384935e-05, + "loss": 1.2344, + "step": 2824 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809792904108453e-05, + "loss": 1.1631, + "step": 2825 + }, + { + "epoch": 0.09, + "learning_rate": 1.980959178911934e-05, + "loss": 1.2007, + "step": 2826 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809390568884156e-05, + "loss": 1.2388, + "step": 2827 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809189243405075e-05, + "loss": 1.041, + "step": 2828 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808987812684247e-05, + "loss": 1.0376, + "step": 2829 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808786276723832e-05, + "loss": 1.1885, + "step": 2830 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808584635526e-05, + "loss": 1.104, + "step": 2831 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808382889092913e-05, + "loss": 1.1318, + "step": 2832 + }, + { + "epoch": 0.09, + "learning_rate": 1.980818103742673e-05, + "loss": 1.0327, + "step": 2833 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807979080529618e-05, + "loss": 1.1133, + "step": 2834 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807777018403746e-05, + "loss": 1.1772, + "step": 2835 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807574851051284e-05, + "loss": 1.0908, + "step": 2836 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807372578474395e-05, + "loss": 1.3281, + "step": 2837 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807170200675255e-05, + "loss": 1.2266, + "step": 2838 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806967717656032e-05, + "loss": 1.0698, + "step": 2839 + }, + { + "epoch": 0.09, + "learning_rate": 1.98067651294189e-05, + "loss": 1.0552, + "step": 2840 + }, + { + "epoch": 0.09, + "learning_rate": 1.980656243596603e-05, + "loss": 1.0439, + "step": 2841 + }, + { + "epoch": 0.09, + "learning_rate": 1.98063596372996e-05, + "loss": 1.0, + "step": 2842 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806156733421783e-05, + "loss": 1.1851, + "step": 2843 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805953724334758e-05, + "loss": 1.1328, + "step": 2844 + }, + { + "epoch": 0.09, + "learning_rate": 1.98057506100407e-05, + "loss": 1.1035, + "step": 2845 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805547390541793e-05, + "loss": 1.2119, + "step": 2846 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805344065840212e-05, + "loss": 1.1621, + "step": 2847 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805140635938148e-05, + "loss": 1.2773, + "step": 2848 + }, + { + "epoch": 0.09, + "learning_rate": 1.980493710083777e-05, + "loss": 1.2031, + "step": 2849 + }, + { + "epoch": 0.09, + "learning_rate": 1.980473346054127e-05, + "loss": 1.001, + "step": 2850 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804529715050834e-05, + "loss": 1.2725, + "step": 2851 + }, + { + "epoch": 0.09, + "learning_rate": 1.980432586436864e-05, + "loss": 1.1152, + "step": 2852 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804121908496887e-05, + "loss": 1.0215, + "step": 2853 + }, + { + "epoch": 0.09, + "learning_rate": 1.980391784743775e-05, + "loss": 1.022, + "step": 2854 + }, + { + "epoch": 0.09, + "learning_rate": 1.980371368119343e-05, + "loss": 1.0664, + "step": 2855 + }, + { + "epoch": 0.09, + "learning_rate": 1.980350940976611e-05, + "loss": 1.1377, + "step": 2856 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803305033157983e-05, + "loss": 1.1313, + "step": 2857 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803100551371245e-05, + "loss": 0.9854, + "step": 2858 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802895964408088e-05, + "loss": 0.4373, + "step": 2859 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802691272270706e-05, + "loss": 1.0562, + "step": 2860 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802486474961295e-05, + "loss": 1.0767, + "step": 2861 + }, + { + "epoch": 0.09, + "learning_rate": 1.980228157248205e-05, + "loss": 1.0747, + "step": 2862 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802076564835175e-05, + "loss": 0.9966, + "step": 2863 + }, + { + "epoch": 0.09, + "learning_rate": 1.980187145202287e-05, + "loss": 1.2617, + "step": 2864 + }, + { + "epoch": 0.09, + "learning_rate": 1.980166623404733e-05, + "loss": 1.1558, + "step": 2865 + }, + { + "epoch": 0.09, + "learning_rate": 1.980146091091076e-05, + "loss": 1.0952, + "step": 2866 + }, + { + "epoch": 0.09, + "learning_rate": 1.980125548261536e-05, + "loss": 1.0259, + "step": 2867 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801049949163338e-05, + "loss": 1.1362, + "step": 2868 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800844310556897e-05, + "loss": 1.2261, + "step": 2869 + }, + { + "epoch": 0.09, + "learning_rate": 1.980063856679824e-05, + "loss": 1.0308, + "step": 2870 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800432717889583e-05, + "loss": 1.0962, + "step": 2871 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800226763833128e-05, + "loss": 1.209, + "step": 2872 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800020704631088e-05, + "loss": 1.0771, + "step": 2873 + }, + { + "epoch": 0.09, + "learning_rate": 1.979981454028567e-05, + "loss": 1.0669, + "step": 2874 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799608270799084e-05, + "loss": 0.4067, + "step": 2875 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799401896173555e-05, + "loss": 1.1338, + "step": 2876 + }, + { + "epoch": 0.09, + "learning_rate": 1.979919541641128e-05, + "loss": 1.0439, + "step": 2877 + }, + { + "epoch": 0.09, + "learning_rate": 1.979898883151449e-05, + "loss": 1.1079, + "step": 2878 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798782141485393e-05, + "loss": 1.1929, + "step": 2879 + }, + { + "epoch": 0.09, + "learning_rate": 1.979857534632621e-05, + "loss": 1.1221, + "step": 2880 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798368446039157e-05, + "loss": 1.1812, + "step": 2881 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798161440626454e-05, + "loss": 0.9907, + "step": 2882 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797954330090322e-05, + "loss": 1.104, + "step": 2883 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797747114432986e-05, + "loss": 1.1797, + "step": 2884 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797539793656666e-05, + "loss": 1.0635, + "step": 2885 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797332367763588e-05, + "loss": 1.1797, + "step": 2886 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797124836755977e-05, + "loss": 1.1812, + "step": 2887 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796917200636057e-05, + "loss": 1.0186, + "step": 2888 + }, + { + "epoch": 0.09, + "learning_rate": 1.979670945940606e-05, + "loss": 1.0752, + "step": 2889 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796501613068212e-05, + "loss": 1.1572, + "step": 2890 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796293661624746e-05, + "loss": 1.0679, + "step": 2891 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796085605077892e-05, + "loss": 1.1138, + "step": 2892 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795877443429878e-05, + "loss": 1.1406, + "step": 2893 + }, + { + "epoch": 0.09, + "learning_rate": 1.979566917668294e-05, + "loss": 0.9897, + "step": 2894 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795460804839313e-05, + "loss": 1.165, + "step": 2895 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795252327901235e-05, + "loss": 1.0845, + "step": 2896 + }, + { + "epoch": 0.09, + "learning_rate": 1.979504374587094e-05, + "loss": 1.0342, + "step": 2897 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794835058750664e-05, + "loss": 1.248, + "step": 2898 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794626266542654e-05, + "loss": 1.0684, + "step": 2899 + }, + { + "epoch": 0.09, + "learning_rate": 1.979441736924914e-05, + "loss": 0.5203, + "step": 2900 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794208366872368e-05, + "loss": 1.2212, + "step": 2901 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793999259414578e-05, + "loss": 0.9922, + "step": 2902 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793790046878018e-05, + "loss": 1.2305, + "step": 2903 + }, + { + "epoch": 0.09, + "learning_rate": 1.979358072926493e-05, + "loss": 1.1182, + "step": 2904 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793371306577558e-05, + "loss": 1.251, + "step": 2905 + }, + { + "epoch": 0.09, + "learning_rate": 1.979316177881815e-05, + "loss": 1.064, + "step": 2906 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792952145988958e-05, + "loss": 1.0396, + "step": 2907 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792742408092227e-05, + "loss": 1.0444, + "step": 2908 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792532565130206e-05, + "loss": 1.1729, + "step": 2909 + }, + { + "epoch": 0.09, + "learning_rate": 1.979232261710515e-05, + "loss": 1.1514, + "step": 2910 + }, + { + "epoch": 0.09, + "learning_rate": 1.979211256401931e-05, + "loss": 1.2524, + "step": 2911 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791902405874937e-05, + "loss": 1.1182, + "step": 2912 + }, + { + "epoch": 0.09, + "learning_rate": 1.979169214267429e-05, + "loss": 1.0889, + "step": 2913 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791481774419625e-05, + "loss": 1.1592, + "step": 2914 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791271301113198e-05, + "loss": 1.1094, + "step": 2915 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791060722757264e-05, + "loss": 1.0908, + "step": 2916 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790850039354083e-05, + "loss": 0.9824, + "step": 2917 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790639250905918e-05, + "loss": 0.4636, + "step": 2918 + }, + { + "epoch": 0.09, + "learning_rate": 1.979042835741503e-05, + "loss": 1.1143, + "step": 2919 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790217358883684e-05, + "loss": 1.1099, + "step": 2920 + }, + { + "epoch": 0.09, + "learning_rate": 1.979000625531414e-05, + "loss": 1.106, + "step": 2921 + }, + { + "epoch": 0.09, + "learning_rate": 1.978979504670866e-05, + "loss": 1.2148, + "step": 2922 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789583733069517e-05, + "loss": 1.1699, + "step": 2923 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789372314398976e-05, + "loss": 1.2881, + "step": 2924 + }, + { + "epoch": 0.09, + "learning_rate": 1.97891607906993e-05, + "loss": 1.1226, + "step": 2925 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788949161972768e-05, + "loss": 0.4539, + "step": 2926 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788737428221644e-05, + "loss": 0.9844, + "step": 2927 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788525589448202e-05, + "loss": 1.1831, + "step": 2928 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788313645654713e-05, + "loss": 1.063, + "step": 2929 + }, + { + "epoch": 0.09, + "learning_rate": 1.978810159684345e-05, + "loss": 1.0493, + "step": 2930 + }, + { + "epoch": 0.09, + "learning_rate": 1.978788944301669e-05, + "loss": 1.2773, + "step": 2931 + }, + { + "epoch": 0.09, + "learning_rate": 1.978767718417671e-05, + "loss": 1.1035, + "step": 2932 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787464820325786e-05, + "loss": 1.0435, + "step": 2933 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787252351466196e-05, + "loss": 1.0884, + "step": 2934 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787039777600222e-05, + "loss": 1.1128, + "step": 2935 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786827098730142e-05, + "loss": 1.2188, + "step": 2936 + }, + { + "epoch": 0.09, + "learning_rate": 1.978661431485824e-05, + "loss": 1.147, + "step": 2937 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786401425986798e-05, + "loss": 1.2529, + "step": 2938 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786188432118094e-05, + "loss": 1.1597, + "step": 2939 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785975333254425e-05, + "loss": 1.1348, + "step": 2940 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785762129398068e-05, + "loss": 1.0967, + "step": 2941 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785548820551316e-05, + "loss": 1.0405, + "step": 2942 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785335406716454e-05, + "loss": 1.1914, + "step": 2943 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785121887895774e-05, + "loss": 1.0698, + "step": 2944 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784908264091564e-05, + "loss": 1.1226, + "step": 2945 + }, + { + "epoch": 0.09, + "learning_rate": 1.978469453530612e-05, + "loss": 0.9722, + "step": 2946 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784480701541732e-05, + "loss": 1.0244, + "step": 2947 + }, + { + "epoch": 0.09, + "learning_rate": 1.978426676280069e-05, + "loss": 1.0903, + "step": 2948 + }, + { + "epoch": 0.09, + "learning_rate": 1.97840527190853e-05, + "loss": 1.1455, + "step": 2949 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783838570397854e-05, + "loss": 1.1621, + "step": 2950 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783624316740646e-05, + "loss": 1.0356, + "step": 2951 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783409958115974e-05, + "loss": 1.2075, + "step": 2952 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783195494526146e-05, + "loss": 1.2329, + "step": 2953 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782980925973455e-05, + "loss": 1.1396, + "step": 2954 + }, + { + "epoch": 0.09, + "learning_rate": 1.978276625246021e-05, + "loss": 1.2168, + "step": 2955 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782551473988706e-05, + "loss": 1.1479, + "step": 2956 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782336590561254e-05, + "loss": 1.147, + "step": 2957 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782121602180153e-05, + "loss": 1.1348, + "step": 2958 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781906508847715e-05, + "loss": 1.1162, + "step": 2959 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781691310566246e-05, + "loss": 1.0498, + "step": 2960 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781476007338058e-05, + "loss": 1.0679, + "step": 2961 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781260599165456e-05, + "loss": 1.0254, + "step": 2962 + }, + { + "epoch": 0.09, + "learning_rate": 1.978104508605075e-05, + "loss": 1.0361, + "step": 2963 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780829467996262e-05, + "loss": 1.0723, + "step": 2964 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780613745004295e-05, + "loss": 0.9805, + "step": 2965 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780397917077168e-05, + "loss": 1.0742, + "step": 2966 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780181984217196e-05, + "loss": 1.1694, + "step": 2967 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779965946426695e-05, + "loss": 1.1807, + "step": 2968 + }, + { + "epoch": 0.09, + "learning_rate": 1.977974980370798e-05, + "loss": 1.1631, + "step": 2969 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779533556063377e-05, + "loss": 1.1582, + "step": 2970 + }, + { + "epoch": 0.1, + "learning_rate": 1.9779317203495202e-05, + "loss": 0.9819, + "step": 2971 + }, + { + "epoch": 0.1, + "learning_rate": 1.977910074600578e-05, + "loss": 1.082, + "step": 2972 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778884183597425e-05, + "loss": 1.0596, + "step": 2973 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778667516272464e-05, + "loss": 1.0996, + "step": 2974 + }, + { + "epoch": 0.1, + "learning_rate": 1.977845074403323e-05, + "loss": 0.9927, + "step": 2975 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778233866882036e-05, + "loss": 0.9751, + "step": 2976 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778016884821215e-05, + "loss": 1.0366, + "step": 2977 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777799797853096e-05, + "loss": 1.1079, + "step": 2978 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777582605980007e-05, + "loss": 1.1074, + "step": 2979 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777365309204277e-05, + "loss": 1.0063, + "step": 2980 + }, + { + "epoch": 0.1, + "learning_rate": 1.977714790752824e-05, + "loss": 0.46, + "step": 2981 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776930400954225e-05, + "loss": 1.1206, + "step": 2982 + }, + { + "epoch": 0.1, + "learning_rate": 1.977671278948457e-05, + "loss": 1.1855, + "step": 2983 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776495073121606e-05, + "loss": 1.0791, + "step": 2984 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776277251867665e-05, + "loss": 1.1787, + "step": 2985 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776059325725093e-05, + "loss": 1.1704, + "step": 2986 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775841294696227e-05, + "loss": 1.0293, + "step": 2987 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775623158783397e-05, + "loss": 1.0767, + "step": 2988 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775404917988953e-05, + "loss": 1.1128, + "step": 2989 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775186572315234e-05, + "loss": 1.0762, + "step": 2990 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774968121764583e-05, + "loss": 1.1406, + "step": 2991 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774749566339337e-05, + "loss": 1.1221, + "step": 2992 + }, + { + "epoch": 0.1, + "learning_rate": 1.977453090604185e-05, + "loss": 1.1914, + "step": 2993 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774312140874462e-05, + "loss": 1.0796, + "step": 2994 + }, + { + "epoch": 0.1, + "learning_rate": 1.977409327083953e-05, + "loss": 1.2275, + "step": 2995 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773874295939386e-05, + "loss": 1.2134, + "step": 2996 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773655216176388e-05, + "loss": 1.0903, + "step": 2997 + }, + { + "epoch": 0.1, + "learning_rate": 1.977343603155289e-05, + "loss": 1.1558, + "step": 2998 + }, + { + "epoch": 0.1, + "learning_rate": 1.977321674207124e-05, + "loss": 1.1729, + "step": 2999 + }, + { + "epoch": 0.1, + "learning_rate": 1.977299734773379e-05, + "loss": 1.0679, + "step": 3000 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772777848542894e-05, + "loss": 1.0542, + "step": 3001 + }, + { + "epoch": 0.1, + "learning_rate": 1.977255824450091e-05, + "loss": 1.1406, + "step": 3002 + }, + { + "epoch": 0.1, + "learning_rate": 1.977233853561019e-05, + "loss": 1.1553, + "step": 3003 + }, + { + "epoch": 0.1, + "learning_rate": 1.977211872187309e-05, + "loss": 1.1987, + "step": 3004 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771898803291977e-05, + "loss": 1.0791, + "step": 3005 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771678779869203e-05, + "loss": 1.0161, + "step": 3006 + }, + { + "epoch": 0.1, + "learning_rate": 1.977145865160713e-05, + "loss": 1.0396, + "step": 3007 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771238418508122e-05, + "loss": 1.1426, + "step": 3008 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771018080574534e-05, + "loss": 1.2031, + "step": 3009 + }, + { + "epoch": 0.1, + "learning_rate": 1.977079763780874e-05, + "loss": 1.0986, + "step": 3010 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770577090213102e-05, + "loss": 1.1182, + "step": 3011 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770356437789984e-05, + "loss": 1.1396, + "step": 3012 + }, + { + "epoch": 0.1, + "learning_rate": 1.977013568054176e-05, + "loss": 1.0601, + "step": 3013 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769914818470786e-05, + "loss": 1.0654, + "step": 3014 + }, + { + "epoch": 0.1, + "learning_rate": 1.976969385157944e-05, + "loss": 1.0923, + "step": 3015 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769472779870093e-05, + "loss": 1.0171, + "step": 3016 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769251603345116e-05, + "loss": 1.1582, + "step": 3017 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769030322006882e-05, + "loss": 1.0806, + "step": 3018 + }, + { + "epoch": 0.1, + "learning_rate": 1.976880893585776e-05, + "loss": 1.1738, + "step": 3019 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768587444900132e-05, + "loss": 1.0566, + "step": 3020 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768365849136377e-05, + "loss": 1.1084, + "step": 3021 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768144148568862e-05, + "loss": 1.0664, + "step": 3022 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767922343199974e-05, + "loss": 1.0791, + "step": 3023 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767700433032086e-05, + "loss": 1.0312, + "step": 3024 + }, + { + "epoch": 0.1, + "learning_rate": 1.976747841806759e-05, + "loss": 1.2197, + "step": 3025 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767256298308852e-05, + "loss": 1.1177, + "step": 3026 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767034073758266e-05, + "loss": 1.0781, + "step": 3027 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766811744418218e-05, + "loss": 1.0723, + "step": 3028 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766589310291087e-05, + "loss": 1.1299, + "step": 3029 + }, + { + "epoch": 0.1, + "learning_rate": 1.976636677137926e-05, + "loss": 1.0332, + "step": 3030 + }, + { + "epoch": 0.1, + "learning_rate": 1.976614412768513e-05, + "loss": 1.1963, + "step": 3031 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765921379211078e-05, + "loss": 1.1484, + "step": 3032 + }, + { + "epoch": 0.1, + "learning_rate": 1.97656985259595e-05, + "loss": 1.0669, + "step": 3033 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765475567932786e-05, + "loss": 1.0288, + "step": 3034 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765252505133327e-05, + "loss": 1.1333, + "step": 3035 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765029337563515e-05, + "loss": 1.1323, + "step": 3036 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764806065225746e-05, + "loss": 0.9702, + "step": 3037 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764582688122415e-05, + "loss": 1.0933, + "step": 3038 + }, + { + "epoch": 0.1, + "learning_rate": 1.976435920625592e-05, + "loss": 1.2627, + "step": 3039 + }, + { + "epoch": 0.1, + "learning_rate": 1.976413561962866e-05, + "loss": 0.4663, + "step": 3040 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763911928243026e-05, + "loss": 1.0767, + "step": 3041 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763688132101424e-05, + "loss": 1.1104, + "step": 3042 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763464231206252e-05, + "loss": 1.1367, + "step": 3043 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763240225559917e-05, + "loss": 1.1909, + "step": 3044 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763016115164823e-05, + "loss": 1.0884, + "step": 3045 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762791900023364e-05, + "loss": 1.0791, + "step": 3046 + }, + { + "epoch": 0.1, + "learning_rate": 1.976256758013796e-05, + "loss": 1.0283, + "step": 3047 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762343155511007e-05, + "loss": 1.103, + "step": 3048 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762118626144916e-05, + "loss": 1.082, + "step": 3049 + }, + { + "epoch": 0.1, + "learning_rate": 1.97618939920421e-05, + "loss": 1.2305, + "step": 3050 + }, + { + "epoch": 0.1, + "learning_rate": 1.976166925320496e-05, + "loss": 1.0752, + "step": 3051 + }, + { + "epoch": 0.1, + "learning_rate": 1.976144440963592e-05, + "loss": 1.1055, + "step": 3052 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761219461337377e-05, + "loss": 1.1387, + "step": 3053 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760994408311757e-05, + "loss": 1.187, + "step": 3054 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760769250561467e-05, + "loss": 1.103, + "step": 3055 + }, + { + "epoch": 0.1, + "learning_rate": 1.976054398808893e-05, + "loss": 1.0693, + "step": 3056 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760318620896557e-05, + "loss": 1.1226, + "step": 3057 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760093148986765e-05, + "loss": 1.2231, + "step": 3058 + }, + { + "epoch": 0.1, + "learning_rate": 1.975986757236198e-05, + "loss": 1.0371, + "step": 3059 + }, + { + "epoch": 0.1, + "learning_rate": 1.975964189102461e-05, + "loss": 1.0991, + "step": 3060 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759416104977086e-05, + "loss": 1.0938, + "step": 3061 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759190214221834e-05, + "loss": 1.2017, + "step": 3062 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758964218761268e-05, + "loss": 1.0898, + "step": 3063 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758738118597818e-05, + "loss": 1.2998, + "step": 3064 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758511913733906e-05, + "loss": 1.0967, + "step": 3065 + }, + { + "epoch": 0.1, + "learning_rate": 1.975828560417196e-05, + "loss": 1.041, + "step": 3066 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758059189914417e-05, + "loss": 1.0391, + "step": 3067 + }, + { + "epoch": 0.1, + "learning_rate": 1.975783267096369e-05, + "loss": 0.9966, + "step": 3068 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757606047322225e-05, + "loss": 1.1377, + "step": 3069 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757379318992443e-05, + "loss": 1.0981, + "step": 3070 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757152485976778e-05, + "loss": 1.2075, + "step": 3071 + }, + { + "epoch": 0.1, + "learning_rate": 1.975692554827767e-05, + "loss": 0.9927, + "step": 3072 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756698505897548e-05, + "loss": 1.0908, + "step": 3073 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756471358838847e-05, + "loss": 0.5229, + "step": 3074 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756244107104005e-05, + "loss": 1.1406, + "step": 3075 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756016750695464e-05, + "loss": 1.0522, + "step": 3076 + }, + { + "epoch": 0.1, + "learning_rate": 1.975578928961566e-05, + "loss": 1.0825, + "step": 3077 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755561723867033e-05, + "loss": 1.0938, + "step": 3078 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755334053452028e-05, + "loss": 1.1196, + "step": 3079 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755106278373085e-05, + "loss": 1.1055, + "step": 3080 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754878398632647e-05, + "loss": 1.1748, + "step": 3081 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754650414233158e-05, + "loss": 1.0591, + "step": 3082 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754422325177072e-05, + "loss": 1.2021, + "step": 3083 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754194131466823e-05, + "loss": 1.1035, + "step": 3084 + }, + { + "epoch": 0.1, + "learning_rate": 1.975396583310487e-05, + "loss": 0.9902, + "step": 3085 + }, + { + "epoch": 0.1, + "learning_rate": 1.975373743009366e-05, + "loss": 1.1138, + "step": 3086 + }, + { + "epoch": 0.1, + "learning_rate": 1.975350892243564e-05, + "loss": 1.0205, + "step": 3087 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753280310133265e-05, + "loss": 1.1885, + "step": 3088 + }, + { + "epoch": 0.1, + "learning_rate": 1.975305159318899e-05, + "loss": 1.1792, + "step": 3089 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752822771605258e-05, + "loss": 1.1616, + "step": 3090 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752593845384535e-05, + "loss": 1.1406, + "step": 3091 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752364814529273e-05, + "loss": 1.1987, + "step": 3092 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752135679041932e-05, + "loss": 1.103, + "step": 3093 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751906438924968e-05, + "loss": 1.0381, + "step": 3094 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751677094180838e-05, + "loss": 1.2061, + "step": 3095 + }, + { + "epoch": 0.1, + "learning_rate": 1.975144764481201e-05, + "loss": 1.0859, + "step": 3096 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751218090820937e-05, + "loss": 1.1553, + "step": 3097 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750988432210086e-05, + "loss": 1.1191, + "step": 3098 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750758668981925e-05, + "loss": 1.249, + "step": 3099 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750528801138913e-05, + "loss": 1.0771, + "step": 3100 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750298828683517e-05, + "loss": 1.0693, + "step": 3101 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750068751618208e-05, + "loss": 1.0586, + "step": 3102 + }, + { + "epoch": 0.1, + "learning_rate": 1.974983856994545e-05, + "loss": 1.1392, + "step": 3103 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749608283667715e-05, + "loss": 1.0903, + "step": 3104 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749377892787476e-05, + "loss": 1.2012, + "step": 3105 + }, + { + "epoch": 0.1, + "learning_rate": 1.97491473973072e-05, + "loss": 1.2163, + "step": 3106 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748916797229362e-05, + "loss": 1.123, + "step": 3107 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748686092556437e-05, + "loss": 1.1514, + "step": 3108 + }, + { + "epoch": 0.1, + "learning_rate": 1.97484552832909e-05, + "loss": 1.0859, + "step": 3109 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748224369435227e-05, + "loss": 1.0679, + "step": 3110 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747993350991895e-05, + "loss": 1.0396, + "step": 3111 + }, + { + "epoch": 0.1, + "learning_rate": 1.974776222796338e-05, + "loss": 0.9492, + "step": 3112 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747531000352168e-05, + "loss": 1.127, + "step": 3113 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747299668160735e-05, + "loss": 1.0903, + "step": 3114 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747068231391565e-05, + "loss": 1.1167, + "step": 3115 + }, + { + "epoch": 0.1, + "learning_rate": 1.974683669004714e-05, + "loss": 1.0283, + "step": 3116 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746605044129944e-05, + "loss": 1.1538, + "step": 3117 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746373293642464e-05, + "loss": 1.1738, + "step": 3118 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746141438587185e-05, + "loss": 1.021, + "step": 3119 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745909478966596e-05, + "loss": 1.1606, + "step": 3120 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745677414783185e-05, + "loss": 0.4705, + "step": 3121 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745445246039438e-05, + "loss": 1.0044, + "step": 3122 + }, + { + "epoch": 0.1, + "learning_rate": 1.974521297273785e-05, + "loss": 1.2314, + "step": 3123 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744980594880912e-05, + "loss": 1.1943, + "step": 3124 + }, + { + "epoch": 0.1, + "learning_rate": 1.974474811247112e-05, + "loss": 1.0449, + "step": 3125 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744515525510965e-05, + "loss": 0.8657, + "step": 3126 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744282834002942e-05, + "loss": 1.1699, + "step": 3127 + }, + { + "epoch": 0.1, + "learning_rate": 1.974405003794955e-05, + "loss": 1.1597, + "step": 3128 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743817137353283e-05, + "loss": 1.2153, + "step": 3129 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743584132216644e-05, + "loss": 1.0586, + "step": 3130 + }, + { + "epoch": 0.1, + "learning_rate": 1.974335102254213e-05, + "loss": 0.9858, + "step": 3131 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743117808332243e-05, + "loss": 1.0239, + "step": 3132 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742884489589485e-05, + "loss": 1.1006, + "step": 3133 + }, + { + "epoch": 0.1, + "learning_rate": 1.974265106631636e-05, + "loss": 1.1807, + "step": 3134 + }, + { + "epoch": 0.1, + "learning_rate": 1.974241753851537e-05, + "loss": 1.1172, + "step": 3135 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742183906189023e-05, + "loss": 1.063, + "step": 3136 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741950169339828e-05, + "loss": 1.1538, + "step": 3137 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741716327970285e-05, + "loss": 1.1323, + "step": 3138 + }, + { + "epoch": 0.1, + "learning_rate": 1.974148238208291e-05, + "loss": 1.0664, + "step": 3139 + }, + { + "epoch": 0.1, + "learning_rate": 1.974124833168021e-05, + "loss": 1.0435, + "step": 3140 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741014176764698e-05, + "loss": 1.0723, + "step": 3141 + }, + { + "epoch": 0.1, + "learning_rate": 1.974077991733888e-05, + "loss": 1.0439, + "step": 3142 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740545553405278e-05, + "loss": 1.1025, + "step": 3143 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740311084966398e-05, + "loss": 1.1553, + "step": 3144 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740076512024768e-05, + "loss": 1.1997, + "step": 3145 + }, + { + "epoch": 0.1, + "learning_rate": 1.973984183458289e-05, + "loss": 1.1118, + "step": 3146 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739607052643293e-05, + "loss": 1.1436, + "step": 3147 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739372166208487e-05, + "loss": 1.0771, + "step": 3148 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739137175281e-05, + "loss": 1.1289, + "step": 3149 + }, + { + "epoch": 0.1, + "learning_rate": 1.973890207986335e-05, + "loss": 1.1108, + "step": 3150 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738666879958058e-05, + "loss": 0.9966, + "step": 3151 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738431575567648e-05, + "loss": 1.0112, + "step": 3152 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738196166694647e-05, + "loss": 1.1431, + "step": 3153 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737960653341575e-05, + "loss": 1.0947, + "step": 3154 + }, + { + "epoch": 0.1, + "learning_rate": 1.973772503551097e-05, + "loss": 1.1792, + "step": 3155 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737489313205345e-05, + "loss": 1.0156, + "step": 3156 + }, + { + "epoch": 0.1, + "learning_rate": 1.973725348642724e-05, + "loss": 1.1064, + "step": 3157 + }, + { + "epoch": 0.1, + "learning_rate": 1.973701755517918e-05, + "loss": 1.0342, + "step": 3158 + }, + { + "epoch": 0.1, + "learning_rate": 1.97367815194637e-05, + "loss": 1.2139, + "step": 3159 + }, + { + "epoch": 0.1, + "learning_rate": 1.973654537928333e-05, + "loss": 1.0522, + "step": 3160 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736309134640604e-05, + "loss": 1.1826, + "step": 3161 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736072785538053e-05, + "loss": 1.1787, + "step": 3162 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735836331978223e-05, + "loss": 1.1157, + "step": 3163 + }, + { + "epoch": 0.1, + "learning_rate": 1.973559977396364e-05, + "loss": 1.1001, + "step": 3164 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735363111496847e-05, + "loss": 1.0518, + "step": 3165 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735126344580384e-05, + "loss": 1.1172, + "step": 3166 + }, + { + "epoch": 0.1, + "learning_rate": 1.973488947321679e-05, + "loss": 1.1362, + "step": 3167 + }, + { + "epoch": 0.1, + "learning_rate": 1.97346524974086e-05, + "loss": 1.083, + "step": 3168 + }, + { + "epoch": 0.1, + "learning_rate": 1.973441541715837e-05, + "loss": 1.0132, + "step": 3169 + }, + { + "epoch": 0.1, + "learning_rate": 1.973417823246863e-05, + "loss": 1.1934, + "step": 3170 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733940943341935e-05, + "loss": 1.2021, + "step": 3171 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733703549780824e-05, + "loss": 1.1274, + "step": 3172 + }, + { + "epoch": 0.1, + "learning_rate": 1.973346605178785e-05, + "loss": 1.0752, + "step": 3173 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733228449365554e-05, + "loss": 1.1865, + "step": 3174 + }, + { + "epoch": 0.1, + "learning_rate": 1.973299074251649e-05, + "loss": 1.1152, + "step": 3175 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732752931243205e-05, + "loss": 1.127, + "step": 3176 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732515015548258e-05, + "loss": 0.9624, + "step": 3177 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732276995434195e-05, + "loss": 1.1255, + "step": 3178 + }, + { + "epoch": 0.1, + "learning_rate": 1.973203887090357e-05, + "loss": 1.1113, + "step": 3179 + }, + { + "epoch": 0.1, + "learning_rate": 1.973180064195894e-05, + "loss": 1.0889, + "step": 3180 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731562308602856e-05, + "loss": 1.082, + "step": 3181 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731323870837878e-05, + "loss": 1.0591, + "step": 3182 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731085328666572e-05, + "loss": 1.0913, + "step": 3183 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730846682091485e-05, + "loss": 1.1094, + "step": 3184 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730607931115183e-05, + "loss": 1.02, + "step": 3185 + }, + { + "epoch": 0.1, + "learning_rate": 1.973036907574023e-05, + "loss": 1.1069, + "step": 3186 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730130115969183e-05, + "loss": 1.0698, + "step": 3187 + }, + { + "epoch": 0.1, + "learning_rate": 1.972989105180461e-05, + "loss": 1.0044, + "step": 3188 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729651883249075e-05, + "loss": 0.5024, + "step": 3189 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729412610305142e-05, + "loss": 1.105, + "step": 3190 + }, + { + "epoch": 0.1, + "learning_rate": 1.972917323297538e-05, + "loss": 1.2148, + "step": 3191 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728933751262358e-05, + "loss": 1.0874, + "step": 3192 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728694165168646e-05, + "loss": 0.9121, + "step": 3193 + }, + { + "epoch": 0.1, + "learning_rate": 1.972845447469681e-05, + "loss": 1.0933, + "step": 3194 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728214679849423e-05, + "loss": 1.2617, + "step": 3195 + }, + { + "epoch": 0.1, + "learning_rate": 1.972797478062906e-05, + "loss": 1.1143, + "step": 3196 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727734777038295e-05, + "loss": 1.0757, + "step": 3197 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727494669079705e-05, + "loss": 1.208, + "step": 3198 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727254456755857e-05, + "loss": 1.2139, + "step": 3199 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727014140069337e-05, + "loss": 1.2617, + "step": 3200 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726773719022723e-05, + "loss": 1.0513, + "step": 3201 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726533193618588e-05, + "loss": 1.166, + "step": 3202 + }, + { + "epoch": 0.1, + "learning_rate": 1.972629256385952e-05, + "loss": 1.1279, + "step": 3203 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726051829748095e-05, + "loss": 1.0884, + "step": 3204 + }, + { + "epoch": 0.1, + "learning_rate": 1.97258109912869e-05, + "loss": 1.1621, + "step": 3205 + }, + { + "epoch": 0.1, + "learning_rate": 1.972557004847852e-05, + "loss": 1.1211, + "step": 3206 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725329001325527e-05, + "loss": 1.1953, + "step": 3207 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725087849830524e-05, + "loss": 1.1255, + "step": 3208 + }, + { + "epoch": 0.1, + "learning_rate": 1.972484659399609e-05, + "loss": 1.0923, + "step": 3209 + }, + { + "epoch": 0.1, + "learning_rate": 1.972460523382482e-05, + "loss": 1.1348, + "step": 3210 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724363769319293e-05, + "loss": 0.9814, + "step": 3211 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724122200482107e-05, + "loss": 1.1382, + "step": 3212 + }, + { + "epoch": 0.1, + "learning_rate": 1.972388052731585e-05, + "loss": 1.2153, + "step": 3213 + }, + { + "epoch": 0.1, + "learning_rate": 1.972363874982312e-05, + "loss": 1.1206, + "step": 3214 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723396868006507e-05, + "loss": 1.0869, + "step": 3215 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723154881868606e-05, + "loss": 1.0903, + "step": 3216 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722912791412014e-05, + "loss": 1.2139, + "step": 3217 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722670596639327e-05, + "loss": 0.9814, + "step": 3218 + }, + { + "epoch": 0.1, + "learning_rate": 1.972242829755315e-05, + "loss": 1.1035, + "step": 3219 + }, + { + "epoch": 0.1, + "learning_rate": 1.972218589415607e-05, + "loss": 1.0913, + "step": 3220 + }, + { + "epoch": 0.1, + "learning_rate": 1.97219433864507e-05, + "loss": 1.1162, + "step": 3221 + }, + { + "epoch": 0.1, + "learning_rate": 1.972170077443964e-05, + "loss": 1.147, + "step": 3222 + }, + { + "epoch": 0.1, + "learning_rate": 1.9721458058125487e-05, + "loss": 1.0757, + "step": 3223 + }, + { + "epoch": 0.1, + "learning_rate": 1.972121523751085e-05, + "loss": 0.9595, + "step": 3224 + }, + { + "epoch": 0.1, + "learning_rate": 1.972097231259833e-05, + "loss": 1.1104, + "step": 3225 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720729283390534e-05, + "loss": 1.1846, + "step": 3226 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720486149890073e-05, + "loss": 1.0625, + "step": 3227 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720242912099558e-05, + "loss": 1.0449, + "step": 3228 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719999570021587e-05, + "loss": 1.144, + "step": 3229 + }, + { + "epoch": 0.1, + "learning_rate": 1.971975612365878e-05, + "loss": 1.146, + "step": 3230 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719512573013753e-05, + "loss": 1.1025, + "step": 3231 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719268918089107e-05, + "loss": 1.0361, + "step": 3232 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719025158887467e-05, + "loss": 1.1387, + "step": 3233 + }, + { + "epoch": 0.1, + "learning_rate": 1.971878129541144e-05, + "loss": 1.0688, + "step": 3234 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718537327663646e-05, + "loss": 0.9619, + "step": 3235 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718293255646705e-05, + "loss": 0.9922, + "step": 3236 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718049079363227e-05, + "loss": 1.1035, + "step": 3237 + }, + { + "epoch": 0.1, + "learning_rate": 1.971780479881584e-05, + "loss": 0.49, + "step": 3238 + }, + { + "epoch": 0.1, + "learning_rate": 1.971756041400717e-05, + "loss": 1.2446, + "step": 3239 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717315924939825e-05, + "loss": 1.1182, + "step": 3240 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717071331616437e-05, + "loss": 1.0039, + "step": 3241 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716826634039626e-05, + "loss": 1.1426, + "step": 3242 + }, + { + "epoch": 0.1, + "learning_rate": 1.971658183221202e-05, + "loss": 1.2373, + "step": 3243 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716336926136244e-05, + "loss": 1.0845, + "step": 3244 + }, + { + "epoch": 0.1, + "learning_rate": 1.971609191581493e-05, + "loss": 1.103, + "step": 3245 + }, + { + "epoch": 0.1, + "learning_rate": 1.97158468012507e-05, + "loss": 1.0981, + "step": 3246 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715601582446188e-05, + "loss": 1.2153, + "step": 3247 + }, + { + "epoch": 0.1, + "learning_rate": 1.971535625940402e-05, + "loss": 1.1846, + "step": 3248 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715110832126835e-05, + "loss": 1.1118, + "step": 3249 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714865300617262e-05, + "loss": 1.1851, + "step": 3250 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714619664877937e-05, + "loss": 1.0522, + "step": 3251 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714373924911493e-05, + "loss": 1.127, + "step": 3252 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714128080720568e-05, + "loss": 0.9434, + "step": 3253 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713882132307805e-05, + "loss": 1.1523, + "step": 3254 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713636079675835e-05, + "loss": 1.0869, + "step": 3255 + }, + { + "epoch": 0.1, + "learning_rate": 1.97133899228273e-05, + "loss": 1.0503, + "step": 3256 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713143661764836e-05, + "loss": 1.105, + "step": 3257 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712897296491094e-05, + "loss": 1.1782, + "step": 3258 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712650827008714e-05, + "loss": 1.0864, + "step": 3259 + }, + { + "epoch": 0.1, + "learning_rate": 1.971240425332034e-05, + "loss": 1.0942, + "step": 3260 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712157575428617e-05, + "loss": 0.9917, + "step": 3261 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711910793336196e-05, + "loss": 1.1011, + "step": 3262 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711663907045718e-05, + "loss": 1.1538, + "step": 3263 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711416916559837e-05, + "loss": 1.1519, + "step": 3264 + }, + { + "epoch": 0.1, + "learning_rate": 1.97111698218812e-05, + "loss": 1.0972, + "step": 3265 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710922623012453e-05, + "loss": 1.0273, + "step": 3266 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710675319956256e-05, + "loss": 1.0381, + "step": 3267 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710427912715264e-05, + "loss": 0.9512, + "step": 3268 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710180401292127e-05, + "loss": 1.1602, + "step": 3269 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709932785689496e-05, + "loss": 1.248, + "step": 3270 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709685065910037e-05, + "loss": 1.209, + "step": 3271 + }, + { + "epoch": 0.1, + "learning_rate": 1.97094372419564e-05, + "loss": 0.9736, + "step": 3272 + }, + { + "epoch": 0.1, + "learning_rate": 1.970918931383125e-05, + "loss": 1.1343, + "step": 3273 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708941281537244e-05, + "loss": 1.1899, + "step": 3274 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708693145077046e-05, + "loss": 1.2275, + "step": 3275 + }, + { + "epoch": 0.1, + "learning_rate": 1.970844490445331e-05, + "loss": 1.0308, + "step": 3276 + }, + { + "epoch": 0.1, + "learning_rate": 1.970819655966871e-05, + "loss": 1.0908, + "step": 3277 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707948110725898e-05, + "loss": 1.1436, + "step": 3278 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707699557627554e-05, + "loss": 1.1592, + "step": 3279 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707450900376335e-05, + "loss": 1.0322, + "step": 3280 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707202138974914e-05, + "loss": 1.1333, + "step": 3281 + }, + { + "epoch": 0.1, + "learning_rate": 1.9706953273425957e-05, + "loss": 1.082, + "step": 3282 + }, + { + "epoch": 0.1, + "learning_rate": 1.970670430373213e-05, + "loss": 1.2163, + "step": 3283 + }, + { + "epoch": 0.11, + "learning_rate": 1.970645522989611e-05, + "loss": 1.1479, + "step": 3284 + }, + { + "epoch": 0.11, + "learning_rate": 1.9706206051920573e-05, + "loss": 1.1055, + "step": 3285 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705956769808185e-05, + "loss": 1.0596, + "step": 3286 + }, + { + "epoch": 0.11, + "learning_rate": 1.970570738356162e-05, + "loss": 1.1138, + "step": 3287 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705457893183556e-05, + "loss": 1.0469, + "step": 3288 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705208298676676e-05, + "loss": 1.083, + "step": 3289 + }, + { + "epoch": 0.11, + "learning_rate": 1.970495860004365e-05, + "loss": 1.1577, + "step": 3290 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704708797287156e-05, + "loss": 1.0679, + "step": 3291 + }, + { + "epoch": 0.11, + "learning_rate": 1.970445889040988e-05, + "loss": 1.0664, + "step": 3292 + }, + { + "epoch": 0.11, + "learning_rate": 1.97042088794145e-05, + "loss": 1.1182, + "step": 3293 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703958764303698e-05, + "loss": 0.5093, + "step": 3294 + }, + { + "epoch": 0.11, + "learning_rate": 1.970370854508016e-05, + "loss": 1.2651, + "step": 3295 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703458221746565e-05, + "loss": 1.1509, + "step": 3296 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703207794305608e-05, + "loss": 1.0166, + "step": 3297 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702957262759964e-05, + "loss": 1.0347, + "step": 3298 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702706627112335e-05, + "loss": 1.0791, + "step": 3299 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702455887365393e-05, + "loss": 1.0737, + "step": 3300 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702205043521843e-05, + "loss": 1.0864, + "step": 3301 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701954095584374e-05, + "loss": 1.1113, + "step": 3302 + }, + { + "epoch": 0.11, + "learning_rate": 1.970170304355567e-05, + "loss": 1.0391, + "step": 3303 + }, + { + "epoch": 0.11, + "learning_rate": 1.970145188743843e-05, + "loss": 1.1914, + "step": 3304 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701200627235352e-05, + "loss": 1.123, + "step": 3305 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700949262949124e-05, + "loss": 1.2256, + "step": 3306 + }, + { + "epoch": 0.11, + "learning_rate": 1.970069779458245e-05, + "loss": 1.1885, + "step": 3307 + }, + { + "epoch": 0.11, + "learning_rate": 1.970044622213802e-05, + "loss": 1.1284, + "step": 3308 + }, + { + "epoch": 0.11, + "learning_rate": 1.970019454561854e-05, + "loss": 1.1021, + "step": 3309 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699942765026713e-05, + "loss": 1.0825, + "step": 3310 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699690880365232e-05, + "loss": 1.2183, + "step": 3311 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699438891636805e-05, + "loss": 1.0176, + "step": 3312 + }, + { + "epoch": 0.11, + "learning_rate": 1.969918679884413e-05, + "loss": 1.0195, + "step": 3313 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698934601989916e-05, + "loss": 1.0298, + "step": 3314 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698682301076867e-05, + "loss": 1.1592, + "step": 3315 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698429896107692e-05, + "loss": 1.0918, + "step": 3316 + }, + { + "epoch": 0.11, + "learning_rate": 1.96981773870851e-05, + "loss": 1.1226, + "step": 3317 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697924774011798e-05, + "loss": 1.0283, + "step": 3318 + }, + { + "epoch": 0.11, + "learning_rate": 1.969767205689049e-05, + "loss": 1.1372, + "step": 3319 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697419235723903e-05, + "loss": 1.1074, + "step": 3320 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697166310514735e-05, + "loss": 1.1704, + "step": 3321 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696913281265703e-05, + "loss": 1.2754, + "step": 3322 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696660147979523e-05, + "loss": 1.167, + "step": 3323 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696406910658918e-05, + "loss": 1.1523, + "step": 3324 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696153569306596e-05, + "loss": 1.0806, + "step": 3325 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695900123925273e-05, + "loss": 1.1787, + "step": 3326 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695646574517675e-05, + "loss": 1.0337, + "step": 3327 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695392921086523e-05, + "loss": 0.9897, + "step": 3328 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695139163634528e-05, + "loss": 1.1689, + "step": 3329 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694885302164427e-05, + "loss": 1.0176, + "step": 3330 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694631336678935e-05, + "loss": 1.1689, + "step": 3331 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694377267180774e-05, + "loss": 0.4922, + "step": 3332 + }, + { + "epoch": 0.11, + "learning_rate": 1.969412309367268e-05, + "loss": 1.1592, + "step": 3333 + }, + { + "epoch": 0.11, + "learning_rate": 1.969386881615737e-05, + "loss": 1.0991, + "step": 3334 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693614434637575e-05, + "loss": 1.1074, + "step": 3335 + }, + { + "epoch": 0.11, + "learning_rate": 1.969335994911603e-05, + "loss": 1.1445, + "step": 3336 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693105359595457e-05, + "loss": 1.0571, + "step": 3337 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692850666078594e-05, + "loss": 1.1895, + "step": 3338 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692595868568172e-05, + "loss": 1.1733, + "step": 3339 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692340967066922e-05, + "loss": 1.1133, + "step": 3340 + }, + { + "epoch": 0.11, + "learning_rate": 1.969208596157758e-05, + "loss": 0.9995, + "step": 3341 + }, + { + "epoch": 0.11, + "learning_rate": 1.969183085210288e-05, + "loss": 1.145, + "step": 3342 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691575638645567e-05, + "loss": 1.0415, + "step": 3343 + }, + { + "epoch": 0.11, + "learning_rate": 1.969132032120837e-05, + "loss": 1.106, + "step": 3344 + }, + { + "epoch": 0.11, + "learning_rate": 1.969106489979403e-05, + "loss": 1.2896, + "step": 3345 + }, + { + "epoch": 0.11, + "learning_rate": 1.969080937440529e-05, + "loss": 1.0664, + "step": 3346 + }, + { + "epoch": 0.11, + "learning_rate": 1.969055374504489e-05, + "loss": 1.1465, + "step": 3347 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690298011715575e-05, + "loss": 1.1636, + "step": 3348 + }, + { + "epoch": 0.11, + "learning_rate": 1.969004217442009e-05, + "loss": 1.0181, + "step": 3349 + }, + { + "epoch": 0.11, + "learning_rate": 1.968978623316117e-05, + "loss": 1.043, + "step": 3350 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689530187941575e-05, + "loss": 1.0981, + "step": 3351 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689274038764043e-05, + "loss": 1.1357, + "step": 3352 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689017785631323e-05, + "loss": 1.1626, + "step": 3353 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688761428546162e-05, + "loss": 0.9766, + "step": 3354 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688504967511316e-05, + "loss": 1.1445, + "step": 3355 + }, + { + "epoch": 0.11, + "learning_rate": 1.968824840252954e-05, + "loss": 1.0947, + "step": 3356 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687991733603574e-05, + "loss": 1.2378, + "step": 3357 + }, + { + "epoch": 0.11, + "learning_rate": 1.968773496073618e-05, + "loss": 1.2656, + "step": 3358 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687478083930114e-05, + "loss": 1.1084, + "step": 3359 + }, + { + "epoch": 0.11, + "learning_rate": 1.968722110318813e-05, + "loss": 1.2334, + "step": 3360 + }, + { + "epoch": 0.11, + "learning_rate": 1.968696401851298e-05, + "loss": 1.1567, + "step": 3361 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686706829907434e-05, + "loss": 1.2559, + "step": 3362 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686449537374242e-05, + "loss": 1.1499, + "step": 3363 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686192140916165e-05, + "loss": 1.0996, + "step": 3364 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685934640535968e-05, + "loss": 1.1294, + "step": 3365 + }, + { + "epoch": 0.11, + "learning_rate": 1.968567703623641e-05, + "loss": 1.1748, + "step": 3366 + }, + { + "epoch": 0.11, + "learning_rate": 1.968541932802026e-05, + "loss": 1.083, + "step": 3367 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685161515890278e-05, + "loss": 1.186, + "step": 3368 + }, + { + "epoch": 0.11, + "learning_rate": 1.968490359984923e-05, + "loss": 1.1597, + "step": 3369 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684645579899892e-05, + "loss": 1.124, + "step": 3370 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684387456045017e-05, + "loss": 1.1641, + "step": 3371 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684129228287386e-05, + "loss": 1.1294, + "step": 3372 + }, + { + "epoch": 0.11, + "learning_rate": 1.968387089662977e-05, + "loss": 1.1548, + "step": 3373 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683612461074936e-05, + "loss": 1.0806, + "step": 3374 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683353921625654e-05, + "loss": 1.084, + "step": 3375 + }, + { + "epoch": 0.11, + "learning_rate": 1.96830952782847e-05, + "loss": 1.123, + "step": 3376 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682836531054854e-05, + "loss": 1.0234, + "step": 3377 + }, + { + "epoch": 0.11, + "learning_rate": 1.968257767993889e-05, + "loss": 1.2197, + "step": 3378 + }, + { + "epoch": 0.11, + "learning_rate": 1.968231872493958e-05, + "loss": 1.2314, + "step": 3379 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682059666059707e-05, + "loss": 1.1157, + "step": 3380 + }, + { + "epoch": 0.11, + "learning_rate": 1.968180050330205e-05, + "loss": 1.0562, + "step": 3381 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681541236669388e-05, + "loss": 1.1392, + "step": 3382 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681281866164507e-05, + "loss": 1.0557, + "step": 3383 + }, + { + "epoch": 0.11, + "learning_rate": 1.968102239179018e-05, + "loss": 1.1279, + "step": 3384 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680762813549202e-05, + "loss": 1.1777, + "step": 3385 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680503131444352e-05, + "loss": 0.998, + "step": 3386 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680243345478416e-05, + "loss": 1.0444, + "step": 3387 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679983455654182e-05, + "loss": 1.1406, + "step": 3388 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679723461974443e-05, + "loss": 0.9829, + "step": 3389 + }, + { + "epoch": 0.11, + "learning_rate": 1.967946336444198e-05, + "loss": 1.2217, + "step": 3390 + }, + { + "epoch": 0.11, + "learning_rate": 1.967920316305959e-05, + "loss": 1.0957, + "step": 3391 + }, + { + "epoch": 0.11, + "learning_rate": 1.967894285783006e-05, + "loss": 1.2954, + "step": 3392 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678682448756188e-05, + "loss": 1.1167, + "step": 3393 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678421935840762e-05, + "loss": 1.0767, + "step": 3394 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678161319086585e-05, + "loss": 1.1694, + "step": 3395 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677900598496445e-05, + "loss": 1.1465, + "step": 3396 + }, + { + "epoch": 0.11, + "learning_rate": 1.967763977407314e-05, + "loss": 1.0605, + "step": 3397 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677378845819473e-05, + "loss": 1.1196, + "step": 3398 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677117813738245e-05, + "loss": 1.2148, + "step": 3399 + }, + { + "epoch": 0.11, + "learning_rate": 1.967685667783225e-05, + "loss": 1.0474, + "step": 3400 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676595438104292e-05, + "loss": 1.0962, + "step": 3401 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676334094557178e-05, + "loss": 1.1123, + "step": 3402 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676072647193704e-05, + "loss": 1.064, + "step": 3403 + }, + { + "epoch": 0.11, + "learning_rate": 1.9675811096016682e-05, + "loss": 1.0933, + "step": 3404 + }, + { + "epoch": 0.11, + "learning_rate": 1.9675549441028913e-05, + "loss": 1.1875, + "step": 3405 + }, + { + "epoch": 0.11, + "learning_rate": 1.967528768223321e-05, + "loss": 1.1362, + "step": 3406 + }, + { + "epoch": 0.11, + "learning_rate": 1.967502581963238e-05, + "loss": 1.0879, + "step": 3407 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674763853229226e-05, + "loss": 1.0913, + "step": 3408 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674501783026566e-05, + "loss": 1.1592, + "step": 3409 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674239609027208e-05, + "loss": 1.0542, + "step": 3410 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673977331233964e-05, + "loss": 1.0752, + "step": 3411 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673714949649653e-05, + "loss": 0.9517, + "step": 3412 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673452464277084e-05, + "loss": 1.0942, + "step": 3413 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673189875119082e-05, + "loss": 1.1025, + "step": 3414 + }, + { + "epoch": 0.11, + "learning_rate": 1.967292718217845e-05, + "loss": 1.042, + "step": 3415 + }, + { + "epoch": 0.11, + "learning_rate": 1.967266438545802e-05, + "loss": 1.0552, + "step": 3416 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672401484960607e-05, + "loss": 1.2402, + "step": 3417 + }, + { + "epoch": 0.11, + "learning_rate": 1.967213848068903e-05, + "loss": 1.084, + "step": 3418 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671875372646113e-05, + "loss": 1.1328, + "step": 3419 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671612160834673e-05, + "loss": 1.1167, + "step": 3420 + }, + { + "epoch": 0.11, + "learning_rate": 1.967134884525754e-05, + "loss": 1.1445, + "step": 3421 + }, + { + "epoch": 0.11, + "learning_rate": 1.967108542591754e-05, + "loss": 1.1558, + "step": 3422 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670821902817496e-05, + "loss": 1.0762, + "step": 3423 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670558275960237e-05, + "loss": 1.0439, + "step": 3424 + }, + { + "epoch": 0.11, + "learning_rate": 1.967029454534859e-05, + "loss": 1.2085, + "step": 3425 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670030710985383e-05, + "loss": 1.1714, + "step": 3426 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669766772873447e-05, + "loss": 1.145, + "step": 3427 + }, + { + "epoch": 0.11, + "learning_rate": 1.966950273101562e-05, + "loss": 1.1216, + "step": 3428 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669238585414728e-05, + "loss": 1.0093, + "step": 3429 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668974336073613e-05, + "loss": 1.1016, + "step": 3430 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668709982995096e-05, + "loss": 1.0254, + "step": 3431 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668445526182026e-05, + "loss": 1.0547, + "step": 3432 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668180965637236e-05, + "loss": 1.0518, + "step": 3433 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667916301363564e-05, + "loss": 1.0181, + "step": 3434 + }, + { + "epoch": 0.11, + "learning_rate": 1.966765153336385e-05, + "loss": 1.1069, + "step": 3435 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667386661640934e-05, + "loss": 1.1821, + "step": 3436 + }, + { + "epoch": 0.11, + "learning_rate": 1.966712168619766e-05, + "loss": 1.1738, + "step": 3437 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666856607036867e-05, + "loss": 1.2314, + "step": 3438 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666591424161402e-05, + "loss": 1.0776, + "step": 3439 + }, + { + "epoch": 0.11, + "learning_rate": 1.966632613757411e-05, + "loss": 1.0825, + "step": 3440 + }, + { + "epoch": 0.11, + "learning_rate": 1.966606074727784e-05, + "loss": 1.0146, + "step": 3441 + }, + { + "epoch": 0.11, + "learning_rate": 1.966579525327543e-05, + "loss": 1.1064, + "step": 3442 + }, + { + "epoch": 0.11, + "learning_rate": 1.966552965556974e-05, + "loss": 1.0532, + "step": 3443 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665263954163613e-05, + "loss": 1.1392, + "step": 3444 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664998149059896e-05, + "loss": 1.0171, + "step": 3445 + }, + { + "epoch": 0.11, + "learning_rate": 1.966473224026145e-05, + "loss": 1.0469, + "step": 3446 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664466227771122e-05, + "loss": 1.1108, + "step": 3447 + }, + { + "epoch": 0.11, + "learning_rate": 1.966420011159177e-05, + "loss": 1.0264, + "step": 3448 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663933891726244e-05, + "loss": 1.1353, + "step": 3449 + }, + { + "epoch": 0.11, + "learning_rate": 1.96636675681774e-05, + "loss": 1.0488, + "step": 3450 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663401140948103e-05, + "loss": 0.999, + "step": 3451 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663134610041207e-05, + "loss": 1.1089, + "step": 3452 + }, + { + "epoch": 0.11, + "learning_rate": 1.966286797545957e-05, + "loss": 1.0464, + "step": 3453 + }, + { + "epoch": 0.11, + "learning_rate": 1.966260123720605e-05, + "loss": 1.0493, + "step": 3454 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662334395283516e-05, + "loss": 1.0984, + "step": 3455 + }, + { + "epoch": 0.11, + "learning_rate": 1.966206744969483e-05, + "loss": 1.0308, + "step": 3456 + }, + { + "epoch": 0.11, + "learning_rate": 1.966180040044285e-05, + "loss": 1.0693, + "step": 3457 + }, + { + "epoch": 0.11, + "learning_rate": 1.966153324753045e-05, + "loss": 1.1592, + "step": 3458 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661265990960486e-05, + "loss": 0.4856, + "step": 3459 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660998630735832e-05, + "loss": 1.1113, + "step": 3460 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660731166859356e-05, + "loss": 1.21, + "step": 3461 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660463599333925e-05, + "loss": 1.0996, + "step": 3462 + }, + { + "epoch": 0.11, + "learning_rate": 1.966019592816241e-05, + "loss": 0.5281, + "step": 3463 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659928153347686e-05, + "loss": 1.2886, + "step": 3464 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659660274892625e-05, + "loss": 1.0098, + "step": 3465 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659392292800098e-05, + "loss": 1.1348, + "step": 3466 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659124207072984e-05, + "loss": 1.1162, + "step": 3467 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658856017714158e-05, + "loss": 1.2212, + "step": 3468 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658587724726495e-05, + "loss": 1.0444, + "step": 3469 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658319328112872e-05, + "loss": 1.1045, + "step": 3470 + }, + { + "epoch": 0.11, + "learning_rate": 1.965805082787618e-05, + "loss": 0.9839, + "step": 3471 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657782224019284e-05, + "loss": 1.2217, + "step": 3472 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657513516545077e-05, + "loss": 1.1611, + "step": 3473 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657244705456438e-05, + "loss": 1.1245, + "step": 3474 + }, + { + "epoch": 0.11, + "learning_rate": 1.965697579075625e-05, + "loss": 1.0386, + "step": 3475 + }, + { + "epoch": 0.11, + "learning_rate": 1.96567067724474e-05, + "loss": 1.1924, + "step": 3476 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656437650532777e-05, + "loss": 1.103, + "step": 3477 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656168425015264e-05, + "loss": 1.1299, + "step": 3478 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655899095897746e-05, + "loss": 1.1489, + "step": 3479 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655629663183126e-05, + "loss": 1.1533, + "step": 3480 + }, + { + "epoch": 0.11, + "learning_rate": 1.965536012687428e-05, + "loss": 1.0278, + "step": 3481 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655090486974105e-05, + "loss": 1.0806, + "step": 3482 + }, + { + "epoch": 0.11, + "learning_rate": 1.96548207434855e-05, + "loss": 1.2188, + "step": 3483 + }, + { + "epoch": 0.11, + "learning_rate": 1.965455089641135e-05, + "loss": 0.9893, + "step": 3484 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654280945754554e-05, + "loss": 1.0107, + "step": 3485 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654010891518006e-05, + "loss": 1.0415, + "step": 3486 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653740733704612e-05, + "loss": 1.1006, + "step": 3487 + }, + { + "epoch": 0.11, + "learning_rate": 1.965347047231726e-05, + "loss": 1.1797, + "step": 3488 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653200107358855e-05, + "loss": 1.1558, + "step": 3489 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652929638832295e-05, + "loss": 0.9634, + "step": 3490 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652659066740484e-05, + "loss": 1.1816, + "step": 3491 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652388391086322e-05, + "loss": 1.0962, + "step": 3492 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652117611872722e-05, + "loss": 1.1011, + "step": 3493 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651846729102576e-05, + "loss": 1.1099, + "step": 3494 + }, + { + "epoch": 0.11, + "learning_rate": 1.96515757427788e-05, + "loss": 1.0923, + "step": 3495 + }, + { + "epoch": 0.11, + "learning_rate": 1.96513046529043e-05, + "loss": 1.1133, + "step": 3496 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651033459481976e-05, + "loss": 1.0732, + "step": 3497 + }, + { + "epoch": 0.11, + "learning_rate": 1.965076216251475e-05, + "loss": 1.0386, + "step": 3498 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650490762005524e-05, + "loss": 1.1128, + "step": 3499 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650219257957212e-05, + "loss": 0.9824, + "step": 3500 + }, + { + "epoch": 0.11, + "learning_rate": 1.964994765037273e-05, + "loss": 1.1611, + "step": 3501 + }, + { + "epoch": 0.11, + "learning_rate": 1.964967593925499e-05, + "loss": 1.0273, + "step": 3502 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649404124606905e-05, + "loss": 1.0054, + "step": 3503 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649132206431395e-05, + "loss": 1.2207, + "step": 3504 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648860184731372e-05, + "loss": 1.1611, + "step": 3505 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648588059509764e-05, + "loss": 1.1187, + "step": 3506 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648315830769478e-05, + "loss": 1.0454, + "step": 3507 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648043498513445e-05, + "loss": 0.999, + "step": 3508 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647771062744585e-05, + "loss": 1.3369, + "step": 3509 + }, + { + "epoch": 0.11, + "learning_rate": 1.964749852346582e-05, + "loss": 1.1719, + "step": 3510 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647225880680066e-05, + "loss": 1.0498, + "step": 3511 + }, + { + "epoch": 0.11, + "learning_rate": 1.964695313439026e-05, + "loss": 1.2812, + "step": 3512 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646680284599324e-05, + "loss": 1.1333, + "step": 3513 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646407331310186e-05, + "loss": 1.1279, + "step": 3514 + }, + { + "epoch": 0.11, + "learning_rate": 1.964613427452577e-05, + "loss": 1.1318, + "step": 3515 + }, + { + "epoch": 0.11, + "learning_rate": 1.964586111424901e-05, + "loss": 1.0835, + "step": 3516 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645587850482836e-05, + "loss": 1.1748, + "step": 3517 + }, + { + "epoch": 0.11, + "learning_rate": 1.964531448323018e-05, + "loss": 1.1499, + "step": 3518 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645041012493974e-05, + "loss": 1.1182, + "step": 3519 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644767438277153e-05, + "loss": 1.0464, + "step": 3520 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644493760582653e-05, + "loss": 1.0908, + "step": 3521 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644219979413405e-05, + "loss": 1.0776, + "step": 3522 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643946094772354e-05, + "loss": 1.1177, + "step": 3523 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643672106662434e-05, + "loss": 1.0835, + "step": 3524 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643398015086585e-05, + "loss": 0.9902, + "step": 3525 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643123820047746e-05, + "loss": 1.0625, + "step": 3526 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642849521548867e-05, + "loss": 1.0625, + "step": 3527 + }, + { + "epoch": 0.11, + "learning_rate": 1.964257511959288e-05, + "loss": 1.0635, + "step": 3528 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642300614182733e-05, + "loss": 0.9546, + "step": 3529 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642026005321373e-05, + "loss": 1.0728, + "step": 3530 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641751293011747e-05, + "loss": 1.1826, + "step": 3531 + }, + { + "epoch": 0.11, + "learning_rate": 1.96414764772568e-05, + "loss": 1.0498, + "step": 3532 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641201558059482e-05, + "loss": 1.1348, + "step": 3533 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640926535422738e-05, + "loss": 1.1333, + "step": 3534 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640651409349525e-05, + "loss": 1.0356, + "step": 3535 + }, + { + "epoch": 0.11, + "learning_rate": 1.964037617984279e-05, + "loss": 1.1372, + "step": 3536 + }, + { + "epoch": 0.11, + "learning_rate": 1.964010084690549e-05, + "loss": 1.2051, + "step": 3537 + }, + { + "epoch": 0.11, + "learning_rate": 1.963982541054058e-05, + "loss": 0.9668, + "step": 3538 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639549870751006e-05, + "loss": 0.9141, + "step": 3539 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639274227539734e-05, + "loss": 1.1738, + "step": 3540 + }, + { + "epoch": 0.11, + "learning_rate": 1.963899848090972e-05, + "loss": 1.0923, + "step": 3541 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638722630863914e-05, + "loss": 1.1182, + "step": 3542 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638446677405286e-05, + "loss": 1.1655, + "step": 3543 + }, + { + "epoch": 0.11, + "learning_rate": 1.963817062053679e-05, + "loss": 1.1128, + "step": 3544 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637894460261394e-05, + "loss": 1.1221, + "step": 3545 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637618196582055e-05, + "loss": 1.1802, + "step": 3546 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637341829501738e-05, + "loss": 1.0352, + "step": 3547 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637065359023412e-05, + "loss": 1.0029, + "step": 3548 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636788785150037e-05, + "loss": 1.1685, + "step": 3549 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636512107884584e-05, + "loss": 1.0742, + "step": 3550 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636235327230023e-05, + "loss": 1.2065, + "step": 3551 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635958443189324e-05, + "loss": 1.083, + "step": 3552 + }, + { + "epoch": 0.11, + "learning_rate": 1.963568145576545e-05, + "loss": 1.1055, + "step": 3553 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635404364961383e-05, + "loss": 1.1548, + "step": 3554 + }, + { + "epoch": 0.11, + "learning_rate": 1.963512717078009e-05, + "loss": 1.1226, + "step": 3555 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634849873224545e-05, + "loss": 1.0645, + "step": 3556 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634572472297725e-05, + "loss": 1.0308, + "step": 3557 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634294968002604e-05, + "loss": 1.1436, + "step": 3558 + }, + { + "epoch": 0.11, + "learning_rate": 1.963401736034216e-05, + "loss": 0.4414, + "step": 3559 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633739649319375e-05, + "loss": 1.0205, + "step": 3560 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633461834937226e-05, + "loss": 1.1646, + "step": 3561 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633183917198694e-05, + "loss": 1.1089, + "step": 3562 + }, + { + "epoch": 0.11, + "learning_rate": 1.9632905896106757e-05, + "loss": 1.1509, + "step": 3563 + }, + { + "epoch": 0.11, + "learning_rate": 1.9632627771664402e-05, + "loss": 1.2241, + "step": 3564 + }, + { + "epoch": 0.11, + "learning_rate": 1.9632349543874613e-05, + "loss": 1.1636, + "step": 3565 + }, + { + "epoch": 0.11, + "learning_rate": 1.963207121274037e-05, + "loss": 1.0879, + "step": 3566 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631792778264665e-05, + "loss": 1.1548, + "step": 3567 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631514240450483e-05, + "loss": 1.1768, + "step": 3568 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631235599300816e-05, + "loss": 1.0669, + "step": 3569 + }, + { + "epoch": 0.11, + "learning_rate": 1.963095685481865e-05, + "loss": 1.1279, + "step": 3570 + }, + { + "epoch": 0.11, + "learning_rate": 1.963067800700697e-05, + "loss": 1.2261, + "step": 3571 + }, + { + "epoch": 0.11, + "learning_rate": 1.963039905586878e-05, + "loss": 1.1069, + "step": 3572 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630120001407066e-05, + "loss": 1.0967, + "step": 3573 + }, + { + "epoch": 0.11, + "learning_rate": 1.962984084362482e-05, + "loss": 1.1089, + "step": 3574 + }, + { + "epoch": 0.11, + "learning_rate": 1.962956158252504e-05, + "loss": 1.0376, + "step": 3575 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629282218110722e-05, + "loss": 1.0596, + "step": 3576 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629002750384865e-05, + "loss": 1.1582, + "step": 3577 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628723179350465e-05, + "loss": 1.2422, + "step": 3578 + }, + { + "epoch": 0.11, + "learning_rate": 1.962844350501052e-05, + "loss": 1.1689, + "step": 3579 + }, + { + "epoch": 0.11, + "learning_rate": 1.962816372736803e-05, + "loss": 1.0723, + "step": 3580 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627883846426003e-05, + "loss": 1.1157, + "step": 3581 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627603862187437e-05, + "loss": 1.1685, + "step": 3582 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627323774655334e-05, + "loss": 1.1201, + "step": 3583 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627043583832707e-05, + "loss": 1.1016, + "step": 3584 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626763289722553e-05, + "loss": 1.0552, + "step": 3585 + }, + { + "epoch": 0.11, + "learning_rate": 1.962648289232789e-05, + "loss": 1.1313, + "step": 3586 + }, + { + "epoch": 0.11, + "learning_rate": 1.962620239165171e-05, + "loss": 1.0913, + "step": 3587 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625921787697038e-05, + "loss": 1.1147, + "step": 3588 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625641080466876e-05, + "loss": 1.1836, + "step": 3589 + }, + { + "epoch": 0.11, + "learning_rate": 1.962536026996424e-05, + "loss": 1.0557, + "step": 3590 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625079356192142e-05, + "loss": 0.981, + "step": 3591 + }, + { + "epoch": 0.11, + "learning_rate": 1.962479833915359e-05, + "loss": 1.0635, + "step": 3592 + }, + { + "epoch": 0.11, + "learning_rate": 1.962451721885161e-05, + "loss": 1.1636, + "step": 3593 + }, + { + "epoch": 0.11, + "learning_rate": 1.9624235995289212e-05, + "loss": 1.0322, + "step": 3594 + }, + { + "epoch": 0.11, + "learning_rate": 1.962395466846941e-05, + "loss": 1.0454, + "step": 3595 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623673238395227e-05, + "loss": 1.2124, + "step": 3596 + }, + { + "epoch": 0.12, + "learning_rate": 1.962339170506968e-05, + "loss": 1.1523, + "step": 3597 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623110068495793e-05, + "loss": 1.0337, + "step": 3598 + }, + { + "epoch": 0.12, + "learning_rate": 1.962282832867658e-05, + "loss": 0.9639, + "step": 3599 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622546485615078e-05, + "loss": 1.2285, + "step": 3600 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622264539314295e-05, + "loss": 1.0117, + "step": 3601 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621982489777266e-05, + "loss": 1.1987, + "step": 3602 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621700337007012e-05, + "loss": 1.0293, + "step": 3603 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621418081006563e-05, + "loss": 0.9487, + "step": 3604 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621135721778946e-05, + "loss": 1.1138, + "step": 3605 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620853259327192e-05, + "loss": 1.0737, + "step": 3606 + }, + { + "epoch": 0.12, + "learning_rate": 1.962057069365433e-05, + "loss": 1.1431, + "step": 3607 + }, + { + "epoch": 0.12, + "learning_rate": 1.962028802476339e-05, + "loss": 1.0654, + "step": 3608 + }, + { + "epoch": 0.12, + "learning_rate": 1.962000525265741e-05, + "loss": 1.1523, + "step": 3609 + }, + { + "epoch": 0.12, + "learning_rate": 1.961972237733942e-05, + "loss": 1.1206, + "step": 3610 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619439398812453e-05, + "loss": 1.0317, + "step": 3611 + }, + { + "epoch": 0.12, + "learning_rate": 1.961915631707955e-05, + "loss": 1.062, + "step": 3612 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618873132143742e-05, + "loss": 1.0913, + "step": 3613 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618589844008076e-05, + "loss": 1.1475, + "step": 3614 + }, + { + "epoch": 0.12, + "learning_rate": 1.961830645267558e-05, + "loss": 1.1182, + "step": 3615 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618022958149305e-05, + "loss": 1.1948, + "step": 3616 + }, + { + "epoch": 0.12, + "learning_rate": 1.961773936043229e-05, + "loss": 1.1274, + "step": 3617 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617455659527574e-05, + "loss": 1.0957, + "step": 3618 + }, + { + "epoch": 0.12, + "learning_rate": 1.96171718554382e-05, + "loss": 1.2393, + "step": 3619 + }, + { + "epoch": 0.12, + "learning_rate": 1.961688794816722e-05, + "loss": 1.2021, + "step": 3620 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616603937717676e-05, + "loss": 1.0806, + "step": 3621 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616319824092614e-05, + "loss": 1.1562, + "step": 3622 + }, + { + "epoch": 0.12, + "learning_rate": 1.961603560729508e-05, + "loss": 1.1401, + "step": 3623 + }, + { + "epoch": 0.12, + "learning_rate": 1.961575128732813e-05, + "loss": 1.1265, + "step": 3624 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615466864194806e-05, + "loss": 1.1113, + "step": 3625 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615182337898172e-05, + "loss": 1.1323, + "step": 3626 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614897708441267e-05, + "loss": 0.9932, + "step": 3627 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614612975827153e-05, + "loss": 0.4978, + "step": 3628 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614328140058886e-05, + "loss": 1.0947, + "step": 3629 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614043201139513e-05, + "loss": 1.0474, + "step": 3630 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613758159072097e-05, + "loss": 1.1514, + "step": 3631 + }, + { + "epoch": 0.12, + "learning_rate": 1.96134730138597e-05, + "loss": 1.2705, + "step": 3632 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613187765505374e-05, + "loss": 1.0605, + "step": 3633 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612902414012183e-05, + "loss": 1.1733, + "step": 3634 + }, + { + "epoch": 0.12, + "learning_rate": 1.961261695938319e-05, + "loss": 1.0664, + "step": 3635 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612331401621455e-05, + "loss": 0.9976, + "step": 3636 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612045740730042e-05, + "loss": 0.5149, + "step": 3637 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611759976712016e-05, + "loss": 1.1206, + "step": 3638 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611474109570446e-05, + "loss": 1.0391, + "step": 3639 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611188139308392e-05, + "loss": 1.2021, + "step": 3640 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610902065928926e-05, + "loss": 1.1167, + "step": 3641 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610615889435124e-05, + "loss": 1.0376, + "step": 3642 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610329609830044e-05, + "loss": 1.1216, + "step": 3643 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610043227116765e-05, + "loss": 1.041, + "step": 3644 + }, + { + "epoch": 0.12, + "learning_rate": 1.960975674129836e-05, + "loss": 1.001, + "step": 3645 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609470152377897e-05, + "loss": 1.04, + "step": 3646 + }, + { + "epoch": 0.12, + "learning_rate": 1.960918346035846e-05, + "loss": 1.0283, + "step": 3647 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608896665243113e-05, + "loss": 1.0234, + "step": 3648 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608609767034943e-05, + "loss": 0.5027, + "step": 3649 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608322765737026e-05, + "loss": 1.126, + "step": 3650 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608035661352437e-05, + "loss": 1.1606, + "step": 3651 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607748453884258e-05, + "loss": 1.0713, + "step": 3652 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607461143335574e-05, + "loss": 1.1362, + "step": 3653 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607173729709462e-05, + "loss": 0.4573, + "step": 3654 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606886213009012e-05, + "loss": 1.1567, + "step": 3655 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606598593237303e-05, + "loss": 1.0947, + "step": 3656 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606310870397426e-05, + "loss": 1.0449, + "step": 3657 + }, + { + "epoch": 0.12, + "learning_rate": 1.960602304449246e-05, + "loss": 1.1836, + "step": 3658 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605735115525502e-05, + "loss": 1.1851, + "step": 3659 + }, + { + "epoch": 0.12, + "learning_rate": 1.960544708349964e-05, + "loss": 1.084, + "step": 3660 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605158948417955e-05, + "loss": 1.0161, + "step": 3661 + }, + { + "epoch": 0.12, + "learning_rate": 1.960487071028355e-05, + "loss": 1.0928, + "step": 3662 + }, + { + "epoch": 0.12, + "learning_rate": 1.960458236909951e-05, + "loss": 1.1035, + "step": 3663 + }, + { + "epoch": 0.12, + "learning_rate": 1.960429392486893e-05, + "loss": 1.127, + "step": 3664 + }, + { + "epoch": 0.12, + "learning_rate": 1.960400537759491e-05, + "loss": 1.1899, + "step": 3665 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603716727280538e-05, + "loss": 1.1299, + "step": 3666 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603427973928915e-05, + "loss": 1.0454, + "step": 3667 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603139117543137e-05, + "loss": 1.0293, + "step": 3668 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602850158126308e-05, + "loss": 1.0703, + "step": 3669 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602561095681524e-05, + "loss": 1.1309, + "step": 3670 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602271930211882e-05, + "loss": 1.0332, + "step": 3671 + }, + { + "epoch": 0.12, + "learning_rate": 1.9601982661720496e-05, + "loss": 1.2393, + "step": 3672 + }, + { + "epoch": 0.12, + "learning_rate": 1.960169329021046e-05, + "loss": 1.1196, + "step": 3673 + }, + { + "epoch": 0.12, + "learning_rate": 1.960140381568488e-05, + "loss": 1.001, + "step": 3674 + }, + { + "epoch": 0.12, + "learning_rate": 1.9601114238146867e-05, + "loss": 1.2231, + "step": 3675 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600824557599522e-05, + "loss": 1.1943, + "step": 3676 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600534774045954e-05, + "loss": 1.0308, + "step": 3677 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600244887489273e-05, + "loss": 1.0781, + "step": 3678 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599954897932593e-05, + "loss": 1.1436, + "step": 3679 + }, + { + "epoch": 0.12, + "learning_rate": 1.959966480537902e-05, + "loss": 1.1372, + "step": 3680 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599374609831667e-05, + "loss": 1.0698, + "step": 3681 + }, + { + "epoch": 0.12, + "learning_rate": 1.959908431129365e-05, + "loss": 1.1782, + "step": 3682 + }, + { + "epoch": 0.12, + "learning_rate": 1.959879390976808e-05, + "loss": 1.0396, + "step": 3683 + }, + { + "epoch": 0.12, + "learning_rate": 1.959850340525808e-05, + "loss": 1.1787, + "step": 3684 + }, + { + "epoch": 0.12, + "learning_rate": 1.9598212797766754e-05, + "loss": 1.1001, + "step": 3685 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597922087297233e-05, + "loss": 1.1553, + "step": 3686 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597631273852628e-05, + "loss": 1.1392, + "step": 3687 + }, + { + "epoch": 0.12, + "learning_rate": 1.959734035743606e-05, + "loss": 1.1104, + "step": 3688 + }, + { + "epoch": 0.12, + "learning_rate": 1.959704933805066e-05, + "loss": 1.1143, + "step": 3689 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596758215699532e-05, + "loss": 1.0591, + "step": 3690 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596466990385814e-05, + "loss": 1.1289, + "step": 3691 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596175662112626e-05, + "loss": 1.1675, + "step": 3692 + }, + { + "epoch": 0.12, + "learning_rate": 1.959588423088309e-05, + "loss": 1.1401, + "step": 3693 + }, + { + "epoch": 0.12, + "learning_rate": 1.959559269670034e-05, + "loss": 1.2134, + "step": 3694 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595301059567502e-05, + "loss": 1.0293, + "step": 3695 + }, + { + "epoch": 0.12, + "learning_rate": 1.95950093194877e-05, + "loss": 1.0078, + "step": 3696 + }, + { + "epoch": 0.12, + "learning_rate": 1.959471747646407e-05, + "loss": 1.1294, + "step": 3697 + }, + { + "epoch": 0.12, + "learning_rate": 1.959442553049974e-05, + "loss": 1.166, + "step": 3698 + }, + { + "epoch": 0.12, + "learning_rate": 1.9594133481597837e-05, + "loss": 1.0698, + "step": 3699 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593841329761508e-05, + "loss": 0.9766, + "step": 3700 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593549074993877e-05, + "loss": 0.9819, + "step": 3701 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593256717298078e-05, + "loss": 1.2334, + "step": 3702 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592964256677253e-05, + "loss": 1.1787, + "step": 3703 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592671693134542e-05, + "loss": 1.2192, + "step": 3704 + }, + { + "epoch": 0.12, + "learning_rate": 1.959237902667308e-05, + "loss": 1.0815, + "step": 3705 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592086257296005e-05, + "loss": 1.0, + "step": 3706 + }, + { + "epoch": 0.12, + "learning_rate": 1.959179338500646e-05, + "loss": 1.0269, + "step": 3707 + }, + { + "epoch": 0.12, + "learning_rate": 1.959150040980759e-05, + "loss": 1.2065, + "step": 3708 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591207331702536e-05, + "loss": 1.1826, + "step": 3709 + }, + { + "epoch": 0.12, + "learning_rate": 1.959091415069444e-05, + "loss": 1.084, + "step": 3710 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590620866786453e-05, + "loss": 0.938, + "step": 3711 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590327479981714e-05, + "loss": 1.125, + "step": 3712 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590033990283377e-05, + "loss": 1.0405, + "step": 3713 + }, + { + "epoch": 0.12, + "learning_rate": 1.958974039769459e-05, + "loss": 1.0093, + "step": 3714 + }, + { + "epoch": 0.12, + "learning_rate": 1.95894467022185e-05, + "loss": 1.1548, + "step": 3715 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589152903858264e-05, + "loss": 1.1641, + "step": 3716 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588859002617026e-05, + "loss": 1.2168, + "step": 3717 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588564998497945e-05, + "loss": 1.2256, + "step": 3718 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588270891504174e-05, + "loss": 1.0342, + "step": 3719 + }, + { + "epoch": 0.12, + "learning_rate": 1.958797668163887e-05, + "loss": 1.1836, + "step": 3720 + }, + { + "epoch": 0.12, + "learning_rate": 1.9587682368905186e-05, + "loss": 1.2139, + "step": 3721 + }, + { + "epoch": 0.12, + "learning_rate": 1.958738795330628e-05, + "loss": 1.1909, + "step": 3722 + }, + { + "epoch": 0.12, + "learning_rate": 1.958709343484531e-05, + "loss": 1.1875, + "step": 3723 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586798813525443e-05, + "loss": 1.2383, + "step": 3724 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586504089349833e-05, + "loss": 1.3213, + "step": 3725 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586209262321642e-05, + "loss": 1.0479, + "step": 3726 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585914332444037e-05, + "loss": 1.1299, + "step": 3727 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585619299720182e-05, + "loss": 1.1523, + "step": 3728 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585324164153236e-05, + "loss": 1.1177, + "step": 3729 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585028925746374e-05, + "loss": 1.1133, + "step": 3730 + }, + { + "epoch": 0.12, + "learning_rate": 1.958473358450276e-05, + "loss": 1.0771, + "step": 3731 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584438140425563e-05, + "loss": 1.0449, + "step": 3732 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584142593517947e-05, + "loss": 1.1113, + "step": 3733 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583846943783093e-05, + "loss": 1.0801, + "step": 3734 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583551191224165e-05, + "loss": 1.2046, + "step": 3735 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583255335844345e-05, + "loss": 1.0596, + "step": 3736 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582959377646797e-05, + "loss": 1.2002, + "step": 3737 + }, + { + "epoch": 0.12, + "learning_rate": 1.95826633166347e-05, + "loss": 1.0972, + "step": 3738 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582367152811232e-05, + "loss": 1.2129, + "step": 3739 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582070886179572e-05, + "loss": 1.0229, + "step": 3740 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581774516742893e-05, + "loss": 1.0459, + "step": 3741 + }, + { + "epoch": 0.12, + "learning_rate": 1.958147804450438e-05, + "loss": 1.0869, + "step": 3742 + }, + { + "epoch": 0.12, + "learning_rate": 1.958118146946721e-05, + "loss": 1.2275, + "step": 3743 + }, + { + "epoch": 0.12, + "learning_rate": 1.958088479163457e-05, + "loss": 1.1968, + "step": 3744 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580588011009637e-05, + "loss": 1.0298, + "step": 3745 + }, + { + "epoch": 0.12, + "learning_rate": 1.95802911275956e-05, + "loss": 1.1348, + "step": 3746 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579994141395643e-05, + "loss": 1.1401, + "step": 3747 + }, + { + "epoch": 0.12, + "learning_rate": 1.957969705241295e-05, + "loss": 0.5542, + "step": 3748 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579399860650708e-05, + "loss": 0.9863, + "step": 3749 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579102566112108e-05, + "loss": 1.1548, + "step": 3750 + }, + { + "epoch": 0.12, + "learning_rate": 1.957880516880034e-05, + "loss": 1.1646, + "step": 3751 + }, + { + "epoch": 0.12, + "learning_rate": 1.9578507668718595e-05, + "loss": 1.1953, + "step": 3752 + }, + { + "epoch": 0.12, + "learning_rate": 1.957821006587006e-05, + "loss": 1.1509, + "step": 3753 + }, + { + "epoch": 0.12, + "learning_rate": 1.9577912360257936e-05, + "loss": 0.5515, + "step": 3754 + }, + { + "epoch": 0.12, + "learning_rate": 1.957761455188541e-05, + "loss": 1.1646, + "step": 3755 + }, + { + "epoch": 0.12, + "learning_rate": 1.957731664075568e-05, + "loss": 1.1255, + "step": 3756 + }, + { + "epoch": 0.12, + "learning_rate": 1.9577018626871945e-05, + "loss": 1.1528, + "step": 3757 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576720510237397e-05, + "loss": 1.1401, + "step": 3758 + }, + { + "epoch": 0.12, + "learning_rate": 1.957642229085524e-05, + "loss": 1.147, + "step": 3759 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576123968728665e-05, + "loss": 1.1611, + "step": 3760 + }, + { + "epoch": 0.12, + "learning_rate": 1.957582554386088e-05, + "loss": 1.043, + "step": 3761 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575527016255087e-05, + "loss": 0.9961, + "step": 3762 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575228385914484e-05, + "loss": 1.0718, + "step": 3763 + }, + { + "epoch": 0.12, + "learning_rate": 1.957492965284228e-05, + "loss": 1.127, + "step": 3764 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574630817041678e-05, + "loss": 1.0366, + "step": 3765 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574331878515883e-05, + "loss": 1.1309, + "step": 3766 + }, + { + "epoch": 0.12, + "learning_rate": 1.95740328372681e-05, + "loss": 0.4988, + "step": 3767 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573733693301545e-05, + "loss": 1.1626, + "step": 3768 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573434446619424e-05, + "loss": 1.1255, + "step": 3769 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573135097224945e-05, + "loss": 1.1626, + "step": 3770 + }, + { + "epoch": 0.12, + "learning_rate": 1.9572835645121322e-05, + "loss": 1.1699, + "step": 3771 + }, + { + "epoch": 0.12, + "learning_rate": 1.9572536090311765e-05, + "loss": 1.1621, + "step": 3772 + }, + { + "epoch": 0.12, + "learning_rate": 1.957223643279949e-05, + "loss": 1.2432, + "step": 3773 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571936672587718e-05, + "loss": 1.0781, + "step": 3774 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571636809679654e-05, + "loss": 1.0659, + "step": 3775 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571336844078522e-05, + "loss": 0.9736, + "step": 3776 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571036775787537e-05, + "loss": 1.2471, + "step": 3777 + }, + { + "epoch": 0.12, + "learning_rate": 1.957073660480992e-05, + "loss": 1.1299, + "step": 3778 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570436331148893e-05, + "loss": 1.1758, + "step": 3779 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570135954807675e-05, + "loss": 1.0986, + "step": 3780 + }, + { + "epoch": 0.12, + "learning_rate": 1.9569835475789492e-05, + "loss": 1.0493, + "step": 3781 + }, + { + "epoch": 0.12, + "learning_rate": 1.9569534894097563e-05, + "loss": 1.0386, + "step": 3782 + }, + { + "epoch": 0.12, + "learning_rate": 1.956923420973512e-05, + "loss": 1.0962, + "step": 3783 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568933422705377e-05, + "loss": 1.1445, + "step": 3784 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568632533011578e-05, + "loss": 1.1172, + "step": 3785 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568331540656938e-05, + "loss": 1.1953, + "step": 3786 + }, + { + "epoch": 0.12, + "learning_rate": 1.956803044564469e-05, + "loss": 1.146, + "step": 3787 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567729247978062e-05, + "loss": 1.1855, + "step": 3788 + }, + { + "epoch": 0.12, + "learning_rate": 1.956742794766029e-05, + "loss": 1.1299, + "step": 3789 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567126544694605e-05, + "loss": 1.2959, + "step": 3790 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566825039084244e-05, + "loss": 1.2432, + "step": 3791 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566523430832432e-05, + "loss": 1.104, + "step": 3792 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566221719942416e-05, + "loss": 1.1948, + "step": 3793 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565919906417428e-05, + "loss": 1.1411, + "step": 3794 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565617990260707e-05, + "loss": 1.1406, + "step": 3795 + }, + { + "epoch": 0.12, + "learning_rate": 1.956531597147549e-05, + "loss": 1.1694, + "step": 3796 + }, + { + "epoch": 0.12, + "learning_rate": 1.956501385006502e-05, + "loss": 1.0615, + "step": 3797 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564711626032533e-05, + "loss": 1.0811, + "step": 3798 + }, + { + "epoch": 0.12, + "learning_rate": 1.956440929938128e-05, + "loss": 0.583, + "step": 3799 + }, + { + "epoch": 0.12, + "learning_rate": 1.95641068701145e-05, + "loss": 1.1943, + "step": 3800 + }, + { + "epoch": 0.12, + "learning_rate": 1.956380433823544e-05, + "loss": 1.1006, + "step": 3801 + }, + { + "epoch": 0.12, + "learning_rate": 1.9563501703747342e-05, + "loss": 1.0303, + "step": 3802 + }, + { + "epoch": 0.12, + "learning_rate": 1.9563198966653455e-05, + "loss": 1.1631, + "step": 3803 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562896126957026e-05, + "loss": 1.0239, + "step": 3804 + }, + { + "epoch": 0.12, + "learning_rate": 1.956259318466131e-05, + "loss": 1.1055, + "step": 3805 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562290139769547e-05, + "loss": 0.5559, + "step": 3806 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561986992284998e-05, + "loss": 0.9482, + "step": 3807 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561683742210912e-05, + "loss": 1.0293, + "step": 3808 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561380389550538e-05, + "loss": 1.1587, + "step": 3809 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561076934307135e-05, + "loss": 1.0674, + "step": 3810 + }, + { + "epoch": 0.12, + "learning_rate": 1.956077337648396e-05, + "loss": 1.0161, + "step": 3811 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560469716084265e-05, + "loss": 1.2139, + "step": 3812 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560165953111312e-05, + "loss": 1.0933, + "step": 3813 + }, + { + "epoch": 0.12, + "learning_rate": 1.955986208756836e-05, + "loss": 1.0967, + "step": 3814 + }, + { + "epoch": 0.12, + "learning_rate": 1.955955811945867e-05, + "loss": 1.0649, + "step": 3815 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559254048785498e-05, + "loss": 1.1499, + "step": 3816 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558949875552114e-05, + "loss": 1.085, + "step": 3817 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558645599761772e-05, + "loss": 0.9937, + "step": 3818 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558341221417744e-05, + "loss": 1.1235, + "step": 3819 + }, + { + "epoch": 0.12, + "learning_rate": 1.955803674052329e-05, + "loss": 1.0366, + "step": 3820 + }, + { + "epoch": 0.12, + "learning_rate": 1.9557732157081686e-05, + "loss": 1.1499, + "step": 3821 + }, + { + "epoch": 0.12, + "learning_rate": 1.9557427471096192e-05, + "loss": 1.1309, + "step": 3822 + }, + { + "epoch": 0.12, + "learning_rate": 1.955712268257008e-05, + "loss": 1.2305, + "step": 3823 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556817791506616e-05, + "loss": 1.1504, + "step": 3824 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556512797909076e-05, + "loss": 1.1826, + "step": 3825 + }, + { + "epoch": 0.12, + "learning_rate": 1.955620770178073e-05, + "loss": 1.1177, + "step": 3826 + }, + { + "epoch": 0.12, + "learning_rate": 1.955590250312485e-05, + "loss": 1.1494, + "step": 3827 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555597201944714e-05, + "loss": 1.1011, + "step": 3828 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555291798243593e-05, + "loss": 1.1006, + "step": 3829 + }, + { + "epoch": 0.12, + "learning_rate": 1.955498629202477e-05, + "loss": 1.0977, + "step": 3830 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554680683291517e-05, + "loss": 1.0791, + "step": 3831 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554374972047112e-05, + "loss": 1.1274, + "step": 3832 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554069158294844e-05, + "loss": 1.0811, + "step": 3833 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553763242037985e-05, + "loss": 1.0698, + "step": 3834 + }, + { + "epoch": 0.12, + "learning_rate": 1.955345722327982e-05, + "loss": 1.2549, + "step": 3835 + }, + { + "epoch": 0.12, + "learning_rate": 1.955315110202363e-05, + "loss": 1.1172, + "step": 3836 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552844878272702e-05, + "loss": 1.1973, + "step": 3837 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552538552030325e-05, + "loss": 1.0278, + "step": 3838 + }, + { + "epoch": 0.12, + "learning_rate": 1.955223212329978e-05, + "loss": 1.1665, + "step": 3839 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551925592084355e-05, + "loss": 1.2539, + "step": 3840 + }, + { + "epoch": 0.12, + "learning_rate": 1.955161895838734e-05, + "loss": 1.1343, + "step": 3841 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551312222212022e-05, + "loss": 1.1377, + "step": 3842 + }, + { + "epoch": 0.12, + "learning_rate": 1.95510053835617e-05, + "loss": 1.002, + "step": 3843 + }, + { + "epoch": 0.12, + "learning_rate": 1.9550698442439657e-05, + "loss": 1.1196, + "step": 3844 + }, + { + "epoch": 0.12, + "learning_rate": 1.955039139884919e-05, + "loss": 1.1489, + "step": 3845 + }, + { + "epoch": 0.12, + "learning_rate": 1.95500842527936e-05, + "loss": 0.481, + "step": 3846 + }, + { + "epoch": 0.12, + "learning_rate": 1.9549777004276165e-05, + "loss": 1.0161, + "step": 3847 + }, + { + "epoch": 0.12, + "learning_rate": 1.9549469653300197e-05, + "loss": 1.1055, + "step": 3848 + }, + { + "epoch": 0.12, + "learning_rate": 1.954916219986899e-05, + "loss": 1.0269, + "step": 3849 + }, + { + "epoch": 0.12, + "learning_rate": 1.954885464398584e-05, + "loss": 1.3027, + "step": 3850 + }, + { + "epoch": 0.12, + "learning_rate": 1.9548546985654046e-05, + "loss": 1.0962, + "step": 3851 + }, + { + "epoch": 0.12, + "learning_rate": 1.9548239224876914e-05, + "loss": 1.1069, + "step": 3852 + }, + { + "epoch": 0.12, + "learning_rate": 1.954793136165774e-05, + "loss": 1.2075, + "step": 3853 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547623395999828e-05, + "loss": 1.1968, + "step": 3854 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547315327906487e-05, + "loss": 0.9419, + "step": 3855 + }, + { + "epoch": 0.12, + "learning_rate": 1.9547007157381016e-05, + "loss": 1.0669, + "step": 3856 + }, + { + "epoch": 0.12, + "learning_rate": 1.954669888442673e-05, + "loss": 1.1475, + "step": 3857 + }, + { + "epoch": 0.12, + "learning_rate": 1.9546390509046926e-05, + "loss": 1.1802, + "step": 3858 + }, + { + "epoch": 0.12, + "learning_rate": 1.9546082031244918e-05, + "loss": 1.1069, + "step": 3859 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545773451024016e-05, + "loss": 1.0801, + "step": 3860 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545464768387526e-05, + "loss": 1.1221, + "step": 3861 + }, + { + "epoch": 0.12, + "learning_rate": 1.9545155983338767e-05, + "loss": 1.0151, + "step": 3862 + }, + { + "epoch": 0.12, + "learning_rate": 1.9544847095881048e-05, + "loss": 1.0981, + "step": 3863 + }, + { + "epoch": 0.12, + "learning_rate": 1.9544538106017682e-05, + "loss": 1.0723, + "step": 3864 + }, + { + "epoch": 0.12, + "learning_rate": 1.9544229013751987e-05, + "loss": 1.1172, + "step": 3865 + }, + { + "epoch": 0.12, + "learning_rate": 1.9543919819087277e-05, + "loss": 1.0957, + "step": 3866 + }, + { + "epoch": 0.12, + "learning_rate": 1.9543610522026867e-05, + "loss": 1.0527, + "step": 3867 + }, + { + "epoch": 0.12, + "learning_rate": 1.9543301122574083e-05, + "loss": 1.0908, + "step": 3868 + }, + { + "epoch": 0.12, + "learning_rate": 1.954299162073224e-05, + "loss": 1.0586, + "step": 3869 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542682016504656e-05, + "loss": 1.1431, + "step": 3870 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542372309894655e-05, + "loss": 1.1084, + "step": 3871 + }, + { + "epoch": 0.12, + "learning_rate": 1.9542062500905564e-05, + "loss": 1.0767, + "step": 3872 + }, + { + "epoch": 0.12, + "learning_rate": 1.95417525895407e-05, + "loss": 1.019, + "step": 3873 + }, + { + "epoch": 0.12, + "learning_rate": 1.954144257580339e-05, + "loss": 1.0254, + "step": 3874 + }, + { + "epoch": 0.12, + "learning_rate": 1.9541132459696963e-05, + "loss": 1.1904, + "step": 3875 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540822241224742e-05, + "loss": 1.1411, + "step": 3876 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540511920390063e-05, + "loss": 1.105, + "step": 3877 + }, + { + "epoch": 0.12, + "learning_rate": 1.9540201497196244e-05, + "loss": 1.1191, + "step": 3878 + }, + { + "epoch": 0.12, + "learning_rate": 1.9539890971646624e-05, + "loss": 1.0869, + "step": 3879 + }, + { + "epoch": 0.12, + "learning_rate": 1.9539580343744534e-05, + "loss": 1.0835, + "step": 3880 + }, + { + "epoch": 0.12, + "learning_rate": 1.9539269613493302e-05, + "loss": 1.0723, + "step": 3881 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538958780896265e-05, + "loss": 1.1587, + "step": 3882 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538647845956756e-05, + "loss": 1.1689, + "step": 3883 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538336808678117e-05, + "loss": 1.1309, + "step": 3884 + }, + { + "epoch": 0.12, + "learning_rate": 1.9538025669063678e-05, + "loss": 1.1333, + "step": 3885 + }, + { + "epoch": 0.12, + "learning_rate": 1.953771442711678e-05, + "loss": 1.1865, + "step": 3886 + }, + { + "epoch": 0.12, + "learning_rate": 1.953740308284076e-05, + "loss": 1.0293, + "step": 3887 + }, + { + "epoch": 0.12, + "learning_rate": 1.9537091636238964e-05, + "loss": 1.1006, + "step": 3888 + }, + { + "epoch": 0.12, + "learning_rate": 1.9536780087314725e-05, + "loss": 1.124, + "step": 3889 + }, + { + "epoch": 0.12, + "learning_rate": 1.9536468436071393e-05, + "loss": 1.0454, + "step": 3890 + }, + { + "epoch": 0.12, + "learning_rate": 1.9536156682512312e-05, + "loss": 1.1392, + "step": 3891 + }, + { + "epoch": 0.12, + "learning_rate": 1.953584482664082e-05, + "loss": 1.1426, + "step": 3892 + }, + { + "epoch": 0.12, + "learning_rate": 1.953553286846027e-05, + "loss": 1.2256, + "step": 3893 + }, + { + "epoch": 0.12, + "learning_rate": 1.9535220807974002e-05, + "loss": 1.0547, + "step": 3894 + }, + { + "epoch": 0.12, + "learning_rate": 1.953490864518537e-05, + "loss": 1.1611, + "step": 3895 + }, + { + "epoch": 0.12, + "learning_rate": 1.953459638009772e-05, + "loss": 1.1333, + "step": 3896 + }, + { + "epoch": 0.12, + "learning_rate": 1.9534284012714405e-05, + "loss": 1.1987, + "step": 3897 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533971543038772e-05, + "loss": 1.085, + "step": 3898 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533658971074177e-05, + "loss": 1.1353, + "step": 3899 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533346296823974e-05, + "loss": 1.2329, + "step": 3900 + }, + { + "epoch": 0.12, + "learning_rate": 1.9533033520291517e-05, + "loss": 1.1074, + "step": 3901 + }, + { + "epoch": 0.12, + "learning_rate": 1.9532720641480157e-05, + "loss": 1.0127, + "step": 3902 + }, + { + "epoch": 0.12, + "learning_rate": 1.9532407660393258e-05, + "loss": 1.0562, + "step": 3903 + }, + { + "epoch": 0.12, + "learning_rate": 1.9532094577034174e-05, + "loss": 1.0283, + "step": 3904 + }, + { + "epoch": 0.12, + "learning_rate": 1.9531781391406264e-05, + "loss": 1.0664, + "step": 3905 + }, + { + "epoch": 0.12, + "learning_rate": 1.9531468103512888e-05, + "loss": 1.0308, + "step": 3906 + }, + { + "epoch": 0.12, + "learning_rate": 1.953115471335741e-05, + "loss": 1.1099, + "step": 3907 + }, + { + "epoch": 0.12, + "learning_rate": 1.953084122094319e-05, + "loss": 1.1123, + "step": 3908 + }, + { + "epoch": 0.13, + "learning_rate": 1.9530527626273592e-05, + "loss": 1.0479, + "step": 3909 + }, + { + "epoch": 0.13, + "learning_rate": 1.953021392935198e-05, + "loss": 1.0762, + "step": 3910 + }, + { + "epoch": 0.13, + "learning_rate": 1.952990013018172e-05, + "loss": 1.21, + "step": 3911 + }, + { + "epoch": 0.13, + "learning_rate": 1.9529586228766183e-05, + "loss": 1.1284, + "step": 3912 + }, + { + "epoch": 0.13, + "learning_rate": 1.952927222510873e-05, + "loss": 1.1973, + "step": 3913 + }, + { + "epoch": 0.13, + "learning_rate": 1.952895811921273e-05, + "loss": 1.0967, + "step": 3914 + }, + { + "epoch": 0.13, + "learning_rate": 1.9528643911081556e-05, + "loss": 1.1538, + "step": 3915 + }, + { + "epoch": 0.13, + "learning_rate": 1.952832960071858e-05, + "loss": 1.0435, + "step": 3916 + }, + { + "epoch": 0.13, + "learning_rate": 1.952801518812717e-05, + "loss": 1.0298, + "step": 3917 + }, + { + "epoch": 0.13, + "learning_rate": 1.9527700673310704e-05, + "loss": 1.2363, + "step": 3918 + }, + { + "epoch": 0.13, + "learning_rate": 1.9527386056272556e-05, + "loss": 1.0864, + "step": 3919 + }, + { + "epoch": 0.13, + "learning_rate": 1.95270713370161e-05, + "loss": 1.144, + "step": 3920 + }, + { + "epoch": 0.13, + "learning_rate": 1.952675651554471e-05, + "loss": 1.0303, + "step": 3921 + }, + { + "epoch": 0.13, + "learning_rate": 1.952644159186177e-05, + "loss": 1.0947, + "step": 3922 + }, + { + "epoch": 0.13, + "learning_rate": 1.952612656597065e-05, + "loss": 1.0723, + "step": 3923 + }, + { + "epoch": 0.13, + "learning_rate": 1.952581143787474e-05, + "loss": 1.1343, + "step": 3924 + }, + { + "epoch": 0.13, + "learning_rate": 1.952549620757741e-05, + "loss": 1.1641, + "step": 3925 + }, + { + "epoch": 0.13, + "learning_rate": 1.9525180875082056e-05, + "loss": 1.1421, + "step": 3926 + }, + { + "epoch": 0.13, + "learning_rate": 1.9524865440392048e-05, + "loss": 1.106, + "step": 3927 + }, + { + "epoch": 0.13, + "learning_rate": 1.9524549903510776e-05, + "loss": 1.0903, + "step": 3928 + }, + { + "epoch": 0.13, + "learning_rate": 1.9524234264441624e-05, + "loss": 1.106, + "step": 3929 + }, + { + "epoch": 0.13, + "learning_rate": 1.9523918523187983e-05, + "loss": 1.1724, + "step": 3930 + }, + { + "epoch": 0.13, + "learning_rate": 1.952360267975323e-05, + "loss": 1.1265, + "step": 3931 + }, + { + "epoch": 0.13, + "learning_rate": 1.9523286734140767e-05, + "loss": 0.5505, + "step": 3932 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522970686353976e-05, + "loss": 1.083, + "step": 3933 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522654536396248e-05, + "loss": 0.9683, + "step": 3934 + }, + { + "epoch": 0.13, + "learning_rate": 1.9522338284270974e-05, + "loss": 1.0635, + "step": 3935 + }, + { + "epoch": 0.13, + "learning_rate": 1.952202192998155e-05, + "loss": 1.2065, + "step": 3936 + }, + { + "epoch": 0.13, + "learning_rate": 1.952170547353137e-05, + "loss": 1.1104, + "step": 3937 + }, + { + "epoch": 0.13, + "learning_rate": 1.952138891492383e-05, + "loss": 1.1768, + "step": 3938 + }, + { + "epoch": 0.13, + "learning_rate": 1.9521072254162324e-05, + "loss": 1.189, + "step": 3939 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520755491250247e-05, + "loss": 1.2246, + "step": 3940 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520438626191e-05, + "loss": 1.1206, + "step": 3941 + }, + { + "epoch": 0.13, + "learning_rate": 1.9520121658987988e-05, + "loss": 1.1548, + "step": 3942 + }, + { + "epoch": 0.13, + "learning_rate": 1.9519804589644605e-05, + "loss": 1.249, + "step": 3943 + }, + { + "epoch": 0.13, + "learning_rate": 1.9519487418164255e-05, + "loss": 1.0889, + "step": 3944 + }, + { + "epoch": 0.13, + "learning_rate": 1.9519170144550342e-05, + "loss": 1.1567, + "step": 3945 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518852768806265e-05, + "loss": 1.2212, + "step": 3946 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518535290935436e-05, + "loss": 1.1118, + "step": 3947 + }, + { + "epoch": 0.13, + "learning_rate": 1.9518217710941257e-05, + "loss": 1.1084, + "step": 3948 + }, + { + "epoch": 0.13, + "learning_rate": 1.9517900028827135e-05, + "loss": 1.1572, + "step": 3949 + }, + { + "epoch": 0.13, + "learning_rate": 1.951758224459648e-05, + "loss": 1.1479, + "step": 3950 + }, + { + "epoch": 0.13, + "learning_rate": 1.9517264358252702e-05, + "loss": 1.0034, + "step": 3951 + }, + { + "epoch": 0.13, + "learning_rate": 1.951694636979921e-05, + "loss": 1.0557, + "step": 3952 + }, + { + "epoch": 0.13, + "learning_rate": 1.9516628279239416e-05, + "loss": 1.0942, + "step": 3953 + }, + { + "epoch": 0.13, + "learning_rate": 1.9516310086576734e-05, + "loss": 1.1108, + "step": 3954 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515991791814578e-05, + "loss": 1.0874, + "step": 3955 + }, + { + "epoch": 0.13, + "learning_rate": 1.951567339495636e-05, + "loss": 1.1665, + "step": 3956 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515354896005496e-05, + "loss": 0.501, + "step": 3957 + }, + { + "epoch": 0.13, + "learning_rate": 1.9515036294965408e-05, + "loss": 1.1958, + "step": 3958 + }, + { + "epoch": 0.13, + "learning_rate": 1.951471759183951e-05, + "loss": 1.0215, + "step": 3959 + }, + { + "epoch": 0.13, + "learning_rate": 1.9514398786631223e-05, + "loss": 1.1455, + "step": 3960 + }, + { + "epoch": 0.13, + "learning_rate": 1.9514079879343967e-05, + "loss": 0.9805, + "step": 3961 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513760869981164e-05, + "loss": 1.1123, + "step": 3962 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513441758546235e-05, + "loss": 1.1094, + "step": 3963 + }, + { + "epoch": 0.13, + "learning_rate": 1.9513122545042606e-05, + "loss": 1.0532, + "step": 3964 + }, + { + "epoch": 0.13, + "learning_rate": 1.95128032294737e-05, + "loss": 1.1626, + "step": 3965 + }, + { + "epoch": 0.13, + "learning_rate": 1.9512483811842943e-05, + "loss": 1.0864, + "step": 3966 + }, + { + "epoch": 0.13, + "learning_rate": 1.9512164292153763e-05, + "loss": 1.166, + "step": 3967 + }, + { + "epoch": 0.13, + "learning_rate": 1.951184467040959e-05, + "loss": 1.1313, + "step": 3968 + }, + { + "epoch": 0.13, + "learning_rate": 1.951152494661385e-05, + "loss": 1.0454, + "step": 3969 + }, + { + "epoch": 0.13, + "learning_rate": 1.9511205120769972e-05, + "loss": 1.0391, + "step": 3970 + }, + { + "epoch": 0.13, + "learning_rate": 1.951088519288139e-05, + "loss": 1.168, + "step": 3971 + }, + { + "epoch": 0.13, + "learning_rate": 1.9510565162951538e-05, + "loss": 1.0327, + "step": 3972 + }, + { + "epoch": 0.13, + "learning_rate": 1.9510245030983847e-05, + "loss": 0.979, + "step": 3973 + }, + { + "epoch": 0.13, + "learning_rate": 1.950992479698175e-05, + "loss": 1.0605, + "step": 3974 + }, + { + "epoch": 0.13, + "learning_rate": 1.9509604460948687e-05, + "loss": 0.9292, + "step": 3975 + }, + { + "epoch": 0.13, + "learning_rate": 1.9509284022888093e-05, + "loss": 1.0601, + "step": 3976 + }, + { + "epoch": 0.13, + "learning_rate": 1.9508963482803404e-05, + "loss": 1.02, + "step": 3977 + }, + { + "epoch": 0.13, + "learning_rate": 1.9508642840698062e-05, + "loss": 1.2197, + "step": 3978 + }, + { + "epoch": 0.13, + "learning_rate": 1.9508322096575506e-05, + "loss": 1.0669, + "step": 3979 + }, + { + "epoch": 0.13, + "learning_rate": 1.950800125043918e-05, + "loss": 1.1587, + "step": 3980 + }, + { + "epoch": 0.13, + "learning_rate": 1.9507680302292518e-05, + "loss": 1.0322, + "step": 3981 + }, + { + "epoch": 0.13, + "learning_rate": 1.950735925213897e-05, + "loss": 1.1484, + "step": 3982 + }, + { + "epoch": 0.13, + "learning_rate": 1.9507038099981982e-05, + "loss": 1.1299, + "step": 3983 + }, + { + "epoch": 0.13, + "learning_rate": 1.9506716845824998e-05, + "loss": 1.1621, + "step": 3984 + }, + { + "epoch": 0.13, + "learning_rate": 1.950639548967146e-05, + "loss": 1.1396, + "step": 3985 + }, + { + "epoch": 0.13, + "learning_rate": 1.9506074031524823e-05, + "loss": 1.1738, + "step": 3986 + }, + { + "epoch": 0.13, + "learning_rate": 1.950575247138853e-05, + "loss": 1.0898, + "step": 3987 + }, + { + "epoch": 0.13, + "learning_rate": 1.9505430809266035e-05, + "loss": 1.0532, + "step": 3988 + }, + { + "epoch": 0.13, + "learning_rate": 1.9505109045160787e-05, + "loss": 1.0815, + "step": 3989 + }, + { + "epoch": 0.13, + "learning_rate": 1.9504787179076237e-05, + "loss": 1.0151, + "step": 3990 + }, + { + "epoch": 0.13, + "learning_rate": 1.9504465211015843e-05, + "loss": 1.0356, + "step": 3991 + }, + { + "epoch": 0.13, + "learning_rate": 1.9504143140983057e-05, + "loss": 0.4717, + "step": 3992 + }, + { + "epoch": 0.13, + "learning_rate": 1.950382096898133e-05, + "loss": 1.2012, + "step": 3993 + }, + { + "epoch": 0.13, + "learning_rate": 1.9503498695014123e-05, + "loss": 0.9844, + "step": 3994 + }, + { + "epoch": 0.13, + "learning_rate": 1.9503176319084895e-05, + "loss": 1.0645, + "step": 3995 + }, + { + "epoch": 0.13, + "learning_rate": 1.95028538411971e-05, + "loss": 1.1602, + "step": 3996 + }, + { + "epoch": 0.13, + "learning_rate": 1.9502531261354204e-05, + "loss": 0.9214, + "step": 3997 + }, + { + "epoch": 0.13, + "learning_rate": 1.950220857955966e-05, + "loss": 1.0454, + "step": 3998 + }, + { + "epoch": 0.13, + "learning_rate": 1.9501885795816937e-05, + "loss": 1.2451, + "step": 3999 + }, + { + "epoch": 0.13, + "learning_rate": 1.9501562910129498e-05, + "loss": 1.1758, + "step": 4000 + }, + { + "epoch": 0.13, + "learning_rate": 1.95012399225008e-05, + "loss": 1.0923, + "step": 4001 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500916832934316e-05, + "loss": 1.0332, + "step": 4002 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500593641433506e-05, + "loss": 1.0786, + "step": 4003 + }, + { + "epoch": 0.13, + "learning_rate": 1.9500270348001845e-05, + "loss": 1.0708, + "step": 4004 + }, + { + "epoch": 0.13, + "learning_rate": 1.94999469526428e-05, + "loss": 1.1636, + "step": 4005 + }, + { + "epoch": 0.13, + "learning_rate": 1.9499623455359832e-05, + "loss": 1.0728, + "step": 4006 + }, + { + "epoch": 0.13, + "learning_rate": 1.9499299856156423e-05, + "loss": 1.0566, + "step": 4007 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498976155036037e-05, + "loss": 1.2295, + "step": 4008 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498652352002148e-05, + "loss": 1.1519, + "step": 4009 + }, + { + "epoch": 0.13, + "learning_rate": 1.9498328447058236e-05, + "loss": 1.0459, + "step": 4010 + }, + { + "epoch": 0.13, + "learning_rate": 1.949800444020777e-05, + "loss": 1.0703, + "step": 4011 + }, + { + "epoch": 0.13, + "learning_rate": 1.9497680331454228e-05, + "loss": 1.0117, + "step": 4012 + }, + { + "epoch": 0.13, + "learning_rate": 1.9497356120801085e-05, + "loss": 1.0361, + "step": 4013 + }, + { + "epoch": 0.13, + "learning_rate": 1.9497031808251826e-05, + "loss": 1.1855, + "step": 4014 + }, + { + "epoch": 0.13, + "learning_rate": 1.9496707393809925e-05, + "loss": 1.1255, + "step": 4015 + }, + { + "epoch": 0.13, + "learning_rate": 1.9496382877478865e-05, + "loss": 1.0742, + "step": 4016 + }, + { + "epoch": 0.13, + "learning_rate": 1.9496058259262124e-05, + "loss": 1.0889, + "step": 4017 + }, + { + "epoch": 0.13, + "learning_rate": 1.949573353916319e-05, + "loss": 1.1318, + "step": 4018 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495408717185546e-05, + "loss": 1.1362, + "step": 4019 + }, + { + "epoch": 0.13, + "learning_rate": 1.9495083793332674e-05, + "loss": 1.1128, + "step": 4020 + }, + { + "epoch": 0.13, + "learning_rate": 1.949475876760806e-05, + "loss": 1.0073, + "step": 4021 + }, + { + "epoch": 0.13, + "learning_rate": 1.9494433640015196e-05, + "loss": 1.1396, + "step": 4022 + }, + { + "epoch": 0.13, + "learning_rate": 1.9494108410557563e-05, + "loss": 1.2671, + "step": 4023 + }, + { + "epoch": 0.13, + "learning_rate": 1.9493783079238658e-05, + "loss": 1.1802, + "step": 4024 + }, + { + "epoch": 0.13, + "learning_rate": 1.9493457646061968e-05, + "loss": 1.0962, + "step": 4025 + }, + { + "epoch": 0.13, + "learning_rate": 1.949313211103098e-05, + "loss": 1.0337, + "step": 4026 + }, + { + "epoch": 0.13, + "learning_rate": 1.9492806474149193e-05, + "loss": 1.0283, + "step": 4027 + }, + { + "epoch": 0.13, + "learning_rate": 1.94924807354201e-05, + "loss": 1.1401, + "step": 4028 + }, + { + "epoch": 0.13, + "learning_rate": 1.9492154894847192e-05, + "loss": 1.188, + "step": 4029 + }, + { + "epoch": 0.13, + "learning_rate": 1.9491828952433968e-05, + "loss": 1.0806, + "step": 4030 + }, + { + "epoch": 0.13, + "learning_rate": 1.9491502908183925e-05, + "loss": 1.2275, + "step": 4031 + }, + { + "epoch": 0.13, + "learning_rate": 1.9491176762100562e-05, + "loss": 1.0903, + "step": 4032 + }, + { + "epoch": 0.13, + "learning_rate": 1.9490850514187376e-05, + "loss": 1.186, + "step": 4033 + }, + { + "epoch": 0.13, + "learning_rate": 1.9490524164447867e-05, + "loss": 1.2422, + "step": 4034 + }, + { + "epoch": 0.13, + "learning_rate": 1.9490197712885538e-05, + "loss": 0.4485, + "step": 4035 + }, + { + "epoch": 0.13, + "learning_rate": 1.948987115950389e-05, + "loss": 1.123, + "step": 4036 + }, + { + "epoch": 0.13, + "learning_rate": 1.948954450430643e-05, + "loss": 1.1846, + "step": 4037 + }, + { + "epoch": 0.13, + "learning_rate": 1.9489217747296654e-05, + "loss": 1.1001, + "step": 4038 + }, + { + "epoch": 0.13, + "learning_rate": 1.9488890888478078e-05, + "loss": 0.9707, + "step": 4039 + }, + { + "epoch": 0.13, + "learning_rate": 1.9488563927854202e-05, + "loss": 1.1201, + "step": 4040 + }, + { + "epoch": 0.13, + "learning_rate": 1.9488236865428537e-05, + "loss": 1.0732, + "step": 4041 + }, + { + "epoch": 0.13, + "learning_rate": 1.9487909701204595e-05, + "loss": 1.1045, + "step": 4042 + }, + { + "epoch": 0.13, + "learning_rate": 1.948758243518588e-05, + "loss": 1.1001, + "step": 4043 + }, + { + "epoch": 0.13, + "learning_rate": 1.9487255067375907e-05, + "loss": 1.1421, + "step": 4044 + }, + { + "epoch": 0.13, + "learning_rate": 1.9486927597778183e-05, + "loss": 1.1924, + "step": 4045 + }, + { + "epoch": 0.13, + "learning_rate": 1.948660002639623e-05, + "loss": 1.146, + "step": 4046 + }, + { + "epoch": 0.13, + "learning_rate": 1.9486272353233556e-05, + "loss": 1.1353, + "step": 4047 + }, + { + "epoch": 0.13, + "learning_rate": 1.9485944578293682e-05, + "loss": 1.1333, + "step": 4048 + }, + { + "epoch": 0.13, + "learning_rate": 1.948561670158012e-05, + "loss": 0.9922, + "step": 4049 + }, + { + "epoch": 0.13, + "learning_rate": 1.9485288723096386e-05, + "loss": 1.0923, + "step": 4050 + }, + { + "epoch": 0.13, + "learning_rate": 1.9484960642846004e-05, + "loss": 1.0464, + "step": 4051 + }, + { + "epoch": 0.13, + "learning_rate": 1.948463246083249e-05, + "loss": 1.0718, + "step": 4052 + }, + { + "epoch": 0.13, + "learning_rate": 1.9484304177059366e-05, + "loss": 1.0298, + "step": 4053 + }, + { + "epoch": 0.13, + "learning_rate": 1.9483975791530158e-05, + "loss": 0.9941, + "step": 4054 + }, + { + "epoch": 0.13, + "learning_rate": 1.9483647304248386e-05, + "loss": 1.0566, + "step": 4055 + }, + { + "epoch": 0.13, + "learning_rate": 1.948331871521757e-05, + "loss": 1.0835, + "step": 4056 + }, + { + "epoch": 0.13, + "learning_rate": 1.9482990024441246e-05, + "loss": 1.0547, + "step": 4057 + }, + { + "epoch": 0.13, + "learning_rate": 1.948266123192293e-05, + "loss": 1.1118, + "step": 4058 + }, + { + "epoch": 0.13, + "learning_rate": 1.9482332337666156e-05, + "loss": 0.9624, + "step": 4059 + }, + { + "epoch": 0.13, + "learning_rate": 1.948200334167445e-05, + "loss": 1.0947, + "step": 4060 + }, + { + "epoch": 0.13, + "learning_rate": 1.948167424395134e-05, + "loss": 1.0791, + "step": 4061 + }, + { + "epoch": 0.13, + "learning_rate": 1.9481345044500362e-05, + "loss": 1.1763, + "step": 4062 + }, + { + "epoch": 0.13, + "learning_rate": 1.9481015743325045e-05, + "loss": 1.0884, + "step": 4063 + }, + { + "epoch": 0.13, + "learning_rate": 1.948068634042892e-05, + "loss": 1.146, + "step": 4064 + }, + { + "epoch": 0.13, + "learning_rate": 1.9480356835815524e-05, + "loss": 0.9395, + "step": 4065 + }, + { + "epoch": 0.13, + "learning_rate": 1.9480027229488395e-05, + "loss": 1.1562, + "step": 4066 + }, + { + "epoch": 0.13, + "learning_rate": 1.947969752145106e-05, + "loss": 1.2339, + "step": 4067 + }, + { + "epoch": 0.13, + "learning_rate": 1.9479367711707065e-05, + "loss": 1.1655, + "step": 4068 + }, + { + "epoch": 0.13, + "learning_rate": 1.947903780025995e-05, + "loss": 1.1094, + "step": 4069 + }, + { + "epoch": 0.13, + "learning_rate": 1.9478707787113243e-05, + "loss": 1.0703, + "step": 4070 + }, + { + "epoch": 0.13, + "learning_rate": 1.9478377672270494e-05, + "loss": 1.2168, + "step": 4071 + }, + { + "epoch": 0.13, + "learning_rate": 1.9478047455735246e-05, + "loss": 1.1006, + "step": 4072 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477717137511037e-05, + "loss": 1.082, + "step": 4073 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477386717601413e-05, + "loss": 0.5322, + "step": 4074 + }, + { + "epoch": 0.13, + "learning_rate": 1.9477056196009916e-05, + "loss": 1.1577, + "step": 4075 + }, + { + "epoch": 0.13, + "learning_rate": 1.9476725572740095e-05, + "loss": 1.0879, + "step": 4076 + }, + { + "epoch": 0.13, + "learning_rate": 1.94763948477955e-05, + "loss": 1.1011, + "step": 4077 + }, + { + "epoch": 0.13, + "learning_rate": 1.9476064021179672e-05, + "loss": 1.1255, + "step": 4078 + }, + { + "epoch": 0.13, + "learning_rate": 1.9475733092896168e-05, + "loss": 1.1001, + "step": 4079 + }, + { + "epoch": 0.13, + "learning_rate": 1.9475402062948533e-05, + "loss": 0.9619, + "step": 4080 + }, + { + "epoch": 0.13, + "learning_rate": 1.9475070931340322e-05, + "loss": 1.1255, + "step": 4081 + }, + { + "epoch": 0.13, + "learning_rate": 1.9474739698075087e-05, + "loss": 1.0439, + "step": 4082 + }, + { + "epoch": 0.13, + "learning_rate": 1.947440836315638e-05, + "loss": 1.1514, + "step": 4083 + }, + { + "epoch": 0.13, + "learning_rate": 1.947407692658776e-05, + "loss": 1.0918, + "step": 4084 + }, + { + "epoch": 0.13, + "learning_rate": 1.9473745388372774e-05, + "loss": 1.04, + "step": 4085 + }, + { + "epoch": 0.13, + "learning_rate": 1.947341374851499e-05, + "loss": 1.0083, + "step": 4086 + }, + { + "epoch": 0.13, + "learning_rate": 1.947308200701796e-05, + "loss": 1.0938, + "step": 4087 + }, + { + "epoch": 0.13, + "learning_rate": 1.9472750163885243e-05, + "loss": 1.0869, + "step": 4088 + }, + { + "epoch": 0.13, + "learning_rate": 1.9472418219120403e-05, + "loss": 1.0532, + "step": 4089 + }, + { + "epoch": 0.13, + "learning_rate": 1.9472086172726997e-05, + "loss": 1.167, + "step": 4090 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471754024708588e-05, + "loss": 1.1709, + "step": 4091 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471421775068746e-05, + "loss": 1.1172, + "step": 4092 + }, + { + "epoch": 0.13, + "learning_rate": 1.9471089423811028e-05, + "loss": 1.1138, + "step": 4093 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470756970939e-05, + "loss": 1.0488, + "step": 4094 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470424416456235e-05, + "loss": 1.0771, + "step": 4095 + }, + { + "epoch": 0.13, + "learning_rate": 1.9470091760366295e-05, + "loss": 1.1245, + "step": 4096 + }, + { + "epoch": 0.13, + "learning_rate": 1.9469759002672754e-05, + "loss": 1.0762, + "step": 4097 + }, + { + "epoch": 0.13, + "learning_rate": 1.946942614337918e-05, + "loss": 1.147, + "step": 4098 + }, + { + "epoch": 0.13, + "learning_rate": 1.9469093182489143e-05, + "loss": 0.9795, + "step": 4099 + }, + { + "epoch": 0.13, + "learning_rate": 1.946876012000621e-05, + "loss": 1.1602, + "step": 4100 + }, + { + "epoch": 0.13, + "learning_rate": 1.946842695593397e-05, + "loss": 1.0518, + "step": 4101 + }, + { + "epoch": 0.13, + "learning_rate": 1.9468093690275983e-05, + "loss": 1.1108, + "step": 4102 + }, + { + "epoch": 0.13, + "learning_rate": 1.946776032303583e-05, + "loss": 1.228, + "step": 4103 + }, + { + "epoch": 0.13, + "learning_rate": 1.9467426854217085e-05, + "loss": 1.1885, + "step": 4104 + }, + { + "epoch": 0.13, + "learning_rate": 1.9467093283823332e-05, + "loss": 1.064, + "step": 4105 + }, + { + "epoch": 0.13, + "learning_rate": 1.946675961185814e-05, + "loss": 1.104, + "step": 4106 + }, + { + "epoch": 0.13, + "learning_rate": 1.9466425838325102e-05, + "loss": 1.0186, + "step": 4107 + }, + { + "epoch": 0.13, + "learning_rate": 1.9466091963227787e-05, + "loss": 1.1934, + "step": 4108 + }, + { + "epoch": 0.13, + "learning_rate": 1.9465757986569782e-05, + "loss": 1.1025, + "step": 4109 + }, + { + "epoch": 0.13, + "learning_rate": 1.946542390835467e-05, + "loss": 0.9897, + "step": 4110 + }, + { + "epoch": 0.13, + "learning_rate": 1.9465089728586036e-05, + "loss": 1.1655, + "step": 4111 + }, + { + "epoch": 0.13, + "learning_rate": 1.9464755447267464e-05, + "loss": 1.061, + "step": 4112 + }, + { + "epoch": 0.13, + "learning_rate": 1.9464421064402544e-05, + "loss": 1.2178, + "step": 4113 + }, + { + "epoch": 0.13, + "learning_rate": 1.9464086579994857e-05, + "loss": 1.0854, + "step": 4114 + }, + { + "epoch": 0.13, + "learning_rate": 1.9463751994047998e-05, + "loss": 1.1147, + "step": 4115 + }, + { + "epoch": 0.13, + "learning_rate": 1.9463417306565554e-05, + "loss": 1.1807, + "step": 4116 + }, + { + "epoch": 0.13, + "learning_rate": 1.9463082517551114e-05, + "loss": 1.1655, + "step": 4117 + }, + { + "epoch": 0.13, + "learning_rate": 1.9462747627008273e-05, + "loss": 1.0728, + "step": 4118 + }, + { + "epoch": 0.13, + "learning_rate": 1.946241263494062e-05, + "loss": 1.0645, + "step": 4119 + }, + { + "epoch": 0.13, + "learning_rate": 1.9462077541351755e-05, + "loss": 1.0342, + "step": 4120 + }, + { + "epoch": 0.13, + "learning_rate": 1.9461742346245268e-05, + "loss": 1.0303, + "step": 4121 + }, + { + "epoch": 0.13, + "learning_rate": 1.946140704962476e-05, + "loss": 1.1396, + "step": 4122 + }, + { + "epoch": 0.13, + "learning_rate": 1.9461071651493822e-05, + "loss": 1.123, + "step": 4123 + }, + { + "epoch": 0.13, + "learning_rate": 1.946073615185606e-05, + "loss": 1.1934, + "step": 4124 + }, + { + "epoch": 0.13, + "learning_rate": 1.946040055071507e-05, + "loss": 0.9619, + "step": 4125 + }, + { + "epoch": 0.13, + "learning_rate": 1.946006484807445e-05, + "loss": 1.0586, + "step": 4126 + }, + { + "epoch": 0.13, + "learning_rate": 1.9459729043937807e-05, + "loss": 1.0884, + "step": 4127 + }, + { + "epoch": 0.13, + "learning_rate": 1.9459393138308737e-05, + "loss": 1.0923, + "step": 4128 + }, + { + "epoch": 0.13, + "learning_rate": 1.945905713119085e-05, + "loss": 1.0264, + "step": 4129 + }, + { + "epoch": 0.13, + "learning_rate": 1.9458721022587744e-05, + "loss": 1.1934, + "step": 4130 + }, + { + "epoch": 0.13, + "learning_rate": 1.945838481250304e-05, + "loss": 1.0869, + "step": 4131 + }, + { + "epoch": 0.13, + "learning_rate": 1.9458048500940327e-05, + "loss": 1.103, + "step": 4132 + }, + { + "epoch": 0.13, + "learning_rate": 1.9457712087903224e-05, + "loss": 1.145, + "step": 4133 + }, + { + "epoch": 0.13, + "learning_rate": 1.9457375573395334e-05, + "loss": 1.1816, + "step": 4134 + }, + { + "epoch": 0.13, + "learning_rate": 1.9457038957420274e-05, + "loss": 1.0874, + "step": 4135 + }, + { + "epoch": 0.13, + "learning_rate": 1.9456702239981654e-05, + "loss": 1.1357, + "step": 4136 + }, + { + "epoch": 0.13, + "learning_rate": 1.945636542108308e-05, + "loss": 1.0972, + "step": 4137 + }, + { + "epoch": 0.13, + "learning_rate": 1.9456028500728177e-05, + "loss": 1.2256, + "step": 4138 + }, + { + "epoch": 0.13, + "learning_rate": 1.945569147892055e-05, + "loss": 1.0801, + "step": 4139 + }, + { + "epoch": 0.13, + "learning_rate": 1.945535435566382e-05, + "loss": 1.1494, + "step": 4140 + }, + { + "epoch": 0.13, + "learning_rate": 1.9455017130961603e-05, + "loss": 0.998, + "step": 4141 + }, + { + "epoch": 0.13, + "learning_rate": 1.9454679804817514e-05, + "loss": 1.062, + "step": 4142 + }, + { + "epoch": 0.13, + "learning_rate": 1.9454342377235175e-05, + "loss": 1.082, + "step": 4143 + }, + { + "epoch": 0.13, + "learning_rate": 1.9454004848218205e-05, + "loss": 1.2896, + "step": 4144 + }, + { + "epoch": 0.13, + "learning_rate": 1.945366721777023e-05, + "loss": 1.1001, + "step": 4145 + }, + { + "epoch": 0.13, + "learning_rate": 1.9453329485894865e-05, + "loss": 1.0625, + "step": 4146 + }, + { + "epoch": 0.13, + "learning_rate": 1.945299165259574e-05, + "loss": 1.1113, + "step": 4147 + }, + { + "epoch": 0.13, + "learning_rate": 1.9452653717876475e-05, + "loss": 1.0488, + "step": 4148 + }, + { + "epoch": 0.13, + "learning_rate": 1.9452315681740693e-05, + "loss": 0.9761, + "step": 4149 + }, + { + "epoch": 0.13, + "learning_rate": 1.945197754419203e-05, + "loss": 1.2793, + "step": 4150 + }, + { + "epoch": 0.13, + "learning_rate": 1.9451639305234108e-05, + "loss": 1.0928, + "step": 4151 + }, + { + "epoch": 0.13, + "learning_rate": 1.945130096487056e-05, + "loss": 1.0474, + "step": 4152 + }, + { + "epoch": 0.13, + "learning_rate": 1.945096252310501e-05, + "loss": 1.1055, + "step": 4153 + }, + { + "epoch": 0.13, + "learning_rate": 1.945062397994109e-05, + "loss": 1.1914, + "step": 4154 + }, + { + "epoch": 0.13, + "learning_rate": 1.9450285335382436e-05, + "loss": 1.2451, + "step": 4155 + }, + { + "epoch": 0.13, + "learning_rate": 1.944994658943268e-05, + "loss": 1.0723, + "step": 4156 + }, + { + "epoch": 0.13, + "learning_rate": 1.9449607742095455e-05, + "loss": 1.0532, + "step": 4157 + }, + { + "epoch": 0.13, + "learning_rate": 1.9449268793374395e-05, + "loss": 1.0342, + "step": 4158 + }, + { + "epoch": 0.13, + "learning_rate": 1.9448929743273145e-05, + "loss": 1.1074, + "step": 4159 + }, + { + "epoch": 0.13, + "learning_rate": 1.944859059179533e-05, + "loss": 1.0942, + "step": 4160 + }, + { + "epoch": 0.13, + "learning_rate": 1.94482513389446e-05, + "loss": 1.0444, + "step": 4161 + }, + { + "epoch": 0.13, + "learning_rate": 1.9447911984724584e-05, + "loss": 1.0356, + "step": 4162 + }, + { + "epoch": 0.13, + "learning_rate": 1.9447572529138938e-05, + "loss": 1.0127, + "step": 4163 + }, + { + "epoch": 0.13, + "learning_rate": 1.9447232972191288e-05, + "loss": 1.1978, + "step": 4164 + }, + { + "epoch": 0.13, + "learning_rate": 1.9446893313885283e-05, + "loss": 1.1978, + "step": 4165 + }, + { + "epoch": 0.13, + "learning_rate": 1.944655355422457e-05, + "loss": 1.0908, + "step": 4166 + }, + { + "epoch": 0.13, + "learning_rate": 1.9446213693212793e-05, + "loss": 1.0625, + "step": 4167 + }, + { + "epoch": 0.13, + "learning_rate": 1.9445873730853596e-05, + "loss": 1.1138, + "step": 4168 + }, + { + "epoch": 0.13, + "learning_rate": 1.944553366715063e-05, + "loss": 1.0713, + "step": 4169 + }, + { + "epoch": 0.13, + "learning_rate": 1.944519350210754e-05, + "loss": 1.292, + "step": 4170 + }, + { + "epoch": 0.13, + "learning_rate": 1.9444853235727977e-05, + "loss": 1.1504, + "step": 4171 + }, + { + "epoch": 0.13, + "learning_rate": 1.9444512868015596e-05, + "loss": 1.1499, + "step": 4172 + }, + { + "epoch": 0.13, + "learning_rate": 1.944417239897404e-05, + "loss": 1.2017, + "step": 4173 + }, + { + "epoch": 0.13, + "learning_rate": 1.944383182860697e-05, + "loss": 1.1699, + "step": 4174 + }, + { + "epoch": 0.13, + "learning_rate": 1.9443491156918034e-05, + "loss": 1.2524, + "step": 4175 + }, + { + "epoch": 0.13, + "learning_rate": 1.9443150383910887e-05, + "loss": 1.1709, + "step": 4176 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442809509589188e-05, + "loss": 1.0908, + "step": 4177 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442468533956598e-05, + "loss": 1.1089, + "step": 4178 + }, + { + "epoch": 0.13, + "learning_rate": 1.9442127457016768e-05, + "loss": 1.2148, + "step": 4179 + }, + { + "epoch": 0.13, + "learning_rate": 1.9441786278773362e-05, + "loss": 1.0806, + "step": 4180 + }, + { + "epoch": 0.13, + "learning_rate": 1.9441444999230037e-05, + "loss": 1.0317, + "step": 4181 + }, + { + "epoch": 0.13, + "learning_rate": 1.9441103618390458e-05, + "loss": 1.1016, + "step": 4182 + }, + { + "epoch": 0.13, + "learning_rate": 1.9440762136258285e-05, + "loss": 1.1396, + "step": 4183 + }, + { + "epoch": 0.13, + "learning_rate": 1.9440420552837184e-05, + "loss": 1.1133, + "step": 4184 + }, + { + "epoch": 0.13, + "learning_rate": 1.9440078868130817e-05, + "loss": 1.1802, + "step": 4185 + }, + { + "epoch": 0.13, + "learning_rate": 1.9439737082142854e-05, + "loss": 1.0439, + "step": 4186 + }, + { + "epoch": 0.13, + "learning_rate": 1.9439395194876957e-05, + "loss": 1.0923, + "step": 4187 + }, + { + "epoch": 0.13, + "learning_rate": 1.94390532063368e-05, + "loss": 1.1323, + "step": 4188 + }, + { + "epoch": 0.13, + "learning_rate": 1.9438711116526042e-05, + "loss": 1.1274, + "step": 4189 + }, + { + "epoch": 0.13, + "learning_rate": 1.9438368925448367e-05, + "loss": 1.0254, + "step": 4190 + }, + { + "epoch": 0.13, + "learning_rate": 1.9438026633107435e-05, + "loss": 1.1094, + "step": 4191 + }, + { + "epoch": 0.13, + "learning_rate": 1.9437684239506926e-05, + "loss": 1.0371, + "step": 4192 + }, + { + "epoch": 0.13, + "learning_rate": 1.943734174465051e-05, + "loss": 1.0894, + "step": 4193 + }, + { + "epoch": 0.13, + "learning_rate": 1.9436999148541858e-05, + "loss": 1.0786, + "step": 4194 + }, + { + "epoch": 0.13, + "learning_rate": 1.9436656451184652e-05, + "loss": 1.1479, + "step": 4195 + }, + { + "epoch": 0.13, + "learning_rate": 1.9436313652582566e-05, + "loss": 1.1885, + "step": 4196 + }, + { + "epoch": 0.13, + "learning_rate": 1.943597075273928e-05, + "loss": 1.1089, + "step": 4197 + }, + { + "epoch": 0.13, + "learning_rate": 1.943562775165847e-05, + "loss": 1.1006, + "step": 4198 + }, + { + "epoch": 0.13, + "learning_rate": 1.9435284649343816e-05, + "loss": 1.0054, + "step": 4199 + }, + { + "epoch": 0.13, + "learning_rate": 1.9434941445799e-05, + "loss": 1.1611, + "step": 4200 + }, + { + "epoch": 0.13, + "learning_rate": 1.9434598141027706e-05, + "loss": 1.1616, + "step": 4201 + }, + { + "epoch": 0.13, + "learning_rate": 1.943425473503361e-05, + "loss": 1.1094, + "step": 4202 + }, + { + "epoch": 0.13, + "learning_rate": 1.9433911227820408e-05, + "loss": 1.084, + "step": 4203 + }, + { + "epoch": 0.13, + "learning_rate": 1.9433567619391778e-05, + "loss": 1.0488, + "step": 4204 + }, + { + "epoch": 0.13, + "learning_rate": 1.943322390975141e-05, + "loss": 1.2036, + "step": 4205 + }, + { + "epoch": 0.13, + "learning_rate": 1.9432880098902986e-05, + "loss": 0.999, + "step": 4206 + }, + { + "epoch": 0.13, + "learning_rate": 1.9432536186850202e-05, + "loss": 1.0884, + "step": 4207 + }, + { + "epoch": 0.13, + "learning_rate": 1.943219217359674e-05, + "loss": 1.1206, + "step": 4208 + }, + { + "epoch": 0.13, + "learning_rate": 1.94318480591463e-05, + "loss": 1.1641, + "step": 4209 + }, + { + "epoch": 0.13, + "learning_rate": 1.9431503843502565e-05, + "loss": 1.1953, + "step": 4210 + }, + { + "epoch": 0.13, + "learning_rate": 1.943115952666923e-05, + "loss": 1.1465, + "step": 4211 + }, + { + "epoch": 0.13, + "learning_rate": 1.9430815108649995e-05, + "loss": 1.0908, + "step": 4212 + }, + { + "epoch": 0.13, + "learning_rate": 1.9430470589448552e-05, + "loss": 1.1162, + "step": 4213 + }, + { + "epoch": 0.13, + "learning_rate": 1.9430125969068593e-05, + "loss": 1.0801, + "step": 4214 + }, + { + "epoch": 0.13, + "learning_rate": 1.9429781247513825e-05, + "loss": 1.0586, + "step": 4215 + }, + { + "epoch": 0.13, + "learning_rate": 1.9429436424787934e-05, + "loss": 1.0693, + "step": 4216 + }, + { + "epoch": 0.13, + "learning_rate": 1.942909150089463e-05, + "loss": 1.0146, + "step": 4217 + }, + { + "epoch": 0.13, + "learning_rate": 1.942874647583761e-05, + "loss": 1.0874, + "step": 4218 + }, + { + "epoch": 0.13, + "learning_rate": 1.9428401349620575e-05, + "loss": 1.1318, + "step": 4219 + }, + { + "epoch": 0.13, + "learning_rate": 1.9428056122247225e-05, + "loss": 0.915, + "step": 4220 + }, + { + "epoch": 0.13, + "learning_rate": 1.942771079372127e-05, + "loss": 1.1885, + "step": 4221 + }, + { + "epoch": 0.14, + "learning_rate": 1.9427365364046413e-05, + "loss": 1.1406, + "step": 4222 + }, + { + "epoch": 0.14, + "learning_rate": 1.942701983322636e-05, + "loss": 1.0654, + "step": 4223 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426674201264814e-05, + "loss": 1.1069, + "step": 4224 + }, + { + "epoch": 0.14, + "learning_rate": 1.9426328468165492e-05, + "loss": 1.0366, + "step": 4225 + }, + { + "epoch": 0.14, + "learning_rate": 1.9425982633932093e-05, + "loss": 0.9585, + "step": 4226 + }, + { + "epoch": 0.14, + "learning_rate": 1.9425636698568337e-05, + "loss": 1.1499, + "step": 4227 + }, + { + "epoch": 0.14, + "learning_rate": 1.9425290662077927e-05, + "loss": 1.1973, + "step": 4228 + }, + { + "epoch": 0.14, + "learning_rate": 1.942494452446458e-05, + "loss": 1.1885, + "step": 4229 + }, + { + "epoch": 0.14, + "learning_rate": 1.942459828573201e-05, + "loss": 1.0664, + "step": 4230 + }, + { + "epoch": 0.14, + "learning_rate": 1.942425194588393e-05, + "loss": 1.209, + "step": 4231 + }, + { + "epoch": 0.14, + "learning_rate": 1.942390550492406e-05, + "loss": 1.1133, + "step": 4232 + }, + { + "epoch": 0.14, + "learning_rate": 1.9423558962856114e-05, + "loss": 1.3408, + "step": 4233 + }, + { + "epoch": 0.14, + "learning_rate": 1.9423212319683807e-05, + "loss": 1.063, + "step": 4234 + }, + { + "epoch": 0.14, + "learning_rate": 1.942286557541086e-05, + "loss": 1.1426, + "step": 4235 + }, + { + "epoch": 0.14, + "learning_rate": 1.9422518730040995e-05, + "loss": 1.2041, + "step": 4236 + }, + { + "epoch": 0.14, + "learning_rate": 1.9422171783577934e-05, + "loss": 1.0493, + "step": 4237 + }, + { + "epoch": 0.14, + "learning_rate": 1.94218247360254e-05, + "loss": 1.1069, + "step": 4238 + }, + { + "epoch": 0.14, + "learning_rate": 1.942147758738711e-05, + "loss": 1.0396, + "step": 4239 + }, + { + "epoch": 0.14, + "learning_rate": 1.9421130337666792e-05, + "loss": 1.105, + "step": 4240 + }, + { + "epoch": 0.14, + "learning_rate": 1.9420782986868173e-05, + "loss": 1.0503, + "step": 4241 + }, + { + "epoch": 0.14, + "learning_rate": 1.942043553499498e-05, + "loss": 0.998, + "step": 4242 + }, + { + "epoch": 0.14, + "learning_rate": 1.9420087982050943e-05, + "loss": 1.2012, + "step": 4243 + }, + { + "epoch": 0.14, + "learning_rate": 1.9419740328039783e-05, + "loss": 1.1611, + "step": 4244 + }, + { + "epoch": 0.14, + "learning_rate": 1.9419392572965236e-05, + "loss": 1.1724, + "step": 4245 + }, + { + "epoch": 0.14, + "learning_rate": 1.941904471683103e-05, + "loss": 1.0713, + "step": 4246 + }, + { + "epoch": 0.14, + "learning_rate": 1.94186967596409e-05, + "loss": 1.0527, + "step": 4247 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418348701398576e-05, + "loss": 1.0278, + "step": 4248 + }, + { + "epoch": 0.14, + "learning_rate": 1.9418000542107798e-05, + "loss": 1.0371, + "step": 4249 + }, + { + "epoch": 0.14, + "learning_rate": 1.9417652281772294e-05, + "loss": 1.0918, + "step": 4250 + }, + { + "epoch": 0.14, + "learning_rate": 1.941730392039581e-05, + "loss": 1.1074, + "step": 4251 + }, + { + "epoch": 0.14, + "learning_rate": 1.941695545798207e-05, + "loss": 1.0884, + "step": 4252 + }, + { + "epoch": 0.14, + "learning_rate": 1.9416606894534822e-05, + "loss": 1.2139, + "step": 4253 + }, + { + "epoch": 0.14, + "learning_rate": 1.941625823005781e-05, + "loss": 1.1055, + "step": 4254 + }, + { + "epoch": 0.14, + "learning_rate": 1.9415909464554762e-05, + "loss": 1.1187, + "step": 4255 + }, + { + "epoch": 0.14, + "learning_rate": 1.9415560598029426e-05, + "loss": 1.1445, + "step": 4256 + }, + { + "epoch": 0.14, + "learning_rate": 1.9415211630485546e-05, + "loss": 1.0908, + "step": 4257 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414862561926867e-05, + "loss": 1.1299, + "step": 4258 + }, + { + "epoch": 0.14, + "learning_rate": 1.9414513392357134e-05, + "loss": 1.1641, + "step": 4259 + }, + { + "epoch": 0.14, + "learning_rate": 1.941416412178009e-05, + "loss": 1.0376, + "step": 4260 + }, + { + "epoch": 0.14, + "learning_rate": 1.9413814750199482e-05, + "loss": 1.0059, + "step": 4261 + }, + { + "epoch": 0.14, + "learning_rate": 1.9413465277619064e-05, + "loss": 0.5447, + "step": 4262 + }, + { + "epoch": 0.14, + "learning_rate": 1.941311570404258e-05, + "loss": 1.124, + "step": 4263 + }, + { + "epoch": 0.14, + "learning_rate": 1.941276602947378e-05, + "loss": 1.1982, + "step": 4264 + }, + { + "epoch": 0.14, + "learning_rate": 1.9412416253916423e-05, + "loss": 1.0415, + "step": 4265 + }, + { + "epoch": 0.14, + "learning_rate": 1.9412066377374256e-05, + "loss": 1.0752, + "step": 4266 + }, + { + "epoch": 0.14, + "learning_rate": 1.941171639985103e-05, + "loss": 0.9932, + "step": 4267 + }, + { + "epoch": 0.14, + "learning_rate": 1.941136632135051e-05, + "loss": 1.1406, + "step": 4268 + }, + { + "epoch": 0.14, + "learning_rate": 1.9411016141876438e-05, + "loss": 1.104, + "step": 4269 + }, + { + "epoch": 0.14, + "learning_rate": 1.9410665861432583e-05, + "loss": 1.0874, + "step": 4270 + }, + { + "epoch": 0.14, + "learning_rate": 1.9410315480022696e-05, + "loss": 1.0703, + "step": 4271 + }, + { + "epoch": 0.14, + "learning_rate": 1.940996499765054e-05, + "loss": 1.1094, + "step": 4272 + }, + { + "epoch": 0.14, + "learning_rate": 1.9409614414319874e-05, + "loss": 1.1025, + "step": 4273 + }, + { + "epoch": 0.14, + "learning_rate": 1.9409263730034457e-05, + "loss": 1.0835, + "step": 4274 + }, + { + "epoch": 0.14, + "learning_rate": 1.940891294479806e-05, + "loss": 1.1533, + "step": 4275 + }, + { + "epoch": 0.14, + "learning_rate": 1.9408562058614435e-05, + "loss": 1.1348, + "step": 4276 + }, + { + "epoch": 0.14, + "learning_rate": 1.940821107148735e-05, + "loss": 1.2158, + "step": 4277 + }, + { + "epoch": 0.14, + "learning_rate": 1.9407859983420575e-05, + "loss": 1.1597, + "step": 4278 + }, + { + "epoch": 0.14, + "learning_rate": 1.9407508794417875e-05, + "loss": 1.2178, + "step": 4279 + }, + { + "epoch": 0.14, + "learning_rate": 1.940715750448302e-05, + "loss": 1.0273, + "step": 4280 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406806113619775e-05, + "loss": 1.1035, + "step": 4281 + }, + { + "epoch": 0.14, + "learning_rate": 1.940645462183191e-05, + "loss": 1.2974, + "step": 4282 + }, + { + "epoch": 0.14, + "learning_rate": 1.9406103029123197e-05, + "loss": 0.9717, + "step": 4283 + }, + { + "epoch": 0.14, + "learning_rate": 1.940575133549741e-05, + "loss": 1.0449, + "step": 4284 + }, + { + "epoch": 0.14, + "learning_rate": 1.940539954095832e-05, + "loss": 1.0859, + "step": 4285 + }, + { + "epoch": 0.14, + "learning_rate": 1.94050476455097e-05, + "loss": 1.0513, + "step": 4286 + }, + { + "epoch": 0.14, + "learning_rate": 1.9404695649155335e-05, + "loss": 1.2402, + "step": 4287 + }, + { + "epoch": 0.14, + "learning_rate": 1.9404343551898988e-05, + "loss": 1.2002, + "step": 4288 + }, + { + "epoch": 0.14, + "learning_rate": 1.9403991353744443e-05, + "loss": 1.0161, + "step": 4289 + }, + { + "epoch": 0.14, + "learning_rate": 1.940363905469548e-05, + "loss": 1.0713, + "step": 4290 + }, + { + "epoch": 0.14, + "learning_rate": 1.9403286654755878e-05, + "loss": 1.145, + "step": 4291 + }, + { + "epoch": 0.14, + "learning_rate": 1.9402934153929417e-05, + "loss": 1.2075, + "step": 4292 + }, + { + "epoch": 0.14, + "learning_rate": 1.9402581552219878e-05, + "loss": 1.1064, + "step": 4293 + }, + { + "epoch": 0.14, + "learning_rate": 1.940222884963105e-05, + "loss": 1.2041, + "step": 4294 + }, + { + "epoch": 0.14, + "learning_rate": 1.9401876046166706e-05, + "loss": 1.0405, + "step": 4295 + }, + { + "epoch": 0.14, + "learning_rate": 1.940152314183064e-05, + "loss": 1.2705, + "step": 4296 + }, + { + "epoch": 0.14, + "learning_rate": 1.9401170136626636e-05, + "loss": 1.3271, + "step": 4297 + }, + { + "epoch": 0.14, + "learning_rate": 1.940081703055848e-05, + "loss": 1.1528, + "step": 4298 + }, + { + "epoch": 0.14, + "learning_rate": 1.9400463823629963e-05, + "loss": 1.0107, + "step": 4299 + }, + { + "epoch": 0.14, + "learning_rate": 1.9400110515844872e-05, + "loss": 1.1348, + "step": 4300 + }, + { + "epoch": 0.14, + "learning_rate": 1.9399757107207e-05, + "loss": 1.1675, + "step": 4301 + }, + { + "epoch": 0.14, + "learning_rate": 1.9399403597720137e-05, + "loss": 1.0698, + "step": 4302 + }, + { + "epoch": 0.14, + "learning_rate": 1.9399049987388078e-05, + "loss": 1.0781, + "step": 4303 + }, + { + "epoch": 0.14, + "learning_rate": 1.9398696276214612e-05, + "loss": 1.1313, + "step": 4304 + }, + { + "epoch": 0.14, + "learning_rate": 1.9398342464203537e-05, + "loss": 1.0254, + "step": 4305 + }, + { + "epoch": 0.14, + "learning_rate": 1.939798855135865e-05, + "loss": 1.1279, + "step": 4306 + }, + { + "epoch": 0.14, + "learning_rate": 1.9397634537683747e-05, + "loss": 0.9819, + "step": 4307 + }, + { + "epoch": 0.14, + "learning_rate": 1.9397280423182624e-05, + "loss": 0.9756, + "step": 4308 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396926207859085e-05, + "loss": 0.9644, + "step": 4309 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396571891716928e-05, + "loss": 1.0, + "step": 4310 + }, + { + "epoch": 0.14, + "learning_rate": 1.9396217474759953e-05, + "loss": 1.2412, + "step": 4311 + }, + { + "epoch": 0.14, + "learning_rate": 1.9395862956991964e-05, + "loss": 1.1787, + "step": 4312 + }, + { + "epoch": 0.14, + "learning_rate": 1.9395508338416764e-05, + "loss": 0.5525, + "step": 4313 + }, + { + "epoch": 0.14, + "learning_rate": 1.9395153619038158e-05, + "loss": 1.127, + "step": 4314 + }, + { + "epoch": 0.14, + "learning_rate": 1.939479879885995e-05, + "loss": 1.1528, + "step": 4315 + }, + { + "epoch": 0.14, + "learning_rate": 1.9394443877885952e-05, + "loss": 1.2036, + "step": 4316 + }, + { + "epoch": 0.14, + "learning_rate": 1.939408885611997e-05, + "loss": 1.1934, + "step": 4317 + }, + { + "epoch": 0.14, + "learning_rate": 1.9393733733565807e-05, + "loss": 0.9663, + "step": 4318 + }, + { + "epoch": 0.14, + "learning_rate": 1.939337851022728e-05, + "loss": 1.1147, + "step": 4319 + }, + { + "epoch": 0.14, + "learning_rate": 1.9393023186108194e-05, + "loss": 1.0654, + "step": 4320 + }, + { + "epoch": 0.14, + "learning_rate": 1.939266776121237e-05, + "loss": 1.0493, + "step": 4321 + }, + { + "epoch": 0.14, + "learning_rate": 1.9392312235543613e-05, + "loss": 1.1113, + "step": 4322 + }, + { + "epoch": 0.14, + "learning_rate": 1.9391956609105742e-05, + "loss": 0.512, + "step": 4323 + }, + { + "epoch": 0.14, + "learning_rate": 1.939160088190257e-05, + "loss": 1.0034, + "step": 4324 + }, + { + "epoch": 0.14, + "learning_rate": 1.9391245053937914e-05, + "loss": 0.9941, + "step": 4325 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390889125215594e-05, + "loss": 1.1289, + "step": 4326 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390533095739426e-05, + "loss": 1.0625, + "step": 4327 + }, + { + "epoch": 0.14, + "learning_rate": 1.9390176965513233e-05, + "loss": 1.1455, + "step": 4328 + }, + { + "epoch": 0.14, + "learning_rate": 1.938982073454083e-05, + "loss": 1.0854, + "step": 4329 + }, + { + "epoch": 0.14, + "learning_rate": 1.9389464402826043e-05, + "loss": 1.0649, + "step": 4330 + }, + { + "epoch": 0.14, + "learning_rate": 1.9389107970372697e-05, + "loss": 1.1587, + "step": 4331 + }, + { + "epoch": 0.14, + "learning_rate": 1.9388751437184613e-05, + "loss": 1.2046, + "step": 4332 + }, + { + "epoch": 0.14, + "learning_rate": 1.938839480326562e-05, + "loss": 1.3423, + "step": 4333 + }, + { + "epoch": 0.14, + "learning_rate": 1.9388038068619537e-05, + "loss": 1.1479, + "step": 4334 + }, + { + "epoch": 0.14, + "learning_rate": 1.93876812332502e-05, + "loss": 0.96, + "step": 4335 + }, + { + "epoch": 0.14, + "learning_rate": 1.9387324297161427e-05, + "loss": 1.0571, + "step": 4336 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386967260357056e-05, + "loss": 1.0854, + "step": 4337 + }, + { + "epoch": 0.14, + "learning_rate": 1.9386610122840915e-05, + "loss": 1.1631, + "step": 4338 + }, + { + "epoch": 0.14, + "learning_rate": 1.938625288461684e-05, + "loss": 1.1025, + "step": 4339 + }, + { + "epoch": 0.14, + "learning_rate": 1.9385895545688657e-05, + "loss": 1.0293, + "step": 4340 + }, + { + "epoch": 0.14, + "learning_rate": 1.93855381060602e-05, + "loss": 1.0986, + "step": 4341 + }, + { + "epoch": 0.14, + "learning_rate": 1.9385180565735312e-05, + "loss": 1.229, + "step": 4342 + }, + { + "epoch": 0.14, + "learning_rate": 1.938482292471782e-05, + "loss": 1.0376, + "step": 4343 + }, + { + "epoch": 0.14, + "learning_rate": 1.9384465183011563e-05, + "loss": 1.0986, + "step": 4344 + }, + { + "epoch": 0.14, + "learning_rate": 1.9384107340620383e-05, + "loss": 1.127, + "step": 4345 + }, + { + "epoch": 0.14, + "learning_rate": 1.938374939754812e-05, + "loss": 1.0762, + "step": 4346 + }, + { + "epoch": 0.14, + "learning_rate": 1.9383391353798605e-05, + "loss": 1.1196, + "step": 4347 + }, + { + "epoch": 0.14, + "learning_rate": 1.938303320937569e-05, + "loss": 1.0889, + "step": 4348 + }, + { + "epoch": 0.14, + "learning_rate": 1.938267496428321e-05, + "loss": 1.1143, + "step": 4349 + }, + { + "epoch": 0.14, + "learning_rate": 1.9382316618525016e-05, + "loss": 0.498, + "step": 4350 + }, + { + "epoch": 0.14, + "learning_rate": 1.938195817210495e-05, + "loss": 1.0952, + "step": 4351 + }, + { + "epoch": 0.14, + "learning_rate": 1.938159962502685e-05, + "loss": 1.1123, + "step": 4352 + }, + { + "epoch": 0.14, + "learning_rate": 1.9381240977294573e-05, + "loss": 1.1777, + "step": 4353 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380882228911962e-05, + "loss": 1.1001, + "step": 4354 + }, + { + "epoch": 0.14, + "learning_rate": 1.9380523379882867e-05, + "loss": 1.1602, + "step": 4355 + }, + { + "epoch": 0.14, + "learning_rate": 1.938016443021114e-05, + "loss": 1.1191, + "step": 4356 + }, + { + "epoch": 0.14, + "learning_rate": 1.9379805379900627e-05, + "loss": 1.1157, + "step": 4357 + }, + { + "epoch": 0.14, + "learning_rate": 1.9379446228955183e-05, + "loss": 1.1543, + "step": 4358 + }, + { + "epoch": 0.14, + "learning_rate": 1.9379086977378664e-05, + "loss": 1.0435, + "step": 4359 + }, + { + "epoch": 0.14, + "learning_rate": 1.937872762517492e-05, + "loss": 1.2446, + "step": 4360 + }, + { + "epoch": 0.14, + "learning_rate": 1.9378368172347813e-05, + "loss": 1.1113, + "step": 4361 + }, + { + "epoch": 0.14, + "learning_rate": 1.937800861890119e-05, + "loss": 1.1177, + "step": 4362 + }, + { + "epoch": 0.14, + "learning_rate": 1.9377648964838913e-05, + "loss": 1.1143, + "step": 4363 + }, + { + "epoch": 0.14, + "learning_rate": 1.9377289210164844e-05, + "loss": 1.0103, + "step": 4364 + }, + { + "epoch": 0.14, + "learning_rate": 1.937692935488284e-05, + "loss": 1.1479, + "step": 4365 + }, + { + "epoch": 0.14, + "learning_rate": 1.9376569398996762e-05, + "loss": 1.0, + "step": 4366 + }, + { + "epoch": 0.14, + "learning_rate": 1.937620934251047e-05, + "loss": 0.999, + "step": 4367 + }, + { + "epoch": 0.14, + "learning_rate": 1.937584918542783e-05, + "loss": 1.0918, + "step": 4368 + }, + { + "epoch": 0.14, + "learning_rate": 1.9375488927752705e-05, + "loss": 1.063, + "step": 4369 + }, + { + "epoch": 0.14, + "learning_rate": 1.9375128569488958e-05, + "loss": 1.1924, + "step": 4370 + }, + { + "epoch": 0.14, + "learning_rate": 1.9374768110640462e-05, + "loss": 1.0967, + "step": 4371 + }, + { + "epoch": 0.14, + "learning_rate": 1.9374407551211076e-05, + "loss": 1.2236, + "step": 4372 + }, + { + "epoch": 0.14, + "learning_rate": 1.9374046891204676e-05, + "loss": 1.1733, + "step": 4373 + }, + { + "epoch": 0.14, + "learning_rate": 1.9373686130625124e-05, + "loss": 0.9644, + "step": 4374 + }, + { + "epoch": 0.14, + "learning_rate": 1.9373325269476293e-05, + "loss": 1.1543, + "step": 4375 + }, + { + "epoch": 0.14, + "learning_rate": 1.937296430776206e-05, + "loss": 1.1528, + "step": 4376 + }, + { + "epoch": 0.14, + "learning_rate": 1.937260324548629e-05, + "loss": 0.9619, + "step": 4377 + }, + { + "epoch": 0.14, + "learning_rate": 1.937224208265286e-05, + "loss": 1.0483, + "step": 4378 + }, + { + "epoch": 0.14, + "learning_rate": 1.9371880819265644e-05, + "loss": 1.0132, + "step": 4379 + }, + { + "epoch": 0.14, + "learning_rate": 1.937151945532852e-05, + "loss": 0.9995, + "step": 4380 + }, + { + "epoch": 0.14, + "learning_rate": 1.937115799084537e-05, + "loss": 0.9995, + "step": 4381 + }, + { + "epoch": 0.14, + "learning_rate": 1.937079642582006e-05, + "loss": 1.1001, + "step": 4382 + }, + { + "epoch": 0.14, + "learning_rate": 1.9370434760256475e-05, + "loss": 1.1709, + "step": 4383 + }, + { + "epoch": 0.14, + "learning_rate": 1.93700729941585e-05, + "loss": 1.1084, + "step": 4384 + }, + { + "epoch": 0.14, + "learning_rate": 1.936971112753001e-05, + "loss": 1.0835, + "step": 4385 + }, + { + "epoch": 0.14, + "learning_rate": 1.936934916037489e-05, + "loss": 1.0474, + "step": 4386 + }, + { + "epoch": 0.14, + "learning_rate": 1.936898709269702e-05, + "loss": 1.2183, + "step": 4387 + }, + { + "epoch": 0.14, + "learning_rate": 1.936862492450029e-05, + "loss": 1.1743, + "step": 4388 + }, + { + "epoch": 0.14, + "learning_rate": 1.9368262655788586e-05, + "loss": 1.1621, + "step": 4389 + }, + { + "epoch": 0.14, + "learning_rate": 1.936790028656579e-05, + "loss": 1.084, + "step": 4390 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367537816835787e-05, + "loss": 1.2383, + "step": 4391 + }, + { + "epoch": 0.14, + "learning_rate": 1.9367175246602477e-05, + "loss": 1.1348, + "step": 4392 + }, + { + "epoch": 0.14, + "learning_rate": 1.9366812575869744e-05, + "loss": 1.0464, + "step": 4393 + }, + { + "epoch": 0.14, + "learning_rate": 1.936644980464147e-05, + "loss": 1.2798, + "step": 4394 + }, + { + "epoch": 0.14, + "learning_rate": 1.9366086932921565e-05, + "loss": 1.2158, + "step": 4395 + }, + { + "epoch": 0.14, + "learning_rate": 1.936572396071391e-05, + "loss": 1.1909, + "step": 4396 + }, + { + "epoch": 0.14, + "learning_rate": 1.9365360888022403e-05, + "loss": 1.0532, + "step": 4397 + }, + { + "epoch": 0.14, + "learning_rate": 1.936499771485094e-05, + "loss": 1.2209, + "step": 4398 + }, + { + "epoch": 0.14, + "learning_rate": 1.936463444120341e-05, + "loss": 1.0015, + "step": 4399 + }, + { + "epoch": 0.14, + "learning_rate": 1.9364271067083723e-05, + "loss": 1.0894, + "step": 4400 + }, + { + "epoch": 0.14, + "learning_rate": 1.936390759249577e-05, + "loss": 1.022, + "step": 4401 + }, + { + "epoch": 0.14, + "learning_rate": 1.9363544017443453e-05, + "loss": 1.2222, + "step": 4402 + }, + { + "epoch": 0.14, + "learning_rate": 1.9363180341930667e-05, + "loss": 0.9756, + "step": 4403 + }, + { + "epoch": 0.14, + "learning_rate": 1.936281656596132e-05, + "loss": 1.167, + "step": 4404 + }, + { + "epoch": 0.14, + "learning_rate": 1.9362452689539316e-05, + "loss": 1.1016, + "step": 4405 + }, + { + "epoch": 0.14, + "learning_rate": 1.9362088712668557e-05, + "loss": 1.2251, + "step": 4406 + }, + { + "epoch": 0.14, + "learning_rate": 1.9361724635352946e-05, + "loss": 1.3018, + "step": 4407 + }, + { + "epoch": 0.14, + "learning_rate": 1.936136045759639e-05, + "loss": 1.1211, + "step": 4408 + }, + { + "epoch": 0.14, + "learning_rate": 1.93609961794028e-05, + "loss": 1.1455, + "step": 4409 + }, + { + "epoch": 0.14, + "learning_rate": 1.936063180077608e-05, + "loss": 1.0332, + "step": 4410 + }, + { + "epoch": 0.14, + "learning_rate": 1.936026732172014e-05, + "loss": 1.1729, + "step": 4411 + }, + { + "epoch": 0.14, + "learning_rate": 1.935990274223889e-05, + "loss": 1.0078, + "step": 4412 + }, + { + "epoch": 0.14, + "learning_rate": 1.935953806233624e-05, + "loss": 1.0972, + "step": 4413 + }, + { + "epoch": 0.14, + "learning_rate": 1.9359173282016112e-05, + "loss": 0.5273, + "step": 4414 + }, + { + "epoch": 0.14, + "learning_rate": 1.9358808401282408e-05, + "loss": 1.1475, + "step": 4415 + }, + { + "epoch": 0.14, + "learning_rate": 1.9358443420139047e-05, + "loss": 1.1284, + "step": 4416 + }, + { + "epoch": 0.14, + "learning_rate": 1.935807833858995e-05, + "loss": 1.2251, + "step": 4417 + }, + { + "epoch": 0.14, + "learning_rate": 1.9357713156639026e-05, + "loss": 1.1328, + "step": 4418 + }, + { + "epoch": 0.14, + "learning_rate": 1.93573478742902e-05, + "loss": 1.0488, + "step": 4419 + }, + { + "epoch": 0.14, + "learning_rate": 1.9356982491547384e-05, + "loss": 1.0537, + "step": 4420 + }, + { + "epoch": 0.14, + "learning_rate": 1.9356617008414503e-05, + "loss": 1.0122, + "step": 4421 + }, + { + "epoch": 0.14, + "learning_rate": 1.9356251424895476e-05, + "loss": 1.0337, + "step": 4422 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355885740994225e-05, + "loss": 1.1113, + "step": 4423 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355519956714678e-05, + "loss": 1.1792, + "step": 4424 + }, + { + "epoch": 0.14, + "learning_rate": 1.9355154072060753e-05, + "loss": 1.0063, + "step": 4425 + }, + { + "epoch": 0.14, + "learning_rate": 1.9354788087036383e-05, + "loss": 1.0503, + "step": 4426 + }, + { + "epoch": 0.14, + "learning_rate": 1.9354422001645486e-05, + "loss": 1.0908, + "step": 4427 + }, + { + "epoch": 0.14, + "learning_rate": 1.9354055815891993e-05, + "loss": 1.2197, + "step": 4428 + }, + { + "epoch": 0.14, + "learning_rate": 1.9353689529779837e-05, + "loss": 0.9863, + "step": 4429 + }, + { + "epoch": 0.14, + "learning_rate": 1.9353323143312943e-05, + "loss": 1.1392, + "step": 4430 + }, + { + "epoch": 0.14, + "learning_rate": 1.9352956656495246e-05, + "loss": 1.1172, + "step": 4431 + }, + { + "epoch": 0.14, + "learning_rate": 1.935259006933067e-05, + "loss": 1.1172, + "step": 4432 + }, + { + "epoch": 0.14, + "learning_rate": 1.9352223381823158e-05, + "loss": 1.1201, + "step": 4433 + }, + { + "epoch": 0.14, + "learning_rate": 1.935185659397664e-05, + "loss": 1.2188, + "step": 4434 + }, + { + "epoch": 0.14, + "learning_rate": 1.9351489705795048e-05, + "loss": 1.0508, + "step": 4435 + }, + { + "epoch": 0.14, + "learning_rate": 1.935112271728232e-05, + "loss": 1.2056, + "step": 4436 + }, + { + "epoch": 0.14, + "learning_rate": 1.93507556284424e-05, + "loss": 1.0137, + "step": 4437 + }, + { + "epoch": 0.14, + "learning_rate": 1.9350388439279215e-05, + "loss": 1.1602, + "step": 4438 + }, + { + "epoch": 0.14, + "learning_rate": 1.9350021149796716e-05, + "loss": 1.1753, + "step": 4439 + }, + { + "epoch": 0.14, + "learning_rate": 1.9349653759998835e-05, + "loss": 1.1357, + "step": 4440 + }, + { + "epoch": 0.14, + "learning_rate": 1.9349286269889518e-05, + "loss": 1.1108, + "step": 4441 + }, + { + "epoch": 0.14, + "learning_rate": 1.9348918679472706e-05, + "loss": 1.0337, + "step": 4442 + }, + { + "epoch": 0.14, + "learning_rate": 1.934855098875234e-05, + "loss": 1.2998, + "step": 4443 + }, + { + "epoch": 0.14, + "learning_rate": 1.9348183197732373e-05, + "loss": 0.9824, + "step": 4444 + }, + { + "epoch": 0.14, + "learning_rate": 1.9347815306416743e-05, + "loss": 0.9888, + "step": 4445 + }, + { + "epoch": 0.14, + "learning_rate": 1.9347447314809406e-05, + "loss": 1.1846, + "step": 4446 + }, + { + "epoch": 0.14, + "learning_rate": 1.93470792229143e-05, + "loss": 1.1836, + "step": 4447 + }, + { + "epoch": 0.14, + "learning_rate": 1.9346711030735382e-05, + "loss": 1.0332, + "step": 4448 + }, + { + "epoch": 0.14, + "learning_rate": 1.9346342738276593e-05, + "loss": 1.1045, + "step": 4449 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345974345541896e-05, + "loss": 1.1768, + "step": 4450 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345605852535236e-05, + "loss": 1.1816, + "step": 4451 + }, + { + "epoch": 0.14, + "learning_rate": 1.9345237259260566e-05, + "loss": 1.1235, + "step": 4452 + }, + { + "epoch": 0.14, + "learning_rate": 1.9344868565721848e-05, + "loss": 0.9946, + "step": 4453 + }, + { + "epoch": 0.14, + "learning_rate": 1.934449977192303e-05, + "loss": 1.061, + "step": 4454 + }, + { + "epoch": 0.14, + "learning_rate": 1.9344130877868072e-05, + "loss": 1.1104, + "step": 4455 + }, + { + "epoch": 0.14, + "learning_rate": 1.934376188356093e-05, + "loss": 1.0669, + "step": 4456 + }, + { + "epoch": 0.14, + "learning_rate": 1.9343392789005566e-05, + "loss": 1.0181, + "step": 4457 + }, + { + "epoch": 0.14, + "learning_rate": 1.9343023594205936e-05, + "loss": 1.0791, + "step": 4458 + }, + { + "epoch": 0.14, + "learning_rate": 1.9342654299166007e-05, + "loss": 0.96, + "step": 4459 + }, + { + "epoch": 0.14, + "learning_rate": 1.9342284903889735e-05, + "loss": 1.0815, + "step": 4460 + }, + { + "epoch": 0.14, + "learning_rate": 1.9341915408381085e-05, + "loss": 1.126, + "step": 4461 + }, + { + "epoch": 0.14, + "learning_rate": 1.934154581264402e-05, + "loss": 1.0278, + "step": 4462 + }, + { + "epoch": 0.14, + "learning_rate": 1.934117611668251e-05, + "loss": 1.1743, + "step": 4463 + }, + { + "epoch": 0.14, + "learning_rate": 1.934080632050052e-05, + "loss": 1.1479, + "step": 4464 + }, + { + "epoch": 0.14, + "learning_rate": 1.934043642410201e-05, + "loss": 1.0137, + "step": 4465 + }, + { + "epoch": 0.14, + "learning_rate": 1.934006642749096e-05, + "loss": 1.0596, + "step": 4466 + }, + { + "epoch": 0.14, + "learning_rate": 1.9339696330671335e-05, + "loss": 1.2036, + "step": 4467 + }, + { + "epoch": 0.14, + "learning_rate": 1.9339326133647102e-05, + "loss": 1.167, + "step": 4468 + }, + { + "epoch": 0.14, + "learning_rate": 1.9338955836422236e-05, + "loss": 1.124, + "step": 4469 + }, + { + "epoch": 0.14, + "learning_rate": 1.933858543900071e-05, + "loss": 0.9995, + "step": 4470 + }, + { + "epoch": 0.14, + "learning_rate": 1.9338214941386497e-05, + "loss": 1.1211, + "step": 4471 + }, + { + "epoch": 0.14, + "learning_rate": 1.9337844343583577e-05, + "loss": 1.1045, + "step": 4472 + }, + { + "epoch": 0.14, + "learning_rate": 1.9337473645595917e-05, + "loss": 1.1543, + "step": 4473 + }, + { + "epoch": 0.14, + "learning_rate": 1.93371028474275e-05, + "loss": 1.1318, + "step": 4474 + }, + { + "epoch": 0.14, + "learning_rate": 1.9336731949082303e-05, + "loss": 0.9951, + "step": 4475 + }, + { + "epoch": 0.14, + "learning_rate": 1.9336360950564308e-05, + "loss": 1.1812, + "step": 4476 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335989851877494e-05, + "loss": 0.9849, + "step": 4477 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335618653025835e-05, + "loss": 1.1304, + "step": 4478 + }, + { + "epoch": 0.14, + "learning_rate": 1.9335247354013323e-05, + "loss": 1.0889, + "step": 4479 + }, + { + "epoch": 0.14, + "learning_rate": 1.9334875954843943e-05, + "loss": 1.2412, + "step": 4480 + }, + { + "epoch": 0.14, + "learning_rate": 1.933450445552167e-05, + "loss": 1.084, + "step": 4481 + }, + { + "epoch": 0.14, + "learning_rate": 1.9334132856050497e-05, + "loss": 1.0293, + "step": 4482 + }, + { + "epoch": 0.14, + "learning_rate": 1.9333761156434406e-05, + "loss": 0.8838, + "step": 4483 + }, + { + "epoch": 0.14, + "learning_rate": 1.9333389356677395e-05, + "loss": 1.0938, + "step": 4484 + }, + { + "epoch": 0.14, + "learning_rate": 1.9333017456783438e-05, + "loss": 1.1577, + "step": 4485 + }, + { + "epoch": 0.14, + "learning_rate": 1.933264545675654e-05, + "loss": 1.1011, + "step": 4486 + }, + { + "epoch": 0.14, + "learning_rate": 1.933227335660068e-05, + "loss": 1.0293, + "step": 4487 + }, + { + "epoch": 0.14, + "learning_rate": 1.9331901156319856e-05, + "loss": 0.9912, + "step": 4488 + }, + { + "epoch": 0.14, + "learning_rate": 1.933152885591806e-05, + "loss": 0.9844, + "step": 4489 + }, + { + "epoch": 0.14, + "learning_rate": 1.933115645539929e-05, + "loss": 1.0371, + "step": 4490 + }, + { + "epoch": 0.14, + "learning_rate": 1.9330783954767537e-05, + "loss": 1.1899, + "step": 4491 + }, + { + "epoch": 0.14, + "learning_rate": 1.9330411354026803e-05, + "loss": 1.1904, + "step": 4492 + }, + { + "epoch": 0.14, + "learning_rate": 1.9330038653181072e-05, + "loss": 0.9775, + "step": 4493 + }, + { + "epoch": 0.14, + "learning_rate": 1.932966585223436e-05, + "loss": 1.0391, + "step": 4494 + }, + { + "epoch": 0.14, + "learning_rate": 1.932929295119066e-05, + "loss": 1.2461, + "step": 4495 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328919950053966e-05, + "loss": 1.0693, + "step": 4496 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328546848828287e-05, + "loss": 1.104, + "step": 4497 + }, + { + "epoch": 0.14, + "learning_rate": 1.9328173647517628e-05, + "loss": 1.1138, + "step": 4498 + }, + { + "epoch": 0.14, + "learning_rate": 1.9327800346125985e-05, + "loss": 1.1426, + "step": 4499 + }, + { + "epoch": 0.14, + "learning_rate": 1.9327426944657372e-05, + "loss": 1.1665, + "step": 4500 + }, + { + "epoch": 0.14, + "learning_rate": 1.9327053443115787e-05, + "loss": 1.1768, + "step": 4501 + }, + { + "epoch": 0.14, + "learning_rate": 1.9326679841505246e-05, + "loss": 1.1138, + "step": 4502 + }, + { + "epoch": 0.14, + "learning_rate": 1.932630613982975e-05, + "loss": 1.1313, + "step": 4503 + }, + { + "epoch": 0.14, + "learning_rate": 1.932593233809331e-05, + "loss": 1.1201, + "step": 4504 + }, + { + "epoch": 0.14, + "learning_rate": 1.9325558436299937e-05, + "loss": 1.1807, + "step": 4505 + }, + { + "epoch": 0.14, + "learning_rate": 1.9325184434453646e-05, + "loss": 1.1997, + "step": 4506 + }, + { + "epoch": 0.14, + "learning_rate": 1.9324810332558446e-05, + "loss": 1.0977, + "step": 4507 + }, + { + "epoch": 0.14, + "learning_rate": 1.9324436130618348e-05, + "loss": 0.5608, + "step": 4508 + }, + { + "epoch": 0.14, + "learning_rate": 1.9324061828637374e-05, + "loss": 1.1958, + "step": 4509 + }, + { + "epoch": 0.14, + "learning_rate": 1.932368742661954e-05, + "loss": 1.1626, + "step": 4510 + }, + { + "epoch": 0.14, + "learning_rate": 1.932331292456885e-05, + "loss": 1.0737, + "step": 4511 + }, + { + "epoch": 0.14, + "learning_rate": 1.9322938322489334e-05, + "loss": 1.0952, + "step": 4512 + }, + { + "epoch": 0.14, + "learning_rate": 1.932256362038501e-05, + "loss": 1.1802, + "step": 4513 + }, + { + "epoch": 0.14, + "learning_rate": 1.9322188818259892e-05, + "loss": 1.2324, + "step": 4514 + }, + { + "epoch": 0.14, + "learning_rate": 1.932181391611801e-05, + "loss": 1.0581, + "step": 4515 + }, + { + "epoch": 0.14, + "learning_rate": 1.9321438913963378e-05, + "loss": 1.0801, + "step": 4516 + }, + { + "epoch": 0.14, + "learning_rate": 1.932106381180002e-05, + "loss": 1.1904, + "step": 4517 + }, + { + "epoch": 0.14, + "learning_rate": 1.932068860963197e-05, + "loss": 1.0327, + "step": 4518 + }, + { + "epoch": 0.14, + "learning_rate": 1.9320313307463244e-05, + "loss": 1.021, + "step": 4519 + }, + { + "epoch": 0.14, + "learning_rate": 1.931993790529787e-05, + "loss": 1.1274, + "step": 4520 + }, + { + "epoch": 0.14, + "learning_rate": 1.931956240313988e-05, + "loss": 1.1792, + "step": 4521 + }, + { + "epoch": 0.14, + "learning_rate": 1.9319186800993296e-05, + "loss": 1.0332, + "step": 4522 + }, + { + "epoch": 0.14, + "learning_rate": 1.9318811098862157e-05, + "loss": 1.1323, + "step": 4523 + }, + { + "epoch": 0.14, + "learning_rate": 1.9318435296750485e-05, + "loss": 1.1099, + "step": 4524 + }, + { + "epoch": 0.14, + "learning_rate": 1.9318059394662315e-05, + "loss": 0.5286, + "step": 4525 + }, + { + "epoch": 0.14, + "learning_rate": 1.9317683392601682e-05, + "loss": 1.1475, + "step": 4526 + }, + { + "epoch": 0.14, + "learning_rate": 1.9317307290572618e-05, + "loss": 1.0898, + "step": 4527 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316931088579158e-05, + "loss": 1.1543, + "step": 4528 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316554786625337e-05, + "loss": 1.0854, + "step": 4529 + }, + { + "epoch": 0.14, + "learning_rate": 1.9316178384715195e-05, + "loss": 1.1094, + "step": 4530 + }, + { + "epoch": 0.14, + "learning_rate": 1.9315801882852773e-05, + "loss": 1.1777, + "step": 4531 + }, + { + "epoch": 0.14, + "learning_rate": 1.9315425281042106e-05, + "loss": 1.1885, + "step": 4532 + }, + { + "epoch": 0.14, + "learning_rate": 1.9315048579287234e-05, + "loss": 1.0801, + "step": 4533 + }, + { + "epoch": 0.14, + "learning_rate": 1.93146717775922e-05, + "loss": 0.9648, + "step": 4534 + }, + { + "epoch": 0.15, + "learning_rate": 1.931429487596105e-05, + "loss": 1.1475, + "step": 4535 + }, + { + "epoch": 0.15, + "learning_rate": 1.9313917874397822e-05, + "loss": 1.0244, + "step": 4536 + }, + { + "epoch": 0.15, + "learning_rate": 1.931354077290656e-05, + "loss": 1.1953, + "step": 4537 + }, + { + "epoch": 0.15, + "learning_rate": 1.931316357149132e-05, + "loss": 1.1167, + "step": 4538 + }, + { + "epoch": 0.15, + "learning_rate": 1.9312786270156135e-05, + "loss": 1.1338, + "step": 4539 + }, + { + "epoch": 0.15, + "learning_rate": 1.9312408868905066e-05, + "loss": 1.2603, + "step": 4540 + }, + { + "epoch": 0.15, + "learning_rate": 1.9312031367742155e-05, + "loss": 1.1743, + "step": 4541 + }, + { + "epoch": 0.15, + "learning_rate": 1.9311653766671454e-05, + "loss": 1.0918, + "step": 4542 + }, + { + "epoch": 0.15, + "learning_rate": 1.931127606569701e-05, + "loss": 1.0991, + "step": 4543 + }, + { + "epoch": 0.15, + "learning_rate": 1.9310898264822882e-05, + "loss": 1.1313, + "step": 4544 + }, + { + "epoch": 0.15, + "learning_rate": 1.931052036405312e-05, + "loss": 1.2085, + "step": 4545 + }, + { + "epoch": 0.15, + "learning_rate": 1.931014236339178e-05, + "loss": 1.0322, + "step": 4546 + }, + { + "epoch": 0.15, + "learning_rate": 1.9309764262842913e-05, + "loss": 1.2148, + "step": 4547 + }, + { + "epoch": 0.15, + "learning_rate": 1.9309386062410584e-05, + "loss": 1.1299, + "step": 4548 + }, + { + "epoch": 0.15, + "learning_rate": 1.9309007762098843e-05, + "loss": 1.0303, + "step": 4549 + }, + { + "epoch": 0.15, + "learning_rate": 1.930862936191175e-05, + "loss": 1.0991, + "step": 4550 + }, + { + "epoch": 0.15, + "learning_rate": 1.9308250861853367e-05, + "loss": 1.0776, + "step": 4551 + }, + { + "epoch": 0.15, + "learning_rate": 1.9307872261927753e-05, + "loss": 1.105, + "step": 4552 + }, + { + "epoch": 0.15, + "learning_rate": 1.9307493562138974e-05, + "loss": 1.0127, + "step": 4553 + }, + { + "epoch": 0.15, + "learning_rate": 1.930711476249109e-05, + "loss": 1.085, + "step": 4554 + }, + { + "epoch": 0.15, + "learning_rate": 1.9306735862988163e-05, + "loss": 0.9263, + "step": 4555 + }, + { + "epoch": 0.15, + "learning_rate": 1.9306356863634263e-05, + "loss": 1.0205, + "step": 4556 + }, + { + "epoch": 0.15, + "learning_rate": 1.930597776443345e-05, + "loss": 1.064, + "step": 4557 + }, + { + "epoch": 0.15, + "learning_rate": 1.9305598565389794e-05, + "loss": 0.8887, + "step": 4558 + }, + { + "epoch": 0.15, + "learning_rate": 1.9305219266507368e-05, + "loss": 1.0796, + "step": 4559 + }, + { + "epoch": 0.15, + "learning_rate": 1.930483986779024e-05, + "loss": 1.0015, + "step": 4560 + }, + { + "epoch": 0.15, + "learning_rate": 1.9304460369242473e-05, + "loss": 1.0669, + "step": 4561 + }, + { + "epoch": 0.15, + "learning_rate": 1.9304080770868147e-05, + "loss": 1.0293, + "step": 4562 + }, + { + "epoch": 0.15, + "learning_rate": 1.930370107267133e-05, + "loss": 1.0972, + "step": 4563 + }, + { + "epoch": 0.15, + "learning_rate": 1.93033212746561e-05, + "loss": 1.0342, + "step": 4564 + }, + { + "epoch": 0.15, + "learning_rate": 1.9302941376826527e-05, + "loss": 1.2158, + "step": 4565 + }, + { + "epoch": 0.15, + "learning_rate": 1.930256137918669e-05, + "loss": 1.0205, + "step": 4566 + }, + { + "epoch": 0.15, + "learning_rate": 1.9302181281740662e-05, + "loss": 1.0479, + "step": 4567 + }, + { + "epoch": 0.15, + "learning_rate": 1.9301801084492527e-05, + "loss": 1.165, + "step": 4568 + }, + { + "epoch": 0.15, + "learning_rate": 1.930142078744636e-05, + "loss": 1.21, + "step": 4569 + }, + { + "epoch": 0.15, + "learning_rate": 1.930104039060624e-05, + "loss": 1.1079, + "step": 4570 + }, + { + "epoch": 0.15, + "learning_rate": 1.9300659893976256e-05, + "loss": 1.147, + "step": 4571 + }, + { + "epoch": 0.15, + "learning_rate": 1.9300279297560484e-05, + "loss": 1.1113, + "step": 4572 + }, + { + "epoch": 0.15, + "learning_rate": 1.9299898601363004e-05, + "loss": 1.0796, + "step": 4573 + }, + { + "epoch": 0.15, + "learning_rate": 1.9299517805387905e-05, + "loss": 1.0107, + "step": 4574 + }, + { + "epoch": 0.15, + "learning_rate": 1.9299136909639275e-05, + "loss": 1.103, + "step": 4575 + }, + { + "epoch": 0.15, + "learning_rate": 1.9298755914121198e-05, + "loss": 1.1729, + "step": 4576 + }, + { + "epoch": 0.15, + "learning_rate": 1.929837481883776e-05, + "loss": 1.1162, + "step": 4577 + }, + { + "epoch": 0.15, + "learning_rate": 1.929799362379305e-05, + "loss": 0.8823, + "step": 4578 + }, + { + "epoch": 0.15, + "learning_rate": 1.929761232899116e-05, + "loss": 1.0122, + "step": 4579 + }, + { + "epoch": 0.15, + "learning_rate": 1.929723093443618e-05, + "loss": 1.1592, + "step": 4580 + }, + { + "epoch": 0.15, + "learning_rate": 1.9296849440132203e-05, + "loss": 1.1431, + "step": 4581 + }, + { + "epoch": 0.15, + "learning_rate": 1.929646784608332e-05, + "loss": 1.0347, + "step": 4582 + }, + { + "epoch": 0.15, + "learning_rate": 1.9296086152293622e-05, + "loss": 1.1309, + "step": 4583 + }, + { + "epoch": 0.15, + "learning_rate": 1.929570435876721e-05, + "loss": 1.1025, + "step": 4584 + }, + { + "epoch": 0.15, + "learning_rate": 1.929532246550818e-05, + "loss": 1.188, + "step": 4585 + }, + { + "epoch": 0.15, + "learning_rate": 1.9294940472520626e-05, + "loss": 1.1597, + "step": 4586 + }, + { + "epoch": 0.15, + "learning_rate": 1.929455837980865e-05, + "loss": 1.1455, + "step": 4587 + }, + { + "epoch": 0.15, + "learning_rate": 1.929417618737635e-05, + "loss": 1.0767, + "step": 4588 + }, + { + "epoch": 0.15, + "learning_rate": 1.929379389522782e-05, + "loss": 1.125, + "step": 4589 + }, + { + "epoch": 0.15, + "learning_rate": 1.929341150336717e-05, + "loss": 1.106, + "step": 4590 + }, + { + "epoch": 0.15, + "learning_rate": 1.9293029011798504e-05, + "loss": 1.207, + "step": 4591 + }, + { + "epoch": 0.15, + "learning_rate": 1.9292646420525917e-05, + "loss": 1.0664, + "step": 4592 + }, + { + "epoch": 0.15, + "learning_rate": 1.9292263729553523e-05, + "loss": 1.1436, + "step": 4593 + }, + { + "epoch": 0.15, + "learning_rate": 1.929188093888542e-05, + "loss": 1.0645, + "step": 4594 + }, + { + "epoch": 0.15, + "learning_rate": 1.929149804852572e-05, + "loss": 1.2549, + "step": 4595 + }, + { + "epoch": 0.15, + "learning_rate": 1.9291115058478528e-05, + "loss": 0.5142, + "step": 4596 + }, + { + "epoch": 0.15, + "learning_rate": 1.9290731968747953e-05, + "loss": 1.1382, + "step": 4597 + }, + { + "epoch": 0.15, + "learning_rate": 1.929034877933811e-05, + "loss": 1.1104, + "step": 4598 + }, + { + "epoch": 0.15, + "learning_rate": 1.9289965490253103e-05, + "loss": 1.0732, + "step": 4599 + }, + { + "epoch": 0.15, + "learning_rate": 1.9289582101497048e-05, + "loss": 1.1157, + "step": 4600 + }, + { + "epoch": 0.15, + "learning_rate": 1.928919861307406e-05, + "loss": 1.0586, + "step": 4601 + }, + { + "epoch": 0.15, + "learning_rate": 1.9288815024988253e-05, + "loss": 1.0762, + "step": 4602 + }, + { + "epoch": 0.15, + "learning_rate": 1.9288431337243737e-05, + "loss": 1.0361, + "step": 4603 + }, + { + "epoch": 0.15, + "learning_rate": 1.9288047549844635e-05, + "loss": 1.0752, + "step": 4604 + }, + { + "epoch": 0.15, + "learning_rate": 1.9287663662795065e-05, + "loss": 1.1401, + "step": 4605 + }, + { + "epoch": 0.15, + "learning_rate": 1.928727967609914e-05, + "loss": 1.083, + "step": 4606 + }, + { + "epoch": 0.15, + "learning_rate": 1.9286895589760984e-05, + "loss": 1.0454, + "step": 4607 + }, + { + "epoch": 0.15, + "learning_rate": 1.9286511403784715e-05, + "loss": 1.064, + "step": 4608 + }, + { + "epoch": 0.15, + "learning_rate": 1.9286127118174458e-05, + "loss": 0.9766, + "step": 4609 + }, + { + "epoch": 0.15, + "learning_rate": 1.9285742732934332e-05, + "loss": 1.0552, + "step": 4610 + }, + { + "epoch": 0.15, + "learning_rate": 1.9285358248068468e-05, + "loss": 1.0854, + "step": 4611 + }, + { + "epoch": 0.15, + "learning_rate": 1.928497366358098e-05, + "loss": 1.0337, + "step": 4612 + }, + { + "epoch": 0.15, + "learning_rate": 1.9284588979476004e-05, + "loss": 1.2959, + "step": 4613 + }, + { + "epoch": 0.15, + "learning_rate": 1.9284204195757664e-05, + "loss": 1.1284, + "step": 4614 + }, + { + "epoch": 0.15, + "learning_rate": 1.928381931243009e-05, + "loss": 1.1021, + "step": 4615 + }, + { + "epoch": 0.15, + "learning_rate": 1.9283434329497408e-05, + "loss": 1.1436, + "step": 4616 + }, + { + "epoch": 0.15, + "learning_rate": 1.928304924696375e-05, + "loss": 1.0859, + "step": 4617 + }, + { + "epoch": 0.15, + "learning_rate": 1.9282664064833246e-05, + "loss": 1.1006, + "step": 4618 + }, + { + "epoch": 0.15, + "learning_rate": 1.9282278783110034e-05, + "loss": 1.0435, + "step": 4619 + }, + { + "epoch": 0.15, + "learning_rate": 1.928189340179824e-05, + "loss": 1.1304, + "step": 4620 + }, + { + "epoch": 0.15, + "learning_rate": 1.9281507920902004e-05, + "loss": 1.1104, + "step": 4621 + }, + { + "epoch": 0.15, + "learning_rate": 1.928112234042546e-05, + "loss": 1.1753, + "step": 4622 + }, + { + "epoch": 0.15, + "learning_rate": 1.928073666037275e-05, + "loss": 1.248, + "step": 4623 + }, + { + "epoch": 0.15, + "learning_rate": 1.9280350880748e-05, + "loss": 1.0908, + "step": 4624 + }, + { + "epoch": 0.15, + "learning_rate": 1.9279965001555355e-05, + "loss": 1.0898, + "step": 4625 + }, + { + "epoch": 0.15, + "learning_rate": 1.927957902279896e-05, + "loss": 1.0571, + "step": 4626 + }, + { + "epoch": 0.15, + "learning_rate": 1.9279192944482954e-05, + "loss": 1.168, + "step": 4627 + }, + { + "epoch": 0.15, + "learning_rate": 1.9278806766611474e-05, + "loss": 1.0723, + "step": 4628 + }, + { + "epoch": 0.15, + "learning_rate": 1.927842048918867e-05, + "loss": 1.1162, + "step": 4629 + }, + { + "epoch": 0.15, + "learning_rate": 1.9278034112218676e-05, + "loss": 1.1602, + "step": 4630 + }, + { + "epoch": 0.15, + "learning_rate": 1.9277647635705652e-05, + "loss": 1.043, + "step": 4631 + }, + { + "epoch": 0.15, + "learning_rate": 1.9277261059653734e-05, + "loss": 1.0908, + "step": 4632 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276874384067073e-05, + "loss": 1.0381, + "step": 4633 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276487608949816e-05, + "loss": 1.1577, + "step": 4634 + }, + { + "epoch": 0.15, + "learning_rate": 1.9276100734306116e-05, + "loss": 1.0767, + "step": 4635 + }, + { + "epoch": 0.15, + "learning_rate": 1.9275713760140118e-05, + "loss": 1.0679, + "step": 4636 + }, + { + "epoch": 0.15, + "learning_rate": 1.927532668645598e-05, + "loss": 1.0728, + "step": 4637 + }, + { + "epoch": 0.15, + "learning_rate": 1.927493951325785e-05, + "loss": 1.0532, + "step": 4638 + }, + { + "epoch": 0.15, + "learning_rate": 1.9274552240549885e-05, + "loss": 0.9893, + "step": 4639 + }, + { + "epoch": 0.15, + "learning_rate": 1.9274164868336237e-05, + "loss": 1.0234, + "step": 4640 + }, + { + "epoch": 0.15, + "learning_rate": 1.927377739662107e-05, + "loss": 0.574, + "step": 4641 + }, + { + "epoch": 0.15, + "learning_rate": 1.927338982540853e-05, + "loss": 1.1377, + "step": 4642 + }, + { + "epoch": 0.15, + "learning_rate": 1.927300215470278e-05, + "loss": 1.0137, + "step": 4643 + }, + { + "epoch": 0.15, + "learning_rate": 1.927261438450798e-05, + "loss": 1.1621, + "step": 4644 + }, + { + "epoch": 0.15, + "learning_rate": 1.927222651482829e-05, + "loss": 1.0405, + "step": 4645 + }, + { + "epoch": 0.15, + "learning_rate": 1.9271838545667876e-05, + "loss": 1.1777, + "step": 4646 + }, + { + "epoch": 0.15, + "learning_rate": 1.9271450477030894e-05, + "loss": 1.1631, + "step": 4647 + }, + { + "epoch": 0.15, + "learning_rate": 1.9271062308921507e-05, + "loss": 1.0972, + "step": 4648 + }, + { + "epoch": 0.15, + "learning_rate": 1.9270674041343885e-05, + "loss": 1.1279, + "step": 4649 + }, + { + "epoch": 0.15, + "learning_rate": 1.927028567430219e-05, + "loss": 1.0312, + "step": 4650 + }, + { + "epoch": 0.15, + "learning_rate": 1.926989720780059e-05, + "loss": 1.0859, + "step": 4651 + }, + { + "epoch": 0.15, + "learning_rate": 1.926950864184325e-05, + "loss": 1.0674, + "step": 4652 + }, + { + "epoch": 0.15, + "learning_rate": 1.926911997643434e-05, + "loss": 1.1099, + "step": 4653 + }, + { + "epoch": 0.15, + "learning_rate": 1.9268731211578034e-05, + "loss": 1.1597, + "step": 4654 + }, + { + "epoch": 0.15, + "learning_rate": 1.9268342347278502e-05, + "loss": 1.019, + "step": 4655 + }, + { + "epoch": 0.15, + "learning_rate": 1.926795338353991e-05, + "loss": 1.0928, + "step": 4656 + }, + { + "epoch": 0.15, + "learning_rate": 1.9267564320366438e-05, + "loss": 1.1279, + "step": 4657 + }, + { + "epoch": 0.15, + "learning_rate": 1.926717515776226e-05, + "loss": 1.0825, + "step": 4658 + }, + { + "epoch": 0.15, + "learning_rate": 1.9266785895731543e-05, + "loss": 1.1128, + "step": 4659 + }, + { + "epoch": 0.15, + "learning_rate": 1.9266396534278474e-05, + "loss": 1.1289, + "step": 4660 + }, + { + "epoch": 0.15, + "learning_rate": 1.9266007073407223e-05, + "loss": 1.2451, + "step": 4661 + }, + { + "epoch": 0.15, + "learning_rate": 1.9265617513121975e-05, + "loss": 0.5164, + "step": 4662 + }, + { + "epoch": 0.15, + "learning_rate": 1.92652278534269e-05, + "loss": 1.1802, + "step": 4663 + }, + { + "epoch": 0.15, + "learning_rate": 1.926483809432619e-05, + "loss": 0.9634, + "step": 4664 + }, + { + "epoch": 0.15, + "learning_rate": 1.926444823582402e-05, + "loss": 1.1104, + "step": 4665 + }, + { + "epoch": 0.15, + "learning_rate": 1.9264058277924574e-05, + "loss": 1.1582, + "step": 4666 + }, + { + "epoch": 0.15, + "learning_rate": 1.926366822063203e-05, + "loss": 1.0356, + "step": 4667 + }, + { + "epoch": 0.15, + "learning_rate": 1.9263278063950587e-05, + "loss": 0.9932, + "step": 4668 + }, + { + "epoch": 0.15, + "learning_rate": 1.9262887807884415e-05, + "loss": 1.1631, + "step": 4669 + }, + { + "epoch": 0.15, + "learning_rate": 1.926249745243771e-05, + "loss": 1.1548, + "step": 4670 + }, + { + "epoch": 0.15, + "learning_rate": 1.9262106997614663e-05, + "loss": 1.0684, + "step": 4671 + }, + { + "epoch": 0.15, + "learning_rate": 1.9261716443419457e-05, + "loss": 1.1812, + "step": 4672 + }, + { + "epoch": 0.15, + "learning_rate": 1.926132578985628e-05, + "loss": 1.0679, + "step": 4673 + }, + { + "epoch": 0.15, + "learning_rate": 1.926093503692933e-05, + "loss": 1.0889, + "step": 4674 + }, + { + "epoch": 0.15, + "learning_rate": 1.92605441846428e-05, + "loss": 1.1572, + "step": 4675 + }, + { + "epoch": 0.15, + "learning_rate": 1.9260153233000875e-05, + "loss": 1.1865, + "step": 4676 + }, + { + "epoch": 0.15, + "learning_rate": 1.9259762182007757e-05, + "loss": 1.0298, + "step": 4677 + }, + { + "epoch": 0.15, + "learning_rate": 1.9259371031667636e-05, + "loss": 1.0957, + "step": 4678 + }, + { + "epoch": 0.15, + "learning_rate": 1.9258979781984716e-05, + "loss": 1.0981, + "step": 4679 + }, + { + "epoch": 0.15, + "learning_rate": 1.9258588432963184e-05, + "loss": 1.061, + "step": 4680 + }, + { + "epoch": 0.15, + "learning_rate": 1.9258196984607253e-05, + "loss": 1.1284, + "step": 4681 + }, + { + "epoch": 0.15, + "learning_rate": 1.925780543692111e-05, + "loss": 1.1855, + "step": 4682 + }, + { + "epoch": 0.15, + "learning_rate": 1.9257413789908962e-05, + "loss": 1.1479, + "step": 4683 + }, + { + "epoch": 0.15, + "learning_rate": 1.9257022043575006e-05, + "loss": 1.2144, + "step": 4684 + }, + { + "epoch": 0.15, + "learning_rate": 1.9256630197923455e-05, + "loss": 1.1543, + "step": 4685 + }, + { + "epoch": 0.15, + "learning_rate": 1.9256238252958503e-05, + "loss": 1.1279, + "step": 4686 + }, + { + "epoch": 0.15, + "learning_rate": 1.925584620868436e-05, + "loss": 1.0825, + "step": 4687 + }, + { + "epoch": 0.15, + "learning_rate": 1.925545406510523e-05, + "loss": 0.9995, + "step": 4688 + }, + { + "epoch": 0.15, + "learning_rate": 1.9255061822225322e-05, + "loss": 1.1768, + "step": 4689 + }, + { + "epoch": 0.15, + "learning_rate": 1.925466948004884e-05, + "loss": 1.0972, + "step": 4690 + }, + { + "epoch": 0.15, + "learning_rate": 1.925427703858e-05, + "loss": 1.1357, + "step": 4691 + }, + { + "epoch": 0.15, + "learning_rate": 1.9253884497823013e-05, + "loss": 1.0938, + "step": 4692 + }, + { + "epoch": 0.15, + "learning_rate": 1.9253491857782083e-05, + "loss": 1.1499, + "step": 4693 + }, + { + "epoch": 0.15, + "learning_rate": 1.9253099118461428e-05, + "loss": 1.0493, + "step": 4694 + }, + { + "epoch": 0.15, + "learning_rate": 1.925270627986526e-05, + "loss": 1.0332, + "step": 4695 + }, + { + "epoch": 0.15, + "learning_rate": 1.9252313341997793e-05, + "loss": 1.1123, + "step": 4696 + }, + { + "epoch": 0.15, + "learning_rate": 1.9251920304863244e-05, + "loss": 1.0317, + "step": 4697 + }, + { + "epoch": 0.15, + "learning_rate": 1.9251527168465828e-05, + "loss": 1.0444, + "step": 4698 + }, + { + "epoch": 0.15, + "learning_rate": 1.9251133932809768e-05, + "loss": 1.0547, + "step": 4699 + }, + { + "epoch": 0.15, + "learning_rate": 1.9250740597899278e-05, + "loss": 1.1348, + "step": 4700 + }, + { + "epoch": 0.15, + "learning_rate": 1.925034716373858e-05, + "loss": 1.1426, + "step": 4701 + }, + { + "epoch": 0.15, + "learning_rate": 1.9249953630331896e-05, + "loss": 1.0444, + "step": 4702 + }, + { + "epoch": 0.15, + "learning_rate": 1.9249559997683446e-05, + "loss": 1.0957, + "step": 4703 + }, + { + "epoch": 0.15, + "learning_rate": 1.924916626579745e-05, + "loss": 1.124, + "step": 4704 + }, + { + "epoch": 0.15, + "learning_rate": 1.9248772434678143e-05, + "loss": 1.1714, + "step": 4705 + }, + { + "epoch": 0.15, + "learning_rate": 1.9248378504329742e-05, + "loss": 1.0332, + "step": 4706 + }, + { + "epoch": 0.15, + "learning_rate": 1.9247984474756474e-05, + "loss": 1.1152, + "step": 4707 + }, + { + "epoch": 0.15, + "learning_rate": 1.9247590345962565e-05, + "loss": 1.1846, + "step": 4708 + }, + { + "epoch": 0.15, + "learning_rate": 1.924719611795225e-05, + "loss": 1.0962, + "step": 4709 + }, + { + "epoch": 0.15, + "learning_rate": 1.9246801790729755e-05, + "loss": 0.552, + "step": 4710 + }, + { + "epoch": 0.15, + "learning_rate": 1.924640736429931e-05, + "loss": 0.9839, + "step": 4711 + }, + { + "epoch": 0.15, + "learning_rate": 1.9246012838665147e-05, + "loss": 1.1685, + "step": 4712 + }, + { + "epoch": 0.15, + "learning_rate": 1.92456182138315e-05, + "loss": 1.1128, + "step": 4713 + }, + { + "epoch": 0.15, + "learning_rate": 1.9245223489802602e-05, + "loss": 1.1152, + "step": 4714 + }, + { + "epoch": 0.15, + "learning_rate": 1.9244828666582687e-05, + "loss": 1.0986, + "step": 4715 + }, + { + "epoch": 0.15, + "learning_rate": 1.9244433744175993e-05, + "loss": 1.125, + "step": 4716 + }, + { + "epoch": 0.15, + "learning_rate": 1.9244038722586756e-05, + "loss": 1.1587, + "step": 4717 + }, + { + "epoch": 0.15, + "learning_rate": 1.9243643601819214e-05, + "loss": 1.1631, + "step": 4718 + }, + { + "epoch": 0.15, + "learning_rate": 1.9243248381877605e-05, + "loss": 1.0796, + "step": 4719 + }, + { + "epoch": 0.15, + "learning_rate": 1.9242853062766175e-05, + "loss": 1.0459, + "step": 4720 + }, + { + "epoch": 0.15, + "learning_rate": 1.9242457644489155e-05, + "loss": 0.9478, + "step": 4721 + }, + { + "epoch": 0.15, + "learning_rate": 1.9242062127050798e-05, + "loss": 0.9575, + "step": 4722 + }, + { + "epoch": 0.15, + "learning_rate": 1.9241666510455342e-05, + "loss": 1.2056, + "step": 4723 + }, + { + "epoch": 0.15, + "learning_rate": 1.9241270794707036e-05, + "loss": 1.1099, + "step": 4724 + }, + { + "epoch": 0.15, + "learning_rate": 1.9240874979810115e-05, + "loss": 1.0762, + "step": 4725 + }, + { + "epoch": 0.15, + "learning_rate": 1.9240479065768838e-05, + "loss": 1.0718, + "step": 4726 + }, + { + "epoch": 0.15, + "learning_rate": 1.9240083052587443e-05, + "loss": 0.9951, + "step": 4727 + }, + { + "epoch": 0.15, + "learning_rate": 1.9239686940270187e-05, + "loss": 1.0527, + "step": 4728 + }, + { + "epoch": 0.15, + "learning_rate": 1.9239290728821317e-05, + "loss": 1.0322, + "step": 4729 + }, + { + "epoch": 0.15, + "learning_rate": 1.9238894418245083e-05, + "loss": 1.0796, + "step": 4730 + }, + { + "epoch": 0.15, + "learning_rate": 1.9238498008545732e-05, + "loss": 1.1318, + "step": 4731 + }, + { + "epoch": 0.15, + "learning_rate": 1.9238101499727525e-05, + "loss": 1.1782, + "step": 4732 + }, + { + "epoch": 0.15, + "learning_rate": 1.9237704891794716e-05, + "loss": 1.2031, + "step": 4733 + }, + { + "epoch": 0.15, + "learning_rate": 1.923730818475155e-05, + "loss": 1.1689, + "step": 4734 + }, + { + "epoch": 0.15, + "learning_rate": 1.92369113786023e-05, + "loss": 0.9102, + "step": 4735 + }, + { + "epoch": 0.15, + "learning_rate": 1.9236514473351206e-05, + "loss": 1.1099, + "step": 4736 + }, + { + "epoch": 0.15, + "learning_rate": 1.923611746900254e-05, + "loss": 0.9814, + "step": 4737 + }, + { + "epoch": 0.15, + "learning_rate": 1.9235720365560552e-05, + "loss": 1.1431, + "step": 4738 + }, + { + "epoch": 0.15, + "learning_rate": 1.9235323163029507e-05, + "loss": 1.1318, + "step": 4739 + }, + { + "epoch": 0.15, + "learning_rate": 1.9234925861413665e-05, + "loss": 1.0679, + "step": 4740 + }, + { + "epoch": 0.15, + "learning_rate": 1.923452846071729e-05, + "loss": 1.1387, + "step": 4741 + }, + { + "epoch": 0.15, + "learning_rate": 1.9234130960944643e-05, + "loss": 1.043, + "step": 4742 + }, + { + "epoch": 0.15, + "learning_rate": 1.9233733362099994e-05, + "loss": 1.1699, + "step": 4743 + }, + { + "epoch": 0.15, + "learning_rate": 1.92333356641876e-05, + "loss": 1.0703, + "step": 4744 + }, + { + "epoch": 0.15, + "learning_rate": 1.923293786721174e-05, + "loss": 1.1152, + "step": 4745 + }, + { + "epoch": 0.15, + "learning_rate": 1.923253997117667e-05, + "loss": 1.1646, + "step": 4746 + }, + { + "epoch": 0.15, + "learning_rate": 1.9232141976086667e-05, + "loss": 1.0483, + "step": 4747 + }, + { + "epoch": 0.15, + "learning_rate": 1.9231743881946e-05, + "loss": 1.0547, + "step": 4748 + }, + { + "epoch": 0.15, + "learning_rate": 1.9231345688758932e-05, + "loss": 1.1782, + "step": 4749 + }, + { + "epoch": 0.15, + "learning_rate": 1.9230947396529747e-05, + "loss": 1.1309, + "step": 4750 + }, + { + "epoch": 0.15, + "learning_rate": 1.923054900526271e-05, + "loss": 1.209, + "step": 4751 + }, + { + "epoch": 0.15, + "learning_rate": 1.92301505149621e-05, + "loss": 1.0039, + "step": 4752 + }, + { + "epoch": 0.15, + "learning_rate": 1.922975192563219e-05, + "loss": 1.2041, + "step": 4753 + }, + { + "epoch": 0.15, + "learning_rate": 1.9229353237277256e-05, + "loss": 1.1235, + "step": 4754 + }, + { + "epoch": 0.15, + "learning_rate": 1.9228954449901576e-05, + "loss": 1.1597, + "step": 4755 + }, + { + "epoch": 0.15, + "learning_rate": 1.9228555563509432e-05, + "loss": 1.1836, + "step": 4756 + }, + { + "epoch": 0.15, + "learning_rate": 1.92281565781051e-05, + "loss": 0.9761, + "step": 4757 + }, + { + "epoch": 0.15, + "learning_rate": 1.9227757493692857e-05, + "loss": 1.1147, + "step": 4758 + }, + { + "epoch": 0.15, + "learning_rate": 1.922735831027699e-05, + "loss": 1.1709, + "step": 4759 + }, + { + "epoch": 0.15, + "learning_rate": 1.922695902786178e-05, + "loss": 0.4824, + "step": 4760 + }, + { + "epoch": 0.15, + "learning_rate": 1.9226559646451515e-05, + "loss": 1.0068, + "step": 4761 + }, + { + "epoch": 0.15, + "learning_rate": 1.9226160166050475e-05, + "loss": 1.0693, + "step": 4762 + }, + { + "epoch": 0.15, + "learning_rate": 1.9225760586662946e-05, + "loss": 1.1191, + "step": 4763 + }, + { + "epoch": 0.15, + "learning_rate": 1.9225360908293217e-05, + "loss": 1.0835, + "step": 4764 + }, + { + "epoch": 0.15, + "learning_rate": 1.9224961130945578e-05, + "loss": 1.2432, + "step": 4765 + }, + { + "epoch": 0.15, + "learning_rate": 1.9224561254624313e-05, + "loss": 1.1108, + "step": 4766 + }, + { + "epoch": 0.15, + "learning_rate": 1.9224161279333714e-05, + "loss": 1.0532, + "step": 4767 + }, + { + "epoch": 0.15, + "learning_rate": 1.9223761205078078e-05, + "loss": 0.5059, + "step": 4768 + }, + { + "epoch": 0.15, + "learning_rate": 1.9223361031861686e-05, + "loss": 1.1484, + "step": 4769 + }, + { + "epoch": 0.15, + "learning_rate": 1.9222960759688844e-05, + "loss": 1.0684, + "step": 4770 + }, + { + "epoch": 0.15, + "learning_rate": 1.9222560388563837e-05, + "loss": 1.167, + "step": 4771 + }, + { + "epoch": 0.15, + "learning_rate": 1.922215991849097e-05, + "loss": 1.0732, + "step": 4772 + }, + { + "epoch": 0.15, + "learning_rate": 1.9221759349474526e-05, + "loss": 1.1753, + "step": 4773 + }, + { + "epoch": 0.15, + "learning_rate": 1.922135868151881e-05, + "loss": 1.041, + "step": 4774 + }, + { + "epoch": 0.15, + "learning_rate": 1.9220957914628128e-05, + "loss": 1.1143, + "step": 4775 + }, + { + "epoch": 0.15, + "learning_rate": 1.922055704880677e-05, + "loss": 1.0737, + "step": 4776 + }, + { + "epoch": 0.15, + "learning_rate": 1.9220156084059037e-05, + "loss": 1.0542, + "step": 4777 + }, + { + "epoch": 0.15, + "learning_rate": 1.9219755020389234e-05, + "loss": 1.0742, + "step": 4778 + }, + { + "epoch": 0.15, + "learning_rate": 1.9219353857801666e-05, + "loss": 1.1528, + "step": 4779 + }, + { + "epoch": 0.15, + "learning_rate": 1.9218952596300636e-05, + "loss": 1.168, + "step": 4780 + }, + { + "epoch": 0.15, + "learning_rate": 1.9218551235890445e-05, + "loss": 1.0566, + "step": 4781 + }, + { + "epoch": 0.15, + "learning_rate": 1.92181497765754e-05, + "loss": 1.0981, + "step": 4782 + }, + { + "epoch": 0.15, + "learning_rate": 1.9217748218359814e-05, + "loss": 1.1147, + "step": 4783 + }, + { + "epoch": 0.15, + "learning_rate": 1.9217346561247987e-05, + "loss": 1.0859, + "step": 4784 + }, + { + "epoch": 0.15, + "learning_rate": 1.9216944805244234e-05, + "loss": 1.0264, + "step": 4785 + }, + { + "epoch": 0.15, + "learning_rate": 1.9216542950352867e-05, + "loss": 1.064, + "step": 4786 + }, + { + "epoch": 0.15, + "learning_rate": 1.9216140996578193e-05, + "loss": 1.1138, + "step": 4787 + }, + { + "epoch": 0.15, + "learning_rate": 1.9215738943924525e-05, + "loss": 1.0967, + "step": 4788 + }, + { + "epoch": 0.15, + "learning_rate": 1.921533679239618e-05, + "loss": 1.0596, + "step": 4789 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214934541997468e-05, + "loss": 1.0454, + "step": 4790 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214532192732706e-05, + "loss": 1.1323, + "step": 4791 + }, + { + "epoch": 0.15, + "learning_rate": 1.9214129744606212e-05, + "loss": 1.1206, + "step": 4792 + }, + { + "epoch": 0.15, + "learning_rate": 1.9213727197622306e-05, + "loss": 1.1157, + "step": 4793 + }, + { + "epoch": 0.15, + "learning_rate": 1.9213324551785302e-05, + "loss": 1.1157, + "step": 4794 + }, + { + "epoch": 0.15, + "learning_rate": 1.9212921807099522e-05, + "loss": 1.1353, + "step": 4795 + }, + { + "epoch": 0.15, + "learning_rate": 1.921251896356929e-05, + "loss": 1.1753, + "step": 4796 + }, + { + "epoch": 0.15, + "learning_rate": 1.9212116021198923e-05, + "loss": 0.9976, + "step": 4797 + }, + { + "epoch": 0.15, + "learning_rate": 1.921171297999275e-05, + "loss": 1.1143, + "step": 4798 + }, + { + "epoch": 0.15, + "learning_rate": 1.921130983995509e-05, + "loss": 1.0986, + "step": 4799 + }, + { + "epoch": 0.15, + "learning_rate": 1.921090660109027e-05, + "loss": 1.0537, + "step": 4800 + }, + { + "epoch": 0.15, + "learning_rate": 1.9210503263402615e-05, + "loss": 1.0591, + "step": 4801 + }, + { + "epoch": 0.15, + "learning_rate": 1.9210099826896455e-05, + "loss": 1.0161, + "step": 4802 + }, + { + "epoch": 0.15, + "learning_rate": 1.9209696291576116e-05, + "loss": 1.0952, + "step": 4803 + }, + { + "epoch": 0.15, + "learning_rate": 1.9209292657445927e-05, + "loss": 0.9868, + "step": 4804 + }, + { + "epoch": 0.15, + "learning_rate": 1.9208888924510223e-05, + "loss": 1.1416, + "step": 4805 + }, + { + "epoch": 0.15, + "learning_rate": 1.9208485092773333e-05, + "loss": 1.1699, + "step": 4806 + }, + { + "epoch": 0.15, + "learning_rate": 1.9208081162239588e-05, + "loss": 1.0718, + "step": 4807 + }, + { + "epoch": 0.15, + "learning_rate": 1.9207677132913326e-05, + "loss": 0.9941, + "step": 4808 + }, + { + "epoch": 0.15, + "learning_rate": 1.9207273004798873e-05, + "loss": 1.0737, + "step": 4809 + }, + { + "epoch": 0.15, + "learning_rate": 1.9206868777900574e-05, + "loss": 1.2217, + "step": 4810 + }, + { + "epoch": 0.15, + "learning_rate": 1.9206464452222767e-05, + "loss": 1.105, + "step": 4811 + }, + { + "epoch": 0.15, + "learning_rate": 1.9206060027769778e-05, + "loss": 1.1191, + "step": 4812 + }, + { + "epoch": 0.15, + "learning_rate": 1.9205655504545958e-05, + "loss": 1.1411, + "step": 4813 + }, + { + "epoch": 0.15, + "learning_rate": 1.9205250882555643e-05, + "loss": 1.0859, + "step": 4814 + }, + { + "epoch": 0.15, + "learning_rate": 1.9204846161803173e-05, + "loss": 1.0894, + "step": 4815 + }, + { + "epoch": 0.15, + "learning_rate": 1.920444134229289e-05, + "loss": 0.9468, + "step": 4816 + }, + { + "epoch": 0.15, + "learning_rate": 1.920403642402914e-05, + "loss": 1.0547, + "step": 4817 + }, + { + "epoch": 0.15, + "learning_rate": 1.9203631407016267e-05, + "loss": 1.0391, + "step": 4818 + }, + { + "epoch": 0.15, + "learning_rate": 1.920322629125861e-05, + "loss": 0.8779, + "step": 4819 + }, + { + "epoch": 0.15, + "learning_rate": 1.9202821076760524e-05, + "loss": 1.0762, + "step": 4820 + }, + { + "epoch": 0.15, + "learning_rate": 1.9202415763526353e-05, + "loss": 1.0356, + "step": 4821 + }, + { + "epoch": 0.15, + "learning_rate": 1.9202010351560446e-05, + "loss": 1.0215, + "step": 4822 + }, + { + "epoch": 0.15, + "learning_rate": 1.920160484086715e-05, + "loss": 1.0503, + "step": 4823 + }, + { + "epoch": 0.15, + "learning_rate": 1.920119923145082e-05, + "loss": 1.0962, + "step": 4824 + }, + { + "epoch": 0.15, + "learning_rate": 1.9200793523315805e-05, + "loss": 1.0723, + "step": 4825 + }, + { + "epoch": 0.15, + "learning_rate": 1.920038771646646e-05, + "loss": 1.0781, + "step": 4826 + }, + { + "epoch": 0.15, + "learning_rate": 1.9199981810907135e-05, + "loss": 1.2007, + "step": 4827 + }, + { + "epoch": 0.15, + "learning_rate": 1.9199575806642188e-05, + "loss": 0.9668, + "step": 4828 + }, + { + "epoch": 0.15, + "learning_rate": 1.9199169703675974e-05, + "loss": 1.0308, + "step": 4829 + }, + { + "epoch": 0.15, + "learning_rate": 1.919876350201285e-05, + "loss": 0.9673, + "step": 4830 + }, + { + "epoch": 0.15, + "learning_rate": 1.9198357201657174e-05, + "loss": 1.1147, + "step": 4831 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197950802613304e-05, + "loss": 1.0371, + "step": 4832 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197544304885604e-05, + "loss": 1.0806, + "step": 4833 + }, + { + "epoch": 0.15, + "learning_rate": 1.9197137708478433e-05, + "loss": 1.1929, + "step": 4834 + }, + { + "epoch": 0.15, + "learning_rate": 1.9196731013396155e-05, + "loss": 0.5078, + "step": 4835 + }, + { + "epoch": 0.15, + "learning_rate": 1.9196324219643128e-05, + "loss": 1.0371, + "step": 4836 + }, + { + "epoch": 0.15, + "learning_rate": 1.9195917327223722e-05, + "loss": 1.188, + "step": 4837 + }, + { + "epoch": 0.15, + "learning_rate": 1.91955103361423e-05, + "loss": 1.2275, + "step": 4838 + }, + { + "epoch": 0.15, + "learning_rate": 1.919510324640323e-05, + "loss": 1.0303, + "step": 4839 + }, + { + "epoch": 0.15, + "learning_rate": 1.9194696058010878e-05, + "loss": 1.3071, + "step": 4840 + }, + { + "epoch": 0.15, + "learning_rate": 1.9194288770969613e-05, + "loss": 1.1104, + "step": 4841 + }, + { + "epoch": 0.15, + "learning_rate": 1.9193881385283806e-05, + "loss": 1.1113, + "step": 4842 + }, + { + "epoch": 0.15, + "learning_rate": 1.9193473900957827e-05, + "loss": 1.0312, + "step": 4843 + }, + { + "epoch": 0.15, + "learning_rate": 1.919306631799605e-05, + "loss": 1.0444, + "step": 4844 + }, + { + "epoch": 0.15, + "learning_rate": 1.919265863640284e-05, + "loss": 1.0161, + "step": 4845 + }, + { + "epoch": 0.15, + "learning_rate": 1.9192250856182583e-05, + "loss": 0.9673, + "step": 4846 + }, + { + "epoch": 0.16, + "learning_rate": 1.9191842977339645e-05, + "loss": 1.231, + "step": 4847 + }, + { + "epoch": 0.16, + "learning_rate": 1.9191434999878403e-05, + "loss": 1.0996, + "step": 4848 + }, + { + "epoch": 0.16, + "learning_rate": 1.919102692380324e-05, + "loss": 1.021, + "step": 4849 + }, + { + "epoch": 0.16, + "learning_rate": 1.919061874911853e-05, + "loss": 0.894, + "step": 4850 + }, + { + "epoch": 0.16, + "learning_rate": 1.9190210475828648e-05, + "loss": 1.168, + "step": 4851 + }, + { + "epoch": 0.16, + "learning_rate": 1.9189802103937986e-05, + "loss": 1.0337, + "step": 4852 + }, + { + "epoch": 0.16, + "learning_rate": 1.918939363345091e-05, + "loss": 1.1772, + "step": 4853 + }, + { + "epoch": 0.16, + "learning_rate": 1.9188985064371818e-05, + "loss": 1.1636, + "step": 4854 + }, + { + "epoch": 0.16, + "learning_rate": 1.9188576396705078e-05, + "loss": 1.0459, + "step": 4855 + }, + { + "epoch": 0.16, + "learning_rate": 1.918816763045509e-05, + "loss": 1.1777, + "step": 4856 + }, + { + "epoch": 0.16, + "learning_rate": 1.9187758765626228e-05, + "loss": 1.0674, + "step": 4857 + }, + { + "epoch": 0.16, + "learning_rate": 1.9187349802222884e-05, + "loss": 1.2007, + "step": 4858 + }, + { + "epoch": 0.16, + "learning_rate": 1.9186940740249445e-05, + "loss": 1.0571, + "step": 4859 + }, + { + "epoch": 0.16, + "learning_rate": 1.9186531579710303e-05, + "loss": 1.1353, + "step": 4860 + }, + { + "epoch": 0.16, + "learning_rate": 1.918612232060984e-05, + "loss": 1.1724, + "step": 4861 + }, + { + "epoch": 0.16, + "learning_rate": 1.9185712962952453e-05, + "loss": 1.064, + "step": 4862 + }, + { + "epoch": 0.16, + "learning_rate": 1.9185303506742528e-05, + "loss": 1.166, + "step": 4863 + }, + { + "epoch": 0.16, + "learning_rate": 1.9184893951984467e-05, + "loss": 1.127, + "step": 4864 + }, + { + "epoch": 0.16, + "learning_rate": 1.9184484298682657e-05, + "loss": 1.146, + "step": 4865 + }, + { + "epoch": 0.16, + "learning_rate": 1.9184074546841496e-05, + "loss": 1.1885, + "step": 4866 + }, + { + "epoch": 0.16, + "learning_rate": 1.918366469646538e-05, + "loss": 1.0811, + "step": 4867 + }, + { + "epoch": 0.16, + "learning_rate": 1.9183254747558706e-05, + "loss": 1.0811, + "step": 4868 + }, + { + "epoch": 0.16, + "learning_rate": 1.918284470012587e-05, + "loss": 1.1509, + "step": 4869 + }, + { + "epoch": 0.16, + "learning_rate": 1.9182434554171276e-05, + "loss": 1.2261, + "step": 4870 + }, + { + "epoch": 0.16, + "learning_rate": 1.9182024309699324e-05, + "loss": 0.9966, + "step": 4871 + }, + { + "epoch": 0.16, + "learning_rate": 1.9181613966714408e-05, + "loss": 1.0146, + "step": 4872 + }, + { + "epoch": 0.16, + "learning_rate": 1.918120352522094e-05, + "loss": 1.0723, + "step": 4873 + }, + { + "epoch": 0.16, + "learning_rate": 1.9180792985223316e-05, + "loss": 0.9556, + "step": 4874 + }, + { + "epoch": 0.16, + "learning_rate": 1.9180382346725945e-05, + "loss": 0.9878, + "step": 4875 + }, + { + "epoch": 0.16, + "learning_rate": 1.917997160973323e-05, + "loss": 1.0503, + "step": 4876 + }, + { + "epoch": 0.16, + "learning_rate": 1.9179560774249584e-05, + "loss": 1.0449, + "step": 4877 + }, + { + "epoch": 0.16, + "learning_rate": 1.9179149840279406e-05, + "loss": 0.999, + "step": 4878 + }, + { + "epoch": 0.16, + "learning_rate": 1.9178738807827113e-05, + "loss": 1.0605, + "step": 4879 + }, + { + "epoch": 0.16, + "learning_rate": 1.917832767689711e-05, + "loss": 1.103, + "step": 4880 + }, + { + "epoch": 0.16, + "learning_rate": 1.9177916447493807e-05, + "loss": 1.0479, + "step": 4881 + }, + { + "epoch": 0.16, + "learning_rate": 1.917750511962162e-05, + "loss": 1.0098, + "step": 4882 + }, + { + "epoch": 0.16, + "learning_rate": 1.9177093693284962e-05, + "loss": 1.085, + "step": 4883 + }, + { + "epoch": 0.16, + "learning_rate": 1.917668216848824e-05, + "loss": 1.1465, + "step": 4884 + }, + { + "epoch": 0.16, + "learning_rate": 1.917627054523588e-05, + "loss": 0.4724, + "step": 4885 + }, + { + "epoch": 0.16, + "learning_rate": 1.9175858823532288e-05, + "loss": 1.1528, + "step": 4886 + }, + { + "epoch": 0.16, + "learning_rate": 1.9175447003381892e-05, + "loss": 1.0537, + "step": 4887 + }, + { + "epoch": 0.16, + "learning_rate": 1.91750350847891e-05, + "loss": 1.0146, + "step": 4888 + }, + { + "epoch": 0.16, + "learning_rate": 1.9174623067758337e-05, + "loss": 1.1978, + "step": 4889 + }, + { + "epoch": 0.16, + "learning_rate": 1.9174210952294024e-05, + "loss": 1.1514, + "step": 4890 + }, + { + "epoch": 0.16, + "learning_rate": 1.9173798738400576e-05, + "loss": 1.1201, + "step": 4891 + }, + { + "epoch": 0.16, + "learning_rate": 1.9173386426082428e-05, + "loss": 1.0933, + "step": 4892 + }, + { + "epoch": 0.16, + "learning_rate": 1.917297401534399e-05, + "loss": 1.0513, + "step": 4893 + }, + { + "epoch": 0.16, + "learning_rate": 1.9172561506189695e-05, + "loss": 1.0415, + "step": 4894 + }, + { + "epoch": 0.16, + "learning_rate": 1.917214889862397e-05, + "loss": 1.0151, + "step": 4895 + }, + { + "epoch": 0.16, + "learning_rate": 1.9171736192651233e-05, + "loss": 1.0698, + "step": 4896 + }, + { + "epoch": 0.16, + "learning_rate": 1.9171323388275922e-05, + "loss": 1.0098, + "step": 4897 + }, + { + "epoch": 0.16, + "learning_rate": 1.9170910485502457e-05, + "loss": 1.02, + "step": 4898 + }, + { + "epoch": 0.16, + "learning_rate": 1.9170497484335276e-05, + "loss": 1.1973, + "step": 4899 + }, + { + "epoch": 0.16, + "learning_rate": 1.9170084384778805e-05, + "loss": 1.1079, + "step": 4900 + }, + { + "epoch": 0.16, + "learning_rate": 1.916967118683748e-05, + "loss": 1.0713, + "step": 4901 + }, + { + "epoch": 0.16, + "learning_rate": 1.9169257890515728e-05, + "loss": 1.208, + "step": 4902 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168844495817992e-05, + "loss": 1.0854, + "step": 4903 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168431002748696e-05, + "loss": 1.0762, + "step": 4904 + }, + { + "epoch": 0.16, + "learning_rate": 1.9168017411312284e-05, + "loss": 1.2607, + "step": 4905 + }, + { + "epoch": 0.16, + "learning_rate": 1.9167603721513192e-05, + "loss": 0.9854, + "step": 4906 + }, + { + "epoch": 0.16, + "learning_rate": 1.9167189933355862e-05, + "loss": 1.0942, + "step": 4907 + }, + { + "epoch": 0.16, + "learning_rate": 1.9166776046844725e-05, + "loss": 1.1143, + "step": 4908 + }, + { + "epoch": 0.16, + "learning_rate": 1.916636206198423e-05, + "loss": 1.2012, + "step": 4909 + }, + { + "epoch": 0.16, + "learning_rate": 1.9165947978778813e-05, + "loss": 0.5166, + "step": 4910 + }, + { + "epoch": 0.16, + "learning_rate": 1.9165533797232917e-05, + "loss": 1.0225, + "step": 4911 + }, + { + "epoch": 0.16, + "learning_rate": 1.916511951735099e-05, + "loss": 1.127, + "step": 4912 + }, + { + "epoch": 0.16, + "learning_rate": 1.916470513913747e-05, + "loss": 1.126, + "step": 4913 + }, + { + "epoch": 0.16, + "learning_rate": 1.9164290662596813e-05, + "loss": 1.0942, + "step": 4914 + }, + { + "epoch": 0.16, + "learning_rate": 1.9163876087733456e-05, + "loss": 1.0566, + "step": 4915 + }, + { + "epoch": 0.16, + "learning_rate": 1.916346141455185e-05, + "loss": 1.1353, + "step": 4916 + }, + { + "epoch": 0.16, + "learning_rate": 1.9163046643056448e-05, + "loss": 1.2056, + "step": 4917 + }, + { + "epoch": 0.16, + "learning_rate": 1.916263177325169e-05, + "loss": 1.1182, + "step": 4918 + }, + { + "epoch": 0.16, + "learning_rate": 1.916221680514204e-05, + "loss": 1.1313, + "step": 4919 + }, + { + "epoch": 0.16, + "learning_rate": 1.9161801738731944e-05, + "loss": 0.9688, + "step": 4920 + }, + { + "epoch": 0.16, + "learning_rate": 1.916138657402585e-05, + "loss": 1.1455, + "step": 4921 + }, + { + "epoch": 0.16, + "learning_rate": 1.9160971311028224e-05, + "loss": 1.103, + "step": 4922 + }, + { + "epoch": 0.16, + "learning_rate": 1.916055594974351e-05, + "loss": 1.0854, + "step": 4923 + }, + { + "epoch": 0.16, + "learning_rate": 1.916014049017617e-05, + "loss": 1.231, + "step": 4924 + }, + { + "epoch": 0.16, + "learning_rate": 1.915972493233066e-05, + "loss": 1.0552, + "step": 4925 + }, + { + "epoch": 0.16, + "learning_rate": 1.915930927621144e-05, + "loss": 1.1655, + "step": 4926 + }, + { + "epoch": 0.16, + "learning_rate": 1.915889352182297e-05, + "loss": 1.0381, + "step": 4927 + }, + { + "epoch": 0.16, + "learning_rate": 1.915847766916971e-05, + "loss": 1.1621, + "step": 4928 + }, + { + "epoch": 0.16, + "learning_rate": 1.9158061718256115e-05, + "loss": 1.0669, + "step": 4929 + }, + { + "epoch": 0.16, + "learning_rate": 1.9157645669086657e-05, + "loss": 1.2114, + "step": 4930 + }, + { + "epoch": 0.16, + "learning_rate": 1.9157229521665796e-05, + "loss": 1.1523, + "step": 4931 + }, + { + "epoch": 0.16, + "learning_rate": 1.9156813275997996e-05, + "loss": 1.0132, + "step": 4932 + }, + { + "epoch": 0.16, + "learning_rate": 1.9156396932087724e-05, + "loss": 1.1172, + "step": 4933 + }, + { + "epoch": 0.16, + "learning_rate": 1.915598048993945e-05, + "loss": 1.1104, + "step": 4934 + }, + { + "epoch": 0.16, + "learning_rate": 1.9155563949557634e-05, + "loss": 1.1587, + "step": 4935 + }, + { + "epoch": 0.16, + "learning_rate": 1.915514731094675e-05, + "loss": 1.1997, + "step": 4936 + }, + { + "epoch": 0.16, + "learning_rate": 1.915473057411127e-05, + "loss": 1.1553, + "step": 4937 + }, + { + "epoch": 0.16, + "learning_rate": 1.915431373905566e-05, + "loss": 1.1714, + "step": 4938 + }, + { + "epoch": 0.16, + "learning_rate": 1.9153896805784396e-05, + "loss": 1.1162, + "step": 4939 + }, + { + "epoch": 0.16, + "learning_rate": 1.9153479774301953e-05, + "loss": 1.2588, + "step": 4940 + }, + { + "epoch": 0.16, + "learning_rate": 1.9153062644612802e-05, + "loss": 1.0874, + "step": 4941 + }, + { + "epoch": 0.16, + "learning_rate": 1.915264541672142e-05, + "loss": 0.9771, + "step": 4942 + }, + { + "epoch": 0.16, + "learning_rate": 1.9152228090632277e-05, + "loss": 1.0679, + "step": 4943 + }, + { + "epoch": 0.16, + "learning_rate": 1.915181066634986e-05, + "loss": 1.0342, + "step": 4944 + }, + { + "epoch": 0.16, + "learning_rate": 1.915139314387864e-05, + "loss": 1.1953, + "step": 4945 + }, + { + "epoch": 0.16, + "learning_rate": 1.9150975523223105e-05, + "loss": 1.0669, + "step": 4946 + }, + { + "epoch": 0.16, + "learning_rate": 1.9150557804387727e-05, + "loss": 1.061, + "step": 4947 + }, + { + "epoch": 0.16, + "learning_rate": 1.9150139987376994e-05, + "loss": 1.0537, + "step": 4948 + }, + { + "epoch": 0.16, + "learning_rate": 1.9149722072195384e-05, + "loss": 1.0469, + "step": 4949 + }, + { + "epoch": 0.16, + "learning_rate": 1.914930405884738e-05, + "loss": 1.084, + "step": 4950 + }, + { + "epoch": 0.16, + "learning_rate": 1.9148885947337472e-05, + "loss": 1.0664, + "step": 4951 + }, + { + "epoch": 0.16, + "learning_rate": 1.9148467737670146e-05, + "loss": 1.1826, + "step": 4952 + }, + { + "epoch": 0.16, + "learning_rate": 1.9148049429849882e-05, + "loss": 1.0737, + "step": 4953 + }, + { + "epoch": 0.16, + "learning_rate": 1.9147631023881173e-05, + "loss": 1.1479, + "step": 4954 + }, + { + "epoch": 0.16, + "learning_rate": 1.914721251976851e-05, + "loss": 1.0093, + "step": 4955 + }, + { + "epoch": 0.16, + "learning_rate": 1.9146793917516377e-05, + "loss": 1.106, + "step": 4956 + }, + { + "epoch": 0.16, + "learning_rate": 1.9146375217129268e-05, + "loss": 1.1934, + "step": 4957 + }, + { + "epoch": 0.16, + "learning_rate": 1.9145956418611674e-05, + "loss": 1.1372, + "step": 4958 + }, + { + "epoch": 0.16, + "learning_rate": 1.9145537521968095e-05, + "loss": 1.0698, + "step": 4959 + }, + { + "epoch": 0.16, + "learning_rate": 1.9145118527203017e-05, + "loss": 1.1162, + "step": 4960 + }, + { + "epoch": 0.16, + "learning_rate": 1.914469943432094e-05, + "loss": 1.1299, + "step": 4961 + }, + { + "epoch": 0.16, + "learning_rate": 1.9144280243326356e-05, + "loss": 1.0391, + "step": 4962 + }, + { + "epoch": 0.16, + "learning_rate": 1.9143860954223766e-05, + "loss": 1.1304, + "step": 4963 + }, + { + "epoch": 0.16, + "learning_rate": 1.914344156701767e-05, + "loss": 1.1484, + "step": 4964 + }, + { + "epoch": 0.16, + "learning_rate": 1.9143022081712567e-05, + "loss": 1.0757, + "step": 4965 + }, + { + "epoch": 0.16, + "learning_rate": 1.9142602498312953e-05, + "loss": 1.1104, + "step": 4966 + }, + { + "epoch": 0.16, + "learning_rate": 1.9142182816823336e-05, + "loss": 0.9961, + "step": 4967 + }, + { + "epoch": 0.16, + "learning_rate": 1.9141763037248212e-05, + "loss": 1.0605, + "step": 4968 + }, + { + "epoch": 0.16, + "learning_rate": 1.914134315959209e-05, + "loss": 0.5015, + "step": 4969 + }, + { + "epoch": 0.16, + "learning_rate": 1.9140923183859473e-05, + "loss": 1.0713, + "step": 4970 + }, + { + "epoch": 0.16, + "learning_rate": 1.9140503110054864e-05, + "loss": 1.123, + "step": 4971 + }, + { + "epoch": 0.16, + "learning_rate": 1.9140082938182777e-05, + "loss": 1.1421, + "step": 4972 + }, + { + "epoch": 0.16, + "learning_rate": 1.9139662668247713e-05, + "loss": 1.0542, + "step": 4973 + }, + { + "epoch": 0.16, + "learning_rate": 1.9139242300254185e-05, + "loss": 1.1265, + "step": 4974 + }, + { + "epoch": 0.16, + "learning_rate": 1.91388218342067e-05, + "loss": 1.0532, + "step": 4975 + }, + { + "epoch": 0.16, + "learning_rate": 1.9138401270109773e-05, + "loss": 1.001, + "step": 4976 + }, + { + "epoch": 0.16, + "learning_rate": 1.9137980607967917e-05, + "loss": 0.5237, + "step": 4977 + }, + { + "epoch": 0.16, + "learning_rate": 1.9137559847785637e-05, + "loss": 1.1631, + "step": 4978 + }, + { + "epoch": 0.16, + "learning_rate": 1.9137138989567458e-05, + "loss": 1.0654, + "step": 4979 + }, + { + "epoch": 0.16, + "learning_rate": 1.9136718033317887e-05, + "loss": 1.0205, + "step": 4980 + }, + { + "epoch": 0.16, + "learning_rate": 1.913629697904144e-05, + "loss": 1.2163, + "step": 4981 + }, + { + "epoch": 0.16, + "learning_rate": 1.9135875826742645e-05, + "loss": 1.0479, + "step": 4982 + }, + { + "epoch": 0.16, + "learning_rate": 1.913545457642601e-05, + "loss": 1.1011, + "step": 4983 + }, + { + "epoch": 0.16, + "learning_rate": 1.9135033228096058e-05, + "loss": 0.9888, + "step": 4984 + }, + { + "epoch": 0.16, + "learning_rate": 1.913461178175731e-05, + "loss": 1.2275, + "step": 4985 + }, + { + "epoch": 0.16, + "learning_rate": 1.913419023741429e-05, + "loss": 1.0273, + "step": 4986 + }, + { + "epoch": 0.16, + "learning_rate": 1.9133768595071513e-05, + "loss": 1.0264, + "step": 4987 + }, + { + "epoch": 0.16, + "learning_rate": 1.913334685473351e-05, + "loss": 1.0386, + "step": 4988 + }, + { + "epoch": 0.16, + "learning_rate": 1.9132925016404805e-05, + "loss": 0.9941, + "step": 4989 + }, + { + "epoch": 0.16, + "learning_rate": 1.9132503080089918e-05, + "loss": 1.1875, + "step": 4990 + }, + { + "epoch": 0.16, + "learning_rate": 1.9132081045793387e-05, + "loss": 1.0903, + "step": 4991 + }, + { + "epoch": 0.16, + "learning_rate": 1.9131658913519728e-05, + "loss": 1.0435, + "step": 4992 + }, + { + "epoch": 0.16, + "learning_rate": 1.9131236683273478e-05, + "loss": 1.3086, + "step": 4993 + }, + { + "epoch": 0.16, + "learning_rate": 1.913081435505916e-05, + "loss": 1.0923, + "step": 4994 + }, + { + "epoch": 0.16, + "learning_rate": 1.9130391928881312e-05, + "loss": 1.1069, + "step": 4995 + }, + { + "epoch": 0.16, + "learning_rate": 1.9129969404744465e-05, + "loss": 1.105, + "step": 4996 + }, + { + "epoch": 0.16, + "learning_rate": 1.912954678265315e-05, + "loss": 1.1006, + "step": 4997 + }, + { + "epoch": 0.16, + "learning_rate": 1.9129124062611905e-05, + "loss": 0.9443, + "step": 4998 + }, + { + "epoch": 0.16, + "learning_rate": 1.9128701244625258e-05, + "loss": 1.0117, + "step": 4999 + }, + { + "epoch": 0.16, + "learning_rate": 1.912827832869775e-05, + "loss": 1.0742, + "step": 5000 + }, + { + "epoch": 0.16, + "learning_rate": 1.912785531483392e-05, + "loss": 1.1304, + "step": 5001 + }, + { + "epoch": 0.16, + "learning_rate": 1.9127432203038307e-05, + "loss": 0.9844, + "step": 5002 + }, + { + "epoch": 0.16, + "learning_rate": 1.9127008993315446e-05, + "loss": 1.0264, + "step": 5003 + }, + { + "epoch": 0.16, + "learning_rate": 1.9126585685669877e-05, + "loss": 1.0942, + "step": 5004 + }, + { + "epoch": 0.16, + "learning_rate": 1.9126162280106148e-05, + "loss": 1.0864, + "step": 5005 + }, + { + "epoch": 0.16, + "learning_rate": 1.9125738776628795e-05, + "loss": 1.145, + "step": 5006 + }, + { + "epoch": 0.16, + "learning_rate": 1.912531517524237e-05, + "loss": 0.9263, + "step": 5007 + }, + { + "epoch": 0.16, + "learning_rate": 1.9124891475951407e-05, + "loss": 1.0347, + "step": 5008 + }, + { + "epoch": 0.16, + "learning_rate": 1.912446767876046e-05, + "loss": 1.1104, + "step": 5009 + }, + { + "epoch": 0.16, + "learning_rate": 1.9124043783674073e-05, + "loss": 1.0718, + "step": 5010 + }, + { + "epoch": 0.16, + "learning_rate": 1.9123619790696796e-05, + "loss": 1.0728, + "step": 5011 + }, + { + "epoch": 0.16, + "learning_rate": 1.9123195699833173e-05, + "loss": 1.0327, + "step": 5012 + }, + { + "epoch": 0.16, + "learning_rate": 1.9122771511087757e-05, + "loss": 1.1226, + "step": 5013 + }, + { + "epoch": 0.16, + "learning_rate": 1.91223472244651e-05, + "loss": 1.0415, + "step": 5014 + }, + { + "epoch": 0.16, + "learning_rate": 1.9121922839969754e-05, + "loss": 1.1616, + "step": 5015 + }, + { + "epoch": 0.16, + "learning_rate": 1.9121498357606273e-05, + "loss": 1.0908, + "step": 5016 + }, + { + "epoch": 0.16, + "learning_rate": 1.9121073777379207e-05, + "loss": 1.0396, + "step": 5017 + }, + { + "epoch": 0.16, + "learning_rate": 1.9120649099293117e-05, + "loss": 1.0947, + "step": 5018 + }, + { + "epoch": 0.16, + "learning_rate": 1.9120224323352553e-05, + "loss": 1.0977, + "step": 5019 + }, + { + "epoch": 0.16, + "learning_rate": 1.9119799449562076e-05, + "loss": 0.9546, + "step": 5020 + }, + { + "epoch": 0.16, + "learning_rate": 1.911937447792625e-05, + "loss": 0.9917, + "step": 5021 + }, + { + "epoch": 0.16, + "learning_rate": 1.9118949408449623e-05, + "loss": 1.0884, + "step": 5022 + }, + { + "epoch": 0.16, + "learning_rate": 1.9118524241136764e-05, + "loss": 1.0825, + "step": 5023 + }, + { + "epoch": 0.16, + "learning_rate": 1.911809897599223e-05, + "loss": 1.0796, + "step": 5024 + }, + { + "epoch": 0.16, + "learning_rate": 1.9117673613020584e-05, + "loss": 1.1191, + "step": 5025 + }, + { + "epoch": 0.16, + "learning_rate": 1.911724815222639e-05, + "loss": 1.1362, + "step": 5026 + }, + { + "epoch": 0.16, + "learning_rate": 1.9116822593614218e-05, + "loss": 1.1919, + "step": 5027 + }, + { + "epoch": 0.16, + "learning_rate": 1.911639693718863e-05, + "loss": 1.1104, + "step": 5028 + }, + { + "epoch": 0.16, + "learning_rate": 1.911597118295419e-05, + "loss": 1.1465, + "step": 5029 + }, + { + "epoch": 0.16, + "learning_rate": 1.9115545330915467e-05, + "loss": 1.0962, + "step": 5030 + }, + { + "epoch": 0.16, + "learning_rate": 1.9115119381077038e-05, + "loss": 1.2852, + "step": 5031 + }, + { + "epoch": 0.16, + "learning_rate": 1.911469333344346e-05, + "loss": 0.9116, + "step": 5032 + }, + { + "epoch": 0.16, + "learning_rate": 1.9114267188019312e-05, + "loss": 1.1587, + "step": 5033 + }, + { + "epoch": 0.16, + "learning_rate": 1.911384094480916e-05, + "loss": 0.9561, + "step": 5034 + }, + { + "epoch": 0.16, + "learning_rate": 1.9113414603817588e-05, + "loss": 1.0845, + "step": 5035 + }, + { + "epoch": 0.16, + "learning_rate": 1.911298816504916e-05, + "loss": 1.0205, + "step": 5036 + }, + { + "epoch": 0.16, + "learning_rate": 1.9112561628508456e-05, + "loss": 0.9805, + "step": 5037 + }, + { + "epoch": 0.16, + "learning_rate": 1.9112134994200052e-05, + "loss": 1.0444, + "step": 5038 + }, + { + "epoch": 0.16, + "learning_rate": 1.911170826212852e-05, + "loss": 1.2866, + "step": 5039 + }, + { + "epoch": 0.16, + "learning_rate": 1.9111281432298446e-05, + "loss": 1.0093, + "step": 5040 + }, + { + "epoch": 0.16, + "learning_rate": 1.911085450471441e-05, + "loss": 0.4788, + "step": 5041 + }, + { + "epoch": 0.16, + "learning_rate": 1.911042747938098e-05, + "loss": 1.1538, + "step": 5042 + }, + { + "epoch": 0.16, + "learning_rate": 1.911000035630275e-05, + "loss": 1.1787, + "step": 5043 + }, + { + "epoch": 0.16, + "learning_rate": 1.91095731354843e-05, + "loss": 1.3262, + "step": 5044 + }, + { + "epoch": 0.16, + "learning_rate": 1.910914581693021e-05, + "loss": 1.1323, + "step": 5045 + }, + { + "epoch": 0.16, + "learning_rate": 1.9108718400645068e-05, + "loss": 1.0498, + "step": 5046 + }, + { + "epoch": 0.16, + "learning_rate": 1.910829088663346e-05, + "loss": 1.1929, + "step": 5047 + }, + { + "epoch": 0.16, + "learning_rate": 1.9107863274899968e-05, + "loss": 1.0503, + "step": 5048 + }, + { + "epoch": 0.16, + "learning_rate": 1.9107435565449185e-05, + "loss": 1.0786, + "step": 5049 + }, + { + "epoch": 0.16, + "learning_rate": 1.9107007758285695e-05, + "loss": 1.0293, + "step": 5050 + }, + { + "epoch": 0.16, + "learning_rate": 1.9106579853414097e-05, + "loss": 1.084, + "step": 5051 + }, + { + "epoch": 0.16, + "learning_rate": 1.910615185083897e-05, + "loss": 1.1367, + "step": 5052 + }, + { + "epoch": 0.16, + "learning_rate": 1.9105723750564916e-05, + "loss": 1.1196, + "step": 5053 + }, + { + "epoch": 0.16, + "learning_rate": 1.910529555259652e-05, + "loss": 1.1201, + "step": 5054 + }, + { + "epoch": 0.16, + "learning_rate": 1.9104867256938385e-05, + "loss": 1.0679, + "step": 5055 + }, + { + "epoch": 0.16, + "learning_rate": 1.9104438863595098e-05, + "loss": 1.0684, + "step": 5056 + }, + { + "epoch": 0.16, + "learning_rate": 1.9104010372571256e-05, + "loss": 1.0508, + "step": 5057 + }, + { + "epoch": 0.16, + "learning_rate": 1.9103581783871465e-05, + "loss": 1.1255, + "step": 5058 + }, + { + "epoch": 0.16, + "learning_rate": 1.910315309750031e-05, + "loss": 1.1133, + "step": 5059 + }, + { + "epoch": 0.16, + "learning_rate": 1.9102724313462403e-05, + "loss": 0.9985, + "step": 5060 + }, + { + "epoch": 0.16, + "learning_rate": 1.9102295431762334e-05, + "loss": 1.1484, + "step": 5061 + }, + { + "epoch": 0.16, + "learning_rate": 1.910186645240471e-05, + "loss": 0.9478, + "step": 5062 + }, + { + "epoch": 0.16, + "learning_rate": 1.9101437375394134e-05, + "loss": 1.1587, + "step": 5063 + }, + { + "epoch": 0.16, + "learning_rate": 1.910100820073521e-05, + "loss": 1.1753, + "step": 5064 + }, + { + "epoch": 0.16, + "learning_rate": 1.9100578928432533e-05, + "loss": 1.0601, + "step": 5065 + }, + { + "epoch": 0.16, + "learning_rate": 1.910014955849072e-05, + "loss": 1.001, + "step": 5066 + }, + { + "epoch": 0.16, + "learning_rate": 1.9099720090914375e-05, + "loss": 1.1069, + "step": 5067 + }, + { + "epoch": 0.16, + "learning_rate": 1.9099290525708106e-05, + "loss": 1.1196, + "step": 5068 + }, + { + "epoch": 0.16, + "learning_rate": 1.9098860862876517e-05, + "loss": 0.5288, + "step": 5069 + }, + { + "epoch": 0.16, + "learning_rate": 1.909843110242422e-05, + "loss": 1.1602, + "step": 5070 + }, + { + "epoch": 0.16, + "learning_rate": 1.909800124435583e-05, + "loss": 1.0444, + "step": 5071 + }, + { + "epoch": 0.16, + "learning_rate": 1.9097571288675956e-05, + "loss": 1.1328, + "step": 5072 + }, + { + "epoch": 0.16, + "learning_rate": 1.909714123538921e-05, + "loss": 1.1699, + "step": 5073 + }, + { + "epoch": 0.16, + "learning_rate": 1.9096711084500206e-05, + "loss": 1.0073, + "step": 5074 + }, + { + "epoch": 0.16, + "learning_rate": 1.9096280836013558e-05, + "loss": 1.0938, + "step": 5075 + }, + { + "epoch": 0.16, + "learning_rate": 1.909585048993389e-05, + "loss": 1.0996, + "step": 5076 + }, + { + "epoch": 0.16, + "learning_rate": 1.9095420046265805e-05, + "loss": 1.2461, + "step": 5077 + }, + { + "epoch": 0.16, + "learning_rate": 1.9094989505013934e-05, + "loss": 1.1494, + "step": 5078 + }, + { + "epoch": 0.16, + "learning_rate": 1.9094558866182892e-05, + "loss": 0.9893, + "step": 5079 + }, + { + "epoch": 0.16, + "learning_rate": 1.90941281297773e-05, + "loss": 1.0518, + "step": 5080 + }, + { + "epoch": 0.16, + "learning_rate": 1.9093697295801772e-05, + "loss": 0.9663, + "step": 5081 + }, + { + "epoch": 0.16, + "learning_rate": 1.909326636426094e-05, + "loss": 1.1567, + "step": 5082 + }, + { + "epoch": 0.16, + "learning_rate": 1.9092835335159425e-05, + "loss": 1.106, + "step": 5083 + }, + { + "epoch": 0.16, + "learning_rate": 1.909240420850185e-05, + "loss": 1.2798, + "step": 5084 + }, + { + "epoch": 0.16, + "learning_rate": 1.909197298429284e-05, + "loss": 1.0601, + "step": 5085 + }, + { + "epoch": 0.16, + "learning_rate": 1.9091541662537022e-05, + "loss": 1.0918, + "step": 5086 + }, + { + "epoch": 0.16, + "learning_rate": 1.9091110243239024e-05, + "loss": 1.1338, + "step": 5087 + }, + { + "epoch": 0.16, + "learning_rate": 1.9090678726403475e-05, + "loss": 1.0605, + "step": 5088 + }, + { + "epoch": 0.16, + "learning_rate": 1.9090247112035006e-05, + "loss": 1.2246, + "step": 5089 + }, + { + "epoch": 0.16, + "learning_rate": 1.908981540013824e-05, + "loss": 1.1499, + "step": 5090 + }, + { + "epoch": 0.16, + "learning_rate": 1.9089383590717824e-05, + "loss": 1.1377, + "step": 5091 + }, + { + "epoch": 0.16, + "learning_rate": 1.9088951683778374e-05, + "loss": 1.1816, + "step": 5092 + }, + { + "epoch": 0.16, + "learning_rate": 1.9088519679324534e-05, + "loss": 1.1758, + "step": 5093 + }, + { + "epoch": 0.16, + "learning_rate": 1.908808757736094e-05, + "loss": 1.1333, + "step": 5094 + }, + { + "epoch": 0.16, + "learning_rate": 1.908765537789222e-05, + "loss": 1.2363, + "step": 5095 + }, + { + "epoch": 0.16, + "learning_rate": 1.9087223080923016e-05, + "loss": 0.9678, + "step": 5096 + }, + { + "epoch": 0.16, + "learning_rate": 1.9086790686457968e-05, + "loss": 0.9629, + "step": 5097 + }, + { + "epoch": 0.16, + "learning_rate": 1.908635819450171e-05, + "loss": 1.0688, + "step": 5098 + }, + { + "epoch": 0.16, + "learning_rate": 1.9085925605058885e-05, + "loss": 1.0425, + "step": 5099 + }, + { + "epoch": 0.16, + "learning_rate": 1.908549291813414e-05, + "loss": 1.0942, + "step": 5100 + }, + { + "epoch": 0.16, + "learning_rate": 1.90850601337321e-05, + "loss": 1.0742, + "step": 5101 + }, + { + "epoch": 0.16, + "learning_rate": 1.908462725185743e-05, + "loss": 1.2075, + "step": 5102 + }, + { + "epoch": 0.16, + "learning_rate": 1.908419427251476e-05, + "loss": 1.0581, + "step": 5103 + }, + { + "epoch": 0.16, + "learning_rate": 1.908376119570874e-05, + "loss": 1.0503, + "step": 5104 + }, + { + "epoch": 0.16, + "learning_rate": 1.9083328021444015e-05, + "loss": 0.9253, + "step": 5105 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082894749725233e-05, + "loss": 0.981, + "step": 5106 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082461380557045e-05, + "loss": 1.0146, + "step": 5107 + }, + { + "epoch": 0.16, + "learning_rate": 1.9082027913944093e-05, + "loss": 1.0767, + "step": 5108 + }, + { + "epoch": 0.16, + "learning_rate": 1.908159434989104e-05, + "loss": 0.5139, + "step": 5109 + }, + { + "epoch": 0.16, + "learning_rate": 1.9081160688402528e-05, + "loss": 1.1162, + "step": 5110 + }, + { + "epoch": 0.16, + "learning_rate": 1.908072692948321e-05, + "loss": 1.2207, + "step": 5111 + }, + { + "epoch": 0.16, + "learning_rate": 1.9080293073137743e-05, + "loss": 1.1758, + "step": 5112 + }, + { + "epoch": 0.16, + "learning_rate": 1.9079859119370782e-05, + "loss": 1.0664, + "step": 5113 + }, + { + "epoch": 0.16, + "learning_rate": 1.907942506818698e-05, + "loss": 1.084, + "step": 5114 + }, + { + "epoch": 0.16, + "learning_rate": 1.9078990919591e-05, + "loss": 1.1167, + "step": 5115 + }, + { + "epoch": 0.16, + "learning_rate": 1.907855667358749e-05, + "loss": 1.0586, + "step": 5116 + }, + { + "epoch": 0.16, + "learning_rate": 1.907812233018112e-05, + "loss": 1.0713, + "step": 5117 + }, + { + "epoch": 0.16, + "learning_rate": 1.907768788937654e-05, + "loss": 1.0371, + "step": 5118 + }, + { + "epoch": 0.16, + "learning_rate": 1.9077253351178416e-05, + "loss": 1.1587, + "step": 5119 + }, + { + "epoch": 0.16, + "learning_rate": 1.907681871559141e-05, + "loss": 1.0308, + "step": 5120 + }, + { + "epoch": 0.16, + "learning_rate": 1.9076383982620187e-05, + "loss": 1.0479, + "step": 5121 + }, + { + "epoch": 0.16, + "learning_rate": 1.907594915226941e-05, + "loss": 1.0312, + "step": 5122 + }, + { + "epoch": 0.16, + "learning_rate": 1.907551422454374e-05, + "loss": 1.1738, + "step": 5123 + }, + { + "epoch": 0.16, + "learning_rate": 1.907507919944785e-05, + "loss": 1.1841, + "step": 5124 + }, + { + "epoch": 0.16, + "learning_rate": 1.90746440769864e-05, + "loss": 1.144, + "step": 5125 + }, + { + "epoch": 0.16, + "learning_rate": 1.9074208857164064e-05, + "loss": 1.1475, + "step": 5126 + }, + { + "epoch": 0.16, + "learning_rate": 1.907377353998551e-05, + "loss": 1.1108, + "step": 5127 + }, + { + "epoch": 0.16, + "learning_rate": 1.9073338125455408e-05, + "loss": 1.0806, + "step": 5128 + }, + { + "epoch": 0.16, + "learning_rate": 1.9072902613578433e-05, + "loss": 1.168, + "step": 5129 + }, + { + "epoch": 0.16, + "learning_rate": 1.9072467004359252e-05, + "loss": 0.9453, + "step": 5130 + }, + { + "epoch": 0.16, + "learning_rate": 1.907203129780254e-05, + "loss": 0.9893, + "step": 5131 + }, + { + "epoch": 0.16, + "learning_rate": 1.907159549391297e-05, + "loss": 1.0801, + "step": 5132 + }, + { + "epoch": 0.16, + "learning_rate": 1.9071159592695224e-05, + "loss": 1.127, + "step": 5133 + }, + { + "epoch": 0.16, + "learning_rate": 1.9070723594153975e-05, + "loss": 1.1299, + "step": 5134 + }, + { + "epoch": 0.16, + "learning_rate": 1.90702874982939e-05, + "loss": 1.061, + "step": 5135 + }, + { + "epoch": 0.16, + "learning_rate": 1.906985130511968e-05, + "loss": 1.064, + "step": 5136 + }, + { + "epoch": 0.16, + "learning_rate": 1.9069415014635993e-05, + "loss": 1.0791, + "step": 5137 + }, + { + "epoch": 0.16, + "learning_rate": 1.9068978626847522e-05, + "loss": 1.0918, + "step": 5138 + }, + { + "epoch": 0.16, + "learning_rate": 1.9068542141758943e-05, + "loss": 1.0913, + "step": 5139 + }, + { + "epoch": 0.16, + "learning_rate": 1.9068105559374946e-05, + "loss": 1.0835, + "step": 5140 + }, + { + "epoch": 0.16, + "learning_rate": 1.9067668879700212e-05, + "loss": 1.0249, + "step": 5141 + }, + { + "epoch": 0.16, + "learning_rate": 1.9067232102739425e-05, + "loss": 1.0605, + "step": 5142 + }, + { + "epoch": 0.16, + "learning_rate": 1.9066795228497276e-05, + "loss": 1.1616, + "step": 5143 + }, + { + "epoch": 0.16, + "learning_rate": 1.9066358256978445e-05, + "loss": 1.1411, + "step": 5144 + }, + { + "epoch": 0.16, + "learning_rate": 1.9065921188187628e-05, + "loss": 1.0806, + "step": 5145 + }, + { + "epoch": 0.16, + "learning_rate": 1.9065484022129507e-05, + "loss": 0.9526, + "step": 5146 + }, + { + "epoch": 0.16, + "learning_rate": 1.9065046758808778e-05, + "loss": 1.1147, + "step": 5147 + }, + { + "epoch": 0.16, + "learning_rate": 1.906460939823013e-05, + "loss": 1.0933, + "step": 5148 + }, + { + "epoch": 0.16, + "learning_rate": 1.9064171940398255e-05, + "loss": 0.9448, + "step": 5149 + }, + { + "epoch": 0.16, + "learning_rate": 1.9063734385317844e-05, + "loss": 1.1436, + "step": 5150 + }, + { + "epoch": 0.16, + "learning_rate": 1.9063296732993598e-05, + "loss": 1.1821, + "step": 5151 + }, + { + "epoch": 0.16, + "learning_rate": 1.9062858983430207e-05, + "loss": 1.0249, + "step": 5152 + }, + { + "epoch": 0.16, + "learning_rate": 1.906242113663237e-05, + "loss": 1.1079, + "step": 5153 + }, + { + "epoch": 0.16, + "learning_rate": 1.9061983192604784e-05, + "loss": 1.1353, + "step": 5154 + }, + { + "epoch": 0.16, + "learning_rate": 1.9061545151352147e-05, + "loss": 1.1807, + "step": 5155 + }, + { + "epoch": 0.16, + "learning_rate": 1.906110701287916e-05, + "loss": 1.0244, + "step": 5156 + }, + { + "epoch": 0.16, + "learning_rate": 1.906066877719052e-05, + "loss": 1.0854, + "step": 5157 + }, + { + "epoch": 0.16, + "learning_rate": 1.9060230444290935e-05, + "loss": 1.0874, + "step": 5158 + }, + { + "epoch": 0.16, + "learning_rate": 1.9059792014185106e-05, + "loss": 1.0234, + "step": 5159 + }, + { + "epoch": 0.17, + "learning_rate": 1.9059353486877734e-05, + "loss": 1.0059, + "step": 5160 + }, + { + "epoch": 0.17, + "learning_rate": 1.9058914862373527e-05, + "loss": 1.0439, + "step": 5161 + }, + { + "epoch": 0.17, + "learning_rate": 1.905847614067719e-05, + "loss": 1.0684, + "step": 5162 + }, + { + "epoch": 0.17, + "learning_rate": 1.9058037321793425e-05, + "loss": 1.0493, + "step": 5163 + }, + { + "epoch": 0.17, + "learning_rate": 1.9057598405726945e-05, + "loss": 1.001, + "step": 5164 + }, + { + "epoch": 0.17, + "learning_rate": 1.905715939248246e-05, + "loss": 1.2031, + "step": 5165 + }, + { + "epoch": 0.17, + "learning_rate": 1.905672028206468e-05, + "loss": 1.0059, + "step": 5166 + }, + { + "epoch": 0.17, + "learning_rate": 1.9056281074478313e-05, + "loss": 1.0034, + "step": 5167 + }, + { + "epoch": 0.17, + "learning_rate": 1.9055841769728076e-05, + "loss": 1.0884, + "step": 5168 + }, + { + "epoch": 0.17, + "learning_rate": 1.9055402367818673e-05, + "loss": 1.2031, + "step": 5169 + }, + { + "epoch": 0.17, + "learning_rate": 1.905496286875483e-05, + "loss": 1.0195, + "step": 5170 + }, + { + "epoch": 0.17, + "learning_rate": 1.9054523272541255e-05, + "loss": 0.4795, + "step": 5171 + }, + { + "epoch": 0.17, + "learning_rate": 1.905408357918267e-05, + "loss": 1.2144, + "step": 5172 + }, + { + "epoch": 0.17, + "learning_rate": 1.9053643788683786e-05, + "loss": 1.1167, + "step": 5173 + }, + { + "epoch": 0.17, + "learning_rate": 1.9053203901049323e-05, + "loss": 1.1304, + "step": 5174 + }, + { + "epoch": 0.17, + "learning_rate": 1.9052763916284003e-05, + "loss": 1.1177, + "step": 5175 + }, + { + "epoch": 0.17, + "learning_rate": 1.9052323834392546e-05, + "loss": 1.0825, + "step": 5176 + }, + { + "epoch": 0.17, + "learning_rate": 1.905188365537967e-05, + "loss": 1.0439, + "step": 5177 + }, + { + "epoch": 0.17, + "learning_rate": 1.9051443379250104e-05, + "loss": 1.1694, + "step": 5178 + }, + { + "epoch": 0.17, + "learning_rate": 1.9051003006008566e-05, + "loss": 1.062, + "step": 5179 + }, + { + "epoch": 0.17, + "learning_rate": 1.9050562535659783e-05, + "loss": 1.0698, + "step": 5180 + }, + { + "epoch": 0.17, + "learning_rate": 1.9050121968208484e-05, + "loss": 1.0757, + "step": 5181 + }, + { + "epoch": 0.17, + "learning_rate": 1.904968130365939e-05, + "loss": 1.2217, + "step": 5182 + }, + { + "epoch": 0.17, + "learning_rate": 1.904924054201723e-05, + "loss": 1.0391, + "step": 5183 + }, + { + "epoch": 0.17, + "learning_rate": 1.9048799683286734e-05, + "loss": 1.0708, + "step": 5184 + }, + { + "epoch": 0.17, + "learning_rate": 1.9048358727472635e-05, + "loss": 1.0322, + "step": 5185 + }, + { + "epoch": 0.17, + "learning_rate": 1.9047917674579656e-05, + "loss": 1.1426, + "step": 5186 + }, + { + "epoch": 0.17, + "learning_rate": 1.904747652461254e-05, + "loss": 1.0801, + "step": 5187 + }, + { + "epoch": 0.17, + "learning_rate": 1.9047035277576012e-05, + "loss": 0.9834, + "step": 5188 + }, + { + "epoch": 0.17, + "learning_rate": 1.9046593933474807e-05, + "loss": 1.1201, + "step": 5189 + }, + { + "epoch": 0.17, + "learning_rate": 1.904615249231366e-05, + "loss": 0.9102, + "step": 5190 + }, + { + "epoch": 0.17, + "learning_rate": 1.9045710954097312e-05, + "loss": 1.0317, + "step": 5191 + }, + { + "epoch": 0.17, + "learning_rate": 1.9045269318830497e-05, + "loss": 1.0596, + "step": 5192 + }, + { + "epoch": 0.17, + "learning_rate": 1.904482758651795e-05, + "loss": 1.1489, + "step": 5193 + }, + { + "epoch": 0.17, + "learning_rate": 1.9044385757164418e-05, + "loss": 1.0576, + "step": 5194 + }, + { + "epoch": 0.17, + "learning_rate": 1.904394383077463e-05, + "loss": 1.019, + "step": 5195 + }, + { + "epoch": 0.17, + "learning_rate": 1.904350180735334e-05, + "loss": 1.064, + "step": 5196 + }, + { + "epoch": 0.17, + "learning_rate": 1.904305968690528e-05, + "loss": 0.5203, + "step": 5197 + }, + { + "epoch": 0.17, + "learning_rate": 1.90426174694352e-05, + "loss": 1.0464, + "step": 5198 + }, + { + "epoch": 0.17, + "learning_rate": 1.9042175154947842e-05, + "loss": 0.5452, + "step": 5199 + }, + { + "epoch": 0.17, + "learning_rate": 1.9041732743447955e-05, + "loss": 1.0366, + "step": 5200 + }, + { + "epoch": 0.17, + "learning_rate": 1.904129023494028e-05, + "loss": 1.0278, + "step": 5201 + }, + { + "epoch": 0.17, + "learning_rate": 1.9040847629429567e-05, + "loss": 0.502, + "step": 5202 + }, + { + "epoch": 0.17, + "learning_rate": 1.9040404926920564e-05, + "loss": 1.1826, + "step": 5203 + }, + { + "epoch": 0.17, + "learning_rate": 1.9039962127418024e-05, + "loss": 1.0601, + "step": 5204 + }, + { + "epoch": 0.17, + "learning_rate": 1.9039519230926692e-05, + "loss": 0.5232, + "step": 5205 + }, + { + "epoch": 0.17, + "learning_rate": 1.9039076237451324e-05, + "loss": 1.1294, + "step": 5206 + }, + { + "epoch": 0.17, + "learning_rate": 1.9038633146996674e-05, + "loss": 1.0659, + "step": 5207 + }, + { + "epoch": 0.17, + "learning_rate": 1.903818995956749e-05, + "loss": 1.1821, + "step": 5208 + }, + { + "epoch": 0.17, + "learning_rate": 1.9037746675168537e-05, + "loss": 1.0483, + "step": 5209 + }, + { + "epoch": 0.17, + "learning_rate": 1.9037303293804558e-05, + "loss": 1.1211, + "step": 5210 + }, + { + "epoch": 0.17, + "learning_rate": 1.903685981548032e-05, + "loss": 1.0195, + "step": 5211 + }, + { + "epoch": 0.17, + "learning_rate": 1.903641624020058e-05, + "loss": 1.0894, + "step": 5212 + }, + { + "epoch": 0.17, + "learning_rate": 1.903597256797009e-05, + "loss": 1.002, + "step": 5213 + }, + { + "epoch": 0.17, + "learning_rate": 1.903552879879362e-05, + "loss": 1.0977, + "step": 5214 + }, + { + "epoch": 0.17, + "learning_rate": 1.903508493267592e-05, + "loss": 1.1221, + "step": 5215 + }, + { + "epoch": 0.17, + "learning_rate": 1.9034640969621764e-05, + "loss": 1.0259, + "step": 5216 + }, + { + "epoch": 0.17, + "learning_rate": 1.9034196909635906e-05, + "loss": 1.0356, + "step": 5217 + }, + { + "epoch": 0.17, + "learning_rate": 1.9033752752723112e-05, + "loss": 1.1069, + "step": 5218 + }, + { + "epoch": 0.17, + "learning_rate": 1.9033308498888154e-05, + "loss": 1.1665, + "step": 5219 + }, + { + "epoch": 0.17, + "learning_rate": 1.9032864148135786e-05, + "loss": 1.1543, + "step": 5220 + }, + { + "epoch": 0.17, + "learning_rate": 1.903241970047079e-05, + "loss": 1.0781, + "step": 5221 + }, + { + "epoch": 0.17, + "learning_rate": 1.9031975155897923e-05, + "loss": 1.0762, + "step": 5222 + }, + { + "epoch": 0.17, + "learning_rate": 1.903153051442196e-05, + "loss": 1.021, + "step": 5223 + }, + { + "epoch": 0.17, + "learning_rate": 1.9031085776047673e-05, + "loss": 1.1494, + "step": 5224 + }, + { + "epoch": 0.17, + "learning_rate": 1.9030640940779826e-05, + "loss": 1.1777, + "step": 5225 + }, + { + "epoch": 0.17, + "learning_rate": 1.90301960086232e-05, + "loss": 1.1055, + "step": 5226 + }, + { + "epoch": 0.17, + "learning_rate": 1.9029750979582563e-05, + "loss": 1.0737, + "step": 5227 + }, + { + "epoch": 0.17, + "learning_rate": 1.9029305853662692e-05, + "loss": 1.0366, + "step": 5228 + }, + { + "epoch": 0.17, + "learning_rate": 1.9028860630868364e-05, + "loss": 1.1802, + "step": 5229 + }, + { + "epoch": 0.17, + "learning_rate": 1.9028415311204353e-05, + "loss": 0.5098, + "step": 5230 + }, + { + "epoch": 0.17, + "learning_rate": 1.9027969894675437e-05, + "loss": 1.0767, + "step": 5231 + }, + { + "epoch": 0.17, + "learning_rate": 1.90275243812864e-05, + "loss": 1.0132, + "step": 5232 + }, + { + "epoch": 0.17, + "learning_rate": 1.902707877104201e-05, + "loss": 1.0327, + "step": 5233 + }, + { + "epoch": 0.17, + "learning_rate": 1.9026633063947056e-05, + "loss": 1.041, + "step": 5234 + }, + { + "epoch": 0.17, + "learning_rate": 1.9026187260006326e-05, + "loss": 1.0249, + "step": 5235 + }, + { + "epoch": 0.17, + "learning_rate": 1.902574135922459e-05, + "loss": 1.2505, + "step": 5236 + }, + { + "epoch": 0.17, + "learning_rate": 1.9025295361606644e-05, + "loss": 1.1475, + "step": 5237 + }, + { + "epoch": 0.17, + "learning_rate": 1.9024849267157263e-05, + "loss": 1.1118, + "step": 5238 + }, + { + "epoch": 0.17, + "learning_rate": 1.9024403075881242e-05, + "loss": 1.1733, + "step": 5239 + }, + { + "epoch": 0.17, + "learning_rate": 1.902395678778336e-05, + "loss": 1.0674, + "step": 5240 + }, + { + "epoch": 0.17, + "learning_rate": 1.902351040286841e-05, + "loss": 0.959, + "step": 5241 + }, + { + "epoch": 0.17, + "learning_rate": 1.902306392114118e-05, + "loss": 1.1509, + "step": 5242 + }, + { + "epoch": 0.17, + "learning_rate": 1.902261734260646e-05, + "loss": 1.0322, + "step": 5243 + }, + { + "epoch": 0.17, + "learning_rate": 1.9022170667269046e-05, + "loss": 1.1455, + "step": 5244 + }, + { + "epoch": 0.17, + "learning_rate": 1.902172389513372e-05, + "loss": 1.0298, + "step": 5245 + }, + { + "epoch": 0.17, + "learning_rate": 1.9021277026205282e-05, + "loss": 1.0142, + "step": 5246 + }, + { + "epoch": 0.17, + "learning_rate": 1.9020830060488528e-05, + "loss": 1.1147, + "step": 5247 + }, + { + "epoch": 0.17, + "learning_rate": 1.902038299798825e-05, + "loss": 1.0186, + "step": 5248 + }, + { + "epoch": 0.17, + "learning_rate": 1.9019935838709246e-05, + "loss": 1.1079, + "step": 5249 + }, + { + "epoch": 0.17, + "learning_rate": 1.9019488582656313e-05, + "loss": 1.0762, + "step": 5250 + }, + { + "epoch": 0.17, + "learning_rate": 1.901904122983425e-05, + "loss": 1.1582, + "step": 5251 + }, + { + "epoch": 0.17, + "learning_rate": 1.9018593780247857e-05, + "loss": 1.0356, + "step": 5252 + }, + { + "epoch": 0.17, + "learning_rate": 1.9018146233901934e-05, + "loss": 1.0693, + "step": 5253 + }, + { + "epoch": 0.17, + "learning_rate": 1.9017698590801283e-05, + "loss": 0.5493, + "step": 5254 + }, + { + "epoch": 0.17, + "learning_rate": 1.9017250850950707e-05, + "loss": 1.0552, + "step": 5255 + }, + { + "epoch": 0.17, + "learning_rate": 1.9016803014355005e-05, + "loss": 1.103, + "step": 5256 + }, + { + "epoch": 0.17, + "learning_rate": 1.901635508101899e-05, + "loss": 1.019, + "step": 5257 + }, + { + "epoch": 0.17, + "learning_rate": 1.901590705094746e-05, + "loss": 1.0698, + "step": 5258 + }, + { + "epoch": 0.17, + "learning_rate": 1.901545892414523e-05, + "loss": 1.0835, + "step": 5259 + }, + { + "epoch": 0.17, + "learning_rate": 1.9015010700617103e-05, + "loss": 1.0713, + "step": 5260 + }, + { + "epoch": 0.17, + "learning_rate": 1.9014562380367885e-05, + "loss": 1.0786, + "step": 5261 + }, + { + "epoch": 0.17, + "learning_rate": 1.9014113963402393e-05, + "loss": 1.043, + "step": 5262 + }, + { + "epoch": 0.17, + "learning_rate": 1.9013665449725433e-05, + "loss": 1.1533, + "step": 5263 + }, + { + "epoch": 0.17, + "learning_rate": 1.901321683934182e-05, + "loss": 1.0122, + "step": 5264 + }, + { + "epoch": 0.17, + "learning_rate": 1.9012768132256366e-05, + "loss": 1.0103, + "step": 5265 + }, + { + "epoch": 0.17, + "learning_rate": 1.901231932847388e-05, + "loss": 1.1484, + "step": 5266 + }, + { + "epoch": 0.17, + "learning_rate": 1.9011870427999187e-05, + "loss": 1.0952, + "step": 5267 + }, + { + "epoch": 0.17, + "learning_rate": 1.9011421430837095e-05, + "loss": 0.9946, + "step": 5268 + }, + { + "epoch": 0.17, + "learning_rate": 1.9010972336992426e-05, + "loss": 0.9385, + "step": 5269 + }, + { + "epoch": 0.17, + "learning_rate": 1.9010523146469998e-05, + "loss": 1.0557, + "step": 5270 + }, + { + "epoch": 0.17, + "learning_rate": 1.9010073859274625e-05, + "loss": 1.2695, + "step": 5271 + }, + { + "epoch": 0.17, + "learning_rate": 1.9009624475411134e-05, + "loss": 1.0674, + "step": 5272 + }, + { + "epoch": 0.17, + "learning_rate": 1.9009174994884344e-05, + "loss": 1.0576, + "step": 5273 + }, + { + "epoch": 0.17, + "learning_rate": 1.9008725417699077e-05, + "loss": 1.1274, + "step": 5274 + }, + { + "epoch": 0.17, + "learning_rate": 1.9008275743860157e-05, + "loss": 1.1309, + "step": 5275 + }, + { + "epoch": 0.17, + "learning_rate": 1.9007825973372407e-05, + "loss": 1.0952, + "step": 5276 + }, + { + "epoch": 0.17, + "learning_rate": 1.9007376106240654e-05, + "loss": 1.0181, + "step": 5277 + }, + { + "epoch": 0.17, + "learning_rate": 1.9006926142469722e-05, + "loss": 1.1587, + "step": 5278 + }, + { + "epoch": 0.17, + "learning_rate": 1.9006476082064448e-05, + "loss": 1.1279, + "step": 5279 + }, + { + "epoch": 0.17, + "learning_rate": 1.9006025925029648e-05, + "loss": 1.0483, + "step": 5280 + }, + { + "epoch": 0.17, + "learning_rate": 1.9005575671370157e-05, + "loss": 1.1284, + "step": 5281 + }, + { + "epoch": 0.17, + "learning_rate": 1.9005125321090804e-05, + "loss": 1.0103, + "step": 5282 + }, + { + "epoch": 0.17, + "learning_rate": 1.9004674874196427e-05, + "loss": 1.0986, + "step": 5283 + }, + { + "epoch": 0.17, + "learning_rate": 1.9004224330691853e-05, + "loss": 1.1455, + "step": 5284 + }, + { + "epoch": 0.17, + "learning_rate": 1.900377369058192e-05, + "loss": 1.0322, + "step": 5285 + }, + { + "epoch": 0.17, + "learning_rate": 1.9003322953871454e-05, + "loss": 1.0508, + "step": 5286 + }, + { + "epoch": 0.17, + "learning_rate": 1.90028721205653e-05, + "loss": 1.0405, + "step": 5287 + }, + { + "epoch": 0.17, + "learning_rate": 1.9002421190668296e-05, + "loss": 1.1392, + "step": 5288 + }, + { + "epoch": 0.17, + "learning_rate": 1.900197016418527e-05, + "loss": 1.2603, + "step": 5289 + }, + { + "epoch": 0.17, + "learning_rate": 1.9001519041121074e-05, + "loss": 1.062, + "step": 5290 + }, + { + "epoch": 0.17, + "learning_rate": 1.900106782148053e-05, + "loss": 1.0459, + "step": 5291 + }, + { + "epoch": 0.17, + "learning_rate": 1.90006165052685e-05, + "loss": 1.0703, + "step": 5292 + }, + { + "epoch": 0.17, + "learning_rate": 1.9000165092489814e-05, + "loss": 1.2095, + "step": 5293 + }, + { + "epoch": 0.17, + "learning_rate": 1.8999713583149315e-05, + "loss": 1.0645, + "step": 5294 + }, + { + "epoch": 0.17, + "learning_rate": 1.899926197725185e-05, + "loss": 1.0449, + "step": 5295 + }, + { + "epoch": 0.17, + "learning_rate": 1.8998810274802268e-05, + "loss": 1.0972, + "step": 5296 + }, + { + "epoch": 0.17, + "learning_rate": 1.899835847580541e-05, + "loss": 1.1895, + "step": 5297 + }, + { + "epoch": 0.17, + "learning_rate": 1.899790658026612e-05, + "loss": 1.0654, + "step": 5298 + }, + { + "epoch": 0.17, + "learning_rate": 1.8997454588189253e-05, + "loss": 1.127, + "step": 5299 + }, + { + "epoch": 0.17, + "learning_rate": 1.8997002499579656e-05, + "loss": 1.0869, + "step": 5300 + }, + { + "epoch": 0.17, + "learning_rate": 1.899655031444218e-05, + "loss": 1.0513, + "step": 5301 + }, + { + "epoch": 0.17, + "learning_rate": 1.899609803278168e-05, + "loss": 1.1885, + "step": 5302 + }, + { + "epoch": 0.17, + "learning_rate": 1.8995645654602997e-05, + "loss": 1.1782, + "step": 5303 + }, + { + "epoch": 0.17, + "learning_rate": 1.8995193179911e-05, + "loss": 0.9883, + "step": 5304 + }, + { + "epoch": 0.17, + "learning_rate": 1.899474060871053e-05, + "loss": 0.9458, + "step": 5305 + }, + { + "epoch": 0.17, + "learning_rate": 1.8994287941006448e-05, + "loss": 1.0386, + "step": 5306 + }, + { + "epoch": 0.17, + "learning_rate": 1.8993835176803613e-05, + "loss": 1.0439, + "step": 5307 + }, + { + "epoch": 0.17, + "learning_rate": 1.899338231610688e-05, + "loss": 1.0093, + "step": 5308 + }, + { + "epoch": 0.17, + "learning_rate": 1.899292935892111e-05, + "loss": 1.1167, + "step": 5309 + }, + { + "epoch": 0.17, + "learning_rate": 1.8992476305251158e-05, + "loss": 1.1602, + "step": 5310 + }, + { + "epoch": 0.17, + "learning_rate": 1.8992023155101887e-05, + "loss": 1.1216, + "step": 5311 + }, + { + "epoch": 0.17, + "learning_rate": 1.899156990847816e-05, + "loss": 1.0542, + "step": 5312 + }, + { + "epoch": 0.17, + "learning_rate": 1.8991116565384842e-05, + "loss": 1.1455, + "step": 5313 + }, + { + "epoch": 0.17, + "learning_rate": 1.8990663125826788e-05, + "loss": 1.0342, + "step": 5314 + }, + { + "epoch": 0.17, + "learning_rate": 1.8990209589808873e-05, + "loss": 1.1147, + "step": 5315 + }, + { + "epoch": 0.17, + "learning_rate": 1.8989755957335962e-05, + "loss": 1.1177, + "step": 5316 + }, + { + "epoch": 0.17, + "learning_rate": 1.8989302228412917e-05, + "loss": 1.1426, + "step": 5317 + }, + { + "epoch": 0.17, + "learning_rate": 1.8988848403044607e-05, + "loss": 0.4954, + "step": 5318 + }, + { + "epoch": 0.17, + "learning_rate": 1.89883944812359e-05, + "loss": 1.1089, + "step": 5319 + }, + { + "epoch": 0.17, + "learning_rate": 1.8987940462991673e-05, + "loss": 1.0195, + "step": 5320 + }, + { + "epoch": 0.17, + "learning_rate": 1.898748634831679e-05, + "loss": 1.1128, + "step": 5321 + }, + { + "epoch": 0.17, + "learning_rate": 1.8987032137216123e-05, + "loss": 1.0854, + "step": 5322 + }, + { + "epoch": 0.17, + "learning_rate": 1.8986577829694548e-05, + "loss": 1.166, + "step": 5323 + }, + { + "epoch": 0.17, + "learning_rate": 1.898612342575694e-05, + "loss": 1.1387, + "step": 5324 + }, + { + "epoch": 0.17, + "learning_rate": 1.8985668925408173e-05, + "loss": 1.2402, + "step": 5325 + }, + { + "epoch": 0.17, + "learning_rate": 1.8985214328653124e-05, + "loss": 1.2222, + "step": 5326 + }, + { + "epoch": 0.17, + "learning_rate": 1.8984759635496666e-05, + "loss": 0.9824, + "step": 5327 + }, + { + "epoch": 0.17, + "learning_rate": 1.8984304845943685e-05, + "loss": 1.1445, + "step": 5328 + }, + { + "epoch": 0.17, + "learning_rate": 1.898384995999906e-05, + "loss": 1.0586, + "step": 5329 + }, + { + "epoch": 0.17, + "learning_rate": 1.898339497766766e-05, + "loss": 1.0625, + "step": 5330 + }, + { + "epoch": 0.17, + "learning_rate": 1.8982939898954377e-05, + "loss": 0.9858, + "step": 5331 + }, + { + "epoch": 0.17, + "learning_rate": 1.898248472386409e-05, + "loss": 1.0234, + "step": 5332 + }, + { + "epoch": 0.17, + "learning_rate": 1.8982029452401684e-05, + "loss": 1.1958, + "step": 5333 + }, + { + "epoch": 0.17, + "learning_rate": 1.8981574084572042e-05, + "loss": 1.0127, + "step": 5334 + }, + { + "epoch": 0.17, + "learning_rate": 1.8981118620380052e-05, + "loss": 1.2373, + "step": 5335 + }, + { + "epoch": 0.17, + "learning_rate": 1.89806630598306e-05, + "loss": 1.0845, + "step": 5336 + }, + { + "epoch": 0.17, + "learning_rate": 1.8980207402928566e-05, + "loss": 1.0259, + "step": 5337 + }, + { + "epoch": 0.17, + "learning_rate": 1.8979751649678855e-05, + "loss": 0.9614, + "step": 5338 + }, + { + "epoch": 0.17, + "learning_rate": 1.8979295800086342e-05, + "loss": 1.0957, + "step": 5339 + }, + { + "epoch": 0.17, + "learning_rate": 1.897883985415592e-05, + "loss": 1.0659, + "step": 5340 + }, + { + "epoch": 0.17, + "learning_rate": 1.897838381189249e-05, + "loss": 1.1289, + "step": 5341 + }, + { + "epoch": 0.17, + "learning_rate": 1.8977927673300934e-05, + "loss": 1.0366, + "step": 5342 + }, + { + "epoch": 0.17, + "learning_rate": 1.8977471438386154e-05, + "loss": 1.0869, + "step": 5343 + }, + { + "epoch": 0.17, + "learning_rate": 1.8977015107153035e-05, + "loss": 1.062, + "step": 5344 + }, + { + "epoch": 0.17, + "learning_rate": 1.8976558679606486e-05, + "loss": 1.1216, + "step": 5345 + }, + { + "epoch": 0.17, + "learning_rate": 1.8976102155751393e-05, + "loss": 1.1763, + "step": 5346 + }, + { + "epoch": 0.17, + "learning_rate": 1.897564553559266e-05, + "loss": 0.9941, + "step": 5347 + }, + { + "epoch": 0.17, + "learning_rate": 1.8975188819135183e-05, + "loss": 1.0703, + "step": 5348 + }, + { + "epoch": 0.17, + "learning_rate": 1.897473200638386e-05, + "loss": 1.1621, + "step": 5349 + }, + { + "epoch": 0.17, + "learning_rate": 1.8974275097343602e-05, + "loss": 1.0942, + "step": 5350 + }, + { + "epoch": 0.17, + "learning_rate": 1.89738180920193e-05, + "loss": 1.1138, + "step": 5351 + }, + { + "epoch": 0.17, + "learning_rate": 1.8973360990415863e-05, + "loss": 1.1382, + "step": 5352 + }, + { + "epoch": 0.17, + "learning_rate": 1.8972903792538196e-05, + "loss": 1.0293, + "step": 5353 + }, + { + "epoch": 0.17, + "learning_rate": 1.8972446498391202e-05, + "loss": 1.0649, + "step": 5354 + }, + { + "epoch": 0.17, + "learning_rate": 1.897198910797978e-05, + "loss": 0.9683, + "step": 5355 + }, + { + "epoch": 0.17, + "learning_rate": 1.8971531621308853e-05, + "loss": 1.147, + "step": 5356 + }, + { + "epoch": 0.17, + "learning_rate": 1.8971074038383316e-05, + "loss": 1.0767, + "step": 5357 + }, + { + "epoch": 0.17, + "learning_rate": 1.8970616359208084e-05, + "loss": 1.0894, + "step": 5358 + }, + { + "epoch": 0.17, + "learning_rate": 1.897015858378807e-05, + "loss": 1.1396, + "step": 5359 + }, + { + "epoch": 0.17, + "learning_rate": 1.896970071212818e-05, + "loss": 1.0425, + "step": 5360 + }, + { + "epoch": 0.17, + "learning_rate": 1.896924274423333e-05, + "loss": 1.0474, + "step": 5361 + }, + { + "epoch": 0.17, + "learning_rate": 1.896878468010843e-05, + "loss": 1.0684, + "step": 5362 + }, + { + "epoch": 0.17, + "learning_rate": 1.8968326519758394e-05, + "loss": 1.0664, + "step": 5363 + }, + { + "epoch": 0.17, + "learning_rate": 1.8967868263188144e-05, + "loss": 0.9224, + "step": 5364 + }, + { + "epoch": 0.17, + "learning_rate": 1.896740991040259e-05, + "loss": 1.0312, + "step": 5365 + }, + { + "epoch": 0.17, + "learning_rate": 1.8966951461406652e-05, + "loss": 1.0479, + "step": 5366 + }, + { + "epoch": 0.17, + "learning_rate": 1.8966492916205253e-05, + "loss": 1.106, + "step": 5367 + }, + { + "epoch": 0.17, + "learning_rate": 1.8966034274803306e-05, + "loss": 0.9717, + "step": 5368 + }, + { + "epoch": 0.17, + "learning_rate": 1.8965575537205736e-05, + "loss": 1.1543, + "step": 5369 + }, + { + "epoch": 0.17, + "learning_rate": 1.896511670341746e-05, + "loss": 1.1392, + "step": 5370 + }, + { + "epoch": 0.17, + "learning_rate": 1.8964657773443406e-05, + "loss": 1.043, + "step": 5371 + }, + { + "epoch": 0.17, + "learning_rate": 1.8964198747288497e-05, + "loss": 0.939, + "step": 5372 + }, + { + "epoch": 0.17, + "learning_rate": 1.8963739624957655e-05, + "loss": 1.0376, + "step": 5373 + }, + { + "epoch": 0.17, + "learning_rate": 1.8963280406455808e-05, + "loss": 0.9468, + "step": 5374 + }, + { + "epoch": 0.17, + "learning_rate": 1.8962821091787883e-05, + "loss": 1.1484, + "step": 5375 + }, + { + "epoch": 0.17, + "learning_rate": 1.8962361680958802e-05, + "loss": 1.1489, + "step": 5376 + }, + { + "epoch": 0.17, + "learning_rate": 1.8961902173973504e-05, + "loss": 1.0415, + "step": 5377 + }, + { + "epoch": 0.17, + "learning_rate": 1.8961442570836915e-05, + "loss": 1.1812, + "step": 5378 + }, + { + "epoch": 0.17, + "learning_rate": 1.8960982871553963e-05, + "loss": 0.958, + "step": 5379 + }, + { + "epoch": 0.17, + "learning_rate": 1.896052307612958e-05, + "loss": 1.0825, + "step": 5380 + }, + { + "epoch": 0.17, + "learning_rate": 1.8960063184568707e-05, + "loss": 0.9648, + "step": 5381 + }, + { + "epoch": 0.17, + "learning_rate": 1.895960319687627e-05, + "loss": 1.1055, + "step": 5382 + }, + { + "epoch": 0.17, + "learning_rate": 1.8959143113057205e-05, + "loss": 1.0742, + "step": 5383 + }, + { + "epoch": 0.17, + "learning_rate": 1.895868293311645e-05, + "loss": 1.0845, + "step": 5384 + }, + { + "epoch": 0.17, + "learning_rate": 1.8958222657058945e-05, + "loss": 1.0645, + "step": 5385 + }, + { + "epoch": 0.17, + "learning_rate": 1.895776228488962e-05, + "loss": 1.0859, + "step": 5386 + }, + { + "epoch": 0.17, + "learning_rate": 1.8957301816613428e-05, + "loss": 1.2744, + "step": 5387 + }, + { + "epoch": 0.17, + "learning_rate": 1.8956841252235295e-05, + "loss": 1.0776, + "step": 5388 + }, + { + "epoch": 0.17, + "learning_rate": 1.895638059176017e-05, + "loss": 1.0898, + "step": 5389 + }, + { + "epoch": 0.17, + "learning_rate": 1.8955919835192997e-05, + "loss": 1.0132, + "step": 5390 + }, + { + "epoch": 0.17, + "learning_rate": 1.8955458982538707e-05, + "loss": 1.1538, + "step": 5391 + }, + { + "epoch": 0.17, + "learning_rate": 1.895499803380226e-05, + "loss": 1.123, + "step": 5392 + }, + { + "epoch": 0.17, + "learning_rate": 1.8954536988988596e-05, + "loss": 1.1216, + "step": 5393 + }, + { + "epoch": 0.17, + "learning_rate": 1.895407584810266e-05, + "loss": 1.1538, + "step": 5394 + }, + { + "epoch": 0.17, + "learning_rate": 1.8953614611149395e-05, + "loss": 1.0918, + "step": 5395 + }, + { + "epoch": 0.17, + "learning_rate": 1.895315327813376e-05, + "loss": 1.1294, + "step": 5396 + }, + { + "epoch": 0.17, + "learning_rate": 1.8952691849060694e-05, + "loss": 1.1152, + "step": 5397 + }, + { + "epoch": 0.17, + "learning_rate": 1.895223032393516e-05, + "loss": 0.5188, + "step": 5398 + }, + { + "epoch": 0.17, + "learning_rate": 1.8951768702762097e-05, + "loss": 1.1069, + "step": 5399 + }, + { + "epoch": 0.17, + "learning_rate": 1.895130698554647e-05, + "loss": 0.9727, + "step": 5400 + }, + { + "epoch": 0.17, + "learning_rate": 1.8950845172293215e-05, + "loss": 1.0181, + "step": 5401 + }, + { + "epoch": 0.17, + "learning_rate": 1.8950383263007305e-05, + "loss": 0.959, + "step": 5402 + }, + { + "epoch": 0.17, + "learning_rate": 1.8949921257693688e-05, + "loss": 0.9688, + "step": 5403 + }, + { + "epoch": 0.17, + "learning_rate": 1.894945915635732e-05, + "loss": 1.1489, + "step": 5404 + }, + { + "epoch": 0.17, + "learning_rate": 1.8948996959003158e-05, + "loss": 1.1372, + "step": 5405 + }, + { + "epoch": 0.17, + "learning_rate": 1.8948534665636166e-05, + "loss": 1.1226, + "step": 5406 + }, + { + "epoch": 0.17, + "learning_rate": 1.89480722762613e-05, + "loss": 1.103, + "step": 5407 + }, + { + "epoch": 0.17, + "learning_rate": 1.894760979088352e-05, + "loss": 1.0493, + "step": 5408 + }, + { + "epoch": 0.17, + "learning_rate": 1.894714720950779e-05, + "loss": 1.1138, + "step": 5409 + }, + { + "epoch": 0.17, + "learning_rate": 1.8946684532139073e-05, + "loss": 1.0796, + "step": 5410 + }, + { + "epoch": 0.17, + "learning_rate": 1.8946221758782335e-05, + "loss": 1.0845, + "step": 5411 + }, + { + "epoch": 0.17, + "learning_rate": 1.8945758889442534e-05, + "loss": 1.1348, + "step": 5412 + }, + { + "epoch": 0.17, + "learning_rate": 1.8945295924124644e-05, + "loss": 1.1592, + "step": 5413 + }, + { + "epoch": 0.17, + "learning_rate": 1.894483286283363e-05, + "loss": 1.0386, + "step": 5414 + }, + { + "epoch": 0.17, + "learning_rate": 1.894436970557446e-05, + "loss": 0.5271, + "step": 5415 + }, + { + "epoch": 0.17, + "learning_rate": 1.8943906452352097e-05, + "loss": 1.1353, + "step": 5416 + }, + { + "epoch": 0.17, + "learning_rate": 1.894344310317152e-05, + "loss": 1.2236, + "step": 5417 + }, + { + "epoch": 0.17, + "learning_rate": 1.8942979658037695e-05, + "loss": 1.1221, + "step": 5418 + }, + { + "epoch": 0.17, + "learning_rate": 1.89425161169556e-05, + "loss": 1.1943, + "step": 5419 + }, + { + "epoch": 0.17, + "learning_rate": 1.89420524799302e-05, + "loss": 1.0503, + "step": 5420 + }, + { + "epoch": 0.17, + "learning_rate": 1.8941588746966473e-05, + "loss": 1.0605, + "step": 5421 + }, + { + "epoch": 0.17, + "learning_rate": 1.8941124918069395e-05, + "loss": 1.0068, + "step": 5422 + }, + { + "epoch": 0.17, + "learning_rate": 1.8940660993243943e-05, + "loss": 1.1533, + "step": 5423 + }, + { + "epoch": 0.17, + "learning_rate": 1.8940196972495096e-05, + "loss": 1.0791, + "step": 5424 + }, + { + "epoch": 0.17, + "learning_rate": 1.8939732855827824e-05, + "loss": 1.1528, + "step": 5425 + }, + { + "epoch": 0.17, + "learning_rate": 1.893926864324712e-05, + "loss": 1.1855, + "step": 5426 + }, + { + "epoch": 0.17, + "learning_rate": 1.893880433475795e-05, + "loss": 1.105, + "step": 5427 + }, + { + "epoch": 0.17, + "learning_rate": 1.8938339930365307e-05, + "loss": 1.0615, + "step": 5428 + }, + { + "epoch": 0.17, + "learning_rate": 1.8937875430074167e-05, + "loss": 1.1235, + "step": 5429 + }, + { + "epoch": 0.17, + "learning_rate": 1.8937410833889517e-05, + "loss": 1.0459, + "step": 5430 + }, + { + "epoch": 0.17, + "learning_rate": 1.8936946141816342e-05, + "loss": 1.1719, + "step": 5431 + }, + { + "epoch": 0.17, + "learning_rate": 1.8936481353859624e-05, + "loss": 1.0161, + "step": 5432 + }, + { + "epoch": 0.17, + "learning_rate": 1.893601647002435e-05, + "loss": 1.04, + "step": 5433 + }, + { + "epoch": 0.17, + "learning_rate": 1.893555149031551e-05, + "loss": 1.1431, + "step": 5434 + }, + { + "epoch": 0.17, + "learning_rate": 1.893508641473809e-05, + "loss": 1.1274, + "step": 5435 + }, + { + "epoch": 0.17, + "learning_rate": 1.8934621243297086e-05, + "loss": 1.0981, + "step": 5436 + }, + { + "epoch": 0.17, + "learning_rate": 1.893415597599748e-05, + "loss": 1.0737, + "step": 5437 + }, + { + "epoch": 0.17, + "learning_rate": 1.8933690612844272e-05, + "loss": 1.0825, + "step": 5438 + }, + { + "epoch": 0.17, + "learning_rate": 1.8933225153842446e-05, + "loss": 0.9805, + "step": 5439 + }, + { + "epoch": 0.17, + "learning_rate": 1.8932759598997005e-05, + "loss": 1.1211, + "step": 5440 + }, + { + "epoch": 0.17, + "learning_rate": 1.893229394831294e-05, + "loss": 0.9673, + "step": 5441 + }, + { + "epoch": 0.17, + "learning_rate": 1.8931828201795243e-05, + "loss": 1.063, + "step": 5442 + }, + { + "epoch": 0.17, + "learning_rate": 1.893136235944892e-05, + "loss": 1.1548, + "step": 5443 + }, + { + "epoch": 0.17, + "learning_rate": 1.8930896421278963e-05, + "loss": 1.1865, + "step": 5444 + }, + { + "epoch": 0.17, + "learning_rate": 1.8930430387290365e-05, + "loss": 1.1895, + "step": 5445 + }, + { + "epoch": 0.17, + "learning_rate": 1.8929964257488138e-05, + "loss": 1.0371, + "step": 5446 + }, + { + "epoch": 0.17, + "learning_rate": 1.892949803187728e-05, + "loss": 1.0024, + "step": 5447 + }, + { + "epoch": 0.17, + "learning_rate": 1.8929031710462785e-05, + "loss": 0.9702, + "step": 5448 + }, + { + "epoch": 0.17, + "learning_rate": 1.8928565293249665e-05, + "loss": 1.0967, + "step": 5449 + }, + { + "epoch": 0.17, + "learning_rate": 1.892809878024292e-05, + "loss": 1.1035, + "step": 5450 + }, + { + "epoch": 0.17, + "learning_rate": 1.8927632171447555e-05, + "loss": 1.0088, + "step": 5451 + }, + { + "epoch": 0.17, + "learning_rate": 1.8927165466868584e-05, + "loss": 1.0259, + "step": 5452 + }, + { + "epoch": 0.17, + "learning_rate": 1.8926698666511e-05, + "loss": 1.2041, + "step": 5453 + }, + { + "epoch": 0.17, + "learning_rate": 1.8926231770379823e-05, + "loss": 1.0703, + "step": 5454 + }, + { + "epoch": 0.17, + "learning_rate": 1.892576477848006e-05, + "loss": 1.0698, + "step": 5455 + }, + { + "epoch": 0.17, + "learning_rate": 1.8925297690816712e-05, + "loss": 1.0723, + "step": 5456 + }, + { + "epoch": 0.17, + "learning_rate": 1.8924830507394803e-05, + "loss": 1.1187, + "step": 5457 + }, + { + "epoch": 0.17, + "learning_rate": 1.892436322821934e-05, + "loss": 1.166, + "step": 5458 + }, + { + "epoch": 0.17, + "learning_rate": 1.8923895853295336e-05, + "loss": 1.1421, + "step": 5459 + }, + { + "epoch": 0.17, + "learning_rate": 1.8923428382627807e-05, + "loss": 1.166, + "step": 5460 + }, + { + "epoch": 0.17, + "learning_rate": 1.8922960816221763e-05, + "loss": 1.0664, + "step": 5461 + }, + { + "epoch": 0.17, + "learning_rate": 1.892249315408223e-05, + "loss": 0.9795, + "step": 5462 + }, + { + "epoch": 0.17, + "learning_rate": 1.8922025396214218e-05, + "loss": 1.127, + "step": 5463 + }, + { + "epoch": 0.17, + "learning_rate": 1.8921557542622748e-05, + "loss": 1.0781, + "step": 5464 + }, + { + "epoch": 0.17, + "learning_rate": 1.892108959331284e-05, + "loss": 1.125, + "step": 5465 + }, + { + "epoch": 0.17, + "learning_rate": 1.8920621548289516e-05, + "loss": 1.0806, + "step": 5466 + }, + { + "epoch": 0.17, + "learning_rate": 1.8920153407557794e-05, + "loss": 1.0859, + "step": 5467 + }, + { + "epoch": 0.17, + "learning_rate": 1.89196851711227e-05, + "loss": 1.0894, + "step": 5468 + }, + { + "epoch": 0.17, + "learning_rate": 1.8919216838989252e-05, + "loss": 1.0356, + "step": 5469 + }, + { + "epoch": 0.17, + "learning_rate": 1.8918748411162478e-05, + "loss": 1.0332, + "step": 5470 + }, + { + "epoch": 0.17, + "learning_rate": 1.891827988764741e-05, + "loss": 1.1587, + "step": 5471 + }, + { + "epoch": 0.17, + "learning_rate": 1.8917811268449065e-05, + "loss": 1.0981, + "step": 5472 + }, + { + "epoch": 0.18, + "learning_rate": 1.8917342553572475e-05, + "loss": 1.0718, + "step": 5473 + }, + { + "epoch": 0.18, + "learning_rate": 1.891687374302267e-05, + "loss": 1.1045, + "step": 5474 + }, + { + "epoch": 0.18, + "learning_rate": 1.8916404836804677e-05, + "loss": 1.0454, + "step": 5475 + }, + { + "epoch": 0.18, + "learning_rate": 1.891593583492353e-05, + "loss": 1.1753, + "step": 5476 + }, + { + "epoch": 0.18, + "learning_rate": 1.8915466737384257e-05, + "loss": 1.0635, + "step": 5477 + }, + { + "epoch": 0.18, + "learning_rate": 1.8914997544191895e-05, + "loss": 0.9717, + "step": 5478 + }, + { + "epoch": 0.18, + "learning_rate": 1.8914528255351475e-05, + "loss": 1.0508, + "step": 5479 + }, + { + "epoch": 0.18, + "learning_rate": 1.8914058870868032e-05, + "loss": 1.1587, + "step": 5480 + }, + { + "epoch": 0.18, + "learning_rate": 1.8913589390746607e-05, + "loss": 1.0903, + "step": 5481 + }, + { + "epoch": 0.18, + "learning_rate": 1.8913119814992227e-05, + "loss": 1.1113, + "step": 5482 + }, + { + "epoch": 0.18, + "learning_rate": 1.891265014360994e-05, + "loss": 1.0474, + "step": 5483 + }, + { + "epoch": 0.18, + "learning_rate": 1.8912180376604777e-05, + "loss": 1.0952, + "step": 5484 + }, + { + "epoch": 0.18, + "learning_rate": 1.8911710513981786e-05, + "loss": 1.0806, + "step": 5485 + }, + { + "epoch": 0.18, + "learning_rate": 1.8911240555746007e-05, + "loss": 1.1016, + "step": 5486 + }, + { + "epoch": 0.18, + "learning_rate": 1.8910770501902475e-05, + "loss": 1.1724, + "step": 5487 + }, + { + "epoch": 0.18, + "learning_rate": 1.8910300352456238e-05, + "loss": 1.0239, + "step": 5488 + }, + { + "epoch": 0.18, + "learning_rate": 1.8909830107412338e-05, + "loss": 1.1323, + "step": 5489 + }, + { + "epoch": 0.18, + "learning_rate": 1.8909359766775823e-05, + "loss": 1.1875, + "step": 5490 + }, + { + "epoch": 0.18, + "learning_rate": 1.890888933055174e-05, + "loss": 1.1445, + "step": 5491 + }, + { + "epoch": 0.18, + "learning_rate": 1.8908418798745133e-05, + "loss": 1.1055, + "step": 5492 + }, + { + "epoch": 0.18, + "learning_rate": 1.8907948171361054e-05, + "loss": 1.1187, + "step": 5493 + }, + { + "epoch": 0.18, + "learning_rate": 1.8907477448404546e-05, + "loss": 1.1211, + "step": 5494 + }, + { + "epoch": 0.18, + "learning_rate": 1.8907006629880662e-05, + "loss": 0.999, + "step": 5495 + }, + { + "epoch": 0.18, + "learning_rate": 1.890653571579446e-05, + "loss": 1.0288, + "step": 5496 + }, + { + "epoch": 0.18, + "learning_rate": 1.8906064706150985e-05, + "loss": 1.0454, + "step": 5497 + }, + { + "epoch": 0.18, + "learning_rate": 1.8905593600955292e-05, + "loss": 1.1104, + "step": 5498 + }, + { + "epoch": 0.18, + "learning_rate": 1.8905122400212435e-05, + "loss": 1.1245, + "step": 5499 + }, + { + "epoch": 0.18, + "learning_rate": 1.8904651103927468e-05, + "loss": 1.1597, + "step": 5500 + }, + { + "epoch": 0.18, + "learning_rate": 1.8904179712105458e-05, + "loss": 1.0977, + "step": 5501 + }, + { + "epoch": 0.18, + "learning_rate": 1.8903708224751446e-05, + "loss": 1.147, + "step": 5502 + }, + { + "epoch": 0.18, + "learning_rate": 1.8903236641870502e-05, + "loss": 1.0254, + "step": 5503 + }, + { + "epoch": 0.18, + "learning_rate": 1.8902764963467685e-05, + "loss": 1.0176, + "step": 5504 + }, + { + "epoch": 0.18, + "learning_rate": 1.890229318954805e-05, + "loss": 1.1499, + "step": 5505 + }, + { + "epoch": 0.18, + "learning_rate": 1.890182132011666e-05, + "loss": 1.1616, + "step": 5506 + }, + { + "epoch": 0.18, + "learning_rate": 1.8901349355178583e-05, + "loss": 1.1069, + "step": 5507 + }, + { + "epoch": 0.18, + "learning_rate": 1.890087729473888e-05, + "loss": 1.1797, + "step": 5508 + }, + { + "epoch": 0.18, + "learning_rate": 1.890040513880261e-05, + "loss": 1.0703, + "step": 5509 + }, + { + "epoch": 0.18, + "learning_rate": 1.889993288737485e-05, + "loss": 1.1133, + "step": 5510 + }, + { + "epoch": 0.18, + "learning_rate": 1.8899460540460653e-05, + "loss": 1.0571, + "step": 5511 + }, + { + "epoch": 0.18, + "learning_rate": 1.8898988098065097e-05, + "loss": 0.9829, + "step": 5512 + }, + { + "epoch": 0.18, + "learning_rate": 1.8898515560193246e-05, + "loss": 1.0747, + "step": 5513 + }, + { + "epoch": 0.18, + "learning_rate": 1.8898042926850172e-05, + "loss": 1.1123, + "step": 5514 + }, + { + "epoch": 0.18, + "learning_rate": 1.8897570198040943e-05, + "loss": 1.1523, + "step": 5515 + }, + { + "epoch": 0.18, + "learning_rate": 1.889709737377064e-05, + "loss": 1.0762, + "step": 5516 + }, + { + "epoch": 0.18, + "learning_rate": 1.8896624454044324e-05, + "loss": 0.981, + "step": 5517 + }, + { + "epoch": 0.18, + "learning_rate": 1.8896151438867075e-05, + "loss": 1.1294, + "step": 5518 + }, + { + "epoch": 0.18, + "learning_rate": 1.889567832824397e-05, + "loss": 1.1274, + "step": 5519 + }, + { + "epoch": 0.18, + "learning_rate": 1.8895205122180073e-05, + "loss": 1.147, + "step": 5520 + }, + { + "epoch": 0.18, + "learning_rate": 1.8894731820680477e-05, + "loss": 1.1055, + "step": 5521 + }, + { + "epoch": 0.18, + "learning_rate": 1.8894258423750253e-05, + "loss": 1.0278, + "step": 5522 + }, + { + "epoch": 0.18, + "learning_rate": 1.8893784931394476e-05, + "loss": 1.0518, + "step": 5523 + }, + { + "epoch": 0.18, + "learning_rate": 1.8893311343618232e-05, + "loss": 1.1108, + "step": 5524 + }, + { + "epoch": 0.18, + "learning_rate": 1.88928376604266e-05, + "loss": 0.9932, + "step": 5525 + }, + { + "epoch": 0.18, + "learning_rate": 1.889236388182466e-05, + "loss": 1.0967, + "step": 5526 + }, + { + "epoch": 0.18, + "learning_rate": 1.8891890007817498e-05, + "loss": 1.1382, + "step": 5527 + }, + { + "epoch": 0.18, + "learning_rate": 1.88914160384102e-05, + "loss": 1.041, + "step": 5528 + }, + { + "epoch": 0.18, + "learning_rate": 1.8890941973607843e-05, + "loss": 1.0698, + "step": 5529 + }, + { + "epoch": 0.18, + "learning_rate": 1.8890467813415525e-05, + "loss": 1.1045, + "step": 5530 + }, + { + "epoch": 0.18, + "learning_rate": 1.8889993557838323e-05, + "loss": 1.1567, + "step": 5531 + }, + { + "epoch": 0.18, + "learning_rate": 1.888951920688133e-05, + "loss": 1.1484, + "step": 5532 + }, + { + "epoch": 0.18, + "learning_rate": 1.8889044760549632e-05, + "loss": 1.0684, + "step": 5533 + }, + { + "epoch": 0.18, + "learning_rate": 1.8888570218848325e-05, + "loss": 0.9111, + "step": 5534 + }, + { + "epoch": 0.18, + "learning_rate": 1.88880955817825e-05, + "loss": 1.0034, + "step": 5535 + }, + { + "epoch": 0.18, + "learning_rate": 1.8887620849357236e-05, + "loss": 1.0161, + "step": 5536 + }, + { + "epoch": 0.18, + "learning_rate": 1.8887146021577646e-05, + "loss": 1.0781, + "step": 5537 + }, + { + "epoch": 0.18, + "learning_rate": 1.8886671098448814e-05, + "loss": 0.9893, + "step": 5538 + }, + { + "epoch": 0.18, + "learning_rate": 1.8886196079975832e-05, + "loss": 1.083, + "step": 5539 + }, + { + "epoch": 0.18, + "learning_rate": 1.8885720966163804e-05, + "loss": 1.1538, + "step": 5540 + }, + { + "epoch": 0.18, + "learning_rate": 1.8885245757017824e-05, + "loss": 1.0903, + "step": 5541 + }, + { + "epoch": 0.18, + "learning_rate": 1.8884770452542985e-05, + "loss": 0.9399, + "step": 5542 + }, + { + "epoch": 0.18, + "learning_rate": 1.88842950527444e-05, + "loss": 1.1841, + "step": 5543 + }, + { + "epoch": 0.18, + "learning_rate": 1.8883819557627155e-05, + "loss": 1.1001, + "step": 5544 + }, + { + "epoch": 0.18, + "learning_rate": 1.8883343967196365e-05, + "loss": 1.1216, + "step": 5545 + }, + { + "epoch": 0.18, + "learning_rate": 1.8882868281457123e-05, + "loss": 1.1484, + "step": 5546 + }, + { + "epoch": 0.18, + "learning_rate": 1.8882392500414535e-05, + "loss": 1.1392, + "step": 5547 + }, + { + "epoch": 0.18, + "learning_rate": 1.8881916624073706e-05, + "loss": 1.167, + "step": 5548 + }, + { + "epoch": 0.18, + "learning_rate": 1.8881440652439743e-05, + "loss": 1.0259, + "step": 5549 + }, + { + "epoch": 0.18, + "learning_rate": 1.8880964585517748e-05, + "loss": 1.1064, + "step": 5550 + }, + { + "epoch": 0.18, + "learning_rate": 1.8880488423312834e-05, + "loss": 1.1733, + "step": 5551 + }, + { + "epoch": 0.18, + "learning_rate": 1.888001216583011e-05, + "loss": 0.999, + "step": 5552 + }, + { + "epoch": 0.18, + "learning_rate": 1.887953581307468e-05, + "loss": 1.2095, + "step": 5553 + }, + { + "epoch": 0.18, + "learning_rate": 1.887905936505166e-05, + "loss": 1.0146, + "step": 5554 + }, + { + "epoch": 0.18, + "learning_rate": 1.887858282176616e-05, + "loss": 1.0571, + "step": 5555 + }, + { + "epoch": 0.18, + "learning_rate": 1.8878106183223294e-05, + "loss": 0.9888, + "step": 5556 + }, + { + "epoch": 0.18, + "learning_rate": 1.887762944942817e-05, + "loss": 1.1226, + "step": 5557 + }, + { + "epoch": 0.18, + "learning_rate": 1.887715262038591e-05, + "loss": 1.1963, + "step": 5558 + }, + { + "epoch": 0.18, + "learning_rate": 1.8876675696101627e-05, + "loss": 1.1294, + "step": 5559 + }, + { + "epoch": 0.18, + "learning_rate": 1.8876198676580437e-05, + "loss": 1.084, + "step": 5560 + }, + { + "epoch": 0.18, + "learning_rate": 1.887572156182746e-05, + "loss": 1.1421, + "step": 5561 + }, + { + "epoch": 0.18, + "learning_rate": 1.8875244351847815e-05, + "loss": 1.1089, + "step": 5562 + }, + { + "epoch": 0.18, + "learning_rate": 1.8874767046646617e-05, + "loss": 1.0957, + "step": 5563 + }, + { + "epoch": 0.18, + "learning_rate": 1.8874289646228994e-05, + "loss": 0.6021, + "step": 5564 + }, + { + "epoch": 0.18, + "learning_rate": 1.8873812150600064e-05, + "loss": 0.9629, + "step": 5565 + }, + { + "epoch": 0.18, + "learning_rate": 1.887333455976495e-05, + "loss": 1.1377, + "step": 5566 + }, + { + "epoch": 0.18, + "learning_rate": 1.8872856873728776e-05, + "loss": 1.1377, + "step": 5567 + }, + { + "epoch": 0.18, + "learning_rate": 1.887237909249667e-05, + "loss": 1.0234, + "step": 5568 + }, + { + "epoch": 0.18, + "learning_rate": 1.8871901216073755e-05, + "loss": 1.0322, + "step": 5569 + }, + { + "epoch": 0.18, + "learning_rate": 1.887142324446516e-05, + "loss": 1.1016, + "step": 5570 + }, + { + "epoch": 0.18, + "learning_rate": 1.887094517767601e-05, + "loss": 1.1621, + "step": 5571 + }, + { + "epoch": 0.18, + "learning_rate": 1.8870467015711433e-05, + "loss": 1.0615, + "step": 5572 + }, + { + "epoch": 0.18, + "learning_rate": 1.8869988758576566e-05, + "loss": 1.1553, + "step": 5573 + }, + { + "epoch": 0.18, + "learning_rate": 1.8869510406276538e-05, + "loss": 1.1211, + "step": 5574 + }, + { + "epoch": 0.18, + "learning_rate": 1.8869031958816478e-05, + "loss": 1.0244, + "step": 5575 + }, + { + "epoch": 0.18, + "learning_rate": 1.886855341620152e-05, + "loss": 0.9434, + "step": 5576 + }, + { + "epoch": 0.18, + "learning_rate": 1.88680747784368e-05, + "loss": 1.1436, + "step": 5577 + }, + { + "epoch": 0.18, + "learning_rate": 1.8867596045527453e-05, + "loss": 1.1011, + "step": 5578 + }, + { + "epoch": 0.18, + "learning_rate": 1.8867117217478614e-05, + "loss": 1.0273, + "step": 5579 + }, + { + "epoch": 0.18, + "learning_rate": 1.8866638294295418e-05, + "loss": 1.1841, + "step": 5580 + }, + { + "epoch": 0.18, + "learning_rate": 1.886615927598301e-05, + "loss": 1.1079, + "step": 5581 + }, + { + "epoch": 0.18, + "learning_rate": 1.8865680162546524e-05, + "loss": 1.0742, + "step": 5582 + }, + { + "epoch": 0.18, + "learning_rate": 1.8865200953991105e-05, + "loss": 0.958, + "step": 5583 + }, + { + "epoch": 0.18, + "learning_rate": 1.8864721650321885e-05, + "loss": 1.105, + "step": 5584 + }, + { + "epoch": 0.18, + "learning_rate": 1.8864242251544018e-05, + "loss": 1.0386, + "step": 5585 + }, + { + "epoch": 0.18, + "learning_rate": 1.886376275766264e-05, + "loss": 1.0781, + "step": 5586 + }, + { + "epoch": 0.18, + "learning_rate": 1.8863283168682898e-05, + "loss": 1.1528, + "step": 5587 + }, + { + "epoch": 0.18, + "learning_rate": 1.8862803484609938e-05, + "loss": 1.0259, + "step": 5588 + }, + { + "epoch": 0.18, + "learning_rate": 1.88623237054489e-05, + "loss": 1.1436, + "step": 5589 + }, + { + "epoch": 0.18, + "learning_rate": 1.886184383120494e-05, + "loss": 1.1431, + "step": 5590 + }, + { + "epoch": 0.18, + "learning_rate": 1.8861363861883208e-05, + "loss": 1.1064, + "step": 5591 + }, + { + "epoch": 0.18, + "learning_rate": 1.8860883797488846e-05, + "loss": 0.999, + "step": 5592 + }, + { + "epoch": 0.18, + "learning_rate": 1.8860403638027007e-05, + "loss": 1.1001, + "step": 5593 + }, + { + "epoch": 0.18, + "learning_rate": 1.8859923383502842e-05, + "loss": 1.0425, + "step": 5594 + }, + { + "epoch": 0.18, + "learning_rate": 1.885944303392151e-05, + "loss": 1.0454, + "step": 5595 + }, + { + "epoch": 0.18, + "learning_rate": 1.8858962589288153e-05, + "loss": 1.0225, + "step": 5596 + }, + { + "epoch": 0.18, + "learning_rate": 1.8858482049607935e-05, + "loss": 1.0625, + "step": 5597 + }, + { + "epoch": 0.18, + "learning_rate": 1.885800141488601e-05, + "loss": 1.0366, + "step": 5598 + }, + { + "epoch": 0.18, + "learning_rate": 1.885752068512753e-05, + "loss": 1.0659, + "step": 5599 + }, + { + "epoch": 0.18, + "learning_rate": 1.885703986033766e-05, + "loss": 1.0474, + "step": 5600 + }, + { + "epoch": 0.18, + "learning_rate": 1.8856558940521553e-05, + "loss": 1.1255, + "step": 5601 + }, + { + "epoch": 0.18, + "learning_rate": 1.8856077925684372e-05, + "loss": 1.0137, + "step": 5602 + }, + { + "epoch": 0.18, + "learning_rate": 1.885559681583127e-05, + "loss": 1.1543, + "step": 5603 + }, + { + "epoch": 0.18, + "learning_rate": 1.885511561096742e-05, + "loss": 1.1079, + "step": 5604 + }, + { + "epoch": 0.18, + "learning_rate": 1.885463431109798e-05, + "loss": 1.1602, + "step": 5605 + }, + { + "epoch": 0.18, + "learning_rate": 1.8854152916228114e-05, + "loss": 1.0757, + "step": 5606 + }, + { + "epoch": 0.18, + "learning_rate": 1.8853671426362985e-05, + "loss": 1.0532, + "step": 5607 + }, + { + "epoch": 0.18, + "learning_rate": 1.885318984150776e-05, + "loss": 1.0645, + "step": 5608 + }, + { + "epoch": 0.18, + "learning_rate": 1.8852708161667607e-05, + "loss": 0.9458, + "step": 5609 + }, + { + "epoch": 0.18, + "learning_rate": 1.8852226386847696e-05, + "loss": 0.8828, + "step": 5610 + }, + { + "epoch": 0.18, + "learning_rate": 1.885174451705319e-05, + "loss": 1.1367, + "step": 5611 + }, + { + "epoch": 0.18, + "learning_rate": 1.8851262552289263e-05, + "loss": 1.1694, + "step": 5612 + }, + { + "epoch": 0.18, + "learning_rate": 1.885078049256108e-05, + "loss": 1.1074, + "step": 5613 + }, + { + "epoch": 0.18, + "learning_rate": 1.8850298337873822e-05, + "loss": 1.0859, + "step": 5614 + }, + { + "epoch": 0.18, + "learning_rate": 1.8849816088232658e-05, + "loss": 1.1055, + "step": 5615 + }, + { + "epoch": 0.18, + "learning_rate": 1.8849333743642764e-05, + "loss": 1.1226, + "step": 5616 + }, + { + "epoch": 0.18, + "learning_rate": 1.8848851304109312e-05, + "loss": 1.123, + "step": 5617 + }, + { + "epoch": 0.18, + "learning_rate": 1.8848368769637476e-05, + "loss": 1.0454, + "step": 5618 + }, + { + "epoch": 0.18, + "learning_rate": 1.8847886140232438e-05, + "loss": 1.1738, + "step": 5619 + }, + { + "epoch": 0.18, + "learning_rate": 1.8847403415899374e-05, + "loss": 1.1592, + "step": 5620 + }, + { + "epoch": 0.18, + "learning_rate": 1.8846920596643464e-05, + "loss": 0.998, + "step": 5621 + }, + { + "epoch": 0.18, + "learning_rate": 1.8846437682469886e-05, + "loss": 1.0522, + "step": 5622 + }, + { + "epoch": 0.18, + "learning_rate": 1.884595467338382e-05, + "loss": 1.0811, + "step": 5623 + }, + { + "epoch": 0.18, + "learning_rate": 1.8845471569390453e-05, + "loss": 1.0908, + "step": 5624 + }, + { + "epoch": 0.18, + "learning_rate": 1.8844988370494966e-05, + "loss": 0.9961, + "step": 5625 + }, + { + "epoch": 0.18, + "learning_rate": 1.8844505076702543e-05, + "loss": 1.1401, + "step": 5626 + }, + { + "epoch": 0.18, + "learning_rate": 1.8844021688018366e-05, + "loss": 1.0947, + "step": 5627 + }, + { + "epoch": 0.18, + "learning_rate": 1.8843538204447628e-05, + "loss": 0.8652, + "step": 5628 + }, + { + "epoch": 0.18, + "learning_rate": 1.8843054625995515e-05, + "loss": 1.0156, + "step": 5629 + }, + { + "epoch": 0.18, + "learning_rate": 1.8842570952667208e-05, + "loss": 1.1294, + "step": 5630 + }, + { + "epoch": 0.18, + "learning_rate": 1.8842087184467904e-05, + "loss": 0.9033, + "step": 5631 + }, + { + "epoch": 0.18, + "learning_rate": 1.8841603321402787e-05, + "loss": 1.208, + "step": 5632 + }, + { + "epoch": 0.18, + "learning_rate": 1.8841119363477057e-05, + "loss": 1.1519, + "step": 5633 + }, + { + "epoch": 0.18, + "learning_rate": 1.8840635310695896e-05, + "loss": 0.9062, + "step": 5634 + }, + { + "epoch": 0.18, + "learning_rate": 1.8840151163064507e-05, + "loss": 1.1372, + "step": 5635 + }, + { + "epoch": 0.18, + "learning_rate": 1.8839666920588074e-05, + "loss": 1.0229, + "step": 5636 + }, + { + "epoch": 0.18, + "learning_rate": 1.8839182583271802e-05, + "loss": 1.0103, + "step": 5637 + }, + { + "epoch": 0.18, + "learning_rate": 1.8838698151120884e-05, + "loss": 1.0537, + "step": 5638 + }, + { + "epoch": 0.18, + "learning_rate": 1.8838213624140514e-05, + "loss": 1.1284, + "step": 5639 + }, + { + "epoch": 0.18, + "learning_rate": 1.88377290023359e-05, + "loss": 1.1313, + "step": 5640 + }, + { + "epoch": 0.18, + "learning_rate": 1.883724428571223e-05, + "loss": 1.1118, + "step": 5641 + }, + { + "epoch": 0.18, + "learning_rate": 1.883675947427471e-05, + "loss": 1.0049, + "step": 5642 + }, + { + "epoch": 0.18, + "learning_rate": 1.883627456802854e-05, + "loss": 0.9751, + "step": 5643 + }, + { + "epoch": 0.18, + "learning_rate": 1.8835789566978925e-05, + "loss": 1.0845, + "step": 5644 + }, + { + "epoch": 0.18, + "learning_rate": 1.8835304471131065e-05, + "loss": 0.9897, + "step": 5645 + }, + { + "epoch": 0.18, + "learning_rate": 1.883481928049017e-05, + "loss": 0.9971, + "step": 5646 + }, + { + "epoch": 0.18, + "learning_rate": 1.883433399506144e-05, + "loss": 0.9536, + "step": 5647 + }, + { + "epoch": 0.18, + "learning_rate": 1.8833848614850083e-05, + "loss": 0.9521, + "step": 5648 + }, + { + "epoch": 0.18, + "learning_rate": 1.8833363139861305e-05, + "loss": 1.0146, + "step": 5649 + }, + { + "epoch": 0.18, + "learning_rate": 1.883287757010032e-05, + "loss": 1.0576, + "step": 5650 + }, + { + "epoch": 0.18, + "learning_rate": 1.8832391905572334e-05, + "loss": 1.0249, + "step": 5651 + }, + { + "epoch": 0.18, + "learning_rate": 1.8831906146282558e-05, + "loss": 1.1807, + "step": 5652 + }, + { + "epoch": 0.18, + "learning_rate": 1.8831420292236205e-05, + "loss": 1.1191, + "step": 5653 + }, + { + "epoch": 0.18, + "learning_rate": 1.8830934343438483e-05, + "loss": 1.1055, + "step": 5654 + }, + { + "epoch": 0.18, + "learning_rate": 1.8830448299894613e-05, + "loss": 1.1538, + "step": 5655 + }, + { + "epoch": 0.18, + "learning_rate": 1.8829962161609804e-05, + "loss": 0.9849, + "step": 5656 + }, + { + "epoch": 0.18, + "learning_rate": 1.8829475928589272e-05, + "loss": 1.0776, + "step": 5657 + }, + { + "epoch": 0.18, + "learning_rate": 1.8828989600838236e-05, + "loss": 1.2148, + "step": 5658 + }, + { + "epoch": 0.18, + "learning_rate": 1.8828503178361912e-05, + "loss": 1.0928, + "step": 5659 + }, + { + "epoch": 0.18, + "learning_rate": 1.8828016661165523e-05, + "loss": 1.0605, + "step": 5660 + }, + { + "epoch": 0.18, + "learning_rate": 1.8827530049254282e-05, + "loss": 1.2402, + "step": 5661 + }, + { + "epoch": 0.18, + "learning_rate": 1.8827043342633415e-05, + "loss": 1.1489, + "step": 5662 + }, + { + "epoch": 0.18, + "learning_rate": 1.882655654130814e-05, + "loss": 0.9492, + "step": 5663 + }, + { + "epoch": 0.18, + "learning_rate": 1.882606964528369e-05, + "loss": 1.0923, + "step": 5664 + }, + { + "epoch": 0.18, + "learning_rate": 1.8825582654565275e-05, + "loss": 0.937, + "step": 5665 + }, + { + "epoch": 0.18, + "learning_rate": 1.8825095569158124e-05, + "loss": 0.9985, + "step": 5666 + }, + { + "epoch": 0.18, + "learning_rate": 1.882460838906747e-05, + "loss": 1.0967, + "step": 5667 + }, + { + "epoch": 0.18, + "learning_rate": 1.882412111429853e-05, + "loss": 0.5767, + "step": 5668 + }, + { + "epoch": 0.18, + "learning_rate": 1.882363374485654e-05, + "loss": 1.0942, + "step": 5669 + }, + { + "epoch": 0.18, + "learning_rate": 1.882314628074672e-05, + "loss": 1.2246, + "step": 5670 + }, + { + "epoch": 0.18, + "learning_rate": 1.8822658721974313e-05, + "loss": 1.0151, + "step": 5671 + }, + { + "epoch": 0.18, + "learning_rate": 1.8822171068544537e-05, + "loss": 1.0464, + "step": 5672 + }, + { + "epoch": 0.18, + "learning_rate": 1.882168332046263e-05, + "loss": 0.4968, + "step": 5673 + }, + { + "epoch": 0.18, + "learning_rate": 1.8821195477733824e-05, + "loss": 1.0908, + "step": 5674 + }, + { + "epoch": 0.18, + "learning_rate": 1.8820707540363354e-05, + "loss": 1.0938, + "step": 5675 + }, + { + "epoch": 0.18, + "learning_rate": 1.8820219508356454e-05, + "loss": 0.9995, + "step": 5676 + }, + { + "epoch": 0.18, + "learning_rate": 1.8819731381718363e-05, + "loss": 1.124, + "step": 5677 + }, + { + "epoch": 0.18, + "learning_rate": 1.8819243160454314e-05, + "loss": 1.0605, + "step": 5678 + }, + { + "epoch": 0.18, + "learning_rate": 1.8818754844569546e-05, + "loss": 1.1846, + "step": 5679 + }, + { + "epoch": 0.18, + "learning_rate": 1.8818266434069297e-05, + "loss": 1.1558, + "step": 5680 + }, + { + "epoch": 0.18, + "learning_rate": 1.8817777928958813e-05, + "loss": 1.0806, + "step": 5681 + }, + { + "epoch": 0.18, + "learning_rate": 1.8817289329243323e-05, + "loss": 1.0225, + "step": 5682 + }, + { + "epoch": 0.18, + "learning_rate": 1.8816800634928087e-05, + "loss": 1.2026, + "step": 5683 + }, + { + "epoch": 0.18, + "learning_rate": 1.8816311846018332e-05, + "loss": 1.0542, + "step": 5684 + }, + { + "epoch": 0.18, + "learning_rate": 1.881582296251931e-05, + "loss": 1.1182, + "step": 5685 + }, + { + "epoch": 0.18, + "learning_rate": 1.8815333984436263e-05, + "loss": 0.9995, + "step": 5686 + }, + { + "epoch": 0.18, + "learning_rate": 1.8814844911774442e-05, + "loss": 1.1299, + "step": 5687 + }, + { + "epoch": 0.18, + "learning_rate": 1.881435574453909e-05, + "loss": 1.2451, + "step": 5688 + }, + { + "epoch": 0.18, + "learning_rate": 1.8813866482735454e-05, + "loss": 0.9888, + "step": 5689 + }, + { + "epoch": 0.18, + "learning_rate": 1.8813377126368783e-05, + "loss": 1.0518, + "step": 5690 + }, + { + "epoch": 0.18, + "learning_rate": 1.8812887675444335e-05, + "loss": 1.0425, + "step": 5691 + }, + { + "epoch": 0.18, + "learning_rate": 1.8812398129967354e-05, + "loss": 1.1748, + "step": 5692 + }, + { + "epoch": 0.18, + "learning_rate": 1.8811908489943095e-05, + "loss": 1.0591, + "step": 5693 + }, + { + "epoch": 0.18, + "learning_rate": 1.8811418755376806e-05, + "loss": 1.0063, + "step": 5694 + }, + { + "epoch": 0.18, + "learning_rate": 1.881092892627375e-05, + "loss": 1.0454, + "step": 5695 + }, + { + "epoch": 0.18, + "learning_rate": 1.8810439002639177e-05, + "loss": 1.1045, + "step": 5696 + }, + { + "epoch": 0.18, + "learning_rate": 1.8809948984478343e-05, + "loss": 1.251, + "step": 5697 + }, + { + "epoch": 0.18, + "learning_rate": 1.880945887179651e-05, + "loss": 1.0503, + "step": 5698 + }, + { + "epoch": 0.18, + "learning_rate": 1.880896866459893e-05, + "loss": 1.0024, + "step": 5699 + }, + { + "epoch": 0.18, + "learning_rate": 1.8808478362890867e-05, + "loss": 1.0952, + "step": 5700 + }, + { + "epoch": 0.18, + "learning_rate": 1.8807987966677582e-05, + "loss": 1.1289, + "step": 5701 + }, + { + "epoch": 0.18, + "learning_rate": 1.880749747596433e-05, + "loss": 1.1753, + "step": 5702 + }, + { + "epoch": 0.18, + "learning_rate": 1.880700689075638e-05, + "loss": 1.1309, + "step": 5703 + }, + { + "epoch": 0.18, + "learning_rate": 1.8806516211058995e-05, + "loss": 1.1035, + "step": 5704 + }, + { + "epoch": 0.18, + "learning_rate": 1.8806025436877437e-05, + "loss": 1.1069, + "step": 5705 + }, + { + "epoch": 0.18, + "learning_rate": 1.8805534568216966e-05, + "loss": 1.0718, + "step": 5706 + }, + { + "epoch": 0.18, + "learning_rate": 1.8805043605082863e-05, + "loss": 1.1177, + "step": 5707 + }, + { + "epoch": 0.18, + "learning_rate": 1.880455254748038e-05, + "loss": 1.1172, + "step": 5708 + }, + { + "epoch": 0.18, + "learning_rate": 1.8804061395414795e-05, + "loss": 1.083, + "step": 5709 + }, + { + "epoch": 0.18, + "learning_rate": 1.8803570148891376e-05, + "loss": 1.1738, + "step": 5710 + }, + { + "epoch": 0.18, + "learning_rate": 1.8803078807915396e-05, + "loss": 0.9878, + "step": 5711 + }, + { + "epoch": 0.18, + "learning_rate": 1.8802587372492117e-05, + "loss": 1.0796, + "step": 5712 + }, + { + "epoch": 0.18, + "learning_rate": 1.880209584262682e-05, + "loss": 0.9678, + "step": 5713 + }, + { + "epoch": 0.18, + "learning_rate": 1.880160421832478e-05, + "loss": 1.0513, + "step": 5714 + }, + { + "epoch": 0.18, + "learning_rate": 1.8801112499591262e-05, + "loss": 1.1934, + "step": 5715 + }, + { + "epoch": 0.18, + "learning_rate": 1.880062068643155e-05, + "loss": 1.1162, + "step": 5716 + }, + { + "epoch": 0.18, + "learning_rate": 1.8800128778850918e-05, + "loss": 1.0464, + "step": 5717 + }, + { + "epoch": 0.18, + "learning_rate": 1.8799636776854645e-05, + "loss": 1.2632, + "step": 5718 + }, + { + "epoch": 0.18, + "learning_rate": 1.8799144680448005e-05, + "loss": 1.0249, + "step": 5719 + }, + { + "epoch": 0.18, + "learning_rate": 1.8798652489636288e-05, + "loss": 1.2061, + "step": 5720 + }, + { + "epoch": 0.18, + "learning_rate": 1.879816020442476e-05, + "loss": 1.0566, + "step": 5721 + }, + { + "epoch": 0.18, + "learning_rate": 1.8797667824818716e-05, + "loss": 1.1182, + "step": 5722 + }, + { + "epoch": 0.18, + "learning_rate": 1.879717535082343e-05, + "loss": 0.9585, + "step": 5723 + }, + { + "epoch": 0.18, + "learning_rate": 1.8796682782444188e-05, + "loss": 1.0015, + "step": 5724 + }, + { + "epoch": 0.18, + "learning_rate": 1.8796190119686275e-05, + "loss": 1.1875, + "step": 5725 + }, + { + "epoch": 0.18, + "learning_rate": 1.879569736255498e-05, + "loss": 1.1396, + "step": 5726 + }, + { + "epoch": 0.18, + "learning_rate": 1.8795204511055584e-05, + "loss": 1.0356, + "step": 5727 + }, + { + "epoch": 0.18, + "learning_rate": 1.8794711565193378e-05, + "loss": 1.0542, + "step": 5728 + }, + { + "epoch": 0.18, + "learning_rate": 1.879421852497365e-05, + "loss": 0.9448, + "step": 5729 + }, + { + "epoch": 0.18, + "learning_rate": 1.8793725390401695e-05, + "loss": 1.061, + "step": 5730 + }, + { + "epoch": 0.18, + "learning_rate": 1.8793232161482795e-05, + "loss": 1.0562, + "step": 5731 + }, + { + "epoch": 0.18, + "learning_rate": 1.8792738838222247e-05, + "loss": 1.0864, + "step": 5732 + }, + { + "epoch": 0.18, + "learning_rate": 1.8792245420625343e-05, + "loss": 0.9888, + "step": 5733 + }, + { + "epoch": 0.18, + "learning_rate": 1.8791751908697374e-05, + "loss": 1.0903, + "step": 5734 + }, + { + "epoch": 0.18, + "learning_rate": 1.8791258302443637e-05, + "loss": 0.9878, + "step": 5735 + }, + { + "epoch": 0.18, + "learning_rate": 1.879076460186943e-05, + "loss": 0.9561, + "step": 5736 + }, + { + "epoch": 0.18, + "learning_rate": 1.879027080698005e-05, + "loss": 1.1011, + "step": 5737 + }, + { + "epoch": 0.18, + "learning_rate": 1.878977691778079e-05, + "loss": 1.0938, + "step": 5738 + }, + { + "epoch": 0.18, + "learning_rate": 1.8789282934276952e-05, + "loss": 1.0674, + "step": 5739 + }, + { + "epoch": 0.18, + "learning_rate": 1.8788788856473833e-05, + "loss": 1.1968, + "step": 5740 + }, + { + "epoch": 0.18, + "learning_rate": 1.878829468437674e-05, + "loss": 1.0542, + "step": 5741 + }, + { + "epoch": 0.18, + "learning_rate": 1.8787800417990973e-05, + "loss": 1.0713, + "step": 5742 + }, + { + "epoch": 0.18, + "learning_rate": 1.8787306057321832e-05, + "loss": 1.2139, + "step": 5743 + }, + { + "epoch": 0.18, + "learning_rate": 1.878681160237462e-05, + "loss": 1.0137, + "step": 5744 + }, + { + "epoch": 0.18, + "learning_rate": 1.8786317053154644e-05, + "loss": 1.1782, + "step": 5745 + }, + { + "epoch": 0.18, + "learning_rate": 1.8785822409667213e-05, + "loss": 0.9692, + "step": 5746 + }, + { + "epoch": 0.18, + "learning_rate": 1.8785327671917632e-05, + "loss": 1.04, + "step": 5747 + }, + { + "epoch": 0.18, + "learning_rate": 1.8784832839911207e-05, + "loss": 1.1143, + "step": 5748 + }, + { + "epoch": 0.18, + "learning_rate": 1.8784337913653244e-05, + "loss": 1.0439, + "step": 5749 + }, + { + "epoch": 0.18, + "learning_rate": 1.8783842893149063e-05, + "loss": 1.0732, + "step": 5750 + }, + { + "epoch": 0.18, + "learning_rate": 1.8783347778403965e-05, + "loss": 1.0776, + "step": 5751 + }, + { + "epoch": 0.18, + "learning_rate": 1.878285256942327e-05, + "loss": 1.0073, + "step": 5752 + }, + { + "epoch": 0.18, + "learning_rate": 1.8782357266212285e-05, + "loss": 1.0918, + "step": 5753 + }, + { + "epoch": 0.18, + "learning_rate": 1.8781861868776328e-05, + "loss": 1.0864, + "step": 5754 + }, + { + "epoch": 0.18, + "learning_rate": 1.878136637712071e-05, + "loss": 0.897, + "step": 5755 + }, + { + "epoch": 0.18, + "learning_rate": 1.8780870791250752e-05, + "loss": 1.0752, + "step": 5756 + }, + { + "epoch": 0.18, + "learning_rate": 1.8780375111171765e-05, + "loss": 1.0537, + "step": 5757 + }, + { + "epoch": 0.18, + "learning_rate": 1.8779879336889078e-05, + "loss": 1.1123, + "step": 5758 + }, + { + "epoch": 0.18, + "learning_rate": 1.8779383468407995e-05, + "loss": 1.0044, + "step": 5759 + }, + { + "epoch": 0.18, + "learning_rate": 1.877888750573385e-05, + "loss": 1.0776, + "step": 5760 + }, + { + "epoch": 0.18, + "learning_rate": 1.8778391448871953e-05, + "loss": 1.083, + "step": 5761 + }, + { + "epoch": 0.18, + "learning_rate": 1.8777895297827635e-05, + "loss": 0.9595, + "step": 5762 + }, + { + "epoch": 0.18, + "learning_rate": 1.8777399052606213e-05, + "loss": 0.981, + "step": 5763 + }, + { + "epoch": 0.18, + "learning_rate": 1.8776902713213012e-05, + "loss": 1.1611, + "step": 5764 + }, + { + "epoch": 0.18, + "learning_rate": 1.877640627965336e-05, + "loss": 1.1035, + "step": 5765 + }, + { + "epoch": 0.18, + "learning_rate": 1.8775909751932583e-05, + "loss": 1.1865, + "step": 5766 + }, + { + "epoch": 0.18, + "learning_rate": 1.8775413130056006e-05, + "loss": 1.0396, + "step": 5767 + }, + { + "epoch": 0.18, + "learning_rate": 1.877491641402896e-05, + "loss": 1.0078, + "step": 5768 + }, + { + "epoch": 0.18, + "learning_rate": 1.877441960385677e-05, + "loss": 1.0737, + "step": 5769 + }, + { + "epoch": 0.18, + "learning_rate": 1.8773922699544772e-05, + "loss": 1.0903, + "step": 5770 + }, + { + "epoch": 0.18, + "learning_rate": 1.8773425701098292e-05, + "loss": 1.0801, + "step": 5771 + }, + { + "epoch": 0.18, + "learning_rate": 1.8772928608522663e-05, + "loss": 1.1279, + "step": 5772 + }, + { + "epoch": 0.18, + "learning_rate": 1.877243142182322e-05, + "loss": 0.9932, + "step": 5773 + }, + { + "epoch": 0.18, + "learning_rate": 1.87719341410053e-05, + "loss": 1.1099, + "step": 5774 + }, + { + "epoch": 0.18, + "learning_rate": 1.8771436766074232e-05, + "loss": 1.0029, + "step": 5775 + }, + { + "epoch": 0.18, + "learning_rate": 1.8770939297035356e-05, + "loss": 1.1694, + "step": 5776 + }, + { + "epoch": 0.18, + "learning_rate": 1.8770441733894006e-05, + "loss": 1.1694, + "step": 5777 + }, + { + "epoch": 0.18, + "learning_rate": 1.8769944076655525e-05, + "loss": 1.0249, + "step": 5778 + }, + { + "epoch": 0.18, + "learning_rate": 1.8769446325325253e-05, + "loss": 1.0513, + "step": 5779 + }, + { + "epoch": 0.18, + "learning_rate": 1.8768948479908523e-05, + "loss": 1.0396, + "step": 5780 + }, + { + "epoch": 0.18, + "learning_rate": 1.876845054041068e-05, + "loss": 1.1494, + "step": 5781 + }, + { + "epoch": 0.18, + "learning_rate": 1.876795250683707e-05, + "loss": 1.0996, + "step": 5782 + }, + { + "epoch": 0.18, + "learning_rate": 1.8767454379193032e-05, + "loss": 1.0444, + "step": 5783 + }, + { + "epoch": 0.18, + "learning_rate": 1.8766956157483914e-05, + "loss": 1.0747, + "step": 5784 + }, + { + "epoch": 0.19, + "learning_rate": 1.876645784171506e-05, + "loss": 0.5532, + "step": 5785 + }, + { + "epoch": 0.19, + "learning_rate": 1.876595943189181e-05, + "loss": 1.0552, + "step": 5786 + }, + { + "epoch": 0.19, + "learning_rate": 1.8765460928019525e-05, + "loss": 0.9971, + "step": 5787 + }, + { + "epoch": 0.19, + "learning_rate": 1.876496233010354e-05, + "loss": 1.0811, + "step": 5788 + }, + { + "epoch": 0.19, + "learning_rate": 1.876446363814921e-05, + "loss": 1.022, + "step": 5789 + }, + { + "epoch": 0.19, + "learning_rate": 1.8763964852161884e-05, + "loss": 1.0615, + "step": 5790 + }, + { + "epoch": 0.19, + "learning_rate": 1.876346597214692e-05, + "loss": 0.9951, + "step": 5791 + }, + { + "epoch": 0.19, + "learning_rate": 1.876296699810966e-05, + "loss": 1.0596, + "step": 5792 + }, + { + "epoch": 0.19, + "learning_rate": 1.8762467930055463e-05, + "loss": 1.0396, + "step": 5793 + }, + { + "epoch": 0.19, + "learning_rate": 1.8761968767989686e-05, + "loss": 1.0918, + "step": 5794 + }, + { + "epoch": 0.19, + "learning_rate": 1.876146951191768e-05, + "loss": 1.0, + "step": 5795 + }, + { + "epoch": 0.19, + "learning_rate": 1.87609701618448e-05, + "loss": 1.04, + "step": 5796 + }, + { + "epoch": 0.19, + "learning_rate": 1.876047071777641e-05, + "loss": 1.0479, + "step": 5797 + }, + { + "epoch": 0.19, + "learning_rate": 1.875997117971786e-05, + "loss": 1.0845, + "step": 5798 + }, + { + "epoch": 0.19, + "learning_rate": 1.875947154767452e-05, + "loss": 0.9421, + "step": 5799 + }, + { + "epoch": 0.19, + "learning_rate": 1.8758971821651742e-05, + "loss": 1.1343, + "step": 5800 + }, + { + "epoch": 0.19, + "learning_rate": 1.875847200165489e-05, + "loss": 1.0664, + "step": 5801 + }, + { + "epoch": 0.19, + "learning_rate": 1.875797208768933e-05, + "loss": 0.9922, + "step": 5802 + }, + { + "epoch": 0.19, + "learning_rate": 1.875747207976042e-05, + "loss": 1.104, + "step": 5803 + }, + { + "epoch": 0.19, + "learning_rate": 1.8756971977873524e-05, + "loss": 1.0527, + "step": 5804 + }, + { + "epoch": 0.19, + "learning_rate": 1.8756471782034014e-05, + "loss": 1.0122, + "step": 5805 + }, + { + "epoch": 0.19, + "learning_rate": 1.875597149224725e-05, + "loss": 1.0596, + "step": 5806 + }, + { + "epoch": 0.19, + "learning_rate": 1.875547110851861e-05, + "loss": 1.0522, + "step": 5807 + }, + { + "epoch": 0.19, + "learning_rate": 1.875497063085345e-05, + "loss": 1.0688, + "step": 5808 + }, + { + "epoch": 0.19, + "learning_rate": 1.875447005925714e-05, + "loss": 1.1377, + "step": 5809 + }, + { + "epoch": 0.19, + "learning_rate": 1.8753969393735062e-05, + "loss": 1.0972, + "step": 5810 + }, + { + "epoch": 0.19, + "learning_rate": 1.8753468634292577e-05, + "loss": 1.1104, + "step": 5811 + }, + { + "epoch": 0.19, + "learning_rate": 1.8752967780935065e-05, + "loss": 1.2227, + "step": 5812 + }, + { + "epoch": 0.19, + "learning_rate": 1.8752466833667893e-05, + "loss": 1.0571, + "step": 5813 + }, + { + "epoch": 0.19, + "learning_rate": 1.875196579249644e-05, + "loss": 1.1353, + "step": 5814 + }, + { + "epoch": 0.19, + "learning_rate": 1.875146465742608e-05, + "loss": 1.0967, + "step": 5815 + }, + { + "epoch": 0.19, + "learning_rate": 1.8750963428462187e-05, + "loss": 1.1396, + "step": 5816 + }, + { + "epoch": 0.19, + "learning_rate": 1.8750462105610147e-05, + "loss": 1.1401, + "step": 5817 + }, + { + "epoch": 0.19, + "learning_rate": 1.8749960688875327e-05, + "loss": 0.5459, + "step": 5818 + }, + { + "epoch": 0.19, + "learning_rate": 1.8749459178263117e-05, + "loss": 1.165, + "step": 5819 + }, + { + "epoch": 0.19, + "learning_rate": 1.8748957573778894e-05, + "loss": 1.1016, + "step": 5820 + }, + { + "epoch": 0.19, + "learning_rate": 1.8748455875428035e-05, + "loss": 1.144, + "step": 5821 + }, + { + "epoch": 0.19, + "learning_rate": 1.874795408321593e-05, + "loss": 1.1743, + "step": 5822 + }, + { + "epoch": 0.19, + "learning_rate": 1.874745219714796e-05, + "loss": 1.0698, + "step": 5823 + }, + { + "epoch": 0.19, + "learning_rate": 1.8746950217229506e-05, + "loss": 0.5181, + "step": 5824 + }, + { + "epoch": 0.19, + "learning_rate": 1.874644814346596e-05, + "loss": 1.1353, + "step": 5825 + }, + { + "epoch": 0.19, + "learning_rate": 1.8745945975862704e-05, + "loss": 1.0601, + "step": 5826 + }, + { + "epoch": 0.19, + "learning_rate": 1.8745443714425128e-05, + "loss": 1.0811, + "step": 5827 + }, + { + "epoch": 0.19, + "learning_rate": 1.874494135915862e-05, + "loss": 1.0449, + "step": 5828 + }, + { + "epoch": 0.19, + "learning_rate": 1.8744438910068567e-05, + "loss": 1.2471, + "step": 5829 + }, + { + "epoch": 0.19, + "learning_rate": 1.8743936367160363e-05, + "loss": 1.1948, + "step": 5830 + }, + { + "epoch": 0.19, + "learning_rate": 1.8743433730439405e-05, + "loss": 1.0923, + "step": 5831 + }, + { + "epoch": 0.19, + "learning_rate": 1.8742930999911073e-05, + "loss": 1.0923, + "step": 5832 + }, + { + "epoch": 0.19, + "learning_rate": 1.874242817558077e-05, + "loss": 1.1172, + "step": 5833 + }, + { + "epoch": 0.19, + "learning_rate": 1.874192525745389e-05, + "loss": 1.1567, + "step": 5834 + }, + { + "epoch": 0.19, + "learning_rate": 1.8741422245535824e-05, + "loss": 1.0581, + "step": 5835 + }, + { + "epoch": 0.19, + "learning_rate": 1.874091913983197e-05, + "loss": 1.1021, + "step": 5836 + }, + { + "epoch": 0.19, + "learning_rate": 1.8740415940347733e-05, + "loss": 1.1675, + "step": 5837 + }, + { + "epoch": 0.19, + "learning_rate": 1.8739912647088505e-05, + "loss": 1.0991, + "step": 5838 + }, + { + "epoch": 0.19, + "learning_rate": 1.8739409260059688e-05, + "loss": 0.9341, + "step": 5839 + }, + { + "epoch": 0.19, + "learning_rate": 1.8738905779266677e-05, + "loss": 1.001, + "step": 5840 + }, + { + "epoch": 0.19, + "learning_rate": 1.8738402204714882e-05, + "loss": 0.9873, + "step": 5841 + }, + { + "epoch": 0.19, + "learning_rate": 1.8737898536409706e-05, + "loss": 1.1641, + "step": 5842 + }, + { + "epoch": 0.19, + "learning_rate": 1.8737394774356543e-05, + "loss": 1.0645, + "step": 5843 + }, + { + "epoch": 0.19, + "learning_rate": 1.873689091856081e-05, + "loss": 0.531, + "step": 5844 + }, + { + "epoch": 0.19, + "learning_rate": 1.8736386969027898e-05, + "loss": 1.0171, + "step": 5845 + }, + { + "epoch": 0.19, + "learning_rate": 1.8735882925763232e-05, + "loss": 0.897, + "step": 5846 + }, + { + "epoch": 0.19, + "learning_rate": 1.873537878877221e-05, + "loss": 0.5332, + "step": 5847 + }, + { + "epoch": 0.19, + "learning_rate": 1.8734874558060235e-05, + "loss": 1.0947, + "step": 5848 + }, + { + "epoch": 0.19, + "learning_rate": 1.873437023363273e-05, + "loss": 0.9863, + "step": 5849 + }, + { + "epoch": 0.19, + "learning_rate": 1.8733865815495093e-05, + "loss": 1.1831, + "step": 5850 + }, + { + "epoch": 0.19, + "learning_rate": 1.8733361303652746e-05, + "loss": 1.1089, + "step": 5851 + }, + { + "epoch": 0.19, + "learning_rate": 1.8732856698111095e-05, + "loss": 1.2119, + "step": 5852 + }, + { + "epoch": 0.19, + "learning_rate": 1.873235199887556e-05, + "loss": 1.2109, + "step": 5853 + }, + { + "epoch": 0.19, + "learning_rate": 1.873184720595155e-05, + "loss": 1.1465, + "step": 5854 + }, + { + "epoch": 0.19, + "learning_rate": 1.8731342319344485e-05, + "loss": 0.5144, + "step": 5855 + }, + { + "epoch": 0.19, + "learning_rate": 1.873083733905978e-05, + "loss": 1.1743, + "step": 5856 + }, + { + "epoch": 0.19, + "learning_rate": 1.8730332265102854e-05, + "loss": 1.1431, + "step": 5857 + }, + { + "epoch": 0.19, + "learning_rate": 1.8729827097479122e-05, + "loss": 1.1689, + "step": 5858 + }, + { + "epoch": 0.19, + "learning_rate": 1.872932183619401e-05, + "loss": 0.9873, + "step": 5859 + }, + { + "epoch": 0.19, + "learning_rate": 1.8728816481252936e-05, + "loss": 1.0269, + "step": 5860 + }, + { + "epoch": 0.19, + "learning_rate": 1.8728311032661318e-05, + "loss": 0.9966, + "step": 5861 + }, + { + "epoch": 0.19, + "learning_rate": 1.8727805490424586e-05, + "loss": 1.1265, + "step": 5862 + }, + { + "epoch": 0.19, + "learning_rate": 1.872729985454816e-05, + "loss": 1.1191, + "step": 5863 + }, + { + "epoch": 0.19, + "learning_rate": 1.8726794125037467e-05, + "loss": 0.9832, + "step": 5864 + }, + { + "epoch": 0.19, + "learning_rate": 1.872628830189793e-05, + "loss": 1.1514, + "step": 5865 + }, + { + "epoch": 0.19, + "learning_rate": 1.8725782385134975e-05, + "loss": 1.0449, + "step": 5866 + }, + { + "epoch": 0.19, + "learning_rate": 1.8725276374754033e-05, + "loss": 1.0806, + "step": 5867 + }, + { + "epoch": 0.19, + "learning_rate": 1.872477027076053e-05, + "loss": 0.4871, + "step": 5868 + }, + { + "epoch": 0.19, + "learning_rate": 1.8724264073159904e-05, + "loss": 1.0493, + "step": 5869 + }, + { + "epoch": 0.19, + "learning_rate": 1.8723757781957575e-05, + "loss": 1.0405, + "step": 5870 + }, + { + "epoch": 0.19, + "learning_rate": 1.8723251397158987e-05, + "loss": 1.0269, + "step": 5871 + }, + { + "epoch": 0.19, + "learning_rate": 1.8722744918769557e-05, + "loss": 1.2827, + "step": 5872 + }, + { + "epoch": 0.19, + "learning_rate": 1.8722238346794735e-05, + "loss": 1.0938, + "step": 5873 + }, + { + "epoch": 0.19, + "learning_rate": 1.8721731681239945e-05, + "loss": 1.103, + "step": 5874 + }, + { + "epoch": 0.19, + "learning_rate": 1.872122492211063e-05, + "loss": 1.1592, + "step": 5875 + }, + { + "epoch": 0.19, + "learning_rate": 1.872071806941222e-05, + "loss": 1.1362, + "step": 5876 + }, + { + "epoch": 0.19, + "learning_rate": 1.872021112315016e-05, + "loss": 1.103, + "step": 5877 + }, + { + "epoch": 0.19, + "learning_rate": 1.8719704083329882e-05, + "loss": 1.0391, + "step": 5878 + }, + { + "epoch": 0.19, + "learning_rate": 1.871919694995683e-05, + "loss": 1.0679, + "step": 5879 + }, + { + "epoch": 0.19, + "learning_rate": 1.871868972303645e-05, + "loss": 1.1172, + "step": 5880 + }, + { + "epoch": 0.19, + "learning_rate": 1.871818240257417e-05, + "loss": 1.0723, + "step": 5881 + }, + { + "epoch": 0.19, + "learning_rate": 1.871767498857545e-05, + "loss": 1.0474, + "step": 5882 + }, + { + "epoch": 0.19, + "learning_rate": 1.871716748104572e-05, + "loss": 0.9805, + "step": 5883 + }, + { + "epoch": 0.19, + "learning_rate": 1.8716659879990434e-05, + "loss": 0.9849, + "step": 5884 + }, + { + "epoch": 0.19, + "learning_rate": 1.8716152185415032e-05, + "loss": 0.9917, + "step": 5885 + }, + { + "epoch": 0.19, + "learning_rate": 1.8715644397324966e-05, + "loss": 1.1021, + "step": 5886 + }, + { + "epoch": 0.19, + "learning_rate": 1.8715136515725678e-05, + "loss": 0.491, + "step": 5887 + }, + { + "epoch": 0.19, + "learning_rate": 1.8714628540622626e-05, + "loss": 1.1099, + "step": 5888 + }, + { + "epoch": 0.19, + "learning_rate": 1.8714120472021252e-05, + "loss": 1.0171, + "step": 5889 + }, + { + "epoch": 0.19, + "learning_rate": 1.871361230992701e-05, + "loss": 1.0435, + "step": 5890 + }, + { + "epoch": 0.19, + "learning_rate": 1.871310405434535e-05, + "loss": 0.978, + "step": 5891 + }, + { + "epoch": 0.19, + "learning_rate": 1.8712595705281727e-05, + "loss": 1.1323, + "step": 5892 + }, + { + "epoch": 0.19, + "learning_rate": 1.8712087262741595e-05, + "loss": 1.0977, + "step": 5893 + }, + { + "epoch": 0.19, + "learning_rate": 1.8711578726730408e-05, + "loss": 1.1069, + "step": 5894 + }, + { + "epoch": 0.19, + "learning_rate": 1.8711070097253625e-05, + "loss": 1.1582, + "step": 5895 + }, + { + "epoch": 0.19, + "learning_rate": 1.87105613743167e-05, + "loss": 1.0767, + "step": 5896 + }, + { + "epoch": 0.19, + "learning_rate": 1.8710052557925093e-05, + "loss": 1.0732, + "step": 5897 + }, + { + "epoch": 0.19, + "learning_rate": 1.870954364808426e-05, + "loss": 0.9858, + "step": 5898 + }, + { + "epoch": 0.19, + "learning_rate": 1.8709034644799663e-05, + "loss": 1.1465, + "step": 5899 + }, + { + "epoch": 0.19, + "learning_rate": 1.8708525548076762e-05, + "loss": 1.1094, + "step": 5900 + }, + { + "epoch": 0.19, + "learning_rate": 1.870801635792102e-05, + "loss": 1.1191, + "step": 5901 + }, + { + "epoch": 0.19, + "learning_rate": 1.8707507074337904e-05, + "loss": 1.1113, + "step": 5902 + }, + { + "epoch": 0.19, + "learning_rate": 1.870699769733287e-05, + "loss": 1.0073, + "step": 5903 + }, + { + "epoch": 0.19, + "learning_rate": 1.870648822691139e-05, + "loss": 1.0361, + "step": 5904 + }, + { + "epoch": 0.19, + "learning_rate": 1.8705978663078926e-05, + "loss": 1.0728, + "step": 5905 + }, + { + "epoch": 0.19, + "learning_rate": 1.8705469005840947e-05, + "loss": 0.9624, + "step": 5906 + }, + { + "epoch": 0.19, + "learning_rate": 1.870495925520292e-05, + "loss": 1.1187, + "step": 5907 + }, + { + "epoch": 0.19, + "learning_rate": 1.8704449411170312e-05, + "loss": 1.0718, + "step": 5908 + }, + { + "epoch": 0.19, + "learning_rate": 1.8703939473748596e-05, + "loss": 1.1006, + "step": 5909 + }, + { + "epoch": 0.19, + "learning_rate": 1.8703429442943243e-05, + "loss": 1.127, + "step": 5910 + }, + { + "epoch": 0.19, + "learning_rate": 1.8702919318759723e-05, + "loss": 1.1729, + "step": 5911 + }, + { + "epoch": 0.19, + "learning_rate": 1.8702409101203517e-05, + "loss": 1.1704, + "step": 5912 + }, + { + "epoch": 0.19, + "learning_rate": 1.8701898790280086e-05, + "loss": 1.1333, + "step": 5913 + }, + { + "epoch": 0.19, + "learning_rate": 1.8701388385994914e-05, + "loss": 1.084, + "step": 5914 + }, + { + "epoch": 0.19, + "learning_rate": 1.8700877888353477e-05, + "loss": 1.1099, + "step": 5915 + }, + { + "epoch": 0.19, + "learning_rate": 1.8700367297361248e-05, + "loss": 1.0596, + "step": 5916 + }, + { + "epoch": 0.19, + "learning_rate": 1.8699856613023707e-05, + "loss": 1.1504, + "step": 5917 + }, + { + "epoch": 0.19, + "learning_rate": 1.869934583534633e-05, + "loss": 0.9146, + "step": 5918 + }, + { + "epoch": 0.19, + "learning_rate": 1.8698834964334604e-05, + "loss": 1.1069, + "step": 5919 + }, + { + "epoch": 0.19, + "learning_rate": 1.8698323999994006e-05, + "loss": 1.0596, + "step": 5920 + }, + { + "epoch": 0.19, + "learning_rate": 1.8697812942330017e-05, + "loss": 0.9492, + "step": 5921 + }, + { + "epoch": 0.19, + "learning_rate": 1.8697301791348125e-05, + "loss": 1.126, + "step": 5922 + }, + { + "epoch": 0.19, + "learning_rate": 1.8696790547053806e-05, + "loss": 1.1167, + "step": 5923 + }, + { + "epoch": 0.19, + "learning_rate": 1.8696279209452554e-05, + "loss": 0.9897, + "step": 5924 + }, + { + "epoch": 0.19, + "learning_rate": 1.8695767778549843e-05, + "loss": 1.1543, + "step": 5925 + }, + { + "epoch": 0.19, + "learning_rate": 1.8695256254351173e-05, + "loss": 1.0562, + "step": 5926 + }, + { + "epoch": 0.19, + "learning_rate": 1.869474463686203e-05, + "loss": 1.0093, + "step": 5927 + }, + { + "epoch": 0.19, + "learning_rate": 1.8694232926087893e-05, + "loss": 1.1724, + "step": 5928 + }, + { + "epoch": 0.19, + "learning_rate": 1.8693721122034264e-05, + "loss": 1.1577, + "step": 5929 + }, + { + "epoch": 0.19, + "learning_rate": 1.8693209224706627e-05, + "loss": 1.0352, + "step": 5930 + }, + { + "epoch": 0.19, + "learning_rate": 1.8692697234110474e-05, + "loss": 1.0674, + "step": 5931 + }, + { + "epoch": 0.19, + "learning_rate": 1.8692185150251305e-05, + "loss": 1.0767, + "step": 5932 + }, + { + "epoch": 0.19, + "learning_rate": 1.8691672973134607e-05, + "loss": 1.1357, + "step": 5933 + }, + { + "epoch": 0.19, + "learning_rate": 1.8691160702765878e-05, + "loss": 1.064, + "step": 5934 + }, + { + "epoch": 0.19, + "learning_rate": 1.869064833915061e-05, + "loss": 1.1221, + "step": 5935 + }, + { + "epoch": 0.19, + "learning_rate": 1.869013588229431e-05, + "loss": 1.2026, + "step": 5936 + }, + { + "epoch": 0.19, + "learning_rate": 1.8689623332202464e-05, + "loss": 1.0723, + "step": 5937 + }, + { + "epoch": 0.19, + "learning_rate": 1.868911068888058e-05, + "loss": 1.0498, + "step": 5938 + }, + { + "epoch": 0.19, + "learning_rate": 1.8688597952334156e-05, + "loss": 1.0884, + "step": 5939 + }, + { + "epoch": 0.19, + "learning_rate": 1.868808512256869e-05, + "loss": 1.0317, + "step": 5940 + }, + { + "epoch": 0.19, + "learning_rate": 1.8687572199589684e-05, + "loss": 1.144, + "step": 5941 + }, + { + "epoch": 0.19, + "learning_rate": 1.8687059183402647e-05, + "loss": 1.1055, + "step": 5942 + }, + { + "epoch": 0.19, + "learning_rate": 1.868654607401308e-05, + "loss": 1.0234, + "step": 5943 + }, + { + "epoch": 0.19, + "learning_rate": 1.8686032871426484e-05, + "loss": 1.1074, + "step": 5944 + }, + { + "epoch": 0.19, + "learning_rate": 1.868551957564837e-05, + "loss": 1.1216, + "step": 5945 + }, + { + "epoch": 0.19, + "learning_rate": 1.8685006186684243e-05, + "loss": 1.1768, + "step": 5946 + }, + { + "epoch": 0.19, + "learning_rate": 1.8684492704539615e-05, + "loss": 0.9971, + "step": 5947 + }, + { + "epoch": 0.19, + "learning_rate": 1.8683979129219987e-05, + "loss": 1.0723, + "step": 5948 + }, + { + "epoch": 0.19, + "learning_rate": 1.8683465460730876e-05, + "loss": 1.0264, + "step": 5949 + }, + { + "epoch": 0.19, + "learning_rate": 1.8682951699077792e-05, + "loss": 1.1528, + "step": 5950 + }, + { + "epoch": 0.19, + "learning_rate": 1.8682437844266243e-05, + "loss": 1.1504, + "step": 5951 + }, + { + "epoch": 0.19, + "learning_rate": 1.868192389630175e-05, + "loss": 1.104, + "step": 5952 + }, + { + "epoch": 0.19, + "learning_rate": 1.868140985518982e-05, + "loss": 0.9458, + "step": 5953 + }, + { + "epoch": 0.19, + "learning_rate": 1.8680895720935968e-05, + "loss": 1.0444, + "step": 5954 + }, + { + "epoch": 0.19, + "learning_rate": 1.8680381493545715e-05, + "loss": 0.5315, + "step": 5955 + }, + { + "epoch": 0.19, + "learning_rate": 1.8679867173024576e-05, + "loss": 1.2153, + "step": 5956 + }, + { + "epoch": 0.19, + "learning_rate": 1.8679352759378066e-05, + "loss": 1.0884, + "step": 5957 + }, + { + "epoch": 0.19, + "learning_rate": 1.867883825261171e-05, + "loss": 1.0757, + "step": 5958 + }, + { + "epoch": 0.19, + "learning_rate": 1.867832365273102e-05, + "loss": 1.0879, + "step": 5959 + }, + { + "epoch": 0.19, + "learning_rate": 1.8677808959741524e-05, + "loss": 1.1021, + "step": 5960 + }, + { + "epoch": 0.19, + "learning_rate": 1.8677294173648746e-05, + "loss": 1.1123, + "step": 5961 + }, + { + "epoch": 0.19, + "learning_rate": 1.86767792944582e-05, + "loss": 1.0693, + "step": 5962 + }, + { + "epoch": 0.19, + "learning_rate": 1.867626432217542e-05, + "loss": 1.0923, + "step": 5963 + }, + { + "epoch": 0.19, + "learning_rate": 1.867574925680592e-05, + "loss": 1.1167, + "step": 5964 + }, + { + "epoch": 0.19, + "learning_rate": 1.8675234098355236e-05, + "loss": 1.2007, + "step": 5965 + }, + { + "epoch": 0.19, + "learning_rate": 1.8674718846828896e-05, + "loss": 1.0425, + "step": 5966 + }, + { + "epoch": 0.19, + "learning_rate": 1.8674203502232414e-05, + "loss": 1.041, + "step": 5967 + }, + { + "epoch": 0.19, + "learning_rate": 1.8673688064571335e-05, + "loss": 1.0591, + "step": 5968 + }, + { + "epoch": 0.19, + "learning_rate": 1.867317253385118e-05, + "loss": 1.1626, + "step": 5969 + }, + { + "epoch": 0.19, + "learning_rate": 1.8672656910077485e-05, + "loss": 1.1802, + "step": 5970 + }, + { + "epoch": 0.19, + "learning_rate": 1.8672141193255777e-05, + "loss": 1.1416, + "step": 5971 + }, + { + "epoch": 0.19, + "learning_rate": 1.8671625383391596e-05, + "loss": 1.0527, + "step": 5972 + }, + { + "epoch": 0.19, + "learning_rate": 1.8671109480490467e-05, + "loss": 1.166, + "step": 5973 + }, + { + "epoch": 0.19, + "learning_rate": 1.867059348455793e-05, + "loss": 1.1597, + "step": 5974 + }, + { + "epoch": 0.19, + "learning_rate": 1.8670077395599528e-05, + "loss": 0.9751, + "step": 5975 + }, + { + "epoch": 0.19, + "learning_rate": 1.8669561213620783e-05, + "loss": 1.0894, + "step": 5976 + }, + { + "epoch": 0.19, + "learning_rate": 1.8669044938627245e-05, + "loss": 1.0303, + "step": 5977 + }, + { + "epoch": 0.19, + "learning_rate": 1.866852857062445e-05, + "loss": 1.0874, + "step": 5978 + }, + { + "epoch": 0.19, + "learning_rate": 1.8668012109617933e-05, + "loss": 1.0889, + "step": 5979 + }, + { + "epoch": 0.19, + "learning_rate": 1.866749555561324e-05, + "loss": 1.0264, + "step": 5980 + }, + { + "epoch": 0.19, + "learning_rate": 1.8666978908615914e-05, + "loss": 1.1758, + "step": 5981 + }, + { + "epoch": 0.19, + "learning_rate": 1.86664621686315e-05, + "loss": 1.1182, + "step": 5982 + }, + { + "epoch": 0.19, + "learning_rate": 1.8665945335665537e-05, + "loss": 1.0713, + "step": 5983 + }, + { + "epoch": 0.19, + "learning_rate": 1.8665428409723568e-05, + "loss": 1.0005, + "step": 5984 + }, + { + "epoch": 0.19, + "learning_rate": 1.8664911390811147e-05, + "loss": 1.0654, + "step": 5985 + }, + { + "epoch": 0.19, + "learning_rate": 1.866439427893381e-05, + "loss": 1.1782, + "step": 5986 + }, + { + "epoch": 0.19, + "learning_rate": 1.8663877074097117e-05, + "loss": 1.1274, + "step": 5987 + }, + { + "epoch": 0.19, + "learning_rate": 1.8663359776306614e-05, + "loss": 1.1226, + "step": 5988 + }, + { + "epoch": 0.19, + "learning_rate": 1.8662842385567844e-05, + "loss": 1.0674, + "step": 5989 + }, + { + "epoch": 0.19, + "learning_rate": 1.8662324901886365e-05, + "loss": 1.0137, + "step": 5990 + }, + { + "epoch": 0.19, + "learning_rate": 1.8661807325267727e-05, + "loss": 0.9829, + "step": 5991 + }, + { + "epoch": 0.19, + "learning_rate": 1.866128965571748e-05, + "loss": 1.0498, + "step": 5992 + }, + { + "epoch": 0.19, + "learning_rate": 1.8660771893241183e-05, + "loss": 0.9922, + "step": 5993 + }, + { + "epoch": 0.19, + "learning_rate": 1.866025403784439e-05, + "loss": 1.063, + "step": 5994 + }, + { + "epoch": 0.19, + "learning_rate": 1.8659736089532652e-05, + "loss": 0.9932, + "step": 5995 + }, + { + "epoch": 0.19, + "learning_rate": 1.8659218048311532e-05, + "loss": 1.104, + "step": 5996 + }, + { + "epoch": 0.19, + "learning_rate": 1.8658699914186585e-05, + "loss": 0.9478, + "step": 5997 + }, + { + "epoch": 0.19, + "learning_rate": 1.8658181687163375e-05, + "loss": 1.0757, + "step": 5998 + }, + { + "epoch": 0.19, + "learning_rate": 1.8657663367247453e-05, + "loss": 0.9307, + "step": 5999 + }, + { + "epoch": 0.19, + "learning_rate": 1.8657144954444387e-05, + "loss": 1.1704, + "step": 6000 + }, + { + "epoch": 0.19, + "learning_rate": 1.8656626448759737e-05, + "loss": 1.0864, + "step": 6001 + }, + { + "epoch": 0.19, + "learning_rate": 1.8656107850199067e-05, + "loss": 1.0811, + "step": 6002 + }, + { + "epoch": 0.19, + "learning_rate": 1.8655589158767938e-05, + "loss": 1.1562, + "step": 6003 + }, + { + "epoch": 0.19, + "learning_rate": 1.865507037447192e-05, + "loss": 1.0938, + "step": 6004 + }, + { + "epoch": 0.19, + "learning_rate": 1.865455149731657e-05, + "loss": 0.5125, + "step": 6005 + }, + { + "epoch": 0.19, + "learning_rate": 1.8654032527307465e-05, + "loss": 1.0674, + "step": 6006 + }, + { + "epoch": 0.19, + "learning_rate": 1.8653513464450167e-05, + "loss": 1.0586, + "step": 6007 + }, + { + "epoch": 0.19, + "learning_rate": 1.8652994308750244e-05, + "loss": 0.999, + "step": 6008 + }, + { + "epoch": 0.19, + "learning_rate": 1.8652475060213275e-05, + "loss": 1.0659, + "step": 6009 + }, + { + "epoch": 0.19, + "learning_rate": 1.865195571884482e-05, + "loss": 1.1387, + "step": 6010 + }, + { + "epoch": 0.19, + "learning_rate": 1.865143628465046e-05, + "loss": 1.0454, + "step": 6011 + }, + { + "epoch": 0.19, + "learning_rate": 1.865091675763576e-05, + "loss": 1.0459, + "step": 6012 + }, + { + "epoch": 0.19, + "learning_rate": 1.86503971378063e-05, + "loss": 1.1177, + "step": 6013 + }, + { + "epoch": 0.19, + "learning_rate": 1.8649877425167652e-05, + "loss": 1.0884, + "step": 6014 + }, + { + "epoch": 0.19, + "learning_rate": 1.864935761972539e-05, + "loss": 1.0894, + "step": 6015 + }, + { + "epoch": 0.19, + "learning_rate": 1.8648837721485095e-05, + "loss": 1.1655, + "step": 6016 + }, + { + "epoch": 0.19, + "learning_rate": 1.8648317730452342e-05, + "loss": 1.0303, + "step": 6017 + }, + { + "epoch": 0.19, + "learning_rate": 1.8647797646632714e-05, + "loss": 1.0303, + "step": 6018 + }, + { + "epoch": 0.19, + "learning_rate": 1.8647277470031787e-05, + "loss": 1.0425, + "step": 6019 + }, + { + "epoch": 0.19, + "learning_rate": 1.864675720065514e-05, + "loss": 1.0591, + "step": 6020 + }, + { + "epoch": 0.19, + "learning_rate": 1.8646236838508363e-05, + "loss": 1.0215, + "step": 6021 + }, + { + "epoch": 0.19, + "learning_rate": 1.864571638359703e-05, + "loss": 1.0337, + "step": 6022 + }, + { + "epoch": 0.19, + "learning_rate": 1.864519583592673e-05, + "loss": 1.0767, + "step": 6023 + }, + { + "epoch": 0.19, + "learning_rate": 1.864467519550305e-05, + "loss": 1.0264, + "step": 6024 + }, + { + "epoch": 0.19, + "learning_rate": 1.864415446233157e-05, + "loss": 1.0488, + "step": 6025 + }, + { + "epoch": 0.19, + "learning_rate": 1.864363363641788e-05, + "loss": 1.1196, + "step": 6026 + }, + { + "epoch": 0.19, + "learning_rate": 1.8643112717767565e-05, + "loss": 1.019, + "step": 6027 + }, + { + "epoch": 0.19, + "learning_rate": 1.864259170638622e-05, + "loss": 1.0327, + "step": 6028 + }, + { + "epoch": 0.19, + "learning_rate": 1.8642070602279427e-05, + "loss": 1.1709, + "step": 6029 + }, + { + "epoch": 0.19, + "learning_rate": 1.8641549405452783e-05, + "loss": 1.1201, + "step": 6030 + }, + { + "epoch": 0.19, + "learning_rate": 1.8641028115911878e-05, + "loss": 1.0962, + "step": 6031 + }, + { + "epoch": 0.19, + "learning_rate": 1.8640506733662305e-05, + "loss": 1.0654, + "step": 6032 + }, + { + "epoch": 0.19, + "learning_rate": 1.8639985258709656e-05, + "loss": 1.209, + "step": 6033 + }, + { + "epoch": 0.19, + "learning_rate": 1.863946369105953e-05, + "loss": 1.1255, + "step": 6034 + }, + { + "epoch": 0.19, + "learning_rate": 1.8638942030717518e-05, + "loss": 1.1738, + "step": 6035 + }, + { + "epoch": 0.19, + "learning_rate": 1.863842027768922e-05, + "loss": 1.0586, + "step": 6036 + }, + { + "epoch": 0.19, + "learning_rate": 1.8637898431980235e-05, + "loss": 1.1074, + "step": 6037 + }, + { + "epoch": 0.19, + "learning_rate": 1.8637376493596158e-05, + "loss": 1.0913, + "step": 6038 + }, + { + "epoch": 0.19, + "learning_rate": 1.863685446254259e-05, + "loss": 1.0435, + "step": 6039 + }, + { + "epoch": 0.19, + "learning_rate": 1.8636332338825133e-05, + "loss": 1.147, + "step": 6040 + }, + { + "epoch": 0.19, + "learning_rate": 1.863581012244939e-05, + "loss": 1.085, + "step": 6041 + }, + { + "epoch": 0.19, + "learning_rate": 1.8635287813420957e-05, + "loss": 1.1011, + "step": 6042 + }, + { + "epoch": 0.19, + "learning_rate": 1.8634765411745447e-05, + "loss": 1.1216, + "step": 6043 + }, + { + "epoch": 0.19, + "learning_rate": 1.8634242917428456e-05, + "loss": 0.9814, + "step": 6044 + }, + { + "epoch": 0.19, + "learning_rate": 1.8633720330475602e-05, + "loss": 1.1216, + "step": 6045 + }, + { + "epoch": 0.19, + "learning_rate": 1.8633197650892478e-05, + "loss": 1.0996, + "step": 6046 + }, + { + "epoch": 0.19, + "learning_rate": 1.8632674878684703e-05, + "loss": 0.8931, + "step": 6047 + }, + { + "epoch": 0.19, + "learning_rate": 1.8632152013857875e-05, + "loss": 1.0337, + "step": 6048 + }, + { + "epoch": 0.19, + "learning_rate": 1.8631629056417614e-05, + "loss": 1.1064, + "step": 6049 + }, + { + "epoch": 0.19, + "learning_rate": 1.8631106006369527e-05, + "loss": 1.0137, + "step": 6050 + }, + { + "epoch": 0.19, + "learning_rate": 1.863058286371922e-05, + "loss": 1.1284, + "step": 6051 + }, + { + "epoch": 0.19, + "learning_rate": 1.8630059628472312e-05, + "loss": 1.0063, + "step": 6052 + }, + { + "epoch": 0.19, + "learning_rate": 1.862953630063442e-05, + "loss": 1.1084, + "step": 6053 + }, + { + "epoch": 0.19, + "learning_rate": 1.862901288021115e-05, + "loss": 1.167, + "step": 6054 + }, + { + "epoch": 0.19, + "learning_rate": 1.862848936720812e-05, + "loss": 1.166, + "step": 6055 + }, + { + "epoch": 0.19, + "learning_rate": 1.8627965761630955e-05, + "loss": 1.1714, + "step": 6056 + }, + { + "epoch": 0.19, + "learning_rate": 1.862744206348526e-05, + "loss": 1.0005, + "step": 6057 + }, + { + "epoch": 0.19, + "learning_rate": 1.8626918272776663e-05, + "loss": 0.9937, + "step": 6058 + }, + { + "epoch": 0.19, + "learning_rate": 1.862639438951078e-05, + "loss": 1.082, + "step": 6059 + }, + { + "epoch": 0.19, + "learning_rate": 1.8625870413693235e-05, + "loss": 1.1221, + "step": 6060 + }, + { + "epoch": 0.19, + "learning_rate": 1.8625346345329646e-05, + "loss": 0.9609, + "step": 6061 + }, + { + "epoch": 0.19, + "learning_rate": 1.8624822184425634e-05, + "loss": 0.9739, + "step": 6062 + }, + { + "epoch": 0.19, + "learning_rate": 1.8624297930986827e-05, + "loss": 1.1587, + "step": 6063 + }, + { + "epoch": 0.19, + "learning_rate": 1.862377358501885e-05, + "loss": 1.0576, + "step": 6064 + }, + { + "epoch": 0.19, + "learning_rate": 1.8623249146527324e-05, + "loss": 1.0972, + "step": 6065 + }, + { + "epoch": 0.19, + "learning_rate": 1.862272461551788e-05, + "loss": 1.0986, + "step": 6066 + }, + { + "epoch": 0.19, + "learning_rate": 1.8622199991996144e-05, + "loss": 0.9941, + "step": 6067 + }, + { + "epoch": 0.19, + "learning_rate": 1.8621675275967742e-05, + "loss": 1.127, + "step": 6068 + }, + { + "epoch": 0.19, + "learning_rate": 1.862115046743831e-05, + "loss": 1.1729, + "step": 6069 + }, + { + "epoch": 0.19, + "learning_rate": 1.8620625566413474e-05, + "loss": 1.0962, + "step": 6070 + }, + { + "epoch": 0.19, + "learning_rate": 1.8620100572898864e-05, + "loss": 1.1421, + "step": 6071 + }, + { + "epoch": 0.19, + "learning_rate": 1.861957548690012e-05, + "loss": 1.0977, + "step": 6072 + }, + { + "epoch": 0.19, + "learning_rate": 1.861905030842287e-05, + "loss": 1.0298, + "step": 6073 + }, + { + "epoch": 0.19, + "learning_rate": 1.8618525037472744e-05, + "loss": 0.9082, + "step": 6074 + }, + { + "epoch": 0.19, + "learning_rate": 1.8617999674055386e-05, + "loss": 1.0845, + "step": 6075 + }, + { + "epoch": 0.19, + "learning_rate": 1.861747421817643e-05, + "loss": 1.1235, + "step": 6076 + }, + { + "epoch": 0.19, + "learning_rate": 1.861694866984151e-05, + "loss": 1.0864, + "step": 6077 + }, + { + "epoch": 0.19, + "learning_rate": 1.8616423029056272e-05, + "loss": 0.9941, + "step": 6078 + }, + { + "epoch": 0.19, + "learning_rate": 1.861589729582635e-05, + "loss": 1.1221, + "step": 6079 + }, + { + "epoch": 0.19, + "learning_rate": 1.8615371470157383e-05, + "loss": 1.0303, + "step": 6080 + }, + { + "epoch": 0.19, + "learning_rate": 1.8614845552055017e-05, + "loss": 1.0229, + "step": 6081 + }, + { + "epoch": 0.19, + "learning_rate": 1.861431954152489e-05, + "loss": 1.105, + "step": 6082 + }, + { + "epoch": 0.19, + "learning_rate": 1.8613793438572652e-05, + "loss": 1.0811, + "step": 6083 + }, + { + "epoch": 0.19, + "learning_rate": 1.8613267243203942e-05, + "loss": 1.167, + "step": 6084 + }, + { + "epoch": 0.19, + "learning_rate": 1.8612740955424408e-05, + "loss": 1.0972, + "step": 6085 + }, + { + "epoch": 0.19, + "learning_rate": 1.8612214575239695e-05, + "loss": 1.1226, + "step": 6086 + }, + { + "epoch": 0.19, + "learning_rate": 1.861168810265545e-05, + "loss": 1.0132, + "step": 6087 + }, + { + "epoch": 0.19, + "learning_rate": 1.8611161537677323e-05, + "loss": 0.9961, + "step": 6088 + }, + { + "epoch": 0.19, + "learning_rate": 1.8610634880310968e-05, + "loss": 1.0176, + "step": 6089 + }, + { + "epoch": 0.19, + "learning_rate": 1.8610108130562023e-05, + "loss": 1.1353, + "step": 6090 + }, + { + "epoch": 0.19, + "learning_rate": 1.8609581288436152e-05, + "loss": 1.0518, + "step": 6091 + }, + { + "epoch": 0.19, + "learning_rate": 1.8609054353939e-05, + "loss": 1.335, + "step": 6092 + }, + { + "epoch": 0.19, + "learning_rate": 1.8608527327076223e-05, + "loss": 1.0112, + "step": 6093 + }, + { + "epoch": 0.19, + "learning_rate": 1.8608000207853475e-05, + "loss": 1.0781, + "step": 6094 + }, + { + "epoch": 0.19, + "learning_rate": 1.8607472996276413e-05, + "loss": 1.1846, + "step": 6095 + }, + { + "epoch": 0.19, + "learning_rate": 1.8606945692350693e-05, + "loss": 1.0806, + "step": 6096 + }, + { + "epoch": 0.19, + "learning_rate": 1.860641829608197e-05, + "loss": 0.5747, + "step": 6097 + }, + { + "epoch": 0.2, + "learning_rate": 1.8605890807475902e-05, + "loss": 1.0312, + "step": 6098 + }, + { + "epoch": 0.2, + "learning_rate": 1.860536322653815e-05, + "loss": 1.1919, + "step": 6099 + }, + { + "epoch": 0.2, + "learning_rate": 1.860483555327438e-05, + "loss": 1.0425, + "step": 6100 + }, + { + "epoch": 0.2, + "learning_rate": 1.8604307787690242e-05, + "loss": 1.0972, + "step": 6101 + }, + { + "epoch": 0.2, + "learning_rate": 1.860377992979141e-05, + "loss": 1.0747, + "step": 6102 + }, + { + "epoch": 0.2, + "learning_rate": 1.8603251979583535e-05, + "loss": 0.9292, + "step": 6103 + }, + { + "epoch": 0.2, + "learning_rate": 1.8602723937072292e-05, + "loss": 1.2026, + "step": 6104 + }, + { + "epoch": 0.2, + "learning_rate": 1.8602195802263342e-05, + "loss": 1.0854, + "step": 6105 + }, + { + "epoch": 0.2, + "learning_rate": 1.860166757516235e-05, + "loss": 1.1636, + "step": 6106 + }, + { + "epoch": 0.2, + "learning_rate": 1.8601139255774985e-05, + "loss": 1.209, + "step": 6107 + }, + { + "epoch": 0.2, + "learning_rate": 1.8600610844106916e-05, + "loss": 1.0337, + "step": 6108 + }, + { + "epoch": 0.2, + "learning_rate": 1.8600082340163806e-05, + "loss": 1.0928, + "step": 6109 + }, + { + "epoch": 0.2, + "learning_rate": 1.8599553743951337e-05, + "loss": 1.0928, + "step": 6110 + }, + { + "epoch": 0.2, + "learning_rate": 1.8599025055475174e-05, + "loss": 1.0718, + "step": 6111 + }, + { + "epoch": 0.2, + "learning_rate": 1.859849627474099e-05, + "loss": 1.0649, + "step": 6112 + }, + { + "epoch": 0.2, + "learning_rate": 1.8597967401754453e-05, + "loss": 0.9604, + "step": 6113 + }, + { + "epoch": 0.2, + "learning_rate": 1.859743843652124e-05, + "loss": 1.1411, + "step": 6114 + }, + { + "epoch": 0.2, + "learning_rate": 1.859690937904703e-05, + "loss": 1.0195, + "step": 6115 + }, + { + "epoch": 0.2, + "learning_rate": 1.85963802293375e-05, + "loss": 1.0322, + "step": 6116 + }, + { + "epoch": 0.2, + "learning_rate": 1.859585098739832e-05, + "loss": 1.085, + "step": 6117 + }, + { + "epoch": 0.2, + "learning_rate": 1.8595321653235175e-05, + "loss": 1.0991, + "step": 6118 + }, + { + "epoch": 0.2, + "learning_rate": 1.859479222685374e-05, + "loss": 1.0557, + "step": 6119 + }, + { + "epoch": 0.2, + "learning_rate": 1.85942627082597e-05, + "loss": 1.1543, + "step": 6120 + }, + { + "epoch": 0.2, + "learning_rate": 1.8593733097458728e-05, + "loss": 1.0708, + "step": 6121 + }, + { + "epoch": 0.2, + "learning_rate": 1.8593203394456513e-05, + "loss": 0.9458, + "step": 6122 + }, + { + "epoch": 0.2, + "learning_rate": 1.8592673599258736e-05, + "loss": 1.1479, + "step": 6123 + }, + { + "epoch": 0.2, + "learning_rate": 1.859214371187108e-05, + "loss": 1.0142, + "step": 6124 + }, + { + "epoch": 0.2, + "learning_rate": 1.8591613732299233e-05, + "loss": 0.5571, + "step": 6125 + }, + { + "epoch": 0.2, + "learning_rate": 1.859108366054888e-05, + "loss": 1.0825, + "step": 6126 + }, + { + "epoch": 0.2, + "learning_rate": 1.8590553496625706e-05, + "loss": 1.083, + "step": 6127 + }, + { + "epoch": 0.2, + "learning_rate": 1.85900232405354e-05, + "loss": 0.9912, + "step": 6128 + }, + { + "epoch": 0.2, + "learning_rate": 1.8589492892283653e-05, + "loss": 1.0654, + "step": 6129 + }, + { + "epoch": 0.2, + "learning_rate": 1.8588962451876154e-05, + "loss": 1.0107, + "step": 6130 + }, + { + "epoch": 0.2, + "learning_rate": 1.858843191931859e-05, + "loss": 1.0571, + "step": 6131 + }, + { + "epoch": 0.2, + "learning_rate": 1.8587901294616665e-05, + "loss": 1.1143, + "step": 6132 + }, + { + "epoch": 0.2, + "learning_rate": 1.8587370577776056e-05, + "loss": 1.1572, + "step": 6133 + }, + { + "epoch": 0.2, + "learning_rate": 1.8586839768802468e-05, + "loss": 1.0918, + "step": 6134 + }, + { + "epoch": 0.2, + "learning_rate": 1.8586308867701592e-05, + "loss": 1.0239, + "step": 6135 + }, + { + "epoch": 0.2, + "learning_rate": 1.8585777874479123e-05, + "loss": 1.0649, + "step": 6136 + }, + { + "epoch": 0.2, + "learning_rate": 1.858524678914076e-05, + "loss": 0.9722, + "step": 6137 + }, + { + "epoch": 0.2, + "learning_rate": 1.8584715611692205e-05, + "loss": 1.082, + "step": 6138 + }, + { + "epoch": 0.2, + "learning_rate": 1.8584184342139146e-05, + "loss": 1.1768, + "step": 6139 + }, + { + "epoch": 0.2, + "learning_rate": 1.8583652980487296e-05, + "loss": 1.0903, + "step": 6140 + }, + { + "epoch": 0.2, + "learning_rate": 1.8583121526742345e-05, + "loss": 1.0732, + "step": 6141 + }, + { + "epoch": 0.2, + "learning_rate": 1.8582589980909998e-05, + "loss": 1.1411, + "step": 6142 + }, + { + "epoch": 0.2, + "learning_rate": 1.8582058342995964e-05, + "loss": 1.1045, + "step": 6143 + }, + { + "epoch": 0.2, + "learning_rate": 1.858152661300594e-05, + "loss": 1.0918, + "step": 6144 + }, + { + "epoch": 0.2, + "learning_rate": 1.8580994790945635e-05, + "loss": 1.1465, + "step": 6145 + }, + { + "epoch": 0.2, + "learning_rate": 1.8580462876820747e-05, + "loss": 1.082, + "step": 6146 + }, + { + "epoch": 0.2, + "learning_rate": 1.8579930870636994e-05, + "loss": 1.0576, + "step": 6147 + }, + { + "epoch": 0.2, + "learning_rate": 1.857939877240008e-05, + "loss": 1.064, + "step": 6148 + }, + { + "epoch": 0.2, + "learning_rate": 1.857886658211571e-05, + "loss": 1.0864, + "step": 6149 + }, + { + "epoch": 0.2, + "learning_rate": 1.8578334299789593e-05, + "loss": 1.0918, + "step": 6150 + }, + { + "epoch": 0.2, + "learning_rate": 1.8577801925427445e-05, + "loss": 1.085, + "step": 6151 + }, + { + "epoch": 0.2, + "learning_rate": 1.857726945903498e-05, + "loss": 1.1094, + "step": 6152 + }, + { + "epoch": 0.2, + "learning_rate": 1.85767369006179e-05, + "loss": 1.2065, + "step": 6153 + }, + { + "epoch": 0.2, + "learning_rate": 1.857620425018193e-05, + "loss": 1.2539, + "step": 6154 + }, + { + "epoch": 0.2, + "learning_rate": 1.857567150773278e-05, + "loss": 1.1235, + "step": 6155 + }, + { + "epoch": 0.2, + "learning_rate": 1.857513867327616e-05, + "loss": 1.0527, + "step": 6156 + }, + { + "epoch": 0.2, + "learning_rate": 1.8574605746817796e-05, + "loss": 1.0757, + "step": 6157 + }, + { + "epoch": 0.2, + "learning_rate": 1.8574072728363407e-05, + "loss": 1.0586, + "step": 6158 + }, + { + "epoch": 0.2, + "learning_rate": 1.85735396179187e-05, + "loss": 1.019, + "step": 6159 + }, + { + "epoch": 0.2, + "learning_rate": 1.8573006415489406e-05, + "loss": 1.0425, + "step": 6160 + }, + { + "epoch": 0.2, + "learning_rate": 1.857247312108124e-05, + "loss": 0.9863, + "step": 6161 + }, + { + "epoch": 0.2, + "learning_rate": 1.8571939734699923e-05, + "loss": 1.1436, + "step": 6162 + }, + { + "epoch": 0.2, + "learning_rate": 1.8571406256351184e-05, + "loss": 0.9785, + "step": 6163 + }, + { + "epoch": 0.2, + "learning_rate": 1.8570872686040738e-05, + "loss": 1.147, + "step": 6164 + }, + { + "epoch": 0.2, + "learning_rate": 1.8570339023774315e-05, + "loss": 1.188, + "step": 6165 + }, + { + "epoch": 0.2, + "learning_rate": 1.856980526955764e-05, + "loss": 1.0981, + "step": 6166 + }, + { + "epoch": 0.2, + "learning_rate": 1.856927142339644e-05, + "loss": 1.1445, + "step": 6167 + }, + { + "epoch": 0.2, + "learning_rate": 1.856873748529644e-05, + "loss": 1.0083, + "step": 6168 + }, + { + "epoch": 0.2, + "learning_rate": 1.856820345526337e-05, + "loss": 1.1011, + "step": 6169 + }, + { + "epoch": 0.2, + "learning_rate": 1.8567669333302962e-05, + "loss": 1.106, + "step": 6170 + }, + { + "epoch": 0.2, + "learning_rate": 1.8567135119420942e-05, + "loss": 1.0234, + "step": 6171 + }, + { + "epoch": 0.2, + "learning_rate": 1.8566600813623047e-05, + "loss": 0.8823, + "step": 6172 + }, + { + "epoch": 0.2, + "learning_rate": 1.8566066415915002e-05, + "loss": 1.1558, + "step": 6173 + }, + { + "epoch": 0.2, + "learning_rate": 1.8565531926302548e-05, + "loss": 1.0977, + "step": 6174 + }, + { + "epoch": 0.2, + "learning_rate": 1.8564997344791417e-05, + "loss": 1.2173, + "step": 6175 + }, + { + "epoch": 0.2, + "learning_rate": 1.856446267138734e-05, + "loss": 1.1938, + "step": 6176 + }, + { + "epoch": 0.2, + "learning_rate": 1.856392790609606e-05, + "loss": 1.1152, + "step": 6177 + }, + { + "epoch": 0.2, + "learning_rate": 1.856339304892331e-05, + "loss": 1.041, + "step": 6178 + }, + { + "epoch": 0.2, + "learning_rate": 1.856285809987483e-05, + "loss": 1.1382, + "step": 6179 + }, + { + "epoch": 0.2, + "learning_rate": 1.856232305895636e-05, + "loss": 1.1313, + "step": 6180 + }, + { + "epoch": 0.2, + "learning_rate": 1.856178792617364e-05, + "loss": 1.0273, + "step": 6181 + }, + { + "epoch": 0.2, + "learning_rate": 1.8561252701532413e-05, + "loss": 1.0933, + "step": 6182 + }, + { + "epoch": 0.2, + "learning_rate": 1.856071738503842e-05, + "loss": 1.1748, + "step": 6183 + }, + { + "epoch": 0.2, + "learning_rate": 1.85601819766974e-05, + "loss": 1.0708, + "step": 6184 + }, + { + "epoch": 0.2, + "learning_rate": 1.85596464765151e-05, + "loss": 0.9683, + "step": 6185 + }, + { + "epoch": 0.2, + "learning_rate": 1.8559110884497273e-05, + "loss": 1.0825, + "step": 6186 + }, + { + "epoch": 0.2, + "learning_rate": 1.8558575200649654e-05, + "loss": 1.2378, + "step": 6187 + }, + { + "epoch": 0.2, + "learning_rate": 1.8558039424978e-05, + "loss": 1.0811, + "step": 6188 + }, + { + "epoch": 0.2, + "learning_rate": 1.855750355748805e-05, + "loss": 0.9922, + "step": 6189 + }, + { + "epoch": 0.2, + "learning_rate": 1.8556967598185562e-05, + "loss": 0.9941, + "step": 6190 + }, + { + "epoch": 0.2, + "learning_rate": 1.855643154707628e-05, + "loss": 1.188, + "step": 6191 + }, + { + "epoch": 0.2, + "learning_rate": 1.8555895404165954e-05, + "loss": 1.0986, + "step": 6192 + }, + { + "epoch": 0.2, + "learning_rate": 1.8555359169460342e-05, + "loss": 1.0117, + "step": 6193 + }, + { + "epoch": 0.2, + "learning_rate": 1.8554822842965198e-05, + "loss": 0.9683, + "step": 6194 + }, + { + "epoch": 0.2, + "learning_rate": 1.8554286424686268e-05, + "loss": 1.0605, + "step": 6195 + }, + { + "epoch": 0.2, + "learning_rate": 1.8553749914629316e-05, + "loss": 1.084, + "step": 6196 + }, + { + "epoch": 0.2, + "learning_rate": 1.8553213312800092e-05, + "loss": 1.061, + "step": 6197 + }, + { + "epoch": 0.2, + "learning_rate": 1.8552676619204355e-05, + "loss": 1.1387, + "step": 6198 + }, + { + "epoch": 0.2, + "learning_rate": 1.8552139833847868e-05, + "loss": 1.1484, + "step": 6199 + }, + { + "epoch": 0.2, + "learning_rate": 1.8551602956736378e-05, + "loss": 1.0693, + "step": 6200 + }, + { + "epoch": 0.2, + "learning_rate": 1.8551065987875655e-05, + "loss": 1.0747, + "step": 6201 + }, + { + "epoch": 0.2, + "learning_rate": 1.855052892727146e-05, + "loss": 1.1323, + "step": 6202 + }, + { + "epoch": 0.2, + "learning_rate": 1.8549991774929555e-05, + "loss": 1.1055, + "step": 6203 + }, + { + "epoch": 0.2, + "learning_rate": 1.8549454530855697e-05, + "loss": 1.0132, + "step": 6204 + }, + { + "epoch": 0.2, + "learning_rate": 1.8548917195055656e-05, + "loss": 1.168, + "step": 6205 + }, + { + "epoch": 0.2, + "learning_rate": 1.8548379767535195e-05, + "loss": 1.041, + "step": 6206 + }, + { + "epoch": 0.2, + "learning_rate": 1.8547842248300075e-05, + "loss": 1.0518, + "step": 6207 + }, + { + "epoch": 0.2, + "learning_rate": 1.8547304637356072e-05, + "loss": 1.0239, + "step": 6208 + }, + { + "epoch": 0.2, + "learning_rate": 1.8546766934708952e-05, + "loss": 1.167, + "step": 6209 + }, + { + "epoch": 0.2, + "learning_rate": 1.854622914036448e-05, + "loss": 1.0674, + "step": 6210 + }, + { + "epoch": 0.2, + "learning_rate": 1.854569125432843e-05, + "loss": 1.0928, + "step": 6211 + }, + { + "epoch": 0.2, + "learning_rate": 1.8545153276606566e-05, + "loss": 1.0352, + "step": 6212 + }, + { + "epoch": 0.2, + "learning_rate": 1.854461520720467e-05, + "loss": 1.0674, + "step": 6213 + }, + { + "epoch": 0.2, + "learning_rate": 1.8544077046128504e-05, + "loss": 1.1699, + "step": 6214 + }, + { + "epoch": 0.2, + "learning_rate": 1.854353879338385e-05, + "loss": 1.0215, + "step": 6215 + }, + { + "epoch": 0.2, + "learning_rate": 1.8543000448976483e-05, + "loss": 1.1567, + "step": 6216 + }, + { + "epoch": 0.2, + "learning_rate": 1.854246201291217e-05, + "loss": 1.0186, + "step": 6217 + }, + { + "epoch": 0.2, + "learning_rate": 1.85419234851967e-05, + "loss": 1.1816, + "step": 6218 + }, + { + "epoch": 0.2, + "learning_rate": 1.854138486583584e-05, + "loss": 1.1167, + "step": 6219 + }, + { + "epoch": 0.2, + "learning_rate": 1.8540846154835376e-05, + "loss": 1.0366, + "step": 6220 + }, + { + "epoch": 0.2, + "learning_rate": 1.8540307352201085e-05, + "loss": 0.9897, + "step": 6221 + }, + { + "epoch": 0.2, + "learning_rate": 1.8539768457938748e-05, + "loss": 1.0151, + "step": 6222 + }, + { + "epoch": 0.2, + "learning_rate": 1.8539229472054148e-05, + "loss": 1.1548, + "step": 6223 + }, + { + "epoch": 0.2, + "learning_rate": 1.8538690394553065e-05, + "loss": 1.2451, + "step": 6224 + }, + { + "epoch": 0.2, + "learning_rate": 1.8538151225441282e-05, + "loss": 1.1172, + "step": 6225 + }, + { + "epoch": 0.2, + "learning_rate": 1.853761196472459e-05, + "loss": 1.1914, + "step": 6226 + }, + { + "epoch": 0.2, + "learning_rate": 1.853707261240877e-05, + "loss": 1.1772, + "step": 6227 + }, + { + "epoch": 0.2, + "learning_rate": 1.8536533168499605e-05, + "loss": 1.085, + "step": 6228 + }, + { + "epoch": 0.2, + "learning_rate": 1.853599363300289e-05, + "loss": 1.1523, + "step": 6229 + }, + { + "epoch": 0.2, + "learning_rate": 1.8535454005924413e-05, + "loss": 1.0962, + "step": 6230 + }, + { + "epoch": 0.2, + "learning_rate": 1.853491428726996e-05, + "loss": 1.0708, + "step": 6231 + }, + { + "epoch": 0.2, + "learning_rate": 1.853437447704532e-05, + "loss": 1.0186, + "step": 6232 + }, + { + "epoch": 0.2, + "learning_rate": 1.853383457525629e-05, + "loss": 1.0103, + "step": 6233 + }, + { + "epoch": 0.2, + "learning_rate": 1.8533294581908658e-05, + "loss": 1.0103, + "step": 6234 + }, + { + "epoch": 0.2, + "learning_rate": 1.8532754497008217e-05, + "loss": 1.0547, + "step": 6235 + }, + { + "epoch": 0.2, + "learning_rate": 1.8532214320560768e-05, + "loss": 1.0615, + "step": 6236 + }, + { + "epoch": 0.2, + "learning_rate": 1.8531674052572102e-05, + "loss": 1.1299, + "step": 6237 + }, + { + "epoch": 0.2, + "learning_rate": 1.8531133693048012e-05, + "loss": 0.979, + "step": 6238 + }, + { + "epoch": 0.2, + "learning_rate": 1.8530593241994307e-05, + "loss": 1.0825, + "step": 6239 + }, + { + "epoch": 0.2, + "learning_rate": 1.853005269941677e-05, + "loss": 1.124, + "step": 6240 + }, + { + "epoch": 0.2, + "learning_rate": 1.852951206532121e-05, + "loss": 1.0806, + "step": 6241 + }, + { + "epoch": 0.2, + "learning_rate": 1.8528971339713425e-05, + "loss": 1.125, + "step": 6242 + }, + { + "epoch": 0.2, + "learning_rate": 1.852843052259922e-05, + "loss": 1.0923, + "step": 6243 + }, + { + "epoch": 0.2, + "learning_rate": 1.852788961398439e-05, + "loss": 1.0225, + "step": 6244 + }, + { + "epoch": 0.2, + "learning_rate": 1.8527348613874748e-05, + "loss": 0.9248, + "step": 6245 + }, + { + "epoch": 0.2, + "learning_rate": 1.852680752227609e-05, + "loss": 1.2065, + "step": 6246 + }, + { + "epoch": 0.2, + "learning_rate": 1.8526266339194225e-05, + "loss": 1.0381, + "step": 6247 + }, + { + "epoch": 0.2, + "learning_rate": 1.8525725064634956e-05, + "loss": 1.1426, + "step": 6248 + }, + { + "epoch": 0.2, + "learning_rate": 1.8525183698604098e-05, + "loss": 1.0723, + "step": 6249 + }, + { + "epoch": 0.2, + "learning_rate": 1.8524642241107448e-05, + "loss": 1.147, + "step": 6250 + }, + { + "epoch": 0.2, + "learning_rate": 1.8524100692150826e-05, + "loss": 1.0269, + "step": 6251 + }, + { + "epoch": 0.2, + "learning_rate": 1.8523559051740036e-05, + "loss": 1.2461, + "step": 6252 + }, + { + "epoch": 0.2, + "learning_rate": 1.852301731988089e-05, + "loss": 1.0361, + "step": 6253 + }, + { + "epoch": 0.2, + "learning_rate": 1.85224754965792e-05, + "loss": 1.2114, + "step": 6254 + }, + { + "epoch": 0.2, + "learning_rate": 1.8521933581840782e-05, + "loss": 1.1064, + "step": 6255 + }, + { + "epoch": 0.2, + "learning_rate": 1.852139157567145e-05, + "loss": 1.0342, + "step": 6256 + }, + { + "epoch": 0.2, + "learning_rate": 1.8520849478077015e-05, + "loss": 1.0488, + "step": 6257 + }, + { + "epoch": 0.2, + "learning_rate": 1.8520307289063297e-05, + "loss": 1.0332, + "step": 6258 + }, + { + "epoch": 0.2, + "learning_rate": 1.8519765008636107e-05, + "loss": 1.0874, + "step": 6259 + }, + { + "epoch": 0.2, + "learning_rate": 1.851922263680127e-05, + "loss": 1.1128, + "step": 6260 + }, + { + "epoch": 0.2, + "learning_rate": 1.8518680173564608e-05, + "loss": 1.0825, + "step": 6261 + }, + { + "epoch": 0.2, + "learning_rate": 1.851813761893193e-05, + "loss": 1.147, + "step": 6262 + }, + { + "epoch": 0.2, + "learning_rate": 1.8517594972909067e-05, + "loss": 1.0562, + "step": 6263 + }, + { + "epoch": 0.2, + "learning_rate": 1.8517052235501832e-05, + "loss": 0.9805, + "step": 6264 + }, + { + "epoch": 0.2, + "learning_rate": 1.8516509406716056e-05, + "loss": 1.1934, + "step": 6265 + }, + { + "epoch": 0.2, + "learning_rate": 1.851596648655756e-05, + "loss": 1.0542, + "step": 6266 + }, + { + "epoch": 0.2, + "learning_rate": 1.8515423475032166e-05, + "loss": 1.1514, + "step": 6267 + }, + { + "epoch": 0.2, + "learning_rate": 1.8514880372145704e-05, + "loss": 1.1055, + "step": 6268 + }, + { + "epoch": 0.2, + "learning_rate": 1.8514337177903998e-05, + "loss": 1.1421, + "step": 6269 + }, + { + "epoch": 0.2, + "learning_rate": 1.851379389231288e-05, + "loss": 1.0811, + "step": 6270 + }, + { + "epoch": 0.2, + "learning_rate": 1.8513250515378174e-05, + "loss": 1.1133, + "step": 6271 + }, + { + "epoch": 0.2, + "learning_rate": 1.8512707047105714e-05, + "loss": 0.5088, + "step": 6272 + }, + { + "epoch": 0.2, + "learning_rate": 1.8512163487501323e-05, + "loss": 1.1377, + "step": 6273 + }, + { + "epoch": 0.2, + "learning_rate": 1.8511619836570846e-05, + "loss": 1.0059, + "step": 6274 + }, + { + "epoch": 0.2, + "learning_rate": 1.8511076094320108e-05, + "loss": 1.0801, + "step": 6275 + }, + { + "epoch": 0.2, + "learning_rate": 1.8510532260754935e-05, + "loss": 0.9326, + "step": 6276 + }, + { + "epoch": 0.2, + "learning_rate": 1.8509988335881176e-05, + "loss": 1.0679, + "step": 6277 + }, + { + "epoch": 0.2, + "learning_rate": 1.850944431970466e-05, + "loss": 1.1997, + "step": 6278 + }, + { + "epoch": 0.2, + "learning_rate": 1.8508900212231225e-05, + "loss": 1.1597, + "step": 6279 + }, + { + "epoch": 0.2, + "learning_rate": 1.850835601346671e-05, + "loss": 1.0264, + "step": 6280 + }, + { + "epoch": 0.2, + "learning_rate": 1.8507811723416945e-05, + "loss": 1.1826, + "step": 6281 + }, + { + "epoch": 0.2, + "learning_rate": 1.850726734208778e-05, + "loss": 1.1245, + "step": 6282 + }, + { + "epoch": 0.2, + "learning_rate": 1.8506722869485053e-05, + "loss": 0.9624, + "step": 6283 + }, + { + "epoch": 0.2, + "learning_rate": 1.8506178305614602e-05, + "loss": 1.0991, + "step": 6284 + }, + { + "epoch": 0.2, + "learning_rate": 1.850563365048227e-05, + "loss": 1.1626, + "step": 6285 + }, + { + "epoch": 0.2, + "learning_rate": 1.850508890409391e-05, + "loss": 1.1431, + "step": 6286 + }, + { + "epoch": 0.2, + "learning_rate": 1.850454406645535e-05, + "loss": 0.9585, + "step": 6287 + }, + { + "epoch": 0.2, + "learning_rate": 1.850399913757245e-05, + "loss": 1.1084, + "step": 6288 + }, + { + "epoch": 0.2, + "learning_rate": 1.850345411745105e-05, + "loss": 1.1265, + "step": 6289 + }, + { + "epoch": 0.2, + "learning_rate": 1.8502909006096997e-05, + "loss": 0.9785, + "step": 6290 + }, + { + "epoch": 0.2, + "learning_rate": 1.8502363803516145e-05, + "loss": 1.1123, + "step": 6291 + }, + { + "epoch": 0.2, + "learning_rate": 1.8501818509714337e-05, + "loss": 1.0918, + "step": 6292 + }, + { + "epoch": 0.2, + "learning_rate": 1.8501273124697424e-05, + "loss": 1.0645, + "step": 6293 + }, + { + "epoch": 0.2, + "learning_rate": 1.8500727648471258e-05, + "loss": 0.5386, + "step": 6294 + }, + { + "epoch": 0.2, + "learning_rate": 1.8500182081041697e-05, + "loss": 1.0845, + "step": 6295 + }, + { + "epoch": 0.2, + "learning_rate": 1.8499636422414585e-05, + "loss": 0.4834, + "step": 6296 + }, + { + "epoch": 0.2, + "learning_rate": 1.8499090672595785e-05, + "loss": 1.1719, + "step": 6297 + }, + { + "epoch": 0.2, + "learning_rate": 1.8498544831591147e-05, + "loss": 1.1108, + "step": 6298 + }, + { + "epoch": 0.2, + "learning_rate": 1.849799889940653e-05, + "loss": 1.1362, + "step": 6299 + }, + { + "epoch": 0.2, + "learning_rate": 1.849745287604779e-05, + "loss": 1.146, + "step": 6300 + }, + { + "epoch": 0.2, + "learning_rate": 1.8496906761520782e-05, + "loss": 1.1729, + "step": 6301 + }, + { + "epoch": 0.2, + "learning_rate": 1.849636055583137e-05, + "loss": 1.0024, + "step": 6302 + }, + { + "epoch": 0.2, + "learning_rate": 1.8495814258985414e-05, + "loss": 1.0771, + "step": 6303 + }, + { + "epoch": 0.2, + "learning_rate": 1.8495267870988772e-05, + "loss": 1.1562, + "step": 6304 + }, + { + "epoch": 0.2, + "learning_rate": 1.849472139184731e-05, + "loss": 1.1084, + "step": 6305 + }, + { + "epoch": 0.2, + "learning_rate": 1.8494174821566887e-05, + "loss": 1.0171, + "step": 6306 + }, + { + "epoch": 0.2, + "learning_rate": 1.849362816015337e-05, + "loss": 1.1401, + "step": 6307 + }, + { + "epoch": 0.2, + "learning_rate": 1.849308140761262e-05, + "loss": 0.9243, + "step": 6308 + }, + { + "epoch": 0.2, + "learning_rate": 1.849253456395051e-05, + "loss": 1.168, + "step": 6309 + }, + { + "epoch": 0.2, + "learning_rate": 1.84919876291729e-05, + "loss": 0.8306, + "step": 6310 + }, + { + "epoch": 0.2, + "learning_rate": 1.8491440603285667e-05, + "loss": 1.1523, + "step": 6311 + }, + { + "epoch": 0.2, + "learning_rate": 1.8490893486294666e-05, + "loss": 0.9551, + "step": 6312 + }, + { + "epoch": 0.2, + "learning_rate": 1.8490346278205783e-05, + "loss": 0.9346, + "step": 6313 + }, + { + "epoch": 0.2, + "learning_rate": 1.8489798979024877e-05, + "loss": 1.0264, + "step": 6314 + }, + { + "epoch": 0.2, + "learning_rate": 1.8489251588757825e-05, + "loss": 1.0923, + "step": 6315 + }, + { + "epoch": 0.2, + "learning_rate": 1.8488704107410495e-05, + "loss": 1.1484, + "step": 6316 + }, + { + "epoch": 0.2, + "learning_rate": 1.848815653498877e-05, + "loss": 1.04, + "step": 6317 + }, + { + "epoch": 0.2, + "learning_rate": 1.848760887149852e-05, + "loss": 1.1504, + "step": 6318 + }, + { + "epoch": 0.2, + "learning_rate": 1.848706111694562e-05, + "loss": 1.2295, + "step": 6319 + }, + { + "epoch": 0.2, + "learning_rate": 1.848651327133594e-05, + "loss": 1.082, + "step": 6320 + }, + { + "epoch": 0.2, + "learning_rate": 1.8485965334675372e-05, + "loss": 1.1904, + "step": 6321 + }, + { + "epoch": 0.2, + "learning_rate": 1.8485417306969786e-05, + "loss": 1.0249, + "step": 6322 + }, + { + "epoch": 0.2, + "learning_rate": 1.8484869188225065e-05, + "loss": 1.0718, + "step": 6323 + }, + { + "epoch": 0.2, + "learning_rate": 1.8484320978447087e-05, + "loss": 0.9883, + "step": 6324 + }, + { + "epoch": 0.2, + "learning_rate": 1.8483772677641734e-05, + "loss": 1.0894, + "step": 6325 + }, + { + "epoch": 0.2, + "learning_rate": 1.848322428581489e-05, + "loss": 0.9971, + "step": 6326 + }, + { + "epoch": 0.2, + "learning_rate": 1.8482675802972438e-05, + "loss": 1.1064, + "step": 6327 + }, + { + "epoch": 0.2, + "learning_rate": 1.8482127229120265e-05, + "loss": 0.9751, + "step": 6328 + }, + { + "epoch": 0.2, + "learning_rate": 1.848157856426425e-05, + "loss": 1.1274, + "step": 6329 + }, + { + "epoch": 0.2, + "learning_rate": 1.848102980841029e-05, + "loss": 0.9287, + "step": 6330 + }, + { + "epoch": 0.2, + "learning_rate": 1.848048096156426e-05, + "loss": 1.1016, + "step": 6331 + }, + { + "epoch": 0.2, + "learning_rate": 1.847993202373206e-05, + "loss": 1.0137, + "step": 6332 + }, + { + "epoch": 0.2, + "learning_rate": 1.8479382994919572e-05, + "loss": 1.1025, + "step": 6333 + }, + { + "epoch": 0.2, + "learning_rate": 1.847883387513269e-05, + "loss": 1.1157, + "step": 6334 + }, + { + "epoch": 0.2, + "learning_rate": 1.8478284664377304e-05, + "loss": 1.0581, + "step": 6335 + }, + { + "epoch": 0.2, + "learning_rate": 1.8477735362659304e-05, + "loss": 1.04, + "step": 6336 + }, + { + "epoch": 0.2, + "learning_rate": 1.8477185969984593e-05, + "loss": 1.0454, + "step": 6337 + }, + { + "epoch": 0.2, + "learning_rate": 1.8476636486359055e-05, + "loss": 1.0283, + "step": 6338 + }, + { + "epoch": 0.2, + "learning_rate": 1.8476086911788588e-05, + "loss": 1.166, + "step": 6339 + }, + { + "epoch": 0.2, + "learning_rate": 1.847553724627909e-05, + "loss": 1.0352, + "step": 6340 + }, + { + "epoch": 0.2, + "learning_rate": 1.847498748983646e-05, + "loss": 1.0645, + "step": 6341 + }, + { + "epoch": 0.2, + "learning_rate": 1.8474437642466588e-05, + "loss": 1.0439, + "step": 6342 + }, + { + "epoch": 0.2, + "learning_rate": 1.8473887704175383e-05, + "loss": 1.1167, + "step": 6343 + }, + { + "epoch": 0.2, + "learning_rate": 1.847333767496874e-05, + "loss": 1.1021, + "step": 6344 + }, + { + "epoch": 0.2, + "learning_rate": 1.8472787554852566e-05, + "loss": 0.9341, + "step": 6345 + }, + { + "epoch": 0.2, + "learning_rate": 1.8472237343832754e-05, + "loss": 1.0845, + "step": 6346 + }, + { + "epoch": 0.2, + "learning_rate": 1.8471687041915214e-05, + "loss": 1.1348, + "step": 6347 + }, + { + "epoch": 0.2, + "learning_rate": 1.847113664910585e-05, + "loss": 1.104, + "step": 6348 + }, + { + "epoch": 0.2, + "learning_rate": 1.8470586165410556e-05, + "loss": 0.9629, + "step": 6349 + }, + { + "epoch": 0.2, + "learning_rate": 1.847003559083526e-05, + "loss": 1.0264, + "step": 6350 + }, + { + "epoch": 0.2, + "learning_rate": 1.8469484925385846e-05, + "loss": 1.0396, + "step": 6351 + }, + { + "epoch": 0.2, + "learning_rate": 1.8468934169068235e-05, + "loss": 1.0771, + "step": 6352 + }, + { + "epoch": 0.2, + "learning_rate": 1.8468383321888337e-05, + "loss": 1.0356, + "step": 6353 + }, + { + "epoch": 0.2, + "learning_rate": 1.8467832383852056e-05, + "loss": 1.0322, + "step": 6354 + }, + { + "epoch": 0.2, + "learning_rate": 1.8467281354965304e-05, + "loss": 0.9487, + "step": 6355 + }, + { + "epoch": 0.2, + "learning_rate": 1.8466730235233995e-05, + "loss": 1.0244, + "step": 6356 + }, + { + "epoch": 0.2, + "learning_rate": 1.8466179024664043e-05, + "loss": 1.0601, + "step": 6357 + }, + { + "epoch": 0.2, + "learning_rate": 1.8465627723261358e-05, + "loss": 1.0249, + "step": 6358 + }, + { + "epoch": 0.2, + "learning_rate": 1.8465076331031858e-05, + "loss": 1.1108, + "step": 6359 + }, + { + "epoch": 0.2, + "learning_rate": 1.8464524847981456e-05, + "loss": 1.2314, + "step": 6360 + }, + { + "epoch": 0.2, + "learning_rate": 1.846397327411607e-05, + "loss": 1.104, + "step": 6361 + }, + { + "epoch": 0.2, + "learning_rate": 1.846342160944162e-05, + "loss": 1.2075, + "step": 6362 + }, + { + "epoch": 0.2, + "learning_rate": 1.8462869853964022e-05, + "loss": 1.2285, + "step": 6363 + }, + { + "epoch": 0.2, + "learning_rate": 1.8462318007689196e-05, + "loss": 1.1162, + "step": 6364 + }, + { + "epoch": 0.2, + "learning_rate": 1.8461766070623066e-05, + "loss": 1.0444, + "step": 6365 + }, + { + "epoch": 0.2, + "learning_rate": 1.8461214042771552e-05, + "loss": 1.0488, + "step": 6366 + }, + { + "epoch": 0.2, + "learning_rate": 1.8460661924140573e-05, + "loss": 1.0942, + "step": 6367 + }, + { + "epoch": 0.2, + "learning_rate": 1.8460109714736058e-05, + "loss": 1.1958, + "step": 6368 + }, + { + "epoch": 0.2, + "learning_rate": 1.8459557414563926e-05, + "loss": 1.0332, + "step": 6369 + }, + { + "epoch": 0.2, + "learning_rate": 1.8459005023630103e-05, + "loss": 1.127, + "step": 6370 + }, + { + "epoch": 0.2, + "learning_rate": 1.8458452541940524e-05, + "loss": 1.0957, + "step": 6371 + }, + { + "epoch": 0.2, + "learning_rate": 1.8457899969501107e-05, + "loss": 1.1816, + "step": 6372 + }, + { + "epoch": 0.2, + "learning_rate": 1.8457347306317785e-05, + "loss": 1.0908, + "step": 6373 + }, + { + "epoch": 0.2, + "learning_rate": 1.8456794552396487e-05, + "loss": 1.002, + "step": 6374 + }, + { + "epoch": 0.2, + "learning_rate": 1.8456241707743143e-05, + "loss": 1.144, + "step": 6375 + }, + { + "epoch": 0.2, + "learning_rate": 1.8455688772363683e-05, + "loss": 1.1265, + "step": 6376 + }, + { + "epoch": 0.2, + "learning_rate": 1.845513574626404e-05, + "loss": 1.125, + "step": 6377 + }, + { + "epoch": 0.2, + "learning_rate": 1.845458262945015e-05, + "loss": 1.1323, + "step": 6378 + }, + { + "epoch": 0.2, + "learning_rate": 1.8454029421927947e-05, + "loss": 1.1138, + "step": 6379 + }, + { + "epoch": 0.2, + "learning_rate": 1.845347612370336e-05, + "loss": 1.106, + "step": 6380 + }, + { + "epoch": 0.2, + "learning_rate": 1.8452922734782336e-05, + "loss": 1.021, + "step": 6381 + }, + { + "epoch": 0.2, + "learning_rate": 1.8452369255170805e-05, + "loss": 1.0664, + "step": 6382 + }, + { + "epoch": 0.2, + "learning_rate": 1.8451815684874703e-05, + "loss": 1.0752, + "step": 6383 + }, + { + "epoch": 0.2, + "learning_rate": 1.8451262023899973e-05, + "loss": 1.1382, + "step": 6384 + }, + { + "epoch": 0.2, + "learning_rate": 1.845070827225256e-05, + "loss": 0.9575, + "step": 6385 + }, + { + "epoch": 0.2, + "learning_rate": 1.8450154429938395e-05, + "loss": 1.041, + "step": 6386 + }, + { + "epoch": 0.2, + "learning_rate": 1.8449600496963426e-05, + "loss": 1.0391, + "step": 6387 + }, + { + "epoch": 0.2, + "learning_rate": 1.8449046473333597e-05, + "loss": 1.0059, + "step": 6388 + }, + { + "epoch": 0.2, + "learning_rate": 1.8448492359054847e-05, + "loss": 1.0366, + "step": 6389 + }, + { + "epoch": 0.2, + "learning_rate": 1.8447938154133126e-05, + "loss": 1.022, + "step": 6390 + }, + { + "epoch": 0.2, + "learning_rate": 1.844738385857438e-05, + "loss": 1.1113, + "step": 6391 + }, + { + "epoch": 0.2, + "learning_rate": 1.8446829472384552e-05, + "loss": 0.9951, + "step": 6392 + }, + { + "epoch": 0.2, + "learning_rate": 1.8446274995569596e-05, + "loss": 1.1592, + "step": 6393 + }, + { + "epoch": 0.2, + "learning_rate": 1.8445720428135453e-05, + "loss": 0.5354, + "step": 6394 + }, + { + "epoch": 0.2, + "learning_rate": 1.844516577008808e-05, + "loss": 1.1489, + "step": 6395 + }, + { + "epoch": 0.2, + "learning_rate": 1.8444611021433423e-05, + "loss": 1.0508, + "step": 6396 + }, + { + "epoch": 0.2, + "learning_rate": 1.8444056182177435e-05, + "loss": 1.1421, + "step": 6397 + }, + { + "epoch": 0.2, + "learning_rate": 1.844350125232607e-05, + "loss": 1.0166, + "step": 6398 + }, + { + "epoch": 0.2, + "learning_rate": 1.8442946231885283e-05, + "loss": 1.0034, + "step": 6399 + }, + { + "epoch": 0.2, + "learning_rate": 1.8442391120861024e-05, + "loss": 1.0938, + "step": 6400 + }, + { + "epoch": 0.2, + "learning_rate": 1.8441835919259253e-05, + "loss": 1.0669, + "step": 6401 + }, + { + "epoch": 0.2, + "learning_rate": 1.844128062708593e-05, + "loss": 1.0278, + "step": 6402 + }, + { + "epoch": 0.2, + "learning_rate": 1.8440725244347006e-05, + "loss": 1.1221, + "step": 6403 + }, + { + "epoch": 0.2, + "learning_rate": 1.8440169771048436e-05, + "loss": 1.0728, + "step": 6404 + }, + { + "epoch": 0.2, + "learning_rate": 1.8439614207196192e-05, + "loss": 0.9595, + "step": 6405 + }, + { + "epoch": 0.2, + "learning_rate": 1.843905855279623e-05, + "loss": 1.1328, + "step": 6406 + }, + { + "epoch": 0.2, + "learning_rate": 1.8438502807854504e-05, + "loss": 1.0933, + "step": 6407 + }, + { + "epoch": 0.2, + "learning_rate": 1.8437946972376982e-05, + "loss": 1.2075, + "step": 6408 + }, + { + "epoch": 0.2, + "learning_rate": 1.843739104636963e-05, + "loss": 1.0239, + "step": 6409 + }, + { + "epoch": 0.2, + "learning_rate": 1.843683502983841e-05, + "loss": 1.0356, + "step": 6410 + }, + { + "epoch": 0.21, + "learning_rate": 1.843627892278929e-05, + "loss": 1.1318, + "step": 6411 + }, + { + "epoch": 0.21, + "learning_rate": 1.843572272522823e-05, + "loss": 1.0107, + "step": 6412 + }, + { + "epoch": 0.21, + "learning_rate": 1.84351664371612e-05, + "loss": 1.1377, + "step": 6413 + }, + { + "epoch": 0.21, + "learning_rate": 1.8434610058594177e-05, + "loss": 1.1084, + "step": 6414 + }, + { + "epoch": 0.21, + "learning_rate": 1.8434053589533117e-05, + "loss": 1.1055, + "step": 6415 + }, + { + "epoch": 0.21, + "learning_rate": 1.8433497029983995e-05, + "loss": 1.0376, + "step": 6416 + }, + { + "epoch": 0.21, + "learning_rate": 1.8432940379952787e-05, + "loss": 1.1069, + "step": 6417 + }, + { + "epoch": 0.21, + "learning_rate": 1.843238363944546e-05, + "loss": 1.0874, + "step": 6418 + }, + { + "epoch": 0.21, + "learning_rate": 1.843182680846799e-05, + "loss": 1.0161, + "step": 6419 + }, + { + "epoch": 0.21, + "learning_rate": 1.8431269887026353e-05, + "loss": 1.064, + "step": 6420 + }, + { + "epoch": 0.21, + "learning_rate": 1.843071287512652e-05, + "loss": 1.0737, + "step": 6421 + }, + { + "epoch": 0.21, + "learning_rate": 1.843015577277447e-05, + "loss": 1.019, + "step": 6422 + }, + { + "epoch": 0.21, + "learning_rate": 1.842959857997617e-05, + "loss": 1.0049, + "step": 6423 + }, + { + "epoch": 0.21, + "learning_rate": 1.8429041296737615e-05, + "loss": 1.1235, + "step": 6424 + }, + { + "epoch": 0.21, + "learning_rate": 1.8428483923064774e-05, + "loss": 1.1074, + "step": 6425 + }, + { + "epoch": 0.21, + "learning_rate": 1.8427926458963628e-05, + "loss": 1.0698, + "step": 6426 + }, + { + "epoch": 0.21, + "learning_rate": 1.8427368904440156e-05, + "loss": 1.144, + "step": 6427 + }, + { + "epoch": 0.21, + "learning_rate": 1.842681125950035e-05, + "loss": 1.1431, + "step": 6428 + }, + { + "epoch": 0.21, + "learning_rate": 1.8426253524150176e-05, + "loss": 1.0728, + "step": 6429 + }, + { + "epoch": 0.21, + "learning_rate": 1.8425695698395635e-05, + "loss": 1.0474, + "step": 6430 + }, + { + "epoch": 0.21, + "learning_rate": 1.8425137782242698e-05, + "loss": 1.1597, + "step": 6431 + }, + { + "epoch": 0.21, + "learning_rate": 1.8424579775697363e-05, + "loss": 1.2402, + "step": 6432 + }, + { + "epoch": 0.21, + "learning_rate": 1.8424021678765606e-05, + "loss": 1.1338, + "step": 6433 + }, + { + "epoch": 0.21, + "learning_rate": 1.8423463491453425e-05, + "loss": 1.0781, + "step": 6434 + }, + { + "epoch": 0.21, + "learning_rate": 1.84229052137668e-05, + "loss": 0.9844, + "step": 6435 + }, + { + "epoch": 0.21, + "learning_rate": 1.8422346845711722e-05, + "loss": 1.0571, + "step": 6436 + }, + { + "epoch": 0.21, + "learning_rate": 1.8421788387294188e-05, + "loss": 1.1626, + "step": 6437 + }, + { + "epoch": 0.21, + "learning_rate": 1.8421229838520183e-05, + "loss": 1.0972, + "step": 6438 + }, + { + "epoch": 0.21, + "learning_rate": 1.8420671199395705e-05, + "loss": 1.1401, + "step": 6439 + }, + { + "epoch": 0.21, + "learning_rate": 1.842011246992674e-05, + "loss": 1.126, + "step": 6440 + }, + { + "epoch": 0.21, + "learning_rate": 1.841955365011929e-05, + "loss": 0.9141, + "step": 6441 + }, + { + "epoch": 0.21, + "learning_rate": 1.8418994739979344e-05, + "loss": 1.063, + "step": 6442 + }, + { + "epoch": 0.21, + "learning_rate": 1.8418435739512908e-05, + "loss": 1.1562, + "step": 6443 + }, + { + "epoch": 0.21, + "learning_rate": 1.841787664872597e-05, + "loss": 1.0249, + "step": 6444 + }, + { + "epoch": 0.21, + "learning_rate": 1.8417317467624532e-05, + "loss": 0.9746, + "step": 6445 + }, + { + "epoch": 0.21, + "learning_rate": 1.8416758196214594e-05, + "loss": 1.0732, + "step": 6446 + }, + { + "epoch": 0.21, + "learning_rate": 1.8416198834502155e-05, + "loss": 1.0835, + "step": 6447 + }, + { + "epoch": 0.21, + "learning_rate": 1.8415639382493218e-05, + "loss": 1.1367, + "step": 6448 + }, + { + "epoch": 0.21, + "learning_rate": 1.8415079840193785e-05, + "loss": 0.9521, + "step": 6449 + }, + { + "epoch": 0.21, + "learning_rate": 1.841452020760986e-05, + "loss": 0.9731, + "step": 6450 + }, + { + "epoch": 0.21, + "learning_rate": 1.841396048474744e-05, + "loss": 1.1499, + "step": 6451 + }, + { + "epoch": 0.21, + "learning_rate": 1.841340067161254e-05, + "loss": 1.1138, + "step": 6452 + }, + { + "epoch": 0.21, + "learning_rate": 1.8412840768211162e-05, + "loss": 1.1294, + "step": 6453 + }, + { + "epoch": 0.21, + "learning_rate": 1.8412280774549318e-05, + "loss": 1.103, + "step": 6454 + }, + { + "epoch": 0.21, + "learning_rate": 1.8411720690633003e-05, + "loss": 1.0215, + "step": 6455 + }, + { + "epoch": 0.21, + "learning_rate": 1.8411160516468238e-05, + "loss": 1.0859, + "step": 6456 + }, + { + "epoch": 0.21, + "learning_rate": 1.8410600252061033e-05, + "loss": 1.1113, + "step": 6457 + }, + { + "epoch": 0.21, + "learning_rate": 1.8410039897417393e-05, + "loss": 1.127, + "step": 6458 + }, + { + "epoch": 0.21, + "learning_rate": 1.840947945254333e-05, + "loss": 1.0977, + "step": 6459 + }, + { + "epoch": 0.21, + "learning_rate": 1.8408918917444863e-05, + "loss": 1.0723, + "step": 6460 + }, + { + "epoch": 0.21, + "learning_rate": 1.8408358292128002e-05, + "loss": 1.1523, + "step": 6461 + }, + { + "epoch": 0.21, + "learning_rate": 1.840779757659876e-05, + "loss": 1.0425, + "step": 6462 + }, + { + "epoch": 0.21, + "learning_rate": 1.840723677086316e-05, + "loss": 1.2246, + "step": 6463 + }, + { + "epoch": 0.21, + "learning_rate": 1.8406675874927212e-05, + "loss": 1.1245, + "step": 6464 + }, + { + "epoch": 0.21, + "learning_rate": 1.8406114888796938e-05, + "loss": 1.0396, + "step": 6465 + }, + { + "epoch": 0.21, + "learning_rate": 1.840555381247835e-05, + "loss": 1.0771, + "step": 6466 + }, + { + "epoch": 0.21, + "learning_rate": 1.8404992645977475e-05, + "loss": 1.0688, + "step": 6467 + }, + { + "epoch": 0.21, + "learning_rate": 1.840443138930033e-05, + "loss": 1.2656, + "step": 6468 + }, + { + "epoch": 0.21, + "learning_rate": 1.8403870042452942e-05, + "loss": 0.9248, + "step": 6469 + }, + { + "epoch": 0.21, + "learning_rate": 1.8403308605441325e-05, + "loss": 0.8994, + "step": 6470 + }, + { + "epoch": 0.21, + "learning_rate": 1.8402747078271507e-05, + "loss": 1.2036, + "step": 6471 + }, + { + "epoch": 0.21, + "learning_rate": 1.8402185460949513e-05, + "loss": 1.1738, + "step": 6472 + }, + { + "epoch": 0.21, + "learning_rate": 1.840162375348137e-05, + "loss": 1.1104, + "step": 6473 + }, + { + "epoch": 0.21, + "learning_rate": 1.8401061955873102e-05, + "loss": 1.1284, + "step": 6474 + }, + { + "epoch": 0.21, + "learning_rate": 1.8400500068130737e-05, + "loss": 1.1523, + "step": 6475 + }, + { + "epoch": 0.21, + "learning_rate": 1.8399938090260302e-05, + "loss": 1.0068, + "step": 6476 + }, + { + "epoch": 0.21, + "learning_rate": 1.839937602226783e-05, + "loss": 1.002, + "step": 6477 + }, + { + "epoch": 0.21, + "learning_rate": 1.8398813864159347e-05, + "loss": 1.1118, + "step": 6478 + }, + { + "epoch": 0.21, + "learning_rate": 1.8398251615940888e-05, + "loss": 1.1738, + "step": 6479 + }, + { + "epoch": 0.21, + "learning_rate": 1.8397689277618482e-05, + "loss": 1.1865, + "step": 6480 + }, + { + "epoch": 0.21, + "learning_rate": 1.8397126849198165e-05, + "loss": 1.1797, + "step": 6481 + }, + { + "epoch": 0.21, + "learning_rate": 1.839656433068597e-05, + "loss": 1.144, + "step": 6482 + }, + { + "epoch": 0.21, + "learning_rate": 1.8396001722087935e-05, + "loss": 1.1499, + "step": 6483 + }, + { + "epoch": 0.21, + "learning_rate": 1.839543902341009e-05, + "loss": 1.0449, + "step": 6484 + }, + { + "epoch": 0.21, + "learning_rate": 1.839487623465848e-05, + "loss": 1.1113, + "step": 6485 + }, + { + "epoch": 0.21, + "learning_rate": 1.8394313355839134e-05, + "loss": 0.9819, + "step": 6486 + }, + { + "epoch": 0.21, + "learning_rate": 1.83937503869581e-05, + "loss": 1.1079, + "step": 6487 + }, + { + "epoch": 0.21, + "learning_rate": 1.8393187328021413e-05, + "loss": 1.187, + "step": 6488 + }, + { + "epoch": 0.21, + "learning_rate": 1.8392624179035118e-05, + "loss": 1.1421, + "step": 6489 + }, + { + "epoch": 0.21, + "learning_rate": 1.839206094000525e-05, + "loss": 1.0962, + "step": 6490 + }, + { + "epoch": 0.21, + "learning_rate": 1.8391497610937858e-05, + "loss": 0.9585, + "step": 6491 + }, + { + "epoch": 0.21, + "learning_rate": 1.8390934191838983e-05, + "loss": 0.9634, + "step": 6492 + }, + { + "epoch": 0.21, + "learning_rate": 1.839037068271467e-05, + "loss": 1.1001, + "step": 6493 + }, + { + "epoch": 0.21, + "learning_rate": 1.8389807083570968e-05, + "loss": 1.0591, + "step": 6494 + }, + { + "epoch": 0.21, + "learning_rate": 1.838924339441392e-05, + "loss": 1.0659, + "step": 6495 + }, + { + "epoch": 0.21, + "learning_rate": 1.8388679615249576e-05, + "loss": 1.0947, + "step": 6496 + }, + { + "epoch": 0.21, + "learning_rate": 1.838811574608398e-05, + "loss": 0.9937, + "step": 6497 + }, + { + "epoch": 0.21, + "learning_rate": 1.838755178692319e-05, + "loss": 1.1689, + "step": 6498 + }, + { + "epoch": 0.21, + "learning_rate": 1.838698773777325e-05, + "loss": 1.1372, + "step": 6499 + }, + { + "epoch": 0.21, + "learning_rate": 1.8386423598640214e-05, + "loss": 1.1528, + "step": 6500 + }, + { + "epoch": 0.21, + "learning_rate": 1.838585936953014e-05, + "loss": 1.1289, + "step": 6501 + }, + { + "epoch": 0.21, + "learning_rate": 1.8385295050449067e-05, + "loss": 1.0625, + "step": 6502 + }, + { + "epoch": 0.21, + "learning_rate": 1.838473064140306e-05, + "loss": 0.9966, + "step": 6503 + }, + { + "epoch": 0.21, + "learning_rate": 1.8384166142398178e-05, + "loss": 0.9512, + "step": 6504 + }, + { + "epoch": 0.21, + "learning_rate": 1.838360155344047e-05, + "loss": 1.1069, + "step": 6505 + }, + { + "epoch": 0.21, + "learning_rate": 1.8383036874535997e-05, + "loss": 1.0933, + "step": 6506 + }, + { + "epoch": 0.21, + "learning_rate": 1.8382472105690815e-05, + "loss": 1.0864, + "step": 6507 + }, + { + "epoch": 0.21, + "learning_rate": 1.8381907246910985e-05, + "loss": 1.1465, + "step": 6508 + }, + { + "epoch": 0.21, + "learning_rate": 1.8381342298202565e-05, + "loss": 1.1099, + "step": 6509 + }, + { + "epoch": 0.21, + "learning_rate": 1.838077725957162e-05, + "loss": 1.0962, + "step": 6510 + }, + { + "epoch": 0.21, + "learning_rate": 1.838021213102421e-05, + "loss": 0.9995, + "step": 6511 + }, + { + "epoch": 0.21, + "learning_rate": 1.8379646912566394e-05, + "loss": 1.0381, + "step": 6512 + }, + { + "epoch": 0.21, + "learning_rate": 1.8379081604204247e-05, + "loss": 1.0264, + "step": 6513 + }, + { + "epoch": 0.21, + "learning_rate": 1.8378516205943827e-05, + "loss": 1.0415, + "step": 6514 + }, + { + "epoch": 0.21, + "learning_rate": 1.83779507177912e-05, + "loss": 1.0439, + "step": 6515 + }, + { + "epoch": 0.21, + "learning_rate": 1.8377385139752434e-05, + "loss": 1.0449, + "step": 6516 + }, + { + "epoch": 0.21, + "learning_rate": 1.8376819471833596e-05, + "loss": 1.0854, + "step": 6517 + }, + { + "epoch": 0.21, + "learning_rate": 1.837625371404076e-05, + "loss": 1.144, + "step": 6518 + }, + { + "epoch": 0.21, + "learning_rate": 1.8375687866379988e-05, + "loss": 1.1514, + "step": 6519 + }, + { + "epoch": 0.21, + "learning_rate": 1.8375121928857357e-05, + "loss": 1.0972, + "step": 6520 + }, + { + "epoch": 0.21, + "learning_rate": 1.8374555901478936e-05, + "loss": 0.9399, + "step": 6521 + }, + { + "epoch": 0.21, + "learning_rate": 1.8373989784250798e-05, + "loss": 1.0825, + "step": 6522 + }, + { + "epoch": 0.21, + "learning_rate": 1.837342357717902e-05, + "loss": 1.0806, + "step": 6523 + }, + { + "epoch": 0.21, + "learning_rate": 1.8372857280269674e-05, + "loss": 1.0923, + "step": 6524 + }, + { + "epoch": 0.21, + "learning_rate": 1.837229089352883e-05, + "loss": 1.1807, + "step": 6525 + }, + { + "epoch": 0.21, + "learning_rate": 1.8371724416962578e-05, + "loss": 1.0903, + "step": 6526 + }, + { + "epoch": 0.21, + "learning_rate": 1.8371157850576984e-05, + "loss": 1.187, + "step": 6527 + }, + { + "epoch": 0.21, + "learning_rate": 1.8370591194378136e-05, + "loss": 1.0361, + "step": 6528 + }, + { + "epoch": 0.21, + "learning_rate": 1.8370024448372104e-05, + "loss": 1.0054, + "step": 6529 + }, + { + "epoch": 0.21, + "learning_rate": 1.8369457612564974e-05, + "loss": 1.1543, + "step": 6530 + }, + { + "epoch": 0.21, + "learning_rate": 1.8368890686962828e-05, + "loss": 1.1934, + "step": 6531 + }, + { + "epoch": 0.21, + "learning_rate": 1.8368323671571744e-05, + "loss": 0.9546, + "step": 6532 + }, + { + "epoch": 0.21, + "learning_rate": 1.8367756566397813e-05, + "loss": 1.1094, + "step": 6533 + }, + { + "epoch": 0.21, + "learning_rate": 1.836718937144711e-05, + "loss": 0.6055, + "step": 6534 + }, + { + "epoch": 0.21, + "learning_rate": 1.8366622086725733e-05, + "loss": 1.1582, + "step": 6535 + }, + { + "epoch": 0.21, + "learning_rate": 1.836605471223975e-05, + "loss": 0.9829, + "step": 6536 + }, + { + "epoch": 0.21, + "learning_rate": 1.836548724799527e-05, + "loss": 1.1396, + "step": 6537 + }, + { + "epoch": 0.21, + "learning_rate": 1.8364919693998362e-05, + "loss": 1.1553, + "step": 6538 + }, + { + "epoch": 0.21, + "learning_rate": 1.8364352050255127e-05, + "loss": 1.1602, + "step": 6539 + }, + { + "epoch": 0.21, + "learning_rate": 1.8363784316771652e-05, + "loss": 1.0732, + "step": 6540 + }, + { + "epoch": 0.21, + "learning_rate": 1.8363216493554028e-05, + "loss": 1.0122, + "step": 6541 + }, + { + "epoch": 0.21, + "learning_rate": 1.8362648580608346e-05, + "loss": 1.0742, + "step": 6542 + }, + { + "epoch": 0.21, + "learning_rate": 1.83620805779407e-05, + "loss": 1.0361, + "step": 6543 + }, + { + "epoch": 0.21, + "learning_rate": 1.8361512485557185e-05, + "loss": 1.1743, + "step": 6544 + }, + { + "epoch": 0.21, + "learning_rate": 1.8360944303463895e-05, + "loss": 1.0308, + "step": 6545 + }, + { + "epoch": 0.21, + "learning_rate": 1.8360376031666925e-05, + "loss": 1.0288, + "step": 6546 + }, + { + "epoch": 0.21, + "learning_rate": 1.8359807670172372e-05, + "loss": 1.1968, + "step": 6547 + }, + { + "epoch": 0.21, + "learning_rate": 1.8359239218986337e-05, + "loss": 1.0986, + "step": 6548 + }, + { + "epoch": 0.21, + "learning_rate": 1.8358670678114915e-05, + "loss": 1.1294, + "step": 6549 + }, + { + "epoch": 0.21, + "learning_rate": 1.8358102047564212e-05, + "loss": 1.0967, + "step": 6550 + }, + { + "epoch": 0.21, + "learning_rate": 1.835753332734032e-05, + "loss": 1.0229, + "step": 6551 + }, + { + "epoch": 0.21, + "learning_rate": 1.8356964517449346e-05, + "loss": 0.9912, + "step": 6552 + }, + { + "epoch": 0.21, + "learning_rate": 1.8356395617897395e-05, + "loss": 1.1929, + "step": 6553 + }, + { + "epoch": 0.21, + "learning_rate": 1.8355826628690563e-05, + "loss": 1.0806, + "step": 6554 + }, + { + "epoch": 0.21, + "learning_rate": 1.835525754983496e-05, + "loss": 1.1475, + "step": 6555 + }, + { + "epoch": 0.21, + "learning_rate": 1.835468838133669e-05, + "loss": 1.1128, + "step": 6556 + }, + { + "epoch": 0.21, + "learning_rate": 1.835411912320186e-05, + "loss": 1.103, + "step": 6557 + }, + { + "epoch": 0.21, + "learning_rate": 1.835354977543658e-05, + "loss": 1.0757, + "step": 6558 + }, + { + "epoch": 0.21, + "learning_rate": 1.8352980338046953e-05, + "loss": 1.0933, + "step": 6559 + }, + { + "epoch": 0.21, + "learning_rate": 1.8352410811039097e-05, + "loss": 1.0815, + "step": 6560 + }, + { + "epoch": 0.21, + "learning_rate": 1.835184119441911e-05, + "loss": 1.0757, + "step": 6561 + }, + { + "epoch": 0.21, + "learning_rate": 1.8351271488193117e-05, + "loss": 0.9043, + "step": 6562 + }, + { + "epoch": 0.21, + "learning_rate": 1.835070169236722e-05, + "loss": 1.0776, + "step": 6563 + }, + { + "epoch": 0.21, + "learning_rate": 1.8350131806947537e-05, + "loss": 1.1465, + "step": 6564 + }, + { + "epoch": 0.21, + "learning_rate": 1.834956183194018e-05, + "loss": 1.0308, + "step": 6565 + }, + { + "epoch": 0.21, + "learning_rate": 1.834899176735127e-05, + "loss": 0.9985, + "step": 6566 + }, + { + "epoch": 0.21, + "learning_rate": 1.834842161318691e-05, + "loss": 0.9282, + "step": 6567 + }, + { + "epoch": 0.21, + "learning_rate": 1.834785136945323e-05, + "loss": 1.2046, + "step": 6568 + }, + { + "epoch": 0.21, + "learning_rate": 1.8347281036156344e-05, + "loss": 1.1274, + "step": 6569 + }, + { + "epoch": 0.21, + "learning_rate": 1.8346710613302373e-05, + "loss": 1.0435, + "step": 6570 + }, + { + "epoch": 0.21, + "learning_rate": 1.8346140100897433e-05, + "loss": 0.8911, + "step": 6571 + }, + { + "epoch": 0.21, + "learning_rate": 1.8345569498947645e-05, + "loss": 1.085, + "step": 6572 + }, + { + "epoch": 0.21, + "learning_rate": 1.8344998807459133e-05, + "loss": 1.0493, + "step": 6573 + }, + { + "epoch": 0.21, + "learning_rate": 1.8344428026438023e-05, + "loss": 1.0581, + "step": 6574 + }, + { + "epoch": 0.21, + "learning_rate": 1.8343857155890432e-05, + "loss": 1.1094, + "step": 6575 + }, + { + "epoch": 0.21, + "learning_rate": 1.834328619582249e-05, + "loss": 1.1069, + "step": 6576 + }, + { + "epoch": 0.21, + "learning_rate": 1.834271514624032e-05, + "loss": 1.0479, + "step": 6577 + }, + { + "epoch": 0.21, + "learning_rate": 1.834214400715005e-05, + "loss": 0.5857, + "step": 6578 + }, + { + "epoch": 0.21, + "learning_rate": 1.834157277855781e-05, + "loss": 1.1211, + "step": 6579 + }, + { + "epoch": 0.21, + "learning_rate": 1.834100146046972e-05, + "loss": 1.1265, + "step": 6580 + }, + { + "epoch": 0.21, + "learning_rate": 1.8340430052891922e-05, + "loss": 1.1577, + "step": 6581 + }, + { + "epoch": 0.21, + "learning_rate": 1.8339858555830537e-05, + "loss": 1.0288, + "step": 6582 + }, + { + "epoch": 0.21, + "learning_rate": 1.8339286969291698e-05, + "loss": 1.0024, + "step": 6583 + }, + { + "epoch": 0.21, + "learning_rate": 1.8338715293281545e-05, + "loss": 1.0684, + "step": 6584 + }, + { + "epoch": 0.21, + "learning_rate": 1.8338143527806203e-05, + "loss": 1.1035, + "step": 6585 + }, + { + "epoch": 0.21, + "learning_rate": 1.8337571672871805e-05, + "loss": 0.9385, + "step": 6586 + }, + { + "epoch": 0.21, + "learning_rate": 1.8336999728484498e-05, + "loss": 1.0586, + "step": 6587 + }, + { + "epoch": 0.21, + "learning_rate": 1.8336427694650405e-05, + "loss": 1.0869, + "step": 6588 + }, + { + "epoch": 0.21, + "learning_rate": 1.8335855571375672e-05, + "loss": 0.9624, + "step": 6589 + }, + { + "epoch": 0.21, + "learning_rate": 1.8335283358666432e-05, + "loss": 0.9922, + "step": 6590 + }, + { + "epoch": 0.21, + "learning_rate": 1.833471105652883e-05, + "loss": 1.0566, + "step": 6591 + }, + { + "epoch": 0.21, + "learning_rate": 1.8334138664969e-05, + "loss": 1.0908, + "step": 6592 + }, + { + "epoch": 0.21, + "learning_rate": 1.8333566183993087e-05, + "loss": 1.084, + "step": 6593 + }, + { + "epoch": 0.21, + "learning_rate": 1.833299361360723e-05, + "loss": 1.0479, + "step": 6594 + }, + { + "epoch": 0.21, + "learning_rate": 1.8332420953817577e-05, + "loss": 0.9297, + "step": 6595 + }, + { + "epoch": 0.21, + "learning_rate": 1.8331848204630266e-05, + "loss": 0.9541, + "step": 6596 + }, + { + "epoch": 0.21, + "learning_rate": 1.8331275366051448e-05, + "loss": 1.0981, + "step": 6597 + }, + { + "epoch": 0.21, + "learning_rate": 1.8330702438087266e-05, + "loss": 1.1182, + "step": 6598 + }, + { + "epoch": 0.21, + "learning_rate": 1.8330129420743862e-05, + "loss": 1.0, + "step": 6599 + }, + { + "epoch": 0.21, + "learning_rate": 1.8329556314027393e-05, + "loss": 0.9932, + "step": 6600 + }, + { + "epoch": 0.21, + "learning_rate": 1.8328983117944005e-05, + "loss": 0.9727, + "step": 6601 + }, + { + "epoch": 0.21, + "learning_rate": 1.8328409832499844e-05, + "loss": 1.0938, + "step": 6602 + }, + { + "epoch": 0.21, + "learning_rate": 1.832783645770106e-05, + "loss": 1.0649, + "step": 6603 + }, + { + "epoch": 0.21, + "learning_rate": 1.832726299355381e-05, + "loss": 1.1313, + "step": 6604 + }, + { + "epoch": 0.21, + "learning_rate": 1.8326689440064247e-05, + "loss": 1.0547, + "step": 6605 + }, + { + "epoch": 0.21, + "learning_rate": 1.8326115797238518e-05, + "loss": 1.0859, + "step": 6606 + }, + { + "epoch": 0.21, + "learning_rate": 1.8325542065082784e-05, + "loss": 1.0288, + "step": 6607 + }, + { + "epoch": 0.21, + "learning_rate": 1.8324968243603197e-05, + "loss": 1.0312, + "step": 6608 + }, + { + "epoch": 0.21, + "learning_rate": 1.8324394332805913e-05, + "loss": 1.0532, + "step": 6609 + }, + { + "epoch": 0.21, + "learning_rate": 1.832382033269709e-05, + "loss": 1.0557, + "step": 6610 + }, + { + "epoch": 0.21, + "learning_rate": 1.832324624328289e-05, + "loss": 0.9814, + "step": 6611 + }, + { + "epoch": 0.21, + "learning_rate": 1.8322672064569467e-05, + "loss": 1.1055, + "step": 6612 + }, + { + "epoch": 0.21, + "learning_rate": 1.8322097796562988e-05, + "loss": 1.0957, + "step": 6613 + }, + { + "epoch": 0.21, + "learning_rate": 1.8321523439269606e-05, + "loss": 1.1772, + "step": 6614 + }, + { + "epoch": 0.21, + "learning_rate": 1.832094899269549e-05, + "loss": 1.1489, + "step": 6615 + }, + { + "epoch": 0.21, + "learning_rate": 1.8320374456846797e-05, + "loss": 1.1323, + "step": 6616 + }, + { + "epoch": 0.21, + "learning_rate": 1.83197998317297e-05, + "loss": 0.9741, + "step": 6617 + }, + { + "epoch": 0.21, + "learning_rate": 1.8319225117350354e-05, + "loss": 1.0532, + "step": 6618 + }, + { + "epoch": 0.21, + "learning_rate": 1.8318650313714933e-05, + "loss": 1.0898, + "step": 6619 + }, + { + "epoch": 0.21, + "learning_rate": 1.83180754208296e-05, + "loss": 1.1392, + "step": 6620 + }, + { + "epoch": 0.21, + "learning_rate": 1.831750043870052e-05, + "loss": 1.0552, + "step": 6621 + }, + { + "epoch": 0.21, + "learning_rate": 1.8316925367333865e-05, + "loss": 1.1577, + "step": 6622 + }, + { + "epoch": 0.21, + "learning_rate": 1.8316350206735812e-05, + "loss": 1.0063, + "step": 6623 + }, + { + "epoch": 0.21, + "learning_rate": 1.8315774956912525e-05, + "loss": 1.1094, + "step": 6624 + }, + { + "epoch": 0.21, + "learning_rate": 1.831519961787017e-05, + "loss": 1.1123, + "step": 6625 + }, + { + "epoch": 0.21, + "learning_rate": 1.831462418961493e-05, + "loss": 1.1343, + "step": 6626 + }, + { + "epoch": 0.21, + "learning_rate": 1.8314048672152976e-05, + "loss": 1.1221, + "step": 6627 + }, + { + "epoch": 0.21, + "learning_rate": 1.8313473065490476e-05, + "loss": 1.1929, + "step": 6628 + }, + { + "epoch": 0.21, + "learning_rate": 1.8312897369633615e-05, + "loss": 0.9507, + "step": 6629 + }, + { + "epoch": 0.21, + "learning_rate": 1.8312321584588565e-05, + "loss": 1.0435, + "step": 6630 + }, + { + "epoch": 0.21, + "learning_rate": 1.8311745710361504e-05, + "loss": 1.0591, + "step": 6631 + }, + { + "epoch": 0.21, + "learning_rate": 1.831116974695861e-05, + "loss": 0.9956, + "step": 6632 + }, + { + "epoch": 0.21, + "learning_rate": 1.831059369438606e-05, + "loss": 1.0103, + "step": 6633 + }, + { + "epoch": 0.21, + "learning_rate": 1.8310017552650044e-05, + "loss": 1.0913, + "step": 6634 + }, + { + "epoch": 0.21, + "learning_rate": 1.8309441321756734e-05, + "loss": 1.0, + "step": 6635 + }, + { + "epoch": 0.21, + "learning_rate": 1.8308865001712312e-05, + "loss": 1.0786, + "step": 6636 + }, + { + "epoch": 0.21, + "learning_rate": 1.8308288592522968e-05, + "loss": 1.105, + "step": 6637 + }, + { + "epoch": 0.21, + "learning_rate": 1.8307712094194882e-05, + "loss": 0.9834, + "step": 6638 + }, + { + "epoch": 0.21, + "learning_rate": 1.8307135506734237e-05, + "loss": 1.1309, + "step": 6639 + }, + { + "epoch": 0.21, + "learning_rate": 1.8306558830147224e-05, + "loss": 0.8823, + "step": 6640 + }, + { + "epoch": 0.21, + "learning_rate": 1.8305982064440025e-05, + "loss": 1.1309, + "step": 6641 + }, + { + "epoch": 0.21, + "learning_rate": 1.8305405209618834e-05, + "loss": 1.0664, + "step": 6642 + }, + { + "epoch": 0.21, + "learning_rate": 1.830482826568984e-05, + "loss": 1.0879, + "step": 6643 + }, + { + "epoch": 0.21, + "learning_rate": 1.830425123265923e-05, + "loss": 1.064, + "step": 6644 + }, + { + "epoch": 0.21, + "learning_rate": 1.830367411053319e-05, + "loss": 0.9321, + "step": 6645 + }, + { + "epoch": 0.21, + "learning_rate": 1.830309689931792e-05, + "loss": 0.9917, + "step": 6646 + }, + { + "epoch": 0.21, + "learning_rate": 1.8302519599019605e-05, + "loss": 1.1377, + "step": 6647 + }, + { + "epoch": 0.21, + "learning_rate": 1.830194220964445e-05, + "loss": 1.0693, + "step": 6648 + }, + { + "epoch": 0.21, + "learning_rate": 1.830136473119864e-05, + "loss": 1.104, + "step": 6649 + }, + { + "epoch": 0.21, + "learning_rate": 1.8300787163688372e-05, + "loss": 1.0405, + "step": 6650 + }, + { + "epoch": 0.21, + "learning_rate": 1.8300209507119845e-05, + "loss": 1.1255, + "step": 6651 + }, + { + "epoch": 0.21, + "learning_rate": 1.829963176149926e-05, + "loss": 1.1138, + "step": 6652 + }, + { + "epoch": 0.21, + "learning_rate": 1.829905392683281e-05, + "loss": 1.0674, + "step": 6653 + }, + { + "epoch": 0.21, + "learning_rate": 1.8298476003126695e-05, + "loss": 1.0737, + "step": 6654 + }, + { + "epoch": 0.21, + "learning_rate": 1.8297897990387116e-05, + "loss": 1.0254, + "step": 6655 + }, + { + "epoch": 0.21, + "learning_rate": 1.8297319888620277e-05, + "loss": 1.2363, + "step": 6656 + }, + { + "epoch": 0.21, + "learning_rate": 1.8296741697832382e-05, + "loss": 1.083, + "step": 6657 + }, + { + "epoch": 0.21, + "learning_rate": 1.8296163418029626e-05, + "loss": 1.0234, + "step": 6658 + }, + { + "epoch": 0.21, + "learning_rate": 1.829558504921822e-05, + "loss": 1.0933, + "step": 6659 + }, + { + "epoch": 0.21, + "learning_rate": 1.8295006591404367e-05, + "loss": 1.0571, + "step": 6660 + }, + { + "epoch": 0.21, + "learning_rate": 1.8294428044594274e-05, + "loss": 1.0171, + "step": 6661 + }, + { + "epoch": 0.21, + "learning_rate": 1.8293849408794148e-05, + "loss": 1.0244, + "step": 6662 + }, + { + "epoch": 0.21, + "learning_rate": 1.82932706840102e-05, + "loss": 0.9946, + "step": 6663 + }, + { + "epoch": 0.21, + "learning_rate": 1.829269187024863e-05, + "loss": 1.0674, + "step": 6664 + }, + { + "epoch": 0.21, + "learning_rate": 1.8292112967515656e-05, + "loss": 1.2148, + "step": 6665 + }, + { + "epoch": 0.21, + "learning_rate": 1.829153397581749e-05, + "loss": 1.0845, + "step": 6666 + }, + { + "epoch": 0.21, + "learning_rate": 1.829095489516034e-05, + "loss": 0.9961, + "step": 6667 + }, + { + "epoch": 0.21, + "learning_rate": 1.8290375725550417e-05, + "loss": 1.0381, + "step": 6668 + }, + { + "epoch": 0.21, + "learning_rate": 1.828979646699394e-05, + "loss": 1.0498, + "step": 6669 + }, + { + "epoch": 0.21, + "learning_rate": 1.8289217119497125e-05, + "loss": 1.1206, + "step": 6670 + }, + { + "epoch": 0.21, + "learning_rate": 1.828863768306618e-05, + "loss": 1.125, + "step": 6671 + }, + { + "epoch": 0.21, + "learning_rate": 1.8288058157707327e-05, + "loss": 1.1328, + "step": 6672 + }, + { + "epoch": 0.21, + "learning_rate": 1.828747854342678e-05, + "loss": 1.106, + "step": 6673 + }, + { + "epoch": 0.21, + "learning_rate": 1.8286898840230765e-05, + "loss": 1.0557, + "step": 6674 + }, + { + "epoch": 0.21, + "learning_rate": 1.8286319048125495e-05, + "loss": 1.0591, + "step": 6675 + }, + { + "epoch": 0.21, + "learning_rate": 1.8285739167117195e-05, + "loss": 1.1558, + "step": 6676 + }, + { + "epoch": 0.21, + "learning_rate": 1.8285159197212083e-05, + "loss": 1.1265, + "step": 6677 + }, + { + "epoch": 0.21, + "learning_rate": 1.8284579138416378e-05, + "loss": 1.0967, + "step": 6678 + }, + { + "epoch": 0.21, + "learning_rate": 1.8283998990736313e-05, + "loss": 1.0737, + "step": 6679 + }, + { + "epoch": 0.21, + "learning_rate": 1.8283418754178103e-05, + "loss": 1.0532, + "step": 6680 + }, + { + "epoch": 0.21, + "learning_rate": 1.828283842874798e-05, + "loss": 1.0371, + "step": 6681 + }, + { + "epoch": 0.21, + "learning_rate": 1.8282258014452165e-05, + "loss": 0.979, + "step": 6682 + }, + { + "epoch": 0.21, + "learning_rate": 1.828167751129689e-05, + "loss": 1.105, + "step": 6683 + }, + { + "epoch": 0.21, + "learning_rate": 1.828109691928838e-05, + "loss": 1.1104, + "step": 6684 + }, + { + "epoch": 0.21, + "learning_rate": 1.828051623843287e-05, + "loss": 1.0542, + "step": 6685 + }, + { + "epoch": 0.21, + "learning_rate": 1.8279935468736578e-05, + "loss": 0.9546, + "step": 6686 + }, + { + "epoch": 0.21, + "learning_rate": 1.8279354610205744e-05, + "loss": 0.9478, + "step": 6687 + }, + { + "epoch": 0.21, + "learning_rate": 1.82787736628466e-05, + "loss": 1.0225, + "step": 6688 + }, + { + "epoch": 0.21, + "learning_rate": 1.8278192626665378e-05, + "loss": 1.1147, + "step": 6689 + }, + { + "epoch": 0.21, + "learning_rate": 1.827761150166831e-05, + "loss": 0.9893, + "step": 6690 + }, + { + "epoch": 0.21, + "learning_rate": 1.827703028786163e-05, + "loss": 1.0518, + "step": 6691 + }, + { + "epoch": 0.21, + "learning_rate": 1.8276448985251578e-05, + "loss": 1.1167, + "step": 6692 + }, + { + "epoch": 0.21, + "learning_rate": 1.8275867593844386e-05, + "loss": 1.1211, + "step": 6693 + }, + { + "epoch": 0.21, + "learning_rate": 1.8275286113646295e-05, + "loss": 1.0122, + "step": 6694 + }, + { + "epoch": 0.21, + "learning_rate": 1.8274704544663545e-05, + "loss": 1.251, + "step": 6695 + }, + { + "epoch": 0.21, + "learning_rate": 1.827412288690237e-05, + "loss": 0.999, + "step": 6696 + }, + { + "epoch": 0.21, + "learning_rate": 1.8273541140369012e-05, + "loss": 1.0332, + "step": 6697 + }, + { + "epoch": 0.21, + "learning_rate": 1.827295930506972e-05, + "loss": 1.1616, + "step": 6698 + }, + { + "epoch": 0.21, + "learning_rate": 1.8272377381010726e-05, + "loss": 1.1426, + "step": 6699 + }, + { + "epoch": 0.21, + "learning_rate": 1.827179536819828e-05, + "loss": 1.084, + "step": 6700 + }, + { + "epoch": 0.21, + "learning_rate": 1.8271213266638624e-05, + "loss": 1.0586, + "step": 6701 + }, + { + "epoch": 0.21, + "learning_rate": 1.8270631076338006e-05, + "loss": 1.1079, + "step": 6702 + }, + { + "epoch": 0.21, + "learning_rate": 1.8270048797302668e-05, + "loss": 1.1362, + "step": 6703 + }, + { + "epoch": 0.21, + "learning_rate": 1.8269466429538856e-05, + "loss": 0.9912, + "step": 6704 + }, + { + "epoch": 0.21, + "learning_rate": 1.8268883973052826e-05, + "loss": 1.0537, + "step": 6705 + }, + { + "epoch": 0.21, + "learning_rate": 1.8268301427850825e-05, + "loss": 1.0454, + "step": 6706 + }, + { + "epoch": 0.21, + "learning_rate": 1.8267718793939093e-05, + "loss": 1.1895, + "step": 6707 + }, + { + "epoch": 0.21, + "learning_rate": 1.8267136071323893e-05, + "loss": 1.0522, + "step": 6708 + }, + { + "epoch": 0.21, + "learning_rate": 1.826655326001147e-05, + "loss": 1.0762, + "step": 6709 + }, + { + "epoch": 0.21, + "learning_rate": 1.8265970360008083e-05, + "loss": 0.9824, + "step": 6710 + }, + { + "epoch": 0.21, + "learning_rate": 1.8265387371319983e-05, + "loss": 1.0391, + "step": 6711 + }, + { + "epoch": 0.21, + "learning_rate": 1.8264804293953422e-05, + "loss": 1.1924, + "step": 6712 + }, + { + "epoch": 0.21, + "learning_rate": 1.8264221127914658e-05, + "loss": 1.1016, + "step": 6713 + }, + { + "epoch": 0.21, + "learning_rate": 1.8263637873209945e-05, + "loss": 0.9873, + "step": 6714 + }, + { + "epoch": 0.21, + "learning_rate": 1.826305452984555e-05, + "loss": 1.0747, + "step": 6715 + }, + { + "epoch": 0.21, + "learning_rate": 1.826247109782772e-05, + "loss": 1.0171, + "step": 6716 + }, + { + "epoch": 0.21, + "learning_rate": 1.8261887577162717e-05, + "loss": 1.0591, + "step": 6717 + }, + { + "epoch": 0.21, + "learning_rate": 1.826130396785681e-05, + "loss": 1.1572, + "step": 6718 + }, + { + "epoch": 0.21, + "learning_rate": 1.8260720269916252e-05, + "loss": 1.1582, + "step": 6719 + }, + { + "epoch": 0.21, + "learning_rate": 1.8260136483347307e-05, + "loss": 1.04, + "step": 6720 + }, + { + "epoch": 0.21, + "learning_rate": 1.8259552608156245e-05, + "loss": 1.021, + "step": 6721 + }, + { + "epoch": 0.21, + "learning_rate": 1.8258968644349318e-05, + "loss": 1.0542, + "step": 6722 + }, + { + "epoch": 0.22, + "learning_rate": 1.82583845919328e-05, + "loss": 1.1162, + "step": 6723 + }, + { + "epoch": 0.22, + "learning_rate": 1.8257800450912958e-05, + "loss": 1.1201, + "step": 6724 + }, + { + "epoch": 0.22, + "learning_rate": 1.825721622129605e-05, + "loss": 0.9917, + "step": 6725 + }, + { + "epoch": 0.22, + "learning_rate": 1.8256631903088357e-05, + "loss": 1.0596, + "step": 6726 + }, + { + "epoch": 0.22, + "learning_rate": 1.8256047496296138e-05, + "loss": 0.9824, + "step": 6727 + }, + { + "epoch": 0.22, + "learning_rate": 1.8255463000925672e-05, + "loss": 1.1426, + "step": 6728 + }, + { + "epoch": 0.22, + "learning_rate": 1.8254878416983217e-05, + "loss": 1.0161, + "step": 6729 + }, + { + "epoch": 0.22, + "learning_rate": 1.825429374447506e-05, + "loss": 0.9683, + "step": 6730 + }, + { + "epoch": 0.22, + "learning_rate": 1.825370898340746e-05, + "loss": 0.9971, + "step": 6731 + }, + { + "epoch": 0.22, + "learning_rate": 1.82531241337867e-05, + "loss": 1.0518, + "step": 6732 + }, + { + "epoch": 0.22, + "learning_rate": 1.8252539195619056e-05, + "loss": 1.0366, + "step": 6733 + }, + { + "epoch": 0.22, + "learning_rate": 1.8251954168910796e-05, + "loss": 0.9932, + "step": 6734 + }, + { + "epoch": 0.22, + "learning_rate": 1.82513690536682e-05, + "loss": 1.106, + "step": 6735 + }, + { + "epoch": 0.22, + "learning_rate": 1.8250783849897547e-05, + "loss": 1.0435, + "step": 6736 + }, + { + "epoch": 0.22, + "learning_rate": 1.825019855760511e-05, + "loss": 1.2002, + "step": 6737 + }, + { + "epoch": 0.22, + "learning_rate": 1.8249613176797183e-05, + "loss": 1.2085, + "step": 6738 + }, + { + "epoch": 0.22, + "learning_rate": 1.8249027707480032e-05, + "loss": 1.0815, + "step": 6739 + }, + { + "epoch": 0.22, + "learning_rate": 1.8248442149659942e-05, + "loss": 1.0308, + "step": 6740 + }, + { + "epoch": 0.22, + "learning_rate": 1.8247856503343198e-05, + "loss": 1.0469, + "step": 6741 + }, + { + "epoch": 0.22, + "learning_rate": 1.824727076853608e-05, + "loss": 1.0786, + "step": 6742 + }, + { + "epoch": 0.22, + "learning_rate": 1.8246684945244878e-05, + "loss": 1.0542, + "step": 6743 + }, + { + "epoch": 0.22, + "learning_rate": 1.8246099033475872e-05, + "loss": 1.0405, + "step": 6744 + }, + { + "epoch": 0.22, + "learning_rate": 1.8245513033235353e-05, + "loss": 1.0513, + "step": 6745 + }, + { + "epoch": 0.22, + "learning_rate": 1.82449269445296e-05, + "loss": 1.0391, + "step": 6746 + }, + { + "epoch": 0.22, + "learning_rate": 1.8244340767364908e-05, + "loss": 1.0122, + "step": 6747 + }, + { + "epoch": 0.22, + "learning_rate": 1.8243754501747563e-05, + "loss": 1.1528, + "step": 6748 + }, + { + "epoch": 0.22, + "learning_rate": 1.8243168147683854e-05, + "loss": 1.1138, + "step": 6749 + }, + { + "epoch": 0.22, + "learning_rate": 1.8242581705180077e-05, + "loss": 1.1396, + "step": 6750 + }, + { + "epoch": 0.22, + "learning_rate": 1.8241995174242523e-05, + "loss": 1.0459, + "step": 6751 + }, + { + "epoch": 0.22, + "learning_rate": 1.8241408554877477e-05, + "loss": 1.1274, + "step": 6752 + }, + { + "epoch": 0.22, + "learning_rate": 1.8240821847091237e-05, + "loss": 0.9141, + "step": 6753 + }, + { + "epoch": 0.22, + "learning_rate": 1.8240235050890106e-05, + "loss": 1.0386, + "step": 6754 + }, + { + "epoch": 0.22, + "learning_rate": 1.823964816628037e-05, + "loss": 1.1055, + "step": 6755 + }, + { + "epoch": 0.22, + "learning_rate": 1.8239061193268327e-05, + "loss": 0.9663, + "step": 6756 + }, + { + "epoch": 0.22, + "learning_rate": 1.823847413186028e-05, + "loss": 0.9888, + "step": 6757 + }, + { + "epoch": 0.22, + "learning_rate": 1.8237886982062517e-05, + "loss": 1.0044, + "step": 6758 + }, + { + "epoch": 0.22, + "learning_rate": 1.823729974388135e-05, + "loss": 1.1201, + "step": 6759 + }, + { + "epoch": 0.22, + "learning_rate": 1.8236712417323068e-05, + "loss": 1.0752, + "step": 6760 + }, + { + "epoch": 0.22, + "learning_rate": 1.823612500239398e-05, + "loss": 0.9995, + "step": 6761 + }, + { + "epoch": 0.22, + "learning_rate": 1.8235537499100388e-05, + "loss": 1.0879, + "step": 6762 + }, + { + "epoch": 0.22, + "learning_rate": 1.823494990744859e-05, + "loss": 1.0801, + "step": 6763 + }, + { + "epoch": 0.22, + "learning_rate": 1.8234362227444896e-05, + "loss": 1.0972, + "step": 6764 + }, + { + "epoch": 0.22, + "learning_rate": 1.823377445909561e-05, + "loss": 1.0981, + "step": 6765 + }, + { + "epoch": 0.22, + "learning_rate": 1.8233186602407032e-05, + "loss": 1.0386, + "step": 6766 + }, + { + "epoch": 0.22, + "learning_rate": 1.8232598657385476e-05, + "loss": 1.1411, + "step": 6767 + }, + { + "epoch": 0.22, + "learning_rate": 1.8232010624037248e-05, + "loss": 1.0688, + "step": 6768 + }, + { + "epoch": 0.22, + "learning_rate": 1.8231422502368658e-05, + "loss": 1.1846, + "step": 6769 + }, + { + "epoch": 0.22, + "learning_rate": 1.8230834292386014e-05, + "loss": 1.1367, + "step": 6770 + }, + { + "epoch": 0.22, + "learning_rate": 1.8230245994095624e-05, + "loss": 1.0718, + "step": 6771 + }, + { + "epoch": 0.22, + "learning_rate": 1.822965760750381e-05, + "loss": 1.1772, + "step": 6772 + }, + { + "epoch": 0.22, + "learning_rate": 1.822906913261687e-05, + "loss": 0.9717, + "step": 6773 + }, + { + "epoch": 0.22, + "learning_rate": 1.822848056944113e-05, + "loss": 1.0259, + "step": 6774 + }, + { + "epoch": 0.22, + "learning_rate": 1.82278919179829e-05, + "loss": 1.0391, + "step": 6775 + }, + { + "epoch": 0.22, + "learning_rate": 1.8227303178248498e-05, + "loss": 0.5459, + "step": 6776 + }, + { + "epoch": 0.22, + "learning_rate": 1.8226714350244235e-05, + "loss": 1.042, + "step": 6777 + }, + { + "epoch": 0.22, + "learning_rate": 1.8226125433976433e-05, + "loss": 0.9272, + "step": 6778 + }, + { + "epoch": 0.22, + "learning_rate": 1.8225536429451412e-05, + "loss": 0.5366, + "step": 6779 + }, + { + "epoch": 0.22, + "learning_rate": 1.8224947336675485e-05, + "loss": 1.0625, + "step": 6780 + }, + { + "epoch": 0.22, + "learning_rate": 1.822435815565498e-05, + "loss": 0.5815, + "step": 6781 + }, + { + "epoch": 0.22, + "learning_rate": 1.822376888639621e-05, + "loss": 0.9785, + "step": 6782 + }, + { + "epoch": 0.22, + "learning_rate": 1.82231795289055e-05, + "loss": 1.063, + "step": 6783 + }, + { + "epoch": 0.22, + "learning_rate": 1.822259008318918e-05, + "loss": 0.5256, + "step": 6784 + }, + { + "epoch": 0.22, + "learning_rate": 1.8222000549253567e-05, + "loss": 1.0767, + "step": 6785 + }, + { + "epoch": 0.22, + "learning_rate": 1.8221410927104986e-05, + "loss": 1.2705, + "step": 6786 + }, + { + "epoch": 0.22, + "learning_rate": 1.8220821216749766e-05, + "loss": 1.1338, + "step": 6787 + }, + { + "epoch": 0.22, + "learning_rate": 1.822023141819423e-05, + "loss": 1.0386, + "step": 6788 + }, + { + "epoch": 0.22, + "learning_rate": 1.8219641531444713e-05, + "loss": 1.0864, + "step": 6789 + }, + { + "epoch": 0.22, + "learning_rate": 1.821905155650754e-05, + "loss": 1.1343, + "step": 6790 + }, + { + "epoch": 0.22, + "learning_rate": 1.821846149338904e-05, + "loss": 1.0342, + "step": 6791 + }, + { + "epoch": 0.22, + "learning_rate": 1.8217871342095538e-05, + "loss": 1.0103, + "step": 6792 + }, + { + "epoch": 0.22, + "learning_rate": 1.8217281102633373e-05, + "loss": 0.9648, + "step": 6793 + }, + { + "epoch": 0.22, + "learning_rate": 1.8216690775008883e-05, + "loss": 1.1299, + "step": 6794 + }, + { + "epoch": 0.22, + "learning_rate": 1.8216100359228387e-05, + "loss": 0.979, + "step": 6795 + }, + { + "epoch": 0.22, + "learning_rate": 1.8215509855298234e-05, + "loss": 1.0327, + "step": 6796 + }, + { + "epoch": 0.22, + "learning_rate": 1.8214919263224747e-05, + "loss": 1.0713, + "step": 6797 + }, + { + "epoch": 0.22, + "learning_rate": 1.821432858301427e-05, + "loss": 1.2109, + "step": 6798 + }, + { + "epoch": 0.22, + "learning_rate": 1.821373781467314e-05, + "loss": 1.1714, + "step": 6799 + }, + { + "epoch": 0.22, + "learning_rate": 1.821314695820769e-05, + "loss": 1.0742, + "step": 6800 + }, + { + "epoch": 0.22, + "learning_rate": 1.821255601362427e-05, + "loss": 0.9966, + "step": 6801 + }, + { + "epoch": 0.22, + "learning_rate": 1.8211964980929208e-05, + "loss": 1.1528, + "step": 6802 + }, + { + "epoch": 0.22, + "learning_rate": 1.8211373860128852e-05, + "loss": 1.208, + "step": 6803 + }, + { + "epoch": 0.22, + "learning_rate": 1.821078265122954e-05, + "loss": 1.0728, + "step": 6804 + }, + { + "epoch": 0.22, + "learning_rate": 1.8210191354237618e-05, + "loss": 0.9966, + "step": 6805 + }, + { + "epoch": 0.22, + "learning_rate": 1.820959996915943e-05, + "loss": 1.001, + "step": 6806 + }, + { + "epoch": 0.22, + "learning_rate": 1.820900849600132e-05, + "loss": 1.1519, + "step": 6807 + }, + { + "epoch": 0.22, + "learning_rate": 1.8208416934769637e-05, + "loss": 0.9888, + "step": 6808 + }, + { + "epoch": 0.22, + "learning_rate": 1.8207825285470724e-05, + "loss": 0.5239, + "step": 6809 + }, + { + "epoch": 0.22, + "learning_rate": 1.820723354811093e-05, + "loss": 0.9004, + "step": 6810 + }, + { + "epoch": 0.22, + "learning_rate": 1.82066417226966e-05, + "loss": 1.082, + "step": 6811 + }, + { + "epoch": 0.22, + "learning_rate": 1.8206049809234087e-05, + "loss": 1.0439, + "step": 6812 + }, + { + "epoch": 0.22, + "learning_rate": 1.8205457807729744e-05, + "loss": 1.125, + "step": 6813 + }, + { + "epoch": 0.22, + "learning_rate": 1.820486571818992e-05, + "loss": 1.0288, + "step": 6814 + }, + { + "epoch": 0.22, + "learning_rate": 1.8204273540620968e-05, + "loss": 1.0449, + "step": 6815 + }, + { + "epoch": 0.22, + "learning_rate": 1.8203681275029238e-05, + "loss": 1.0903, + "step": 6816 + }, + { + "epoch": 0.22, + "learning_rate": 1.820308892142109e-05, + "loss": 1.1128, + "step": 6817 + }, + { + "epoch": 0.22, + "learning_rate": 1.8202496479802873e-05, + "loss": 1.0732, + "step": 6818 + }, + { + "epoch": 0.22, + "learning_rate": 1.820190395018095e-05, + "loss": 1.0894, + "step": 6819 + }, + { + "epoch": 0.22, + "learning_rate": 1.8201311332561674e-05, + "loss": 1.1445, + "step": 6820 + }, + { + "epoch": 0.22, + "learning_rate": 1.8200718626951405e-05, + "loss": 0.9326, + "step": 6821 + }, + { + "epoch": 0.22, + "learning_rate": 1.8200125833356503e-05, + "loss": 1.1592, + "step": 6822 + }, + { + "epoch": 0.22, + "learning_rate": 1.819953295178332e-05, + "loss": 1.1138, + "step": 6823 + }, + { + "epoch": 0.22, + "learning_rate": 1.819893998223823e-05, + "loss": 0.9487, + "step": 6824 + }, + { + "epoch": 0.22, + "learning_rate": 1.8198346924727587e-05, + "loss": 1.1396, + "step": 6825 + }, + { + "epoch": 0.22, + "learning_rate": 1.819775377925775e-05, + "loss": 1.0225, + "step": 6826 + }, + { + "epoch": 0.22, + "learning_rate": 1.8197160545835095e-05, + "loss": 1.1738, + "step": 6827 + }, + { + "epoch": 0.22, + "learning_rate": 1.8196567224465975e-05, + "loss": 1.0493, + "step": 6828 + }, + { + "epoch": 0.22, + "learning_rate": 1.819597381515676e-05, + "loss": 1.1841, + "step": 6829 + }, + { + "epoch": 0.22, + "learning_rate": 1.8195380317913817e-05, + "loss": 1.0342, + "step": 6830 + }, + { + "epoch": 0.22, + "learning_rate": 1.8194786732743516e-05, + "loss": 1.0908, + "step": 6831 + }, + { + "epoch": 0.22, + "learning_rate": 1.819419305965222e-05, + "loss": 0.9561, + "step": 6832 + }, + { + "epoch": 0.22, + "learning_rate": 1.8193599298646303e-05, + "loss": 1.1782, + "step": 6833 + }, + { + "epoch": 0.22, + "learning_rate": 1.8193005449732134e-05, + "loss": 1.1284, + "step": 6834 + }, + { + "epoch": 0.22, + "learning_rate": 1.819241151291608e-05, + "loss": 1.0713, + "step": 6835 + }, + { + "epoch": 0.22, + "learning_rate": 1.8191817488204523e-05, + "loss": 1.002, + "step": 6836 + }, + { + "epoch": 0.22, + "learning_rate": 1.8191223375603827e-05, + "loss": 1.1069, + "step": 6837 + }, + { + "epoch": 0.22, + "learning_rate": 1.8190629175120373e-05, + "loss": 1.0391, + "step": 6838 + }, + { + "epoch": 0.22, + "learning_rate": 1.819003488676053e-05, + "loss": 1.1411, + "step": 6839 + }, + { + "epoch": 0.22, + "learning_rate": 1.818944051053068e-05, + "loss": 1.0713, + "step": 6840 + }, + { + "epoch": 0.22, + "learning_rate": 1.8188846046437192e-05, + "loss": 1.0957, + "step": 6841 + }, + { + "epoch": 0.22, + "learning_rate": 1.8188251494486453e-05, + "loss": 1.1714, + "step": 6842 + }, + { + "epoch": 0.22, + "learning_rate": 1.8187656854684838e-05, + "loss": 0.999, + "step": 6843 + }, + { + "epoch": 0.22, + "learning_rate": 1.8187062127038723e-05, + "loss": 1.04, + "step": 6844 + }, + { + "epoch": 0.22, + "learning_rate": 1.8186467311554496e-05, + "loss": 1.0811, + "step": 6845 + }, + { + "epoch": 0.22, + "learning_rate": 1.8185872408238532e-05, + "loss": 1.0654, + "step": 6846 + }, + { + "epoch": 0.22, + "learning_rate": 1.8185277417097218e-05, + "loss": 1.1074, + "step": 6847 + }, + { + "epoch": 0.22, + "learning_rate": 1.8184682338136938e-05, + "loss": 1.1489, + "step": 6848 + }, + { + "epoch": 0.22, + "learning_rate": 1.8184087171364074e-05, + "loss": 1.0562, + "step": 6849 + }, + { + "epoch": 0.22, + "learning_rate": 1.818349191678501e-05, + "loss": 1.0874, + "step": 6850 + }, + { + "epoch": 0.22, + "learning_rate": 1.8182896574406138e-05, + "loss": 0.9956, + "step": 6851 + }, + { + "epoch": 0.22, + "learning_rate": 1.818230114423384e-05, + "loss": 1.1821, + "step": 6852 + }, + { + "epoch": 0.22, + "learning_rate": 1.8181705626274507e-05, + "loss": 1.0288, + "step": 6853 + }, + { + "epoch": 0.22, + "learning_rate": 1.8181110020534527e-05, + "loss": 1.1611, + "step": 6854 + }, + { + "epoch": 0.22, + "learning_rate": 1.8180514327020293e-05, + "loss": 1.1069, + "step": 6855 + }, + { + "epoch": 0.22, + "learning_rate": 1.8179918545738193e-05, + "loss": 1.019, + "step": 6856 + }, + { + "epoch": 0.22, + "learning_rate": 1.817932267669462e-05, + "loss": 1.2476, + "step": 6857 + }, + { + "epoch": 0.22, + "learning_rate": 1.817872671989597e-05, + "loss": 1.0771, + "step": 6858 + }, + { + "epoch": 0.22, + "learning_rate": 1.817813067534863e-05, + "loss": 1.0215, + "step": 6859 + }, + { + "epoch": 0.22, + "learning_rate": 1.8177534543059e-05, + "loss": 1.0615, + "step": 6860 + }, + { + "epoch": 0.22, + "learning_rate": 1.817693832303347e-05, + "loss": 1.1299, + "step": 6861 + }, + { + "epoch": 0.22, + "learning_rate": 1.8176342015278447e-05, + "loss": 0.9873, + "step": 6862 + }, + { + "epoch": 0.22, + "learning_rate": 1.8175745619800326e-05, + "loss": 1.1118, + "step": 6863 + }, + { + "epoch": 0.22, + "learning_rate": 1.8175149136605495e-05, + "loss": 1.0308, + "step": 6864 + }, + { + "epoch": 0.22, + "learning_rate": 1.817455256570037e-05, + "loss": 1.02, + "step": 6865 + }, + { + "epoch": 0.22, + "learning_rate": 1.817395590709134e-05, + "loss": 1.0068, + "step": 6866 + }, + { + "epoch": 0.22, + "learning_rate": 1.8173359160784806e-05, + "loss": 1.085, + "step": 6867 + }, + { + "epoch": 0.22, + "learning_rate": 1.817276232678718e-05, + "loss": 1.1646, + "step": 6868 + }, + { + "epoch": 0.22, + "learning_rate": 1.8172165405104857e-05, + "loss": 1.0845, + "step": 6869 + }, + { + "epoch": 0.22, + "learning_rate": 1.8171568395744242e-05, + "loss": 1.1069, + "step": 6870 + }, + { + "epoch": 0.22, + "learning_rate": 1.8170971298711742e-05, + "loss": 1.0269, + "step": 6871 + }, + { + "epoch": 0.22, + "learning_rate": 1.8170374114013762e-05, + "loss": 1.1001, + "step": 6872 + }, + { + "epoch": 0.22, + "learning_rate": 1.8169776841656715e-05, + "loss": 1.1714, + "step": 6873 + }, + { + "epoch": 0.22, + "learning_rate": 1.8169179481647e-05, + "loss": 1.1982, + "step": 6874 + }, + { + "epoch": 0.22, + "learning_rate": 1.8168582033991034e-05, + "loss": 1.104, + "step": 6875 + }, + { + "epoch": 0.22, + "learning_rate": 1.816798449869522e-05, + "loss": 1.1919, + "step": 6876 + }, + { + "epoch": 0.22, + "learning_rate": 1.8167386875765975e-05, + "loss": 1.1577, + "step": 6877 + }, + { + "epoch": 0.22, + "learning_rate": 1.8166789165209707e-05, + "loss": 1.0894, + "step": 6878 + }, + { + "epoch": 0.22, + "learning_rate": 1.8166191367032828e-05, + "loss": 1.0679, + "step": 6879 + }, + { + "epoch": 0.22, + "learning_rate": 1.8165593481241757e-05, + "loss": 0.9805, + "step": 6880 + }, + { + "epoch": 0.22, + "learning_rate": 1.8164995507842902e-05, + "loss": 1.0361, + "step": 6881 + }, + { + "epoch": 0.22, + "learning_rate": 1.8164397446842684e-05, + "loss": 1.1011, + "step": 6882 + }, + { + "epoch": 0.22, + "learning_rate": 1.8163799298247515e-05, + "loss": 1.1216, + "step": 6883 + }, + { + "epoch": 0.22, + "learning_rate": 1.8163201062063817e-05, + "loss": 0.9941, + "step": 6884 + }, + { + "epoch": 0.22, + "learning_rate": 1.8162602738298006e-05, + "loss": 1.0449, + "step": 6885 + }, + { + "epoch": 0.22, + "learning_rate": 1.81620043269565e-05, + "loss": 1.2783, + "step": 6886 + }, + { + "epoch": 0.22, + "learning_rate": 1.816140582804572e-05, + "loss": 1.0322, + "step": 6887 + }, + { + "epoch": 0.22, + "learning_rate": 1.816080724157209e-05, + "loss": 1.0156, + "step": 6888 + }, + { + "epoch": 0.22, + "learning_rate": 1.816020856754203e-05, + "loss": 1.0845, + "step": 6889 + }, + { + "epoch": 0.22, + "learning_rate": 1.8159609805961966e-05, + "loss": 1.1909, + "step": 6890 + }, + { + "epoch": 0.22, + "learning_rate": 1.8159010956838315e-05, + "loss": 0.9092, + "step": 6891 + }, + { + "epoch": 0.22, + "learning_rate": 1.815841202017751e-05, + "loss": 1.0459, + "step": 6892 + }, + { + "epoch": 0.22, + "learning_rate": 1.8157812995985968e-05, + "loss": 1.1074, + "step": 6893 + }, + { + "epoch": 0.22, + "learning_rate": 1.8157213884270123e-05, + "loss": 1.022, + "step": 6894 + }, + { + "epoch": 0.22, + "learning_rate": 1.8156614685036405e-05, + "loss": 1.1035, + "step": 6895 + }, + { + "epoch": 0.22, + "learning_rate": 1.8156015398291234e-05, + "loss": 1.0562, + "step": 6896 + }, + { + "epoch": 0.22, + "learning_rate": 1.815541602404105e-05, + "loss": 1.1074, + "step": 6897 + }, + { + "epoch": 0.22, + "learning_rate": 1.8154816562292273e-05, + "loss": 0.9932, + "step": 6898 + }, + { + "epoch": 0.22, + "learning_rate": 1.8154217013051344e-05, + "loss": 1.0737, + "step": 6899 + }, + { + "epoch": 0.22, + "learning_rate": 1.815361737632469e-05, + "loss": 1.1685, + "step": 6900 + }, + { + "epoch": 0.22, + "learning_rate": 1.8153017652118737e-05, + "loss": 1.1147, + "step": 6901 + }, + { + "epoch": 0.22, + "learning_rate": 1.8152417840439937e-05, + "loss": 0.96, + "step": 6902 + }, + { + "epoch": 0.22, + "learning_rate": 1.8151817941294715e-05, + "loss": 1.147, + "step": 6903 + }, + { + "epoch": 0.22, + "learning_rate": 1.815121795468951e-05, + "loss": 1.1636, + "step": 6904 + }, + { + "epoch": 0.22, + "learning_rate": 1.815061788063075e-05, + "loss": 1.1299, + "step": 6905 + }, + { + "epoch": 0.22, + "learning_rate": 1.8150017719124888e-05, + "loss": 0.9409, + "step": 6906 + }, + { + "epoch": 0.22, + "learning_rate": 1.8149417470178356e-05, + "loss": 1.0186, + "step": 6907 + }, + { + "epoch": 0.22, + "learning_rate": 1.814881713379759e-05, + "loss": 1.1396, + "step": 6908 + }, + { + "epoch": 0.22, + "learning_rate": 1.8148216709989034e-05, + "loss": 1.0278, + "step": 6909 + }, + { + "epoch": 0.22, + "learning_rate": 1.8147616198759137e-05, + "loss": 1.0679, + "step": 6910 + }, + { + "epoch": 0.22, + "learning_rate": 1.814701560011433e-05, + "loss": 0.5581, + "step": 6911 + }, + { + "epoch": 0.22, + "learning_rate": 1.8146414914061064e-05, + "loss": 1.0962, + "step": 6912 + }, + { + "epoch": 0.22, + "learning_rate": 1.8145814140605782e-05, + "loss": 1.1758, + "step": 6913 + }, + { + "epoch": 0.22, + "learning_rate": 1.814521327975493e-05, + "loss": 1.1133, + "step": 6914 + }, + { + "epoch": 0.22, + "learning_rate": 1.8144612331514952e-05, + "loss": 1.1128, + "step": 6915 + }, + { + "epoch": 0.22, + "learning_rate": 1.81440112958923e-05, + "loss": 1.1128, + "step": 6916 + }, + { + "epoch": 0.22, + "learning_rate": 1.814341017289342e-05, + "loss": 1.0498, + "step": 6917 + }, + { + "epoch": 0.22, + "learning_rate": 1.8142808962524762e-05, + "loss": 1.1387, + "step": 6918 + }, + { + "epoch": 0.22, + "learning_rate": 1.8142207664792775e-05, + "loss": 1.1914, + "step": 6919 + }, + { + "epoch": 0.22, + "learning_rate": 1.8141606279703912e-05, + "loss": 1.0591, + "step": 6920 + }, + { + "epoch": 0.22, + "learning_rate": 1.814100480726462e-05, + "loss": 1.2158, + "step": 6921 + }, + { + "epoch": 0.22, + "learning_rate": 1.814040324748136e-05, + "loss": 1.0771, + "step": 6922 + }, + { + "epoch": 0.22, + "learning_rate": 1.8139801600360586e-05, + "loss": 1.0796, + "step": 6923 + }, + { + "epoch": 0.22, + "learning_rate": 1.8139199865908742e-05, + "loss": 1.0029, + "step": 6924 + }, + { + "epoch": 0.22, + "learning_rate": 1.8138598044132298e-05, + "loss": 1.0942, + "step": 6925 + }, + { + "epoch": 0.22, + "learning_rate": 1.81379961350377e-05, + "loss": 1.0835, + "step": 6926 + }, + { + "epoch": 0.22, + "learning_rate": 1.8137394138631413e-05, + "loss": 1.0767, + "step": 6927 + }, + { + "epoch": 0.22, + "learning_rate": 1.8136792054919892e-05, + "loss": 1.1172, + "step": 6928 + }, + { + "epoch": 0.22, + "learning_rate": 1.8136189883909598e-05, + "loss": 1.106, + "step": 6929 + }, + { + "epoch": 0.22, + "learning_rate": 1.813558762560699e-05, + "loss": 1.0947, + "step": 6930 + }, + { + "epoch": 0.22, + "learning_rate": 1.8134985280018532e-05, + "loss": 1.0713, + "step": 6931 + }, + { + "epoch": 0.22, + "learning_rate": 1.8134382847150683e-05, + "loss": 1.1167, + "step": 6932 + }, + { + "epoch": 0.22, + "learning_rate": 1.813378032700991e-05, + "loss": 1.1157, + "step": 6933 + }, + { + "epoch": 0.22, + "learning_rate": 1.8133177719602676e-05, + "loss": 0.9912, + "step": 6934 + }, + { + "epoch": 0.22, + "learning_rate": 1.813257502493545e-05, + "loss": 1.0938, + "step": 6935 + }, + { + "epoch": 0.22, + "learning_rate": 1.813197224301469e-05, + "loss": 1.1504, + "step": 6936 + }, + { + "epoch": 0.22, + "learning_rate": 1.813136937384687e-05, + "loss": 1.1382, + "step": 6937 + }, + { + "epoch": 0.22, + "learning_rate": 1.8130766417438456e-05, + "loss": 1.0146, + "step": 6938 + }, + { + "epoch": 0.22, + "learning_rate": 1.8130163373795918e-05, + "loss": 1.1992, + "step": 6939 + }, + { + "epoch": 0.22, + "learning_rate": 1.8129560242925723e-05, + "loss": 0.8486, + "step": 6940 + }, + { + "epoch": 0.22, + "learning_rate": 1.8128957024834343e-05, + "loss": 1.0933, + "step": 6941 + }, + { + "epoch": 0.22, + "learning_rate": 1.8128353719528253e-05, + "loss": 1.1436, + "step": 6942 + }, + { + "epoch": 0.22, + "learning_rate": 1.812775032701392e-05, + "loss": 1.0791, + "step": 6943 + }, + { + "epoch": 0.22, + "learning_rate": 1.8127146847297825e-05, + "loss": 1.1294, + "step": 6944 + }, + { + "epoch": 0.22, + "learning_rate": 1.812654328038644e-05, + "loss": 0.9878, + "step": 6945 + }, + { + "epoch": 0.22, + "learning_rate": 1.8125939626286236e-05, + "loss": 1.0605, + "step": 6946 + }, + { + "epoch": 0.22, + "learning_rate": 1.8125335885003696e-05, + "loss": 1.1929, + "step": 6947 + }, + { + "epoch": 0.22, + "learning_rate": 1.8124732056545294e-05, + "loss": 1.0718, + "step": 6948 + }, + { + "epoch": 0.22, + "learning_rate": 1.812412814091751e-05, + "loss": 0.9062, + "step": 6949 + }, + { + "epoch": 0.22, + "learning_rate": 1.8123524138126817e-05, + "loss": 1.0703, + "step": 6950 + }, + { + "epoch": 0.22, + "learning_rate": 1.8122920048179705e-05, + "loss": 1.0142, + "step": 6951 + }, + { + "epoch": 0.22, + "learning_rate": 1.812231587108265e-05, + "loss": 1.022, + "step": 6952 + }, + { + "epoch": 0.22, + "learning_rate": 1.812171160684214e-05, + "loss": 1.0859, + "step": 6953 + }, + { + "epoch": 0.22, + "learning_rate": 1.8121107255464646e-05, + "loss": 0.9863, + "step": 6954 + }, + { + "epoch": 0.22, + "learning_rate": 1.8120502816956663e-05, + "loss": 1.1147, + "step": 6955 + }, + { + "epoch": 0.22, + "learning_rate": 1.811989829132467e-05, + "loss": 1.0542, + "step": 6956 + }, + { + "epoch": 0.22, + "learning_rate": 1.8119293678575155e-05, + "loss": 1.0811, + "step": 6957 + }, + { + "epoch": 0.22, + "learning_rate": 1.8118688978714604e-05, + "loss": 1.0591, + "step": 6958 + }, + { + "epoch": 0.22, + "learning_rate": 1.811808419174951e-05, + "loss": 1.1514, + "step": 6959 + }, + { + "epoch": 0.22, + "learning_rate": 1.8117479317686354e-05, + "loss": 1.0161, + "step": 6960 + }, + { + "epoch": 0.22, + "learning_rate": 1.811687435653163e-05, + "loss": 1.0723, + "step": 6961 + }, + { + "epoch": 0.22, + "learning_rate": 1.8116269308291827e-05, + "loss": 1.1655, + "step": 6962 + }, + { + "epoch": 0.22, + "learning_rate": 1.8115664172973437e-05, + "loss": 0.9419, + "step": 6963 + }, + { + "epoch": 0.22, + "learning_rate": 1.8115058950582953e-05, + "loss": 1.0913, + "step": 6964 + }, + { + "epoch": 0.22, + "learning_rate": 1.8114453641126867e-05, + "loss": 1.0488, + "step": 6965 + }, + { + "epoch": 0.22, + "learning_rate": 1.8113848244611674e-05, + "loss": 1.0747, + "step": 6966 + }, + { + "epoch": 0.22, + "learning_rate": 1.811324276104387e-05, + "loss": 1.0249, + "step": 6967 + }, + { + "epoch": 0.22, + "learning_rate": 1.811263719042995e-05, + "loss": 1.0425, + "step": 6968 + }, + { + "epoch": 0.22, + "learning_rate": 1.811203153277641e-05, + "loss": 0.9775, + "step": 6969 + }, + { + "epoch": 0.22, + "learning_rate": 1.8111425788089753e-05, + "loss": 1.0596, + "step": 6970 + }, + { + "epoch": 0.22, + "learning_rate": 1.8110819956376473e-05, + "loss": 1.1313, + "step": 6971 + }, + { + "epoch": 0.22, + "learning_rate": 1.811021403764307e-05, + "loss": 1.2119, + "step": 6972 + }, + { + "epoch": 0.22, + "learning_rate": 1.8109608031896048e-05, + "loss": 1.019, + "step": 6973 + }, + { + "epoch": 0.22, + "learning_rate": 1.8109001939141907e-05, + "loss": 1.0762, + "step": 6974 + }, + { + "epoch": 0.22, + "learning_rate": 1.810839575938715e-05, + "loss": 1.2842, + "step": 6975 + }, + { + "epoch": 0.22, + "learning_rate": 1.810778949263828e-05, + "loss": 0.9751, + "step": 6976 + }, + { + "epoch": 0.22, + "learning_rate": 1.81071831389018e-05, + "loss": 1.1226, + "step": 6977 + }, + { + "epoch": 0.22, + "learning_rate": 1.8106576698184218e-05, + "loss": 1.0728, + "step": 6978 + }, + { + "epoch": 0.22, + "learning_rate": 1.8105970170492042e-05, + "loss": 1.0581, + "step": 6979 + }, + { + "epoch": 0.22, + "learning_rate": 1.8105363555831775e-05, + "loss": 1.1348, + "step": 6980 + }, + { + "epoch": 0.22, + "learning_rate": 1.810475685420993e-05, + "loss": 0.9888, + "step": 6981 + }, + { + "epoch": 0.22, + "learning_rate": 1.8104150065633014e-05, + "loss": 0.9155, + "step": 6982 + }, + { + "epoch": 0.22, + "learning_rate": 1.8103543190107537e-05, + "loss": 1.0732, + "step": 6983 + }, + { + "epoch": 0.22, + "learning_rate": 1.8102936227640007e-05, + "loss": 1.0947, + "step": 6984 + }, + { + "epoch": 0.22, + "learning_rate": 1.8102329178236942e-05, + "loss": 1.1729, + "step": 6985 + }, + { + "epoch": 0.22, + "learning_rate": 1.810172204190485e-05, + "loss": 0.9106, + "step": 6986 + }, + { + "epoch": 0.22, + "learning_rate": 1.8101114818650252e-05, + "loss": 1.0146, + "step": 6987 + }, + { + "epoch": 0.22, + "learning_rate": 1.810050750847965e-05, + "loss": 1.127, + "step": 6988 + }, + { + "epoch": 0.22, + "learning_rate": 1.8099900111399577e-05, + "loss": 1.0059, + "step": 6989 + }, + { + "epoch": 0.22, + "learning_rate": 1.8099292627416535e-05, + "loss": 1.1084, + "step": 6990 + }, + { + "epoch": 0.22, + "learning_rate": 1.8098685056537046e-05, + "loss": 1.1938, + "step": 6991 + }, + { + "epoch": 0.22, + "learning_rate": 1.809807739876763e-05, + "loss": 1.0396, + "step": 6992 + }, + { + "epoch": 0.22, + "learning_rate": 1.8097469654114806e-05, + "loss": 0.9805, + "step": 6993 + }, + { + "epoch": 0.22, + "learning_rate": 1.8096861822585096e-05, + "loss": 1.019, + "step": 6994 + }, + { + "epoch": 0.22, + "learning_rate": 1.8096253904185018e-05, + "loss": 1.063, + "step": 6995 + }, + { + "epoch": 0.22, + "learning_rate": 1.8095645898921098e-05, + "loss": 1.1274, + "step": 6996 + }, + { + "epoch": 0.22, + "learning_rate": 1.8095037806799858e-05, + "loss": 1.0483, + "step": 6997 + }, + { + "epoch": 0.22, + "learning_rate": 1.8094429627827817e-05, + "loss": 1.1162, + "step": 6998 + }, + { + "epoch": 0.22, + "learning_rate": 1.809382136201151e-05, + "loss": 1.123, + "step": 6999 + }, + { + "epoch": 0.22, + "learning_rate": 1.8093213009357455e-05, + "loss": 1.0557, + "step": 7000 + }, + { + "epoch": 0.22, + "learning_rate": 1.809260456987218e-05, + "loss": 0.9502, + "step": 7001 + }, + { + "epoch": 0.22, + "learning_rate": 1.8091996043562218e-05, + "loss": 1.0127, + "step": 7002 + }, + { + "epoch": 0.22, + "learning_rate": 1.8091387430434092e-05, + "loss": 1.0488, + "step": 7003 + }, + { + "epoch": 0.22, + "learning_rate": 1.8090778730494335e-05, + "loss": 1.1089, + "step": 7004 + }, + { + "epoch": 0.22, + "learning_rate": 1.8090169943749477e-05, + "loss": 1.0176, + "step": 7005 + }, + { + "epoch": 0.22, + "learning_rate": 1.808956107020605e-05, + "loss": 1.1597, + "step": 7006 + }, + { + "epoch": 0.22, + "learning_rate": 1.8088952109870583e-05, + "loss": 1.0103, + "step": 7007 + }, + { + "epoch": 0.22, + "learning_rate": 1.8088343062749612e-05, + "loss": 1.0811, + "step": 7008 + }, + { + "epoch": 0.22, + "learning_rate": 1.8087733928849672e-05, + "loss": 1.0361, + "step": 7009 + }, + { + "epoch": 0.22, + "learning_rate": 1.80871247081773e-05, + "loss": 1.0586, + "step": 7010 + }, + { + "epoch": 0.22, + "learning_rate": 1.8086515400739027e-05, + "loss": 1.1309, + "step": 7011 + }, + { + "epoch": 0.22, + "learning_rate": 1.8085906006541397e-05, + "loss": 1.0464, + "step": 7012 + }, + { + "epoch": 0.22, + "learning_rate": 1.8085296525590944e-05, + "loss": 1.0239, + "step": 7013 + }, + { + "epoch": 0.22, + "learning_rate": 1.8084686957894207e-05, + "loss": 1.0142, + "step": 7014 + }, + { + "epoch": 0.22, + "learning_rate": 1.8084077303457728e-05, + "loss": 1.1235, + "step": 7015 + }, + { + "epoch": 0.22, + "learning_rate": 1.8083467562288043e-05, + "loss": 1.0977, + "step": 7016 + }, + { + "epoch": 0.22, + "learning_rate": 1.80828577343917e-05, + "loss": 0.9653, + "step": 7017 + }, + { + "epoch": 0.22, + "learning_rate": 1.8082247819775237e-05, + "loss": 1.1113, + "step": 7018 + }, + { + "epoch": 0.22, + "learning_rate": 1.8081637818445203e-05, + "loss": 1.1011, + "step": 7019 + }, + { + "epoch": 0.22, + "learning_rate": 1.8081027730408138e-05, + "loss": 1.1011, + "step": 7020 + }, + { + "epoch": 0.22, + "learning_rate": 1.808041755567059e-05, + "loss": 0.9897, + "step": 7021 + }, + { + "epoch": 0.22, + "learning_rate": 1.8079807294239102e-05, + "loss": 1.084, + "step": 7022 + }, + { + "epoch": 0.22, + "learning_rate": 1.8079196946120227e-05, + "loss": 1.0664, + "step": 7023 + }, + { + "epoch": 0.22, + "learning_rate": 1.8078586511320507e-05, + "loss": 1.2227, + "step": 7024 + }, + { + "epoch": 0.22, + "learning_rate": 1.80779759898465e-05, + "loss": 1.0425, + "step": 7025 + }, + { + "epoch": 0.22, + "learning_rate": 1.8077365381704747e-05, + "loss": 1.0498, + "step": 7026 + }, + { + "epoch": 0.22, + "learning_rate": 1.8076754686901803e-05, + "loss": 0.9673, + "step": 7027 + }, + { + "epoch": 0.22, + "learning_rate": 1.807614390544422e-05, + "loss": 1.0552, + "step": 7028 + }, + { + "epoch": 0.22, + "learning_rate": 1.8075533037338554e-05, + "loss": 1.0039, + "step": 7029 + }, + { + "epoch": 0.22, + "learning_rate": 1.8074922082591352e-05, + "loss": 0.9575, + "step": 7030 + }, + { + "epoch": 0.22, + "learning_rate": 1.8074311041209177e-05, + "loss": 1.1108, + "step": 7031 + }, + { + "epoch": 0.22, + "learning_rate": 1.807369991319858e-05, + "loss": 0.9102, + "step": 7032 + }, + { + "epoch": 0.22, + "learning_rate": 1.8073088698566115e-05, + "loss": 0.958, + "step": 7033 + }, + { + "epoch": 0.22, + "learning_rate": 1.8072477397318345e-05, + "loss": 1.0195, + "step": 7034 + }, + { + "epoch": 0.22, + "learning_rate": 1.8071866009461828e-05, + "loss": 1.1152, + "step": 7035 + }, + { + "epoch": 0.23, + "learning_rate": 1.807125453500312e-05, + "loss": 1.0195, + "step": 7036 + }, + { + "epoch": 0.23, + "learning_rate": 1.8070642973948787e-05, + "loss": 0.9058, + "step": 7037 + }, + { + "epoch": 0.23, + "learning_rate": 1.8070031326305382e-05, + "loss": 0.9058, + "step": 7038 + }, + { + "epoch": 0.23, + "learning_rate": 1.8069419592079475e-05, + "loss": 1.0415, + "step": 7039 + }, + { + "epoch": 0.23, + "learning_rate": 1.806880777127763e-05, + "loss": 1.1289, + "step": 7040 + }, + { + "epoch": 0.23, + "learning_rate": 1.80681958639064e-05, + "loss": 0.9714, + "step": 7041 + }, + { + "epoch": 0.23, + "learning_rate": 1.8067583869972364e-05, + "loss": 0.936, + "step": 7042 + }, + { + "epoch": 0.23, + "learning_rate": 1.8066971789482082e-05, + "loss": 1.0801, + "step": 7043 + }, + { + "epoch": 0.23, + "learning_rate": 1.806635962244212e-05, + "loss": 1.0708, + "step": 7044 + }, + { + "epoch": 0.23, + "learning_rate": 1.8065747368859044e-05, + "loss": 1.0635, + "step": 7045 + }, + { + "epoch": 0.23, + "learning_rate": 1.8065135028739428e-05, + "loss": 0.9824, + "step": 7046 + }, + { + "epoch": 0.23, + "learning_rate": 1.8064522602089836e-05, + "loss": 1.1724, + "step": 7047 + }, + { + "epoch": 0.23, + "learning_rate": 1.8063910088916846e-05, + "loss": 1.0596, + "step": 7048 + }, + { + "epoch": 0.23, + "learning_rate": 1.8063297489227025e-05, + "loss": 0.9619, + "step": 7049 + }, + { + "epoch": 0.23, + "learning_rate": 1.8062684803026945e-05, + "loss": 1.1084, + "step": 7050 + }, + { + "epoch": 0.23, + "learning_rate": 1.8062072030323178e-05, + "loss": 0.5171, + "step": 7051 + }, + { + "epoch": 0.23, + "learning_rate": 1.8061459171122303e-05, + "loss": 1.0693, + "step": 7052 + }, + { + "epoch": 0.23, + "learning_rate": 1.806084622543089e-05, + "loss": 1.0259, + "step": 7053 + }, + { + "epoch": 0.23, + "learning_rate": 1.8060233193255524e-05, + "loss": 1.043, + "step": 7054 + }, + { + "epoch": 0.23, + "learning_rate": 1.8059620074602773e-05, + "loss": 0.9912, + "step": 7055 + }, + { + "epoch": 0.23, + "learning_rate": 1.805900686947922e-05, + "loss": 1.0034, + "step": 7056 + }, + { + "epoch": 0.23, + "learning_rate": 1.805839357789144e-05, + "loss": 1.0137, + "step": 7057 + }, + { + "epoch": 0.23, + "learning_rate": 1.8057780199846016e-05, + "loss": 1.1191, + "step": 7058 + }, + { + "epoch": 0.23, + "learning_rate": 1.8057166735349533e-05, + "loss": 1.0283, + "step": 7059 + }, + { + "epoch": 0.23, + "learning_rate": 1.8056553184408563e-05, + "loss": 1.0195, + "step": 7060 + }, + { + "epoch": 0.23, + "learning_rate": 1.8055939547029694e-05, + "loss": 0.8999, + "step": 7061 + }, + { + "epoch": 0.23, + "learning_rate": 1.8055325823219512e-05, + "loss": 1.0352, + "step": 7062 + }, + { + "epoch": 0.23, + "learning_rate": 1.80547120129846e-05, + "loss": 1.0742, + "step": 7063 + }, + { + "epoch": 0.23, + "learning_rate": 1.805409811633154e-05, + "loss": 1.1353, + "step": 7064 + }, + { + "epoch": 0.23, + "learning_rate": 1.805348413326692e-05, + "loss": 1.0073, + "step": 7065 + }, + { + "epoch": 0.23, + "learning_rate": 1.805287006379733e-05, + "loss": 1.0752, + "step": 7066 + }, + { + "epoch": 0.23, + "learning_rate": 1.8052255907929358e-05, + "loss": 1.083, + "step": 7067 + }, + { + "epoch": 0.23, + "learning_rate": 1.8051641665669594e-05, + "loss": 1.1743, + "step": 7068 + }, + { + "epoch": 0.23, + "learning_rate": 1.805102733702462e-05, + "loss": 1.1118, + "step": 7069 + }, + { + "epoch": 0.23, + "learning_rate": 1.8050412922001037e-05, + "loss": 1.1533, + "step": 7070 + }, + { + "epoch": 0.23, + "learning_rate": 1.804979842060543e-05, + "loss": 0.9414, + "step": 7071 + }, + { + "epoch": 0.23, + "learning_rate": 1.80491838328444e-05, + "loss": 1.103, + "step": 7072 + }, + { + "epoch": 0.23, + "learning_rate": 1.8048569158724532e-05, + "loss": 1.0786, + "step": 7073 + }, + { + "epoch": 0.23, + "learning_rate": 1.8047954398252427e-05, + "loss": 1.0835, + "step": 7074 + }, + { + "epoch": 0.23, + "learning_rate": 1.8047339551434674e-05, + "loss": 1.1089, + "step": 7075 + }, + { + "epoch": 0.23, + "learning_rate": 1.8046724618277876e-05, + "loss": 1.0601, + "step": 7076 + }, + { + "epoch": 0.23, + "learning_rate": 1.8046109598788632e-05, + "loss": 1.1191, + "step": 7077 + }, + { + "epoch": 0.23, + "learning_rate": 1.8045494492973533e-05, + "loss": 1.0928, + "step": 7078 + }, + { + "epoch": 0.23, + "learning_rate": 1.8044879300839184e-05, + "loss": 1.0244, + "step": 7079 + }, + { + "epoch": 0.23, + "learning_rate": 1.8044264022392184e-05, + "loss": 1.1279, + "step": 7080 + }, + { + "epoch": 0.23, + "learning_rate": 1.8043648657639134e-05, + "loss": 1.1855, + "step": 7081 + }, + { + "epoch": 0.23, + "learning_rate": 1.8043033206586637e-05, + "loss": 1.042, + "step": 7082 + }, + { + "epoch": 0.23, + "learning_rate": 1.8042417669241296e-05, + "loss": 1.1353, + "step": 7083 + }, + { + "epoch": 0.23, + "learning_rate": 1.804180204560971e-05, + "loss": 1.0112, + "step": 7084 + }, + { + "epoch": 0.23, + "learning_rate": 1.804118633569849e-05, + "loss": 0.9536, + "step": 7085 + }, + { + "epoch": 0.23, + "learning_rate": 1.804057053951424e-05, + "loss": 0.5315, + "step": 7086 + }, + { + "epoch": 0.23, + "learning_rate": 1.8039954657063567e-05, + "loss": 1.1089, + "step": 7087 + }, + { + "epoch": 0.23, + "learning_rate": 1.803933868835308e-05, + "loss": 1.0757, + "step": 7088 + }, + { + "epoch": 0.23, + "learning_rate": 1.803872263338939e-05, + "loss": 0.9937, + "step": 7089 + }, + { + "epoch": 0.23, + "learning_rate": 1.8038106492179098e-05, + "loss": 0.9893, + "step": 7090 + }, + { + "epoch": 0.23, + "learning_rate": 1.803749026472882e-05, + "loss": 1.1709, + "step": 7091 + }, + { + "epoch": 0.23, + "learning_rate": 1.8036873951045167e-05, + "loss": 1.2031, + "step": 7092 + }, + { + "epoch": 0.23, + "learning_rate": 1.803625755113475e-05, + "loss": 0.9634, + "step": 7093 + }, + { + "epoch": 0.23, + "learning_rate": 1.8035641065004184e-05, + "loss": 0.9849, + "step": 7094 + }, + { + "epoch": 0.23, + "learning_rate": 1.8035024492660086e-05, + "loss": 1.0537, + "step": 7095 + }, + { + "epoch": 0.23, + "learning_rate": 1.8034407834109068e-05, + "loss": 1.1885, + "step": 7096 + }, + { + "epoch": 0.23, + "learning_rate": 1.803379108935774e-05, + "loss": 1.0547, + "step": 7097 + }, + { + "epoch": 0.23, + "learning_rate": 1.8033174258412728e-05, + "loss": 1.063, + "step": 7098 + }, + { + "epoch": 0.23, + "learning_rate": 1.8032557341280647e-05, + "loss": 1.0874, + "step": 7099 + }, + { + "epoch": 0.23, + "learning_rate": 1.8031940337968117e-05, + "loss": 1.0703, + "step": 7100 + }, + { + "epoch": 0.23, + "learning_rate": 1.8031323248481758e-05, + "loss": 0.9648, + "step": 7101 + }, + { + "epoch": 0.23, + "learning_rate": 1.8030706072828187e-05, + "loss": 1.0107, + "step": 7102 + }, + { + "epoch": 0.23, + "learning_rate": 1.8030088811014026e-05, + "loss": 1.0874, + "step": 7103 + }, + { + "epoch": 0.23, + "learning_rate": 1.8029471463045904e-05, + "loss": 1.0869, + "step": 7104 + }, + { + "epoch": 0.23, + "learning_rate": 1.8028854028930434e-05, + "loss": 1.1255, + "step": 7105 + }, + { + "epoch": 0.23, + "learning_rate": 1.8028236508674252e-05, + "loss": 1.0723, + "step": 7106 + }, + { + "epoch": 0.23, + "learning_rate": 1.802761890228397e-05, + "loss": 1.0044, + "step": 7107 + }, + { + "epoch": 0.23, + "learning_rate": 1.8027001209766227e-05, + "loss": 1.0718, + "step": 7108 + }, + { + "epoch": 0.23, + "learning_rate": 1.8026383431127645e-05, + "loss": 1.123, + "step": 7109 + }, + { + "epoch": 0.23, + "learning_rate": 1.8025765566374852e-05, + "loss": 1.0898, + "step": 7110 + }, + { + "epoch": 0.23, + "learning_rate": 1.8025147615514475e-05, + "loss": 0.999, + "step": 7111 + }, + { + "epoch": 0.23, + "learning_rate": 1.802452957855315e-05, + "loss": 1.1528, + "step": 7112 + }, + { + "epoch": 0.23, + "learning_rate": 1.80239114554975e-05, + "loss": 1.0239, + "step": 7113 + }, + { + "epoch": 0.23, + "learning_rate": 1.8023293246354157e-05, + "loss": 1.1084, + "step": 7114 + }, + { + "epoch": 0.23, + "learning_rate": 1.8022674951129764e-05, + "loss": 0.9888, + "step": 7115 + }, + { + "epoch": 0.23, + "learning_rate": 1.8022056569830947e-05, + "loss": 1.0107, + "step": 7116 + }, + { + "epoch": 0.23, + "learning_rate": 1.8021438102464336e-05, + "loss": 0.998, + "step": 7117 + }, + { + "epoch": 0.23, + "learning_rate": 1.8020819549036576e-05, + "loss": 1.0713, + "step": 7118 + }, + { + "epoch": 0.23, + "learning_rate": 1.80202009095543e-05, + "loss": 1.1245, + "step": 7119 + }, + { + "epoch": 0.23, + "learning_rate": 1.801958218402414e-05, + "loss": 0.5454, + "step": 7120 + }, + { + "epoch": 0.23, + "learning_rate": 1.8018963372452744e-05, + "loss": 1.1655, + "step": 7121 + }, + { + "epoch": 0.23, + "learning_rate": 1.801834447484674e-05, + "loss": 1.0801, + "step": 7122 + }, + { + "epoch": 0.23, + "learning_rate": 1.801772549121278e-05, + "loss": 1.1675, + "step": 7123 + }, + { + "epoch": 0.23, + "learning_rate": 1.80171064215575e-05, + "loss": 1.0527, + "step": 7124 + }, + { + "epoch": 0.23, + "learning_rate": 1.8016487265887536e-05, + "loss": 1.061, + "step": 7125 + }, + { + "epoch": 0.23, + "learning_rate": 1.8015868024209537e-05, + "loss": 1.083, + "step": 7126 + }, + { + "epoch": 0.23, + "learning_rate": 1.801524869653015e-05, + "loss": 1.165, + "step": 7127 + }, + { + "epoch": 0.23, + "learning_rate": 1.801462928285601e-05, + "loss": 0.8799, + "step": 7128 + }, + { + "epoch": 0.23, + "learning_rate": 1.801400978319377e-05, + "loss": 1.0005, + "step": 7129 + }, + { + "epoch": 0.23, + "learning_rate": 1.8013390197550074e-05, + "loss": 1.0728, + "step": 7130 + }, + { + "epoch": 0.23, + "learning_rate": 1.8012770525931572e-05, + "loss": 1.0469, + "step": 7131 + }, + { + "epoch": 0.23, + "learning_rate": 1.8012150768344908e-05, + "loss": 1.1553, + "step": 7132 + }, + { + "epoch": 0.23, + "learning_rate": 1.8011530924796732e-05, + "loss": 1.0791, + "step": 7133 + }, + { + "epoch": 0.23, + "learning_rate": 1.80109109952937e-05, + "loss": 1.0156, + "step": 7134 + }, + { + "epoch": 0.23, + "learning_rate": 1.8010290979842454e-05, + "loss": 1.1123, + "step": 7135 + }, + { + "epoch": 0.23, + "learning_rate": 1.8009670878449656e-05, + "loss": 1.2031, + "step": 7136 + }, + { + "epoch": 0.23, + "learning_rate": 1.8009050691121954e-05, + "loss": 1.0874, + "step": 7137 + }, + { + "epoch": 0.23, + "learning_rate": 1.8008430417866002e-05, + "loss": 1.1738, + "step": 7138 + }, + { + "epoch": 0.23, + "learning_rate": 1.800781005868845e-05, + "loss": 1.0425, + "step": 7139 + }, + { + "epoch": 0.23, + "learning_rate": 1.8007189613595962e-05, + "loss": 1.1372, + "step": 7140 + }, + { + "epoch": 0.23, + "learning_rate": 1.8006569082595192e-05, + "loss": 0.9858, + "step": 7141 + }, + { + "epoch": 0.23, + "learning_rate": 1.8005948465692796e-05, + "loss": 1.0811, + "step": 7142 + }, + { + "epoch": 0.23, + "learning_rate": 1.8005327762895436e-05, + "loss": 1.0898, + "step": 7143 + }, + { + "epoch": 0.23, + "learning_rate": 1.8004706974209764e-05, + "loss": 1.103, + "step": 7144 + }, + { + "epoch": 0.23, + "learning_rate": 1.800408609964245e-05, + "loss": 1.0591, + "step": 7145 + }, + { + "epoch": 0.23, + "learning_rate": 1.800346513920015e-05, + "loss": 1.1528, + "step": 7146 + }, + { + "epoch": 0.23, + "learning_rate": 1.8002844092889526e-05, + "loss": 1.1128, + "step": 7147 + }, + { + "epoch": 0.23, + "learning_rate": 1.800222296071724e-05, + "loss": 1.0786, + "step": 7148 + }, + { + "epoch": 0.23, + "learning_rate": 1.800160174268996e-05, + "loss": 1.0103, + "step": 7149 + }, + { + "epoch": 0.23, + "learning_rate": 1.8000980438814347e-05, + "loss": 1.1274, + "step": 7150 + }, + { + "epoch": 0.23, + "learning_rate": 1.800035904909707e-05, + "loss": 1.0581, + "step": 7151 + }, + { + "epoch": 0.23, + "learning_rate": 1.7999737573544797e-05, + "loss": 1.0088, + "step": 7152 + }, + { + "epoch": 0.23, + "learning_rate": 1.799911601216419e-05, + "loss": 1.0151, + "step": 7153 + }, + { + "epoch": 0.23, + "learning_rate": 1.7998494364961927e-05, + "loss": 1.1616, + "step": 7154 + }, + { + "epoch": 0.23, + "learning_rate": 1.7997872631944667e-05, + "loss": 1.1006, + "step": 7155 + }, + { + "epoch": 0.23, + "learning_rate": 1.7997250813119088e-05, + "loss": 1.02, + "step": 7156 + }, + { + "epoch": 0.23, + "learning_rate": 1.7996628908491855e-05, + "loss": 1.062, + "step": 7157 + }, + { + "epoch": 0.23, + "learning_rate": 1.7996006918069645e-05, + "loss": 1.124, + "step": 7158 + }, + { + "epoch": 0.23, + "learning_rate": 1.7995384841859135e-05, + "loss": 1.0288, + "step": 7159 + }, + { + "epoch": 0.23, + "learning_rate": 1.799476267986699e-05, + "loss": 1.0327, + "step": 7160 + }, + { + "epoch": 0.23, + "learning_rate": 1.799414043209989e-05, + "loss": 1.0698, + "step": 7161 + }, + { + "epoch": 0.23, + "learning_rate": 1.799351809856451e-05, + "loss": 1.0625, + "step": 7162 + }, + { + "epoch": 0.23, + "learning_rate": 1.799289567926753e-05, + "loss": 1.0576, + "step": 7163 + }, + { + "epoch": 0.23, + "learning_rate": 1.7992273174215622e-05, + "loss": 1.1084, + "step": 7164 + }, + { + "epoch": 0.23, + "learning_rate": 1.7991650583415473e-05, + "loss": 1.1675, + "step": 7165 + }, + { + "epoch": 0.23, + "learning_rate": 1.7991027906873753e-05, + "loss": 1.0396, + "step": 7166 + }, + { + "epoch": 0.23, + "learning_rate": 1.799040514459715e-05, + "loss": 1.0469, + "step": 7167 + }, + { + "epoch": 0.23, + "learning_rate": 1.7989782296592343e-05, + "loss": 1.0522, + "step": 7168 + }, + { + "epoch": 0.23, + "learning_rate": 1.7989159362866013e-05, + "loss": 1.1138, + "step": 7169 + }, + { + "epoch": 0.23, + "learning_rate": 1.7988536343424847e-05, + "loss": 1.1934, + "step": 7170 + }, + { + "epoch": 0.23, + "learning_rate": 1.7987913238275528e-05, + "loss": 1.1309, + "step": 7171 + }, + { + "epoch": 0.23, + "learning_rate": 1.798729004742474e-05, + "loss": 1.0576, + "step": 7172 + }, + { + "epoch": 0.23, + "learning_rate": 1.7986666770879168e-05, + "loss": 1.0898, + "step": 7173 + }, + { + "epoch": 0.23, + "learning_rate": 1.7986043408645504e-05, + "loss": 1.0312, + "step": 7174 + }, + { + "epoch": 0.23, + "learning_rate": 1.798541996073043e-05, + "loss": 1.1294, + "step": 7175 + }, + { + "epoch": 0.23, + "learning_rate": 1.798479642714064e-05, + "loss": 0.533, + "step": 7176 + }, + { + "epoch": 0.23, + "learning_rate": 1.7984172807882817e-05, + "loss": 0.9185, + "step": 7177 + }, + { + "epoch": 0.23, + "learning_rate": 1.798354910296366e-05, + "loss": 0.9858, + "step": 7178 + }, + { + "epoch": 0.23, + "learning_rate": 1.7982925312389856e-05, + "loss": 1.2515, + "step": 7179 + }, + { + "epoch": 0.23, + "learning_rate": 1.7982301436168098e-05, + "loss": 1.1553, + "step": 7180 + }, + { + "epoch": 0.23, + "learning_rate": 1.7981677474305082e-05, + "loss": 1.1523, + "step": 7181 + }, + { + "epoch": 0.23, + "learning_rate": 1.79810534268075e-05, + "loss": 1.0786, + "step": 7182 + }, + { + "epoch": 0.23, + "learning_rate": 1.7980429293682045e-05, + "loss": 1.0562, + "step": 7183 + }, + { + "epoch": 0.23, + "learning_rate": 1.7979805074935416e-05, + "loss": 1.1279, + "step": 7184 + }, + { + "epoch": 0.23, + "learning_rate": 1.7979180770574313e-05, + "loss": 1.0547, + "step": 7185 + }, + { + "epoch": 0.23, + "learning_rate": 1.7978556380605426e-05, + "loss": 1.0508, + "step": 7186 + }, + { + "epoch": 0.23, + "learning_rate": 1.7977931905035464e-05, + "loss": 1.0225, + "step": 7187 + }, + { + "epoch": 0.23, + "learning_rate": 1.797730734387112e-05, + "loss": 1.0977, + "step": 7188 + }, + { + "epoch": 0.23, + "learning_rate": 1.79766826971191e-05, + "loss": 1.1235, + "step": 7189 + }, + { + "epoch": 0.23, + "learning_rate": 1.7976057964786103e-05, + "loss": 0.979, + "step": 7190 + }, + { + "epoch": 0.23, + "learning_rate": 1.7975433146878828e-05, + "loss": 1.0752, + "step": 7191 + }, + { + "epoch": 0.23, + "learning_rate": 1.7974808243403985e-05, + "loss": 1.083, + "step": 7192 + }, + { + "epoch": 0.23, + "learning_rate": 1.7974183254368275e-05, + "loss": 1.0454, + "step": 7193 + }, + { + "epoch": 0.23, + "learning_rate": 1.79735581797784e-05, + "loss": 1.0381, + "step": 7194 + }, + { + "epoch": 0.23, + "learning_rate": 1.7972933019641073e-05, + "loss": 1.1733, + "step": 7195 + }, + { + "epoch": 0.23, + "learning_rate": 1.7972307773963003e-05, + "loss": 1.1396, + "step": 7196 + }, + { + "epoch": 0.23, + "learning_rate": 1.7971682442750888e-05, + "loss": 1.1938, + "step": 7197 + }, + { + "epoch": 0.23, + "learning_rate": 1.7971057026011447e-05, + "loss": 1.085, + "step": 7198 + }, + { + "epoch": 0.23, + "learning_rate": 1.7970431523751383e-05, + "loss": 0.5623, + "step": 7199 + }, + { + "epoch": 0.23, + "learning_rate": 1.7969805935977413e-05, + "loss": 1.0474, + "step": 7200 + }, + { + "epoch": 0.23, + "learning_rate": 1.7969180262696243e-05, + "loss": 0.999, + "step": 7201 + }, + { + "epoch": 0.23, + "learning_rate": 1.796855450391459e-05, + "loss": 1.1733, + "step": 7202 + }, + { + "epoch": 0.23, + "learning_rate": 1.7967928659639167e-05, + "loss": 0.5188, + "step": 7203 + }, + { + "epoch": 0.23, + "learning_rate": 1.796730272987669e-05, + "loss": 1.1265, + "step": 7204 + }, + { + "epoch": 0.23, + "learning_rate": 1.7966676714633867e-05, + "loss": 1.165, + "step": 7205 + }, + { + "epoch": 0.23, + "learning_rate": 1.7966050613917422e-05, + "loss": 0.9951, + "step": 7206 + }, + { + "epoch": 0.23, + "learning_rate": 1.796542442773407e-05, + "loss": 1.0713, + "step": 7207 + }, + { + "epoch": 0.23, + "learning_rate": 1.7964798156090533e-05, + "loss": 1.0903, + "step": 7208 + }, + { + "epoch": 0.23, + "learning_rate": 1.7964171798993525e-05, + "loss": 1.0293, + "step": 7209 + }, + { + "epoch": 0.23, + "learning_rate": 1.7963545356449765e-05, + "loss": 1.0532, + "step": 7210 + }, + { + "epoch": 0.23, + "learning_rate": 1.796291882846598e-05, + "loss": 0.8906, + "step": 7211 + }, + { + "epoch": 0.23, + "learning_rate": 1.796229221504889e-05, + "loss": 1.0225, + "step": 7212 + }, + { + "epoch": 0.23, + "learning_rate": 1.7961665516205214e-05, + "loss": 1.0142, + "step": 7213 + }, + { + "epoch": 0.23, + "learning_rate": 1.796103873194168e-05, + "loss": 0.9854, + "step": 7214 + }, + { + "epoch": 0.23, + "learning_rate": 1.7960411862265007e-05, + "loss": 1.1206, + "step": 7215 + }, + { + "epoch": 0.23, + "learning_rate": 1.7959784907181932e-05, + "loss": 1.1221, + "step": 7216 + }, + { + "epoch": 0.23, + "learning_rate": 1.795915786669917e-05, + "loss": 1.0825, + "step": 7217 + }, + { + "epoch": 0.23, + "learning_rate": 1.7958530740823455e-05, + "loss": 0.9053, + "step": 7218 + }, + { + "epoch": 0.23, + "learning_rate": 1.795790352956151e-05, + "loss": 1.0098, + "step": 7219 + }, + { + "epoch": 0.23, + "learning_rate": 1.795727623292007e-05, + "loss": 0.877, + "step": 7220 + }, + { + "epoch": 0.23, + "learning_rate": 1.795664885090586e-05, + "loss": 1.0571, + "step": 7221 + }, + { + "epoch": 0.23, + "learning_rate": 1.795602138352562e-05, + "loss": 1.0718, + "step": 7222 + }, + { + "epoch": 0.23, + "learning_rate": 1.795539383078607e-05, + "loss": 1.1421, + "step": 7223 + }, + { + "epoch": 0.23, + "learning_rate": 1.7954766192693948e-05, + "loss": 1.0249, + "step": 7224 + }, + { + "epoch": 0.23, + "learning_rate": 1.7954138469255994e-05, + "loss": 1.1084, + "step": 7225 + }, + { + "epoch": 0.23, + "learning_rate": 1.7953510660478936e-05, + "loss": 1.0762, + "step": 7226 + }, + { + "epoch": 0.23, + "learning_rate": 1.7952882766369508e-05, + "loss": 1.1377, + "step": 7227 + }, + { + "epoch": 0.23, + "learning_rate": 1.7952254786934454e-05, + "loss": 1.0894, + "step": 7228 + }, + { + "epoch": 0.23, + "learning_rate": 1.79516267221805e-05, + "loss": 1.1128, + "step": 7229 + }, + { + "epoch": 0.23, + "learning_rate": 1.79509985721144e-05, + "loss": 1.0811, + "step": 7230 + }, + { + "epoch": 0.23, + "learning_rate": 1.7950370336742883e-05, + "loss": 1.0337, + "step": 7231 + }, + { + "epoch": 0.23, + "learning_rate": 1.7949742016072688e-05, + "loss": 1.0312, + "step": 7232 + }, + { + "epoch": 0.23, + "learning_rate": 1.794911361011056e-05, + "loss": 1.0781, + "step": 7233 + }, + { + "epoch": 0.23, + "learning_rate": 1.794848511886324e-05, + "loss": 1.0405, + "step": 7234 + }, + { + "epoch": 0.23, + "learning_rate": 1.7947856542337477e-05, + "loss": 1.0347, + "step": 7235 + }, + { + "epoch": 0.23, + "learning_rate": 1.7947227880540006e-05, + "loss": 1.1133, + "step": 7236 + }, + { + "epoch": 0.23, + "learning_rate": 1.7946599133477575e-05, + "loss": 1.0142, + "step": 7237 + }, + { + "epoch": 0.23, + "learning_rate": 1.7945970301156932e-05, + "loss": 0.9805, + "step": 7238 + }, + { + "epoch": 0.23, + "learning_rate": 1.7945341383584818e-05, + "loss": 1.0415, + "step": 7239 + }, + { + "epoch": 0.23, + "learning_rate": 1.794471238076799e-05, + "loss": 0.9736, + "step": 7240 + }, + { + "epoch": 0.23, + "learning_rate": 1.7944083292713187e-05, + "loss": 1.1538, + "step": 7241 + }, + { + "epoch": 0.23, + "learning_rate": 1.7943454119427162e-05, + "loss": 1.0522, + "step": 7242 + }, + { + "epoch": 0.23, + "learning_rate": 1.7942824860916667e-05, + "loss": 1.063, + "step": 7243 + }, + { + "epoch": 0.23, + "learning_rate": 1.7942195517188453e-05, + "loss": 0.9351, + "step": 7244 + }, + { + "epoch": 0.23, + "learning_rate": 1.794156608824927e-05, + "loss": 1.0718, + "step": 7245 + }, + { + "epoch": 0.23, + "learning_rate": 1.7940936574105872e-05, + "loss": 1.1328, + "step": 7246 + }, + { + "epoch": 0.23, + "learning_rate": 1.7940306974765012e-05, + "loss": 1.2451, + "step": 7247 + }, + { + "epoch": 0.23, + "learning_rate": 1.7939677290233448e-05, + "loss": 1.022, + "step": 7248 + }, + { + "epoch": 0.23, + "learning_rate": 1.7939047520517935e-05, + "loss": 1.082, + "step": 7249 + }, + { + "epoch": 0.23, + "learning_rate": 1.7938417665625226e-05, + "loss": 1.1284, + "step": 7250 + }, + { + "epoch": 0.23, + "learning_rate": 1.793778772556208e-05, + "loss": 1.1182, + "step": 7251 + }, + { + "epoch": 0.23, + "learning_rate": 1.7937157700335258e-05, + "loss": 1.0229, + "step": 7252 + }, + { + "epoch": 0.23, + "learning_rate": 1.793652758995152e-05, + "loss": 0.9775, + "step": 7253 + }, + { + "epoch": 0.23, + "learning_rate": 1.7935897394417626e-05, + "loss": 1.1021, + "step": 7254 + }, + { + "epoch": 0.23, + "learning_rate": 1.793526711374033e-05, + "loss": 1.0005, + "step": 7255 + }, + { + "epoch": 0.23, + "learning_rate": 1.7934636747926407e-05, + "loss": 1.1924, + "step": 7256 + }, + { + "epoch": 0.23, + "learning_rate": 1.793400629698261e-05, + "loss": 1.0161, + "step": 7257 + }, + { + "epoch": 0.23, + "learning_rate": 1.793337576091571e-05, + "loss": 1.0664, + "step": 7258 + }, + { + "epoch": 0.23, + "learning_rate": 1.7932745139732462e-05, + "loss": 1.0757, + "step": 7259 + }, + { + "epoch": 0.23, + "learning_rate": 1.7932114433439643e-05, + "loss": 1.084, + "step": 7260 + }, + { + "epoch": 0.23, + "learning_rate": 1.793148364204401e-05, + "loss": 1.0938, + "step": 7261 + }, + { + "epoch": 0.23, + "learning_rate": 1.793085276555234e-05, + "loss": 1.0132, + "step": 7262 + }, + { + "epoch": 0.23, + "learning_rate": 1.7930221803971395e-05, + "loss": 1.1313, + "step": 7263 + }, + { + "epoch": 0.23, + "learning_rate": 1.7929590757307948e-05, + "loss": 1.0342, + "step": 7264 + }, + { + "epoch": 0.23, + "learning_rate": 1.792895962556877e-05, + "loss": 1.0254, + "step": 7265 + }, + { + "epoch": 0.23, + "learning_rate": 1.7928328408760627e-05, + "loss": 1.002, + "step": 7266 + }, + { + "epoch": 0.23, + "learning_rate": 1.7927697106890296e-05, + "loss": 1.2041, + "step": 7267 + }, + { + "epoch": 0.23, + "learning_rate": 1.7927065719964548e-05, + "loss": 1.0366, + "step": 7268 + }, + { + "epoch": 0.23, + "learning_rate": 1.792643424799016e-05, + "loss": 1.0386, + "step": 7269 + }, + { + "epoch": 0.23, + "learning_rate": 1.7925802690973902e-05, + "loss": 1.1011, + "step": 7270 + }, + { + "epoch": 0.23, + "learning_rate": 1.792517104892256e-05, + "loss": 0.9971, + "step": 7271 + }, + { + "epoch": 0.23, + "learning_rate": 1.79245393218429e-05, + "loss": 1.1025, + "step": 7272 + }, + { + "epoch": 0.23, + "learning_rate": 1.79239075097417e-05, + "loss": 1.0396, + "step": 7273 + }, + { + "epoch": 0.23, + "learning_rate": 1.7923275612625747e-05, + "loss": 1.0586, + "step": 7274 + }, + { + "epoch": 0.23, + "learning_rate": 1.792264363050181e-05, + "loss": 1.0591, + "step": 7275 + }, + { + "epoch": 0.23, + "learning_rate": 1.792201156337668e-05, + "loss": 1.1489, + "step": 7276 + }, + { + "epoch": 0.23, + "learning_rate": 1.7921379411257132e-05, + "loss": 1.0005, + "step": 7277 + }, + { + "epoch": 0.23, + "learning_rate": 1.792074717414995e-05, + "loss": 1.0396, + "step": 7278 + }, + { + "epoch": 0.23, + "learning_rate": 1.7920114852061917e-05, + "loss": 0.9419, + "step": 7279 + }, + { + "epoch": 0.23, + "learning_rate": 1.791948244499982e-05, + "loss": 1.0386, + "step": 7280 + }, + { + "epoch": 0.23, + "learning_rate": 1.791884995297044e-05, + "loss": 1.0244, + "step": 7281 + }, + { + "epoch": 0.23, + "learning_rate": 1.791821737598056e-05, + "loss": 0.5513, + "step": 7282 + }, + { + "epoch": 0.23, + "learning_rate": 1.7917584714036976e-05, + "loss": 1.0552, + "step": 7283 + }, + { + "epoch": 0.23, + "learning_rate": 1.791695196714647e-05, + "loss": 1.1602, + "step": 7284 + }, + { + "epoch": 0.23, + "learning_rate": 1.7916319135315826e-05, + "loss": 1.0195, + "step": 7285 + }, + { + "epoch": 0.23, + "learning_rate": 1.7915686218551845e-05, + "loss": 1.0498, + "step": 7286 + }, + { + "epoch": 0.23, + "learning_rate": 1.7915053216861307e-05, + "loss": 0.9893, + "step": 7287 + }, + { + "epoch": 0.23, + "learning_rate": 1.791442013025101e-05, + "loss": 1.1328, + "step": 7288 + }, + { + "epoch": 0.23, + "learning_rate": 1.791378695872775e-05, + "loss": 0.9907, + "step": 7289 + }, + { + "epoch": 0.23, + "learning_rate": 1.7913153702298308e-05, + "loss": 1.0967, + "step": 7290 + }, + { + "epoch": 0.23, + "learning_rate": 1.7912520360969487e-05, + "loss": 1.2139, + "step": 7291 + }, + { + "epoch": 0.23, + "learning_rate": 1.7911886934748076e-05, + "loss": 1.1553, + "step": 7292 + }, + { + "epoch": 0.23, + "learning_rate": 1.7911253423640876e-05, + "loss": 0.9961, + "step": 7293 + }, + { + "epoch": 0.23, + "learning_rate": 1.7910619827654686e-05, + "loss": 1.0244, + "step": 7294 + }, + { + "epoch": 0.23, + "learning_rate": 1.7909986146796296e-05, + "loss": 1.0776, + "step": 7295 + }, + { + "epoch": 0.23, + "learning_rate": 1.790935238107251e-05, + "loss": 1.0713, + "step": 7296 + }, + { + "epoch": 0.23, + "learning_rate": 1.790871853049013e-05, + "loss": 1.2637, + "step": 7297 + }, + { + "epoch": 0.23, + "learning_rate": 1.790808459505595e-05, + "loss": 1.1846, + "step": 7298 + }, + { + "epoch": 0.23, + "learning_rate": 1.7907450574776776e-05, + "loss": 1.083, + "step": 7299 + }, + { + "epoch": 0.23, + "learning_rate": 1.7906816469659406e-05, + "loss": 1.125, + "step": 7300 + }, + { + "epoch": 0.23, + "learning_rate": 1.790618227971065e-05, + "loss": 1.0659, + "step": 7301 + }, + { + "epoch": 0.23, + "learning_rate": 1.790554800493731e-05, + "loss": 1.127, + "step": 7302 + }, + { + "epoch": 0.23, + "learning_rate": 1.7904913645346184e-05, + "loss": 1.0195, + "step": 7303 + }, + { + "epoch": 0.23, + "learning_rate": 1.7904279200944082e-05, + "loss": 1.104, + "step": 7304 + }, + { + "epoch": 0.23, + "learning_rate": 1.7903644671737818e-05, + "loss": 0.9458, + "step": 7305 + }, + { + "epoch": 0.23, + "learning_rate": 1.790301005773419e-05, + "loss": 1.0815, + "step": 7306 + }, + { + "epoch": 0.23, + "learning_rate": 1.790237535894001e-05, + "loss": 0.9941, + "step": 7307 + }, + { + "epoch": 0.23, + "learning_rate": 1.7901740575362093e-05, + "loss": 0.9624, + "step": 7308 + }, + { + "epoch": 0.23, + "learning_rate": 1.790110570700724e-05, + "loss": 1.1528, + "step": 7309 + }, + { + "epoch": 0.23, + "learning_rate": 1.790047075388227e-05, + "loss": 1.0825, + "step": 7310 + }, + { + "epoch": 0.23, + "learning_rate": 1.7899835715993993e-05, + "loss": 1.0718, + "step": 7311 + }, + { + "epoch": 0.23, + "learning_rate": 1.789920059334922e-05, + "loss": 0.9883, + "step": 7312 + }, + { + "epoch": 0.23, + "learning_rate": 1.789856538595477e-05, + "loss": 1.0957, + "step": 7313 + }, + { + "epoch": 0.23, + "learning_rate": 1.7897930093817453e-05, + "loss": 1.0522, + "step": 7314 + }, + { + "epoch": 0.23, + "learning_rate": 1.7897294716944088e-05, + "loss": 1.0918, + "step": 7315 + }, + { + "epoch": 0.23, + "learning_rate": 1.789665925534149e-05, + "loss": 1.0801, + "step": 7316 + }, + { + "epoch": 0.23, + "learning_rate": 1.7896023709016476e-05, + "loss": 1.0166, + "step": 7317 + }, + { + "epoch": 0.23, + "learning_rate": 1.789538807797587e-05, + "loss": 0.9272, + "step": 7318 + }, + { + "epoch": 0.23, + "learning_rate": 1.7894752362226486e-05, + "loss": 0.5054, + "step": 7319 + }, + { + "epoch": 0.23, + "learning_rate": 1.7894116561775147e-05, + "loss": 1.249, + "step": 7320 + }, + { + "epoch": 0.23, + "learning_rate": 1.7893480676628674e-05, + "loss": 1.1787, + "step": 7321 + }, + { + "epoch": 0.23, + "learning_rate": 1.789284470679389e-05, + "loss": 1.1816, + "step": 7322 + }, + { + "epoch": 0.23, + "learning_rate": 1.789220865227762e-05, + "loss": 0.9648, + "step": 7323 + }, + { + "epoch": 0.23, + "learning_rate": 1.789157251308668e-05, + "loss": 1.1289, + "step": 7324 + }, + { + "epoch": 0.23, + "learning_rate": 1.7890936289227904e-05, + "loss": 1.1465, + "step": 7325 + }, + { + "epoch": 0.23, + "learning_rate": 1.7890299980708114e-05, + "loss": 1.1392, + "step": 7326 + }, + { + "epoch": 0.23, + "learning_rate": 1.788966358753414e-05, + "loss": 1.0415, + "step": 7327 + }, + { + "epoch": 0.23, + "learning_rate": 1.7889027109712808e-05, + "loss": 1.1484, + "step": 7328 + }, + { + "epoch": 0.23, + "learning_rate": 1.7888390547250944e-05, + "loss": 1.0474, + "step": 7329 + }, + { + "epoch": 0.23, + "learning_rate": 1.788775390015538e-05, + "loss": 1.0645, + "step": 7330 + }, + { + "epoch": 0.23, + "learning_rate": 1.7887117168432942e-05, + "loss": 1.0752, + "step": 7331 + }, + { + "epoch": 0.23, + "learning_rate": 1.7886480352090475e-05, + "loss": 1.0625, + "step": 7332 + }, + { + "epoch": 0.23, + "learning_rate": 1.7885843451134794e-05, + "loss": 0.9766, + "step": 7333 + }, + { + "epoch": 0.23, + "learning_rate": 1.7885206465572746e-05, + "loss": 1.1538, + "step": 7334 + }, + { + "epoch": 0.23, + "learning_rate": 1.7884569395411158e-05, + "loss": 0.9629, + "step": 7335 + }, + { + "epoch": 0.23, + "learning_rate": 1.7883932240656867e-05, + "loss": 0.9878, + "step": 7336 + }, + { + "epoch": 0.23, + "learning_rate": 1.7883295001316704e-05, + "loss": 1.0928, + "step": 7337 + }, + { + "epoch": 0.23, + "learning_rate": 1.7882657677397516e-05, + "loss": 1.0337, + "step": 7338 + }, + { + "epoch": 0.23, + "learning_rate": 1.7882020268906134e-05, + "loss": 1.019, + "step": 7339 + }, + { + "epoch": 0.23, + "learning_rate": 1.7881382775849395e-05, + "loss": 1.1313, + "step": 7340 + }, + { + "epoch": 0.23, + "learning_rate": 1.7880745198234146e-05, + "loss": 1.1553, + "step": 7341 + }, + { + "epoch": 0.23, + "learning_rate": 1.788010753606722e-05, + "loss": 0.9917, + "step": 7342 + }, + { + "epoch": 0.23, + "learning_rate": 1.7879469789355464e-05, + "loss": 1.0386, + "step": 7343 + }, + { + "epoch": 0.23, + "learning_rate": 1.7878831958105716e-05, + "loss": 1.1421, + "step": 7344 + }, + { + "epoch": 0.23, + "learning_rate": 1.787819404232482e-05, + "loss": 0.9263, + "step": 7345 + }, + { + "epoch": 0.23, + "learning_rate": 1.787755604201962e-05, + "loss": 1.0332, + "step": 7346 + }, + { + "epoch": 0.23, + "learning_rate": 1.7876917957196966e-05, + "loss": 1.1514, + "step": 7347 + }, + { + "epoch": 0.23, + "learning_rate": 1.78762797878637e-05, + "loss": 1.1538, + "step": 7348 + }, + { + "epoch": 0.24, + "learning_rate": 1.7875641534026664e-05, + "loss": 1.0688, + "step": 7349 + }, + { + "epoch": 0.24, + "learning_rate": 1.7875003195692714e-05, + "loss": 1.0054, + "step": 7350 + }, + { + "epoch": 0.24, + "learning_rate": 1.78743647728687e-05, + "loss": 1.0972, + "step": 7351 + }, + { + "epoch": 0.24, + "learning_rate": 1.7873726265561457e-05, + "loss": 0.9878, + "step": 7352 + }, + { + "epoch": 0.24, + "learning_rate": 1.787308767377785e-05, + "loss": 1.2656, + "step": 7353 + }, + { + "epoch": 0.24, + "learning_rate": 1.787244899752473e-05, + "loss": 1.0879, + "step": 7354 + }, + { + "epoch": 0.24, + "learning_rate": 1.7871810236808936e-05, + "loss": 1.0981, + "step": 7355 + }, + { + "epoch": 0.24, + "learning_rate": 1.7871171391637338e-05, + "loss": 1.2559, + "step": 7356 + }, + { + "epoch": 0.24, + "learning_rate": 1.787053246201678e-05, + "loss": 1.0601, + "step": 7357 + }, + { + "epoch": 0.24, + "learning_rate": 1.786989344795412e-05, + "loss": 1.1567, + "step": 7358 + }, + { + "epoch": 0.24, + "learning_rate": 1.786925434945621e-05, + "loss": 1.1187, + "step": 7359 + }, + { + "epoch": 0.24, + "learning_rate": 1.786861516652991e-05, + "loss": 1.1548, + "step": 7360 + }, + { + "epoch": 0.24, + "learning_rate": 1.7867975899182083e-05, + "loss": 0.9214, + "step": 7361 + }, + { + "epoch": 0.24, + "learning_rate": 1.7867336547419582e-05, + "loss": 1.1006, + "step": 7362 + }, + { + "epoch": 0.24, + "learning_rate": 1.786669711124926e-05, + "loss": 1.0527, + "step": 7363 + }, + { + "epoch": 0.24, + "learning_rate": 1.7866057590677993e-05, + "loss": 1.1084, + "step": 7364 + }, + { + "epoch": 0.24, + "learning_rate": 1.786541798571263e-05, + "loss": 0.9741, + "step": 7365 + }, + { + "epoch": 0.24, + "learning_rate": 1.7864778296360035e-05, + "loss": 1.1074, + "step": 7366 + }, + { + "epoch": 0.24, + "learning_rate": 1.786413852262708e-05, + "loss": 1.0649, + "step": 7367 + }, + { + "epoch": 0.24, + "learning_rate": 1.7863498664520614e-05, + "loss": 1.0098, + "step": 7368 + }, + { + "epoch": 0.24, + "learning_rate": 1.7862858722047513e-05, + "loss": 1.1592, + "step": 7369 + }, + { + "epoch": 0.24, + "learning_rate": 1.786221869521464e-05, + "loss": 1.2695, + "step": 7370 + }, + { + "epoch": 0.24, + "learning_rate": 1.7861578584028863e-05, + "loss": 1.1777, + "step": 7371 + }, + { + "epoch": 0.24, + "learning_rate": 1.786093838849705e-05, + "loss": 1.106, + "step": 7372 + }, + { + "epoch": 0.24, + "learning_rate": 1.7860298108626063e-05, + "loss": 1.0942, + "step": 7373 + }, + { + "epoch": 0.24, + "learning_rate": 1.785965774442278e-05, + "loss": 1.1543, + "step": 7374 + }, + { + "epoch": 0.24, + "learning_rate": 1.7859017295894066e-05, + "loss": 1.127, + "step": 7375 + }, + { + "epoch": 0.24, + "learning_rate": 1.7858376763046797e-05, + "loss": 0.978, + "step": 7376 + }, + { + "epoch": 0.24, + "learning_rate": 1.7857736145887844e-05, + "loss": 1.083, + "step": 7377 + }, + { + "epoch": 0.24, + "learning_rate": 1.7857095444424078e-05, + "loss": 1.105, + "step": 7378 + }, + { + "epoch": 0.24, + "learning_rate": 1.785645465866237e-05, + "loss": 0.9648, + "step": 7379 + }, + { + "epoch": 0.24, + "learning_rate": 1.78558137886096e-05, + "loss": 1.0635, + "step": 7380 + }, + { + "epoch": 0.24, + "learning_rate": 1.7855172834272645e-05, + "loss": 1.0254, + "step": 7381 + }, + { + "epoch": 0.24, + "learning_rate": 1.7854531795658378e-05, + "loss": 1.2842, + "step": 7382 + }, + { + "epoch": 0.24, + "learning_rate": 1.7853890672773677e-05, + "loss": 1.04, + "step": 7383 + }, + { + "epoch": 0.24, + "learning_rate": 1.785324946562542e-05, + "loss": 1.0981, + "step": 7384 + }, + { + "epoch": 0.24, + "learning_rate": 1.785260817422049e-05, + "loss": 1.147, + "step": 7385 + }, + { + "epoch": 0.24, + "learning_rate": 1.7851966798565764e-05, + "loss": 0.9307, + "step": 7386 + }, + { + "epoch": 0.24, + "learning_rate": 1.7851325338668128e-05, + "loss": 1.082, + "step": 7387 + }, + { + "epoch": 0.24, + "learning_rate": 1.7850683794534457e-05, + "loss": 1.0068, + "step": 7388 + }, + { + "epoch": 0.24, + "learning_rate": 1.7850042166171637e-05, + "loss": 1.082, + "step": 7389 + }, + { + "epoch": 0.24, + "learning_rate": 1.7849400453586555e-05, + "loss": 0.9297, + "step": 7390 + }, + { + "epoch": 0.24, + "learning_rate": 1.784875865678609e-05, + "loss": 1.0947, + "step": 7391 + }, + { + "epoch": 0.24, + "learning_rate": 1.784811677577713e-05, + "loss": 1.0986, + "step": 7392 + }, + { + "epoch": 0.24, + "learning_rate": 1.7847474810566566e-05, + "loss": 1.0366, + "step": 7393 + }, + { + "epoch": 0.24, + "learning_rate": 1.7846832761161287e-05, + "loss": 1.0879, + "step": 7394 + }, + { + "epoch": 0.24, + "learning_rate": 1.784619062756817e-05, + "loss": 1.0723, + "step": 7395 + }, + { + "epoch": 0.24, + "learning_rate": 1.7845548409794115e-05, + "loss": 1.1265, + "step": 7396 + }, + { + "epoch": 0.24, + "learning_rate": 1.7844906107846006e-05, + "loss": 0.9775, + "step": 7397 + }, + { + "epoch": 0.24, + "learning_rate": 1.7844263721730737e-05, + "loss": 1.0596, + "step": 7398 + }, + { + "epoch": 0.24, + "learning_rate": 1.7843621251455202e-05, + "loss": 1.0288, + "step": 7399 + }, + { + "epoch": 0.24, + "learning_rate": 1.784297869702629e-05, + "loss": 1.1265, + "step": 7400 + }, + { + "epoch": 0.24, + "learning_rate": 1.7842336058450896e-05, + "loss": 1.0703, + "step": 7401 + }, + { + "epoch": 0.24, + "learning_rate": 1.7841693335735918e-05, + "loss": 1.0947, + "step": 7402 + }, + { + "epoch": 0.24, + "learning_rate": 1.7841050528888247e-05, + "loss": 0.9653, + "step": 7403 + }, + { + "epoch": 0.24, + "learning_rate": 1.7840407637914785e-05, + "loss": 1.1328, + "step": 7404 + }, + { + "epoch": 0.24, + "learning_rate": 1.7839764662822424e-05, + "loss": 0.9614, + "step": 7405 + }, + { + "epoch": 0.24, + "learning_rate": 1.7839121603618064e-05, + "loss": 0.9463, + "step": 7406 + }, + { + "epoch": 0.24, + "learning_rate": 1.783847846030861e-05, + "loss": 1.043, + "step": 7407 + }, + { + "epoch": 0.24, + "learning_rate": 1.7837835232900955e-05, + "loss": 0.9639, + "step": 7408 + }, + { + "epoch": 0.24, + "learning_rate": 1.7837191921402e-05, + "loss": 1.1011, + "step": 7409 + }, + { + "epoch": 0.24, + "learning_rate": 1.783654852581865e-05, + "loss": 0.9922, + "step": 7410 + }, + { + "epoch": 0.24, + "learning_rate": 1.7835905046157807e-05, + "loss": 1.1216, + "step": 7411 + }, + { + "epoch": 0.24, + "learning_rate": 1.7835261482426377e-05, + "loss": 1.0386, + "step": 7412 + }, + { + "epoch": 0.24, + "learning_rate": 1.783461783463126e-05, + "loss": 1.1318, + "step": 7413 + }, + { + "epoch": 0.24, + "learning_rate": 1.783397410277937e-05, + "loss": 0.9385, + "step": 7414 + }, + { + "epoch": 0.24, + "learning_rate": 1.78333302868776e-05, + "loss": 1.1929, + "step": 7415 + }, + { + "epoch": 0.24, + "learning_rate": 1.7832686386932875e-05, + "loss": 1.1445, + "step": 7416 + }, + { + "epoch": 0.24, + "learning_rate": 1.783204240295209e-05, + "loss": 1.0361, + "step": 7417 + }, + { + "epoch": 0.24, + "learning_rate": 1.7831398334942155e-05, + "loss": 1.1812, + "step": 7418 + }, + { + "epoch": 0.24, + "learning_rate": 1.7830754182909985e-05, + "loss": 1.0649, + "step": 7419 + }, + { + "epoch": 0.24, + "learning_rate": 1.783010994686249e-05, + "loss": 1.1055, + "step": 7420 + }, + { + "epoch": 0.24, + "learning_rate": 1.782946562680658e-05, + "loss": 1.0576, + "step": 7421 + }, + { + "epoch": 0.24, + "learning_rate": 1.7828821222749172e-05, + "loss": 0.9839, + "step": 7422 + }, + { + "epoch": 0.24, + "learning_rate": 1.7828176734697172e-05, + "loss": 1.0864, + "step": 7423 + }, + { + "epoch": 0.24, + "learning_rate": 1.7827532162657505e-05, + "loss": 1.0435, + "step": 7424 + }, + { + "epoch": 0.24, + "learning_rate": 1.7826887506637075e-05, + "loss": 1.0654, + "step": 7425 + }, + { + "epoch": 0.24, + "learning_rate": 1.782624276664281e-05, + "loss": 1.1206, + "step": 7426 + }, + { + "epoch": 0.24, + "learning_rate": 1.7825597942681618e-05, + "loss": 1.0522, + "step": 7427 + }, + { + "epoch": 0.24, + "learning_rate": 1.7824953034760424e-05, + "loss": 1.1104, + "step": 7428 + }, + { + "epoch": 0.24, + "learning_rate": 1.7824308042886142e-05, + "loss": 1.062, + "step": 7429 + }, + { + "epoch": 0.24, + "learning_rate": 1.7823662967065694e-05, + "loss": 1.0425, + "step": 7430 + }, + { + "epoch": 0.24, + "learning_rate": 1.7823017807306003e-05, + "loss": 0.5713, + "step": 7431 + }, + { + "epoch": 0.24, + "learning_rate": 1.782237256361399e-05, + "loss": 1.0903, + "step": 7432 + }, + { + "epoch": 0.24, + "learning_rate": 1.7821727235996572e-05, + "loss": 1.0464, + "step": 7433 + }, + { + "epoch": 0.24, + "learning_rate": 1.782108182446068e-05, + "loss": 1.0962, + "step": 7434 + }, + { + "epoch": 0.24, + "learning_rate": 1.7820436329013236e-05, + "loss": 1.02, + "step": 7435 + }, + { + "epoch": 0.24, + "learning_rate": 1.7819790749661168e-05, + "loss": 1.0425, + "step": 7436 + }, + { + "epoch": 0.24, + "learning_rate": 1.7819145086411394e-05, + "loss": 1.0498, + "step": 7437 + }, + { + "epoch": 0.24, + "learning_rate": 1.781849933927085e-05, + "loss": 1.1416, + "step": 7438 + }, + { + "epoch": 0.24, + "learning_rate": 1.7817853508246462e-05, + "loss": 1.1665, + "step": 7439 + }, + { + "epoch": 0.24, + "learning_rate": 1.781720759334516e-05, + "loss": 1.1084, + "step": 7440 + }, + { + "epoch": 0.24, + "learning_rate": 1.7816561594573867e-05, + "loss": 1.0942, + "step": 7441 + }, + { + "epoch": 0.24, + "learning_rate": 1.7815915511939525e-05, + "loss": 1.0796, + "step": 7442 + }, + { + "epoch": 0.24, + "learning_rate": 1.7815269345449056e-05, + "loss": 1.2188, + "step": 7443 + }, + { + "epoch": 0.24, + "learning_rate": 1.7814623095109396e-05, + "loss": 0.979, + "step": 7444 + }, + { + "epoch": 0.24, + "learning_rate": 1.781397676092748e-05, + "loss": 1.0073, + "step": 7445 + }, + { + "epoch": 0.24, + "learning_rate": 1.7813330342910243e-05, + "loss": 1.0869, + "step": 7446 + }, + { + "epoch": 0.24, + "learning_rate": 1.7812683841064617e-05, + "loss": 1.0664, + "step": 7447 + }, + { + "epoch": 0.24, + "learning_rate": 1.781203725539754e-05, + "loss": 1.0913, + "step": 7448 + }, + { + "epoch": 0.24, + "learning_rate": 1.781139058591595e-05, + "loss": 1.041, + "step": 7449 + }, + { + "epoch": 0.24, + "learning_rate": 1.7810743832626783e-05, + "loss": 0.9443, + "step": 7450 + }, + { + "epoch": 0.24, + "learning_rate": 1.7810096995536983e-05, + "loss": 1.0254, + "step": 7451 + }, + { + "epoch": 0.24, + "learning_rate": 1.7809450074653482e-05, + "loss": 1.1206, + "step": 7452 + }, + { + "epoch": 0.24, + "learning_rate": 1.7808803069983224e-05, + "loss": 0.96, + "step": 7453 + }, + { + "epoch": 0.24, + "learning_rate": 1.7808155981533157e-05, + "loss": 1.0723, + "step": 7454 + }, + { + "epoch": 0.24, + "learning_rate": 1.7807508809310213e-05, + "loss": 1.0576, + "step": 7455 + }, + { + "epoch": 0.24, + "learning_rate": 1.780686155332134e-05, + "loss": 1.1431, + "step": 7456 + }, + { + "epoch": 0.24, + "learning_rate": 1.7806214213573483e-05, + "loss": 1.0562, + "step": 7457 + }, + { + "epoch": 0.24, + "learning_rate": 1.7805566790073588e-05, + "loss": 1.0874, + "step": 7458 + }, + { + "epoch": 0.24, + "learning_rate": 1.7804919282828602e-05, + "loss": 1.0615, + "step": 7459 + }, + { + "epoch": 0.24, + "learning_rate": 1.780427169184547e-05, + "loss": 1.1973, + "step": 7460 + }, + { + "epoch": 0.24, + "learning_rate": 1.780362401713114e-05, + "loss": 1.0356, + "step": 7461 + }, + { + "epoch": 0.24, + "learning_rate": 1.7802976258692556e-05, + "loss": 1.085, + "step": 7462 + }, + { + "epoch": 0.24, + "learning_rate": 1.7802328416536675e-05, + "loss": 1.1978, + "step": 7463 + }, + { + "epoch": 0.24, + "learning_rate": 1.780168049067045e-05, + "loss": 1.0776, + "step": 7464 + }, + { + "epoch": 0.24, + "learning_rate": 1.7801032481100825e-05, + "loss": 1.0811, + "step": 7465 + }, + { + "epoch": 0.24, + "learning_rate": 1.7800384387834753e-05, + "loss": 1.0225, + "step": 7466 + }, + { + "epoch": 0.24, + "learning_rate": 1.779973621087919e-05, + "loss": 1.1309, + "step": 7467 + }, + { + "epoch": 0.24, + "learning_rate": 1.779908795024109e-05, + "loss": 1.0767, + "step": 7468 + }, + { + "epoch": 0.24, + "learning_rate": 1.779843960592741e-05, + "loss": 1.0801, + "step": 7469 + }, + { + "epoch": 0.24, + "learning_rate": 1.77977911779451e-05, + "loss": 0.8955, + "step": 7470 + }, + { + "epoch": 0.24, + "learning_rate": 1.7797142666301125e-05, + "loss": 1.1543, + "step": 7471 + }, + { + "epoch": 0.24, + "learning_rate": 1.779649407100244e-05, + "loss": 0.9434, + "step": 7472 + }, + { + "epoch": 0.24, + "learning_rate": 1.7795845392055996e-05, + "loss": 1.0205, + "step": 7473 + }, + { + "epoch": 0.24, + "learning_rate": 1.7795196629468765e-05, + "loss": 1.1553, + "step": 7474 + }, + { + "epoch": 0.24, + "learning_rate": 1.77945477832477e-05, + "loss": 1.064, + "step": 7475 + }, + { + "epoch": 0.24, + "learning_rate": 1.7793898853399762e-05, + "loss": 1.2036, + "step": 7476 + }, + { + "epoch": 0.24, + "learning_rate": 1.7793249839931913e-05, + "loss": 1.1606, + "step": 7477 + }, + { + "epoch": 0.24, + "learning_rate": 1.7792600742851123e-05, + "loss": 1.1968, + "step": 7478 + }, + { + "epoch": 0.24, + "learning_rate": 1.779195156216435e-05, + "loss": 0.9932, + "step": 7479 + }, + { + "epoch": 0.24, + "learning_rate": 1.7791302297878562e-05, + "loss": 1.0161, + "step": 7480 + }, + { + "epoch": 0.24, + "learning_rate": 1.7790652950000723e-05, + "loss": 1.0967, + "step": 7481 + }, + { + "epoch": 0.24, + "learning_rate": 1.7790003518537794e-05, + "loss": 1.062, + "step": 7482 + }, + { + "epoch": 0.24, + "learning_rate": 1.7789354003496756e-05, + "loss": 1.0405, + "step": 7483 + }, + { + "epoch": 0.24, + "learning_rate": 1.7788704404884567e-05, + "loss": 1.1646, + "step": 7484 + }, + { + "epoch": 0.24, + "learning_rate": 1.77880547227082e-05, + "loss": 0.5132, + "step": 7485 + }, + { + "epoch": 0.24, + "learning_rate": 1.7787404956974628e-05, + "loss": 1.1597, + "step": 7486 + }, + { + "epoch": 0.24, + "learning_rate": 1.7786755107690818e-05, + "loss": 1.1245, + "step": 7487 + }, + { + "epoch": 0.24, + "learning_rate": 1.778610517486374e-05, + "loss": 1.0854, + "step": 7488 + }, + { + "epoch": 0.24, + "learning_rate": 1.7785455158500374e-05, + "loss": 1.0239, + "step": 7489 + }, + { + "epoch": 0.24, + "learning_rate": 1.778480505860769e-05, + "loss": 1.0337, + "step": 7490 + }, + { + "epoch": 0.24, + "learning_rate": 1.7784154875192665e-05, + "loss": 1.0322, + "step": 7491 + }, + { + "epoch": 0.24, + "learning_rate": 1.7783504608262272e-05, + "loss": 1.1431, + "step": 7492 + }, + { + "epoch": 0.24, + "learning_rate": 1.778285425782349e-05, + "loss": 1.0503, + "step": 7493 + }, + { + "epoch": 0.24, + "learning_rate": 1.778220382388329e-05, + "loss": 1.1729, + "step": 7494 + }, + { + "epoch": 0.24, + "learning_rate": 1.7781553306448664e-05, + "loss": 1.0225, + "step": 7495 + }, + { + "epoch": 0.24, + "learning_rate": 1.778090270552658e-05, + "loss": 1.1011, + "step": 7496 + }, + { + "epoch": 0.24, + "learning_rate": 1.778025202112402e-05, + "loss": 0.9956, + "step": 7497 + }, + { + "epoch": 0.24, + "learning_rate": 1.7779601253247968e-05, + "loss": 1.1396, + "step": 7498 + }, + { + "epoch": 0.24, + "learning_rate": 1.7778950401905404e-05, + "loss": 1.0825, + "step": 7499 + }, + { + "epoch": 0.24, + "learning_rate": 1.777829946710331e-05, + "loss": 1.0254, + "step": 7500 + }, + { + "epoch": 0.24, + "learning_rate": 1.7777648448848673e-05, + "loss": 0.5334, + "step": 7501 + }, + { + "epoch": 0.24, + "learning_rate": 1.777699734714848e-05, + "loss": 0.9175, + "step": 7502 + }, + { + "epoch": 0.24, + "learning_rate": 1.777634616200971e-05, + "loss": 1.064, + "step": 7503 + }, + { + "epoch": 0.24, + "learning_rate": 1.7775694893439348e-05, + "loss": 1.0327, + "step": 7504 + }, + { + "epoch": 0.24, + "learning_rate": 1.7775043541444388e-05, + "loss": 1.0781, + "step": 7505 + }, + { + "epoch": 0.24, + "learning_rate": 1.7774392106031817e-05, + "loss": 1.0938, + "step": 7506 + }, + { + "epoch": 0.24, + "learning_rate": 1.777374058720862e-05, + "loss": 1.0898, + "step": 7507 + }, + { + "epoch": 0.24, + "learning_rate": 1.777308898498179e-05, + "loss": 1.03, + "step": 7508 + }, + { + "epoch": 0.24, + "learning_rate": 1.7772437299358324e-05, + "loss": 0.9424, + "step": 7509 + }, + { + "epoch": 0.24, + "learning_rate": 1.77717855303452e-05, + "loss": 1.0933, + "step": 7510 + }, + { + "epoch": 0.24, + "learning_rate": 1.7771133677949427e-05, + "loss": 1.0806, + "step": 7511 + }, + { + "epoch": 0.24, + "learning_rate": 1.7770481742177987e-05, + "loss": 1.0059, + "step": 7512 + }, + { + "epoch": 0.24, + "learning_rate": 1.7769829723037878e-05, + "loss": 0.9644, + "step": 7513 + }, + { + "epoch": 0.24, + "learning_rate": 1.7769177620536093e-05, + "loss": 1.0586, + "step": 7514 + }, + { + "epoch": 0.24, + "learning_rate": 1.7768525434679632e-05, + "loss": 1.0996, + "step": 7515 + }, + { + "epoch": 0.24, + "learning_rate": 1.776787316547549e-05, + "loss": 1.106, + "step": 7516 + }, + { + "epoch": 0.24, + "learning_rate": 1.7767220812930665e-05, + "loss": 1.1084, + "step": 7517 + }, + { + "epoch": 0.24, + "learning_rate": 1.776656837705216e-05, + "loss": 1.1396, + "step": 7518 + }, + { + "epoch": 0.24, + "learning_rate": 1.7765915857846973e-05, + "loss": 1.1646, + "step": 7519 + }, + { + "epoch": 0.24, + "learning_rate": 1.77652632553221e-05, + "loss": 1.1499, + "step": 7520 + }, + { + "epoch": 0.24, + "learning_rate": 1.7764610569484547e-05, + "loss": 1.1367, + "step": 7521 + }, + { + "epoch": 0.24, + "learning_rate": 1.7763957800341317e-05, + "loss": 1.106, + "step": 7522 + }, + { + "epoch": 0.24, + "learning_rate": 1.776330494789941e-05, + "loss": 1.0288, + "step": 7523 + }, + { + "epoch": 0.24, + "learning_rate": 1.7762652012165837e-05, + "loss": 1.0034, + "step": 7524 + }, + { + "epoch": 0.24, + "learning_rate": 1.77619989931476e-05, + "loss": 1.0576, + "step": 7525 + }, + { + "epoch": 0.24, + "learning_rate": 1.77613458908517e-05, + "loss": 1.043, + "step": 7526 + }, + { + "epoch": 0.24, + "learning_rate": 1.776069270528515e-05, + "loss": 0.5564, + "step": 7527 + }, + { + "epoch": 0.24, + "learning_rate": 1.7760039436454955e-05, + "loss": 0.9009, + "step": 7528 + }, + { + "epoch": 0.24, + "learning_rate": 1.7759386084368127e-05, + "loss": 1.0366, + "step": 7529 + }, + { + "epoch": 0.24, + "learning_rate": 1.7758732649031675e-05, + "loss": 0.9492, + "step": 7530 + }, + { + "epoch": 0.24, + "learning_rate": 1.7758079130452606e-05, + "loss": 1.1416, + "step": 7531 + }, + { + "epoch": 0.24, + "learning_rate": 1.7757425528637933e-05, + "loss": 1.0996, + "step": 7532 + }, + { + "epoch": 0.24, + "learning_rate": 1.7756771843594674e-05, + "loss": 1.186, + "step": 7533 + }, + { + "epoch": 0.24, + "learning_rate": 1.7756118075329835e-05, + "loss": 1.105, + "step": 7534 + }, + { + "epoch": 0.24, + "learning_rate": 1.7755464223850436e-05, + "loss": 1.123, + "step": 7535 + }, + { + "epoch": 0.24, + "learning_rate": 1.7754810289163488e-05, + "loss": 1.1196, + "step": 7536 + }, + { + "epoch": 0.24, + "learning_rate": 1.7754156271276006e-05, + "loss": 1.0942, + "step": 7537 + }, + { + "epoch": 0.24, + "learning_rate": 1.7753502170195012e-05, + "loss": 1.1284, + "step": 7538 + }, + { + "epoch": 0.24, + "learning_rate": 1.775284798592752e-05, + "loss": 1.021, + "step": 7539 + }, + { + "epoch": 0.24, + "learning_rate": 1.7752193718480546e-05, + "loss": 1.1001, + "step": 7540 + }, + { + "epoch": 0.24, + "learning_rate": 1.7751539367861118e-05, + "loss": 0.9604, + "step": 7541 + }, + { + "epoch": 0.24, + "learning_rate": 1.7750884934076248e-05, + "loss": 1.0049, + "step": 7542 + }, + { + "epoch": 0.24, + "learning_rate": 1.7750230417132963e-05, + "loss": 1.0425, + "step": 7543 + }, + { + "epoch": 0.24, + "learning_rate": 1.7749575817038283e-05, + "loss": 1.0469, + "step": 7544 + }, + { + "epoch": 0.24, + "learning_rate": 1.774892113379923e-05, + "loss": 1.166, + "step": 7545 + }, + { + "epoch": 0.24, + "learning_rate": 1.774826636742283e-05, + "loss": 1.0874, + "step": 7546 + }, + { + "epoch": 0.24, + "learning_rate": 1.7747611517916106e-05, + "loss": 1.0342, + "step": 7547 + }, + { + "epoch": 0.24, + "learning_rate": 1.774695658528609e-05, + "loss": 1.0371, + "step": 7548 + }, + { + "epoch": 0.24, + "learning_rate": 1.77463015695398e-05, + "loss": 1.1587, + "step": 7549 + }, + { + "epoch": 0.24, + "learning_rate": 1.7745646470684266e-05, + "loss": 1.0122, + "step": 7550 + }, + { + "epoch": 0.24, + "learning_rate": 1.774499128872652e-05, + "loss": 1.0337, + "step": 7551 + }, + { + "epoch": 0.24, + "learning_rate": 1.774433602367359e-05, + "loss": 1.0254, + "step": 7552 + }, + { + "epoch": 0.24, + "learning_rate": 1.7743680675532502e-05, + "loss": 1.105, + "step": 7553 + }, + { + "epoch": 0.24, + "learning_rate": 1.7743025244310293e-05, + "loss": 0.9858, + "step": 7554 + }, + { + "epoch": 0.24, + "learning_rate": 1.7742369730013994e-05, + "loss": 0.9937, + "step": 7555 + }, + { + "epoch": 0.24, + "learning_rate": 1.7741714132650636e-05, + "loss": 1.0327, + "step": 7556 + }, + { + "epoch": 0.24, + "learning_rate": 1.7741058452227252e-05, + "loss": 1.0884, + "step": 7557 + }, + { + "epoch": 0.24, + "learning_rate": 1.774040268875088e-05, + "loss": 1.2339, + "step": 7558 + }, + { + "epoch": 0.24, + "learning_rate": 1.773974684222855e-05, + "loss": 1.0552, + "step": 7559 + }, + { + "epoch": 0.24, + "learning_rate": 1.7739090912667305e-05, + "loss": 0.9165, + "step": 7560 + }, + { + "epoch": 0.24, + "learning_rate": 1.7738434900074183e-05, + "loss": 1.04, + "step": 7561 + }, + { + "epoch": 0.24, + "learning_rate": 1.7737778804456214e-05, + "loss": 1.0581, + "step": 7562 + }, + { + "epoch": 0.24, + "learning_rate": 1.7737122625820444e-05, + "loss": 1.0625, + "step": 7563 + }, + { + "epoch": 0.24, + "learning_rate": 1.773646636417391e-05, + "loss": 1.0229, + "step": 7564 + }, + { + "epoch": 0.24, + "learning_rate": 1.7735810019523657e-05, + "loss": 1.1265, + "step": 7565 + }, + { + "epoch": 0.24, + "learning_rate": 1.773515359187672e-05, + "loss": 1.1558, + "step": 7566 + }, + { + "epoch": 0.24, + "learning_rate": 1.773449708124015e-05, + "loss": 1.1245, + "step": 7567 + }, + { + "epoch": 0.24, + "learning_rate": 1.773384048762098e-05, + "loss": 1.1206, + "step": 7568 + }, + { + "epoch": 0.24, + "learning_rate": 1.7733183811026266e-05, + "loss": 1.0088, + "step": 7569 + }, + { + "epoch": 0.24, + "learning_rate": 1.773252705146305e-05, + "loss": 1.1387, + "step": 7570 + }, + { + "epoch": 0.24, + "learning_rate": 1.7731870208938376e-05, + "loss": 1.0239, + "step": 7571 + }, + { + "epoch": 0.24, + "learning_rate": 1.773121328345929e-05, + "loss": 0.9224, + "step": 7572 + }, + { + "epoch": 0.24, + "learning_rate": 1.773055627503284e-05, + "loss": 0.998, + "step": 7573 + }, + { + "epoch": 0.24, + "learning_rate": 1.7729899183666083e-05, + "loss": 1.0254, + "step": 7574 + }, + { + "epoch": 0.24, + "learning_rate": 1.7729242009366058e-05, + "loss": 1.0591, + "step": 7575 + }, + { + "epoch": 0.24, + "learning_rate": 1.772858475213982e-05, + "loss": 1.0264, + "step": 7576 + }, + { + "epoch": 0.24, + "learning_rate": 1.7727927411994424e-05, + "loss": 1.0913, + "step": 7577 + }, + { + "epoch": 0.24, + "learning_rate": 1.772726998893692e-05, + "loss": 1.0669, + "step": 7578 + }, + { + "epoch": 0.24, + "learning_rate": 1.7726612482974358e-05, + "loss": 0.9849, + "step": 7579 + }, + { + "epoch": 0.24, + "learning_rate": 1.7725954894113793e-05, + "loss": 1.0791, + "step": 7580 + }, + { + "epoch": 0.24, + "learning_rate": 1.772529722236229e-05, + "loss": 1.0376, + "step": 7581 + }, + { + "epoch": 0.24, + "learning_rate": 1.7724639467726894e-05, + "loss": 1.1431, + "step": 7582 + }, + { + "epoch": 0.24, + "learning_rate": 1.7723981630214667e-05, + "loss": 1.0278, + "step": 7583 + }, + { + "epoch": 0.24, + "learning_rate": 1.7723323709832664e-05, + "loss": 1.0835, + "step": 7584 + }, + { + "epoch": 0.24, + "learning_rate": 1.7722665706587946e-05, + "loss": 0.9702, + "step": 7585 + }, + { + "epoch": 0.24, + "learning_rate": 1.7722007620487575e-05, + "loss": 1.0928, + "step": 7586 + }, + { + "epoch": 0.24, + "learning_rate": 1.7721349451538604e-05, + "loss": 1.0171, + "step": 7587 + }, + { + "epoch": 0.24, + "learning_rate": 1.77206911997481e-05, + "loss": 1.2178, + "step": 7588 + }, + { + "epoch": 0.24, + "learning_rate": 1.7720032865123125e-05, + "loss": 1.0615, + "step": 7589 + }, + { + "epoch": 0.24, + "learning_rate": 1.771937444767074e-05, + "loss": 1.0732, + "step": 7590 + }, + { + "epoch": 0.24, + "learning_rate": 1.7718715947398015e-05, + "loss": 1.1899, + "step": 7591 + }, + { + "epoch": 0.24, + "learning_rate": 1.7718057364312006e-05, + "loss": 0.9531, + "step": 7592 + }, + { + "epoch": 0.24, + "learning_rate": 1.7717398698419782e-05, + "loss": 0.9678, + "step": 7593 + }, + { + "epoch": 0.24, + "learning_rate": 1.7716739949728417e-05, + "loss": 1.0098, + "step": 7594 + }, + { + "epoch": 0.24, + "learning_rate": 1.771608111824497e-05, + "loss": 1.0503, + "step": 7595 + }, + { + "epoch": 0.24, + "learning_rate": 1.771542220397651e-05, + "loss": 1.0503, + "step": 7596 + }, + { + "epoch": 0.24, + "learning_rate": 1.7714763206930112e-05, + "loss": 1.1006, + "step": 7597 + }, + { + "epoch": 0.24, + "learning_rate": 1.771410412711284e-05, + "loss": 0.9116, + "step": 7598 + }, + { + "epoch": 0.24, + "learning_rate": 1.771344496453177e-05, + "loss": 0.9741, + "step": 7599 + }, + { + "epoch": 0.24, + "learning_rate": 1.7712785719193974e-05, + "loss": 1.0425, + "step": 7600 + }, + { + "epoch": 0.24, + "learning_rate": 1.7712126391106524e-05, + "loss": 1.0825, + "step": 7601 + }, + { + "epoch": 0.24, + "learning_rate": 1.771146698027649e-05, + "loss": 0.9932, + "step": 7602 + }, + { + "epoch": 0.24, + "learning_rate": 1.771080748671095e-05, + "loss": 1.0103, + "step": 7603 + }, + { + "epoch": 0.24, + "learning_rate": 1.7710147910416982e-05, + "loss": 1.0947, + "step": 7604 + }, + { + "epoch": 0.24, + "learning_rate": 1.770948825140166e-05, + "loss": 1.0601, + "step": 7605 + }, + { + "epoch": 0.24, + "learning_rate": 1.770882850967206e-05, + "loss": 1.0654, + "step": 7606 + }, + { + "epoch": 0.24, + "learning_rate": 1.770816868523526e-05, + "loss": 1.1016, + "step": 7607 + }, + { + "epoch": 0.24, + "learning_rate": 1.7707508778098344e-05, + "loss": 1.0303, + "step": 7608 + }, + { + "epoch": 0.24, + "learning_rate": 1.7706848788268386e-05, + "loss": 1.1401, + "step": 7609 + }, + { + "epoch": 0.24, + "learning_rate": 1.7706188715752473e-05, + "loss": 1.1221, + "step": 7610 + }, + { + "epoch": 0.24, + "learning_rate": 1.7705528560557682e-05, + "loss": 0.9888, + "step": 7611 + }, + { + "epoch": 0.24, + "learning_rate": 1.77048683226911e-05, + "loss": 1.1143, + "step": 7612 + }, + { + "epoch": 0.24, + "learning_rate": 1.7704208002159803e-05, + "loss": 0.9985, + "step": 7613 + }, + { + "epoch": 0.24, + "learning_rate": 1.7703547598970888e-05, + "loss": 0.8945, + "step": 7614 + }, + { + "epoch": 0.24, + "learning_rate": 1.770288711313143e-05, + "loss": 1.0723, + "step": 7615 + }, + { + "epoch": 0.24, + "learning_rate": 1.7702226544648517e-05, + "loss": 1.1304, + "step": 7616 + }, + { + "epoch": 0.24, + "learning_rate": 1.7701565893529236e-05, + "loss": 0.9224, + "step": 7617 + }, + { + "epoch": 0.24, + "learning_rate": 1.7700905159780682e-05, + "loss": 1.1357, + "step": 7618 + }, + { + "epoch": 0.24, + "learning_rate": 1.7700244343409933e-05, + "loss": 0.895, + "step": 7619 + }, + { + "epoch": 0.24, + "learning_rate": 1.769958344442409e-05, + "loss": 1.1465, + "step": 7620 + }, + { + "epoch": 0.24, + "learning_rate": 1.7698922462830235e-05, + "loss": 1.1284, + "step": 7621 + }, + { + "epoch": 0.24, + "learning_rate": 1.7698261398635463e-05, + "loss": 0.9956, + "step": 7622 + }, + { + "epoch": 0.24, + "learning_rate": 1.7697600251846868e-05, + "loss": 0.9321, + "step": 7623 + }, + { + "epoch": 0.24, + "learning_rate": 1.7696939022471537e-05, + "loss": 1.0542, + "step": 7624 + }, + { + "epoch": 0.24, + "learning_rate": 1.769627771051657e-05, + "loss": 1.127, + "step": 7625 + }, + { + "epoch": 0.24, + "learning_rate": 1.7695616315989064e-05, + "loss": 1.0562, + "step": 7626 + }, + { + "epoch": 0.24, + "learning_rate": 1.769495483889611e-05, + "loss": 0.9941, + "step": 7627 + }, + { + "epoch": 0.24, + "learning_rate": 1.7694293279244806e-05, + "loss": 1.0327, + "step": 7628 + }, + { + "epoch": 0.24, + "learning_rate": 1.7693631637042256e-05, + "loss": 1.0332, + "step": 7629 + }, + { + "epoch": 0.24, + "learning_rate": 1.769296991229555e-05, + "loss": 1.0796, + "step": 7630 + }, + { + "epoch": 0.24, + "learning_rate": 1.769230810501179e-05, + "loss": 1.1855, + "step": 7631 + }, + { + "epoch": 0.24, + "learning_rate": 1.7691646215198076e-05, + "loss": 1.0308, + "step": 7632 + }, + { + "epoch": 0.24, + "learning_rate": 1.7690984242861514e-05, + "loss": 1.0244, + "step": 7633 + }, + { + "epoch": 0.24, + "learning_rate": 1.7690322188009204e-05, + "loss": 1.1011, + "step": 7634 + }, + { + "epoch": 0.24, + "learning_rate": 1.7689660050648245e-05, + "loss": 1.0366, + "step": 7635 + }, + { + "epoch": 0.24, + "learning_rate": 1.7688997830785742e-05, + "loss": 1.0957, + "step": 7636 + }, + { + "epoch": 0.24, + "learning_rate": 1.768833552842881e-05, + "loss": 0.957, + "step": 7637 + }, + { + "epoch": 0.24, + "learning_rate": 1.7687673143584542e-05, + "loss": 1.0044, + "step": 7638 + }, + { + "epoch": 0.24, + "learning_rate": 1.768701067626005e-05, + "loss": 0.9087, + "step": 7639 + }, + { + "epoch": 0.24, + "learning_rate": 1.7686348126462437e-05, + "loss": 1.0874, + "step": 7640 + }, + { + "epoch": 0.24, + "learning_rate": 1.768568549419882e-05, + "loss": 1.1177, + "step": 7641 + }, + { + "epoch": 0.24, + "learning_rate": 1.7685022779476304e-05, + "loss": 1.2295, + "step": 7642 + }, + { + "epoch": 0.24, + "learning_rate": 1.7684359982302e-05, + "loss": 1.0254, + "step": 7643 + }, + { + "epoch": 0.24, + "learning_rate": 1.7683697102683012e-05, + "loss": 1.0718, + "step": 7644 + }, + { + "epoch": 0.24, + "learning_rate": 1.7683034140626463e-05, + "loss": 1.1709, + "step": 7645 + }, + { + "epoch": 0.24, + "learning_rate": 1.768237109613946e-05, + "loss": 1.0552, + "step": 7646 + }, + { + "epoch": 0.24, + "learning_rate": 1.7681707969229115e-05, + "loss": 1.1245, + "step": 7647 + }, + { + "epoch": 0.24, + "learning_rate": 1.7681044759902547e-05, + "loss": 1.0127, + "step": 7648 + }, + { + "epoch": 0.24, + "learning_rate": 1.7680381468166873e-05, + "loss": 1.104, + "step": 7649 + }, + { + "epoch": 0.24, + "learning_rate": 1.76797180940292e-05, + "loss": 0.979, + "step": 7650 + }, + { + "epoch": 0.24, + "learning_rate": 1.7679054637496656e-05, + "loss": 1.1406, + "step": 7651 + }, + { + "epoch": 0.24, + "learning_rate": 1.767839109857635e-05, + "loss": 1.1704, + "step": 7652 + }, + { + "epoch": 0.24, + "learning_rate": 1.7677727477275414e-05, + "loss": 1.1577, + "step": 7653 + }, + { + "epoch": 0.24, + "learning_rate": 1.767706377360095e-05, + "loss": 1.0562, + "step": 7654 + }, + { + "epoch": 0.24, + "learning_rate": 1.7676399987560094e-05, + "loss": 1.0112, + "step": 7655 + }, + { + "epoch": 0.24, + "learning_rate": 1.7675736119159957e-05, + "loss": 1.0996, + "step": 7656 + }, + { + "epoch": 0.24, + "learning_rate": 1.7675072168407673e-05, + "loss": 1.1152, + "step": 7657 + }, + { + "epoch": 0.24, + "learning_rate": 1.7674408135310356e-05, + "loss": 1.0986, + "step": 7658 + }, + { + "epoch": 0.24, + "learning_rate": 1.767374401987513e-05, + "loss": 1.0962, + "step": 7659 + }, + { + "epoch": 0.24, + "learning_rate": 1.767307982210913e-05, + "loss": 1.1108, + "step": 7660 + }, + { + "epoch": 0.25, + "learning_rate": 1.767241554201947e-05, + "loss": 1.0605, + "step": 7661 + }, + { + "epoch": 0.25, + "learning_rate": 1.7671751179613284e-05, + "loss": 0.5273, + "step": 7662 + }, + { + "epoch": 0.25, + "learning_rate": 1.7671086734897704e-05, + "loss": 1.0361, + "step": 7663 + }, + { + "epoch": 0.25, + "learning_rate": 1.7670422207879845e-05, + "loss": 0.8809, + "step": 7664 + }, + { + "epoch": 0.25, + "learning_rate": 1.7669757598566848e-05, + "loss": 0.9897, + "step": 7665 + }, + { + "epoch": 0.25, + "learning_rate": 1.7669092906965842e-05, + "loss": 1.0527, + "step": 7666 + }, + { + "epoch": 0.25, + "learning_rate": 1.7668428133083955e-05, + "loss": 1.0518, + "step": 7667 + }, + { + "epoch": 0.25, + "learning_rate": 1.766776327692832e-05, + "loss": 0.9292, + "step": 7668 + }, + { + "epoch": 0.25, + "learning_rate": 1.7667098338506075e-05, + "loss": 1.0005, + "step": 7669 + }, + { + "epoch": 0.25, + "learning_rate": 1.7666433317824347e-05, + "loss": 1.084, + "step": 7670 + }, + { + "epoch": 0.25, + "learning_rate": 1.766576821489028e-05, + "loss": 1.0845, + "step": 7671 + }, + { + "epoch": 0.25, + "learning_rate": 1.7665103029710998e-05, + "loss": 0.9106, + "step": 7672 + }, + { + "epoch": 0.25, + "learning_rate": 1.7664437762293645e-05, + "loss": 1.0811, + "step": 7673 + }, + { + "epoch": 0.25, + "learning_rate": 1.766377241264536e-05, + "loss": 0.5793, + "step": 7674 + }, + { + "epoch": 0.25, + "learning_rate": 1.7663106980773273e-05, + "loss": 1.1704, + "step": 7675 + }, + { + "epoch": 0.25, + "learning_rate": 1.7662441466684533e-05, + "loss": 1.1084, + "step": 7676 + }, + { + "epoch": 0.25, + "learning_rate": 1.766177587038628e-05, + "loss": 0.9341, + "step": 7677 + }, + { + "epoch": 0.25, + "learning_rate": 1.7661110191885648e-05, + "loss": 1.0884, + "step": 7678 + }, + { + "epoch": 0.25, + "learning_rate": 1.766044443118978e-05, + "loss": 1.0645, + "step": 7679 + }, + { + "epoch": 0.25, + "learning_rate": 1.765977858830583e-05, + "loss": 1.0054, + "step": 7680 + }, + { + "epoch": 0.25, + "learning_rate": 1.7659112663240924e-05, + "loss": 1.0645, + "step": 7681 + }, + { + "epoch": 0.25, + "learning_rate": 1.7658446656002216e-05, + "loss": 1.0503, + "step": 7682 + }, + { + "epoch": 0.25, + "learning_rate": 1.7657780566596856e-05, + "loss": 1.0449, + "step": 7683 + }, + { + "epoch": 0.25, + "learning_rate": 1.7657114395031985e-05, + "loss": 1.0078, + "step": 7684 + }, + { + "epoch": 0.25, + "learning_rate": 1.765644814131475e-05, + "loss": 1.1621, + "step": 7685 + }, + { + "epoch": 0.25, + "learning_rate": 1.7655781805452302e-05, + "loss": 1.2393, + "step": 7686 + }, + { + "epoch": 0.25, + "learning_rate": 1.7655115387451783e-05, + "loss": 1.002, + "step": 7687 + }, + { + "epoch": 0.25, + "learning_rate": 1.765444888732035e-05, + "loss": 1.0791, + "step": 7688 + }, + { + "epoch": 0.25, + "learning_rate": 1.7653782305065158e-05, + "loss": 0.9756, + "step": 7689 + }, + { + "epoch": 0.25, + "learning_rate": 1.765311564069335e-05, + "loss": 1.0156, + "step": 7690 + }, + { + "epoch": 0.25, + "learning_rate": 1.7652448894212078e-05, + "loss": 1.0767, + "step": 7691 + }, + { + "epoch": 0.25, + "learning_rate": 1.76517820656285e-05, + "loss": 1.0269, + "step": 7692 + }, + { + "epoch": 0.25, + "learning_rate": 1.765111515494977e-05, + "loss": 1.0469, + "step": 7693 + }, + { + "epoch": 0.25, + "learning_rate": 1.765044816218304e-05, + "loss": 1.0278, + "step": 7694 + }, + { + "epoch": 0.25, + "learning_rate": 1.764978108733547e-05, + "loss": 1.1826, + "step": 7695 + }, + { + "epoch": 0.25, + "learning_rate": 1.7649113930414216e-05, + "loss": 1.0308, + "step": 7696 + }, + { + "epoch": 0.25, + "learning_rate": 1.7648446691426433e-05, + "loss": 0.6152, + "step": 7697 + }, + { + "epoch": 0.25, + "learning_rate": 1.7647779370379285e-05, + "loss": 1.0859, + "step": 7698 + }, + { + "epoch": 0.25, + "learning_rate": 1.7647111967279926e-05, + "loss": 1.0552, + "step": 7699 + }, + { + "epoch": 0.25, + "learning_rate": 1.764644448213552e-05, + "loss": 1.0825, + "step": 7700 + }, + { + "epoch": 0.25, + "learning_rate": 1.7645776914953227e-05, + "loss": 1.0977, + "step": 7701 + }, + { + "epoch": 0.25, + "learning_rate": 1.764510926574021e-05, + "loss": 0.9946, + "step": 7702 + }, + { + "epoch": 0.25, + "learning_rate": 1.764444153450363e-05, + "loss": 1.0273, + "step": 7703 + }, + { + "epoch": 0.25, + "learning_rate": 1.7643773721250653e-05, + "loss": 1.1499, + "step": 7704 + }, + { + "epoch": 0.25, + "learning_rate": 1.7643105825988445e-05, + "loss": 1.0747, + "step": 7705 + }, + { + "epoch": 0.25, + "learning_rate": 1.764243784872417e-05, + "loss": 1.0356, + "step": 7706 + }, + { + "epoch": 0.25, + "learning_rate": 1.7641769789464996e-05, + "loss": 1.0181, + "step": 7707 + }, + { + "epoch": 0.25, + "learning_rate": 1.7641101648218088e-05, + "loss": 1.0405, + "step": 7708 + }, + { + "epoch": 0.25, + "learning_rate": 1.7640433424990618e-05, + "loss": 1.1079, + "step": 7709 + }, + { + "epoch": 0.25, + "learning_rate": 1.7639765119789755e-05, + "loss": 1.1426, + "step": 7710 + }, + { + "epoch": 0.25, + "learning_rate": 1.7639096732622666e-05, + "loss": 1.0884, + "step": 7711 + }, + { + "epoch": 0.25, + "learning_rate": 1.7638428263496517e-05, + "loss": 1.0273, + "step": 7712 + }, + { + "epoch": 0.25, + "learning_rate": 1.7637759712418497e-05, + "loss": 1.1089, + "step": 7713 + }, + { + "epoch": 0.25, + "learning_rate": 1.7637091079395762e-05, + "loss": 0.959, + "step": 7714 + }, + { + "epoch": 0.25, + "learning_rate": 1.763642236443549e-05, + "loss": 1.1587, + "step": 7715 + }, + { + "epoch": 0.25, + "learning_rate": 1.7635753567544865e-05, + "loss": 1.1304, + "step": 7716 + }, + { + "epoch": 0.25, + "learning_rate": 1.763508468873105e-05, + "loss": 1.0728, + "step": 7717 + }, + { + "epoch": 0.25, + "learning_rate": 1.763441572800123e-05, + "loss": 1.0845, + "step": 7718 + }, + { + "epoch": 0.25, + "learning_rate": 1.7633746685362573e-05, + "loss": 1.0513, + "step": 7719 + }, + { + "epoch": 0.25, + "learning_rate": 1.7633077560822265e-05, + "loss": 0.9844, + "step": 7720 + }, + { + "epoch": 0.25, + "learning_rate": 1.763240835438749e-05, + "loss": 1.0317, + "step": 7721 + }, + { + "epoch": 0.25, + "learning_rate": 1.7631739066065408e-05, + "loss": 1.0244, + "step": 7722 + }, + { + "epoch": 0.25, + "learning_rate": 1.763106969586322e-05, + "loss": 1.0493, + "step": 7723 + }, + { + "epoch": 0.25, + "learning_rate": 1.7630400243788096e-05, + "loss": 1.1475, + "step": 7724 + }, + { + "epoch": 0.25, + "learning_rate": 1.7629730709847227e-05, + "loss": 1.1802, + "step": 7725 + }, + { + "epoch": 0.25, + "learning_rate": 1.762906109404779e-05, + "loss": 1.1646, + "step": 7726 + }, + { + "epoch": 0.25, + "learning_rate": 1.762839139639697e-05, + "loss": 1.0527, + "step": 7727 + }, + { + "epoch": 0.25, + "learning_rate": 1.762772161690195e-05, + "loss": 1.0586, + "step": 7728 + }, + { + "epoch": 0.25, + "learning_rate": 1.762705175556992e-05, + "loss": 1.1948, + "step": 7729 + }, + { + "epoch": 0.25, + "learning_rate": 1.7626381812408073e-05, + "loss": 1.0518, + "step": 7730 + }, + { + "epoch": 0.25, + "learning_rate": 1.7625711787423583e-05, + "loss": 0.999, + "step": 7731 + }, + { + "epoch": 0.25, + "learning_rate": 1.7625041680623645e-05, + "loss": 1.1025, + "step": 7732 + }, + { + "epoch": 0.25, + "learning_rate": 1.762437149201545e-05, + "loss": 0.9883, + "step": 7733 + }, + { + "epoch": 0.25, + "learning_rate": 1.762370122160619e-05, + "loss": 1.0098, + "step": 7734 + }, + { + "epoch": 0.25, + "learning_rate": 1.762303086940305e-05, + "loss": 1.0615, + "step": 7735 + }, + { + "epoch": 0.25, + "learning_rate": 1.7622360435413224e-05, + "loss": 1.188, + "step": 7736 + }, + { + "epoch": 0.25, + "learning_rate": 1.762168991964391e-05, + "loss": 1.0767, + "step": 7737 + }, + { + "epoch": 0.25, + "learning_rate": 1.7621019322102298e-05, + "loss": 0.9824, + "step": 7738 + }, + { + "epoch": 0.25, + "learning_rate": 1.7620348642795583e-05, + "loss": 1.0259, + "step": 7739 + }, + { + "epoch": 0.25, + "learning_rate": 1.7619677881730958e-05, + "loss": 1.0859, + "step": 7740 + }, + { + "epoch": 0.25, + "learning_rate": 1.7619007038915625e-05, + "loss": 0.9946, + "step": 7741 + }, + { + "epoch": 0.25, + "learning_rate": 1.761833611435678e-05, + "loss": 1.0278, + "step": 7742 + }, + { + "epoch": 0.25, + "learning_rate": 1.7617665108061617e-05, + "loss": 0.9985, + "step": 7743 + }, + { + "epoch": 0.25, + "learning_rate": 1.761699402003734e-05, + "loss": 1.0923, + "step": 7744 + }, + { + "epoch": 0.25, + "learning_rate": 1.7616322850291147e-05, + "loss": 1.0674, + "step": 7745 + }, + { + "epoch": 0.25, + "learning_rate": 1.7615651598830238e-05, + "loss": 1.125, + "step": 7746 + }, + { + "epoch": 0.25, + "learning_rate": 1.761498026566182e-05, + "loss": 1.0986, + "step": 7747 + }, + { + "epoch": 0.25, + "learning_rate": 1.7614308850793086e-05, + "loss": 1.0405, + "step": 7748 + }, + { + "epoch": 0.25, + "learning_rate": 1.761363735423125e-05, + "loss": 0.9419, + "step": 7749 + }, + { + "epoch": 0.25, + "learning_rate": 1.761296577598351e-05, + "loss": 1.0996, + "step": 7750 + }, + { + "epoch": 0.25, + "learning_rate": 1.761229411605707e-05, + "loss": 1.0732, + "step": 7751 + }, + { + "epoch": 0.25, + "learning_rate": 1.761162237445914e-05, + "loss": 1.1187, + "step": 7752 + }, + { + "epoch": 0.25, + "learning_rate": 1.761095055119693e-05, + "loss": 1.3525, + "step": 7753 + }, + { + "epoch": 0.25, + "learning_rate": 1.761027864627764e-05, + "loss": 1.125, + "step": 7754 + }, + { + "epoch": 0.25, + "learning_rate": 1.7609606659708482e-05, + "loss": 1.0884, + "step": 7755 + }, + { + "epoch": 0.25, + "learning_rate": 1.7608934591496673e-05, + "loss": 1.0923, + "step": 7756 + }, + { + "epoch": 0.25, + "learning_rate": 1.760826244164941e-05, + "loss": 1.0278, + "step": 7757 + }, + { + "epoch": 0.25, + "learning_rate": 1.7607590210173912e-05, + "loss": 1.0698, + "step": 7758 + }, + { + "epoch": 0.25, + "learning_rate": 1.7606917897077394e-05, + "loss": 1.1445, + "step": 7759 + }, + { + "epoch": 0.25, + "learning_rate": 1.7606245502367063e-05, + "loss": 1.0708, + "step": 7760 + }, + { + "epoch": 0.25, + "learning_rate": 1.7605573026050138e-05, + "loss": 1.0298, + "step": 7761 + }, + { + "epoch": 0.25, + "learning_rate": 1.760490046813383e-05, + "loss": 1.0059, + "step": 7762 + }, + { + "epoch": 0.25, + "learning_rate": 1.760422782862536e-05, + "loss": 1.0601, + "step": 7763 + }, + { + "epoch": 0.25, + "learning_rate": 1.760355510753194e-05, + "loss": 1.1191, + "step": 7764 + }, + { + "epoch": 0.25, + "learning_rate": 1.7602882304860786e-05, + "loss": 1.0005, + "step": 7765 + }, + { + "epoch": 0.25, + "learning_rate": 1.760220942061912e-05, + "loss": 0.9673, + "step": 7766 + }, + { + "epoch": 0.25, + "learning_rate": 1.7601536454814164e-05, + "loss": 1.1953, + "step": 7767 + }, + { + "epoch": 0.25, + "learning_rate": 1.7600863407453132e-05, + "loss": 1.0967, + "step": 7768 + }, + { + "epoch": 0.25, + "learning_rate": 1.760019027854325e-05, + "loss": 1.124, + "step": 7769 + }, + { + "epoch": 0.25, + "learning_rate": 1.7599517068091735e-05, + "loss": 0.9514, + "step": 7770 + }, + { + "epoch": 0.25, + "learning_rate": 1.7598843776105816e-05, + "loss": 1.1494, + "step": 7771 + }, + { + "epoch": 0.25, + "learning_rate": 1.759817040259271e-05, + "loss": 0.9946, + "step": 7772 + }, + { + "epoch": 0.25, + "learning_rate": 1.7597496947559645e-05, + "loss": 1.0874, + "step": 7773 + }, + { + "epoch": 0.25, + "learning_rate": 1.7596823411013847e-05, + "loss": 1.1631, + "step": 7774 + }, + { + "epoch": 0.25, + "learning_rate": 1.7596149792962545e-05, + "loss": 0.936, + "step": 7775 + }, + { + "epoch": 0.25, + "learning_rate": 1.7595476093412957e-05, + "loss": 1.1406, + "step": 7776 + }, + { + "epoch": 0.25, + "learning_rate": 1.7594802312372322e-05, + "loss": 1.1343, + "step": 7777 + }, + { + "epoch": 0.25, + "learning_rate": 1.759412844984786e-05, + "loss": 1.0815, + "step": 7778 + }, + { + "epoch": 0.25, + "learning_rate": 1.7593454505846807e-05, + "loss": 0.96, + "step": 7779 + }, + { + "epoch": 0.25, + "learning_rate": 1.7592780480376387e-05, + "loss": 1.1787, + "step": 7780 + }, + { + "epoch": 0.25, + "learning_rate": 1.759210637344384e-05, + "loss": 1.1562, + "step": 7781 + }, + { + "epoch": 0.25, + "learning_rate": 1.759143218505639e-05, + "loss": 0.9849, + "step": 7782 + }, + { + "epoch": 0.25, + "learning_rate": 1.7590757915221278e-05, + "loss": 0.9336, + "step": 7783 + }, + { + "epoch": 0.25, + "learning_rate": 1.7590083563945734e-05, + "loss": 1.1128, + "step": 7784 + }, + { + "epoch": 0.25, + "learning_rate": 1.7589409131236994e-05, + "loss": 0.9399, + "step": 7785 + }, + { + "epoch": 0.25, + "learning_rate": 1.7588734617102292e-05, + "loss": 0.9893, + "step": 7786 + }, + { + "epoch": 0.25, + "learning_rate": 1.758806002154887e-05, + "loss": 1.0654, + "step": 7787 + }, + { + "epoch": 0.25, + "learning_rate": 1.758738534458396e-05, + "loss": 0.5789, + "step": 7788 + }, + { + "epoch": 0.25, + "learning_rate": 1.75867105862148e-05, + "loss": 1.084, + "step": 7789 + }, + { + "epoch": 0.25, + "learning_rate": 1.758603574644863e-05, + "loss": 1.2422, + "step": 7790 + }, + { + "epoch": 0.25, + "learning_rate": 1.75853608252927e-05, + "loss": 1.1621, + "step": 7791 + }, + { + "epoch": 0.25, + "learning_rate": 1.758468582275424e-05, + "loss": 0.9951, + "step": 7792 + }, + { + "epoch": 0.25, + "learning_rate": 1.7584010738840494e-05, + "loss": 0.9253, + "step": 7793 + }, + { + "epoch": 0.25, + "learning_rate": 1.7583335573558705e-05, + "loss": 1.0713, + "step": 7794 + }, + { + "epoch": 0.25, + "learning_rate": 1.758266032691612e-05, + "loss": 1.2871, + "step": 7795 + }, + { + "epoch": 0.25, + "learning_rate": 1.7581984998919983e-05, + "loss": 1.0645, + "step": 7796 + }, + { + "epoch": 0.25, + "learning_rate": 1.758130958957754e-05, + "loss": 0.9131, + "step": 7797 + }, + { + "epoch": 0.25, + "learning_rate": 1.7580634098896032e-05, + "loss": 1.1382, + "step": 7798 + }, + { + "epoch": 0.25, + "learning_rate": 1.757995852688271e-05, + "loss": 0.9917, + "step": 7799 + }, + { + "epoch": 0.25, + "learning_rate": 1.757928287354482e-05, + "loss": 1.208, + "step": 7800 + }, + { + "epoch": 0.25, + "learning_rate": 1.757860713888962e-05, + "loss": 1.0249, + "step": 7801 + }, + { + "epoch": 0.25, + "learning_rate": 1.757793132292435e-05, + "loss": 1.1572, + "step": 7802 + }, + { + "epoch": 0.25, + "learning_rate": 1.757725542565626e-05, + "loss": 1.104, + "step": 7803 + }, + { + "epoch": 0.25, + "learning_rate": 1.757657944709261e-05, + "loss": 1.0396, + "step": 7804 + }, + { + "epoch": 0.25, + "learning_rate": 1.7575903387240645e-05, + "loss": 0.957, + "step": 7805 + }, + { + "epoch": 0.25, + "learning_rate": 1.7575227246107623e-05, + "loss": 1.0591, + "step": 7806 + }, + { + "epoch": 0.25, + "learning_rate": 1.7574551023700797e-05, + "loss": 1.0806, + "step": 7807 + }, + { + "epoch": 0.25, + "learning_rate": 1.757387472002742e-05, + "loss": 1.1719, + "step": 7808 + }, + { + "epoch": 0.25, + "learning_rate": 1.7573198335094753e-05, + "loss": 1.0474, + "step": 7809 + }, + { + "epoch": 0.25, + "learning_rate": 1.7572521868910045e-05, + "loss": 0.9897, + "step": 7810 + }, + { + "epoch": 0.25, + "learning_rate": 1.757184532148056e-05, + "loss": 0.9648, + "step": 7811 + }, + { + "epoch": 0.25, + "learning_rate": 1.7571168692813555e-05, + "loss": 0.9346, + "step": 7812 + }, + { + "epoch": 0.25, + "learning_rate": 1.757049198291629e-05, + "loss": 1.0254, + "step": 7813 + }, + { + "epoch": 0.25, + "learning_rate": 1.7569815191796022e-05, + "loss": 1.1074, + "step": 7814 + }, + { + "epoch": 0.25, + "learning_rate": 1.7569138319460016e-05, + "loss": 0.9849, + "step": 7815 + }, + { + "epoch": 0.25, + "learning_rate": 1.7568461365915532e-05, + "loss": 0.9014, + "step": 7816 + }, + { + "epoch": 0.25, + "learning_rate": 1.7567784331169836e-05, + "loss": 1.0869, + "step": 7817 + }, + { + "epoch": 0.25, + "learning_rate": 1.7567107215230187e-05, + "loss": 1.0088, + "step": 7818 + }, + { + "epoch": 0.25, + "learning_rate": 1.7566430018103856e-05, + "loss": 1.0649, + "step": 7819 + }, + { + "epoch": 0.25, + "learning_rate": 1.7565752739798104e-05, + "loss": 1.0615, + "step": 7820 + }, + { + "epoch": 0.25, + "learning_rate": 1.75650753803202e-05, + "loss": 1.0962, + "step": 7821 + }, + { + "epoch": 0.25, + "learning_rate": 1.7564397939677405e-05, + "loss": 1.1196, + "step": 7822 + }, + { + "epoch": 0.25, + "learning_rate": 1.7563720417876993e-05, + "loss": 1.0801, + "step": 7823 + }, + { + "epoch": 0.25, + "learning_rate": 1.7563042814926237e-05, + "loss": 1.1392, + "step": 7824 + }, + { + "epoch": 0.25, + "learning_rate": 1.7562365130832398e-05, + "loss": 1.0635, + "step": 7825 + }, + { + "epoch": 0.25, + "learning_rate": 1.756168736560275e-05, + "loss": 0.998, + "step": 7826 + }, + { + "epoch": 0.25, + "learning_rate": 1.7561009519244568e-05, + "loss": 0.5132, + "step": 7827 + }, + { + "epoch": 0.25, + "learning_rate": 1.756033159176512e-05, + "loss": 0.9692, + "step": 7828 + }, + { + "epoch": 0.25, + "learning_rate": 1.7559653583171683e-05, + "loss": 1.0605, + "step": 7829 + }, + { + "epoch": 0.25, + "learning_rate": 1.755897549347153e-05, + "loss": 0.9868, + "step": 7830 + }, + { + "epoch": 0.25, + "learning_rate": 1.7558297322671932e-05, + "loss": 0.9993, + "step": 7831 + }, + { + "epoch": 0.25, + "learning_rate": 1.7557619070780172e-05, + "loss": 1.1714, + "step": 7832 + }, + { + "epoch": 0.25, + "learning_rate": 1.7556940737803523e-05, + "loss": 1.0625, + "step": 7833 + }, + { + "epoch": 0.25, + "learning_rate": 1.7556262323749263e-05, + "loss": 1.1211, + "step": 7834 + }, + { + "epoch": 0.25, + "learning_rate": 1.7555583828624674e-05, + "loss": 1.0298, + "step": 7835 + }, + { + "epoch": 0.25, + "learning_rate": 1.755490525243703e-05, + "loss": 1.0615, + "step": 7836 + }, + { + "epoch": 0.25, + "learning_rate": 1.7554226595193613e-05, + "loss": 1.186, + "step": 7837 + }, + { + "epoch": 0.25, + "learning_rate": 1.7553547856901707e-05, + "loss": 1.1304, + "step": 7838 + }, + { + "epoch": 0.25, + "learning_rate": 1.7552869037568592e-05, + "loss": 1.1982, + "step": 7839 + }, + { + "epoch": 0.25, + "learning_rate": 1.7552190137201554e-05, + "loss": 1.0991, + "step": 7840 + }, + { + "epoch": 0.25, + "learning_rate": 1.755151115580787e-05, + "loss": 1.0757, + "step": 7841 + }, + { + "epoch": 0.25, + "learning_rate": 1.755083209339483e-05, + "loss": 1.0825, + "step": 7842 + }, + { + "epoch": 0.25, + "learning_rate": 1.7550152949969716e-05, + "loss": 1.0801, + "step": 7843 + }, + { + "epoch": 0.25, + "learning_rate": 1.754947372553982e-05, + "loss": 1.0215, + "step": 7844 + }, + { + "epoch": 0.25, + "learning_rate": 1.754879442011243e-05, + "loss": 1.0381, + "step": 7845 + }, + { + "epoch": 0.25, + "learning_rate": 1.7548115033694825e-05, + "loss": 1.167, + "step": 7846 + }, + { + "epoch": 0.25, + "learning_rate": 1.7547435566294302e-05, + "loss": 1.0913, + "step": 7847 + }, + { + "epoch": 0.25, + "learning_rate": 1.7546756017918144e-05, + "loss": 1.0352, + "step": 7848 + }, + { + "epoch": 0.25, + "learning_rate": 1.7546076388573653e-05, + "loss": 1.0503, + "step": 7849 + }, + { + "epoch": 0.25, + "learning_rate": 1.754539667826811e-05, + "loss": 1.0459, + "step": 7850 + }, + { + "epoch": 0.25, + "learning_rate": 1.754471688700881e-05, + "loss": 1.147, + "step": 7851 + }, + { + "epoch": 0.25, + "learning_rate": 1.7544037014803054e-05, + "loss": 1.062, + "step": 7852 + }, + { + "epoch": 0.25, + "learning_rate": 1.7543357061658125e-05, + "loss": 1.0493, + "step": 7853 + }, + { + "epoch": 0.25, + "learning_rate": 1.754267702758132e-05, + "loss": 0.9775, + "step": 7854 + }, + { + "epoch": 0.25, + "learning_rate": 1.754199691257995e-05, + "loss": 1.1128, + "step": 7855 + }, + { + "epoch": 0.25, + "learning_rate": 1.754131671666129e-05, + "loss": 1.1143, + "step": 7856 + }, + { + "epoch": 0.25, + "learning_rate": 1.754063643983265e-05, + "loss": 1.0596, + "step": 7857 + }, + { + "epoch": 0.25, + "learning_rate": 1.753995608210133e-05, + "loss": 1.0518, + "step": 7858 + }, + { + "epoch": 0.25, + "learning_rate": 1.753927564347462e-05, + "loss": 1.1787, + "step": 7859 + }, + { + "epoch": 0.25, + "learning_rate": 1.753859512395983e-05, + "loss": 0.9507, + "step": 7860 + }, + { + "epoch": 0.25, + "learning_rate": 1.753791452356426e-05, + "loss": 1.0376, + "step": 7861 + }, + { + "epoch": 0.25, + "learning_rate": 1.7537233842295206e-05, + "loss": 1.1763, + "step": 7862 + }, + { + "epoch": 0.25, + "learning_rate": 1.7536553080159978e-05, + "loss": 1.0332, + "step": 7863 + }, + { + "epoch": 0.25, + "learning_rate": 1.7535872237165874e-05, + "loss": 1.0625, + "step": 7864 + }, + { + "epoch": 0.25, + "learning_rate": 1.75351913133202e-05, + "loss": 1.1479, + "step": 7865 + }, + { + "epoch": 0.25, + "learning_rate": 1.7534510308630265e-05, + "loss": 1.0586, + "step": 7866 + }, + { + "epoch": 0.25, + "learning_rate": 1.7533829223103373e-05, + "loss": 1.0444, + "step": 7867 + }, + { + "epoch": 0.25, + "learning_rate": 1.753314805674683e-05, + "loss": 1.2139, + "step": 7868 + }, + { + "epoch": 0.25, + "learning_rate": 1.753246680956795e-05, + "loss": 1.1084, + "step": 7869 + }, + { + "epoch": 0.25, + "learning_rate": 1.7531785481574032e-05, + "loss": 1.1006, + "step": 7870 + }, + { + "epoch": 0.25, + "learning_rate": 1.7531104072772395e-05, + "loss": 1.1812, + "step": 7871 + }, + { + "epoch": 0.25, + "learning_rate": 1.7530422583170344e-05, + "loss": 0.9976, + "step": 7872 + }, + { + "epoch": 0.25, + "learning_rate": 1.7529741012775193e-05, + "loss": 0.8887, + "step": 7873 + }, + { + "epoch": 0.25, + "learning_rate": 1.7529059361594258e-05, + "loss": 1.0332, + "step": 7874 + }, + { + "epoch": 0.25, + "learning_rate": 1.7528377629634847e-05, + "loss": 1.083, + "step": 7875 + }, + { + "epoch": 0.25, + "learning_rate": 1.7527695816904275e-05, + "loss": 0.9922, + "step": 7876 + }, + { + "epoch": 0.25, + "learning_rate": 1.7527013923409856e-05, + "loss": 1.0991, + "step": 7877 + }, + { + "epoch": 0.25, + "learning_rate": 1.7526331949158914e-05, + "loss": 1.0327, + "step": 7878 + }, + { + "epoch": 0.25, + "learning_rate": 1.7525649894158755e-05, + "loss": 0.9795, + "step": 7879 + }, + { + "epoch": 0.25, + "learning_rate": 1.7524967758416705e-05, + "loss": 0.9307, + "step": 7880 + }, + { + "epoch": 0.25, + "learning_rate": 1.7524285541940075e-05, + "loss": 1.0908, + "step": 7881 + }, + { + "epoch": 0.25, + "learning_rate": 1.752360324473619e-05, + "loss": 1.1514, + "step": 7882 + }, + { + "epoch": 0.25, + "learning_rate": 1.7522920866812372e-05, + "loss": 1.0552, + "step": 7883 + }, + { + "epoch": 0.25, + "learning_rate": 1.7522238408175938e-05, + "loss": 1.0532, + "step": 7884 + }, + { + "epoch": 0.25, + "learning_rate": 1.752155586883421e-05, + "loss": 0.9678, + "step": 7885 + }, + { + "epoch": 0.25, + "learning_rate": 1.752087324879451e-05, + "loss": 1.0996, + "step": 7886 + }, + { + "epoch": 0.25, + "learning_rate": 1.7520190548064166e-05, + "loss": 1.0557, + "step": 7887 + }, + { + "epoch": 0.25, + "learning_rate": 1.75195077666505e-05, + "loss": 1.1187, + "step": 7888 + }, + { + "epoch": 0.25, + "learning_rate": 1.7518824904560838e-05, + "loss": 1.0664, + "step": 7889 + }, + { + "epoch": 0.25, + "learning_rate": 1.7518141961802506e-05, + "loss": 1.2446, + "step": 7890 + }, + { + "epoch": 0.25, + "learning_rate": 1.7517458938382833e-05, + "loss": 1.0605, + "step": 7891 + }, + { + "epoch": 0.25, + "learning_rate": 1.7516775834309143e-05, + "loss": 1.1201, + "step": 7892 + }, + { + "epoch": 0.25, + "learning_rate": 1.751609264958877e-05, + "loss": 1.0381, + "step": 7893 + }, + { + "epoch": 0.25, + "learning_rate": 1.751540938422904e-05, + "loss": 1.1294, + "step": 7894 + }, + { + "epoch": 0.25, + "learning_rate": 1.751472603823728e-05, + "loss": 1.1221, + "step": 7895 + }, + { + "epoch": 0.25, + "learning_rate": 1.7514042611620835e-05, + "loss": 1.1455, + "step": 7896 + }, + { + "epoch": 0.25, + "learning_rate": 1.7513359104387024e-05, + "loss": 1.103, + "step": 7897 + }, + { + "epoch": 0.25, + "learning_rate": 1.7512675516543186e-05, + "loss": 1.0591, + "step": 7898 + }, + { + "epoch": 0.25, + "learning_rate": 1.7511991848096658e-05, + "loss": 1.0024, + "step": 7899 + }, + { + "epoch": 0.25, + "learning_rate": 1.7511308099054765e-05, + "loss": 1.0635, + "step": 7900 + }, + { + "epoch": 0.25, + "learning_rate": 1.7510624269424853e-05, + "loss": 1.019, + "step": 7901 + }, + { + "epoch": 0.25, + "learning_rate": 1.7509940359214255e-05, + "loss": 1.0371, + "step": 7902 + }, + { + "epoch": 0.25, + "learning_rate": 1.7509256368430304e-05, + "loss": 1.0469, + "step": 7903 + }, + { + "epoch": 0.25, + "learning_rate": 1.7508572297080347e-05, + "loss": 1.2305, + "step": 7904 + }, + { + "epoch": 0.25, + "learning_rate": 1.750788814517172e-05, + "loss": 1.1992, + "step": 7905 + }, + { + "epoch": 0.25, + "learning_rate": 1.7507203912711762e-05, + "loss": 1.0869, + "step": 7906 + }, + { + "epoch": 0.25, + "learning_rate": 1.7506519599707812e-05, + "loss": 1.0771, + "step": 7907 + }, + { + "epoch": 0.25, + "learning_rate": 1.750583520616722e-05, + "loss": 1.1743, + "step": 7908 + }, + { + "epoch": 0.25, + "learning_rate": 1.7505150732097318e-05, + "loss": 1.248, + "step": 7909 + }, + { + "epoch": 0.25, + "learning_rate": 1.7504466177505456e-05, + "loss": 1.1274, + "step": 7910 + }, + { + "epoch": 0.25, + "learning_rate": 1.750378154239898e-05, + "loss": 1.0083, + "step": 7911 + }, + { + "epoch": 0.25, + "learning_rate": 1.750309682678523e-05, + "loss": 1.1113, + "step": 7912 + }, + { + "epoch": 0.25, + "learning_rate": 1.7502412030671558e-05, + "loss": 0.9531, + "step": 7913 + }, + { + "epoch": 0.25, + "learning_rate": 1.7501727154065304e-05, + "loss": 1.0649, + "step": 7914 + }, + { + "epoch": 0.25, + "learning_rate": 1.7501042196973824e-05, + "loss": 1.0425, + "step": 7915 + }, + { + "epoch": 0.25, + "learning_rate": 1.750035715940446e-05, + "loss": 1.0015, + "step": 7916 + }, + { + "epoch": 0.25, + "learning_rate": 1.7499672041364567e-05, + "loss": 0.9834, + "step": 7917 + }, + { + "epoch": 0.25, + "learning_rate": 1.749898684286149e-05, + "loss": 1.0049, + "step": 7918 + }, + { + "epoch": 0.25, + "learning_rate": 1.749830156390259e-05, + "loss": 1.0913, + "step": 7919 + }, + { + "epoch": 0.25, + "learning_rate": 1.7497616204495208e-05, + "loss": 0.5608, + "step": 7920 + }, + { + "epoch": 0.25, + "learning_rate": 1.74969307646467e-05, + "loss": 1.0723, + "step": 7921 + }, + { + "epoch": 0.25, + "learning_rate": 1.7496245244364428e-05, + "loss": 1.1802, + "step": 7922 + }, + { + "epoch": 0.25, + "learning_rate": 1.7495559643655736e-05, + "loss": 1.0425, + "step": 7923 + }, + { + "epoch": 0.25, + "learning_rate": 1.7494873962527984e-05, + "loss": 1.0117, + "step": 7924 + }, + { + "epoch": 0.25, + "learning_rate": 1.7494188200988535e-05, + "loss": 0.9609, + "step": 7925 + }, + { + "epoch": 0.25, + "learning_rate": 1.7493502359044737e-05, + "loss": 1.2139, + "step": 7926 + }, + { + "epoch": 0.25, + "learning_rate": 1.7492816436703953e-05, + "loss": 1.1289, + "step": 7927 + }, + { + "epoch": 0.25, + "learning_rate": 1.7492130433973543e-05, + "loss": 0.9976, + "step": 7928 + }, + { + "epoch": 0.25, + "learning_rate": 1.749144435086086e-05, + "loss": 1.0278, + "step": 7929 + }, + { + "epoch": 0.25, + "learning_rate": 1.7490758187373277e-05, + "loss": 1.0352, + "step": 7930 + }, + { + "epoch": 0.25, + "learning_rate": 1.7490071943518143e-05, + "loss": 1.0342, + "step": 7931 + }, + { + "epoch": 0.25, + "learning_rate": 1.748938561930283e-05, + "loss": 1.1182, + "step": 7932 + }, + { + "epoch": 0.25, + "learning_rate": 1.74886992147347e-05, + "loss": 1.0195, + "step": 7933 + }, + { + "epoch": 0.25, + "learning_rate": 1.748801272982111e-05, + "loss": 1.084, + "step": 7934 + }, + { + "epoch": 0.25, + "learning_rate": 1.7487326164569434e-05, + "loss": 0.9907, + "step": 7935 + }, + { + "epoch": 0.25, + "learning_rate": 1.7486639518987034e-05, + "loss": 1.0537, + "step": 7936 + }, + { + "epoch": 0.25, + "learning_rate": 1.748595279308128e-05, + "loss": 1.0381, + "step": 7937 + }, + { + "epoch": 0.25, + "learning_rate": 1.7485265986859534e-05, + "loss": 1.1157, + "step": 7938 + }, + { + "epoch": 0.25, + "learning_rate": 1.748457910032917e-05, + "loss": 1.002, + "step": 7939 + }, + { + "epoch": 0.25, + "learning_rate": 1.7483892133497558e-05, + "loss": 1.0547, + "step": 7940 + }, + { + "epoch": 0.25, + "learning_rate": 1.7483205086372064e-05, + "loss": 0.9683, + "step": 7941 + }, + { + "epoch": 0.25, + "learning_rate": 1.7482517958960058e-05, + "loss": 1.0854, + "step": 7942 + }, + { + "epoch": 0.25, + "learning_rate": 1.748183075126892e-05, + "loss": 0.9468, + "step": 7943 + }, + { + "epoch": 0.25, + "learning_rate": 1.7481143463306018e-05, + "loss": 1.1079, + "step": 7944 + }, + { + "epoch": 0.25, + "learning_rate": 1.7480456095078726e-05, + "loss": 1.0186, + "step": 7945 + }, + { + "epoch": 0.25, + "learning_rate": 1.747976864659442e-05, + "loss": 1.063, + "step": 7946 + }, + { + "epoch": 0.25, + "learning_rate": 1.7479081117860475e-05, + "loss": 1.0557, + "step": 7947 + }, + { + "epoch": 0.25, + "learning_rate": 1.7478393508884265e-05, + "loss": 1.0229, + "step": 7948 + }, + { + "epoch": 0.25, + "learning_rate": 1.7477705819673167e-05, + "loss": 1.0752, + "step": 7949 + }, + { + "epoch": 0.25, + "learning_rate": 1.7477018050234568e-05, + "loss": 1.2041, + "step": 7950 + }, + { + "epoch": 0.25, + "learning_rate": 1.7476330200575836e-05, + "loss": 1.0908, + "step": 7951 + }, + { + "epoch": 0.25, + "learning_rate": 1.7475642270704355e-05, + "loss": 1.0337, + "step": 7952 + }, + { + "epoch": 0.25, + "learning_rate": 1.7474954260627507e-05, + "loss": 1.0776, + "step": 7953 + }, + { + "epoch": 0.25, + "learning_rate": 1.7474266170352674e-05, + "loss": 0.9727, + "step": 7954 + }, + { + "epoch": 0.25, + "learning_rate": 1.7473577999887234e-05, + "loss": 1.2915, + "step": 7955 + }, + { + "epoch": 0.25, + "learning_rate": 1.7472889749238575e-05, + "loss": 1.1714, + "step": 7956 + }, + { + "epoch": 0.25, + "learning_rate": 1.747220141841408e-05, + "loss": 1.0684, + "step": 7957 + }, + { + "epoch": 0.25, + "learning_rate": 1.7471513007421136e-05, + "loss": 1.0562, + "step": 7958 + }, + { + "epoch": 0.25, + "learning_rate": 1.7470824516267125e-05, + "loss": 1.0845, + "step": 7959 + }, + { + "epoch": 0.25, + "learning_rate": 1.7470135944959432e-05, + "loss": 1.019, + "step": 7960 + }, + { + "epoch": 0.25, + "learning_rate": 1.7469447293505452e-05, + "loss": 1.1143, + "step": 7961 + }, + { + "epoch": 0.25, + "learning_rate": 1.7468758561912566e-05, + "loss": 1.1294, + "step": 7962 + }, + { + "epoch": 0.25, + "learning_rate": 1.746806975018817e-05, + "loss": 1.0649, + "step": 7963 + }, + { + "epoch": 0.25, + "learning_rate": 1.7467380858339648e-05, + "loss": 1.0542, + "step": 7964 + }, + { + "epoch": 0.25, + "learning_rate": 1.7466691886374394e-05, + "loss": 1.1094, + "step": 7965 + }, + { + "epoch": 0.25, + "learning_rate": 1.74660028342998e-05, + "loss": 1.165, + "step": 7966 + }, + { + "epoch": 0.25, + "learning_rate": 1.7465313702123258e-05, + "loss": 1.0977, + "step": 7967 + }, + { + "epoch": 0.25, + "learning_rate": 1.7464624489852165e-05, + "loss": 1.0552, + "step": 7968 + }, + { + "epoch": 0.25, + "learning_rate": 1.746393519749391e-05, + "loss": 1.167, + "step": 7969 + }, + { + "epoch": 0.25, + "learning_rate": 1.7463245825055894e-05, + "loss": 1.147, + "step": 7970 + }, + { + "epoch": 0.25, + "learning_rate": 1.7462556372545508e-05, + "loss": 1.0303, + "step": 7971 + }, + { + "epoch": 0.25, + "learning_rate": 1.746186683997015e-05, + "loss": 1.0537, + "step": 7972 + }, + { + "epoch": 0.25, + "learning_rate": 1.746117722733722e-05, + "loss": 1.083, + "step": 7973 + }, + { + "epoch": 0.26, + "learning_rate": 1.746048753465412e-05, + "loss": 0.9683, + "step": 7974 + }, + { + "epoch": 0.26, + "learning_rate": 1.745979776192824e-05, + "loss": 1.0264, + "step": 7975 + }, + { + "epoch": 0.26, + "learning_rate": 1.7459107909166988e-05, + "loss": 0.897, + "step": 7976 + }, + { + "epoch": 0.26, + "learning_rate": 1.7458417976377764e-05, + "loss": 1.064, + "step": 7977 + }, + { + "epoch": 0.26, + "learning_rate": 1.7457727963567968e-05, + "loss": 1.1792, + "step": 7978 + }, + { + "epoch": 0.26, + "learning_rate": 1.7457037870745008e-05, + "loss": 0.9941, + "step": 7979 + }, + { + "epoch": 0.26, + "learning_rate": 1.7456347697916283e-05, + "loss": 1.0381, + "step": 7980 + }, + { + "epoch": 0.26, + "learning_rate": 1.74556574450892e-05, + "loss": 1.1411, + "step": 7981 + }, + { + "epoch": 0.26, + "learning_rate": 1.7454967112271162e-05, + "loss": 1.1084, + "step": 7982 + }, + { + "epoch": 0.26, + "learning_rate": 1.745427669946958e-05, + "loss": 1.0908, + "step": 7983 + }, + { + "epoch": 0.26, + "learning_rate": 1.745358620669186e-05, + "loss": 0.9856, + "step": 7984 + }, + { + "epoch": 0.26, + "learning_rate": 1.7452895633945407e-05, + "loss": 1.0571, + "step": 7985 + }, + { + "epoch": 0.26, + "learning_rate": 1.745220498123763e-05, + "loss": 1.2485, + "step": 7986 + }, + { + "epoch": 0.26, + "learning_rate": 1.7451514248575946e-05, + "loss": 0.9858, + "step": 7987 + }, + { + "epoch": 0.26, + "learning_rate": 1.7450823435967758e-05, + "loss": 1.1123, + "step": 7988 + }, + { + "epoch": 0.26, + "learning_rate": 1.745013254342048e-05, + "loss": 1.0786, + "step": 7989 + }, + { + "epoch": 0.26, + "learning_rate": 1.7449441570941526e-05, + "loss": 1.0859, + "step": 7990 + }, + { + "epoch": 0.26, + "learning_rate": 1.744875051853831e-05, + "loss": 1.063, + "step": 7991 + }, + { + "epoch": 0.26, + "learning_rate": 1.7448059386218246e-05, + "loss": 0.9751, + "step": 7992 + }, + { + "epoch": 0.26, + "learning_rate": 1.7447368173988747e-05, + "loss": 1.0771, + "step": 7993 + }, + { + "epoch": 0.26, + "learning_rate": 1.744667688185723e-05, + "loss": 1.019, + "step": 7994 + }, + { + "epoch": 0.26, + "learning_rate": 1.744598550983111e-05, + "loss": 1.0688, + "step": 7995 + }, + { + "epoch": 0.26, + "learning_rate": 1.7445294057917806e-05, + "loss": 1.0679, + "step": 7996 + }, + { + "epoch": 0.26, + "learning_rate": 1.7444602526124738e-05, + "loss": 0.9912, + "step": 7997 + }, + { + "epoch": 0.26, + "learning_rate": 1.7443910914459326e-05, + "loss": 1.0391, + "step": 7998 + }, + { + "epoch": 0.26, + "learning_rate": 1.7443219222928983e-05, + "loss": 0.9458, + "step": 7999 + }, + { + "epoch": 0.26, + "learning_rate": 1.744252745154114e-05, + "loss": 1.1323, + "step": 8000 + }, + { + "epoch": 0.26, + "learning_rate": 1.7441835600303216e-05, + "loss": 1.0874, + "step": 8001 + }, + { + "epoch": 0.26, + "learning_rate": 1.7441143669222628e-05, + "loss": 1.0493, + "step": 8002 + }, + { + "epoch": 0.26, + "learning_rate": 1.7440451658306803e-05, + "loss": 1.0596, + "step": 8003 + }, + { + "epoch": 0.26, + "learning_rate": 1.743975956756317e-05, + "loss": 1.166, + "step": 8004 + }, + { + "epoch": 0.26, + "learning_rate": 1.743906739699915e-05, + "loss": 1.0444, + "step": 8005 + }, + { + "epoch": 0.26, + "learning_rate": 1.743837514662217e-05, + "loss": 1.082, + "step": 8006 + }, + { + "epoch": 0.26, + "learning_rate": 1.7437682816439654e-05, + "loss": 1.1816, + "step": 8007 + }, + { + "epoch": 0.26, + "learning_rate": 1.7436990406459034e-05, + "loss": 1.0664, + "step": 8008 + }, + { + "epoch": 0.26, + "learning_rate": 1.7436297916687742e-05, + "loss": 1.0332, + "step": 8009 + }, + { + "epoch": 0.26, + "learning_rate": 1.74356053471332e-05, + "loss": 1.0435, + "step": 8010 + }, + { + "epoch": 0.26, + "learning_rate": 1.7434912697802843e-05, + "loss": 1.0542, + "step": 8011 + }, + { + "epoch": 0.26, + "learning_rate": 1.74342199687041e-05, + "loss": 1.0811, + "step": 8012 + }, + { + "epoch": 0.26, + "learning_rate": 1.7433527159844406e-05, + "loss": 1.1704, + "step": 8013 + }, + { + "epoch": 0.26, + "learning_rate": 1.7432834271231195e-05, + "loss": 1.041, + "step": 8014 + }, + { + "epoch": 0.26, + "learning_rate": 1.7432141302871895e-05, + "loss": 0.5674, + "step": 8015 + }, + { + "epoch": 0.26, + "learning_rate": 1.7431448254773943e-05, + "loss": 1.1484, + "step": 8016 + }, + { + "epoch": 0.26, + "learning_rate": 1.743075512694478e-05, + "loss": 1.0176, + "step": 8017 + }, + { + "epoch": 0.26, + "learning_rate": 1.7430061919391836e-05, + "loss": 1.0137, + "step": 8018 + }, + { + "epoch": 0.26, + "learning_rate": 1.742936863212255e-05, + "loss": 1.0068, + "step": 8019 + }, + { + "epoch": 0.26, + "learning_rate": 1.742867526514436e-05, + "loss": 1.0996, + "step": 8020 + }, + { + "epoch": 0.26, + "learning_rate": 1.742798181846471e-05, + "loss": 1.2598, + "step": 8021 + }, + { + "epoch": 0.26, + "learning_rate": 1.7427288292091033e-05, + "loss": 1.1094, + "step": 8022 + }, + { + "epoch": 0.26, + "learning_rate": 1.7426594686030775e-05, + "loss": 1.0581, + "step": 8023 + }, + { + "epoch": 0.26, + "learning_rate": 1.7425901000291374e-05, + "loss": 1.106, + "step": 8024 + }, + { + "epoch": 0.26, + "learning_rate": 1.7425207234880278e-05, + "loss": 1.1162, + "step": 8025 + }, + { + "epoch": 0.26, + "learning_rate": 1.742451338980492e-05, + "loss": 1.1226, + "step": 8026 + }, + { + "epoch": 0.26, + "learning_rate": 1.7423819465072753e-05, + "loss": 0.9595, + "step": 8027 + }, + { + "epoch": 0.26, + "learning_rate": 1.7423125460691216e-05, + "loss": 1.2119, + "step": 8028 + }, + { + "epoch": 0.26, + "learning_rate": 1.7422431376667765e-05, + "loss": 1.0732, + "step": 8029 + }, + { + "epoch": 0.26, + "learning_rate": 1.7421737213009834e-05, + "loss": 1.0757, + "step": 8030 + }, + { + "epoch": 0.26, + "learning_rate": 1.742104296972488e-05, + "loss": 1.0371, + "step": 8031 + }, + { + "epoch": 0.26, + "learning_rate": 1.742034864682035e-05, + "loss": 1.002, + "step": 8032 + }, + { + "epoch": 0.26, + "learning_rate": 1.7419654244303685e-05, + "loss": 1.0107, + "step": 8033 + }, + { + "epoch": 0.26, + "learning_rate": 1.741895976218235e-05, + "loss": 1.0913, + "step": 8034 + }, + { + "epoch": 0.26, + "learning_rate": 1.7418265200463777e-05, + "loss": 1.1836, + "step": 8035 + }, + { + "epoch": 0.26, + "learning_rate": 1.7417570559155436e-05, + "loss": 1.0293, + "step": 8036 + }, + { + "epoch": 0.26, + "learning_rate": 1.741687583826477e-05, + "loss": 1.0444, + "step": 8037 + }, + { + "epoch": 0.26, + "learning_rate": 1.7416181037799237e-05, + "loss": 1.0073, + "step": 8038 + }, + { + "epoch": 0.26, + "learning_rate": 1.7415486157766288e-05, + "loss": 0.8613, + "step": 8039 + }, + { + "epoch": 0.26, + "learning_rate": 1.741479119817338e-05, + "loss": 1.0991, + "step": 8040 + }, + { + "epoch": 0.26, + "learning_rate": 1.7414096159027968e-05, + "loss": 1.0737, + "step": 8041 + }, + { + "epoch": 0.26, + "learning_rate": 1.741340104033751e-05, + "loss": 1.1289, + "step": 8042 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412705842109464e-05, + "loss": 1.0811, + "step": 8043 + }, + { + "epoch": 0.26, + "learning_rate": 1.7412010564351286e-05, + "loss": 0.9673, + "step": 8044 + }, + { + "epoch": 0.26, + "learning_rate": 1.741131520707044e-05, + "loss": 1.2627, + "step": 8045 + }, + { + "epoch": 0.26, + "learning_rate": 1.7410619770274385e-05, + "loss": 1.1328, + "step": 8046 + }, + { + "epoch": 0.26, + "learning_rate": 1.740992425397058e-05, + "loss": 1.0815, + "step": 8047 + }, + { + "epoch": 0.26, + "learning_rate": 1.740922865816649e-05, + "loss": 0.5601, + "step": 8048 + }, + { + "epoch": 0.26, + "learning_rate": 1.7408532982869573e-05, + "loss": 1.0986, + "step": 8049 + }, + { + "epoch": 0.26, + "learning_rate": 1.74078372280873e-05, + "loss": 1.0271, + "step": 8050 + }, + { + "epoch": 0.26, + "learning_rate": 1.7407141393827132e-05, + "loss": 1.0249, + "step": 8051 + }, + { + "epoch": 0.26, + "learning_rate": 1.7406445480096533e-05, + "loss": 1.1509, + "step": 8052 + }, + { + "epoch": 0.26, + "learning_rate": 1.7405749486902974e-05, + "loss": 1.1128, + "step": 8053 + }, + { + "epoch": 0.26, + "learning_rate": 1.740505341425392e-05, + "loss": 1.0845, + "step": 8054 + }, + { + "epoch": 0.26, + "learning_rate": 1.7404357262156834e-05, + "loss": 0.9976, + "step": 8055 + }, + { + "epoch": 0.26, + "learning_rate": 1.740366103061919e-05, + "loss": 1.0645, + "step": 8056 + }, + { + "epoch": 0.26, + "learning_rate": 1.740296471964846e-05, + "loss": 1.103, + "step": 8057 + }, + { + "epoch": 0.26, + "learning_rate": 1.740226832925211e-05, + "loss": 0.9736, + "step": 8058 + }, + { + "epoch": 0.26, + "learning_rate": 1.740157185943761e-05, + "loss": 1.2363, + "step": 8059 + }, + { + "epoch": 0.26, + "learning_rate": 1.740087531021244e-05, + "loss": 1.0898, + "step": 8060 + }, + { + "epoch": 0.26, + "learning_rate": 1.7400178681584065e-05, + "loss": 0.6033, + "step": 8061 + }, + { + "epoch": 0.26, + "learning_rate": 1.7399481973559967e-05, + "loss": 1.0674, + "step": 8062 + }, + { + "epoch": 0.26, + "learning_rate": 1.7398785186147615e-05, + "loss": 0.9863, + "step": 8063 + }, + { + "epoch": 0.26, + "learning_rate": 1.7398088319354487e-05, + "loss": 1.1304, + "step": 8064 + }, + { + "epoch": 0.26, + "learning_rate": 1.739739137318806e-05, + "loss": 1.0029, + "step": 8065 + }, + { + "epoch": 0.26, + "learning_rate": 1.7396694347655806e-05, + "loss": 1.1406, + "step": 8066 + }, + { + "epoch": 0.26, + "learning_rate": 1.7395997242765215e-05, + "loss": 1.1548, + "step": 8067 + }, + { + "epoch": 0.26, + "learning_rate": 1.7395300058523755e-05, + "loss": 1.1079, + "step": 8068 + }, + { + "epoch": 0.26, + "learning_rate": 1.739460279493891e-05, + "loss": 0.9932, + "step": 8069 + }, + { + "epoch": 0.26, + "learning_rate": 1.739390545201816e-05, + "loss": 0.967, + "step": 8070 + }, + { + "epoch": 0.26, + "learning_rate": 1.739320802976899e-05, + "loss": 1.1929, + "step": 8071 + }, + { + "epoch": 0.26, + "learning_rate": 1.739251052819888e-05, + "loss": 1.064, + "step": 8072 + }, + { + "epoch": 0.26, + "learning_rate": 1.739181294731531e-05, + "loss": 1.0034, + "step": 8073 + }, + { + "epoch": 0.26, + "learning_rate": 1.7391115287125773e-05, + "loss": 1.1211, + "step": 8074 + }, + { + "epoch": 0.26, + "learning_rate": 1.7390417547637746e-05, + "loss": 1.0864, + "step": 8075 + }, + { + "epoch": 0.26, + "learning_rate": 1.738971972885872e-05, + "loss": 1.0889, + "step": 8076 + }, + { + "epoch": 0.26, + "learning_rate": 1.738902183079618e-05, + "loss": 1.001, + "step": 8077 + }, + { + "epoch": 0.26, + "learning_rate": 1.738832385345761e-05, + "loss": 1.1177, + "step": 8078 + }, + { + "epoch": 0.26, + "learning_rate": 1.738762579685051e-05, + "loss": 1.022, + "step": 8079 + }, + { + "epoch": 0.26, + "learning_rate": 1.7386927660982352e-05, + "loss": 1.1299, + "step": 8080 + }, + { + "epoch": 0.26, + "learning_rate": 1.738622944586064e-05, + "loss": 0.9873, + "step": 8081 + }, + { + "epoch": 0.26, + "learning_rate": 1.738553115149286e-05, + "loss": 1.1143, + "step": 8082 + }, + { + "epoch": 0.26, + "learning_rate": 1.7384832777886504e-05, + "loss": 0.8428, + "step": 8083 + }, + { + "epoch": 0.26, + "learning_rate": 1.7384134325049068e-05, + "loss": 1.1909, + "step": 8084 + }, + { + "epoch": 0.26, + "learning_rate": 1.738343579298804e-05, + "loss": 1.189, + "step": 8085 + }, + { + "epoch": 0.26, + "learning_rate": 1.738273718171092e-05, + "loss": 1.1567, + "step": 8086 + }, + { + "epoch": 0.26, + "learning_rate": 1.73820384912252e-05, + "loss": 1.1074, + "step": 8087 + }, + { + "epoch": 0.26, + "learning_rate": 1.7381339721538374e-05, + "loss": 1.0981, + "step": 8088 + }, + { + "epoch": 0.26, + "learning_rate": 1.7380640872657947e-05, + "loss": 1.0132, + "step": 8089 + }, + { + "epoch": 0.26, + "learning_rate": 1.737994194459141e-05, + "loss": 0.5327, + "step": 8090 + }, + { + "epoch": 0.26, + "learning_rate": 1.7379242937346263e-05, + "loss": 1.1943, + "step": 8091 + }, + { + "epoch": 0.26, + "learning_rate": 1.7378543850930004e-05, + "loss": 1.1494, + "step": 8092 + }, + { + "epoch": 0.26, + "learning_rate": 1.737784468535014e-05, + "loss": 1.0659, + "step": 8093 + }, + { + "epoch": 0.26, + "learning_rate": 1.7377145440614165e-05, + "loss": 0.5762, + "step": 8094 + }, + { + "epoch": 0.26, + "learning_rate": 1.7376446116729588e-05, + "loss": 1.1943, + "step": 8095 + }, + { + "epoch": 0.26, + "learning_rate": 1.7375746713703906e-05, + "loss": 1.0659, + "step": 8096 + }, + { + "epoch": 0.26, + "learning_rate": 1.7375047231544622e-05, + "loss": 1.1157, + "step": 8097 + }, + { + "epoch": 0.26, + "learning_rate": 1.7374347670259247e-05, + "loss": 1.2798, + "step": 8098 + }, + { + "epoch": 0.26, + "learning_rate": 1.737364802985528e-05, + "loss": 1.0391, + "step": 8099 + }, + { + "epoch": 0.26, + "learning_rate": 1.737294831034023e-05, + "loss": 1.0542, + "step": 8100 + }, + { + "epoch": 0.26, + "learning_rate": 1.7372248511721607e-05, + "loss": 1.002, + "step": 8101 + }, + { + "epoch": 0.26, + "learning_rate": 1.7371548634006918e-05, + "loss": 1.1367, + "step": 8102 + }, + { + "epoch": 0.26, + "learning_rate": 1.7370848677203666e-05, + "loss": 0.5544, + "step": 8103 + }, + { + "epoch": 0.26, + "learning_rate": 1.737014864131937e-05, + "loss": 1.0234, + "step": 8104 + }, + { + "epoch": 0.26, + "learning_rate": 1.736944852636153e-05, + "loss": 0.9814, + "step": 8105 + }, + { + "epoch": 0.26, + "learning_rate": 1.736874833233767e-05, + "loss": 1.0688, + "step": 8106 + }, + { + "epoch": 0.26, + "learning_rate": 1.7368048059255293e-05, + "loss": 1.293, + "step": 8107 + }, + { + "epoch": 0.26, + "learning_rate": 1.7367347707121916e-05, + "loss": 1.0078, + "step": 8108 + }, + { + "epoch": 0.26, + "learning_rate": 1.736664727594505e-05, + "loss": 1.0596, + "step": 8109 + }, + { + "epoch": 0.26, + "learning_rate": 1.7365946765732213e-05, + "loss": 1.1362, + "step": 8110 + }, + { + "epoch": 0.26, + "learning_rate": 1.736524617649092e-05, + "loss": 1.0879, + "step": 8111 + }, + { + "epoch": 0.26, + "learning_rate": 1.7364545508228685e-05, + "loss": 0.5457, + "step": 8112 + }, + { + "epoch": 0.26, + "learning_rate": 1.736384476095303e-05, + "loss": 1.1538, + "step": 8113 + }, + { + "epoch": 0.26, + "learning_rate": 1.736314393467147e-05, + "loss": 1.1665, + "step": 8114 + }, + { + "epoch": 0.26, + "learning_rate": 1.7362443029391524e-05, + "loss": 1.0527, + "step": 8115 + }, + { + "epoch": 0.26, + "learning_rate": 1.736174204512071e-05, + "loss": 1.04, + "step": 8116 + }, + { + "epoch": 0.26, + "learning_rate": 1.7361040981866553e-05, + "loss": 1.0737, + "step": 8117 + }, + { + "epoch": 0.26, + "learning_rate": 1.7360339839636575e-05, + "loss": 1.1475, + "step": 8118 + }, + { + "epoch": 0.26, + "learning_rate": 1.7359638618438297e-05, + "loss": 0.9858, + "step": 8119 + }, + { + "epoch": 0.26, + "learning_rate": 1.7358937318279242e-05, + "loss": 1.0996, + "step": 8120 + }, + { + "epoch": 0.26, + "learning_rate": 1.7358235939166928e-05, + "loss": 1.0249, + "step": 8121 + }, + { + "epoch": 0.26, + "learning_rate": 1.7357534481108894e-05, + "loss": 1.0928, + "step": 8122 + }, + { + "epoch": 0.26, + "learning_rate": 1.735683294411265e-05, + "loss": 0.9551, + "step": 8123 + }, + { + "epoch": 0.26, + "learning_rate": 1.7356131328185737e-05, + "loss": 1.1143, + "step": 8124 + }, + { + "epoch": 0.26, + "learning_rate": 1.7355429633335675e-05, + "loss": 1.0181, + "step": 8125 + }, + { + "epoch": 0.26, + "learning_rate": 1.7354727859569993e-05, + "loss": 1.1187, + "step": 8126 + }, + { + "epoch": 0.26, + "learning_rate": 1.7354026006896222e-05, + "loss": 1.1387, + "step": 8127 + }, + { + "epoch": 0.26, + "learning_rate": 1.7353324075321887e-05, + "loss": 1.0146, + "step": 8128 + }, + { + "epoch": 0.26, + "learning_rate": 1.735262206485453e-05, + "loss": 1.0669, + "step": 8129 + }, + { + "epoch": 0.26, + "learning_rate": 1.735191997550167e-05, + "loss": 1.0576, + "step": 8130 + }, + { + "epoch": 0.26, + "learning_rate": 1.7351217807270843e-05, + "loss": 1.0498, + "step": 8131 + }, + { + "epoch": 0.26, + "learning_rate": 1.735051556016959e-05, + "loss": 1.0254, + "step": 8132 + }, + { + "epoch": 0.26, + "learning_rate": 1.734981323420544e-05, + "loss": 1.1226, + "step": 8133 + }, + { + "epoch": 0.26, + "learning_rate": 1.7349110829385928e-05, + "loss": 1.1025, + "step": 8134 + }, + { + "epoch": 0.26, + "learning_rate": 1.7348408345718588e-05, + "loss": 1.0571, + "step": 8135 + }, + { + "epoch": 0.26, + "learning_rate": 1.7347705783210963e-05, + "loss": 1.1421, + "step": 8136 + }, + { + "epoch": 0.26, + "learning_rate": 1.734700314187058e-05, + "loss": 0.9414, + "step": 8137 + }, + { + "epoch": 0.26, + "learning_rate": 1.7346300421704992e-05, + "loss": 1.0635, + "step": 8138 + }, + { + "epoch": 0.26, + "learning_rate": 1.7345597622721727e-05, + "loss": 1.2236, + "step": 8139 + }, + { + "epoch": 0.26, + "learning_rate": 1.734489474492833e-05, + "loss": 1.0884, + "step": 8140 + }, + { + "epoch": 0.26, + "learning_rate": 1.7344191788332342e-05, + "loss": 1.1729, + "step": 8141 + }, + { + "epoch": 0.26, + "learning_rate": 1.7343488752941303e-05, + "loss": 1.0356, + "step": 8142 + }, + { + "epoch": 0.26, + "learning_rate": 1.734278563876276e-05, + "loss": 1.083, + "step": 8143 + }, + { + "epoch": 0.26, + "learning_rate": 1.734208244580425e-05, + "loss": 1.1089, + "step": 8144 + }, + { + "epoch": 0.26, + "learning_rate": 1.7341379174073327e-05, + "loss": 1.0967, + "step": 8145 + }, + { + "epoch": 0.26, + "learning_rate": 1.7340675823577524e-05, + "loss": 1.0483, + "step": 8146 + }, + { + "epoch": 0.26, + "learning_rate": 1.7339972394324398e-05, + "loss": 1.1416, + "step": 8147 + }, + { + "epoch": 0.26, + "learning_rate": 1.7339268886321487e-05, + "loss": 0.5193, + "step": 8148 + }, + { + "epoch": 0.26, + "learning_rate": 1.7338565299576347e-05, + "loss": 1.2529, + "step": 8149 + }, + { + "epoch": 0.26, + "learning_rate": 1.7337861634096524e-05, + "loss": 0.9795, + "step": 8150 + }, + { + "epoch": 0.26, + "learning_rate": 1.7337157889889565e-05, + "loss": 1.105, + "step": 8151 + }, + { + "epoch": 0.26, + "learning_rate": 1.7336454066963023e-05, + "loss": 1.0444, + "step": 8152 + }, + { + "epoch": 0.26, + "learning_rate": 1.733575016532445e-05, + "loss": 1.0576, + "step": 8153 + }, + { + "epoch": 0.26, + "learning_rate": 1.7335046184981393e-05, + "loss": 1.147, + "step": 8154 + }, + { + "epoch": 0.26, + "learning_rate": 1.733434212594141e-05, + "loss": 1.084, + "step": 8155 + }, + { + "epoch": 0.26, + "learning_rate": 1.7333637988212053e-05, + "loss": 0.9819, + "step": 8156 + }, + { + "epoch": 0.26, + "learning_rate": 1.733293377180088e-05, + "loss": 0.9824, + "step": 8157 + }, + { + "epoch": 0.26, + "learning_rate": 1.7332229476715438e-05, + "loss": 1.0317, + "step": 8158 + }, + { + "epoch": 0.26, + "learning_rate": 1.7331525102963292e-05, + "loss": 1.1362, + "step": 8159 + }, + { + "epoch": 0.26, + "learning_rate": 1.7330820650551995e-05, + "loss": 1.127, + "step": 8160 + }, + { + "epoch": 0.26, + "learning_rate": 1.7330116119489104e-05, + "loss": 0.574, + "step": 8161 + }, + { + "epoch": 0.26, + "learning_rate": 1.7329411509782182e-05, + "loss": 0.8975, + "step": 8162 + }, + { + "epoch": 0.26, + "learning_rate": 1.7328706821438785e-05, + "loss": 1.1626, + "step": 8163 + }, + { + "epoch": 0.26, + "learning_rate": 1.7328002054466475e-05, + "loss": 1.0698, + "step": 8164 + }, + { + "epoch": 0.26, + "learning_rate": 1.7327297208872817e-05, + "loss": 1.0171, + "step": 8165 + }, + { + "epoch": 0.26, + "learning_rate": 1.7326592284665364e-05, + "loss": 1.0669, + "step": 8166 + }, + { + "epoch": 0.26, + "learning_rate": 1.7325887281851684e-05, + "loss": 1.1099, + "step": 8167 + }, + { + "epoch": 0.26, + "learning_rate": 1.7325182200439345e-05, + "loss": 1.1489, + "step": 8168 + }, + { + "epoch": 0.26, + "learning_rate": 1.7324477040435906e-05, + "loss": 1.0903, + "step": 8169 + }, + { + "epoch": 0.26, + "learning_rate": 1.7323771801848936e-05, + "loss": 1.104, + "step": 8170 + }, + { + "epoch": 0.26, + "learning_rate": 1.7323066484686002e-05, + "loss": 1.0356, + "step": 8171 + }, + { + "epoch": 0.26, + "learning_rate": 1.7322361088954667e-05, + "loss": 1.0361, + "step": 8172 + }, + { + "epoch": 0.26, + "learning_rate": 1.73216556146625e-05, + "loss": 0.9209, + "step": 8173 + }, + { + "epoch": 0.26, + "learning_rate": 1.7320950061817073e-05, + "loss": 1.0566, + "step": 8174 + }, + { + "epoch": 0.26, + "learning_rate": 1.7320244430425953e-05, + "loss": 1.0181, + "step": 8175 + }, + { + "epoch": 0.26, + "learning_rate": 1.7319538720496714e-05, + "loss": 1.2446, + "step": 8176 + }, + { + "epoch": 0.26, + "learning_rate": 1.7318832932036924e-05, + "loss": 0.959, + "step": 8177 + }, + { + "epoch": 0.26, + "learning_rate": 1.7318127065054156e-05, + "loss": 0.9814, + "step": 8178 + }, + { + "epoch": 0.26, + "learning_rate": 1.731742111955599e-05, + "loss": 1.1284, + "step": 8179 + }, + { + "epoch": 0.26, + "learning_rate": 1.7316715095549986e-05, + "loss": 1.1069, + "step": 8180 + }, + { + "epoch": 0.26, + "learning_rate": 1.731600899304373e-05, + "loss": 1.0503, + "step": 8181 + }, + { + "epoch": 0.26, + "learning_rate": 1.7315302812044797e-05, + "loss": 1.125, + "step": 8182 + }, + { + "epoch": 0.26, + "learning_rate": 1.7314596552560758e-05, + "loss": 1.0571, + "step": 8183 + }, + { + "epoch": 0.26, + "learning_rate": 1.7313890214599195e-05, + "loss": 1.1162, + "step": 8184 + }, + { + "epoch": 0.26, + "learning_rate": 1.731318379816768e-05, + "loss": 1.0459, + "step": 8185 + }, + { + "epoch": 0.26, + "learning_rate": 1.7312477303273802e-05, + "loss": 1.2061, + "step": 8186 + }, + { + "epoch": 0.26, + "learning_rate": 1.7311770729925134e-05, + "loss": 1.04, + "step": 8187 + }, + { + "epoch": 0.26, + "learning_rate": 1.7311064078129258e-05, + "loss": 1.166, + "step": 8188 + }, + { + "epoch": 0.26, + "learning_rate": 1.7310357347893757e-05, + "loss": 1.2275, + "step": 8189 + }, + { + "epoch": 0.26, + "learning_rate": 1.7309650539226207e-05, + "loss": 1.1636, + "step": 8190 + }, + { + "epoch": 0.26, + "learning_rate": 1.7308943652134204e-05, + "loss": 1.2158, + "step": 8191 + }, + { + "epoch": 0.26, + "learning_rate": 1.730823668662532e-05, + "loss": 0.9531, + "step": 8192 + }, + { + "epoch": 0.26, + "learning_rate": 1.730752964270715e-05, + "loss": 0.9854, + "step": 8193 + }, + { + "epoch": 0.26, + "learning_rate": 1.7306822520387266e-05, + "loss": 1.0557, + "step": 8194 + }, + { + "epoch": 0.26, + "learning_rate": 1.7306115319673268e-05, + "loss": 1.0869, + "step": 8195 + }, + { + "epoch": 0.26, + "learning_rate": 1.7305408040572743e-05, + "loss": 1.0728, + "step": 8196 + }, + { + "epoch": 0.26, + "learning_rate": 1.7304700683093268e-05, + "loss": 1.0459, + "step": 8197 + }, + { + "epoch": 0.26, + "learning_rate": 1.7303993247242442e-05, + "loss": 1.0508, + "step": 8198 + }, + { + "epoch": 0.26, + "learning_rate": 1.7303285733027852e-05, + "loss": 0.9873, + "step": 8199 + }, + { + "epoch": 0.26, + "learning_rate": 1.7302578140457086e-05, + "loss": 1.0596, + "step": 8200 + }, + { + "epoch": 0.26, + "learning_rate": 1.730187046953774e-05, + "loss": 1.1343, + "step": 8201 + }, + { + "epoch": 0.26, + "learning_rate": 1.730116272027741e-05, + "loss": 1.127, + "step": 8202 + }, + { + "epoch": 0.26, + "learning_rate": 1.7300454892683682e-05, + "loss": 1.082, + "step": 8203 + }, + { + "epoch": 0.26, + "learning_rate": 1.729974698676415e-05, + "loss": 1.0342, + "step": 8204 + }, + { + "epoch": 0.26, + "learning_rate": 1.7299039002526413e-05, + "loss": 1.0815, + "step": 8205 + }, + { + "epoch": 0.26, + "learning_rate": 1.7298330939978068e-05, + "loss": 1.0454, + "step": 8206 + }, + { + "epoch": 0.26, + "learning_rate": 1.7297622799126704e-05, + "loss": 0.9897, + "step": 8207 + }, + { + "epoch": 0.26, + "learning_rate": 1.729691457997993e-05, + "loss": 1.084, + "step": 8208 + }, + { + "epoch": 0.26, + "learning_rate": 1.7296206282545334e-05, + "loss": 1.0215, + "step": 8209 + }, + { + "epoch": 0.26, + "learning_rate": 1.7295497906830525e-05, + "loss": 1.1367, + "step": 8210 + }, + { + "epoch": 0.26, + "learning_rate": 1.7294789452843095e-05, + "loss": 0.981, + "step": 8211 + }, + { + "epoch": 0.26, + "learning_rate": 1.7294080920590647e-05, + "loss": 0.9644, + "step": 8212 + }, + { + "epoch": 0.26, + "learning_rate": 1.7293372310080784e-05, + "loss": 1.0547, + "step": 8213 + }, + { + "epoch": 0.26, + "learning_rate": 1.7292663621321106e-05, + "loss": 1.0283, + "step": 8214 + }, + { + "epoch": 0.26, + "learning_rate": 1.729195485431922e-05, + "loss": 1.1392, + "step": 8215 + }, + { + "epoch": 0.26, + "learning_rate": 1.7291246009082728e-05, + "loss": 1.1152, + "step": 8216 + }, + { + "epoch": 0.26, + "learning_rate": 1.729053708561924e-05, + "loss": 1.0698, + "step": 8217 + }, + { + "epoch": 0.26, + "learning_rate": 1.7289828083936352e-05, + "loss": 0.9351, + "step": 8218 + }, + { + "epoch": 0.26, + "learning_rate": 1.7289119004041678e-05, + "loss": 1.1646, + "step": 8219 + }, + { + "epoch": 0.26, + "learning_rate": 1.7288409845942826e-05, + "loss": 1.1348, + "step": 8220 + }, + { + "epoch": 0.26, + "learning_rate": 1.7287700609647404e-05, + "loss": 1.0146, + "step": 8221 + }, + { + "epoch": 0.26, + "learning_rate": 1.7286991295163018e-05, + "loss": 1.1367, + "step": 8222 + }, + { + "epoch": 0.26, + "learning_rate": 1.728628190249728e-05, + "loss": 1.0825, + "step": 8223 + }, + { + "epoch": 0.26, + "learning_rate": 1.72855724316578e-05, + "loss": 1.1592, + "step": 8224 + }, + { + "epoch": 0.26, + "learning_rate": 1.7284862882652194e-05, + "loss": 0.9766, + "step": 8225 + }, + { + "epoch": 0.26, + "learning_rate": 1.728415325548807e-05, + "loss": 1.0698, + "step": 8226 + }, + { + "epoch": 0.26, + "learning_rate": 1.7283443550173045e-05, + "loss": 1.0908, + "step": 8227 + }, + { + "epoch": 0.26, + "learning_rate": 1.728273376671473e-05, + "loss": 1.062, + "step": 8228 + }, + { + "epoch": 0.26, + "learning_rate": 1.7282023905120743e-05, + "loss": 1.0464, + "step": 8229 + }, + { + "epoch": 0.26, + "learning_rate": 1.72813139653987e-05, + "loss": 1.1851, + "step": 8230 + }, + { + "epoch": 0.26, + "learning_rate": 1.7280603947556212e-05, + "loss": 0.9966, + "step": 8231 + }, + { + "epoch": 0.26, + "learning_rate": 1.7279893851600906e-05, + "loss": 1.0508, + "step": 8232 + }, + { + "epoch": 0.26, + "learning_rate": 1.7279183677540393e-05, + "loss": 0.9995, + "step": 8233 + }, + { + "epoch": 0.26, + "learning_rate": 1.7278473425382298e-05, + "loss": 1.0303, + "step": 8234 + }, + { + "epoch": 0.26, + "learning_rate": 1.7277763095134236e-05, + "loss": 1.0996, + "step": 8235 + }, + { + "epoch": 0.26, + "learning_rate": 1.727705268680383e-05, + "loss": 1.1826, + "step": 8236 + }, + { + "epoch": 0.26, + "learning_rate": 1.7276342200398706e-05, + "loss": 0.9302, + "step": 8237 + }, + { + "epoch": 0.26, + "learning_rate": 1.727563163592648e-05, + "loss": 1.1553, + "step": 8238 + }, + { + "epoch": 0.26, + "learning_rate": 1.727492099339478e-05, + "loss": 1.145, + "step": 8239 + }, + { + "epoch": 0.26, + "learning_rate": 1.727421027281123e-05, + "loss": 1.0684, + "step": 8240 + }, + { + "epoch": 0.26, + "learning_rate": 1.7273499474183454e-05, + "loss": 1.1025, + "step": 8241 + }, + { + "epoch": 0.26, + "learning_rate": 1.7272788597519076e-05, + "loss": 1.208, + "step": 8242 + }, + { + "epoch": 0.26, + "learning_rate": 1.727207764282573e-05, + "loss": 1.064, + "step": 8243 + }, + { + "epoch": 0.26, + "learning_rate": 1.7271366610111034e-05, + "loss": 1.147, + "step": 8244 + }, + { + "epoch": 0.26, + "learning_rate": 1.7270655499382627e-05, + "loss": 0.9639, + "step": 8245 + }, + { + "epoch": 0.26, + "learning_rate": 1.7269944310648133e-05, + "loss": 1.1655, + "step": 8246 + }, + { + "epoch": 0.26, + "learning_rate": 1.726923304391518e-05, + "loss": 0.9727, + "step": 8247 + }, + { + "epoch": 0.26, + "learning_rate": 1.7268521699191403e-05, + "loss": 1.0669, + "step": 8248 + }, + { + "epoch": 0.26, + "learning_rate": 1.726781027648443e-05, + "loss": 1.0493, + "step": 8249 + }, + { + "epoch": 0.26, + "learning_rate": 1.7267098775801902e-05, + "loss": 1.123, + "step": 8250 + }, + { + "epoch": 0.26, + "learning_rate": 1.726638719715144e-05, + "loss": 1.0967, + "step": 8251 + }, + { + "epoch": 0.26, + "learning_rate": 1.726567554054069e-05, + "loss": 1.062, + "step": 8252 + }, + { + "epoch": 0.26, + "learning_rate": 1.726496380597728e-05, + "loss": 1.1147, + "step": 8253 + }, + { + "epoch": 0.26, + "learning_rate": 1.7264251993468853e-05, + "loss": 1.1602, + "step": 8254 + }, + { + "epoch": 0.26, + "learning_rate": 1.7263540103023036e-05, + "loss": 1.0425, + "step": 8255 + }, + { + "epoch": 0.26, + "learning_rate": 1.7262828134647478e-05, + "loss": 1.1016, + "step": 8256 + }, + { + "epoch": 0.26, + "learning_rate": 1.726211608834981e-05, + "loss": 1.2607, + "step": 8257 + }, + { + "epoch": 0.26, + "learning_rate": 1.7261403964137673e-05, + "loss": 1.0347, + "step": 8258 + }, + { + "epoch": 0.26, + "learning_rate": 1.7260691762018708e-05, + "loss": 1.0381, + "step": 8259 + }, + { + "epoch": 0.26, + "learning_rate": 1.7259979482000558e-05, + "loss": 0.98, + "step": 8260 + }, + { + "epoch": 0.26, + "learning_rate": 1.725926712409086e-05, + "loss": 1.0791, + "step": 8261 + }, + { + "epoch": 0.26, + "learning_rate": 1.725855468829726e-05, + "loss": 1.1426, + "step": 8262 + }, + { + "epoch": 0.26, + "learning_rate": 1.7257842174627407e-05, + "loss": 0.5557, + "step": 8263 + }, + { + "epoch": 0.26, + "learning_rate": 1.7257129583088933e-05, + "loss": 1.1562, + "step": 8264 + }, + { + "epoch": 0.26, + "learning_rate": 1.7256416913689497e-05, + "loss": 1.0469, + "step": 8265 + }, + { + "epoch": 0.26, + "learning_rate": 1.7255704166436732e-05, + "loss": 0.9463, + "step": 8266 + }, + { + "epoch": 0.26, + "learning_rate": 1.7254991341338298e-05, + "loss": 0.9453, + "step": 8267 + }, + { + "epoch": 0.26, + "learning_rate": 1.725427843840183e-05, + "loss": 1.0249, + "step": 8268 + }, + { + "epoch": 0.26, + "learning_rate": 1.7253565457634986e-05, + "loss": 1.1079, + "step": 8269 + }, + { + "epoch": 0.26, + "learning_rate": 1.725285239904541e-05, + "loss": 1.1055, + "step": 8270 + }, + { + "epoch": 0.26, + "learning_rate": 1.725213926264076e-05, + "loss": 1.126, + "step": 8271 + }, + { + "epoch": 0.26, + "learning_rate": 1.725142604842868e-05, + "loss": 1.0474, + "step": 8272 + }, + { + "epoch": 0.26, + "learning_rate": 1.7250712756416824e-05, + "loss": 1.001, + "step": 8273 + }, + { + "epoch": 0.26, + "learning_rate": 1.7249999386612844e-05, + "loss": 1.1553, + "step": 8274 + }, + { + "epoch": 0.26, + "learning_rate": 1.7249285939024395e-05, + "loss": 1.0557, + "step": 8275 + }, + { + "epoch": 0.26, + "learning_rate": 1.7248572413659127e-05, + "loss": 1.1875, + "step": 8276 + }, + { + "epoch": 0.26, + "learning_rate": 1.7247858810524706e-05, + "loss": 1.0869, + "step": 8277 + }, + { + "epoch": 0.26, + "learning_rate": 1.7247145129628776e-05, + "loss": 0.9585, + "step": 8278 + }, + { + "epoch": 0.26, + "learning_rate": 1.7246431370979004e-05, + "loss": 0.9834, + "step": 8279 + }, + { + "epoch": 0.26, + "learning_rate": 1.7245717534583042e-05, + "loss": 1.02, + "step": 8280 + }, + { + "epoch": 0.26, + "learning_rate": 1.724500362044855e-05, + "loss": 1.1201, + "step": 8281 + }, + { + "epoch": 0.26, + "learning_rate": 1.7244289628583188e-05, + "loss": 0.9556, + "step": 8282 + }, + { + "epoch": 0.26, + "learning_rate": 1.7243575558994613e-05, + "loss": 1.0483, + "step": 8283 + }, + { + "epoch": 0.26, + "learning_rate": 1.7242861411690488e-05, + "loss": 1.145, + "step": 8284 + }, + { + "epoch": 0.26, + "learning_rate": 1.724214718667848e-05, + "loss": 1.0635, + "step": 8285 + }, + { + "epoch": 0.26, + "learning_rate": 1.724143288396625e-05, + "loss": 1.1001, + "step": 8286 + }, + { + "epoch": 0.27, + "learning_rate": 1.7240718503561455e-05, + "loss": 1.1611, + "step": 8287 + }, + { + "epoch": 0.27, + "learning_rate": 1.7240004045471763e-05, + "loss": 1.0337, + "step": 8288 + }, + { + "epoch": 0.27, + "learning_rate": 1.7239289509704843e-05, + "loss": 1.0356, + "step": 8289 + }, + { + "epoch": 0.27, + "learning_rate": 1.7238574896268357e-05, + "loss": 0.9517, + "step": 8290 + }, + { + "epoch": 0.27, + "learning_rate": 1.7237860205169978e-05, + "loss": 0.9727, + "step": 8291 + }, + { + "epoch": 0.27, + "learning_rate": 1.7237145436417363e-05, + "loss": 0.978, + "step": 8292 + }, + { + "epoch": 0.27, + "learning_rate": 1.7236430590018193e-05, + "loss": 1.123, + "step": 8293 + }, + { + "epoch": 0.27, + "learning_rate": 1.723571566598013e-05, + "loss": 1.0317, + "step": 8294 + }, + { + "epoch": 0.27, + "learning_rate": 1.7235000664310844e-05, + "loss": 1.0415, + "step": 8295 + }, + { + "epoch": 0.27, + "learning_rate": 1.723428558501801e-05, + "loss": 1.1304, + "step": 8296 + }, + { + "epoch": 0.27, + "learning_rate": 1.7233570428109295e-05, + "loss": 1.1318, + "step": 8297 + }, + { + "epoch": 0.27, + "learning_rate": 1.7232855193592378e-05, + "loss": 1.1187, + "step": 8298 + }, + { + "epoch": 0.27, + "learning_rate": 1.723213988147493e-05, + "loss": 1.0293, + "step": 8299 + }, + { + "epoch": 0.27, + "learning_rate": 1.7231424491764627e-05, + "loss": 1.0894, + "step": 8300 + }, + { + "epoch": 0.27, + "learning_rate": 1.7230709024469137e-05, + "loss": 1.0146, + "step": 8301 + }, + { + "epoch": 0.27, + "learning_rate": 1.7229993479596147e-05, + "loss": 1.1133, + "step": 8302 + }, + { + "epoch": 0.27, + "learning_rate": 1.7229277857153326e-05, + "loss": 0.9678, + "step": 8303 + }, + { + "epoch": 0.27, + "learning_rate": 1.7228562157148355e-05, + "loss": 1.0547, + "step": 8304 + }, + { + "epoch": 0.27, + "learning_rate": 1.7227846379588912e-05, + "loss": 1.0264, + "step": 8305 + }, + { + "epoch": 0.27, + "learning_rate": 1.7227130524482677e-05, + "loss": 0.9541, + "step": 8306 + }, + { + "epoch": 0.27, + "learning_rate": 1.722641459183733e-05, + "loss": 1.0596, + "step": 8307 + }, + { + "epoch": 0.27, + "learning_rate": 1.7225698581660552e-05, + "loss": 1.0645, + "step": 8308 + }, + { + "epoch": 0.27, + "learning_rate": 1.7224982493960026e-05, + "loss": 1.0874, + "step": 8309 + }, + { + "epoch": 0.27, + "learning_rate": 1.7224266328743432e-05, + "loss": 1.0142, + "step": 8310 + }, + { + "epoch": 0.27, + "learning_rate": 1.7223550086018458e-05, + "loss": 1.0303, + "step": 8311 + }, + { + "epoch": 0.27, + "learning_rate": 1.7222833765792785e-05, + "loss": 0.9902, + "step": 8312 + }, + { + "epoch": 0.27, + "learning_rate": 1.72221173680741e-05, + "loss": 1.1064, + "step": 8313 + }, + { + "epoch": 0.27, + "learning_rate": 1.7221400892870092e-05, + "loss": 1.0703, + "step": 8314 + }, + { + "epoch": 0.27, + "learning_rate": 1.7220684340188437e-05, + "loss": 0.9595, + "step": 8315 + }, + { + "epoch": 0.27, + "learning_rate": 1.7219967710036834e-05, + "loss": 1.0571, + "step": 8316 + }, + { + "epoch": 0.27, + "learning_rate": 1.721925100242297e-05, + "loss": 0.9448, + "step": 8317 + }, + { + "epoch": 0.27, + "learning_rate": 1.7218534217354534e-05, + "loss": 0.915, + "step": 8318 + }, + { + "epoch": 0.27, + "learning_rate": 1.721781735483921e-05, + "loss": 0.9604, + "step": 8319 + }, + { + "epoch": 0.27, + "learning_rate": 1.72171004148847e-05, + "loss": 0.957, + "step": 8320 + }, + { + "epoch": 0.27, + "learning_rate": 1.7216383397498685e-05, + "loss": 1.1543, + "step": 8321 + }, + { + "epoch": 0.27, + "learning_rate": 1.7215666302688866e-05, + "loss": 1.1704, + "step": 8322 + }, + { + "epoch": 0.27, + "learning_rate": 1.7214949130462933e-05, + "loss": 1.145, + "step": 8323 + }, + { + "epoch": 0.27, + "learning_rate": 1.721423188082858e-05, + "loss": 1.0298, + "step": 8324 + }, + { + "epoch": 0.27, + "learning_rate": 1.72135145537935e-05, + "loss": 1.1099, + "step": 8325 + }, + { + "epoch": 0.27, + "learning_rate": 1.7212797149365398e-05, + "loss": 1.1333, + "step": 8326 + }, + { + "epoch": 0.27, + "learning_rate": 1.721207966755196e-05, + "loss": 0.9512, + "step": 8327 + }, + { + "epoch": 0.27, + "learning_rate": 1.7211362108360893e-05, + "loss": 1.0459, + "step": 8328 + }, + { + "epoch": 0.27, + "learning_rate": 1.721064447179989e-05, + "loss": 0.9893, + "step": 8329 + }, + { + "epoch": 0.27, + "learning_rate": 1.7209926757876655e-05, + "loss": 1.0322, + "step": 8330 + }, + { + "epoch": 0.27, + "learning_rate": 1.720920896659888e-05, + "loss": 1.0391, + "step": 8331 + }, + { + "epoch": 0.27, + "learning_rate": 1.7208491097974277e-05, + "loss": 1.0537, + "step": 8332 + }, + { + "epoch": 0.27, + "learning_rate": 1.720777315201054e-05, + "loss": 0.9941, + "step": 8333 + }, + { + "epoch": 0.27, + "learning_rate": 1.7207055128715374e-05, + "loss": 0.8462, + "step": 8334 + }, + { + "epoch": 0.27, + "learning_rate": 1.7206337028096482e-05, + "loss": 0.5474, + "step": 8335 + }, + { + "epoch": 0.27, + "learning_rate": 1.7205618850161575e-05, + "loss": 1.2646, + "step": 8336 + }, + { + "epoch": 0.27, + "learning_rate": 1.7204900594918346e-05, + "loss": 0.9458, + "step": 8337 + }, + { + "epoch": 0.27, + "learning_rate": 1.7204182262374508e-05, + "loss": 1.0591, + "step": 8338 + }, + { + "epoch": 0.27, + "learning_rate": 1.720346385253777e-05, + "loss": 1.1401, + "step": 8339 + }, + { + "epoch": 0.27, + "learning_rate": 1.7202745365415834e-05, + "loss": 1.0376, + "step": 8340 + }, + { + "epoch": 0.27, + "learning_rate": 1.720202680101642e-05, + "loss": 1.0073, + "step": 8341 + }, + { + "epoch": 0.27, + "learning_rate": 1.7201308159347222e-05, + "loss": 1.0625, + "step": 8342 + }, + { + "epoch": 0.27, + "learning_rate": 1.720058944041596e-05, + "loss": 1.022, + "step": 8343 + }, + { + "epoch": 0.27, + "learning_rate": 1.719987064423034e-05, + "loss": 1.2012, + "step": 8344 + }, + { + "epoch": 0.27, + "learning_rate": 1.7199151770798077e-05, + "loss": 1.0747, + "step": 8345 + }, + { + "epoch": 0.27, + "learning_rate": 1.7198432820126883e-05, + "loss": 1.1226, + "step": 8346 + }, + { + "epoch": 0.27, + "learning_rate": 1.7197713792224476e-05, + "loss": 1.0542, + "step": 8347 + }, + { + "epoch": 0.27, + "learning_rate": 1.7196994687098563e-05, + "loss": 0.9434, + "step": 8348 + }, + { + "epoch": 0.27, + "learning_rate": 1.7196275504756862e-05, + "loss": 1.1318, + "step": 8349 + }, + { + "epoch": 0.27, + "learning_rate": 1.7195556245207088e-05, + "loss": 1.1553, + "step": 8350 + }, + { + "epoch": 0.27, + "learning_rate": 1.7194836908456963e-05, + "loss": 1.1035, + "step": 8351 + }, + { + "epoch": 0.27, + "learning_rate": 1.7194117494514198e-05, + "loss": 1.1182, + "step": 8352 + }, + { + "epoch": 0.27, + "learning_rate": 1.7193398003386514e-05, + "loss": 0.9248, + "step": 8353 + }, + { + "epoch": 0.27, + "learning_rate": 1.719267843508163e-05, + "loss": 1.1636, + "step": 8354 + }, + { + "epoch": 0.27, + "learning_rate": 1.7191958789607267e-05, + "loss": 1.1035, + "step": 8355 + }, + { + "epoch": 0.27, + "learning_rate": 1.7191239066971147e-05, + "loss": 1.0493, + "step": 8356 + }, + { + "epoch": 0.27, + "learning_rate": 1.719051926718099e-05, + "loss": 1.0679, + "step": 8357 + }, + { + "epoch": 0.27, + "learning_rate": 1.718979939024452e-05, + "loss": 1.0732, + "step": 8358 + }, + { + "epoch": 0.27, + "learning_rate": 1.7189079436169462e-05, + "loss": 1.1421, + "step": 8359 + }, + { + "epoch": 0.27, + "learning_rate": 1.7188359404963536e-05, + "loss": 1.1113, + "step": 8360 + }, + { + "epoch": 0.27, + "learning_rate": 1.718763929663447e-05, + "loss": 1.0747, + "step": 8361 + }, + { + "epoch": 0.27, + "learning_rate": 1.718691911118999e-05, + "loss": 1.0376, + "step": 8362 + }, + { + "epoch": 0.27, + "learning_rate": 1.718619884863782e-05, + "loss": 1.0415, + "step": 8363 + }, + { + "epoch": 0.27, + "learning_rate": 1.718547850898569e-05, + "loss": 1.062, + "step": 8364 + }, + { + "epoch": 0.27, + "learning_rate": 1.7184758092241326e-05, + "loss": 1.0093, + "step": 8365 + }, + { + "epoch": 0.27, + "learning_rate": 1.7184037598412465e-05, + "loss": 1.0942, + "step": 8366 + }, + { + "epoch": 0.27, + "learning_rate": 1.7183317027506828e-05, + "loss": 1.0205, + "step": 8367 + }, + { + "epoch": 0.27, + "learning_rate": 1.718259637953215e-05, + "loss": 1.0562, + "step": 8368 + }, + { + "epoch": 0.27, + "learning_rate": 1.7181875654496162e-05, + "loss": 0.9453, + "step": 8369 + }, + { + "epoch": 0.27, + "learning_rate": 1.7181154852406594e-05, + "loss": 1.0708, + "step": 8370 + }, + { + "epoch": 0.27, + "learning_rate": 1.7180433973271183e-05, + "loss": 1.1201, + "step": 8371 + }, + { + "epoch": 0.27, + "learning_rate": 1.7179713017097665e-05, + "loss": 1.0386, + "step": 8372 + }, + { + "epoch": 0.27, + "learning_rate": 1.717899198389377e-05, + "loss": 1.1431, + "step": 8373 + }, + { + "epoch": 0.27, + "learning_rate": 1.717827087366724e-05, + "loss": 1.0005, + "step": 8374 + }, + { + "epoch": 0.27, + "learning_rate": 1.7177549686425804e-05, + "loss": 1.1929, + "step": 8375 + }, + { + "epoch": 0.27, + "learning_rate": 1.7176828422177204e-05, + "loss": 1.2451, + "step": 8376 + }, + { + "epoch": 0.27, + "learning_rate": 1.7176107080929182e-05, + "loss": 1.0859, + "step": 8377 + }, + { + "epoch": 0.27, + "learning_rate": 1.717538566268947e-05, + "loss": 1.0239, + "step": 8378 + }, + { + "epoch": 0.27, + "learning_rate": 1.717466416746581e-05, + "loss": 1.0532, + "step": 8379 + }, + { + "epoch": 0.27, + "learning_rate": 1.7173942595265946e-05, + "loss": 1.0171, + "step": 8380 + }, + { + "epoch": 0.27, + "learning_rate": 1.7173220946097616e-05, + "loss": 1.0742, + "step": 8381 + }, + { + "epoch": 0.27, + "learning_rate": 1.7172499219968567e-05, + "loss": 1.085, + "step": 8382 + }, + { + "epoch": 0.27, + "learning_rate": 1.7171777416886538e-05, + "loss": 1.0928, + "step": 8383 + }, + { + "epoch": 0.27, + "learning_rate": 1.7171055536859276e-05, + "loss": 0.9951, + "step": 8384 + }, + { + "epoch": 0.27, + "learning_rate": 1.717033357989452e-05, + "loss": 1.0049, + "step": 8385 + }, + { + "epoch": 0.27, + "learning_rate": 1.7169611546000025e-05, + "loss": 1.0049, + "step": 8386 + }, + { + "epoch": 0.27, + "learning_rate": 1.7168889435183533e-05, + "loss": 0.9912, + "step": 8387 + }, + { + "epoch": 0.27, + "learning_rate": 1.716816724745279e-05, + "loss": 0.8813, + "step": 8388 + }, + { + "epoch": 0.27, + "learning_rate": 1.7167444982815545e-05, + "loss": 0.8672, + "step": 8389 + }, + { + "epoch": 0.27, + "learning_rate": 1.716672264127955e-05, + "loss": 1.1655, + "step": 8390 + }, + { + "epoch": 0.27, + "learning_rate": 1.7166000222852554e-05, + "loss": 1.0103, + "step": 8391 + }, + { + "epoch": 0.27, + "learning_rate": 1.71652777275423e-05, + "loss": 1.1416, + "step": 8392 + }, + { + "epoch": 0.27, + "learning_rate": 1.716455515535655e-05, + "loss": 1.0908, + "step": 8393 + }, + { + "epoch": 0.27, + "learning_rate": 1.7163832506303056e-05, + "loss": 0.9556, + "step": 8394 + }, + { + "epoch": 0.27, + "learning_rate": 1.7163109780389564e-05, + "loss": 1.1357, + "step": 8395 + }, + { + "epoch": 0.27, + "learning_rate": 1.716238697762383e-05, + "loss": 1.082, + "step": 8396 + }, + { + "epoch": 0.27, + "learning_rate": 1.716166409801361e-05, + "loss": 1.123, + "step": 8397 + }, + { + "epoch": 0.27, + "learning_rate": 1.716094114156667e-05, + "loss": 1.0806, + "step": 8398 + }, + { + "epoch": 0.27, + "learning_rate": 1.7160218108290748e-05, + "loss": 1.0483, + "step": 8399 + }, + { + "epoch": 0.27, + "learning_rate": 1.7159494998193614e-05, + "loss": 1.1377, + "step": 8400 + }, + { + "epoch": 0.27, + "learning_rate": 1.7158771811283017e-05, + "loss": 1.061, + "step": 8401 + }, + { + "epoch": 0.27, + "learning_rate": 1.7158048547566724e-05, + "loss": 1.0518, + "step": 8402 + }, + { + "epoch": 0.27, + "learning_rate": 1.71573252070525e-05, + "loss": 1.002, + "step": 8403 + }, + { + "epoch": 0.27, + "learning_rate": 1.7156601789748086e-05, + "loss": 1.1797, + "step": 8404 + }, + { + "epoch": 0.27, + "learning_rate": 1.7155878295661264e-05, + "loss": 0.5698, + "step": 8405 + }, + { + "epoch": 0.27, + "learning_rate": 1.7155154724799783e-05, + "loss": 1.0376, + "step": 8406 + }, + { + "epoch": 0.27, + "learning_rate": 1.715443107717141e-05, + "loss": 1.0312, + "step": 8407 + }, + { + "epoch": 0.27, + "learning_rate": 1.7153707352783914e-05, + "loss": 1.083, + "step": 8408 + }, + { + "epoch": 0.27, + "learning_rate": 1.7152983551645054e-05, + "loss": 1.0894, + "step": 8409 + }, + { + "epoch": 0.27, + "learning_rate": 1.7152259673762596e-05, + "loss": 1.0073, + "step": 8410 + }, + { + "epoch": 0.27, + "learning_rate": 1.7151535719144307e-05, + "loss": 0.9595, + "step": 8411 + }, + { + "epoch": 0.27, + "learning_rate": 1.7150811687797955e-05, + "loss": 1.0293, + "step": 8412 + }, + { + "epoch": 0.27, + "learning_rate": 1.7150087579731306e-05, + "loss": 1.0342, + "step": 8413 + }, + { + "epoch": 0.27, + "learning_rate": 1.7149363394952135e-05, + "loss": 1.104, + "step": 8414 + }, + { + "epoch": 0.27, + "learning_rate": 1.7148639133468205e-05, + "loss": 1.0049, + "step": 8415 + }, + { + "epoch": 0.27, + "learning_rate": 1.714791479528729e-05, + "loss": 1.022, + "step": 8416 + }, + { + "epoch": 0.27, + "learning_rate": 1.7147190380417157e-05, + "loss": 1.0479, + "step": 8417 + }, + { + "epoch": 0.27, + "learning_rate": 1.7146465888865584e-05, + "loss": 0.9922, + "step": 8418 + }, + { + "epoch": 0.27, + "learning_rate": 1.7145741320640337e-05, + "loss": 0.9629, + "step": 8419 + }, + { + "epoch": 0.27, + "learning_rate": 1.71450166757492e-05, + "loss": 1.0845, + "step": 8420 + }, + { + "epoch": 0.27, + "learning_rate": 1.7144291954199942e-05, + "loss": 0.9521, + "step": 8421 + }, + { + "epoch": 0.27, + "learning_rate": 1.7143567156000334e-05, + "loss": 1.0752, + "step": 8422 + }, + { + "epoch": 0.27, + "learning_rate": 1.714284228115816e-05, + "loss": 0.9668, + "step": 8423 + }, + { + "epoch": 0.27, + "learning_rate": 1.714211732968119e-05, + "loss": 1.1221, + "step": 8424 + }, + { + "epoch": 0.27, + "learning_rate": 1.7141392301577207e-05, + "loss": 0.5452, + "step": 8425 + }, + { + "epoch": 0.27, + "learning_rate": 1.7140667196853987e-05, + "loss": 0.9751, + "step": 8426 + }, + { + "epoch": 0.27, + "learning_rate": 1.7139942015519314e-05, + "loss": 1.0127, + "step": 8427 + }, + { + "epoch": 0.27, + "learning_rate": 1.7139216757580963e-05, + "loss": 1.1045, + "step": 8428 + }, + { + "epoch": 0.27, + "learning_rate": 1.713849142304672e-05, + "loss": 1.0547, + "step": 8429 + }, + { + "epoch": 0.27, + "learning_rate": 1.713776601192436e-05, + "loss": 1.1626, + "step": 8430 + }, + { + "epoch": 0.27, + "learning_rate": 1.7137040524221676e-05, + "loss": 0.9888, + "step": 8431 + }, + { + "epoch": 0.27, + "learning_rate": 1.7136314959946443e-05, + "loss": 1.0645, + "step": 8432 + }, + { + "epoch": 0.27, + "learning_rate": 1.7135589319106448e-05, + "loss": 1.0464, + "step": 8433 + }, + { + "epoch": 0.27, + "learning_rate": 1.7134863601709478e-05, + "loss": 0.9561, + "step": 8434 + }, + { + "epoch": 0.27, + "learning_rate": 1.7134137807763318e-05, + "loss": 0.979, + "step": 8435 + }, + { + "epoch": 0.27, + "learning_rate": 1.7133411937275757e-05, + "loss": 1.1465, + "step": 8436 + }, + { + "epoch": 0.27, + "learning_rate": 1.7132685990254575e-05, + "loss": 1.1255, + "step": 8437 + }, + { + "epoch": 0.27, + "learning_rate": 1.7131959966707572e-05, + "loss": 1.0405, + "step": 8438 + }, + { + "epoch": 0.27, + "learning_rate": 1.713123386664253e-05, + "loss": 1.0327, + "step": 8439 + }, + { + "epoch": 0.27, + "learning_rate": 1.7130507690067243e-05, + "loss": 0.9932, + "step": 8440 + }, + { + "epoch": 0.27, + "learning_rate": 1.7129781436989493e-05, + "loss": 1.0186, + "step": 8441 + }, + { + "epoch": 0.27, + "learning_rate": 1.7129055107417088e-05, + "loss": 0.9458, + "step": 8442 + }, + { + "epoch": 0.27, + "learning_rate": 1.7128328701357805e-05, + "loss": 1.126, + "step": 8443 + }, + { + "epoch": 0.27, + "learning_rate": 1.7127602218819446e-05, + "loss": 1.0737, + "step": 8444 + }, + { + "epoch": 0.27, + "learning_rate": 1.7126875659809808e-05, + "loss": 0.9951, + "step": 8445 + }, + { + "epoch": 0.27, + "learning_rate": 1.712614902433668e-05, + "loss": 1.063, + "step": 8446 + }, + { + "epoch": 0.27, + "learning_rate": 1.7125422312407856e-05, + "loss": 1.1118, + "step": 8447 + }, + { + "epoch": 0.27, + "learning_rate": 1.7124695524031138e-05, + "loss": 1.0435, + "step": 8448 + }, + { + "epoch": 0.27, + "learning_rate": 1.712396865921432e-05, + "loss": 1.1899, + "step": 8449 + }, + { + "epoch": 0.27, + "learning_rate": 1.712324171796521e-05, + "loss": 1.0083, + "step": 8450 + }, + { + "epoch": 0.27, + "learning_rate": 1.7122514700291593e-05, + "loss": 1.0713, + "step": 8451 + }, + { + "epoch": 0.27, + "learning_rate": 1.712178760620128e-05, + "loss": 0.9697, + "step": 8452 + }, + { + "epoch": 0.27, + "learning_rate": 1.7121060435702068e-05, + "loss": 1.041, + "step": 8453 + }, + { + "epoch": 0.27, + "learning_rate": 1.7120333188801756e-05, + "loss": 0.5386, + "step": 8454 + }, + { + "epoch": 0.27, + "learning_rate": 1.711960586550815e-05, + "loss": 0.9355, + "step": 8455 + }, + { + "epoch": 0.27, + "learning_rate": 1.7118878465829055e-05, + "loss": 1.0059, + "step": 8456 + }, + { + "epoch": 0.27, + "learning_rate": 1.711815098977227e-05, + "loss": 1.083, + "step": 8457 + }, + { + "epoch": 0.27, + "learning_rate": 1.7117423437345607e-05, + "loss": 1.0605, + "step": 8458 + }, + { + "epoch": 0.27, + "learning_rate": 1.7116695808556863e-05, + "loss": 1.0918, + "step": 8459 + }, + { + "epoch": 0.27, + "learning_rate": 1.7115968103413852e-05, + "loss": 1.1182, + "step": 8460 + }, + { + "epoch": 0.27, + "learning_rate": 1.711524032192438e-05, + "loss": 0.9653, + "step": 8461 + }, + { + "epoch": 0.27, + "learning_rate": 1.711451246409625e-05, + "loss": 1.0024, + "step": 8462 + }, + { + "epoch": 0.27, + "learning_rate": 1.7113784529937278e-05, + "loss": 1.1201, + "step": 8463 + }, + { + "epoch": 0.27, + "learning_rate": 1.711305651945527e-05, + "loss": 1.0386, + "step": 8464 + }, + { + "epoch": 0.27, + "learning_rate": 1.7112328432658038e-05, + "loss": 1.1313, + "step": 8465 + }, + { + "epoch": 0.27, + "learning_rate": 1.7111600269553396e-05, + "loss": 1.0049, + "step": 8466 + }, + { + "epoch": 0.27, + "learning_rate": 1.711087203014915e-05, + "loss": 1.0586, + "step": 8467 + }, + { + "epoch": 0.27, + "learning_rate": 1.711014371445312e-05, + "loss": 1.0186, + "step": 8468 + }, + { + "epoch": 0.27, + "learning_rate": 1.7109415322473117e-05, + "loss": 1.0776, + "step": 8469 + }, + { + "epoch": 0.27, + "learning_rate": 1.7108686854216955e-05, + "loss": 1.0288, + "step": 8470 + }, + { + "epoch": 0.27, + "learning_rate": 1.710795830969245e-05, + "loss": 1.124, + "step": 8471 + }, + { + "epoch": 0.27, + "learning_rate": 1.7107229688907417e-05, + "loss": 1.022, + "step": 8472 + }, + { + "epoch": 0.27, + "learning_rate": 1.710650099186968e-05, + "loss": 0.9019, + "step": 8473 + }, + { + "epoch": 0.27, + "learning_rate": 1.710577221858705e-05, + "loss": 1.0078, + "step": 8474 + }, + { + "epoch": 0.27, + "learning_rate": 1.7105043369067347e-05, + "loss": 1.125, + "step": 8475 + }, + { + "epoch": 0.27, + "learning_rate": 1.7104314443318393e-05, + "loss": 1.0771, + "step": 8476 + }, + { + "epoch": 0.27, + "learning_rate": 1.7103585441348007e-05, + "loss": 1.0947, + "step": 8477 + }, + { + "epoch": 0.27, + "learning_rate": 1.7102856363164012e-05, + "loss": 1.0713, + "step": 8478 + }, + { + "epoch": 0.27, + "learning_rate": 1.7102127208774232e-05, + "loss": 1.0981, + "step": 8479 + }, + { + "epoch": 0.27, + "learning_rate": 1.7101397978186486e-05, + "loss": 1.2065, + "step": 8480 + }, + { + "epoch": 0.27, + "learning_rate": 1.7100668671408597e-05, + "loss": 1.0542, + "step": 8481 + }, + { + "epoch": 0.27, + "learning_rate": 1.7099939288448392e-05, + "loss": 1.0527, + "step": 8482 + }, + { + "epoch": 0.27, + "learning_rate": 1.7099209829313697e-05, + "loss": 0.9824, + "step": 8483 + }, + { + "epoch": 0.27, + "learning_rate": 1.7098480294012337e-05, + "loss": 0.9927, + "step": 8484 + }, + { + "epoch": 0.27, + "learning_rate": 1.709775068255214e-05, + "loss": 1.0669, + "step": 8485 + }, + { + "epoch": 0.27, + "learning_rate": 1.7097020994940936e-05, + "loss": 1.063, + "step": 8486 + }, + { + "epoch": 0.27, + "learning_rate": 1.709629123118655e-05, + "loss": 1.0923, + "step": 8487 + }, + { + "epoch": 0.27, + "learning_rate": 1.709556139129681e-05, + "loss": 1.0063, + "step": 8488 + }, + { + "epoch": 0.27, + "learning_rate": 1.709483147527955e-05, + "loss": 1.0571, + "step": 8489 + }, + { + "epoch": 0.27, + "learning_rate": 1.7094101483142605e-05, + "loss": 1.1006, + "step": 8490 + }, + { + "epoch": 0.27, + "learning_rate": 1.70933714148938e-05, + "loss": 1.1558, + "step": 8491 + }, + { + "epoch": 0.27, + "learning_rate": 1.709264127054097e-05, + "loss": 1.1182, + "step": 8492 + }, + { + "epoch": 0.27, + "learning_rate": 1.7091911050091944e-05, + "loss": 1.0376, + "step": 8493 + }, + { + "epoch": 0.27, + "learning_rate": 1.709118075355457e-05, + "loss": 1.1118, + "step": 8494 + }, + { + "epoch": 0.27, + "learning_rate": 1.7090450380936668e-05, + "loss": 1.1104, + "step": 8495 + }, + { + "epoch": 0.27, + "learning_rate": 1.7089719932246083e-05, + "loss": 1.0, + "step": 8496 + }, + { + "epoch": 0.27, + "learning_rate": 1.708898940749065e-05, + "loss": 1.0859, + "step": 8497 + }, + { + "epoch": 0.27, + "learning_rate": 1.7088258806678206e-05, + "loss": 1.0425, + "step": 8498 + }, + { + "epoch": 0.27, + "learning_rate": 1.708752812981659e-05, + "loss": 1.1987, + "step": 8499 + }, + { + "epoch": 0.27, + "learning_rate": 1.708679737691364e-05, + "loss": 1.1851, + "step": 8500 + }, + { + "epoch": 0.27, + "learning_rate": 1.70860665479772e-05, + "loss": 1.0405, + "step": 8501 + }, + { + "epoch": 0.27, + "learning_rate": 1.7085335643015107e-05, + "loss": 0.9468, + "step": 8502 + }, + { + "epoch": 0.27, + "learning_rate": 1.7084604662035205e-05, + "loss": 1.0757, + "step": 8503 + }, + { + "epoch": 0.27, + "learning_rate": 1.7083873605045335e-05, + "loss": 0.979, + "step": 8504 + }, + { + "epoch": 0.27, + "learning_rate": 1.7083142472053343e-05, + "loss": 1.0156, + "step": 8505 + }, + { + "epoch": 0.27, + "learning_rate": 1.7082411263067066e-05, + "loss": 1.1035, + "step": 8506 + }, + { + "epoch": 0.27, + "learning_rate": 1.708167997809436e-05, + "loss": 1.1445, + "step": 8507 + }, + { + "epoch": 0.27, + "learning_rate": 1.7080948617143065e-05, + "loss": 1.061, + "step": 8508 + }, + { + "epoch": 0.27, + "learning_rate": 1.7080217180221024e-05, + "loss": 1.0508, + "step": 8509 + }, + { + "epoch": 0.27, + "learning_rate": 1.7079485667336094e-05, + "loss": 1.0884, + "step": 8510 + }, + { + "epoch": 0.27, + "learning_rate": 1.7078754078496116e-05, + "loss": 0.9351, + "step": 8511 + }, + { + "epoch": 0.27, + "learning_rate": 1.707802241370894e-05, + "loss": 1.0864, + "step": 8512 + }, + { + "epoch": 0.27, + "learning_rate": 1.7077290672982418e-05, + "loss": 1.0596, + "step": 8513 + }, + { + "epoch": 0.27, + "learning_rate": 1.70765588563244e-05, + "loss": 1.0151, + "step": 8514 + }, + { + "epoch": 0.27, + "learning_rate": 1.707582696374274e-05, + "loss": 1.1577, + "step": 8515 + }, + { + "epoch": 0.27, + "learning_rate": 1.7075094995245284e-05, + "loss": 1.1128, + "step": 8516 + }, + { + "epoch": 0.27, + "learning_rate": 1.7074362950839887e-05, + "loss": 1.0542, + "step": 8517 + }, + { + "epoch": 0.27, + "learning_rate": 1.707363083053441e-05, + "loss": 1.144, + "step": 8518 + }, + { + "epoch": 0.27, + "learning_rate": 1.7072898634336704e-05, + "loss": 1.0098, + "step": 8519 + }, + { + "epoch": 0.27, + "learning_rate": 1.7072166362254615e-05, + "loss": 1.0898, + "step": 8520 + }, + { + "epoch": 0.27, + "learning_rate": 1.7071434014296016e-05, + "loss": 1.0508, + "step": 8521 + }, + { + "epoch": 0.27, + "learning_rate": 1.7070701590468753e-05, + "loss": 1.0869, + "step": 8522 + }, + { + "epoch": 0.27, + "learning_rate": 1.7069969090780688e-05, + "loss": 1.0728, + "step": 8523 + }, + { + "epoch": 0.27, + "learning_rate": 1.706923651523968e-05, + "loss": 1.064, + "step": 8524 + }, + { + "epoch": 0.27, + "learning_rate": 1.7068503863853588e-05, + "loss": 1.0962, + "step": 8525 + }, + { + "epoch": 0.27, + "learning_rate": 1.7067771136630274e-05, + "loss": 1.0244, + "step": 8526 + }, + { + "epoch": 0.27, + "learning_rate": 1.7067038333577595e-05, + "loss": 0.9629, + "step": 8527 + }, + { + "epoch": 0.27, + "learning_rate": 1.7066305454703417e-05, + "loss": 1.1255, + "step": 8528 + }, + { + "epoch": 0.27, + "learning_rate": 1.7065572500015602e-05, + "loss": 1.1157, + "step": 8529 + }, + { + "epoch": 0.27, + "learning_rate": 1.7064839469522014e-05, + "loss": 1.0835, + "step": 8530 + }, + { + "epoch": 0.27, + "learning_rate": 1.7064106363230518e-05, + "loss": 1.0127, + "step": 8531 + }, + { + "epoch": 0.27, + "learning_rate": 1.706337318114898e-05, + "loss": 1.1494, + "step": 8532 + }, + { + "epoch": 0.27, + "learning_rate": 1.7062639923285264e-05, + "loss": 1.1353, + "step": 8533 + }, + { + "epoch": 0.27, + "learning_rate": 1.7061906589647238e-05, + "loss": 1.0273, + "step": 8534 + }, + { + "epoch": 0.27, + "learning_rate": 1.7061173180242768e-05, + "loss": 1.0503, + "step": 8535 + }, + { + "epoch": 0.27, + "learning_rate": 1.7060439695079726e-05, + "loss": 1.0957, + "step": 8536 + }, + { + "epoch": 0.27, + "learning_rate": 1.7059706134165982e-05, + "loss": 1.1113, + "step": 8537 + }, + { + "epoch": 0.27, + "learning_rate": 1.7058972497509403e-05, + "loss": 1.1177, + "step": 8538 + }, + { + "epoch": 0.27, + "learning_rate": 1.7058238785117864e-05, + "loss": 1.0596, + "step": 8539 + }, + { + "epoch": 0.27, + "learning_rate": 1.705750499699923e-05, + "loss": 0.5637, + "step": 8540 + }, + { + "epoch": 0.27, + "learning_rate": 1.7056771133161382e-05, + "loss": 1.1514, + "step": 8541 + }, + { + "epoch": 0.27, + "learning_rate": 1.7056037193612184e-05, + "loss": 1.1318, + "step": 8542 + }, + { + "epoch": 0.27, + "learning_rate": 1.7055303178359525e-05, + "loss": 1.0591, + "step": 8543 + }, + { + "epoch": 0.27, + "learning_rate": 1.7054569087411262e-05, + "loss": 0.9863, + "step": 8544 + }, + { + "epoch": 0.27, + "learning_rate": 1.7053834920775286e-05, + "loss": 1.0425, + "step": 8545 + }, + { + "epoch": 0.27, + "learning_rate": 1.7053100678459467e-05, + "loss": 1.0645, + "step": 8546 + }, + { + "epoch": 0.27, + "learning_rate": 1.7052366360471686e-05, + "loss": 1.0171, + "step": 8547 + }, + { + "epoch": 0.27, + "learning_rate": 1.7051631966819814e-05, + "loss": 1.0127, + "step": 8548 + }, + { + "epoch": 0.27, + "learning_rate": 1.705089749751174e-05, + "loss": 1.002, + "step": 8549 + }, + { + "epoch": 0.27, + "learning_rate": 1.7050162952555337e-05, + "loss": 1.0776, + "step": 8550 + }, + { + "epoch": 0.27, + "learning_rate": 1.704942833195849e-05, + "loss": 1.1006, + "step": 8551 + }, + { + "epoch": 0.27, + "learning_rate": 1.704869363572908e-05, + "loss": 1.0166, + "step": 8552 + }, + { + "epoch": 0.27, + "learning_rate": 1.7047958863874986e-05, + "loss": 1.0674, + "step": 8553 + }, + { + "epoch": 0.27, + "learning_rate": 1.704722401640409e-05, + "loss": 1.0688, + "step": 8554 + }, + { + "epoch": 0.27, + "learning_rate": 1.7046489093324284e-05, + "loss": 0.9385, + "step": 8555 + }, + { + "epoch": 0.27, + "learning_rate": 1.704575409464345e-05, + "loss": 1.1226, + "step": 8556 + }, + { + "epoch": 0.27, + "learning_rate": 1.704501902036947e-05, + "loss": 1.0884, + "step": 8557 + }, + { + "epoch": 0.27, + "learning_rate": 1.7044283870510233e-05, + "loss": 1.0405, + "step": 8558 + }, + { + "epoch": 0.27, + "learning_rate": 1.704354864507363e-05, + "loss": 1.1362, + "step": 8559 + }, + { + "epoch": 0.27, + "learning_rate": 1.7042813344067542e-05, + "loss": 0.9946, + "step": 8560 + }, + { + "epoch": 0.27, + "learning_rate": 1.7042077967499858e-05, + "loss": 1.0454, + "step": 8561 + }, + { + "epoch": 0.27, + "learning_rate": 1.7041342515378478e-05, + "loss": 1.1333, + "step": 8562 + }, + { + "epoch": 0.27, + "learning_rate": 1.7040606987711283e-05, + "loss": 1.0166, + "step": 8563 + }, + { + "epoch": 0.27, + "learning_rate": 1.7039871384506165e-05, + "loss": 1.1279, + "step": 8564 + }, + { + "epoch": 0.27, + "learning_rate": 1.7039135705771023e-05, + "loss": 1.0425, + "step": 8565 + }, + { + "epoch": 0.27, + "learning_rate": 1.7038399951513743e-05, + "loss": 0.9507, + "step": 8566 + }, + { + "epoch": 0.27, + "learning_rate": 1.7037664121742222e-05, + "loss": 1.1934, + "step": 8567 + }, + { + "epoch": 0.27, + "learning_rate": 1.703692821646435e-05, + "loss": 1.0679, + "step": 8568 + }, + { + "epoch": 0.27, + "learning_rate": 1.703619223568803e-05, + "loss": 1.0732, + "step": 8569 + }, + { + "epoch": 0.27, + "learning_rate": 1.7035456179421154e-05, + "loss": 1.0698, + "step": 8570 + }, + { + "epoch": 0.27, + "learning_rate": 1.703472004767162e-05, + "loss": 1.1665, + "step": 8571 + }, + { + "epoch": 0.27, + "learning_rate": 1.7033983840447327e-05, + "loss": 1.0601, + "step": 8572 + }, + { + "epoch": 0.27, + "learning_rate": 1.703324755775617e-05, + "loss": 1.083, + "step": 8573 + }, + { + "epoch": 0.27, + "learning_rate": 1.703251119960605e-05, + "loss": 1.0132, + "step": 8574 + }, + { + "epoch": 0.27, + "learning_rate": 1.7031774766004868e-05, + "loss": 1.0659, + "step": 8575 + }, + { + "epoch": 0.27, + "learning_rate": 1.7031038256960527e-05, + "loss": 1.0303, + "step": 8576 + }, + { + "epoch": 0.27, + "learning_rate": 1.7030301672480925e-05, + "loss": 1.1382, + "step": 8577 + }, + { + "epoch": 0.27, + "learning_rate": 1.7029565012573968e-05, + "loss": 1.3115, + "step": 8578 + }, + { + "epoch": 0.27, + "learning_rate": 1.7028828277247563e-05, + "loss": 1.0576, + "step": 8579 + }, + { + "epoch": 0.27, + "learning_rate": 1.7028091466509602e-05, + "loss": 1.0557, + "step": 8580 + }, + { + "epoch": 0.27, + "learning_rate": 1.7027354580368002e-05, + "loss": 1.02, + "step": 8581 + }, + { + "epoch": 0.27, + "learning_rate": 1.7026617618830664e-05, + "loss": 1.168, + "step": 8582 + }, + { + "epoch": 0.27, + "learning_rate": 1.7025880581905493e-05, + "loss": 1.0151, + "step": 8583 + }, + { + "epoch": 0.27, + "learning_rate": 1.7025143469600404e-05, + "loss": 1.1074, + "step": 8584 + }, + { + "epoch": 0.27, + "learning_rate": 1.70244062819233e-05, + "loss": 0.9941, + "step": 8585 + }, + { + "epoch": 0.27, + "learning_rate": 1.702366901888209e-05, + "loss": 1.0107, + "step": 8586 + }, + { + "epoch": 0.27, + "learning_rate": 1.702293168048468e-05, + "loss": 1.0747, + "step": 8587 + }, + { + "epoch": 0.27, + "learning_rate": 1.702219426673899e-05, + "loss": 1.0884, + "step": 8588 + }, + { + "epoch": 0.27, + "learning_rate": 1.702145677765293e-05, + "loss": 1.0679, + "step": 8589 + }, + { + "epoch": 0.27, + "learning_rate": 1.7020719213234405e-05, + "loss": 0.9434, + "step": 8590 + }, + { + "epoch": 0.27, + "learning_rate": 1.7019981573491336e-05, + "loss": 1.02, + "step": 8591 + }, + { + "epoch": 0.27, + "learning_rate": 1.701924385843163e-05, + "loss": 1.0254, + "step": 8592 + }, + { + "epoch": 0.27, + "learning_rate": 1.701850606806321e-05, + "loss": 1.1562, + "step": 8593 + }, + { + "epoch": 0.27, + "learning_rate": 1.7017768202393983e-05, + "loss": 1.0845, + "step": 8594 + }, + { + "epoch": 0.27, + "learning_rate": 1.7017030261431876e-05, + "loss": 1.0093, + "step": 8595 + }, + { + "epoch": 0.27, + "learning_rate": 1.70162922451848e-05, + "loss": 0.959, + "step": 8596 + }, + { + "epoch": 0.27, + "learning_rate": 1.7015554153660668e-05, + "loss": 1.0728, + "step": 8597 + }, + { + "epoch": 0.27, + "learning_rate": 1.701481598686741e-05, + "loss": 1.0342, + "step": 8598 + }, + { + "epoch": 0.28, + "learning_rate": 1.7014077744812937e-05, + "loss": 0.9878, + "step": 8599 + }, + { + "epoch": 0.28, + "learning_rate": 1.7013339427505174e-05, + "loss": 0.9893, + "step": 8600 + }, + { + "epoch": 0.28, + "learning_rate": 1.701260103495204e-05, + "loss": 0.5564, + "step": 8601 + }, + { + "epoch": 0.28, + "learning_rate": 1.7011862567161455e-05, + "loss": 1.1836, + "step": 8602 + }, + { + "epoch": 0.28, + "learning_rate": 1.7011124024141353e-05, + "loss": 1.1235, + "step": 8603 + }, + { + "epoch": 0.28, + "learning_rate": 1.701038540589964e-05, + "loss": 1.1333, + "step": 8604 + }, + { + "epoch": 0.28, + "learning_rate": 1.7009646712444257e-05, + "loss": 0.9683, + "step": 8605 + }, + { + "epoch": 0.28, + "learning_rate": 1.700890794378312e-05, + "loss": 1.0405, + "step": 8606 + }, + { + "epoch": 0.28, + "learning_rate": 1.7008169099924158e-05, + "loss": 0.96, + "step": 8607 + }, + { + "epoch": 0.28, + "learning_rate": 1.7007430180875298e-05, + "loss": 1.0405, + "step": 8608 + }, + { + "epoch": 0.28, + "learning_rate": 1.700669118664447e-05, + "loss": 1.0625, + "step": 8609 + }, + { + "epoch": 0.28, + "learning_rate": 1.7005952117239597e-05, + "loss": 1.0312, + "step": 8610 + }, + { + "epoch": 0.28, + "learning_rate": 1.7005212972668615e-05, + "loss": 0.9697, + "step": 8611 + }, + { + "epoch": 0.28, + "learning_rate": 1.7004473752939448e-05, + "loss": 1.0605, + "step": 8612 + }, + { + "epoch": 0.28, + "learning_rate": 1.700373445806003e-05, + "loss": 1.1074, + "step": 8613 + }, + { + "epoch": 0.28, + "learning_rate": 1.7002995088038293e-05, + "loss": 1.0366, + "step": 8614 + }, + { + "epoch": 0.28, + "learning_rate": 1.700225564288217e-05, + "loss": 0.9292, + "step": 8615 + }, + { + "epoch": 0.28, + "learning_rate": 1.7001516122599593e-05, + "loss": 1.0093, + "step": 8616 + }, + { + "epoch": 0.28, + "learning_rate": 1.7000776527198496e-05, + "loss": 0.9282, + "step": 8617 + }, + { + "epoch": 0.28, + "learning_rate": 1.7000036856686817e-05, + "loss": 0.9985, + "step": 8618 + }, + { + "epoch": 0.28, + "learning_rate": 1.6999297111072488e-05, + "loss": 1.082, + "step": 8619 + }, + { + "epoch": 0.28, + "learning_rate": 1.699855729036345e-05, + "loss": 1.0767, + "step": 8620 + }, + { + "epoch": 0.28, + "learning_rate": 1.6997817394567634e-05, + "loss": 1.0142, + "step": 8621 + }, + { + "epoch": 0.28, + "learning_rate": 1.6997077423692986e-05, + "loss": 1.1279, + "step": 8622 + }, + { + "epoch": 0.28, + "learning_rate": 1.6996337377747438e-05, + "loss": 0.9697, + "step": 8623 + }, + { + "epoch": 0.28, + "learning_rate": 1.6995597256738934e-05, + "loss": 1.0542, + "step": 8624 + }, + { + "epoch": 0.28, + "learning_rate": 1.6994857060675414e-05, + "loss": 1.062, + "step": 8625 + }, + { + "epoch": 0.28, + "learning_rate": 1.6994116789564822e-05, + "loss": 1.145, + "step": 8626 + }, + { + "epoch": 0.28, + "learning_rate": 1.6993376443415094e-05, + "loss": 1.1475, + "step": 8627 + }, + { + "epoch": 0.28, + "learning_rate": 1.6992636022234177e-05, + "loss": 1.1401, + "step": 8628 + }, + { + "epoch": 0.28, + "learning_rate": 1.6991895526030014e-05, + "loss": 0.9189, + "step": 8629 + }, + { + "epoch": 0.28, + "learning_rate": 1.6991154954810553e-05, + "loss": 1.0693, + "step": 8630 + }, + { + "epoch": 0.28, + "learning_rate": 1.6990414308583733e-05, + "loss": 1.0986, + "step": 8631 + }, + { + "epoch": 0.28, + "learning_rate": 1.6989673587357508e-05, + "loss": 1.0371, + "step": 8632 + }, + { + "epoch": 0.28, + "learning_rate": 1.6988932791139818e-05, + "loss": 0.9849, + "step": 8633 + }, + { + "epoch": 0.28, + "learning_rate": 1.6988191919938618e-05, + "loss": 1.0278, + "step": 8634 + }, + { + "epoch": 0.28, + "learning_rate": 1.698745097376185e-05, + "loss": 1.188, + "step": 8635 + }, + { + "epoch": 0.28, + "learning_rate": 1.6986709952617467e-05, + "loss": 1.0444, + "step": 8636 + }, + { + "epoch": 0.28, + "learning_rate": 1.6985968856513417e-05, + "loss": 0.9761, + "step": 8637 + }, + { + "epoch": 0.28, + "learning_rate": 1.6985227685457654e-05, + "loss": 1.0032, + "step": 8638 + }, + { + "epoch": 0.28, + "learning_rate": 1.698448643945813e-05, + "loss": 1.1504, + "step": 8639 + }, + { + "epoch": 0.28, + "learning_rate": 1.6983745118522797e-05, + "loss": 1.0645, + "step": 8640 + }, + { + "epoch": 0.28, + "learning_rate": 1.6983003722659604e-05, + "loss": 0.9756, + "step": 8641 + }, + { + "epoch": 0.28, + "learning_rate": 1.6982262251876513e-05, + "loss": 1.0283, + "step": 8642 + }, + { + "epoch": 0.28, + "learning_rate": 1.6981520706181476e-05, + "loss": 1.0063, + "step": 8643 + }, + { + "epoch": 0.28, + "learning_rate": 1.6980779085582448e-05, + "loss": 1.0996, + "step": 8644 + }, + { + "epoch": 0.28, + "learning_rate": 1.6980037390087388e-05, + "loss": 1.124, + "step": 8645 + }, + { + "epoch": 0.28, + "learning_rate": 1.6979295619704247e-05, + "loss": 1.1343, + "step": 8646 + }, + { + "epoch": 0.28, + "learning_rate": 1.6978553774440992e-05, + "loss": 1.0508, + "step": 8647 + }, + { + "epoch": 0.28, + "learning_rate": 1.6977811854305578e-05, + "loss": 1.0889, + "step": 8648 + }, + { + "epoch": 0.28, + "learning_rate": 1.6977069859305966e-05, + "loss": 0.9785, + "step": 8649 + }, + { + "epoch": 0.28, + "learning_rate": 1.6976327789450112e-05, + "loss": 0.958, + "step": 8650 + }, + { + "epoch": 0.28, + "learning_rate": 1.6975585644745985e-05, + "loss": 1.085, + "step": 8651 + }, + { + "epoch": 0.28, + "learning_rate": 1.6974843425201547e-05, + "loss": 0.9619, + "step": 8652 + }, + { + "epoch": 0.28, + "learning_rate": 1.6974101130824757e-05, + "loss": 1.0352, + "step": 8653 + }, + { + "epoch": 0.28, + "learning_rate": 1.6973358761623577e-05, + "loss": 1.0376, + "step": 8654 + }, + { + "epoch": 0.28, + "learning_rate": 1.6972616317605976e-05, + "loss": 1.1157, + "step": 8655 + }, + { + "epoch": 0.28, + "learning_rate": 1.6971873798779922e-05, + "loss": 0.9775, + "step": 8656 + }, + { + "epoch": 0.28, + "learning_rate": 1.6971131205153378e-05, + "loss": 1.0195, + "step": 8657 + }, + { + "epoch": 0.28, + "learning_rate": 1.697038853673431e-05, + "loss": 1.0151, + "step": 8658 + }, + { + "epoch": 0.28, + "learning_rate": 1.6969645793530688e-05, + "loss": 1.0215, + "step": 8659 + }, + { + "epoch": 0.28, + "learning_rate": 1.6968902975550476e-05, + "loss": 0.8906, + "step": 8660 + }, + { + "epoch": 0.28, + "learning_rate": 1.6968160082801653e-05, + "loss": 1.0337, + "step": 8661 + }, + { + "epoch": 0.28, + "learning_rate": 1.696741711529218e-05, + "loss": 1.0786, + "step": 8662 + }, + { + "epoch": 0.28, + "learning_rate": 1.6966674073030036e-05, + "loss": 1.1567, + "step": 8663 + }, + { + "epoch": 0.28, + "learning_rate": 1.6965930956023192e-05, + "loss": 0.9404, + "step": 8664 + }, + { + "epoch": 0.28, + "learning_rate": 1.6965187764279613e-05, + "loss": 0.9683, + "step": 8665 + }, + { + "epoch": 0.28, + "learning_rate": 1.696444449780728e-05, + "loss": 0.8896, + "step": 8666 + }, + { + "epoch": 0.28, + "learning_rate": 1.6963701156614167e-05, + "loss": 1.0562, + "step": 8667 + }, + { + "epoch": 0.28, + "learning_rate": 1.6962957740708245e-05, + "loss": 1.1055, + "step": 8668 + }, + { + "epoch": 0.28, + "learning_rate": 1.6962214250097492e-05, + "loss": 0.9805, + "step": 8669 + }, + { + "epoch": 0.28, + "learning_rate": 1.696147068478989e-05, + "loss": 0.9639, + "step": 8670 + }, + { + "epoch": 0.28, + "learning_rate": 1.6960727044793412e-05, + "loss": 1.0171, + "step": 8671 + }, + { + "epoch": 0.28, + "learning_rate": 1.695998333011603e-05, + "loss": 1.105, + "step": 8672 + }, + { + "epoch": 0.28, + "learning_rate": 1.6959239540765734e-05, + "loss": 1.0693, + "step": 8673 + }, + { + "epoch": 0.28, + "learning_rate": 1.6958495676750502e-05, + "loss": 1.0542, + "step": 8674 + }, + { + "epoch": 0.28, + "learning_rate": 1.695775173807831e-05, + "loss": 0.9507, + "step": 8675 + }, + { + "epoch": 0.28, + "learning_rate": 1.695700772475714e-05, + "loss": 0.9751, + "step": 8676 + }, + { + "epoch": 0.28, + "learning_rate": 1.695626363679498e-05, + "loss": 0.8359, + "step": 8677 + }, + { + "epoch": 0.28, + "learning_rate": 1.695551947419981e-05, + "loss": 1.2065, + "step": 8678 + }, + { + "epoch": 0.28, + "learning_rate": 1.6954775236979616e-05, + "loss": 1.0757, + "step": 8679 + }, + { + "epoch": 0.28, + "learning_rate": 1.6954030925142374e-05, + "loss": 1.0835, + "step": 8680 + }, + { + "epoch": 0.28, + "learning_rate": 1.6953286538696082e-05, + "loss": 0.9507, + "step": 8681 + }, + { + "epoch": 0.28, + "learning_rate": 1.6952542077648722e-05, + "loss": 1.0894, + "step": 8682 + }, + { + "epoch": 0.28, + "learning_rate": 1.6951797542008276e-05, + "loss": 1.1001, + "step": 8683 + }, + { + "epoch": 0.28, + "learning_rate": 1.695105293178274e-05, + "loss": 0.9336, + "step": 8684 + }, + { + "epoch": 0.28, + "learning_rate": 1.6950308246980097e-05, + "loss": 0.9824, + "step": 8685 + }, + { + "epoch": 0.28, + "learning_rate": 1.694956348760834e-05, + "loss": 0.9888, + "step": 8686 + }, + { + "epoch": 0.28, + "learning_rate": 1.6948818653675456e-05, + "loss": 1.042, + "step": 8687 + }, + { + "epoch": 0.28, + "learning_rate": 1.694807374518944e-05, + "loss": 0.9805, + "step": 8688 + }, + { + "epoch": 0.28, + "learning_rate": 1.694732876215828e-05, + "loss": 0.5591, + "step": 8689 + }, + { + "epoch": 0.28, + "learning_rate": 1.6946583704589973e-05, + "loss": 1.0259, + "step": 8690 + }, + { + "epoch": 0.28, + "learning_rate": 1.6945838572492513e-05, + "loss": 1.0967, + "step": 8691 + }, + { + "epoch": 0.28, + "learning_rate": 1.694509336587389e-05, + "loss": 1.0464, + "step": 8692 + }, + { + "epoch": 0.28, + "learning_rate": 1.69443480847421e-05, + "loss": 1.2119, + "step": 8693 + }, + { + "epoch": 0.28, + "learning_rate": 1.6943602729105148e-05, + "loss": 1.0669, + "step": 8694 + }, + { + "epoch": 0.28, + "learning_rate": 1.6942857298971015e-05, + "loss": 1.166, + "step": 8695 + }, + { + "epoch": 0.28, + "learning_rate": 1.6942111794347713e-05, + "loss": 1.0454, + "step": 8696 + }, + { + "epoch": 0.28, + "learning_rate": 1.6941366215243232e-05, + "loss": 1.1172, + "step": 8697 + }, + { + "epoch": 0.28, + "learning_rate": 1.6940620561665575e-05, + "loss": 1.0815, + "step": 8698 + }, + { + "epoch": 0.28, + "learning_rate": 1.6939874833622738e-05, + "loss": 1.0776, + "step": 8699 + }, + { + "epoch": 0.28, + "learning_rate": 1.6939129031122726e-05, + "loss": 1.0884, + "step": 8700 + }, + { + "epoch": 0.28, + "learning_rate": 1.6938383154173543e-05, + "loss": 1.0845, + "step": 8701 + }, + { + "epoch": 0.28, + "learning_rate": 1.693763720278318e-05, + "loss": 1.0273, + "step": 8702 + }, + { + "epoch": 0.28, + "learning_rate": 1.6936891176959656e-05, + "loss": 1.1514, + "step": 8703 + }, + { + "epoch": 0.28, + "learning_rate": 1.6936145076710965e-05, + "loss": 1.0127, + "step": 8704 + }, + { + "epoch": 0.28, + "learning_rate": 1.693539890204511e-05, + "loss": 1.061, + "step": 8705 + }, + { + "epoch": 0.28, + "learning_rate": 1.6934652652970104e-05, + "loss": 1.0762, + "step": 8706 + }, + { + "epoch": 0.28, + "learning_rate": 1.693390632949395e-05, + "loss": 1.125, + "step": 8707 + }, + { + "epoch": 0.28, + "learning_rate": 1.6933159931624654e-05, + "loss": 0.9971, + "step": 8708 + }, + { + "epoch": 0.28, + "learning_rate": 1.6932413459370226e-05, + "loss": 1.1313, + "step": 8709 + }, + { + "epoch": 0.28, + "learning_rate": 1.6931666912738675e-05, + "loss": 1.1309, + "step": 8710 + }, + { + "epoch": 0.28, + "learning_rate": 1.6930920291738004e-05, + "loss": 1.0894, + "step": 8711 + }, + { + "epoch": 0.28, + "learning_rate": 1.6930173596376236e-05, + "loss": 1.0996, + "step": 8712 + }, + { + "epoch": 0.28, + "learning_rate": 1.6929426826661373e-05, + "loss": 1.0659, + "step": 8713 + }, + { + "epoch": 0.28, + "learning_rate": 1.692867998260143e-05, + "loss": 1.002, + "step": 8714 + }, + { + "epoch": 0.28, + "learning_rate": 1.6927933064204414e-05, + "loss": 0.8843, + "step": 8715 + }, + { + "epoch": 0.28, + "learning_rate": 1.692718607147835e-05, + "loss": 1.0908, + "step": 8716 + }, + { + "epoch": 0.28, + "learning_rate": 1.692643900443124e-05, + "loss": 1.0713, + "step": 8717 + }, + { + "epoch": 0.28, + "learning_rate": 1.692569186307111e-05, + "loss": 0.8901, + "step": 8718 + }, + { + "epoch": 0.28, + "learning_rate": 1.692494464740597e-05, + "loss": 1.1147, + "step": 8719 + }, + { + "epoch": 0.28, + "learning_rate": 1.692419735744384e-05, + "loss": 1.1147, + "step": 8720 + }, + { + "epoch": 0.28, + "learning_rate": 1.692344999319273e-05, + "loss": 0.9829, + "step": 8721 + }, + { + "epoch": 0.28, + "learning_rate": 1.692270255466067e-05, + "loss": 1.0605, + "step": 8722 + }, + { + "epoch": 0.28, + "learning_rate": 1.692195504185567e-05, + "loss": 1.0986, + "step": 8723 + }, + { + "epoch": 0.28, + "learning_rate": 1.6921207454785754e-05, + "loss": 1.0698, + "step": 8724 + }, + { + "epoch": 0.28, + "learning_rate": 1.6920459793458946e-05, + "loss": 1.2017, + "step": 8725 + }, + { + "epoch": 0.28, + "learning_rate": 1.6919712057883256e-05, + "loss": 1.0146, + "step": 8726 + }, + { + "epoch": 0.28, + "learning_rate": 1.691896424806672e-05, + "loss": 1.1138, + "step": 8727 + }, + { + "epoch": 0.28, + "learning_rate": 1.6918216364017356e-05, + "loss": 1.0098, + "step": 8728 + }, + { + "epoch": 0.28, + "learning_rate": 1.691746840574318e-05, + "loss": 1.0581, + "step": 8729 + }, + { + "epoch": 0.28, + "learning_rate": 1.6916720373252232e-05, + "loss": 1.0859, + "step": 8730 + }, + { + "epoch": 0.28, + "learning_rate": 1.6915972266552522e-05, + "loss": 1.0498, + "step": 8731 + }, + { + "epoch": 0.28, + "learning_rate": 1.691522408565209e-05, + "loss": 1.0093, + "step": 8732 + }, + { + "epoch": 0.28, + "learning_rate": 1.6914475830558953e-05, + "loss": 1.0693, + "step": 8733 + }, + { + "epoch": 0.28, + "learning_rate": 1.6913727501281143e-05, + "loss": 1.1177, + "step": 8734 + }, + { + "epoch": 0.28, + "learning_rate": 1.6912979097826693e-05, + "loss": 1.0347, + "step": 8735 + }, + { + "epoch": 0.28, + "learning_rate": 1.6912230620203622e-05, + "loss": 1.0713, + "step": 8736 + }, + { + "epoch": 0.28, + "learning_rate": 1.691148206841997e-05, + "loss": 0.9292, + "step": 8737 + }, + { + "epoch": 0.28, + "learning_rate": 1.6910733442483765e-05, + "loss": 1.0786, + "step": 8738 + }, + { + "epoch": 0.28, + "learning_rate": 1.690998474240304e-05, + "loss": 1.0835, + "step": 8739 + }, + { + "epoch": 0.28, + "learning_rate": 1.690923596818582e-05, + "loss": 1.0464, + "step": 8740 + }, + { + "epoch": 0.28, + "learning_rate": 1.690848711984015e-05, + "loss": 1.0371, + "step": 8741 + }, + { + "epoch": 0.28, + "learning_rate": 1.6907738197374057e-05, + "loss": 0.5786, + "step": 8742 + }, + { + "epoch": 0.28, + "learning_rate": 1.6906989200795583e-05, + "loss": 1.02, + "step": 8743 + }, + { + "epoch": 0.28, + "learning_rate": 1.6906240130112755e-05, + "loss": 1.002, + "step": 8744 + }, + { + "epoch": 0.28, + "learning_rate": 1.6905490985333612e-05, + "loss": 1.2798, + "step": 8745 + }, + { + "epoch": 0.28, + "learning_rate": 1.69047417664662e-05, + "loss": 0.9316, + "step": 8746 + }, + { + "epoch": 0.28, + "learning_rate": 1.6903992473518544e-05, + "loss": 1.1543, + "step": 8747 + }, + { + "epoch": 0.28, + "learning_rate": 1.690324310649869e-05, + "loss": 0.998, + "step": 8748 + }, + { + "epoch": 0.28, + "learning_rate": 1.690249366541468e-05, + "loss": 1.062, + "step": 8749 + }, + { + "epoch": 0.28, + "learning_rate": 1.6901744150274553e-05, + "loss": 0.9297, + "step": 8750 + }, + { + "epoch": 0.28, + "learning_rate": 1.6900994561086347e-05, + "loss": 1.1396, + "step": 8751 + }, + { + "epoch": 0.28, + "learning_rate": 1.690024489785811e-05, + "loss": 1.1572, + "step": 8752 + }, + { + "epoch": 0.28, + "learning_rate": 1.689949516059788e-05, + "loss": 1.0718, + "step": 8753 + }, + { + "epoch": 0.28, + "learning_rate": 1.6898745349313705e-05, + "loss": 1.0244, + "step": 8754 + }, + { + "epoch": 0.28, + "learning_rate": 1.6897995464013627e-05, + "loss": 1.0044, + "step": 8755 + }, + { + "epoch": 0.28, + "learning_rate": 1.6897245504705687e-05, + "loss": 1.0947, + "step": 8756 + }, + { + "epoch": 0.28, + "learning_rate": 1.689649547139794e-05, + "loss": 1.082, + "step": 8757 + }, + { + "epoch": 0.28, + "learning_rate": 1.689574536409843e-05, + "loss": 1.043, + "step": 8758 + }, + { + "epoch": 0.28, + "learning_rate": 1.6894995182815205e-05, + "loss": 1.0732, + "step": 8759 + }, + { + "epoch": 0.28, + "learning_rate": 1.6894244927556313e-05, + "loss": 1.0459, + "step": 8760 + }, + { + "epoch": 0.28, + "learning_rate": 1.68934945983298e-05, + "loss": 0.9741, + "step": 8761 + }, + { + "epoch": 0.28, + "learning_rate": 1.689274419514372e-05, + "loss": 1.0703, + "step": 8762 + }, + { + "epoch": 0.28, + "learning_rate": 1.6891993718006127e-05, + "loss": 1.002, + "step": 8763 + }, + { + "epoch": 0.28, + "learning_rate": 1.6891243166925067e-05, + "loss": 0.9653, + "step": 8764 + }, + { + "epoch": 0.28, + "learning_rate": 1.689049254190859e-05, + "loss": 1.0059, + "step": 8765 + }, + { + "epoch": 0.28, + "learning_rate": 1.688974184296476e-05, + "loss": 1.0894, + "step": 8766 + }, + { + "epoch": 0.28, + "learning_rate": 1.6888991070101625e-05, + "loss": 1.0298, + "step": 8767 + }, + { + "epoch": 0.28, + "learning_rate": 1.6888240223327238e-05, + "loss": 1.002, + "step": 8768 + }, + { + "epoch": 0.28, + "learning_rate": 1.6887489302649657e-05, + "loss": 1.0723, + "step": 8769 + }, + { + "epoch": 0.28, + "learning_rate": 1.6886738308076938e-05, + "loss": 1.0298, + "step": 8770 + }, + { + "epoch": 0.28, + "learning_rate": 1.688598723961714e-05, + "loss": 1.1118, + "step": 8771 + }, + { + "epoch": 0.28, + "learning_rate": 1.688523609727832e-05, + "loss": 1.0396, + "step": 8772 + }, + { + "epoch": 0.28, + "learning_rate": 1.6884484881068534e-05, + "loss": 0.9727, + "step": 8773 + }, + { + "epoch": 0.28, + "learning_rate": 1.688373359099585e-05, + "loss": 1.1006, + "step": 8774 + }, + { + "epoch": 0.28, + "learning_rate": 1.6882982227068318e-05, + "loss": 1.1035, + "step": 8775 + }, + { + "epoch": 0.28, + "learning_rate": 1.6882230789294006e-05, + "loss": 1.0747, + "step": 8776 + }, + { + "epoch": 0.28, + "learning_rate": 1.6881479277680975e-05, + "loss": 0.9939, + "step": 8777 + }, + { + "epoch": 0.28, + "learning_rate": 1.688072769223729e-05, + "loss": 1.0576, + "step": 8778 + }, + { + "epoch": 0.28, + "learning_rate": 1.6879976032971007e-05, + "loss": 1.0229, + "step": 8779 + }, + { + "epoch": 0.28, + "learning_rate": 1.6879224299890197e-05, + "loss": 0.9893, + "step": 8780 + }, + { + "epoch": 0.28, + "learning_rate": 1.6878472493002928e-05, + "loss": 1.0796, + "step": 8781 + }, + { + "epoch": 0.28, + "learning_rate": 1.6877720612317255e-05, + "loss": 1.0728, + "step": 8782 + }, + { + "epoch": 0.28, + "learning_rate": 1.6876968657841258e-05, + "loss": 0.981, + "step": 8783 + }, + { + "epoch": 0.28, + "learning_rate": 1.6876216629582995e-05, + "loss": 0.8892, + "step": 8784 + }, + { + "epoch": 0.28, + "learning_rate": 1.6875464527550537e-05, + "loss": 0.9785, + "step": 8785 + }, + { + "epoch": 0.28, + "learning_rate": 1.6874712351751955e-05, + "loss": 0.9995, + "step": 8786 + }, + { + "epoch": 0.28, + "learning_rate": 1.687396010219532e-05, + "loss": 1.147, + "step": 8787 + }, + { + "epoch": 0.28, + "learning_rate": 1.68732077788887e-05, + "loss": 1.1367, + "step": 8788 + }, + { + "epoch": 0.28, + "learning_rate": 1.6872455381840164e-05, + "loss": 1.0176, + "step": 8789 + }, + { + "epoch": 0.28, + "learning_rate": 1.687170291105779e-05, + "loss": 1.0327, + "step": 8790 + }, + { + "epoch": 0.28, + "learning_rate": 1.6870950366549654e-05, + "loss": 1.1006, + "step": 8791 + }, + { + "epoch": 0.28, + "learning_rate": 1.687019774832382e-05, + "loss": 1.1162, + "step": 8792 + }, + { + "epoch": 0.28, + "learning_rate": 1.6869445056388367e-05, + "loss": 1.0757, + "step": 8793 + }, + { + "epoch": 0.28, + "learning_rate": 1.6868692290751372e-05, + "loss": 0.9966, + "step": 8794 + }, + { + "epoch": 0.28, + "learning_rate": 1.6867939451420912e-05, + "loss": 1.1089, + "step": 8795 + }, + { + "epoch": 0.28, + "learning_rate": 1.6867186538405063e-05, + "loss": 1.0522, + "step": 8796 + }, + { + "epoch": 0.28, + "learning_rate": 1.68664335517119e-05, + "loss": 0.9917, + "step": 8797 + }, + { + "epoch": 0.28, + "learning_rate": 1.6865680491349507e-05, + "loss": 1.1704, + "step": 8798 + }, + { + "epoch": 0.28, + "learning_rate": 1.686492735732596e-05, + "loss": 1.0215, + "step": 8799 + }, + { + "epoch": 0.28, + "learning_rate": 1.686417414964934e-05, + "loss": 1.1484, + "step": 8800 + }, + { + "epoch": 0.28, + "learning_rate": 1.6863420868327728e-05, + "loss": 1.0332, + "step": 8801 + }, + { + "epoch": 0.28, + "learning_rate": 1.6862667513369207e-05, + "loss": 0.9893, + "step": 8802 + }, + { + "epoch": 0.28, + "learning_rate": 1.6861914084781856e-05, + "loss": 0.9951, + "step": 8803 + }, + { + "epoch": 0.28, + "learning_rate": 1.6861160582573765e-05, + "loss": 0.8311, + "step": 8804 + }, + { + "epoch": 0.28, + "learning_rate": 1.6860407006753012e-05, + "loss": 1.0601, + "step": 8805 + }, + { + "epoch": 0.28, + "learning_rate": 1.6859653357327685e-05, + "loss": 1.0317, + "step": 8806 + }, + { + "epoch": 0.28, + "learning_rate": 1.685889963430587e-05, + "loss": 0.9883, + "step": 8807 + }, + { + "epoch": 0.28, + "learning_rate": 1.6858145837695653e-05, + "loss": 1.0342, + "step": 8808 + }, + { + "epoch": 0.28, + "learning_rate": 1.685739196750512e-05, + "loss": 1.0, + "step": 8809 + }, + { + "epoch": 0.28, + "learning_rate": 1.6856638023742365e-05, + "loss": 0.9858, + "step": 8810 + }, + { + "epoch": 0.28, + "learning_rate": 1.6855884006415467e-05, + "loss": 1.0498, + "step": 8811 + }, + { + "epoch": 0.28, + "learning_rate": 1.6855129915532525e-05, + "loss": 1.0205, + "step": 8812 + }, + { + "epoch": 0.28, + "learning_rate": 1.6854375751101625e-05, + "loss": 1.0991, + "step": 8813 + }, + { + "epoch": 0.28, + "learning_rate": 1.6853621513130857e-05, + "loss": 1.0078, + "step": 8814 + }, + { + "epoch": 0.28, + "learning_rate": 1.6852867201628318e-05, + "loss": 1.0767, + "step": 8815 + }, + { + "epoch": 0.28, + "learning_rate": 1.6852112816602096e-05, + "loss": 0.9014, + "step": 8816 + }, + { + "epoch": 0.28, + "learning_rate": 1.685135835806029e-05, + "loss": 1.0234, + "step": 8817 + }, + { + "epoch": 0.28, + "learning_rate": 1.6850603826010988e-05, + "loss": 1.0508, + "step": 8818 + }, + { + "epoch": 0.28, + "learning_rate": 1.6849849220462293e-05, + "loss": 0.9648, + "step": 8819 + }, + { + "epoch": 0.28, + "learning_rate": 1.6849094541422296e-05, + "loss": 0.9253, + "step": 8820 + }, + { + "epoch": 0.28, + "learning_rate": 1.684833978889909e-05, + "loss": 0.9756, + "step": 8821 + }, + { + "epoch": 0.28, + "learning_rate": 1.6847584962900784e-05, + "loss": 1.0405, + "step": 8822 + }, + { + "epoch": 0.28, + "learning_rate": 1.6846830063435468e-05, + "loss": 1.0889, + "step": 8823 + }, + { + "epoch": 0.28, + "learning_rate": 1.684607509051124e-05, + "loss": 1.043, + "step": 8824 + }, + { + "epoch": 0.28, + "learning_rate": 1.6845320044136204e-05, + "loss": 1.1812, + "step": 8825 + }, + { + "epoch": 0.28, + "learning_rate": 1.6844564924318462e-05, + "loss": 1.0137, + "step": 8826 + }, + { + "epoch": 0.28, + "learning_rate": 1.6843809731066115e-05, + "loss": 1.125, + "step": 8827 + }, + { + "epoch": 0.28, + "learning_rate": 1.684305446438726e-05, + "loss": 1.0918, + "step": 8828 + }, + { + "epoch": 0.28, + "learning_rate": 1.6842299124290003e-05, + "loss": 1.166, + "step": 8829 + }, + { + "epoch": 0.28, + "learning_rate": 1.6841543710782455e-05, + "loss": 0.9497, + "step": 8830 + }, + { + "epoch": 0.28, + "learning_rate": 1.6840788223872712e-05, + "loss": 1.0591, + "step": 8831 + }, + { + "epoch": 0.28, + "learning_rate": 1.684003266356888e-05, + "loss": 0.9932, + "step": 8832 + }, + { + "epoch": 0.28, + "learning_rate": 1.683927702987907e-05, + "loss": 0.9766, + "step": 8833 + }, + { + "epoch": 0.28, + "learning_rate": 1.6838521322811387e-05, + "loss": 1.1562, + "step": 8834 + }, + { + "epoch": 0.28, + "learning_rate": 1.6837765542373937e-05, + "loss": 0.9424, + "step": 8835 + }, + { + "epoch": 0.28, + "learning_rate": 1.683700968857483e-05, + "loss": 1.1387, + "step": 8836 + }, + { + "epoch": 0.28, + "learning_rate": 1.683625376142218e-05, + "loss": 0.9976, + "step": 8837 + }, + { + "epoch": 0.28, + "learning_rate": 1.683549776092409e-05, + "loss": 0.9043, + "step": 8838 + }, + { + "epoch": 0.28, + "learning_rate": 1.6834741687088673e-05, + "loss": 1.0249, + "step": 8839 + }, + { + "epoch": 0.28, + "learning_rate": 1.6833985539924046e-05, + "loss": 1.2065, + "step": 8840 + }, + { + "epoch": 0.28, + "learning_rate": 1.6833229319438316e-05, + "loss": 1.1113, + "step": 8841 + }, + { + "epoch": 0.28, + "learning_rate": 1.6832473025639597e-05, + "loss": 1.0908, + "step": 8842 + }, + { + "epoch": 0.28, + "learning_rate": 1.6831716658536004e-05, + "loss": 0.9863, + "step": 8843 + }, + { + "epoch": 0.28, + "learning_rate": 1.6830960218135652e-05, + "loss": 1.0205, + "step": 8844 + }, + { + "epoch": 0.28, + "learning_rate": 1.683020370444666e-05, + "loss": 1.0718, + "step": 8845 + }, + { + "epoch": 0.28, + "learning_rate": 1.6829447117477138e-05, + "loss": 1.0166, + "step": 8846 + }, + { + "epoch": 0.28, + "learning_rate": 1.6828690457235208e-05, + "loss": 1.0679, + "step": 8847 + }, + { + "epoch": 0.28, + "learning_rate": 1.6827933723728988e-05, + "loss": 1.0288, + "step": 8848 + }, + { + "epoch": 0.28, + "learning_rate": 1.6827176916966594e-05, + "loss": 1.0942, + "step": 8849 + }, + { + "epoch": 0.28, + "learning_rate": 1.682642003695615e-05, + "loss": 1.022, + "step": 8850 + }, + { + "epoch": 0.28, + "learning_rate": 1.6825663083705774e-05, + "loss": 0.5535, + "step": 8851 + }, + { + "epoch": 0.28, + "learning_rate": 1.6824906057223586e-05, + "loss": 0.9624, + "step": 8852 + }, + { + "epoch": 0.28, + "learning_rate": 1.682414895751771e-05, + "loss": 0.9702, + "step": 8853 + }, + { + "epoch": 0.28, + "learning_rate": 1.682339178459627e-05, + "loss": 1.1318, + "step": 8854 + }, + { + "epoch": 0.28, + "learning_rate": 1.6822634538467383e-05, + "loss": 1.0518, + "step": 8855 + }, + { + "epoch": 0.28, + "learning_rate": 1.6821877219139183e-05, + "loss": 1.1323, + "step": 8856 + }, + { + "epoch": 0.28, + "learning_rate": 1.6821119826619788e-05, + "loss": 0.9688, + "step": 8857 + }, + { + "epoch": 0.28, + "learning_rate": 1.6820362360917325e-05, + "loss": 0.9878, + "step": 8858 + }, + { + "epoch": 0.28, + "learning_rate": 1.6819604822039924e-05, + "loss": 1.0117, + "step": 8859 + }, + { + "epoch": 0.28, + "learning_rate": 1.6818847209995715e-05, + "loss": 1.0796, + "step": 8860 + }, + { + "epoch": 0.28, + "learning_rate": 1.681808952479282e-05, + "loss": 1.0771, + "step": 8861 + }, + { + "epoch": 0.28, + "learning_rate": 1.6817331766439365e-05, + "loss": 1.062, + "step": 8862 + }, + { + "epoch": 0.28, + "learning_rate": 1.681657393494349e-05, + "loss": 1.1172, + "step": 8863 + }, + { + "epoch": 0.28, + "learning_rate": 1.681581603031332e-05, + "loss": 1.1602, + "step": 8864 + }, + { + "epoch": 0.28, + "learning_rate": 1.6815058052556988e-05, + "loss": 1.1021, + "step": 8865 + }, + { + "epoch": 0.28, + "learning_rate": 1.6814300001682623e-05, + "loss": 0.9409, + "step": 8866 + }, + { + "epoch": 0.28, + "learning_rate": 1.6813541877698365e-05, + "loss": 0.9458, + "step": 8867 + }, + { + "epoch": 0.28, + "learning_rate": 1.681278368061234e-05, + "loss": 1.0649, + "step": 8868 + }, + { + "epoch": 0.28, + "learning_rate": 1.6812025410432687e-05, + "loss": 1.1279, + "step": 8869 + }, + { + "epoch": 0.28, + "learning_rate": 1.681126706716754e-05, + "loss": 1.0264, + "step": 8870 + }, + { + "epoch": 0.28, + "learning_rate": 1.681050865082504e-05, + "loss": 1.1426, + "step": 8871 + }, + { + "epoch": 0.28, + "learning_rate": 1.6809750161413315e-05, + "loss": 1.1011, + "step": 8872 + }, + { + "epoch": 0.28, + "learning_rate": 1.6808991598940507e-05, + "loss": 0.8975, + "step": 8873 + }, + { + "epoch": 0.28, + "learning_rate": 1.6808232963414756e-05, + "loss": 1.1138, + "step": 8874 + }, + { + "epoch": 0.28, + "learning_rate": 1.6807474254844204e-05, + "loss": 1.1118, + "step": 8875 + }, + { + "epoch": 0.28, + "learning_rate": 1.6806715473236983e-05, + "loss": 0.9785, + "step": 8876 + }, + { + "epoch": 0.28, + "learning_rate": 1.680595661860124e-05, + "loss": 1.1318, + "step": 8877 + }, + { + "epoch": 0.28, + "learning_rate": 1.6805197690945112e-05, + "loss": 1.0566, + "step": 8878 + }, + { + "epoch": 0.28, + "learning_rate": 1.680443869027675e-05, + "loss": 1.1826, + "step": 8879 + }, + { + "epoch": 0.28, + "learning_rate": 1.6803679616604287e-05, + "loss": 1.0332, + "step": 8880 + }, + { + "epoch": 0.28, + "learning_rate": 1.6802920469935874e-05, + "loss": 0.9517, + "step": 8881 + }, + { + "epoch": 0.28, + "learning_rate": 1.6802161250279652e-05, + "loss": 1.0415, + "step": 8882 + }, + { + "epoch": 0.28, + "learning_rate": 1.680140195764377e-05, + "loss": 0.9805, + "step": 8883 + }, + { + "epoch": 0.28, + "learning_rate": 1.680064259203637e-05, + "loss": 1.0225, + "step": 8884 + }, + { + "epoch": 0.28, + "learning_rate": 1.67998831534656e-05, + "loss": 1.061, + "step": 8885 + }, + { + "epoch": 0.28, + "learning_rate": 1.6799123641939616e-05, + "loss": 1.0571, + "step": 8886 + }, + { + "epoch": 0.28, + "learning_rate": 1.6798364057466554e-05, + "loss": 1.0215, + "step": 8887 + }, + { + "epoch": 0.28, + "learning_rate": 1.6797604400054573e-05, + "loss": 1.0146, + "step": 8888 + }, + { + "epoch": 0.28, + "learning_rate": 1.679684466971182e-05, + "loss": 0.9028, + "step": 8889 + }, + { + "epoch": 0.28, + "learning_rate": 1.6796084866446444e-05, + "loss": 0.9834, + "step": 8890 + }, + { + "epoch": 0.28, + "learning_rate": 1.67953249902666e-05, + "loss": 0.9346, + "step": 8891 + }, + { + "epoch": 0.28, + "learning_rate": 1.679456504118044e-05, + "loss": 1.0386, + "step": 8892 + }, + { + "epoch": 0.28, + "learning_rate": 1.6793805019196113e-05, + "loss": 1.1499, + "step": 8893 + }, + { + "epoch": 0.28, + "learning_rate": 1.6793044924321783e-05, + "loss": 1.0962, + "step": 8894 + }, + { + "epoch": 0.28, + "learning_rate": 1.6792284756565598e-05, + "loss": 1.0615, + "step": 8895 + }, + { + "epoch": 0.28, + "learning_rate": 1.6791524515935712e-05, + "loss": 1.0234, + "step": 8896 + }, + { + "epoch": 0.28, + "learning_rate": 1.679076420244029e-05, + "loss": 1.0146, + "step": 8897 + }, + { + "epoch": 0.28, + "learning_rate": 1.6790003816087478e-05, + "loss": 1.1021, + "step": 8898 + }, + { + "epoch": 0.28, + "learning_rate": 1.678924335688544e-05, + "loss": 1.0229, + "step": 8899 + }, + { + "epoch": 0.28, + "learning_rate": 1.678848282484234e-05, + "loss": 1.0845, + "step": 8900 + }, + { + "epoch": 0.28, + "learning_rate": 1.6787722219966327e-05, + "loss": 0.9883, + "step": 8901 + }, + { + "epoch": 0.28, + "learning_rate": 1.678696154226557e-05, + "loss": 1.1489, + "step": 8902 + }, + { + "epoch": 0.28, + "learning_rate": 1.6786200791748223e-05, + "loss": 1.0957, + "step": 8903 + }, + { + "epoch": 0.28, + "learning_rate": 1.6785439968422456e-05, + "loss": 0.876, + "step": 8904 + }, + { + "epoch": 0.28, + "learning_rate": 1.6784679072296425e-05, + "loss": 1.1543, + "step": 8905 + }, + { + "epoch": 0.28, + "learning_rate": 1.67839181033783e-05, + "loss": 1.0288, + "step": 8906 + }, + { + "epoch": 0.28, + "learning_rate": 1.678315706167624e-05, + "loss": 1.0264, + "step": 8907 + }, + { + "epoch": 0.28, + "learning_rate": 1.6782395947198412e-05, + "loss": 0.9487, + "step": 8908 + }, + { + "epoch": 0.28, + "learning_rate": 1.678163475995298e-05, + "loss": 1.0854, + "step": 8909 + }, + { + "epoch": 0.28, + "learning_rate": 1.6780873499948116e-05, + "loss": 1.0669, + "step": 8910 + }, + { + "epoch": 0.28, + "learning_rate": 1.6780112167191983e-05, + "loss": 1.0771, + "step": 8911 + }, + { + "epoch": 0.29, + "learning_rate": 1.677935076169275e-05, + "loss": 1.0957, + "step": 8912 + }, + { + "epoch": 0.29, + "learning_rate": 1.6778589283458585e-05, + "loss": 1.1338, + "step": 8913 + }, + { + "epoch": 0.29, + "learning_rate": 1.677782773249766e-05, + "loss": 1.0664, + "step": 8914 + }, + { + "epoch": 0.29, + "learning_rate": 1.6777066108818146e-05, + "loss": 1.0195, + "step": 8915 + }, + { + "epoch": 0.29, + "learning_rate": 1.6776304412428214e-05, + "loss": 1.1147, + "step": 8916 + }, + { + "epoch": 0.29, + "learning_rate": 1.6775542643336032e-05, + "loss": 0.5789, + "step": 8917 + }, + { + "epoch": 0.29, + "learning_rate": 1.6774780801549778e-05, + "loss": 1.0425, + "step": 8918 + }, + { + "epoch": 0.29, + "learning_rate": 1.6774018887077624e-05, + "loss": 1.0044, + "step": 8919 + }, + { + "epoch": 0.29, + "learning_rate": 1.6773256899927746e-05, + "loss": 1.0171, + "step": 8920 + }, + { + "epoch": 0.29, + "learning_rate": 1.6772494840108317e-05, + "loss": 1.0176, + "step": 8921 + }, + { + "epoch": 0.29, + "learning_rate": 1.677173270762751e-05, + "loss": 0.98, + "step": 8922 + }, + { + "epoch": 0.29, + "learning_rate": 1.6770970502493508e-05, + "loss": 1.1875, + "step": 8923 + }, + { + "epoch": 0.29, + "learning_rate": 1.677020822471449e-05, + "loss": 0.9863, + "step": 8924 + }, + { + "epoch": 0.29, + "learning_rate": 1.6769445874298625e-05, + "loss": 1.1846, + "step": 8925 + }, + { + "epoch": 0.29, + "learning_rate": 1.67686834512541e-05, + "loss": 1.0562, + "step": 8926 + }, + { + "epoch": 0.29, + "learning_rate": 1.6767920955589094e-05, + "loss": 1.0747, + "step": 8927 + }, + { + "epoch": 0.29, + "learning_rate": 1.6767158387311784e-05, + "loss": 1.0742, + "step": 8928 + }, + { + "epoch": 0.29, + "learning_rate": 1.6766395746430355e-05, + "loss": 1.001, + "step": 8929 + }, + { + "epoch": 0.29, + "learning_rate": 1.6765633032952988e-05, + "loss": 1.0103, + "step": 8930 + }, + { + "epoch": 0.29, + "learning_rate": 1.676487024688787e-05, + "loss": 1.0342, + "step": 8931 + }, + { + "epoch": 0.29, + "learning_rate": 1.6764107388243176e-05, + "loss": 1.0264, + "step": 8932 + }, + { + "epoch": 0.29, + "learning_rate": 1.67633444570271e-05, + "loss": 1.0249, + "step": 8933 + }, + { + "epoch": 0.29, + "learning_rate": 1.676258145324782e-05, + "loss": 1.147, + "step": 8934 + }, + { + "epoch": 0.29, + "learning_rate": 1.676181837691353e-05, + "loss": 1.0488, + "step": 8935 + }, + { + "epoch": 0.29, + "learning_rate": 1.6761055228032404e-05, + "loss": 1.105, + "step": 8936 + }, + { + "epoch": 0.29, + "learning_rate": 1.6760292006612645e-05, + "loss": 1.0664, + "step": 8937 + }, + { + "epoch": 0.29, + "learning_rate": 1.6759528712662433e-05, + "loss": 1.103, + "step": 8938 + }, + { + "epoch": 0.29, + "learning_rate": 1.675876534618996e-05, + "loss": 1.0615, + "step": 8939 + }, + { + "epoch": 0.29, + "learning_rate": 1.6758001907203417e-05, + "loss": 1.042, + "step": 8940 + }, + { + "epoch": 0.29, + "learning_rate": 1.6757238395710987e-05, + "loss": 1.1323, + "step": 8941 + }, + { + "epoch": 0.29, + "learning_rate": 1.6756474811720872e-05, + "loss": 0.9824, + "step": 8942 + }, + { + "epoch": 0.29, + "learning_rate": 1.6755711155241257e-05, + "loss": 1.0, + "step": 8943 + }, + { + "epoch": 0.29, + "learning_rate": 1.6754947426280342e-05, + "loss": 1.0425, + "step": 8944 + }, + { + "epoch": 0.29, + "learning_rate": 1.6754183624846315e-05, + "loss": 1.292, + "step": 8945 + }, + { + "epoch": 0.29, + "learning_rate": 1.675341975094737e-05, + "loss": 1.1631, + "step": 8946 + }, + { + "epoch": 0.29, + "learning_rate": 1.675265580459171e-05, + "loss": 0.9775, + "step": 8947 + }, + { + "epoch": 0.29, + "learning_rate": 1.6751891785787525e-05, + "loss": 0.8804, + "step": 8948 + }, + { + "epoch": 0.29, + "learning_rate": 1.6751127694543012e-05, + "loss": 1.0972, + "step": 8949 + }, + { + "epoch": 0.29, + "learning_rate": 1.6750363530866373e-05, + "loss": 1.0132, + "step": 8950 + }, + { + "epoch": 0.29, + "learning_rate": 1.6749599294765806e-05, + "loss": 1.1172, + "step": 8951 + }, + { + "epoch": 0.29, + "learning_rate": 1.6748834986249505e-05, + "loss": 1.1172, + "step": 8952 + }, + { + "epoch": 0.29, + "learning_rate": 1.6748070605325676e-05, + "loss": 1.0381, + "step": 8953 + }, + { + "epoch": 0.29, + "learning_rate": 1.6747306152002517e-05, + "loss": 1.0293, + "step": 8954 + }, + { + "epoch": 0.29, + "learning_rate": 1.6746541626288228e-05, + "loss": 1.1143, + "step": 8955 + }, + { + "epoch": 0.29, + "learning_rate": 1.674577702819102e-05, + "loss": 1.0454, + "step": 8956 + }, + { + "epoch": 0.29, + "learning_rate": 1.6745012357719086e-05, + "loss": 1.0073, + "step": 8957 + }, + { + "epoch": 0.29, + "learning_rate": 1.6744247614880637e-05, + "loss": 1.042, + "step": 8958 + }, + { + "epoch": 0.29, + "learning_rate": 1.674348279968387e-05, + "loss": 0.9863, + "step": 8959 + }, + { + "epoch": 0.29, + "learning_rate": 1.6742717912137002e-05, + "loss": 1.0444, + "step": 8960 + }, + { + "epoch": 0.29, + "learning_rate": 1.6741952952248228e-05, + "loss": 1.1514, + "step": 8961 + }, + { + "epoch": 0.29, + "learning_rate": 1.6741187920025767e-05, + "loss": 0.9751, + "step": 8962 + }, + { + "epoch": 0.29, + "learning_rate": 1.6740422815477812e-05, + "loss": 1.1445, + "step": 8963 + }, + { + "epoch": 0.29, + "learning_rate": 1.6739657638612587e-05, + "loss": 1.1519, + "step": 8964 + }, + { + "epoch": 0.29, + "learning_rate": 1.6738892389438293e-05, + "loss": 1.0923, + "step": 8965 + }, + { + "epoch": 0.29, + "learning_rate": 1.673812706796314e-05, + "loss": 1.1177, + "step": 8966 + }, + { + "epoch": 0.29, + "learning_rate": 1.673736167419534e-05, + "loss": 0.9243, + "step": 8967 + }, + { + "epoch": 0.29, + "learning_rate": 1.6736596208143108e-05, + "loss": 0.5635, + "step": 8968 + }, + { + "epoch": 0.29, + "learning_rate": 1.6735830669814656e-05, + "loss": 0.9795, + "step": 8969 + }, + { + "epoch": 0.29, + "learning_rate": 1.6735065059218194e-05, + "loss": 1.1763, + "step": 8970 + }, + { + "epoch": 0.29, + "learning_rate": 1.6734299376361937e-05, + "loss": 1.0439, + "step": 8971 + }, + { + "epoch": 0.29, + "learning_rate": 1.6733533621254098e-05, + "loss": 1.0566, + "step": 8972 + }, + { + "epoch": 0.29, + "learning_rate": 1.6732767793902904e-05, + "loss": 1.0767, + "step": 8973 + }, + { + "epoch": 0.29, + "learning_rate": 1.6732001894316557e-05, + "loss": 1.1333, + "step": 8974 + }, + { + "epoch": 0.29, + "learning_rate": 1.673123592250328e-05, + "loss": 0.9692, + "step": 8975 + }, + { + "epoch": 0.29, + "learning_rate": 1.6730469878471294e-05, + "loss": 1.0059, + "step": 8976 + }, + { + "epoch": 0.29, + "learning_rate": 1.6729703762228813e-05, + "loss": 1.0679, + "step": 8977 + }, + { + "epoch": 0.29, + "learning_rate": 1.6728937573784062e-05, + "loss": 1.0952, + "step": 8978 + }, + { + "epoch": 0.29, + "learning_rate": 1.6728171313145252e-05, + "loss": 1.1543, + "step": 8979 + }, + { + "epoch": 0.29, + "learning_rate": 1.6727404980320615e-05, + "loss": 1.1465, + "step": 8980 + }, + { + "epoch": 0.29, + "learning_rate": 1.6726638575318368e-05, + "loss": 0.9814, + "step": 8981 + }, + { + "epoch": 0.29, + "learning_rate": 1.6725872098146733e-05, + "loss": 0.9565, + "step": 8982 + }, + { + "epoch": 0.29, + "learning_rate": 1.6725105548813935e-05, + "loss": 0.583, + "step": 8983 + }, + { + "epoch": 0.29, + "learning_rate": 1.6724338927328197e-05, + "loss": 1.022, + "step": 8984 + }, + { + "epoch": 0.29, + "learning_rate": 1.6723572233697742e-05, + "loss": 0.9941, + "step": 8985 + }, + { + "epoch": 0.29, + "learning_rate": 1.67228054679308e-05, + "loss": 0.9688, + "step": 8986 + }, + { + "epoch": 0.29, + "learning_rate": 1.6722038630035596e-05, + "loss": 1.061, + "step": 8987 + }, + { + "epoch": 0.29, + "learning_rate": 1.672127172002036e-05, + "loss": 1.0581, + "step": 8988 + }, + { + "epoch": 0.29, + "learning_rate": 1.672050473789331e-05, + "loss": 1.0459, + "step": 8989 + }, + { + "epoch": 0.29, + "learning_rate": 1.671973768366269e-05, + "loss": 0.9893, + "step": 8990 + }, + { + "epoch": 0.29, + "learning_rate": 1.6718970557336717e-05, + "loss": 1.0732, + "step": 8991 + }, + { + "epoch": 0.29, + "learning_rate": 1.6718203358923625e-05, + "loss": 1.0762, + "step": 8992 + }, + { + "epoch": 0.29, + "learning_rate": 1.6717436088431648e-05, + "loss": 0.958, + "step": 8993 + }, + { + "epoch": 0.29, + "learning_rate": 1.671666874586902e-05, + "loss": 1.0015, + "step": 8994 + }, + { + "epoch": 0.29, + "learning_rate": 1.6715901331243963e-05, + "loss": 1.1758, + "step": 8995 + }, + { + "epoch": 0.29, + "learning_rate": 1.6715133844564723e-05, + "loss": 0.9521, + "step": 8996 + }, + { + "epoch": 0.29, + "learning_rate": 1.6714366285839526e-05, + "loss": 1.0693, + "step": 8997 + }, + { + "epoch": 0.29, + "learning_rate": 1.671359865507661e-05, + "loss": 1.0044, + "step": 8998 + }, + { + "epoch": 0.29, + "learning_rate": 1.671283095228421e-05, + "loss": 1.1104, + "step": 8999 + }, + { + "epoch": 0.29, + "learning_rate": 1.6712063177470564e-05, + "loss": 1.0679, + "step": 9000 + }, + { + "epoch": 0.29, + "learning_rate": 1.6711295330643912e-05, + "loss": 1.0967, + "step": 9001 + }, + { + "epoch": 0.29, + "learning_rate": 1.6710527411812483e-05, + "loss": 0.9692, + "step": 9002 + }, + { + "epoch": 0.29, + "learning_rate": 1.6709759420984528e-05, + "loss": 1.0342, + "step": 9003 + }, + { + "epoch": 0.29, + "learning_rate": 1.6708991358168276e-05, + "loss": 0.9365, + "step": 9004 + }, + { + "epoch": 0.29, + "learning_rate": 1.6708223223371973e-05, + "loss": 1.125, + "step": 9005 + }, + { + "epoch": 0.29, + "learning_rate": 1.670745501660386e-05, + "loss": 1.0669, + "step": 9006 + }, + { + "epoch": 0.29, + "learning_rate": 1.6706686737872178e-05, + "loss": 0.5247, + "step": 9007 + }, + { + "epoch": 0.29, + "learning_rate": 1.670591838718517e-05, + "loss": 1.0767, + "step": 9008 + }, + { + "epoch": 0.29, + "learning_rate": 1.670514996455108e-05, + "loss": 0.9385, + "step": 9009 + }, + { + "epoch": 0.29, + "learning_rate": 1.670438146997815e-05, + "loss": 1.1084, + "step": 9010 + }, + { + "epoch": 0.29, + "learning_rate": 1.6703612903474625e-05, + "loss": 1.0381, + "step": 9011 + }, + { + "epoch": 0.29, + "learning_rate": 1.6702844265048755e-05, + "loss": 1.0142, + "step": 9012 + }, + { + "epoch": 0.29, + "learning_rate": 1.6702075554708784e-05, + "loss": 1.1318, + "step": 9013 + }, + { + "epoch": 0.29, + "learning_rate": 1.670130677246296e-05, + "loss": 0.8936, + "step": 9014 + }, + { + "epoch": 0.29, + "learning_rate": 1.6700537918319533e-05, + "loss": 1.1328, + "step": 9015 + }, + { + "epoch": 0.29, + "learning_rate": 1.6699768992286746e-05, + "loss": 1.0894, + "step": 9016 + }, + { + "epoch": 0.29, + "learning_rate": 1.6698999994372856e-05, + "loss": 1.1611, + "step": 9017 + }, + { + "epoch": 0.29, + "learning_rate": 1.6698230924586106e-05, + "loss": 0.5732, + "step": 9018 + }, + { + "epoch": 0.29, + "learning_rate": 1.6697461782934756e-05, + "loss": 1.0728, + "step": 9019 + }, + { + "epoch": 0.29, + "learning_rate": 1.6696692569427054e-05, + "loss": 0.9619, + "step": 9020 + }, + { + "epoch": 0.29, + "learning_rate": 1.669592328407125e-05, + "loss": 1.0977, + "step": 9021 + }, + { + "epoch": 0.29, + "learning_rate": 1.6695153926875598e-05, + "loss": 1.0083, + "step": 9022 + }, + { + "epoch": 0.29, + "learning_rate": 1.669438449784836e-05, + "loss": 1.0024, + "step": 9023 + }, + { + "epoch": 0.29, + "learning_rate": 1.6693614996997776e-05, + "loss": 1.0425, + "step": 9024 + }, + { + "epoch": 0.29, + "learning_rate": 1.6692845424332116e-05, + "loss": 0.9819, + "step": 9025 + }, + { + "epoch": 0.29, + "learning_rate": 1.6692075779859634e-05, + "loss": 1.0737, + "step": 9026 + }, + { + "epoch": 0.29, + "learning_rate": 1.6691306063588583e-05, + "loss": 0.9365, + "step": 9027 + }, + { + "epoch": 0.29, + "learning_rate": 1.6690536275527225e-05, + "loss": 1.0308, + "step": 9028 + }, + { + "epoch": 0.29, + "learning_rate": 1.6689766415683815e-05, + "loss": 1.0098, + "step": 9029 + }, + { + "epoch": 0.29, + "learning_rate": 1.668899648406662e-05, + "loss": 1.0562, + "step": 9030 + }, + { + "epoch": 0.29, + "learning_rate": 1.668822648068389e-05, + "loss": 0.9976, + "step": 9031 + }, + { + "epoch": 0.29, + "learning_rate": 1.6687456405543897e-05, + "loss": 0.9575, + "step": 9032 + }, + { + "epoch": 0.29, + "learning_rate": 1.6686686258654896e-05, + "loss": 0.9834, + "step": 9033 + }, + { + "epoch": 0.29, + "learning_rate": 1.668591604002515e-05, + "loss": 1.0063, + "step": 9034 + }, + { + "epoch": 0.29, + "learning_rate": 1.6685145749662926e-05, + "loss": 0.9551, + "step": 9035 + }, + { + "epoch": 0.29, + "learning_rate": 1.6684375387576487e-05, + "loss": 1.0933, + "step": 9036 + }, + { + "epoch": 0.29, + "learning_rate": 1.66836049537741e-05, + "loss": 0.9512, + "step": 9037 + }, + { + "epoch": 0.29, + "learning_rate": 1.6682834448264027e-05, + "loss": 1.1064, + "step": 9038 + }, + { + "epoch": 0.29, + "learning_rate": 1.6682063871054534e-05, + "loss": 1.1938, + "step": 9039 + }, + { + "epoch": 0.29, + "learning_rate": 1.6681293222153896e-05, + "loss": 0.9707, + "step": 9040 + }, + { + "epoch": 0.29, + "learning_rate": 1.6680522501570374e-05, + "loss": 1.0649, + "step": 9041 + }, + { + "epoch": 0.29, + "learning_rate": 1.6679751709312236e-05, + "loss": 1.0742, + "step": 9042 + }, + { + "epoch": 0.29, + "learning_rate": 1.6678980845387758e-05, + "loss": 0.957, + "step": 9043 + }, + { + "epoch": 0.29, + "learning_rate": 1.6678209909805207e-05, + "loss": 1.0767, + "step": 9044 + }, + { + "epoch": 0.29, + "learning_rate": 1.6677438902572855e-05, + "loss": 1.0142, + "step": 9045 + }, + { + "epoch": 0.29, + "learning_rate": 1.6676667823698974e-05, + "loss": 1.0161, + "step": 9046 + }, + { + "epoch": 0.29, + "learning_rate": 1.6675896673191835e-05, + "loss": 1.0527, + "step": 9047 + }, + { + "epoch": 0.29, + "learning_rate": 1.6675125451059716e-05, + "loss": 1.1553, + "step": 9048 + }, + { + "epoch": 0.29, + "learning_rate": 1.667435415731089e-05, + "loss": 0.9517, + "step": 9049 + }, + { + "epoch": 0.29, + "learning_rate": 1.6673582791953627e-05, + "loss": 1.0938, + "step": 9050 + }, + { + "epoch": 0.29, + "learning_rate": 1.6672811354996205e-05, + "loss": 1.0317, + "step": 9051 + }, + { + "epoch": 0.29, + "learning_rate": 1.6672039846446908e-05, + "loss": 1.1089, + "step": 9052 + }, + { + "epoch": 0.29, + "learning_rate": 1.6671268266314006e-05, + "loss": 1.127, + "step": 9053 + }, + { + "epoch": 0.29, + "learning_rate": 1.6670496614605776e-05, + "loss": 1.0986, + "step": 9054 + }, + { + "epoch": 0.29, + "learning_rate": 1.66697248913305e-05, + "loss": 1.0732, + "step": 9055 + }, + { + "epoch": 0.29, + "learning_rate": 1.6668953096496457e-05, + "loss": 0.9126, + "step": 9056 + }, + { + "epoch": 0.29, + "learning_rate": 1.6668181230111932e-05, + "loss": 1.0142, + "step": 9057 + }, + { + "epoch": 0.29, + "learning_rate": 1.6667409292185198e-05, + "loss": 0.9243, + "step": 9058 + }, + { + "epoch": 0.29, + "learning_rate": 1.6666637282724546e-05, + "loss": 1.0317, + "step": 9059 + }, + { + "epoch": 0.29, + "learning_rate": 1.666586520173825e-05, + "loss": 0.9644, + "step": 9060 + }, + { + "epoch": 0.29, + "learning_rate": 1.66650930492346e-05, + "loss": 1.0298, + "step": 9061 + }, + { + "epoch": 0.29, + "learning_rate": 1.666432082522188e-05, + "loss": 0.9565, + "step": 9062 + }, + { + "epoch": 0.29, + "learning_rate": 1.666354852970837e-05, + "loss": 1.0508, + "step": 9063 + }, + { + "epoch": 0.29, + "learning_rate": 1.666277616270236e-05, + "loss": 1.1143, + "step": 9064 + }, + { + "epoch": 0.29, + "learning_rate": 1.666200372421214e-05, + "loss": 1.0493, + "step": 9065 + }, + { + "epoch": 0.29, + "learning_rate": 1.6661231214245987e-05, + "loss": 1.123, + "step": 9066 + }, + { + "epoch": 0.29, + "learning_rate": 1.66604586328122e-05, + "loss": 1.2095, + "step": 9067 + }, + { + "epoch": 0.29, + "learning_rate": 1.665968597991906e-05, + "loss": 1.061, + "step": 9068 + }, + { + "epoch": 0.29, + "learning_rate": 1.6658913255574858e-05, + "loss": 1.043, + "step": 9069 + }, + { + "epoch": 0.29, + "learning_rate": 1.665814045978789e-05, + "loss": 1.0396, + "step": 9070 + }, + { + "epoch": 0.29, + "learning_rate": 1.6657367592566446e-05, + "loss": 1.1108, + "step": 9071 + }, + { + "epoch": 0.29, + "learning_rate": 1.665659465391881e-05, + "loss": 1.0537, + "step": 9072 + }, + { + "epoch": 0.29, + "learning_rate": 1.6655821643853282e-05, + "loss": 1.0278, + "step": 9073 + }, + { + "epoch": 0.29, + "learning_rate": 1.6655048562378157e-05, + "loss": 1.103, + "step": 9074 + }, + { + "epoch": 0.29, + "learning_rate": 1.6654275409501726e-05, + "loss": 1.0991, + "step": 9075 + }, + { + "epoch": 0.29, + "learning_rate": 1.6653502185232283e-05, + "loss": 1.1826, + "step": 9076 + }, + { + "epoch": 0.29, + "learning_rate": 1.6652728889578125e-05, + "loss": 1.1431, + "step": 9077 + }, + { + "epoch": 0.29, + "learning_rate": 1.665195552254755e-05, + "loss": 1.0327, + "step": 9078 + }, + { + "epoch": 0.29, + "learning_rate": 1.6651182084148852e-05, + "loss": 1.0327, + "step": 9079 + }, + { + "epoch": 0.29, + "learning_rate": 1.6650408574390335e-05, + "loss": 1.0454, + "step": 9080 + }, + { + "epoch": 0.29, + "learning_rate": 1.664963499328029e-05, + "loss": 1.0249, + "step": 9081 + }, + { + "epoch": 0.29, + "learning_rate": 1.6648861340827026e-05, + "loss": 1.0947, + "step": 9082 + }, + { + "epoch": 0.29, + "learning_rate": 1.6648087617038835e-05, + "loss": 1.1191, + "step": 9083 + }, + { + "epoch": 0.29, + "learning_rate": 1.6647313821924022e-05, + "loss": 0.9473, + "step": 9084 + }, + { + "epoch": 0.29, + "learning_rate": 1.664653995549089e-05, + "loss": 1.0972, + "step": 9085 + }, + { + "epoch": 0.29, + "learning_rate": 1.664576601774774e-05, + "loss": 1.1162, + "step": 9086 + }, + { + "epoch": 0.29, + "learning_rate": 1.6644992008702873e-05, + "loss": 0.9053, + "step": 9087 + }, + { + "epoch": 0.29, + "learning_rate": 1.66442179283646e-05, + "loss": 1.1689, + "step": 9088 + }, + { + "epoch": 0.29, + "learning_rate": 1.664344377674122e-05, + "loss": 1.0234, + "step": 9089 + }, + { + "epoch": 0.29, + "learning_rate": 1.664266955384104e-05, + "loss": 0.9868, + "step": 9090 + }, + { + "epoch": 0.29, + "learning_rate": 1.664189525967237e-05, + "loss": 1.0015, + "step": 9091 + }, + { + "epoch": 0.29, + "learning_rate": 1.664112089424351e-05, + "loss": 1.0508, + "step": 9092 + }, + { + "epoch": 0.29, + "learning_rate": 1.664034645756278e-05, + "loss": 1.0835, + "step": 9093 + }, + { + "epoch": 0.29, + "learning_rate": 1.6639571949638476e-05, + "loss": 1.0513, + "step": 9094 + }, + { + "epoch": 0.29, + "learning_rate": 1.6638797370478915e-05, + "loss": 0.561, + "step": 9095 + }, + { + "epoch": 0.29, + "learning_rate": 1.6638022720092405e-05, + "loss": 1.04, + "step": 9096 + }, + { + "epoch": 0.29, + "learning_rate": 1.663724799848726e-05, + "loss": 1.0229, + "step": 9097 + }, + { + "epoch": 0.29, + "learning_rate": 1.663647320567179e-05, + "loss": 1.0254, + "step": 9098 + }, + { + "epoch": 0.29, + "learning_rate": 1.6635698341654305e-05, + "loss": 1.0225, + "step": 9099 + }, + { + "epoch": 0.29, + "learning_rate": 1.663492340644312e-05, + "loss": 0.9373, + "step": 9100 + }, + { + "epoch": 0.29, + "learning_rate": 1.663414840004655e-05, + "loss": 1.1055, + "step": 9101 + }, + { + "epoch": 0.29, + "learning_rate": 1.6633373322472916e-05, + "loss": 1.0405, + "step": 9102 + }, + { + "epoch": 0.29, + "learning_rate": 1.6632598173730522e-05, + "loss": 1.1787, + "step": 9103 + }, + { + "epoch": 0.29, + "learning_rate": 1.663182295382769e-05, + "loss": 0.938, + "step": 9104 + }, + { + "epoch": 0.29, + "learning_rate": 1.663104766277274e-05, + "loss": 0.9487, + "step": 9105 + }, + { + "epoch": 0.29, + "learning_rate": 1.6630272300573987e-05, + "loss": 1.0674, + "step": 9106 + }, + { + "epoch": 0.29, + "learning_rate": 1.6629496867239748e-05, + "loss": 1.0703, + "step": 9107 + }, + { + "epoch": 0.29, + "learning_rate": 1.6628721362778346e-05, + "loss": 1.2144, + "step": 9108 + }, + { + "epoch": 0.29, + "learning_rate": 1.66279457871981e-05, + "loss": 1.0747, + "step": 9109 + }, + { + "epoch": 0.29, + "learning_rate": 1.6627170140507333e-05, + "loss": 1.0791, + "step": 9110 + }, + { + "epoch": 0.29, + "learning_rate": 1.662639442271436e-05, + "loss": 1.1729, + "step": 9111 + }, + { + "epoch": 0.29, + "learning_rate": 1.6625618633827515e-05, + "loss": 0.957, + "step": 9112 + }, + { + "epoch": 0.29, + "learning_rate": 1.6624842773855113e-05, + "loss": 0.9395, + "step": 9113 + }, + { + "epoch": 0.29, + "learning_rate": 1.6624066842805476e-05, + "loss": 1.0347, + "step": 9114 + }, + { + "epoch": 0.29, + "learning_rate": 1.6623290840686936e-05, + "loss": 1.0088, + "step": 9115 + }, + { + "epoch": 0.29, + "learning_rate": 1.662251476750782e-05, + "loss": 0.5706, + "step": 9116 + }, + { + "epoch": 0.29, + "learning_rate": 1.6621738623276442e-05, + "loss": 0.998, + "step": 9117 + }, + { + "epoch": 0.29, + "learning_rate": 1.6620962408001142e-05, + "loss": 0.96, + "step": 9118 + }, + { + "epoch": 0.29, + "learning_rate": 1.662018612169024e-05, + "loss": 1.0259, + "step": 9119 + }, + { + "epoch": 0.29, + "learning_rate": 1.661940976435207e-05, + "loss": 0.9702, + "step": 9120 + }, + { + "epoch": 0.29, + "learning_rate": 1.661863333599496e-05, + "loss": 1.0342, + "step": 9121 + }, + { + "epoch": 0.29, + "learning_rate": 1.6617856836627238e-05, + "loss": 1.0376, + "step": 9122 + }, + { + "epoch": 0.29, + "learning_rate": 1.6617080266257234e-05, + "loss": 1.0513, + "step": 9123 + }, + { + "epoch": 0.29, + "learning_rate": 1.6616303624893285e-05, + "loss": 0.9941, + "step": 9124 + }, + { + "epoch": 0.29, + "learning_rate": 1.6615526912543724e-05, + "loss": 1.1519, + "step": 9125 + }, + { + "epoch": 0.29, + "learning_rate": 1.6614750129216877e-05, + "loss": 1.043, + "step": 9126 + }, + { + "epoch": 0.29, + "learning_rate": 1.6613973274921084e-05, + "loss": 1.106, + "step": 9127 + }, + { + "epoch": 0.29, + "learning_rate": 1.661319634966467e-05, + "loss": 1.0493, + "step": 9128 + }, + { + "epoch": 0.29, + "learning_rate": 1.661241935345599e-05, + "loss": 1.0552, + "step": 9129 + }, + { + "epoch": 0.29, + "learning_rate": 1.661164228630336e-05, + "loss": 0.9307, + "step": 9130 + }, + { + "epoch": 0.29, + "learning_rate": 1.661086514821513e-05, + "loss": 0.9885, + "step": 9131 + }, + { + "epoch": 0.29, + "learning_rate": 1.6610087939199632e-05, + "loss": 1.0249, + "step": 9132 + }, + { + "epoch": 0.29, + "learning_rate": 1.6609310659265205e-05, + "loss": 0.9966, + "step": 9133 + }, + { + "epoch": 0.29, + "learning_rate": 1.6608533308420192e-05, + "loss": 1.0742, + "step": 9134 + }, + { + "epoch": 0.29, + "learning_rate": 1.6607755886672928e-05, + "loss": 1.1089, + "step": 9135 + }, + { + "epoch": 0.29, + "learning_rate": 1.660697839403176e-05, + "loss": 1.0576, + "step": 9136 + }, + { + "epoch": 0.29, + "learning_rate": 1.6606200830505024e-05, + "loss": 1.2285, + "step": 9137 + }, + { + "epoch": 0.29, + "learning_rate": 1.660542319610106e-05, + "loss": 1.1299, + "step": 9138 + }, + { + "epoch": 0.29, + "learning_rate": 1.6604645490828224e-05, + "loss": 1.0571, + "step": 9139 + }, + { + "epoch": 0.29, + "learning_rate": 1.660386771469485e-05, + "loss": 0.5403, + "step": 9140 + }, + { + "epoch": 0.29, + "learning_rate": 1.660308986770928e-05, + "loss": 1.1167, + "step": 9141 + }, + { + "epoch": 0.29, + "learning_rate": 1.6602311949879863e-05, + "loss": 1.0806, + "step": 9142 + }, + { + "epoch": 0.29, + "learning_rate": 1.660153396121495e-05, + "loss": 1.0649, + "step": 9143 + }, + { + "epoch": 0.29, + "learning_rate": 1.660075590172288e-05, + "loss": 1.2466, + "step": 9144 + }, + { + "epoch": 0.29, + "learning_rate": 1.6599977771412005e-05, + "loss": 0.9893, + "step": 9145 + }, + { + "epoch": 0.29, + "learning_rate": 1.6599199570290677e-05, + "loss": 1.1729, + "step": 9146 + }, + { + "epoch": 0.29, + "learning_rate": 1.659842129836724e-05, + "loss": 1.1284, + "step": 9147 + }, + { + "epoch": 0.29, + "learning_rate": 1.6597642955650042e-05, + "loss": 1.061, + "step": 9148 + }, + { + "epoch": 0.29, + "learning_rate": 1.659686454214744e-05, + "loss": 0.5544, + "step": 9149 + }, + { + "epoch": 0.29, + "learning_rate": 1.6596086057867784e-05, + "loss": 1.0234, + "step": 9150 + }, + { + "epoch": 0.29, + "learning_rate": 1.659530750281942e-05, + "loss": 1.0703, + "step": 9151 + }, + { + "epoch": 0.29, + "learning_rate": 1.659452887701071e-05, + "loss": 1.1152, + "step": 9152 + }, + { + "epoch": 0.29, + "learning_rate": 1.659375018045e-05, + "loss": 1.0684, + "step": 9153 + }, + { + "epoch": 0.29, + "learning_rate": 1.6592971413145652e-05, + "loss": 0.9868, + "step": 9154 + }, + { + "epoch": 0.29, + "learning_rate": 1.659219257510601e-05, + "loss": 1.0513, + "step": 9155 + }, + { + "epoch": 0.29, + "learning_rate": 1.6591413666339448e-05, + "loss": 1.2344, + "step": 9156 + }, + { + "epoch": 0.29, + "learning_rate": 1.6590634686854304e-05, + "loss": 1.1421, + "step": 9157 + }, + { + "epoch": 0.29, + "learning_rate": 1.658985563665895e-05, + "loss": 0.5547, + "step": 9158 + }, + { + "epoch": 0.29, + "learning_rate": 1.6589076515761734e-05, + "loss": 1.1094, + "step": 9159 + }, + { + "epoch": 0.29, + "learning_rate": 1.6588297324171022e-05, + "loss": 0.543, + "step": 9160 + }, + { + "epoch": 0.29, + "learning_rate": 1.6587518061895167e-05, + "loss": 1.1372, + "step": 9161 + }, + { + "epoch": 0.29, + "learning_rate": 1.658673872894254e-05, + "loss": 1.083, + "step": 9162 + }, + { + "epoch": 0.29, + "learning_rate": 1.6585959325321494e-05, + "loss": 0.9512, + "step": 9163 + }, + { + "epoch": 0.29, + "learning_rate": 1.6585179851040393e-05, + "loss": 1.0757, + "step": 9164 + }, + { + "epoch": 0.29, + "learning_rate": 1.65844003061076e-05, + "loss": 1.1118, + "step": 9165 + }, + { + "epoch": 0.29, + "learning_rate": 1.658362069053148e-05, + "loss": 1.0391, + "step": 9166 + }, + { + "epoch": 0.29, + "learning_rate": 1.6582841004320395e-05, + "loss": 1.1426, + "step": 9167 + }, + { + "epoch": 0.29, + "learning_rate": 1.6582061247482714e-05, + "loss": 1.1558, + "step": 9168 + }, + { + "epoch": 0.29, + "learning_rate": 1.6581281420026794e-05, + "loss": 1.0151, + "step": 9169 + }, + { + "epoch": 0.29, + "learning_rate": 1.6580501521961015e-05, + "loss": 1.0698, + "step": 9170 + }, + { + "epoch": 0.29, + "learning_rate": 1.6579721553293736e-05, + "loss": 1.042, + "step": 9171 + }, + { + "epoch": 0.29, + "learning_rate": 1.6578941514033325e-05, + "loss": 1.0879, + "step": 9172 + }, + { + "epoch": 0.29, + "learning_rate": 1.6578161404188154e-05, + "loss": 1.0537, + "step": 9173 + }, + { + "epoch": 0.29, + "learning_rate": 1.6577381223766592e-05, + "loss": 1.0347, + "step": 9174 + }, + { + "epoch": 0.29, + "learning_rate": 1.6576600972777007e-05, + "loss": 1.0474, + "step": 9175 + }, + { + "epoch": 0.29, + "learning_rate": 1.6575820651227772e-05, + "loss": 1.1177, + "step": 9176 + }, + { + "epoch": 0.29, + "learning_rate": 1.6575040259127263e-05, + "loss": 0.9292, + "step": 9177 + }, + { + "epoch": 0.29, + "learning_rate": 1.6574259796483847e-05, + "loss": 1.0518, + "step": 9178 + }, + { + "epoch": 0.29, + "learning_rate": 1.65734792633059e-05, + "loss": 0.9844, + "step": 9179 + }, + { + "epoch": 0.29, + "learning_rate": 1.657269865960179e-05, + "loss": 1.1157, + "step": 9180 + }, + { + "epoch": 0.29, + "learning_rate": 1.6571917985379905e-05, + "loss": 1.0625, + "step": 9181 + }, + { + "epoch": 0.29, + "learning_rate": 1.6571137240648612e-05, + "loss": 1.0776, + "step": 9182 + }, + { + "epoch": 0.29, + "learning_rate": 1.6570356425416287e-05, + "loss": 1.0024, + "step": 9183 + }, + { + "epoch": 0.29, + "learning_rate": 1.656957553969131e-05, + "loss": 1.0195, + "step": 9184 + }, + { + "epoch": 0.29, + "learning_rate": 1.6568794583482057e-05, + "loss": 1.084, + "step": 9185 + }, + { + "epoch": 0.29, + "learning_rate": 1.6568013556796908e-05, + "loss": 1.0186, + "step": 9186 + }, + { + "epoch": 0.29, + "learning_rate": 1.6567232459644245e-05, + "loss": 0.8701, + "step": 9187 + }, + { + "epoch": 0.29, + "learning_rate": 1.6566451292032448e-05, + "loss": 0.9878, + "step": 9188 + }, + { + "epoch": 0.29, + "learning_rate": 1.656567005396989e-05, + "loss": 0.979, + "step": 9189 + }, + { + "epoch": 0.29, + "learning_rate": 1.6564888745464964e-05, + "loss": 0.9521, + "step": 9190 + }, + { + "epoch": 0.29, + "learning_rate": 1.6564107366526046e-05, + "loss": 0.9771, + "step": 9191 + }, + { + "epoch": 0.29, + "learning_rate": 1.6563325917161522e-05, + "loss": 1.084, + "step": 9192 + }, + { + "epoch": 0.29, + "learning_rate": 1.656254439737977e-05, + "loss": 0.9688, + "step": 9193 + }, + { + "epoch": 0.29, + "learning_rate": 1.6561762807189182e-05, + "loss": 0.9976, + "step": 9194 + }, + { + "epoch": 0.29, + "learning_rate": 1.6560981146598145e-05, + "loss": 0.917, + "step": 9195 + }, + { + "epoch": 0.29, + "learning_rate": 1.6560199415615037e-05, + "loss": 1.0776, + "step": 9196 + }, + { + "epoch": 0.29, + "learning_rate": 1.6559417614248252e-05, + "loss": 0.981, + "step": 9197 + }, + { + "epoch": 0.29, + "learning_rate": 1.6558635742506175e-05, + "loss": 0.9492, + "step": 9198 + }, + { + "epoch": 0.29, + "learning_rate": 1.6557853800397198e-05, + "loss": 1.0732, + "step": 9199 + }, + { + "epoch": 0.29, + "learning_rate": 1.65570717879297e-05, + "loss": 1.0225, + "step": 9200 + }, + { + "epoch": 0.29, + "learning_rate": 1.6556289705112085e-05, + "loss": 0.9985, + "step": 9201 + }, + { + "epoch": 0.29, + "learning_rate": 1.6555507551952733e-05, + "loss": 1.0361, + "step": 9202 + }, + { + "epoch": 0.29, + "learning_rate": 1.6554725328460045e-05, + "loss": 1.04, + "step": 9203 + }, + { + "epoch": 0.29, + "learning_rate": 1.655394303464241e-05, + "loss": 1.1113, + "step": 9204 + }, + { + "epoch": 0.29, + "learning_rate": 1.6553160670508215e-05, + "loss": 1.1206, + "step": 9205 + }, + { + "epoch": 0.29, + "learning_rate": 1.6552378236065857e-05, + "loss": 1.0645, + "step": 9206 + }, + { + "epoch": 0.29, + "learning_rate": 1.655159573132374e-05, + "loss": 0.9849, + "step": 9207 + }, + { + "epoch": 0.29, + "learning_rate": 1.6550813156290247e-05, + "loss": 1.0469, + "step": 9208 + }, + { + "epoch": 0.29, + "learning_rate": 1.655003051097378e-05, + "loss": 1.0039, + "step": 9209 + }, + { + "epoch": 0.29, + "learning_rate": 1.6549247795382734e-05, + "loss": 1.064, + "step": 9210 + }, + { + "epoch": 0.29, + "learning_rate": 1.6548465009525506e-05, + "loss": 1.0337, + "step": 9211 + }, + { + "epoch": 0.29, + "learning_rate": 1.65476821534105e-05, + "loss": 0.9824, + "step": 9212 + }, + { + "epoch": 0.29, + "learning_rate": 1.6546899227046107e-05, + "loss": 0.4978, + "step": 9213 + }, + { + "epoch": 0.29, + "learning_rate": 1.6546116230440735e-05, + "loss": 1.2295, + "step": 9214 + }, + { + "epoch": 0.29, + "learning_rate": 1.6545333163602777e-05, + "loss": 0.9185, + "step": 9215 + }, + { + "epoch": 0.29, + "learning_rate": 1.654455002654064e-05, + "loss": 1.2324, + "step": 9216 + }, + { + "epoch": 0.29, + "learning_rate": 1.6543766819262724e-05, + "loss": 1.1006, + "step": 9217 + }, + { + "epoch": 0.29, + "learning_rate": 1.6542983541777432e-05, + "loss": 0.9517, + "step": 9218 + }, + { + "epoch": 0.29, + "learning_rate": 1.654220019409317e-05, + "loss": 1.127, + "step": 9219 + }, + { + "epoch": 0.29, + "learning_rate": 1.6541416776218338e-05, + "loss": 1.0288, + "step": 9220 + }, + { + "epoch": 0.29, + "learning_rate": 1.6540633288161343e-05, + "loss": 1.0405, + "step": 9221 + }, + { + "epoch": 0.29, + "learning_rate": 1.6539849729930596e-05, + "loss": 1.0806, + "step": 9222 + }, + { + "epoch": 0.29, + "learning_rate": 1.6539066101534493e-05, + "loss": 0.8711, + "step": 9223 + }, + { + "epoch": 0.29, + "learning_rate": 1.6538282402981455e-05, + "loss": 1.0459, + "step": 9224 + }, + { + "epoch": 0.3, + "learning_rate": 1.6537498634279877e-05, + "loss": 0.9697, + "step": 9225 + }, + { + "epoch": 0.3, + "learning_rate": 1.6536714795438177e-05, + "loss": 1.104, + "step": 9226 + }, + { + "epoch": 0.3, + "learning_rate": 1.653593088646476e-05, + "loss": 1.0142, + "step": 9227 + }, + { + "epoch": 0.3, + "learning_rate": 1.653514690736804e-05, + "loss": 0.9863, + "step": 9228 + }, + { + "epoch": 0.3, + "learning_rate": 1.6534362858156426e-05, + "loss": 1.0, + "step": 9229 + }, + { + "epoch": 0.3, + "learning_rate": 1.6533578738838332e-05, + "loss": 0.9512, + "step": 9230 + }, + { + "epoch": 0.3, + "learning_rate": 1.6532794549422164e-05, + "loss": 1.0664, + "step": 9231 + }, + { + "epoch": 0.3, + "learning_rate": 1.6532010289916346e-05, + "loss": 1.019, + "step": 9232 + }, + { + "epoch": 0.3, + "learning_rate": 1.6531225960329286e-05, + "loss": 0.9858, + "step": 9233 + }, + { + "epoch": 0.3, + "learning_rate": 1.6530441560669396e-05, + "loss": 1.0166, + "step": 9234 + }, + { + "epoch": 0.3, + "learning_rate": 1.65296570909451e-05, + "loss": 1.0093, + "step": 9235 + }, + { + "epoch": 0.3, + "learning_rate": 1.6528872551164807e-05, + "loss": 1.1084, + "step": 9236 + }, + { + "epoch": 0.3, + "learning_rate": 1.652808794133694e-05, + "loss": 0.9438, + "step": 9237 + }, + { + "epoch": 0.3, + "learning_rate": 1.6527303261469914e-05, + "loss": 1.0918, + "step": 9238 + }, + { + "epoch": 0.3, + "learning_rate": 1.6526518511572146e-05, + "loss": 1.022, + "step": 9239 + }, + { + "epoch": 0.3, + "learning_rate": 1.6525733691652055e-05, + "loss": 0.9458, + "step": 9240 + }, + { + "epoch": 0.3, + "learning_rate": 1.652494880171807e-05, + "loss": 0.9971, + "step": 9241 + }, + { + "epoch": 0.3, + "learning_rate": 1.6524163841778604e-05, + "loss": 1.062, + "step": 9242 + }, + { + "epoch": 0.3, + "learning_rate": 1.6523378811842077e-05, + "loss": 1.084, + "step": 9243 + }, + { + "epoch": 0.3, + "learning_rate": 1.652259371191692e-05, + "loss": 1.0981, + "step": 9244 + }, + { + "epoch": 0.3, + "learning_rate": 1.6521808542011548e-05, + "loss": 1.0132, + "step": 9245 + }, + { + "epoch": 0.3, + "learning_rate": 1.652102330213439e-05, + "loss": 1.0659, + "step": 9246 + }, + { + "epoch": 0.3, + "learning_rate": 1.6520237992293866e-05, + "loss": 1.0151, + "step": 9247 + }, + { + "epoch": 0.3, + "learning_rate": 1.6519452612498406e-05, + "loss": 1.0107, + "step": 9248 + }, + { + "epoch": 0.3, + "learning_rate": 1.651866716275644e-05, + "loss": 0.8877, + "step": 9249 + }, + { + "epoch": 0.3, + "learning_rate": 1.6517881643076387e-05, + "loss": 1.0605, + "step": 9250 + }, + { + "epoch": 0.3, + "learning_rate": 1.6517096053466675e-05, + "loss": 1.0176, + "step": 9251 + }, + { + "epoch": 0.3, + "learning_rate": 1.6516310393935736e-05, + "loss": 1.0142, + "step": 9252 + }, + { + "epoch": 0.3, + "learning_rate": 1.6515524664492e-05, + "loss": 0.9858, + "step": 9253 + }, + { + "epoch": 0.3, + "learning_rate": 1.6514738865143894e-05, + "loss": 1.0225, + "step": 9254 + }, + { + "epoch": 0.3, + "learning_rate": 1.651395299589985e-05, + "loss": 1.0371, + "step": 9255 + }, + { + "epoch": 0.3, + "learning_rate": 1.65131670567683e-05, + "loss": 1.1294, + "step": 9256 + }, + { + "epoch": 0.3, + "learning_rate": 1.6512381047757675e-05, + "loss": 0.5342, + "step": 9257 + }, + { + "epoch": 0.3, + "learning_rate": 1.651159496887641e-05, + "loss": 1.1328, + "step": 9258 + }, + { + "epoch": 0.3, + "learning_rate": 1.651080882013294e-05, + "loss": 0.9785, + "step": 9259 + }, + { + "epoch": 0.3, + "learning_rate": 1.6510022601535694e-05, + "loss": 0.9946, + "step": 9260 + }, + { + "epoch": 0.3, + "learning_rate": 1.6509236313093108e-05, + "loss": 0.9775, + "step": 9261 + }, + { + "epoch": 0.3, + "learning_rate": 1.6508449954813624e-05, + "loss": 1.1592, + "step": 9262 + }, + { + "epoch": 0.3, + "learning_rate": 1.6507663526705676e-05, + "loss": 1.1187, + "step": 9263 + }, + { + "epoch": 0.3, + "learning_rate": 1.65068770287777e-05, + "loss": 1.0244, + "step": 9264 + }, + { + "epoch": 0.3, + "learning_rate": 1.6506090461038133e-05, + "loss": 1.0938, + "step": 9265 + }, + { + "epoch": 0.3, + "learning_rate": 1.6505303823495417e-05, + "loss": 1.144, + "step": 9266 + }, + { + "epoch": 0.3, + "learning_rate": 1.6504517116157993e-05, + "loss": 1.1216, + "step": 9267 + }, + { + "epoch": 0.3, + "learning_rate": 1.6503730339034296e-05, + "loss": 1.1216, + "step": 9268 + }, + { + "epoch": 0.3, + "learning_rate": 1.650294349213277e-05, + "loss": 1.1108, + "step": 9269 + }, + { + "epoch": 0.3, + "learning_rate": 1.6502156575461857e-05, + "loss": 1.021, + "step": 9270 + }, + { + "epoch": 0.3, + "learning_rate": 1.6501369589030004e-05, + "loss": 0.9712, + "step": 9271 + }, + { + "epoch": 0.3, + "learning_rate": 1.6500582532845652e-05, + "loss": 1.0894, + "step": 9272 + }, + { + "epoch": 0.3, + "learning_rate": 1.6499795406917237e-05, + "loss": 0.9341, + "step": 9273 + }, + { + "epoch": 0.3, + "learning_rate": 1.6499008211253216e-05, + "loss": 1.0195, + "step": 9274 + }, + { + "epoch": 0.3, + "learning_rate": 1.649822094586203e-05, + "loss": 1.0693, + "step": 9275 + }, + { + "epoch": 0.3, + "learning_rate": 1.649743361075212e-05, + "loss": 1.0244, + "step": 9276 + }, + { + "epoch": 0.3, + "learning_rate": 1.6496646205931944e-05, + "loss": 1.0688, + "step": 9277 + }, + { + "epoch": 0.3, + "learning_rate": 1.6495858731409942e-05, + "loss": 1.0811, + "step": 9278 + }, + { + "epoch": 0.3, + "learning_rate": 1.6495071187194564e-05, + "loss": 1.0352, + "step": 9279 + }, + { + "epoch": 0.3, + "learning_rate": 1.6494283573294263e-05, + "loss": 1.0249, + "step": 9280 + }, + { + "epoch": 0.3, + "learning_rate": 1.6493495889717486e-05, + "loss": 1.127, + "step": 9281 + }, + { + "epoch": 0.3, + "learning_rate": 1.6492708136472685e-05, + "loss": 1.0322, + "step": 9282 + }, + { + "epoch": 0.3, + "learning_rate": 1.649192031356831e-05, + "loss": 1.1226, + "step": 9283 + }, + { + "epoch": 0.3, + "learning_rate": 1.6491132421012817e-05, + "loss": 0.9785, + "step": 9284 + }, + { + "epoch": 0.3, + "learning_rate": 1.6490344458814655e-05, + "loss": 1.126, + "step": 9285 + }, + { + "epoch": 0.3, + "learning_rate": 1.648955642698228e-05, + "loss": 1.0698, + "step": 9286 + }, + { + "epoch": 0.3, + "learning_rate": 1.6488768325524146e-05, + "loss": 0.9312, + "step": 9287 + }, + { + "epoch": 0.3, + "learning_rate": 1.6487980154448713e-05, + "loss": 1.0396, + "step": 9288 + }, + { + "epoch": 0.3, + "learning_rate": 1.648719191376443e-05, + "loss": 1.0093, + "step": 9289 + }, + { + "epoch": 0.3, + "learning_rate": 1.648640360347976e-05, + "loss": 1.0562, + "step": 9290 + }, + { + "epoch": 0.3, + "learning_rate": 1.6485615223603157e-05, + "loss": 1.0293, + "step": 9291 + }, + { + "epoch": 0.3, + "learning_rate": 1.648482677414308e-05, + "loss": 1.0205, + "step": 9292 + }, + { + "epoch": 0.3, + "learning_rate": 1.648403825510799e-05, + "loss": 1.0508, + "step": 9293 + }, + { + "epoch": 0.3, + "learning_rate": 1.6483249666506346e-05, + "loss": 0.9385, + "step": 9294 + }, + { + "epoch": 0.3, + "learning_rate": 1.6482461008346605e-05, + "loss": 0.5369, + "step": 9295 + }, + { + "epoch": 0.3, + "learning_rate": 1.6481672280637237e-05, + "loss": 0.9243, + "step": 9296 + }, + { + "epoch": 0.3, + "learning_rate": 1.6480883483386697e-05, + "loss": 0.9407, + "step": 9297 + }, + { + "epoch": 0.3, + "learning_rate": 1.6480094616603447e-05, + "loss": 1.042, + "step": 9298 + }, + { + "epoch": 0.3, + "learning_rate": 1.6479305680295955e-05, + "loss": 1.0083, + "step": 9299 + }, + { + "epoch": 0.3, + "learning_rate": 1.6478516674472687e-05, + "loss": 1.145, + "step": 9300 + }, + { + "epoch": 0.3, + "learning_rate": 1.64777275991421e-05, + "loss": 0.9292, + "step": 9301 + }, + { + "epoch": 0.3, + "learning_rate": 1.6476938454312673e-05, + "loss": 0.9194, + "step": 9302 + }, + { + "epoch": 0.3, + "learning_rate": 1.6476149239992857e-05, + "loss": 1.0566, + "step": 9303 + }, + { + "epoch": 0.3, + "learning_rate": 1.647535995619113e-05, + "loss": 0.978, + "step": 9304 + }, + { + "epoch": 0.3, + "learning_rate": 1.6474570602915956e-05, + "loss": 1.1348, + "step": 9305 + }, + { + "epoch": 0.3, + "learning_rate": 1.6473781180175806e-05, + "loss": 1.0459, + "step": 9306 + }, + { + "epoch": 0.3, + "learning_rate": 1.647299168797915e-05, + "loss": 0.9214, + "step": 9307 + }, + { + "epoch": 0.3, + "learning_rate": 1.6472202126334456e-05, + "loss": 0.9624, + "step": 9308 + }, + { + "epoch": 0.3, + "learning_rate": 1.6471412495250195e-05, + "loss": 0.9634, + "step": 9309 + }, + { + "epoch": 0.3, + "learning_rate": 1.6470622794734844e-05, + "loss": 1.0591, + "step": 9310 + }, + { + "epoch": 0.3, + "learning_rate": 1.646983302479687e-05, + "loss": 1.021, + "step": 9311 + }, + { + "epoch": 0.3, + "learning_rate": 1.6469043185444744e-05, + "loss": 1.0425, + "step": 9312 + }, + { + "epoch": 0.3, + "learning_rate": 1.646825327668695e-05, + "loss": 1.0352, + "step": 9313 + }, + { + "epoch": 0.3, + "learning_rate": 1.6467463298531954e-05, + "loss": 1.1597, + "step": 9314 + }, + { + "epoch": 0.3, + "learning_rate": 1.6466673250988235e-05, + "loss": 1.0664, + "step": 9315 + }, + { + "epoch": 0.3, + "learning_rate": 1.646588313406427e-05, + "loss": 1.0601, + "step": 9316 + }, + { + "epoch": 0.3, + "learning_rate": 1.6465092947768533e-05, + "loss": 0.9648, + "step": 9317 + }, + { + "epoch": 0.3, + "learning_rate": 1.6464302692109506e-05, + "loss": 1.1602, + "step": 9318 + }, + { + "epoch": 0.3, + "learning_rate": 1.6463512367095665e-05, + "loss": 1.1689, + "step": 9319 + }, + { + "epoch": 0.3, + "learning_rate": 1.6462721972735488e-05, + "loss": 0.918, + "step": 9320 + }, + { + "epoch": 0.3, + "learning_rate": 1.6461931509037457e-05, + "loss": 1.1074, + "step": 9321 + }, + { + "epoch": 0.3, + "learning_rate": 1.6461140976010055e-05, + "loss": 1.0015, + "step": 9322 + }, + { + "epoch": 0.3, + "learning_rate": 1.646035037366176e-05, + "loss": 0.9639, + "step": 9323 + }, + { + "epoch": 0.3, + "learning_rate": 1.6459559702001054e-05, + "loss": 1.0674, + "step": 9324 + }, + { + "epoch": 0.3, + "learning_rate": 1.6458768961036424e-05, + "loss": 1.1157, + "step": 9325 + }, + { + "epoch": 0.3, + "learning_rate": 1.6457978150776348e-05, + "loss": 1.0137, + "step": 9326 + }, + { + "epoch": 0.3, + "learning_rate": 1.6457187271229318e-05, + "loss": 1.1011, + "step": 9327 + }, + { + "epoch": 0.3, + "learning_rate": 1.6456396322403812e-05, + "loss": 0.5449, + "step": 9328 + }, + { + "epoch": 0.3, + "learning_rate": 1.645560530430832e-05, + "loss": 1.126, + "step": 9329 + }, + { + "epoch": 0.3, + "learning_rate": 1.645481421695133e-05, + "loss": 1.1104, + "step": 9330 + }, + { + "epoch": 0.3, + "learning_rate": 1.645402306034132e-05, + "loss": 1.0137, + "step": 9331 + }, + { + "epoch": 0.3, + "learning_rate": 1.6453231834486795e-05, + "loss": 1.0518, + "step": 9332 + }, + { + "epoch": 0.3, + "learning_rate": 1.645244053939623e-05, + "loss": 0.958, + "step": 9333 + }, + { + "epoch": 0.3, + "learning_rate": 1.645164917507812e-05, + "loss": 1.0913, + "step": 9334 + }, + { + "epoch": 0.3, + "learning_rate": 1.6450857741540954e-05, + "loss": 1.0508, + "step": 9335 + }, + { + "epoch": 0.3, + "learning_rate": 1.6450066238793222e-05, + "loss": 1.1699, + "step": 9336 + }, + { + "epoch": 0.3, + "learning_rate": 1.644927466684342e-05, + "loss": 1.0054, + "step": 9337 + }, + { + "epoch": 0.3, + "learning_rate": 1.644848302570004e-05, + "loss": 0.9771, + "step": 9338 + }, + { + "epoch": 0.3, + "learning_rate": 1.644769131537157e-05, + "loss": 1.0835, + "step": 9339 + }, + { + "epoch": 0.3, + "learning_rate": 1.644689953586651e-05, + "loss": 0.5076, + "step": 9340 + }, + { + "epoch": 0.3, + "learning_rate": 1.6446107687193355e-05, + "loss": 0.9229, + "step": 9341 + }, + { + "epoch": 0.3, + "learning_rate": 1.6445315769360596e-05, + "loss": 0.5405, + "step": 9342 + }, + { + "epoch": 0.3, + "learning_rate": 1.6444523782376735e-05, + "loss": 1.0518, + "step": 9343 + }, + { + "epoch": 0.3, + "learning_rate": 1.6443731726250262e-05, + "loss": 1.1172, + "step": 9344 + }, + { + "epoch": 0.3, + "learning_rate": 1.644293960098968e-05, + "loss": 1.0811, + "step": 9345 + }, + { + "epoch": 0.3, + "learning_rate": 1.644214740660349e-05, + "loss": 0.9697, + "step": 9346 + }, + { + "epoch": 0.3, + "learning_rate": 1.6441355143100187e-05, + "loss": 0.896, + "step": 9347 + }, + { + "epoch": 0.3, + "learning_rate": 1.6440562810488276e-05, + "loss": 1.1699, + "step": 9348 + }, + { + "epoch": 0.3, + "learning_rate": 1.6439770408776246e-05, + "loss": 0.5688, + "step": 9349 + }, + { + "epoch": 0.3, + "learning_rate": 1.6438977937972614e-05, + "loss": 0.9878, + "step": 9350 + }, + { + "epoch": 0.3, + "learning_rate": 1.6438185398085868e-05, + "loss": 1.1245, + "step": 9351 + }, + { + "epoch": 0.3, + "learning_rate": 1.643739278912452e-05, + "loss": 1.0869, + "step": 9352 + }, + { + "epoch": 0.3, + "learning_rate": 1.6436600111097076e-05, + "loss": 1.0156, + "step": 9353 + }, + { + "epoch": 0.3, + "learning_rate": 1.6435807364012035e-05, + "loss": 0.9883, + "step": 9354 + }, + { + "epoch": 0.3, + "learning_rate": 1.6435014547877904e-05, + "loss": 1.1074, + "step": 9355 + }, + { + "epoch": 0.3, + "learning_rate": 1.6434221662703186e-05, + "loss": 1.0264, + "step": 9356 + }, + { + "epoch": 0.3, + "learning_rate": 1.6433428708496393e-05, + "loss": 1.0034, + "step": 9357 + }, + { + "epoch": 0.3, + "learning_rate": 1.643263568526603e-05, + "loss": 1.0337, + "step": 9358 + }, + { + "epoch": 0.3, + "learning_rate": 1.6431842593020603e-05, + "loss": 0.9033, + "step": 9359 + }, + { + "epoch": 0.3, + "learning_rate": 1.6431049431768628e-05, + "loss": 1.0039, + "step": 9360 + }, + { + "epoch": 0.3, + "learning_rate": 1.6430256201518605e-05, + "loss": 1.0659, + "step": 9361 + }, + { + "epoch": 0.3, + "learning_rate": 1.6429462902279054e-05, + "loss": 1.0103, + "step": 9362 + }, + { + "epoch": 0.3, + "learning_rate": 1.642866953405848e-05, + "loss": 1.0747, + "step": 9363 + }, + { + "epoch": 0.3, + "learning_rate": 1.6427876096865394e-05, + "loss": 0.9136, + "step": 9364 + }, + { + "epoch": 0.3, + "learning_rate": 1.6427082590708314e-05, + "loss": 0.9956, + "step": 9365 + }, + { + "epoch": 0.3, + "learning_rate": 1.6426289015595752e-05, + "loss": 0.9707, + "step": 9366 + }, + { + "epoch": 0.3, + "learning_rate": 1.642549537153622e-05, + "loss": 1.1289, + "step": 9367 + }, + { + "epoch": 0.3, + "learning_rate": 1.6424701658538235e-05, + "loss": 1.0781, + "step": 9368 + }, + { + "epoch": 0.3, + "learning_rate": 1.642390787661031e-05, + "loss": 0.9336, + "step": 9369 + }, + { + "epoch": 0.3, + "learning_rate": 1.6423114025760966e-05, + "loss": 1.0005, + "step": 9370 + }, + { + "epoch": 0.3, + "learning_rate": 1.642232010599871e-05, + "loss": 1.0474, + "step": 9371 + }, + { + "epoch": 0.3, + "learning_rate": 1.6421526117332078e-05, + "loss": 1.1167, + "step": 9372 + }, + { + "epoch": 0.3, + "learning_rate": 1.6420732059769573e-05, + "loss": 1.1582, + "step": 9373 + }, + { + "epoch": 0.3, + "learning_rate": 1.6419937933319716e-05, + "loss": 1.0918, + "step": 9374 + }, + { + "epoch": 0.3, + "learning_rate": 1.6419143737991033e-05, + "loss": 1.0747, + "step": 9375 + }, + { + "epoch": 0.3, + "learning_rate": 1.641834947379204e-05, + "loss": 0.9307, + "step": 9376 + }, + { + "epoch": 0.3, + "learning_rate": 1.6417555140731262e-05, + "loss": 1.0557, + "step": 9377 + }, + { + "epoch": 0.3, + "learning_rate": 1.641676073881722e-05, + "loss": 0.9263, + "step": 9378 + }, + { + "epoch": 0.3, + "learning_rate": 1.641596626805844e-05, + "loss": 1.064, + "step": 9379 + }, + { + "epoch": 0.3, + "learning_rate": 1.6415171728463438e-05, + "loss": 1.1191, + "step": 9380 + }, + { + "epoch": 0.3, + "learning_rate": 1.6414377120040747e-05, + "loss": 1.0532, + "step": 9381 + }, + { + "epoch": 0.3, + "learning_rate": 1.6413582442798886e-05, + "loss": 1.0371, + "step": 9382 + }, + { + "epoch": 0.3, + "learning_rate": 1.6412787696746383e-05, + "loss": 0.9585, + "step": 9383 + }, + { + "epoch": 0.3, + "learning_rate": 1.641199288189177e-05, + "loss": 1.1943, + "step": 9384 + }, + { + "epoch": 0.3, + "learning_rate": 1.6411197998243563e-05, + "loss": 1.0425, + "step": 9385 + }, + { + "epoch": 0.3, + "learning_rate": 1.6410403045810303e-05, + "loss": 1.0723, + "step": 9386 + }, + { + "epoch": 0.3, + "learning_rate": 1.6409608024600508e-05, + "loss": 1.1938, + "step": 9387 + }, + { + "epoch": 0.3, + "learning_rate": 1.6408812934622717e-05, + "loss": 1.1396, + "step": 9388 + }, + { + "epoch": 0.3, + "learning_rate": 1.6408017775885455e-05, + "loss": 1.0854, + "step": 9389 + }, + { + "epoch": 0.3, + "learning_rate": 1.6407222548397253e-05, + "loss": 0.8794, + "step": 9390 + }, + { + "epoch": 0.3, + "learning_rate": 1.6406427252166645e-05, + "loss": 1.0234, + "step": 9391 + }, + { + "epoch": 0.3, + "learning_rate": 1.6405631887202162e-05, + "loss": 0.9644, + "step": 9392 + }, + { + "epoch": 0.3, + "learning_rate": 1.640483645351234e-05, + "loss": 0.9595, + "step": 9393 + }, + { + "epoch": 0.3, + "learning_rate": 1.640404095110571e-05, + "loss": 1.0605, + "step": 9394 + }, + { + "epoch": 0.3, + "learning_rate": 1.640324537999081e-05, + "loss": 0.9287, + "step": 9395 + }, + { + "epoch": 0.3, + "learning_rate": 1.6402449740176175e-05, + "loss": 1.0181, + "step": 9396 + }, + { + "epoch": 0.3, + "learning_rate": 1.640165403167033e-05, + "loss": 0.939, + "step": 9397 + }, + { + "epoch": 0.3, + "learning_rate": 1.6400858254481833e-05, + "loss": 1.0254, + "step": 9398 + }, + { + "epoch": 0.3, + "learning_rate": 1.640006240861921e-05, + "loss": 0.9824, + "step": 9399 + }, + { + "epoch": 0.3, + "learning_rate": 1.6399266494091e-05, + "loss": 1.0093, + "step": 9400 + }, + { + "epoch": 0.3, + "learning_rate": 1.639847051090574e-05, + "loss": 1.0615, + "step": 9401 + }, + { + "epoch": 0.3, + "learning_rate": 1.6397674459071975e-05, + "loss": 1.2012, + "step": 9402 + }, + { + "epoch": 0.3, + "learning_rate": 1.6396878338598245e-05, + "loss": 0.8931, + "step": 9403 + }, + { + "epoch": 0.3, + "learning_rate": 1.6396082149493087e-05, + "loss": 0.8921, + "step": 9404 + }, + { + "epoch": 0.3, + "learning_rate": 1.6395285891765046e-05, + "loss": 1.0352, + "step": 9405 + }, + { + "epoch": 0.3, + "learning_rate": 1.6394489565422668e-05, + "loss": 1.0527, + "step": 9406 + }, + { + "epoch": 0.3, + "learning_rate": 1.639369317047449e-05, + "loss": 1.1108, + "step": 9407 + }, + { + "epoch": 0.3, + "learning_rate": 1.6392896706929065e-05, + "loss": 1.0107, + "step": 9408 + }, + { + "epoch": 0.3, + "learning_rate": 1.639210017479493e-05, + "loss": 0.9731, + "step": 9409 + }, + { + "epoch": 0.3, + "learning_rate": 1.639130357408064e-05, + "loss": 1.0547, + "step": 9410 + }, + { + "epoch": 0.3, + "learning_rate": 1.639050690479473e-05, + "loss": 1.0703, + "step": 9411 + }, + { + "epoch": 0.3, + "learning_rate": 1.638971016694576e-05, + "loss": 0.9658, + "step": 9412 + }, + { + "epoch": 0.3, + "learning_rate": 1.6388913360542265e-05, + "loss": 1.1016, + "step": 9413 + }, + { + "epoch": 0.3, + "learning_rate": 1.63881164855928e-05, + "loss": 1.0698, + "step": 9414 + }, + { + "epoch": 0.3, + "learning_rate": 1.6387319542105925e-05, + "loss": 1.0845, + "step": 9415 + }, + { + "epoch": 0.3, + "learning_rate": 1.6386522530090173e-05, + "loss": 1.1475, + "step": 9416 + }, + { + "epoch": 0.3, + "learning_rate": 1.6385725449554104e-05, + "loss": 1.0771, + "step": 9417 + }, + { + "epoch": 0.3, + "learning_rate": 1.638492830050627e-05, + "loss": 1.0132, + "step": 9418 + }, + { + "epoch": 0.3, + "learning_rate": 1.6384131082955218e-05, + "loss": 1.1943, + "step": 9419 + }, + { + "epoch": 0.3, + "learning_rate": 1.6383333796909505e-05, + "loss": 1.0806, + "step": 9420 + }, + { + "epoch": 0.3, + "learning_rate": 1.6382536442377688e-05, + "loss": 1.0537, + "step": 9421 + }, + { + "epoch": 0.3, + "learning_rate": 1.6381739019368316e-05, + "loss": 1.0854, + "step": 9422 + }, + { + "epoch": 0.3, + "learning_rate": 1.638094152788995e-05, + "loss": 1.1177, + "step": 9423 + }, + { + "epoch": 0.3, + "learning_rate": 1.6380143967951143e-05, + "loss": 1.0679, + "step": 9424 + }, + { + "epoch": 0.3, + "learning_rate": 1.6379346339560452e-05, + "loss": 0.9956, + "step": 9425 + }, + { + "epoch": 0.3, + "learning_rate": 1.6378548642726432e-05, + "loss": 1.001, + "step": 9426 + }, + { + "epoch": 0.3, + "learning_rate": 1.637775087745765e-05, + "loss": 1.0205, + "step": 9427 + }, + { + "epoch": 0.3, + "learning_rate": 1.6376953043762656e-05, + "loss": 1.0576, + "step": 9428 + }, + { + "epoch": 0.3, + "learning_rate": 1.6376155141650015e-05, + "loss": 1.1641, + "step": 9429 + }, + { + "epoch": 0.3, + "learning_rate": 1.6375357171128287e-05, + "loss": 1.0537, + "step": 9430 + }, + { + "epoch": 0.3, + "learning_rate": 1.637455913220603e-05, + "loss": 0.9004, + "step": 9431 + }, + { + "epoch": 0.3, + "learning_rate": 1.637376102489181e-05, + "loss": 1.0903, + "step": 9432 + }, + { + "epoch": 0.3, + "learning_rate": 1.637296284919419e-05, + "loss": 0.9937, + "step": 9433 + }, + { + "epoch": 0.3, + "learning_rate": 1.637216460512173e-05, + "loss": 1.0615, + "step": 9434 + }, + { + "epoch": 0.3, + "learning_rate": 1.6371366292682993e-05, + "loss": 1.1211, + "step": 9435 + }, + { + "epoch": 0.3, + "learning_rate": 1.637056791188655e-05, + "loss": 1.0366, + "step": 9436 + }, + { + "epoch": 0.3, + "learning_rate": 1.636976946274097e-05, + "loss": 1.0654, + "step": 9437 + }, + { + "epoch": 0.3, + "learning_rate": 1.6368970945254803e-05, + "loss": 1.0532, + "step": 9438 + }, + { + "epoch": 0.3, + "learning_rate": 1.6368172359436632e-05, + "loss": 1.0122, + "step": 9439 + }, + { + "epoch": 0.3, + "learning_rate": 1.6367373705295016e-05, + "loss": 0.9287, + "step": 9440 + }, + { + "epoch": 0.3, + "learning_rate": 1.636657498283853e-05, + "loss": 1.0249, + "step": 9441 + }, + { + "epoch": 0.3, + "learning_rate": 1.6365776192075742e-05, + "loss": 1.0464, + "step": 9442 + }, + { + "epoch": 0.3, + "learning_rate": 1.636497733301522e-05, + "loss": 1.0244, + "step": 9443 + }, + { + "epoch": 0.3, + "learning_rate": 1.6364178405665534e-05, + "loss": 1.0171, + "step": 9444 + }, + { + "epoch": 0.3, + "learning_rate": 1.636337941003526e-05, + "loss": 0.998, + "step": 9445 + }, + { + "epoch": 0.3, + "learning_rate": 1.6362580346132967e-05, + "loss": 1.1567, + "step": 9446 + }, + { + "epoch": 0.3, + "learning_rate": 1.6361781213967222e-05, + "loss": 1.0645, + "step": 9447 + }, + { + "epoch": 0.3, + "learning_rate": 1.636098201354661e-05, + "loss": 1.0322, + "step": 9448 + }, + { + "epoch": 0.3, + "learning_rate": 1.6360182744879707e-05, + "loss": 0.9355, + "step": 9449 + }, + { + "epoch": 0.3, + "learning_rate": 1.6359383407975077e-05, + "loss": 1.0186, + "step": 9450 + }, + { + "epoch": 0.3, + "learning_rate": 1.6358584002841297e-05, + "loss": 0.9834, + "step": 9451 + }, + { + "epoch": 0.3, + "learning_rate": 1.6357784529486954e-05, + "loss": 1.1846, + "step": 9452 + }, + { + "epoch": 0.3, + "learning_rate": 1.6356984987920617e-05, + "loss": 1.0122, + "step": 9453 + }, + { + "epoch": 0.3, + "learning_rate": 1.635618537815087e-05, + "loss": 1.0288, + "step": 9454 + }, + { + "epoch": 0.3, + "learning_rate": 1.6355385700186282e-05, + "loss": 1.2188, + "step": 9455 + }, + { + "epoch": 0.3, + "learning_rate": 1.6354585954035445e-05, + "loss": 1.1104, + "step": 9456 + }, + { + "epoch": 0.3, + "learning_rate": 1.6353786139706932e-05, + "loss": 0.9941, + "step": 9457 + }, + { + "epoch": 0.3, + "learning_rate": 1.6352986257209326e-05, + "loss": 1.1763, + "step": 9458 + }, + { + "epoch": 0.3, + "learning_rate": 1.6352186306551206e-05, + "loss": 1.0127, + "step": 9459 + }, + { + "epoch": 0.3, + "learning_rate": 1.635138628774116e-05, + "loss": 1.0996, + "step": 9460 + }, + { + "epoch": 0.3, + "learning_rate": 1.6350586200787768e-05, + "loss": 1.0034, + "step": 9461 + }, + { + "epoch": 0.3, + "learning_rate": 1.6349786045699612e-05, + "loss": 0.9648, + "step": 9462 + }, + { + "epoch": 0.3, + "learning_rate": 1.6348985822485283e-05, + "loss": 1.0444, + "step": 9463 + }, + { + "epoch": 0.3, + "learning_rate": 1.634818553115336e-05, + "loss": 1.0586, + "step": 9464 + }, + { + "epoch": 0.3, + "learning_rate": 1.6347385171712434e-05, + "loss": 1.0439, + "step": 9465 + }, + { + "epoch": 0.3, + "learning_rate": 1.634658474417109e-05, + "loss": 1.0547, + "step": 9466 + }, + { + "epoch": 0.3, + "learning_rate": 1.6345784248537917e-05, + "loss": 1.0908, + "step": 9467 + }, + { + "epoch": 0.3, + "learning_rate": 1.63449836848215e-05, + "loss": 1.0073, + "step": 9468 + }, + { + "epoch": 0.3, + "learning_rate": 1.634418305303043e-05, + "loss": 1.0708, + "step": 9469 + }, + { + "epoch": 0.3, + "learning_rate": 1.63433823531733e-05, + "loss": 1.0967, + "step": 9470 + }, + { + "epoch": 0.3, + "learning_rate": 1.6342581585258695e-05, + "loss": 0.9141, + "step": 9471 + }, + { + "epoch": 0.3, + "learning_rate": 1.634178074929521e-05, + "loss": 1.1025, + "step": 9472 + }, + { + "epoch": 0.3, + "learning_rate": 1.6340979845291435e-05, + "loss": 1.0186, + "step": 9473 + }, + { + "epoch": 0.3, + "learning_rate": 1.6340178873255965e-05, + "loss": 1.0469, + "step": 9474 + }, + { + "epoch": 0.3, + "learning_rate": 1.6339377833197394e-05, + "loss": 1.1973, + "step": 9475 + }, + { + "epoch": 0.3, + "learning_rate": 1.6338576725124314e-05, + "loss": 1.0562, + "step": 9476 + }, + { + "epoch": 0.3, + "learning_rate": 1.633777554904532e-05, + "loss": 1.1353, + "step": 9477 + }, + { + "epoch": 0.3, + "learning_rate": 1.633697430496901e-05, + "loss": 1.0425, + "step": 9478 + }, + { + "epoch": 0.3, + "learning_rate": 1.633617299290398e-05, + "loss": 0.9546, + "step": 9479 + }, + { + "epoch": 0.3, + "learning_rate": 1.6335371612858827e-05, + "loss": 1.1919, + "step": 9480 + }, + { + "epoch": 0.3, + "learning_rate": 1.633457016484214e-05, + "loss": 1.2114, + "step": 9481 + }, + { + "epoch": 0.3, + "learning_rate": 1.6333768648862537e-05, + "loss": 1.0142, + "step": 9482 + }, + { + "epoch": 0.3, + "learning_rate": 1.63329670649286e-05, + "loss": 1.1084, + "step": 9483 + }, + { + "epoch": 0.3, + "learning_rate": 1.6332165413048936e-05, + "loss": 0.9966, + "step": 9484 + }, + { + "epoch": 0.3, + "learning_rate": 1.6331363693232147e-05, + "loss": 1.0327, + "step": 9485 + }, + { + "epoch": 0.3, + "learning_rate": 1.6330561905486832e-05, + "loss": 0.9448, + "step": 9486 + }, + { + "epoch": 0.3, + "learning_rate": 1.6329760049821595e-05, + "loss": 0.9458, + "step": 9487 + }, + { + "epoch": 0.3, + "learning_rate": 1.6328958126245037e-05, + "loss": 1.167, + "step": 9488 + }, + { + "epoch": 0.3, + "learning_rate": 1.632815613476576e-05, + "loss": 1.0518, + "step": 9489 + }, + { + "epoch": 0.3, + "learning_rate": 1.6327354075392376e-05, + "loss": 1.0049, + "step": 9490 + }, + { + "epoch": 0.3, + "learning_rate": 1.6326551948133485e-05, + "loss": 1.0186, + "step": 9491 + }, + { + "epoch": 0.3, + "learning_rate": 1.632574975299769e-05, + "loss": 1.0405, + "step": 9492 + }, + { + "epoch": 0.3, + "learning_rate": 1.6324947489993603e-05, + "loss": 1.1045, + "step": 9493 + }, + { + "epoch": 0.3, + "learning_rate": 1.6324145159129835e-05, + "loss": 1.0078, + "step": 9494 + }, + { + "epoch": 0.3, + "learning_rate": 1.632334276041498e-05, + "loss": 1.0381, + "step": 9495 + }, + { + "epoch": 0.3, + "learning_rate": 1.6322540293857662e-05, + "loss": 1.042, + "step": 9496 + }, + { + "epoch": 0.3, + "learning_rate": 1.632173775946648e-05, + "loss": 0.9819, + "step": 9497 + }, + { + "epoch": 0.3, + "learning_rate": 1.6320935157250048e-05, + "loss": 1.0649, + "step": 9498 + }, + { + "epoch": 0.3, + "learning_rate": 1.6320132487216977e-05, + "loss": 1.1436, + "step": 9499 + }, + { + "epoch": 0.3, + "learning_rate": 1.6319329749375884e-05, + "loss": 1.1475, + "step": 9500 + }, + { + "epoch": 0.3, + "learning_rate": 1.6318526943735373e-05, + "loss": 1.1147, + "step": 9501 + }, + { + "epoch": 0.3, + "learning_rate": 1.631772407030406e-05, + "loss": 1.0454, + "step": 9502 + }, + { + "epoch": 0.3, + "learning_rate": 1.6316921129090563e-05, + "loss": 1.0518, + "step": 9503 + }, + { + "epoch": 0.3, + "learning_rate": 1.6316118120103492e-05, + "loss": 1.0796, + "step": 9504 + }, + { + "epoch": 0.3, + "learning_rate": 1.6315315043351465e-05, + "loss": 1.0762, + "step": 9505 + }, + { + "epoch": 0.3, + "learning_rate": 1.6314511898843094e-05, + "loss": 0.9731, + "step": 9506 + }, + { + "epoch": 0.3, + "learning_rate": 1.6313708686587002e-05, + "loss": 1.1021, + "step": 9507 + }, + { + "epoch": 0.3, + "learning_rate": 1.63129054065918e-05, + "loss": 1.0596, + "step": 9508 + }, + { + "epoch": 0.3, + "learning_rate": 1.6312102058866115e-05, + "loss": 0.958, + "step": 9509 + }, + { + "epoch": 0.3, + "learning_rate": 1.6311298643418555e-05, + "loss": 1.0352, + "step": 9510 + }, + { + "epoch": 0.3, + "learning_rate": 1.631049516025775e-05, + "loss": 1.0991, + "step": 9511 + }, + { + "epoch": 0.3, + "learning_rate": 1.6309691609392314e-05, + "loss": 0.9663, + "step": 9512 + }, + { + "epoch": 0.3, + "learning_rate": 1.630888799083087e-05, + "loss": 1.1553, + "step": 9513 + }, + { + "epoch": 0.3, + "learning_rate": 1.6308084304582043e-05, + "loss": 1.0312, + "step": 9514 + }, + { + "epoch": 0.3, + "learning_rate": 1.6307280550654453e-05, + "loss": 1.1846, + "step": 9515 + }, + { + "epoch": 0.3, + "learning_rate": 1.630647672905672e-05, + "loss": 1.0991, + "step": 9516 + }, + { + "epoch": 0.3, + "learning_rate": 1.6305672839797472e-05, + "loss": 1.0166, + "step": 9517 + }, + { + "epoch": 0.3, + "learning_rate": 1.6304868882885334e-05, + "loss": 1.021, + "step": 9518 + }, + { + "epoch": 0.3, + "learning_rate": 1.6304064858328928e-05, + "loss": 1.1309, + "step": 9519 + }, + { + "epoch": 0.3, + "learning_rate": 1.6303260766136887e-05, + "loss": 1.0181, + "step": 9520 + }, + { + "epoch": 0.3, + "learning_rate": 1.630245660631783e-05, + "loss": 0.9463, + "step": 9521 + }, + { + "epoch": 0.3, + "learning_rate": 1.6301652378880394e-05, + "loss": 1.0933, + "step": 9522 + }, + { + "epoch": 0.3, + "learning_rate": 1.6300848083833198e-05, + "loss": 1.0454, + "step": 9523 + }, + { + "epoch": 0.3, + "learning_rate": 1.630004372118488e-05, + "loss": 1.0479, + "step": 9524 + }, + { + "epoch": 0.3, + "learning_rate": 1.6299239290944062e-05, + "loss": 1.02, + "step": 9525 + }, + { + "epoch": 0.3, + "learning_rate": 1.6298434793119374e-05, + "loss": 1.0933, + "step": 9526 + }, + { + "epoch": 0.3, + "learning_rate": 1.6297630227719458e-05, + "loss": 1.1123, + "step": 9527 + }, + { + "epoch": 0.3, + "learning_rate": 1.629682559475294e-05, + "loss": 1.0273, + "step": 9528 + }, + { + "epoch": 0.3, + "learning_rate": 1.629602089422845e-05, + "loss": 1.0371, + "step": 9529 + }, + { + "epoch": 0.3, + "learning_rate": 1.6295216126154624e-05, + "loss": 1.1035, + "step": 9530 + }, + { + "epoch": 0.3, + "learning_rate": 1.629441129054009e-05, + "loss": 1.0825, + "step": 9531 + }, + { + "epoch": 0.3, + "learning_rate": 1.6293606387393502e-05, + "loss": 0.9897, + "step": 9532 + }, + { + "epoch": 0.3, + "learning_rate": 1.6292801416723473e-05, + "loss": 0.9016, + "step": 9533 + }, + { + "epoch": 0.3, + "learning_rate": 1.6291996378538653e-05, + "loss": 0.9517, + "step": 9534 + }, + { + "epoch": 0.3, + "learning_rate": 1.6291191272847676e-05, + "loss": 1.1001, + "step": 9535 + }, + { + "epoch": 0.3, + "learning_rate": 1.629038609965918e-05, + "loss": 1.0176, + "step": 9536 + }, + { + "epoch": 0.3, + "learning_rate": 1.6289580858981796e-05, + "loss": 1.1514, + "step": 9537 + }, + { + "epoch": 0.31, + "learning_rate": 1.6288775550824178e-05, + "loss": 1.1089, + "step": 9538 + }, + { + "epoch": 0.31, + "learning_rate": 1.6287970175194957e-05, + "loss": 0.9126, + "step": 9539 + }, + { + "epoch": 0.31, + "learning_rate": 1.6287164732102774e-05, + "loss": 1.0317, + "step": 9540 + }, + { + "epoch": 0.31, + "learning_rate": 1.6286359221556272e-05, + "loss": 1.043, + "step": 9541 + }, + { + "epoch": 0.31, + "learning_rate": 1.628555364356409e-05, + "loss": 0.9604, + "step": 9542 + }, + { + "epoch": 0.31, + "learning_rate": 1.628474799813488e-05, + "loss": 0.9297, + "step": 9543 + }, + { + "epoch": 0.31, + "learning_rate": 1.6283942285277278e-05, + "loss": 1.0654, + "step": 9544 + }, + { + "epoch": 0.31, + "learning_rate": 1.6283136504999927e-05, + "loss": 0.5293, + "step": 9545 + }, + { + "epoch": 0.31, + "learning_rate": 1.6282330657311476e-05, + "loss": 1.0684, + "step": 9546 + }, + { + "epoch": 0.31, + "learning_rate": 1.6281524742220567e-05, + "loss": 1.0532, + "step": 9547 + }, + { + "epoch": 0.31, + "learning_rate": 1.6280718759735855e-05, + "loss": 0.9995, + "step": 9548 + }, + { + "epoch": 0.31, + "learning_rate": 1.627991270986598e-05, + "loss": 1.0386, + "step": 9549 + }, + { + "epoch": 0.31, + "learning_rate": 1.6279106592619593e-05, + "loss": 0.9136, + "step": 9550 + }, + { + "epoch": 0.31, + "learning_rate": 1.627830040800534e-05, + "loss": 1.1167, + "step": 9551 + }, + { + "epoch": 0.31, + "learning_rate": 1.6277494156031873e-05, + "loss": 1.2104, + "step": 9552 + }, + { + "epoch": 0.31, + "learning_rate": 1.627668783670784e-05, + "loss": 1.0854, + "step": 9553 + }, + { + "epoch": 0.31, + "learning_rate": 1.6275881450041896e-05, + "loss": 1.0249, + "step": 9554 + }, + { + "epoch": 0.31, + "learning_rate": 1.6275074996042687e-05, + "loss": 1.1914, + "step": 9555 + }, + { + "epoch": 0.31, + "learning_rate": 1.627426847471887e-05, + "loss": 1.0791, + "step": 9556 + }, + { + "epoch": 0.31, + "learning_rate": 1.6273461886079095e-05, + "loss": 1.0264, + "step": 9557 + }, + { + "epoch": 0.31, + "learning_rate": 1.6272655230132016e-05, + "loss": 1.0151, + "step": 9558 + }, + { + "epoch": 0.31, + "learning_rate": 1.627184850688629e-05, + "loss": 0.9941, + "step": 9559 + }, + { + "epoch": 0.31, + "learning_rate": 1.6271041716350572e-05, + "loss": 1.0864, + "step": 9560 + }, + { + "epoch": 0.31, + "learning_rate": 1.6270234858533512e-05, + "loss": 1.083, + "step": 9561 + }, + { + "epoch": 0.31, + "learning_rate": 1.626942793344378e-05, + "loss": 1.0601, + "step": 9562 + }, + { + "epoch": 0.31, + "learning_rate": 1.626862094109002e-05, + "loss": 1.1074, + "step": 9563 + }, + { + "epoch": 0.31, + "learning_rate": 1.6267813881480892e-05, + "loss": 1.0537, + "step": 9564 + }, + { + "epoch": 0.31, + "learning_rate": 1.626700675462506e-05, + "loss": 1.1099, + "step": 9565 + }, + { + "epoch": 0.31, + "learning_rate": 1.6266199560531186e-05, + "loss": 1.0908, + "step": 9566 + }, + { + "epoch": 0.31, + "learning_rate": 1.626539229920792e-05, + "loss": 0.9536, + "step": 9567 + }, + { + "epoch": 0.31, + "learning_rate": 1.626458497066393e-05, + "loss": 0.9736, + "step": 9568 + }, + { + "epoch": 0.31, + "learning_rate": 1.626377757490788e-05, + "loss": 1.0088, + "step": 9569 + }, + { + "epoch": 0.31, + "learning_rate": 1.626297011194843e-05, + "loss": 1.0894, + "step": 9570 + }, + { + "epoch": 0.31, + "learning_rate": 1.6262162581794237e-05, + "loss": 0.9966, + "step": 9571 + }, + { + "epoch": 0.31, + "learning_rate": 1.626135498445397e-05, + "loss": 1.0371, + "step": 9572 + }, + { + "epoch": 0.31, + "learning_rate": 1.6260547319936297e-05, + "loss": 1.0933, + "step": 9573 + }, + { + "epoch": 0.31, + "learning_rate": 1.625973958824988e-05, + "loss": 0.9053, + "step": 9574 + }, + { + "epoch": 0.31, + "learning_rate": 1.6258931789403386e-05, + "loss": 1.0576, + "step": 9575 + }, + { + "epoch": 0.31, + "learning_rate": 1.625812392340548e-05, + "loss": 0.8589, + "step": 9576 + }, + { + "epoch": 0.31, + "learning_rate": 1.6257315990264828e-05, + "loss": 1.061, + "step": 9577 + }, + { + "epoch": 0.31, + "learning_rate": 1.62565079899901e-05, + "loss": 0.9834, + "step": 9578 + }, + { + "epoch": 0.31, + "learning_rate": 1.6255699922589968e-05, + "loss": 1.1538, + "step": 9579 + }, + { + "epoch": 0.31, + "learning_rate": 1.62548917880731e-05, + "loss": 1.0752, + "step": 9580 + }, + { + "epoch": 0.31, + "learning_rate": 1.6254083586448165e-05, + "loss": 1.0117, + "step": 9581 + }, + { + "epoch": 0.31, + "learning_rate": 1.6253275317723835e-05, + "loss": 0.8789, + "step": 9582 + }, + { + "epoch": 0.31, + "learning_rate": 1.6252466981908783e-05, + "loss": 1.2271, + "step": 9583 + }, + { + "epoch": 0.31, + "learning_rate": 1.625165857901168e-05, + "loss": 1.0137, + "step": 9584 + }, + { + "epoch": 0.31, + "learning_rate": 1.6250850109041194e-05, + "loss": 1.0552, + "step": 9585 + }, + { + "epoch": 0.31, + "learning_rate": 1.625004157200601e-05, + "loss": 0.853, + "step": 9586 + }, + { + "epoch": 0.31, + "learning_rate": 1.6249232967914796e-05, + "loss": 1.0142, + "step": 9587 + }, + { + "epoch": 0.31, + "learning_rate": 1.6248424296776225e-05, + "loss": 1.1445, + "step": 9588 + }, + { + "epoch": 0.31, + "learning_rate": 1.6247615558598982e-05, + "loss": 1.0381, + "step": 9589 + }, + { + "epoch": 0.31, + "learning_rate": 1.624680675339174e-05, + "loss": 1.0732, + "step": 9590 + }, + { + "epoch": 0.31, + "learning_rate": 1.6245997881163166e-05, + "loss": 1.1157, + "step": 9591 + }, + { + "epoch": 0.31, + "learning_rate": 1.6245188941921958e-05, + "loss": 0.9878, + "step": 9592 + }, + { + "epoch": 0.31, + "learning_rate": 1.6244379935676774e-05, + "loss": 0.9229, + "step": 9593 + }, + { + "epoch": 0.31, + "learning_rate": 1.6243570862436312e-05, + "loss": 1.1841, + "step": 9594 + }, + { + "epoch": 0.31, + "learning_rate": 1.6242761722209242e-05, + "loss": 0.9849, + "step": 9595 + }, + { + "epoch": 0.31, + "learning_rate": 1.6241952515004248e-05, + "loss": 0.9282, + "step": 9596 + }, + { + "epoch": 0.31, + "learning_rate": 1.6241143240830014e-05, + "loss": 0.9268, + "step": 9597 + }, + { + "epoch": 0.31, + "learning_rate": 1.6240333899695216e-05, + "loss": 0.5454, + "step": 9598 + }, + { + "epoch": 0.31, + "learning_rate": 1.6239524491608546e-05, + "loss": 1.2432, + "step": 9599 + }, + { + "epoch": 0.31, + "learning_rate": 1.623871501657868e-05, + "loss": 1.1416, + "step": 9600 + }, + { + "epoch": 0.31, + "learning_rate": 1.623790547461431e-05, + "loss": 0.9424, + "step": 9601 + }, + { + "epoch": 0.31, + "learning_rate": 1.6237095865724115e-05, + "loss": 1.0493, + "step": 9602 + }, + { + "epoch": 0.31, + "learning_rate": 1.6236286189916785e-05, + "loss": 0.979, + "step": 9603 + }, + { + "epoch": 0.31, + "learning_rate": 1.6235476447201007e-05, + "loss": 0.8564, + "step": 9604 + }, + { + "epoch": 0.31, + "learning_rate": 1.623466663758547e-05, + "loss": 1.0518, + "step": 9605 + }, + { + "epoch": 0.31, + "learning_rate": 1.6233856761078856e-05, + "loss": 0.8469, + "step": 9606 + }, + { + "epoch": 0.31, + "learning_rate": 1.623304681768986e-05, + "loss": 1.1172, + "step": 9607 + }, + { + "epoch": 0.31, + "learning_rate": 1.623223680742717e-05, + "loss": 1.1328, + "step": 9608 + }, + { + "epoch": 0.31, + "learning_rate": 1.6231426730299474e-05, + "loss": 1.0913, + "step": 9609 + }, + { + "epoch": 0.31, + "learning_rate": 1.623061658631547e-05, + "loss": 1.0664, + "step": 9610 + }, + { + "epoch": 0.31, + "learning_rate": 1.6229806375483846e-05, + "loss": 1.1045, + "step": 9611 + }, + { + "epoch": 0.31, + "learning_rate": 1.6228996097813294e-05, + "loss": 1.0264, + "step": 9612 + }, + { + "epoch": 0.31, + "learning_rate": 1.6228185753312503e-05, + "loss": 1.0742, + "step": 9613 + }, + { + "epoch": 0.31, + "learning_rate": 1.6227375341990177e-05, + "loss": 0.9131, + "step": 9614 + }, + { + "epoch": 0.31, + "learning_rate": 1.6226564863855004e-05, + "loss": 0.9771, + "step": 9615 + }, + { + "epoch": 0.31, + "learning_rate": 1.6225754318915684e-05, + "loss": 1.0674, + "step": 9616 + }, + { + "epoch": 0.31, + "learning_rate": 1.6224943707180905e-05, + "loss": 1.0166, + "step": 9617 + }, + { + "epoch": 0.31, + "learning_rate": 1.6224133028659373e-05, + "loss": 1.0527, + "step": 9618 + }, + { + "epoch": 0.31, + "learning_rate": 1.622332228335978e-05, + "loss": 1.103, + "step": 9619 + }, + { + "epoch": 0.31, + "learning_rate": 1.622251147129083e-05, + "loss": 1.0493, + "step": 9620 + }, + { + "epoch": 0.31, + "learning_rate": 1.6221700592461214e-05, + "loss": 1.1265, + "step": 9621 + }, + { + "epoch": 0.31, + "learning_rate": 1.622088964687964e-05, + "loss": 0.9858, + "step": 9622 + }, + { + "epoch": 0.31, + "learning_rate": 1.6220078634554804e-05, + "loss": 1.1201, + "step": 9623 + }, + { + "epoch": 0.31, + "learning_rate": 1.6219267555495407e-05, + "loss": 1.0273, + "step": 9624 + }, + { + "epoch": 0.31, + "learning_rate": 1.6218456409710152e-05, + "loss": 1.0078, + "step": 9625 + }, + { + "epoch": 0.31, + "learning_rate": 1.6217645197207744e-05, + "loss": 0.9888, + "step": 9626 + }, + { + "epoch": 0.31, + "learning_rate": 1.6216833917996884e-05, + "loss": 1.001, + "step": 9627 + }, + { + "epoch": 0.31, + "learning_rate": 1.6216022572086275e-05, + "loss": 1.0103, + "step": 9628 + }, + { + "epoch": 0.31, + "learning_rate": 1.6215211159484625e-05, + "loss": 1.0718, + "step": 9629 + }, + { + "epoch": 0.31, + "learning_rate": 1.621439968020064e-05, + "loss": 0.916, + "step": 9630 + }, + { + "epoch": 0.31, + "learning_rate": 1.621358813424302e-05, + "loss": 1.0576, + "step": 9631 + }, + { + "epoch": 0.31, + "learning_rate": 1.621277652162048e-05, + "loss": 1.1108, + "step": 9632 + }, + { + "epoch": 0.31, + "learning_rate": 1.621196484234172e-05, + "loss": 1.1533, + "step": 9633 + }, + { + "epoch": 0.31, + "learning_rate": 1.6211153096415456e-05, + "loss": 0.9146, + "step": 9634 + }, + { + "epoch": 0.31, + "learning_rate": 1.6210341283850392e-05, + "loss": 1.0503, + "step": 9635 + }, + { + "epoch": 0.31, + "learning_rate": 1.620952940465524e-05, + "loss": 0.9771, + "step": 9636 + }, + { + "epoch": 0.31, + "learning_rate": 1.6208717458838706e-05, + "loss": 1.1743, + "step": 9637 + }, + { + "epoch": 0.31, + "learning_rate": 1.620790544640951e-05, + "loss": 1.1338, + "step": 9638 + }, + { + "epoch": 0.31, + "learning_rate": 1.6207093367376353e-05, + "loss": 1.02, + "step": 9639 + }, + { + "epoch": 0.31, + "learning_rate": 1.6206281221747963e-05, + "loss": 1.0713, + "step": 9640 + }, + { + "epoch": 0.31, + "learning_rate": 1.620546900953304e-05, + "loss": 0.9495, + "step": 9641 + }, + { + "epoch": 0.31, + "learning_rate": 1.6204656730740298e-05, + "loss": 0.9297, + "step": 9642 + }, + { + "epoch": 0.31, + "learning_rate": 1.620384438537846e-05, + "loss": 1.147, + "step": 9643 + }, + { + "epoch": 0.31, + "learning_rate": 1.620303197345624e-05, + "loss": 1.1074, + "step": 9644 + }, + { + "epoch": 0.31, + "learning_rate": 1.620221949498235e-05, + "loss": 1.0576, + "step": 9645 + }, + { + "epoch": 0.31, + "learning_rate": 1.6201406949965506e-05, + "loss": 0.9219, + "step": 9646 + }, + { + "epoch": 0.31, + "learning_rate": 1.6200594338414432e-05, + "loss": 0.9199, + "step": 9647 + }, + { + "epoch": 0.31, + "learning_rate": 1.619978166033784e-05, + "loss": 0.9678, + "step": 9648 + }, + { + "epoch": 0.31, + "learning_rate": 1.6198968915744458e-05, + "loss": 1.064, + "step": 9649 + }, + { + "epoch": 0.31, + "learning_rate": 1.6198156104642996e-05, + "loss": 1.0342, + "step": 9650 + }, + { + "epoch": 0.31, + "learning_rate": 1.619734322704218e-05, + "loss": 0.9575, + "step": 9651 + }, + { + "epoch": 0.31, + "learning_rate": 1.619653028295073e-05, + "loss": 1.0342, + "step": 9652 + }, + { + "epoch": 0.31, + "learning_rate": 1.619571727237737e-05, + "loss": 0.9414, + "step": 9653 + }, + { + "epoch": 0.31, + "learning_rate": 1.6194904195330816e-05, + "loss": 1.1069, + "step": 9654 + }, + { + "epoch": 0.31, + "learning_rate": 1.6194091051819802e-05, + "loss": 1.0112, + "step": 9655 + }, + { + "epoch": 0.31, + "learning_rate": 1.6193277841853043e-05, + "loss": 1.0293, + "step": 9656 + }, + { + "epoch": 0.31, + "learning_rate": 1.619246456543927e-05, + "loss": 1.0107, + "step": 9657 + }, + { + "epoch": 0.31, + "learning_rate": 1.6191651222587203e-05, + "loss": 1.1484, + "step": 9658 + }, + { + "epoch": 0.31, + "learning_rate": 1.6190837813305568e-05, + "loss": 1.0713, + "step": 9659 + }, + { + "epoch": 0.31, + "learning_rate": 1.6190024337603102e-05, + "loss": 1.1367, + "step": 9660 + }, + { + "epoch": 0.31, + "learning_rate": 1.618921079548852e-05, + "loss": 0.9771, + "step": 9661 + }, + { + "epoch": 0.31, + "learning_rate": 1.6188397186970557e-05, + "loss": 1.1035, + "step": 9662 + }, + { + "epoch": 0.31, + "learning_rate": 1.618758351205794e-05, + "loss": 0.959, + "step": 9663 + }, + { + "epoch": 0.31, + "learning_rate": 1.6186769770759407e-05, + "loss": 1.0239, + "step": 9664 + }, + { + "epoch": 0.31, + "learning_rate": 1.6185955963083672e-05, + "loss": 1.1211, + "step": 9665 + }, + { + "epoch": 0.31, + "learning_rate": 1.6185142089039484e-05, + "loss": 1.0029, + "step": 9666 + }, + { + "epoch": 0.31, + "learning_rate": 1.6184328148635563e-05, + "loss": 1.1279, + "step": 9667 + }, + { + "epoch": 0.31, + "learning_rate": 1.6183514141880645e-05, + "loss": 0.9937, + "step": 9668 + }, + { + "epoch": 0.31, + "learning_rate": 1.6182700068783463e-05, + "loss": 1.0171, + "step": 9669 + }, + { + "epoch": 0.31, + "learning_rate": 1.6181885929352753e-05, + "loss": 1.0083, + "step": 9670 + }, + { + "epoch": 0.31, + "learning_rate": 1.618107172359725e-05, + "loss": 1.0454, + "step": 9671 + }, + { + "epoch": 0.31, + "learning_rate": 1.618025745152569e-05, + "loss": 1.0449, + "step": 9672 + }, + { + "epoch": 0.31, + "learning_rate": 1.61794431131468e-05, + "loss": 0.9727, + "step": 9673 + }, + { + "epoch": 0.31, + "learning_rate": 1.617862870846933e-05, + "loss": 0.9976, + "step": 9674 + }, + { + "epoch": 0.31, + "learning_rate": 1.6177814237502012e-05, + "loss": 1.1191, + "step": 9675 + }, + { + "epoch": 0.31, + "learning_rate": 1.6176999700253584e-05, + "loss": 1.1084, + "step": 9676 + }, + { + "epoch": 0.31, + "learning_rate": 1.6176185096732787e-05, + "loss": 1.0327, + "step": 9677 + }, + { + "epoch": 0.31, + "learning_rate": 1.617537042694836e-05, + "loss": 1.0083, + "step": 9678 + }, + { + "epoch": 0.31, + "learning_rate": 1.6174555690909037e-05, + "loss": 1.123, + "step": 9679 + }, + { + "epoch": 0.31, + "learning_rate": 1.617374088862357e-05, + "loss": 1.0552, + "step": 9680 + }, + { + "epoch": 0.31, + "learning_rate": 1.6172926020100695e-05, + "loss": 0.937, + "step": 9681 + }, + { + "epoch": 0.31, + "learning_rate": 1.6172111085349155e-05, + "loss": 1.1392, + "step": 9682 + }, + { + "epoch": 0.31, + "learning_rate": 1.6171296084377693e-05, + "loss": 1.0127, + "step": 9683 + }, + { + "epoch": 0.31, + "learning_rate": 1.617048101719506e-05, + "loss": 1.0117, + "step": 9684 + }, + { + "epoch": 0.31, + "learning_rate": 1.616966588380999e-05, + "loss": 0.9829, + "step": 9685 + }, + { + "epoch": 0.31, + "learning_rate": 1.6168850684231233e-05, + "loss": 1.1279, + "step": 9686 + }, + { + "epoch": 0.31, + "learning_rate": 1.6168035418467535e-05, + "loss": 1.0742, + "step": 9687 + }, + { + "epoch": 0.31, + "learning_rate": 1.6167220086527642e-05, + "loss": 0.9717, + "step": 9688 + }, + { + "epoch": 0.31, + "learning_rate": 1.6166404688420308e-05, + "loss": 0.9639, + "step": 9689 + }, + { + "epoch": 0.31, + "learning_rate": 1.6165589224154276e-05, + "loss": 1.0029, + "step": 9690 + }, + { + "epoch": 0.31, + "learning_rate": 1.616477369373829e-05, + "loss": 1.082, + "step": 9691 + }, + { + "epoch": 0.31, + "learning_rate": 1.6163958097181108e-05, + "loss": 1.0664, + "step": 9692 + }, + { + "epoch": 0.31, + "learning_rate": 1.616314243449148e-05, + "loss": 1.0669, + "step": 9693 + }, + { + "epoch": 0.31, + "learning_rate": 1.616232670567815e-05, + "loss": 1.1021, + "step": 9694 + }, + { + "epoch": 0.31, + "learning_rate": 1.616151091074988e-05, + "loss": 0.999, + "step": 9695 + }, + { + "epoch": 0.31, + "learning_rate": 1.6160695049715415e-05, + "loss": 1.0781, + "step": 9696 + }, + { + "epoch": 0.31, + "learning_rate": 1.615987912258351e-05, + "loss": 1.0664, + "step": 9697 + }, + { + "epoch": 0.31, + "learning_rate": 1.615906312936292e-05, + "loss": 1.0439, + "step": 9698 + }, + { + "epoch": 0.31, + "learning_rate": 1.6158247070062397e-05, + "loss": 1.0239, + "step": 9699 + }, + { + "epoch": 0.31, + "learning_rate": 1.61574309446907e-05, + "loss": 0.9365, + "step": 9700 + }, + { + "epoch": 0.31, + "learning_rate": 1.6156614753256583e-05, + "loss": 0.9868, + "step": 9701 + }, + { + "epoch": 0.31, + "learning_rate": 1.6155798495768804e-05, + "loss": 1.0273, + "step": 9702 + }, + { + "epoch": 0.31, + "learning_rate": 1.615498217223612e-05, + "loss": 0.9521, + "step": 9703 + }, + { + "epoch": 0.31, + "learning_rate": 1.615416578266729e-05, + "loss": 1.084, + "step": 9704 + }, + { + "epoch": 0.31, + "learning_rate": 1.6153349327071074e-05, + "loss": 1.0767, + "step": 9705 + }, + { + "epoch": 0.31, + "learning_rate": 1.6152532805456226e-05, + "loss": 0.5525, + "step": 9706 + }, + { + "epoch": 0.31, + "learning_rate": 1.615171621783151e-05, + "loss": 0.9424, + "step": 9707 + }, + { + "epoch": 0.31, + "learning_rate": 1.6150899564205694e-05, + "loss": 1.1196, + "step": 9708 + }, + { + "epoch": 0.31, + "learning_rate": 1.6150082844587526e-05, + "loss": 1.0664, + "step": 9709 + }, + { + "epoch": 0.31, + "learning_rate": 1.6149266058985777e-05, + "loss": 1.106, + "step": 9710 + }, + { + "epoch": 0.31, + "learning_rate": 1.614844920740921e-05, + "loss": 1.1187, + "step": 9711 + }, + { + "epoch": 0.31, + "learning_rate": 1.614763228986659e-05, + "loss": 0.9731, + "step": 9712 + }, + { + "epoch": 0.31, + "learning_rate": 1.6146815306366678e-05, + "loss": 0.9194, + "step": 9713 + }, + { + "epoch": 0.31, + "learning_rate": 1.6145998256918238e-05, + "loss": 1.061, + "step": 9714 + }, + { + "epoch": 0.31, + "learning_rate": 1.614518114153004e-05, + "loss": 1.2324, + "step": 9715 + }, + { + "epoch": 0.31, + "learning_rate": 1.614436396021085e-05, + "loss": 1.061, + "step": 9716 + }, + { + "epoch": 0.31, + "learning_rate": 1.6143546712969436e-05, + "loss": 1.002, + "step": 9717 + }, + { + "epoch": 0.31, + "learning_rate": 1.6142729399814564e-05, + "loss": 1.1895, + "step": 9718 + }, + { + "epoch": 0.31, + "learning_rate": 1.6141912020755002e-05, + "loss": 0.6016, + "step": 9719 + }, + { + "epoch": 0.31, + "learning_rate": 1.6141094575799527e-05, + "loss": 0.9194, + "step": 9720 + }, + { + "epoch": 0.31, + "learning_rate": 1.6140277064956903e-05, + "loss": 0.7935, + "step": 9721 + }, + { + "epoch": 0.31, + "learning_rate": 1.6139459488235898e-05, + "loss": 1.1499, + "step": 9722 + }, + { + "epoch": 0.31, + "learning_rate": 1.6138641845645288e-05, + "loss": 0.9648, + "step": 9723 + }, + { + "epoch": 0.31, + "learning_rate": 1.6137824137193847e-05, + "loss": 1.1738, + "step": 9724 + }, + { + "epoch": 0.31, + "learning_rate": 1.6137006362890342e-05, + "loss": 1.0273, + "step": 9725 + }, + { + "epoch": 0.31, + "learning_rate": 1.6136188522743555e-05, + "loss": 1.2393, + "step": 9726 + }, + { + "epoch": 0.31, + "learning_rate": 1.6135370616762256e-05, + "loss": 1.123, + "step": 9727 + }, + { + "epoch": 0.31, + "learning_rate": 1.6134552644955218e-05, + "loss": 1.0498, + "step": 9728 + }, + { + "epoch": 0.31, + "learning_rate": 1.6133734607331223e-05, + "loss": 1.0894, + "step": 9729 + }, + { + "epoch": 0.31, + "learning_rate": 1.613291650389904e-05, + "loss": 0.5625, + "step": 9730 + }, + { + "epoch": 0.31, + "learning_rate": 1.6132098334667453e-05, + "loss": 0.9434, + "step": 9731 + }, + { + "epoch": 0.31, + "learning_rate": 1.613128009964524e-05, + "loss": 1.041, + "step": 9732 + }, + { + "epoch": 0.31, + "learning_rate": 1.613046179884117e-05, + "loss": 1.1021, + "step": 9733 + }, + { + "epoch": 0.31, + "learning_rate": 1.6129643432264037e-05, + "loss": 1.1416, + "step": 9734 + }, + { + "epoch": 0.31, + "learning_rate": 1.612882499992261e-05, + "loss": 0.519, + "step": 9735 + }, + { + "epoch": 0.31, + "learning_rate": 1.6128006501825674e-05, + "loss": 0.9346, + "step": 9736 + }, + { + "epoch": 0.31, + "learning_rate": 1.612718793798201e-05, + "loss": 1.0322, + "step": 9737 + }, + { + "epoch": 0.31, + "learning_rate": 1.6126369308400404e-05, + "loss": 1.0796, + "step": 9738 + }, + { + "epoch": 0.31, + "learning_rate": 1.6125550613089637e-05, + "loss": 0.9072, + "step": 9739 + }, + { + "epoch": 0.31, + "learning_rate": 1.6124731852058485e-05, + "loss": 1.0098, + "step": 9740 + }, + { + "epoch": 0.31, + "learning_rate": 1.6123913025315743e-05, + "loss": 0.9248, + "step": 9741 + }, + { + "epoch": 0.31, + "learning_rate": 1.612309413287019e-05, + "loss": 1.0884, + "step": 9742 + }, + { + "epoch": 0.31, + "learning_rate": 1.6122275174730616e-05, + "loss": 0.9893, + "step": 9743 + }, + { + "epoch": 0.31, + "learning_rate": 1.6121456150905802e-05, + "loss": 1.1401, + "step": 9744 + }, + { + "epoch": 0.31, + "learning_rate": 1.612063706140454e-05, + "loss": 1.0142, + "step": 9745 + }, + { + "epoch": 0.31, + "learning_rate": 1.6119817906235615e-05, + "loss": 1.1602, + "step": 9746 + }, + { + "epoch": 0.31, + "learning_rate": 1.611899868540782e-05, + "loss": 0.9678, + "step": 9747 + }, + { + "epoch": 0.31, + "learning_rate": 1.611817939892994e-05, + "loss": 0.9785, + "step": 9748 + }, + { + "epoch": 0.31, + "learning_rate": 1.6117360046810767e-05, + "loss": 1.0269, + "step": 9749 + }, + { + "epoch": 0.31, + "learning_rate": 1.6116540629059088e-05, + "loss": 0.9893, + "step": 9750 + }, + { + "epoch": 0.31, + "learning_rate": 1.6115721145683697e-05, + "loss": 1.0635, + "step": 9751 + }, + { + "epoch": 0.31, + "learning_rate": 1.611490159669339e-05, + "loss": 1.0205, + "step": 9752 + }, + { + "epoch": 0.31, + "learning_rate": 1.6114081982096955e-05, + "loss": 0.9639, + "step": 9753 + }, + { + "epoch": 0.31, + "learning_rate": 1.611326230190319e-05, + "loss": 1.0049, + "step": 9754 + }, + { + "epoch": 0.31, + "learning_rate": 1.611244255612088e-05, + "loss": 1.0166, + "step": 9755 + }, + { + "epoch": 0.31, + "learning_rate": 1.6111622744758832e-05, + "loss": 1.0767, + "step": 9756 + }, + { + "epoch": 0.31, + "learning_rate": 1.6110802867825835e-05, + "loss": 1.0239, + "step": 9757 + }, + { + "epoch": 0.31, + "learning_rate": 1.6109982925330685e-05, + "loss": 1.0679, + "step": 9758 + }, + { + "epoch": 0.31, + "learning_rate": 1.610916291728218e-05, + "loss": 1.022, + "step": 9759 + }, + { + "epoch": 0.31, + "learning_rate": 1.610834284368912e-05, + "loss": 1.0635, + "step": 9760 + }, + { + "epoch": 0.31, + "learning_rate": 1.61075227045603e-05, + "loss": 1.0654, + "step": 9761 + }, + { + "epoch": 0.31, + "learning_rate": 1.6106702499904522e-05, + "loss": 1.1392, + "step": 9762 + }, + { + "epoch": 0.31, + "learning_rate": 1.6105882229730584e-05, + "loss": 1.1621, + "step": 9763 + }, + { + "epoch": 0.31, + "learning_rate": 1.610506189404729e-05, + "loss": 1.1597, + "step": 9764 + }, + { + "epoch": 0.31, + "learning_rate": 1.6104241492863437e-05, + "loss": 1.1323, + "step": 9765 + }, + { + "epoch": 0.31, + "learning_rate": 1.6103421026187826e-05, + "loss": 1.145, + "step": 9766 + }, + { + "epoch": 0.31, + "learning_rate": 1.6102600494029263e-05, + "loss": 0.5591, + "step": 9767 + }, + { + "epoch": 0.31, + "learning_rate": 1.610177989639655e-05, + "loss": 1.0503, + "step": 9768 + }, + { + "epoch": 0.31, + "learning_rate": 1.6100959233298496e-05, + "loss": 0.9956, + "step": 9769 + }, + { + "epoch": 0.31, + "learning_rate": 1.61001385047439e-05, + "loss": 1.0718, + "step": 9770 + }, + { + "epoch": 0.31, + "learning_rate": 1.6099317710741565e-05, + "loss": 1.0786, + "step": 9771 + }, + { + "epoch": 0.31, + "learning_rate": 1.6098496851300307e-05, + "loss": 1.0381, + "step": 9772 + }, + { + "epoch": 0.31, + "learning_rate": 1.6097675926428925e-05, + "loss": 1.0908, + "step": 9773 + }, + { + "epoch": 0.31, + "learning_rate": 1.609685493613623e-05, + "loss": 0.9883, + "step": 9774 + }, + { + "epoch": 0.31, + "learning_rate": 1.609603388043103e-05, + "loss": 1.0225, + "step": 9775 + }, + { + "epoch": 0.31, + "learning_rate": 1.6095212759322127e-05, + "loss": 1.0752, + "step": 9776 + }, + { + "epoch": 0.31, + "learning_rate": 1.6094391572818345e-05, + "loss": 1.0059, + "step": 9777 + }, + { + "epoch": 0.31, + "learning_rate": 1.609357032092848e-05, + "loss": 1.123, + "step": 9778 + }, + { + "epoch": 0.31, + "learning_rate": 1.609274900366135e-05, + "loss": 1.1362, + "step": 9779 + }, + { + "epoch": 0.31, + "learning_rate": 1.6091927621025768e-05, + "loss": 1.0815, + "step": 9780 + }, + { + "epoch": 0.31, + "learning_rate": 1.6091106173030548e-05, + "loss": 1.1113, + "step": 9781 + }, + { + "epoch": 0.31, + "learning_rate": 1.60902846596845e-05, + "loss": 1.0698, + "step": 9782 + }, + { + "epoch": 0.31, + "learning_rate": 1.6089463080996432e-05, + "loss": 1.0532, + "step": 9783 + }, + { + "epoch": 0.31, + "learning_rate": 1.6088641436975172e-05, + "loss": 1.0669, + "step": 9784 + }, + { + "epoch": 0.31, + "learning_rate": 1.6087819727629522e-05, + "loss": 1.0093, + "step": 9785 + }, + { + "epoch": 0.31, + "learning_rate": 1.6086997952968307e-05, + "loss": 1.0542, + "step": 9786 + }, + { + "epoch": 0.31, + "learning_rate": 1.6086176113000337e-05, + "loss": 0.9951, + "step": 9787 + }, + { + "epoch": 0.31, + "learning_rate": 1.6085354207734435e-05, + "loss": 1.123, + "step": 9788 + }, + { + "epoch": 0.31, + "learning_rate": 1.608453223717942e-05, + "loss": 1.0186, + "step": 9789 + }, + { + "epoch": 0.31, + "learning_rate": 1.6083710201344106e-05, + "loss": 1.0522, + "step": 9790 + }, + { + "epoch": 0.31, + "learning_rate": 1.6082888100237315e-05, + "loss": 1.0166, + "step": 9791 + }, + { + "epoch": 0.31, + "learning_rate": 1.608206593386787e-05, + "loss": 0.9858, + "step": 9792 + }, + { + "epoch": 0.31, + "learning_rate": 1.6081243702244583e-05, + "loss": 1.0942, + "step": 9793 + }, + { + "epoch": 0.31, + "learning_rate": 1.6080421405376284e-05, + "loss": 1.1333, + "step": 9794 + }, + { + "epoch": 0.31, + "learning_rate": 1.607959904327179e-05, + "loss": 0.9893, + "step": 9795 + }, + { + "epoch": 0.31, + "learning_rate": 1.6078776615939933e-05, + "loss": 0.8638, + "step": 9796 + }, + { + "epoch": 0.31, + "learning_rate": 1.6077954123389527e-05, + "loss": 1.0269, + "step": 9797 + }, + { + "epoch": 0.31, + "learning_rate": 1.60771315656294e-05, + "loss": 1.1274, + "step": 9798 + }, + { + "epoch": 0.31, + "learning_rate": 1.6076308942668377e-05, + "loss": 1.062, + "step": 9799 + }, + { + "epoch": 0.31, + "learning_rate": 1.6075486254515285e-05, + "loss": 1.1187, + "step": 9800 + }, + { + "epoch": 0.31, + "learning_rate": 1.607466350117895e-05, + "loss": 0.9614, + "step": 9801 + }, + { + "epoch": 0.31, + "learning_rate": 1.6073840682668197e-05, + "loss": 1.02, + "step": 9802 + }, + { + "epoch": 0.31, + "learning_rate": 1.6073017798991857e-05, + "loss": 1.0693, + "step": 9803 + }, + { + "epoch": 0.31, + "learning_rate": 1.6072194850158755e-05, + "loss": 0.9844, + "step": 9804 + }, + { + "epoch": 0.31, + "learning_rate": 1.6071371836177726e-05, + "loss": 1.0107, + "step": 9805 + }, + { + "epoch": 0.31, + "learning_rate": 1.6070548757057593e-05, + "loss": 1.0752, + "step": 9806 + }, + { + "epoch": 0.31, + "learning_rate": 1.6069725612807193e-05, + "loss": 1.1426, + "step": 9807 + }, + { + "epoch": 0.31, + "learning_rate": 1.6068902403435352e-05, + "loss": 1.0518, + "step": 9808 + }, + { + "epoch": 0.31, + "learning_rate": 1.6068079128950912e-05, + "loss": 1.0933, + "step": 9809 + }, + { + "epoch": 0.31, + "learning_rate": 1.606725578936269e-05, + "loss": 1.1162, + "step": 9810 + }, + { + "epoch": 0.31, + "learning_rate": 1.6066432384679535e-05, + "loss": 1.0396, + "step": 9811 + }, + { + "epoch": 0.31, + "learning_rate": 1.606560891491027e-05, + "loss": 1.106, + "step": 9812 + }, + { + "epoch": 0.31, + "learning_rate": 1.6064785380063736e-05, + "loss": 1.0913, + "step": 9813 + }, + { + "epoch": 0.31, + "learning_rate": 1.6063961780148763e-05, + "loss": 0.9839, + "step": 9814 + }, + { + "epoch": 0.31, + "learning_rate": 1.60631381151742e-05, + "loss": 0.9604, + "step": 9815 + }, + { + "epoch": 0.31, + "learning_rate": 1.6062314385148868e-05, + "loss": 1.063, + "step": 9816 + }, + { + "epoch": 0.31, + "learning_rate": 1.6061490590081618e-05, + "loss": 0.9585, + "step": 9817 + }, + { + "epoch": 0.31, + "learning_rate": 1.606066672998128e-05, + "loss": 1.0566, + "step": 9818 + }, + { + "epoch": 0.31, + "learning_rate": 1.6059842804856693e-05, + "loss": 0.9624, + "step": 9819 + }, + { + "epoch": 0.31, + "learning_rate": 1.60590188147167e-05, + "loss": 1.0508, + "step": 9820 + }, + { + "epoch": 0.31, + "learning_rate": 1.6058194759570143e-05, + "loss": 1.1357, + "step": 9821 + }, + { + "epoch": 0.31, + "learning_rate": 1.605737063942586e-05, + "loss": 1.0015, + "step": 9822 + }, + { + "epoch": 0.31, + "learning_rate": 1.6056546454292698e-05, + "loss": 0.999, + "step": 9823 + }, + { + "epoch": 0.31, + "learning_rate": 1.605572220417949e-05, + "loss": 1.0894, + "step": 9824 + }, + { + "epoch": 0.31, + "learning_rate": 1.605489788909509e-05, + "loss": 1.0493, + "step": 9825 + }, + { + "epoch": 0.31, + "learning_rate": 1.6054073509048328e-05, + "loss": 1.0815, + "step": 9826 + }, + { + "epoch": 0.31, + "learning_rate": 1.6053249064048066e-05, + "loss": 1.1953, + "step": 9827 + }, + { + "epoch": 0.31, + "learning_rate": 1.6052424554103138e-05, + "loss": 0.8965, + "step": 9828 + }, + { + "epoch": 0.31, + "learning_rate": 1.605159997922239e-05, + "loss": 1.1406, + "step": 9829 + }, + { + "epoch": 0.31, + "learning_rate": 1.6050775339414678e-05, + "loss": 0.9727, + "step": 9830 + }, + { + "epoch": 0.31, + "learning_rate": 1.6049950634688842e-05, + "loss": 1.0635, + "step": 9831 + }, + { + "epoch": 0.31, + "learning_rate": 1.604912586505373e-05, + "loss": 0.9468, + "step": 9832 + }, + { + "epoch": 0.31, + "learning_rate": 1.604830103051819e-05, + "loss": 1.1016, + "step": 9833 + }, + { + "epoch": 0.31, + "learning_rate": 1.6047476131091077e-05, + "loss": 1.0864, + "step": 9834 + }, + { + "epoch": 0.31, + "learning_rate": 1.6046651166781237e-05, + "loss": 0.9683, + "step": 9835 + }, + { + "epoch": 0.31, + "learning_rate": 1.6045826137597523e-05, + "loss": 0.8628, + "step": 9836 + }, + { + "epoch": 0.31, + "learning_rate": 1.6045001043548785e-05, + "loss": 1.0308, + "step": 9837 + }, + { + "epoch": 0.31, + "learning_rate": 1.6044175884643876e-05, + "loss": 1.0298, + "step": 9838 + }, + { + "epoch": 0.31, + "learning_rate": 1.604335066089165e-05, + "loss": 1.1592, + "step": 9839 + }, + { + "epoch": 0.31, + "learning_rate": 1.6042525372300958e-05, + "loss": 0.957, + "step": 9840 + }, + { + "epoch": 0.31, + "learning_rate": 1.604170001888066e-05, + "loss": 1.1724, + "step": 9841 + }, + { + "epoch": 0.31, + "learning_rate": 1.60408746006396e-05, + "loss": 1.0366, + "step": 9842 + }, + { + "epoch": 0.31, + "learning_rate": 1.6040049117586648e-05, + "loss": 1.0542, + "step": 9843 + }, + { + "epoch": 0.31, + "learning_rate": 1.6039223569730647e-05, + "loss": 0.9985, + "step": 9844 + }, + { + "epoch": 0.31, + "learning_rate": 1.6038397957080468e-05, + "loss": 0.5498, + "step": 9845 + }, + { + "epoch": 0.31, + "learning_rate": 1.6037572279644963e-05, + "loss": 1.1328, + "step": 9846 + }, + { + "epoch": 0.31, + "learning_rate": 1.6036746537432984e-05, + "loss": 1.0381, + "step": 9847 + }, + { + "epoch": 0.31, + "learning_rate": 1.60359207304534e-05, + "loss": 1.1274, + "step": 9848 + }, + { + "epoch": 0.31, + "learning_rate": 1.6035094858715065e-05, + "loss": 1.0977, + "step": 9849 + }, + { + "epoch": 0.32, + "learning_rate": 1.6034268922226843e-05, + "loss": 1.042, + "step": 9850 + }, + { + "epoch": 0.32, + "learning_rate": 1.603344292099759e-05, + "loss": 1.0273, + "step": 9851 + }, + { + "epoch": 0.32, + "learning_rate": 1.6032616855036175e-05, + "loss": 0.9619, + "step": 9852 + }, + { + "epoch": 0.32, + "learning_rate": 1.6031790724351454e-05, + "loss": 1.0786, + "step": 9853 + }, + { + "epoch": 0.32, + "learning_rate": 1.6030964528952297e-05, + "loss": 1.0249, + "step": 9854 + }, + { + "epoch": 0.32, + "learning_rate": 1.6030138268847564e-05, + "loss": 1.0181, + "step": 9855 + }, + { + "epoch": 0.32, + "learning_rate": 1.6029311944046124e-05, + "loss": 1.1289, + "step": 9856 + }, + { + "epoch": 0.32, + "learning_rate": 1.6028485554556838e-05, + "loss": 0.9697, + "step": 9857 + }, + { + "epoch": 0.32, + "learning_rate": 1.6027659100388573e-05, + "loss": 1.0591, + "step": 9858 + }, + { + "epoch": 0.32, + "learning_rate": 1.6026832581550197e-05, + "loss": 0.9868, + "step": 9859 + }, + { + "epoch": 0.32, + "learning_rate": 1.602600599805058e-05, + "loss": 1.0938, + "step": 9860 + }, + { + "epoch": 0.32, + "learning_rate": 1.6025179349898587e-05, + "loss": 1.064, + "step": 9861 + }, + { + "epoch": 0.32, + "learning_rate": 1.6024352637103086e-05, + "loss": 0.9731, + "step": 9862 + }, + { + "epoch": 0.32, + "learning_rate": 1.602352585967295e-05, + "loss": 0.9712, + "step": 9863 + }, + { + "epoch": 0.32, + "learning_rate": 1.602269901761705e-05, + "loss": 1.0747, + "step": 9864 + }, + { + "epoch": 0.32, + "learning_rate": 1.6021872110944253e-05, + "loss": 0.9922, + "step": 9865 + }, + { + "epoch": 0.32, + "learning_rate": 1.6021045139663432e-05, + "loss": 1.1182, + "step": 9866 + }, + { + "epoch": 0.32, + "learning_rate": 1.6020218103783463e-05, + "loss": 1.0615, + "step": 9867 + }, + { + "epoch": 0.32, + "learning_rate": 1.6019391003313213e-05, + "loss": 1.0884, + "step": 9868 + }, + { + "epoch": 0.32, + "learning_rate": 1.6018563838261564e-05, + "loss": 1.1646, + "step": 9869 + }, + { + "epoch": 0.32, + "learning_rate": 1.6017736608637384e-05, + "loss": 0.8721, + "step": 9870 + }, + { + "epoch": 0.32, + "learning_rate": 1.601690931444955e-05, + "loss": 0.9585, + "step": 9871 + }, + { + "epoch": 0.32, + "learning_rate": 1.6016081955706938e-05, + "loss": 1.1216, + "step": 9872 + }, + { + "epoch": 0.32, + "learning_rate": 1.6015254532418425e-05, + "loss": 0.9092, + "step": 9873 + }, + { + "epoch": 0.32, + "learning_rate": 1.6014427044592887e-05, + "loss": 1.0718, + "step": 9874 + }, + { + "epoch": 0.32, + "learning_rate": 1.6013599492239204e-05, + "loss": 0.9761, + "step": 9875 + }, + { + "epoch": 0.32, + "learning_rate": 1.6012771875366254e-05, + "loss": 1.0156, + "step": 9876 + }, + { + "epoch": 0.32, + "learning_rate": 1.6011944193982917e-05, + "loss": 1.1128, + "step": 9877 + }, + { + "epoch": 0.32, + "learning_rate": 1.6011116448098075e-05, + "loss": 1.1245, + "step": 9878 + }, + { + "epoch": 0.32, + "learning_rate": 1.6010288637720605e-05, + "loss": 1.0371, + "step": 9879 + }, + { + "epoch": 0.32, + "learning_rate": 1.6009460762859387e-05, + "loss": 0.9219, + "step": 9880 + }, + { + "epoch": 0.32, + "learning_rate": 1.6008632823523306e-05, + "loss": 1.0015, + "step": 9881 + }, + { + "epoch": 0.32, + "learning_rate": 1.6007804819721244e-05, + "loss": 0.9419, + "step": 9882 + }, + { + "epoch": 0.32, + "learning_rate": 1.600697675146209e-05, + "loss": 1.0767, + "step": 9883 + }, + { + "epoch": 0.32, + "learning_rate": 1.6006148618754722e-05, + "loss": 1.0581, + "step": 9884 + }, + { + "epoch": 0.32, + "learning_rate": 1.6005320421608022e-05, + "loss": 1.0156, + "step": 9885 + }, + { + "epoch": 0.32, + "learning_rate": 1.6004492160030887e-05, + "loss": 1.0088, + "step": 9886 + }, + { + "epoch": 0.32, + "learning_rate": 1.6003663834032194e-05, + "loss": 1.0122, + "step": 9887 + }, + { + "epoch": 0.32, + "learning_rate": 1.6002835443620832e-05, + "loss": 1.0908, + "step": 9888 + }, + { + "epoch": 0.32, + "learning_rate": 1.6002006988805693e-05, + "loss": 1.0107, + "step": 9889 + }, + { + "epoch": 0.32, + "learning_rate": 1.6001178469595657e-05, + "loss": 0.999, + "step": 9890 + }, + { + "epoch": 0.32, + "learning_rate": 1.6000349885999618e-05, + "loss": 1.0454, + "step": 9891 + }, + { + "epoch": 0.32, + "learning_rate": 1.5999521238026466e-05, + "loss": 1.0059, + "step": 9892 + }, + { + "epoch": 0.32, + "learning_rate": 1.5998692525685094e-05, + "loss": 0.998, + "step": 9893 + }, + { + "epoch": 0.32, + "learning_rate": 1.5997863748984384e-05, + "loss": 1.1201, + "step": 9894 + }, + { + "epoch": 0.32, + "learning_rate": 1.5997034907933243e-05, + "loss": 1.0303, + "step": 9895 + }, + { + "epoch": 0.32, + "learning_rate": 1.5996206002540547e-05, + "loss": 1.0771, + "step": 9896 + }, + { + "epoch": 0.32, + "learning_rate": 1.59953770328152e-05, + "loss": 1.1504, + "step": 9897 + }, + { + "epoch": 0.32, + "learning_rate": 1.5994547998766092e-05, + "loss": 1.1831, + "step": 9898 + }, + { + "epoch": 0.32, + "learning_rate": 1.599371890040212e-05, + "loss": 1.0088, + "step": 9899 + }, + { + "epoch": 0.32, + "learning_rate": 1.5992889737732172e-05, + "loss": 1.0225, + "step": 9900 + }, + { + "epoch": 0.32, + "learning_rate": 1.5992060510765156e-05, + "loss": 0.9194, + "step": 9901 + }, + { + "epoch": 0.32, + "learning_rate": 1.5991231219509962e-05, + "loss": 1.0835, + "step": 9902 + }, + { + "epoch": 0.32, + "learning_rate": 1.5990401863975488e-05, + "loss": 1.2334, + "step": 9903 + }, + { + "epoch": 0.32, + "learning_rate": 1.5989572444170627e-05, + "loss": 1.0068, + "step": 9904 + }, + { + "epoch": 0.32, + "learning_rate": 1.5988742960104287e-05, + "loss": 0.9829, + "step": 9905 + }, + { + "epoch": 0.32, + "learning_rate": 1.598791341178536e-05, + "loss": 0.9355, + "step": 9906 + }, + { + "epoch": 0.32, + "learning_rate": 1.5987083799222754e-05, + "loss": 1.0986, + "step": 9907 + }, + { + "epoch": 0.32, + "learning_rate": 1.598625412242536e-05, + "loss": 0.9692, + "step": 9908 + }, + { + "epoch": 0.32, + "learning_rate": 1.5985424381402087e-05, + "loss": 1.0269, + "step": 9909 + }, + { + "epoch": 0.32, + "learning_rate": 1.5984594576161832e-05, + "loss": 1.1655, + "step": 9910 + }, + { + "epoch": 0.32, + "learning_rate": 1.5983764706713506e-05, + "loss": 0.9229, + "step": 9911 + }, + { + "epoch": 0.32, + "learning_rate": 1.5982934773066004e-05, + "loss": 1.252, + "step": 9912 + }, + { + "epoch": 0.32, + "learning_rate": 1.5982104775228235e-05, + "loss": 0.9741, + "step": 9913 + }, + { + "epoch": 0.32, + "learning_rate": 1.59812747132091e-05, + "loss": 0.9829, + "step": 9914 + }, + { + "epoch": 0.32, + "learning_rate": 1.5980444587017506e-05, + "loss": 0.9844, + "step": 9915 + }, + { + "epoch": 0.32, + "learning_rate": 1.597961439666236e-05, + "loss": 0.9541, + "step": 9916 + }, + { + "epoch": 0.32, + "learning_rate": 1.5978784142152576e-05, + "loss": 1.0933, + "step": 9917 + }, + { + "epoch": 0.32, + "learning_rate": 1.597795382349705e-05, + "loss": 1.0693, + "step": 9918 + }, + { + "epoch": 0.32, + "learning_rate": 1.5977123440704696e-05, + "loss": 1.0566, + "step": 9919 + }, + { + "epoch": 0.32, + "learning_rate": 1.5976292993784427e-05, + "loss": 1.0762, + "step": 9920 + }, + { + "epoch": 0.32, + "learning_rate": 1.5975462482745143e-05, + "loss": 1.1313, + "step": 9921 + }, + { + "epoch": 0.32, + "learning_rate": 1.597463190759576e-05, + "loss": 1.1021, + "step": 9922 + }, + { + "epoch": 0.32, + "learning_rate": 1.597380126834519e-05, + "loss": 1.125, + "step": 9923 + }, + { + "epoch": 0.32, + "learning_rate": 1.5972970565002345e-05, + "loss": 1.0591, + "step": 9924 + }, + { + "epoch": 0.32, + "learning_rate": 1.597213979757614e-05, + "loss": 1.0317, + "step": 9925 + }, + { + "epoch": 0.32, + "learning_rate": 1.5971308966075475e-05, + "loss": 1.0356, + "step": 9926 + }, + { + "epoch": 0.32, + "learning_rate": 1.5970478070509278e-05, + "loss": 1.1157, + "step": 9927 + }, + { + "epoch": 0.32, + "learning_rate": 1.5969647110886464e-05, + "loss": 1.0903, + "step": 9928 + }, + { + "epoch": 0.32, + "learning_rate": 1.596881608721594e-05, + "loss": 0.981, + "step": 9929 + }, + { + "epoch": 0.32, + "learning_rate": 1.5967984999506623e-05, + "loss": 1.0532, + "step": 9930 + }, + { + "epoch": 0.32, + "learning_rate": 1.5967153847767433e-05, + "loss": 0.9917, + "step": 9931 + }, + { + "epoch": 0.32, + "learning_rate": 1.596632263200729e-05, + "loss": 1.0552, + "step": 9932 + }, + { + "epoch": 0.32, + "learning_rate": 1.5965491352235104e-05, + "loss": 1.1846, + "step": 9933 + }, + { + "epoch": 0.32, + "learning_rate": 1.59646600084598e-05, + "loss": 1.0181, + "step": 9934 + }, + { + "epoch": 0.32, + "learning_rate": 1.5963828600690295e-05, + "loss": 0.8965, + "step": 9935 + }, + { + "epoch": 0.32, + "learning_rate": 1.596299712893551e-05, + "loss": 1.1226, + "step": 9936 + }, + { + "epoch": 0.32, + "learning_rate": 1.5962165593204365e-05, + "loss": 0.9629, + "step": 9937 + }, + { + "epoch": 0.32, + "learning_rate": 1.596133399350578e-05, + "loss": 1.0332, + "step": 9938 + }, + { + "epoch": 0.32, + "learning_rate": 1.5960502329848683e-05, + "loss": 1.0605, + "step": 9939 + }, + { + "epoch": 0.32, + "learning_rate": 1.5959670602241988e-05, + "loss": 1.0195, + "step": 9940 + }, + { + "epoch": 0.32, + "learning_rate": 1.5958838810694625e-05, + "loss": 1.0781, + "step": 9941 + }, + { + "epoch": 0.32, + "learning_rate": 1.5958006955215516e-05, + "loss": 1.0337, + "step": 9942 + }, + { + "epoch": 0.32, + "learning_rate": 1.5957175035813592e-05, + "loss": 1.0977, + "step": 9943 + }, + { + "epoch": 0.32, + "learning_rate": 1.5956343052497766e-05, + "loss": 1.1099, + "step": 9944 + }, + { + "epoch": 0.32, + "learning_rate": 1.5955511005276975e-05, + "loss": 1.0552, + "step": 9945 + }, + { + "epoch": 0.32, + "learning_rate": 1.5954678894160137e-05, + "loss": 1.0498, + "step": 9946 + }, + { + "epoch": 0.32, + "learning_rate": 1.595384671915619e-05, + "loss": 1.0586, + "step": 9947 + }, + { + "epoch": 0.32, + "learning_rate": 1.5953014480274057e-05, + "loss": 1.2227, + "step": 9948 + }, + { + "epoch": 0.32, + "learning_rate": 1.5952182177522663e-05, + "loss": 1.0684, + "step": 9949 + }, + { + "epoch": 0.32, + "learning_rate": 1.5951349810910944e-05, + "loss": 0.9683, + "step": 9950 + }, + { + "epoch": 0.32, + "learning_rate": 1.595051738044783e-05, + "loss": 1.0054, + "step": 9951 + }, + { + "epoch": 0.32, + "learning_rate": 1.5949684886142248e-05, + "loss": 1.1553, + "step": 9952 + }, + { + "epoch": 0.32, + "learning_rate": 1.5948852328003135e-05, + "loss": 1.1919, + "step": 9953 + }, + { + "epoch": 0.32, + "learning_rate": 1.5948019706039416e-05, + "loss": 0.9092, + "step": 9954 + }, + { + "epoch": 0.32, + "learning_rate": 1.5947187020260032e-05, + "loss": 1.0049, + "step": 9955 + }, + { + "epoch": 0.32, + "learning_rate": 1.5946354270673913e-05, + "loss": 1.0103, + "step": 9956 + }, + { + "epoch": 0.32, + "learning_rate": 1.594552145728999e-05, + "loss": 0.8647, + "step": 9957 + }, + { + "epoch": 0.32, + "learning_rate": 1.5944688580117206e-05, + "loss": 0.8945, + "step": 9958 + }, + { + "epoch": 0.32, + "learning_rate": 1.594385563916449e-05, + "loss": 1.0107, + "step": 9959 + }, + { + "epoch": 0.32, + "learning_rate": 1.5943022634440786e-05, + "loss": 1.0527, + "step": 9960 + }, + { + "epoch": 0.32, + "learning_rate": 1.5942189565955023e-05, + "loss": 0.9194, + "step": 9961 + }, + { + "epoch": 0.32, + "learning_rate": 1.5941356433716145e-05, + "loss": 1.1396, + "step": 9962 + }, + { + "epoch": 0.32, + "learning_rate": 1.5940523237733082e-05, + "loss": 1.0078, + "step": 9963 + }, + { + "epoch": 0.32, + "learning_rate": 1.5939689978014784e-05, + "loss": 1.0254, + "step": 9964 + }, + { + "epoch": 0.32, + "learning_rate": 1.5938856654570186e-05, + "loss": 1.1011, + "step": 9965 + }, + { + "epoch": 0.32, + "learning_rate": 1.5938023267408225e-05, + "loss": 0.9951, + "step": 9966 + }, + { + "epoch": 0.32, + "learning_rate": 1.5937189816537854e-05, + "loss": 1.0771, + "step": 9967 + }, + { + "epoch": 0.32, + "learning_rate": 1.5936356301968e-05, + "loss": 1.0366, + "step": 9968 + }, + { + "epoch": 0.32, + "learning_rate": 1.5935522723707617e-05, + "loss": 1.1372, + "step": 9969 + }, + { + "epoch": 0.32, + "learning_rate": 1.5934689081765644e-05, + "loss": 0.9771, + "step": 9970 + }, + { + "epoch": 0.32, + "learning_rate": 1.5933855376151025e-05, + "loss": 1.1387, + "step": 9971 + }, + { + "epoch": 0.32, + "learning_rate": 1.5933021606872702e-05, + "loss": 0.9624, + "step": 9972 + }, + { + "epoch": 0.32, + "learning_rate": 1.5932187773939626e-05, + "loss": 1.0693, + "step": 9973 + }, + { + "epoch": 0.32, + "learning_rate": 1.593135387736074e-05, + "loss": 1.1284, + "step": 9974 + }, + { + "epoch": 0.32, + "learning_rate": 1.593051991714499e-05, + "loss": 1.0562, + "step": 9975 + }, + { + "epoch": 0.32, + "learning_rate": 1.592968589330133e-05, + "loss": 1.0415, + "step": 9976 + }, + { + "epoch": 0.32, + "learning_rate": 1.5928851805838698e-05, + "loss": 1.0474, + "step": 9977 + }, + { + "epoch": 0.32, + "learning_rate": 1.592801765476605e-05, + "loss": 1.0205, + "step": 9978 + }, + { + "epoch": 0.32, + "learning_rate": 1.592718344009233e-05, + "loss": 1.1006, + "step": 9979 + }, + { + "epoch": 0.32, + "learning_rate": 1.5926349161826496e-05, + "loss": 1.1069, + "step": 9980 + }, + { + "epoch": 0.32, + "learning_rate": 1.5925514819977495e-05, + "loss": 1.0117, + "step": 9981 + }, + { + "epoch": 0.32, + "learning_rate": 1.5924680414554278e-05, + "loss": 1.0771, + "step": 9982 + }, + { + "epoch": 0.32, + "learning_rate": 1.592384594556579e-05, + "loss": 0.998, + "step": 9983 + }, + { + "epoch": 0.32, + "learning_rate": 1.5923011413021e-05, + "loss": 1.1401, + "step": 9984 + }, + { + "epoch": 0.32, + "learning_rate": 1.592217681692885e-05, + "loss": 0.9702, + "step": 9985 + }, + { + "epoch": 0.32, + "learning_rate": 1.5921342157298295e-05, + "loss": 1.0444, + "step": 9986 + }, + { + "epoch": 0.32, + "learning_rate": 1.5920507434138294e-05, + "loss": 0.9902, + "step": 9987 + }, + { + "epoch": 0.32, + "learning_rate": 1.59196726474578e-05, + "loss": 1.0186, + "step": 9988 + }, + { + "epoch": 0.32, + "learning_rate": 1.591883779726577e-05, + "loss": 1.0757, + "step": 9989 + }, + { + "epoch": 0.32, + "learning_rate": 1.591800288357116e-05, + "loss": 1.061, + "step": 9990 + }, + { + "epoch": 0.32, + "learning_rate": 1.591716790638293e-05, + "loss": 1.0225, + "step": 9991 + }, + { + "epoch": 0.32, + "learning_rate": 1.5916332865710037e-05, + "loss": 0.939, + "step": 9992 + }, + { + "epoch": 0.32, + "learning_rate": 1.5915497761561443e-05, + "loss": 1.1523, + "step": 9993 + }, + { + "epoch": 0.32, + "learning_rate": 1.59146625939461e-05, + "loss": 0.9883, + "step": 9994 + }, + { + "epoch": 0.32, + "learning_rate": 1.5913827362872978e-05, + "loss": 1.0576, + "step": 9995 + }, + { + "epoch": 0.32, + "learning_rate": 1.5912992068351032e-05, + "loss": 1.0244, + "step": 9996 + }, + { + "epoch": 0.32, + "learning_rate": 1.5912156710389222e-05, + "loss": 0.8491, + "step": 9997 + }, + { + "epoch": 0.32, + "learning_rate": 1.5911321288996517e-05, + "loss": 1.084, + "step": 9998 + }, + { + "epoch": 0.32, + "learning_rate": 1.5910485804181878e-05, + "loss": 1.0625, + "step": 9999 + }, + { + "epoch": 0.32, + "learning_rate": 1.5909650255954266e-05, + "loss": 1.0542, + "step": 10000 + }, + { + "epoch": 0.32, + "learning_rate": 1.5908814644322646e-05, + "loss": 1.0767, + "step": 10001 + }, + { + "epoch": 0.32, + "learning_rate": 1.5907978969295984e-05, + "loss": 1.064, + "step": 10002 + }, + { + "epoch": 0.32, + "learning_rate": 1.590714323088325e-05, + "loss": 1.1401, + "step": 10003 + }, + { + "epoch": 0.32, + "learning_rate": 1.5906307429093403e-05, + "loss": 1.0425, + "step": 10004 + }, + { + "epoch": 0.32, + "learning_rate": 1.5905471563935416e-05, + "loss": 1.105, + "step": 10005 + }, + { + "epoch": 0.32, + "learning_rate": 1.5904635635418255e-05, + "loss": 1.1655, + "step": 10006 + }, + { + "epoch": 0.32, + "learning_rate": 1.590379964355089e-05, + "loss": 1.1421, + "step": 10007 + }, + { + "epoch": 0.32, + "learning_rate": 1.5902963588342287e-05, + "loss": 0.5894, + "step": 10008 + }, + { + "epoch": 0.32, + "learning_rate": 1.5902127469801415e-05, + "loss": 0.9492, + "step": 10009 + }, + { + "epoch": 0.32, + "learning_rate": 1.590129128793725e-05, + "loss": 1.1611, + "step": 10010 + }, + { + "epoch": 0.32, + "learning_rate": 1.5900455042758763e-05, + "loss": 0.9902, + "step": 10011 + }, + { + "epoch": 0.32, + "learning_rate": 1.589961873427492e-05, + "loss": 1.1118, + "step": 10012 + }, + { + "epoch": 0.32, + "learning_rate": 1.58987823624947e-05, + "loss": 1.1265, + "step": 10013 + }, + { + "epoch": 0.32, + "learning_rate": 1.5897945927427076e-05, + "loss": 1.1162, + "step": 10014 + }, + { + "epoch": 0.32, + "learning_rate": 1.5897109429081014e-05, + "loss": 0.9741, + "step": 10015 + }, + { + "epoch": 0.32, + "learning_rate": 1.58962728674655e-05, + "loss": 1.0337, + "step": 10016 + }, + { + "epoch": 0.32, + "learning_rate": 1.58954362425895e-05, + "loss": 1.1738, + "step": 10017 + }, + { + "epoch": 0.32, + "learning_rate": 1.5894599554461993e-05, + "loss": 0.9634, + "step": 10018 + }, + { + "epoch": 0.32, + "learning_rate": 1.5893762803091957e-05, + "loss": 1.0522, + "step": 10019 + }, + { + "epoch": 0.32, + "learning_rate": 1.589292598848837e-05, + "loss": 0.9771, + "step": 10020 + }, + { + "epoch": 0.32, + "learning_rate": 1.5892089110660213e-05, + "loss": 1.0215, + "step": 10021 + }, + { + "epoch": 0.32, + "learning_rate": 1.5891252169616457e-05, + "loss": 0.9937, + "step": 10022 + }, + { + "epoch": 0.32, + "learning_rate": 1.5890415165366084e-05, + "loss": 0.9561, + "step": 10023 + }, + { + "epoch": 0.32, + "learning_rate": 1.5889578097918077e-05, + "loss": 1.0293, + "step": 10024 + }, + { + "epoch": 0.32, + "learning_rate": 1.5888740967281412e-05, + "loss": 0.9683, + "step": 10025 + }, + { + "epoch": 0.32, + "learning_rate": 1.5887903773465082e-05, + "loss": 1.0723, + "step": 10026 + }, + { + "epoch": 0.32, + "learning_rate": 1.5887066516478054e-05, + "loss": 1.02, + "step": 10027 + }, + { + "epoch": 0.32, + "learning_rate": 1.5886229196329318e-05, + "loss": 1.1187, + "step": 10028 + }, + { + "epoch": 0.32, + "learning_rate": 1.588539181302786e-05, + "loss": 0.9858, + "step": 10029 + }, + { + "epoch": 0.32, + "learning_rate": 1.5884554366582664e-05, + "loss": 1.1147, + "step": 10030 + }, + { + "epoch": 0.32, + "learning_rate": 1.5883716857002703e-05, + "loss": 1.1504, + "step": 10031 + }, + { + "epoch": 0.32, + "learning_rate": 1.588287928429698e-05, + "loss": 1.1362, + "step": 10032 + }, + { + "epoch": 0.32, + "learning_rate": 1.5882041648474472e-05, + "loss": 1.1523, + "step": 10033 + }, + { + "epoch": 0.32, + "learning_rate": 1.5881203949544162e-05, + "loss": 1.0308, + "step": 10034 + }, + { + "epoch": 0.32, + "learning_rate": 1.588036618751505e-05, + "loss": 0.9668, + "step": 10035 + }, + { + "epoch": 0.32, + "learning_rate": 1.5879528362396114e-05, + "loss": 0.9878, + "step": 10036 + }, + { + "epoch": 0.32, + "learning_rate": 1.5878690474196346e-05, + "loss": 1.0039, + "step": 10037 + }, + { + "epoch": 0.32, + "learning_rate": 1.5877852522924733e-05, + "loss": 1.0068, + "step": 10038 + }, + { + "epoch": 0.32, + "learning_rate": 1.5877014508590266e-05, + "loss": 0.8872, + "step": 10039 + }, + { + "epoch": 0.32, + "learning_rate": 1.5876176431201942e-05, + "loss": 1.04, + "step": 10040 + }, + { + "epoch": 0.32, + "learning_rate": 1.587533829076875e-05, + "loss": 0.9316, + "step": 10041 + }, + { + "epoch": 0.32, + "learning_rate": 1.587450008729968e-05, + "loss": 1.0781, + "step": 10042 + }, + { + "epoch": 0.32, + "learning_rate": 1.5873661820803722e-05, + "loss": 0.978, + "step": 10043 + }, + { + "epoch": 0.32, + "learning_rate": 1.587282349128987e-05, + "loss": 1.1826, + "step": 10044 + }, + { + "epoch": 0.32, + "learning_rate": 1.587198509876713e-05, + "loss": 0.9033, + "step": 10045 + }, + { + "epoch": 0.32, + "learning_rate": 1.5871146643244484e-05, + "loss": 1.1436, + "step": 10046 + }, + { + "epoch": 0.32, + "learning_rate": 1.5870308124730936e-05, + "loss": 1.1147, + "step": 10047 + }, + { + "epoch": 0.32, + "learning_rate": 1.5869469543235476e-05, + "loss": 1.0312, + "step": 10048 + }, + { + "epoch": 0.32, + "learning_rate": 1.5868630898767102e-05, + "loss": 1.0342, + "step": 10049 + }, + { + "epoch": 0.32, + "learning_rate": 1.5867792191334818e-05, + "loss": 1.0581, + "step": 10050 + }, + { + "epoch": 0.32, + "learning_rate": 1.5866953420947616e-05, + "loss": 1.106, + "step": 10051 + }, + { + "epoch": 0.32, + "learning_rate": 1.58661145876145e-05, + "loss": 1.104, + "step": 10052 + }, + { + "epoch": 0.32, + "learning_rate": 1.586527569134446e-05, + "loss": 1.0244, + "step": 10053 + }, + { + "epoch": 0.32, + "learning_rate": 1.5864436732146506e-05, + "loss": 1.0576, + "step": 10054 + }, + { + "epoch": 0.32, + "learning_rate": 1.5863597710029637e-05, + "loss": 1.1157, + "step": 10055 + }, + { + "epoch": 0.32, + "learning_rate": 1.5862758625002855e-05, + "loss": 1.1367, + "step": 10056 + }, + { + "epoch": 0.32, + "learning_rate": 1.586191947707516e-05, + "loss": 1.1348, + "step": 10057 + }, + { + "epoch": 0.32, + "learning_rate": 1.5861080266255556e-05, + "loss": 0.9155, + "step": 10058 + }, + { + "epoch": 0.32, + "learning_rate": 1.5860240992553052e-05, + "loss": 1.105, + "step": 10059 + }, + { + "epoch": 0.32, + "learning_rate": 1.585940165597664e-05, + "loss": 1.0654, + "step": 10060 + }, + { + "epoch": 0.32, + "learning_rate": 1.585856225653534e-05, + "loss": 0.981, + "step": 10061 + }, + { + "epoch": 0.32, + "learning_rate": 1.585772279423815e-05, + "loss": 1.0039, + "step": 10062 + }, + { + "epoch": 0.32, + "learning_rate": 1.5856883269094074e-05, + "loss": 1.0352, + "step": 10063 + }, + { + "epoch": 0.32, + "learning_rate": 1.5856043681112125e-05, + "loss": 0.9512, + "step": 10064 + }, + { + "epoch": 0.32, + "learning_rate": 1.585520403030131e-05, + "loss": 1.0312, + "step": 10065 + }, + { + "epoch": 0.32, + "learning_rate": 1.5854364316670633e-05, + "loss": 0.9897, + "step": 10066 + }, + { + "epoch": 0.32, + "learning_rate": 1.5853524540229105e-05, + "loss": 1.0991, + "step": 10067 + }, + { + "epoch": 0.32, + "learning_rate": 1.5852684700985742e-05, + "loss": 1.0259, + "step": 10068 + }, + { + "epoch": 0.32, + "learning_rate": 1.5851844798949546e-05, + "loss": 1.0732, + "step": 10069 + }, + { + "epoch": 0.32, + "learning_rate": 1.585100483412953e-05, + "loss": 1.1841, + "step": 10070 + }, + { + "epoch": 0.32, + "learning_rate": 1.5850164806534714e-05, + "loss": 0.8354, + "step": 10071 + }, + { + "epoch": 0.32, + "learning_rate": 1.58493247161741e-05, + "loss": 0.9468, + "step": 10072 + }, + { + "epoch": 0.32, + "learning_rate": 1.5848484563056707e-05, + "loss": 1.0767, + "step": 10073 + }, + { + "epoch": 0.32, + "learning_rate": 1.5847644347191545e-05, + "loss": 1.0488, + "step": 10074 + }, + { + "epoch": 0.32, + "learning_rate": 1.5846804068587632e-05, + "loss": 1.1362, + "step": 10075 + }, + { + "epoch": 0.32, + "learning_rate": 1.5845963727253987e-05, + "loss": 1.0415, + "step": 10076 + }, + { + "epoch": 0.32, + "learning_rate": 1.5845123323199615e-05, + "loss": 1.0337, + "step": 10077 + }, + { + "epoch": 0.32, + "learning_rate": 1.584428285643354e-05, + "loss": 0.8833, + "step": 10078 + }, + { + "epoch": 0.32, + "learning_rate": 1.584344232696478e-05, + "loss": 1.0049, + "step": 10079 + }, + { + "epoch": 0.32, + "learning_rate": 1.5842601734802348e-05, + "loss": 1.0054, + "step": 10080 + }, + { + "epoch": 0.32, + "learning_rate": 1.584176107995527e-05, + "loss": 1.0083, + "step": 10081 + }, + { + "epoch": 0.32, + "learning_rate": 1.5840920362432558e-05, + "loss": 0.9849, + "step": 10082 + }, + { + "epoch": 0.32, + "learning_rate": 1.5840079582243235e-05, + "loss": 1.2646, + "step": 10083 + }, + { + "epoch": 0.32, + "learning_rate": 1.5839238739396323e-05, + "loss": 0.5454, + "step": 10084 + }, + { + "epoch": 0.32, + "learning_rate": 1.583839783390084e-05, + "loss": 1.0137, + "step": 10085 + }, + { + "epoch": 0.32, + "learning_rate": 1.583755686576581e-05, + "loss": 1.0381, + "step": 10086 + }, + { + "epoch": 0.32, + "learning_rate": 1.583671583500026e-05, + "loss": 1.0352, + "step": 10087 + }, + { + "epoch": 0.32, + "learning_rate": 1.5835874741613203e-05, + "loss": 0.9751, + "step": 10088 + }, + { + "epoch": 0.32, + "learning_rate": 1.583503358561367e-05, + "loss": 1.1021, + "step": 10089 + }, + { + "epoch": 0.32, + "learning_rate": 1.583419236701069e-05, + "loss": 1.1045, + "step": 10090 + }, + { + "epoch": 0.32, + "learning_rate": 1.5833351085813276e-05, + "loss": 1.1782, + "step": 10091 + }, + { + "epoch": 0.32, + "learning_rate": 1.5832509742030463e-05, + "loss": 1.0278, + "step": 10092 + }, + { + "epoch": 0.32, + "learning_rate": 1.583166833567127e-05, + "loss": 0.9478, + "step": 10093 + }, + { + "epoch": 0.32, + "learning_rate": 1.583082686674474e-05, + "loss": 0.9453, + "step": 10094 + }, + { + "epoch": 0.32, + "learning_rate": 1.5829985335259887e-05, + "loss": 1.0508, + "step": 10095 + }, + { + "epoch": 0.32, + "learning_rate": 1.5829143741225744e-05, + "loss": 0.9053, + "step": 10096 + }, + { + "epoch": 0.32, + "learning_rate": 1.5828302084651335e-05, + "loss": 1.02, + "step": 10097 + }, + { + "epoch": 0.32, + "learning_rate": 1.5827460365545702e-05, + "loss": 0.9927, + "step": 10098 + }, + { + "epoch": 0.32, + "learning_rate": 1.5826618583917867e-05, + "loss": 1.0913, + "step": 10099 + }, + { + "epoch": 0.32, + "learning_rate": 1.5825776739776863e-05, + "loss": 1.1299, + "step": 10100 + }, + { + "epoch": 0.32, + "learning_rate": 1.582493483313172e-05, + "loss": 1.0605, + "step": 10101 + }, + { + "epoch": 0.32, + "learning_rate": 1.5824092863991477e-05, + "loss": 0.9263, + "step": 10102 + }, + { + "epoch": 0.32, + "learning_rate": 1.582325083236516e-05, + "loss": 1.1426, + "step": 10103 + }, + { + "epoch": 0.32, + "learning_rate": 1.582240873826181e-05, + "loss": 1.0557, + "step": 10104 + }, + { + "epoch": 0.32, + "learning_rate": 1.5821566581690456e-05, + "loss": 1.0342, + "step": 10105 + }, + { + "epoch": 0.32, + "learning_rate": 1.5820724362660133e-05, + "loss": 1.0942, + "step": 10106 + }, + { + "epoch": 0.32, + "learning_rate": 1.5819882081179883e-05, + "loss": 1.1113, + "step": 10107 + }, + { + "epoch": 0.32, + "learning_rate": 1.5819039737258736e-05, + "loss": 1.0303, + "step": 10108 + }, + { + "epoch": 0.32, + "learning_rate": 1.581819733090574e-05, + "loss": 1.0527, + "step": 10109 + }, + { + "epoch": 0.32, + "learning_rate": 1.581735486212992e-05, + "loss": 0.9551, + "step": 10110 + }, + { + "epoch": 0.32, + "learning_rate": 1.581651233094032e-05, + "loss": 1.0278, + "step": 10111 + }, + { + "epoch": 0.32, + "learning_rate": 1.581566973734598e-05, + "loss": 1.1753, + "step": 10112 + }, + { + "epoch": 0.32, + "learning_rate": 1.5814827081355943e-05, + "loss": 0.9888, + "step": 10113 + }, + { + "epoch": 0.32, + "learning_rate": 1.5813984362979247e-05, + "loss": 1.0542, + "step": 10114 + }, + { + "epoch": 0.32, + "learning_rate": 1.581314158222493e-05, + "loss": 1.0264, + "step": 10115 + }, + { + "epoch": 0.32, + "learning_rate": 1.581229873910204e-05, + "loss": 1.0308, + "step": 10116 + }, + { + "epoch": 0.32, + "learning_rate": 1.5811455833619618e-05, + "loss": 1.1099, + "step": 10117 + }, + { + "epoch": 0.32, + "learning_rate": 1.5810612865786704e-05, + "loss": 1.0957, + "step": 10118 + }, + { + "epoch": 0.32, + "learning_rate": 1.580976983561235e-05, + "loss": 1.2104, + "step": 10119 + }, + { + "epoch": 0.32, + "learning_rate": 1.580892674310559e-05, + "loss": 1.041, + "step": 10120 + }, + { + "epoch": 0.32, + "learning_rate": 1.5808083588275477e-05, + "loss": 1.0586, + "step": 10121 + }, + { + "epoch": 0.32, + "learning_rate": 1.5807240371131056e-05, + "loss": 1.1367, + "step": 10122 + }, + { + "epoch": 0.32, + "learning_rate": 1.5806397091681372e-05, + "loss": 0.999, + "step": 10123 + }, + { + "epoch": 0.32, + "learning_rate": 1.5805553749935475e-05, + "loss": 1.0259, + "step": 10124 + }, + { + "epoch": 0.32, + "learning_rate": 1.5804710345902406e-05, + "loss": 1.1045, + "step": 10125 + }, + { + "epoch": 0.32, + "learning_rate": 1.5803866879591227e-05, + "loss": 1.0513, + "step": 10126 + }, + { + "epoch": 0.32, + "learning_rate": 1.5803023351010976e-05, + "loss": 1.001, + "step": 10127 + }, + { + "epoch": 0.32, + "learning_rate": 1.5802179760170706e-05, + "loss": 1.0879, + "step": 10128 + }, + { + "epoch": 0.32, + "learning_rate": 1.580133610707947e-05, + "loss": 1.0776, + "step": 10129 + }, + { + "epoch": 0.32, + "learning_rate": 1.5800492391746318e-05, + "loss": 1.0835, + "step": 10130 + }, + { + "epoch": 0.32, + "learning_rate": 1.5799648614180302e-05, + "loss": 1.0791, + "step": 10131 + }, + { + "epoch": 0.32, + "learning_rate": 1.5798804774390476e-05, + "loss": 1.1616, + "step": 10132 + }, + { + "epoch": 0.32, + "learning_rate": 1.5797960872385894e-05, + "loss": 1.186, + "step": 10133 + }, + { + "epoch": 0.32, + "learning_rate": 1.5797116908175606e-05, + "loss": 1.0054, + "step": 10134 + }, + { + "epoch": 0.32, + "learning_rate": 1.579627288176867e-05, + "loss": 0.9023, + "step": 10135 + }, + { + "epoch": 0.32, + "learning_rate": 1.579542879317414e-05, + "loss": 1.0488, + "step": 10136 + }, + { + "epoch": 0.32, + "learning_rate": 1.5794584642401076e-05, + "loss": 1.0269, + "step": 10137 + }, + { + "epoch": 0.32, + "learning_rate": 1.579374042945853e-05, + "loss": 0.9922, + "step": 10138 + }, + { + "epoch": 0.32, + "learning_rate": 1.579289615435556e-05, + "loss": 1.0088, + "step": 10139 + }, + { + "epoch": 0.32, + "learning_rate": 1.579205181710123e-05, + "loss": 1.041, + "step": 10140 + }, + { + "epoch": 0.32, + "learning_rate": 1.579120741770459e-05, + "loss": 1.0693, + "step": 10141 + }, + { + "epoch": 0.32, + "learning_rate": 1.5790362956174704e-05, + "loss": 1.0347, + "step": 10142 + }, + { + "epoch": 0.32, + "learning_rate": 1.5789518432520635e-05, + "loss": 1.1367, + "step": 10143 + }, + { + "epoch": 0.32, + "learning_rate": 1.578867384675144e-05, + "loss": 1.0205, + "step": 10144 + }, + { + "epoch": 0.32, + "learning_rate": 1.5787829198876178e-05, + "loss": 1.0469, + "step": 10145 + }, + { + "epoch": 0.32, + "learning_rate": 1.5786984488903917e-05, + "loss": 1.0444, + "step": 10146 + }, + { + "epoch": 0.32, + "learning_rate": 1.5786139716843717e-05, + "loss": 0.9736, + "step": 10147 + }, + { + "epoch": 0.32, + "learning_rate": 1.5785294882704642e-05, + "loss": 0.894, + "step": 10148 + }, + { + "epoch": 0.32, + "learning_rate": 1.5784449986495755e-05, + "loss": 0.9863, + "step": 10149 + }, + { + "epoch": 0.32, + "learning_rate": 1.5783605028226122e-05, + "loss": 1.0742, + "step": 10150 + }, + { + "epoch": 0.32, + "learning_rate": 1.578276000790481e-05, + "loss": 0.9849, + "step": 10151 + }, + { + "epoch": 0.32, + "learning_rate": 1.5781914925540877e-05, + "loss": 0.9883, + "step": 10152 + }, + { + "epoch": 0.32, + "learning_rate": 1.57810697811434e-05, + "loss": 1.0605, + "step": 10153 + }, + { + "epoch": 0.32, + "learning_rate": 1.5780224574721444e-05, + "loss": 0.8955, + "step": 10154 + }, + { + "epoch": 0.32, + "learning_rate": 1.5779379306284074e-05, + "loss": 0.9712, + "step": 10155 + }, + { + "epoch": 0.32, + "learning_rate": 1.5778533975840358e-05, + "loss": 0.8394, + "step": 10156 + }, + { + "epoch": 0.32, + "learning_rate": 1.5777688583399367e-05, + "loss": 1.104, + "step": 10157 + }, + { + "epoch": 0.32, + "learning_rate": 1.5776843128970177e-05, + "loss": 1.0688, + "step": 10158 + }, + { + "epoch": 0.32, + "learning_rate": 1.5775997612561852e-05, + "loss": 0.9629, + "step": 10159 + }, + { + "epoch": 0.32, + "learning_rate": 1.5775152034183464e-05, + "loss": 1.0449, + "step": 10160 + }, + { + "epoch": 0.32, + "learning_rate": 1.577430639384409e-05, + "loss": 0.9971, + "step": 10161 + }, + { + "epoch": 0.32, + "learning_rate": 1.5773460691552797e-05, + "loss": 1.1265, + "step": 10162 + }, + { + "epoch": 0.33, + "learning_rate": 1.5772614927318657e-05, + "loss": 1.0693, + "step": 10163 + }, + { + "epoch": 0.33, + "learning_rate": 1.5771769101150752e-05, + "loss": 1.1836, + "step": 10164 + }, + { + "epoch": 0.33, + "learning_rate": 1.577092321305815e-05, + "loss": 1.1274, + "step": 10165 + }, + { + "epoch": 0.33, + "learning_rate": 1.5770077263049928e-05, + "loss": 1.0034, + "step": 10166 + }, + { + "epoch": 0.33, + "learning_rate": 1.5769231251135165e-05, + "loss": 1.1372, + "step": 10167 + }, + { + "epoch": 0.33, + "learning_rate": 1.5768385177322934e-05, + "loss": 1.1592, + "step": 10168 + }, + { + "epoch": 0.33, + "learning_rate": 1.5767539041622316e-05, + "loss": 1.0005, + "step": 10169 + }, + { + "epoch": 0.33, + "learning_rate": 1.576669284404239e-05, + "loss": 0.9116, + "step": 10170 + }, + { + "epoch": 0.33, + "learning_rate": 1.576584658459223e-05, + "loss": 1.0508, + "step": 10171 + }, + { + "epoch": 0.33, + "learning_rate": 1.5765000263280912e-05, + "loss": 1.0479, + "step": 10172 + }, + { + "epoch": 0.33, + "learning_rate": 1.5764153880117528e-05, + "loss": 1.001, + "step": 10173 + }, + { + "epoch": 0.33, + "learning_rate": 1.576330743511115e-05, + "loss": 1.1372, + "step": 10174 + }, + { + "epoch": 0.33, + "learning_rate": 1.5762460928270858e-05, + "loss": 1.0811, + "step": 10175 + }, + { + "epoch": 0.33, + "learning_rate": 1.5761614359605744e-05, + "loss": 1.103, + "step": 10176 + }, + { + "epoch": 0.33, + "learning_rate": 1.5760767729124883e-05, + "loss": 0.9722, + "step": 10177 + }, + { + "epoch": 0.33, + "learning_rate": 1.5759921036837357e-05, + "loss": 0.9414, + "step": 10178 + }, + { + "epoch": 0.33, + "learning_rate": 1.5759074282752254e-05, + "loss": 1.1553, + "step": 10179 + }, + { + "epoch": 0.33, + "learning_rate": 1.5758227466878663e-05, + "loss": 0.9756, + "step": 10180 + }, + { + "epoch": 0.33, + "learning_rate": 1.575738058922566e-05, + "loss": 1.0415, + "step": 10181 + }, + { + "epoch": 0.33, + "learning_rate": 1.5756533649802336e-05, + "loss": 1.0752, + "step": 10182 + }, + { + "epoch": 0.33, + "learning_rate": 1.5755686648617774e-05, + "loss": 1.0234, + "step": 10183 + }, + { + "epoch": 0.33, + "learning_rate": 1.5754839585681064e-05, + "loss": 0.9917, + "step": 10184 + }, + { + "epoch": 0.33, + "learning_rate": 1.57539924610013e-05, + "loss": 1.0732, + "step": 10185 + }, + { + "epoch": 0.33, + "learning_rate": 1.5753145274587566e-05, + "loss": 1.0537, + "step": 10186 + }, + { + "epoch": 0.33, + "learning_rate": 1.5752298026448945e-05, + "loss": 0.9487, + "step": 10187 + }, + { + "epoch": 0.33, + "learning_rate": 1.5751450716594535e-05, + "loss": 1.0703, + "step": 10188 + }, + { + "epoch": 0.33, + "learning_rate": 1.5750603345033422e-05, + "loss": 0.9609, + "step": 10189 + }, + { + "epoch": 0.33, + "learning_rate": 1.57497559117747e-05, + "loss": 1.0483, + "step": 10190 + }, + { + "epoch": 0.33, + "learning_rate": 1.574890841682746e-05, + "loss": 1.0991, + "step": 10191 + }, + { + "epoch": 0.33, + "learning_rate": 1.57480608602008e-05, + "loss": 1.0586, + "step": 10192 + }, + { + "epoch": 0.33, + "learning_rate": 1.5747213241903803e-05, + "loss": 1.0068, + "step": 10193 + }, + { + "epoch": 0.33, + "learning_rate": 1.574636556194557e-05, + "loss": 1.0728, + "step": 10194 + }, + { + "epoch": 0.33, + "learning_rate": 1.5745517820335197e-05, + "loss": 0.999, + "step": 10195 + }, + { + "epoch": 0.33, + "learning_rate": 1.5744670017081774e-05, + "loss": 1.0693, + "step": 10196 + }, + { + "epoch": 0.33, + "learning_rate": 1.5743822152194402e-05, + "loss": 1.084, + "step": 10197 + }, + { + "epoch": 0.33, + "learning_rate": 1.5742974225682175e-05, + "loss": 1.0308, + "step": 10198 + }, + { + "epoch": 0.33, + "learning_rate": 1.574212623755419e-05, + "loss": 1.0156, + "step": 10199 + }, + { + "epoch": 0.33, + "learning_rate": 1.5741278187819542e-05, + "loss": 1.0034, + "step": 10200 + }, + { + "epoch": 0.33, + "learning_rate": 1.5740430076487337e-05, + "loss": 1.2524, + "step": 10201 + }, + { + "epoch": 0.33, + "learning_rate": 1.5739581903566666e-05, + "loss": 1.0571, + "step": 10202 + }, + { + "epoch": 0.33, + "learning_rate": 1.5738733669066642e-05, + "loss": 0.9819, + "step": 10203 + }, + { + "epoch": 0.33, + "learning_rate": 1.5737885372996352e-05, + "loss": 1.0767, + "step": 10204 + }, + { + "epoch": 0.33, + "learning_rate": 1.57370370153649e-05, + "loss": 0.9062, + "step": 10205 + }, + { + "epoch": 0.33, + "learning_rate": 1.5736188596181393e-05, + "loss": 0.9956, + "step": 10206 + }, + { + "epoch": 0.33, + "learning_rate": 1.5735340115454932e-05, + "loss": 1.0195, + "step": 10207 + }, + { + "epoch": 0.33, + "learning_rate": 1.5734491573194617e-05, + "loss": 1.0352, + "step": 10208 + }, + { + "epoch": 0.33, + "learning_rate": 1.5733642969409553e-05, + "loss": 0.979, + "step": 10209 + }, + { + "epoch": 0.33, + "learning_rate": 1.5732794304108846e-05, + "loss": 0.9961, + "step": 10210 + }, + { + "epoch": 0.33, + "learning_rate": 1.5731945577301604e-05, + "loss": 1.1401, + "step": 10211 + }, + { + "epoch": 0.33, + "learning_rate": 1.5731096788996928e-05, + "loss": 1.0996, + "step": 10212 + }, + { + "epoch": 0.33, + "learning_rate": 1.5730247939203928e-05, + "loss": 0.9951, + "step": 10213 + }, + { + "epoch": 0.33, + "learning_rate": 1.5729399027931704e-05, + "loss": 1.0381, + "step": 10214 + }, + { + "epoch": 0.33, + "learning_rate": 1.5728550055189374e-05, + "loss": 1.0137, + "step": 10215 + }, + { + "epoch": 0.33, + "learning_rate": 1.5727701020986042e-05, + "loss": 0.9458, + "step": 10216 + }, + { + "epoch": 0.33, + "learning_rate": 1.5726851925330816e-05, + "loss": 0.9995, + "step": 10217 + }, + { + "epoch": 0.33, + "learning_rate": 1.5726002768232805e-05, + "loss": 1.0991, + "step": 10218 + }, + { + "epoch": 0.33, + "learning_rate": 1.5725153549701128e-05, + "loss": 1.0166, + "step": 10219 + }, + { + "epoch": 0.33, + "learning_rate": 1.5724304269744885e-05, + "loss": 1.1768, + "step": 10220 + }, + { + "epoch": 0.33, + "learning_rate": 1.572345492837319e-05, + "loss": 0.9727, + "step": 10221 + }, + { + "epoch": 0.33, + "learning_rate": 1.5722605525595166e-05, + "loss": 1.126, + "step": 10222 + }, + { + "epoch": 0.33, + "learning_rate": 1.572175606141991e-05, + "loss": 1.1948, + "step": 10223 + }, + { + "epoch": 0.33, + "learning_rate": 1.572090653585655e-05, + "loss": 1.0166, + "step": 10224 + }, + { + "epoch": 0.33, + "learning_rate": 1.5720056948914187e-05, + "loss": 1.0615, + "step": 10225 + }, + { + "epoch": 0.33, + "learning_rate": 1.571920730060195e-05, + "loss": 1.0068, + "step": 10226 + }, + { + "epoch": 0.33, + "learning_rate": 1.5718357590928945e-05, + "loss": 1.0493, + "step": 10227 + }, + { + "epoch": 0.33, + "learning_rate": 1.5717507819904294e-05, + "loss": 0.9482, + "step": 10228 + }, + { + "epoch": 0.33, + "learning_rate": 1.5716657987537113e-05, + "loss": 1.062, + "step": 10229 + }, + { + "epoch": 0.33, + "learning_rate": 1.5715808093836517e-05, + "loss": 0.9443, + "step": 10230 + }, + { + "epoch": 0.33, + "learning_rate": 1.5714958138811623e-05, + "loss": 1.0649, + "step": 10231 + }, + { + "epoch": 0.33, + "learning_rate": 1.571410812247156e-05, + "loss": 1.0161, + "step": 10232 + }, + { + "epoch": 0.33, + "learning_rate": 1.5713258044825437e-05, + "loss": 1.1362, + "step": 10233 + }, + { + "epoch": 0.33, + "learning_rate": 1.5712407905882375e-05, + "loss": 1.021, + "step": 10234 + }, + { + "epoch": 0.33, + "learning_rate": 1.5711557705651506e-05, + "loss": 0.957, + "step": 10235 + }, + { + "epoch": 0.33, + "learning_rate": 1.5710707444141937e-05, + "loss": 1.1689, + "step": 10236 + }, + { + "epoch": 0.33, + "learning_rate": 1.5709857121362802e-05, + "loss": 1.0381, + "step": 10237 + }, + { + "epoch": 0.33, + "learning_rate": 1.5709006737323215e-05, + "loss": 0.9199, + "step": 10238 + }, + { + "epoch": 0.33, + "learning_rate": 1.5708156292032308e-05, + "loss": 0.9756, + "step": 10239 + }, + { + "epoch": 0.33, + "learning_rate": 1.57073057854992e-05, + "loss": 1.0557, + "step": 10240 + }, + { + "epoch": 0.33, + "learning_rate": 1.5706455217733017e-05, + "loss": 1.1006, + "step": 10241 + }, + { + "epoch": 0.33, + "learning_rate": 1.570560458874289e-05, + "loss": 1.0308, + "step": 10242 + }, + { + "epoch": 0.33, + "learning_rate": 1.5704753898537935e-05, + "loss": 1.04, + "step": 10243 + }, + { + "epoch": 0.33, + "learning_rate": 1.5703903147127284e-05, + "loss": 1.0508, + "step": 10244 + }, + { + "epoch": 0.33, + "learning_rate": 1.5703052334520068e-05, + "loss": 0.5996, + "step": 10245 + }, + { + "epoch": 0.33, + "learning_rate": 1.570220146072541e-05, + "loss": 0.9165, + "step": 10246 + }, + { + "epoch": 0.33, + "learning_rate": 1.5701350525752444e-05, + "loss": 0.9482, + "step": 10247 + }, + { + "epoch": 0.33, + "learning_rate": 1.5700499529610295e-05, + "loss": 1.0601, + "step": 10248 + }, + { + "epoch": 0.33, + "learning_rate": 1.5699648472308095e-05, + "loss": 1.0034, + "step": 10249 + }, + { + "epoch": 0.33, + "learning_rate": 1.5698797353854974e-05, + "loss": 1.1504, + "step": 10250 + }, + { + "epoch": 0.33, + "learning_rate": 1.5697946174260065e-05, + "loss": 1.1558, + "step": 10251 + }, + { + "epoch": 0.33, + "learning_rate": 1.56970949335325e-05, + "loss": 1.125, + "step": 10252 + }, + { + "epoch": 0.33, + "learning_rate": 1.5696243631681412e-05, + "loss": 1.1768, + "step": 10253 + }, + { + "epoch": 0.33, + "learning_rate": 1.5695392268715934e-05, + "loss": 1.0459, + "step": 10254 + }, + { + "epoch": 0.33, + "learning_rate": 1.56945408446452e-05, + "loss": 1.0737, + "step": 10255 + }, + { + "epoch": 0.33, + "learning_rate": 1.5693689359478345e-05, + "loss": 0.9902, + "step": 10256 + }, + { + "epoch": 0.33, + "learning_rate": 1.5692837813224502e-05, + "loss": 1.0547, + "step": 10257 + }, + { + "epoch": 0.33, + "learning_rate": 1.5691986205892815e-05, + "loss": 1.002, + "step": 10258 + }, + { + "epoch": 0.33, + "learning_rate": 1.5691134537492413e-05, + "loss": 1.2153, + "step": 10259 + }, + { + "epoch": 0.33, + "learning_rate": 1.569028280803244e-05, + "loss": 0.98, + "step": 10260 + }, + { + "epoch": 0.33, + "learning_rate": 1.5689431017522022e-05, + "loss": 1.041, + "step": 10261 + }, + { + "epoch": 0.33, + "learning_rate": 1.5688579165970312e-05, + "loss": 0.9907, + "step": 10262 + }, + { + "epoch": 0.33, + "learning_rate": 1.5687727253386443e-05, + "loss": 1.0093, + "step": 10263 + }, + { + "epoch": 0.33, + "learning_rate": 1.5686875279779555e-05, + "loss": 1.0791, + "step": 10264 + }, + { + "epoch": 0.33, + "learning_rate": 1.5686023245158787e-05, + "loss": 1.0981, + "step": 10265 + }, + { + "epoch": 0.33, + "learning_rate": 1.5685171149533283e-05, + "loss": 1.0049, + "step": 10266 + }, + { + "epoch": 0.33, + "learning_rate": 1.5684318992912188e-05, + "loss": 1.1289, + "step": 10267 + }, + { + "epoch": 0.33, + "learning_rate": 1.568346677530464e-05, + "loss": 0.9624, + "step": 10268 + }, + { + "epoch": 0.33, + "learning_rate": 1.568261449671978e-05, + "loss": 1.0356, + "step": 10269 + }, + { + "epoch": 0.33, + "learning_rate": 1.568176215716676e-05, + "loss": 1.1035, + "step": 10270 + }, + { + "epoch": 0.33, + "learning_rate": 1.5680909756654717e-05, + "loss": 1.0771, + "step": 10271 + }, + { + "epoch": 0.33, + "learning_rate": 1.56800572951928e-05, + "loss": 1.0322, + "step": 10272 + }, + { + "epoch": 0.33, + "learning_rate": 1.5679204772790155e-05, + "loss": 0.9917, + "step": 10273 + }, + { + "epoch": 0.33, + "learning_rate": 1.567835218945593e-05, + "loss": 1.1113, + "step": 10274 + }, + { + "epoch": 0.33, + "learning_rate": 1.567749954519927e-05, + "loss": 1.1216, + "step": 10275 + }, + { + "epoch": 0.33, + "learning_rate": 1.5676646840029324e-05, + "loss": 1.064, + "step": 10276 + }, + { + "epoch": 0.33, + "learning_rate": 1.567579407395524e-05, + "loss": 0.9775, + "step": 10277 + }, + { + "epoch": 0.33, + "learning_rate": 1.5674941246986164e-05, + "loss": 0.9937, + "step": 10278 + }, + { + "epoch": 0.33, + "learning_rate": 1.5674088359131253e-05, + "loss": 1.0952, + "step": 10279 + }, + { + "epoch": 0.33, + "learning_rate": 1.567323541039965e-05, + "loss": 1.0112, + "step": 10280 + }, + { + "epoch": 0.33, + "learning_rate": 1.5672382400800514e-05, + "loss": 1.1504, + "step": 10281 + }, + { + "epoch": 0.33, + "learning_rate": 1.5671529330342988e-05, + "loss": 0.9868, + "step": 10282 + }, + { + "epoch": 0.33, + "learning_rate": 1.5670676199036236e-05, + "loss": 1.1641, + "step": 10283 + }, + { + "epoch": 0.33, + "learning_rate": 1.5669823006889397e-05, + "loss": 1.0513, + "step": 10284 + }, + { + "epoch": 0.33, + "learning_rate": 1.5668969753911633e-05, + "loss": 1.0903, + "step": 10285 + }, + { + "epoch": 0.33, + "learning_rate": 1.5668116440112103e-05, + "loss": 1.0322, + "step": 10286 + }, + { + "epoch": 0.33, + "learning_rate": 1.5667263065499954e-05, + "loss": 0.9834, + "step": 10287 + }, + { + "epoch": 0.33, + "learning_rate": 1.5666409630084344e-05, + "loss": 0.9551, + "step": 10288 + }, + { + "epoch": 0.33, + "learning_rate": 1.566555613387443e-05, + "loss": 1.0864, + "step": 10289 + }, + { + "epoch": 0.33, + "learning_rate": 1.5664702576879373e-05, + "loss": 1.0439, + "step": 10290 + }, + { + "epoch": 0.33, + "learning_rate": 1.566384895910832e-05, + "loss": 1.0156, + "step": 10291 + }, + { + "epoch": 0.33, + "learning_rate": 1.566299528057044e-05, + "loss": 1.0615, + "step": 10292 + }, + { + "epoch": 0.33, + "learning_rate": 1.5662141541274886e-05, + "loss": 1.0713, + "step": 10293 + }, + { + "epoch": 0.33, + "learning_rate": 1.566128774123082e-05, + "loss": 0.9395, + "step": 10294 + }, + { + "epoch": 0.33, + "learning_rate": 1.5660433880447405e-05, + "loss": 1.0654, + "step": 10295 + }, + { + "epoch": 0.33, + "learning_rate": 1.5659579958933797e-05, + "loss": 1.0137, + "step": 10296 + }, + { + "epoch": 0.33, + "learning_rate": 1.5658725976699158e-05, + "loss": 0.9028, + "step": 10297 + }, + { + "epoch": 0.33, + "learning_rate": 1.5657871933752657e-05, + "loss": 1.1162, + "step": 10298 + }, + { + "epoch": 0.33, + "learning_rate": 1.5657017830103448e-05, + "loss": 1.0469, + "step": 10299 + }, + { + "epoch": 0.33, + "learning_rate": 1.56561636657607e-05, + "loss": 1.0825, + "step": 10300 + }, + { + "epoch": 0.33, + "learning_rate": 1.5655309440733574e-05, + "loss": 1.0728, + "step": 10301 + }, + { + "epoch": 0.33, + "learning_rate": 1.565445515503124e-05, + "loss": 0.5347, + "step": 10302 + }, + { + "epoch": 0.33, + "learning_rate": 1.5653600808662858e-05, + "loss": 1.0239, + "step": 10303 + }, + { + "epoch": 0.33, + "learning_rate": 1.5652746401637595e-05, + "loss": 0.9302, + "step": 10304 + }, + { + "epoch": 0.33, + "learning_rate": 1.5651891933964623e-05, + "loss": 1.0059, + "step": 10305 + }, + { + "epoch": 0.33, + "learning_rate": 1.5651037405653102e-05, + "loss": 1.0742, + "step": 10306 + }, + { + "epoch": 0.33, + "learning_rate": 1.5650182816712206e-05, + "loss": 0.9678, + "step": 10307 + }, + { + "epoch": 0.33, + "learning_rate": 1.5649328167151103e-05, + "loss": 1.022, + "step": 10308 + }, + { + "epoch": 0.33, + "learning_rate": 1.5648473456978958e-05, + "loss": 0.999, + "step": 10309 + }, + { + "epoch": 0.33, + "learning_rate": 1.5647618686204947e-05, + "loss": 1.0259, + "step": 10310 + }, + { + "epoch": 0.33, + "learning_rate": 1.5646763854838233e-05, + "loss": 0.9736, + "step": 10311 + }, + { + "epoch": 0.33, + "learning_rate": 1.5645908962887997e-05, + "loss": 1.0137, + "step": 10312 + }, + { + "epoch": 0.33, + "learning_rate": 1.564505401036341e-05, + "loss": 1.0688, + "step": 10313 + }, + { + "epoch": 0.33, + "learning_rate": 1.5644198997273634e-05, + "loss": 0.978, + "step": 10314 + }, + { + "epoch": 0.33, + "learning_rate": 1.564334392362785e-05, + "loss": 1.0425, + "step": 10315 + }, + { + "epoch": 0.33, + "learning_rate": 1.5642488789435234e-05, + "loss": 1.0127, + "step": 10316 + }, + { + "epoch": 0.33, + "learning_rate": 1.564163359470496e-05, + "loss": 1.1079, + "step": 10317 + }, + { + "epoch": 0.33, + "learning_rate": 1.5640778339446198e-05, + "loss": 1.0225, + "step": 10318 + }, + { + "epoch": 0.33, + "learning_rate": 1.5639923023668125e-05, + "loss": 1.0933, + "step": 10319 + }, + { + "epoch": 0.33, + "learning_rate": 1.5639067647379923e-05, + "loss": 0.9858, + "step": 10320 + }, + { + "epoch": 0.33, + "learning_rate": 1.563821221059077e-05, + "loss": 1.019, + "step": 10321 + }, + { + "epoch": 0.33, + "learning_rate": 1.5637356713309833e-05, + "loss": 1.0239, + "step": 10322 + }, + { + "epoch": 0.33, + "learning_rate": 1.5636501155546303e-05, + "loss": 1.0039, + "step": 10323 + }, + { + "epoch": 0.33, + "learning_rate": 1.5635645537309353e-05, + "loss": 0.8765, + "step": 10324 + }, + { + "epoch": 0.33, + "learning_rate": 1.5634789858608162e-05, + "loss": 1.0122, + "step": 10325 + }, + { + "epoch": 0.33, + "learning_rate": 1.563393411945191e-05, + "loss": 0.9678, + "step": 10326 + }, + { + "epoch": 0.33, + "learning_rate": 1.5633078319849783e-05, + "loss": 1.0488, + "step": 10327 + }, + { + "epoch": 0.33, + "learning_rate": 1.563222245981096e-05, + "loss": 1.1162, + "step": 10328 + }, + { + "epoch": 0.33, + "learning_rate": 1.5631366539344622e-05, + "loss": 0.9707, + "step": 10329 + }, + { + "epoch": 0.33, + "learning_rate": 1.563051055845996e-05, + "loss": 1.1313, + "step": 10330 + }, + { + "epoch": 0.33, + "learning_rate": 1.5629654517166142e-05, + "loss": 1.0518, + "step": 10331 + }, + { + "epoch": 0.33, + "learning_rate": 1.5628798415472367e-05, + "loss": 1.0088, + "step": 10332 + }, + { + "epoch": 0.33, + "learning_rate": 1.562794225338781e-05, + "loss": 1.0586, + "step": 10333 + }, + { + "epoch": 0.33, + "learning_rate": 1.5627086030921666e-05, + "loss": 0.9897, + "step": 10334 + }, + { + "epoch": 0.33, + "learning_rate": 1.5626229748083116e-05, + "loss": 1.0747, + "step": 10335 + }, + { + "epoch": 0.33, + "learning_rate": 1.5625373404881346e-05, + "loss": 1.0103, + "step": 10336 + }, + { + "epoch": 0.33, + "learning_rate": 1.5624517001325545e-05, + "loss": 1.0713, + "step": 10337 + }, + { + "epoch": 0.33, + "learning_rate": 1.5623660537424903e-05, + "loss": 1.0479, + "step": 10338 + }, + { + "epoch": 0.33, + "learning_rate": 1.5622804013188608e-05, + "loss": 1.02, + "step": 10339 + }, + { + "epoch": 0.33, + "learning_rate": 1.5621947428625846e-05, + "loss": 0.9868, + "step": 10340 + }, + { + "epoch": 0.33, + "learning_rate": 1.562109078374581e-05, + "loss": 1.0684, + "step": 10341 + }, + { + "epoch": 0.33, + "learning_rate": 1.5620234078557694e-05, + "loss": 1.0596, + "step": 10342 + }, + { + "epoch": 0.33, + "learning_rate": 1.5619377313070688e-05, + "loss": 0.9287, + "step": 10343 + }, + { + "epoch": 0.33, + "learning_rate": 1.561852048729398e-05, + "loss": 1.1226, + "step": 10344 + }, + { + "epoch": 0.33, + "learning_rate": 1.5617663601236763e-05, + "loss": 1.0376, + "step": 10345 + }, + { + "epoch": 0.33, + "learning_rate": 1.5616806654908235e-05, + "loss": 1.1177, + "step": 10346 + }, + { + "epoch": 0.33, + "learning_rate": 1.561594964831759e-05, + "loss": 1.0537, + "step": 10347 + }, + { + "epoch": 0.33, + "learning_rate": 1.5615092581474017e-05, + "loss": 1.0908, + "step": 10348 + }, + { + "epoch": 0.33, + "learning_rate": 1.5614235454386718e-05, + "loss": 0.9922, + "step": 10349 + }, + { + "epoch": 0.33, + "learning_rate": 1.5613378267064886e-05, + "loss": 1.0015, + "step": 10350 + }, + { + "epoch": 0.33, + "learning_rate": 1.5612521019517717e-05, + "loss": 1.0586, + "step": 10351 + }, + { + "epoch": 0.33, + "learning_rate": 1.5611663711754407e-05, + "loss": 1.1558, + "step": 10352 + }, + { + "epoch": 0.33, + "learning_rate": 1.561080634378416e-05, + "loss": 0.9902, + "step": 10353 + }, + { + "epoch": 0.33, + "learning_rate": 1.560994891561617e-05, + "loss": 0.8901, + "step": 10354 + }, + { + "epoch": 0.33, + "learning_rate": 1.5609091427259636e-05, + "loss": 0.5571, + "step": 10355 + }, + { + "epoch": 0.33, + "learning_rate": 1.5608233878723758e-05, + "loss": 0.8848, + "step": 10356 + }, + { + "epoch": 0.33, + "learning_rate": 1.560737627001774e-05, + "loss": 1.0386, + "step": 10357 + }, + { + "epoch": 0.33, + "learning_rate": 1.5606518601150775e-05, + "loss": 1.0024, + "step": 10358 + }, + { + "epoch": 0.33, + "learning_rate": 1.5605660872132074e-05, + "loss": 1.0376, + "step": 10359 + }, + { + "epoch": 0.33, + "learning_rate": 1.5604803082970836e-05, + "loss": 1.0762, + "step": 10360 + }, + { + "epoch": 0.33, + "learning_rate": 1.5603945233676263e-05, + "loss": 0.9648, + "step": 10361 + }, + { + "epoch": 0.33, + "learning_rate": 1.5603087324257562e-05, + "loss": 1.165, + "step": 10362 + }, + { + "epoch": 0.33, + "learning_rate": 1.5602229354723934e-05, + "loss": 1.0542, + "step": 10363 + }, + { + "epoch": 0.33, + "learning_rate": 1.5601371325084584e-05, + "loss": 1.0381, + "step": 10364 + }, + { + "epoch": 0.33, + "learning_rate": 1.560051323534872e-05, + "loss": 1.1196, + "step": 10365 + }, + { + "epoch": 0.33, + "learning_rate": 1.559965508552555e-05, + "loss": 0.9482, + "step": 10366 + }, + { + "epoch": 0.33, + "learning_rate": 1.5598796875624273e-05, + "loss": 0.5698, + "step": 10367 + }, + { + "epoch": 0.33, + "learning_rate": 1.5597938605654103e-05, + "loss": 1.0762, + "step": 10368 + }, + { + "epoch": 0.33, + "learning_rate": 1.559708027562425e-05, + "loss": 1.0381, + "step": 10369 + }, + { + "epoch": 0.33, + "learning_rate": 1.559622188554392e-05, + "loss": 0.8828, + "step": 10370 + }, + { + "epoch": 0.33, + "learning_rate": 1.5595363435422323e-05, + "loss": 0.9922, + "step": 10371 + }, + { + "epoch": 0.33, + "learning_rate": 1.5594504925268666e-05, + "loss": 0.9595, + "step": 10372 + }, + { + "epoch": 0.33, + "learning_rate": 1.5593646355092165e-05, + "loss": 1.0747, + "step": 10373 + }, + { + "epoch": 0.33, + "learning_rate": 1.559278772490203e-05, + "loss": 0.8867, + "step": 10374 + }, + { + "epoch": 0.33, + "learning_rate": 1.5591929034707468e-05, + "loss": 1.0103, + "step": 10375 + }, + { + "epoch": 0.33, + "learning_rate": 1.55910702845177e-05, + "loss": 1.1367, + "step": 10376 + }, + { + "epoch": 0.33, + "learning_rate": 1.559021147434194e-05, + "loss": 1.0991, + "step": 10377 + }, + { + "epoch": 0.33, + "learning_rate": 1.558935260418939e-05, + "loss": 1.1011, + "step": 10378 + }, + { + "epoch": 0.33, + "learning_rate": 1.5588493674069276e-05, + "loss": 1.0225, + "step": 10379 + }, + { + "epoch": 0.33, + "learning_rate": 1.558763468399081e-05, + "loss": 1.0039, + "step": 10380 + }, + { + "epoch": 0.33, + "learning_rate": 1.5586775633963208e-05, + "loss": 1.0376, + "step": 10381 + }, + { + "epoch": 0.33, + "learning_rate": 1.5585916523995686e-05, + "loss": 1.0747, + "step": 10382 + }, + { + "epoch": 0.33, + "learning_rate": 1.5585057354097465e-05, + "loss": 1.0195, + "step": 10383 + }, + { + "epoch": 0.33, + "learning_rate": 1.5584198124277755e-05, + "loss": 1.1069, + "step": 10384 + }, + { + "epoch": 0.33, + "learning_rate": 1.5583338834545782e-05, + "loss": 0.9517, + "step": 10385 + }, + { + "epoch": 0.33, + "learning_rate": 1.558247948491076e-05, + "loss": 0.9492, + "step": 10386 + }, + { + "epoch": 0.33, + "learning_rate": 1.5581620075381912e-05, + "loss": 0.9888, + "step": 10387 + }, + { + "epoch": 0.33, + "learning_rate": 1.558076060596846e-05, + "loss": 1.123, + "step": 10388 + }, + { + "epoch": 0.33, + "learning_rate": 1.5579901076679625e-05, + "loss": 0.958, + "step": 10389 + }, + { + "epoch": 0.33, + "learning_rate": 1.5579041487524626e-05, + "loss": 1.0391, + "step": 10390 + }, + { + "epoch": 0.33, + "learning_rate": 1.5578181838512684e-05, + "loss": 1.0366, + "step": 10391 + }, + { + "epoch": 0.33, + "learning_rate": 1.5577322129653027e-05, + "loss": 1.0674, + "step": 10392 + }, + { + "epoch": 0.33, + "learning_rate": 1.5576462360954873e-05, + "loss": 1.063, + "step": 10393 + }, + { + "epoch": 0.33, + "learning_rate": 1.557560253242745e-05, + "loss": 1.0767, + "step": 10394 + }, + { + "epoch": 0.33, + "learning_rate": 1.5574742644079984e-05, + "loss": 0.998, + "step": 10395 + }, + { + "epoch": 0.33, + "learning_rate": 1.55738826959217e-05, + "loss": 0.936, + "step": 10396 + }, + { + "epoch": 0.33, + "learning_rate": 1.557302268796182e-05, + "loss": 1.1162, + "step": 10397 + }, + { + "epoch": 0.33, + "learning_rate": 1.5572162620209578e-05, + "loss": 1.1025, + "step": 10398 + }, + { + "epoch": 0.33, + "learning_rate": 1.55713024926742e-05, + "loss": 1.0186, + "step": 10399 + }, + { + "epoch": 0.33, + "learning_rate": 1.557044230536491e-05, + "loss": 1.0166, + "step": 10400 + }, + { + "epoch": 0.33, + "learning_rate": 1.556958205829094e-05, + "loss": 1.1172, + "step": 10401 + }, + { + "epoch": 0.33, + "learning_rate": 1.5568721751461518e-05, + "loss": 1.1851, + "step": 10402 + }, + { + "epoch": 0.33, + "learning_rate": 1.5567861384885873e-05, + "loss": 0.9536, + "step": 10403 + }, + { + "epoch": 0.33, + "learning_rate": 1.5567000958573236e-05, + "loss": 1.1084, + "step": 10404 + }, + { + "epoch": 0.33, + "learning_rate": 1.5566140472532843e-05, + "loss": 0.9414, + "step": 10405 + }, + { + "epoch": 0.33, + "learning_rate": 1.5565279926773924e-05, + "loss": 0.8052, + "step": 10406 + }, + { + "epoch": 0.33, + "learning_rate": 1.556441932130571e-05, + "loss": 0.9727, + "step": 10407 + }, + { + "epoch": 0.33, + "learning_rate": 1.5563558656137433e-05, + "loss": 1.3188, + "step": 10408 + }, + { + "epoch": 0.33, + "learning_rate": 1.5562697931278334e-05, + "loss": 0.9707, + "step": 10409 + }, + { + "epoch": 0.33, + "learning_rate": 1.556183714673764e-05, + "loss": 1.0796, + "step": 10410 + }, + { + "epoch": 0.33, + "learning_rate": 1.5560976302524587e-05, + "loss": 1.0981, + "step": 10411 + }, + { + "epoch": 0.33, + "learning_rate": 1.556011539864841e-05, + "loss": 0.9829, + "step": 10412 + }, + { + "epoch": 0.33, + "learning_rate": 1.5559254435118354e-05, + "loss": 1.0576, + "step": 10413 + }, + { + "epoch": 0.33, + "learning_rate": 1.555839341194365e-05, + "loss": 1.0142, + "step": 10414 + }, + { + "epoch": 0.33, + "learning_rate": 1.5557532329133536e-05, + "loss": 0.9653, + "step": 10415 + }, + { + "epoch": 0.33, + "learning_rate": 1.555667118669725e-05, + "loss": 1.0259, + "step": 10416 + }, + { + "epoch": 0.33, + "learning_rate": 1.555580998464403e-05, + "loss": 0.9487, + "step": 10417 + }, + { + "epoch": 0.33, + "learning_rate": 1.555494872298312e-05, + "loss": 1.1348, + "step": 10418 + }, + { + "epoch": 0.33, + "learning_rate": 1.5554087401723758e-05, + "loss": 0.7952, + "step": 10419 + }, + { + "epoch": 0.33, + "learning_rate": 1.5553226020875184e-05, + "loss": 1.0107, + "step": 10420 + }, + { + "epoch": 0.33, + "learning_rate": 1.5552364580446642e-05, + "loss": 0.9775, + "step": 10421 + }, + { + "epoch": 0.33, + "learning_rate": 1.5551503080447378e-05, + "loss": 0.96, + "step": 10422 + }, + { + "epoch": 0.33, + "learning_rate": 1.5550641520886623e-05, + "loss": 1.0142, + "step": 10423 + }, + { + "epoch": 0.33, + "learning_rate": 1.5549779901773628e-05, + "loss": 1.0479, + "step": 10424 + }, + { + "epoch": 0.33, + "learning_rate": 1.5548918223117642e-05, + "loss": 1.0669, + "step": 10425 + }, + { + "epoch": 0.33, + "learning_rate": 1.55480564849279e-05, + "loss": 0.9849, + "step": 10426 + }, + { + "epoch": 0.33, + "learning_rate": 1.5547194687213652e-05, + "loss": 1.0405, + "step": 10427 + }, + { + "epoch": 0.33, + "learning_rate": 1.5546332829984147e-05, + "loss": 0.8823, + "step": 10428 + }, + { + "epoch": 0.33, + "learning_rate": 1.5545470913248627e-05, + "loss": 1.0869, + "step": 10429 + }, + { + "epoch": 0.33, + "learning_rate": 1.5544608937016346e-05, + "loss": 1.0454, + "step": 10430 + }, + { + "epoch": 0.33, + "learning_rate": 1.5543746901296545e-05, + "loss": 1.0234, + "step": 10431 + }, + { + "epoch": 0.33, + "learning_rate": 1.5542884806098474e-05, + "loss": 1.1064, + "step": 10432 + }, + { + "epoch": 0.33, + "learning_rate": 1.5542022651431384e-05, + "loss": 1.1133, + "step": 10433 + }, + { + "epoch": 0.33, + "learning_rate": 1.5541160437304524e-05, + "loss": 1.0581, + "step": 10434 + }, + { + "epoch": 0.33, + "learning_rate": 1.5540298163727144e-05, + "loss": 1.0161, + "step": 10435 + }, + { + "epoch": 0.33, + "learning_rate": 1.55394358307085e-05, + "loss": 1.1108, + "step": 10436 + }, + { + "epoch": 0.33, + "learning_rate": 1.553857343825784e-05, + "loss": 1.0039, + "step": 10437 + }, + { + "epoch": 0.33, + "learning_rate": 1.553771098638441e-05, + "loss": 1.0884, + "step": 10438 + }, + { + "epoch": 0.33, + "learning_rate": 1.5536848475097478e-05, + "loss": 1.1138, + "step": 10439 + }, + { + "epoch": 0.33, + "learning_rate": 1.5535985904406284e-05, + "loss": 1.0249, + "step": 10440 + }, + { + "epoch": 0.33, + "learning_rate": 1.5535123274320093e-05, + "loss": 1.001, + "step": 10441 + }, + { + "epoch": 0.33, + "learning_rate": 1.5534260584848154e-05, + "loss": 1.0283, + "step": 10442 + }, + { + "epoch": 0.33, + "learning_rate": 1.5533397835999724e-05, + "loss": 1.0146, + "step": 10443 + }, + { + "epoch": 0.33, + "learning_rate": 1.553253502778406e-05, + "loss": 1.1001, + "step": 10444 + }, + { + "epoch": 0.33, + "learning_rate": 1.5531672160210413e-05, + "loss": 1.0879, + "step": 10445 + }, + { + "epoch": 0.33, + "learning_rate": 1.553080923328805e-05, + "loss": 1.0225, + "step": 10446 + }, + { + "epoch": 0.33, + "learning_rate": 1.5529946247026227e-05, + "loss": 1.0352, + "step": 10447 + }, + { + "epoch": 0.33, + "learning_rate": 1.55290832014342e-05, + "loss": 0.9893, + "step": 10448 + }, + { + "epoch": 0.33, + "learning_rate": 1.5528220096521227e-05, + "loss": 1.04, + "step": 10449 + }, + { + "epoch": 0.33, + "learning_rate": 1.552735693229657e-05, + "loss": 1.1108, + "step": 10450 + }, + { + "epoch": 0.33, + "learning_rate": 1.5526493708769495e-05, + "loss": 1.1084, + "step": 10451 + }, + { + "epoch": 0.33, + "learning_rate": 1.5525630425949255e-05, + "loss": 1.0396, + "step": 10452 + }, + { + "epoch": 0.33, + "learning_rate": 1.552476708384512e-05, + "loss": 1.0898, + "step": 10453 + }, + { + "epoch": 0.33, + "learning_rate": 1.5523903682466345e-05, + "loss": 1.1074, + "step": 10454 + }, + { + "epoch": 0.33, + "learning_rate": 1.5523040221822197e-05, + "loss": 1.0845, + "step": 10455 + }, + { + "epoch": 0.33, + "learning_rate": 1.5522176701921942e-05, + "loss": 1.0576, + "step": 10456 + }, + { + "epoch": 0.33, + "learning_rate": 1.5521313122774843e-05, + "loss": 1.127, + "step": 10457 + }, + { + "epoch": 0.33, + "learning_rate": 1.5520449484390162e-05, + "loss": 0.9497, + "step": 10458 + }, + { + "epoch": 0.33, + "learning_rate": 1.551958578677717e-05, + "loss": 0.9683, + "step": 10459 + }, + { + "epoch": 0.33, + "learning_rate": 1.551872202994513e-05, + "loss": 1.106, + "step": 10460 + }, + { + "epoch": 0.33, + "learning_rate": 1.551785821390331e-05, + "loss": 0.9961, + "step": 10461 + }, + { + "epoch": 0.33, + "learning_rate": 1.551699433866098e-05, + "loss": 0.9712, + "step": 10462 + }, + { + "epoch": 0.33, + "learning_rate": 1.5516130404227406e-05, + "loss": 1.0635, + "step": 10463 + }, + { + "epoch": 0.33, + "learning_rate": 1.551526641061186e-05, + "loss": 1.0093, + "step": 10464 + }, + { + "epoch": 0.33, + "learning_rate": 1.5514402357823607e-05, + "loss": 0.9238, + "step": 10465 + }, + { + "epoch": 0.33, + "learning_rate": 1.5513538245871917e-05, + "loss": 1.1562, + "step": 10466 + }, + { + "epoch": 0.33, + "learning_rate": 1.551267407476607e-05, + "loss": 1.1265, + "step": 10467 + }, + { + "epoch": 0.33, + "learning_rate": 1.551180984451533e-05, + "loss": 0.5449, + "step": 10468 + }, + { + "epoch": 0.33, + "learning_rate": 1.5510945555128965e-05, + "loss": 1.0688, + "step": 10469 + }, + { + "epoch": 0.33, + "learning_rate": 1.5510081206616256e-05, + "loss": 1.1274, + "step": 10470 + }, + { + "epoch": 0.33, + "learning_rate": 1.5509216798986478e-05, + "loss": 1.0684, + "step": 10471 + }, + { + "epoch": 0.33, + "learning_rate": 1.5508352332248895e-05, + "loss": 0.9922, + "step": 10472 + }, + { + "epoch": 0.33, + "learning_rate": 1.5507487806412792e-05, + "loss": 0.9575, + "step": 10473 + }, + { + "epoch": 0.33, + "learning_rate": 1.5506623221487438e-05, + "loss": 1.1553, + "step": 10474 + }, + { + "epoch": 0.33, + "learning_rate": 1.550575857748211e-05, + "loss": 1.0342, + "step": 10475 + }, + { + "epoch": 0.34, + "learning_rate": 1.550489387440609e-05, + "loss": 1.0713, + "step": 10476 + }, + { + "epoch": 0.34, + "learning_rate": 1.550402911226865e-05, + "loss": 1.0083, + "step": 10477 + }, + { + "epoch": 0.34, + "learning_rate": 1.5503164291079067e-05, + "loss": 0.9888, + "step": 10478 + }, + { + "epoch": 0.34, + "learning_rate": 1.5502299410846626e-05, + "loss": 0.9395, + "step": 10479 + }, + { + "epoch": 0.34, + "learning_rate": 1.55014344715806e-05, + "loss": 0.9946, + "step": 10480 + }, + { + "epoch": 0.34, + "learning_rate": 1.5500569473290268e-05, + "loss": 1.0449, + "step": 10481 + }, + { + "epoch": 0.34, + "learning_rate": 1.5499704415984912e-05, + "loss": 1.0581, + "step": 10482 + }, + { + "epoch": 0.34, + "learning_rate": 1.549883929967382e-05, + "loss": 0.979, + "step": 10483 + }, + { + "epoch": 0.34, + "learning_rate": 1.5497974124366265e-05, + "loss": 0.8462, + "step": 10484 + }, + { + "epoch": 0.34, + "learning_rate": 1.5497108890071535e-05, + "loss": 0.9749, + "step": 10485 + }, + { + "epoch": 0.34, + "learning_rate": 1.549624359679891e-05, + "loss": 1.0737, + "step": 10486 + }, + { + "epoch": 0.34, + "learning_rate": 1.5495378244557674e-05, + "loss": 1.1328, + "step": 10487 + }, + { + "epoch": 0.34, + "learning_rate": 1.549451283335711e-05, + "loss": 1.0791, + "step": 10488 + }, + { + "epoch": 0.34, + "learning_rate": 1.5493647363206503e-05, + "loss": 1.0908, + "step": 10489 + }, + { + "epoch": 0.34, + "learning_rate": 1.5492781834115142e-05, + "loss": 1.1089, + "step": 10490 + }, + { + "epoch": 0.34, + "learning_rate": 1.549191624609231e-05, + "loss": 1.1162, + "step": 10491 + }, + { + "epoch": 0.34, + "learning_rate": 1.5491050599147296e-05, + "loss": 1.0273, + "step": 10492 + }, + { + "epoch": 0.34, + "learning_rate": 1.5490184893289385e-05, + "loss": 1.0996, + "step": 10493 + }, + { + "epoch": 0.34, + "learning_rate": 1.5489319128527868e-05, + "loss": 1.0376, + "step": 10494 + }, + { + "epoch": 0.34, + "learning_rate": 1.548845330487203e-05, + "loss": 1.0527, + "step": 10495 + }, + { + "epoch": 0.34, + "learning_rate": 1.5487587422331165e-05, + "loss": 1.0962, + "step": 10496 + }, + { + "epoch": 0.34, + "learning_rate": 1.5486721480914556e-05, + "loss": 1.145, + "step": 10497 + }, + { + "epoch": 0.34, + "learning_rate": 1.5485855480631503e-05, + "loss": 1.0269, + "step": 10498 + }, + { + "epoch": 0.34, + "learning_rate": 1.548498942149129e-05, + "loss": 1.1504, + "step": 10499 + }, + { + "epoch": 0.34, + "learning_rate": 1.548412330350321e-05, + "loss": 0.9224, + "step": 10500 + }, + { + "epoch": 0.34, + "learning_rate": 1.5483257126676556e-05, + "loss": 0.9878, + "step": 10501 + }, + { + "epoch": 0.34, + "learning_rate": 1.5482390891020624e-05, + "loss": 1.1123, + "step": 10502 + }, + { + "epoch": 0.34, + "learning_rate": 1.5481524596544705e-05, + "loss": 0.9932, + "step": 10503 + }, + { + "epoch": 0.34, + "learning_rate": 1.5480658243258092e-05, + "loss": 0.5486, + "step": 10504 + }, + { + "epoch": 0.34, + "learning_rate": 1.5479791831170083e-05, + "loss": 1.0483, + "step": 10505 + }, + { + "epoch": 0.34, + "learning_rate": 1.5478925360289974e-05, + "loss": 1.0532, + "step": 10506 + }, + { + "epoch": 0.34, + "learning_rate": 1.5478058830627056e-05, + "loss": 0.9365, + "step": 10507 + }, + { + "epoch": 0.34, + "learning_rate": 1.5477192242190633e-05, + "loss": 0.9795, + "step": 10508 + }, + { + "epoch": 0.34, + "learning_rate": 1.5476325594989998e-05, + "loss": 1.0317, + "step": 10509 + }, + { + "epoch": 0.34, + "learning_rate": 1.5475458889034448e-05, + "loss": 1.1182, + "step": 10510 + }, + { + "epoch": 0.34, + "learning_rate": 1.5474592124333284e-05, + "loss": 0.9902, + "step": 10511 + }, + { + "epoch": 0.34, + "learning_rate": 1.547372530089581e-05, + "loss": 1.085, + "step": 10512 + }, + { + "epoch": 0.34, + "learning_rate": 1.5472858418731318e-05, + "loss": 0.5339, + "step": 10513 + }, + { + "epoch": 0.34, + "learning_rate": 1.5471991477849113e-05, + "loss": 1.0337, + "step": 10514 + }, + { + "epoch": 0.34, + "learning_rate": 1.5471124478258496e-05, + "loss": 1.1108, + "step": 10515 + }, + { + "epoch": 0.34, + "learning_rate": 1.5470257419968767e-05, + "loss": 1.0342, + "step": 10516 + }, + { + "epoch": 0.34, + "learning_rate": 1.5469390302989232e-05, + "loss": 0.9722, + "step": 10517 + }, + { + "epoch": 0.34, + "learning_rate": 1.546852312732919e-05, + "loss": 0.8955, + "step": 10518 + }, + { + "epoch": 0.34, + "learning_rate": 1.5467655892997947e-05, + "loss": 1.0186, + "step": 10519 + }, + { + "epoch": 0.34, + "learning_rate": 1.546678860000481e-05, + "loss": 0.9888, + "step": 10520 + }, + { + "epoch": 0.34, + "learning_rate": 1.546592124835908e-05, + "loss": 0.8916, + "step": 10521 + }, + { + "epoch": 0.34, + "learning_rate": 1.5465053838070065e-05, + "loss": 1.1309, + "step": 10522 + }, + { + "epoch": 0.34, + "learning_rate": 1.546418636914707e-05, + "loss": 1.0127, + "step": 10523 + }, + { + "epoch": 0.34, + "learning_rate": 1.5463318841599408e-05, + "loss": 0.9595, + "step": 10524 + }, + { + "epoch": 0.34, + "learning_rate": 1.5462451255436374e-05, + "loss": 0.9399, + "step": 10525 + }, + { + "epoch": 0.34, + "learning_rate": 1.5461583610667287e-05, + "loss": 1.022, + "step": 10526 + }, + { + "epoch": 0.34, + "learning_rate": 1.5460715907301453e-05, + "loss": 1.0308, + "step": 10527 + }, + { + "epoch": 0.34, + "learning_rate": 1.545984814534818e-05, + "loss": 1.0591, + "step": 10528 + }, + { + "epoch": 0.34, + "learning_rate": 1.5458980324816777e-05, + "loss": 1.1567, + "step": 10529 + }, + { + "epoch": 0.34, + "learning_rate": 1.545811244571656e-05, + "loss": 1.0498, + "step": 10530 + }, + { + "epoch": 0.34, + "learning_rate": 1.5457244508056835e-05, + "loss": 1.0371, + "step": 10531 + }, + { + "epoch": 0.34, + "learning_rate": 1.5456376511846917e-05, + "loss": 0.9185, + "step": 10532 + }, + { + "epoch": 0.34, + "learning_rate": 1.545550845709612e-05, + "loss": 1.0332, + "step": 10533 + }, + { + "epoch": 0.34, + "learning_rate": 1.5454640343813752e-05, + "loss": 1.1118, + "step": 10534 + }, + { + "epoch": 0.34, + "learning_rate": 1.545377217200913e-05, + "loss": 1.0396, + "step": 10535 + }, + { + "epoch": 0.34, + "learning_rate": 1.545290394169157e-05, + "loss": 1.0034, + "step": 10536 + }, + { + "epoch": 0.34, + "learning_rate": 1.5452035652870386e-05, + "loss": 1.0771, + "step": 10537 + }, + { + "epoch": 0.34, + "learning_rate": 1.5451167305554893e-05, + "loss": 0.489, + "step": 10538 + }, + { + "epoch": 0.34, + "learning_rate": 1.5450298899754407e-05, + "loss": 1.0488, + "step": 10539 + }, + { + "epoch": 0.34, + "learning_rate": 1.5449430435478246e-05, + "loss": 1.2256, + "step": 10540 + }, + { + "epoch": 0.34, + "learning_rate": 1.5448561912735727e-05, + "loss": 0.9956, + "step": 10541 + }, + { + "epoch": 0.34, + "learning_rate": 1.544769333153617e-05, + "loss": 1.0791, + "step": 10542 + }, + { + "epoch": 0.34, + "learning_rate": 1.5446824691888897e-05, + "loss": 1.0645, + "step": 10543 + }, + { + "epoch": 0.34, + "learning_rate": 1.5445955993803216e-05, + "loss": 1.0991, + "step": 10544 + }, + { + "epoch": 0.34, + "learning_rate": 1.5445087237288457e-05, + "loss": 0.9209, + "step": 10545 + }, + { + "epoch": 0.34, + "learning_rate": 1.5444218422353937e-05, + "loss": 1.0601, + "step": 10546 + }, + { + "epoch": 0.34, + "learning_rate": 1.544334954900898e-05, + "loss": 0.9517, + "step": 10547 + }, + { + "epoch": 0.34, + "learning_rate": 1.544248061726291e-05, + "loss": 0.9307, + "step": 10548 + }, + { + "epoch": 0.34, + "learning_rate": 1.5441611627125045e-05, + "loss": 1.0493, + "step": 10549 + }, + { + "epoch": 0.34, + "learning_rate": 1.5440742578604707e-05, + "loss": 0.9756, + "step": 10550 + }, + { + "epoch": 0.34, + "learning_rate": 1.5439873471711222e-05, + "loss": 0.9873, + "step": 10551 + }, + { + "epoch": 0.34, + "learning_rate": 1.5439004306453918e-05, + "loss": 1.0439, + "step": 10552 + }, + { + "epoch": 0.34, + "learning_rate": 1.5438135082842116e-05, + "loss": 1.1021, + "step": 10553 + }, + { + "epoch": 0.34, + "learning_rate": 1.5437265800885145e-05, + "loss": 1.0449, + "step": 10554 + }, + { + "epoch": 0.34, + "learning_rate": 1.5436396460592328e-05, + "loss": 1.0679, + "step": 10555 + }, + { + "epoch": 0.34, + "learning_rate": 1.5435527061972995e-05, + "loss": 0.9531, + "step": 10556 + }, + { + "epoch": 0.34, + "learning_rate": 1.543465760503647e-05, + "loss": 1.0649, + "step": 10557 + }, + { + "epoch": 0.34, + "learning_rate": 1.5433788089792087e-05, + "loss": 0.5212, + "step": 10558 + }, + { + "epoch": 0.34, + "learning_rate": 1.5432918516249168e-05, + "loss": 0.8057, + "step": 10559 + }, + { + "epoch": 0.34, + "learning_rate": 1.5432048884417047e-05, + "loss": 1.0581, + "step": 10560 + }, + { + "epoch": 0.34, + "learning_rate": 1.5431179194305053e-05, + "loss": 1.0112, + "step": 10561 + }, + { + "epoch": 0.34, + "learning_rate": 1.543030944592252e-05, + "loss": 1.0713, + "step": 10562 + }, + { + "epoch": 0.34, + "learning_rate": 1.5429439639278772e-05, + "loss": 1.1016, + "step": 10563 + }, + { + "epoch": 0.34, + "learning_rate": 1.5428569774383153e-05, + "loss": 0.9551, + "step": 10564 + }, + { + "epoch": 0.34, + "learning_rate": 1.5427699851244983e-05, + "loss": 1.0249, + "step": 10565 + }, + { + "epoch": 0.34, + "learning_rate": 1.54268298698736e-05, + "loss": 1.0225, + "step": 10566 + }, + { + "epoch": 0.34, + "learning_rate": 1.5425959830278343e-05, + "loss": 1.0771, + "step": 10567 + }, + { + "epoch": 0.34, + "learning_rate": 1.542508973246854e-05, + "loss": 0.9976, + "step": 10568 + }, + { + "epoch": 0.34, + "learning_rate": 1.5424219576453526e-05, + "loss": 1.0674, + "step": 10569 + }, + { + "epoch": 0.34, + "learning_rate": 1.542334936224264e-05, + "loss": 1.0151, + "step": 10570 + }, + { + "epoch": 0.34, + "learning_rate": 1.542247908984522e-05, + "loss": 1.0425, + "step": 10571 + }, + { + "epoch": 0.34, + "learning_rate": 1.54216087592706e-05, + "loss": 1.0015, + "step": 10572 + }, + { + "epoch": 0.34, + "learning_rate": 1.542073837052812e-05, + "loss": 1.0107, + "step": 10573 + }, + { + "epoch": 0.34, + "learning_rate": 1.5419867923627112e-05, + "loss": 1.1206, + "step": 10574 + }, + { + "epoch": 0.34, + "learning_rate": 1.541899741857692e-05, + "loss": 1.0669, + "step": 10575 + }, + { + "epoch": 0.34, + "learning_rate": 1.5418126855386886e-05, + "loss": 1.1987, + "step": 10576 + }, + { + "epoch": 0.34, + "learning_rate": 1.541725623406635e-05, + "loss": 0.9443, + "step": 10577 + }, + { + "epoch": 0.34, + "learning_rate": 1.5416385554624644e-05, + "loss": 1.1304, + "step": 10578 + }, + { + "epoch": 0.34, + "learning_rate": 1.541551481707112e-05, + "loss": 0.9419, + "step": 10579 + }, + { + "epoch": 0.34, + "learning_rate": 1.541464402141511e-05, + "loss": 1.0605, + "step": 10580 + }, + { + "epoch": 0.34, + "learning_rate": 1.5413773167665968e-05, + "loss": 0.9585, + "step": 10581 + }, + { + "epoch": 0.34, + "learning_rate": 1.541290225583303e-05, + "loss": 0.9985, + "step": 10582 + }, + { + "epoch": 0.34, + "learning_rate": 1.541203128592564e-05, + "loss": 1.0215, + "step": 10583 + }, + { + "epoch": 0.34, + "learning_rate": 1.5411160257953143e-05, + "loss": 1.0269, + "step": 10584 + }, + { + "epoch": 0.34, + "learning_rate": 1.541028917192489e-05, + "loss": 0.9663, + "step": 10585 + }, + { + "epoch": 0.34, + "learning_rate": 1.5409418027850217e-05, + "loss": 1.0371, + "step": 10586 + }, + { + "epoch": 0.34, + "learning_rate": 1.5408546825738477e-05, + "loss": 0.9863, + "step": 10587 + }, + { + "epoch": 0.34, + "learning_rate": 1.540767556559902e-05, + "loss": 1.0132, + "step": 10588 + }, + { + "epoch": 0.34, + "learning_rate": 1.5406804247441184e-05, + "loss": 1.0928, + "step": 10589 + }, + { + "epoch": 0.34, + "learning_rate": 1.5405932871274324e-05, + "loss": 1.0107, + "step": 10590 + }, + { + "epoch": 0.34, + "learning_rate": 1.5405061437107787e-05, + "loss": 1.064, + "step": 10591 + }, + { + "epoch": 0.34, + "learning_rate": 1.5404189944950922e-05, + "loss": 1.0649, + "step": 10592 + }, + { + "epoch": 0.34, + "learning_rate": 1.540331839481308e-05, + "loss": 1.0376, + "step": 10593 + }, + { + "epoch": 0.34, + "learning_rate": 1.5402446786703614e-05, + "loss": 1.0762, + "step": 10594 + }, + { + "epoch": 0.34, + "learning_rate": 1.5401575120631868e-05, + "loss": 0.9829, + "step": 10595 + }, + { + "epoch": 0.34, + "learning_rate": 1.5400703396607207e-05, + "loss": 0.9844, + "step": 10596 + }, + { + "epoch": 0.34, + "learning_rate": 1.5399831614638968e-05, + "loss": 1.043, + "step": 10597 + }, + { + "epoch": 0.34, + "learning_rate": 1.5398959774736514e-05, + "loss": 0.979, + "step": 10598 + }, + { + "epoch": 0.34, + "learning_rate": 1.53980878769092e-05, + "loss": 1.0645, + "step": 10599 + }, + { + "epoch": 0.34, + "learning_rate": 1.539721592116638e-05, + "loss": 1.1152, + "step": 10600 + }, + { + "epoch": 0.34, + "learning_rate": 1.53963439075174e-05, + "loss": 1.1211, + "step": 10601 + }, + { + "epoch": 0.34, + "learning_rate": 1.539547183597162e-05, + "loss": 1.0278, + "step": 10602 + }, + { + "epoch": 0.34, + "learning_rate": 1.5394599706538405e-05, + "loss": 0.5659, + "step": 10603 + }, + { + "epoch": 0.34, + "learning_rate": 1.5393727519227106e-05, + "loss": 1.0176, + "step": 10604 + }, + { + "epoch": 0.34, + "learning_rate": 1.5392855274047074e-05, + "loss": 1.0767, + "step": 10605 + }, + { + "epoch": 0.34, + "learning_rate": 1.539198297100768e-05, + "loss": 1.0244, + "step": 10606 + }, + { + "epoch": 0.34, + "learning_rate": 1.5391110610118276e-05, + "loss": 1.0435, + "step": 10607 + }, + { + "epoch": 0.34, + "learning_rate": 1.5390238191388217e-05, + "loss": 1.0005, + "step": 10608 + }, + { + "epoch": 0.34, + "learning_rate": 1.5389365714826872e-05, + "loss": 1.0713, + "step": 10609 + }, + { + "epoch": 0.34, + "learning_rate": 1.5388493180443592e-05, + "loss": 1.0625, + "step": 10610 + }, + { + "epoch": 0.34, + "learning_rate": 1.5387620588247752e-05, + "loss": 0.9927, + "step": 10611 + }, + { + "epoch": 0.34, + "learning_rate": 1.5386747938248697e-05, + "loss": 1.0947, + "step": 10612 + }, + { + "epoch": 0.34, + "learning_rate": 1.5385875230455804e-05, + "loss": 0.9048, + "step": 10613 + }, + { + "epoch": 0.34, + "learning_rate": 1.538500246487843e-05, + "loss": 1.0005, + "step": 10614 + }, + { + "epoch": 0.34, + "learning_rate": 1.538412964152594e-05, + "loss": 1.0835, + "step": 10615 + }, + { + "epoch": 0.34, + "learning_rate": 1.5383256760407693e-05, + "loss": 1.0933, + "step": 10616 + }, + { + "epoch": 0.34, + "learning_rate": 1.538238382153306e-05, + "loss": 0.5652, + "step": 10617 + }, + { + "epoch": 0.34, + "learning_rate": 1.538151082491141e-05, + "loss": 1.0107, + "step": 10618 + }, + { + "epoch": 0.34, + "learning_rate": 1.5380637770552104e-05, + "loss": 0.9849, + "step": 10619 + }, + { + "epoch": 0.34, + "learning_rate": 1.537976465846451e-05, + "loss": 1.1055, + "step": 10620 + }, + { + "epoch": 0.34, + "learning_rate": 1.537889148865799e-05, + "loss": 1.0396, + "step": 10621 + }, + { + "epoch": 0.34, + "learning_rate": 1.537801826114192e-05, + "loss": 0.8813, + "step": 10622 + }, + { + "epoch": 0.34, + "learning_rate": 1.5377144975925672e-05, + "loss": 0.811, + "step": 10623 + }, + { + "epoch": 0.34, + "learning_rate": 1.5376271633018602e-05, + "loss": 0.9468, + "step": 10624 + }, + { + "epoch": 0.34, + "learning_rate": 1.537539823243009e-05, + "loss": 0.9839, + "step": 10625 + }, + { + "epoch": 0.34, + "learning_rate": 1.5374524774169505e-05, + "loss": 0.9741, + "step": 10626 + }, + { + "epoch": 0.34, + "learning_rate": 1.5373651258246218e-05, + "loss": 0.9902, + "step": 10627 + }, + { + "epoch": 0.34, + "learning_rate": 1.5372777684669596e-05, + "loss": 1.0337, + "step": 10628 + }, + { + "epoch": 0.34, + "learning_rate": 1.537190405344902e-05, + "loss": 1.0015, + "step": 10629 + }, + { + "epoch": 0.34, + "learning_rate": 1.537103036459386e-05, + "loss": 0.8916, + "step": 10630 + }, + { + "epoch": 0.34, + "learning_rate": 1.5370156618113484e-05, + "loss": 1.0259, + "step": 10631 + }, + { + "epoch": 0.34, + "learning_rate": 1.5369282814017273e-05, + "loss": 0.9565, + "step": 10632 + }, + { + "epoch": 0.34, + "learning_rate": 1.53684089523146e-05, + "loss": 0.9741, + "step": 10633 + }, + { + "epoch": 0.34, + "learning_rate": 1.536753503301484e-05, + "loss": 1.0703, + "step": 10634 + }, + { + "epoch": 0.34, + "learning_rate": 1.536666105612737e-05, + "loss": 1.0112, + "step": 10635 + }, + { + "epoch": 0.34, + "learning_rate": 1.5365787021661565e-05, + "loss": 1.0938, + "step": 10636 + }, + { + "epoch": 0.34, + "learning_rate": 1.5364912929626807e-05, + "loss": 1.0562, + "step": 10637 + }, + { + "epoch": 0.34, + "learning_rate": 1.536403878003247e-05, + "loss": 0.9302, + "step": 10638 + }, + { + "epoch": 0.34, + "learning_rate": 1.536316457288793e-05, + "loss": 1.019, + "step": 10639 + }, + { + "epoch": 0.34, + "learning_rate": 1.5362290308202573e-05, + "loss": 1.1611, + "step": 10640 + }, + { + "epoch": 0.34, + "learning_rate": 1.5361415985985778e-05, + "loss": 0.96, + "step": 10641 + }, + { + "epoch": 0.34, + "learning_rate": 1.5360541606246917e-05, + "loss": 0.8633, + "step": 10642 + }, + { + "epoch": 0.34, + "learning_rate": 1.535966716899538e-05, + "loss": 1.0801, + "step": 10643 + }, + { + "epoch": 0.34, + "learning_rate": 1.5358792674240548e-05, + "loss": 1.062, + "step": 10644 + }, + { + "epoch": 0.34, + "learning_rate": 1.53579181219918e-05, + "loss": 1.084, + "step": 10645 + }, + { + "epoch": 0.34, + "learning_rate": 1.535704351225852e-05, + "loss": 1.0244, + "step": 10646 + }, + { + "epoch": 0.34, + "learning_rate": 1.5356168845050094e-05, + "loss": 1.0508, + "step": 10647 + }, + { + "epoch": 0.34, + "learning_rate": 1.53552941203759e-05, + "loss": 1.0415, + "step": 10648 + }, + { + "epoch": 0.34, + "learning_rate": 1.535441933824533e-05, + "loss": 1.0493, + "step": 10649 + }, + { + "epoch": 0.34, + "learning_rate": 1.5353544498667766e-05, + "loss": 0.9609, + "step": 10650 + }, + { + "epoch": 0.34, + "learning_rate": 1.5352669601652593e-05, + "loss": 1.0928, + "step": 10651 + }, + { + "epoch": 0.34, + "learning_rate": 1.5351794647209203e-05, + "loss": 1.1304, + "step": 10652 + }, + { + "epoch": 0.34, + "learning_rate": 1.535091963534698e-05, + "loss": 1.0366, + "step": 10653 + }, + { + "epoch": 0.34, + "learning_rate": 1.5350044566075308e-05, + "loss": 1.1294, + "step": 10654 + }, + { + "epoch": 0.34, + "learning_rate": 1.5349169439403577e-05, + "loss": 1.0181, + "step": 10655 + }, + { + "epoch": 0.34, + "learning_rate": 1.534829425534118e-05, + "loss": 1.2021, + "step": 10656 + }, + { + "epoch": 0.34, + "learning_rate": 1.5347419013897507e-05, + "loss": 1.0513, + "step": 10657 + }, + { + "epoch": 0.34, + "learning_rate": 1.5346543715081942e-05, + "loss": 0.8984, + "step": 10658 + }, + { + "epoch": 0.34, + "learning_rate": 1.5345668358903886e-05, + "loss": 1.1519, + "step": 10659 + }, + { + "epoch": 0.34, + "learning_rate": 1.534479294537272e-05, + "loss": 1.0088, + "step": 10660 + }, + { + "epoch": 0.34, + "learning_rate": 1.5343917474497842e-05, + "loss": 0.5674, + "step": 10661 + }, + { + "epoch": 0.34, + "learning_rate": 1.5343041946288645e-05, + "loss": 1.0366, + "step": 10662 + }, + { + "epoch": 0.34, + "learning_rate": 1.534216636075452e-05, + "loss": 0.9976, + "step": 10663 + }, + { + "epoch": 0.34, + "learning_rate": 1.5341290717904866e-05, + "loss": 1.0479, + "step": 10664 + }, + { + "epoch": 0.34, + "learning_rate": 1.534041501774907e-05, + "loss": 1.04, + "step": 10665 + }, + { + "epoch": 0.34, + "learning_rate": 1.5339539260296533e-05, + "loss": 1.0488, + "step": 10666 + }, + { + "epoch": 0.34, + "learning_rate": 1.533866344555665e-05, + "loss": 1.061, + "step": 10667 + }, + { + "epoch": 0.34, + "learning_rate": 1.5337787573538818e-05, + "loss": 1.0522, + "step": 10668 + }, + { + "epoch": 0.34, + "learning_rate": 1.533691164425243e-05, + "loss": 1.0132, + "step": 10669 + }, + { + "epoch": 0.34, + "learning_rate": 1.5336035657706883e-05, + "loss": 1.0581, + "step": 10670 + }, + { + "epoch": 0.34, + "learning_rate": 1.5335159613911587e-05, + "loss": 1.104, + "step": 10671 + }, + { + "epoch": 0.34, + "learning_rate": 1.5334283512875925e-05, + "loss": 1.0049, + "step": 10672 + }, + { + "epoch": 0.34, + "learning_rate": 1.533340735460931e-05, + "loss": 1.0366, + "step": 10673 + }, + { + "epoch": 0.34, + "learning_rate": 1.5332531139121137e-05, + "loss": 0.96, + "step": 10674 + }, + { + "epoch": 0.34, + "learning_rate": 1.5331654866420808e-05, + "loss": 1.0825, + "step": 10675 + }, + { + "epoch": 0.34, + "learning_rate": 1.5330778536517718e-05, + "loss": 1.0449, + "step": 10676 + }, + { + "epoch": 0.34, + "learning_rate": 1.5329902149421277e-05, + "loss": 0.5376, + "step": 10677 + }, + { + "epoch": 0.34, + "learning_rate": 1.532902570514088e-05, + "loss": 1.0684, + "step": 10678 + }, + { + "epoch": 0.34, + "learning_rate": 1.5328149203685946e-05, + "loss": 1.0928, + "step": 10679 + }, + { + "epoch": 0.34, + "learning_rate": 1.5327272645065857e-05, + "loss": 1.0132, + "step": 10680 + }, + { + "epoch": 0.34, + "learning_rate": 1.5326396029290035e-05, + "loss": 1.0581, + "step": 10681 + }, + { + "epoch": 0.34, + "learning_rate": 1.5325519356367875e-05, + "loss": 1.083, + "step": 10682 + }, + { + "epoch": 0.34, + "learning_rate": 1.532464262630879e-05, + "loss": 1.0269, + "step": 10683 + }, + { + "epoch": 0.34, + "learning_rate": 1.532376583912218e-05, + "loss": 0.8667, + "step": 10684 + }, + { + "epoch": 0.34, + "learning_rate": 1.532288899481745e-05, + "loss": 0.9429, + "step": 10685 + }, + { + "epoch": 0.34, + "learning_rate": 1.5322012093404023e-05, + "loss": 1.1338, + "step": 10686 + }, + { + "epoch": 0.34, + "learning_rate": 1.5321135134891288e-05, + "loss": 1.0225, + "step": 10687 + }, + { + "epoch": 0.34, + "learning_rate": 1.5320258119288664e-05, + "loss": 1.0615, + "step": 10688 + }, + { + "epoch": 0.34, + "learning_rate": 1.531938104660556e-05, + "loss": 0.9717, + "step": 10689 + }, + { + "epoch": 0.34, + "learning_rate": 1.531850391685138e-05, + "loss": 1.1562, + "step": 10690 + }, + { + "epoch": 0.34, + "learning_rate": 1.5317626730035547e-05, + "loss": 0.9961, + "step": 10691 + }, + { + "epoch": 0.34, + "learning_rate": 1.5316749486167458e-05, + "loss": 0.9434, + "step": 10692 + }, + { + "epoch": 0.34, + "learning_rate": 1.5315872185256534e-05, + "loss": 0.98, + "step": 10693 + }, + { + "epoch": 0.34, + "learning_rate": 1.531499482731219e-05, + "loss": 0.9673, + "step": 10694 + }, + { + "epoch": 0.34, + "learning_rate": 1.5314117412343824e-05, + "loss": 1.0186, + "step": 10695 + }, + { + "epoch": 0.34, + "learning_rate": 1.5313239940360865e-05, + "loss": 1.0142, + "step": 10696 + }, + { + "epoch": 0.34, + "learning_rate": 1.531236241137272e-05, + "loss": 1.002, + "step": 10697 + }, + { + "epoch": 0.34, + "learning_rate": 1.5311484825388806e-05, + "loss": 1.0298, + "step": 10698 + }, + { + "epoch": 0.34, + "learning_rate": 1.5310607182418538e-05, + "loss": 1.0449, + "step": 10699 + }, + { + "epoch": 0.34, + "learning_rate": 1.530972948247133e-05, + "loss": 1.1289, + "step": 10700 + }, + { + "epoch": 0.34, + "learning_rate": 1.5308851725556606e-05, + "loss": 1.105, + "step": 10701 + }, + { + "epoch": 0.34, + "learning_rate": 1.530797391168378e-05, + "loss": 1.1948, + "step": 10702 + }, + { + "epoch": 0.34, + "learning_rate": 1.5307096040862262e-05, + "loss": 0.9551, + "step": 10703 + }, + { + "epoch": 0.34, + "learning_rate": 1.5306218113101482e-05, + "loss": 0.9609, + "step": 10704 + }, + { + "epoch": 0.34, + "learning_rate": 1.5305340128410856e-05, + "loss": 1.001, + "step": 10705 + }, + { + "epoch": 0.34, + "learning_rate": 1.5304462086799798e-05, + "loss": 0.999, + "step": 10706 + }, + { + "epoch": 0.34, + "learning_rate": 1.530358398827773e-05, + "loss": 1.0537, + "step": 10707 + }, + { + "epoch": 0.34, + "learning_rate": 1.530270583285408e-05, + "loss": 1.0752, + "step": 10708 + }, + { + "epoch": 0.34, + "learning_rate": 1.5301827620538262e-05, + "loss": 1.0703, + "step": 10709 + }, + { + "epoch": 0.34, + "learning_rate": 1.5300949351339704e-05, + "loss": 0.9849, + "step": 10710 + }, + { + "epoch": 0.34, + "learning_rate": 1.5300071025267826e-05, + "loss": 1.1221, + "step": 10711 + }, + { + "epoch": 0.34, + "learning_rate": 1.529919264233205e-05, + "loss": 1.0874, + "step": 10712 + }, + { + "epoch": 0.34, + "learning_rate": 1.5298314202541807e-05, + "loss": 1.1118, + "step": 10713 + }, + { + "epoch": 0.34, + "learning_rate": 1.529743570590651e-05, + "loss": 0.9341, + "step": 10714 + }, + { + "epoch": 0.34, + "learning_rate": 1.529655715243559e-05, + "loss": 1.0532, + "step": 10715 + }, + { + "epoch": 0.34, + "learning_rate": 1.5295678542138477e-05, + "loss": 0.9634, + "step": 10716 + }, + { + "epoch": 0.34, + "learning_rate": 1.5294799875024596e-05, + "loss": 1.0366, + "step": 10717 + }, + { + "epoch": 0.34, + "learning_rate": 1.529392115110337e-05, + "loss": 0.8452, + "step": 10718 + }, + { + "epoch": 0.34, + "learning_rate": 1.529304237038423e-05, + "loss": 1.0591, + "step": 10719 + }, + { + "epoch": 0.34, + "learning_rate": 1.5292163532876604e-05, + "loss": 1.1318, + "step": 10720 + }, + { + "epoch": 0.34, + "learning_rate": 1.529128463858992e-05, + "loss": 1.0576, + "step": 10721 + }, + { + "epoch": 0.34, + "learning_rate": 1.5290405687533606e-05, + "loss": 1.0015, + "step": 10722 + }, + { + "epoch": 0.34, + "learning_rate": 1.5289526679717094e-05, + "loss": 1.1572, + "step": 10723 + }, + { + "epoch": 0.34, + "learning_rate": 1.5288647615149817e-05, + "loss": 1.0713, + "step": 10724 + }, + { + "epoch": 0.34, + "learning_rate": 1.5287768493841207e-05, + "loss": 0.8638, + "step": 10725 + }, + { + "epoch": 0.34, + "learning_rate": 1.5286889315800687e-05, + "loss": 1.0386, + "step": 10726 + }, + { + "epoch": 0.34, + "learning_rate": 1.52860100810377e-05, + "loss": 0.9658, + "step": 10727 + }, + { + "epoch": 0.34, + "learning_rate": 1.5285130789561676e-05, + "loss": 1.0049, + "step": 10728 + }, + { + "epoch": 0.34, + "learning_rate": 1.5284251441382046e-05, + "loss": 1.0332, + "step": 10729 + }, + { + "epoch": 0.34, + "learning_rate": 1.5283372036508245e-05, + "loss": 0.9497, + "step": 10730 + }, + { + "epoch": 0.34, + "learning_rate": 1.528249257494971e-05, + "loss": 1.0947, + "step": 10731 + }, + { + "epoch": 0.34, + "learning_rate": 1.5281613056715874e-05, + "loss": 0.9551, + "step": 10732 + }, + { + "epoch": 0.34, + "learning_rate": 1.5280733481816184e-05, + "loss": 1.0098, + "step": 10733 + }, + { + "epoch": 0.34, + "learning_rate": 1.527985385026006e-05, + "loss": 1.0435, + "step": 10734 + }, + { + "epoch": 0.34, + "learning_rate": 1.527897416205695e-05, + "loss": 1.0718, + "step": 10735 + }, + { + "epoch": 0.34, + "learning_rate": 1.527809441721629e-05, + "loss": 1.1455, + "step": 10736 + }, + { + "epoch": 0.34, + "learning_rate": 1.5277214615747516e-05, + "loss": 1.0801, + "step": 10737 + }, + { + "epoch": 0.34, + "learning_rate": 1.5276334757660073e-05, + "loss": 0.9451, + "step": 10738 + }, + { + "epoch": 0.34, + "learning_rate": 1.5275454842963395e-05, + "loss": 1.1084, + "step": 10739 + }, + { + "epoch": 0.34, + "learning_rate": 1.527457487166693e-05, + "loss": 1.0488, + "step": 10740 + }, + { + "epoch": 0.34, + "learning_rate": 1.5273694843780106e-05, + "loss": 0.9907, + "step": 10741 + }, + { + "epoch": 0.34, + "learning_rate": 1.5272814759312378e-05, + "loss": 1.1045, + "step": 10742 + }, + { + "epoch": 0.34, + "learning_rate": 1.5271934618273182e-05, + "loss": 1.1104, + "step": 10743 + }, + { + "epoch": 0.34, + "learning_rate": 1.5271054420671965e-05, + "loss": 1.0024, + "step": 10744 + }, + { + "epoch": 0.34, + "learning_rate": 1.5270174166518167e-05, + "loss": 0.9585, + "step": 10745 + }, + { + "epoch": 0.34, + "learning_rate": 1.5269293855821226e-05, + "loss": 1.0439, + "step": 10746 + }, + { + "epoch": 0.34, + "learning_rate": 1.5268413488590603e-05, + "loss": 1.125, + "step": 10747 + }, + { + "epoch": 0.34, + "learning_rate": 1.526753306483573e-05, + "loss": 1.0264, + "step": 10748 + }, + { + "epoch": 0.34, + "learning_rate": 1.5266652584566056e-05, + "loss": 0.9956, + "step": 10749 + }, + { + "epoch": 0.34, + "learning_rate": 1.526577204779103e-05, + "loss": 0.9741, + "step": 10750 + }, + { + "epoch": 0.34, + "learning_rate": 1.5264891454520098e-05, + "loss": 1.1504, + "step": 10751 + }, + { + "epoch": 0.34, + "learning_rate": 1.526401080476271e-05, + "loss": 0.9451, + "step": 10752 + }, + { + "epoch": 0.34, + "learning_rate": 1.5263130098528307e-05, + "loss": 0.938, + "step": 10753 + }, + { + "epoch": 0.34, + "learning_rate": 1.526224933582635e-05, + "loss": 0.9507, + "step": 10754 + }, + { + "epoch": 0.34, + "learning_rate": 1.5261368516666274e-05, + "loss": 1.001, + "step": 10755 + }, + { + "epoch": 0.34, + "learning_rate": 1.526048764105754e-05, + "loss": 0.936, + "step": 10756 + }, + { + "epoch": 0.34, + "learning_rate": 1.5259606709009594e-05, + "loss": 0.9111, + "step": 10757 + }, + { + "epoch": 0.34, + "learning_rate": 1.5258725720531892e-05, + "loss": 0.9946, + "step": 10758 + }, + { + "epoch": 0.34, + "learning_rate": 1.5257844675633882e-05, + "loss": 1.1167, + "step": 10759 + }, + { + "epoch": 0.34, + "learning_rate": 1.5256963574325017e-05, + "loss": 1.0029, + "step": 10760 + }, + { + "epoch": 0.34, + "learning_rate": 1.525608241661475e-05, + "loss": 0.6077, + "step": 10761 + }, + { + "epoch": 0.34, + "learning_rate": 1.5255201202512537e-05, + "loss": 0.8325, + "step": 10762 + }, + { + "epoch": 0.34, + "learning_rate": 1.5254319932027832e-05, + "loss": 1.1416, + "step": 10763 + }, + { + "epoch": 0.34, + "learning_rate": 1.5253438605170088e-05, + "loss": 1.0703, + "step": 10764 + }, + { + "epoch": 0.34, + "learning_rate": 1.5252557221948762e-05, + "loss": 1.0918, + "step": 10765 + }, + { + "epoch": 0.34, + "learning_rate": 1.5251675782373312e-05, + "loss": 1.0117, + "step": 10766 + }, + { + "epoch": 0.34, + "learning_rate": 1.5250794286453192e-05, + "loss": 0.8999, + "step": 10767 + }, + { + "epoch": 0.34, + "learning_rate": 1.5249912734197861e-05, + "loss": 1.0439, + "step": 10768 + }, + { + "epoch": 0.34, + "learning_rate": 1.5249031125616777e-05, + "loss": 1.1436, + "step": 10769 + }, + { + "epoch": 0.34, + "learning_rate": 1.5248149460719397e-05, + "loss": 1.0444, + "step": 10770 + }, + { + "epoch": 0.34, + "learning_rate": 1.5247267739515183e-05, + "loss": 1.0669, + "step": 10771 + }, + { + "epoch": 0.34, + "learning_rate": 1.5246385962013595e-05, + "loss": 1.0898, + "step": 10772 + }, + { + "epoch": 0.34, + "learning_rate": 1.5245504128224089e-05, + "loss": 1.083, + "step": 10773 + }, + { + "epoch": 0.34, + "learning_rate": 1.524462223815613e-05, + "loss": 1.0312, + "step": 10774 + }, + { + "epoch": 0.34, + "learning_rate": 1.5243740291819178e-05, + "loss": 1.063, + "step": 10775 + }, + { + "epoch": 0.34, + "learning_rate": 1.52428582892227e-05, + "loss": 1.0654, + "step": 10776 + }, + { + "epoch": 0.34, + "learning_rate": 1.524197623037615e-05, + "loss": 0.9985, + "step": 10777 + }, + { + "epoch": 0.34, + "learning_rate": 1.5241094115289001e-05, + "loss": 1.064, + "step": 10778 + }, + { + "epoch": 0.34, + "learning_rate": 1.5240211943970708e-05, + "loss": 1.0181, + "step": 10779 + }, + { + "epoch": 0.34, + "learning_rate": 1.5239329716430744e-05, + "loss": 1.0747, + "step": 10780 + }, + { + "epoch": 0.34, + "learning_rate": 1.5238447432678571e-05, + "loss": 0.9683, + "step": 10781 + }, + { + "epoch": 0.34, + "learning_rate": 1.5237565092723652e-05, + "loss": 0.9937, + "step": 10782 + }, + { + "epoch": 0.34, + "learning_rate": 1.5236682696575456e-05, + "loss": 1.1157, + "step": 10783 + }, + { + "epoch": 0.34, + "learning_rate": 1.5235800244243452e-05, + "loss": 0.8096, + "step": 10784 + }, + { + "epoch": 0.34, + "learning_rate": 1.5234917735737108e-05, + "loss": 1.145, + "step": 10785 + }, + { + "epoch": 0.34, + "learning_rate": 1.5234035171065889e-05, + "loss": 1.0537, + "step": 10786 + }, + { + "epoch": 0.34, + "learning_rate": 1.5233152550239262e-05, + "loss": 0.9932, + "step": 10787 + }, + { + "epoch": 0.35, + "learning_rate": 1.5232269873266705e-05, + "loss": 1.106, + "step": 10788 + }, + { + "epoch": 0.35, + "learning_rate": 1.5231387140157678e-05, + "loss": 0.9087, + "step": 10789 + }, + { + "epoch": 0.35, + "learning_rate": 1.5230504350921659e-05, + "loss": 1.1162, + "step": 10790 + }, + { + "epoch": 0.35, + "learning_rate": 1.5229621505568118e-05, + "loss": 1.0239, + "step": 10791 + }, + { + "epoch": 0.35, + "learning_rate": 1.5228738604106528e-05, + "loss": 0.9746, + "step": 10792 + }, + { + "epoch": 0.35, + "learning_rate": 1.5227855646546355e-05, + "loss": 0.9434, + "step": 10793 + }, + { + "epoch": 0.35, + "learning_rate": 1.5226972632897079e-05, + "loss": 1.1045, + "step": 10794 + }, + { + "epoch": 0.35, + "learning_rate": 1.5226089563168171e-05, + "loss": 0.9985, + "step": 10795 + }, + { + "epoch": 0.35, + "learning_rate": 1.5225206437369109e-05, + "loss": 0.9995, + "step": 10796 + }, + { + "epoch": 0.35, + "learning_rate": 1.5224323255509361e-05, + "loss": 1.0537, + "step": 10797 + }, + { + "epoch": 0.35, + "learning_rate": 1.5223440017598407e-05, + "loss": 1.02, + "step": 10798 + }, + { + "epoch": 0.35, + "learning_rate": 1.5222556723645726e-05, + "loss": 0.551, + "step": 10799 + }, + { + "epoch": 0.35, + "learning_rate": 1.5221673373660788e-05, + "loss": 1.1992, + "step": 10800 + }, + { + "epoch": 0.35, + "learning_rate": 1.5220789967653073e-05, + "loss": 0.9902, + "step": 10801 + }, + { + "epoch": 0.35, + "learning_rate": 1.5219906505632064e-05, + "loss": 0.9897, + "step": 10802 + }, + { + "epoch": 0.35, + "learning_rate": 1.5219022987607233e-05, + "loss": 0.9033, + "step": 10803 + }, + { + "epoch": 0.35, + "learning_rate": 1.521813941358806e-05, + "loss": 1.022, + "step": 10804 + }, + { + "epoch": 0.35, + "learning_rate": 1.5217255783584028e-05, + "loss": 1.083, + "step": 10805 + }, + { + "epoch": 0.35, + "learning_rate": 1.5216372097604615e-05, + "loss": 1.0278, + "step": 10806 + }, + { + "epoch": 0.35, + "learning_rate": 1.5215488355659305e-05, + "loss": 1.085, + "step": 10807 + }, + { + "epoch": 0.35, + "learning_rate": 1.5214604557757573e-05, + "loss": 1.0708, + "step": 10808 + }, + { + "epoch": 0.35, + "learning_rate": 1.5213720703908905e-05, + "loss": 0.9199, + "step": 10809 + }, + { + "epoch": 0.35, + "learning_rate": 1.5212836794122786e-05, + "loss": 0.9761, + "step": 10810 + }, + { + "epoch": 0.35, + "learning_rate": 1.52119528284087e-05, + "loss": 1.0664, + "step": 10811 + }, + { + "epoch": 0.35, + "learning_rate": 1.5211068806776127e-05, + "loss": 1.04, + "step": 10812 + }, + { + "epoch": 0.35, + "learning_rate": 1.5210184729234552e-05, + "loss": 1.0278, + "step": 10813 + }, + { + "epoch": 0.35, + "learning_rate": 1.5209300595793463e-05, + "loss": 1.0811, + "step": 10814 + }, + { + "epoch": 0.35, + "learning_rate": 1.5208416406462341e-05, + "loss": 0.9727, + "step": 10815 + }, + { + "epoch": 0.35, + "learning_rate": 1.5207532161250677e-05, + "loss": 1.0498, + "step": 10816 + }, + { + "epoch": 0.35, + "learning_rate": 1.5206647860167958e-05, + "loss": 1.0781, + "step": 10817 + }, + { + "epoch": 0.35, + "learning_rate": 1.5205763503223665e-05, + "loss": 0.9961, + "step": 10818 + }, + { + "epoch": 0.35, + "learning_rate": 1.5204879090427298e-05, + "loss": 1.0747, + "step": 10819 + }, + { + "epoch": 0.35, + "learning_rate": 1.5203994621788334e-05, + "loss": 0.9419, + "step": 10820 + }, + { + "epoch": 0.35, + "learning_rate": 1.5203110097316267e-05, + "loss": 0.9268, + "step": 10821 + }, + { + "epoch": 0.35, + "learning_rate": 1.520222551702059e-05, + "loss": 0.9971, + "step": 10822 + }, + { + "epoch": 0.35, + "learning_rate": 1.5201340880910788e-05, + "loss": 1.0532, + "step": 10823 + }, + { + "epoch": 0.35, + "learning_rate": 1.5200456188996355e-05, + "loss": 1.0908, + "step": 10824 + }, + { + "epoch": 0.35, + "learning_rate": 1.5199571441286784e-05, + "loss": 0.9556, + "step": 10825 + }, + { + "epoch": 0.35, + "learning_rate": 1.5198686637791566e-05, + "loss": 0.9365, + "step": 10826 + }, + { + "epoch": 0.35, + "learning_rate": 1.5197801778520194e-05, + "loss": 0.9795, + "step": 10827 + }, + { + "epoch": 0.35, + "learning_rate": 1.519691686348216e-05, + "loss": 1.021, + "step": 10828 + }, + { + "epoch": 0.35, + "learning_rate": 1.519603189268696e-05, + "loss": 0.9976, + "step": 10829 + }, + { + "epoch": 0.35, + "learning_rate": 1.5195146866144093e-05, + "loss": 0.8398, + "step": 10830 + }, + { + "epoch": 0.35, + "learning_rate": 1.5194261783863044e-05, + "loss": 1.0972, + "step": 10831 + }, + { + "epoch": 0.35, + "learning_rate": 1.5193376645853317e-05, + "loss": 1.2295, + "step": 10832 + }, + { + "epoch": 0.35, + "learning_rate": 1.5192491452124408e-05, + "loss": 1.0186, + "step": 10833 + }, + { + "epoch": 0.35, + "learning_rate": 1.519160620268581e-05, + "loss": 1.0366, + "step": 10834 + }, + { + "epoch": 0.35, + "learning_rate": 1.5190720897547024e-05, + "loss": 1.105, + "step": 10835 + }, + { + "epoch": 0.35, + "learning_rate": 1.5189835536717547e-05, + "loss": 0.9834, + "step": 10836 + }, + { + "epoch": 0.35, + "learning_rate": 1.5188950120206879e-05, + "loss": 1.0254, + "step": 10837 + }, + { + "epoch": 0.35, + "learning_rate": 1.518806464802452e-05, + "loss": 1.1304, + "step": 10838 + }, + { + "epoch": 0.35, + "learning_rate": 1.5187179120179969e-05, + "loss": 0.8711, + "step": 10839 + }, + { + "epoch": 0.35, + "learning_rate": 1.5186293536682726e-05, + "loss": 1.0981, + "step": 10840 + }, + { + "epoch": 0.35, + "learning_rate": 1.5185407897542296e-05, + "loss": 1.2158, + "step": 10841 + }, + { + "epoch": 0.35, + "learning_rate": 1.5184522202768173e-05, + "loss": 0.9897, + "step": 10842 + }, + { + "epoch": 0.35, + "learning_rate": 1.5183636452369868e-05, + "loss": 0.9727, + "step": 10843 + }, + { + "epoch": 0.35, + "learning_rate": 1.5182750646356881e-05, + "loss": 1.0063, + "step": 10844 + }, + { + "epoch": 0.35, + "learning_rate": 1.5181864784738713e-05, + "loss": 1.043, + "step": 10845 + }, + { + "epoch": 0.35, + "learning_rate": 1.5180978867524874e-05, + "loss": 1.0762, + "step": 10846 + }, + { + "epoch": 0.35, + "learning_rate": 1.5180092894724863e-05, + "loss": 1.0693, + "step": 10847 + }, + { + "epoch": 0.35, + "learning_rate": 1.517920686634819e-05, + "loss": 0.8999, + "step": 10848 + }, + { + "epoch": 0.35, + "learning_rate": 1.517832078240436e-05, + "loss": 0.9756, + "step": 10849 + }, + { + "epoch": 0.35, + "learning_rate": 1.517743464290288e-05, + "loss": 1.1074, + "step": 10850 + }, + { + "epoch": 0.35, + "learning_rate": 1.5176548447853252e-05, + "loss": 1.106, + "step": 10851 + }, + { + "epoch": 0.35, + "learning_rate": 1.517566219726499e-05, + "loss": 1.1128, + "step": 10852 + }, + { + "epoch": 0.35, + "learning_rate": 1.5174775891147605e-05, + "loss": 1.0649, + "step": 10853 + }, + { + "epoch": 0.35, + "learning_rate": 1.5173889529510598e-05, + "loss": 1.1025, + "step": 10854 + }, + { + "epoch": 0.35, + "learning_rate": 1.5173003112363481e-05, + "loss": 0.9214, + "step": 10855 + }, + { + "epoch": 0.35, + "learning_rate": 1.517211663971577e-05, + "loss": 1.0649, + "step": 10856 + }, + { + "epoch": 0.35, + "learning_rate": 1.5171230111576969e-05, + "loss": 1.0938, + "step": 10857 + }, + { + "epoch": 0.35, + "learning_rate": 1.5170343527956592e-05, + "loss": 0.8823, + "step": 10858 + }, + { + "epoch": 0.35, + "learning_rate": 1.5169456888864153e-05, + "loss": 1.0635, + "step": 10859 + }, + { + "epoch": 0.35, + "learning_rate": 1.516857019430916e-05, + "loss": 1.0347, + "step": 10860 + }, + { + "epoch": 0.35, + "learning_rate": 1.5167683444301133e-05, + "loss": 1.166, + "step": 10861 + }, + { + "epoch": 0.35, + "learning_rate": 1.5166796638849582e-05, + "loss": 1.1899, + "step": 10862 + }, + { + "epoch": 0.35, + "learning_rate": 1.5165909777964017e-05, + "loss": 0.9609, + "step": 10863 + }, + { + "epoch": 0.35, + "learning_rate": 1.5165022861653959e-05, + "loss": 0.8845, + "step": 10864 + }, + { + "epoch": 0.35, + "learning_rate": 1.5164135889928924e-05, + "loss": 1.0493, + "step": 10865 + }, + { + "epoch": 0.35, + "learning_rate": 1.5163248862798425e-05, + "loss": 1.063, + "step": 10866 + }, + { + "epoch": 0.35, + "learning_rate": 1.516236178027198e-05, + "loss": 1.0439, + "step": 10867 + }, + { + "epoch": 0.35, + "learning_rate": 1.5161474642359107e-05, + "loss": 0.9634, + "step": 10868 + }, + { + "epoch": 0.35, + "learning_rate": 1.5160587449069321e-05, + "loss": 1.04, + "step": 10869 + }, + { + "epoch": 0.35, + "learning_rate": 1.5159700200412146e-05, + "loss": 1.0703, + "step": 10870 + }, + { + "epoch": 0.35, + "learning_rate": 1.51588128963971e-05, + "loss": 0.8398, + "step": 10871 + }, + { + "epoch": 0.35, + "learning_rate": 1.5157925537033697e-05, + "loss": 0.9629, + "step": 10872 + }, + { + "epoch": 0.35, + "learning_rate": 1.5157038122331463e-05, + "loss": 0.9985, + "step": 10873 + }, + { + "epoch": 0.35, + "learning_rate": 1.5156150652299917e-05, + "loss": 0.8823, + "step": 10874 + }, + { + "epoch": 0.35, + "learning_rate": 1.5155263126948583e-05, + "loss": 1.0454, + "step": 10875 + }, + { + "epoch": 0.35, + "learning_rate": 1.5154375546286979e-05, + "loss": 1.082, + "step": 10876 + }, + { + "epoch": 0.35, + "learning_rate": 1.515348791032463e-05, + "loss": 1.1152, + "step": 10877 + }, + { + "epoch": 0.35, + "learning_rate": 1.5152600219071061e-05, + "loss": 1.0356, + "step": 10878 + }, + { + "epoch": 0.35, + "learning_rate": 1.5151712472535794e-05, + "loss": 0.9766, + "step": 10879 + }, + { + "epoch": 0.35, + "learning_rate": 1.5150824670728353e-05, + "loss": 1.0811, + "step": 10880 + }, + { + "epoch": 0.35, + "learning_rate": 1.5149936813658264e-05, + "loss": 0.9238, + "step": 10881 + }, + { + "epoch": 0.35, + "learning_rate": 1.5149048901335054e-05, + "loss": 1.1611, + "step": 10882 + }, + { + "epoch": 0.35, + "learning_rate": 1.5148160933768246e-05, + "loss": 1.0342, + "step": 10883 + }, + { + "epoch": 0.35, + "learning_rate": 1.5147272910967368e-05, + "loss": 1.042, + "step": 10884 + }, + { + "epoch": 0.35, + "learning_rate": 1.5146384832941952e-05, + "loss": 0.9395, + "step": 10885 + }, + { + "epoch": 0.35, + "learning_rate": 1.5145496699701519e-05, + "loss": 1.0581, + "step": 10886 + }, + { + "epoch": 0.35, + "learning_rate": 1.5144608511255605e-05, + "loss": 0.9932, + "step": 10887 + }, + { + "epoch": 0.35, + "learning_rate": 1.514372026761373e-05, + "loss": 1.0684, + "step": 10888 + }, + { + "epoch": 0.35, + "learning_rate": 1.5142831968785432e-05, + "loss": 1.1519, + "step": 10889 + }, + { + "epoch": 0.35, + "learning_rate": 1.5141943614780244e-05, + "loss": 0.8496, + "step": 10890 + }, + { + "epoch": 0.35, + "learning_rate": 1.5141055205607685e-05, + "loss": 0.9473, + "step": 10891 + }, + { + "epoch": 0.35, + "learning_rate": 1.5140166741277296e-05, + "loss": 1.0835, + "step": 10892 + }, + { + "epoch": 0.35, + "learning_rate": 1.5139278221798605e-05, + "loss": 1.0796, + "step": 10893 + }, + { + "epoch": 0.35, + "learning_rate": 1.5138389647181148e-05, + "loss": 1.0132, + "step": 10894 + }, + { + "epoch": 0.35, + "learning_rate": 1.5137501017434457e-05, + "loss": 1.0239, + "step": 10895 + }, + { + "epoch": 0.35, + "learning_rate": 1.5136612332568064e-05, + "loss": 0.9917, + "step": 10896 + }, + { + "epoch": 0.35, + "learning_rate": 1.5135723592591508e-05, + "loss": 1.1094, + "step": 10897 + }, + { + "epoch": 0.35, + "learning_rate": 1.5134834797514325e-05, + "loss": 0.9897, + "step": 10898 + }, + { + "epoch": 0.35, + "learning_rate": 1.513394594734604e-05, + "loss": 1.1562, + "step": 10899 + }, + { + "epoch": 0.35, + "learning_rate": 1.5133057042096204e-05, + "loss": 0.9272, + "step": 10900 + }, + { + "epoch": 0.35, + "learning_rate": 1.5132168081774343e-05, + "loss": 0.9673, + "step": 10901 + }, + { + "epoch": 0.35, + "learning_rate": 1.5131279066389998e-05, + "loss": 0.9648, + "step": 10902 + }, + { + "epoch": 0.35, + "learning_rate": 1.513038999595271e-05, + "loss": 1.0127, + "step": 10903 + }, + { + "epoch": 0.35, + "learning_rate": 1.5129500870472014e-05, + "loss": 1.064, + "step": 10904 + }, + { + "epoch": 0.35, + "learning_rate": 1.5128611689957451e-05, + "loss": 1.0674, + "step": 10905 + }, + { + "epoch": 0.35, + "learning_rate": 1.5127722454418561e-05, + "loss": 1.0283, + "step": 10906 + }, + { + "epoch": 0.35, + "learning_rate": 1.5126833163864882e-05, + "loss": 1.0884, + "step": 10907 + }, + { + "epoch": 0.35, + "learning_rate": 1.512594381830596e-05, + "loss": 1.0493, + "step": 10908 + }, + { + "epoch": 0.35, + "learning_rate": 1.5125054417751335e-05, + "loss": 0.9854, + "step": 10909 + }, + { + "epoch": 0.35, + "learning_rate": 1.5124164962210544e-05, + "loss": 1.0786, + "step": 10910 + }, + { + "epoch": 0.35, + "learning_rate": 1.5123275451693136e-05, + "loss": 1.0571, + "step": 10911 + }, + { + "epoch": 0.35, + "learning_rate": 1.512238588620865e-05, + "loss": 0.9639, + "step": 10912 + }, + { + "epoch": 0.35, + "learning_rate": 1.512149626576664e-05, + "loss": 1.0176, + "step": 10913 + }, + { + "epoch": 0.35, + "learning_rate": 1.5120606590376637e-05, + "loss": 1.0151, + "step": 10914 + }, + { + "epoch": 0.35, + "learning_rate": 1.5119716860048192e-05, + "loss": 1.0112, + "step": 10915 + }, + { + "epoch": 0.35, + "learning_rate": 1.5118827074790853e-05, + "loss": 0.9648, + "step": 10916 + }, + { + "epoch": 0.35, + "learning_rate": 1.5117937234614167e-05, + "loss": 0.9844, + "step": 10917 + }, + { + "epoch": 0.35, + "learning_rate": 1.5117047339527675e-05, + "loss": 0.9697, + "step": 10918 + }, + { + "epoch": 0.35, + "learning_rate": 1.511615738954093e-05, + "loss": 0.897, + "step": 10919 + }, + { + "epoch": 0.35, + "learning_rate": 1.511526738466348e-05, + "loss": 0.9888, + "step": 10920 + }, + { + "epoch": 0.35, + "learning_rate": 1.5114377324904872e-05, + "loss": 0.5676, + "step": 10921 + }, + { + "epoch": 0.35, + "learning_rate": 1.5113487210274653e-05, + "loss": 1.0112, + "step": 10922 + }, + { + "epoch": 0.35, + "learning_rate": 1.5112597040782377e-05, + "loss": 1.0269, + "step": 10923 + }, + { + "epoch": 0.35, + "learning_rate": 1.5111706816437595e-05, + "loss": 0.9229, + "step": 10924 + }, + { + "epoch": 0.35, + "learning_rate": 1.5110816537249852e-05, + "loss": 1.0078, + "step": 10925 + }, + { + "epoch": 0.35, + "learning_rate": 1.5109926203228708e-05, + "loss": 0.9917, + "step": 10926 + }, + { + "epoch": 0.35, + "learning_rate": 1.5109035814383708e-05, + "loss": 0.9521, + "step": 10927 + }, + { + "epoch": 0.35, + "learning_rate": 1.5108145370724412e-05, + "loss": 1.0249, + "step": 10928 + }, + { + "epoch": 0.35, + "learning_rate": 1.5107254872260366e-05, + "loss": 1.062, + "step": 10929 + }, + { + "epoch": 0.35, + "learning_rate": 1.510636431900113e-05, + "loss": 0.9863, + "step": 10930 + }, + { + "epoch": 0.35, + "learning_rate": 1.5105473710956253e-05, + "loss": 1.2314, + "step": 10931 + }, + { + "epoch": 0.35, + "learning_rate": 1.5104583048135298e-05, + "loss": 1.0156, + "step": 10932 + }, + { + "epoch": 0.35, + "learning_rate": 1.5103692330547812e-05, + "loss": 1.0718, + "step": 10933 + }, + { + "epoch": 0.35, + "learning_rate": 1.5102801558203357e-05, + "loss": 1.1123, + "step": 10934 + }, + { + "epoch": 0.35, + "learning_rate": 1.510191073111149e-05, + "loss": 1.0278, + "step": 10935 + }, + { + "epoch": 0.35, + "learning_rate": 1.5101019849281769e-05, + "loss": 1.0137, + "step": 10936 + }, + { + "epoch": 0.35, + "learning_rate": 1.5100128912723748e-05, + "loss": 1.1353, + "step": 10937 + }, + { + "epoch": 0.35, + "learning_rate": 1.509923792144699e-05, + "loss": 0.8472, + "step": 10938 + }, + { + "epoch": 0.35, + "learning_rate": 1.5098346875461052e-05, + "loss": 1.1841, + "step": 10939 + }, + { + "epoch": 0.35, + "learning_rate": 1.5097455774775491e-05, + "loss": 1.0273, + "step": 10940 + }, + { + "epoch": 0.35, + "learning_rate": 1.5096564619399877e-05, + "loss": 1.0439, + "step": 10941 + }, + { + "epoch": 0.35, + "learning_rate": 1.5095673409343762e-05, + "loss": 1.0474, + "step": 10942 + }, + { + "epoch": 0.35, + "learning_rate": 1.5094782144616712e-05, + "loss": 0.9326, + "step": 10943 + }, + { + "epoch": 0.35, + "learning_rate": 1.5093890825228289e-05, + "loss": 0.9658, + "step": 10944 + }, + { + "epoch": 0.35, + "learning_rate": 1.5092999451188052e-05, + "loss": 0.9795, + "step": 10945 + }, + { + "epoch": 0.35, + "learning_rate": 1.509210802250557e-05, + "loss": 0.9429, + "step": 10946 + }, + { + "epoch": 0.35, + "learning_rate": 1.5091216539190402e-05, + "loss": 0.9707, + "step": 10947 + }, + { + "epoch": 0.35, + "learning_rate": 1.509032500125212e-05, + "loss": 1.0386, + "step": 10948 + }, + { + "epoch": 0.35, + "learning_rate": 1.508943340870028e-05, + "loss": 1.0103, + "step": 10949 + }, + { + "epoch": 0.35, + "learning_rate": 1.5088541761544453e-05, + "loss": 1.1348, + "step": 10950 + }, + { + "epoch": 0.35, + "learning_rate": 1.5087650059794208e-05, + "loss": 1.0327, + "step": 10951 + }, + { + "epoch": 0.35, + "learning_rate": 1.5086758303459106e-05, + "loss": 0.9766, + "step": 10952 + }, + { + "epoch": 0.35, + "learning_rate": 1.5085866492548715e-05, + "loss": 0.896, + "step": 10953 + }, + { + "epoch": 0.35, + "learning_rate": 1.5084974627072608e-05, + "loss": 1.0312, + "step": 10954 + }, + { + "epoch": 0.35, + "learning_rate": 1.508408270704035e-05, + "loss": 0.9883, + "step": 10955 + }, + { + "epoch": 0.35, + "learning_rate": 1.5083190732461512e-05, + "loss": 0.9937, + "step": 10956 + }, + { + "epoch": 0.35, + "learning_rate": 1.5082298703345664e-05, + "loss": 1.0, + "step": 10957 + }, + { + "epoch": 0.35, + "learning_rate": 1.5081406619702375e-05, + "loss": 0.9956, + "step": 10958 + }, + { + "epoch": 0.35, + "learning_rate": 1.5080514481541214e-05, + "loss": 1.0435, + "step": 10959 + }, + { + "epoch": 0.35, + "learning_rate": 1.507962228887176e-05, + "loss": 0.854, + "step": 10960 + }, + { + "epoch": 0.35, + "learning_rate": 1.5078730041703578e-05, + "loss": 1.0342, + "step": 10961 + }, + { + "epoch": 0.35, + "learning_rate": 1.5077837740046245e-05, + "loss": 1.0542, + "step": 10962 + }, + { + "epoch": 0.35, + "learning_rate": 1.5076945383909331e-05, + "loss": 1.0347, + "step": 10963 + }, + { + "epoch": 0.35, + "learning_rate": 1.5076052973302414e-05, + "loss": 0.9292, + "step": 10964 + }, + { + "epoch": 0.35, + "learning_rate": 1.5075160508235062e-05, + "loss": 1.1318, + "step": 10965 + }, + { + "epoch": 0.35, + "learning_rate": 1.507426798871686e-05, + "loss": 0.9966, + "step": 10966 + }, + { + "epoch": 0.35, + "learning_rate": 1.5073375414757379e-05, + "loss": 1.0498, + "step": 10967 + }, + { + "epoch": 0.35, + "learning_rate": 1.5072482786366189e-05, + "loss": 1.0898, + "step": 10968 + }, + { + "epoch": 0.35, + "learning_rate": 1.5071590103552875e-05, + "loss": 1.0371, + "step": 10969 + }, + { + "epoch": 0.35, + "learning_rate": 1.5070697366327016e-05, + "loss": 1.0742, + "step": 10970 + }, + { + "epoch": 0.35, + "learning_rate": 1.5069804574698182e-05, + "loss": 0.9561, + "step": 10971 + }, + { + "epoch": 0.35, + "learning_rate": 1.5068911728675957e-05, + "loss": 1.0073, + "step": 10972 + }, + { + "epoch": 0.35, + "learning_rate": 1.5068018828269922e-05, + "loss": 0.9697, + "step": 10973 + }, + { + "epoch": 0.35, + "learning_rate": 1.506712587348965e-05, + "loss": 1.0039, + "step": 10974 + }, + { + "epoch": 0.35, + "learning_rate": 1.5066232864344726e-05, + "loss": 1.0146, + "step": 10975 + }, + { + "epoch": 0.35, + "learning_rate": 1.506533980084473e-05, + "loss": 0.9805, + "step": 10976 + }, + { + "epoch": 0.35, + "learning_rate": 1.5064446682999245e-05, + "loss": 0.5564, + "step": 10977 + }, + { + "epoch": 0.35, + "learning_rate": 1.5063553510817853e-05, + "loss": 1.1865, + "step": 10978 + }, + { + "epoch": 0.35, + "learning_rate": 1.5062660284310135e-05, + "loss": 0.8804, + "step": 10979 + }, + { + "epoch": 0.35, + "learning_rate": 1.5061767003485674e-05, + "loss": 1.0327, + "step": 10980 + }, + { + "epoch": 0.35, + "learning_rate": 1.5060873668354059e-05, + "loss": 1.0986, + "step": 10981 + }, + { + "epoch": 0.35, + "learning_rate": 1.5059980278924866e-05, + "loss": 1.1323, + "step": 10982 + }, + { + "epoch": 0.35, + "learning_rate": 1.5059086835207688e-05, + "loss": 0.9795, + "step": 10983 + }, + { + "epoch": 0.35, + "learning_rate": 1.5058193337212105e-05, + "loss": 1.019, + "step": 10984 + }, + { + "epoch": 0.35, + "learning_rate": 1.5057299784947707e-05, + "loss": 0.9463, + "step": 10985 + }, + { + "epoch": 0.35, + "learning_rate": 1.505640617842408e-05, + "loss": 0.9521, + "step": 10986 + }, + { + "epoch": 0.35, + "learning_rate": 1.5055512517650812e-05, + "loss": 1.0127, + "step": 10987 + }, + { + "epoch": 0.35, + "learning_rate": 1.5054618802637486e-05, + "loss": 1.0342, + "step": 10988 + }, + { + "epoch": 0.35, + "learning_rate": 1.5053725033393698e-05, + "loss": 0.895, + "step": 10989 + }, + { + "epoch": 0.35, + "learning_rate": 1.5052831209929033e-05, + "loss": 1.0205, + "step": 10990 + }, + { + "epoch": 0.35, + "learning_rate": 1.505193733225308e-05, + "loss": 1.0044, + "step": 10991 + }, + { + "epoch": 0.35, + "learning_rate": 1.5051043400375432e-05, + "loss": 1.0142, + "step": 10992 + }, + { + "epoch": 0.35, + "learning_rate": 1.5050149414305677e-05, + "loss": 0.9116, + "step": 10993 + }, + { + "epoch": 0.35, + "learning_rate": 1.5049255374053409e-05, + "loss": 1.0176, + "step": 10994 + }, + { + "epoch": 0.35, + "learning_rate": 1.5048361279628217e-05, + "loss": 1.1689, + "step": 10995 + }, + { + "epoch": 0.35, + "learning_rate": 1.5047467131039698e-05, + "loss": 1.1387, + "step": 10996 + }, + { + "epoch": 0.35, + "learning_rate": 1.504657292829744e-05, + "loss": 1.0547, + "step": 10997 + }, + { + "epoch": 0.35, + "learning_rate": 1.5045678671411044e-05, + "loss": 1.0322, + "step": 10998 + }, + { + "epoch": 0.35, + "learning_rate": 1.5044784360390096e-05, + "loss": 0.5342, + "step": 10999 + }, + { + "epoch": 0.35, + "learning_rate": 1.5043889995244199e-05, + "loss": 0.9917, + "step": 11000 + }, + { + "epoch": 0.35, + "learning_rate": 1.5042995575982944e-05, + "loss": 0.9224, + "step": 11001 + }, + { + "epoch": 0.35, + "learning_rate": 1.5042101102615925e-05, + "loss": 1.0493, + "step": 11002 + }, + { + "epoch": 0.35, + "learning_rate": 1.5041206575152743e-05, + "loss": 1.0913, + "step": 11003 + }, + { + "epoch": 0.35, + "learning_rate": 1.5040311993602994e-05, + "loss": 1.084, + "step": 11004 + }, + { + "epoch": 0.35, + "learning_rate": 1.5039417357976276e-05, + "loss": 0.9722, + "step": 11005 + }, + { + "epoch": 0.35, + "learning_rate": 1.5038522668282185e-05, + "loss": 0.9736, + "step": 11006 + }, + { + "epoch": 0.35, + "learning_rate": 1.5037627924530326e-05, + "loss": 1.0972, + "step": 11007 + }, + { + "epoch": 0.35, + "learning_rate": 1.5036733126730291e-05, + "loss": 1.1338, + "step": 11008 + }, + { + "epoch": 0.35, + "learning_rate": 1.5035838274891688e-05, + "loss": 1.1196, + "step": 11009 + }, + { + "epoch": 0.35, + "learning_rate": 1.5034943369024108e-05, + "loss": 1.2236, + "step": 11010 + }, + { + "epoch": 0.35, + "learning_rate": 1.5034048409137163e-05, + "loss": 0.9434, + "step": 11011 + }, + { + "epoch": 0.35, + "learning_rate": 1.5033153395240447e-05, + "loss": 1.1328, + "step": 11012 + }, + { + "epoch": 0.35, + "learning_rate": 1.5032258327343567e-05, + "loss": 1.0835, + "step": 11013 + }, + { + "epoch": 0.35, + "learning_rate": 1.5031363205456124e-05, + "loss": 0.8999, + "step": 11014 + }, + { + "epoch": 0.35, + "learning_rate": 1.5030468029587723e-05, + "loss": 1.062, + "step": 11015 + }, + { + "epoch": 0.35, + "learning_rate": 1.5029572799747968e-05, + "loss": 1.106, + "step": 11016 + }, + { + "epoch": 0.35, + "learning_rate": 1.5028677515946462e-05, + "loss": 1.1597, + "step": 11017 + }, + { + "epoch": 0.35, + "learning_rate": 1.502778217819281e-05, + "loss": 0.917, + "step": 11018 + }, + { + "epoch": 0.35, + "learning_rate": 1.5026886786496624e-05, + "loss": 1.0879, + "step": 11019 + }, + { + "epoch": 0.35, + "learning_rate": 1.5025991340867503e-05, + "loss": 1.0854, + "step": 11020 + }, + { + "epoch": 0.35, + "learning_rate": 1.5025095841315059e-05, + "loss": 1.0352, + "step": 11021 + }, + { + "epoch": 0.35, + "learning_rate": 1.5024200287848898e-05, + "loss": 0.9878, + "step": 11022 + }, + { + "epoch": 0.35, + "learning_rate": 1.5023304680478628e-05, + "loss": 1.0474, + "step": 11023 + }, + { + "epoch": 0.35, + "learning_rate": 1.5022409019213856e-05, + "loss": 1.0205, + "step": 11024 + }, + { + "epoch": 0.35, + "learning_rate": 1.5021513304064197e-05, + "loss": 0.5491, + "step": 11025 + }, + { + "epoch": 0.35, + "learning_rate": 1.5020617535039259e-05, + "loss": 1.0796, + "step": 11026 + }, + { + "epoch": 0.35, + "learning_rate": 1.5019721712148648e-05, + "loss": 1.0815, + "step": 11027 + }, + { + "epoch": 0.35, + "learning_rate": 1.501882583540198e-05, + "loss": 1.1226, + "step": 11028 + }, + { + "epoch": 0.35, + "learning_rate": 1.5017929904808863e-05, + "loss": 1.0317, + "step": 11029 + }, + { + "epoch": 0.35, + "learning_rate": 1.5017033920378917e-05, + "loss": 1.1489, + "step": 11030 + }, + { + "epoch": 0.35, + "learning_rate": 1.5016137882121746e-05, + "loss": 1.1553, + "step": 11031 + }, + { + "epoch": 0.35, + "learning_rate": 1.5015241790046968e-05, + "loss": 1.0, + "step": 11032 + }, + { + "epoch": 0.35, + "learning_rate": 1.5014345644164195e-05, + "loss": 1.0366, + "step": 11033 + }, + { + "epoch": 0.35, + "learning_rate": 1.5013449444483043e-05, + "loss": 1.0024, + "step": 11034 + }, + { + "epoch": 0.35, + "learning_rate": 1.5012553191013128e-05, + "loss": 0.9619, + "step": 11035 + }, + { + "epoch": 0.35, + "learning_rate": 1.5011656883764065e-05, + "loss": 1.0107, + "step": 11036 + }, + { + "epoch": 0.35, + "learning_rate": 1.501076052274547e-05, + "loss": 1.0015, + "step": 11037 + }, + { + "epoch": 0.35, + "learning_rate": 1.500986410796696e-05, + "loss": 0.936, + "step": 11038 + }, + { + "epoch": 0.35, + "learning_rate": 1.5008967639438153e-05, + "loss": 1.0444, + "step": 11039 + }, + { + "epoch": 0.35, + "learning_rate": 1.5008071117168665e-05, + "loss": 1.0283, + "step": 11040 + }, + { + "epoch": 0.35, + "learning_rate": 1.500717454116812e-05, + "loss": 1.1787, + "step": 11041 + }, + { + "epoch": 0.35, + "learning_rate": 1.5006277911446133e-05, + "loss": 1.0508, + "step": 11042 + }, + { + "epoch": 0.35, + "learning_rate": 1.5005381228012321e-05, + "loss": 1.0405, + "step": 11043 + }, + { + "epoch": 0.35, + "learning_rate": 1.500448449087631e-05, + "loss": 1.1904, + "step": 11044 + }, + { + "epoch": 0.35, + "learning_rate": 1.500358770004772e-05, + "loss": 1.1963, + "step": 11045 + }, + { + "epoch": 0.35, + "learning_rate": 1.5002690855536173e-05, + "loss": 1.1357, + "step": 11046 + }, + { + "epoch": 0.35, + "learning_rate": 1.5001793957351288e-05, + "loss": 1.0386, + "step": 11047 + }, + { + "epoch": 0.35, + "learning_rate": 1.5000897005502691e-05, + "loss": 0.9956, + "step": 11048 + }, + { + "epoch": 0.35, + "learning_rate": 1.5000000000000002e-05, + "loss": 0.9668, + "step": 11049 + }, + { + "epoch": 0.35, + "learning_rate": 1.4999102940852848e-05, + "loss": 1.0347, + "step": 11050 + }, + { + "epoch": 0.35, + "learning_rate": 1.4998205828070851e-05, + "loss": 0.9932, + "step": 11051 + }, + { + "epoch": 0.35, + "learning_rate": 1.4997308661663637e-05, + "loss": 0.9111, + "step": 11052 + }, + { + "epoch": 0.35, + "learning_rate": 1.4996411441640835e-05, + "loss": 0.9805, + "step": 11053 + }, + { + "epoch": 0.35, + "learning_rate": 1.4995514168012064e-05, + "loss": 1.0767, + "step": 11054 + }, + { + "epoch": 0.35, + "learning_rate": 1.4994616840786956e-05, + "loss": 1.0742, + "step": 11055 + }, + { + "epoch": 0.35, + "learning_rate": 1.4993719459975139e-05, + "loss": 1.0205, + "step": 11056 + }, + { + "epoch": 0.35, + "learning_rate": 1.4992822025586238e-05, + "loss": 1.1016, + "step": 11057 + }, + { + "epoch": 0.35, + "learning_rate": 1.499192453762988e-05, + "loss": 0.5542, + "step": 11058 + }, + { + "epoch": 0.35, + "learning_rate": 1.4991026996115698e-05, + "loss": 1.0928, + "step": 11059 + }, + { + "epoch": 0.35, + "learning_rate": 1.499012940105332e-05, + "loss": 1.0991, + "step": 11060 + }, + { + "epoch": 0.35, + "learning_rate": 1.4989231752452378e-05, + "loss": 0.9614, + "step": 11061 + }, + { + "epoch": 0.35, + "learning_rate": 1.49883340503225e-05, + "loss": 0.5371, + "step": 11062 + }, + { + "epoch": 0.35, + "learning_rate": 1.4987436294673318e-05, + "loss": 1.0713, + "step": 11063 + }, + { + "epoch": 0.35, + "learning_rate": 1.4986538485514466e-05, + "loss": 0.9663, + "step": 11064 + }, + { + "epoch": 0.35, + "learning_rate": 1.4985640622855573e-05, + "loss": 1.0425, + "step": 11065 + }, + { + "epoch": 0.35, + "learning_rate": 1.4984742706706275e-05, + "loss": 1.0337, + "step": 11066 + }, + { + "epoch": 0.35, + "learning_rate": 1.4983844737076205e-05, + "loss": 1.0635, + "step": 11067 + }, + { + "epoch": 0.35, + "learning_rate": 1.4982946713974996e-05, + "loss": 0.9834, + "step": 11068 + }, + { + "epoch": 0.35, + "learning_rate": 1.4982048637412282e-05, + "loss": 1.0376, + "step": 11069 + }, + { + "epoch": 0.35, + "learning_rate": 1.4981150507397704e-05, + "loss": 0.9238, + "step": 11070 + }, + { + "epoch": 0.35, + "learning_rate": 1.4980252323940891e-05, + "loss": 1.0195, + "step": 11071 + }, + { + "epoch": 0.35, + "learning_rate": 1.4979354087051484e-05, + "loss": 1.0425, + "step": 11072 + }, + { + "epoch": 0.35, + "learning_rate": 1.4978455796739117e-05, + "loss": 0.9624, + "step": 11073 + }, + { + "epoch": 0.35, + "learning_rate": 1.497755745301343e-05, + "loss": 1.0542, + "step": 11074 + }, + { + "epoch": 0.35, + "learning_rate": 1.4976659055884062e-05, + "loss": 1.0132, + "step": 11075 + }, + { + "epoch": 0.35, + "learning_rate": 1.4975760605360648e-05, + "loss": 0.8906, + "step": 11076 + }, + { + "epoch": 0.35, + "learning_rate": 1.497486210145283e-05, + "loss": 1.0156, + "step": 11077 + }, + { + "epoch": 0.35, + "learning_rate": 1.4973963544170245e-05, + "loss": 0.8984, + "step": 11078 + }, + { + "epoch": 0.35, + "learning_rate": 1.497306493352254e-05, + "loss": 1.0405, + "step": 11079 + }, + { + "epoch": 0.35, + "learning_rate": 1.497216626951935e-05, + "loss": 1.1035, + "step": 11080 + }, + { + "epoch": 0.35, + "learning_rate": 1.4971267552170319e-05, + "loss": 1.168, + "step": 11081 + }, + { + "epoch": 0.35, + "learning_rate": 1.497036878148509e-05, + "loss": 1.0537, + "step": 11082 + }, + { + "epoch": 0.35, + "learning_rate": 1.4969469957473306e-05, + "loss": 1.1436, + "step": 11083 + }, + { + "epoch": 0.35, + "learning_rate": 1.4968571080144607e-05, + "loss": 1.0479, + "step": 11084 + }, + { + "epoch": 0.35, + "learning_rate": 1.4967672149508641e-05, + "loss": 0.9805, + "step": 11085 + }, + { + "epoch": 0.35, + "learning_rate": 1.4966773165575048e-05, + "loss": 1.0562, + "step": 11086 + }, + { + "epoch": 0.35, + "learning_rate": 1.496587412835348e-05, + "loss": 1.0493, + "step": 11087 + }, + { + "epoch": 0.35, + "learning_rate": 1.4964975037853575e-05, + "loss": 1.1182, + "step": 11088 + }, + { + "epoch": 0.35, + "learning_rate": 1.4964075894084983e-05, + "loss": 0.9165, + "step": 11089 + }, + { + "epoch": 0.35, + "learning_rate": 1.4963176697057353e-05, + "loss": 1.3037, + "step": 11090 + }, + { + "epoch": 0.35, + "learning_rate": 1.496227744678033e-05, + "loss": 1.0503, + "step": 11091 + }, + { + "epoch": 0.35, + "learning_rate": 1.4961378143263558e-05, + "loss": 1.0229, + "step": 11092 + }, + { + "epoch": 0.35, + "learning_rate": 1.4960478786516692e-05, + "loss": 1.1221, + "step": 11093 + }, + { + "epoch": 0.35, + "learning_rate": 1.4959579376549382e-05, + "loss": 0.9961, + "step": 11094 + }, + { + "epoch": 0.35, + "learning_rate": 1.495867991337127e-05, + "loss": 1.0454, + "step": 11095 + }, + { + "epoch": 0.35, + "learning_rate": 1.4957780396992013e-05, + "loss": 1.0913, + "step": 11096 + }, + { + "epoch": 0.35, + "learning_rate": 1.4956880827421256e-05, + "loss": 0.9146, + "step": 11097 + }, + { + "epoch": 0.35, + "learning_rate": 1.4955981204668658e-05, + "loss": 0.9956, + "step": 11098 + }, + { + "epoch": 0.35, + "learning_rate": 1.4955081528743863e-05, + "loss": 0.9766, + "step": 11099 + }, + { + "epoch": 0.35, + "learning_rate": 1.4954181799656529e-05, + "loss": 0.9956, + "step": 11100 + }, + { + "epoch": 0.36, + "learning_rate": 1.4953282017416306e-05, + "loss": 0.5466, + "step": 11101 + }, + { + "epoch": 0.36, + "learning_rate": 1.4952382182032852e-05, + "loss": 1.0215, + "step": 11102 + }, + { + "epoch": 0.36, + "learning_rate": 1.4951482293515815e-05, + "loss": 1.0562, + "step": 11103 + }, + { + "epoch": 0.36, + "learning_rate": 1.4950582351874853e-05, + "loss": 0.9331, + "step": 11104 + }, + { + "epoch": 0.36, + "learning_rate": 1.4949682357119625e-05, + "loss": 1.1123, + "step": 11105 + }, + { + "epoch": 0.36, + "learning_rate": 1.494878230925978e-05, + "loss": 1.0454, + "step": 11106 + }, + { + "epoch": 0.36, + "learning_rate": 1.4947882208304979e-05, + "loss": 0.9634, + "step": 11107 + }, + { + "epoch": 0.36, + "learning_rate": 1.4946982054264876e-05, + "loss": 1.0449, + "step": 11108 + }, + { + "epoch": 0.36, + "learning_rate": 1.4946081847149134e-05, + "loss": 1.1113, + "step": 11109 + }, + { + "epoch": 0.36, + "learning_rate": 1.4945181586967405e-05, + "loss": 1.021, + "step": 11110 + }, + { + "epoch": 0.36, + "learning_rate": 1.4944281273729352e-05, + "loss": 1.0176, + "step": 11111 + }, + { + "epoch": 0.36, + "learning_rate": 1.4943380907444635e-05, + "loss": 1.1294, + "step": 11112 + }, + { + "epoch": 0.36, + "learning_rate": 1.4942480488122909e-05, + "loss": 1.1309, + "step": 11113 + }, + { + "epoch": 0.36, + "learning_rate": 1.494158001577384e-05, + "loss": 1.0044, + "step": 11114 + }, + { + "epoch": 0.36, + "learning_rate": 1.4940679490407084e-05, + "loss": 1.0625, + "step": 11115 + }, + { + "epoch": 0.36, + "learning_rate": 1.4939778912032306e-05, + "loss": 1.0801, + "step": 11116 + }, + { + "epoch": 0.36, + "learning_rate": 1.4938878280659168e-05, + "loss": 1.0444, + "step": 11117 + }, + { + "epoch": 0.36, + "learning_rate": 1.4937977596297331e-05, + "loss": 0.9297, + "step": 11118 + }, + { + "epoch": 0.36, + "learning_rate": 1.4937076858956458e-05, + "loss": 0.8657, + "step": 11119 + }, + { + "epoch": 0.36, + "learning_rate": 1.4936176068646215e-05, + "loss": 1.061, + "step": 11120 + }, + { + "epoch": 0.36, + "learning_rate": 1.493527522537627e-05, + "loss": 1.0903, + "step": 11121 + }, + { + "epoch": 0.36, + "learning_rate": 1.4934374329156281e-05, + "loss": 1.0249, + "step": 11122 + }, + { + "epoch": 0.36, + "learning_rate": 1.4933473379995915e-05, + "loss": 1.1299, + "step": 11123 + }, + { + "epoch": 0.36, + "learning_rate": 1.4932572377904842e-05, + "loss": 0.9766, + "step": 11124 + }, + { + "epoch": 0.36, + "learning_rate": 1.4931671322892724e-05, + "loss": 1.0332, + "step": 11125 + }, + { + "epoch": 0.36, + "learning_rate": 1.4930770214969234e-05, + "loss": 1.1362, + "step": 11126 + }, + { + "epoch": 0.36, + "learning_rate": 1.4929869054144033e-05, + "loss": 0.9126, + "step": 11127 + }, + { + "epoch": 0.36, + "learning_rate": 1.4928967840426799e-05, + "loss": 0.9048, + "step": 11128 + }, + { + "epoch": 0.36, + "learning_rate": 1.4928066573827189e-05, + "loss": 0.937, + "step": 11129 + }, + { + "epoch": 0.36, + "learning_rate": 1.4927165254354883e-05, + "loss": 1.0166, + "step": 11130 + }, + { + "epoch": 0.36, + "learning_rate": 1.4926263882019543e-05, + "loss": 1.0195, + "step": 11131 + }, + { + "epoch": 0.36, + "learning_rate": 1.4925362456830847e-05, + "loss": 0.9487, + "step": 11132 + }, + { + "epoch": 0.36, + "learning_rate": 1.4924460978798463e-05, + "loss": 1.0923, + "step": 11133 + }, + { + "epoch": 0.36, + "learning_rate": 1.492355944793206e-05, + "loss": 0.9741, + "step": 11134 + }, + { + "epoch": 0.36, + "learning_rate": 1.4922657864241314e-05, + "loss": 0.5681, + "step": 11135 + }, + { + "epoch": 0.36, + "learning_rate": 1.49217562277359e-05, + "loss": 1.103, + "step": 11136 + }, + { + "epoch": 0.36, + "learning_rate": 1.4920854538425487e-05, + "loss": 0.9253, + "step": 11137 + }, + { + "epoch": 0.36, + "learning_rate": 1.491995279631975e-05, + "loss": 1.0342, + "step": 11138 + }, + { + "epoch": 0.36, + "learning_rate": 1.4919051001428369e-05, + "loss": 0.9263, + "step": 11139 + }, + { + "epoch": 0.36, + "learning_rate": 1.491814915376101e-05, + "loss": 1.0073, + "step": 11140 + }, + { + "epoch": 0.36, + "learning_rate": 1.4917247253327353e-05, + "loss": 1.0825, + "step": 11141 + }, + { + "epoch": 0.36, + "learning_rate": 1.4916345300137076e-05, + "loss": 1.0229, + "step": 11142 + }, + { + "epoch": 0.36, + "learning_rate": 1.491544329419986e-05, + "loss": 1.1304, + "step": 11143 + }, + { + "epoch": 0.36, + "learning_rate": 1.4914541235525373e-05, + "loss": 0.896, + "step": 11144 + }, + { + "epoch": 0.36, + "learning_rate": 1.4913639124123297e-05, + "loss": 1.0332, + "step": 11145 + }, + { + "epoch": 0.36, + "learning_rate": 1.4912736960003314e-05, + "loss": 1.0337, + "step": 11146 + }, + { + "epoch": 0.36, + "learning_rate": 1.49118347431751e-05, + "loss": 1.1416, + "step": 11147 + }, + { + "epoch": 0.36, + "learning_rate": 1.4910932473648334e-05, + "loss": 1.0239, + "step": 11148 + }, + { + "epoch": 0.36, + "learning_rate": 1.49100301514327e-05, + "loss": 1.0376, + "step": 11149 + }, + { + "epoch": 0.36, + "learning_rate": 1.4909127776537873e-05, + "loss": 1.1006, + "step": 11150 + }, + { + "epoch": 0.36, + "learning_rate": 1.4908225348973541e-05, + "loss": 0.9634, + "step": 11151 + }, + { + "epoch": 0.36, + "learning_rate": 1.4907322868749384e-05, + "loss": 0.8887, + "step": 11152 + }, + { + "epoch": 0.36, + "learning_rate": 1.490642033587508e-05, + "loss": 1.0317, + "step": 11153 + }, + { + "epoch": 0.36, + "learning_rate": 1.4905517750360321e-05, + "loss": 1.0469, + "step": 11154 + }, + { + "epoch": 0.36, + "learning_rate": 1.4904615112214781e-05, + "loss": 1.124, + "step": 11155 + }, + { + "epoch": 0.36, + "learning_rate": 1.4903712421448153e-05, + "loss": 1.0474, + "step": 11156 + }, + { + "epoch": 0.36, + "learning_rate": 1.4902809678070115e-05, + "loss": 1.0215, + "step": 11157 + }, + { + "epoch": 0.36, + "learning_rate": 1.4901906882090355e-05, + "loss": 0.9668, + "step": 11158 + }, + { + "epoch": 0.36, + "learning_rate": 1.4901004033518565e-05, + "loss": 1.041, + "step": 11159 + }, + { + "epoch": 0.36, + "learning_rate": 1.490010113236442e-05, + "loss": 1.0981, + "step": 11160 + }, + { + "epoch": 0.36, + "learning_rate": 1.4899198178637613e-05, + "loss": 1.0249, + "step": 11161 + }, + { + "epoch": 0.36, + "learning_rate": 1.4898295172347834e-05, + "loss": 1.105, + "step": 11162 + }, + { + "epoch": 0.36, + "learning_rate": 1.4897392113504767e-05, + "loss": 1.0264, + "step": 11163 + }, + { + "epoch": 0.36, + "learning_rate": 1.4896489002118104e-05, + "loss": 1.1738, + "step": 11164 + }, + { + "epoch": 0.36, + "learning_rate": 1.4895585838197532e-05, + "loss": 1.1006, + "step": 11165 + }, + { + "epoch": 0.36, + "learning_rate": 1.4894682621752745e-05, + "loss": 1.043, + "step": 11166 + }, + { + "epoch": 0.36, + "learning_rate": 1.4893779352793428e-05, + "loss": 0.8726, + "step": 11167 + }, + { + "epoch": 0.36, + "learning_rate": 1.4892876031329276e-05, + "loss": 1.0229, + "step": 11168 + }, + { + "epoch": 0.36, + "learning_rate": 1.4891972657369977e-05, + "loss": 0.999, + "step": 11169 + }, + { + "epoch": 0.36, + "learning_rate": 1.4891069230925231e-05, + "loss": 0.9932, + "step": 11170 + }, + { + "epoch": 0.36, + "learning_rate": 1.489016575200472e-05, + "loss": 1.0654, + "step": 11171 + }, + { + "epoch": 0.36, + "learning_rate": 1.4889262220618143e-05, + "loss": 1.0742, + "step": 11172 + }, + { + "epoch": 0.36, + "learning_rate": 1.4888358636775197e-05, + "loss": 1.1504, + "step": 11173 + }, + { + "epoch": 0.36, + "learning_rate": 1.4887455000485571e-05, + "loss": 1.021, + "step": 11174 + }, + { + "epoch": 0.36, + "learning_rate": 1.4886551311758962e-05, + "loss": 1.0142, + "step": 11175 + }, + { + "epoch": 0.36, + "learning_rate": 1.4885647570605062e-05, + "loss": 1.0562, + "step": 11176 + }, + { + "epoch": 0.36, + "learning_rate": 1.4884743777033576e-05, + "loss": 1.0361, + "step": 11177 + }, + { + "epoch": 0.36, + "learning_rate": 1.4883839931054194e-05, + "loss": 1.0015, + "step": 11178 + }, + { + "epoch": 0.36, + "learning_rate": 1.4882936032676613e-05, + "loss": 0.9951, + "step": 11179 + }, + { + "epoch": 0.36, + "learning_rate": 1.4882032081910534e-05, + "loss": 0.9712, + "step": 11180 + }, + { + "epoch": 0.36, + "learning_rate": 1.4881128078765654e-05, + "loss": 1.0801, + "step": 11181 + }, + { + "epoch": 0.36, + "learning_rate": 1.4880224023251672e-05, + "loss": 1.0474, + "step": 11182 + }, + { + "epoch": 0.36, + "learning_rate": 1.4879319915378288e-05, + "loss": 0.9487, + "step": 11183 + }, + { + "epoch": 0.36, + "learning_rate": 1.4878415755155198e-05, + "loss": 1.1035, + "step": 11184 + }, + { + "epoch": 0.36, + "learning_rate": 1.487751154259211e-05, + "loss": 0.9946, + "step": 11185 + }, + { + "epoch": 0.36, + "learning_rate": 1.487660727769872e-05, + "loss": 0.897, + "step": 11186 + }, + { + "epoch": 0.36, + "learning_rate": 1.4875702960484731e-05, + "loss": 0.9956, + "step": 11187 + }, + { + "epoch": 0.36, + "learning_rate": 1.4874798590959846e-05, + "loss": 0.9487, + "step": 11188 + }, + { + "epoch": 0.36, + "learning_rate": 1.4873894169133767e-05, + "loss": 0.9331, + "step": 11189 + }, + { + "epoch": 0.36, + "learning_rate": 1.4872989695016198e-05, + "loss": 1.0518, + "step": 11190 + }, + { + "epoch": 0.36, + "learning_rate": 1.4872085168616841e-05, + "loss": 1.1118, + "step": 11191 + }, + { + "epoch": 0.36, + "learning_rate": 1.4871180589945405e-05, + "loss": 1.0771, + "step": 11192 + }, + { + "epoch": 0.36, + "learning_rate": 1.4870275959011591e-05, + "loss": 1.1465, + "step": 11193 + }, + { + "epoch": 0.36, + "learning_rate": 1.4869371275825105e-05, + "loss": 1.0352, + "step": 11194 + }, + { + "epoch": 0.36, + "learning_rate": 1.4868466540395657e-05, + "loss": 1.0391, + "step": 11195 + }, + { + "epoch": 0.36, + "learning_rate": 1.4867561752732949e-05, + "loss": 1.0659, + "step": 11196 + }, + { + "epoch": 0.36, + "learning_rate": 1.4866656912846691e-05, + "loss": 1.0864, + "step": 11197 + }, + { + "epoch": 0.36, + "learning_rate": 1.4865752020746591e-05, + "loss": 0.9429, + "step": 11198 + }, + { + "epoch": 0.36, + "learning_rate": 1.4864847076442358e-05, + "loss": 1.0137, + "step": 11199 + }, + { + "epoch": 0.36, + "learning_rate": 1.4863942079943696e-05, + "loss": 1.061, + "step": 11200 + }, + { + "epoch": 0.36, + "learning_rate": 1.4863037031260321e-05, + "loss": 1.0605, + "step": 11201 + }, + { + "epoch": 0.36, + "learning_rate": 1.4862131930401943e-05, + "loss": 0.9985, + "step": 11202 + }, + { + "epoch": 0.36, + "learning_rate": 1.4861226777378269e-05, + "loss": 1.0347, + "step": 11203 + }, + { + "epoch": 0.36, + "learning_rate": 1.4860321572199011e-05, + "loss": 0.9756, + "step": 11204 + }, + { + "epoch": 0.36, + "learning_rate": 1.4859416314873882e-05, + "loss": 0.572, + "step": 11205 + }, + { + "epoch": 0.36, + "learning_rate": 1.4858511005412593e-05, + "loss": 1.1147, + "step": 11206 + }, + { + "epoch": 0.36, + "learning_rate": 1.485760564382486e-05, + "loss": 0.9673, + "step": 11207 + }, + { + "epoch": 0.36, + "learning_rate": 1.4856700230120393e-05, + "loss": 1.0811, + "step": 11208 + }, + { + "epoch": 0.36, + "learning_rate": 1.4855794764308909e-05, + "loss": 0.9839, + "step": 11209 + }, + { + "epoch": 0.36, + "learning_rate": 1.485488924640012e-05, + "loss": 0.917, + "step": 11210 + }, + { + "epoch": 0.36, + "learning_rate": 1.4853983676403743e-05, + "loss": 0.999, + "step": 11211 + }, + { + "epoch": 0.36, + "learning_rate": 1.4853078054329493e-05, + "loss": 0.9834, + "step": 11212 + }, + { + "epoch": 0.36, + "learning_rate": 1.4852172380187086e-05, + "loss": 1.0044, + "step": 11213 + }, + { + "epoch": 0.36, + "learning_rate": 1.485126665398624e-05, + "loss": 1.1514, + "step": 11214 + }, + { + "epoch": 0.36, + "learning_rate": 1.4850360875736676e-05, + "loss": 0.8413, + "step": 11215 + }, + { + "epoch": 0.36, + "learning_rate": 1.4849455045448103e-05, + "loss": 1.1128, + "step": 11216 + }, + { + "epoch": 0.36, + "learning_rate": 1.4848549163130246e-05, + "loss": 0.8848, + "step": 11217 + }, + { + "epoch": 0.36, + "learning_rate": 1.484764322879282e-05, + "loss": 0.9502, + "step": 11218 + }, + { + "epoch": 0.36, + "learning_rate": 1.4846737242445552e-05, + "loss": 0.9976, + "step": 11219 + }, + { + "epoch": 0.36, + "learning_rate": 1.4845831204098153e-05, + "loss": 1.0859, + "step": 11220 + }, + { + "epoch": 0.36, + "learning_rate": 1.4844925113760351e-05, + "loss": 1.0229, + "step": 11221 + }, + { + "epoch": 0.36, + "learning_rate": 1.4844018971441867e-05, + "loss": 0.9453, + "step": 11222 + }, + { + "epoch": 0.36, + "learning_rate": 1.4843112777152415e-05, + "loss": 0.9912, + "step": 11223 + }, + { + "epoch": 0.36, + "learning_rate": 1.4842206530901726e-05, + "loss": 1.0938, + "step": 11224 + }, + { + "epoch": 0.36, + "learning_rate": 1.4841300232699518e-05, + "loss": 0.9819, + "step": 11225 + }, + { + "epoch": 0.36, + "learning_rate": 1.4840393882555522e-05, + "loss": 0.9429, + "step": 11226 + }, + { + "epoch": 0.36, + "learning_rate": 1.4839487480479451e-05, + "loss": 1.0249, + "step": 11227 + }, + { + "epoch": 0.36, + "learning_rate": 1.4838581026481035e-05, + "loss": 0.9902, + "step": 11228 + }, + { + "epoch": 0.36, + "learning_rate": 1.4837674520570003e-05, + "loss": 1.0088, + "step": 11229 + }, + { + "epoch": 0.36, + "learning_rate": 1.4836767962756077e-05, + "loss": 1.0698, + "step": 11230 + }, + { + "epoch": 0.36, + "learning_rate": 1.4835861353048983e-05, + "loss": 0.9731, + "step": 11231 + }, + { + "epoch": 0.36, + "learning_rate": 1.483495469145845e-05, + "loss": 1.0176, + "step": 11232 + }, + { + "epoch": 0.36, + "learning_rate": 1.4834047977994201e-05, + "loss": 1.0425, + "step": 11233 + }, + { + "epoch": 0.36, + "learning_rate": 1.4833141212665974e-05, + "loss": 1.02, + "step": 11234 + }, + { + "epoch": 0.36, + "learning_rate": 1.4832234395483484e-05, + "loss": 1.0898, + "step": 11235 + }, + { + "epoch": 0.36, + "learning_rate": 1.483132752645647e-05, + "loss": 0.542, + "step": 11236 + }, + { + "epoch": 0.36, + "learning_rate": 1.4830420605594658e-05, + "loss": 1.0728, + "step": 11237 + }, + { + "epoch": 0.36, + "learning_rate": 1.482951363290778e-05, + "loss": 1.0249, + "step": 11238 + }, + { + "epoch": 0.36, + "learning_rate": 1.4828606608405562e-05, + "loss": 0.9136, + "step": 11239 + }, + { + "epoch": 0.36, + "learning_rate": 1.4827699532097745e-05, + "loss": 1.0674, + "step": 11240 + }, + { + "epoch": 0.36, + "learning_rate": 1.482679240399405e-05, + "loss": 0.9194, + "step": 11241 + }, + { + "epoch": 0.36, + "learning_rate": 1.482588522410422e-05, + "loss": 1.0962, + "step": 11242 + }, + { + "epoch": 0.36, + "learning_rate": 1.4824977992437977e-05, + "loss": 0.915, + "step": 11243 + }, + { + "epoch": 0.36, + "learning_rate": 1.4824070709005063e-05, + "loss": 0.9951, + "step": 11244 + }, + { + "epoch": 0.36, + "learning_rate": 1.4823163373815209e-05, + "loss": 0.9155, + "step": 11245 + }, + { + "epoch": 0.36, + "learning_rate": 1.4822255986878148e-05, + "loss": 0.8911, + "step": 11246 + }, + { + "epoch": 0.36, + "learning_rate": 1.482134854820362e-05, + "loss": 1.0786, + "step": 11247 + }, + { + "epoch": 0.36, + "learning_rate": 1.4820441057801356e-05, + "loss": 1.0435, + "step": 11248 + }, + { + "epoch": 0.36, + "learning_rate": 1.4819533515681097e-05, + "loss": 0.8474, + "step": 11249 + }, + { + "epoch": 0.36, + "learning_rate": 1.4818625921852574e-05, + "loss": 1.0522, + "step": 11250 + }, + { + "epoch": 0.36, + "learning_rate": 1.481771827632553e-05, + "loss": 1.0063, + "step": 11251 + }, + { + "epoch": 0.36, + "learning_rate": 1.48168105791097e-05, + "loss": 1.0371, + "step": 11252 + }, + { + "epoch": 0.36, + "learning_rate": 1.4815902830214826e-05, + "loss": 1.0518, + "step": 11253 + }, + { + "epoch": 0.36, + "learning_rate": 1.481499502965064e-05, + "loss": 1.0571, + "step": 11254 + }, + { + "epoch": 0.36, + "learning_rate": 1.4814087177426887e-05, + "loss": 1.0889, + "step": 11255 + }, + { + "epoch": 0.36, + "learning_rate": 1.481317927355331e-05, + "loss": 1.0825, + "step": 11256 + }, + { + "epoch": 0.36, + "learning_rate": 1.4812271318039646e-05, + "loss": 0.9097, + "step": 11257 + }, + { + "epoch": 0.36, + "learning_rate": 1.4811363310895634e-05, + "loss": 1.2949, + "step": 11258 + }, + { + "epoch": 0.36, + "learning_rate": 1.4810455252131019e-05, + "loss": 0.9883, + "step": 11259 + }, + { + "epoch": 0.36, + "learning_rate": 1.4809547141755548e-05, + "loss": 1.1201, + "step": 11260 + }, + { + "epoch": 0.36, + "learning_rate": 1.4808638979778953e-05, + "loss": 1.0649, + "step": 11261 + }, + { + "epoch": 0.36, + "learning_rate": 1.4807730766210985e-05, + "loss": 1.0659, + "step": 11262 + }, + { + "epoch": 0.36, + "learning_rate": 1.4806822501061388e-05, + "loss": 0.957, + "step": 11263 + }, + { + "epoch": 0.36, + "learning_rate": 1.4805914184339907e-05, + "loss": 1.0742, + "step": 11264 + }, + { + "epoch": 0.36, + "learning_rate": 1.4805005816056283e-05, + "loss": 1.1597, + "step": 11265 + }, + { + "epoch": 0.36, + "learning_rate": 1.4804097396220266e-05, + "loss": 0.9438, + "step": 11266 + }, + { + "epoch": 0.36, + "learning_rate": 1.4803188924841598e-05, + "loss": 1.0991, + "step": 11267 + }, + { + "epoch": 0.36, + "learning_rate": 1.4802280401930033e-05, + "loss": 0.8604, + "step": 11268 + }, + { + "epoch": 0.36, + "learning_rate": 1.4801371827495312e-05, + "loss": 1.0288, + "step": 11269 + }, + { + "epoch": 0.36, + "learning_rate": 1.4800463201547186e-05, + "loss": 0.9673, + "step": 11270 + }, + { + "epoch": 0.36, + "learning_rate": 1.47995545240954e-05, + "loss": 1.0195, + "step": 11271 + }, + { + "epoch": 0.36, + "learning_rate": 1.4798645795149708e-05, + "loss": 1.1104, + "step": 11272 + }, + { + "epoch": 0.36, + "learning_rate": 1.4797737014719859e-05, + "loss": 1.124, + "step": 11273 + }, + { + "epoch": 0.36, + "learning_rate": 1.4796828182815599e-05, + "loss": 1.0898, + "step": 11274 + }, + { + "epoch": 0.36, + "learning_rate": 1.4795919299446686e-05, + "loss": 1.0791, + "step": 11275 + }, + { + "epoch": 0.36, + "learning_rate": 1.4795010364622861e-05, + "loss": 1.0176, + "step": 11276 + }, + { + "epoch": 0.36, + "learning_rate": 1.4794101378353884e-05, + "loss": 0.9756, + "step": 11277 + }, + { + "epoch": 0.36, + "learning_rate": 1.4793192340649505e-05, + "loss": 1.0737, + "step": 11278 + }, + { + "epoch": 0.36, + "learning_rate": 1.4792283251519479e-05, + "loss": 0.9678, + "step": 11279 + }, + { + "epoch": 0.36, + "learning_rate": 1.4791374110973555e-05, + "loss": 0.9907, + "step": 11280 + }, + { + "epoch": 0.36, + "learning_rate": 1.479046491902149e-05, + "loss": 1.041, + "step": 11281 + }, + { + "epoch": 0.36, + "learning_rate": 1.4789555675673039e-05, + "loss": 1.0225, + "step": 11282 + }, + { + "epoch": 0.36, + "learning_rate": 1.4788646380937957e-05, + "loss": 0.9341, + "step": 11283 + }, + { + "epoch": 0.36, + "learning_rate": 1.4787737034825997e-05, + "loss": 0.9424, + "step": 11284 + }, + { + "epoch": 0.36, + "learning_rate": 1.4786827637346922e-05, + "loss": 1.0557, + "step": 11285 + }, + { + "epoch": 0.36, + "learning_rate": 1.4785918188510477e-05, + "loss": 0.9956, + "step": 11286 + }, + { + "epoch": 0.36, + "learning_rate": 1.4785008688326432e-05, + "loss": 1.0947, + "step": 11287 + }, + { + "epoch": 0.36, + "learning_rate": 1.4784099136804537e-05, + "loss": 1.0303, + "step": 11288 + }, + { + "epoch": 0.36, + "learning_rate": 1.4783189533954555e-05, + "loss": 1.022, + "step": 11289 + }, + { + "epoch": 0.36, + "learning_rate": 1.4782279879786243e-05, + "loss": 1.0391, + "step": 11290 + }, + { + "epoch": 0.36, + "learning_rate": 1.478137017430936e-05, + "loss": 1.0474, + "step": 11291 + }, + { + "epoch": 0.36, + "learning_rate": 1.4780460417533666e-05, + "loss": 0.9678, + "step": 11292 + }, + { + "epoch": 0.36, + "learning_rate": 1.4779550609468921e-05, + "loss": 1.0044, + "step": 11293 + }, + { + "epoch": 0.36, + "learning_rate": 1.4778640750124893e-05, + "loss": 1.0625, + "step": 11294 + }, + { + "epoch": 0.36, + "learning_rate": 1.4777730839511335e-05, + "loss": 0.9478, + "step": 11295 + }, + { + "epoch": 0.36, + "learning_rate": 1.4776820877638011e-05, + "loss": 1.0659, + "step": 11296 + }, + { + "epoch": 0.36, + "learning_rate": 1.4775910864514688e-05, + "loss": 0.9824, + "step": 11297 + }, + { + "epoch": 0.36, + "learning_rate": 1.4775000800151129e-05, + "loss": 1.0308, + "step": 11298 + }, + { + "epoch": 0.36, + "learning_rate": 1.4774090684557094e-05, + "loss": 1.0352, + "step": 11299 + }, + { + "epoch": 0.36, + "learning_rate": 1.4773180517742346e-05, + "loss": 0.9385, + "step": 11300 + }, + { + "epoch": 0.36, + "learning_rate": 1.4772270299716658e-05, + "loss": 0.592, + "step": 11301 + }, + { + "epoch": 0.36, + "learning_rate": 1.4771360030489792e-05, + "loss": 0.9219, + "step": 11302 + }, + { + "epoch": 0.36, + "learning_rate": 1.4770449710071513e-05, + "loss": 0.9902, + "step": 11303 + }, + { + "epoch": 0.36, + "learning_rate": 1.4769539338471587e-05, + "loss": 1.0845, + "step": 11304 + }, + { + "epoch": 0.36, + "learning_rate": 1.4768628915699783e-05, + "loss": 1.0088, + "step": 11305 + }, + { + "epoch": 0.36, + "learning_rate": 1.476771844176587e-05, + "loss": 1.1294, + "step": 11306 + }, + { + "epoch": 0.36, + "learning_rate": 1.4766807916679613e-05, + "loss": 1.186, + "step": 11307 + }, + { + "epoch": 0.36, + "learning_rate": 1.4765897340450784e-05, + "loss": 0.9897, + "step": 11308 + }, + { + "epoch": 0.36, + "learning_rate": 1.476498671308915e-05, + "loss": 0.9614, + "step": 11309 + }, + { + "epoch": 0.36, + "learning_rate": 1.4764076034604482e-05, + "loss": 1.0312, + "step": 11310 + }, + { + "epoch": 0.36, + "learning_rate": 1.4763165305006551e-05, + "loss": 1.021, + "step": 11311 + }, + { + "epoch": 0.36, + "learning_rate": 1.4762254524305126e-05, + "loss": 1.0654, + "step": 11312 + }, + { + "epoch": 0.36, + "learning_rate": 1.4761343692509983e-05, + "loss": 1.1055, + "step": 11313 + }, + { + "epoch": 0.36, + "learning_rate": 1.4760432809630894e-05, + "loss": 0.9912, + "step": 11314 + }, + { + "epoch": 0.36, + "learning_rate": 1.4759521875677625e-05, + "loss": 1.0454, + "step": 11315 + }, + { + "epoch": 0.36, + "learning_rate": 1.4758610890659957e-05, + "loss": 1.0347, + "step": 11316 + }, + { + "epoch": 0.36, + "learning_rate": 1.475769985458766e-05, + "loss": 0.999, + "step": 11317 + }, + { + "epoch": 0.36, + "learning_rate": 1.4756788767470507e-05, + "loss": 1.0625, + "step": 11318 + }, + { + "epoch": 0.36, + "learning_rate": 1.4755877629318276e-05, + "loss": 0.9858, + "step": 11319 + }, + { + "epoch": 0.36, + "learning_rate": 1.4754966440140743e-05, + "loss": 1.0371, + "step": 11320 + }, + { + "epoch": 0.36, + "learning_rate": 1.4754055199947684e-05, + "loss": 1.0435, + "step": 11321 + }, + { + "epoch": 0.36, + "learning_rate": 1.4753143908748871e-05, + "loss": 1.0166, + "step": 11322 + }, + { + "epoch": 0.36, + "learning_rate": 1.4752232566554087e-05, + "loss": 0.9702, + "step": 11323 + }, + { + "epoch": 0.36, + "learning_rate": 1.4751321173373105e-05, + "loss": 0.9712, + "step": 11324 + }, + { + "epoch": 0.36, + "learning_rate": 1.4750409729215704e-05, + "loss": 0.9565, + "step": 11325 + }, + { + "epoch": 0.36, + "learning_rate": 1.4749498234091668e-05, + "loss": 1.1182, + "step": 11326 + }, + { + "epoch": 0.36, + "learning_rate": 1.4748586688010771e-05, + "loss": 1.0371, + "step": 11327 + }, + { + "epoch": 0.36, + "learning_rate": 1.4747675090982791e-05, + "loss": 1.0063, + "step": 11328 + }, + { + "epoch": 0.36, + "learning_rate": 1.4746763443017518e-05, + "loss": 0.9351, + "step": 11329 + }, + { + "epoch": 0.36, + "learning_rate": 1.4745851744124722e-05, + "loss": 1.0459, + "step": 11330 + }, + { + "epoch": 0.36, + "learning_rate": 1.474493999431419e-05, + "loss": 1.1768, + "step": 11331 + }, + { + "epoch": 0.36, + "learning_rate": 1.4744028193595705e-05, + "loss": 1.1187, + "step": 11332 + }, + { + "epoch": 0.36, + "learning_rate": 1.4743116341979047e-05, + "loss": 1.0142, + "step": 11333 + }, + { + "epoch": 0.36, + "learning_rate": 1.4742204439473999e-05, + "loss": 1.0371, + "step": 11334 + }, + { + "epoch": 0.36, + "learning_rate": 1.4741292486090347e-05, + "loss": 1.0225, + "step": 11335 + }, + { + "epoch": 0.36, + "learning_rate": 1.4740380481837874e-05, + "loss": 1.0835, + "step": 11336 + }, + { + "epoch": 0.36, + "learning_rate": 1.4739468426726365e-05, + "loss": 0.9775, + "step": 11337 + }, + { + "epoch": 0.36, + "learning_rate": 1.4738556320765603e-05, + "loss": 1.0825, + "step": 11338 + }, + { + "epoch": 0.36, + "learning_rate": 1.4737644163965378e-05, + "loss": 1.0469, + "step": 11339 + }, + { + "epoch": 0.36, + "learning_rate": 1.4736731956335475e-05, + "loss": 1.041, + "step": 11340 + }, + { + "epoch": 0.36, + "learning_rate": 1.4735819697885679e-05, + "loss": 1.0522, + "step": 11341 + }, + { + "epoch": 0.36, + "learning_rate": 1.4734907388625778e-05, + "loss": 1.1108, + "step": 11342 + }, + { + "epoch": 0.36, + "learning_rate": 1.4733995028565564e-05, + "loss": 1.0942, + "step": 11343 + }, + { + "epoch": 0.36, + "learning_rate": 1.473308261771482e-05, + "loss": 0.9277, + "step": 11344 + }, + { + "epoch": 0.36, + "learning_rate": 1.4732170156083338e-05, + "loss": 1.0586, + "step": 11345 + }, + { + "epoch": 0.36, + "learning_rate": 1.4731257643680909e-05, + "loss": 1.0586, + "step": 11346 + }, + { + "epoch": 0.36, + "learning_rate": 1.4730345080517324e-05, + "loss": 0.9751, + "step": 11347 + }, + { + "epoch": 0.36, + "learning_rate": 1.4729432466602367e-05, + "loss": 1.0449, + "step": 11348 + }, + { + "epoch": 0.36, + "learning_rate": 1.4728519801945836e-05, + "loss": 1.1772, + "step": 11349 + }, + { + "epoch": 0.36, + "learning_rate": 1.4727607086557521e-05, + "loss": 1.0972, + "step": 11350 + }, + { + "epoch": 0.36, + "learning_rate": 1.4726694320447215e-05, + "loss": 1.0913, + "step": 11351 + }, + { + "epoch": 0.36, + "learning_rate": 1.4725781503624708e-05, + "loss": 1.0752, + "step": 11352 + }, + { + "epoch": 0.36, + "learning_rate": 1.4724868636099796e-05, + "loss": 0.9707, + "step": 11353 + }, + { + "epoch": 0.36, + "learning_rate": 1.4723955717882273e-05, + "loss": 1.0073, + "step": 11354 + }, + { + "epoch": 0.36, + "learning_rate": 1.4723042748981936e-05, + "loss": 1.0981, + "step": 11355 + }, + { + "epoch": 0.36, + "learning_rate": 1.4722129729408574e-05, + "loss": 1.1006, + "step": 11356 + }, + { + "epoch": 0.36, + "learning_rate": 1.4721216659171988e-05, + "loss": 1.1597, + "step": 11357 + }, + { + "epoch": 0.36, + "learning_rate": 1.4720303538281972e-05, + "loss": 1.0293, + "step": 11358 + }, + { + "epoch": 0.36, + "learning_rate": 1.4719390366748323e-05, + "loss": 1.0806, + "step": 11359 + }, + { + "epoch": 0.36, + "learning_rate": 1.4718477144580838e-05, + "loss": 0.9146, + "step": 11360 + }, + { + "epoch": 0.36, + "learning_rate": 1.4717563871789318e-05, + "loss": 1.082, + "step": 11361 + }, + { + "epoch": 0.36, + "learning_rate": 1.4716650548383556e-05, + "loss": 1.1001, + "step": 11362 + }, + { + "epoch": 0.36, + "learning_rate": 1.4715737174373355e-05, + "loss": 0.8955, + "step": 11363 + }, + { + "epoch": 0.36, + "learning_rate": 1.4714823749768513e-05, + "loss": 1.106, + "step": 11364 + }, + { + "epoch": 0.36, + "learning_rate": 1.471391027457883e-05, + "loss": 0.9316, + "step": 11365 + }, + { + "epoch": 0.36, + "learning_rate": 1.4712996748814107e-05, + "loss": 1.0977, + "step": 11366 + }, + { + "epoch": 0.36, + "learning_rate": 1.4712083172484146e-05, + "loss": 0.9487, + "step": 11367 + }, + { + "epoch": 0.36, + "learning_rate": 1.4711169545598747e-05, + "loss": 1.0635, + "step": 11368 + }, + { + "epoch": 0.36, + "learning_rate": 1.4710255868167713e-05, + "loss": 1.1206, + "step": 11369 + }, + { + "epoch": 0.36, + "learning_rate": 1.4709342140200844e-05, + "loss": 1.1357, + "step": 11370 + }, + { + "epoch": 0.36, + "learning_rate": 1.4708428361707949e-05, + "loss": 1.0093, + "step": 11371 + }, + { + "epoch": 0.36, + "learning_rate": 1.4707514532698828e-05, + "loss": 1.0884, + "step": 11372 + }, + { + "epoch": 0.36, + "learning_rate": 1.4706600653183287e-05, + "loss": 0.9868, + "step": 11373 + }, + { + "epoch": 0.36, + "learning_rate": 1.4705686723171128e-05, + "loss": 1.0708, + "step": 11374 + }, + { + "epoch": 0.36, + "learning_rate": 1.470477274267216e-05, + "loss": 1.0396, + "step": 11375 + }, + { + "epoch": 0.36, + "learning_rate": 1.4703858711696186e-05, + "loss": 1.0469, + "step": 11376 + }, + { + "epoch": 0.36, + "learning_rate": 1.4702944630253016e-05, + "loss": 0.9976, + "step": 11377 + }, + { + "epoch": 0.36, + "learning_rate": 1.4702030498352452e-05, + "loss": 1.0957, + "step": 11378 + }, + { + "epoch": 0.36, + "learning_rate": 1.4701116316004307e-05, + "loss": 0.9722, + "step": 11379 + }, + { + "epoch": 0.36, + "learning_rate": 1.4700202083218384e-05, + "loss": 0.9766, + "step": 11380 + }, + { + "epoch": 0.36, + "learning_rate": 1.46992878000045e-05, + "loss": 0.9697, + "step": 11381 + }, + { + "epoch": 0.36, + "learning_rate": 1.4698373466372455e-05, + "loss": 1.0503, + "step": 11382 + }, + { + "epoch": 0.36, + "learning_rate": 1.4697459082332062e-05, + "loss": 1.0898, + "step": 11383 + }, + { + "epoch": 0.36, + "learning_rate": 1.4696544647893134e-05, + "loss": 0.9634, + "step": 11384 + }, + { + "epoch": 0.36, + "learning_rate": 1.469563016306548e-05, + "loss": 1.0493, + "step": 11385 + }, + { + "epoch": 0.36, + "learning_rate": 1.469471562785891e-05, + "loss": 0.9985, + "step": 11386 + }, + { + "epoch": 0.36, + "learning_rate": 1.4693801042283235e-05, + "loss": 1.061, + "step": 11387 + }, + { + "epoch": 0.36, + "learning_rate": 1.4692886406348272e-05, + "loss": 1.1357, + "step": 11388 + }, + { + "epoch": 0.36, + "learning_rate": 1.4691971720063833e-05, + "loss": 0.998, + "step": 11389 + }, + { + "epoch": 0.36, + "learning_rate": 1.4691056983439728e-05, + "loss": 1.0332, + "step": 11390 + }, + { + "epoch": 0.36, + "learning_rate": 1.4690142196485775e-05, + "loss": 1.0752, + "step": 11391 + }, + { + "epoch": 0.36, + "learning_rate": 1.4689227359211788e-05, + "loss": 1.0684, + "step": 11392 + }, + { + "epoch": 0.36, + "learning_rate": 1.468831247162758e-05, + "loss": 0.9556, + "step": 11393 + }, + { + "epoch": 0.36, + "learning_rate": 1.4687397533742966e-05, + "loss": 0.8892, + "step": 11394 + }, + { + "epoch": 0.36, + "learning_rate": 1.4686482545567767e-05, + "loss": 0.9692, + "step": 11395 + }, + { + "epoch": 0.36, + "learning_rate": 1.4685567507111795e-05, + "loss": 1.0674, + "step": 11396 + }, + { + "epoch": 0.36, + "learning_rate": 1.468465241838487e-05, + "loss": 1.0625, + "step": 11397 + }, + { + "epoch": 0.36, + "learning_rate": 1.468373727939681e-05, + "loss": 1.0322, + "step": 11398 + }, + { + "epoch": 0.36, + "learning_rate": 1.4682822090157431e-05, + "loss": 0.979, + "step": 11399 + }, + { + "epoch": 0.36, + "learning_rate": 1.4681906850676557e-05, + "loss": 0.8965, + "step": 11400 + }, + { + "epoch": 0.36, + "learning_rate": 1.4680991560964003e-05, + "loss": 1.1113, + "step": 11401 + }, + { + "epoch": 0.36, + "learning_rate": 1.4680076221029586e-05, + "loss": 1.082, + "step": 11402 + }, + { + "epoch": 0.36, + "learning_rate": 1.4679160830883135e-05, + "loss": 1.083, + "step": 11403 + }, + { + "epoch": 0.36, + "learning_rate": 1.4678245390534467e-05, + "loss": 0.9966, + "step": 11404 + }, + { + "epoch": 0.36, + "learning_rate": 1.4677329899993403e-05, + "loss": 1.0425, + "step": 11405 + }, + { + "epoch": 0.36, + "learning_rate": 1.4676414359269764e-05, + "loss": 1.1367, + "step": 11406 + }, + { + "epoch": 0.36, + "learning_rate": 1.4675498768373378e-05, + "loss": 0.9409, + "step": 11407 + }, + { + "epoch": 0.36, + "learning_rate": 1.4674583127314062e-05, + "loss": 1.0088, + "step": 11408 + }, + { + "epoch": 0.36, + "learning_rate": 1.4673667436101641e-05, + "loss": 1.0254, + "step": 11409 + }, + { + "epoch": 0.36, + "learning_rate": 1.4672751694745946e-05, + "loss": 1.0142, + "step": 11410 + }, + { + "epoch": 0.36, + "learning_rate": 1.4671835903256792e-05, + "loss": 0.9575, + "step": 11411 + }, + { + "epoch": 0.36, + "learning_rate": 1.4670920061644014e-05, + "loss": 0.958, + "step": 11412 + }, + { + "epoch": 0.36, + "learning_rate": 1.4670004169917431e-05, + "loss": 1.0884, + "step": 11413 + }, + { + "epoch": 0.37, + "learning_rate": 1.4669088228086869e-05, + "loss": 1.0381, + "step": 11414 + }, + { + "epoch": 0.37, + "learning_rate": 1.4668172236162162e-05, + "loss": 0.9966, + "step": 11415 + }, + { + "epoch": 0.37, + "learning_rate": 1.4667256194153132e-05, + "loss": 1.0586, + "step": 11416 + }, + { + "epoch": 0.37, + "learning_rate": 1.4666340102069608e-05, + "loss": 1.0278, + "step": 11417 + }, + { + "epoch": 0.37, + "learning_rate": 1.4665423959921419e-05, + "loss": 0.939, + "step": 11418 + }, + { + "epoch": 0.37, + "learning_rate": 1.4664507767718394e-05, + "loss": 0.9814, + "step": 11419 + }, + { + "epoch": 0.37, + "learning_rate": 1.4663591525470363e-05, + "loss": 1.1284, + "step": 11420 + }, + { + "epoch": 0.37, + "learning_rate": 1.4662675233187158e-05, + "loss": 1.168, + "step": 11421 + }, + { + "epoch": 0.37, + "learning_rate": 1.4661758890878606e-05, + "loss": 0.9653, + "step": 11422 + }, + { + "epoch": 0.37, + "learning_rate": 1.4660842498554543e-05, + "loss": 1.1299, + "step": 11423 + }, + { + "epoch": 0.37, + "learning_rate": 1.4659926056224798e-05, + "loss": 0.5503, + "step": 11424 + }, + { + "epoch": 0.37, + "learning_rate": 1.4659009563899202e-05, + "loss": 0.9351, + "step": 11425 + }, + { + "epoch": 0.37, + "learning_rate": 1.4658093021587592e-05, + "loss": 1.0986, + "step": 11426 + }, + { + "epoch": 0.37, + "learning_rate": 1.4657176429299798e-05, + "loss": 1.0146, + "step": 11427 + }, + { + "epoch": 0.37, + "learning_rate": 1.4656259787045652e-05, + "loss": 1.0176, + "step": 11428 + }, + { + "epoch": 0.37, + "learning_rate": 1.4655343094834996e-05, + "loss": 1.0928, + "step": 11429 + }, + { + "epoch": 0.37, + "learning_rate": 1.4654426352677659e-05, + "loss": 1.1421, + "step": 11430 + }, + { + "epoch": 0.37, + "learning_rate": 1.4653509560583478e-05, + "loss": 1.0117, + "step": 11431 + }, + { + "epoch": 0.37, + "learning_rate": 1.465259271856229e-05, + "loss": 1.0801, + "step": 11432 + }, + { + "epoch": 0.37, + "learning_rate": 1.4651675826623933e-05, + "loss": 1.0156, + "step": 11433 + }, + { + "epoch": 0.37, + "learning_rate": 1.4650758884778242e-05, + "loss": 1.0771, + "step": 11434 + }, + { + "epoch": 0.37, + "learning_rate": 1.4649841893035055e-05, + "loss": 1.0967, + "step": 11435 + }, + { + "epoch": 0.37, + "learning_rate": 1.4648924851404212e-05, + "loss": 0.8916, + "step": 11436 + }, + { + "epoch": 0.37, + "learning_rate": 1.4648007759895548e-05, + "loss": 1.0293, + "step": 11437 + }, + { + "epoch": 0.37, + "learning_rate": 1.4647090618518906e-05, + "loss": 1.1582, + "step": 11438 + }, + { + "epoch": 0.37, + "learning_rate": 1.4646173427284125e-05, + "loss": 1.1021, + "step": 11439 + }, + { + "epoch": 0.37, + "learning_rate": 1.4645256186201045e-05, + "loss": 0.9492, + "step": 11440 + }, + { + "epoch": 0.37, + "learning_rate": 1.4644338895279509e-05, + "loss": 0.96, + "step": 11441 + }, + { + "epoch": 0.37, + "learning_rate": 1.4643421554529354e-05, + "loss": 1.0288, + "step": 11442 + }, + { + "epoch": 0.37, + "learning_rate": 1.4642504163960425e-05, + "loss": 0.9775, + "step": 11443 + }, + { + "epoch": 0.37, + "learning_rate": 1.4641586723582566e-05, + "loss": 1.1226, + "step": 11444 + }, + { + "epoch": 0.37, + "learning_rate": 1.464066923340562e-05, + "loss": 0.5664, + "step": 11445 + }, + { + "epoch": 0.37, + "learning_rate": 1.4639751693439425e-05, + "loss": 0.9019, + "step": 11446 + }, + { + "epoch": 0.37, + "learning_rate": 1.463883410369383e-05, + "loss": 1.0776, + "step": 11447 + }, + { + "epoch": 0.37, + "learning_rate": 1.4637916464178683e-05, + "loss": 0.9849, + "step": 11448 + }, + { + "epoch": 0.37, + "learning_rate": 1.4636998774903823e-05, + "loss": 1.0488, + "step": 11449 + }, + { + "epoch": 0.37, + "learning_rate": 1.4636081035879098e-05, + "loss": 1.1641, + "step": 11450 + }, + { + "epoch": 0.37, + "learning_rate": 1.4635163247114355e-05, + "loss": 0.9932, + "step": 11451 + }, + { + "epoch": 0.37, + "learning_rate": 1.4634245408619439e-05, + "loss": 1.0376, + "step": 11452 + }, + { + "epoch": 0.37, + "learning_rate": 1.46333275204042e-05, + "loss": 0.999, + "step": 11453 + }, + { + "epoch": 0.37, + "learning_rate": 1.4632409582478485e-05, + "loss": 1.0044, + "step": 11454 + }, + { + "epoch": 0.37, + "learning_rate": 1.463149159485214e-05, + "loss": 1.1152, + "step": 11455 + }, + { + "epoch": 0.37, + "learning_rate": 1.4630573557535017e-05, + "loss": 1.0991, + "step": 11456 + }, + { + "epoch": 0.37, + "learning_rate": 1.4629655470536965e-05, + "loss": 1.0205, + "step": 11457 + }, + { + "epoch": 0.37, + "learning_rate": 1.4628737333867835e-05, + "loss": 1.0786, + "step": 11458 + }, + { + "epoch": 0.37, + "learning_rate": 1.4627819147537472e-05, + "loss": 1.0269, + "step": 11459 + }, + { + "epoch": 0.37, + "learning_rate": 1.4626900911555737e-05, + "loss": 1.0981, + "step": 11460 + }, + { + "epoch": 0.37, + "learning_rate": 1.4625982625932472e-05, + "loss": 1.0718, + "step": 11461 + }, + { + "epoch": 0.37, + "learning_rate": 1.4625064290677533e-05, + "loss": 1.0688, + "step": 11462 + }, + { + "epoch": 0.37, + "learning_rate": 1.4624145905800773e-05, + "loss": 1.106, + "step": 11463 + }, + { + "epoch": 0.37, + "learning_rate": 1.4623227471312049e-05, + "loss": 1.0513, + "step": 11464 + }, + { + "epoch": 0.37, + "learning_rate": 1.4622308987221207e-05, + "loss": 1.0635, + "step": 11465 + }, + { + "epoch": 0.37, + "learning_rate": 1.4621390453538108e-05, + "loss": 1.0332, + "step": 11466 + }, + { + "epoch": 0.37, + "learning_rate": 1.4620471870272601e-05, + "loss": 1.0, + "step": 11467 + }, + { + "epoch": 0.37, + "learning_rate": 1.4619553237434548e-05, + "loss": 1.0405, + "step": 11468 + }, + { + "epoch": 0.37, + "learning_rate": 1.46186345550338e-05, + "loss": 1.0522, + "step": 11469 + }, + { + "epoch": 0.37, + "learning_rate": 1.4617715823080212e-05, + "loss": 1.0078, + "step": 11470 + }, + { + "epoch": 0.37, + "learning_rate": 1.4616797041583646e-05, + "loss": 0.8574, + "step": 11471 + }, + { + "epoch": 0.37, + "learning_rate": 1.461587821055396e-05, + "loss": 1.0664, + "step": 11472 + }, + { + "epoch": 0.37, + "learning_rate": 1.4614959330001007e-05, + "loss": 1.0, + "step": 11473 + }, + { + "epoch": 0.37, + "learning_rate": 1.4614040399934647e-05, + "loss": 0.998, + "step": 11474 + }, + { + "epoch": 0.37, + "learning_rate": 1.4613121420364742e-05, + "loss": 0.9717, + "step": 11475 + }, + { + "epoch": 0.37, + "learning_rate": 1.4612202391301149e-05, + "loss": 1.0483, + "step": 11476 + }, + { + "epoch": 0.37, + "learning_rate": 1.4611283312753728e-05, + "loss": 0.9834, + "step": 11477 + }, + { + "epoch": 0.37, + "learning_rate": 1.4610364184732343e-05, + "loss": 0.8789, + "step": 11478 + }, + { + "epoch": 0.37, + "learning_rate": 1.4609445007246849e-05, + "loss": 1.0459, + "step": 11479 + }, + { + "epoch": 0.37, + "learning_rate": 1.4608525780307115e-05, + "loss": 0.9561, + "step": 11480 + }, + { + "epoch": 0.37, + "learning_rate": 1.4607606503922998e-05, + "loss": 1.0571, + "step": 11481 + }, + { + "epoch": 0.37, + "learning_rate": 1.4606687178104362e-05, + "loss": 1.1245, + "step": 11482 + }, + { + "epoch": 0.37, + "learning_rate": 1.4605767802861075e-05, + "loss": 1.0093, + "step": 11483 + }, + { + "epoch": 0.37, + "learning_rate": 1.4604848378202992e-05, + "loss": 1.0132, + "step": 11484 + }, + { + "epoch": 0.37, + "learning_rate": 1.4603928904139984e-05, + "loss": 1.0083, + "step": 11485 + }, + { + "epoch": 0.37, + "learning_rate": 1.4603009380681914e-05, + "loss": 1.0073, + "step": 11486 + }, + { + "epoch": 0.37, + "learning_rate": 1.4602089807838651e-05, + "loss": 1.0557, + "step": 11487 + }, + { + "epoch": 0.37, + "learning_rate": 1.4601170185620053e-05, + "loss": 0.9844, + "step": 11488 + }, + { + "epoch": 0.37, + "learning_rate": 1.4600250514035994e-05, + "loss": 1.0278, + "step": 11489 + }, + { + "epoch": 0.37, + "learning_rate": 1.4599330793096338e-05, + "loss": 1.0454, + "step": 11490 + }, + { + "epoch": 0.37, + "learning_rate": 1.4598411022810952e-05, + "loss": 0.9868, + "step": 11491 + }, + { + "epoch": 0.37, + "learning_rate": 1.4597491203189705e-05, + "loss": 0.918, + "step": 11492 + }, + { + "epoch": 0.37, + "learning_rate": 1.4596571334242468e-05, + "loss": 1.0269, + "step": 11493 + }, + { + "epoch": 0.37, + "learning_rate": 1.4595651415979104e-05, + "loss": 1.0547, + "step": 11494 + }, + { + "epoch": 0.37, + "learning_rate": 1.4594731448409489e-05, + "loss": 1.1221, + "step": 11495 + }, + { + "epoch": 0.37, + "learning_rate": 1.4593811431543492e-05, + "loss": 1.0122, + "step": 11496 + }, + { + "epoch": 0.37, + "learning_rate": 1.459289136539098e-05, + "loss": 1.0796, + "step": 11497 + }, + { + "epoch": 0.37, + "learning_rate": 1.4591971249961828e-05, + "loss": 1.0903, + "step": 11498 + }, + { + "epoch": 0.37, + "learning_rate": 1.4591051085265907e-05, + "loss": 1.0171, + "step": 11499 + }, + { + "epoch": 0.37, + "learning_rate": 1.4590130871313086e-05, + "loss": 1.0356, + "step": 11500 + }, + { + "epoch": 0.37, + "learning_rate": 1.4589210608113242e-05, + "loss": 0.9873, + "step": 11501 + }, + { + "epoch": 0.37, + "learning_rate": 1.458829029567625e-05, + "loss": 0.9185, + "step": 11502 + }, + { + "epoch": 0.37, + "learning_rate": 1.458736993401198e-05, + "loss": 1.0981, + "step": 11503 + }, + { + "epoch": 0.37, + "learning_rate": 1.4586449523130307e-05, + "loss": 1.0449, + "step": 11504 + }, + { + "epoch": 0.37, + "learning_rate": 1.4585529063041107e-05, + "loss": 0.9487, + "step": 11505 + }, + { + "epoch": 0.37, + "learning_rate": 1.4584608553754257e-05, + "loss": 1.106, + "step": 11506 + }, + { + "epoch": 0.37, + "learning_rate": 1.4583687995279627e-05, + "loss": 1.0493, + "step": 11507 + }, + { + "epoch": 0.37, + "learning_rate": 1.45827673876271e-05, + "loss": 1.104, + "step": 11508 + }, + { + "epoch": 0.37, + "learning_rate": 1.458184673080655e-05, + "loss": 1.0171, + "step": 11509 + }, + { + "epoch": 0.37, + "learning_rate": 1.4580926024827858e-05, + "loss": 1.0454, + "step": 11510 + }, + { + "epoch": 0.37, + "learning_rate": 1.4580005269700896e-05, + "loss": 0.8682, + "step": 11511 + }, + { + "epoch": 0.37, + "learning_rate": 1.4579084465435548e-05, + "loss": 0.9463, + "step": 11512 + }, + { + "epoch": 0.37, + "learning_rate": 1.4578163612041694e-05, + "loss": 1.1387, + "step": 11513 + }, + { + "epoch": 0.37, + "learning_rate": 1.457724270952921e-05, + "loss": 1.0327, + "step": 11514 + }, + { + "epoch": 0.37, + "learning_rate": 1.4576321757907973e-05, + "loss": 1.0649, + "step": 11515 + }, + { + "epoch": 0.37, + "learning_rate": 1.4575400757187872e-05, + "loss": 1.0337, + "step": 11516 + }, + { + "epoch": 0.37, + "learning_rate": 1.4574479707378788e-05, + "loss": 1.104, + "step": 11517 + }, + { + "epoch": 0.37, + "learning_rate": 1.4573558608490594e-05, + "loss": 1.061, + "step": 11518 + }, + { + "epoch": 0.37, + "learning_rate": 1.4572637460533182e-05, + "loss": 1.0322, + "step": 11519 + }, + { + "epoch": 0.37, + "learning_rate": 1.4571716263516426e-05, + "loss": 1.0757, + "step": 11520 + }, + { + "epoch": 0.37, + "learning_rate": 1.457079501745022e-05, + "loss": 1.0479, + "step": 11521 + }, + { + "epoch": 0.37, + "learning_rate": 1.456987372234444e-05, + "loss": 0.9673, + "step": 11522 + }, + { + "epoch": 0.37, + "learning_rate": 1.4568952378208972e-05, + "loss": 0.9302, + "step": 11523 + }, + { + "epoch": 0.37, + "learning_rate": 1.45680309850537e-05, + "loss": 1.085, + "step": 11524 + }, + { + "epoch": 0.37, + "learning_rate": 1.4567109542888513e-05, + "loss": 1.0005, + "step": 11525 + }, + { + "epoch": 0.37, + "learning_rate": 1.4566188051723296e-05, + "loss": 0.9995, + "step": 11526 + }, + { + "epoch": 0.37, + "learning_rate": 1.4565266511567933e-05, + "loss": 0.9673, + "step": 11527 + }, + { + "epoch": 0.37, + "learning_rate": 1.4564344922432314e-05, + "loss": 0.9805, + "step": 11528 + }, + { + "epoch": 0.37, + "learning_rate": 1.4563423284326326e-05, + "loss": 1.0981, + "step": 11529 + }, + { + "epoch": 0.37, + "learning_rate": 1.4562501597259854e-05, + "loss": 0.9243, + "step": 11530 + }, + { + "epoch": 0.37, + "learning_rate": 1.4561579861242791e-05, + "loss": 1.1465, + "step": 11531 + }, + { + "epoch": 0.37, + "learning_rate": 1.4560658076285028e-05, + "loss": 0.9702, + "step": 11532 + }, + { + "epoch": 0.37, + "learning_rate": 1.4559736242396448e-05, + "loss": 1.0054, + "step": 11533 + }, + { + "epoch": 0.37, + "learning_rate": 1.4558814359586943e-05, + "loss": 1.0796, + "step": 11534 + }, + { + "epoch": 0.37, + "learning_rate": 1.455789242786641e-05, + "loss": 0.9805, + "step": 11535 + }, + { + "epoch": 0.37, + "learning_rate": 1.4556970447244732e-05, + "loss": 1.085, + "step": 11536 + }, + { + "epoch": 0.37, + "learning_rate": 1.4556048417731805e-05, + "loss": 1.2217, + "step": 11537 + }, + { + "epoch": 0.37, + "learning_rate": 1.4555126339337525e-05, + "loss": 0.9609, + "step": 11538 + }, + { + "epoch": 0.37, + "learning_rate": 1.4554204212071774e-05, + "loss": 0.9121, + "step": 11539 + }, + { + "epoch": 0.37, + "learning_rate": 1.4553282035944456e-05, + "loss": 1.0078, + "step": 11540 + }, + { + "epoch": 0.37, + "learning_rate": 1.455235981096546e-05, + "loss": 0.9639, + "step": 11541 + }, + { + "epoch": 0.37, + "learning_rate": 1.4551437537144682e-05, + "loss": 0.9878, + "step": 11542 + }, + { + "epoch": 0.37, + "learning_rate": 1.4550515214492019e-05, + "loss": 0.9141, + "step": 11543 + }, + { + "epoch": 0.37, + "learning_rate": 1.4549592843017359e-05, + "loss": 1.0024, + "step": 11544 + }, + { + "epoch": 0.37, + "learning_rate": 1.4548670422730606e-05, + "loss": 1.0615, + "step": 11545 + }, + { + "epoch": 0.37, + "learning_rate": 1.4547747953641654e-05, + "loss": 1.2041, + "step": 11546 + }, + { + "epoch": 0.37, + "learning_rate": 1.4546825435760399e-05, + "loss": 1.0166, + "step": 11547 + }, + { + "epoch": 0.37, + "learning_rate": 1.4545902869096737e-05, + "loss": 1.0044, + "step": 11548 + }, + { + "epoch": 0.37, + "learning_rate": 1.454498025366057e-05, + "loss": 1.0454, + "step": 11549 + }, + { + "epoch": 0.37, + "learning_rate": 1.4544057589461796e-05, + "loss": 0.5823, + "step": 11550 + }, + { + "epoch": 0.37, + "learning_rate": 1.4543134876510312e-05, + "loss": 1.1753, + "step": 11551 + }, + { + "epoch": 0.37, + "learning_rate": 1.4542212114816021e-05, + "loss": 1.0264, + "step": 11552 + }, + { + "epoch": 0.37, + "learning_rate": 1.4541289304388819e-05, + "loss": 1.022, + "step": 11553 + }, + { + "epoch": 0.37, + "learning_rate": 1.4540366445238607e-05, + "loss": 0.9702, + "step": 11554 + }, + { + "epoch": 0.37, + "learning_rate": 1.4539443537375292e-05, + "loss": 0.9741, + "step": 11555 + }, + { + "epoch": 0.37, + "learning_rate": 1.453852058080877e-05, + "loss": 1.0913, + "step": 11556 + }, + { + "epoch": 0.37, + "learning_rate": 1.4537597575548945e-05, + "loss": 1.0449, + "step": 11557 + }, + { + "epoch": 0.37, + "learning_rate": 1.4536674521605721e-05, + "loss": 0.9795, + "step": 11558 + }, + { + "epoch": 0.37, + "learning_rate": 1.4535751418989e-05, + "loss": 1.1523, + "step": 11559 + }, + { + "epoch": 0.37, + "learning_rate": 1.4534828267708685e-05, + "loss": 1.0615, + "step": 11560 + }, + { + "epoch": 0.37, + "learning_rate": 1.4533905067774683e-05, + "loss": 1.0762, + "step": 11561 + }, + { + "epoch": 0.37, + "learning_rate": 1.4532981819196896e-05, + "loss": 1.145, + "step": 11562 + }, + { + "epoch": 0.37, + "learning_rate": 1.4532058521985232e-05, + "loss": 1.0903, + "step": 11563 + }, + { + "epoch": 0.37, + "learning_rate": 1.4531135176149596e-05, + "loss": 1.0112, + "step": 11564 + }, + { + "epoch": 0.37, + "learning_rate": 1.4530211781699892e-05, + "loss": 0.9658, + "step": 11565 + }, + { + "epoch": 0.37, + "learning_rate": 1.4529288338646034e-05, + "loss": 1.043, + "step": 11566 + }, + { + "epoch": 0.37, + "learning_rate": 1.4528364846997922e-05, + "loss": 0.9131, + "step": 11567 + }, + { + "epoch": 0.37, + "learning_rate": 1.4527441306765467e-05, + "loss": 0.9482, + "step": 11568 + }, + { + "epoch": 0.37, + "learning_rate": 1.4526517717958577e-05, + "loss": 0.9331, + "step": 11569 + }, + { + "epoch": 0.37, + "learning_rate": 1.4525594080587164e-05, + "loss": 0.8604, + "step": 11570 + }, + { + "epoch": 0.37, + "learning_rate": 1.4524670394661134e-05, + "loss": 1.2021, + "step": 11571 + }, + { + "epoch": 0.37, + "learning_rate": 1.4523746660190397e-05, + "loss": 0.9355, + "step": 11572 + }, + { + "epoch": 0.37, + "learning_rate": 1.4522822877184868e-05, + "loss": 1.1108, + "step": 11573 + }, + { + "epoch": 0.37, + "learning_rate": 1.452189904565445e-05, + "loss": 0.9956, + "step": 11574 + }, + { + "epoch": 0.37, + "learning_rate": 1.4520975165609063e-05, + "loss": 0.9927, + "step": 11575 + }, + { + "epoch": 0.37, + "learning_rate": 1.4520051237058614e-05, + "loss": 1.0762, + "step": 11576 + }, + { + "epoch": 0.37, + "learning_rate": 1.4519127260013025e-05, + "loss": 1.1719, + "step": 11577 + }, + { + "epoch": 0.37, + "learning_rate": 1.4518203234482194e-05, + "loss": 0.9585, + "step": 11578 + }, + { + "epoch": 0.37, + "learning_rate": 1.4517279160476045e-05, + "loss": 1.1006, + "step": 11579 + }, + { + "epoch": 0.37, + "learning_rate": 1.4516355038004492e-05, + "loss": 1.0244, + "step": 11580 + }, + { + "epoch": 0.37, + "learning_rate": 1.4515430867077447e-05, + "loss": 0.918, + "step": 11581 + }, + { + "epoch": 0.37, + "learning_rate": 1.4514506647704827e-05, + "loss": 1.0581, + "step": 11582 + }, + { + "epoch": 0.37, + "learning_rate": 1.4513582379896545e-05, + "loss": 1.0151, + "step": 11583 + }, + { + "epoch": 0.37, + "learning_rate": 1.451265806366252e-05, + "loss": 1.1182, + "step": 11584 + }, + { + "epoch": 0.37, + "learning_rate": 1.4511733699012668e-05, + "loss": 1.1953, + "step": 11585 + }, + { + "epoch": 0.37, + "learning_rate": 1.4510809285956908e-05, + "loss": 1.0371, + "step": 11586 + }, + { + "epoch": 0.37, + "learning_rate": 1.4509884824505153e-05, + "loss": 0.9238, + "step": 11587 + }, + { + "epoch": 0.37, + "learning_rate": 1.450896031466733e-05, + "loss": 0.9517, + "step": 11588 + }, + { + "epoch": 0.37, + "learning_rate": 1.4508035756453351e-05, + "loss": 1.0688, + "step": 11589 + }, + { + "epoch": 0.37, + "learning_rate": 1.450711114987314e-05, + "loss": 1.0317, + "step": 11590 + }, + { + "epoch": 0.37, + "learning_rate": 1.4506186494936611e-05, + "loss": 0.9917, + "step": 11591 + }, + { + "epoch": 0.37, + "learning_rate": 1.450526179165369e-05, + "loss": 1.0425, + "step": 11592 + }, + { + "epoch": 0.37, + "learning_rate": 1.4504337040034295e-05, + "loss": 1.1255, + "step": 11593 + }, + { + "epoch": 0.37, + "learning_rate": 1.450341224008835e-05, + "loss": 0.9888, + "step": 11594 + }, + { + "epoch": 0.37, + "learning_rate": 1.4502487391825773e-05, + "loss": 1.1045, + "step": 11595 + }, + { + "epoch": 0.37, + "learning_rate": 1.4501562495256494e-05, + "loss": 1.104, + "step": 11596 + }, + { + "epoch": 0.37, + "learning_rate": 1.4500637550390428e-05, + "loss": 1.0244, + "step": 11597 + }, + { + "epoch": 0.37, + "learning_rate": 1.4499712557237505e-05, + "loss": 0.9395, + "step": 11598 + }, + { + "epoch": 0.37, + "learning_rate": 1.4498787515807642e-05, + "loss": 1.0522, + "step": 11599 + }, + { + "epoch": 0.37, + "learning_rate": 1.4497862426110771e-05, + "loss": 1.1147, + "step": 11600 + }, + { + "epoch": 0.37, + "learning_rate": 1.4496937288156814e-05, + "loss": 1.0029, + "step": 11601 + }, + { + "epoch": 0.37, + "learning_rate": 1.4496012101955697e-05, + "loss": 1.0605, + "step": 11602 + }, + { + "epoch": 0.37, + "learning_rate": 1.4495086867517344e-05, + "loss": 0.5781, + "step": 11603 + }, + { + "epoch": 0.37, + "learning_rate": 1.4494161584851687e-05, + "loss": 1.0327, + "step": 11604 + }, + { + "epoch": 0.37, + "learning_rate": 1.4493236253968647e-05, + "loss": 1.0278, + "step": 11605 + }, + { + "epoch": 0.37, + "learning_rate": 1.4492310874878157e-05, + "loss": 1.1274, + "step": 11606 + }, + { + "epoch": 0.37, + "learning_rate": 1.4491385447590143e-05, + "loss": 1.0415, + "step": 11607 + }, + { + "epoch": 0.37, + "learning_rate": 1.4490459972114534e-05, + "loss": 1.124, + "step": 11608 + }, + { + "epoch": 0.37, + "learning_rate": 1.4489534448461257e-05, + "loss": 1.043, + "step": 11609 + }, + { + "epoch": 0.37, + "learning_rate": 1.4488608876640244e-05, + "loss": 0.917, + "step": 11610 + }, + { + "epoch": 0.37, + "learning_rate": 1.448768325666143e-05, + "loss": 0.9829, + "step": 11611 + }, + { + "epoch": 0.37, + "learning_rate": 1.4486757588534739e-05, + "loss": 0.9736, + "step": 11612 + }, + { + "epoch": 0.37, + "learning_rate": 1.4485831872270103e-05, + "loss": 1.1113, + "step": 11613 + }, + { + "epoch": 0.37, + "learning_rate": 1.4484906107877456e-05, + "loss": 1.0493, + "step": 11614 + }, + { + "epoch": 0.37, + "learning_rate": 1.4483980295366732e-05, + "loss": 1.0972, + "step": 11615 + }, + { + "epoch": 0.37, + "learning_rate": 1.4483054434747862e-05, + "loss": 0.9453, + "step": 11616 + }, + { + "epoch": 0.37, + "learning_rate": 1.4482128526030779e-05, + "loss": 1.0293, + "step": 11617 + }, + { + "epoch": 0.37, + "learning_rate": 1.4481202569225417e-05, + "loss": 0.8794, + "step": 11618 + }, + { + "epoch": 0.37, + "learning_rate": 1.4480276564341714e-05, + "loss": 1.1777, + "step": 11619 + }, + { + "epoch": 0.37, + "learning_rate": 1.4479350511389598e-05, + "loss": 1.1309, + "step": 11620 + }, + { + "epoch": 0.37, + "learning_rate": 1.447842441037901e-05, + "loss": 0.9614, + "step": 11621 + }, + { + "epoch": 0.37, + "learning_rate": 1.4477498261319884e-05, + "loss": 1.0527, + "step": 11622 + }, + { + "epoch": 0.37, + "learning_rate": 1.4476572064222158e-05, + "loss": 1.0225, + "step": 11623 + }, + { + "epoch": 0.37, + "learning_rate": 1.447564581909577e-05, + "loss": 1.0874, + "step": 11624 + }, + { + "epoch": 0.37, + "learning_rate": 1.4474719525950654e-05, + "loss": 0.9419, + "step": 11625 + }, + { + "epoch": 0.37, + "learning_rate": 1.4473793184796747e-05, + "loss": 1.0688, + "step": 11626 + }, + { + "epoch": 0.37, + "learning_rate": 1.4472866795643994e-05, + "loss": 0.9004, + "step": 11627 + }, + { + "epoch": 0.37, + "learning_rate": 1.4471940358502328e-05, + "loss": 0.9771, + "step": 11628 + }, + { + "epoch": 0.37, + "learning_rate": 1.4471013873381692e-05, + "loss": 0.8901, + "step": 11629 + }, + { + "epoch": 0.37, + "learning_rate": 1.4470087340292029e-05, + "loss": 0.979, + "step": 11630 + }, + { + "epoch": 0.37, + "learning_rate": 1.4469160759243272e-05, + "loss": 1.0562, + "step": 11631 + }, + { + "epoch": 0.37, + "learning_rate": 1.4468234130245367e-05, + "loss": 1.0454, + "step": 11632 + }, + { + "epoch": 0.37, + "learning_rate": 1.4467307453308254e-05, + "loss": 1.0093, + "step": 11633 + }, + { + "epoch": 0.37, + "learning_rate": 1.446638072844188e-05, + "loss": 1.1318, + "step": 11634 + }, + { + "epoch": 0.37, + "learning_rate": 1.4465453955656182e-05, + "loss": 1.0215, + "step": 11635 + }, + { + "epoch": 0.37, + "learning_rate": 1.4464527134961104e-05, + "loss": 1.0093, + "step": 11636 + }, + { + "epoch": 0.37, + "learning_rate": 1.4463600266366591e-05, + "loss": 0.9297, + "step": 11637 + }, + { + "epoch": 0.37, + "learning_rate": 1.4462673349882587e-05, + "loss": 1.0942, + "step": 11638 + }, + { + "epoch": 0.37, + "learning_rate": 1.446174638551904e-05, + "loss": 0.9985, + "step": 11639 + }, + { + "epoch": 0.37, + "learning_rate": 1.4460819373285888e-05, + "loss": 1.0083, + "step": 11640 + }, + { + "epoch": 0.37, + "learning_rate": 1.4459892313193085e-05, + "loss": 1.1025, + "step": 11641 + }, + { + "epoch": 0.37, + "learning_rate": 1.4458965205250572e-05, + "loss": 1.1021, + "step": 11642 + }, + { + "epoch": 0.37, + "learning_rate": 1.4458038049468295e-05, + "loss": 1.0479, + "step": 11643 + }, + { + "epoch": 0.37, + "learning_rate": 1.4457110845856207e-05, + "loss": 1.0215, + "step": 11644 + }, + { + "epoch": 0.37, + "learning_rate": 1.4456183594424252e-05, + "loss": 1.1211, + "step": 11645 + }, + { + "epoch": 0.37, + "learning_rate": 1.4455256295182376e-05, + "loss": 0.9121, + "step": 11646 + }, + { + "epoch": 0.37, + "learning_rate": 1.4454328948140534e-05, + "loss": 0.9219, + "step": 11647 + }, + { + "epoch": 0.37, + "learning_rate": 1.445340155330867e-05, + "loss": 1.1328, + "step": 11648 + }, + { + "epoch": 0.37, + "learning_rate": 1.4452474110696738e-05, + "loss": 0.8926, + "step": 11649 + }, + { + "epoch": 0.37, + "learning_rate": 1.4451546620314685e-05, + "loss": 0.9062, + "step": 11650 + }, + { + "epoch": 0.37, + "learning_rate": 1.4450619082172464e-05, + "loss": 0.9727, + "step": 11651 + }, + { + "epoch": 0.37, + "learning_rate": 1.4449691496280024e-05, + "loss": 1.0737, + "step": 11652 + }, + { + "epoch": 0.37, + "learning_rate": 1.4448763862647324e-05, + "loss": 0.9683, + "step": 11653 + }, + { + "epoch": 0.37, + "learning_rate": 1.444783618128431e-05, + "loss": 1.0093, + "step": 11654 + }, + { + "epoch": 0.37, + "learning_rate": 1.4446908452200932e-05, + "loss": 1.0889, + "step": 11655 + }, + { + "epoch": 0.37, + "learning_rate": 1.4445980675407149e-05, + "loss": 0.9639, + "step": 11656 + }, + { + "epoch": 0.37, + "learning_rate": 1.444505285091292e-05, + "loss": 1.0454, + "step": 11657 + }, + { + "epoch": 0.37, + "learning_rate": 1.4444124978728187e-05, + "loss": 1.1216, + "step": 11658 + }, + { + "epoch": 0.37, + "learning_rate": 1.4443197058862914e-05, + "loss": 1.0244, + "step": 11659 + }, + { + "epoch": 0.37, + "learning_rate": 1.4442269091327054e-05, + "loss": 1.0405, + "step": 11660 + }, + { + "epoch": 0.37, + "learning_rate": 1.4441341076130565e-05, + "loss": 0.9258, + "step": 11661 + }, + { + "epoch": 0.37, + "learning_rate": 1.4440413013283396e-05, + "loss": 1.0234, + "step": 11662 + }, + { + "epoch": 0.37, + "learning_rate": 1.4439484902795516e-05, + "loss": 0.9189, + "step": 11663 + }, + { + "epoch": 0.37, + "learning_rate": 1.4438556744676873e-05, + "loss": 1.104, + "step": 11664 + }, + { + "epoch": 0.37, + "learning_rate": 1.4437628538937427e-05, + "loss": 1.0278, + "step": 11665 + }, + { + "epoch": 0.37, + "learning_rate": 1.4436700285587141e-05, + "loss": 0.5376, + "step": 11666 + }, + { + "epoch": 0.37, + "learning_rate": 1.4435771984635971e-05, + "loss": 1.0415, + "step": 11667 + }, + { + "epoch": 0.37, + "learning_rate": 1.4434843636093878e-05, + "loss": 1.0713, + "step": 11668 + }, + { + "epoch": 0.37, + "learning_rate": 1.4433915239970819e-05, + "loss": 0.9551, + "step": 11669 + }, + { + "epoch": 0.37, + "learning_rate": 1.4432986796276755e-05, + "loss": 0.5569, + "step": 11670 + }, + { + "epoch": 0.37, + "learning_rate": 1.4432058305021652e-05, + "loss": 1.0039, + "step": 11671 + }, + { + "epoch": 0.37, + "learning_rate": 1.4431129766215468e-05, + "loss": 1.0146, + "step": 11672 + }, + { + "epoch": 0.37, + "learning_rate": 1.4430201179868167e-05, + "loss": 0.9922, + "step": 11673 + }, + { + "epoch": 0.37, + "learning_rate": 1.4429272545989705e-05, + "loss": 1.1245, + "step": 11674 + }, + { + "epoch": 0.37, + "learning_rate": 1.4428343864590058e-05, + "loss": 0.9858, + "step": 11675 + }, + { + "epoch": 0.37, + "learning_rate": 1.4427415135679177e-05, + "loss": 1.0039, + "step": 11676 + }, + { + "epoch": 0.37, + "learning_rate": 1.442648635926703e-05, + "loss": 0.8672, + "step": 11677 + }, + { + "epoch": 0.37, + "learning_rate": 1.4425557535363588e-05, + "loss": 0.9268, + "step": 11678 + }, + { + "epoch": 0.37, + "learning_rate": 1.4424628663978813e-05, + "loss": 1.1187, + "step": 11679 + }, + { + "epoch": 0.37, + "learning_rate": 1.4423699745122666e-05, + "loss": 0.9609, + "step": 11680 + }, + { + "epoch": 0.37, + "learning_rate": 1.4422770778805116e-05, + "loss": 0.9727, + "step": 11681 + }, + { + "epoch": 0.37, + "learning_rate": 1.4421841765036131e-05, + "loss": 1.0391, + "step": 11682 + }, + { + "epoch": 0.37, + "learning_rate": 1.442091270382568e-05, + "loss": 1.0884, + "step": 11683 + }, + { + "epoch": 0.37, + "learning_rate": 1.4419983595183726e-05, + "loss": 0.96, + "step": 11684 + }, + { + "epoch": 0.37, + "learning_rate": 1.441905443912024e-05, + "loss": 0.9434, + "step": 11685 + }, + { + "epoch": 0.37, + "learning_rate": 1.4418125235645191e-05, + "loss": 1.0039, + "step": 11686 + }, + { + "epoch": 0.37, + "learning_rate": 1.441719598476855e-05, + "loss": 1.0962, + "step": 11687 + }, + { + "epoch": 0.37, + "learning_rate": 1.4416266686500281e-05, + "loss": 0.9697, + "step": 11688 + }, + { + "epoch": 0.37, + "learning_rate": 1.4415337340850361e-05, + "loss": 1.0176, + "step": 11689 + }, + { + "epoch": 0.37, + "learning_rate": 1.441440794782876e-05, + "loss": 1.0439, + "step": 11690 + }, + { + "epoch": 0.37, + "learning_rate": 1.4413478507445441e-05, + "loss": 1.1836, + "step": 11691 + }, + { + "epoch": 0.37, + "learning_rate": 1.4412549019710388e-05, + "loss": 1.0166, + "step": 11692 + }, + { + "epoch": 0.37, + "learning_rate": 1.4411619484633563e-05, + "loss": 1.062, + "step": 11693 + }, + { + "epoch": 0.37, + "learning_rate": 1.4410689902224947e-05, + "loss": 0.9907, + "step": 11694 + }, + { + "epoch": 0.37, + "learning_rate": 1.4409760272494508e-05, + "loss": 0.9595, + "step": 11695 + }, + { + "epoch": 0.37, + "learning_rate": 1.4408830595452223e-05, + "loss": 0.9849, + "step": 11696 + }, + { + "epoch": 0.37, + "learning_rate": 1.4407900871108064e-05, + "loss": 1.0366, + "step": 11697 + }, + { + "epoch": 0.37, + "learning_rate": 1.4406971099472009e-05, + "loss": 1.1504, + "step": 11698 + }, + { + "epoch": 0.37, + "learning_rate": 1.440604128055403e-05, + "loss": 1.0664, + "step": 11699 + }, + { + "epoch": 0.37, + "learning_rate": 1.4405111414364104e-05, + "loss": 0.9814, + "step": 11700 + }, + { + "epoch": 0.37, + "learning_rate": 1.4404181500912207e-05, + "loss": 0.9268, + "step": 11701 + }, + { + "epoch": 0.37, + "learning_rate": 1.440325154020832e-05, + "loss": 0.9668, + "step": 11702 + }, + { + "epoch": 0.37, + "learning_rate": 1.4402321532262416e-05, + "loss": 1.0977, + "step": 11703 + }, + { + "epoch": 0.37, + "learning_rate": 1.4401391477084471e-05, + "loss": 0.9888, + "step": 11704 + }, + { + "epoch": 0.37, + "learning_rate": 1.440046137468447e-05, + "loss": 0.978, + "step": 11705 + }, + { + "epoch": 0.37, + "learning_rate": 1.439953122507239e-05, + "loss": 0.9702, + "step": 11706 + }, + { + "epoch": 0.37, + "learning_rate": 1.4398601028258207e-05, + "loss": 0.9946, + "step": 11707 + }, + { + "epoch": 0.37, + "learning_rate": 1.4397670784251904e-05, + "loss": 0.9619, + "step": 11708 + }, + { + "epoch": 0.37, + "learning_rate": 1.4396740493063459e-05, + "loss": 1.0552, + "step": 11709 + }, + { + "epoch": 0.37, + "learning_rate": 1.4395810154702856e-05, + "loss": 1.061, + "step": 11710 + }, + { + "epoch": 0.37, + "learning_rate": 1.4394879769180075e-05, + "loss": 1.0996, + "step": 11711 + }, + { + "epoch": 0.37, + "learning_rate": 1.4393949336505097e-05, + "loss": 1.0093, + "step": 11712 + }, + { + "epoch": 0.37, + "learning_rate": 1.4393018856687908e-05, + "loss": 1.0972, + "step": 11713 + }, + { + "epoch": 0.37, + "learning_rate": 1.4392088329738486e-05, + "loss": 0.9775, + "step": 11714 + }, + { + "epoch": 0.37, + "learning_rate": 1.4391157755666818e-05, + "loss": 1.022, + "step": 11715 + }, + { + "epoch": 0.37, + "learning_rate": 1.4390227134482886e-05, + "loss": 1.1069, + "step": 11716 + }, + { + "epoch": 0.37, + "learning_rate": 1.4389296466196679e-05, + "loss": 0.9365, + "step": 11717 + }, + { + "epoch": 0.37, + "learning_rate": 1.4388365750818177e-05, + "loss": 1.0923, + "step": 11718 + }, + { + "epoch": 0.37, + "learning_rate": 1.4387434988357366e-05, + "loss": 1.062, + "step": 11719 + }, + { + "epoch": 0.37, + "learning_rate": 1.4386504178824232e-05, + "loss": 1.0547, + "step": 11720 + }, + { + "epoch": 0.37, + "learning_rate": 1.4385573322228768e-05, + "loss": 1.1196, + "step": 11721 + }, + { + "epoch": 0.37, + "learning_rate": 1.4384642418580953e-05, + "loss": 1.0962, + "step": 11722 + }, + { + "epoch": 0.37, + "learning_rate": 1.4383711467890776e-05, + "loss": 0.5332, + "step": 11723 + }, + { + "epoch": 0.37, + "learning_rate": 1.4382780470168229e-05, + "loss": 1.0356, + "step": 11724 + }, + { + "epoch": 0.37, + "learning_rate": 1.4381849425423298e-05, + "loss": 1.0869, + "step": 11725 + }, + { + "epoch": 0.38, + "learning_rate": 1.438091833366597e-05, + "loss": 0.9419, + "step": 11726 + }, + { + "epoch": 0.38, + "learning_rate": 1.4379987194906237e-05, + "loss": 1.0918, + "step": 11727 + }, + { + "epoch": 0.38, + "learning_rate": 1.4379056009154092e-05, + "loss": 1.0981, + "step": 11728 + }, + { + "epoch": 0.38, + "learning_rate": 1.4378124776419522e-05, + "loss": 0.9272, + "step": 11729 + }, + { + "epoch": 0.38, + "learning_rate": 1.4377193496712517e-05, + "loss": 1.1001, + "step": 11730 + }, + { + "epoch": 0.38, + "learning_rate": 1.437626217004307e-05, + "loss": 1.0669, + "step": 11731 + }, + { + "epoch": 0.38, + "learning_rate": 1.4375330796421177e-05, + "loss": 0.9824, + "step": 11732 + }, + { + "epoch": 0.38, + "learning_rate": 1.4374399375856823e-05, + "loss": 1.0835, + "step": 11733 + }, + { + "epoch": 0.38, + "learning_rate": 1.4373467908360003e-05, + "loss": 0.8789, + "step": 11734 + }, + { + "epoch": 0.38, + "learning_rate": 1.4372536393940718e-05, + "loss": 1.0522, + "step": 11735 + }, + { + "epoch": 0.38, + "learning_rate": 1.4371604832608954e-05, + "loss": 1.0142, + "step": 11736 + }, + { + "epoch": 0.38, + "learning_rate": 1.437067322437471e-05, + "loss": 0.9302, + "step": 11737 + }, + { + "epoch": 0.38, + "learning_rate": 1.4369741569247977e-05, + "loss": 1.0771, + "step": 11738 + }, + { + "epoch": 0.38, + "learning_rate": 1.4368809867238754e-05, + "loss": 1.0483, + "step": 11739 + }, + { + "epoch": 0.38, + "learning_rate": 1.4367878118357037e-05, + "loss": 1.1084, + "step": 11740 + }, + { + "epoch": 0.38, + "learning_rate": 1.4366946322612822e-05, + "loss": 0.9785, + "step": 11741 + }, + { + "epoch": 0.38, + "learning_rate": 1.4366014480016107e-05, + "loss": 1.2163, + "step": 11742 + }, + { + "epoch": 0.38, + "learning_rate": 1.4365082590576887e-05, + "loss": 0.9751, + "step": 11743 + }, + { + "epoch": 0.38, + "learning_rate": 1.4364150654305163e-05, + "loss": 0.9741, + "step": 11744 + }, + { + "epoch": 0.38, + "learning_rate": 1.436321867121093e-05, + "loss": 0.9854, + "step": 11745 + }, + { + "epoch": 0.38, + "learning_rate": 1.436228664130419e-05, + "loss": 0.9238, + "step": 11746 + }, + { + "epoch": 0.38, + "learning_rate": 1.4361354564594948e-05, + "loss": 1.0298, + "step": 11747 + }, + { + "epoch": 0.38, + "learning_rate": 1.436042244109319e-05, + "loss": 0.9756, + "step": 11748 + }, + { + "epoch": 0.38, + "learning_rate": 1.4359490270808933e-05, + "loss": 0.9214, + "step": 11749 + }, + { + "epoch": 0.38, + "learning_rate": 1.4358558053752166e-05, + "loss": 0.8599, + "step": 11750 + }, + { + "epoch": 0.38, + "learning_rate": 1.4357625789932894e-05, + "loss": 1.0801, + "step": 11751 + }, + { + "epoch": 0.38, + "learning_rate": 1.4356693479361124e-05, + "loss": 1.0093, + "step": 11752 + }, + { + "epoch": 0.38, + "learning_rate": 1.4355761122046852e-05, + "loss": 1.0229, + "step": 11753 + }, + { + "epoch": 0.38, + "learning_rate": 1.4354828718000082e-05, + "loss": 0.9067, + "step": 11754 + }, + { + "epoch": 0.38, + "learning_rate": 1.4353896267230822e-05, + "loss": 1.0552, + "step": 11755 + }, + { + "epoch": 0.38, + "learning_rate": 1.4352963769749075e-05, + "loss": 1.1826, + "step": 11756 + }, + { + "epoch": 0.38, + "learning_rate": 1.4352031225564842e-05, + "loss": 0.957, + "step": 11757 + }, + { + "epoch": 0.38, + "learning_rate": 1.4351098634688132e-05, + "loss": 0.9644, + "step": 11758 + }, + { + "epoch": 0.38, + "learning_rate": 1.4350165997128949e-05, + "loss": 0.9902, + "step": 11759 + }, + { + "epoch": 0.38, + "learning_rate": 1.4349233312897298e-05, + "loss": 1.0059, + "step": 11760 + }, + { + "epoch": 0.38, + "learning_rate": 1.4348300582003187e-05, + "loss": 1.1021, + "step": 11761 + }, + { + "epoch": 0.38, + "learning_rate": 1.4347367804456625e-05, + "loss": 1.001, + "step": 11762 + }, + { + "epoch": 0.38, + "learning_rate": 1.4346434980267616e-05, + "loss": 1.1655, + "step": 11763 + }, + { + "epoch": 0.38, + "learning_rate": 1.4345502109446173e-05, + "loss": 0.9038, + "step": 11764 + }, + { + "epoch": 0.38, + "learning_rate": 1.4344569192002298e-05, + "loss": 1.0752, + "step": 11765 + }, + { + "epoch": 0.38, + "learning_rate": 1.4343636227946007e-05, + "loss": 1.0474, + "step": 11766 + }, + { + "epoch": 0.38, + "learning_rate": 1.4342703217287304e-05, + "loss": 1.1582, + "step": 11767 + }, + { + "epoch": 0.38, + "learning_rate": 1.4341770160036203e-05, + "loss": 1.0156, + "step": 11768 + }, + { + "epoch": 0.38, + "learning_rate": 1.4340837056202712e-05, + "loss": 0.9985, + "step": 11769 + }, + { + "epoch": 0.38, + "learning_rate": 1.4339903905796845e-05, + "loss": 0.9756, + "step": 11770 + }, + { + "epoch": 0.38, + "learning_rate": 1.4338970708828615e-05, + "loss": 1.0044, + "step": 11771 + }, + { + "epoch": 0.38, + "learning_rate": 1.4338037465308026e-05, + "loss": 1.0049, + "step": 11772 + }, + { + "epoch": 0.38, + "learning_rate": 1.4337104175245097e-05, + "loss": 1.0659, + "step": 11773 + }, + { + "epoch": 0.38, + "learning_rate": 1.4336170838649845e-05, + "loss": 1.0332, + "step": 11774 + }, + { + "epoch": 0.38, + "learning_rate": 1.4335237455532277e-05, + "loss": 0.9902, + "step": 11775 + }, + { + "epoch": 0.38, + "learning_rate": 1.4334304025902405e-05, + "loss": 1.0918, + "step": 11776 + }, + { + "epoch": 0.38, + "learning_rate": 1.4333370549770253e-05, + "loss": 1.1016, + "step": 11777 + }, + { + "epoch": 0.38, + "learning_rate": 1.4332437027145829e-05, + "loss": 1.0015, + "step": 11778 + }, + { + "epoch": 0.38, + "learning_rate": 1.433150345803915e-05, + "loss": 0.9756, + "step": 11779 + }, + { + "epoch": 0.38, + "learning_rate": 1.4330569842460231e-05, + "loss": 1.0425, + "step": 11780 + }, + { + "epoch": 0.38, + "learning_rate": 1.4329636180419094e-05, + "loss": 1.0103, + "step": 11781 + }, + { + "epoch": 0.38, + "learning_rate": 1.4328702471925751e-05, + "loss": 0.9995, + "step": 11782 + }, + { + "epoch": 0.38, + "learning_rate": 1.432776871699022e-05, + "loss": 1.0381, + "step": 11783 + }, + { + "epoch": 0.38, + "learning_rate": 1.4326834915622522e-05, + "loss": 1.0122, + "step": 11784 + }, + { + "epoch": 0.38, + "learning_rate": 1.4325901067832674e-05, + "loss": 0.9258, + "step": 11785 + }, + { + "epoch": 0.38, + "learning_rate": 1.4324967173630694e-05, + "loss": 0.9746, + "step": 11786 + }, + { + "epoch": 0.38, + "learning_rate": 1.4324033233026606e-05, + "loss": 1.0586, + "step": 11787 + }, + { + "epoch": 0.38, + "learning_rate": 1.4323099246030422e-05, + "loss": 1.1211, + "step": 11788 + }, + { + "epoch": 0.38, + "learning_rate": 1.4322165212652172e-05, + "loss": 0.9761, + "step": 11789 + }, + { + "epoch": 0.38, + "learning_rate": 1.4321231132901874e-05, + "loss": 0.9233, + "step": 11790 + }, + { + "epoch": 0.38, + "learning_rate": 1.4320297006789546e-05, + "loss": 1.0083, + "step": 11791 + }, + { + "epoch": 0.38, + "learning_rate": 1.4319362834325212e-05, + "loss": 0.9316, + "step": 11792 + }, + { + "epoch": 0.38, + "learning_rate": 1.4318428615518895e-05, + "loss": 0.8945, + "step": 11793 + }, + { + "epoch": 0.38, + "learning_rate": 1.431749435038062e-05, + "loss": 0.8433, + "step": 11794 + }, + { + "epoch": 0.38, + "learning_rate": 1.4316560038920408e-05, + "loss": 0.9893, + "step": 11795 + }, + { + "epoch": 0.38, + "learning_rate": 1.4315625681148284e-05, + "loss": 0.9971, + "step": 11796 + }, + { + "epoch": 0.38, + "learning_rate": 1.4314691277074273e-05, + "loss": 1.0576, + "step": 11797 + }, + { + "epoch": 0.38, + "learning_rate": 1.43137568267084e-05, + "loss": 0.9966, + "step": 11798 + }, + { + "epoch": 0.38, + "learning_rate": 1.4312822330060689e-05, + "loss": 1.0967, + "step": 11799 + }, + { + "epoch": 0.38, + "learning_rate": 1.4311887787141169e-05, + "loss": 0.9243, + "step": 11800 + }, + { + "epoch": 0.38, + "learning_rate": 1.4310953197959867e-05, + "loss": 1.0342, + "step": 11801 + }, + { + "epoch": 0.38, + "learning_rate": 1.4310018562526806e-05, + "loss": 1.0142, + "step": 11802 + }, + { + "epoch": 0.38, + "learning_rate": 1.4309083880852015e-05, + "loss": 1.0449, + "step": 11803 + }, + { + "epoch": 0.38, + "learning_rate": 1.4308149152945526e-05, + "loss": 1.165, + "step": 11804 + }, + { + "epoch": 0.38, + "learning_rate": 1.4307214378817362e-05, + "loss": 0.9316, + "step": 11805 + }, + { + "epoch": 0.38, + "learning_rate": 1.4306279558477557e-05, + "loss": 0.9233, + "step": 11806 + }, + { + "epoch": 0.38, + "learning_rate": 1.4305344691936136e-05, + "loss": 1.043, + "step": 11807 + }, + { + "epoch": 0.38, + "learning_rate": 1.4304409779203133e-05, + "loss": 1.0889, + "step": 11808 + }, + { + "epoch": 0.38, + "learning_rate": 1.4303474820288575e-05, + "loss": 0.9785, + "step": 11809 + }, + { + "epoch": 0.38, + "learning_rate": 1.4302539815202497e-05, + "loss": 0.9229, + "step": 11810 + }, + { + "epoch": 0.38, + "learning_rate": 1.4301604763954929e-05, + "loss": 0.9209, + "step": 11811 + }, + { + "epoch": 0.38, + "learning_rate": 1.43006696665559e-05, + "loss": 1.1089, + "step": 11812 + }, + { + "epoch": 0.38, + "learning_rate": 1.4299734523015444e-05, + "loss": 1.0796, + "step": 11813 + }, + { + "epoch": 0.38, + "learning_rate": 1.4298799333343597e-05, + "loss": 1.0493, + "step": 11814 + }, + { + "epoch": 0.38, + "learning_rate": 1.4297864097550397e-05, + "loss": 1.0166, + "step": 11815 + }, + { + "epoch": 0.38, + "learning_rate": 1.4296928815645865e-05, + "loss": 1.0562, + "step": 11816 + }, + { + "epoch": 0.38, + "learning_rate": 1.4295993487640043e-05, + "loss": 1.0649, + "step": 11817 + }, + { + "epoch": 0.38, + "learning_rate": 1.4295058113542966e-05, + "loss": 1.1523, + "step": 11818 + }, + { + "epoch": 0.38, + "learning_rate": 1.4294122693364672e-05, + "loss": 0.9136, + "step": 11819 + }, + { + "epoch": 0.38, + "learning_rate": 1.429318722711519e-05, + "loss": 0.9966, + "step": 11820 + }, + { + "epoch": 0.38, + "learning_rate": 1.4292251714804561e-05, + "loss": 1.001, + "step": 11821 + }, + { + "epoch": 0.38, + "learning_rate": 1.4291316156442823e-05, + "loss": 1.0117, + "step": 11822 + }, + { + "epoch": 0.38, + "learning_rate": 1.4290380552040013e-05, + "loss": 1.103, + "step": 11823 + }, + { + "epoch": 0.38, + "learning_rate": 1.4289444901606165e-05, + "loss": 1.0586, + "step": 11824 + }, + { + "epoch": 0.38, + "learning_rate": 1.4288509205151322e-05, + "loss": 0.9854, + "step": 11825 + }, + { + "epoch": 0.38, + "learning_rate": 1.4287573462685521e-05, + "loss": 1.1006, + "step": 11826 + }, + { + "epoch": 0.38, + "learning_rate": 1.4286637674218805e-05, + "loss": 1.0542, + "step": 11827 + }, + { + "epoch": 0.38, + "learning_rate": 1.4285701839761207e-05, + "loss": 1.1152, + "step": 11828 + }, + { + "epoch": 0.38, + "learning_rate": 1.4284765959322772e-05, + "loss": 1.0591, + "step": 11829 + }, + { + "epoch": 0.38, + "learning_rate": 1.4283830032913545e-05, + "loss": 0.9238, + "step": 11830 + }, + { + "epoch": 0.38, + "learning_rate": 1.4282894060543557e-05, + "loss": 0.6597, + "step": 11831 + }, + { + "epoch": 0.38, + "learning_rate": 1.428195804222286e-05, + "loss": 0.9897, + "step": 11832 + }, + { + "epoch": 0.38, + "learning_rate": 1.428102197796149e-05, + "loss": 1.0981, + "step": 11833 + }, + { + "epoch": 0.38, + "learning_rate": 1.428008586776949e-05, + "loss": 1.0562, + "step": 11834 + }, + { + "epoch": 0.38, + "learning_rate": 1.427914971165691e-05, + "loss": 0.9883, + "step": 11835 + }, + { + "epoch": 0.38, + "learning_rate": 1.4278213509633787e-05, + "loss": 1.1255, + "step": 11836 + }, + { + "epoch": 0.38, + "learning_rate": 1.427727726171017e-05, + "loss": 1.084, + "step": 11837 + }, + { + "epoch": 0.38, + "learning_rate": 1.42763409678961e-05, + "loss": 1.0923, + "step": 11838 + }, + { + "epoch": 0.38, + "learning_rate": 1.4275404628201626e-05, + "loss": 0.9849, + "step": 11839 + }, + { + "epoch": 0.38, + "learning_rate": 1.427446824263679e-05, + "loss": 1.0229, + "step": 11840 + }, + { + "epoch": 0.38, + "learning_rate": 1.4273531811211642e-05, + "loss": 1.0615, + "step": 11841 + }, + { + "epoch": 0.38, + "learning_rate": 1.4272595333936225e-05, + "loss": 1.0098, + "step": 11842 + }, + { + "epoch": 0.38, + "learning_rate": 1.4271658810820592e-05, + "loss": 1.0283, + "step": 11843 + }, + { + "epoch": 0.38, + "learning_rate": 1.4270722241874784e-05, + "loss": 0.9932, + "step": 11844 + }, + { + "epoch": 0.38, + "learning_rate": 1.4269785627108856e-05, + "loss": 1.0107, + "step": 11845 + }, + { + "epoch": 0.38, + "learning_rate": 1.4268848966532853e-05, + "loss": 1.0146, + "step": 11846 + }, + { + "epoch": 0.38, + "learning_rate": 1.4267912260156824e-05, + "loss": 1.002, + "step": 11847 + }, + { + "epoch": 0.38, + "learning_rate": 1.4266975507990821e-05, + "loss": 1.0493, + "step": 11848 + }, + { + "epoch": 0.38, + "learning_rate": 1.4266038710044893e-05, + "loss": 0.9829, + "step": 11849 + }, + { + "epoch": 0.38, + "learning_rate": 1.4265101866329092e-05, + "loss": 0.8977, + "step": 11850 + }, + { + "epoch": 0.38, + "learning_rate": 1.4264164976853467e-05, + "loss": 0.9082, + "step": 11851 + }, + { + "epoch": 0.38, + "learning_rate": 1.426322804162807e-05, + "loss": 0.9609, + "step": 11852 + }, + { + "epoch": 0.38, + "learning_rate": 1.4262291060662959e-05, + "loss": 1.0469, + "step": 11853 + }, + { + "epoch": 0.38, + "learning_rate": 1.4261354033968177e-05, + "loss": 0.9102, + "step": 11854 + }, + { + "epoch": 0.38, + "learning_rate": 1.4260416961553784e-05, + "loss": 0.9692, + "step": 11855 + }, + { + "epoch": 0.38, + "learning_rate": 1.4259479843429834e-05, + "loss": 1.0405, + "step": 11856 + }, + { + "epoch": 0.38, + "learning_rate": 1.425854267960638e-05, + "loss": 0.9683, + "step": 11857 + }, + { + "epoch": 0.38, + "learning_rate": 1.4257605470093475e-05, + "loss": 1.0581, + "step": 11858 + }, + { + "epoch": 0.38, + "learning_rate": 1.4256668214901172e-05, + "loss": 0.9185, + "step": 11859 + }, + { + "epoch": 0.38, + "learning_rate": 1.4255730914039533e-05, + "loss": 1.1206, + "step": 11860 + }, + { + "epoch": 0.38, + "learning_rate": 1.4254793567518612e-05, + "loss": 1.0493, + "step": 11861 + }, + { + "epoch": 0.38, + "learning_rate": 1.4253856175348463e-05, + "loss": 1.0698, + "step": 11862 + }, + { + "epoch": 0.38, + "learning_rate": 1.4252918737539148e-05, + "loss": 1.0923, + "step": 11863 + }, + { + "epoch": 0.38, + "learning_rate": 1.4251981254100722e-05, + "loss": 0.7996, + "step": 11864 + }, + { + "epoch": 0.38, + "learning_rate": 1.425104372504324e-05, + "loss": 1.1128, + "step": 11865 + }, + { + "epoch": 0.38, + "learning_rate": 1.4250106150376763e-05, + "loss": 0.5657, + "step": 11866 + }, + { + "epoch": 0.38, + "learning_rate": 1.4249168530111352e-05, + "loss": 0.9658, + "step": 11867 + }, + { + "epoch": 0.38, + "learning_rate": 1.4248230864257068e-05, + "loss": 1.0171, + "step": 11868 + }, + { + "epoch": 0.38, + "learning_rate": 1.4247293152823967e-05, + "loss": 1.062, + "step": 11869 + }, + { + "epoch": 0.38, + "learning_rate": 1.4246355395822111e-05, + "loss": 1.1016, + "step": 11870 + }, + { + "epoch": 0.38, + "learning_rate": 1.4245417593261558e-05, + "loss": 1.1187, + "step": 11871 + }, + { + "epoch": 0.38, + "learning_rate": 1.424447974515238e-05, + "loss": 1.0103, + "step": 11872 + }, + { + "epoch": 0.38, + "learning_rate": 1.4243541851504628e-05, + "loss": 1.0317, + "step": 11873 + }, + { + "epoch": 0.38, + "learning_rate": 1.4242603912328367e-05, + "loss": 0.9155, + "step": 11874 + }, + { + "epoch": 0.38, + "learning_rate": 1.4241665927633664e-05, + "loss": 0.9824, + "step": 11875 + }, + { + "epoch": 0.38, + "learning_rate": 1.4240727897430577e-05, + "loss": 1.0957, + "step": 11876 + }, + { + "epoch": 0.38, + "learning_rate": 1.4239789821729177e-05, + "loss": 0.9932, + "step": 11877 + }, + { + "epoch": 0.38, + "learning_rate": 1.4238851700539522e-05, + "loss": 1.0542, + "step": 11878 + }, + { + "epoch": 0.38, + "learning_rate": 1.4237913533871679e-05, + "loss": 1.0073, + "step": 11879 + }, + { + "epoch": 0.38, + "learning_rate": 1.4236975321735718e-05, + "loss": 0.8208, + "step": 11880 + }, + { + "epoch": 0.38, + "learning_rate": 1.4236037064141697e-05, + "loss": 1.1079, + "step": 11881 + }, + { + "epoch": 0.38, + "learning_rate": 1.4235098761099686e-05, + "loss": 0.958, + "step": 11882 + }, + { + "epoch": 0.38, + "learning_rate": 1.4234160412619758e-05, + "loss": 1.1104, + "step": 11883 + }, + { + "epoch": 0.38, + "learning_rate": 1.4233222018711969e-05, + "loss": 1.1196, + "step": 11884 + }, + { + "epoch": 0.38, + "learning_rate": 1.4232283579386395e-05, + "loss": 1.0918, + "step": 11885 + }, + { + "epoch": 0.38, + "learning_rate": 1.4231345094653101e-05, + "loss": 1.0518, + "step": 11886 + }, + { + "epoch": 0.38, + "learning_rate": 1.4230406564522158e-05, + "loss": 1.1431, + "step": 11887 + }, + { + "epoch": 0.38, + "learning_rate": 1.4229467989003636e-05, + "loss": 1.0942, + "step": 11888 + }, + { + "epoch": 0.38, + "learning_rate": 1.42285293681076e-05, + "loss": 1.0723, + "step": 11889 + }, + { + "epoch": 0.38, + "learning_rate": 1.4227590701844128e-05, + "loss": 1.1631, + "step": 11890 + }, + { + "epoch": 0.38, + "learning_rate": 1.4226651990223284e-05, + "loss": 1.0449, + "step": 11891 + }, + { + "epoch": 0.38, + "learning_rate": 1.4225713233255142e-05, + "loss": 1.0547, + "step": 11892 + }, + { + "epoch": 0.38, + "learning_rate": 1.4224774430949774e-05, + "loss": 1.0371, + "step": 11893 + }, + { + "epoch": 0.38, + "learning_rate": 1.4223835583317254e-05, + "loss": 0.9321, + "step": 11894 + }, + { + "epoch": 0.38, + "learning_rate": 1.422289669036765e-05, + "loss": 1.0459, + "step": 11895 + }, + { + "epoch": 0.38, + "learning_rate": 1.422195775211104e-05, + "loss": 1.0566, + "step": 11896 + }, + { + "epoch": 0.38, + "learning_rate": 1.4221018768557495e-05, + "loss": 0.9888, + "step": 11897 + }, + { + "epoch": 0.38, + "learning_rate": 1.4220079739717092e-05, + "loss": 0.9976, + "step": 11898 + }, + { + "epoch": 0.38, + "learning_rate": 1.4219140665599901e-05, + "loss": 0.9814, + "step": 11899 + }, + { + "epoch": 0.38, + "learning_rate": 1.4218201546216004e-05, + "loss": 0.9595, + "step": 11900 + }, + { + "epoch": 0.38, + "learning_rate": 1.4217262381575469e-05, + "loss": 0.9976, + "step": 11901 + }, + { + "epoch": 0.38, + "learning_rate": 1.421632317168838e-05, + "loss": 1.0459, + "step": 11902 + }, + { + "epoch": 0.38, + "learning_rate": 1.4215383916564807e-05, + "loss": 0.9199, + "step": 11903 + }, + { + "epoch": 0.38, + "learning_rate": 1.421444461621483e-05, + "loss": 1.0132, + "step": 11904 + }, + { + "epoch": 0.38, + "learning_rate": 1.4213505270648527e-05, + "loss": 1.085, + "step": 11905 + }, + { + "epoch": 0.38, + "learning_rate": 1.4212565879875977e-05, + "loss": 1.0405, + "step": 11906 + }, + { + "epoch": 0.38, + "learning_rate": 1.4211626443907256e-05, + "loss": 1.019, + "step": 11907 + }, + { + "epoch": 0.38, + "learning_rate": 1.4210686962752443e-05, + "loss": 1.0649, + "step": 11908 + }, + { + "epoch": 0.38, + "learning_rate": 1.4209747436421624e-05, + "loss": 1.1147, + "step": 11909 + }, + { + "epoch": 0.38, + "learning_rate": 1.4208807864924868e-05, + "loss": 1.0352, + "step": 11910 + }, + { + "epoch": 0.38, + "learning_rate": 1.4207868248272266e-05, + "loss": 0.9634, + "step": 11911 + }, + { + "epoch": 0.38, + "learning_rate": 1.420692858647389e-05, + "loss": 1.1851, + "step": 11912 + }, + { + "epoch": 0.38, + "learning_rate": 1.4205988879539832e-05, + "loss": 1.0566, + "step": 11913 + }, + { + "epoch": 0.38, + "learning_rate": 1.4205049127480166e-05, + "loss": 0.9995, + "step": 11914 + }, + { + "epoch": 0.38, + "learning_rate": 1.4204109330304977e-05, + "loss": 0.9805, + "step": 11915 + }, + { + "epoch": 0.38, + "learning_rate": 1.4203169488024346e-05, + "loss": 1.126, + "step": 11916 + }, + { + "epoch": 0.38, + "learning_rate": 1.4202229600648361e-05, + "loss": 1.1377, + "step": 11917 + }, + { + "epoch": 0.38, + "learning_rate": 1.4201289668187101e-05, + "loss": 1.1201, + "step": 11918 + }, + { + "epoch": 0.38, + "learning_rate": 1.4200349690650654e-05, + "loss": 0.9697, + "step": 11919 + }, + { + "epoch": 0.38, + "learning_rate": 1.4199409668049102e-05, + "loss": 0.9595, + "step": 11920 + }, + { + "epoch": 0.38, + "learning_rate": 1.4198469600392533e-05, + "loss": 1.0674, + "step": 11921 + }, + { + "epoch": 0.38, + "learning_rate": 1.4197529487691032e-05, + "loss": 1.0347, + "step": 11922 + }, + { + "epoch": 0.38, + "learning_rate": 1.4196589329954687e-05, + "loss": 1.0244, + "step": 11923 + }, + { + "epoch": 0.38, + "learning_rate": 1.4195649127193578e-05, + "loss": 1.0195, + "step": 11924 + }, + { + "epoch": 0.38, + "learning_rate": 1.4194708879417802e-05, + "loss": 1.0894, + "step": 11925 + }, + { + "epoch": 0.38, + "learning_rate": 1.419376858663744e-05, + "loss": 1.0918, + "step": 11926 + }, + { + "epoch": 0.38, + "learning_rate": 1.4192828248862582e-05, + "loss": 0.9697, + "step": 11927 + }, + { + "epoch": 0.38, + "learning_rate": 1.419188786610332e-05, + "loss": 0.8867, + "step": 11928 + }, + { + "epoch": 0.38, + "learning_rate": 1.4190947438369737e-05, + "loss": 1.0073, + "step": 11929 + }, + { + "epoch": 0.38, + "learning_rate": 1.419000696567193e-05, + "loss": 1.0601, + "step": 11930 + }, + { + "epoch": 0.38, + "learning_rate": 1.4189066448019981e-05, + "loss": 1.1592, + "step": 11931 + }, + { + "epoch": 0.38, + "learning_rate": 1.4188125885423991e-05, + "loss": 0.9751, + "step": 11932 + }, + { + "epoch": 0.38, + "learning_rate": 1.418718527789404e-05, + "loss": 0.9673, + "step": 11933 + }, + { + "epoch": 0.38, + "learning_rate": 1.4186244625440228e-05, + "loss": 1.0127, + "step": 11934 + }, + { + "epoch": 0.38, + "learning_rate": 1.4185303928072643e-05, + "loss": 0.9717, + "step": 11935 + }, + { + "epoch": 0.38, + "learning_rate": 1.4184363185801381e-05, + "loss": 1.0879, + "step": 11936 + }, + { + "epoch": 0.38, + "learning_rate": 1.418342239863653e-05, + "loss": 0.9409, + "step": 11937 + }, + { + "epoch": 0.38, + "learning_rate": 1.4182481566588189e-05, + "loss": 0.9844, + "step": 11938 + }, + { + "epoch": 0.38, + "learning_rate": 1.4181540689666449e-05, + "loss": 1.0435, + "step": 11939 + }, + { + "epoch": 0.38, + "learning_rate": 1.4180599767881406e-05, + "loss": 1.022, + "step": 11940 + }, + { + "epoch": 0.38, + "learning_rate": 1.4179658801243155e-05, + "loss": 0.8896, + "step": 11941 + }, + { + "epoch": 0.38, + "learning_rate": 1.4178717789761787e-05, + "loss": 1.0581, + "step": 11942 + }, + { + "epoch": 0.38, + "learning_rate": 1.4177776733447406e-05, + "loss": 0.9995, + "step": 11943 + }, + { + "epoch": 0.38, + "learning_rate": 1.4176835632310102e-05, + "loss": 1.0679, + "step": 11944 + }, + { + "epoch": 0.38, + "learning_rate": 1.4175894486359975e-05, + "loss": 0.9453, + "step": 11945 + }, + { + "epoch": 0.38, + "learning_rate": 1.4174953295607121e-05, + "loss": 1.0303, + "step": 11946 + }, + { + "epoch": 0.38, + "learning_rate": 1.4174012060061642e-05, + "loss": 1.002, + "step": 11947 + }, + { + "epoch": 0.38, + "learning_rate": 1.4173070779733631e-05, + "loss": 1.0127, + "step": 11948 + }, + { + "epoch": 0.38, + "learning_rate": 1.4172129454633188e-05, + "loss": 0.6113, + "step": 11949 + }, + { + "epoch": 0.38, + "learning_rate": 1.4171188084770413e-05, + "loss": 1.103, + "step": 11950 + }, + { + "epoch": 0.38, + "learning_rate": 1.417024667015541e-05, + "loss": 0.9858, + "step": 11951 + }, + { + "epoch": 0.38, + "learning_rate": 1.4169305210798273e-05, + "loss": 1.0205, + "step": 11952 + }, + { + "epoch": 0.38, + "learning_rate": 1.4168363706709105e-05, + "loss": 1.0703, + "step": 11953 + }, + { + "epoch": 0.38, + "learning_rate": 1.4167422157898008e-05, + "loss": 0.8638, + "step": 11954 + }, + { + "epoch": 0.38, + "learning_rate": 1.4166480564375084e-05, + "loss": 0.9253, + "step": 11955 + }, + { + "epoch": 0.38, + "learning_rate": 1.4165538926150435e-05, + "loss": 1.0503, + "step": 11956 + }, + { + "epoch": 0.38, + "learning_rate": 1.4164597243234164e-05, + "loss": 1.0718, + "step": 11957 + }, + { + "epoch": 0.38, + "learning_rate": 1.4163655515636373e-05, + "loss": 1.0513, + "step": 11958 + }, + { + "epoch": 0.38, + "learning_rate": 1.4162713743367166e-05, + "loss": 1.0918, + "step": 11959 + }, + { + "epoch": 0.38, + "learning_rate": 1.416177192643665e-05, + "loss": 0.979, + "step": 11960 + }, + { + "epoch": 0.38, + "learning_rate": 1.4160830064854925e-05, + "loss": 1.0044, + "step": 11961 + }, + { + "epoch": 0.38, + "learning_rate": 1.4159888158632099e-05, + "loss": 0.9731, + "step": 11962 + }, + { + "epoch": 0.38, + "learning_rate": 1.4158946207778275e-05, + "loss": 1.1431, + "step": 11963 + }, + { + "epoch": 0.38, + "learning_rate": 1.4158004212303565e-05, + "loss": 0.5007, + "step": 11964 + }, + { + "epoch": 0.38, + "learning_rate": 1.4157062172218068e-05, + "loss": 1.1128, + "step": 11965 + }, + { + "epoch": 0.38, + "learning_rate": 1.4156120087531896e-05, + "loss": 0.9771, + "step": 11966 + }, + { + "epoch": 0.38, + "learning_rate": 1.4155177958255158e-05, + "loss": 1.041, + "step": 11967 + }, + { + "epoch": 0.38, + "learning_rate": 1.4154235784397954e-05, + "loss": 1.0479, + "step": 11968 + }, + { + "epoch": 0.38, + "learning_rate": 1.4153293565970402e-05, + "loss": 1.0869, + "step": 11969 + }, + { + "epoch": 0.38, + "learning_rate": 1.4152351302982606e-05, + "loss": 1.0449, + "step": 11970 + }, + { + "epoch": 0.38, + "learning_rate": 1.4151408995444673e-05, + "loss": 0.8506, + "step": 11971 + }, + { + "epoch": 0.38, + "learning_rate": 1.4150466643366718e-05, + "loss": 0.9526, + "step": 11972 + }, + { + "epoch": 0.38, + "learning_rate": 1.4149524246758847e-05, + "loss": 0.9932, + "step": 11973 + }, + { + "epoch": 0.38, + "learning_rate": 1.4148581805631178e-05, + "loss": 1.0562, + "step": 11974 + }, + { + "epoch": 0.38, + "learning_rate": 1.4147639319993816e-05, + "loss": 1.0312, + "step": 11975 + }, + { + "epoch": 0.38, + "learning_rate": 1.4146696789856873e-05, + "loss": 1.0991, + "step": 11976 + }, + { + "epoch": 0.38, + "learning_rate": 1.4145754215230466e-05, + "loss": 0.9219, + "step": 11977 + }, + { + "epoch": 0.38, + "learning_rate": 1.41448115961247e-05, + "loss": 1.0361, + "step": 11978 + }, + { + "epoch": 0.38, + "learning_rate": 1.4143868932549696e-05, + "loss": 1.0742, + "step": 11979 + }, + { + "epoch": 0.38, + "learning_rate": 1.4142926224515564e-05, + "loss": 1.0347, + "step": 11980 + }, + { + "epoch": 0.38, + "learning_rate": 1.4141983472032418e-05, + "loss": 0.9697, + "step": 11981 + }, + { + "epoch": 0.38, + "learning_rate": 1.4141040675110374e-05, + "loss": 1.0884, + "step": 11982 + }, + { + "epoch": 0.38, + "learning_rate": 1.4140097833759547e-05, + "loss": 0.5371, + "step": 11983 + }, + { + "epoch": 0.38, + "learning_rate": 1.413915494799005e-05, + "loss": 1.0552, + "step": 11984 + }, + { + "epoch": 0.38, + "learning_rate": 1.4138212017812006e-05, + "loss": 0.9502, + "step": 11985 + }, + { + "epoch": 0.38, + "learning_rate": 1.4137269043235524e-05, + "loss": 0.9712, + "step": 11986 + }, + { + "epoch": 0.38, + "learning_rate": 1.4136326024270723e-05, + "loss": 0.8916, + "step": 11987 + }, + { + "epoch": 0.38, + "learning_rate": 1.4135382960927722e-05, + "loss": 1.0166, + "step": 11988 + }, + { + "epoch": 0.38, + "learning_rate": 1.4134439853216642e-05, + "loss": 1.0605, + "step": 11989 + }, + { + "epoch": 0.38, + "learning_rate": 1.4133496701147596e-05, + "loss": 1.0396, + "step": 11990 + }, + { + "epoch": 0.38, + "learning_rate": 1.4132553504730703e-05, + "loss": 1.0186, + "step": 11991 + }, + { + "epoch": 0.38, + "learning_rate": 1.4131610263976088e-05, + "loss": 1.0127, + "step": 11992 + }, + { + "epoch": 0.38, + "learning_rate": 1.4130666978893865e-05, + "loss": 0.9277, + "step": 11993 + }, + { + "epoch": 0.38, + "learning_rate": 1.4129723649494158e-05, + "loss": 0.9429, + "step": 11994 + }, + { + "epoch": 0.38, + "learning_rate": 1.4128780275787087e-05, + "loss": 1.1333, + "step": 11995 + }, + { + "epoch": 0.38, + "learning_rate": 1.4127836857782773e-05, + "loss": 0.9507, + "step": 11996 + }, + { + "epoch": 0.38, + "learning_rate": 1.4126893395491336e-05, + "loss": 1.0762, + "step": 11997 + }, + { + "epoch": 0.38, + "learning_rate": 1.4125949888922903e-05, + "loss": 0.9839, + "step": 11998 + }, + { + "epoch": 0.38, + "learning_rate": 1.4125006338087593e-05, + "loss": 1.0645, + "step": 11999 + }, + { + "epoch": 0.38, + "learning_rate": 1.4124062742995533e-05, + "loss": 1.1616, + "step": 12000 + }, + { + "epoch": 0.38, + "learning_rate": 1.412311910365684e-05, + "loss": 1.0371, + "step": 12001 + }, + { + "epoch": 0.38, + "learning_rate": 1.4122175420081643e-05, + "loss": 0.9243, + "step": 12002 + }, + { + "epoch": 0.38, + "learning_rate": 1.4121231692280066e-05, + "loss": 0.9116, + "step": 12003 + }, + { + "epoch": 0.38, + "learning_rate": 1.4120287920262238e-05, + "loss": 1.0659, + "step": 12004 + }, + { + "epoch": 0.38, + "learning_rate": 1.4119344104038274e-05, + "loss": 0.9932, + "step": 12005 + }, + { + "epoch": 0.38, + "learning_rate": 1.4118400243618312e-05, + "loss": 1.0132, + "step": 12006 + }, + { + "epoch": 0.38, + "learning_rate": 1.4117456339012472e-05, + "loss": 1.0513, + "step": 12007 + }, + { + "epoch": 0.38, + "learning_rate": 1.4116512390230883e-05, + "loss": 1.0171, + "step": 12008 + }, + { + "epoch": 0.38, + "learning_rate": 1.411556839728367e-05, + "loss": 1.0596, + "step": 12009 + }, + { + "epoch": 0.38, + "learning_rate": 1.4114624360180963e-05, + "loss": 0.9868, + "step": 12010 + }, + { + "epoch": 0.38, + "learning_rate": 1.4113680278932895e-05, + "loss": 0.9854, + "step": 12011 + }, + { + "epoch": 0.38, + "learning_rate": 1.4112736153549587e-05, + "loss": 0.8838, + "step": 12012 + }, + { + "epoch": 0.38, + "learning_rate": 1.411179198404117e-05, + "loss": 1.0435, + "step": 12013 + }, + { + "epoch": 0.38, + "learning_rate": 1.4110847770417778e-05, + "loss": 1.0039, + "step": 12014 + }, + { + "epoch": 0.38, + "learning_rate": 1.410990351268954e-05, + "loss": 0.5461, + "step": 12015 + }, + { + "epoch": 0.38, + "learning_rate": 1.4108959210866583e-05, + "loss": 1.0996, + "step": 12016 + }, + { + "epoch": 0.38, + "learning_rate": 1.4108014864959044e-05, + "loss": 1.0425, + "step": 12017 + }, + { + "epoch": 0.38, + "learning_rate": 1.410707047497705e-05, + "loss": 0.9937, + "step": 12018 + }, + { + "epoch": 0.38, + "learning_rate": 1.4106126040930736e-05, + "loss": 0.9839, + "step": 12019 + }, + { + "epoch": 0.38, + "learning_rate": 1.4105181562830235e-05, + "loss": 1.1104, + "step": 12020 + }, + { + "epoch": 0.38, + "learning_rate": 1.4104237040685678e-05, + "loss": 0.5203, + "step": 12021 + }, + { + "epoch": 0.38, + "learning_rate": 1.41032924745072e-05, + "loss": 0.9854, + "step": 12022 + }, + { + "epoch": 0.38, + "learning_rate": 1.4102347864304936e-05, + "loss": 0.917, + "step": 12023 + }, + { + "epoch": 0.38, + "learning_rate": 1.4101403210089018e-05, + "loss": 1.0332, + "step": 12024 + }, + { + "epoch": 0.38, + "learning_rate": 1.4100458511869584e-05, + "loss": 0.9565, + "step": 12025 + }, + { + "epoch": 0.38, + "learning_rate": 1.409951376965677e-05, + "loss": 0.9463, + "step": 12026 + }, + { + "epoch": 0.38, + "learning_rate": 1.4098568983460707e-05, + "loss": 1.042, + "step": 12027 + }, + { + "epoch": 0.38, + "learning_rate": 1.4097624153291536e-05, + "loss": 1.1396, + "step": 12028 + }, + { + "epoch": 0.38, + "learning_rate": 1.409667927915939e-05, + "loss": 1.0645, + "step": 12029 + }, + { + "epoch": 0.38, + "learning_rate": 1.4095734361074415e-05, + "loss": 0.9512, + "step": 12030 + }, + { + "epoch": 0.38, + "learning_rate": 1.4094789399046743e-05, + "loss": 1.104, + "step": 12031 + }, + { + "epoch": 0.38, + "learning_rate": 1.409384439308651e-05, + "loss": 1.0063, + "step": 12032 + }, + { + "epoch": 0.38, + "learning_rate": 1.4092899343203856e-05, + "loss": 1.1011, + "step": 12033 + }, + { + "epoch": 0.38, + "learning_rate": 1.4091954249408923e-05, + "loss": 0.9355, + "step": 12034 + }, + { + "epoch": 0.38, + "learning_rate": 1.409100911171185e-05, + "loss": 0.9565, + "step": 12035 + }, + { + "epoch": 0.38, + "learning_rate": 1.4090063930122778e-05, + "loss": 0.9512, + "step": 12036 + }, + { + "epoch": 0.38, + "learning_rate": 1.4089118704651845e-05, + "loss": 1.0879, + "step": 12037 + }, + { + "epoch": 0.38, + "learning_rate": 1.4088173435309198e-05, + "loss": 1.0171, + "step": 12038 + }, + { + "epoch": 0.39, + "learning_rate": 1.408722812210497e-05, + "loss": 1.0264, + "step": 12039 + }, + { + "epoch": 0.39, + "learning_rate": 1.4086282765049308e-05, + "loss": 1.2051, + "step": 12040 + }, + { + "epoch": 0.39, + "learning_rate": 1.4085337364152356e-05, + "loss": 0.9395, + "step": 12041 + }, + { + "epoch": 0.39, + "learning_rate": 1.4084391919424256e-05, + "loss": 1.0698, + "step": 12042 + }, + { + "epoch": 0.39, + "learning_rate": 1.4083446430875149e-05, + "loss": 0.9307, + "step": 12043 + }, + { + "epoch": 0.39, + "learning_rate": 1.4082500898515184e-05, + "loss": 1.0786, + "step": 12044 + }, + { + "epoch": 0.39, + "learning_rate": 1.4081555322354501e-05, + "loss": 1.1519, + "step": 12045 + }, + { + "epoch": 0.39, + "learning_rate": 1.4080609702403247e-05, + "loss": 0.9937, + "step": 12046 + }, + { + "epoch": 0.39, + "learning_rate": 1.4079664038671567e-05, + "loss": 0.9031, + "step": 12047 + }, + { + "epoch": 0.39, + "learning_rate": 1.4078718331169607e-05, + "loss": 1.0815, + "step": 12048 + }, + { + "epoch": 0.39, + "learning_rate": 1.4077772579907514e-05, + "loss": 0.9678, + "step": 12049 + }, + { + "epoch": 0.39, + "learning_rate": 1.4076826784895433e-05, + "loss": 0.9668, + "step": 12050 + }, + { + "epoch": 0.39, + "learning_rate": 1.4075880946143514e-05, + "loss": 1.1611, + "step": 12051 + }, + { + "epoch": 0.39, + "learning_rate": 1.40749350636619e-05, + "loss": 1.1021, + "step": 12052 + }, + { + "epoch": 0.39, + "learning_rate": 1.4073989137460748e-05, + "loss": 1.0391, + "step": 12053 + }, + { + "epoch": 0.39, + "learning_rate": 1.4073043167550198e-05, + "loss": 1.0005, + "step": 12054 + }, + { + "epoch": 0.39, + "learning_rate": 1.4072097153940404e-05, + "loss": 0.9082, + "step": 12055 + }, + { + "epoch": 0.39, + "learning_rate": 1.4071151096641514e-05, + "loss": 1.082, + "step": 12056 + }, + { + "epoch": 0.39, + "learning_rate": 1.4070204995663677e-05, + "loss": 1.0615, + "step": 12057 + }, + { + "epoch": 0.39, + "learning_rate": 1.4069258851017047e-05, + "loss": 1.0552, + "step": 12058 + }, + { + "epoch": 0.39, + "learning_rate": 1.4068312662711771e-05, + "loss": 1.1099, + "step": 12059 + }, + { + "epoch": 0.39, + "learning_rate": 1.4067366430758004e-05, + "loss": 1.02, + "step": 12060 + }, + { + "epoch": 0.39, + "learning_rate": 1.4066420155165897e-05, + "loss": 0.9043, + "step": 12061 + }, + { + "epoch": 0.39, + "learning_rate": 1.4065473835945602e-05, + "loss": 0.9697, + "step": 12062 + }, + { + "epoch": 0.39, + "learning_rate": 1.406452747310727e-05, + "loss": 0.938, + "step": 12063 + }, + { + "epoch": 0.39, + "learning_rate": 1.4063581066661061e-05, + "loss": 1.019, + "step": 12064 + }, + { + "epoch": 0.39, + "learning_rate": 1.4062634616617122e-05, + "loss": 1.0786, + "step": 12065 + }, + { + "epoch": 0.39, + "learning_rate": 1.4061688122985611e-05, + "loss": 0.8755, + "step": 12066 + }, + { + "epoch": 0.39, + "learning_rate": 1.4060741585776678e-05, + "loss": 1.0918, + "step": 12067 + }, + { + "epoch": 0.39, + "learning_rate": 1.4059795005000488e-05, + "loss": 0.9712, + "step": 12068 + }, + { + "epoch": 0.39, + "learning_rate": 1.4058848380667187e-05, + "loss": 0.9229, + "step": 12069 + }, + { + "epoch": 0.39, + "learning_rate": 1.4057901712786935e-05, + "loss": 1.061, + "step": 12070 + }, + { + "epoch": 0.39, + "learning_rate": 1.4056955001369887e-05, + "loss": 0.9292, + "step": 12071 + }, + { + "epoch": 0.39, + "learning_rate": 1.4056008246426207e-05, + "loss": 1.0498, + "step": 12072 + }, + { + "epoch": 0.39, + "learning_rate": 1.4055061447966043e-05, + "loss": 0.9473, + "step": 12073 + }, + { + "epoch": 0.39, + "learning_rate": 1.4054114605999558e-05, + "loss": 1.0532, + "step": 12074 + }, + { + "epoch": 0.39, + "learning_rate": 1.4053167720536914e-05, + "loss": 1.082, + "step": 12075 + }, + { + "epoch": 0.39, + "learning_rate": 1.4052220791588263e-05, + "loss": 1.0332, + "step": 12076 + }, + { + "epoch": 0.39, + "learning_rate": 1.4051273819163766e-05, + "loss": 1.0557, + "step": 12077 + }, + { + "epoch": 0.39, + "learning_rate": 1.4050326803273586e-05, + "loss": 1.0474, + "step": 12078 + }, + { + "epoch": 0.39, + "learning_rate": 1.4049379743927884e-05, + "loss": 1.0903, + "step": 12079 + }, + { + "epoch": 0.39, + "learning_rate": 1.4048432641136817e-05, + "loss": 1.0903, + "step": 12080 + }, + { + "epoch": 0.39, + "learning_rate": 1.4047485494910551e-05, + "loss": 0.9561, + "step": 12081 + }, + { + "epoch": 0.39, + "learning_rate": 1.4046538305259243e-05, + "loss": 0.9395, + "step": 12082 + }, + { + "epoch": 0.39, + "learning_rate": 1.4045591072193059e-05, + "loss": 1.0659, + "step": 12083 + }, + { + "epoch": 0.39, + "learning_rate": 1.4044643795722156e-05, + "loss": 1.0156, + "step": 12084 + }, + { + "epoch": 0.39, + "learning_rate": 1.4043696475856706e-05, + "loss": 1.0347, + "step": 12085 + }, + { + "epoch": 0.39, + "learning_rate": 1.4042749112606868e-05, + "loss": 1.0454, + "step": 12086 + }, + { + "epoch": 0.39, + "learning_rate": 1.4041801705982804e-05, + "loss": 1.0649, + "step": 12087 + }, + { + "epoch": 0.39, + "learning_rate": 1.404085425599468e-05, + "loss": 0.5774, + "step": 12088 + }, + { + "epoch": 0.39, + "learning_rate": 1.4039906762652666e-05, + "loss": 1.0, + "step": 12089 + }, + { + "epoch": 0.39, + "learning_rate": 1.4038959225966922e-05, + "loss": 0.9756, + "step": 12090 + }, + { + "epoch": 0.39, + "learning_rate": 1.4038011645947615e-05, + "loss": 1.0366, + "step": 12091 + }, + { + "epoch": 0.39, + "learning_rate": 1.4037064022604914e-05, + "loss": 1.0269, + "step": 12092 + }, + { + "epoch": 0.39, + "learning_rate": 1.403611635594898e-05, + "loss": 0.9448, + "step": 12093 + }, + { + "epoch": 0.39, + "learning_rate": 1.4035168645989989e-05, + "loss": 1.0557, + "step": 12094 + }, + { + "epoch": 0.39, + "learning_rate": 1.4034220892738101e-05, + "loss": 1.0234, + "step": 12095 + }, + { + "epoch": 0.39, + "learning_rate": 1.4033273096203491e-05, + "loss": 1.0615, + "step": 12096 + }, + { + "epoch": 0.39, + "learning_rate": 1.403232525639632e-05, + "loss": 1.063, + "step": 12097 + }, + { + "epoch": 0.39, + "learning_rate": 1.4031377373326764e-05, + "loss": 1.1392, + "step": 12098 + }, + { + "epoch": 0.39, + "learning_rate": 1.4030429447004992e-05, + "loss": 1.1162, + "step": 12099 + }, + { + "epoch": 0.39, + "learning_rate": 1.402948147744117e-05, + "loss": 1.0503, + "step": 12100 + }, + { + "epoch": 0.39, + "learning_rate": 1.4028533464645471e-05, + "loss": 1.0415, + "step": 12101 + }, + { + "epoch": 0.39, + "learning_rate": 1.402758540862807e-05, + "loss": 1.0845, + "step": 12102 + }, + { + "epoch": 0.39, + "learning_rate": 1.4026637309399131e-05, + "loss": 0.9517, + "step": 12103 + }, + { + "epoch": 0.39, + "learning_rate": 1.4025689166968834e-05, + "loss": 1.0054, + "step": 12104 + }, + { + "epoch": 0.39, + "learning_rate": 1.4024740981347344e-05, + "loss": 0.8442, + "step": 12105 + }, + { + "epoch": 0.39, + "learning_rate": 1.402379275254484e-05, + "loss": 1.084, + "step": 12106 + }, + { + "epoch": 0.39, + "learning_rate": 1.4022844480571494e-05, + "loss": 0.981, + "step": 12107 + }, + { + "epoch": 0.39, + "learning_rate": 1.4021896165437477e-05, + "loss": 0.9131, + "step": 12108 + }, + { + "epoch": 0.39, + "learning_rate": 1.4020947807152968e-05, + "loss": 1.0078, + "step": 12109 + }, + { + "epoch": 0.39, + "learning_rate": 1.4019999405728135e-05, + "loss": 0.9854, + "step": 12110 + }, + { + "epoch": 0.39, + "learning_rate": 1.401905096117316e-05, + "loss": 1.0791, + "step": 12111 + }, + { + "epoch": 0.39, + "learning_rate": 1.4018102473498213e-05, + "loss": 1.0342, + "step": 12112 + }, + { + "epoch": 0.39, + "learning_rate": 1.4017153942713479e-05, + "loss": 0.9932, + "step": 12113 + }, + { + "epoch": 0.39, + "learning_rate": 1.4016205368829128e-05, + "loss": 0.9697, + "step": 12114 + }, + { + "epoch": 0.39, + "learning_rate": 1.4015256751855335e-05, + "loss": 1.0786, + "step": 12115 + }, + { + "epoch": 0.39, + "learning_rate": 1.4014308091802282e-05, + "loss": 1.0469, + "step": 12116 + }, + { + "epoch": 0.39, + "learning_rate": 1.4013359388680148e-05, + "loss": 1.1016, + "step": 12117 + }, + { + "epoch": 0.39, + "learning_rate": 1.4012410642499108e-05, + "loss": 0.9795, + "step": 12118 + }, + { + "epoch": 0.39, + "learning_rate": 1.4011461853269342e-05, + "loss": 0.9409, + "step": 12119 + }, + { + "epoch": 0.39, + "learning_rate": 1.4010513021001031e-05, + "loss": 1.0015, + "step": 12120 + }, + { + "epoch": 0.39, + "learning_rate": 1.4009564145704353e-05, + "loss": 1.0405, + "step": 12121 + }, + { + "epoch": 0.39, + "learning_rate": 1.400861522738949e-05, + "loss": 0.9824, + "step": 12122 + }, + { + "epoch": 0.39, + "learning_rate": 1.4007666266066622e-05, + "loss": 0.9922, + "step": 12123 + }, + { + "epoch": 0.39, + "learning_rate": 1.4006717261745931e-05, + "loss": 1.1006, + "step": 12124 + }, + { + "epoch": 0.39, + "learning_rate": 1.4005768214437598e-05, + "loss": 0.9629, + "step": 12125 + }, + { + "epoch": 0.39, + "learning_rate": 1.4004819124151805e-05, + "loss": 1.082, + "step": 12126 + }, + { + "epoch": 0.39, + "learning_rate": 1.4003869990898734e-05, + "loss": 0.9507, + "step": 12127 + }, + { + "epoch": 0.39, + "learning_rate": 1.4002920814688571e-05, + "loss": 1.041, + "step": 12128 + }, + { + "epoch": 0.39, + "learning_rate": 1.4001971595531499e-05, + "loss": 1.0815, + "step": 12129 + }, + { + "epoch": 0.39, + "learning_rate": 1.4001022333437697e-05, + "loss": 1.0386, + "step": 12130 + }, + { + "epoch": 0.39, + "learning_rate": 1.4000073028417359e-05, + "loss": 0.9927, + "step": 12131 + }, + { + "epoch": 0.39, + "learning_rate": 1.399912368048066e-05, + "loss": 0.9663, + "step": 12132 + }, + { + "epoch": 0.39, + "learning_rate": 1.399817428963779e-05, + "loss": 0.9497, + "step": 12133 + }, + { + "epoch": 0.39, + "learning_rate": 1.3997224855898939e-05, + "loss": 1.0674, + "step": 12134 + }, + { + "epoch": 0.39, + "learning_rate": 1.3996275379274284e-05, + "loss": 1.0957, + "step": 12135 + }, + { + "epoch": 0.39, + "learning_rate": 1.399532585977402e-05, + "loss": 0.9375, + "step": 12136 + }, + { + "epoch": 0.39, + "learning_rate": 1.3994376297408332e-05, + "loss": 0.9883, + "step": 12137 + }, + { + "epoch": 0.39, + "learning_rate": 1.3993426692187405e-05, + "loss": 1.0078, + "step": 12138 + }, + { + "epoch": 0.39, + "learning_rate": 1.3992477044121429e-05, + "loss": 0.9326, + "step": 12139 + }, + { + "epoch": 0.39, + "learning_rate": 1.3991527353220595e-05, + "loss": 1.0889, + "step": 12140 + }, + { + "epoch": 0.39, + "learning_rate": 1.399057761949509e-05, + "loss": 1.0166, + "step": 12141 + }, + { + "epoch": 0.39, + "learning_rate": 1.39896278429551e-05, + "loss": 1.0562, + "step": 12142 + }, + { + "epoch": 0.39, + "learning_rate": 1.3988678023610822e-05, + "loss": 0.9604, + "step": 12143 + }, + { + "epoch": 0.39, + "learning_rate": 1.3987728161472442e-05, + "loss": 1.0371, + "step": 12144 + }, + { + "epoch": 0.39, + "learning_rate": 1.3986778256550153e-05, + "loss": 1.0713, + "step": 12145 + }, + { + "epoch": 0.39, + "learning_rate": 1.3985828308854146e-05, + "loss": 1.0361, + "step": 12146 + }, + { + "epoch": 0.39, + "learning_rate": 1.3984878318394613e-05, + "loss": 0.9287, + "step": 12147 + }, + { + "epoch": 0.39, + "learning_rate": 1.3983928285181745e-05, + "loss": 0.9634, + "step": 12148 + }, + { + "epoch": 0.39, + "learning_rate": 1.3982978209225735e-05, + "loss": 1.1074, + "step": 12149 + }, + { + "epoch": 0.39, + "learning_rate": 1.3982028090536776e-05, + "loss": 1.0928, + "step": 12150 + }, + { + "epoch": 0.39, + "learning_rate": 1.3981077929125066e-05, + "loss": 0.5723, + "step": 12151 + }, + { + "epoch": 0.39, + "learning_rate": 1.3980127725000794e-05, + "loss": 1.1138, + "step": 12152 + }, + { + "epoch": 0.39, + "learning_rate": 1.3979177478174158e-05, + "loss": 0.8213, + "step": 12153 + }, + { + "epoch": 0.39, + "learning_rate": 1.3978227188655349e-05, + "loss": 0.9927, + "step": 12154 + }, + { + "epoch": 0.39, + "learning_rate": 1.3977276856454566e-05, + "loss": 0.9863, + "step": 12155 + }, + { + "epoch": 0.39, + "learning_rate": 1.3976326481582005e-05, + "loss": 0.9551, + "step": 12156 + }, + { + "epoch": 0.39, + "learning_rate": 1.397537606404786e-05, + "loss": 0.9536, + "step": 12157 + }, + { + "epoch": 0.39, + "learning_rate": 1.3974425603862332e-05, + "loss": 1.0991, + "step": 12158 + }, + { + "epoch": 0.39, + "learning_rate": 1.3973475101035616e-05, + "loss": 1.0054, + "step": 12159 + }, + { + "epoch": 0.39, + "learning_rate": 1.3972524555577907e-05, + "loss": 0.9502, + "step": 12160 + }, + { + "epoch": 0.39, + "learning_rate": 1.3971573967499408e-05, + "loss": 0.9292, + "step": 12161 + }, + { + "epoch": 0.39, + "learning_rate": 1.3970623336810316e-05, + "loss": 1.1143, + "step": 12162 + }, + { + "epoch": 0.39, + "learning_rate": 1.396967266352083e-05, + "loss": 1.1494, + "step": 12163 + }, + { + "epoch": 0.39, + "learning_rate": 1.3968721947641149e-05, + "loss": 0.9814, + "step": 12164 + }, + { + "epoch": 0.39, + "learning_rate": 1.396777118918147e-05, + "loss": 0.5322, + "step": 12165 + }, + { + "epoch": 0.39, + "learning_rate": 1.3966820388152007e-05, + "loss": 1.0479, + "step": 12166 + }, + { + "epoch": 0.39, + "learning_rate": 1.3965869544562945e-05, + "loss": 1.1714, + "step": 12167 + }, + { + "epoch": 0.39, + "learning_rate": 1.3964918658424492e-05, + "loss": 0.8101, + "step": 12168 + }, + { + "epoch": 0.39, + "learning_rate": 1.396396772974685e-05, + "loss": 0.541, + "step": 12169 + }, + { + "epoch": 0.39, + "learning_rate": 1.3963016758540224e-05, + "loss": 1.0791, + "step": 12170 + }, + { + "epoch": 0.39, + "learning_rate": 1.3962065744814813e-05, + "loss": 0.9609, + "step": 12171 + }, + { + "epoch": 0.39, + "learning_rate": 1.396111468858082e-05, + "loss": 0.8745, + "step": 12172 + }, + { + "epoch": 0.39, + "learning_rate": 1.3960163589848451e-05, + "loss": 1.0498, + "step": 12173 + }, + { + "epoch": 0.39, + "learning_rate": 1.3959212448627912e-05, + "loss": 1.0073, + "step": 12174 + }, + { + "epoch": 0.39, + "learning_rate": 1.3958261264929405e-05, + "loss": 0.9521, + "step": 12175 + }, + { + "epoch": 0.39, + "learning_rate": 1.3957310038763131e-05, + "loss": 1.0815, + "step": 12176 + }, + { + "epoch": 0.39, + "learning_rate": 1.3956358770139304e-05, + "loss": 1.0874, + "step": 12177 + }, + { + "epoch": 0.39, + "learning_rate": 1.3955407459068126e-05, + "loss": 0.9092, + "step": 12178 + }, + { + "epoch": 0.39, + "learning_rate": 1.3954456105559802e-05, + "loss": 1.1187, + "step": 12179 + }, + { + "epoch": 0.39, + "learning_rate": 1.395350470962454e-05, + "loss": 0.9678, + "step": 12180 + }, + { + "epoch": 0.39, + "learning_rate": 1.3952553271272551e-05, + "loss": 1.0566, + "step": 12181 + }, + { + "epoch": 0.39, + "learning_rate": 1.3951601790514038e-05, + "loss": 0.9614, + "step": 12182 + }, + { + "epoch": 0.39, + "learning_rate": 1.3950650267359209e-05, + "loss": 0.9927, + "step": 12183 + }, + { + "epoch": 0.39, + "learning_rate": 1.3949698701818277e-05, + "loss": 1.0952, + "step": 12184 + }, + { + "epoch": 0.39, + "learning_rate": 1.3948747093901452e-05, + "loss": 0.9829, + "step": 12185 + }, + { + "epoch": 0.39, + "learning_rate": 1.3947795443618937e-05, + "loss": 0.8359, + "step": 12186 + }, + { + "epoch": 0.39, + "learning_rate": 1.3946843750980947e-05, + "loss": 0.9316, + "step": 12187 + }, + { + "epoch": 0.39, + "learning_rate": 1.394589201599769e-05, + "loss": 0.9751, + "step": 12188 + }, + { + "epoch": 0.39, + "learning_rate": 1.3944940238679384e-05, + "loss": 1.0674, + "step": 12189 + }, + { + "epoch": 0.39, + "learning_rate": 1.3943988419036231e-05, + "loss": 1.1382, + "step": 12190 + }, + { + "epoch": 0.39, + "learning_rate": 1.3943036557078448e-05, + "loss": 0.9399, + "step": 12191 + }, + { + "epoch": 0.39, + "learning_rate": 1.3942084652816247e-05, + "loss": 1.0537, + "step": 12192 + }, + { + "epoch": 0.39, + "learning_rate": 1.3941132706259841e-05, + "loss": 0.9819, + "step": 12193 + }, + { + "epoch": 0.39, + "learning_rate": 1.3940180717419443e-05, + "loss": 0.9399, + "step": 12194 + }, + { + "epoch": 0.39, + "learning_rate": 1.3939228686305266e-05, + "loss": 0.9883, + "step": 12195 + }, + { + "epoch": 0.39, + "learning_rate": 1.3938276612927526e-05, + "loss": 1.1313, + "step": 12196 + }, + { + "epoch": 0.39, + "learning_rate": 1.3937324497296433e-05, + "loss": 1.0093, + "step": 12197 + }, + { + "epoch": 0.39, + "learning_rate": 1.3936372339422208e-05, + "loss": 1.0054, + "step": 12198 + }, + { + "epoch": 0.39, + "learning_rate": 1.3935420139315065e-05, + "loss": 0.9565, + "step": 12199 + }, + { + "epoch": 0.39, + "learning_rate": 1.3934467896985223e-05, + "loss": 1.0503, + "step": 12200 + }, + { + "epoch": 0.39, + "learning_rate": 1.3933515612442889e-05, + "loss": 0.9121, + "step": 12201 + }, + { + "epoch": 0.39, + "learning_rate": 1.3932563285698289e-05, + "loss": 1.0264, + "step": 12202 + }, + { + "epoch": 0.39, + "learning_rate": 1.3931610916761632e-05, + "loss": 1.0444, + "step": 12203 + }, + { + "epoch": 0.39, + "learning_rate": 1.3930658505643149e-05, + "loss": 1.0132, + "step": 12204 + }, + { + "epoch": 0.39, + "learning_rate": 1.3929706052353048e-05, + "loss": 0.9688, + "step": 12205 + }, + { + "epoch": 0.39, + "learning_rate": 1.3928753556901546e-05, + "loss": 0.8745, + "step": 12206 + }, + { + "epoch": 0.39, + "learning_rate": 1.392780101929887e-05, + "loss": 0.999, + "step": 12207 + }, + { + "epoch": 0.39, + "learning_rate": 1.3926848439555239e-05, + "loss": 1.0791, + "step": 12208 + }, + { + "epoch": 0.39, + "learning_rate": 1.3925895817680865e-05, + "loss": 0.9253, + "step": 12209 + }, + { + "epoch": 0.39, + "learning_rate": 1.3924943153685975e-05, + "loss": 1.002, + "step": 12210 + }, + { + "epoch": 0.39, + "learning_rate": 1.392399044758079e-05, + "loss": 0.9946, + "step": 12211 + }, + { + "epoch": 0.39, + "learning_rate": 1.392303769937553e-05, + "loss": 1.0562, + "step": 12212 + }, + { + "epoch": 0.39, + "learning_rate": 1.3922084909080418e-05, + "loss": 1.0015, + "step": 12213 + }, + { + "epoch": 0.39, + "learning_rate": 1.3921132076705674e-05, + "loss": 0.9077, + "step": 12214 + }, + { + "epoch": 0.39, + "learning_rate": 1.3920179202261526e-05, + "loss": 1.0308, + "step": 12215 + }, + { + "epoch": 0.39, + "learning_rate": 1.391922628575819e-05, + "loss": 1.1147, + "step": 12216 + }, + { + "epoch": 0.39, + "learning_rate": 1.3918273327205896e-05, + "loss": 1.0728, + "step": 12217 + }, + { + "epoch": 0.39, + "learning_rate": 1.3917320326614865e-05, + "loss": 0.8379, + "step": 12218 + }, + { + "epoch": 0.39, + "learning_rate": 1.391636728399532e-05, + "loss": 0.9478, + "step": 12219 + }, + { + "epoch": 0.39, + "learning_rate": 1.3915414199357494e-05, + "loss": 1.0591, + "step": 12220 + }, + { + "epoch": 0.39, + "learning_rate": 1.3914461072711602e-05, + "loss": 1.042, + "step": 12221 + }, + { + "epoch": 0.39, + "learning_rate": 1.3913507904067876e-05, + "loss": 1.1133, + "step": 12222 + }, + { + "epoch": 0.39, + "learning_rate": 1.3912554693436543e-05, + "loss": 0.9956, + "step": 12223 + }, + { + "epoch": 0.39, + "learning_rate": 1.3911601440827827e-05, + "loss": 0.9004, + "step": 12224 + }, + { + "epoch": 0.39, + "learning_rate": 1.3910648146251958e-05, + "loss": 1.0078, + "step": 12225 + }, + { + "epoch": 0.39, + "learning_rate": 1.390969480971916e-05, + "loss": 0.9844, + "step": 12226 + }, + { + "epoch": 0.39, + "learning_rate": 1.3908741431239668e-05, + "loss": 1.0796, + "step": 12227 + }, + { + "epoch": 0.39, + "learning_rate": 1.3907788010823702e-05, + "loss": 0.9858, + "step": 12228 + }, + { + "epoch": 0.39, + "learning_rate": 1.3906834548481496e-05, + "loss": 1.0371, + "step": 12229 + }, + { + "epoch": 0.39, + "learning_rate": 1.3905881044223282e-05, + "loss": 0.9155, + "step": 12230 + }, + { + "epoch": 0.39, + "learning_rate": 1.3904927498059285e-05, + "loss": 0.8662, + "step": 12231 + }, + { + "epoch": 0.39, + "learning_rate": 1.3903973909999737e-05, + "loss": 1.0562, + "step": 12232 + }, + { + "epoch": 0.39, + "learning_rate": 1.3903020280054871e-05, + "loss": 0.9604, + "step": 12233 + }, + { + "epoch": 0.39, + "learning_rate": 1.3902066608234919e-05, + "loss": 0.9082, + "step": 12234 + }, + { + "epoch": 0.39, + "learning_rate": 1.3901112894550109e-05, + "loss": 1.1255, + "step": 12235 + }, + { + "epoch": 0.39, + "learning_rate": 1.3900159139010674e-05, + "loss": 1.0337, + "step": 12236 + }, + { + "epoch": 0.39, + "learning_rate": 1.3899205341626847e-05, + "loss": 1.0166, + "step": 12237 + }, + { + "epoch": 0.39, + "learning_rate": 1.3898251502408866e-05, + "loss": 1.0469, + "step": 12238 + }, + { + "epoch": 0.39, + "learning_rate": 1.3897297621366958e-05, + "loss": 0.9844, + "step": 12239 + }, + { + "epoch": 0.39, + "learning_rate": 1.3896343698511362e-05, + "loss": 1.0005, + "step": 12240 + }, + { + "epoch": 0.39, + "learning_rate": 1.3895389733852307e-05, + "loss": 0.9746, + "step": 12241 + }, + { + "epoch": 0.39, + "learning_rate": 1.3894435727400036e-05, + "loss": 0.999, + "step": 12242 + }, + { + "epoch": 0.39, + "learning_rate": 1.3893481679164776e-05, + "loss": 1.0137, + "step": 12243 + }, + { + "epoch": 0.39, + "learning_rate": 1.3892527589156765e-05, + "loss": 0.8931, + "step": 12244 + }, + { + "epoch": 0.39, + "learning_rate": 1.3891573457386247e-05, + "loss": 0.9966, + "step": 12245 + }, + { + "epoch": 0.39, + "learning_rate": 1.3890619283863449e-05, + "loss": 1.0269, + "step": 12246 + }, + { + "epoch": 0.39, + "learning_rate": 1.3889665068598613e-05, + "loss": 1.1045, + "step": 12247 + }, + { + "epoch": 0.39, + "learning_rate": 1.3888710811601973e-05, + "loss": 1.0649, + "step": 12248 + }, + { + "epoch": 0.39, + "learning_rate": 1.3887756512883775e-05, + "loss": 1.0532, + "step": 12249 + }, + { + "epoch": 0.39, + "learning_rate": 1.3886802172454247e-05, + "loss": 0.8784, + "step": 12250 + }, + { + "epoch": 0.39, + "learning_rate": 1.3885847790323636e-05, + "loss": 0.8789, + "step": 12251 + }, + { + "epoch": 0.39, + "learning_rate": 1.3884893366502178e-05, + "loss": 1.0425, + "step": 12252 + }, + { + "epoch": 0.39, + "learning_rate": 1.3883938901000115e-05, + "loss": 0.9849, + "step": 12253 + }, + { + "epoch": 0.39, + "learning_rate": 1.3882984393827688e-05, + "loss": 1.021, + "step": 12254 + }, + { + "epoch": 0.39, + "learning_rate": 1.3882029844995131e-05, + "loss": 1.0083, + "step": 12255 + }, + { + "epoch": 0.39, + "learning_rate": 1.3881075254512694e-05, + "loss": 1.1172, + "step": 12256 + }, + { + "epoch": 0.39, + "learning_rate": 1.3880120622390613e-05, + "loss": 1.1187, + "step": 12257 + }, + { + "epoch": 0.39, + "learning_rate": 1.3879165948639134e-05, + "loss": 0.9004, + "step": 12258 + }, + { + "epoch": 0.39, + "learning_rate": 1.3878211233268497e-05, + "loss": 0.9263, + "step": 12259 + }, + { + "epoch": 0.39, + "learning_rate": 1.3877256476288943e-05, + "loss": 1.085, + "step": 12260 + }, + { + "epoch": 0.39, + "learning_rate": 1.387630167771072e-05, + "loss": 0.5613, + "step": 12261 + }, + { + "epoch": 0.39, + "learning_rate": 1.3875346837544072e-05, + "loss": 0.9829, + "step": 12262 + }, + { + "epoch": 0.39, + "learning_rate": 1.387439195579924e-05, + "loss": 0.9854, + "step": 12263 + }, + { + "epoch": 0.39, + "learning_rate": 1.387343703248647e-05, + "loss": 1.1294, + "step": 12264 + }, + { + "epoch": 0.39, + "learning_rate": 1.3872482067616006e-05, + "loss": 0.8945, + "step": 12265 + }, + { + "epoch": 0.39, + "learning_rate": 1.3871527061198098e-05, + "loss": 1.0161, + "step": 12266 + }, + { + "epoch": 0.39, + "learning_rate": 1.3870572013242986e-05, + "loss": 1.0039, + "step": 12267 + }, + { + "epoch": 0.39, + "learning_rate": 1.3869616923760924e-05, + "loss": 1.0649, + "step": 12268 + }, + { + "epoch": 0.39, + "learning_rate": 1.3868661792762152e-05, + "loss": 0.9873, + "step": 12269 + }, + { + "epoch": 0.39, + "learning_rate": 1.3867706620256922e-05, + "loss": 1.1426, + "step": 12270 + }, + { + "epoch": 0.39, + "learning_rate": 1.386675140625548e-05, + "loss": 0.998, + "step": 12271 + }, + { + "epoch": 0.39, + "learning_rate": 1.3865796150768076e-05, + "loss": 1.0396, + "step": 12272 + }, + { + "epoch": 0.39, + "learning_rate": 1.3864840853804958e-05, + "loss": 1.1387, + "step": 12273 + }, + { + "epoch": 0.39, + "learning_rate": 1.3863885515376372e-05, + "loss": 0.9453, + "step": 12274 + }, + { + "epoch": 0.39, + "learning_rate": 1.3862930135492572e-05, + "loss": 1.0151, + "step": 12275 + }, + { + "epoch": 0.39, + "learning_rate": 1.3861974714163811e-05, + "loss": 1.0127, + "step": 12276 + }, + { + "epoch": 0.39, + "learning_rate": 1.3861019251400331e-05, + "loss": 1.0142, + "step": 12277 + }, + { + "epoch": 0.39, + "learning_rate": 1.386006374721239e-05, + "loss": 0.9062, + "step": 12278 + }, + { + "epoch": 0.39, + "learning_rate": 1.3859108201610236e-05, + "loss": 1.0737, + "step": 12279 + }, + { + "epoch": 0.39, + "learning_rate": 1.385815261460412e-05, + "loss": 1.0654, + "step": 12280 + }, + { + "epoch": 0.39, + "learning_rate": 1.38571969862043e-05, + "loss": 0.9585, + "step": 12281 + }, + { + "epoch": 0.39, + "learning_rate": 1.3856241316421021e-05, + "loss": 0.9592, + "step": 12282 + }, + { + "epoch": 0.39, + "learning_rate": 1.3855285605264547e-05, + "loss": 0.4976, + "step": 12283 + }, + { + "epoch": 0.39, + "learning_rate": 1.385432985274512e-05, + "loss": 1.0635, + "step": 12284 + }, + { + "epoch": 0.39, + "learning_rate": 1.3853374058873002e-05, + "loss": 0.9053, + "step": 12285 + }, + { + "epoch": 0.39, + "learning_rate": 1.3852418223658442e-05, + "loss": 0.9839, + "step": 12286 + }, + { + "epoch": 0.39, + "learning_rate": 1.3851462347111702e-05, + "loss": 0.9326, + "step": 12287 + }, + { + "epoch": 0.39, + "learning_rate": 1.3850506429243032e-05, + "loss": 0.9404, + "step": 12288 + }, + { + "epoch": 0.39, + "learning_rate": 1.3849550470062688e-05, + "loss": 0.9941, + "step": 12289 + }, + { + "epoch": 0.39, + "learning_rate": 1.3848594469580927e-05, + "loss": 0.9673, + "step": 12290 + }, + { + "epoch": 0.39, + "learning_rate": 1.3847638427808012e-05, + "loss": 1.0693, + "step": 12291 + }, + { + "epoch": 0.39, + "learning_rate": 1.3846682344754192e-05, + "loss": 1.0342, + "step": 12292 + }, + { + "epoch": 0.39, + "learning_rate": 1.3845726220429728e-05, + "loss": 0.9814, + "step": 12293 + }, + { + "epoch": 0.39, + "learning_rate": 1.3844770054844876e-05, + "loss": 1.0386, + "step": 12294 + }, + { + "epoch": 0.39, + "learning_rate": 1.38438138480099e-05, + "loss": 0.9683, + "step": 12295 + }, + { + "epoch": 0.39, + "learning_rate": 1.384285759993505e-05, + "loss": 1.0801, + "step": 12296 + }, + { + "epoch": 0.39, + "learning_rate": 1.3841901310630597e-05, + "loss": 0.9893, + "step": 12297 + }, + { + "epoch": 0.39, + "learning_rate": 1.3840944980106792e-05, + "loss": 0.9517, + "step": 12298 + }, + { + "epoch": 0.39, + "learning_rate": 1.38399886083739e-05, + "loss": 1.0854, + "step": 12299 + }, + { + "epoch": 0.39, + "learning_rate": 1.3839032195442177e-05, + "loss": 0.9722, + "step": 12300 + }, + { + "epoch": 0.39, + "learning_rate": 1.3838075741321893e-05, + "loss": 0.8584, + "step": 12301 + }, + { + "epoch": 0.39, + "learning_rate": 1.3837119246023298e-05, + "loss": 1.0742, + "step": 12302 + }, + { + "epoch": 0.39, + "learning_rate": 1.3836162709556665e-05, + "loss": 0.998, + "step": 12303 + }, + { + "epoch": 0.39, + "learning_rate": 1.3835206131932248e-05, + "loss": 1.0747, + "step": 12304 + }, + { + "epoch": 0.39, + "learning_rate": 1.3834249513160316e-05, + "loss": 0.9878, + "step": 12305 + }, + { + "epoch": 0.39, + "learning_rate": 1.383329285325113e-05, + "loss": 1.165, + "step": 12306 + }, + { + "epoch": 0.39, + "learning_rate": 1.3832336152214953e-05, + "loss": 0.9517, + "step": 12307 + }, + { + "epoch": 0.39, + "learning_rate": 1.3831379410062052e-05, + "loss": 0.9595, + "step": 12308 + }, + { + "epoch": 0.39, + "learning_rate": 1.3830422626802691e-05, + "loss": 1.084, + "step": 12309 + }, + { + "epoch": 0.39, + "learning_rate": 1.3829465802447133e-05, + "loss": 0.959, + "step": 12310 + }, + { + "epoch": 0.39, + "learning_rate": 1.3828508937005642e-05, + "loss": 0.9824, + "step": 12311 + }, + { + "epoch": 0.39, + "learning_rate": 1.3827552030488492e-05, + "loss": 0.9268, + "step": 12312 + }, + { + "epoch": 0.39, + "learning_rate": 1.3826595082905945e-05, + "loss": 1.0352, + "step": 12313 + }, + { + "epoch": 0.39, + "learning_rate": 1.3825638094268263e-05, + "loss": 0.9995, + "step": 12314 + }, + { + "epoch": 0.39, + "learning_rate": 1.3824681064585721e-05, + "loss": 1.0918, + "step": 12315 + }, + { + "epoch": 0.39, + "learning_rate": 1.3823723993868584e-05, + "loss": 1.0889, + "step": 12316 + }, + { + "epoch": 0.39, + "learning_rate": 1.3822766882127123e-05, + "loss": 1.1558, + "step": 12317 + }, + { + "epoch": 0.39, + "learning_rate": 1.3821809729371602e-05, + "loss": 1.0054, + "step": 12318 + }, + { + "epoch": 0.39, + "learning_rate": 1.3820852535612293e-05, + "loss": 1.1069, + "step": 12319 + }, + { + "epoch": 0.39, + "learning_rate": 1.3819895300859463e-05, + "loss": 0.9429, + "step": 12320 + }, + { + "epoch": 0.39, + "learning_rate": 1.3818938025123387e-05, + "loss": 0.5701, + "step": 12321 + }, + { + "epoch": 0.39, + "learning_rate": 1.381798070841433e-05, + "loss": 0.9634, + "step": 12322 + }, + { + "epoch": 0.39, + "learning_rate": 1.3817023350742567e-05, + "loss": 1.0776, + "step": 12323 + }, + { + "epoch": 0.39, + "learning_rate": 1.3816065952118368e-05, + "loss": 0.9531, + "step": 12324 + }, + { + "epoch": 0.39, + "learning_rate": 1.3815108512552004e-05, + "loss": 1.1187, + "step": 12325 + }, + { + "epoch": 0.39, + "learning_rate": 1.381415103205375e-05, + "loss": 0.9463, + "step": 12326 + }, + { + "epoch": 0.39, + "learning_rate": 1.3813193510633873e-05, + "loss": 0.9766, + "step": 12327 + }, + { + "epoch": 0.39, + "learning_rate": 1.3812235948302655e-05, + "loss": 1.0229, + "step": 12328 + }, + { + "epoch": 0.39, + "learning_rate": 1.381127834507036e-05, + "loss": 0.9902, + "step": 12329 + }, + { + "epoch": 0.39, + "learning_rate": 1.3810320700947266e-05, + "loss": 0.8975, + "step": 12330 + }, + { + "epoch": 0.39, + "learning_rate": 1.3809363015943651e-05, + "loss": 0.5203, + "step": 12331 + }, + { + "epoch": 0.39, + "learning_rate": 1.3808405290069785e-05, + "loss": 0.9951, + "step": 12332 + }, + { + "epoch": 0.39, + "learning_rate": 1.3807447523335947e-05, + "loss": 0.533, + "step": 12333 + }, + { + "epoch": 0.39, + "learning_rate": 1.3806489715752409e-05, + "loss": 1.0073, + "step": 12334 + }, + { + "epoch": 0.39, + "learning_rate": 1.3805531867329449e-05, + "loss": 0.9458, + "step": 12335 + }, + { + "epoch": 0.39, + "learning_rate": 1.3804573978077346e-05, + "loss": 1.0859, + "step": 12336 + }, + { + "epoch": 0.39, + "learning_rate": 1.3803616048006375e-05, + "loss": 1.0547, + "step": 12337 + }, + { + "epoch": 0.39, + "learning_rate": 1.380265807712681e-05, + "loss": 1.0063, + "step": 12338 + }, + { + "epoch": 0.39, + "learning_rate": 1.3801700065448933e-05, + "loss": 1.0586, + "step": 12339 + }, + { + "epoch": 0.39, + "learning_rate": 1.3800742012983026e-05, + "loss": 0.9233, + "step": 12340 + }, + { + "epoch": 0.39, + "learning_rate": 1.379978391973936e-05, + "loss": 1.1704, + "step": 12341 + }, + { + "epoch": 0.39, + "learning_rate": 1.3798825785728219e-05, + "loss": 0.5071, + "step": 12342 + }, + { + "epoch": 0.39, + "learning_rate": 1.3797867610959883e-05, + "loss": 0.9487, + "step": 12343 + }, + { + "epoch": 0.39, + "learning_rate": 1.3796909395444629e-05, + "loss": 0.936, + "step": 12344 + }, + { + "epoch": 0.39, + "learning_rate": 1.379595113919274e-05, + "loss": 0.9312, + "step": 12345 + }, + { + "epoch": 0.39, + "learning_rate": 1.3794992842214496e-05, + "loss": 1.0503, + "step": 12346 + }, + { + "epoch": 0.39, + "learning_rate": 1.379403450452018e-05, + "loss": 1.0566, + "step": 12347 + }, + { + "epoch": 0.39, + "learning_rate": 1.3793076126120073e-05, + "loss": 1.0444, + "step": 12348 + }, + { + "epoch": 0.39, + "learning_rate": 1.3792117707024455e-05, + "loss": 1.144, + "step": 12349 + }, + { + "epoch": 0.39, + "learning_rate": 1.379115924724361e-05, + "loss": 0.5183, + "step": 12350 + }, + { + "epoch": 0.39, + "learning_rate": 1.3790200746787825e-05, + "loss": 1.0913, + "step": 12351 + }, + { + "epoch": 0.4, + "learning_rate": 1.3789242205667382e-05, + "loss": 0.9956, + "step": 12352 + }, + { + "epoch": 0.4, + "learning_rate": 1.378828362389256e-05, + "loss": 0.9961, + "step": 12353 + }, + { + "epoch": 0.4, + "learning_rate": 1.3787325001473649e-05, + "loss": 1.1055, + "step": 12354 + }, + { + "epoch": 0.4, + "learning_rate": 1.3786366338420934e-05, + "loss": 1.0547, + "step": 12355 + }, + { + "epoch": 0.4, + "learning_rate": 1.3785407634744696e-05, + "loss": 0.9976, + "step": 12356 + }, + { + "epoch": 0.4, + "learning_rate": 1.3784448890455225e-05, + "loss": 1.1011, + "step": 12357 + }, + { + "epoch": 0.4, + "learning_rate": 1.3783490105562805e-05, + "loss": 1.0, + "step": 12358 + }, + { + "epoch": 0.4, + "learning_rate": 1.3782531280077724e-05, + "loss": 0.9814, + "step": 12359 + }, + { + "epoch": 0.4, + "learning_rate": 1.3781572414010269e-05, + "loss": 1.0859, + "step": 12360 + }, + { + "epoch": 0.4, + "learning_rate": 1.3780613507370727e-05, + "loss": 1.0684, + "step": 12361 + }, + { + "epoch": 0.4, + "learning_rate": 1.3779654560169387e-05, + "loss": 1.0747, + "step": 12362 + }, + { + "epoch": 0.4, + "learning_rate": 1.3778695572416535e-05, + "loss": 1.0806, + "step": 12363 + }, + { + "epoch": 0.4, + "learning_rate": 1.3777736544122461e-05, + "loss": 1.1699, + "step": 12364 + }, + { + "epoch": 0.4, + "learning_rate": 1.3776777475297456e-05, + "loss": 1.1196, + "step": 12365 + }, + { + "epoch": 0.4, + "learning_rate": 1.377581836595181e-05, + "loss": 0.9917, + "step": 12366 + }, + { + "epoch": 0.4, + "learning_rate": 1.377485921609581e-05, + "loss": 0.9229, + "step": 12367 + }, + { + "epoch": 0.4, + "learning_rate": 1.3773900025739751e-05, + "loss": 0.9204, + "step": 12368 + }, + { + "epoch": 0.4, + "learning_rate": 1.3772940794893916e-05, + "loss": 0.8296, + "step": 12369 + }, + { + "epoch": 0.4, + "learning_rate": 1.377198152356861e-05, + "loss": 0.9087, + "step": 12370 + }, + { + "epoch": 0.4, + "learning_rate": 1.377102221177411e-05, + "loss": 1.1587, + "step": 12371 + }, + { + "epoch": 0.4, + "learning_rate": 1.3770062859520717e-05, + "loss": 0.9775, + "step": 12372 + }, + { + "epoch": 0.4, + "learning_rate": 1.3769103466818722e-05, + "loss": 1.1753, + "step": 12373 + }, + { + "epoch": 0.4, + "learning_rate": 1.3768144033678421e-05, + "loss": 1.0918, + "step": 12374 + }, + { + "epoch": 0.4, + "learning_rate": 1.3767184560110103e-05, + "loss": 1.0283, + "step": 12375 + }, + { + "epoch": 0.4, + "learning_rate": 1.3766225046124062e-05, + "loss": 1.0327, + "step": 12376 + }, + { + "epoch": 0.4, + "learning_rate": 1.3765265491730598e-05, + "loss": 0.9561, + "step": 12377 + }, + { + "epoch": 0.4, + "learning_rate": 1.376430589694e-05, + "loss": 0.9761, + "step": 12378 + }, + { + "epoch": 0.4, + "learning_rate": 1.3763346261762568e-05, + "loss": 0.9966, + "step": 12379 + }, + { + "epoch": 0.4, + "learning_rate": 1.3762386586208591e-05, + "loss": 0.9185, + "step": 12380 + }, + { + "epoch": 0.4, + "learning_rate": 1.3761426870288377e-05, + "loss": 0.9673, + "step": 12381 + }, + { + "epoch": 0.4, + "learning_rate": 1.376046711401221e-05, + "loss": 0.9873, + "step": 12382 + }, + { + "epoch": 0.4, + "learning_rate": 1.3759507317390394e-05, + "loss": 1.0503, + "step": 12383 + }, + { + "epoch": 0.4, + "learning_rate": 1.375854748043323e-05, + "loss": 1.0352, + "step": 12384 + }, + { + "epoch": 0.4, + "learning_rate": 1.3757587603151006e-05, + "loss": 1.1396, + "step": 12385 + }, + { + "epoch": 0.4, + "learning_rate": 1.3756627685554024e-05, + "loss": 0.9302, + "step": 12386 + }, + { + "epoch": 0.4, + "learning_rate": 1.375566772765259e-05, + "loss": 0.9111, + "step": 12387 + }, + { + "epoch": 0.4, + "learning_rate": 1.3754707729456995e-05, + "loss": 1.0024, + "step": 12388 + }, + { + "epoch": 0.4, + "learning_rate": 1.3753747690977542e-05, + "loss": 1.0801, + "step": 12389 + }, + { + "epoch": 0.4, + "learning_rate": 1.3752787612224533e-05, + "loss": 1.063, + "step": 12390 + }, + { + "epoch": 0.4, + "learning_rate": 1.375182749320826e-05, + "loss": 1.1108, + "step": 12391 + }, + { + "epoch": 0.4, + "learning_rate": 1.3750867333939035e-05, + "loss": 0.9268, + "step": 12392 + }, + { + "epoch": 0.4, + "learning_rate": 1.3749907134427155e-05, + "loss": 1.0557, + "step": 12393 + }, + { + "epoch": 0.4, + "learning_rate": 1.3748946894682922e-05, + "loss": 1.1089, + "step": 12394 + }, + { + "epoch": 0.4, + "learning_rate": 1.3747986614716636e-05, + "loss": 0.9902, + "step": 12395 + }, + { + "epoch": 0.4, + "learning_rate": 1.3747026294538604e-05, + "loss": 1.0322, + "step": 12396 + }, + { + "epoch": 0.4, + "learning_rate": 1.3746065934159123e-05, + "loss": 1.0474, + "step": 12397 + }, + { + "epoch": 0.4, + "learning_rate": 1.3745105533588504e-05, + "loss": 1.1411, + "step": 12398 + }, + { + "epoch": 0.4, + "learning_rate": 1.3744145092837045e-05, + "loss": 0.8721, + "step": 12399 + }, + { + "epoch": 0.4, + "learning_rate": 1.3743184611915055e-05, + "loss": 0.9878, + "step": 12400 + }, + { + "epoch": 0.4, + "learning_rate": 1.3742224090832834e-05, + "loss": 1.0415, + "step": 12401 + }, + { + "epoch": 0.4, + "learning_rate": 1.3741263529600693e-05, + "loss": 1.0337, + "step": 12402 + }, + { + "epoch": 0.4, + "learning_rate": 1.3740302928228933e-05, + "loss": 1.0288, + "step": 12403 + }, + { + "epoch": 0.4, + "learning_rate": 1.3739342286727863e-05, + "loss": 1.1167, + "step": 12404 + }, + { + "epoch": 0.4, + "learning_rate": 1.373838160510779e-05, + "loss": 1.0283, + "step": 12405 + }, + { + "epoch": 0.4, + "learning_rate": 1.3737420883379017e-05, + "loss": 0.9067, + "step": 12406 + }, + { + "epoch": 0.4, + "learning_rate": 1.3736460121551856e-05, + "loss": 0.9902, + "step": 12407 + }, + { + "epoch": 0.4, + "learning_rate": 1.3735499319636615e-05, + "loss": 1.0293, + "step": 12408 + }, + { + "epoch": 0.4, + "learning_rate": 1.3734538477643598e-05, + "loss": 1.0015, + "step": 12409 + }, + { + "epoch": 0.4, + "learning_rate": 1.3733577595583116e-05, + "loss": 0.8975, + "step": 12410 + }, + { + "epoch": 0.4, + "learning_rate": 1.3732616673465481e-05, + "loss": 0.7896, + "step": 12411 + }, + { + "epoch": 0.4, + "learning_rate": 1.3731655711301e-05, + "loss": 0.9385, + "step": 12412 + }, + { + "epoch": 0.4, + "learning_rate": 1.373069470909998e-05, + "loss": 1.2349, + "step": 12413 + }, + { + "epoch": 0.4, + "learning_rate": 1.3729733666872736e-05, + "loss": 0.9395, + "step": 12414 + }, + { + "epoch": 0.4, + "learning_rate": 1.3728772584629582e-05, + "loss": 0.9526, + "step": 12415 + }, + { + "epoch": 0.4, + "learning_rate": 1.372781146238082e-05, + "loss": 1.041, + "step": 12416 + }, + { + "epoch": 0.4, + "learning_rate": 1.3726850300136767e-05, + "loss": 1.0029, + "step": 12417 + }, + { + "epoch": 0.4, + "learning_rate": 1.3725889097907734e-05, + "loss": 0.874, + "step": 12418 + }, + { + "epoch": 0.4, + "learning_rate": 1.372492785570404e-05, + "loss": 0.9058, + "step": 12419 + }, + { + "epoch": 0.4, + "learning_rate": 1.3723966573535988e-05, + "loss": 1.0308, + "step": 12420 + }, + { + "epoch": 0.4, + "learning_rate": 1.3723005251413896e-05, + "loss": 1.0254, + "step": 12421 + }, + { + "epoch": 0.4, + "learning_rate": 1.372204388934808e-05, + "loss": 1.0645, + "step": 12422 + }, + { + "epoch": 0.4, + "learning_rate": 1.3721082487348853e-05, + "loss": 0.8916, + "step": 12423 + }, + { + "epoch": 0.4, + "learning_rate": 1.3720121045426526e-05, + "loss": 0.9619, + "step": 12424 + }, + { + "epoch": 0.4, + "learning_rate": 1.371915956359142e-05, + "loss": 1.0649, + "step": 12425 + }, + { + "epoch": 0.4, + "learning_rate": 1.371819804185385e-05, + "loss": 0.5483, + "step": 12426 + }, + { + "epoch": 0.4, + "learning_rate": 1.3717236480224128e-05, + "loss": 0.533, + "step": 12427 + }, + { + "epoch": 0.4, + "learning_rate": 1.371627487871257e-05, + "loss": 0.9907, + "step": 12428 + }, + { + "epoch": 0.4, + "learning_rate": 1.37153132373295e-05, + "loss": 0.9692, + "step": 12429 + }, + { + "epoch": 0.4, + "learning_rate": 1.3714351556085229e-05, + "loss": 1.0576, + "step": 12430 + }, + { + "epoch": 0.4, + "learning_rate": 1.3713389834990075e-05, + "loss": 1.0654, + "step": 12431 + }, + { + "epoch": 0.4, + "learning_rate": 1.371242807405436e-05, + "loss": 0.981, + "step": 12432 + }, + { + "epoch": 0.4, + "learning_rate": 1.3711466273288398e-05, + "loss": 0.9146, + "step": 12433 + }, + { + "epoch": 0.4, + "learning_rate": 1.3710504432702517e-05, + "loss": 1.0254, + "step": 12434 + }, + { + "epoch": 0.4, + "learning_rate": 1.3709542552307023e-05, + "loss": 0.9429, + "step": 12435 + }, + { + "epoch": 0.4, + "learning_rate": 1.3708580632112245e-05, + "loss": 1.0562, + "step": 12436 + }, + { + "epoch": 0.4, + "learning_rate": 1.37076186721285e-05, + "loss": 0.9702, + "step": 12437 + }, + { + "epoch": 0.4, + "learning_rate": 1.3706656672366112e-05, + "loss": 0.552, + "step": 12438 + }, + { + "epoch": 0.4, + "learning_rate": 1.37056946328354e-05, + "loss": 1.0547, + "step": 12439 + }, + { + "epoch": 0.4, + "learning_rate": 1.3704732553546685e-05, + "loss": 1.0371, + "step": 12440 + }, + { + "epoch": 0.4, + "learning_rate": 1.3703770434510291e-05, + "loss": 1.0371, + "step": 12441 + }, + { + "epoch": 0.4, + "learning_rate": 1.3702808275736538e-05, + "loss": 0.9727, + "step": 12442 + }, + { + "epoch": 0.4, + "learning_rate": 1.3701846077235751e-05, + "loss": 1.1128, + "step": 12443 + }, + { + "epoch": 0.4, + "learning_rate": 1.3700883839018253e-05, + "loss": 0.9619, + "step": 12444 + }, + { + "epoch": 0.4, + "learning_rate": 1.3699921561094367e-05, + "loss": 1.0659, + "step": 12445 + }, + { + "epoch": 0.4, + "learning_rate": 1.3698959243474416e-05, + "loss": 1.1123, + "step": 12446 + }, + { + "epoch": 0.4, + "learning_rate": 1.369799688616873e-05, + "loss": 0.9985, + "step": 12447 + }, + { + "epoch": 0.4, + "learning_rate": 1.3697034489187627e-05, + "loss": 1.0161, + "step": 12448 + }, + { + "epoch": 0.4, + "learning_rate": 1.3696072052541436e-05, + "loss": 1.063, + "step": 12449 + }, + { + "epoch": 0.4, + "learning_rate": 1.3695109576240482e-05, + "loss": 0.5513, + "step": 12450 + }, + { + "epoch": 0.4, + "learning_rate": 1.3694147060295093e-05, + "loss": 1.0264, + "step": 12451 + }, + { + "epoch": 0.4, + "learning_rate": 1.3693184504715594e-05, + "loss": 0.9966, + "step": 12452 + }, + { + "epoch": 0.4, + "learning_rate": 1.3692221909512313e-05, + "loss": 0.9517, + "step": 12453 + }, + { + "epoch": 0.4, + "learning_rate": 1.3691259274695577e-05, + "loss": 0.9351, + "step": 12454 + }, + { + "epoch": 0.4, + "learning_rate": 1.3690296600275715e-05, + "loss": 1.0215, + "step": 12455 + }, + { + "epoch": 0.4, + "learning_rate": 1.3689333886263053e-05, + "loss": 0.9932, + "step": 12456 + }, + { + "epoch": 0.4, + "learning_rate": 1.3688371132667924e-05, + "loss": 0.9878, + "step": 12457 + }, + { + "epoch": 0.4, + "learning_rate": 1.3687408339500652e-05, + "loss": 0.9231, + "step": 12458 + }, + { + "epoch": 0.4, + "learning_rate": 1.368644550677157e-05, + "loss": 0.9766, + "step": 12459 + }, + { + "epoch": 0.4, + "learning_rate": 1.3685482634491012e-05, + "loss": 1.0576, + "step": 12460 + }, + { + "epoch": 0.4, + "learning_rate": 1.3684519722669301e-05, + "loss": 0.9575, + "step": 12461 + }, + { + "epoch": 0.4, + "learning_rate": 1.3683556771316769e-05, + "loss": 0.9019, + "step": 12462 + }, + { + "epoch": 0.4, + "learning_rate": 1.3682593780443751e-05, + "loss": 1.0835, + "step": 12463 + }, + { + "epoch": 0.4, + "learning_rate": 1.368163075006058e-05, + "loss": 0.9429, + "step": 12464 + }, + { + "epoch": 0.4, + "learning_rate": 1.3680667680177581e-05, + "loss": 0.8835, + "step": 12465 + }, + { + "epoch": 0.4, + "learning_rate": 1.3679704570805096e-05, + "loss": 1.1255, + "step": 12466 + }, + { + "epoch": 0.4, + "learning_rate": 1.367874142195345e-05, + "loss": 1.0288, + "step": 12467 + }, + { + "epoch": 0.4, + "learning_rate": 1.3677778233632984e-05, + "loss": 1.0562, + "step": 12468 + }, + { + "epoch": 0.4, + "learning_rate": 1.3676815005854025e-05, + "loss": 0.9927, + "step": 12469 + }, + { + "epoch": 0.4, + "learning_rate": 1.3675851738626911e-05, + "loss": 0.9829, + "step": 12470 + }, + { + "epoch": 0.4, + "learning_rate": 1.3674888431961977e-05, + "loss": 0.98, + "step": 12471 + }, + { + "epoch": 0.4, + "learning_rate": 1.3673925085869555e-05, + "loss": 0.9746, + "step": 12472 + }, + { + "epoch": 0.4, + "learning_rate": 1.3672961700359986e-05, + "loss": 1.0342, + "step": 12473 + }, + { + "epoch": 0.4, + "learning_rate": 1.3671998275443598e-05, + "loss": 1.1069, + "step": 12474 + }, + { + "epoch": 0.4, + "learning_rate": 1.3671034811130737e-05, + "loss": 1.0771, + "step": 12475 + }, + { + "epoch": 0.4, + "learning_rate": 1.3670071307431734e-05, + "loss": 1.0361, + "step": 12476 + }, + { + "epoch": 0.4, + "learning_rate": 1.3669107764356928e-05, + "loss": 1.0288, + "step": 12477 + }, + { + "epoch": 0.4, + "learning_rate": 1.3668144181916655e-05, + "loss": 1.0293, + "step": 12478 + }, + { + "epoch": 0.4, + "learning_rate": 1.3667180560121259e-05, + "loss": 0.9966, + "step": 12479 + }, + { + "epoch": 0.4, + "learning_rate": 1.3666216898981068e-05, + "loss": 0.8621, + "step": 12480 + }, + { + "epoch": 0.4, + "learning_rate": 1.366525319850643e-05, + "loss": 0.9272, + "step": 12481 + }, + { + "epoch": 0.4, + "learning_rate": 1.3664289458707683e-05, + "loss": 1.0605, + "step": 12482 + }, + { + "epoch": 0.4, + "learning_rate": 1.3663325679595166e-05, + "loss": 0.9839, + "step": 12483 + }, + { + "epoch": 0.4, + "learning_rate": 1.3662361861179214e-05, + "loss": 0.9194, + "step": 12484 + }, + { + "epoch": 0.4, + "learning_rate": 1.3661398003470176e-05, + "loss": 1.0532, + "step": 12485 + }, + { + "epoch": 0.4, + "learning_rate": 1.366043410647839e-05, + "loss": 0.9731, + "step": 12486 + }, + { + "epoch": 0.4, + "learning_rate": 1.3659470170214198e-05, + "loss": 1.0483, + "step": 12487 + }, + { + "epoch": 0.4, + "learning_rate": 1.3658506194687937e-05, + "loss": 1.0815, + "step": 12488 + }, + { + "epoch": 0.4, + "learning_rate": 1.3657542179909956e-05, + "loss": 1.0625, + "step": 12489 + }, + { + "epoch": 0.4, + "learning_rate": 1.3656578125890596e-05, + "loss": 1.062, + "step": 12490 + }, + { + "epoch": 0.4, + "learning_rate": 1.3655614032640201e-05, + "loss": 1.0933, + "step": 12491 + }, + { + "epoch": 0.4, + "learning_rate": 1.365464990016911e-05, + "loss": 1.0864, + "step": 12492 + }, + { + "epoch": 0.4, + "learning_rate": 1.3653685728487673e-05, + "loss": 1.0327, + "step": 12493 + }, + { + "epoch": 0.4, + "learning_rate": 1.3652721517606231e-05, + "loss": 1.0459, + "step": 12494 + }, + { + "epoch": 0.4, + "learning_rate": 1.3651757267535129e-05, + "loss": 0.9551, + "step": 12495 + }, + { + "epoch": 0.4, + "learning_rate": 1.3650792978284715e-05, + "loss": 1.1445, + "step": 12496 + }, + { + "epoch": 0.4, + "learning_rate": 1.3649828649865328e-05, + "loss": 1.062, + "step": 12497 + }, + { + "epoch": 0.4, + "learning_rate": 1.3648864282287323e-05, + "loss": 1.0283, + "step": 12498 + }, + { + "epoch": 0.4, + "learning_rate": 1.3647899875561043e-05, + "loss": 1.0127, + "step": 12499 + }, + { + "epoch": 0.4, + "learning_rate": 1.3646935429696835e-05, + "loss": 0.9292, + "step": 12500 + }, + { + "epoch": 0.4, + "learning_rate": 1.3645970944705042e-05, + "loss": 1.02, + "step": 12501 + }, + { + "epoch": 0.4, + "learning_rate": 1.3645006420596021e-05, + "loss": 1.0713, + "step": 12502 + }, + { + "epoch": 0.4, + "learning_rate": 1.3644041857380113e-05, + "loss": 1.0376, + "step": 12503 + }, + { + "epoch": 0.4, + "learning_rate": 1.3643077255067667e-05, + "loss": 1.0698, + "step": 12504 + }, + { + "epoch": 0.4, + "learning_rate": 1.3642112613669037e-05, + "loss": 1.1416, + "step": 12505 + }, + { + "epoch": 0.4, + "learning_rate": 1.364114793319457e-05, + "loss": 1.0845, + "step": 12506 + }, + { + "epoch": 0.4, + "learning_rate": 1.3640183213654612e-05, + "loss": 1.0181, + "step": 12507 + }, + { + "epoch": 0.4, + "learning_rate": 1.3639218455059517e-05, + "loss": 0.9771, + "step": 12508 + }, + { + "epoch": 0.4, + "learning_rate": 1.3638253657419637e-05, + "loss": 1.0366, + "step": 12509 + }, + { + "epoch": 0.4, + "learning_rate": 1.3637288820745325e-05, + "loss": 1.0571, + "step": 12510 + }, + { + "epoch": 0.4, + "learning_rate": 1.3636323945046926e-05, + "loss": 0.9292, + "step": 12511 + }, + { + "epoch": 0.4, + "learning_rate": 1.3635359030334799e-05, + "loss": 1.0659, + "step": 12512 + }, + { + "epoch": 0.4, + "learning_rate": 1.3634394076619288e-05, + "loss": 0.9307, + "step": 12513 + }, + { + "epoch": 0.4, + "learning_rate": 1.3633429083910751e-05, + "loss": 1.0234, + "step": 12514 + }, + { + "epoch": 0.4, + "learning_rate": 1.3632464052219548e-05, + "loss": 1.0518, + "step": 12515 + }, + { + "epoch": 0.4, + "learning_rate": 1.3631498981556019e-05, + "loss": 1.0493, + "step": 12516 + }, + { + "epoch": 0.4, + "learning_rate": 1.3630533871930527e-05, + "loss": 0.9692, + "step": 12517 + }, + { + "epoch": 0.4, + "learning_rate": 1.3629568723353428e-05, + "loss": 1.0869, + "step": 12518 + }, + { + "epoch": 0.4, + "learning_rate": 1.362860353583507e-05, + "loss": 1.063, + "step": 12519 + }, + { + "epoch": 0.4, + "learning_rate": 1.3627638309385811e-05, + "loss": 1.0195, + "step": 12520 + }, + { + "epoch": 0.4, + "learning_rate": 1.362667304401601e-05, + "loss": 0.9287, + "step": 12521 + }, + { + "epoch": 0.4, + "learning_rate": 1.3625707739736021e-05, + "loss": 1.0234, + "step": 12522 + }, + { + "epoch": 0.4, + "learning_rate": 1.3624742396556201e-05, + "loss": 1.0854, + "step": 12523 + }, + { + "epoch": 0.4, + "learning_rate": 1.3623777014486903e-05, + "loss": 1.0132, + "step": 12524 + }, + { + "epoch": 0.4, + "learning_rate": 1.3622811593538495e-05, + "loss": 1.0576, + "step": 12525 + }, + { + "epoch": 0.4, + "learning_rate": 1.3621846133721323e-05, + "loss": 0.9995, + "step": 12526 + }, + { + "epoch": 0.4, + "learning_rate": 1.362088063504575e-05, + "loss": 0.9717, + "step": 12527 + }, + { + "epoch": 0.4, + "learning_rate": 1.361991509752214e-05, + "loss": 0.9805, + "step": 12528 + }, + { + "epoch": 0.4, + "learning_rate": 1.3618949521160843e-05, + "loss": 0.9146, + "step": 12529 + }, + { + "epoch": 0.4, + "learning_rate": 1.3617983905972223e-05, + "loss": 0.8442, + "step": 12530 + }, + { + "epoch": 0.4, + "learning_rate": 1.3617018251966641e-05, + "loss": 0.9697, + "step": 12531 + }, + { + "epoch": 0.4, + "learning_rate": 1.3616052559154458e-05, + "loss": 0.958, + "step": 12532 + }, + { + "epoch": 0.4, + "learning_rate": 1.361508682754603e-05, + "loss": 1.0723, + "step": 12533 + }, + { + "epoch": 0.4, + "learning_rate": 1.3614121057151723e-05, + "loss": 1.0308, + "step": 12534 + }, + { + "epoch": 0.4, + "learning_rate": 1.3613155247981898e-05, + "loss": 1.0664, + "step": 12535 + }, + { + "epoch": 0.4, + "learning_rate": 1.3612189400046913e-05, + "loss": 0.9814, + "step": 12536 + }, + { + "epoch": 0.4, + "learning_rate": 1.3611223513357134e-05, + "loss": 0.9478, + "step": 12537 + }, + { + "epoch": 0.4, + "learning_rate": 1.3610257587922926e-05, + "loss": 1.0386, + "step": 12538 + }, + { + "epoch": 0.4, + "learning_rate": 1.3609291623754648e-05, + "loss": 1.021, + "step": 12539 + }, + { + "epoch": 0.4, + "learning_rate": 1.3608325620862668e-05, + "loss": 0.9766, + "step": 12540 + }, + { + "epoch": 0.4, + "learning_rate": 1.3607359579257344e-05, + "loss": 1.0098, + "step": 12541 + }, + { + "epoch": 0.4, + "learning_rate": 1.3606393498949043e-05, + "loss": 1.0283, + "step": 12542 + }, + { + "epoch": 0.4, + "learning_rate": 1.3605427379948137e-05, + "loss": 1.0703, + "step": 12543 + }, + { + "epoch": 0.4, + "learning_rate": 1.3604461222264982e-05, + "loss": 1.1323, + "step": 12544 + }, + { + "epoch": 0.4, + "learning_rate": 1.3603495025909946e-05, + "loss": 1.0146, + "step": 12545 + }, + { + "epoch": 0.4, + "learning_rate": 1.3602528790893396e-05, + "loss": 1.103, + "step": 12546 + }, + { + "epoch": 0.4, + "learning_rate": 1.3601562517225704e-05, + "loss": 1.0908, + "step": 12547 + }, + { + "epoch": 0.4, + "learning_rate": 1.3600596204917228e-05, + "loss": 1.0737, + "step": 12548 + }, + { + "epoch": 0.4, + "learning_rate": 1.3599629853978342e-05, + "loss": 1.0122, + "step": 12549 + }, + { + "epoch": 0.4, + "learning_rate": 1.3598663464419408e-05, + "loss": 1.064, + "step": 12550 + }, + { + "epoch": 0.4, + "learning_rate": 1.3597697036250803e-05, + "loss": 0.9126, + "step": 12551 + }, + { + "epoch": 0.4, + "learning_rate": 1.3596730569482888e-05, + "loss": 0.937, + "step": 12552 + }, + { + "epoch": 0.4, + "learning_rate": 1.3595764064126033e-05, + "loss": 1.0435, + "step": 12553 + }, + { + "epoch": 0.4, + "learning_rate": 1.3594797520190613e-05, + "loss": 1.0483, + "step": 12554 + }, + { + "epoch": 0.4, + "learning_rate": 1.3593830937686991e-05, + "loss": 0.9644, + "step": 12555 + }, + { + "epoch": 0.4, + "learning_rate": 1.359286431662554e-05, + "loss": 0.9761, + "step": 12556 + }, + { + "epoch": 0.4, + "learning_rate": 1.3591897657016635e-05, + "loss": 1.0303, + "step": 12557 + }, + { + "epoch": 0.4, + "learning_rate": 1.359093095887064e-05, + "loss": 1.0391, + "step": 12558 + }, + { + "epoch": 0.4, + "learning_rate": 1.3589964222197932e-05, + "loss": 0.873, + "step": 12559 + }, + { + "epoch": 0.4, + "learning_rate": 1.358899744700888e-05, + "loss": 0.9897, + "step": 12560 + }, + { + "epoch": 0.4, + "learning_rate": 1.3588030633313858e-05, + "loss": 0.9312, + "step": 12561 + }, + { + "epoch": 0.4, + "learning_rate": 1.358706378112324e-05, + "loss": 0.8892, + "step": 12562 + }, + { + "epoch": 0.4, + "learning_rate": 1.3586096890447395e-05, + "loss": 1.083, + "step": 12563 + }, + { + "epoch": 0.4, + "learning_rate": 1.35851299612967e-05, + "loss": 1.105, + "step": 12564 + }, + { + "epoch": 0.4, + "learning_rate": 1.3584162993681527e-05, + "loss": 0.9209, + "step": 12565 + }, + { + "epoch": 0.4, + "learning_rate": 1.3583195987612255e-05, + "loss": 1.0986, + "step": 12566 + }, + { + "epoch": 0.4, + "learning_rate": 1.3582228943099253e-05, + "loss": 0.9448, + "step": 12567 + }, + { + "epoch": 0.4, + "learning_rate": 1.35812618601529e-05, + "loss": 0.9805, + "step": 12568 + }, + { + "epoch": 0.4, + "learning_rate": 1.3580294738783568e-05, + "loss": 0.9392, + "step": 12569 + }, + { + "epoch": 0.4, + "learning_rate": 1.357932757900164e-05, + "loss": 1.0635, + "step": 12570 + }, + { + "epoch": 0.4, + "learning_rate": 1.3578360380817488e-05, + "loss": 0.9897, + "step": 12571 + }, + { + "epoch": 0.4, + "learning_rate": 1.3577393144241487e-05, + "loss": 0.981, + "step": 12572 + }, + { + "epoch": 0.4, + "learning_rate": 1.3576425869284018e-05, + "loss": 1.126, + "step": 12573 + }, + { + "epoch": 0.4, + "learning_rate": 1.3575458555955461e-05, + "loss": 0.9966, + "step": 12574 + }, + { + "epoch": 0.4, + "learning_rate": 1.3574491204266185e-05, + "loss": 1.019, + "step": 12575 + }, + { + "epoch": 0.4, + "learning_rate": 1.3573523814226576e-05, + "loss": 0.9639, + "step": 12576 + }, + { + "epoch": 0.4, + "learning_rate": 1.3572556385847015e-05, + "loss": 1.0576, + "step": 12577 + }, + { + "epoch": 0.4, + "learning_rate": 1.3571588919137875e-05, + "loss": 0.9561, + "step": 12578 + }, + { + "epoch": 0.4, + "learning_rate": 1.357062141410954e-05, + "loss": 0.8687, + "step": 12579 + }, + { + "epoch": 0.4, + "learning_rate": 1.3569653870772389e-05, + "loss": 1.0186, + "step": 12580 + }, + { + "epoch": 0.4, + "learning_rate": 1.3568686289136802e-05, + "loss": 1.0654, + "step": 12581 + }, + { + "epoch": 0.4, + "learning_rate": 1.3567718669213159e-05, + "loss": 1.0576, + "step": 12582 + }, + { + "epoch": 0.4, + "learning_rate": 1.3566751011011845e-05, + "loss": 1.063, + "step": 12583 + }, + { + "epoch": 0.4, + "learning_rate": 1.3565783314543239e-05, + "loss": 1.0059, + "step": 12584 + }, + { + "epoch": 0.4, + "learning_rate": 1.3564815579817727e-05, + "loss": 1.0405, + "step": 12585 + }, + { + "epoch": 0.4, + "learning_rate": 1.3563847806845687e-05, + "loss": 1.0918, + "step": 12586 + }, + { + "epoch": 0.4, + "learning_rate": 1.3562879995637506e-05, + "loss": 1.1279, + "step": 12587 + }, + { + "epoch": 0.4, + "learning_rate": 1.3561912146203564e-05, + "loss": 0.9941, + "step": 12588 + }, + { + "epoch": 0.4, + "learning_rate": 1.3560944258554249e-05, + "loss": 1.1157, + "step": 12589 + }, + { + "epoch": 0.4, + "learning_rate": 1.355997633269994e-05, + "loss": 0.9468, + "step": 12590 + }, + { + "epoch": 0.4, + "learning_rate": 1.3559008368651028e-05, + "loss": 0.9575, + "step": 12591 + }, + { + "epoch": 0.4, + "learning_rate": 1.3558040366417894e-05, + "loss": 0.9829, + "step": 12592 + }, + { + "epoch": 0.4, + "learning_rate": 1.3557072326010924e-05, + "loss": 0.9102, + "step": 12593 + }, + { + "epoch": 0.4, + "learning_rate": 1.3556104247440504e-05, + "loss": 0.9443, + "step": 12594 + }, + { + "epoch": 0.4, + "learning_rate": 1.3555136130717021e-05, + "loss": 1.0972, + "step": 12595 + }, + { + "epoch": 0.4, + "learning_rate": 1.3554167975850868e-05, + "loss": 1.0898, + "step": 12596 + }, + { + "epoch": 0.4, + "learning_rate": 1.355319978285242e-05, + "loss": 0.9907, + "step": 12597 + }, + { + "epoch": 0.4, + "learning_rate": 1.3552231551732071e-05, + "loss": 1.0977, + "step": 12598 + }, + { + "epoch": 0.4, + "learning_rate": 1.3551263282500213e-05, + "loss": 1.1206, + "step": 12599 + }, + { + "epoch": 0.4, + "learning_rate": 1.3550294975167226e-05, + "loss": 0.9863, + "step": 12600 + }, + { + "epoch": 0.4, + "learning_rate": 1.3549326629743506e-05, + "loss": 1.1055, + "step": 12601 + }, + { + "epoch": 0.4, + "learning_rate": 1.3548358246239438e-05, + "loss": 1.0, + "step": 12602 + }, + { + "epoch": 0.4, + "learning_rate": 1.3547389824665413e-05, + "loss": 0.9399, + "step": 12603 + }, + { + "epoch": 0.4, + "learning_rate": 1.3546421365031824e-05, + "loss": 1.0259, + "step": 12604 + }, + { + "epoch": 0.4, + "learning_rate": 1.3545452867349054e-05, + "loss": 1.165, + "step": 12605 + }, + { + "epoch": 0.4, + "learning_rate": 1.35444843316275e-05, + "loss": 0.9243, + "step": 12606 + }, + { + "epoch": 0.4, + "learning_rate": 1.3543515757877552e-05, + "loss": 0.9561, + "step": 12607 + }, + { + "epoch": 0.4, + "learning_rate": 1.3542547146109605e-05, + "loss": 0.9673, + "step": 12608 + }, + { + "epoch": 0.4, + "learning_rate": 1.3541578496334045e-05, + "loss": 1.0913, + "step": 12609 + }, + { + "epoch": 0.4, + "learning_rate": 1.3540609808561266e-05, + "loss": 0.9517, + "step": 12610 + }, + { + "epoch": 0.4, + "learning_rate": 1.3539641082801667e-05, + "loss": 0.906, + "step": 12611 + }, + { + "epoch": 0.4, + "learning_rate": 1.3538672319065634e-05, + "loss": 0.9658, + "step": 12612 + }, + { + "epoch": 0.4, + "learning_rate": 1.3537703517363561e-05, + "loss": 0.9009, + "step": 12613 + }, + { + "epoch": 0.4, + "learning_rate": 1.3536734677705847e-05, + "loss": 0.917, + "step": 12614 + }, + { + "epoch": 0.4, + "learning_rate": 1.3535765800102885e-05, + "loss": 1.0454, + "step": 12615 + }, + { + "epoch": 0.4, + "learning_rate": 1.3534796884565067e-05, + "loss": 0.9873, + "step": 12616 + }, + { + "epoch": 0.4, + "learning_rate": 1.3533827931102791e-05, + "loss": 1.0005, + "step": 12617 + }, + { + "epoch": 0.4, + "learning_rate": 1.3532858939726453e-05, + "loss": 1.0791, + "step": 12618 + }, + { + "epoch": 0.4, + "learning_rate": 1.353188991044645e-05, + "loss": 0.8877, + "step": 12619 + }, + { + "epoch": 0.4, + "learning_rate": 1.3530920843273174e-05, + "loss": 1.1401, + "step": 12620 + }, + { + "epoch": 0.4, + "learning_rate": 1.3529951738217029e-05, + "loss": 0.9609, + "step": 12621 + }, + { + "epoch": 0.4, + "learning_rate": 1.3528982595288405e-05, + "loss": 0.9277, + "step": 12622 + }, + { + "epoch": 0.4, + "learning_rate": 1.3528013414497708e-05, + "loss": 0.8643, + "step": 12623 + }, + { + "epoch": 0.4, + "learning_rate": 1.3527044195855325e-05, + "loss": 1.1025, + "step": 12624 + }, + { + "epoch": 0.4, + "learning_rate": 1.3526074939371664e-05, + "loss": 1.0308, + "step": 12625 + }, + { + "epoch": 0.4, + "learning_rate": 1.3525105645057122e-05, + "loss": 0.8721, + "step": 12626 + }, + { + "epoch": 0.4, + "learning_rate": 1.35241363129221e-05, + "loss": 0.9673, + "step": 12627 + }, + { + "epoch": 0.4, + "learning_rate": 1.3523166942976996e-05, + "loss": 1.0164, + "step": 12628 + }, + { + "epoch": 0.4, + "learning_rate": 1.3522197535232204e-05, + "loss": 0.8501, + "step": 12629 + }, + { + "epoch": 0.4, + "learning_rate": 1.3521228089698138e-05, + "loss": 1.1338, + "step": 12630 + }, + { + "epoch": 0.4, + "learning_rate": 1.3520258606385186e-05, + "loss": 1.1235, + "step": 12631 + }, + { + "epoch": 0.4, + "learning_rate": 1.3519289085303757e-05, + "loss": 1.1084, + "step": 12632 + }, + { + "epoch": 0.4, + "learning_rate": 1.3518319526464253e-05, + "loss": 0.9478, + "step": 12633 + }, + { + "epoch": 0.4, + "learning_rate": 1.3517349929877073e-05, + "loss": 0.9531, + "step": 12634 + }, + { + "epoch": 0.4, + "learning_rate": 1.3516380295552618e-05, + "loss": 0.9976, + "step": 12635 + }, + { + "epoch": 0.4, + "learning_rate": 1.3515410623501296e-05, + "loss": 1.0801, + "step": 12636 + }, + { + "epoch": 0.4, + "learning_rate": 1.351444091373351e-05, + "loss": 1.021, + "step": 12637 + }, + { + "epoch": 0.4, + "learning_rate": 1.3513471166259665e-05, + "loss": 1.0405, + "step": 12638 + }, + { + "epoch": 0.4, + "learning_rate": 1.3512501381090158e-05, + "loss": 1.0903, + "step": 12639 + }, + { + "epoch": 0.4, + "learning_rate": 1.35115315582354e-05, + "loss": 1.1074, + "step": 12640 + }, + { + "epoch": 0.4, + "learning_rate": 1.3510561697705798e-05, + "loss": 1.0278, + "step": 12641 + }, + { + "epoch": 0.4, + "learning_rate": 1.350959179951175e-05, + "loss": 1.123, + "step": 12642 + }, + { + "epoch": 0.4, + "learning_rate": 1.350862186366367e-05, + "loss": 1.0957, + "step": 12643 + }, + { + "epoch": 0.4, + "learning_rate": 1.3507651890171958e-05, + "loss": 1.0674, + "step": 12644 + }, + { + "epoch": 0.4, + "learning_rate": 1.3506681879047024e-05, + "loss": 1.0474, + "step": 12645 + }, + { + "epoch": 0.4, + "learning_rate": 1.3505711830299275e-05, + "loss": 1.0132, + "step": 12646 + }, + { + "epoch": 0.4, + "learning_rate": 1.3504741743939116e-05, + "loss": 0.9468, + "step": 12647 + }, + { + "epoch": 0.4, + "learning_rate": 1.3503771619976958e-05, + "loss": 0.9517, + "step": 12648 + }, + { + "epoch": 0.4, + "learning_rate": 1.3502801458423211e-05, + "loss": 0.563, + "step": 12649 + }, + { + "epoch": 0.4, + "learning_rate": 1.3501831259288278e-05, + "loss": 0.9663, + "step": 12650 + }, + { + "epoch": 0.4, + "learning_rate": 1.3500861022582569e-05, + "loss": 1.0674, + "step": 12651 + }, + { + "epoch": 0.4, + "learning_rate": 1.34998907483165e-05, + "loss": 0.9761, + "step": 12652 + }, + { + "epoch": 0.4, + "learning_rate": 1.3498920436500476e-05, + "loss": 1.0732, + "step": 12653 + }, + { + "epoch": 0.4, + "learning_rate": 1.3497950087144906e-05, + "loss": 1.0, + "step": 12654 + }, + { + "epoch": 0.4, + "learning_rate": 1.3496979700260204e-05, + "loss": 1.0205, + "step": 12655 + }, + { + "epoch": 0.4, + "learning_rate": 1.3496009275856781e-05, + "loss": 0.98, + "step": 12656 + }, + { + "epoch": 0.4, + "learning_rate": 1.349503881394505e-05, + "loss": 1.043, + "step": 12657 + }, + { + "epoch": 0.4, + "learning_rate": 1.3494068314535416e-05, + "loss": 1.083, + "step": 12658 + }, + { + "epoch": 0.4, + "learning_rate": 1.3493097777638297e-05, + "loss": 0.9722, + "step": 12659 + }, + { + "epoch": 0.4, + "learning_rate": 1.3492127203264107e-05, + "loss": 0.9673, + "step": 12660 + }, + { + "epoch": 0.4, + "learning_rate": 1.3491156591423256e-05, + "loss": 1.0859, + "step": 12661 + }, + { + "epoch": 0.4, + "learning_rate": 1.3490185942126157e-05, + "loss": 1.1626, + "step": 12662 + }, + { + "epoch": 0.4, + "learning_rate": 1.3489215255383226e-05, + "loss": 1.0435, + "step": 12663 + }, + { + "epoch": 0.41, + "learning_rate": 1.348824453120488e-05, + "loss": 0.8979, + "step": 12664 + }, + { + "epoch": 0.41, + "learning_rate": 1.3487273769601527e-05, + "loss": 0.9429, + "step": 12665 + }, + { + "epoch": 0.41, + "learning_rate": 1.3486302970583587e-05, + "loss": 1.0083, + "step": 12666 + }, + { + "epoch": 0.41, + "learning_rate": 1.3485332134161473e-05, + "loss": 1.0669, + "step": 12667 + }, + { + "epoch": 0.41, + "learning_rate": 1.3484361260345606e-05, + "loss": 1.0166, + "step": 12668 + }, + { + "epoch": 0.41, + "learning_rate": 1.34833903491464e-05, + "loss": 0.8906, + "step": 12669 + }, + { + "epoch": 0.41, + "learning_rate": 1.3482419400574265e-05, + "loss": 1.0933, + "step": 12670 + }, + { + "epoch": 0.41, + "learning_rate": 1.3481448414639628e-05, + "loss": 1.0635, + "step": 12671 + }, + { + "epoch": 0.41, + "learning_rate": 1.3480477391352902e-05, + "loss": 1.0273, + "step": 12672 + }, + { + "epoch": 0.41, + "learning_rate": 1.3479506330724506e-05, + "loss": 1.0303, + "step": 12673 + }, + { + "epoch": 0.41, + "learning_rate": 1.3478535232764858e-05, + "loss": 1.0293, + "step": 12674 + }, + { + "epoch": 0.41, + "learning_rate": 1.3477564097484374e-05, + "loss": 0.9185, + "step": 12675 + }, + { + "epoch": 0.41, + "learning_rate": 1.3476592924893482e-05, + "loss": 1.0044, + "step": 12676 + }, + { + "epoch": 0.41, + "learning_rate": 1.3475621715002589e-05, + "loss": 1.0396, + "step": 12677 + }, + { + "epoch": 0.41, + "learning_rate": 1.3474650467822126e-05, + "loss": 1.0044, + "step": 12678 + }, + { + "epoch": 0.41, + "learning_rate": 1.347367918336251e-05, + "loss": 0.9878, + "step": 12679 + }, + { + "epoch": 0.41, + "learning_rate": 1.3472707861634155e-05, + "loss": 1.0381, + "step": 12680 + }, + { + "epoch": 0.41, + "learning_rate": 1.3471736502647493e-05, + "loss": 1.0811, + "step": 12681 + }, + { + "epoch": 0.41, + "learning_rate": 1.347076510641294e-05, + "loss": 1.0566, + "step": 12682 + }, + { + "epoch": 0.41, + "learning_rate": 1.3469793672940919e-05, + "loss": 1.019, + "step": 12683 + }, + { + "epoch": 0.41, + "learning_rate": 1.346882220224185e-05, + "loss": 1.0771, + "step": 12684 + }, + { + "epoch": 0.41, + "learning_rate": 1.346785069432616e-05, + "loss": 0.9785, + "step": 12685 + }, + { + "epoch": 0.41, + "learning_rate": 1.3466879149204268e-05, + "loss": 0.9536, + "step": 12686 + }, + { + "epoch": 0.41, + "learning_rate": 1.3465907566886599e-05, + "loss": 0.8745, + "step": 12687 + }, + { + "epoch": 0.41, + "learning_rate": 1.346493594738358e-05, + "loss": 0.9648, + "step": 12688 + }, + { + "epoch": 0.41, + "learning_rate": 1.3463964290705631e-05, + "loss": 1.0239, + "step": 12689 + }, + { + "epoch": 0.41, + "learning_rate": 1.346299259686318e-05, + "loss": 0.9175, + "step": 12690 + }, + { + "epoch": 0.41, + "learning_rate": 1.3462020865866654e-05, + "loss": 1.1094, + "step": 12691 + }, + { + "epoch": 0.41, + "learning_rate": 1.3461049097726473e-05, + "loss": 0.9819, + "step": 12692 + }, + { + "epoch": 0.41, + "learning_rate": 1.3460077292453064e-05, + "loss": 0.9067, + "step": 12693 + }, + { + "epoch": 0.41, + "learning_rate": 1.345910545005686e-05, + "loss": 1.0903, + "step": 12694 + }, + { + "epoch": 0.41, + "learning_rate": 1.3458133570548277e-05, + "loss": 1.104, + "step": 12695 + }, + { + "epoch": 0.41, + "learning_rate": 1.3457161653937752e-05, + "loss": 1.04, + "step": 12696 + }, + { + "epoch": 0.41, + "learning_rate": 1.3456189700235705e-05, + "loss": 0.8892, + "step": 12697 + }, + { + "epoch": 0.41, + "learning_rate": 1.3455217709452573e-05, + "loss": 1.0308, + "step": 12698 + }, + { + "epoch": 0.41, + "learning_rate": 1.3454245681598775e-05, + "loss": 1.0728, + "step": 12699 + }, + { + "epoch": 0.41, + "learning_rate": 1.3453273616684743e-05, + "loss": 1.1265, + "step": 12700 + }, + { + "epoch": 0.41, + "learning_rate": 1.345230151472091e-05, + "loss": 0.9263, + "step": 12701 + }, + { + "epoch": 0.41, + "learning_rate": 1.3451329375717702e-05, + "loss": 0.9883, + "step": 12702 + }, + { + "epoch": 0.41, + "learning_rate": 1.3450357199685549e-05, + "loss": 0.9849, + "step": 12703 + }, + { + "epoch": 0.41, + "learning_rate": 1.344938498663488e-05, + "loss": 1.061, + "step": 12704 + }, + { + "epoch": 0.41, + "learning_rate": 1.3448412736576128e-05, + "loss": 1.0723, + "step": 12705 + }, + { + "epoch": 0.41, + "learning_rate": 1.3447440449519728e-05, + "loss": 0.9629, + "step": 12706 + }, + { + "epoch": 0.41, + "learning_rate": 1.3446468125476105e-05, + "loss": 1.0112, + "step": 12707 + }, + { + "epoch": 0.41, + "learning_rate": 1.3445495764455692e-05, + "loss": 1.0205, + "step": 12708 + }, + { + "epoch": 0.41, + "learning_rate": 1.3444523366468923e-05, + "loss": 1.0635, + "step": 12709 + }, + { + "epoch": 0.41, + "learning_rate": 1.3443550931526233e-05, + "loss": 0.9556, + "step": 12710 + }, + { + "epoch": 0.41, + "learning_rate": 1.3442578459638048e-05, + "loss": 1.1836, + "step": 12711 + }, + { + "epoch": 0.41, + "learning_rate": 1.3441605950814811e-05, + "loss": 0.9717, + "step": 12712 + }, + { + "epoch": 0.41, + "learning_rate": 1.3440633405066949e-05, + "loss": 0.958, + "step": 12713 + }, + { + "epoch": 0.41, + "learning_rate": 1.3439660822404898e-05, + "loss": 1.0771, + "step": 12714 + }, + { + "epoch": 0.41, + "learning_rate": 1.3438688202839095e-05, + "loss": 1.0254, + "step": 12715 + }, + { + "epoch": 0.41, + "learning_rate": 1.3437715546379973e-05, + "loss": 1.0732, + "step": 12716 + }, + { + "epoch": 0.41, + "learning_rate": 1.3436742853037967e-05, + "loss": 0.9663, + "step": 12717 + }, + { + "epoch": 0.41, + "learning_rate": 1.3435770122823513e-05, + "loss": 1.1152, + "step": 12718 + }, + { + "epoch": 0.41, + "learning_rate": 1.3434797355747049e-05, + "loss": 1.0947, + "step": 12719 + }, + { + "epoch": 0.41, + "learning_rate": 1.343382455181901e-05, + "loss": 0.9087, + "step": 12720 + }, + { + "epoch": 0.41, + "learning_rate": 1.3432851711049836e-05, + "loss": 1.1162, + "step": 12721 + }, + { + "epoch": 0.41, + "learning_rate": 1.343187883344996e-05, + "loss": 1.0454, + "step": 12722 + }, + { + "epoch": 0.41, + "learning_rate": 1.3430905919029824e-05, + "loss": 0.9849, + "step": 12723 + }, + { + "epoch": 0.41, + "learning_rate": 1.3429932967799867e-05, + "loss": 1.1206, + "step": 12724 + }, + { + "epoch": 0.41, + "learning_rate": 1.3428959979770522e-05, + "loss": 1.0625, + "step": 12725 + }, + { + "epoch": 0.41, + "learning_rate": 1.342798695495223e-05, + "loss": 1.0679, + "step": 12726 + }, + { + "epoch": 0.41, + "learning_rate": 1.3427013893355433e-05, + "loss": 1.0034, + "step": 12727 + }, + { + "epoch": 0.41, + "learning_rate": 1.3426040794990571e-05, + "loss": 0.9551, + "step": 12728 + }, + { + "epoch": 0.41, + "learning_rate": 1.3425067659868084e-05, + "loss": 0.9644, + "step": 12729 + }, + { + "epoch": 0.41, + "learning_rate": 1.3424094487998411e-05, + "loss": 0.9492, + "step": 12730 + }, + { + "epoch": 0.41, + "learning_rate": 1.342312127939199e-05, + "loss": 0.9336, + "step": 12731 + }, + { + "epoch": 0.41, + "learning_rate": 1.3422148034059272e-05, + "loss": 1.0952, + "step": 12732 + }, + { + "epoch": 0.41, + "learning_rate": 1.342117475201069e-05, + "loss": 1.0449, + "step": 12733 + }, + { + "epoch": 0.41, + "learning_rate": 1.342020143325669e-05, + "loss": 1.1245, + "step": 12734 + }, + { + "epoch": 0.41, + "learning_rate": 1.3419228077807711e-05, + "loss": 1.02, + "step": 12735 + }, + { + "epoch": 0.41, + "learning_rate": 1.3418254685674203e-05, + "loss": 0.8213, + "step": 12736 + }, + { + "epoch": 0.41, + "learning_rate": 1.3417281256866603e-05, + "loss": 1.0186, + "step": 12737 + }, + { + "epoch": 0.41, + "learning_rate": 1.3416307791395357e-05, + "loss": 1.0322, + "step": 12738 + }, + { + "epoch": 0.41, + "learning_rate": 1.3415334289270911e-05, + "loss": 0.9692, + "step": 12739 + }, + { + "epoch": 0.41, + "learning_rate": 1.3414360750503708e-05, + "loss": 1.0537, + "step": 12740 + }, + { + "epoch": 0.41, + "learning_rate": 1.3413387175104193e-05, + "loss": 1.0259, + "step": 12741 + }, + { + "epoch": 0.41, + "learning_rate": 1.3412413563082808e-05, + "loss": 0.9458, + "step": 12742 + }, + { + "epoch": 0.41, + "learning_rate": 1.3411439914450006e-05, + "loss": 0.9307, + "step": 12743 + }, + { + "epoch": 0.41, + "learning_rate": 1.3410466229216228e-05, + "loss": 1.0947, + "step": 12744 + }, + { + "epoch": 0.41, + "learning_rate": 1.3409492507391923e-05, + "loss": 1.1426, + "step": 12745 + }, + { + "epoch": 0.41, + "learning_rate": 1.3408518748987533e-05, + "loss": 0.9478, + "step": 12746 + }, + { + "epoch": 0.41, + "learning_rate": 1.3407544954013515e-05, + "loss": 1.0273, + "step": 12747 + }, + { + "epoch": 0.41, + "learning_rate": 1.3406571122480307e-05, + "loss": 1.083, + "step": 12748 + }, + { + "epoch": 0.41, + "learning_rate": 1.3405597254398361e-05, + "loss": 0.9805, + "step": 12749 + }, + { + "epoch": 0.41, + "learning_rate": 1.3404623349778127e-05, + "loss": 1.0107, + "step": 12750 + }, + { + "epoch": 0.41, + "learning_rate": 1.3403649408630053e-05, + "loss": 0.9819, + "step": 12751 + }, + { + "epoch": 0.41, + "learning_rate": 1.3402675430964586e-05, + "loss": 0.5566, + "step": 12752 + }, + { + "epoch": 0.41, + "learning_rate": 1.3401701416792179e-05, + "loss": 0.9692, + "step": 12753 + }, + { + "epoch": 0.41, + "learning_rate": 1.3400727366123279e-05, + "loss": 1.0044, + "step": 12754 + }, + { + "epoch": 0.41, + "learning_rate": 1.339975327896834e-05, + "loss": 0.9868, + "step": 12755 + }, + { + "epoch": 0.41, + "learning_rate": 1.3398779155337808e-05, + "loss": 0.9551, + "step": 12756 + }, + { + "epoch": 0.41, + "learning_rate": 1.339780499524214e-05, + "loss": 1.1152, + "step": 12757 + }, + { + "epoch": 0.41, + "learning_rate": 1.3396830798691785e-05, + "loss": 1.062, + "step": 12758 + }, + { + "epoch": 0.41, + "learning_rate": 1.3395856565697194e-05, + "loss": 0.9004, + "step": 12759 + }, + { + "epoch": 0.41, + "learning_rate": 1.3394882296268822e-05, + "loss": 0.5354, + "step": 12760 + }, + { + "epoch": 0.41, + "learning_rate": 1.3393907990417117e-05, + "loss": 1.0718, + "step": 12761 + }, + { + "epoch": 0.41, + "learning_rate": 1.3392933648152537e-05, + "loss": 1.0557, + "step": 12762 + }, + { + "epoch": 0.41, + "learning_rate": 1.3391959269485536e-05, + "loss": 1.144, + "step": 12763 + }, + { + "epoch": 0.41, + "learning_rate": 1.3390984854426563e-05, + "loss": 1.0303, + "step": 12764 + }, + { + "epoch": 0.41, + "learning_rate": 1.3390010402986075e-05, + "loss": 1.0513, + "step": 12765 + }, + { + "epoch": 0.41, + "learning_rate": 1.3389035915174531e-05, + "loss": 1.0698, + "step": 12766 + }, + { + "epoch": 0.41, + "learning_rate": 1.338806139100238e-05, + "loss": 1.0703, + "step": 12767 + }, + { + "epoch": 0.41, + "learning_rate": 1.3387086830480081e-05, + "loss": 1.0215, + "step": 12768 + }, + { + "epoch": 0.41, + "learning_rate": 1.3386112233618089e-05, + "loss": 1.0288, + "step": 12769 + }, + { + "epoch": 0.41, + "learning_rate": 1.3385137600426861e-05, + "loss": 0.9268, + "step": 12770 + }, + { + "epoch": 0.41, + "learning_rate": 1.3384162930916854e-05, + "loss": 1.0474, + "step": 12771 + }, + { + "epoch": 0.41, + "learning_rate": 1.3383188225098521e-05, + "loss": 1.1602, + "step": 12772 + }, + { + "epoch": 0.41, + "learning_rate": 1.3382213482982323e-05, + "loss": 1.0054, + "step": 12773 + }, + { + "epoch": 0.41, + "learning_rate": 1.3381238704578718e-05, + "loss": 0.9424, + "step": 12774 + }, + { + "epoch": 0.41, + "learning_rate": 1.3380263889898164e-05, + "loss": 1.0239, + "step": 12775 + }, + { + "epoch": 0.41, + "learning_rate": 1.3379289038951119e-05, + "loss": 1.0376, + "step": 12776 + }, + { + "epoch": 0.41, + "learning_rate": 1.3378314151748045e-05, + "loss": 1.0996, + "step": 12777 + }, + { + "epoch": 0.41, + "learning_rate": 1.3377339228299398e-05, + "loss": 1.0903, + "step": 12778 + }, + { + "epoch": 0.41, + "learning_rate": 1.3376364268615636e-05, + "loss": 0.5549, + "step": 12779 + }, + { + "epoch": 0.41, + "learning_rate": 1.3375389272707225e-05, + "loss": 1.0571, + "step": 12780 + }, + { + "epoch": 0.41, + "learning_rate": 1.3374414240584623e-05, + "loss": 0.874, + "step": 12781 + }, + { + "epoch": 0.41, + "learning_rate": 1.3373439172258288e-05, + "loss": 1.0469, + "step": 12782 + }, + { + "epoch": 0.41, + "learning_rate": 1.3372464067738686e-05, + "loss": 0.5403, + "step": 12783 + }, + { + "epoch": 0.41, + "learning_rate": 1.3371488927036278e-05, + "loss": 1.0718, + "step": 12784 + }, + { + "epoch": 0.41, + "learning_rate": 1.3370513750161527e-05, + "loss": 1.0122, + "step": 12785 + }, + { + "epoch": 0.41, + "learning_rate": 1.3369538537124888e-05, + "loss": 1.0796, + "step": 12786 + }, + { + "epoch": 0.41, + "learning_rate": 1.3368563287936833e-05, + "loss": 0.9897, + "step": 12787 + }, + { + "epoch": 0.41, + "learning_rate": 1.3367588002607822e-05, + "loss": 0.958, + "step": 12788 + }, + { + "epoch": 0.41, + "learning_rate": 1.336661268114832e-05, + "loss": 0.9263, + "step": 12789 + }, + { + "epoch": 0.41, + "learning_rate": 1.3365637323568787e-05, + "loss": 0.7944, + "step": 12790 + }, + { + "epoch": 0.41, + "learning_rate": 1.3364661929879694e-05, + "loss": 0.5879, + "step": 12791 + }, + { + "epoch": 0.41, + "learning_rate": 1.33636865000915e-05, + "loss": 1.1118, + "step": 12792 + }, + { + "epoch": 0.41, + "learning_rate": 1.3362711034214675e-05, + "loss": 1.0205, + "step": 12793 + }, + { + "epoch": 0.41, + "learning_rate": 1.3361735532259678e-05, + "loss": 0.9414, + "step": 12794 + }, + { + "epoch": 0.41, + "learning_rate": 1.3360759994236982e-05, + "loss": 1.0835, + "step": 12795 + }, + { + "epoch": 0.41, + "learning_rate": 1.3359784420157051e-05, + "loss": 1.0146, + "step": 12796 + }, + { + "epoch": 0.41, + "learning_rate": 1.335880881003035e-05, + "loss": 1.1001, + "step": 12797 + }, + { + "epoch": 0.41, + "learning_rate": 1.335783316386735e-05, + "loss": 1.0923, + "step": 12798 + }, + { + "epoch": 0.41, + "learning_rate": 1.3356857481678514e-05, + "loss": 1.0127, + "step": 12799 + }, + { + "epoch": 0.41, + "learning_rate": 1.3355881763474314e-05, + "loss": 1.0845, + "step": 12800 + }, + { + "epoch": 0.41, + "learning_rate": 1.3354906009265215e-05, + "loss": 0.9473, + "step": 12801 + }, + { + "epoch": 0.41, + "learning_rate": 1.335393021906169e-05, + "loss": 1.0425, + "step": 12802 + }, + { + "epoch": 0.41, + "learning_rate": 1.3352954392874205e-05, + "loss": 1.0073, + "step": 12803 + }, + { + "epoch": 0.41, + "learning_rate": 1.335197853071323e-05, + "loss": 1.0044, + "step": 12804 + }, + { + "epoch": 0.41, + "learning_rate": 1.3351002632589233e-05, + "loss": 1.0215, + "step": 12805 + }, + { + "epoch": 0.41, + "learning_rate": 1.335002669851269e-05, + "loss": 1.0381, + "step": 12806 + }, + { + "epoch": 0.41, + "learning_rate": 1.3349050728494066e-05, + "loss": 1.061, + "step": 12807 + }, + { + "epoch": 0.41, + "learning_rate": 1.3348074722543838e-05, + "loss": 1.0205, + "step": 12808 + }, + { + "epoch": 0.41, + "learning_rate": 1.3347098680672467e-05, + "loss": 1.0054, + "step": 12809 + }, + { + "epoch": 0.41, + "learning_rate": 1.334612260289044e-05, + "loss": 0.9282, + "step": 12810 + }, + { + "epoch": 0.41, + "learning_rate": 1.3345146489208213e-05, + "loss": 1.0039, + "step": 12811 + }, + { + "epoch": 0.41, + "learning_rate": 1.334417033963627e-05, + "loss": 1.0557, + "step": 12812 + }, + { + "epoch": 0.41, + "learning_rate": 1.334319415418508e-05, + "loss": 1.126, + "step": 12813 + }, + { + "epoch": 0.41, + "learning_rate": 1.3342217932865116e-05, + "loss": 1.0581, + "step": 12814 + }, + { + "epoch": 0.41, + "learning_rate": 1.3341241675686853e-05, + "loss": 1.0532, + "step": 12815 + }, + { + "epoch": 0.41, + "learning_rate": 1.3340265382660767e-05, + "loss": 1.0977, + "step": 12816 + }, + { + "epoch": 0.41, + "learning_rate": 1.3339289053797327e-05, + "loss": 1.0234, + "step": 12817 + }, + { + "epoch": 0.41, + "learning_rate": 1.3338312689107012e-05, + "loss": 0.9829, + "step": 12818 + }, + { + "epoch": 0.41, + "learning_rate": 1.3337336288600297e-05, + "loss": 0.9316, + "step": 12819 + }, + { + "epoch": 0.41, + "learning_rate": 1.3336359852287656e-05, + "loss": 1.1162, + "step": 12820 + }, + { + "epoch": 0.41, + "learning_rate": 1.3335383380179567e-05, + "loss": 0.8918, + "step": 12821 + }, + { + "epoch": 0.41, + "learning_rate": 1.3334406872286505e-05, + "loss": 1.1855, + "step": 12822 + }, + { + "epoch": 0.41, + "learning_rate": 1.333343032861895e-05, + "loss": 0.8843, + "step": 12823 + }, + { + "epoch": 0.41, + "learning_rate": 1.3332453749187375e-05, + "loss": 0.917, + "step": 12824 + }, + { + "epoch": 0.41, + "learning_rate": 1.333147713400226e-05, + "loss": 0.9971, + "step": 12825 + }, + { + "epoch": 0.41, + "learning_rate": 1.3330500483074083e-05, + "loss": 1.0869, + "step": 12826 + }, + { + "epoch": 0.41, + "learning_rate": 1.3329523796413321e-05, + "loss": 0.9482, + "step": 12827 + }, + { + "epoch": 0.41, + "learning_rate": 1.3328547074030454e-05, + "loss": 1.0952, + "step": 12828 + }, + { + "epoch": 0.41, + "learning_rate": 1.332757031593596e-05, + "loss": 0.8848, + "step": 12829 + }, + { + "epoch": 0.41, + "learning_rate": 1.332659352214032e-05, + "loss": 0.9722, + "step": 12830 + }, + { + "epoch": 0.41, + "learning_rate": 1.3325616692654014e-05, + "loss": 0.9326, + "step": 12831 + }, + { + "epoch": 0.41, + "learning_rate": 1.332463982748752e-05, + "loss": 1.0781, + "step": 12832 + }, + { + "epoch": 0.41, + "learning_rate": 1.3323662926651319e-05, + "loss": 1.0151, + "step": 12833 + }, + { + "epoch": 0.41, + "learning_rate": 1.3322685990155897e-05, + "loss": 1.1392, + "step": 12834 + }, + { + "epoch": 0.41, + "learning_rate": 1.3321709018011728e-05, + "loss": 0.9678, + "step": 12835 + }, + { + "epoch": 0.41, + "learning_rate": 1.3320732010229299e-05, + "loss": 0.9302, + "step": 12836 + }, + { + "epoch": 0.41, + "learning_rate": 1.331975496681909e-05, + "loss": 1.0327, + "step": 12837 + }, + { + "epoch": 0.41, + "learning_rate": 1.3318777887791587e-05, + "loss": 0.9043, + "step": 12838 + }, + { + "epoch": 0.41, + "learning_rate": 1.3317800773157268e-05, + "loss": 0.9497, + "step": 12839 + }, + { + "epoch": 0.41, + "learning_rate": 1.331682362292662e-05, + "loss": 1.0283, + "step": 12840 + }, + { + "epoch": 0.41, + "learning_rate": 1.3315846437110122e-05, + "loss": 1.0371, + "step": 12841 + }, + { + "epoch": 0.41, + "learning_rate": 1.3314869215718267e-05, + "loss": 1.0083, + "step": 12842 + }, + { + "epoch": 0.41, + "learning_rate": 1.331389195876153e-05, + "loss": 0.9858, + "step": 12843 + }, + { + "epoch": 0.41, + "learning_rate": 1.3312914666250401e-05, + "loss": 1.041, + "step": 12844 + }, + { + "epoch": 0.41, + "learning_rate": 1.3311937338195365e-05, + "loss": 0.9707, + "step": 12845 + }, + { + "epoch": 0.41, + "learning_rate": 1.3310959974606905e-05, + "loss": 0.9312, + "step": 12846 + }, + { + "epoch": 0.41, + "learning_rate": 1.330998257549551e-05, + "loss": 1.1758, + "step": 12847 + }, + { + "epoch": 0.41, + "learning_rate": 1.3309005140871665e-05, + "loss": 1.0024, + "step": 12848 + }, + { + "epoch": 0.41, + "learning_rate": 1.3308027670745858e-05, + "loss": 1.0278, + "step": 12849 + }, + { + "epoch": 0.41, + "learning_rate": 1.3307050165128575e-05, + "loss": 0.8462, + "step": 12850 + }, + { + "epoch": 0.41, + "learning_rate": 1.3306072624030302e-05, + "loss": 1.0688, + "step": 12851 + }, + { + "epoch": 0.41, + "learning_rate": 1.3305095047461532e-05, + "loss": 1.0654, + "step": 12852 + }, + { + "epoch": 0.41, + "learning_rate": 1.3304117435432747e-05, + "loss": 1.0498, + "step": 12853 + }, + { + "epoch": 0.41, + "learning_rate": 1.3303139787954437e-05, + "loss": 1.1133, + "step": 12854 + }, + { + "epoch": 0.41, + "learning_rate": 1.33021621050371e-05, + "loss": 0.9683, + "step": 12855 + }, + { + "epoch": 0.41, + "learning_rate": 1.3301184386691213e-05, + "loss": 1.0293, + "step": 12856 + }, + { + "epoch": 0.41, + "learning_rate": 1.330020663292727e-05, + "loss": 0.9951, + "step": 12857 + }, + { + "epoch": 0.41, + "learning_rate": 1.3299228843755768e-05, + "loss": 1.1724, + "step": 12858 + }, + { + "epoch": 0.41, + "learning_rate": 1.3298251019187185e-05, + "loss": 1.104, + "step": 12859 + }, + { + "epoch": 0.41, + "learning_rate": 1.3297273159232023e-05, + "loss": 0.9814, + "step": 12860 + }, + { + "epoch": 0.41, + "learning_rate": 1.329629526390077e-05, + "loss": 1.1094, + "step": 12861 + }, + { + "epoch": 0.41, + "learning_rate": 1.3295317333203914e-05, + "loss": 1.0195, + "step": 12862 + }, + { + "epoch": 0.41, + "learning_rate": 1.3294339367151951e-05, + "loss": 1.0425, + "step": 12863 + }, + { + "epoch": 0.41, + "learning_rate": 1.3293361365755373e-05, + "loss": 1.0068, + "step": 12864 + }, + { + "epoch": 0.41, + "learning_rate": 1.3292383329024672e-05, + "loss": 1.0254, + "step": 12865 + }, + { + "epoch": 0.41, + "learning_rate": 1.3291405256970341e-05, + "loss": 1.0005, + "step": 12866 + }, + { + "epoch": 0.41, + "learning_rate": 1.3290427149602876e-05, + "loss": 1.105, + "step": 12867 + }, + { + "epoch": 0.41, + "learning_rate": 1.328944900693277e-05, + "loss": 0.9395, + "step": 12868 + }, + { + "epoch": 0.41, + "learning_rate": 1.3288470828970516e-05, + "loss": 0.9854, + "step": 12869 + }, + { + "epoch": 0.41, + "learning_rate": 1.3287492615726607e-05, + "loss": 1.1035, + "step": 12870 + }, + { + "epoch": 0.41, + "learning_rate": 1.3286514367211543e-05, + "loss": 1.0063, + "step": 12871 + }, + { + "epoch": 0.41, + "learning_rate": 1.3285536083435819e-05, + "loss": 1.0542, + "step": 12872 + }, + { + "epoch": 0.41, + "learning_rate": 1.3284557764409927e-05, + "loss": 1.022, + "step": 12873 + }, + { + "epoch": 0.41, + "learning_rate": 1.3283579410144363e-05, + "loss": 1.0669, + "step": 12874 + }, + { + "epoch": 0.41, + "learning_rate": 1.3282601020649628e-05, + "loss": 1.0806, + "step": 12875 + }, + { + "epoch": 0.41, + "learning_rate": 1.3281622595936219e-05, + "loss": 1.043, + "step": 12876 + }, + { + "epoch": 0.41, + "learning_rate": 1.328064413601463e-05, + "loss": 1.0547, + "step": 12877 + }, + { + "epoch": 0.41, + "learning_rate": 1.3279665640895359e-05, + "loss": 1.0391, + "step": 12878 + }, + { + "epoch": 0.41, + "learning_rate": 1.3278687110588908e-05, + "loss": 0.9092, + "step": 12879 + }, + { + "epoch": 0.41, + "learning_rate": 1.3277708545105771e-05, + "loss": 0.9253, + "step": 12880 + }, + { + "epoch": 0.41, + "learning_rate": 1.3276729944456448e-05, + "loss": 1.0278, + "step": 12881 + }, + { + "epoch": 0.41, + "learning_rate": 1.327575130865144e-05, + "loss": 1.0503, + "step": 12882 + }, + { + "epoch": 0.41, + "learning_rate": 1.327477263770125e-05, + "loss": 1.0894, + "step": 12883 + }, + { + "epoch": 0.41, + "learning_rate": 1.3273793931616367e-05, + "loss": 0.9927, + "step": 12884 + }, + { + "epoch": 0.41, + "learning_rate": 1.3272815190407301e-05, + "loss": 1.1172, + "step": 12885 + }, + { + "epoch": 0.41, + "learning_rate": 1.327183641408455e-05, + "loss": 1.0347, + "step": 12886 + }, + { + "epoch": 0.41, + "learning_rate": 1.327085760265862e-05, + "loss": 1.1118, + "step": 12887 + }, + { + "epoch": 0.41, + "learning_rate": 1.3269878756140005e-05, + "loss": 0.9902, + "step": 12888 + }, + { + "epoch": 0.41, + "learning_rate": 1.3268899874539207e-05, + "loss": 0.9551, + "step": 12889 + }, + { + "epoch": 0.41, + "learning_rate": 1.3267920957866733e-05, + "loss": 0.853, + "step": 12890 + }, + { + "epoch": 0.41, + "learning_rate": 1.3266942006133086e-05, + "loss": 0.9717, + "step": 12891 + }, + { + "epoch": 0.41, + "learning_rate": 1.3265963019348764e-05, + "loss": 1.0356, + "step": 12892 + }, + { + "epoch": 0.41, + "learning_rate": 1.3264983997524274e-05, + "loss": 1.0713, + "step": 12893 + }, + { + "epoch": 0.41, + "learning_rate": 1.3264004940670122e-05, + "loss": 1.0796, + "step": 12894 + }, + { + "epoch": 0.41, + "learning_rate": 1.3263025848796807e-05, + "loss": 0.999, + "step": 12895 + }, + { + "epoch": 0.41, + "learning_rate": 1.3262046721914834e-05, + "loss": 0.9922, + "step": 12896 + }, + { + "epoch": 0.41, + "learning_rate": 1.3261067560034712e-05, + "loss": 1.125, + "step": 12897 + }, + { + "epoch": 0.41, + "learning_rate": 1.3260088363166946e-05, + "loss": 1.0337, + "step": 12898 + }, + { + "epoch": 0.41, + "learning_rate": 1.325910913132204e-05, + "loss": 1.0117, + "step": 12899 + }, + { + "epoch": 0.41, + "learning_rate": 1.3258129864510498e-05, + "loss": 1.1396, + "step": 12900 + }, + { + "epoch": 0.41, + "learning_rate": 1.325715056274283e-05, + "loss": 1.0415, + "step": 12901 + }, + { + "epoch": 0.41, + "learning_rate": 1.325617122602954e-05, + "loss": 1.0684, + "step": 12902 + }, + { + "epoch": 0.41, + "learning_rate": 1.3255191854381136e-05, + "loss": 1.0991, + "step": 12903 + }, + { + "epoch": 0.41, + "learning_rate": 1.3254212447808125e-05, + "loss": 1.0581, + "step": 12904 + }, + { + "epoch": 0.41, + "learning_rate": 1.3253233006321016e-05, + "loss": 0.9507, + "step": 12905 + }, + { + "epoch": 0.41, + "learning_rate": 1.325225352993032e-05, + "loss": 0.98, + "step": 12906 + }, + { + "epoch": 0.41, + "learning_rate": 1.3251274018646543e-05, + "loss": 1.0586, + "step": 12907 + }, + { + "epoch": 0.41, + "learning_rate": 1.3250294472480192e-05, + "loss": 1.0181, + "step": 12908 + }, + { + "epoch": 0.41, + "learning_rate": 1.324931489144178e-05, + "loss": 1.0288, + "step": 12909 + }, + { + "epoch": 0.41, + "learning_rate": 1.3248335275541817e-05, + "loss": 1.0981, + "step": 12910 + }, + { + "epoch": 0.41, + "learning_rate": 1.3247355624790807e-05, + "loss": 1.0503, + "step": 12911 + }, + { + "epoch": 0.41, + "learning_rate": 1.3246375939199267e-05, + "loss": 0.9473, + "step": 12912 + }, + { + "epoch": 0.41, + "learning_rate": 1.3245396218777708e-05, + "loss": 0.9922, + "step": 12913 + }, + { + "epoch": 0.41, + "learning_rate": 1.324441646353664e-05, + "loss": 1.0405, + "step": 12914 + }, + { + "epoch": 0.41, + "learning_rate": 1.3243436673486572e-05, + "loss": 1.0825, + "step": 12915 + }, + { + "epoch": 0.41, + "learning_rate": 1.3242456848638017e-05, + "loss": 0.9675, + "step": 12916 + }, + { + "epoch": 0.41, + "learning_rate": 1.3241476989001491e-05, + "loss": 1.0791, + "step": 12917 + }, + { + "epoch": 0.41, + "learning_rate": 1.3240497094587503e-05, + "loss": 0.9258, + "step": 12918 + }, + { + "epoch": 0.41, + "learning_rate": 1.3239517165406569e-05, + "loss": 0.9268, + "step": 12919 + }, + { + "epoch": 0.41, + "learning_rate": 1.3238537201469199e-05, + "loss": 1.0156, + "step": 12920 + }, + { + "epoch": 0.41, + "learning_rate": 1.323755720278591e-05, + "loss": 1.0522, + "step": 12921 + }, + { + "epoch": 0.41, + "learning_rate": 1.3236577169367218e-05, + "loss": 0.958, + "step": 12922 + }, + { + "epoch": 0.41, + "learning_rate": 1.3235597101223631e-05, + "loss": 0.9907, + "step": 12923 + }, + { + "epoch": 0.41, + "learning_rate": 1.323461699836567e-05, + "loss": 0.9355, + "step": 12924 + }, + { + "epoch": 0.41, + "learning_rate": 1.323363686080385e-05, + "loss": 0.9556, + "step": 12925 + }, + { + "epoch": 0.41, + "learning_rate": 1.3232656688548682e-05, + "loss": 1.0571, + "step": 12926 + }, + { + "epoch": 0.41, + "learning_rate": 1.3231676481610687e-05, + "loss": 1.0083, + "step": 12927 + }, + { + "epoch": 0.41, + "learning_rate": 1.323069624000038e-05, + "loss": 1.1934, + "step": 12928 + }, + { + "epoch": 0.41, + "learning_rate": 1.3229715963728277e-05, + "loss": 0.9346, + "step": 12929 + }, + { + "epoch": 0.41, + "learning_rate": 1.3228735652804898e-05, + "loss": 1.0449, + "step": 12930 + }, + { + "epoch": 0.41, + "learning_rate": 1.3227755307240757e-05, + "loss": 0.8384, + "step": 12931 + }, + { + "epoch": 0.41, + "learning_rate": 1.3226774927046377e-05, + "loss": 1.1138, + "step": 12932 + }, + { + "epoch": 0.41, + "learning_rate": 1.322579451223227e-05, + "loss": 1.0327, + "step": 12933 + }, + { + "epoch": 0.41, + "learning_rate": 1.322481406280896e-05, + "loss": 1.085, + "step": 12934 + }, + { + "epoch": 0.41, + "learning_rate": 1.322383357878696e-05, + "loss": 1.0151, + "step": 12935 + }, + { + "epoch": 0.41, + "learning_rate": 1.3222853060176798e-05, + "loss": 0.8755, + "step": 12936 + }, + { + "epoch": 0.41, + "learning_rate": 1.3221872506988988e-05, + "loss": 1.0244, + "step": 12937 + }, + { + "epoch": 0.41, + "learning_rate": 1.3220891919234054e-05, + "loss": 1.0225, + "step": 12938 + }, + { + "epoch": 0.41, + "learning_rate": 1.3219911296922513e-05, + "loss": 0.9717, + "step": 12939 + }, + { + "epoch": 0.41, + "learning_rate": 1.3218930640064887e-05, + "loss": 1.0815, + "step": 12940 + }, + { + "epoch": 0.41, + "learning_rate": 1.3217949948671698e-05, + "loss": 1.0449, + "step": 12941 + }, + { + "epoch": 0.41, + "learning_rate": 1.3216969222753466e-05, + "loss": 1.0337, + "step": 12942 + }, + { + "epoch": 0.41, + "learning_rate": 1.3215988462320717e-05, + "loss": 1.1064, + "step": 12943 + }, + { + "epoch": 0.41, + "learning_rate": 1.3215007667383972e-05, + "loss": 1.0698, + "step": 12944 + }, + { + "epoch": 0.41, + "learning_rate": 1.3214026837953749e-05, + "loss": 1.0444, + "step": 12945 + }, + { + "epoch": 0.41, + "learning_rate": 1.3213045974040578e-05, + "loss": 1.0278, + "step": 12946 + }, + { + "epoch": 0.41, + "learning_rate": 1.3212065075654978e-05, + "loss": 1.0581, + "step": 12947 + }, + { + "epoch": 0.41, + "learning_rate": 1.3211084142807476e-05, + "loss": 0.96, + "step": 12948 + }, + { + "epoch": 0.41, + "learning_rate": 1.3210103175508593e-05, + "loss": 1.0903, + "step": 12949 + }, + { + "epoch": 0.41, + "learning_rate": 1.3209122173768859e-05, + "loss": 1.0469, + "step": 12950 + }, + { + "epoch": 0.41, + "learning_rate": 1.3208141137598793e-05, + "loss": 1.1299, + "step": 12951 + }, + { + "epoch": 0.41, + "learning_rate": 1.3207160067008924e-05, + "loss": 0.9692, + "step": 12952 + }, + { + "epoch": 0.41, + "learning_rate": 1.3206178962009778e-05, + "loss": 1.1284, + "step": 12953 + }, + { + "epoch": 0.41, + "learning_rate": 1.3205197822611876e-05, + "loss": 0.9502, + "step": 12954 + }, + { + "epoch": 0.41, + "learning_rate": 1.3204216648825754e-05, + "loss": 1.0654, + "step": 12955 + }, + { + "epoch": 0.41, + "learning_rate": 1.3203235440661932e-05, + "loss": 1.1978, + "step": 12956 + }, + { + "epoch": 0.41, + "learning_rate": 1.3202254198130937e-05, + "loss": 1.0537, + "step": 12957 + }, + { + "epoch": 0.41, + "learning_rate": 1.32012729212433e-05, + "loss": 1.0273, + "step": 12958 + }, + { + "epoch": 0.41, + "learning_rate": 1.3200291610009547e-05, + "loss": 1.0361, + "step": 12959 + }, + { + "epoch": 0.41, + "learning_rate": 1.3199310264440208e-05, + "loss": 0.9507, + "step": 12960 + }, + { + "epoch": 0.41, + "learning_rate": 1.319832888454581e-05, + "loss": 0.915, + "step": 12961 + }, + { + "epoch": 0.41, + "learning_rate": 1.3197347470336884e-05, + "loss": 1.1025, + "step": 12962 + }, + { + "epoch": 0.41, + "learning_rate": 1.3196366021823956e-05, + "loss": 0.9341, + "step": 12963 + }, + { + "epoch": 0.41, + "learning_rate": 1.319538453901756e-05, + "loss": 1.0049, + "step": 12964 + }, + { + "epoch": 0.41, + "learning_rate": 1.3194403021928225e-05, + "loss": 1.1284, + "step": 12965 + }, + { + "epoch": 0.41, + "learning_rate": 1.319342147056648e-05, + "loss": 0.8945, + "step": 12966 + }, + { + "epoch": 0.41, + "learning_rate": 1.3192439884942856e-05, + "loss": 0.9966, + "step": 12967 + }, + { + "epoch": 0.41, + "learning_rate": 1.3191458265067884e-05, + "loss": 0.9941, + "step": 12968 + }, + { + "epoch": 0.41, + "learning_rate": 1.31904766109521e-05, + "loss": 0.9243, + "step": 12969 + }, + { + "epoch": 0.41, + "learning_rate": 1.3189494922606031e-05, + "loss": 1.062, + "step": 12970 + }, + { + "epoch": 0.41, + "learning_rate": 1.3188513200040212e-05, + "loss": 1.0459, + "step": 12971 + }, + { + "epoch": 0.41, + "learning_rate": 1.3187531443265176e-05, + "loss": 1.0269, + "step": 12972 + }, + { + "epoch": 0.41, + "learning_rate": 1.3186549652291454e-05, + "loss": 0.9814, + "step": 12973 + }, + { + "epoch": 0.41, + "learning_rate": 1.3185567827129583e-05, + "loss": 0.9678, + "step": 12974 + }, + { + "epoch": 0.41, + "learning_rate": 1.3184585967790094e-05, + "loss": 1.0342, + "step": 12975 + }, + { + "epoch": 0.41, + "learning_rate": 1.318360407428352e-05, + "loss": 0.9893, + "step": 12976 + }, + { + "epoch": 0.42, + "learning_rate": 1.31826221466204e-05, + "loss": 1.104, + "step": 12977 + }, + { + "epoch": 0.42, + "learning_rate": 1.3181640184811266e-05, + "loss": 1.1416, + "step": 12978 + }, + { + "epoch": 0.42, + "learning_rate": 1.3180658188866656e-05, + "loss": 0.9775, + "step": 12979 + }, + { + "epoch": 0.42, + "learning_rate": 1.3179676158797103e-05, + "loss": 0.9878, + "step": 12980 + }, + { + "epoch": 0.42, + "learning_rate": 1.3178694094613143e-05, + "loss": 1.0088, + "step": 12981 + }, + { + "epoch": 0.42, + "learning_rate": 1.317771199632531e-05, + "loss": 0.9941, + "step": 12982 + }, + { + "epoch": 0.42, + "learning_rate": 1.3176729863944151e-05, + "loss": 1.0503, + "step": 12983 + }, + { + "epoch": 0.42, + "learning_rate": 1.3175747697480192e-05, + "loss": 0.981, + "step": 12984 + }, + { + "epoch": 0.42, + "learning_rate": 1.3174765496943975e-05, + "loss": 1.022, + "step": 12985 + }, + { + "epoch": 0.42, + "learning_rate": 1.317378326234604e-05, + "loss": 0.9146, + "step": 12986 + }, + { + "epoch": 0.42, + "learning_rate": 1.3172800993696921e-05, + "loss": 0.5339, + "step": 12987 + }, + { + "epoch": 0.42, + "learning_rate": 1.3171818691007162e-05, + "loss": 1.0034, + "step": 12988 + }, + { + "epoch": 0.42, + "learning_rate": 1.3170836354287295e-05, + "loss": 1.0566, + "step": 12989 + }, + { + "epoch": 0.42, + "learning_rate": 1.3169853983547864e-05, + "loss": 1.0161, + "step": 12990 + }, + { + "epoch": 0.42, + "learning_rate": 1.316887157879941e-05, + "loss": 0.8369, + "step": 12991 + }, + { + "epoch": 0.42, + "learning_rate": 1.316788914005247e-05, + "loss": 1.1543, + "step": 12992 + }, + { + "epoch": 0.42, + "learning_rate": 1.3166906667317586e-05, + "loss": 1.0117, + "step": 12993 + }, + { + "epoch": 0.42, + "learning_rate": 1.3165924160605297e-05, + "loss": 1.1333, + "step": 12994 + }, + { + "epoch": 0.42, + "learning_rate": 1.3164941619926145e-05, + "loss": 1.0146, + "step": 12995 + }, + { + "epoch": 0.42, + "learning_rate": 1.3163959045290678e-05, + "loss": 0.9834, + "step": 12996 + }, + { + "epoch": 0.42, + "learning_rate": 1.3162976436709425e-05, + "loss": 1.2344, + "step": 12997 + }, + { + "epoch": 0.42, + "learning_rate": 1.3161993794192938e-05, + "loss": 1.022, + "step": 12998 + }, + { + "epoch": 0.42, + "learning_rate": 1.3161011117751756e-05, + "loss": 0.9683, + "step": 12999 + }, + { + "epoch": 0.42, + "learning_rate": 1.3160028407396428e-05, + "loss": 0.9478, + "step": 13000 + }, + { + "epoch": 0.42, + "learning_rate": 1.3159045663137487e-05, + "loss": 1.0527, + "step": 13001 + }, + { + "epoch": 0.42, + "learning_rate": 1.3158062884985485e-05, + "loss": 1.0239, + "step": 13002 + }, + { + "epoch": 0.42, + "learning_rate": 1.315708007295096e-05, + "loss": 1.0356, + "step": 13003 + }, + { + "epoch": 0.42, + "learning_rate": 1.3156097227044467e-05, + "loss": 1.1021, + "step": 13004 + }, + { + "epoch": 0.42, + "learning_rate": 1.3155114347276538e-05, + "loss": 1.0396, + "step": 13005 + }, + { + "epoch": 0.42, + "learning_rate": 1.3154131433657724e-05, + "loss": 0.8794, + "step": 13006 + }, + { + "epoch": 0.42, + "learning_rate": 1.3153148486198573e-05, + "loss": 0.9062, + "step": 13007 + }, + { + "epoch": 0.42, + "learning_rate": 1.3152165504909626e-05, + "loss": 1.0078, + "step": 13008 + }, + { + "epoch": 0.42, + "learning_rate": 1.3151182489801434e-05, + "loss": 1.0527, + "step": 13009 + }, + { + "epoch": 0.42, + "learning_rate": 1.315019944088454e-05, + "loss": 1.0674, + "step": 13010 + }, + { + "epoch": 0.42, + "learning_rate": 1.3149216358169494e-05, + "loss": 1.0747, + "step": 13011 + }, + { + "epoch": 0.42, + "learning_rate": 1.314823324166684e-05, + "loss": 1.0596, + "step": 13012 + }, + { + "epoch": 0.42, + "learning_rate": 1.3147250091387129e-05, + "loss": 1.0493, + "step": 13013 + }, + { + "epoch": 0.42, + "learning_rate": 1.3146266907340907e-05, + "loss": 0.7512, + "step": 13014 + }, + { + "epoch": 0.42, + "learning_rate": 1.3145283689538725e-05, + "loss": 0.9604, + "step": 13015 + }, + { + "epoch": 0.42, + "learning_rate": 1.3144300437991129e-05, + "loss": 1.0415, + "step": 13016 + }, + { + "epoch": 0.42, + "learning_rate": 1.3143317152708668e-05, + "loss": 1.0171, + "step": 13017 + }, + { + "epoch": 0.42, + "learning_rate": 1.3142333833701892e-05, + "loss": 1.0859, + "step": 13018 + }, + { + "epoch": 0.42, + "learning_rate": 1.3141350480981355e-05, + "loss": 1.0562, + "step": 13019 + }, + { + "epoch": 0.42, + "learning_rate": 1.3140367094557602e-05, + "loss": 0.9272, + "step": 13020 + }, + { + "epoch": 0.42, + "learning_rate": 1.3139383674441187e-05, + "loss": 0.9961, + "step": 13021 + }, + { + "epoch": 0.42, + "learning_rate": 1.3138400220642658e-05, + "loss": 1.0215, + "step": 13022 + }, + { + "epoch": 0.42, + "learning_rate": 1.3137416733172567e-05, + "loss": 1.042, + "step": 13023 + }, + { + "epoch": 0.42, + "learning_rate": 1.3136433212041468e-05, + "loss": 1.1685, + "step": 13024 + }, + { + "epoch": 0.42, + "learning_rate": 1.3135449657259915e-05, + "loss": 1.1953, + "step": 13025 + }, + { + "epoch": 0.42, + "learning_rate": 1.3134466068838454e-05, + "loss": 1.0449, + "step": 13026 + }, + { + "epoch": 0.42, + "learning_rate": 1.3133482446787645e-05, + "loss": 1.0522, + "step": 13027 + }, + { + "epoch": 0.42, + "learning_rate": 1.3132498791118032e-05, + "loss": 0.9424, + "step": 13028 + }, + { + "epoch": 0.42, + "learning_rate": 1.3131515101840176e-05, + "loss": 0.9829, + "step": 13029 + }, + { + "epoch": 0.42, + "learning_rate": 1.3130531378964628e-05, + "loss": 0.998, + "step": 13030 + }, + { + "epoch": 0.42, + "learning_rate": 1.3129547622501944e-05, + "loss": 1.1362, + "step": 13031 + }, + { + "epoch": 0.42, + "learning_rate": 1.3128563832462677e-05, + "loss": 1.126, + "step": 13032 + }, + { + "epoch": 0.42, + "learning_rate": 1.3127580008857384e-05, + "loss": 1.0078, + "step": 13033 + }, + { + "epoch": 0.42, + "learning_rate": 1.3126596151696618e-05, + "loss": 1.1182, + "step": 13034 + }, + { + "epoch": 0.42, + "learning_rate": 1.3125612260990934e-05, + "loss": 1.0142, + "step": 13035 + }, + { + "epoch": 0.42, + "learning_rate": 1.312462833675089e-05, + "loss": 1.1299, + "step": 13036 + }, + { + "epoch": 0.42, + "learning_rate": 1.3123644378987043e-05, + "loss": 0.9829, + "step": 13037 + }, + { + "epoch": 0.42, + "learning_rate": 1.3122660387709951e-05, + "loss": 0.9829, + "step": 13038 + }, + { + "epoch": 0.42, + "learning_rate": 1.3121676362930166e-05, + "loss": 1.0181, + "step": 13039 + }, + { + "epoch": 0.42, + "learning_rate": 1.3120692304658252e-05, + "loss": 1.0244, + "step": 13040 + }, + { + "epoch": 0.42, + "learning_rate": 1.3119708212904758e-05, + "loss": 1.0298, + "step": 13041 + }, + { + "epoch": 0.42, + "learning_rate": 1.3118724087680254e-05, + "loss": 0.8896, + "step": 13042 + }, + { + "epoch": 0.42, + "learning_rate": 1.3117739928995287e-05, + "loss": 1.0703, + "step": 13043 + }, + { + "epoch": 0.42, + "learning_rate": 1.3116755736860422e-05, + "loss": 1.0425, + "step": 13044 + }, + { + "epoch": 0.42, + "learning_rate": 1.3115771511286222e-05, + "loss": 0.9771, + "step": 13045 + }, + { + "epoch": 0.42, + "learning_rate": 1.3114787252283237e-05, + "loss": 0.8755, + "step": 13046 + }, + { + "epoch": 0.42, + "learning_rate": 1.3113802959862032e-05, + "loss": 0.5483, + "step": 13047 + }, + { + "epoch": 0.42, + "learning_rate": 1.3112818634033169e-05, + "loss": 1.0093, + "step": 13048 + }, + { + "epoch": 0.42, + "learning_rate": 1.3111834274807209e-05, + "loss": 0.9976, + "step": 13049 + }, + { + "epoch": 0.42, + "learning_rate": 1.311084988219471e-05, + "loss": 0.9028, + "step": 13050 + }, + { + "epoch": 0.42, + "learning_rate": 1.3109865456206232e-05, + "loss": 1.1802, + "step": 13051 + }, + { + "epoch": 0.42, + "learning_rate": 1.310888099685234e-05, + "loss": 0.9419, + "step": 13052 + }, + { + "epoch": 0.42, + "learning_rate": 1.31078965041436e-05, + "loss": 0.9868, + "step": 13053 + }, + { + "epoch": 0.42, + "learning_rate": 1.3106911978090565e-05, + "loss": 1.0142, + "step": 13054 + }, + { + "epoch": 0.42, + "learning_rate": 1.3105927418703805e-05, + "loss": 1.0278, + "step": 13055 + }, + { + "epoch": 0.42, + "learning_rate": 1.3104942825993883e-05, + "loss": 1.0986, + "step": 13056 + }, + { + "epoch": 0.42, + "learning_rate": 1.310395819997136e-05, + "loss": 1.0864, + "step": 13057 + }, + { + "epoch": 0.42, + "learning_rate": 1.31029735406468e-05, + "loss": 0.8911, + "step": 13058 + }, + { + "epoch": 0.42, + "learning_rate": 1.3101988848030767e-05, + "loss": 1.0503, + "step": 13059 + }, + { + "epoch": 0.42, + "learning_rate": 1.310100412213383e-05, + "loss": 0.9932, + "step": 13060 + }, + { + "epoch": 0.42, + "learning_rate": 1.3100019362966547e-05, + "loss": 1.0029, + "step": 13061 + }, + { + "epoch": 0.42, + "learning_rate": 1.309903457053949e-05, + "loss": 1.1177, + "step": 13062 + }, + { + "epoch": 0.42, + "learning_rate": 1.3098049744863221e-05, + "loss": 1.0312, + "step": 13063 + }, + { + "epoch": 0.42, + "learning_rate": 1.3097064885948308e-05, + "loss": 1.0366, + "step": 13064 + }, + { + "epoch": 0.42, + "learning_rate": 1.3096079993805314e-05, + "loss": 1.0195, + "step": 13065 + }, + { + "epoch": 0.42, + "learning_rate": 1.309509506844481e-05, + "loss": 1.0508, + "step": 13066 + }, + { + "epoch": 0.42, + "learning_rate": 1.3094110109877363e-05, + "loss": 1.0146, + "step": 13067 + }, + { + "epoch": 0.42, + "learning_rate": 1.3093125118113537e-05, + "loss": 1.0317, + "step": 13068 + }, + { + "epoch": 0.42, + "learning_rate": 1.3092140093163903e-05, + "loss": 1.0347, + "step": 13069 + }, + { + "epoch": 0.42, + "learning_rate": 1.3091155035039025e-05, + "loss": 1.0356, + "step": 13070 + }, + { + "epoch": 0.42, + "learning_rate": 1.3090169943749475e-05, + "loss": 1.04, + "step": 13071 + }, + { + "epoch": 0.42, + "learning_rate": 1.3089184819305825e-05, + "loss": 0.9717, + "step": 13072 + }, + { + "epoch": 0.42, + "learning_rate": 1.3088199661718639e-05, + "loss": 0.9766, + "step": 13073 + }, + { + "epoch": 0.42, + "learning_rate": 1.3087214470998487e-05, + "loss": 1.1226, + "step": 13074 + }, + { + "epoch": 0.42, + "learning_rate": 1.3086229247155941e-05, + "loss": 1.0132, + "step": 13075 + }, + { + "epoch": 0.42, + "learning_rate": 1.3085243990201572e-05, + "loss": 1.1104, + "step": 13076 + }, + { + "epoch": 0.42, + "learning_rate": 1.3084258700145948e-05, + "loss": 1.0571, + "step": 13077 + }, + { + "epoch": 0.42, + "learning_rate": 1.3083273376999643e-05, + "loss": 1.0366, + "step": 13078 + }, + { + "epoch": 0.42, + "learning_rate": 1.3082288020773228e-05, + "loss": 1.0469, + "step": 13079 + }, + { + "epoch": 0.42, + "learning_rate": 1.3081302631477272e-05, + "loss": 0.8354, + "step": 13080 + }, + { + "epoch": 0.42, + "learning_rate": 1.308031720912235e-05, + "loss": 0.9839, + "step": 13081 + }, + { + "epoch": 0.42, + "learning_rate": 1.3079331753719034e-05, + "loss": 0.9541, + "step": 13082 + }, + { + "epoch": 0.42, + "learning_rate": 1.3078346265277894e-05, + "loss": 1.001, + "step": 13083 + }, + { + "epoch": 0.42, + "learning_rate": 1.307736074380951e-05, + "loss": 1.0278, + "step": 13084 + }, + { + "epoch": 0.42, + "learning_rate": 1.3076375189324446e-05, + "loss": 0.998, + "step": 13085 + }, + { + "epoch": 0.42, + "learning_rate": 1.3075389601833282e-05, + "loss": 0.9365, + "step": 13086 + }, + { + "epoch": 0.42, + "learning_rate": 1.3074403981346594e-05, + "loss": 1.022, + "step": 13087 + }, + { + "epoch": 0.42, + "learning_rate": 1.3073418327874951e-05, + "loss": 1.1187, + "step": 13088 + }, + { + "epoch": 0.42, + "learning_rate": 1.3072432641428931e-05, + "loss": 0.9429, + "step": 13089 + }, + { + "epoch": 0.42, + "learning_rate": 1.3071446922019111e-05, + "loss": 0.9751, + "step": 13090 + }, + { + "epoch": 0.42, + "learning_rate": 1.3070461169656064e-05, + "loss": 0.9907, + "step": 13091 + }, + { + "epoch": 0.42, + "learning_rate": 1.3069475384350367e-05, + "loss": 1.062, + "step": 13092 + }, + { + "epoch": 0.42, + "learning_rate": 1.3068489566112594e-05, + "loss": 1.1284, + "step": 13093 + }, + { + "epoch": 0.42, + "learning_rate": 1.3067503714953327e-05, + "loss": 1.0386, + "step": 13094 + }, + { + "epoch": 0.42, + "learning_rate": 1.306651783088314e-05, + "loss": 1.0054, + "step": 13095 + }, + { + "epoch": 0.42, + "learning_rate": 1.3065531913912607e-05, + "loss": 1.1035, + "step": 13096 + }, + { + "epoch": 0.42, + "learning_rate": 1.3064545964052314e-05, + "loss": 1.0679, + "step": 13097 + }, + { + "epoch": 0.42, + "learning_rate": 1.3063559981312834e-05, + "loss": 0.9595, + "step": 13098 + }, + { + "epoch": 0.42, + "learning_rate": 1.3062573965704744e-05, + "loss": 1.0054, + "step": 13099 + }, + { + "epoch": 0.42, + "learning_rate": 1.3061587917238626e-05, + "loss": 1.0244, + "step": 13100 + }, + { + "epoch": 0.42, + "learning_rate": 1.3060601835925055e-05, + "loss": 0.9297, + "step": 13101 + }, + { + "epoch": 0.42, + "learning_rate": 1.3059615721774617e-05, + "loss": 1.0264, + "step": 13102 + }, + { + "epoch": 0.42, + "learning_rate": 1.3058629574797888e-05, + "loss": 1.0342, + "step": 13103 + }, + { + "epoch": 0.42, + "learning_rate": 1.3057643395005446e-05, + "loss": 0.9673, + "step": 13104 + }, + { + "epoch": 0.42, + "learning_rate": 1.3056657182407877e-05, + "loss": 0.8865, + "step": 13105 + }, + { + "epoch": 0.42, + "learning_rate": 1.305567093701576e-05, + "loss": 1.1055, + "step": 13106 + }, + { + "epoch": 0.42, + "learning_rate": 1.3054684658839674e-05, + "loss": 0.9331, + "step": 13107 + }, + { + "epoch": 0.42, + "learning_rate": 1.3053698347890205e-05, + "loss": 0.979, + "step": 13108 + }, + { + "epoch": 0.42, + "learning_rate": 1.305271200417793e-05, + "loss": 1.0659, + "step": 13109 + }, + { + "epoch": 0.42, + "learning_rate": 1.3051725627713434e-05, + "loss": 0.646, + "step": 13110 + }, + { + "epoch": 0.42, + "learning_rate": 1.30507392185073e-05, + "loss": 1.0483, + "step": 13111 + }, + { + "epoch": 0.42, + "learning_rate": 1.304975277657011e-05, + "loss": 0.9556, + "step": 13112 + }, + { + "epoch": 0.42, + "learning_rate": 1.3048766301912448e-05, + "loss": 1.0059, + "step": 13113 + }, + { + "epoch": 0.42, + "learning_rate": 1.3047779794544899e-05, + "loss": 1.0874, + "step": 13114 + }, + { + "epoch": 0.42, + "learning_rate": 1.3046793254478045e-05, + "loss": 1.0732, + "step": 13115 + }, + { + "epoch": 0.42, + "learning_rate": 1.304580668172247e-05, + "loss": 0.9531, + "step": 13116 + }, + { + "epoch": 0.42, + "learning_rate": 1.3044820076288765e-05, + "loss": 0.9844, + "step": 13117 + }, + { + "epoch": 0.42, + "learning_rate": 1.3043833438187505e-05, + "loss": 0.9507, + "step": 13118 + }, + { + "epoch": 0.42, + "learning_rate": 1.3042846767429283e-05, + "loss": 0.9292, + "step": 13119 + }, + { + "epoch": 0.42, + "learning_rate": 1.3041860064024683e-05, + "loss": 0.9629, + "step": 13120 + }, + { + "epoch": 0.42, + "learning_rate": 1.3040873327984292e-05, + "loss": 0.8779, + "step": 13121 + }, + { + "epoch": 0.42, + "learning_rate": 1.3039886559318693e-05, + "loss": 1.041, + "step": 13122 + }, + { + "epoch": 0.42, + "learning_rate": 1.3038899758038477e-05, + "loss": 0.9185, + "step": 13123 + }, + { + "epoch": 0.42, + "learning_rate": 1.303791292415423e-05, + "loss": 0.9995, + "step": 13124 + }, + { + "epoch": 0.42, + "learning_rate": 1.3036926057676542e-05, + "loss": 0.9355, + "step": 13125 + }, + { + "epoch": 0.42, + "learning_rate": 1.3035939158615993e-05, + "loss": 1.0371, + "step": 13126 + }, + { + "epoch": 0.42, + "learning_rate": 1.3034952226983181e-05, + "loss": 1.0913, + "step": 13127 + }, + { + "epoch": 0.42, + "learning_rate": 1.303396526278869e-05, + "loss": 0.9258, + "step": 13128 + }, + { + "epoch": 0.42, + "learning_rate": 1.303297826604311e-05, + "loss": 1.0059, + "step": 13129 + }, + { + "epoch": 0.42, + "learning_rate": 1.3031991236757029e-05, + "loss": 0.9229, + "step": 13130 + }, + { + "epoch": 0.42, + "learning_rate": 1.3031004174941035e-05, + "loss": 1.0142, + "step": 13131 + }, + { + "epoch": 0.42, + "learning_rate": 1.3030017080605726e-05, + "loss": 0.958, + "step": 13132 + }, + { + "epoch": 0.42, + "learning_rate": 1.3029029953761683e-05, + "loss": 1.0073, + "step": 13133 + }, + { + "epoch": 0.42, + "learning_rate": 1.3028042794419502e-05, + "loss": 1.0347, + "step": 13134 + }, + { + "epoch": 0.42, + "learning_rate": 1.3027055602589774e-05, + "loss": 1.0132, + "step": 13135 + }, + { + "epoch": 0.42, + "learning_rate": 1.302606837828309e-05, + "loss": 0.9946, + "step": 13136 + }, + { + "epoch": 0.42, + "learning_rate": 1.3025081121510042e-05, + "loss": 1.106, + "step": 13137 + }, + { + "epoch": 0.42, + "learning_rate": 1.3024093832281217e-05, + "loss": 0.9517, + "step": 13138 + }, + { + "epoch": 0.42, + "learning_rate": 1.3023106510607216e-05, + "loss": 1.0103, + "step": 13139 + }, + { + "epoch": 0.42, + "learning_rate": 1.302211915649863e-05, + "loss": 1.0854, + "step": 13140 + }, + { + "epoch": 0.42, + "learning_rate": 1.3021131769966044e-05, + "loss": 0.9805, + "step": 13141 + }, + { + "epoch": 0.42, + "learning_rate": 1.3020144351020063e-05, + "loss": 0.9072, + "step": 13142 + }, + { + "epoch": 0.42, + "learning_rate": 1.3019156899671273e-05, + "loss": 0.981, + "step": 13143 + }, + { + "epoch": 0.42, + "learning_rate": 1.3018169415930272e-05, + "loss": 1.0293, + "step": 13144 + }, + { + "epoch": 0.42, + "learning_rate": 1.3017181899807651e-05, + "loss": 1.0557, + "step": 13145 + }, + { + "epoch": 0.42, + "learning_rate": 1.301619435131401e-05, + "loss": 1.0942, + "step": 13146 + }, + { + "epoch": 0.42, + "learning_rate": 1.3015206770459942e-05, + "loss": 1.0034, + "step": 13147 + }, + { + "epoch": 0.42, + "learning_rate": 1.3014219157256043e-05, + "loss": 0.9395, + "step": 13148 + }, + { + "epoch": 0.42, + "learning_rate": 1.3013231511712908e-05, + "loss": 1.0127, + "step": 13149 + }, + { + "epoch": 0.42, + "learning_rate": 1.3012243833841134e-05, + "loss": 0.9204, + "step": 13150 + }, + { + "epoch": 0.42, + "learning_rate": 1.3011256123651315e-05, + "loss": 1.0479, + "step": 13151 + }, + { + "epoch": 0.42, + "learning_rate": 1.301026838115405e-05, + "loss": 0.9663, + "step": 13152 + }, + { + "epoch": 0.42, + "learning_rate": 1.3009280606359943e-05, + "loss": 1.1021, + "step": 13153 + }, + { + "epoch": 0.42, + "learning_rate": 1.3008292799279578e-05, + "loss": 1.0957, + "step": 13154 + }, + { + "epoch": 0.42, + "learning_rate": 1.3007304959923568e-05, + "loss": 0.8989, + "step": 13155 + }, + { + "epoch": 0.42, + "learning_rate": 1.3006317088302498e-05, + "loss": 1.0332, + "step": 13156 + }, + { + "epoch": 0.42, + "learning_rate": 1.3005329184426976e-05, + "loss": 1.0278, + "step": 13157 + }, + { + "epoch": 0.42, + "learning_rate": 1.3004341248307596e-05, + "loss": 1.0933, + "step": 13158 + }, + { + "epoch": 0.42, + "learning_rate": 1.3003353279954961e-05, + "loss": 0.9736, + "step": 13159 + }, + { + "epoch": 0.42, + "learning_rate": 1.3002365279379667e-05, + "loss": 0.9854, + "step": 13160 + }, + { + "epoch": 0.42, + "learning_rate": 1.3001377246592317e-05, + "loss": 0.9624, + "step": 13161 + }, + { + "epoch": 0.42, + "learning_rate": 1.3000389181603515e-05, + "loss": 1.0352, + "step": 13162 + }, + { + "epoch": 0.42, + "learning_rate": 1.2999401084423854e-05, + "loss": 1.0044, + "step": 13163 + }, + { + "epoch": 0.42, + "learning_rate": 1.299841295506394e-05, + "loss": 0.978, + "step": 13164 + }, + { + "epoch": 0.42, + "learning_rate": 1.2997424793534373e-05, + "loss": 1.0825, + "step": 13165 + }, + { + "epoch": 0.42, + "learning_rate": 1.2996436599845757e-05, + "loss": 1.1377, + "step": 13166 + }, + { + "epoch": 0.42, + "learning_rate": 1.299544837400869e-05, + "loss": 1.0776, + "step": 13167 + }, + { + "epoch": 0.42, + "learning_rate": 1.2994460116033776e-05, + "loss": 1.022, + "step": 13168 + }, + { + "epoch": 0.42, + "learning_rate": 1.2993471825931621e-05, + "loss": 1.0513, + "step": 13169 + }, + { + "epoch": 0.42, + "learning_rate": 1.299248350371283e-05, + "loss": 0.9614, + "step": 13170 + }, + { + "epoch": 0.42, + "learning_rate": 1.2991495149387998e-05, + "loss": 1.1177, + "step": 13171 + }, + { + "epoch": 0.42, + "learning_rate": 1.2990506762967734e-05, + "loss": 0.9912, + "step": 13172 + }, + { + "epoch": 0.42, + "learning_rate": 1.2989518344462644e-05, + "loss": 0.9497, + "step": 13173 + }, + { + "epoch": 0.42, + "learning_rate": 1.2988529893883333e-05, + "loss": 1.082, + "step": 13174 + }, + { + "epoch": 0.42, + "learning_rate": 1.2987541411240399e-05, + "loss": 0.9194, + "step": 13175 + }, + { + "epoch": 0.42, + "learning_rate": 1.2986552896544453e-05, + "loss": 1.0337, + "step": 13176 + }, + { + "epoch": 0.42, + "learning_rate": 1.2985564349806101e-05, + "loss": 1.002, + "step": 13177 + }, + { + "epoch": 0.42, + "learning_rate": 1.298457577103595e-05, + "loss": 0.8511, + "step": 13178 + }, + { + "epoch": 0.42, + "learning_rate": 1.2983587160244602e-05, + "loss": 1.0464, + "step": 13179 + }, + { + "epoch": 0.42, + "learning_rate": 1.2982598517442664e-05, + "loss": 1.0459, + "step": 13180 + }, + { + "epoch": 0.42, + "learning_rate": 1.2981609842640749e-05, + "loss": 1.0298, + "step": 13181 + }, + { + "epoch": 0.42, + "learning_rate": 1.2980621135849456e-05, + "loss": 1.1665, + "step": 13182 + }, + { + "epoch": 0.42, + "learning_rate": 1.29796323970794e-05, + "loss": 1.0225, + "step": 13183 + }, + { + "epoch": 0.42, + "learning_rate": 1.2978643626341183e-05, + "loss": 1.0039, + "step": 13184 + }, + { + "epoch": 0.42, + "learning_rate": 1.2977654823645421e-05, + "loss": 1.0103, + "step": 13185 + }, + { + "epoch": 0.42, + "learning_rate": 1.2976665989002715e-05, + "loss": 0.9546, + "step": 13186 + }, + { + "epoch": 0.42, + "learning_rate": 1.2975677122423678e-05, + "loss": 1.124, + "step": 13187 + }, + { + "epoch": 0.42, + "learning_rate": 1.2974688223918918e-05, + "loss": 0.9771, + "step": 13188 + }, + { + "epoch": 0.42, + "learning_rate": 1.2973699293499047e-05, + "loss": 0.9512, + "step": 13189 + }, + { + "epoch": 0.42, + "learning_rate": 1.2972710331174674e-05, + "loss": 1.0986, + "step": 13190 + }, + { + "epoch": 0.42, + "learning_rate": 1.2971721336956407e-05, + "loss": 0.9614, + "step": 13191 + }, + { + "epoch": 0.42, + "learning_rate": 1.2970732310854861e-05, + "loss": 1.0107, + "step": 13192 + }, + { + "epoch": 0.42, + "learning_rate": 1.2969743252880644e-05, + "loss": 1.0332, + "step": 13193 + }, + { + "epoch": 0.42, + "learning_rate": 1.2968754163044369e-05, + "loss": 0.5273, + "step": 13194 + }, + { + "epoch": 0.42, + "learning_rate": 1.2967765041356649e-05, + "loss": 1.0127, + "step": 13195 + }, + { + "epoch": 0.42, + "learning_rate": 1.2966775887828093e-05, + "loss": 1.041, + "step": 13196 + }, + { + "epoch": 0.42, + "learning_rate": 1.2965786702469316e-05, + "loss": 1.001, + "step": 13197 + }, + { + "epoch": 0.42, + "learning_rate": 1.296479748529093e-05, + "loss": 0.8979, + "step": 13198 + }, + { + "epoch": 0.42, + "learning_rate": 1.2963808236303547e-05, + "loss": 1.0322, + "step": 13199 + }, + { + "epoch": 0.42, + "learning_rate": 1.2962818955517785e-05, + "loss": 1.02, + "step": 13200 + }, + { + "epoch": 0.42, + "learning_rate": 1.2961829642944251e-05, + "loss": 0.9263, + "step": 13201 + }, + { + "epoch": 0.42, + "learning_rate": 1.2960840298593565e-05, + "loss": 0.9409, + "step": 13202 + }, + { + "epoch": 0.42, + "learning_rate": 1.2959850922476336e-05, + "loss": 0.9062, + "step": 13203 + }, + { + "epoch": 0.42, + "learning_rate": 1.2958861514603187e-05, + "loss": 1.1172, + "step": 13204 + }, + { + "epoch": 0.42, + "learning_rate": 1.2957872074984725e-05, + "loss": 1.0298, + "step": 13205 + }, + { + "epoch": 0.42, + "learning_rate": 1.295688260363157e-05, + "loss": 1.0684, + "step": 13206 + }, + { + "epoch": 0.42, + "learning_rate": 1.295589310055434e-05, + "loss": 0.978, + "step": 13207 + }, + { + "epoch": 0.42, + "learning_rate": 1.2954903565763648e-05, + "loss": 0.9927, + "step": 13208 + }, + { + "epoch": 0.42, + "learning_rate": 1.295391399927011e-05, + "loss": 1.019, + "step": 13209 + }, + { + "epoch": 0.42, + "learning_rate": 1.295292440108434e-05, + "loss": 0.9702, + "step": 13210 + }, + { + "epoch": 0.42, + "learning_rate": 1.2951934771216965e-05, + "loss": 1.0327, + "step": 13211 + }, + { + "epoch": 0.42, + "learning_rate": 1.2950945109678596e-05, + "loss": 1.0703, + "step": 13212 + }, + { + "epoch": 0.42, + "learning_rate": 1.2949955416479849e-05, + "loss": 1.0298, + "step": 13213 + }, + { + "epoch": 0.42, + "learning_rate": 1.2948965691631348e-05, + "loss": 0.9951, + "step": 13214 + }, + { + "epoch": 0.42, + "learning_rate": 1.294797593514371e-05, + "loss": 1.0957, + "step": 13215 + }, + { + "epoch": 0.42, + "learning_rate": 1.2946986147027549e-05, + "loss": 1.0342, + "step": 13216 + }, + { + "epoch": 0.42, + "learning_rate": 1.294599632729349e-05, + "loss": 1.0166, + "step": 13217 + }, + { + "epoch": 0.42, + "learning_rate": 1.294500647595215e-05, + "loss": 1.0137, + "step": 13218 + }, + { + "epoch": 0.42, + "learning_rate": 1.2944016593014152e-05, + "loss": 1.0103, + "step": 13219 + }, + { + "epoch": 0.42, + "learning_rate": 1.2943026678490114e-05, + "loss": 1.0005, + "step": 13220 + }, + { + "epoch": 0.42, + "learning_rate": 1.2942036732390656e-05, + "loss": 0.9834, + "step": 13221 + }, + { + "epoch": 0.42, + "learning_rate": 1.2941046754726398e-05, + "loss": 1.0781, + "step": 13222 + }, + { + "epoch": 0.42, + "learning_rate": 1.2940056745507967e-05, + "loss": 0.9683, + "step": 13223 + }, + { + "epoch": 0.42, + "learning_rate": 1.293906670474598e-05, + "loss": 0.9536, + "step": 13224 + }, + { + "epoch": 0.42, + "learning_rate": 1.2938076632451058e-05, + "loss": 0.9521, + "step": 13225 + }, + { + "epoch": 0.42, + "learning_rate": 1.2937086528633827e-05, + "loss": 1.1665, + "step": 13226 + }, + { + "epoch": 0.42, + "learning_rate": 1.293609639330491e-05, + "loss": 1.0386, + "step": 13227 + }, + { + "epoch": 0.42, + "learning_rate": 1.2935106226474923e-05, + "loss": 0.9243, + "step": 13228 + }, + { + "epoch": 0.42, + "learning_rate": 1.2934116028154497e-05, + "loss": 0.9023, + "step": 13229 + }, + { + "epoch": 0.42, + "learning_rate": 1.2933125798354255e-05, + "loss": 1.0347, + "step": 13230 + }, + { + "epoch": 0.42, + "learning_rate": 1.293213553708482e-05, + "loss": 0.896, + "step": 13231 + }, + { + "epoch": 0.42, + "learning_rate": 1.2931145244356813e-05, + "loss": 0.9199, + "step": 13232 + }, + { + "epoch": 0.42, + "learning_rate": 1.2930154920180862e-05, + "loss": 1.1128, + "step": 13233 + }, + { + "epoch": 0.42, + "learning_rate": 1.2929164564567593e-05, + "loss": 0.9233, + "step": 13234 + }, + { + "epoch": 0.42, + "learning_rate": 1.2928174177527631e-05, + "loss": 1.0166, + "step": 13235 + }, + { + "epoch": 0.42, + "learning_rate": 1.29271837590716e-05, + "loss": 1.0332, + "step": 13236 + }, + { + "epoch": 0.42, + "learning_rate": 1.2926193309210127e-05, + "loss": 1.084, + "step": 13237 + }, + { + "epoch": 0.42, + "learning_rate": 1.2925202827953835e-05, + "loss": 0.9053, + "step": 13238 + }, + { + "epoch": 0.42, + "learning_rate": 1.292421231531336e-05, + "loss": 0.9707, + "step": 13239 + }, + { + "epoch": 0.42, + "learning_rate": 1.2923221771299319e-05, + "loss": 1.0542, + "step": 13240 + }, + { + "epoch": 0.42, + "learning_rate": 1.2922231195922344e-05, + "loss": 1.0503, + "step": 13241 + }, + { + "epoch": 0.42, + "learning_rate": 1.2921240589193062e-05, + "loss": 1.0073, + "step": 13242 + }, + { + "epoch": 0.42, + "learning_rate": 1.2920249951122103e-05, + "loss": 0.9648, + "step": 13243 + }, + { + "epoch": 0.42, + "learning_rate": 1.2919259281720093e-05, + "loss": 0.814, + "step": 13244 + }, + { + "epoch": 0.42, + "learning_rate": 1.2918268580997665e-05, + "loss": 1.1426, + "step": 13245 + }, + { + "epoch": 0.42, + "learning_rate": 1.2917277848965442e-05, + "loss": 1.021, + "step": 13246 + }, + { + "epoch": 0.42, + "learning_rate": 1.2916287085634057e-05, + "loss": 1.0903, + "step": 13247 + }, + { + "epoch": 0.42, + "learning_rate": 1.2915296291014136e-05, + "loss": 1.0869, + "step": 13248 + }, + { + "epoch": 0.42, + "learning_rate": 1.2914305465116316e-05, + "loss": 1.0054, + "step": 13249 + }, + { + "epoch": 0.42, + "learning_rate": 1.2913314607951224e-05, + "loss": 0.9517, + "step": 13250 + }, + { + "epoch": 0.42, + "learning_rate": 1.2912323719529489e-05, + "loss": 0.978, + "step": 13251 + }, + { + "epoch": 0.42, + "learning_rate": 1.2911332799861742e-05, + "loss": 0.9268, + "step": 13252 + }, + { + "epoch": 0.42, + "learning_rate": 1.291034184895862e-05, + "loss": 1.0327, + "step": 13253 + }, + { + "epoch": 0.42, + "learning_rate": 1.290935086683075e-05, + "loss": 1.0269, + "step": 13254 + }, + { + "epoch": 0.42, + "learning_rate": 1.2908359853488762e-05, + "loss": 1.0586, + "step": 13255 + }, + { + "epoch": 0.42, + "learning_rate": 1.2907368808943295e-05, + "loss": 0.9556, + "step": 13256 + }, + { + "epoch": 0.42, + "learning_rate": 1.2906377733204978e-05, + "loss": 1.0581, + "step": 13257 + }, + { + "epoch": 0.42, + "learning_rate": 1.2905386626284444e-05, + "loss": 1.0952, + "step": 13258 + }, + { + "epoch": 0.42, + "learning_rate": 1.2904395488192326e-05, + "loss": 1.0845, + "step": 13259 + }, + { + "epoch": 0.42, + "learning_rate": 1.2903404318939259e-05, + "loss": 1.0923, + "step": 13260 + }, + { + "epoch": 0.42, + "learning_rate": 1.2902413118535879e-05, + "loss": 0.9434, + "step": 13261 + }, + { + "epoch": 0.42, + "learning_rate": 1.2901421886992815e-05, + "loss": 0.9414, + "step": 13262 + }, + { + "epoch": 0.42, + "learning_rate": 1.2900430624320707e-05, + "loss": 0.9546, + "step": 13263 + }, + { + "epoch": 0.42, + "learning_rate": 1.2899439330530192e-05, + "loss": 0.8604, + "step": 13264 + }, + { + "epoch": 0.42, + "learning_rate": 1.2898448005631898e-05, + "loss": 1.0015, + "step": 13265 + }, + { + "epoch": 0.42, + "learning_rate": 1.2897456649636463e-05, + "loss": 1.0005, + "step": 13266 + }, + { + "epoch": 0.42, + "learning_rate": 1.289646526255453e-05, + "loss": 1.0347, + "step": 13267 + }, + { + "epoch": 0.42, + "learning_rate": 1.2895473844396728e-05, + "loss": 0.939, + "step": 13268 + }, + { + "epoch": 0.42, + "learning_rate": 1.2894482395173695e-05, + "loss": 1.0864, + "step": 13269 + }, + { + "epoch": 0.42, + "learning_rate": 1.289349091489607e-05, + "loss": 1.0684, + "step": 13270 + }, + { + "epoch": 0.42, + "learning_rate": 1.289249940357449e-05, + "loss": 1.0806, + "step": 13271 + }, + { + "epoch": 0.42, + "learning_rate": 1.2891507861219595e-05, + "loss": 1.0269, + "step": 13272 + }, + { + "epoch": 0.42, + "learning_rate": 1.289051628784202e-05, + "loss": 0.9116, + "step": 13273 + }, + { + "epoch": 0.42, + "learning_rate": 1.28895246834524e-05, + "loss": 1.0186, + "step": 13274 + }, + { + "epoch": 0.42, + "learning_rate": 1.288853304806138e-05, + "loss": 1.104, + "step": 13275 + }, + { + "epoch": 0.42, + "learning_rate": 1.28875413816796e-05, + "loss": 1.001, + "step": 13276 + }, + { + "epoch": 0.42, + "learning_rate": 1.2886549684317695e-05, + "loss": 0.9351, + "step": 13277 + }, + { + "epoch": 0.42, + "learning_rate": 1.2885557955986307e-05, + "loss": 1.0396, + "step": 13278 + }, + { + "epoch": 0.42, + "learning_rate": 1.2884566196696076e-05, + "loss": 0.8438, + "step": 13279 + }, + { + "epoch": 0.42, + "learning_rate": 1.288357440645764e-05, + "loss": 1.0, + "step": 13280 + }, + { + "epoch": 0.42, + "learning_rate": 1.2882582585281647e-05, + "loss": 0.9473, + "step": 13281 + }, + { + "epoch": 0.42, + "learning_rate": 1.2881590733178729e-05, + "loss": 0.9629, + "step": 13282 + }, + { + "epoch": 0.42, + "learning_rate": 1.2880598850159532e-05, + "loss": 1.0801, + "step": 13283 + }, + { + "epoch": 0.42, + "learning_rate": 1.28796069362347e-05, + "loss": 1.0645, + "step": 13284 + }, + { + "epoch": 0.42, + "learning_rate": 1.2878614991414869e-05, + "loss": 1.0088, + "step": 13285 + }, + { + "epoch": 0.42, + "learning_rate": 1.2877623015710686e-05, + "loss": 1.0435, + "step": 13286 + }, + { + "epoch": 0.42, + "learning_rate": 1.2876631009132793e-05, + "loss": 1.0859, + "step": 13287 + }, + { + "epoch": 0.42, + "learning_rate": 1.2875638971691833e-05, + "loss": 1.0142, + "step": 13288 + }, + { + "epoch": 0.42, + "learning_rate": 1.287464690339845e-05, + "loss": 1.0947, + "step": 13289 + }, + { + "epoch": 0.43, + "learning_rate": 1.2873654804263285e-05, + "loss": 1.0791, + "step": 13290 + }, + { + "epoch": 0.43, + "learning_rate": 1.2872662674296987e-05, + "loss": 1.0674, + "step": 13291 + }, + { + "epoch": 0.43, + "learning_rate": 1.2871670513510198e-05, + "loss": 1.0825, + "step": 13292 + }, + { + "epoch": 0.43, + "learning_rate": 1.2870678321913559e-05, + "loss": 1.042, + "step": 13293 + }, + { + "epoch": 0.43, + "learning_rate": 1.2869686099517721e-05, + "loss": 1.0601, + "step": 13294 + }, + { + "epoch": 0.43, + "learning_rate": 1.286869384633333e-05, + "loss": 0.9863, + "step": 13295 + }, + { + "epoch": 0.43, + "learning_rate": 1.2867701562371027e-05, + "loss": 1.0913, + "step": 13296 + }, + { + "epoch": 0.43, + "learning_rate": 1.2866709247641457e-05, + "loss": 0.9238, + "step": 13297 + }, + { + "epoch": 0.43, + "learning_rate": 1.2865716902155275e-05, + "loss": 0.9536, + "step": 13298 + }, + { + "epoch": 0.43, + "learning_rate": 1.2864724525923117e-05, + "loss": 0.999, + "step": 13299 + }, + { + "epoch": 0.43, + "learning_rate": 1.2863732118955638e-05, + "loss": 0.5227, + "step": 13300 + }, + { + "epoch": 0.43, + "learning_rate": 1.2862739681263481e-05, + "loss": 1.0029, + "step": 13301 + }, + { + "epoch": 0.43, + "learning_rate": 1.2861747212857299e-05, + "loss": 0.8984, + "step": 13302 + }, + { + "epoch": 0.43, + "learning_rate": 1.2860754713747734e-05, + "loss": 0.8677, + "step": 13303 + }, + { + "epoch": 0.43, + "learning_rate": 1.2859762183945437e-05, + "loss": 1.1152, + "step": 13304 + }, + { + "epoch": 0.43, + "learning_rate": 1.2858769623461055e-05, + "loss": 0.9902, + "step": 13305 + }, + { + "epoch": 0.43, + "learning_rate": 1.2857777032305245e-05, + "loss": 1.0562, + "step": 13306 + }, + { + "epoch": 0.43, + "learning_rate": 1.2856784410488646e-05, + "loss": 0.9604, + "step": 13307 + }, + { + "epoch": 0.43, + "learning_rate": 1.285579175802191e-05, + "loss": 0.9482, + "step": 13308 + }, + { + "epoch": 0.43, + "learning_rate": 1.2854799074915692e-05, + "loss": 0.9277, + "step": 13309 + }, + { + "epoch": 0.43, + "learning_rate": 1.2853806361180641e-05, + "loss": 1.0215, + "step": 13310 + }, + { + "epoch": 0.43, + "learning_rate": 1.2852813616827404e-05, + "loss": 0.9697, + "step": 13311 + }, + { + "epoch": 0.43, + "learning_rate": 1.2851820841866633e-05, + "loss": 1.0732, + "step": 13312 + }, + { + "epoch": 0.43, + "learning_rate": 1.2850828036308984e-05, + "loss": 1.1191, + "step": 13313 + }, + { + "epoch": 0.43, + "learning_rate": 1.2849835200165104e-05, + "loss": 1.0791, + "step": 13314 + }, + { + "epoch": 0.43, + "learning_rate": 1.2848842333445644e-05, + "loss": 1.0083, + "step": 13315 + }, + { + "epoch": 0.43, + "learning_rate": 1.284784943616126e-05, + "loss": 1.0303, + "step": 13316 + }, + { + "epoch": 0.43, + "learning_rate": 1.2846856508322605e-05, + "loss": 1.0425, + "step": 13317 + }, + { + "epoch": 0.43, + "learning_rate": 1.284586354994033e-05, + "loss": 1.0591, + "step": 13318 + }, + { + "epoch": 0.43, + "learning_rate": 1.2844870561025087e-05, + "loss": 1.0571, + "step": 13319 + }, + { + "epoch": 0.43, + "learning_rate": 1.2843877541587533e-05, + "loss": 0.9375, + "step": 13320 + }, + { + "epoch": 0.43, + "learning_rate": 1.2842884491638319e-05, + "loss": 1.0073, + "step": 13321 + }, + { + "epoch": 0.43, + "learning_rate": 1.28418914111881e-05, + "loss": 0.9785, + "step": 13322 + }, + { + "epoch": 0.43, + "learning_rate": 1.2840898300247536e-05, + "loss": 1.082, + "step": 13323 + }, + { + "epoch": 0.43, + "learning_rate": 1.2839905158827274e-05, + "loss": 1.022, + "step": 13324 + }, + { + "epoch": 0.43, + "learning_rate": 1.2838911986937974e-05, + "loss": 0.9048, + "step": 13325 + }, + { + "epoch": 0.43, + "learning_rate": 1.2837918784590289e-05, + "loss": 1.0117, + "step": 13326 + }, + { + "epoch": 0.43, + "learning_rate": 1.2836925551794876e-05, + "loss": 0.8965, + "step": 13327 + }, + { + "epoch": 0.43, + "learning_rate": 1.2835932288562391e-05, + "loss": 1.103, + "step": 13328 + }, + { + "epoch": 0.43, + "learning_rate": 1.2834938994903494e-05, + "loss": 0.9521, + "step": 13329 + }, + { + "epoch": 0.43, + "learning_rate": 1.2833945670828838e-05, + "loss": 1.1509, + "step": 13330 + }, + { + "epoch": 0.43, + "learning_rate": 1.2832952316349082e-05, + "loss": 1.0957, + "step": 13331 + }, + { + "epoch": 0.43, + "learning_rate": 1.2831958931474882e-05, + "loss": 0.9731, + "step": 13332 + }, + { + "epoch": 0.43, + "learning_rate": 1.2830965516216898e-05, + "loss": 1.0273, + "step": 13333 + }, + { + "epoch": 0.43, + "learning_rate": 1.2829972070585784e-05, + "loss": 0.9985, + "step": 13334 + }, + { + "epoch": 0.43, + "learning_rate": 1.2828978594592203e-05, + "loss": 1.1094, + "step": 13335 + }, + { + "epoch": 0.43, + "learning_rate": 1.2827985088246815e-05, + "loss": 1.0322, + "step": 13336 + }, + { + "epoch": 0.43, + "learning_rate": 1.2826991551560276e-05, + "loss": 0.9409, + "step": 13337 + }, + { + "epoch": 0.43, + "learning_rate": 1.2825997984543243e-05, + "loss": 0.9209, + "step": 13338 + }, + { + "epoch": 0.43, + "learning_rate": 1.2825004387206383e-05, + "loss": 0.9717, + "step": 13339 + }, + { + "epoch": 0.43, + "learning_rate": 1.2824010759560353e-05, + "loss": 1.0762, + "step": 13340 + }, + { + "epoch": 0.43, + "learning_rate": 1.282301710161581e-05, + "loss": 0.9775, + "step": 13341 + }, + { + "epoch": 0.43, + "learning_rate": 1.282202341338342e-05, + "loss": 0.9641, + "step": 13342 + }, + { + "epoch": 0.43, + "learning_rate": 1.2821029694873842e-05, + "loss": 1.0259, + "step": 13343 + }, + { + "epoch": 0.43, + "learning_rate": 1.2820035946097738e-05, + "loss": 1.0444, + "step": 13344 + }, + { + "epoch": 0.43, + "learning_rate": 1.2819042167065767e-05, + "loss": 0.958, + "step": 13345 + }, + { + "epoch": 0.43, + "learning_rate": 1.2818048357788594e-05, + "loss": 1.0015, + "step": 13346 + }, + { + "epoch": 0.43, + "learning_rate": 1.2817054518276884e-05, + "loss": 0.9546, + "step": 13347 + }, + { + "epoch": 0.43, + "learning_rate": 1.2816060648541293e-05, + "loss": 0.959, + "step": 13348 + }, + { + "epoch": 0.43, + "learning_rate": 1.2815066748592488e-05, + "loss": 1.0229, + "step": 13349 + }, + { + "epoch": 0.43, + "learning_rate": 1.2814072818441135e-05, + "loss": 0.9023, + "step": 13350 + }, + { + "epoch": 0.43, + "learning_rate": 1.2813078858097895e-05, + "loss": 1.0278, + "step": 13351 + }, + { + "epoch": 0.43, + "learning_rate": 1.281208486757343e-05, + "loss": 1.0757, + "step": 13352 + }, + { + "epoch": 0.43, + "learning_rate": 1.2811090846878409e-05, + "loss": 1.0303, + "step": 13353 + }, + { + "epoch": 0.43, + "learning_rate": 1.2810096796023492e-05, + "loss": 1.0811, + "step": 13354 + }, + { + "epoch": 0.43, + "learning_rate": 1.2809102715019348e-05, + "loss": 1.0142, + "step": 13355 + }, + { + "epoch": 0.43, + "learning_rate": 1.2808108603876641e-05, + "loss": 1.1016, + "step": 13356 + }, + { + "epoch": 0.43, + "learning_rate": 1.2807114462606036e-05, + "loss": 1.0337, + "step": 13357 + }, + { + "epoch": 0.43, + "learning_rate": 1.2806120291218198e-05, + "loss": 0.9639, + "step": 13358 + }, + { + "epoch": 0.43, + "learning_rate": 1.2805126089723798e-05, + "loss": 1.1538, + "step": 13359 + }, + { + "epoch": 0.43, + "learning_rate": 1.2804131858133497e-05, + "loss": 1.0059, + "step": 13360 + }, + { + "epoch": 0.43, + "learning_rate": 1.2803137596457966e-05, + "loss": 0.8994, + "step": 13361 + }, + { + "epoch": 0.43, + "learning_rate": 1.2802143304707873e-05, + "loss": 1.0254, + "step": 13362 + }, + { + "epoch": 0.43, + "learning_rate": 1.280114898289388e-05, + "loss": 0.9307, + "step": 13363 + }, + { + "epoch": 0.43, + "learning_rate": 1.2800154631026663e-05, + "loss": 1.062, + "step": 13364 + }, + { + "epoch": 0.43, + "learning_rate": 1.2799160249116879e-05, + "loss": 1.1094, + "step": 13365 + }, + { + "epoch": 0.43, + "learning_rate": 1.279816583717521e-05, + "loss": 1.0547, + "step": 13366 + }, + { + "epoch": 0.43, + "learning_rate": 1.2797171395212316e-05, + "loss": 0.9956, + "step": 13367 + }, + { + "epoch": 0.43, + "learning_rate": 1.2796176923238867e-05, + "loss": 0.9922, + "step": 13368 + }, + { + "epoch": 0.43, + "learning_rate": 1.2795182421265536e-05, + "loss": 1.0806, + "step": 13369 + }, + { + "epoch": 0.43, + "learning_rate": 1.2794187889302991e-05, + "loss": 0.9697, + "step": 13370 + }, + { + "epoch": 0.43, + "learning_rate": 1.2793193327361904e-05, + "loss": 0.9692, + "step": 13371 + }, + { + "epoch": 0.43, + "learning_rate": 1.2792198735452938e-05, + "loss": 1.0337, + "step": 13372 + }, + { + "epoch": 0.43, + "learning_rate": 1.2791204113586774e-05, + "loss": 1.0347, + "step": 13373 + }, + { + "epoch": 0.43, + "learning_rate": 1.2790209461774082e-05, + "loss": 0.8799, + "step": 13374 + }, + { + "epoch": 0.43, + "learning_rate": 1.2789214780025527e-05, + "loss": 1.0435, + "step": 13375 + }, + { + "epoch": 0.43, + "learning_rate": 1.2788220068351785e-05, + "loss": 0.8862, + "step": 13376 + }, + { + "epoch": 0.43, + "learning_rate": 1.2787225326763526e-05, + "loss": 0.9907, + "step": 13377 + }, + { + "epoch": 0.43, + "learning_rate": 1.2786230555271425e-05, + "loss": 0.9482, + "step": 13378 + }, + { + "epoch": 0.43, + "learning_rate": 1.2785235753886155e-05, + "loss": 0.9033, + "step": 13379 + }, + { + "epoch": 0.43, + "learning_rate": 1.2784240922618388e-05, + "loss": 0.9517, + "step": 13380 + }, + { + "epoch": 0.43, + "learning_rate": 1.2783246061478796e-05, + "loss": 0.9941, + "step": 13381 + }, + { + "epoch": 0.43, + "learning_rate": 1.2782251170478054e-05, + "loss": 0.9092, + "step": 13382 + }, + { + "epoch": 0.43, + "learning_rate": 1.2781256249626837e-05, + "loss": 0.9844, + "step": 13383 + }, + { + "epoch": 0.43, + "learning_rate": 1.2780261298935818e-05, + "loss": 0.9365, + "step": 13384 + }, + { + "epoch": 0.43, + "learning_rate": 1.2779266318415675e-05, + "loss": 1.0928, + "step": 13385 + }, + { + "epoch": 0.43, + "learning_rate": 1.2778271308077077e-05, + "loss": 0.9194, + "step": 13386 + }, + { + "epoch": 0.43, + "learning_rate": 1.2777276267930702e-05, + "loss": 0.5454, + "step": 13387 + }, + { + "epoch": 0.43, + "learning_rate": 1.2776281197987227e-05, + "loss": 0.9756, + "step": 13388 + }, + { + "epoch": 0.43, + "learning_rate": 1.277528609825733e-05, + "loss": 1.0127, + "step": 13389 + }, + { + "epoch": 0.43, + "learning_rate": 1.2774290968751683e-05, + "loss": 0.8779, + "step": 13390 + }, + { + "epoch": 0.43, + "learning_rate": 1.2773295809480963e-05, + "loss": 0.9692, + "step": 13391 + }, + { + "epoch": 0.43, + "learning_rate": 1.277230062045585e-05, + "loss": 0.8892, + "step": 13392 + }, + { + "epoch": 0.43, + "learning_rate": 1.277130540168702e-05, + "loss": 1.0562, + "step": 13393 + }, + { + "epoch": 0.43, + "learning_rate": 1.277031015318515e-05, + "loss": 1.0161, + "step": 13394 + }, + { + "epoch": 0.43, + "learning_rate": 1.2769314874960916e-05, + "loss": 0.9297, + "step": 13395 + }, + { + "epoch": 0.43, + "learning_rate": 1.2768319567025003e-05, + "loss": 1.0776, + "step": 13396 + }, + { + "epoch": 0.43, + "learning_rate": 1.276732422938808e-05, + "loss": 0.9741, + "step": 13397 + }, + { + "epoch": 0.43, + "learning_rate": 1.2766328862060833e-05, + "loss": 1.0298, + "step": 13398 + }, + { + "epoch": 0.43, + "learning_rate": 1.2765333465053938e-05, + "loss": 1.0566, + "step": 13399 + }, + { + "epoch": 0.43, + "learning_rate": 1.2764338038378077e-05, + "loss": 0.8711, + "step": 13400 + }, + { + "epoch": 0.43, + "learning_rate": 1.2763342582043925e-05, + "loss": 1.0, + "step": 13401 + }, + { + "epoch": 0.43, + "learning_rate": 1.276234709606217e-05, + "loss": 1.0654, + "step": 13402 + }, + { + "epoch": 0.43, + "learning_rate": 1.2761351580443483e-05, + "loss": 0.9751, + "step": 13403 + }, + { + "epoch": 0.43, + "learning_rate": 1.2760356035198553e-05, + "loss": 0.9517, + "step": 13404 + }, + { + "epoch": 0.43, + "learning_rate": 1.2759360460338057e-05, + "loss": 1.0273, + "step": 13405 + }, + { + "epoch": 0.43, + "learning_rate": 1.2758364855872677e-05, + "loss": 0.9473, + "step": 13406 + }, + { + "epoch": 0.43, + "learning_rate": 1.2757369221813097e-05, + "loss": 1.0112, + "step": 13407 + }, + { + "epoch": 0.43, + "learning_rate": 1.2756373558169992e-05, + "loss": 1.0098, + "step": 13408 + }, + { + "epoch": 0.43, + "learning_rate": 1.2755377864954056e-05, + "loss": 1.043, + "step": 13409 + }, + { + "epoch": 0.43, + "learning_rate": 1.2754382142175961e-05, + "loss": 0.9136, + "step": 13410 + }, + { + "epoch": 0.43, + "learning_rate": 1.2753386389846391e-05, + "loss": 1.0381, + "step": 13411 + }, + { + "epoch": 0.43, + "learning_rate": 1.2752390607976037e-05, + "loss": 1.0503, + "step": 13412 + }, + { + "epoch": 0.43, + "learning_rate": 1.2751394796575576e-05, + "loss": 0.9316, + "step": 13413 + }, + { + "epoch": 0.43, + "learning_rate": 1.2750398955655694e-05, + "loss": 1.085, + "step": 13414 + }, + { + "epoch": 0.43, + "learning_rate": 1.274940308522708e-05, + "loss": 1.0396, + "step": 13415 + }, + { + "epoch": 0.43, + "learning_rate": 1.2748407185300406e-05, + "loss": 0.9521, + "step": 13416 + }, + { + "epoch": 0.43, + "learning_rate": 1.2747411255886369e-05, + "loss": 1.0898, + "step": 13417 + }, + { + "epoch": 0.43, + "learning_rate": 1.2746415296995647e-05, + "loss": 1.0898, + "step": 13418 + }, + { + "epoch": 0.43, + "learning_rate": 1.2745419308638933e-05, + "loss": 0.9736, + "step": 13419 + }, + { + "epoch": 0.43, + "learning_rate": 1.2744423290826906e-05, + "loss": 1.0337, + "step": 13420 + }, + { + "epoch": 0.43, + "learning_rate": 1.2743427243570253e-05, + "loss": 1.0918, + "step": 13421 + }, + { + "epoch": 0.43, + "learning_rate": 1.2742431166879663e-05, + "loss": 1.043, + "step": 13422 + }, + { + "epoch": 0.43, + "learning_rate": 1.2741435060765824e-05, + "loss": 1.0552, + "step": 13423 + }, + { + "epoch": 0.43, + "learning_rate": 1.2740438925239418e-05, + "loss": 0.9414, + "step": 13424 + }, + { + "epoch": 0.43, + "learning_rate": 1.2739442760311134e-05, + "loss": 1.0273, + "step": 13425 + }, + { + "epoch": 0.43, + "learning_rate": 1.2738446565991663e-05, + "loss": 1.0054, + "step": 13426 + }, + { + "epoch": 0.43, + "learning_rate": 1.2737450342291694e-05, + "loss": 1.0117, + "step": 13427 + }, + { + "epoch": 0.43, + "learning_rate": 1.2736454089221907e-05, + "loss": 1.1045, + "step": 13428 + }, + { + "epoch": 0.43, + "learning_rate": 1.2735457806793e-05, + "loss": 0.9966, + "step": 13429 + }, + { + "epoch": 0.43, + "learning_rate": 1.2734461495015658e-05, + "loss": 0.9248, + "step": 13430 + }, + { + "epoch": 0.43, + "learning_rate": 1.2733465153900569e-05, + "loss": 0.5327, + "step": 13431 + }, + { + "epoch": 0.43, + "learning_rate": 1.2732468783458423e-05, + "loss": 0.8975, + "step": 13432 + }, + { + "epoch": 0.43, + "learning_rate": 1.2731472383699913e-05, + "loss": 1.0122, + "step": 13433 + }, + { + "epoch": 0.43, + "learning_rate": 1.2730475954635728e-05, + "loss": 0.5347, + "step": 13434 + }, + { + "epoch": 0.43, + "learning_rate": 1.2729479496276558e-05, + "loss": 0.9338, + "step": 13435 + }, + { + "epoch": 0.43, + "learning_rate": 1.2728483008633092e-05, + "loss": 1.0547, + "step": 13436 + }, + { + "epoch": 0.43, + "learning_rate": 1.2727486491716024e-05, + "loss": 0.8628, + "step": 13437 + }, + { + "epoch": 0.43, + "learning_rate": 1.2726489945536044e-05, + "loss": 1.0645, + "step": 13438 + }, + { + "epoch": 0.43, + "learning_rate": 1.2725493370103844e-05, + "loss": 1.0815, + "step": 13439 + }, + { + "epoch": 0.43, + "learning_rate": 1.2724496765430119e-05, + "loss": 0.9644, + "step": 13440 + }, + { + "epoch": 0.43, + "learning_rate": 1.2723500131525558e-05, + "loss": 1.0254, + "step": 13441 + }, + { + "epoch": 0.43, + "learning_rate": 1.2722503468400855e-05, + "loss": 1.0444, + "step": 13442 + }, + { + "epoch": 0.43, + "learning_rate": 1.27215067760667e-05, + "loss": 0.9395, + "step": 13443 + }, + { + "epoch": 0.43, + "learning_rate": 1.2720510054533792e-05, + "loss": 0.9468, + "step": 13444 + }, + { + "epoch": 0.43, + "learning_rate": 1.2719513303812825e-05, + "loss": 0.9307, + "step": 13445 + }, + { + "epoch": 0.43, + "learning_rate": 1.2718516523914487e-05, + "loss": 0.8892, + "step": 13446 + }, + { + "epoch": 0.43, + "learning_rate": 1.2717519714849473e-05, + "loss": 0.959, + "step": 13447 + }, + { + "epoch": 0.43, + "learning_rate": 1.2716522876628483e-05, + "loss": 0.9341, + "step": 13448 + }, + { + "epoch": 0.43, + "learning_rate": 1.2715526009262209e-05, + "loss": 1.041, + "step": 13449 + }, + { + "epoch": 0.43, + "learning_rate": 1.2714529112761345e-05, + "loss": 1.0381, + "step": 13450 + }, + { + "epoch": 0.43, + "learning_rate": 1.2713532187136589e-05, + "loss": 0.936, + "step": 13451 + }, + { + "epoch": 0.43, + "learning_rate": 1.2712535232398635e-05, + "loss": 0.936, + "step": 13452 + }, + { + "epoch": 0.43, + "learning_rate": 1.2711538248558182e-05, + "loss": 0.5029, + "step": 13453 + }, + { + "epoch": 0.43, + "learning_rate": 1.2710541235625921e-05, + "loss": 1.0215, + "step": 13454 + }, + { + "epoch": 0.43, + "learning_rate": 1.2709544193612554e-05, + "loss": 1.0449, + "step": 13455 + }, + { + "epoch": 0.43, + "learning_rate": 1.2708547122528778e-05, + "loss": 1.0337, + "step": 13456 + }, + { + "epoch": 0.43, + "learning_rate": 1.2707550022385288e-05, + "loss": 0.9438, + "step": 13457 + }, + { + "epoch": 0.43, + "learning_rate": 1.2706552893192784e-05, + "loss": 0.9575, + "step": 13458 + }, + { + "epoch": 0.43, + "learning_rate": 1.2705555734961961e-05, + "loss": 0.5388, + "step": 13459 + }, + { + "epoch": 0.43, + "learning_rate": 1.2704558547703518e-05, + "loss": 1.0059, + "step": 13460 + }, + { + "epoch": 0.43, + "learning_rate": 1.2703561331428161e-05, + "loss": 0.9902, + "step": 13461 + }, + { + "epoch": 0.43, + "learning_rate": 1.2702564086146575e-05, + "loss": 0.9277, + "step": 13462 + }, + { + "epoch": 0.43, + "learning_rate": 1.2701566811869473e-05, + "loss": 1.0352, + "step": 13463 + }, + { + "epoch": 0.43, + "learning_rate": 1.270056950860755e-05, + "loss": 1.0518, + "step": 13464 + }, + { + "epoch": 0.43, + "learning_rate": 1.26995721763715e-05, + "loss": 0.958, + "step": 13465 + }, + { + "epoch": 0.43, + "learning_rate": 1.2698574815172031e-05, + "loss": 1.0537, + "step": 13466 + }, + { + "epoch": 0.43, + "learning_rate": 1.269757742501984e-05, + "loss": 0.9561, + "step": 13467 + }, + { + "epoch": 0.43, + "learning_rate": 1.2696580005925632e-05, + "loss": 0.9814, + "step": 13468 + }, + { + "epoch": 0.43, + "learning_rate": 1.26955825579001e-05, + "loss": 1.0239, + "step": 13469 + }, + { + "epoch": 0.43, + "learning_rate": 1.2694585080953954e-05, + "loss": 1.0991, + "step": 13470 + }, + { + "epoch": 0.43, + "learning_rate": 1.269358757509789e-05, + "loss": 1.0273, + "step": 13471 + }, + { + "epoch": 0.43, + "learning_rate": 1.2692590040342616e-05, + "loss": 0.9512, + "step": 13472 + }, + { + "epoch": 0.43, + "learning_rate": 1.2691592476698828e-05, + "loss": 0.9648, + "step": 13473 + }, + { + "epoch": 0.43, + "learning_rate": 1.2690594884177231e-05, + "loss": 1.0171, + "step": 13474 + }, + { + "epoch": 0.43, + "learning_rate": 1.268959726278853e-05, + "loss": 1.0791, + "step": 13475 + }, + { + "epoch": 0.43, + "learning_rate": 1.2688599612543429e-05, + "loss": 1.022, + "step": 13476 + }, + { + "epoch": 0.43, + "learning_rate": 1.2687601933452627e-05, + "loss": 0.9971, + "step": 13477 + }, + { + "epoch": 0.43, + "learning_rate": 1.2686604225526831e-05, + "loss": 0.9014, + "step": 13478 + }, + { + "epoch": 0.43, + "learning_rate": 1.2685606488776748e-05, + "loss": 1.0303, + "step": 13479 + }, + { + "epoch": 0.43, + "learning_rate": 1.2684608723213074e-05, + "loss": 0.9688, + "step": 13480 + }, + { + "epoch": 0.43, + "learning_rate": 1.2683610928846524e-05, + "loss": 1.0083, + "step": 13481 + }, + { + "epoch": 0.43, + "learning_rate": 1.2682613105687796e-05, + "loss": 1.0142, + "step": 13482 + }, + { + "epoch": 0.43, + "learning_rate": 1.2681615253747601e-05, + "loss": 1.0273, + "step": 13483 + }, + { + "epoch": 0.43, + "learning_rate": 1.268061737303664e-05, + "loss": 1.0015, + "step": 13484 + }, + { + "epoch": 0.43, + "learning_rate": 1.2679619463565623e-05, + "loss": 1.0591, + "step": 13485 + }, + { + "epoch": 0.43, + "learning_rate": 1.2678621525345253e-05, + "loss": 0.9629, + "step": 13486 + }, + { + "epoch": 0.43, + "learning_rate": 1.2677623558386243e-05, + "loss": 0.939, + "step": 13487 + }, + { + "epoch": 0.43, + "learning_rate": 1.2676625562699291e-05, + "loss": 1.0254, + "step": 13488 + }, + { + "epoch": 0.43, + "learning_rate": 1.267562753829511e-05, + "loss": 0.9819, + "step": 13489 + }, + { + "epoch": 0.43, + "learning_rate": 1.2674629485184409e-05, + "loss": 1.0371, + "step": 13490 + }, + { + "epoch": 0.43, + "learning_rate": 1.267363140337789e-05, + "loss": 1.0264, + "step": 13491 + }, + { + "epoch": 0.43, + "learning_rate": 1.2672633292886268e-05, + "loss": 0.9497, + "step": 13492 + }, + { + "epoch": 0.43, + "learning_rate": 1.2671635153720249e-05, + "loss": 0.9878, + "step": 13493 + }, + { + "epoch": 0.43, + "learning_rate": 1.2670636985890542e-05, + "loss": 1.0195, + "step": 13494 + }, + { + "epoch": 0.43, + "learning_rate": 1.2669638789407855e-05, + "loss": 0.9609, + "step": 13495 + }, + { + "epoch": 0.43, + "learning_rate": 1.2668640564282897e-05, + "loss": 0.9771, + "step": 13496 + }, + { + "epoch": 0.43, + "learning_rate": 1.2667642310526378e-05, + "loss": 1.0078, + "step": 13497 + }, + { + "epoch": 0.43, + "learning_rate": 1.2666644028149013e-05, + "loss": 1.0576, + "step": 13498 + }, + { + "epoch": 0.43, + "learning_rate": 1.2665645717161508e-05, + "loss": 1.0454, + "step": 13499 + }, + { + "epoch": 0.43, + "learning_rate": 1.2664647377574573e-05, + "loss": 1.0771, + "step": 13500 + }, + { + "epoch": 0.43, + "learning_rate": 1.2663649009398919e-05, + "loss": 0.9683, + "step": 13501 + }, + { + "epoch": 0.43, + "learning_rate": 1.2662650612645265e-05, + "loss": 1.0527, + "step": 13502 + }, + { + "epoch": 0.43, + "learning_rate": 1.2661652187324314e-05, + "loss": 0.9834, + "step": 13503 + }, + { + "epoch": 0.43, + "learning_rate": 1.2660653733446776e-05, + "loss": 0.9468, + "step": 13504 + }, + { + "epoch": 0.43, + "learning_rate": 1.2659655251023374e-05, + "loss": 0.9688, + "step": 13505 + }, + { + "epoch": 0.43, + "learning_rate": 1.2658656740064815e-05, + "loss": 0.9692, + "step": 13506 + }, + { + "epoch": 0.43, + "learning_rate": 1.265765820058181e-05, + "loss": 0.9248, + "step": 13507 + }, + { + "epoch": 0.43, + "learning_rate": 1.265665963258507e-05, + "loss": 1.0083, + "step": 13508 + }, + { + "epoch": 0.43, + "learning_rate": 1.2655661036085318e-05, + "loss": 0.978, + "step": 13509 + }, + { + "epoch": 0.43, + "learning_rate": 1.2654662411093262e-05, + "loss": 0.9775, + "step": 13510 + }, + { + "epoch": 0.43, + "learning_rate": 1.2653663757619615e-05, + "loss": 0.8838, + "step": 13511 + }, + { + "epoch": 0.43, + "learning_rate": 1.265266507567509e-05, + "loss": 0.999, + "step": 13512 + }, + { + "epoch": 0.43, + "learning_rate": 1.2651666365270407e-05, + "loss": 1.0684, + "step": 13513 + }, + { + "epoch": 0.43, + "learning_rate": 1.265066762641628e-05, + "loss": 0.7832, + "step": 13514 + }, + { + "epoch": 0.43, + "learning_rate": 1.264966885912342e-05, + "loss": 1.0537, + "step": 13515 + }, + { + "epoch": 0.43, + "learning_rate": 1.2648670063402548e-05, + "loss": 0.9302, + "step": 13516 + }, + { + "epoch": 0.43, + "learning_rate": 1.2647671239264378e-05, + "loss": 1.0439, + "step": 13517 + }, + { + "epoch": 0.43, + "learning_rate": 1.2646672386719625e-05, + "loss": 1.0117, + "step": 13518 + }, + { + "epoch": 0.43, + "learning_rate": 1.2645673505779007e-05, + "loss": 0.937, + "step": 13519 + }, + { + "epoch": 0.43, + "learning_rate": 1.2644674596453237e-05, + "loss": 1.0894, + "step": 13520 + }, + { + "epoch": 0.43, + "learning_rate": 1.2643675658753041e-05, + "loss": 1.0518, + "step": 13521 + }, + { + "epoch": 0.43, + "learning_rate": 1.2642676692689127e-05, + "loss": 1.0659, + "step": 13522 + }, + { + "epoch": 0.43, + "learning_rate": 1.2641677698272217e-05, + "loss": 0.5957, + "step": 13523 + }, + { + "epoch": 0.43, + "learning_rate": 1.264067867551303e-05, + "loss": 1.0952, + "step": 13524 + }, + { + "epoch": 0.43, + "learning_rate": 1.2639679624422283e-05, + "loss": 1.125, + "step": 13525 + }, + { + "epoch": 0.43, + "learning_rate": 1.2638680545010695e-05, + "loss": 0.9746, + "step": 13526 + }, + { + "epoch": 0.43, + "learning_rate": 1.2637681437288986e-05, + "loss": 1.0742, + "step": 13527 + }, + { + "epoch": 0.43, + "learning_rate": 1.2636682301267872e-05, + "loss": 0.9702, + "step": 13528 + }, + { + "epoch": 0.43, + "learning_rate": 1.2635683136958079e-05, + "loss": 0.8799, + "step": 13529 + }, + { + "epoch": 0.43, + "learning_rate": 1.263468394437032e-05, + "loss": 1.0474, + "step": 13530 + }, + { + "epoch": 0.43, + "learning_rate": 1.263368472351532e-05, + "loss": 1.062, + "step": 13531 + }, + { + "epoch": 0.43, + "learning_rate": 1.2632685474403798e-05, + "loss": 1.0542, + "step": 13532 + }, + { + "epoch": 0.43, + "learning_rate": 1.2631686197046474e-05, + "loss": 0.5945, + "step": 13533 + }, + { + "epoch": 0.43, + "learning_rate": 1.2630686891454072e-05, + "loss": 1.1396, + "step": 13534 + }, + { + "epoch": 0.43, + "learning_rate": 1.2629687557637307e-05, + "loss": 0.9893, + "step": 13535 + }, + { + "epoch": 0.43, + "learning_rate": 1.2628688195606906e-05, + "loss": 1.0752, + "step": 13536 + }, + { + "epoch": 0.43, + "learning_rate": 1.262768880537359e-05, + "loss": 0.9746, + "step": 13537 + }, + { + "epoch": 0.43, + "learning_rate": 1.2626689386948081e-05, + "loss": 0.9927, + "step": 13538 + }, + { + "epoch": 0.43, + "learning_rate": 1.2625689940341102e-05, + "loss": 1.0571, + "step": 13539 + }, + { + "epoch": 0.43, + "learning_rate": 1.2624690465563378e-05, + "loss": 0.8081, + "step": 13540 + }, + { + "epoch": 0.43, + "learning_rate": 1.262369096262563e-05, + "loss": 1.0107, + "step": 13541 + }, + { + "epoch": 0.43, + "learning_rate": 1.2622691431538578e-05, + "loss": 0.9839, + "step": 13542 + }, + { + "epoch": 0.43, + "learning_rate": 1.262169187231295e-05, + "loss": 0.9429, + "step": 13543 + }, + { + "epoch": 0.43, + "learning_rate": 1.2620692284959475e-05, + "loss": 0.998, + "step": 13544 + }, + { + "epoch": 0.43, + "learning_rate": 1.2619692669488868e-05, + "loss": 1.1367, + "step": 13545 + }, + { + "epoch": 0.43, + "learning_rate": 1.2618693025911858e-05, + "loss": 1.021, + "step": 13546 + }, + { + "epoch": 0.43, + "learning_rate": 1.2617693354239174e-05, + "loss": 1.0791, + "step": 13547 + }, + { + "epoch": 0.43, + "learning_rate": 1.2616693654481533e-05, + "loss": 0.998, + "step": 13548 + }, + { + "epoch": 0.43, + "learning_rate": 1.2615693926649666e-05, + "loss": 0.9927, + "step": 13549 + }, + { + "epoch": 0.43, + "learning_rate": 1.2614694170754298e-05, + "loss": 1.0142, + "step": 13550 + }, + { + "epoch": 0.43, + "learning_rate": 1.2613694386806157e-05, + "loss": 1.0723, + "step": 13551 + }, + { + "epoch": 0.43, + "learning_rate": 1.2612694574815968e-05, + "loss": 1.0767, + "step": 13552 + }, + { + "epoch": 0.43, + "learning_rate": 1.2611694734794455e-05, + "loss": 1.1045, + "step": 13553 + }, + { + "epoch": 0.43, + "learning_rate": 1.2610694866752348e-05, + "loss": 0.9502, + "step": 13554 + }, + { + "epoch": 0.43, + "learning_rate": 1.2609694970700378e-05, + "loss": 1.1216, + "step": 13555 + }, + { + "epoch": 0.43, + "learning_rate": 1.2608695046649267e-05, + "loss": 0.918, + "step": 13556 + }, + { + "epoch": 0.43, + "learning_rate": 1.2607695094609744e-05, + "loss": 1.0342, + "step": 13557 + }, + { + "epoch": 0.43, + "learning_rate": 1.260669511459254e-05, + "loss": 0.5176, + "step": 13558 + }, + { + "epoch": 0.43, + "learning_rate": 1.2605695106608385e-05, + "loss": 1.0366, + "step": 13559 + }, + { + "epoch": 0.43, + "learning_rate": 1.2604695070668002e-05, + "loss": 0.9849, + "step": 13560 + }, + { + "epoch": 0.43, + "learning_rate": 1.2603695006782124e-05, + "loss": 1.0718, + "step": 13561 + }, + { + "epoch": 0.43, + "learning_rate": 1.260269491496148e-05, + "loss": 1.1392, + "step": 13562 + }, + { + "epoch": 0.43, + "learning_rate": 1.2601694795216801e-05, + "loss": 0.9248, + "step": 13563 + }, + { + "epoch": 0.43, + "learning_rate": 1.2600694647558817e-05, + "loss": 0.9487, + "step": 13564 + }, + { + "epoch": 0.43, + "learning_rate": 1.2599694471998257e-05, + "loss": 1.0303, + "step": 13565 + }, + { + "epoch": 0.43, + "learning_rate": 1.2598694268545852e-05, + "loss": 0.9126, + "step": 13566 + }, + { + "epoch": 0.43, + "learning_rate": 1.2597694037212334e-05, + "loss": 1.021, + "step": 13567 + }, + { + "epoch": 0.43, + "learning_rate": 1.2596693778008432e-05, + "loss": 1.0459, + "step": 13568 + }, + { + "epoch": 0.43, + "learning_rate": 1.2595693490944883e-05, + "loss": 0.8843, + "step": 13569 + }, + { + "epoch": 0.43, + "learning_rate": 1.2594693176032416e-05, + "loss": 0.9414, + "step": 13570 + }, + { + "epoch": 0.43, + "learning_rate": 1.2593692833281761e-05, + "loss": 1.0493, + "step": 13571 + }, + { + "epoch": 0.43, + "learning_rate": 1.259269246270365e-05, + "loss": 0.9683, + "step": 13572 + }, + { + "epoch": 0.43, + "learning_rate": 1.2591692064308823e-05, + "loss": 0.96, + "step": 13573 + }, + { + "epoch": 0.43, + "learning_rate": 1.2590691638108009e-05, + "loss": 1.1924, + "step": 13574 + }, + { + "epoch": 0.43, + "learning_rate": 1.2589691184111935e-05, + "loss": 0.936, + "step": 13575 + }, + { + "epoch": 0.43, + "learning_rate": 1.2588690702331346e-05, + "loss": 0.9385, + "step": 13576 + }, + { + "epoch": 0.43, + "learning_rate": 1.258769019277697e-05, + "loss": 0.9956, + "step": 13577 + }, + { + "epoch": 0.43, + "learning_rate": 1.2586689655459539e-05, + "loss": 1.1411, + "step": 13578 + }, + { + "epoch": 0.43, + "learning_rate": 1.2585689090389795e-05, + "loss": 0.9702, + "step": 13579 + }, + { + "epoch": 0.43, + "learning_rate": 1.2584688497578465e-05, + "loss": 1.0596, + "step": 13580 + }, + { + "epoch": 0.43, + "learning_rate": 1.258368787703629e-05, + "loss": 0.8457, + "step": 13581 + }, + { + "epoch": 0.43, + "learning_rate": 1.2582687228774004e-05, + "loss": 0.9409, + "step": 13582 + }, + { + "epoch": 0.43, + "learning_rate": 1.2581686552802341e-05, + "loss": 1.0835, + "step": 13583 + }, + { + "epoch": 0.43, + "learning_rate": 1.258068584913204e-05, + "loss": 1.1035, + "step": 13584 + }, + { + "epoch": 0.43, + "learning_rate": 1.2579685117773834e-05, + "loss": 1.1035, + "step": 13585 + }, + { + "epoch": 0.43, + "learning_rate": 1.2578684358738462e-05, + "loss": 0.9434, + "step": 13586 + }, + { + "epoch": 0.43, + "learning_rate": 1.2577683572036664e-05, + "loss": 1.0664, + "step": 13587 + }, + { + "epoch": 0.43, + "learning_rate": 1.2576682757679169e-05, + "loss": 0.96, + "step": 13588 + }, + { + "epoch": 0.43, + "learning_rate": 1.2575681915676725e-05, + "loss": 1.0498, + "step": 13589 + }, + { + "epoch": 0.43, + "learning_rate": 1.257468104604006e-05, + "loss": 0.8872, + "step": 13590 + }, + { + "epoch": 0.43, + "learning_rate": 1.2573680148779919e-05, + "loss": 1.002, + "step": 13591 + }, + { + "epoch": 0.43, + "learning_rate": 1.2572679223907039e-05, + "loss": 0.9688, + "step": 13592 + }, + { + "epoch": 0.43, + "learning_rate": 1.2571678271432157e-05, + "loss": 1.0645, + "step": 13593 + }, + { + "epoch": 0.43, + "learning_rate": 1.2570677291366015e-05, + "loss": 0.9873, + "step": 13594 + }, + { + "epoch": 0.43, + "learning_rate": 1.256967628371935e-05, + "loss": 0.9438, + "step": 13595 + }, + { + "epoch": 0.43, + "learning_rate": 1.25686752485029e-05, + "loss": 0.9648, + "step": 13596 + }, + { + "epoch": 0.43, + "learning_rate": 1.2567674185727412e-05, + "loss": 0.9717, + "step": 13597 + }, + { + "epoch": 0.43, + "learning_rate": 1.2566673095403619e-05, + "loss": 1.1733, + "step": 13598 + }, + { + "epoch": 0.43, + "learning_rate": 1.2565671977542264e-05, + "loss": 1.0205, + "step": 13599 + }, + { + "epoch": 0.43, + "learning_rate": 1.2564670832154092e-05, + "loss": 1.0713, + "step": 13600 + }, + { + "epoch": 0.43, + "learning_rate": 1.2563669659249836e-05, + "loss": 0.9639, + "step": 13601 + }, + { + "epoch": 0.43, + "learning_rate": 1.2562668458840241e-05, + "loss": 0.9663, + "step": 13602 + }, + { + "epoch": 0.44, + "learning_rate": 1.2561667230936052e-05, + "loss": 1.1528, + "step": 13603 + }, + { + "epoch": 0.44, + "learning_rate": 1.256066597554801e-05, + "loss": 0.9878, + "step": 13604 + }, + { + "epoch": 0.44, + "learning_rate": 1.2559664692686855e-05, + "loss": 1.0127, + "step": 13605 + }, + { + "epoch": 0.44, + "learning_rate": 1.255866338236333e-05, + "loss": 1.0518, + "step": 13606 + }, + { + "epoch": 0.44, + "learning_rate": 1.2557662044588176e-05, + "loss": 0.9595, + "step": 13607 + }, + { + "epoch": 0.44, + "learning_rate": 1.2556660679372145e-05, + "loss": 0.9473, + "step": 13608 + }, + { + "epoch": 0.44, + "learning_rate": 1.2555659286725972e-05, + "loss": 0.5808, + "step": 13609 + }, + { + "epoch": 0.44, + "learning_rate": 1.2554657866660402e-05, + "loss": 1.0146, + "step": 13610 + }, + { + "epoch": 0.44, + "learning_rate": 1.255365641918618e-05, + "loss": 0.9219, + "step": 13611 + }, + { + "epoch": 0.44, + "learning_rate": 1.2552654944314053e-05, + "loss": 1.1167, + "step": 13612 + }, + { + "epoch": 0.44, + "learning_rate": 1.2551653442054762e-05, + "loss": 0.9429, + "step": 13613 + }, + { + "epoch": 0.44, + "learning_rate": 1.2550651912419057e-05, + "loss": 1.0278, + "step": 13614 + }, + { + "epoch": 0.44, + "learning_rate": 1.2549650355417677e-05, + "loss": 0.9609, + "step": 13615 + }, + { + "epoch": 0.44, + "learning_rate": 1.2548648771061371e-05, + "loss": 0.8901, + "step": 13616 + }, + { + "epoch": 0.44, + "learning_rate": 1.2547647159360882e-05, + "loss": 1.1265, + "step": 13617 + }, + { + "epoch": 0.44, + "learning_rate": 1.2546645520326963e-05, + "loss": 1.0737, + "step": 13618 + }, + { + "epoch": 0.44, + "learning_rate": 1.2545643853970354e-05, + "loss": 0.6118, + "step": 13619 + }, + { + "epoch": 0.44, + "learning_rate": 1.2544642160301804e-05, + "loss": 1.1533, + "step": 13620 + }, + { + "epoch": 0.44, + "learning_rate": 1.2543640439332058e-05, + "loss": 1.1416, + "step": 13621 + }, + { + "epoch": 0.44, + "learning_rate": 1.2542638691071867e-05, + "loss": 1.0679, + "step": 13622 + }, + { + "epoch": 0.44, + "learning_rate": 1.254163691553198e-05, + "loss": 0.5188, + "step": 13623 + }, + { + "epoch": 0.44, + "learning_rate": 1.2540635112723138e-05, + "loss": 0.7998, + "step": 13624 + }, + { + "epoch": 0.44, + "learning_rate": 1.2539633282656092e-05, + "loss": 0.9932, + "step": 13625 + }, + { + "epoch": 0.44, + "learning_rate": 1.2538631425341593e-05, + "loss": 0.9751, + "step": 13626 + }, + { + "epoch": 0.44, + "learning_rate": 1.253762954079039e-05, + "loss": 0.9043, + "step": 13627 + }, + { + "epoch": 0.44, + "learning_rate": 1.253662762901323e-05, + "loss": 1.0347, + "step": 13628 + }, + { + "epoch": 0.44, + "learning_rate": 1.2535625690020861e-05, + "loss": 1.0518, + "step": 13629 + }, + { + "epoch": 0.44, + "learning_rate": 1.2534623723824036e-05, + "loss": 1.062, + "step": 13630 + }, + { + "epoch": 0.44, + "learning_rate": 1.2533621730433503e-05, + "loss": 1.0312, + "step": 13631 + }, + { + "epoch": 0.44, + "learning_rate": 1.2532619709860012e-05, + "loss": 1.042, + "step": 13632 + }, + { + "epoch": 0.44, + "learning_rate": 1.2531617662114317e-05, + "loss": 0.9502, + "step": 13633 + }, + { + "epoch": 0.44, + "learning_rate": 1.2530615587207167e-05, + "loss": 1.0176, + "step": 13634 + }, + { + "epoch": 0.44, + "learning_rate": 1.2529613485149308e-05, + "loss": 0.981, + "step": 13635 + }, + { + "epoch": 0.44, + "learning_rate": 1.2528611355951499e-05, + "loss": 1.0129, + "step": 13636 + }, + { + "epoch": 0.44, + "learning_rate": 1.2527609199624488e-05, + "loss": 1.1128, + "step": 13637 + }, + { + "epoch": 0.44, + "learning_rate": 1.252660701617903e-05, + "loss": 1.0586, + "step": 13638 + }, + { + "epoch": 0.44, + "learning_rate": 1.252560480562587e-05, + "loss": 0.9146, + "step": 13639 + }, + { + "epoch": 0.44, + "learning_rate": 1.252460256797577e-05, + "loss": 0.9746, + "step": 13640 + }, + { + "epoch": 0.44, + "learning_rate": 1.2523600303239475e-05, + "loss": 0.96, + "step": 13641 + }, + { + "epoch": 0.44, + "learning_rate": 1.2522598011427747e-05, + "loss": 1.0562, + "step": 13642 + }, + { + "epoch": 0.44, + "learning_rate": 1.252159569255133e-05, + "loss": 0.5474, + "step": 13643 + }, + { + "epoch": 0.44, + "learning_rate": 1.2520593346620982e-05, + "loss": 0.5386, + "step": 13644 + }, + { + "epoch": 0.44, + "learning_rate": 1.2519590973647458e-05, + "loss": 1.0195, + "step": 13645 + }, + { + "epoch": 0.44, + "learning_rate": 1.2518588573641513e-05, + "loss": 1.1514, + "step": 13646 + }, + { + "epoch": 0.44, + "learning_rate": 1.2517586146613899e-05, + "loss": 0.9517, + "step": 13647 + }, + { + "epoch": 0.44, + "learning_rate": 1.2516583692575371e-05, + "loss": 0.9189, + "step": 13648 + }, + { + "epoch": 0.44, + "learning_rate": 1.2515581211536687e-05, + "loss": 1.1118, + "step": 13649 + }, + { + "epoch": 0.44, + "learning_rate": 1.2514578703508598e-05, + "loss": 0.9976, + "step": 13650 + }, + { + "epoch": 0.44, + "learning_rate": 1.2513576168501863e-05, + "loss": 1.0703, + "step": 13651 + }, + { + "epoch": 0.44, + "learning_rate": 1.2512573606527238e-05, + "loss": 1.0708, + "step": 13652 + }, + { + "epoch": 0.44, + "learning_rate": 1.2511571017595481e-05, + "loss": 1.0112, + "step": 13653 + }, + { + "epoch": 0.44, + "learning_rate": 1.2510568401717345e-05, + "loss": 0.9648, + "step": 13654 + }, + { + "epoch": 0.44, + "learning_rate": 1.2509565758903588e-05, + "loss": 1.0259, + "step": 13655 + }, + { + "epoch": 0.44, + "learning_rate": 1.2508563089164968e-05, + "loss": 1.0474, + "step": 13656 + }, + { + "epoch": 0.44, + "learning_rate": 1.2507560392512246e-05, + "loss": 1.0708, + "step": 13657 + }, + { + "epoch": 0.44, + "learning_rate": 1.250655766895617e-05, + "loss": 0.979, + "step": 13658 + }, + { + "epoch": 0.44, + "learning_rate": 1.2505554918507507e-05, + "loss": 0.981, + "step": 13659 + }, + { + "epoch": 0.44, + "learning_rate": 1.2504552141177015e-05, + "loss": 0.9888, + "step": 13660 + }, + { + "epoch": 0.44, + "learning_rate": 1.2503549336975448e-05, + "loss": 1.0688, + "step": 13661 + }, + { + "epoch": 0.44, + "learning_rate": 1.250254650591357e-05, + "loss": 0.9927, + "step": 13662 + }, + { + "epoch": 0.44, + "learning_rate": 1.2501543648002134e-05, + "loss": 0.8613, + "step": 13663 + }, + { + "epoch": 0.44, + "learning_rate": 1.2500540763251906e-05, + "loss": 1.1152, + "step": 13664 + }, + { + "epoch": 0.44, + "learning_rate": 1.2499537851673645e-05, + "loss": 0.9966, + "step": 13665 + }, + { + "epoch": 0.44, + "learning_rate": 1.2498534913278104e-05, + "loss": 1.0654, + "step": 13666 + }, + { + "epoch": 0.44, + "learning_rate": 1.2497531948076051e-05, + "loss": 1.0317, + "step": 13667 + }, + { + "epoch": 0.44, + "learning_rate": 1.2496528956078247e-05, + "loss": 1.0518, + "step": 13668 + }, + { + "epoch": 0.44, + "learning_rate": 1.2495525937295446e-05, + "loss": 1.0298, + "step": 13669 + }, + { + "epoch": 0.44, + "learning_rate": 1.2494522891738415e-05, + "loss": 0.9368, + "step": 13670 + }, + { + "epoch": 0.44, + "learning_rate": 1.2493519819417915e-05, + "loss": 1.0449, + "step": 13671 + }, + { + "epoch": 0.44, + "learning_rate": 1.2492516720344706e-05, + "loss": 1.0454, + "step": 13672 + }, + { + "epoch": 0.44, + "learning_rate": 1.2491513594529552e-05, + "loss": 1.0762, + "step": 13673 + }, + { + "epoch": 0.44, + "learning_rate": 1.2490510441983212e-05, + "loss": 0.9556, + "step": 13674 + }, + { + "epoch": 0.44, + "learning_rate": 1.2489507262716454e-05, + "loss": 0.9468, + "step": 13675 + }, + { + "epoch": 0.44, + "learning_rate": 1.2488504056740039e-05, + "loss": 0.9565, + "step": 13676 + }, + { + "epoch": 0.44, + "learning_rate": 1.2487500824064726e-05, + "loss": 0.9907, + "step": 13677 + }, + { + "epoch": 0.44, + "learning_rate": 1.2486497564701286e-05, + "loss": 1.0342, + "step": 13678 + }, + { + "epoch": 0.44, + "learning_rate": 1.2485494278660476e-05, + "loss": 1.0791, + "step": 13679 + }, + { + "epoch": 0.44, + "learning_rate": 1.2484490965953066e-05, + "loss": 0.981, + "step": 13680 + }, + { + "epoch": 0.44, + "learning_rate": 1.2483487626589817e-05, + "loss": 1.061, + "step": 13681 + }, + { + "epoch": 0.44, + "learning_rate": 1.2482484260581493e-05, + "loss": 1.0962, + "step": 13682 + }, + { + "epoch": 0.44, + "learning_rate": 1.2481480867938865e-05, + "loss": 0.939, + "step": 13683 + }, + { + "epoch": 0.44, + "learning_rate": 1.2480477448672687e-05, + "loss": 0.9668, + "step": 13684 + }, + { + "epoch": 0.44, + "learning_rate": 1.2479474002793735e-05, + "loss": 1.0391, + "step": 13685 + }, + { + "epoch": 0.44, + "learning_rate": 1.247847053031277e-05, + "loss": 0.979, + "step": 13686 + }, + { + "epoch": 0.44, + "learning_rate": 1.247746703124056e-05, + "loss": 1.0469, + "step": 13687 + }, + { + "epoch": 0.44, + "learning_rate": 1.2476463505587868e-05, + "loss": 0.9741, + "step": 13688 + }, + { + "epoch": 0.44, + "learning_rate": 1.2475459953365466e-05, + "loss": 1.0654, + "step": 13689 + }, + { + "epoch": 0.44, + "learning_rate": 1.2474456374584116e-05, + "loss": 1.0415, + "step": 13690 + }, + { + "epoch": 0.44, + "learning_rate": 1.2473452769254593e-05, + "loss": 0.897, + "step": 13691 + }, + { + "epoch": 0.44, + "learning_rate": 1.2472449137387655e-05, + "loss": 0.9678, + "step": 13692 + }, + { + "epoch": 0.44, + "learning_rate": 1.2471445478994073e-05, + "loss": 0.8335, + "step": 13693 + }, + { + "epoch": 0.44, + "learning_rate": 1.2470441794084618e-05, + "loss": 1.1421, + "step": 13694 + }, + { + "epoch": 0.44, + "learning_rate": 1.246943808267006e-05, + "loss": 0.9951, + "step": 13695 + }, + { + "epoch": 0.44, + "learning_rate": 1.246843434476116e-05, + "loss": 1.0361, + "step": 13696 + }, + { + "epoch": 0.44, + "learning_rate": 1.2467430580368692e-05, + "loss": 0.853, + "step": 13697 + }, + { + "epoch": 0.44, + "learning_rate": 1.2466426789503428e-05, + "loss": 1.0059, + "step": 13698 + }, + { + "epoch": 0.44, + "learning_rate": 1.246542297217613e-05, + "loss": 1.0029, + "step": 13699 + }, + { + "epoch": 0.44, + "learning_rate": 1.2464419128397573e-05, + "loss": 1.0962, + "step": 13700 + }, + { + "epoch": 0.44, + "learning_rate": 1.2463415258178527e-05, + "loss": 0.98, + "step": 13701 + }, + { + "epoch": 0.44, + "learning_rate": 1.2462411361529762e-05, + "loss": 1.1069, + "step": 13702 + }, + { + "epoch": 0.44, + "learning_rate": 1.246140743846205e-05, + "loss": 1.0093, + "step": 13703 + }, + { + "epoch": 0.44, + "learning_rate": 1.2460403488986159e-05, + "loss": 0.9692, + "step": 13704 + }, + { + "epoch": 0.44, + "learning_rate": 1.245939951311286e-05, + "loss": 1.0293, + "step": 13705 + }, + { + "epoch": 0.44, + "learning_rate": 1.2458395510852928e-05, + "loss": 0.917, + "step": 13706 + }, + { + "epoch": 0.44, + "learning_rate": 1.2457391482217132e-05, + "loss": 0.9995, + "step": 13707 + }, + { + "epoch": 0.44, + "learning_rate": 1.2456387427216245e-05, + "loss": 0.9463, + "step": 13708 + }, + { + "epoch": 0.44, + "learning_rate": 1.2455383345861038e-05, + "loss": 0.9243, + "step": 13709 + }, + { + "epoch": 0.44, + "learning_rate": 1.2454379238162292e-05, + "loss": 1.146, + "step": 13710 + }, + { + "epoch": 0.44, + "learning_rate": 1.2453375104130764e-05, + "loss": 1.0166, + "step": 13711 + }, + { + "epoch": 0.44, + "learning_rate": 1.2452370943777241e-05, + "loss": 1.0674, + "step": 13712 + }, + { + "epoch": 0.44, + "learning_rate": 1.2451366757112491e-05, + "loss": 1.0308, + "step": 13713 + }, + { + "epoch": 0.44, + "learning_rate": 1.2450362544147292e-05, + "loss": 1.0415, + "step": 13714 + }, + { + "epoch": 0.44, + "learning_rate": 1.244935830489241e-05, + "loss": 0.9263, + "step": 13715 + }, + { + "epoch": 0.44, + "learning_rate": 1.2448354039358626e-05, + "loss": 0.9194, + "step": 13716 + }, + { + "epoch": 0.44, + "learning_rate": 1.2447349747556714e-05, + "loss": 0.8936, + "step": 13717 + }, + { + "epoch": 0.44, + "learning_rate": 1.2446345429497446e-05, + "loss": 0.9946, + "step": 13718 + }, + { + "epoch": 0.44, + "learning_rate": 1.24453410851916e-05, + "loss": 0.9731, + "step": 13719 + }, + { + "epoch": 0.44, + "learning_rate": 1.244433671464995e-05, + "loss": 1.0054, + "step": 13720 + }, + { + "epoch": 0.44, + "learning_rate": 1.2443332317883275e-05, + "loss": 1.0366, + "step": 13721 + }, + { + "epoch": 0.44, + "learning_rate": 1.2442327894902343e-05, + "loss": 0.9717, + "step": 13722 + }, + { + "epoch": 0.44, + "learning_rate": 1.244132344571794e-05, + "loss": 1.085, + "step": 13723 + }, + { + "epoch": 0.44, + "learning_rate": 1.2440318970340837e-05, + "loss": 0.9585, + "step": 13724 + }, + { + "epoch": 0.44, + "learning_rate": 1.2439314468781814e-05, + "loss": 1.043, + "step": 13725 + }, + { + "epoch": 0.44, + "learning_rate": 1.2438309941051643e-05, + "loss": 0.8613, + "step": 13726 + }, + { + "epoch": 0.44, + "learning_rate": 1.2437305387161107e-05, + "loss": 1.0508, + "step": 13727 + }, + { + "epoch": 0.44, + "learning_rate": 1.243630080712098e-05, + "loss": 1.1221, + "step": 13728 + }, + { + "epoch": 0.44, + "learning_rate": 1.2435296200942045e-05, + "loss": 1.0098, + "step": 13729 + }, + { + "epoch": 0.44, + "learning_rate": 1.2434291568635075e-05, + "loss": 0.9878, + "step": 13730 + }, + { + "epoch": 0.44, + "learning_rate": 1.243328691021085e-05, + "loss": 1.0122, + "step": 13731 + }, + { + "epoch": 0.44, + "learning_rate": 1.2432282225680152e-05, + "loss": 0.9668, + "step": 13732 + }, + { + "epoch": 0.44, + "learning_rate": 1.2431277515053755e-05, + "loss": 0.9229, + "step": 13733 + }, + { + "epoch": 0.44, + "learning_rate": 1.243027277834244e-05, + "loss": 1.0195, + "step": 13734 + }, + { + "epoch": 0.44, + "learning_rate": 1.242926801555699e-05, + "loss": 0.9263, + "step": 13735 + }, + { + "epoch": 0.44, + "learning_rate": 1.2428263226708183e-05, + "loss": 1.124, + "step": 13736 + }, + { + "epoch": 0.44, + "learning_rate": 1.24272584118068e-05, + "loss": 1.0796, + "step": 13737 + }, + { + "epoch": 0.44, + "learning_rate": 1.2426253570863618e-05, + "loss": 0.9634, + "step": 13738 + }, + { + "epoch": 0.44, + "learning_rate": 1.2425248703889421e-05, + "loss": 1.0688, + "step": 13739 + }, + { + "epoch": 0.44, + "learning_rate": 1.2424243810894991e-05, + "loss": 1.019, + "step": 13740 + }, + { + "epoch": 0.44, + "learning_rate": 1.2423238891891108e-05, + "loss": 0.9634, + "step": 13741 + }, + { + "epoch": 0.44, + "learning_rate": 1.2422233946888551e-05, + "loss": 0.9829, + "step": 13742 + }, + { + "epoch": 0.44, + "learning_rate": 1.2421228975898108e-05, + "loss": 0.9727, + "step": 13743 + }, + { + "epoch": 0.44, + "learning_rate": 1.2420223978930554e-05, + "loss": 1.0366, + "step": 13744 + }, + { + "epoch": 0.44, + "learning_rate": 1.2419218955996677e-05, + "loss": 1.0815, + "step": 13745 + }, + { + "epoch": 0.44, + "learning_rate": 1.2418213907107261e-05, + "loss": 1.021, + "step": 13746 + }, + { + "epoch": 0.44, + "learning_rate": 1.2417208832273082e-05, + "loss": 1.0024, + "step": 13747 + }, + { + "epoch": 0.44, + "learning_rate": 1.241620373150493e-05, + "loss": 0.8994, + "step": 13748 + }, + { + "epoch": 0.44, + "learning_rate": 1.2415198604813585e-05, + "loss": 1.0435, + "step": 13749 + }, + { + "epoch": 0.44, + "learning_rate": 1.2414193452209831e-05, + "loss": 0.9341, + "step": 13750 + }, + { + "epoch": 0.44, + "learning_rate": 1.2413188273704457e-05, + "loss": 1.0181, + "step": 13751 + }, + { + "epoch": 0.44, + "learning_rate": 1.2412183069308238e-05, + "loss": 1.0127, + "step": 13752 + }, + { + "epoch": 0.44, + "learning_rate": 1.2411177839031967e-05, + "loss": 1.0708, + "step": 13753 + }, + { + "epoch": 0.44, + "learning_rate": 1.2410172582886426e-05, + "loss": 0.9355, + "step": 13754 + }, + { + "epoch": 0.44, + "learning_rate": 1.2409167300882403e-05, + "loss": 1.0762, + "step": 13755 + }, + { + "epoch": 0.44, + "learning_rate": 1.240816199303068e-05, + "loss": 1.0601, + "step": 13756 + }, + { + "epoch": 0.44, + "learning_rate": 1.240715665934204e-05, + "loss": 0.9146, + "step": 13757 + }, + { + "epoch": 0.44, + "learning_rate": 1.2406151299827279e-05, + "loss": 0.9849, + "step": 13758 + }, + { + "epoch": 0.44, + "learning_rate": 1.2405145914497174e-05, + "loss": 1.0815, + "step": 13759 + }, + { + "epoch": 0.44, + "learning_rate": 1.2404140503362517e-05, + "loss": 0.9097, + "step": 13760 + }, + { + "epoch": 0.44, + "learning_rate": 1.240313506643409e-05, + "loss": 0.9268, + "step": 13761 + }, + { + "epoch": 0.44, + "learning_rate": 1.2402129603722686e-05, + "loss": 1.0425, + "step": 13762 + }, + { + "epoch": 0.44, + "learning_rate": 1.2401124115239091e-05, + "loss": 0.9043, + "step": 13763 + }, + { + "epoch": 0.44, + "learning_rate": 1.240011860099409e-05, + "loss": 0.8384, + "step": 13764 + }, + { + "epoch": 0.44, + "learning_rate": 1.2399113060998473e-05, + "loss": 1.0215, + "step": 13765 + }, + { + "epoch": 0.44, + "learning_rate": 1.239810749526303e-05, + "loss": 1.0137, + "step": 13766 + }, + { + "epoch": 0.44, + "learning_rate": 1.2397101903798545e-05, + "loss": 0.9346, + "step": 13767 + }, + { + "epoch": 0.44, + "learning_rate": 1.2396096286615809e-05, + "loss": 0.9619, + "step": 13768 + }, + { + "epoch": 0.44, + "learning_rate": 1.2395090643725614e-05, + "loss": 1.1416, + "step": 13769 + }, + { + "epoch": 0.44, + "learning_rate": 1.2394084975138747e-05, + "loss": 0.9429, + "step": 13770 + }, + { + "epoch": 0.44, + "learning_rate": 1.2393079280865996e-05, + "loss": 0.9795, + "step": 13771 + }, + { + "epoch": 0.44, + "learning_rate": 1.2392073560918154e-05, + "loss": 1.0103, + "step": 13772 + }, + { + "epoch": 0.44, + "learning_rate": 1.2391067815306009e-05, + "loss": 0.9907, + "step": 13773 + }, + { + "epoch": 0.44, + "learning_rate": 1.2390062044040354e-05, + "loss": 1.0498, + "step": 13774 + }, + { + "epoch": 0.44, + "learning_rate": 1.2389056247131979e-05, + "loss": 0.8965, + "step": 13775 + }, + { + "epoch": 0.44, + "learning_rate": 1.2388050424591673e-05, + "loss": 0.9263, + "step": 13776 + }, + { + "epoch": 0.44, + "learning_rate": 1.238704457643023e-05, + "loss": 1.0293, + "step": 13777 + }, + { + "epoch": 0.44, + "learning_rate": 1.238603870265844e-05, + "loss": 1.0083, + "step": 13778 + }, + { + "epoch": 0.44, + "learning_rate": 1.2385032803287096e-05, + "loss": 0.937, + "step": 13779 + }, + { + "epoch": 0.44, + "learning_rate": 1.2384026878326988e-05, + "loss": 1.0728, + "step": 13780 + }, + { + "epoch": 0.44, + "learning_rate": 1.2383020927788914e-05, + "loss": 1.1118, + "step": 13781 + }, + { + "epoch": 0.44, + "learning_rate": 1.2382014951683659e-05, + "loss": 1.0737, + "step": 13782 + }, + { + "epoch": 0.44, + "learning_rate": 1.238100895002202e-05, + "loss": 1.0029, + "step": 13783 + }, + { + "epoch": 0.44, + "learning_rate": 1.2380002922814792e-05, + "loss": 1.0366, + "step": 13784 + }, + { + "epoch": 0.44, + "learning_rate": 1.2378996870072765e-05, + "loss": 1.0186, + "step": 13785 + }, + { + "epoch": 0.44, + "learning_rate": 1.2377990791806736e-05, + "loss": 0.9248, + "step": 13786 + }, + { + "epoch": 0.44, + "learning_rate": 1.2376984688027495e-05, + "loss": 1.1343, + "step": 13787 + }, + { + "epoch": 0.44, + "learning_rate": 1.2375978558745842e-05, + "loss": 1.1592, + "step": 13788 + }, + { + "epoch": 0.44, + "learning_rate": 1.237497240397257e-05, + "loss": 1.0293, + "step": 13789 + }, + { + "epoch": 0.44, + "learning_rate": 1.237396622371847e-05, + "loss": 1.0117, + "step": 13790 + }, + { + "epoch": 0.44, + "learning_rate": 1.237296001799434e-05, + "loss": 0.8965, + "step": 13791 + }, + { + "epoch": 0.44, + "learning_rate": 1.2371953786810976e-05, + "loss": 0.9844, + "step": 13792 + }, + { + "epoch": 0.44, + "learning_rate": 1.2370947530179174e-05, + "loss": 1.019, + "step": 13793 + }, + { + "epoch": 0.44, + "learning_rate": 1.2369941248109725e-05, + "loss": 1.021, + "step": 13794 + }, + { + "epoch": 0.44, + "learning_rate": 1.2368934940613431e-05, + "loss": 0.8931, + "step": 13795 + }, + { + "epoch": 0.44, + "learning_rate": 1.2367928607701088e-05, + "loss": 1.0186, + "step": 13796 + }, + { + "epoch": 0.44, + "learning_rate": 1.2366922249383493e-05, + "loss": 0.8657, + "step": 13797 + }, + { + "epoch": 0.44, + "learning_rate": 1.2365915865671439e-05, + "loss": 1.0864, + "step": 13798 + }, + { + "epoch": 0.44, + "learning_rate": 1.2364909456575726e-05, + "loss": 1.1035, + "step": 13799 + }, + { + "epoch": 0.44, + "learning_rate": 1.2363903022107155e-05, + "loss": 0.9639, + "step": 13800 + }, + { + "epoch": 0.44, + "learning_rate": 1.236289656227652e-05, + "loss": 1.0752, + "step": 13801 + }, + { + "epoch": 0.44, + "learning_rate": 1.2361890077094618e-05, + "loss": 0.9951, + "step": 13802 + }, + { + "epoch": 0.44, + "learning_rate": 1.236088356657225e-05, + "loss": 0.8633, + "step": 13803 + }, + { + "epoch": 0.44, + "learning_rate": 1.2359877030720217e-05, + "loss": 1.1172, + "step": 13804 + }, + { + "epoch": 0.44, + "learning_rate": 1.2358870469549312e-05, + "loss": 0.5791, + "step": 13805 + }, + { + "epoch": 0.44, + "learning_rate": 1.2357863883070338e-05, + "loss": 1.0884, + "step": 13806 + }, + { + "epoch": 0.44, + "learning_rate": 1.2356857271294093e-05, + "loss": 1.0371, + "step": 13807 + }, + { + "epoch": 0.44, + "learning_rate": 1.2355850634231384e-05, + "loss": 1.0562, + "step": 13808 + }, + { + "epoch": 0.44, + "learning_rate": 1.2354843971892998e-05, + "loss": 1.04, + "step": 13809 + }, + { + "epoch": 0.44, + "learning_rate": 1.2353837284289744e-05, + "loss": 0.9946, + "step": 13810 + }, + { + "epoch": 0.44, + "learning_rate": 1.2352830571432421e-05, + "loss": 1.0479, + "step": 13811 + }, + { + "epoch": 0.44, + "learning_rate": 1.2351823833331832e-05, + "loss": 0.9448, + "step": 13812 + }, + { + "epoch": 0.44, + "learning_rate": 1.2350817069998778e-05, + "loss": 1.0083, + "step": 13813 + }, + { + "epoch": 0.44, + "learning_rate": 1.2349810281444055e-05, + "loss": 1.0244, + "step": 13814 + }, + { + "epoch": 0.44, + "learning_rate": 1.234880346767847e-05, + "loss": 0.8521, + "step": 13815 + }, + { + "epoch": 0.44, + "learning_rate": 1.234779662871282e-05, + "loss": 0.9854, + "step": 13816 + }, + { + "epoch": 0.44, + "learning_rate": 1.2346789764557913e-05, + "loss": 1.0586, + "step": 13817 + }, + { + "epoch": 0.44, + "learning_rate": 1.2345782875224549e-05, + "loss": 0.9619, + "step": 13818 + }, + { + "epoch": 0.44, + "learning_rate": 1.2344775960723531e-05, + "loss": 1.062, + "step": 13819 + }, + { + "epoch": 0.44, + "learning_rate": 1.2343769021065664e-05, + "loss": 0.9429, + "step": 13820 + }, + { + "epoch": 0.44, + "learning_rate": 1.2342762056261745e-05, + "loss": 0.9766, + "step": 13821 + }, + { + "epoch": 0.44, + "learning_rate": 1.2341755066322584e-05, + "loss": 0.9546, + "step": 13822 + }, + { + "epoch": 0.44, + "learning_rate": 1.2340748051258986e-05, + "loss": 1.082, + "step": 13823 + }, + { + "epoch": 0.44, + "learning_rate": 1.2339741011081748e-05, + "loss": 0.8887, + "step": 13824 + }, + { + "epoch": 0.44, + "learning_rate": 1.233873394580168e-05, + "loss": 0.9272, + "step": 13825 + }, + { + "epoch": 0.44, + "learning_rate": 1.2337726855429583e-05, + "loss": 0.998, + "step": 13826 + }, + { + "epoch": 0.44, + "learning_rate": 1.2336719739976269e-05, + "loss": 1.0, + "step": 13827 + }, + { + "epoch": 0.44, + "learning_rate": 1.2335712599452534e-05, + "loss": 1.0195, + "step": 13828 + }, + { + "epoch": 0.44, + "learning_rate": 1.2334705433869187e-05, + "loss": 0.9058, + "step": 13829 + }, + { + "epoch": 0.44, + "learning_rate": 1.233369824323704e-05, + "loss": 1.0723, + "step": 13830 + }, + { + "epoch": 0.44, + "learning_rate": 1.233269102756689e-05, + "loss": 1.0908, + "step": 13831 + }, + { + "epoch": 0.44, + "learning_rate": 1.2331683786869548e-05, + "loss": 1.0537, + "step": 13832 + }, + { + "epoch": 0.44, + "learning_rate": 1.2330676521155818e-05, + "loss": 1.0552, + "step": 13833 + }, + { + "epoch": 0.44, + "learning_rate": 1.232966923043651e-05, + "loss": 1.0396, + "step": 13834 + }, + { + "epoch": 0.44, + "learning_rate": 1.2328661914722431e-05, + "loss": 0.9927, + "step": 13835 + }, + { + "epoch": 0.44, + "learning_rate": 1.2327654574024384e-05, + "loss": 1.0435, + "step": 13836 + }, + { + "epoch": 0.44, + "learning_rate": 1.232664720835318e-05, + "loss": 1.0371, + "step": 13837 + }, + { + "epoch": 0.44, + "learning_rate": 1.2325639817719628e-05, + "loss": 0.9321, + "step": 13838 + }, + { + "epoch": 0.44, + "learning_rate": 1.2324632402134533e-05, + "loss": 0.8975, + "step": 13839 + }, + { + "epoch": 0.44, + "learning_rate": 1.2323624961608704e-05, + "loss": 1.0288, + "step": 13840 + }, + { + "epoch": 0.44, + "learning_rate": 1.2322617496152952e-05, + "loss": 0.9663, + "step": 13841 + }, + { + "epoch": 0.44, + "learning_rate": 1.2321610005778086e-05, + "loss": 1.062, + "step": 13842 + }, + { + "epoch": 0.44, + "learning_rate": 1.2320602490494913e-05, + "loss": 0.9097, + "step": 13843 + }, + { + "epoch": 0.44, + "learning_rate": 1.2319594950314243e-05, + "loss": 1.0498, + "step": 13844 + }, + { + "epoch": 0.44, + "learning_rate": 1.2318587385246887e-05, + "loss": 1.041, + "step": 13845 + }, + { + "epoch": 0.44, + "learning_rate": 1.2317579795303655e-05, + "loss": 0.9912, + "step": 13846 + }, + { + "epoch": 0.44, + "learning_rate": 1.2316572180495357e-05, + "loss": 1.0874, + "step": 13847 + }, + { + "epoch": 0.44, + "learning_rate": 1.2315564540832803e-05, + "loss": 1.0176, + "step": 13848 + }, + { + "epoch": 0.44, + "learning_rate": 1.2314556876326804e-05, + "loss": 0.9731, + "step": 13849 + }, + { + "epoch": 0.44, + "learning_rate": 1.2313549186988172e-05, + "loss": 1.0288, + "step": 13850 + }, + { + "epoch": 0.44, + "learning_rate": 1.2312541472827715e-05, + "loss": 0.9448, + "step": 13851 + }, + { + "epoch": 0.44, + "learning_rate": 1.231153373385625e-05, + "loss": 0.9531, + "step": 13852 + }, + { + "epoch": 0.44, + "learning_rate": 1.2310525970084587e-05, + "loss": 0.9756, + "step": 13853 + }, + { + "epoch": 0.44, + "learning_rate": 1.2309518181523537e-05, + "loss": 0.9575, + "step": 13854 + }, + { + "epoch": 0.44, + "learning_rate": 1.230851036818391e-05, + "loss": 0.9448, + "step": 13855 + }, + { + "epoch": 0.44, + "learning_rate": 1.2307502530076523e-05, + "loss": 0.9863, + "step": 13856 + }, + { + "epoch": 0.44, + "learning_rate": 1.230649466721219e-05, + "loss": 1.0054, + "step": 13857 + }, + { + "epoch": 0.44, + "learning_rate": 1.2305486779601721e-05, + "loss": 0.9331, + "step": 13858 + }, + { + "epoch": 0.44, + "learning_rate": 1.2304478867255929e-05, + "loss": 1.0249, + "step": 13859 + }, + { + "epoch": 0.44, + "learning_rate": 1.2303470930185628e-05, + "loss": 0.9844, + "step": 13860 + }, + { + "epoch": 0.44, + "learning_rate": 1.230246296840164e-05, + "loss": 0.9595, + "step": 13861 + }, + { + "epoch": 0.44, + "learning_rate": 1.2301454981914765e-05, + "loss": 0.9111, + "step": 13862 + }, + { + "epoch": 0.44, + "learning_rate": 1.2300446970735828e-05, + "loss": 1.1123, + "step": 13863 + }, + { + "epoch": 0.44, + "learning_rate": 1.2299438934875643e-05, + "loss": 1.0371, + "step": 13864 + }, + { + "epoch": 0.44, + "learning_rate": 1.229843087434502e-05, + "loss": 1.0493, + "step": 13865 + }, + { + "epoch": 0.44, + "learning_rate": 1.2297422789154778e-05, + "loss": 1.0449, + "step": 13866 + }, + { + "epoch": 0.44, + "learning_rate": 1.2296414679315734e-05, + "loss": 0.9756, + "step": 13867 + }, + { + "epoch": 0.44, + "learning_rate": 1.2295406544838702e-05, + "loss": 1.0015, + "step": 13868 + }, + { + "epoch": 0.44, + "learning_rate": 1.2294398385734496e-05, + "loss": 1.0444, + "step": 13869 + }, + { + "epoch": 0.44, + "learning_rate": 1.2293390202013935e-05, + "loss": 0.9302, + "step": 13870 + }, + { + "epoch": 0.44, + "learning_rate": 1.2292381993687838e-05, + "loss": 1.0161, + "step": 13871 + }, + { + "epoch": 0.44, + "learning_rate": 1.2291373760767017e-05, + "loss": 1.0254, + "step": 13872 + }, + { + "epoch": 0.44, + "learning_rate": 1.2290365503262292e-05, + "loss": 1.0962, + "step": 13873 + }, + { + "epoch": 0.44, + "learning_rate": 1.228935722118448e-05, + "loss": 0.9761, + "step": 13874 + }, + { + "epoch": 0.44, + "learning_rate": 1.2288348914544399e-05, + "loss": 1.1104, + "step": 13875 + }, + { + "epoch": 0.44, + "learning_rate": 1.2287340583352868e-05, + "loss": 1.0352, + "step": 13876 + }, + { + "epoch": 0.44, + "learning_rate": 1.2286332227620701e-05, + "loss": 0.9556, + "step": 13877 + }, + { + "epoch": 0.44, + "learning_rate": 1.2285323847358722e-05, + "loss": 1.1294, + "step": 13878 + }, + { + "epoch": 0.44, + "learning_rate": 1.2284315442577745e-05, + "loss": 1.02, + "step": 13879 + }, + { + "epoch": 0.44, + "learning_rate": 1.2283307013288594e-05, + "loss": 1.0459, + "step": 13880 + }, + { + "epoch": 0.44, + "learning_rate": 1.2282298559502086e-05, + "loss": 1.0498, + "step": 13881 + }, + { + "epoch": 0.44, + "learning_rate": 1.2281290081229038e-05, + "loss": 0.9238, + "step": 13882 + }, + { + "epoch": 0.44, + "learning_rate": 1.2280281578480275e-05, + "loss": 1.1113, + "step": 13883 + }, + { + "epoch": 0.44, + "learning_rate": 1.2279273051266612e-05, + "loss": 1.0571, + "step": 13884 + }, + { + "epoch": 0.44, + "learning_rate": 1.227826449959887e-05, + "loss": 1.0532, + "step": 13885 + }, + { + "epoch": 0.44, + "learning_rate": 1.2277255923487875e-05, + "loss": 1.0293, + "step": 13886 + }, + { + "epoch": 0.44, + "learning_rate": 1.2276247322944443e-05, + "loss": 1.2373, + "step": 13887 + }, + { + "epoch": 0.44, + "learning_rate": 1.2275238697979397e-05, + "loss": 1.0898, + "step": 13888 + }, + { + "epoch": 0.44, + "learning_rate": 1.2274230048603557e-05, + "loss": 0.939, + "step": 13889 + }, + { + "epoch": 0.44, + "learning_rate": 1.2273221374827746e-05, + "loss": 1.0669, + "step": 13890 + }, + { + "epoch": 0.44, + "learning_rate": 1.2272212676662787e-05, + "loss": 1.1001, + "step": 13891 + }, + { + "epoch": 0.44, + "learning_rate": 1.2271203954119498e-05, + "loss": 0.9971, + "step": 13892 + }, + { + "epoch": 0.44, + "learning_rate": 1.2270195207208705e-05, + "loss": 0.9407, + "step": 13893 + }, + { + "epoch": 0.44, + "learning_rate": 1.226918643594123e-05, + "loss": 0.9443, + "step": 13894 + }, + { + "epoch": 0.44, + "learning_rate": 1.2268177640327898e-05, + "loss": 1.0298, + "step": 13895 + }, + { + "epoch": 0.44, + "learning_rate": 1.2267168820379527e-05, + "loss": 0.9834, + "step": 13896 + }, + { + "epoch": 0.44, + "learning_rate": 1.2266159976106946e-05, + "loss": 0.9302, + "step": 13897 + }, + { + "epoch": 0.44, + "learning_rate": 1.2265151107520978e-05, + "loss": 1.0073, + "step": 13898 + }, + { + "epoch": 0.44, + "learning_rate": 1.2264142214632441e-05, + "loss": 0.9937, + "step": 13899 + }, + { + "epoch": 0.44, + "learning_rate": 1.2263133297452169e-05, + "loss": 0.9971, + "step": 13900 + }, + { + "epoch": 0.44, + "learning_rate": 1.2262124355990976e-05, + "loss": 0.9556, + "step": 13901 + }, + { + "epoch": 0.44, + "learning_rate": 1.2261115390259696e-05, + "loss": 0.9863, + "step": 13902 + }, + { + "epoch": 0.44, + "learning_rate": 1.2260106400269149e-05, + "loss": 1.0317, + "step": 13903 + }, + { + "epoch": 0.44, + "learning_rate": 1.2259097386030161e-05, + "loss": 1.1104, + "step": 13904 + }, + { + "epoch": 0.44, + "learning_rate": 1.2258088347553559e-05, + "loss": 1.041, + "step": 13905 + }, + { + "epoch": 0.44, + "learning_rate": 1.2257079284850171e-05, + "loss": 0.9883, + "step": 13906 + }, + { + "epoch": 0.44, + "learning_rate": 1.2256070197930816e-05, + "loss": 0.9814, + "step": 13907 + }, + { + "epoch": 0.44, + "learning_rate": 1.2255061086806326e-05, + "loss": 0.9072, + "step": 13908 + }, + { + "epoch": 0.44, + "learning_rate": 1.2254051951487527e-05, + "loss": 1.0645, + "step": 13909 + }, + { + "epoch": 0.44, + "learning_rate": 1.2253042791985246e-05, + "loss": 0.9976, + "step": 13910 + }, + { + "epoch": 0.44, + "learning_rate": 1.2252033608310307e-05, + "loss": 0.9326, + "step": 13911 + }, + { + "epoch": 0.44, + "learning_rate": 1.2251024400473541e-05, + "loss": 1.0283, + "step": 13912 + }, + { + "epoch": 0.44, + "learning_rate": 1.2250015168485776e-05, + "loss": 1.1001, + "step": 13913 + }, + { + "epoch": 0.44, + "learning_rate": 1.2249005912357837e-05, + "loss": 1.0479, + "step": 13914 + }, + { + "epoch": 0.45, + "learning_rate": 1.2247996632100554e-05, + "loss": 0.9365, + "step": 13915 + }, + { + "epoch": 0.45, + "learning_rate": 1.2246987327724755e-05, + "loss": 1.0303, + "step": 13916 + }, + { + "epoch": 0.45, + "learning_rate": 1.2245977999241268e-05, + "loss": 0.9854, + "step": 13917 + }, + { + "epoch": 0.45, + "learning_rate": 1.2244968646660923e-05, + "loss": 1.0864, + "step": 13918 + }, + { + "epoch": 0.45, + "learning_rate": 1.2243959269994552e-05, + "loss": 1.0688, + "step": 13919 + }, + { + "epoch": 0.45, + "learning_rate": 1.2242949869252978e-05, + "loss": 1.0532, + "step": 13920 + }, + { + "epoch": 0.45, + "learning_rate": 1.2241940444447035e-05, + "loss": 0.9932, + "step": 13921 + }, + { + "epoch": 0.45, + "learning_rate": 1.2240930995587554e-05, + "loss": 0.8635, + "step": 13922 + }, + { + "epoch": 0.45, + "learning_rate": 1.223992152268536e-05, + "loss": 0.9595, + "step": 13923 + }, + { + "epoch": 0.45, + "learning_rate": 1.2238912025751289e-05, + "loss": 0.9038, + "step": 13924 + }, + { + "epoch": 0.45, + "learning_rate": 1.2237902504796174e-05, + "loss": 1.0142, + "step": 13925 + }, + { + "epoch": 0.45, + "learning_rate": 1.2236892959830835e-05, + "loss": 0.5781, + "step": 13926 + }, + { + "epoch": 0.45, + "learning_rate": 1.2235883390866112e-05, + "loss": 0.9478, + "step": 13927 + }, + { + "epoch": 0.45, + "learning_rate": 1.2234873797912836e-05, + "loss": 1.0684, + "step": 13928 + }, + { + "epoch": 0.45, + "learning_rate": 1.2233864180981838e-05, + "loss": 1.0132, + "step": 13929 + }, + { + "epoch": 0.45, + "learning_rate": 1.2232854540083949e-05, + "loss": 0.9541, + "step": 13930 + }, + { + "epoch": 0.45, + "learning_rate": 1.223184487523e-05, + "loss": 0.9458, + "step": 13931 + }, + { + "epoch": 0.45, + "learning_rate": 1.223083518643083e-05, + "loss": 1.1333, + "step": 13932 + }, + { + "epoch": 0.45, + "learning_rate": 1.2229825473697266e-05, + "loss": 0.9971, + "step": 13933 + }, + { + "epoch": 0.45, + "learning_rate": 1.222881573704014e-05, + "loss": 0.998, + "step": 13934 + }, + { + "epoch": 0.45, + "learning_rate": 1.222780597647029e-05, + "loss": 1.0352, + "step": 13935 + }, + { + "epoch": 0.45, + "learning_rate": 1.222679619199855e-05, + "loss": 0.8906, + "step": 13936 + }, + { + "epoch": 0.45, + "learning_rate": 1.2225786383635748e-05, + "loss": 1.1021, + "step": 13937 + }, + { + "epoch": 0.45, + "learning_rate": 1.222477655139272e-05, + "loss": 1.0005, + "step": 13938 + }, + { + "epoch": 0.45, + "learning_rate": 1.2223766695280303e-05, + "loss": 0.8745, + "step": 13939 + }, + { + "epoch": 0.45, + "learning_rate": 1.2222756815309333e-05, + "loss": 0.8242, + "step": 13940 + }, + { + "epoch": 0.45, + "learning_rate": 1.222174691149064e-05, + "loss": 1.0166, + "step": 13941 + }, + { + "epoch": 0.45, + "learning_rate": 1.2220736983835062e-05, + "loss": 1.0034, + "step": 13942 + }, + { + "epoch": 0.45, + "learning_rate": 1.2219727032353433e-05, + "loss": 1.0103, + "step": 13943 + }, + { + "epoch": 0.45, + "learning_rate": 1.2218717057056592e-05, + "loss": 1.0049, + "step": 13944 + }, + { + "epoch": 0.45, + "learning_rate": 1.221770705795537e-05, + "loss": 0.9297, + "step": 13945 + }, + { + "epoch": 0.45, + "learning_rate": 1.2216697035060608e-05, + "loss": 1.0391, + "step": 13946 + }, + { + "epoch": 0.45, + "learning_rate": 1.2215686988383138e-05, + "loss": 1.0762, + "step": 13947 + }, + { + "epoch": 0.45, + "learning_rate": 1.22146769179338e-05, + "loss": 0.9907, + "step": 13948 + }, + { + "epoch": 0.45, + "learning_rate": 1.221366682372343e-05, + "loss": 1.0356, + "step": 13949 + }, + { + "epoch": 0.45, + "learning_rate": 1.2212656705762865e-05, + "loss": 0.9663, + "step": 13950 + }, + { + "epoch": 0.45, + "learning_rate": 1.2211646564062942e-05, + "loss": 1.0254, + "step": 13951 + }, + { + "epoch": 0.45, + "learning_rate": 1.22106363986345e-05, + "loss": 0.9336, + "step": 13952 + }, + { + "epoch": 0.45, + "learning_rate": 1.2209626209488375e-05, + "loss": 1.0127, + "step": 13953 + }, + { + "epoch": 0.45, + "learning_rate": 1.2208615996635405e-05, + "loss": 0.9487, + "step": 13954 + }, + { + "epoch": 0.45, + "learning_rate": 1.2207605760086432e-05, + "loss": 1.0752, + "step": 13955 + }, + { + "epoch": 0.45, + "learning_rate": 1.2206595499852291e-05, + "loss": 0.8853, + "step": 13956 + }, + { + "epoch": 0.45, + "learning_rate": 1.2205585215943822e-05, + "loss": 0.9438, + "step": 13957 + }, + { + "epoch": 0.45, + "learning_rate": 1.2204574908371869e-05, + "loss": 0.9922, + "step": 13958 + }, + { + "epoch": 0.45, + "learning_rate": 1.2203564577147262e-05, + "loss": 1.0151, + "step": 13959 + }, + { + "epoch": 0.45, + "learning_rate": 1.2202554222280849e-05, + "loss": 0.8892, + "step": 13960 + }, + { + "epoch": 0.45, + "learning_rate": 1.2201543843783466e-05, + "loss": 0.8486, + "step": 13961 + }, + { + "epoch": 0.45, + "learning_rate": 1.2200533441665953e-05, + "loss": 1.0464, + "step": 13962 + }, + { + "epoch": 0.45, + "learning_rate": 1.2199523015939155e-05, + "loss": 0.5488, + "step": 13963 + }, + { + "epoch": 0.45, + "learning_rate": 1.2198512566613904e-05, + "loss": 1.1582, + "step": 13964 + }, + { + "epoch": 0.45, + "learning_rate": 1.2197502093701048e-05, + "loss": 0.9639, + "step": 13965 + }, + { + "epoch": 0.45, + "learning_rate": 1.2196491597211431e-05, + "loss": 0.9346, + "step": 13966 + }, + { + "epoch": 0.45, + "learning_rate": 1.2195481077155886e-05, + "loss": 0.9629, + "step": 13967 + }, + { + "epoch": 0.45, + "learning_rate": 1.219447053354526e-05, + "loss": 0.9097, + "step": 13968 + }, + { + "epoch": 0.45, + "learning_rate": 1.219345996639039e-05, + "loss": 1.0542, + "step": 13969 + }, + { + "epoch": 0.45, + "learning_rate": 1.2192449375702129e-05, + "loss": 0.9409, + "step": 13970 + }, + { + "epoch": 0.45, + "learning_rate": 1.2191438761491307e-05, + "loss": 0.9707, + "step": 13971 + }, + { + "epoch": 0.45, + "learning_rate": 1.2190428123768775e-05, + "loss": 1.0439, + "step": 13972 + }, + { + "epoch": 0.45, + "learning_rate": 1.2189417462545374e-05, + "loss": 0.9844, + "step": 13973 + }, + { + "epoch": 0.45, + "learning_rate": 1.2188406777831948e-05, + "loss": 1.1318, + "step": 13974 + }, + { + "epoch": 0.45, + "learning_rate": 1.2187396069639334e-05, + "loss": 0.9883, + "step": 13975 + }, + { + "epoch": 0.45, + "learning_rate": 1.2186385337978385e-05, + "loss": 1.0601, + "step": 13976 + }, + { + "epoch": 0.45, + "learning_rate": 1.218537458285994e-05, + "loss": 0.96, + "step": 13977 + }, + { + "epoch": 0.45, + "learning_rate": 1.2184363804294845e-05, + "loss": 0.9951, + "step": 13978 + }, + { + "epoch": 0.45, + "learning_rate": 1.2183353002293946e-05, + "loss": 0.9312, + "step": 13979 + }, + { + "epoch": 0.45, + "learning_rate": 1.218234217686808e-05, + "loss": 1.1611, + "step": 13980 + }, + { + "epoch": 0.45, + "learning_rate": 1.2181331328028103e-05, + "loss": 1.1909, + "step": 13981 + }, + { + "epoch": 0.45, + "learning_rate": 1.2180320455784853e-05, + "loss": 1.0156, + "step": 13982 + }, + { + "epoch": 0.45, + "learning_rate": 1.2179309560149178e-05, + "loss": 1.0244, + "step": 13983 + }, + { + "epoch": 0.45, + "learning_rate": 1.2178298641131922e-05, + "loss": 0.8477, + "step": 13984 + }, + { + "epoch": 0.45, + "learning_rate": 1.2177287698743937e-05, + "loss": 0.8862, + "step": 13985 + }, + { + "epoch": 0.45, + "learning_rate": 1.217627673299606e-05, + "loss": 0.9834, + "step": 13986 + }, + { + "epoch": 0.45, + "learning_rate": 1.2175265743899143e-05, + "loss": 1.0952, + "step": 13987 + }, + { + "epoch": 0.45, + "learning_rate": 1.217425473146403e-05, + "loss": 0.9971, + "step": 13988 + }, + { + "epoch": 0.45, + "learning_rate": 1.2173243695701575e-05, + "loss": 1.0151, + "step": 13989 + }, + { + "epoch": 0.45, + "learning_rate": 1.2172232636622617e-05, + "loss": 0.5798, + "step": 13990 + }, + { + "epoch": 0.45, + "learning_rate": 1.2171221554238008e-05, + "loss": 0.936, + "step": 13991 + }, + { + "epoch": 0.45, + "learning_rate": 1.2170210448558594e-05, + "loss": 0.959, + "step": 13992 + }, + { + "epoch": 0.45, + "learning_rate": 1.2169199319595227e-05, + "loss": 1.0386, + "step": 13993 + }, + { + "epoch": 0.45, + "learning_rate": 1.2168188167358747e-05, + "loss": 1.0015, + "step": 13994 + }, + { + "epoch": 0.45, + "learning_rate": 1.216717699186001e-05, + "loss": 0.9761, + "step": 13995 + }, + { + "epoch": 0.45, + "learning_rate": 1.216616579310986e-05, + "loss": 1.0337, + "step": 13996 + }, + { + "epoch": 0.45, + "learning_rate": 1.2165154571119153e-05, + "loss": 1.0439, + "step": 13997 + }, + { + "epoch": 0.45, + "learning_rate": 1.2164143325898732e-05, + "loss": 0.9937, + "step": 13998 + }, + { + "epoch": 0.45, + "learning_rate": 1.2163132057459449e-05, + "loss": 0.9053, + "step": 13999 + }, + { + "epoch": 0.45, + "learning_rate": 1.216212076581215e-05, + "loss": 0.9644, + "step": 14000 + }, + { + "epoch": 0.45, + "learning_rate": 1.2161109450967689e-05, + "loss": 1.0625, + "step": 14001 + }, + { + "epoch": 0.45, + "learning_rate": 1.216009811293692e-05, + "loss": 1.0142, + "step": 14002 + }, + { + "epoch": 0.45, + "learning_rate": 1.2159086751730684e-05, + "loss": 1.0234, + "step": 14003 + }, + { + "epoch": 0.45, + "learning_rate": 1.2158075367359839e-05, + "loss": 1.0439, + "step": 14004 + }, + { + "epoch": 0.45, + "learning_rate": 1.2157063959835233e-05, + "loss": 1.0562, + "step": 14005 + }, + { + "epoch": 0.45, + "learning_rate": 1.2156052529167717e-05, + "loss": 0.9814, + "step": 14006 + }, + { + "epoch": 0.45, + "learning_rate": 1.2155041075368143e-05, + "loss": 1.0024, + "step": 14007 + }, + { + "epoch": 0.45, + "learning_rate": 1.2154029598447368e-05, + "loss": 1.0161, + "step": 14008 + }, + { + "epoch": 0.45, + "learning_rate": 1.2153018098416234e-05, + "loss": 1.022, + "step": 14009 + }, + { + "epoch": 0.45, + "learning_rate": 1.21520065752856e-05, + "loss": 0.8975, + "step": 14010 + }, + { + "epoch": 0.45, + "learning_rate": 1.215099502906632e-05, + "loss": 1.0195, + "step": 14011 + }, + { + "epoch": 0.45, + "learning_rate": 1.214998345976924e-05, + "loss": 1.105, + "step": 14012 + }, + { + "epoch": 0.45, + "learning_rate": 1.2148971867405219e-05, + "loss": 1.0396, + "step": 14013 + }, + { + "epoch": 0.45, + "learning_rate": 1.2147960251985107e-05, + "loss": 1.0347, + "step": 14014 + }, + { + "epoch": 0.45, + "learning_rate": 1.2146948613519762e-05, + "loss": 0.7988, + "step": 14015 + }, + { + "epoch": 0.45, + "learning_rate": 1.2145936952020032e-05, + "loss": 1.042, + "step": 14016 + }, + { + "epoch": 0.45, + "learning_rate": 1.2144925267496772e-05, + "loss": 1.0278, + "step": 14017 + }, + { + "epoch": 0.45, + "learning_rate": 1.214391355996084e-05, + "loss": 1.0283, + "step": 14018 + }, + { + "epoch": 0.45, + "learning_rate": 1.2142901829423091e-05, + "loss": 1.0127, + "step": 14019 + }, + { + "epoch": 0.45, + "learning_rate": 1.2141890075894371e-05, + "loss": 1.0703, + "step": 14020 + }, + { + "epoch": 0.45, + "learning_rate": 1.2140878299385544e-05, + "loss": 0.9702, + "step": 14021 + }, + { + "epoch": 0.45, + "learning_rate": 1.213986649990746e-05, + "loss": 0.9609, + "step": 14022 + }, + { + "epoch": 0.45, + "learning_rate": 1.2138854677470982e-05, + "loss": 1.0786, + "step": 14023 + }, + { + "epoch": 0.45, + "learning_rate": 1.2137842832086957e-05, + "loss": 1.0522, + "step": 14024 + }, + { + "epoch": 0.45, + "learning_rate": 1.2136830963766245e-05, + "loss": 0.979, + "step": 14025 + }, + { + "epoch": 0.45, + "learning_rate": 1.2135819072519698e-05, + "loss": 1.0156, + "step": 14026 + }, + { + "epoch": 0.45, + "learning_rate": 1.2134807158358182e-05, + "loss": 1.0796, + "step": 14027 + }, + { + "epoch": 0.45, + "learning_rate": 1.2133795221292547e-05, + "loss": 0.916, + "step": 14028 + }, + { + "epoch": 0.45, + "learning_rate": 1.2132783261333647e-05, + "loss": 0.9937, + "step": 14029 + }, + { + "epoch": 0.45, + "learning_rate": 1.2131771278492343e-05, + "loss": 0.9731, + "step": 14030 + }, + { + "epoch": 0.45, + "learning_rate": 1.2130759272779497e-05, + "loss": 0.9575, + "step": 14031 + }, + { + "epoch": 0.45, + "learning_rate": 1.212974724420596e-05, + "loss": 1.0215, + "step": 14032 + }, + { + "epoch": 0.45, + "learning_rate": 1.2128735192782592e-05, + "loss": 1.1011, + "step": 14033 + }, + { + "epoch": 0.45, + "learning_rate": 1.2127723118520254e-05, + "loss": 0.9985, + "step": 14034 + }, + { + "epoch": 0.45, + "learning_rate": 1.2126711021429798e-05, + "loss": 1.0059, + "step": 14035 + }, + { + "epoch": 0.45, + "learning_rate": 1.2125698901522086e-05, + "loss": 0.5339, + "step": 14036 + }, + { + "epoch": 0.45, + "learning_rate": 1.212468675880798e-05, + "loss": 1.062, + "step": 14037 + }, + { + "epoch": 0.45, + "learning_rate": 1.2123674593298339e-05, + "loss": 0.9985, + "step": 14038 + }, + { + "epoch": 0.45, + "learning_rate": 1.2122662405004016e-05, + "loss": 1.0352, + "step": 14039 + }, + { + "epoch": 0.45, + "learning_rate": 1.2121650193935874e-05, + "loss": 1.0171, + "step": 14040 + }, + { + "epoch": 0.45, + "learning_rate": 1.2120637960104775e-05, + "loss": 1.0029, + "step": 14041 + }, + { + "epoch": 0.45, + "learning_rate": 1.2119625703521578e-05, + "loss": 1.002, + "step": 14042 + }, + { + "epoch": 0.45, + "learning_rate": 1.2118613424197143e-05, + "loss": 1.0156, + "step": 14043 + }, + { + "epoch": 0.45, + "learning_rate": 1.211760112214233e-05, + "loss": 0.9741, + "step": 14044 + }, + { + "epoch": 0.45, + "learning_rate": 1.2116588797368001e-05, + "loss": 0.4983, + "step": 14045 + }, + { + "epoch": 0.45, + "learning_rate": 1.2115576449885018e-05, + "loss": 0.918, + "step": 14046 + }, + { + "epoch": 0.45, + "learning_rate": 1.211456407970424e-05, + "loss": 1.1777, + "step": 14047 + }, + { + "epoch": 0.45, + "learning_rate": 1.211355168683653e-05, + "loss": 1.0288, + "step": 14048 + }, + { + "epoch": 0.45, + "learning_rate": 1.2112539271292748e-05, + "loss": 1.0142, + "step": 14049 + }, + { + "epoch": 0.45, + "learning_rate": 1.211152683308376e-05, + "loss": 1.0596, + "step": 14050 + }, + { + "epoch": 0.45, + "learning_rate": 1.2110514372220426e-05, + "loss": 1.019, + "step": 14051 + }, + { + "epoch": 0.45, + "learning_rate": 1.2109501888713607e-05, + "loss": 1.0513, + "step": 14052 + }, + { + "epoch": 0.45, + "learning_rate": 1.2108489382574166e-05, + "loss": 1.0513, + "step": 14053 + }, + { + "epoch": 0.45, + "learning_rate": 1.2107476853812972e-05, + "loss": 1.0132, + "step": 14054 + }, + { + "epoch": 0.45, + "learning_rate": 1.210646430244088e-05, + "loss": 1.1084, + "step": 14055 + }, + { + "epoch": 0.45, + "learning_rate": 1.2105451728468757e-05, + "loss": 1.0483, + "step": 14056 + }, + { + "epoch": 0.45, + "learning_rate": 1.210443913190747e-05, + "loss": 0.9961, + "step": 14057 + }, + { + "epoch": 0.45, + "learning_rate": 1.2103426512767879e-05, + "loss": 1.0342, + "step": 14058 + }, + { + "epoch": 0.45, + "learning_rate": 1.2102413871060848e-05, + "loss": 0.8564, + "step": 14059 + }, + { + "epoch": 0.45, + "learning_rate": 1.2101401206797243e-05, + "loss": 1.0073, + "step": 14060 + }, + { + "epoch": 0.45, + "learning_rate": 1.2100388519987933e-05, + "loss": 1.0229, + "step": 14061 + }, + { + "epoch": 0.45, + "learning_rate": 1.2099375810643775e-05, + "loss": 0.8955, + "step": 14062 + }, + { + "epoch": 0.45, + "learning_rate": 1.2098363078775638e-05, + "loss": 0.5273, + "step": 14063 + }, + { + "epoch": 0.45, + "learning_rate": 1.2097350324394388e-05, + "loss": 0.9751, + "step": 14064 + }, + { + "epoch": 0.45, + "learning_rate": 1.2096337547510892e-05, + "loss": 0.8867, + "step": 14065 + }, + { + "epoch": 0.45, + "learning_rate": 1.2095324748136013e-05, + "loss": 0.958, + "step": 14066 + }, + { + "epoch": 0.45, + "learning_rate": 1.2094311926280617e-05, + "loss": 1.0488, + "step": 14067 + }, + { + "epoch": 0.45, + "learning_rate": 1.2093299081955575e-05, + "loss": 1.0024, + "step": 14068 + }, + { + "epoch": 0.45, + "learning_rate": 1.2092286215171747e-05, + "loss": 0.9258, + "step": 14069 + }, + { + "epoch": 0.45, + "learning_rate": 1.2091273325940004e-05, + "loss": 1.0737, + "step": 14070 + }, + { + "epoch": 0.45, + "learning_rate": 1.2090260414271215e-05, + "loss": 1.0322, + "step": 14071 + }, + { + "epoch": 0.45, + "learning_rate": 1.2089247480176245e-05, + "loss": 1.0239, + "step": 14072 + }, + { + "epoch": 0.45, + "learning_rate": 1.208823452366596e-05, + "loss": 1.0376, + "step": 14073 + }, + { + "epoch": 0.45, + "learning_rate": 1.2087221544751231e-05, + "loss": 1.0942, + "step": 14074 + }, + { + "epoch": 0.45, + "learning_rate": 1.2086208543442925e-05, + "loss": 1.0742, + "step": 14075 + }, + { + "epoch": 0.45, + "learning_rate": 1.208519551975191e-05, + "loss": 1.0625, + "step": 14076 + }, + { + "epoch": 0.45, + "learning_rate": 1.2084182473689055e-05, + "loss": 0.9644, + "step": 14077 + }, + { + "epoch": 0.45, + "learning_rate": 1.2083169405265228e-05, + "loss": 1.0488, + "step": 14078 + }, + { + "epoch": 0.45, + "learning_rate": 1.2082156314491298e-05, + "loss": 1.0088, + "step": 14079 + }, + { + "epoch": 0.45, + "learning_rate": 1.2081143201378139e-05, + "loss": 0.9214, + "step": 14080 + }, + { + "epoch": 0.45, + "learning_rate": 1.2080130065936612e-05, + "loss": 1.0586, + "step": 14081 + }, + { + "epoch": 0.45, + "learning_rate": 1.2079116908177592e-05, + "loss": 0.9482, + "step": 14082 + }, + { + "epoch": 0.45, + "learning_rate": 1.2078103728111954e-05, + "loss": 1.0225, + "step": 14083 + }, + { + "epoch": 0.45, + "learning_rate": 1.207709052575056e-05, + "loss": 1.0781, + "step": 14084 + }, + { + "epoch": 0.45, + "learning_rate": 1.2076077301104282e-05, + "loss": 0.9565, + "step": 14085 + }, + { + "epoch": 0.45, + "learning_rate": 1.2075064054183993e-05, + "loss": 1.0459, + "step": 14086 + }, + { + "epoch": 0.45, + "learning_rate": 1.2074050785000562e-05, + "loss": 1.0464, + "step": 14087 + }, + { + "epoch": 0.45, + "learning_rate": 1.2073037493564862e-05, + "loss": 1.0498, + "step": 14088 + }, + { + "epoch": 0.45, + "learning_rate": 1.2072024179887766e-05, + "loss": 1.063, + "step": 14089 + }, + { + "epoch": 0.45, + "learning_rate": 1.2071010843980141e-05, + "loss": 1.0469, + "step": 14090 + }, + { + "epoch": 0.45, + "learning_rate": 1.2069997485852867e-05, + "loss": 1.0117, + "step": 14091 + }, + { + "epoch": 0.45, + "learning_rate": 1.2068984105516806e-05, + "loss": 0.9878, + "step": 14092 + }, + { + "epoch": 0.45, + "learning_rate": 1.2067970702982836e-05, + "loss": 1.0518, + "step": 14093 + }, + { + "epoch": 0.45, + "learning_rate": 1.2066957278261828e-05, + "loss": 0.9409, + "step": 14094 + }, + { + "epoch": 0.45, + "learning_rate": 1.2065943831364659e-05, + "loss": 1.0059, + "step": 14095 + }, + { + "epoch": 0.45, + "learning_rate": 1.2064930362302198e-05, + "loss": 1.0396, + "step": 14096 + }, + { + "epoch": 0.45, + "learning_rate": 1.2063916871085319e-05, + "loss": 1.0352, + "step": 14097 + }, + { + "epoch": 0.45, + "learning_rate": 1.2062903357724898e-05, + "loss": 1.0537, + "step": 14098 + }, + { + "epoch": 0.45, + "learning_rate": 1.2061889822231804e-05, + "loss": 1.0215, + "step": 14099 + }, + { + "epoch": 0.45, + "learning_rate": 1.2060876264616916e-05, + "loss": 1.0059, + "step": 14100 + }, + { + "epoch": 0.45, + "learning_rate": 1.2059862684891106e-05, + "loss": 1.0425, + "step": 14101 + }, + { + "epoch": 0.45, + "learning_rate": 1.2058849083065251e-05, + "loss": 0.8638, + "step": 14102 + }, + { + "epoch": 0.45, + "learning_rate": 1.2057835459150222e-05, + "loss": 0.9932, + "step": 14103 + }, + { + "epoch": 0.45, + "learning_rate": 1.2056821813156896e-05, + "loss": 1.1104, + "step": 14104 + }, + { + "epoch": 0.45, + "learning_rate": 1.2055808145096148e-05, + "loss": 0.9985, + "step": 14105 + }, + { + "epoch": 0.45, + "learning_rate": 1.2054794454978856e-05, + "loss": 1.0054, + "step": 14106 + }, + { + "epoch": 0.45, + "learning_rate": 1.205378074281589e-05, + "loss": 1.249, + "step": 14107 + }, + { + "epoch": 0.45, + "learning_rate": 1.2052767008618134e-05, + "loss": 1.0098, + "step": 14108 + }, + { + "epoch": 0.45, + "learning_rate": 1.2051753252396457e-05, + "loss": 1.0864, + "step": 14109 + }, + { + "epoch": 0.45, + "learning_rate": 1.2050739474161739e-05, + "loss": 1.0713, + "step": 14110 + }, + { + "epoch": 0.45, + "learning_rate": 1.2049725673924857e-05, + "loss": 1.062, + "step": 14111 + }, + { + "epoch": 0.45, + "learning_rate": 1.2048711851696686e-05, + "loss": 1.0474, + "step": 14112 + }, + { + "epoch": 0.45, + "learning_rate": 1.2047698007488103e-05, + "loss": 1.0093, + "step": 14113 + }, + { + "epoch": 0.45, + "learning_rate": 1.204668414130999e-05, + "loss": 0.9766, + "step": 14114 + }, + { + "epoch": 0.45, + "learning_rate": 1.2045670253173218e-05, + "loss": 0.8667, + "step": 14115 + }, + { + "epoch": 0.45, + "learning_rate": 1.204465634308867e-05, + "loss": 0.9014, + "step": 14116 + }, + { + "epoch": 0.45, + "learning_rate": 1.2043642411067225e-05, + "loss": 0.9736, + "step": 14117 + }, + { + "epoch": 0.45, + "learning_rate": 1.2042628457119753e-05, + "loss": 1.1152, + "step": 14118 + }, + { + "epoch": 0.45, + "learning_rate": 1.2041614481257143e-05, + "loss": 0.9312, + "step": 14119 + }, + { + "epoch": 0.45, + "learning_rate": 1.2040600483490265e-05, + "loss": 1.0371, + "step": 14120 + }, + { + "epoch": 0.45, + "learning_rate": 1.2039586463830006e-05, + "loss": 1.0059, + "step": 14121 + }, + { + "epoch": 0.45, + "learning_rate": 1.203857242228724e-05, + "loss": 1.0796, + "step": 14122 + }, + { + "epoch": 0.45, + "learning_rate": 1.2037558358872847e-05, + "loss": 1.1025, + "step": 14123 + }, + { + "epoch": 0.45, + "learning_rate": 1.2036544273597708e-05, + "loss": 0.9912, + "step": 14124 + }, + { + "epoch": 0.45, + "learning_rate": 1.2035530166472707e-05, + "loss": 1.0933, + "step": 14125 + }, + { + "epoch": 0.45, + "learning_rate": 1.2034516037508715e-05, + "loss": 1.0547, + "step": 14126 + }, + { + "epoch": 0.45, + "learning_rate": 1.2033501886716622e-05, + "loss": 1.0957, + "step": 14127 + }, + { + "epoch": 0.45, + "learning_rate": 1.2032487714107301e-05, + "loss": 0.9653, + "step": 14128 + }, + { + "epoch": 0.45, + "learning_rate": 1.2031473519691636e-05, + "loss": 0.9834, + "step": 14129 + }, + { + "epoch": 0.45, + "learning_rate": 1.2030459303480513e-05, + "loss": 1.0151, + "step": 14130 + }, + { + "epoch": 0.45, + "learning_rate": 1.2029445065484803e-05, + "loss": 0.9619, + "step": 14131 + }, + { + "epoch": 0.45, + "learning_rate": 1.2028430805715397e-05, + "loss": 0.9316, + "step": 14132 + }, + { + "epoch": 0.45, + "learning_rate": 1.2027416524183173e-05, + "loss": 0.9624, + "step": 14133 + }, + { + "epoch": 0.45, + "learning_rate": 1.2026402220899014e-05, + "loss": 1.2017, + "step": 14134 + }, + { + "epoch": 0.45, + "learning_rate": 1.2025387895873798e-05, + "loss": 1.0068, + "step": 14135 + }, + { + "epoch": 0.45, + "learning_rate": 1.2024373549118418e-05, + "loss": 0.9268, + "step": 14136 + }, + { + "epoch": 0.45, + "learning_rate": 1.2023359180643746e-05, + "loss": 1.1724, + "step": 14137 + }, + { + "epoch": 0.45, + "learning_rate": 1.202234479046067e-05, + "loss": 0.9922, + "step": 14138 + }, + { + "epoch": 0.45, + "learning_rate": 1.2021330378580071e-05, + "loss": 1.061, + "step": 14139 + }, + { + "epoch": 0.45, + "learning_rate": 1.2020315945012836e-05, + "loss": 1.0151, + "step": 14140 + }, + { + "epoch": 0.45, + "learning_rate": 1.2019301489769847e-05, + "loss": 0.9883, + "step": 14141 + }, + { + "epoch": 0.45, + "learning_rate": 1.2018287012861986e-05, + "loss": 1.0586, + "step": 14142 + }, + { + "epoch": 0.45, + "learning_rate": 1.2017272514300143e-05, + "loss": 1.0786, + "step": 14143 + }, + { + "epoch": 0.45, + "learning_rate": 1.2016257994095195e-05, + "loss": 0.9473, + "step": 14144 + }, + { + "epoch": 0.45, + "learning_rate": 1.2015243452258032e-05, + "loss": 0.9834, + "step": 14145 + }, + { + "epoch": 0.45, + "learning_rate": 1.2014228888799535e-05, + "loss": 1.0723, + "step": 14146 + }, + { + "epoch": 0.45, + "learning_rate": 1.2013214303730594e-05, + "loss": 1.0405, + "step": 14147 + }, + { + "epoch": 0.45, + "learning_rate": 1.201219969706209e-05, + "loss": 1.0986, + "step": 14148 + }, + { + "epoch": 0.45, + "learning_rate": 1.2011185068804912e-05, + "loss": 1.0942, + "step": 14149 + }, + { + "epoch": 0.45, + "learning_rate": 1.2010170418969942e-05, + "loss": 0.5422, + "step": 14150 + }, + { + "epoch": 0.45, + "learning_rate": 1.200915574756807e-05, + "loss": 0.9468, + "step": 14151 + }, + { + "epoch": 0.45, + "learning_rate": 1.2008141054610179e-05, + "loss": 1.0195, + "step": 14152 + }, + { + "epoch": 0.45, + "learning_rate": 1.2007126340107156e-05, + "loss": 1.0259, + "step": 14153 + }, + { + "epoch": 0.45, + "learning_rate": 1.200611160406989e-05, + "loss": 1.0815, + "step": 14154 + }, + { + "epoch": 0.45, + "learning_rate": 1.2005096846509268e-05, + "loss": 1.0571, + "step": 14155 + }, + { + "epoch": 0.45, + "learning_rate": 1.2004082067436174e-05, + "loss": 1.0254, + "step": 14156 + }, + { + "epoch": 0.45, + "learning_rate": 1.2003067266861496e-05, + "loss": 1.1382, + "step": 14157 + }, + { + "epoch": 0.45, + "learning_rate": 1.2002052444796127e-05, + "loss": 1.0615, + "step": 14158 + }, + { + "epoch": 0.45, + "learning_rate": 1.2001037601250949e-05, + "loss": 1.0396, + "step": 14159 + }, + { + "epoch": 0.45, + "learning_rate": 1.2000022736236852e-05, + "loss": 1.1997, + "step": 14160 + }, + { + "epoch": 0.45, + "learning_rate": 1.1999007849764726e-05, + "loss": 1.0181, + "step": 14161 + }, + { + "epoch": 0.45, + "learning_rate": 1.1997992941845456e-05, + "loss": 0.9287, + "step": 14162 + }, + { + "epoch": 0.45, + "learning_rate": 1.1996978012489935e-05, + "loss": 0.9736, + "step": 14163 + }, + { + "epoch": 0.45, + "learning_rate": 1.1995963061709048e-05, + "loss": 0.8477, + "step": 14164 + }, + { + "epoch": 0.45, + "learning_rate": 1.1994948089513687e-05, + "loss": 1.1133, + "step": 14165 + }, + { + "epoch": 0.45, + "learning_rate": 1.1993933095914746e-05, + "loss": 0.9307, + "step": 14166 + }, + { + "epoch": 0.45, + "learning_rate": 1.1992918080923101e-05, + "loss": 0.9624, + "step": 14167 + }, + { + "epoch": 0.45, + "learning_rate": 1.1991903044549655e-05, + "loss": 0.9824, + "step": 14168 + }, + { + "epoch": 0.45, + "learning_rate": 1.1990887986805295e-05, + "loss": 0.5027, + "step": 14169 + }, + { + "epoch": 0.45, + "learning_rate": 1.198987290770091e-05, + "loss": 1.0034, + "step": 14170 + }, + { + "epoch": 0.45, + "learning_rate": 1.198885780724739e-05, + "loss": 0.9927, + "step": 14171 + }, + { + "epoch": 0.45, + "learning_rate": 1.1987842685455626e-05, + "loss": 1.1533, + "step": 14172 + }, + { + "epoch": 0.45, + "learning_rate": 1.1986827542336513e-05, + "loss": 0.8618, + "step": 14173 + }, + { + "epoch": 0.45, + "learning_rate": 1.1985812377900936e-05, + "loss": 0.9756, + "step": 14174 + }, + { + "epoch": 0.45, + "learning_rate": 1.1984797192159791e-05, + "loss": 1.0059, + "step": 14175 + }, + { + "epoch": 0.45, + "learning_rate": 1.198378198512397e-05, + "loss": 1.0386, + "step": 14176 + }, + { + "epoch": 0.45, + "learning_rate": 1.1982766756804362e-05, + "loss": 1.1045, + "step": 14177 + }, + { + "epoch": 0.45, + "learning_rate": 1.1981751507211862e-05, + "loss": 0.9458, + "step": 14178 + }, + { + "epoch": 0.45, + "learning_rate": 1.1980736236357362e-05, + "loss": 0.9087, + "step": 14179 + }, + { + "epoch": 0.45, + "learning_rate": 1.1979720944251753e-05, + "loss": 0.938, + "step": 14180 + }, + { + "epoch": 0.45, + "learning_rate": 1.1978705630905931e-05, + "loss": 1.0854, + "step": 14181 + }, + { + "epoch": 0.45, + "learning_rate": 1.1977690296330788e-05, + "loss": 0.9976, + "step": 14182 + }, + { + "epoch": 0.45, + "learning_rate": 1.1976674940537215e-05, + "loss": 0.9893, + "step": 14183 + }, + { + "epoch": 0.45, + "learning_rate": 1.1975659563536108e-05, + "loss": 0.9526, + "step": 14184 + }, + { + "epoch": 0.45, + "learning_rate": 1.1974644165338362e-05, + "loss": 0.9282, + "step": 14185 + }, + { + "epoch": 0.45, + "learning_rate": 1.1973628745954869e-05, + "loss": 1.0762, + "step": 14186 + }, + { + "epoch": 0.45, + "learning_rate": 1.1972613305396521e-05, + "loss": 0.918, + "step": 14187 + }, + { + "epoch": 0.45, + "learning_rate": 1.1971597843674217e-05, + "loss": 0.9443, + "step": 14188 + }, + { + "epoch": 0.45, + "learning_rate": 1.1970582360798852e-05, + "loss": 0.8481, + "step": 14189 + }, + { + "epoch": 0.45, + "learning_rate": 1.1969566856781319e-05, + "loss": 0.9917, + "step": 14190 + }, + { + "epoch": 0.45, + "learning_rate": 1.196855133163251e-05, + "loss": 1.0591, + "step": 14191 + }, + { + "epoch": 0.45, + "learning_rate": 1.1967535785363327e-05, + "loss": 1.0532, + "step": 14192 + }, + { + "epoch": 0.45, + "learning_rate": 1.1966520217984664e-05, + "loss": 0.9868, + "step": 14193 + }, + { + "epoch": 0.45, + "learning_rate": 1.1965504629507412e-05, + "loss": 1.1055, + "step": 14194 + }, + { + "epoch": 0.45, + "learning_rate": 1.1964489019942471e-05, + "loss": 1.0835, + "step": 14195 + }, + { + "epoch": 0.45, + "learning_rate": 1.1963473389300736e-05, + "loss": 1.0083, + "step": 14196 + }, + { + "epoch": 0.45, + "learning_rate": 1.1962457737593109e-05, + "loss": 0.8237, + "step": 14197 + }, + { + "epoch": 0.45, + "learning_rate": 1.1961442064830479e-05, + "loss": 0.5283, + "step": 14198 + }, + { + "epoch": 0.45, + "learning_rate": 1.1960426371023747e-05, + "loss": 1.0479, + "step": 14199 + }, + { + "epoch": 0.45, + "learning_rate": 1.1959410656183812e-05, + "loss": 1.0093, + "step": 14200 + }, + { + "epoch": 0.45, + "learning_rate": 1.1958394920321568e-05, + "loss": 0.9961, + "step": 14201 + }, + { + "epoch": 0.45, + "learning_rate": 1.195737916344791e-05, + "loss": 1.0459, + "step": 14202 + }, + { + "epoch": 0.45, + "learning_rate": 1.1956363385573745e-05, + "loss": 0.9658, + "step": 14203 + }, + { + "epoch": 0.45, + "learning_rate": 1.1955347586709964e-05, + "loss": 0.9097, + "step": 14204 + }, + { + "epoch": 0.45, + "learning_rate": 1.1954331766867469e-05, + "loss": 1.0093, + "step": 14205 + }, + { + "epoch": 0.45, + "learning_rate": 1.1953315926057154e-05, + "loss": 1.0571, + "step": 14206 + }, + { + "epoch": 0.45, + "learning_rate": 1.1952300064289923e-05, + "loss": 0.9678, + "step": 14207 + }, + { + "epoch": 0.45, + "learning_rate": 1.1951284181576674e-05, + "loss": 1.0669, + "step": 14208 + }, + { + "epoch": 0.45, + "learning_rate": 1.1950268277928304e-05, + "loss": 0.9219, + "step": 14209 + }, + { + "epoch": 0.45, + "learning_rate": 1.1949252353355714e-05, + "loss": 1.0244, + "step": 14210 + }, + { + "epoch": 0.45, + "learning_rate": 1.1948236407869804e-05, + "loss": 0.8184, + "step": 14211 + }, + { + "epoch": 0.45, + "learning_rate": 1.1947220441481473e-05, + "loss": 1.1504, + "step": 14212 + }, + { + "epoch": 0.45, + "learning_rate": 1.194620445420162e-05, + "loss": 1.1099, + "step": 14213 + }, + { + "epoch": 0.45, + "learning_rate": 1.194518844604115e-05, + "loss": 0.8862, + "step": 14214 + }, + { + "epoch": 0.45, + "learning_rate": 1.1944172417010959e-05, + "loss": 1.0322, + "step": 14215 + }, + { + "epoch": 0.45, + "learning_rate": 1.1943156367121952e-05, + "loss": 1.1328, + "step": 14216 + }, + { + "epoch": 0.45, + "learning_rate": 1.1942140296385027e-05, + "loss": 0.979, + "step": 14217 + }, + { + "epoch": 0.45, + "learning_rate": 1.1941124204811085e-05, + "loss": 1.064, + "step": 14218 + }, + { + "epoch": 0.45, + "learning_rate": 1.194010809241103e-05, + "loss": 0.9287, + "step": 14219 + }, + { + "epoch": 0.45, + "learning_rate": 1.193909195919576e-05, + "loss": 0.9312, + "step": 14220 + }, + { + "epoch": 0.45, + "learning_rate": 1.193807580517618e-05, + "loss": 0.9712, + "step": 14221 + }, + { + "epoch": 0.45, + "learning_rate": 1.1937059630363192e-05, + "loss": 0.9834, + "step": 14222 + }, + { + "epoch": 0.45, + "learning_rate": 1.1936043434767699e-05, + "loss": 0.9253, + "step": 14223 + }, + { + "epoch": 0.45, + "learning_rate": 1.1935027218400604e-05, + "loss": 0.9941, + "step": 14224 + }, + { + "epoch": 0.45, + "learning_rate": 1.1934010981272803e-05, + "loss": 1.1553, + "step": 14225 + }, + { + "epoch": 0.45, + "learning_rate": 1.1932994723395208e-05, + "loss": 1.0776, + "step": 14226 + }, + { + "epoch": 0.45, + "learning_rate": 1.193197844477872e-05, + "loss": 0.9194, + "step": 14227 + }, + { + "epoch": 0.46, + "learning_rate": 1.1930962145434242e-05, + "loss": 1.0034, + "step": 14228 + }, + { + "epoch": 0.46, + "learning_rate": 1.1929945825372674e-05, + "loss": 0.9028, + "step": 14229 + }, + { + "epoch": 0.46, + "learning_rate": 1.1928929484604925e-05, + "loss": 0.9912, + "step": 14230 + }, + { + "epoch": 0.46, + "learning_rate": 1.19279131231419e-05, + "loss": 0.8872, + "step": 14231 + }, + { + "epoch": 0.46, + "learning_rate": 1.1926896740994495e-05, + "loss": 1.1465, + "step": 14232 + }, + { + "epoch": 0.46, + "learning_rate": 1.1925880338173624e-05, + "loss": 1.0312, + "step": 14233 + }, + { + "epoch": 0.46, + "learning_rate": 1.192486391469019e-05, + "loss": 0.9561, + "step": 14234 + }, + { + "epoch": 0.46, + "learning_rate": 1.1923847470555094e-05, + "loss": 0.9717, + "step": 14235 + }, + { + "epoch": 0.46, + "learning_rate": 1.1922831005779245e-05, + "loss": 0.9219, + "step": 14236 + }, + { + "epoch": 0.46, + "learning_rate": 1.1921814520373544e-05, + "loss": 0.9116, + "step": 14237 + }, + { + "epoch": 0.46, + "learning_rate": 1.1920798014348906e-05, + "loss": 1.0063, + "step": 14238 + }, + { + "epoch": 0.46, + "learning_rate": 1.1919781487716228e-05, + "loss": 1.1211, + "step": 14239 + }, + { + "epoch": 0.46, + "learning_rate": 1.1918764940486421e-05, + "loss": 0.9561, + "step": 14240 + }, + { + "epoch": 0.46, + "learning_rate": 1.1917748372670386e-05, + "loss": 0.9424, + "step": 14241 + }, + { + "epoch": 0.46, + "learning_rate": 1.1916731784279038e-05, + "loss": 0.8823, + "step": 14242 + }, + { + "epoch": 0.46, + "learning_rate": 1.1915715175323277e-05, + "loss": 0.9961, + "step": 14243 + }, + { + "epoch": 0.46, + "learning_rate": 1.1914698545814012e-05, + "loss": 1.0054, + "step": 14244 + }, + { + "epoch": 0.46, + "learning_rate": 1.191368189576215e-05, + "loss": 0.8721, + "step": 14245 + }, + { + "epoch": 0.46, + "learning_rate": 1.1912665225178601e-05, + "loss": 0.9409, + "step": 14246 + }, + { + "epoch": 0.46, + "learning_rate": 1.1911648534074271e-05, + "loss": 1.0459, + "step": 14247 + }, + { + "epoch": 0.46, + "learning_rate": 1.1910631822460066e-05, + "loss": 0.9551, + "step": 14248 + }, + { + "epoch": 0.46, + "learning_rate": 1.19096150903469e-05, + "loss": 1.0146, + "step": 14249 + }, + { + "epoch": 0.46, + "learning_rate": 1.1908598337745674e-05, + "loss": 0.98, + "step": 14250 + }, + { + "epoch": 0.46, + "learning_rate": 1.1907581564667302e-05, + "loss": 0.9922, + "step": 14251 + }, + { + "epoch": 0.46, + "learning_rate": 1.1906564771122688e-05, + "loss": 0.9272, + "step": 14252 + }, + { + "epoch": 0.46, + "learning_rate": 1.1905547957122748e-05, + "loss": 0.9043, + "step": 14253 + }, + { + "epoch": 0.46, + "learning_rate": 1.1904531122678385e-05, + "loss": 0.9946, + "step": 14254 + }, + { + "epoch": 0.46, + "learning_rate": 1.190351426780051e-05, + "loss": 0.8872, + "step": 14255 + }, + { + "epoch": 0.46, + "learning_rate": 1.1902497392500038e-05, + "loss": 0.9844, + "step": 14256 + }, + { + "epoch": 0.46, + "learning_rate": 1.1901480496787871e-05, + "loss": 1.02, + "step": 14257 + }, + { + "epoch": 0.46, + "learning_rate": 1.1900463580674925e-05, + "loss": 1.0522, + "step": 14258 + }, + { + "epoch": 0.46, + "learning_rate": 1.1899446644172106e-05, + "loss": 0.9419, + "step": 14259 + }, + { + "epoch": 0.46, + "learning_rate": 1.1898429687290327e-05, + "loss": 0.9727, + "step": 14260 + }, + { + "epoch": 0.46, + "learning_rate": 1.1897412710040501e-05, + "loss": 0.918, + "step": 14261 + }, + { + "epoch": 0.46, + "learning_rate": 1.1896395712433533e-05, + "loss": 0.9839, + "step": 14262 + }, + { + "epoch": 0.46, + "learning_rate": 1.1895378694480338e-05, + "loss": 1.0737, + "step": 14263 + }, + { + "epoch": 0.46, + "learning_rate": 1.1894361656191829e-05, + "loss": 1.0239, + "step": 14264 + }, + { + "epoch": 0.46, + "learning_rate": 1.1893344597578917e-05, + "loss": 0.9282, + "step": 14265 + }, + { + "epoch": 0.46, + "learning_rate": 1.189232751865251e-05, + "loss": 0.9893, + "step": 14266 + }, + { + "epoch": 0.46, + "learning_rate": 1.1891310419423526e-05, + "loss": 1.0571, + "step": 14267 + }, + { + "epoch": 0.46, + "learning_rate": 1.1890293299902873e-05, + "loss": 0.9883, + "step": 14268 + }, + { + "epoch": 0.46, + "learning_rate": 1.1889276160101464e-05, + "loss": 0.9741, + "step": 14269 + }, + { + "epoch": 0.46, + "learning_rate": 1.1888259000030214e-05, + "loss": 1.063, + "step": 14270 + }, + { + "epoch": 0.46, + "learning_rate": 1.1887241819700033e-05, + "loss": 1.0508, + "step": 14271 + }, + { + "epoch": 0.46, + "learning_rate": 1.1886224619121838e-05, + "loss": 0.9126, + "step": 14272 + }, + { + "epoch": 0.46, + "learning_rate": 1.188520739830654e-05, + "loss": 1.0972, + "step": 14273 + }, + { + "epoch": 0.46, + "learning_rate": 1.1884190157265051e-05, + "loss": 0.8691, + "step": 14274 + }, + { + "epoch": 0.46, + "learning_rate": 1.1883172896008289e-05, + "loss": 1.0781, + "step": 14275 + }, + { + "epoch": 0.46, + "learning_rate": 1.1882155614547169e-05, + "loss": 0.9546, + "step": 14276 + }, + { + "epoch": 0.46, + "learning_rate": 1.1881138312892598e-05, + "loss": 0.5688, + "step": 14277 + }, + { + "epoch": 0.46, + "learning_rate": 1.1880120991055495e-05, + "loss": 1.0376, + "step": 14278 + }, + { + "epoch": 0.46, + "learning_rate": 1.1879103649046774e-05, + "loss": 1.0034, + "step": 14279 + }, + { + "epoch": 0.46, + "learning_rate": 1.1878086286877353e-05, + "loss": 1.0972, + "step": 14280 + }, + { + "epoch": 0.46, + "learning_rate": 1.1877068904558144e-05, + "loss": 0.9404, + "step": 14281 + }, + { + "epoch": 0.46, + "learning_rate": 1.1876051502100063e-05, + "loss": 1.0972, + "step": 14282 + }, + { + "epoch": 0.46, + "learning_rate": 1.1875034079514026e-05, + "loss": 0.9214, + "step": 14283 + }, + { + "epoch": 0.46, + "learning_rate": 1.1874016636810948e-05, + "loss": 1.0459, + "step": 14284 + }, + { + "epoch": 0.46, + "learning_rate": 1.1872999174001745e-05, + "loss": 0.9463, + "step": 14285 + }, + { + "epoch": 0.46, + "learning_rate": 1.1871981691097333e-05, + "loss": 0.5112, + "step": 14286 + }, + { + "epoch": 0.46, + "learning_rate": 1.1870964188108633e-05, + "loss": 0.8896, + "step": 14287 + }, + { + "epoch": 0.46, + "learning_rate": 1.1869946665046555e-05, + "loss": 0.9678, + "step": 14288 + }, + { + "epoch": 0.46, + "learning_rate": 1.1868929121922019e-05, + "loss": 0.9688, + "step": 14289 + }, + { + "epoch": 0.46, + "learning_rate": 1.1867911558745941e-05, + "loss": 1.0264, + "step": 14290 + }, + { + "epoch": 0.46, + "learning_rate": 1.1866893975529244e-05, + "loss": 1.0049, + "step": 14291 + }, + { + "epoch": 0.46, + "learning_rate": 1.1865876372282836e-05, + "loss": 0.8657, + "step": 14292 + }, + { + "epoch": 0.46, + "learning_rate": 1.1864858749017641e-05, + "loss": 0.9692, + "step": 14293 + }, + { + "epoch": 0.46, + "learning_rate": 1.1863841105744574e-05, + "loss": 0.9355, + "step": 14294 + }, + { + "epoch": 0.46, + "learning_rate": 1.1862823442474557e-05, + "loss": 0.8472, + "step": 14295 + }, + { + "epoch": 0.46, + "learning_rate": 1.1861805759218505e-05, + "loss": 0.9634, + "step": 14296 + }, + { + "epoch": 0.46, + "learning_rate": 1.186078805598734e-05, + "loss": 1.1343, + "step": 14297 + }, + { + "epoch": 0.46, + "learning_rate": 1.1859770332791975e-05, + "loss": 1.0732, + "step": 14298 + }, + { + "epoch": 0.46, + "learning_rate": 1.1858752589643334e-05, + "loss": 1.104, + "step": 14299 + }, + { + "epoch": 0.46, + "learning_rate": 1.1857734826552339e-05, + "loss": 0.8799, + "step": 14300 + }, + { + "epoch": 0.46, + "learning_rate": 1.18567170435299e-05, + "loss": 0.9976, + "step": 14301 + }, + { + "epoch": 0.46, + "learning_rate": 1.1855699240586942e-05, + "loss": 0.9463, + "step": 14302 + }, + { + "epoch": 0.46, + "learning_rate": 1.1854681417734391e-05, + "loss": 1.0684, + "step": 14303 + }, + { + "epoch": 0.46, + "learning_rate": 1.1853663574983154e-05, + "loss": 1.0322, + "step": 14304 + }, + { + "epoch": 0.46, + "learning_rate": 1.185264571234416e-05, + "loss": 0.9932, + "step": 14305 + }, + { + "epoch": 0.46, + "learning_rate": 1.1851627829828332e-05, + "loss": 1.0117, + "step": 14306 + }, + { + "epoch": 0.46, + "learning_rate": 1.1850609927446582e-05, + "loss": 0.9854, + "step": 14307 + }, + { + "epoch": 0.46, + "learning_rate": 1.1849592005209838e-05, + "loss": 0.8862, + "step": 14308 + }, + { + "epoch": 0.46, + "learning_rate": 1.1848574063129017e-05, + "loss": 1.0415, + "step": 14309 + }, + { + "epoch": 0.46, + "learning_rate": 1.1847556101215044e-05, + "loss": 1.0146, + "step": 14310 + }, + { + "epoch": 0.46, + "learning_rate": 1.1846538119478837e-05, + "loss": 0.9858, + "step": 14311 + }, + { + "epoch": 0.46, + "learning_rate": 1.1845520117931322e-05, + "loss": 1.0972, + "step": 14312 + }, + { + "epoch": 0.46, + "learning_rate": 1.1844502096583415e-05, + "loss": 0.9893, + "step": 14313 + }, + { + "epoch": 0.46, + "learning_rate": 1.1843484055446045e-05, + "loss": 0.939, + "step": 14314 + }, + { + "epoch": 0.46, + "learning_rate": 1.184246599453013e-05, + "loss": 0.9023, + "step": 14315 + }, + { + "epoch": 0.46, + "learning_rate": 1.1841447913846596e-05, + "loss": 0.9756, + "step": 14316 + }, + { + "epoch": 0.46, + "learning_rate": 1.1840429813406363e-05, + "loss": 0.9355, + "step": 14317 + }, + { + "epoch": 0.46, + "learning_rate": 1.1839411693220355e-05, + "loss": 0.9536, + "step": 14318 + }, + { + "epoch": 0.46, + "learning_rate": 1.1838393553299495e-05, + "loss": 0.8877, + "step": 14319 + }, + { + "epoch": 0.46, + "learning_rate": 1.1837375393654704e-05, + "loss": 0.9814, + "step": 14320 + }, + { + "epoch": 0.46, + "learning_rate": 1.1836357214296915e-05, + "loss": 0.9028, + "step": 14321 + }, + { + "epoch": 0.46, + "learning_rate": 1.183533901523704e-05, + "loss": 0.9883, + "step": 14322 + }, + { + "epoch": 0.46, + "learning_rate": 1.183432079648601e-05, + "loss": 0.8857, + "step": 14323 + }, + { + "epoch": 0.46, + "learning_rate": 1.1833302558054746e-05, + "loss": 0.9644, + "step": 14324 + }, + { + "epoch": 0.46, + "learning_rate": 1.1832284299954177e-05, + "loss": 0.5603, + "step": 14325 + }, + { + "epoch": 0.46, + "learning_rate": 1.1831266022195226e-05, + "loss": 0.9131, + "step": 14326 + }, + { + "epoch": 0.46, + "learning_rate": 1.1830247724788814e-05, + "loss": 0.5957, + "step": 14327 + }, + { + "epoch": 0.46, + "learning_rate": 1.1829229407745873e-05, + "loss": 0.9902, + "step": 14328 + }, + { + "epoch": 0.46, + "learning_rate": 1.1828211071077326e-05, + "loss": 0.8979, + "step": 14329 + }, + { + "epoch": 0.46, + "learning_rate": 1.1827192714794093e-05, + "loss": 0.5193, + "step": 14330 + }, + { + "epoch": 0.46, + "learning_rate": 1.1826174338907109e-05, + "loss": 0.9604, + "step": 14331 + }, + { + "epoch": 0.46, + "learning_rate": 1.1825155943427294e-05, + "loss": 0.9482, + "step": 14332 + }, + { + "epoch": 0.46, + "learning_rate": 1.1824137528365573e-05, + "loss": 0.8599, + "step": 14333 + }, + { + "epoch": 0.46, + "learning_rate": 1.1823119093732878e-05, + "loss": 0.9746, + "step": 14334 + }, + { + "epoch": 0.46, + "learning_rate": 1.1822100639540131e-05, + "loss": 0.9692, + "step": 14335 + }, + { + "epoch": 0.46, + "learning_rate": 1.1821082165798263e-05, + "loss": 1.0151, + "step": 14336 + }, + { + "epoch": 0.46, + "learning_rate": 1.1820063672518198e-05, + "loss": 0.9072, + "step": 14337 + }, + { + "epoch": 0.46, + "learning_rate": 1.1819045159710862e-05, + "loss": 1.1045, + "step": 14338 + }, + { + "epoch": 0.46, + "learning_rate": 1.181802662738719e-05, + "loss": 1.0928, + "step": 14339 + }, + { + "epoch": 0.46, + "learning_rate": 1.18170080755581e-05, + "loss": 0.936, + "step": 14340 + }, + { + "epoch": 0.46, + "learning_rate": 1.1815989504234524e-05, + "loss": 1.0049, + "step": 14341 + }, + { + "epoch": 0.46, + "learning_rate": 1.1814970913427392e-05, + "loss": 0.9673, + "step": 14342 + }, + { + "epoch": 0.46, + "learning_rate": 1.1813952303147631e-05, + "loss": 1.0737, + "step": 14343 + }, + { + "epoch": 0.46, + "learning_rate": 1.1812933673406169e-05, + "loss": 1.0586, + "step": 14344 + }, + { + "epoch": 0.46, + "learning_rate": 1.1811915024213937e-05, + "loss": 1.0166, + "step": 14345 + }, + { + "epoch": 0.46, + "learning_rate": 1.1810896355581858e-05, + "loss": 0.9731, + "step": 14346 + }, + { + "epoch": 0.46, + "learning_rate": 1.1809877667520866e-05, + "loss": 0.9951, + "step": 14347 + }, + { + "epoch": 0.46, + "learning_rate": 1.180885896004189e-05, + "loss": 0.9336, + "step": 14348 + }, + { + "epoch": 0.46, + "learning_rate": 1.1807840233155863e-05, + "loss": 0.9614, + "step": 14349 + }, + { + "epoch": 0.46, + "learning_rate": 1.1806821486873707e-05, + "loss": 1.0103, + "step": 14350 + }, + { + "epoch": 0.46, + "learning_rate": 1.1805802721206359e-05, + "loss": 1.0142, + "step": 14351 + }, + { + "epoch": 0.46, + "learning_rate": 1.1804783936164745e-05, + "loss": 1.0854, + "step": 14352 + }, + { + "epoch": 0.46, + "learning_rate": 1.1803765131759795e-05, + "loss": 1.1099, + "step": 14353 + }, + { + "epoch": 0.46, + "learning_rate": 1.1802746308002444e-05, + "loss": 0.8711, + "step": 14354 + }, + { + "epoch": 0.46, + "learning_rate": 1.1801727464903621e-05, + "loss": 0.9854, + "step": 14355 + }, + { + "epoch": 0.46, + "learning_rate": 1.1800708602474253e-05, + "loss": 1.0244, + "step": 14356 + }, + { + "epoch": 0.46, + "learning_rate": 1.1799689720725276e-05, + "loss": 0.9844, + "step": 14357 + }, + { + "epoch": 0.46, + "learning_rate": 1.179867081966762e-05, + "loss": 0.8691, + "step": 14358 + }, + { + "epoch": 0.46, + "learning_rate": 1.1797651899312219e-05, + "loss": 0.9834, + "step": 14359 + }, + { + "epoch": 0.46, + "learning_rate": 1.179663295967e-05, + "loss": 1.0962, + "step": 14360 + }, + { + "epoch": 0.46, + "learning_rate": 1.1795614000751899e-05, + "loss": 1.0513, + "step": 14361 + }, + { + "epoch": 0.46, + "learning_rate": 1.1794595022568846e-05, + "loss": 1.0283, + "step": 14362 + }, + { + "epoch": 0.46, + "learning_rate": 1.1793576025131777e-05, + "loss": 0.9785, + "step": 14363 + }, + { + "epoch": 0.46, + "learning_rate": 1.1792557008451619e-05, + "loss": 0.9233, + "step": 14364 + }, + { + "epoch": 0.46, + "learning_rate": 1.1791537972539312e-05, + "loss": 0.9336, + "step": 14365 + }, + { + "epoch": 0.46, + "learning_rate": 1.1790518917405782e-05, + "loss": 0.9556, + "step": 14366 + }, + { + "epoch": 0.46, + "learning_rate": 1.1789499843061967e-05, + "loss": 0.8394, + "step": 14367 + }, + { + "epoch": 0.46, + "learning_rate": 1.1788480749518799e-05, + "loss": 0.9829, + "step": 14368 + }, + { + "epoch": 0.46, + "learning_rate": 1.1787461636787212e-05, + "loss": 1.0122, + "step": 14369 + }, + { + "epoch": 0.46, + "learning_rate": 1.178644250487814e-05, + "loss": 0.9141, + "step": 14370 + }, + { + "epoch": 0.46, + "learning_rate": 1.1785423353802516e-05, + "loss": 0.9946, + "step": 14371 + }, + { + "epoch": 0.46, + "learning_rate": 1.1784404183571277e-05, + "loss": 0.9746, + "step": 14372 + }, + { + "epoch": 0.46, + "learning_rate": 1.1783384994195355e-05, + "loss": 0.9292, + "step": 14373 + }, + { + "epoch": 0.46, + "learning_rate": 1.1782365785685688e-05, + "loss": 1.0884, + "step": 14374 + }, + { + "epoch": 0.46, + "learning_rate": 1.1781346558053206e-05, + "loss": 0.9473, + "step": 14375 + }, + { + "epoch": 0.46, + "learning_rate": 1.1780327311308845e-05, + "loss": 1.0664, + "step": 14376 + }, + { + "epoch": 0.46, + "learning_rate": 1.1779308045463546e-05, + "loss": 0.9775, + "step": 14377 + }, + { + "epoch": 0.46, + "learning_rate": 1.1778288760528243e-05, + "loss": 0.7825, + "step": 14378 + }, + { + "epoch": 0.46, + "learning_rate": 1.1777269456513865e-05, + "loss": 0.9985, + "step": 14379 + }, + { + "epoch": 0.46, + "learning_rate": 1.1776250133431353e-05, + "loss": 0.9521, + "step": 14380 + }, + { + "epoch": 0.46, + "learning_rate": 1.1775230791291648e-05, + "loss": 0.9565, + "step": 14381 + }, + { + "epoch": 0.46, + "learning_rate": 1.1774211430105675e-05, + "loss": 0.9028, + "step": 14382 + }, + { + "epoch": 0.46, + "learning_rate": 1.1773192049884382e-05, + "loss": 1.1509, + "step": 14383 + }, + { + "epoch": 0.46, + "learning_rate": 1.1772172650638698e-05, + "loss": 1.0366, + "step": 14384 + }, + { + "epoch": 0.46, + "learning_rate": 1.1771153232379562e-05, + "loss": 0.9937, + "step": 14385 + }, + { + "epoch": 0.46, + "learning_rate": 1.1770133795117915e-05, + "loss": 0.9839, + "step": 14386 + }, + { + "epoch": 0.46, + "learning_rate": 1.176911433886469e-05, + "loss": 0.9414, + "step": 14387 + }, + { + "epoch": 0.46, + "learning_rate": 1.1768094863630826e-05, + "loss": 1.0493, + "step": 14388 + }, + { + "epoch": 0.46, + "learning_rate": 1.1767075369427261e-05, + "loss": 1.0522, + "step": 14389 + }, + { + "epoch": 0.46, + "learning_rate": 1.1766055856264933e-05, + "loss": 0.9609, + "step": 14390 + }, + { + "epoch": 0.46, + "learning_rate": 1.176503632415478e-05, + "loss": 0.9292, + "step": 14391 + }, + { + "epoch": 0.46, + "learning_rate": 1.1764016773107738e-05, + "loss": 1.0225, + "step": 14392 + }, + { + "epoch": 0.46, + "learning_rate": 1.1762997203134754e-05, + "loss": 0.9045, + "step": 14393 + }, + { + "epoch": 0.46, + "learning_rate": 1.1761977614246757e-05, + "loss": 1.2139, + "step": 14394 + }, + { + "epoch": 0.46, + "learning_rate": 1.1760958006454692e-05, + "loss": 1.0698, + "step": 14395 + }, + { + "epoch": 0.46, + "learning_rate": 1.1759938379769495e-05, + "loss": 1.0493, + "step": 14396 + }, + { + "epoch": 0.46, + "learning_rate": 1.1758918734202109e-05, + "loss": 1.0332, + "step": 14397 + }, + { + "epoch": 0.46, + "learning_rate": 1.175789906976347e-05, + "loss": 0.9692, + "step": 14398 + }, + { + "epoch": 0.46, + "learning_rate": 1.1756879386464518e-05, + "loss": 1.1387, + "step": 14399 + }, + { + "epoch": 0.46, + "learning_rate": 1.17558596843162e-05, + "loss": 1.0757, + "step": 14400 + }, + { + "epoch": 0.46, + "learning_rate": 1.1754839963329447e-05, + "loss": 1.0571, + "step": 14401 + }, + { + "epoch": 0.46, + "learning_rate": 1.1753820223515204e-05, + "loss": 0.9873, + "step": 14402 + }, + { + "epoch": 0.46, + "learning_rate": 1.1752800464884411e-05, + "loss": 0.8413, + "step": 14403 + }, + { + "epoch": 0.46, + "learning_rate": 1.1751780687448008e-05, + "loss": 0.958, + "step": 14404 + }, + { + "epoch": 0.46, + "learning_rate": 1.1750760891216939e-05, + "loss": 1.1636, + "step": 14405 + }, + { + "epoch": 0.46, + "learning_rate": 1.1749741076202141e-05, + "loss": 0.918, + "step": 14406 + }, + { + "epoch": 0.46, + "learning_rate": 1.174872124241456e-05, + "loss": 1.0986, + "step": 14407 + }, + { + "epoch": 0.46, + "learning_rate": 1.1747701389865136e-05, + "loss": 1.1094, + "step": 14408 + }, + { + "epoch": 0.46, + "learning_rate": 1.1746681518564809e-05, + "loss": 0.9683, + "step": 14409 + }, + { + "epoch": 0.46, + "learning_rate": 1.174566162852452e-05, + "loss": 1.0645, + "step": 14410 + }, + { + "epoch": 0.46, + "learning_rate": 1.1744641719755217e-05, + "loss": 1.0142, + "step": 14411 + }, + { + "epoch": 0.46, + "learning_rate": 1.1743621792267839e-05, + "loss": 0.938, + "step": 14412 + }, + { + "epoch": 0.46, + "learning_rate": 1.1742601846073328e-05, + "loss": 0.8994, + "step": 14413 + }, + { + "epoch": 0.46, + "learning_rate": 1.1741581881182626e-05, + "loss": 0.9551, + "step": 14414 + }, + { + "epoch": 0.46, + "learning_rate": 1.174056189760668e-05, + "loss": 1.0801, + "step": 14415 + }, + { + "epoch": 0.46, + "learning_rate": 1.1739541895356432e-05, + "loss": 1.0137, + "step": 14416 + }, + { + "epoch": 0.46, + "learning_rate": 1.1738521874442824e-05, + "loss": 1.0122, + "step": 14417 + }, + { + "epoch": 0.46, + "learning_rate": 1.1737501834876798e-05, + "loss": 0.9702, + "step": 14418 + }, + { + "epoch": 0.46, + "learning_rate": 1.1736481776669307e-05, + "loss": 0.8726, + "step": 14419 + }, + { + "epoch": 0.46, + "learning_rate": 1.173546169983128e-05, + "loss": 1.0342, + "step": 14420 + }, + { + "epoch": 0.46, + "learning_rate": 1.1734441604373674e-05, + "loss": 1.0508, + "step": 14421 + }, + { + "epoch": 0.46, + "learning_rate": 1.1733421490307428e-05, + "loss": 0.9497, + "step": 14422 + }, + { + "epoch": 0.46, + "learning_rate": 1.173240135764349e-05, + "loss": 1.0425, + "step": 14423 + }, + { + "epoch": 0.46, + "learning_rate": 1.1731381206392802e-05, + "loss": 0.9531, + "step": 14424 + }, + { + "epoch": 0.46, + "learning_rate": 1.173036103656631e-05, + "loss": 0.9258, + "step": 14425 + }, + { + "epoch": 0.46, + "learning_rate": 1.1729340848174957e-05, + "loss": 1.0415, + "step": 14426 + }, + { + "epoch": 0.46, + "learning_rate": 1.1728320641229691e-05, + "loss": 1.02, + "step": 14427 + }, + { + "epoch": 0.46, + "learning_rate": 1.172730041574146e-05, + "loss": 1.0396, + "step": 14428 + }, + { + "epoch": 0.46, + "learning_rate": 1.1726280171721205e-05, + "loss": 0.9805, + "step": 14429 + }, + { + "epoch": 0.46, + "learning_rate": 1.1725259909179875e-05, + "loss": 0.9702, + "step": 14430 + }, + { + "epoch": 0.46, + "learning_rate": 1.1724239628128417e-05, + "loss": 0.978, + "step": 14431 + }, + { + "epoch": 0.46, + "learning_rate": 1.1723219328577774e-05, + "loss": 0.9878, + "step": 14432 + }, + { + "epoch": 0.46, + "learning_rate": 1.1722199010538896e-05, + "loss": 1.0078, + "step": 14433 + }, + { + "epoch": 0.46, + "learning_rate": 1.1721178674022728e-05, + "loss": 1.0103, + "step": 14434 + }, + { + "epoch": 0.46, + "learning_rate": 1.1720158319040216e-05, + "loss": 0.9702, + "step": 14435 + }, + { + "epoch": 0.46, + "learning_rate": 1.171913794560231e-05, + "loss": 1.0039, + "step": 14436 + }, + { + "epoch": 0.46, + "learning_rate": 1.1718117553719956e-05, + "loss": 0.9507, + "step": 14437 + }, + { + "epoch": 0.46, + "learning_rate": 1.1717097143404106e-05, + "loss": 0.9878, + "step": 14438 + }, + { + "epoch": 0.46, + "learning_rate": 1.1716076714665701e-05, + "loss": 0.9771, + "step": 14439 + }, + { + "epoch": 0.46, + "learning_rate": 1.1715056267515692e-05, + "loss": 1.0718, + "step": 14440 + }, + { + "epoch": 0.46, + "learning_rate": 1.1714035801965027e-05, + "loss": 0.978, + "step": 14441 + }, + { + "epoch": 0.46, + "learning_rate": 1.1713015318024656e-05, + "loss": 1.0342, + "step": 14442 + }, + { + "epoch": 0.46, + "learning_rate": 1.1711994815705528e-05, + "loss": 1.0137, + "step": 14443 + }, + { + "epoch": 0.46, + "learning_rate": 1.1710974295018586e-05, + "loss": 1.0474, + "step": 14444 + }, + { + "epoch": 0.46, + "learning_rate": 1.1709953755974786e-05, + "loss": 0.9028, + "step": 14445 + }, + { + "epoch": 0.46, + "learning_rate": 1.1708933198585075e-05, + "loss": 0.9731, + "step": 14446 + }, + { + "epoch": 0.46, + "learning_rate": 1.1707912622860402e-05, + "loss": 1.0415, + "step": 14447 + }, + { + "epoch": 0.46, + "learning_rate": 1.1706892028811717e-05, + "loss": 1.0288, + "step": 14448 + }, + { + "epoch": 0.46, + "learning_rate": 1.1705871416449969e-05, + "loss": 1.1147, + "step": 14449 + }, + { + "epoch": 0.46, + "learning_rate": 1.1704850785786108e-05, + "loss": 1.0859, + "step": 14450 + }, + { + "epoch": 0.46, + "learning_rate": 1.1703830136831086e-05, + "loss": 0.9561, + "step": 14451 + }, + { + "epoch": 0.46, + "learning_rate": 1.1702809469595852e-05, + "loss": 1.0601, + "step": 14452 + }, + { + "epoch": 0.46, + "learning_rate": 1.170178878409136e-05, + "loss": 0.9741, + "step": 14453 + }, + { + "epoch": 0.46, + "learning_rate": 1.1700768080328553e-05, + "loss": 1.0459, + "step": 14454 + }, + { + "epoch": 0.46, + "learning_rate": 1.1699747358318387e-05, + "loss": 1.0479, + "step": 14455 + }, + { + "epoch": 0.46, + "learning_rate": 1.1698726618071813e-05, + "loss": 1.0239, + "step": 14456 + }, + { + "epoch": 0.46, + "learning_rate": 1.1697705859599789e-05, + "loss": 1.0068, + "step": 14457 + }, + { + "epoch": 0.46, + "learning_rate": 1.1696685082913253e-05, + "loss": 0.9419, + "step": 14458 + }, + { + "epoch": 0.46, + "learning_rate": 1.1695664288023164e-05, + "loss": 1.1401, + "step": 14459 + }, + { + "epoch": 0.46, + "learning_rate": 1.1694643474940477e-05, + "loss": 1.0239, + "step": 14460 + }, + { + "epoch": 0.46, + "learning_rate": 1.1693622643676142e-05, + "loss": 1.0073, + "step": 14461 + }, + { + "epoch": 0.46, + "learning_rate": 1.1692601794241107e-05, + "loss": 1.0962, + "step": 14462 + }, + { + "epoch": 0.46, + "learning_rate": 1.1691580926646328e-05, + "loss": 0.9126, + "step": 14463 + }, + { + "epoch": 0.46, + "learning_rate": 1.169056004090276e-05, + "loss": 0.9883, + "step": 14464 + }, + { + "epoch": 0.46, + "learning_rate": 1.1689539137021356e-05, + "loss": 1.0435, + "step": 14465 + }, + { + "epoch": 0.46, + "learning_rate": 1.1688518215013063e-05, + "loss": 1.0913, + "step": 14466 + }, + { + "epoch": 0.46, + "learning_rate": 1.1687497274888839e-05, + "loss": 0.98, + "step": 14467 + }, + { + "epoch": 0.46, + "learning_rate": 1.1686476316659637e-05, + "loss": 1.0435, + "step": 14468 + }, + { + "epoch": 0.46, + "learning_rate": 1.168545534033641e-05, + "loss": 0.9409, + "step": 14469 + }, + { + "epoch": 0.46, + "learning_rate": 1.1684434345930115e-05, + "loss": 1.0918, + "step": 14470 + }, + { + "epoch": 0.46, + "learning_rate": 1.16834133334517e-05, + "loss": 1.0474, + "step": 14471 + }, + { + "epoch": 0.46, + "learning_rate": 1.1682392302912126e-05, + "loss": 1.0522, + "step": 14472 + }, + { + "epoch": 0.46, + "learning_rate": 1.1681371254322344e-05, + "loss": 0.9956, + "step": 14473 + }, + { + "epoch": 0.46, + "learning_rate": 1.168035018769331e-05, + "loss": 0.9395, + "step": 14474 + }, + { + "epoch": 0.46, + "learning_rate": 1.1679329103035977e-05, + "loss": 1.0728, + "step": 14475 + }, + { + "epoch": 0.46, + "learning_rate": 1.1678308000361302e-05, + "loss": 1.1191, + "step": 14476 + }, + { + "epoch": 0.46, + "learning_rate": 1.1677286879680239e-05, + "loss": 1.0776, + "step": 14477 + }, + { + "epoch": 0.46, + "learning_rate": 1.1676265741003743e-05, + "loss": 0.877, + "step": 14478 + }, + { + "epoch": 0.46, + "learning_rate": 1.167524458434277e-05, + "loss": 0.937, + "step": 14479 + }, + { + "epoch": 0.46, + "learning_rate": 1.1674223409708284e-05, + "loss": 1.0005, + "step": 14480 + }, + { + "epoch": 0.46, + "learning_rate": 1.1673202217111227e-05, + "loss": 1.1323, + "step": 14481 + }, + { + "epoch": 0.46, + "learning_rate": 1.1672181006562562e-05, + "loss": 1.0332, + "step": 14482 + }, + { + "epoch": 0.46, + "learning_rate": 1.167115977807325e-05, + "loss": 0.9629, + "step": 14483 + }, + { + "epoch": 0.46, + "learning_rate": 1.1670138531654238e-05, + "loss": 1.0474, + "step": 14484 + }, + { + "epoch": 0.46, + "learning_rate": 1.166911726731649e-05, + "loss": 0.9414, + "step": 14485 + }, + { + "epoch": 0.46, + "learning_rate": 1.166809598507096e-05, + "loss": 0.957, + "step": 14486 + }, + { + "epoch": 0.46, + "learning_rate": 1.1667074684928608e-05, + "loss": 1.1191, + "step": 14487 + }, + { + "epoch": 0.46, + "learning_rate": 1.166605336690039e-05, + "loss": 0.9746, + "step": 14488 + }, + { + "epoch": 0.46, + "learning_rate": 1.166503203099726e-05, + "loss": 0.8916, + "step": 14489 + }, + { + "epoch": 0.46, + "learning_rate": 1.166401067723018e-05, + "loss": 0.918, + "step": 14490 + }, + { + "epoch": 0.46, + "learning_rate": 1.166298930561011e-05, + "loss": 0.9243, + "step": 14491 + }, + { + "epoch": 0.46, + "learning_rate": 1.1661967916148004e-05, + "loss": 0.9614, + "step": 14492 + }, + { + "epoch": 0.46, + "learning_rate": 1.166094650885482e-05, + "loss": 1.0601, + "step": 14493 + }, + { + "epoch": 0.46, + "learning_rate": 1.165992508374152e-05, + "loss": 0.8994, + "step": 14494 + }, + { + "epoch": 0.46, + "learning_rate": 1.1658903640819062e-05, + "loss": 0.9346, + "step": 14495 + }, + { + "epoch": 0.46, + "learning_rate": 1.1657882180098401e-05, + "loss": 1.1055, + "step": 14496 + }, + { + "epoch": 0.46, + "learning_rate": 1.16568607015905e-05, + "loss": 0.978, + "step": 14497 + }, + { + "epoch": 0.46, + "learning_rate": 1.1655839205306317e-05, + "loss": 1.0933, + "step": 14498 + }, + { + "epoch": 0.46, + "learning_rate": 1.1654817691256814e-05, + "loss": 1.0298, + "step": 14499 + }, + { + "epoch": 0.46, + "learning_rate": 1.1653796159452945e-05, + "loss": 1.041, + "step": 14500 + }, + { + "epoch": 0.46, + "learning_rate": 1.1652774609905677e-05, + "loss": 0.9336, + "step": 14501 + }, + { + "epoch": 0.46, + "learning_rate": 1.1651753042625966e-05, + "loss": 0.9282, + "step": 14502 + }, + { + "epoch": 0.46, + "learning_rate": 1.1650731457624773e-05, + "loss": 0.9722, + "step": 14503 + }, + { + "epoch": 0.46, + "learning_rate": 1.1649709854913058e-05, + "loss": 1.0493, + "step": 14504 + }, + { + "epoch": 0.46, + "learning_rate": 1.1648688234501781e-05, + "loss": 1.0249, + "step": 14505 + }, + { + "epoch": 0.46, + "learning_rate": 1.1647666596401906e-05, + "loss": 0.9863, + "step": 14506 + }, + { + "epoch": 0.46, + "learning_rate": 1.1646644940624389e-05, + "loss": 0.9453, + "step": 14507 + }, + { + "epoch": 0.46, + "learning_rate": 1.1645623267180198e-05, + "loss": 0.9395, + "step": 14508 + }, + { + "epoch": 0.46, + "learning_rate": 1.164460157608029e-05, + "loss": 1.0337, + "step": 14509 + }, + { + "epoch": 0.46, + "learning_rate": 1.1643579867335623e-05, + "loss": 0.9385, + "step": 14510 + }, + { + "epoch": 0.46, + "learning_rate": 1.1642558140957165e-05, + "loss": 0.9229, + "step": 14511 + }, + { + "epoch": 0.46, + "learning_rate": 1.1641536396955879e-05, + "loss": 0.9429, + "step": 14512 + }, + { + "epoch": 0.46, + "learning_rate": 1.1640514635342724e-05, + "loss": 1.002, + "step": 14513 + }, + { + "epoch": 0.46, + "learning_rate": 1.163949285612866e-05, + "loss": 0.8755, + "step": 14514 + }, + { + "epoch": 0.46, + "learning_rate": 1.1638471059324654e-05, + "loss": 0.5017, + "step": 14515 + }, + { + "epoch": 0.46, + "learning_rate": 1.1637449244941666e-05, + "loss": 1.1069, + "step": 14516 + }, + { + "epoch": 0.46, + "learning_rate": 1.1636427412990661e-05, + "loss": 1.0015, + "step": 14517 + }, + { + "epoch": 0.46, + "learning_rate": 1.1635405563482597e-05, + "loss": 0.9438, + "step": 14518 + }, + { + "epoch": 0.46, + "learning_rate": 1.1634383696428445e-05, + "loss": 1.0547, + "step": 14519 + }, + { + "epoch": 0.46, + "learning_rate": 1.1633361811839161e-05, + "loss": 0.8628, + "step": 14520 + }, + { + "epoch": 0.46, + "learning_rate": 1.1632339909725719e-05, + "loss": 1.0786, + "step": 14521 + }, + { + "epoch": 0.46, + "learning_rate": 1.1631317990099072e-05, + "loss": 0.5786, + "step": 14522 + }, + { + "epoch": 0.46, + "learning_rate": 1.1630296052970186e-05, + "loss": 0.999, + "step": 14523 + }, + { + "epoch": 0.46, + "learning_rate": 1.1629274098350032e-05, + "loss": 1.0537, + "step": 14524 + }, + { + "epoch": 0.46, + "learning_rate": 1.162825212624957e-05, + "loss": 0.936, + "step": 14525 + }, + { + "epoch": 0.46, + "learning_rate": 1.1627230136679764e-05, + "loss": 0.9038, + "step": 14526 + }, + { + "epoch": 0.46, + "learning_rate": 1.1626208129651577e-05, + "loss": 1.0396, + "step": 14527 + }, + { + "epoch": 0.46, + "learning_rate": 1.1625186105175978e-05, + "loss": 1.0098, + "step": 14528 + }, + { + "epoch": 0.46, + "learning_rate": 1.1624164063263931e-05, + "loss": 0.9575, + "step": 14529 + }, + { + "epoch": 0.46, + "learning_rate": 1.1623142003926401e-05, + "loss": 1.0146, + "step": 14530 + }, + { + "epoch": 0.46, + "learning_rate": 1.1622119927174355e-05, + "loss": 0.937, + "step": 14531 + }, + { + "epoch": 0.46, + "learning_rate": 1.1621097833018755e-05, + "loss": 1.0249, + "step": 14532 + }, + { + "epoch": 0.46, + "learning_rate": 1.1620075721470574e-05, + "loss": 1.084, + "step": 14533 + }, + { + "epoch": 0.46, + "learning_rate": 1.161905359254077e-05, + "loss": 1.1421, + "step": 14534 + }, + { + "epoch": 0.46, + "learning_rate": 1.1618031446240312e-05, + "loss": 1.0356, + "step": 14535 + }, + { + "epoch": 0.46, + "learning_rate": 1.1617009282580171e-05, + "loss": 0.894, + "step": 14536 + }, + { + "epoch": 0.46, + "learning_rate": 1.1615987101571307e-05, + "loss": 0.9458, + "step": 14537 + }, + { + "epoch": 0.46, + "learning_rate": 1.161496490322469e-05, + "loss": 1.0425, + "step": 14538 + }, + { + "epoch": 0.46, + "learning_rate": 1.1613942687551287e-05, + "loss": 0.9487, + "step": 14539 + }, + { + "epoch": 0.46, + "learning_rate": 1.1612920454562067e-05, + "loss": 0.9531, + "step": 14540 + }, + { + "epoch": 0.47, + "learning_rate": 1.1611898204267995e-05, + "loss": 0.9082, + "step": 14541 + }, + { + "epoch": 0.47, + "learning_rate": 1.1610875936680037e-05, + "loss": 0.9116, + "step": 14542 + }, + { + "epoch": 0.47, + "learning_rate": 1.1609853651809165e-05, + "loss": 0.9307, + "step": 14543 + }, + { + "epoch": 0.47, + "learning_rate": 1.1608831349666345e-05, + "loss": 0.8936, + "step": 14544 + }, + { + "epoch": 0.47, + "learning_rate": 1.1607809030262545e-05, + "loss": 1.1289, + "step": 14545 + }, + { + "epoch": 0.47, + "learning_rate": 1.1606786693608733e-05, + "loss": 1.0107, + "step": 14546 + }, + { + "epoch": 0.47, + "learning_rate": 1.1605764339715877e-05, + "loss": 0.9634, + "step": 14547 + }, + { + "epoch": 0.47, + "learning_rate": 1.160474196859495e-05, + "loss": 0.5632, + "step": 14548 + }, + { + "epoch": 0.47, + "learning_rate": 1.1603719580256916e-05, + "loss": 0.9199, + "step": 14549 + }, + { + "epoch": 0.47, + "learning_rate": 1.1602697174712744e-05, + "loss": 0.9131, + "step": 14550 + }, + { + "epoch": 0.47, + "learning_rate": 1.160167475197341e-05, + "loss": 1.0308, + "step": 14551 + }, + { + "epoch": 0.47, + "learning_rate": 1.1600652312049875e-05, + "loss": 1.0742, + "step": 14552 + }, + { + "epoch": 0.47, + "learning_rate": 1.1599629854953114e-05, + "loss": 1.1738, + "step": 14553 + }, + { + "epoch": 0.47, + "learning_rate": 1.1598607380694093e-05, + "loss": 1.0034, + "step": 14554 + }, + { + "epoch": 0.47, + "learning_rate": 1.1597584889283784e-05, + "loss": 1.0845, + "step": 14555 + }, + { + "epoch": 0.47, + "learning_rate": 1.1596562380733159e-05, + "loss": 0.9575, + "step": 14556 + }, + { + "epoch": 0.47, + "learning_rate": 1.1595539855053187e-05, + "loss": 1.0425, + "step": 14557 + }, + { + "epoch": 0.47, + "learning_rate": 1.1594517312254834e-05, + "loss": 0.918, + "step": 14558 + }, + { + "epoch": 0.47, + "learning_rate": 1.1593494752349079e-05, + "loss": 1.0674, + "step": 14559 + }, + { + "epoch": 0.47, + "learning_rate": 1.1592472175346887e-05, + "loss": 0.8662, + "step": 14560 + }, + { + "epoch": 0.47, + "learning_rate": 1.1591449581259231e-05, + "loss": 0.9634, + "step": 14561 + }, + { + "epoch": 0.47, + "learning_rate": 1.1590426970097085e-05, + "loss": 0.9951, + "step": 14562 + }, + { + "epoch": 0.47, + "learning_rate": 1.1589404341871416e-05, + "loss": 0.9121, + "step": 14563 + }, + { + "epoch": 0.47, + "learning_rate": 1.1588381696593196e-05, + "loss": 1.0728, + "step": 14564 + }, + { + "epoch": 0.47, + "learning_rate": 1.15873590342734e-05, + "loss": 0.9858, + "step": 14565 + }, + { + "epoch": 0.47, + "learning_rate": 1.1586336354923e-05, + "loss": 0.9756, + "step": 14566 + }, + { + "epoch": 0.47, + "learning_rate": 1.1585313658552964e-05, + "loss": 0.9341, + "step": 14567 + }, + { + "epoch": 0.47, + "learning_rate": 1.1584290945174265e-05, + "loss": 1.0176, + "step": 14568 + }, + { + "epoch": 0.47, + "learning_rate": 1.1583268214797881e-05, + "loss": 1.0249, + "step": 14569 + }, + { + "epoch": 0.47, + "learning_rate": 1.1582245467434783e-05, + "loss": 0.9526, + "step": 14570 + }, + { + "epoch": 0.47, + "learning_rate": 1.1581222703095938e-05, + "loss": 0.9478, + "step": 14571 + }, + { + "epoch": 0.47, + "learning_rate": 1.1580199921792326e-05, + "loss": 0.8633, + "step": 14572 + }, + { + "epoch": 0.47, + "learning_rate": 1.1579177123534915e-05, + "loss": 1.0264, + "step": 14573 + }, + { + "epoch": 0.47, + "learning_rate": 1.1578154308334683e-05, + "loss": 0.981, + "step": 14574 + }, + { + "epoch": 0.47, + "learning_rate": 1.1577131476202603e-05, + "loss": 1.1001, + "step": 14575 + }, + { + "epoch": 0.47, + "learning_rate": 1.1576108627149646e-05, + "loss": 1.0132, + "step": 14576 + }, + { + "epoch": 0.47, + "learning_rate": 1.1575085761186788e-05, + "loss": 1.0679, + "step": 14577 + }, + { + "epoch": 0.47, + "learning_rate": 1.1574062878325005e-05, + "loss": 1.0781, + "step": 14578 + }, + { + "epoch": 0.47, + "learning_rate": 1.1573039978575267e-05, + "loss": 0.5466, + "step": 14579 + }, + { + "epoch": 0.47, + "learning_rate": 1.1572017061948553e-05, + "loss": 1.0015, + "step": 14580 + }, + { + "epoch": 0.47, + "learning_rate": 1.1570994128455833e-05, + "loss": 0.8081, + "step": 14581 + }, + { + "epoch": 0.47, + "learning_rate": 1.1569971178108088e-05, + "loss": 1.1421, + "step": 14582 + }, + { + "epoch": 0.47, + "learning_rate": 1.1568948210916286e-05, + "loss": 0.8618, + "step": 14583 + }, + { + "epoch": 0.47, + "learning_rate": 1.156792522689141e-05, + "loss": 0.5488, + "step": 14584 + }, + { + "epoch": 0.47, + "learning_rate": 1.156690222604443e-05, + "loss": 0.9858, + "step": 14585 + }, + { + "epoch": 0.47, + "learning_rate": 1.1565879208386324e-05, + "loss": 1.04, + "step": 14586 + }, + { + "epoch": 0.47, + "learning_rate": 1.1564856173928065e-05, + "loss": 1.0459, + "step": 14587 + }, + { + "epoch": 0.47, + "learning_rate": 1.1563833122680631e-05, + "loss": 1.0098, + "step": 14588 + }, + { + "epoch": 0.47, + "learning_rate": 1.1562810054655002e-05, + "loss": 0.9644, + "step": 14589 + }, + { + "epoch": 0.47, + "learning_rate": 1.1561786969862147e-05, + "loss": 0.8184, + "step": 14590 + }, + { + "epoch": 0.47, + "learning_rate": 1.1560763868313047e-05, + "loss": 1.0234, + "step": 14591 + }, + { + "epoch": 0.47, + "learning_rate": 1.1559740750018679e-05, + "loss": 0.9434, + "step": 14592 + }, + { + "epoch": 0.47, + "learning_rate": 1.1558717614990021e-05, + "loss": 0.9438, + "step": 14593 + }, + { + "epoch": 0.47, + "learning_rate": 1.1557694463238044e-05, + "loss": 0.9302, + "step": 14594 + }, + { + "epoch": 0.47, + "learning_rate": 1.1556671294773733e-05, + "loss": 1.0674, + "step": 14595 + }, + { + "epoch": 0.47, + "learning_rate": 1.1555648109608062e-05, + "loss": 1.0366, + "step": 14596 + }, + { + "epoch": 0.47, + "learning_rate": 1.1554624907752006e-05, + "loss": 1.0068, + "step": 14597 + }, + { + "epoch": 0.47, + "learning_rate": 1.1553601689216547e-05, + "loss": 0.8569, + "step": 14598 + }, + { + "epoch": 0.47, + "learning_rate": 1.155257845401266e-05, + "loss": 0.9668, + "step": 14599 + }, + { + "epoch": 0.47, + "learning_rate": 1.1551555202151323e-05, + "loss": 1.0303, + "step": 14600 + }, + { + "epoch": 0.47, + "learning_rate": 1.155053193364352e-05, + "loss": 1.1182, + "step": 14601 + }, + { + "epoch": 0.47, + "learning_rate": 1.1549508648500222e-05, + "loss": 1.1455, + "step": 14602 + }, + { + "epoch": 0.47, + "learning_rate": 1.1548485346732412e-05, + "loss": 1.0703, + "step": 14603 + }, + { + "epoch": 0.47, + "learning_rate": 1.154746202835107e-05, + "loss": 1.0532, + "step": 14604 + }, + { + "epoch": 0.47, + "learning_rate": 1.154643869336717e-05, + "loss": 0.9458, + "step": 14605 + }, + { + "epoch": 0.47, + "learning_rate": 1.1545415341791695e-05, + "loss": 1.0312, + "step": 14606 + }, + { + "epoch": 0.47, + "learning_rate": 1.1544391973635624e-05, + "loss": 0.8564, + "step": 14607 + }, + { + "epoch": 0.47, + "learning_rate": 1.1543368588909937e-05, + "loss": 1.002, + "step": 14608 + }, + { + "epoch": 0.47, + "learning_rate": 1.1542345187625614e-05, + "loss": 0.9746, + "step": 14609 + }, + { + "epoch": 0.47, + "learning_rate": 1.1541321769793632e-05, + "loss": 1.0273, + "step": 14610 + }, + { + "epoch": 0.47, + "learning_rate": 1.1540298335424972e-05, + "loss": 0.9321, + "step": 14611 + }, + { + "epoch": 0.47, + "learning_rate": 1.1539274884530622e-05, + "loss": 1.0537, + "step": 14612 + }, + { + "epoch": 0.47, + "learning_rate": 1.1538251417121551e-05, + "loss": 0.9888, + "step": 14613 + }, + { + "epoch": 0.47, + "learning_rate": 1.1537227933208745e-05, + "loss": 0.9937, + "step": 14614 + }, + { + "epoch": 0.47, + "learning_rate": 1.1536204432803184e-05, + "loss": 0.9321, + "step": 14615 + }, + { + "epoch": 0.47, + "learning_rate": 1.1535180915915853e-05, + "loss": 0.9722, + "step": 14616 + }, + { + "epoch": 0.47, + "learning_rate": 1.1534157382557727e-05, + "loss": 1.0566, + "step": 14617 + }, + { + "epoch": 0.47, + "learning_rate": 1.1533133832739791e-05, + "loss": 1.0938, + "step": 14618 + }, + { + "epoch": 0.47, + "learning_rate": 1.1532110266473026e-05, + "loss": 0.9858, + "step": 14619 + }, + { + "epoch": 0.47, + "learning_rate": 1.1531086683768413e-05, + "loss": 1.0366, + "step": 14620 + }, + { + "epoch": 0.47, + "learning_rate": 1.1530063084636935e-05, + "loss": 1.1641, + "step": 14621 + }, + { + "epoch": 0.47, + "learning_rate": 1.1529039469089573e-05, + "loss": 1.0298, + "step": 14622 + }, + { + "epoch": 0.47, + "learning_rate": 1.152801583713731e-05, + "loss": 1.105, + "step": 14623 + }, + { + "epoch": 0.47, + "learning_rate": 1.1526992188791128e-05, + "loss": 1.0166, + "step": 14624 + }, + { + "epoch": 0.47, + "learning_rate": 1.152596852406201e-05, + "loss": 1.1016, + "step": 14625 + }, + { + "epoch": 0.47, + "learning_rate": 1.1524944842960937e-05, + "loss": 1.0161, + "step": 14626 + }, + { + "epoch": 0.47, + "learning_rate": 1.1523921145498897e-05, + "loss": 1.0806, + "step": 14627 + }, + { + "epoch": 0.47, + "learning_rate": 1.1522897431686868e-05, + "loss": 0.875, + "step": 14628 + }, + { + "epoch": 0.47, + "learning_rate": 1.1521873701535833e-05, + "loss": 0.9165, + "step": 14629 + }, + { + "epoch": 0.47, + "learning_rate": 1.1520849955056779e-05, + "loss": 1.0811, + "step": 14630 + }, + { + "epoch": 0.47, + "learning_rate": 1.1519826192260691e-05, + "loss": 0.999, + "step": 14631 + }, + { + "epoch": 0.47, + "learning_rate": 1.1518802413158546e-05, + "loss": 0.9414, + "step": 14632 + }, + { + "epoch": 0.47, + "learning_rate": 1.1517778617761333e-05, + "loss": 0.9707, + "step": 14633 + }, + { + "epoch": 0.47, + "learning_rate": 1.1516754806080036e-05, + "loss": 1.1006, + "step": 14634 + }, + { + "epoch": 0.47, + "learning_rate": 1.1515730978125639e-05, + "loss": 0.8555, + "step": 14635 + }, + { + "epoch": 0.47, + "learning_rate": 1.1514707133909125e-05, + "loss": 0.9478, + "step": 14636 + }, + { + "epoch": 0.47, + "learning_rate": 1.151368327344148e-05, + "loss": 1.0835, + "step": 14637 + }, + { + "epoch": 0.47, + "learning_rate": 1.151265939673369e-05, + "loss": 1.02, + "step": 14638 + }, + { + "epoch": 0.47, + "learning_rate": 1.1511635503796736e-05, + "loss": 0.8496, + "step": 14639 + }, + { + "epoch": 0.47, + "learning_rate": 1.151061159464161e-05, + "loss": 0.9604, + "step": 14640 + }, + { + "epoch": 0.47, + "learning_rate": 1.1509587669279292e-05, + "loss": 0.9453, + "step": 14641 + }, + { + "epoch": 0.47, + "learning_rate": 1.150856372772077e-05, + "loss": 0.9629, + "step": 14642 + }, + { + "epoch": 0.47, + "learning_rate": 1.1507539769977025e-05, + "loss": 1.0234, + "step": 14643 + }, + { + "epoch": 0.47, + "learning_rate": 1.150651579605905e-05, + "loss": 0.9473, + "step": 14644 + }, + { + "epoch": 0.47, + "learning_rate": 1.1505491805977825e-05, + "loss": 0.9917, + "step": 14645 + }, + { + "epoch": 0.47, + "learning_rate": 1.1504467799744344e-05, + "loss": 1.0864, + "step": 14646 + }, + { + "epoch": 0.47, + "learning_rate": 1.1503443777369583e-05, + "loss": 0.9028, + "step": 14647 + }, + { + "epoch": 0.47, + "learning_rate": 1.1502419738864536e-05, + "loss": 0.938, + "step": 14648 + }, + { + "epoch": 0.47, + "learning_rate": 1.1501395684240189e-05, + "loss": 0.9932, + "step": 14649 + }, + { + "epoch": 0.47, + "learning_rate": 1.1500371613507529e-05, + "loss": 1.0273, + "step": 14650 + }, + { + "epoch": 0.47, + "learning_rate": 1.149934752667754e-05, + "loss": 1.0566, + "step": 14651 + }, + { + "epoch": 0.47, + "learning_rate": 1.1498323423761211e-05, + "loss": 1.0459, + "step": 14652 + }, + { + "epoch": 0.47, + "learning_rate": 1.1497299304769532e-05, + "loss": 1.0454, + "step": 14653 + }, + { + "epoch": 0.47, + "learning_rate": 1.1496275169713487e-05, + "loss": 0.9038, + "step": 14654 + }, + { + "epoch": 0.47, + "learning_rate": 1.1495251018604065e-05, + "loss": 1.0132, + "step": 14655 + }, + { + "epoch": 0.47, + "learning_rate": 1.1494226851452256e-05, + "loss": 0.9688, + "step": 14656 + }, + { + "epoch": 0.47, + "learning_rate": 1.1493202668269048e-05, + "loss": 1.0469, + "step": 14657 + }, + { + "epoch": 0.47, + "learning_rate": 1.1492178469065424e-05, + "loss": 0.9224, + "step": 14658 + }, + { + "epoch": 0.47, + "learning_rate": 1.149115425385238e-05, + "loss": 0.5957, + "step": 14659 + }, + { + "epoch": 0.47, + "learning_rate": 1.1490130022640897e-05, + "loss": 1.0269, + "step": 14660 + }, + { + "epoch": 0.47, + "learning_rate": 1.1489105775441973e-05, + "loss": 1.04, + "step": 14661 + }, + { + "epoch": 0.47, + "learning_rate": 1.148808151226659e-05, + "loss": 1.0708, + "step": 14662 + }, + { + "epoch": 0.47, + "learning_rate": 1.1487057233125739e-05, + "loss": 1.001, + "step": 14663 + }, + { + "epoch": 0.47, + "learning_rate": 1.1486032938030409e-05, + "loss": 1.1475, + "step": 14664 + }, + { + "epoch": 0.47, + "learning_rate": 1.1485008626991592e-05, + "loss": 1.0908, + "step": 14665 + }, + { + "epoch": 0.47, + "learning_rate": 1.1483984300020274e-05, + "loss": 0.9619, + "step": 14666 + }, + { + "epoch": 0.47, + "learning_rate": 1.1482959957127448e-05, + "loss": 1.125, + "step": 14667 + }, + { + "epoch": 0.47, + "learning_rate": 1.1481935598324103e-05, + "loss": 1.0215, + "step": 14668 + }, + { + "epoch": 0.47, + "learning_rate": 1.148091122362123e-05, + "loss": 0.9404, + "step": 14669 + }, + { + "epoch": 0.47, + "learning_rate": 1.1479886833029816e-05, + "loss": 0.9512, + "step": 14670 + }, + { + "epoch": 0.47, + "learning_rate": 1.1478862426560855e-05, + "loss": 1.1309, + "step": 14671 + }, + { + "epoch": 0.47, + "learning_rate": 1.1477838004225339e-05, + "loss": 1.0322, + "step": 14672 + }, + { + "epoch": 0.47, + "learning_rate": 1.1476813566034255e-05, + "loss": 1.0088, + "step": 14673 + }, + { + "epoch": 0.47, + "learning_rate": 1.1475789111998597e-05, + "loss": 1.0527, + "step": 14674 + }, + { + "epoch": 0.47, + "learning_rate": 1.1474764642129352e-05, + "loss": 0.8926, + "step": 14675 + }, + { + "epoch": 0.47, + "learning_rate": 1.147374015643752e-05, + "loss": 0.9609, + "step": 14676 + }, + { + "epoch": 0.47, + "learning_rate": 1.1472715654934083e-05, + "loss": 1.0063, + "step": 14677 + }, + { + "epoch": 0.47, + "learning_rate": 1.147169113763004e-05, + "loss": 0.9927, + "step": 14678 + }, + { + "epoch": 0.47, + "learning_rate": 1.1470666604536378e-05, + "loss": 0.9053, + "step": 14679 + }, + { + "epoch": 0.47, + "learning_rate": 1.1469642055664089e-05, + "loss": 1.0752, + "step": 14680 + }, + { + "epoch": 0.47, + "learning_rate": 1.146861749102417e-05, + "loss": 0.9634, + "step": 14681 + }, + { + "epoch": 0.47, + "learning_rate": 1.1467592910627612e-05, + "loss": 0.8818, + "step": 14682 + }, + { + "epoch": 0.47, + "learning_rate": 1.1466568314485402e-05, + "loss": 1.0337, + "step": 14683 + }, + { + "epoch": 0.47, + "learning_rate": 1.146554370260854e-05, + "loss": 0.9668, + "step": 14684 + }, + { + "epoch": 0.47, + "learning_rate": 1.1464519075008015e-05, + "loss": 1.0396, + "step": 14685 + }, + { + "epoch": 0.47, + "learning_rate": 1.1463494431694822e-05, + "loss": 1.0205, + "step": 14686 + }, + { + "epoch": 0.47, + "learning_rate": 1.1462469772679956e-05, + "loss": 1.1001, + "step": 14687 + }, + { + "epoch": 0.47, + "learning_rate": 1.1461445097974403e-05, + "loss": 1.0806, + "step": 14688 + }, + { + "epoch": 0.47, + "learning_rate": 1.1460420407589164e-05, + "loss": 1.0454, + "step": 14689 + }, + { + "epoch": 0.47, + "learning_rate": 1.145939570153523e-05, + "loss": 0.9917, + "step": 14690 + }, + { + "epoch": 0.47, + "learning_rate": 1.1458370979823595e-05, + "loss": 1.0664, + "step": 14691 + }, + { + "epoch": 0.47, + "learning_rate": 1.1457346242465253e-05, + "loss": 0.9355, + "step": 14692 + }, + { + "epoch": 0.47, + "learning_rate": 1.1456321489471199e-05, + "loss": 0.9648, + "step": 14693 + }, + { + "epoch": 0.47, + "learning_rate": 1.1455296720852427e-05, + "loss": 0.9287, + "step": 14694 + }, + { + "epoch": 0.47, + "learning_rate": 1.1454271936619935e-05, + "loss": 0.9619, + "step": 14695 + }, + { + "epoch": 0.47, + "learning_rate": 1.145324713678471e-05, + "loss": 0.9805, + "step": 14696 + }, + { + "epoch": 0.47, + "learning_rate": 1.1452222321357753e-05, + "loss": 1.083, + "step": 14697 + }, + { + "epoch": 0.47, + "learning_rate": 1.1451197490350058e-05, + "loss": 0.5173, + "step": 14698 + }, + { + "epoch": 0.47, + "learning_rate": 1.1450172643772622e-05, + "loss": 1.0034, + "step": 14699 + }, + { + "epoch": 0.47, + "learning_rate": 1.1449147781636436e-05, + "loss": 0.9521, + "step": 14700 + }, + { + "epoch": 0.47, + "learning_rate": 1.1448122903952499e-05, + "loss": 0.9751, + "step": 14701 + }, + { + "epoch": 0.47, + "learning_rate": 1.1447098010731808e-05, + "loss": 1.0073, + "step": 14702 + }, + { + "epoch": 0.47, + "learning_rate": 1.1446073101985354e-05, + "loss": 0.9941, + "step": 14703 + }, + { + "epoch": 0.47, + "learning_rate": 1.1445048177724136e-05, + "loss": 1.0249, + "step": 14704 + }, + { + "epoch": 0.47, + "learning_rate": 1.144402323795915e-05, + "loss": 1.0498, + "step": 14705 + }, + { + "epoch": 0.47, + "learning_rate": 1.1442998282701397e-05, + "loss": 0.9702, + "step": 14706 + }, + { + "epoch": 0.47, + "learning_rate": 1.1441973311961865e-05, + "loss": 0.9775, + "step": 14707 + }, + { + "epoch": 0.47, + "learning_rate": 1.1440948325751555e-05, + "loss": 0.939, + "step": 14708 + }, + { + "epoch": 0.47, + "learning_rate": 1.1439923324081465e-05, + "loss": 1.0435, + "step": 14709 + }, + { + "epoch": 0.47, + "learning_rate": 1.1438898306962596e-05, + "loss": 0.9102, + "step": 14710 + }, + { + "epoch": 0.47, + "learning_rate": 1.1437873274405936e-05, + "loss": 1.0278, + "step": 14711 + }, + { + "epoch": 0.47, + "learning_rate": 1.1436848226422488e-05, + "loss": 1.1172, + "step": 14712 + }, + { + "epoch": 0.47, + "learning_rate": 1.1435823163023247e-05, + "loss": 0.9487, + "step": 14713 + }, + { + "epoch": 0.47, + "learning_rate": 1.1434798084219216e-05, + "loss": 1.1021, + "step": 14714 + }, + { + "epoch": 0.47, + "learning_rate": 1.1433772990021388e-05, + "loss": 0.8525, + "step": 14715 + }, + { + "epoch": 0.47, + "learning_rate": 1.1432747880440762e-05, + "loss": 0.8962, + "step": 14716 + }, + { + "epoch": 0.47, + "learning_rate": 1.143172275548834e-05, + "loss": 1.0083, + "step": 14717 + }, + { + "epoch": 0.47, + "learning_rate": 1.1430697615175116e-05, + "loss": 0.938, + "step": 14718 + }, + { + "epoch": 0.47, + "learning_rate": 1.1429672459512087e-05, + "loss": 0.8345, + "step": 14719 + }, + { + "epoch": 0.47, + "learning_rate": 1.1428647288510255e-05, + "loss": 0.8237, + "step": 14720 + }, + { + "epoch": 0.47, + "learning_rate": 1.1427622102180623e-05, + "loss": 0.9263, + "step": 14721 + }, + { + "epoch": 0.47, + "learning_rate": 1.1426596900534182e-05, + "loss": 0.9844, + "step": 14722 + }, + { + "epoch": 0.47, + "learning_rate": 1.1425571683581938e-05, + "loss": 1.0259, + "step": 14723 + }, + { + "epoch": 0.47, + "learning_rate": 1.1424546451334885e-05, + "loss": 0.9951, + "step": 14724 + }, + { + "epoch": 0.47, + "learning_rate": 1.1423521203804027e-05, + "loss": 0.9028, + "step": 14725 + }, + { + "epoch": 0.47, + "learning_rate": 1.1422495941000362e-05, + "loss": 1.0005, + "step": 14726 + }, + { + "epoch": 0.47, + "learning_rate": 1.1421470662934888e-05, + "loss": 0.8833, + "step": 14727 + }, + { + "epoch": 0.47, + "learning_rate": 1.1420445369618606e-05, + "loss": 0.53, + "step": 14728 + }, + { + "epoch": 0.47, + "learning_rate": 1.1419420061062523e-05, + "loss": 0.9639, + "step": 14729 + }, + { + "epoch": 0.47, + "learning_rate": 1.1418394737277629e-05, + "loss": 0.5161, + "step": 14730 + }, + { + "epoch": 0.47, + "learning_rate": 1.141736939827493e-05, + "loss": 0.9424, + "step": 14731 + }, + { + "epoch": 0.47, + "learning_rate": 1.1416344044065427e-05, + "loss": 1.1519, + "step": 14732 + }, + { + "epoch": 0.47, + "learning_rate": 1.1415318674660118e-05, + "loss": 1.0239, + "step": 14733 + }, + { + "epoch": 0.47, + "learning_rate": 1.1414293290070008e-05, + "loss": 1.0015, + "step": 14734 + }, + { + "epoch": 0.47, + "learning_rate": 1.1413267890306095e-05, + "loss": 0.5625, + "step": 14735 + }, + { + "epoch": 0.47, + "learning_rate": 1.1412242475379385e-05, + "loss": 1.1411, + "step": 14736 + }, + { + "epoch": 0.47, + "learning_rate": 1.141121704530087e-05, + "loss": 1.0005, + "step": 14737 + }, + { + "epoch": 0.47, + "learning_rate": 1.1410191600081562e-05, + "loss": 1.0894, + "step": 14738 + }, + { + "epoch": 0.47, + "learning_rate": 1.1409166139732457e-05, + "loss": 0.8594, + "step": 14739 + }, + { + "epoch": 0.47, + "learning_rate": 1.1408140664264562e-05, + "loss": 1.0474, + "step": 14740 + }, + { + "epoch": 0.47, + "learning_rate": 1.1407115173688874e-05, + "loss": 1.0098, + "step": 14741 + }, + { + "epoch": 0.47, + "learning_rate": 1.1406089668016399e-05, + "loss": 0.9424, + "step": 14742 + }, + { + "epoch": 0.47, + "learning_rate": 1.1405064147258137e-05, + "loss": 1.0098, + "step": 14743 + }, + { + "epoch": 0.47, + "learning_rate": 1.140403861142509e-05, + "loss": 0.939, + "step": 14744 + }, + { + "epoch": 0.47, + "learning_rate": 1.1403013060528266e-05, + "loss": 0.9575, + "step": 14745 + }, + { + "epoch": 0.47, + "learning_rate": 1.1401987494578662e-05, + "loss": 1.0078, + "step": 14746 + }, + { + "epoch": 0.47, + "learning_rate": 1.1400961913587284e-05, + "loss": 0.9253, + "step": 14747 + }, + { + "epoch": 0.47, + "learning_rate": 1.1399936317565136e-05, + "loss": 0.9316, + "step": 14748 + }, + { + "epoch": 0.47, + "learning_rate": 1.1398910706523221e-05, + "loss": 0.9893, + "step": 14749 + }, + { + "epoch": 0.47, + "learning_rate": 1.139788508047254e-05, + "loss": 1.1006, + "step": 14750 + }, + { + "epoch": 0.47, + "learning_rate": 1.1396859439424103e-05, + "loss": 1.1025, + "step": 14751 + }, + { + "epoch": 0.47, + "learning_rate": 1.1395833783388908e-05, + "loss": 1.0039, + "step": 14752 + }, + { + "epoch": 0.47, + "learning_rate": 1.139480811237796e-05, + "loss": 0.5508, + "step": 14753 + }, + { + "epoch": 0.47, + "learning_rate": 1.1393782426402267e-05, + "loss": 1.0547, + "step": 14754 + }, + { + "epoch": 0.47, + "learning_rate": 1.1392756725472832e-05, + "loss": 0.9893, + "step": 14755 + }, + { + "epoch": 0.47, + "learning_rate": 1.1391731009600655e-05, + "loss": 0.9453, + "step": 14756 + }, + { + "epoch": 0.47, + "learning_rate": 1.1390705278796747e-05, + "loss": 1.061, + "step": 14757 + }, + { + "epoch": 0.47, + "learning_rate": 1.138967953307211e-05, + "loss": 0.9292, + "step": 14758 + }, + { + "epoch": 0.47, + "learning_rate": 1.138865377243775e-05, + "loss": 0.9888, + "step": 14759 + }, + { + "epoch": 0.47, + "learning_rate": 1.1387627996904671e-05, + "loss": 1.0771, + "step": 14760 + }, + { + "epoch": 0.47, + "learning_rate": 1.138660220648388e-05, + "loss": 1.02, + "step": 14761 + }, + { + "epoch": 0.47, + "learning_rate": 1.1385576401186384e-05, + "loss": 0.9663, + "step": 14762 + }, + { + "epoch": 0.47, + "learning_rate": 1.1384550581023185e-05, + "loss": 1.0215, + "step": 14763 + }, + { + "epoch": 0.47, + "learning_rate": 1.1383524746005287e-05, + "loss": 1.0806, + "step": 14764 + }, + { + "epoch": 0.47, + "learning_rate": 1.1382498896143705e-05, + "loss": 0.9014, + "step": 14765 + }, + { + "epoch": 0.47, + "learning_rate": 1.138147303144944e-05, + "loss": 0.9775, + "step": 14766 + }, + { + "epoch": 0.47, + "learning_rate": 1.1380447151933493e-05, + "loss": 0.9795, + "step": 14767 + }, + { + "epoch": 0.47, + "learning_rate": 1.1379421257606882e-05, + "loss": 0.9048, + "step": 14768 + }, + { + "epoch": 0.47, + "learning_rate": 1.1378395348480604e-05, + "loss": 0.9814, + "step": 14769 + }, + { + "epoch": 0.47, + "learning_rate": 1.1377369424565672e-05, + "loss": 1.1753, + "step": 14770 + }, + { + "epoch": 0.47, + "learning_rate": 1.1376343485873087e-05, + "loss": 0.9927, + "step": 14771 + }, + { + "epoch": 0.47, + "learning_rate": 1.1375317532413861e-05, + "loss": 1.0479, + "step": 14772 + }, + { + "epoch": 0.47, + "learning_rate": 1.1374291564199001e-05, + "loss": 0.9888, + "step": 14773 + }, + { + "epoch": 0.47, + "learning_rate": 1.1373265581239515e-05, + "loss": 1.0083, + "step": 14774 + }, + { + "epoch": 0.47, + "learning_rate": 1.1372239583546404e-05, + "loss": 0.9888, + "step": 14775 + }, + { + "epoch": 0.47, + "learning_rate": 1.1371213571130683e-05, + "loss": 1.1172, + "step": 14776 + }, + { + "epoch": 0.47, + "learning_rate": 1.1370187544003359e-05, + "loss": 0.8994, + "step": 14777 + }, + { + "epoch": 0.47, + "learning_rate": 1.136916150217544e-05, + "loss": 1.0869, + "step": 14778 + }, + { + "epoch": 0.47, + "learning_rate": 1.1368135445657933e-05, + "loss": 0.9756, + "step": 14779 + }, + { + "epoch": 0.47, + "learning_rate": 1.1367109374461844e-05, + "loss": 0.981, + "step": 14780 + }, + { + "epoch": 0.47, + "learning_rate": 1.1366083288598187e-05, + "loss": 1.085, + "step": 14781 + }, + { + "epoch": 0.47, + "learning_rate": 1.136505718807797e-05, + "loss": 0.9521, + "step": 14782 + }, + { + "epoch": 0.47, + "learning_rate": 1.1364031072912196e-05, + "loss": 0.9785, + "step": 14783 + }, + { + "epoch": 0.47, + "learning_rate": 1.1363004943111881e-05, + "loss": 0.8491, + "step": 14784 + }, + { + "epoch": 0.47, + "learning_rate": 1.136197879868803e-05, + "loss": 0.9556, + "step": 14785 + }, + { + "epoch": 0.47, + "learning_rate": 1.1360952639651657e-05, + "loss": 1.0635, + "step": 14786 + }, + { + "epoch": 0.47, + "learning_rate": 1.1359926466013764e-05, + "loss": 1.0488, + "step": 14787 + }, + { + "epoch": 0.47, + "learning_rate": 1.1358900277785367e-05, + "loss": 0.8589, + "step": 14788 + }, + { + "epoch": 0.47, + "learning_rate": 1.1357874074977475e-05, + "loss": 0.9341, + "step": 14789 + }, + { + "epoch": 0.47, + "learning_rate": 1.1356847857601096e-05, + "loss": 0.9253, + "step": 14790 + }, + { + "epoch": 0.47, + "learning_rate": 1.135582162566724e-05, + "loss": 1.0132, + "step": 14791 + }, + { + "epoch": 0.47, + "learning_rate": 1.1354795379186921e-05, + "loss": 0.9937, + "step": 14792 + }, + { + "epoch": 0.47, + "learning_rate": 1.1353769118171145e-05, + "loss": 1.0781, + "step": 14793 + }, + { + "epoch": 0.47, + "learning_rate": 1.1352742842630926e-05, + "loss": 1.0088, + "step": 14794 + }, + { + "epoch": 0.47, + "learning_rate": 1.1351716552577272e-05, + "loss": 0.9688, + "step": 14795 + }, + { + "epoch": 0.47, + "learning_rate": 1.1350690248021197e-05, + "loss": 0.9395, + "step": 14796 + }, + { + "epoch": 0.47, + "learning_rate": 1.1349663928973711e-05, + "loss": 0.9761, + "step": 14797 + }, + { + "epoch": 0.47, + "learning_rate": 1.1348637595445822e-05, + "loss": 1.0913, + "step": 14798 + }, + { + "epoch": 0.47, + "learning_rate": 1.1347611247448544e-05, + "loss": 0.9194, + "step": 14799 + }, + { + "epoch": 0.47, + "learning_rate": 1.1346584884992893e-05, + "loss": 0.96, + "step": 14800 + }, + { + "epoch": 0.47, + "learning_rate": 1.1345558508089873e-05, + "loss": 0.9399, + "step": 14801 + }, + { + "epoch": 0.47, + "learning_rate": 1.13445321167505e-05, + "loss": 1.0435, + "step": 14802 + }, + { + "epoch": 0.47, + "learning_rate": 1.1343505710985784e-05, + "loss": 1.0103, + "step": 14803 + }, + { + "epoch": 0.47, + "learning_rate": 1.1342479290806742e-05, + "loss": 1.0449, + "step": 14804 + }, + { + "epoch": 0.47, + "learning_rate": 1.1341452856224381e-05, + "loss": 1.0112, + "step": 14805 + }, + { + "epoch": 0.47, + "learning_rate": 1.1340426407249716e-05, + "loss": 1.0215, + "step": 14806 + }, + { + "epoch": 0.47, + "learning_rate": 1.1339399943893761e-05, + "loss": 0.9697, + "step": 14807 + }, + { + "epoch": 0.47, + "learning_rate": 1.1338373466167524e-05, + "loss": 1.0757, + "step": 14808 + }, + { + "epoch": 0.47, + "learning_rate": 1.1337346974082023e-05, + "loss": 1.0225, + "step": 14809 + }, + { + "epoch": 0.47, + "learning_rate": 1.1336320467648269e-05, + "loss": 1.0874, + "step": 14810 + }, + { + "epoch": 0.47, + "learning_rate": 1.1335293946877273e-05, + "loss": 0.9336, + "step": 14811 + }, + { + "epoch": 0.47, + "learning_rate": 1.1334267411780053e-05, + "loss": 0.9194, + "step": 14812 + }, + { + "epoch": 0.47, + "learning_rate": 1.1333240862367621e-05, + "loss": 1.0283, + "step": 14813 + }, + { + "epoch": 0.47, + "learning_rate": 1.1332214298650988e-05, + "loss": 1.0938, + "step": 14814 + }, + { + "epoch": 0.47, + "learning_rate": 1.133118772064117e-05, + "loss": 1.0425, + "step": 14815 + }, + { + "epoch": 0.47, + "learning_rate": 1.1330161128349182e-05, + "loss": 0.9199, + "step": 14816 + }, + { + "epoch": 0.47, + "learning_rate": 1.1329134521786039e-05, + "loss": 0.9922, + "step": 14817 + }, + { + "epoch": 0.47, + "learning_rate": 1.1328107900962749e-05, + "loss": 1.0029, + "step": 14818 + }, + { + "epoch": 0.47, + "learning_rate": 1.1327081265890335e-05, + "loss": 0.5532, + "step": 14819 + }, + { + "epoch": 0.47, + "learning_rate": 1.1326054616579806e-05, + "loss": 1.0142, + "step": 14820 + }, + { + "epoch": 0.47, + "learning_rate": 1.1325027953042178e-05, + "loss": 0.833, + "step": 14821 + }, + { + "epoch": 0.47, + "learning_rate": 1.1324001275288467e-05, + "loss": 0.8577, + "step": 14822 + }, + { + "epoch": 0.47, + "learning_rate": 1.132297458332969e-05, + "loss": 1.0405, + "step": 14823 + }, + { + "epoch": 0.47, + "learning_rate": 1.1321947877176857e-05, + "loss": 0.9629, + "step": 14824 + }, + { + "epoch": 0.47, + "learning_rate": 1.1320921156840987e-05, + "loss": 1.1289, + "step": 14825 + }, + { + "epoch": 0.47, + "learning_rate": 1.1319894422333096e-05, + "loss": 0.9658, + "step": 14826 + }, + { + "epoch": 0.47, + "learning_rate": 1.13188676736642e-05, + "loss": 0.9307, + "step": 14827 + }, + { + "epoch": 0.47, + "learning_rate": 1.131784091084531e-05, + "loss": 1.0288, + "step": 14828 + }, + { + "epoch": 0.47, + "learning_rate": 1.1316814133887447e-05, + "loss": 1.0381, + "step": 14829 + }, + { + "epoch": 0.47, + "learning_rate": 1.1315787342801625e-05, + "loss": 1.064, + "step": 14830 + }, + { + "epoch": 0.47, + "learning_rate": 1.1314760537598863e-05, + "loss": 0.9829, + "step": 14831 + }, + { + "epoch": 0.47, + "learning_rate": 1.1313733718290173e-05, + "loss": 0.998, + "step": 14832 + }, + { + "epoch": 0.47, + "learning_rate": 1.1312706884886574e-05, + "loss": 0.9155, + "step": 14833 + }, + { + "epoch": 0.47, + "learning_rate": 1.1311680037399087e-05, + "loss": 0.9937, + "step": 14834 + }, + { + "epoch": 0.47, + "learning_rate": 1.1310653175838722e-05, + "loss": 0.9214, + "step": 14835 + }, + { + "epoch": 0.47, + "learning_rate": 1.1309626300216498e-05, + "loss": 1.0229, + "step": 14836 + }, + { + "epoch": 0.47, + "learning_rate": 1.1308599410543435e-05, + "loss": 0.9087, + "step": 14837 + }, + { + "epoch": 0.47, + "learning_rate": 1.1307572506830548e-05, + "loss": 0.9116, + "step": 14838 + }, + { + "epoch": 0.47, + "learning_rate": 1.1306545589088857e-05, + "loss": 1.0615, + "step": 14839 + }, + { + "epoch": 0.47, + "learning_rate": 1.1305518657329375e-05, + "loss": 1.0996, + "step": 14840 + }, + { + "epoch": 0.47, + "learning_rate": 1.1304491711563124e-05, + "loss": 0.9404, + "step": 14841 + }, + { + "epoch": 0.47, + "learning_rate": 1.130346475180112e-05, + "loss": 0.9629, + "step": 14842 + }, + { + "epoch": 0.47, + "learning_rate": 1.1302437778054382e-05, + "loss": 0.9668, + "step": 14843 + }, + { + "epoch": 0.47, + "learning_rate": 1.130141079033393e-05, + "loss": 0.8574, + "step": 14844 + }, + { + "epoch": 0.47, + "learning_rate": 1.1300383788650777e-05, + "loss": 1.0454, + "step": 14845 + }, + { + "epoch": 0.47, + "learning_rate": 1.1299356773015951e-05, + "loss": 1.1299, + "step": 14846 + }, + { + "epoch": 0.47, + "learning_rate": 1.129832974344046e-05, + "loss": 0.9746, + "step": 14847 + }, + { + "epoch": 0.47, + "learning_rate": 1.1297302699935327e-05, + "loss": 1.0205, + "step": 14848 + }, + { + "epoch": 0.47, + "learning_rate": 1.1296275642511576e-05, + "loss": 0.9492, + "step": 14849 + }, + { + "epoch": 0.47, + "learning_rate": 1.1295248571180218e-05, + "loss": 0.916, + "step": 14850 + }, + { + "epoch": 0.47, + "learning_rate": 1.1294221485952279e-05, + "loss": 1.0312, + "step": 14851 + }, + { + "epoch": 0.47, + "learning_rate": 1.1293194386838776e-05, + "loss": 0.9961, + "step": 14852 + }, + { + "epoch": 0.48, + "learning_rate": 1.1292167273850725e-05, + "loss": 1.0127, + "step": 14853 + }, + { + "epoch": 0.48, + "learning_rate": 1.1291140146999153e-05, + "loss": 1.0337, + "step": 14854 + }, + { + "epoch": 0.48, + "learning_rate": 1.1290113006295074e-05, + "loss": 1.0815, + "step": 14855 + }, + { + "epoch": 0.48, + "learning_rate": 1.128908585174951e-05, + "loss": 0.9387, + "step": 14856 + }, + { + "epoch": 0.48, + "learning_rate": 1.1288058683373486e-05, + "loss": 0.9287, + "step": 14857 + }, + { + "epoch": 0.48, + "learning_rate": 1.1287031501178011e-05, + "loss": 1.0015, + "step": 14858 + }, + { + "epoch": 0.48, + "learning_rate": 1.1286004305174117e-05, + "loss": 0.9692, + "step": 14859 + }, + { + "epoch": 0.48, + "learning_rate": 1.1284977095372818e-05, + "loss": 0.9893, + "step": 14860 + }, + { + "epoch": 0.48, + "learning_rate": 1.128394987178514e-05, + "loss": 0.5251, + "step": 14861 + }, + { + "epoch": 0.48, + "learning_rate": 1.1282922634422097e-05, + "loss": 0.9043, + "step": 14862 + }, + { + "epoch": 0.48, + "learning_rate": 1.1281895383294717e-05, + "loss": 1.1133, + "step": 14863 + }, + { + "epoch": 0.48, + "learning_rate": 1.1280868118414015e-05, + "loss": 1.1445, + "step": 14864 + }, + { + "epoch": 0.48, + "learning_rate": 1.1279840839791018e-05, + "loss": 0.9526, + "step": 14865 + }, + { + "epoch": 0.48, + "learning_rate": 1.1278813547436745e-05, + "loss": 1.0039, + "step": 14866 + }, + { + "epoch": 0.48, + "learning_rate": 1.1277786241362217e-05, + "loss": 1.0991, + "step": 14867 + }, + { + "epoch": 0.48, + "learning_rate": 1.1276758921578459e-05, + "loss": 0.9585, + "step": 14868 + }, + { + "epoch": 0.48, + "learning_rate": 1.127573158809649e-05, + "loss": 1.0605, + "step": 14869 + }, + { + "epoch": 0.48, + "learning_rate": 1.1274704240927332e-05, + "loss": 0.9238, + "step": 14870 + }, + { + "epoch": 0.48, + "learning_rate": 1.1273676880082007e-05, + "loss": 1.0938, + "step": 14871 + }, + { + "epoch": 0.48, + "learning_rate": 1.1272649505571542e-05, + "loss": 1.019, + "step": 14872 + }, + { + "epoch": 0.48, + "learning_rate": 1.1271622117406955e-05, + "loss": 0.543, + "step": 14873 + }, + { + "epoch": 0.48, + "learning_rate": 1.1270594715599267e-05, + "loss": 0.8911, + "step": 14874 + }, + { + "epoch": 0.48, + "learning_rate": 1.1269567300159507e-05, + "loss": 1.0479, + "step": 14875 + }, + { + "epoch": 0.48, + "learning_rate": 1.1268539871098696e-05, + "loss": 1.0605, + "step": 14876 + }, + { + "epoch": 0.48, + "learning_rate": 1.1267512428427853e-05, + "loss": 0.9009, + "step": 14877 + }, + { + "epoch": 0.48, + "learning_rate": 1.1266484972158007e-05, + "loss": 1.0415, + "step": 14878 + }, + { + "epoch": 0.48, + "learning_rate": 1.1265457502300178e-05, + "loss": 1.002, + "step": 14879 + }, + { + "epoch": 0.48, + "learning_rate": 1.1264430018865391e-05, + "loss": 1.0791, + "step": 14880 + }, + { + "epoch": 0.48, + "learning_rate": 1.1263402521864668e-05, + "loss": 0.9077, + "step": 14881 + }, + { + "epoch": 0.48, + "learning_rate": 1.1262375011309037e-05, + "loss": 0.9907, + "step": 14882 + }, + { + "epoch": 0.48, + "learning_rate": 1.1261347487209515e-05, + "loss": 0.9663, + "step": 14883 + }, + { + "epoch": 0.48, + "learning_rate": 1.1260319949577134e-05, + "loss": 0.9375, + "step": 14884 + }, + { + "epoch": 0.48, + "learning_rate": 1.1259292398422915e-05, + "loss": 0.9858, + "step": 14885 + }, + { + "epoch": 0.48, + "learning_rate": 1.125826483375788e-05, + "loss": 0.979, + "step": 14886 + }, + { + "epoch": 0.48, + "learning_rate": 1.1257237255593058e-05, + "loss": 1.002, + "step": 14887 + }, + { + "epoch": 0.48, + "learning_rate": 1.125620966393947e-05, + "loss": 1.001, + "step": 14888 + }, + { + "epoch": 0.48, + "learning_rate": 1.1255182058808143e-05, + "loss": 1.0142, + "step": 14889 + }, + { + "epoch": 0.48, + "learning_rate": 1.1254154440210101e-05, + "loss": 1.1216, + "step": 14890 + }, + { + "epoch": 0.48, + "learning_rate": 1.1253126808156371e-05, + "loss": 0.9131, + "step": 14891 + }, + { + "epoch": 0.48, + "learning_rate": 1.1252099162657977e-05, + "loss": 0.916, + "step": 14892 + }, + { + "epoch": 0.48, + "learning_rate": 1.1251071503725947e-05, + "loss": 0.9644, + "step": 14893 + }, + { + "epoch": 0.48, + "learning_rate": 1.12500438313713e-05, + "loss": 0.9922, + "step": 14894 + }, + { + "epoch": 0.48, + "learning_rate": 1.124901614560507e-05, + "loss": 0.9873, + "step": 14895 + }, + { + "epoch": 0.48, + "learning_rate": 1.1247988446438278e-05, + "loss": 1.0122, + "step": 14896 + }, + { + "epoch": 0.48, + "learning_rate": 1.124696073388195e-05, + "loss": 1.0225, + "step": 14897 + }, + { + "epoch": 0.48, + "learning_rate": 1.1245933007947113e-05, + "loss": 1.0747, + "step": 14898 + }, + { + "epoch": 0.48, + "learning_rate": 1.1244905268644797e-05, + "loss": 1.1377, + "step": 14899 + }, + { + "epoch": 0.48, + "learning_rate": 1.1243877515986022e-05, + "loss": 1.0186, + "step": 14900 + }, + { + "epoch": 0.48, + "learning_rate": 1.1242849749981817e-05, + "loss": 1.0552, + "step": 14901 + }, + { + "epoch": 0.48, + "learning_rate": 1.1241821970643213e-05, + "loss": 0.8945, + "step": 14902 + }, + { + "epoch": 0.48, + "learning_rate": 1.124079417798123e-05, + "loss": 0.9741, + "step": 14903 + }, + { + "epoch": 0.48, + "learning_rate": 1.1239766372006902e-05, + "loss": 0.959, + "step": 14904 + }, + { + "epoch": 0.48, + "learning_rate": 1.1238738552731252e-05, + "loss": 0.9004, + "step": 14905 + }, + { + "epoch": 0.48, + "learning_rate": 1.123771072016531e-05, + "loss": 1.0977, + "step": 14906 + }, + { + "epoch": 0.48, + "learning_rate": 1.12366828743201e-05, + "loss": 0.9072, + "step": 14907 + }, + { + "epoch": 0.48, + "learning_rate": 1.123565501520665e-05, + "loss": 0.9668, + "step": 14908 + }, + { + "epoch": 0.48, + "learning_rate": 1.123462714283599e-05, + "loss": 0.5247, + "step": 14909 + }, + { + "epoch": 0.48, + "learning_rate": 1.123359925721915e-05, + "loss": 1.0322, + "step": 14910 + }, + { + "epoch": 0.48, + "learning_rate": 1.123257135836715e-05, + "loss": 1.1069, + "step": 14911 + }, + { + "epoch": 0.48, + "learning_rate": 1.1231543446291026e-05, + "loss": 1.0737, + "step": 14912 + }, + { + "epoch": 0.48, + "learning_rate": 1.1230515521001805e-05, + "loss": 1.0864, + "step": 14913 + }, + { + "epoch": 0.48, + "learning_rate": 1.1229487582510515e-05, + "loss": 0.8477, + "step": 14914 + }, + { + "epoch": 0.48, + "learning_rate": 1.1228459630828181e-05, + "loss": 1.0454, + "step": 14915 + }, + { + "epoch": 0.48, + "learning_rate": 1.1227431665965838e-05, + "loss": 0.9658, + "step": 14916 + }, + { + "epoch": 0.48, + "learning_rate": 1.1226403687934511e-05, + "loss": 1.0547, + "step": 14917 + }, + { + "epoch": 0.48, + "learning_rate": 1.1225375696745227e-05, + "loss": 1.0845, + "step": 14918 + }, + { + "epoch": 0.48, + "learning_rate": 1.122434769240902e-05, + "loss": 1.1313, + "step": 14919 + }, + { + "epoch": 0.48, + "learning_rate": 1.1223319674936916e-05, + "loss": 0.9297, + "step": 14920 + }, + { + "epoch": 0.48, + "learning_rate": 1.1222291644339948e-05, + "loss": 0.9353, + "step": 14921 + }, + { + "epoch": 0.48, + "learning_rate": 1.1221263600629143e-05, + "loss": 0.9976, + "step": 14922 + }, + { + "epoch": 0.48, + "learning_rate": 1.122023554381553e-05, + "loss": 0.96, + "step": 14923 + }, + { + "epoch": 0.48, + "learning_rate": 1.1219207473910138e-05, + "loss": 1.0425, + "step": 14924 + }, + { + "epoch": 0.48, + "learning_rate": 1.1218179390924004e-05, + "loss": 0.8955, + "step": 14925 + }, + { + "epoch": 0.48, + "learning_rate": 1.1217151294868152e-05, + "loss": 0.9248, + "step": 14926 + }, + { + "epoch": 0.48, + "learning_rate": 1.1216123185753611e-05, + "loss": 0.9331, + "step": 14927 + }, + { + "epoch": 0.48, + "learning_rate": 1.1215095063591415e-05, + "loss": 1.1045, + "step": 14928 + }, + { + "epoch": 0.48, + "learning_rate": 1.1214066928392597e-05, + "loss": 0.937, + "step": 14929 + }, + { + "epoch": 0.48, + "learning_rate": 1.1213038780168181e-05, + "loss": 1.0698, + "step": 14930 + }, + { + "epoch": 0.48, + "learning_rate": 1.1212010618929203e-05, + "loss": 1.0317, + "step": 14931 + }, + { + "epoch": 0.48, + "learning_rate": 1.1210982444686694e-05, + "loss": 0.5107, + "step": 14932 + }, + { + "epoch": 0.48, + "learning_rate": 1.120995425745168e-05, + "loss": 1.1206, + "step": 14933 + }, + { + "epoch": 0.48, + "learning_rate": 1.1208926057235197e-05, + "loss": 1.0371, + "step": 14934 + }, + { + "epoch": 0.48, + "learning_rate": 1.1207897844048276e-05, + "loss": 0.5334, + "step": 14935 + }, + { + "epoch": 0.48, + "learning_rate": 1.1206869617901949e-05, + "loss": 1.1265, + "step": 14936 + }, + { + "epoch": 0.48, + "learning_rate": 1.1205841378807245e-05, + "loss": 0.9048, + "step": 14937 + }, + { + "epoch": 0.48, + "learning_rate": 1.12048131267752e-05, + "loss": 0.8745, + "step": 14938 + }, + { + "epoch": 0.48, + "learning_rate": 1.120378486181684e-05, + "loss": 0.9414, + "step": 14939 + }, + { + "epoch": 0.48, + "learning_rate": 1.1202756583943205e-05, + "loss": 1.0156, + "step": 14940 + }, + { + "epoch": 0.48, + "learning_rate": 1.120172829316532e-05, + "loss": 1.0234, + "step": 14941 + }, + { + "epoch": 0.48, + "learning_rate": 1.120069998949422e-05, + "loss": 0.9185, + "step": 14942 + }, + { + "epoch": 0.48, + "learning_rate": 1.1199671672940939e-05, + "loss": 1.0752, + "step": 14943 + }, + { + "epoch": 0.48, + "learning_rate": 1.1198643343516508e-05, + "loss": 0.9434, + "step": 14944 + }, + { + "epoch": 0.48, + "learning_rate": 1.1197615001231962e-05, + "loss": 0.9839, + "step": 14945 + }, + { + "epoch": 0.48, + "learning_rate": 1.119658664609833e-05, + "loss": 1.1294, + "step": 14946 + }, + { + "epoch": 0.48, + "learning_rate": 1.119555827812665e-05, + "loss": 1.0176, + "step": 14947 + }, + { + "epoch": 0.48, + "learning_rate": 1.1194529897327953e-05, + "loss": 0.8745, + "step": 14948 + }, + { + "epoch": 0.48, + "learning_rate": 1.1193501503713272e-05, + "loss": 1.063, + "step": 14949 + }, + { + "epoch": 0.48, + "learning_rate": 1.119247309729364e-05, + "loss": 0.9897, + "step": 14950 + }, + { + "epoch": 0.48, + "learning_rate": 1.1191444678080093e-05, + "loss": 0.9058, + "step": 14951 + }, + { + "epoch": 0.48, + "learning_rate": 1.119041624608366e-05, + "loss": 0.9497, + "step": 14952 + }, + { + "epoch": 0.48, + "learning_rate": 1.1189387801315378e-05, + "loss": 1.0615, + "step": 14953 + }, + { + "epoch": 0.48, + "learning_rate": 1.1188359343786283e-05, + "loss": 0.9165, + "step": 14954 + }, + { + "epoch": 0.48, + "learning_rate": 1.118733087350741e-05, + "loss": 1.041, + "step": 14955 + }, + { + "epoch": 0.48, + "learning_rate": 1.1186302390489787e-05, + "loss": 1.0762, + "step": 14956 + }, + { + "epoch": 0.48, + "learning_rate": 1.1185273894744452e-05, + "loss": 0.9751, + "step": 14957 + }, + { + "epoch": 0.48, + "learning_rate": 1.1184245386282442e-05, + "loss": 0.9985, + "step": 14958 + }, + { + "epoch": 0.48, + "learning_rate": 1.1183216865114792e-05, + "loss": 0.9224, + "step": 14959 + }, + { + "epoch": 0.48, + "learning_rate": 1.118218833125253e-05, + "loss": 1.1587, + "step": 14960 + }, + { + "epoch": 0.48, + "learning_rate": 1.1181159784706696e-05, + "loss": 0.5112, + "step": 14961 + }, + { + "epoch": 0.48, + "learning_rate": 1.1180131225488327e-05, + "loss": 0.9668, + "step": 14962 + }, + { + "epoch": 0.48, + "learning_rate": 1.1179102653608456e-05, + "loss": 1.0532, + "step": 14963 + }, + { + "epoch": 0.48, + "learning_rate": 1.1178074069078118e-05, + "loss": 0.9609, + "step": 14964 + }, + { + "epoch": 0.48, + "learning_rate": 1.1177045471908348e-05, + "loss": 1.0591, + "step": 14965 + }, + { + "epoch": 0.48, + "learning_rate": 1.1176016862110181e-05, + "loss": 0.9575, + "step": 14966 + }, + { + "epoch": 0.48, + "learning_rate": 1.117498823969466e-05, + "loss": 1.0254, + "step": 14967 + }, + { + "epoch": 0.48, + "learning_rate": 1.1173959604672812e-05, + "loss": 0.9722, + "step": 14968 + }, + { + "epoch": 0.48, + "learning_rate": 1.1172930957055676e-05, + "loss": 1.123, + "step": 14969 + }, + { + "epoch": 0.48, + "learning_rate": 1.1171902296854293e-05, + "loss": 1.124, + "step": 14970 + }, + { + "epoch": 0.48, + "learning_rate": 1.1170873624079692e-05, + "loss": 0.9468, + "step": 14971 + }, + { + "epoch": 0.48, + "learning_rate": 1.1169844938742913e-05, + "loss": 1.0381, + "step": 14972 + }, + { + "epoch": 0.48, + "learning_rate": 1.1168816240854994e-05, + "loss": 1.0913, + "step": 14973 + }, + { + "epoch": 0.48, + "learning_rate": 1.116778753042697e-05, + "loss": 1.1099, + "step": 14974 + }, + { + "epoch": 0.48, + "learning_rate": 1.1166758807469878e-05, + "loss": 1.0498, + "step": 14975 + }, + { + "epoch": 0.48, + "learning_rate": 1.1165730071994754e-05, + "loss": 1.0532, + "step": 14976 + }, + { + "epoch": 0.48, + "learning_rate": 1.116470132401264e-05, + "loss": 0.9941, + "step": 14977 + }, + { + "epoch": 0.48, + "learning_rate": 1.1163672563534567e-05, + "loss": 0.9312, + "step": 14978 + }, + { + "epoch": 0.48, + "learning_rate": 1.1162643790571574e-05, + "loss": 1.0879, + "step": 14979 + }, + { + "epoch": 0.48, + "learning_rate": 1.1161615005134703e-05, + "loss": 0.8572, + "step": 14980 + }, + { + "epoch": 0.48, + "learning_rate": 1.1160586207234987e-05, + "loss": 1.0225, + "step": 14981 + }, + { + "epoch": 0.48, + "learning_rate": 1.1159557396883467e-05, + "loss": 0.9521, + "step": 14982 + }, + { + "epoch": 0.48, + "learning_rate": 1.1158528574091176e-05, + "loss": 0.9722, + "step": 14983 + }, + { + "epoch": 0.48, + "learning_rate": 1.1157499738869157e-05, + "loss": 0.9082, + "step": 14984 + }, + { + "epoch": 0.48, + "learning_rate": 1.115647089122845e-05, + "loss": 0.937, + "step": 14985 + }, + { + "epoch": 0.48, + "learning_rate": 1.1155442031180086e-05, + "loss": 1.0386, + "step": 14986 + }, + { + "epoch": 0.48, + "learning_rate": 1.115441315873511e-05, + "loss": 0.9082, + "step": 14987 + }, + { + "epoch": 0.48, + "learning_rate": 1.1153384273904556e-05, + "loss": 0.9663, + "step": 14988 + }, + { + "epoch": 0.48, + "learning_rate": 1.115235537669947e-05, + "loss": 1.0215, + "step": 14989 + }, + { + "epoch": 0.48, + "learning_rate": 1.1151326467130883e-05, + "loss": 1.0239, + "step": 14990 + }, + { + "epoch": 0.48, + "learning_rate": 1.1150297545209837e-05, + "loss": 1.0347, + "step": 14991 + }, + { + "epoch": 0.48, + "learning_rate": 1.114926861094737e-05, + "loss": 0.9771, + "step": 14992 + }, + { + "epoch": 0.48, + "learning_rate": 1.1148239664354528e-05, + "loss": 1.0005, + "step": 14993 + }, + { + "epoch": 0.48, + "learning_rate": 1.1147210705442342e-05, + "loss": 0.999, + "step": 14994 + }, + { + "epoch": 0.48, + "learning_rate": 1.1146181734221853e-05, + "loss": 1.001, + "step": 14995 + }, + { + "epoch": 0.48, + "learning_rate": 1.1145152750704105e-05, + "loss": 0.9492, + "step": 14996 + }, + { + "epoch": 0.48, + "learning_rate": 1.1144123754900135e-05, + "loss": 1.0586, + "step": 14997 + }, + { + "epoch": 0.48, + "learning_rate": 1.1143094746820984e-05, + "loss": 0.9346, + "step": 14998 + }, + { + "epoch": 0.48, + "learning_rate": 1.114206572647769e-05, + "loss": 0.9502, + "step": 14999 + }, + { + "epoch": 0.48, + "learning_rate": 1.1141036693881293e-05, + "loss": 1.0781, + "step": 15000 + }, + { + "epoch": 0.48, + "learning_rate": 1.1140007649042839e-05, + "loss": 1.0762, + "step": 15001 + }, + { + "epoch": 0.48, + "learning_rate": 1.1138978591973364e-05, + "loss": 1.0894, + "step": 15002 + }, + { + "epoch": 0.48, + "learning_rate": 1.1137949522683906e-05, + "loss": 0.8813, + "step": 15003 + }, + { + "epoch": 0.48, + "learning_rate": 1.1136920441185515e-05, + "loss": 1.0156, + "step": 15004 + }, + { + "epoch": 0.48, + "learning_rate": 1.113589134748922e-05, + "loss": 1.0098, + "step": 15005 + }, + { + "epoch": 0.48, + "learning_rate": 1.113486224160607e-05, + "loss": 1.0649, + "step": 15006 + }, + { + "epoch": 0.48, + "learning_rate": 1.1133833123547105e-05, + "loss": 1.0288, + "step": 15007 + }, + { + "epoch": 0.48, + "learning_rate": 1.1132803993323365e-05, + "loss": 0.5361, + "step": 15008 + }, + { + "epoch": 0.48, + "learning_rate": 1.1131774850945892e-05, + "loss": 1.0098, + "step": 15009 + }, + { + "epoch": 0.48, + "learning_rate": 1.1130745696425727e-05, + "loss": 1.0029, + "step": 15010 + }, + { + "epoch": 0.48, + "learning_rate": 1.112971652977391e-05, + "loss": 0.9526, + "step": 15011 + }, + { + "epoch": 0.48, + "learning_rate": 1.112868735100149e-05, + "loss": 1.0264, + "step": 15012 + }, + { + "epoch": 0.48, + "learning_rate": 1.11276581601195e-05, + "loss": 1.0674, + "step": 15013 + }, + { + "epoch": 0.48, + "learning_rate": 1.1126628957138988e-05, + "loss": 1.1011, + "step": 15014 + }, + { + "epoch": 0.48, + "learning_rate": 1.1125599742070992e-05, + "loss": 1.0308, + "step": 15015 + }, + { + "epoch": 0.48, + "learning_rate": 1.1124570514926558e-05, + "loss": 1.0454, + "step": 15016 + }, + { + "epoch": 0.48, + "learning_rate": 1.1123541275716727e-05, + "loss": 1.0317, + "step": 15017 + }, + { + "epoch": 0.48, + "learning_rate": 1.112251202445254e-05, + "loss": 1.0317, + "step": 15018 + }, + { + "epoch": 0.48, + "learning_rate": 1.1121482761145046e-05, + "loss": 1.0923, + "step": 15019 + }, + { + "epoch": 0.48, + "learning_rate": 1.1120453485805279e-05, + "loss": 0.9805, + "step": 15020 + }, + { + "epoch": 0.48, + "learning_rate": 1.1119424198444288e-05, + "loss": 0.9741, + "step": 15021 + }, + { + "epoch": 0.48, + "learning_rate": 1.1118394899073112e-05, + "loss": 1.0981, + "step": 15022 + }, + { + "epoch": 0.48, + "learning_rate": 1.1117365587702798e-05, + "loss": 0.8779, + "step": 15023 + }, + { + "epoch": 0.48, + "learning_rate": 1.111633626434439e-05, + "loss": 0.9399, + "step": 15024 + }, + { + "epoch": 0.48, + "learning_rate": 1.1115306929008926e-05, + "loss": 1.0474, + "step": 15025 + }, + { + "epoch": 0.48, + "learning_rate": 1.1114277581707454e-05, + "loss": 0.9204, + "step": 15026 + }, + { + "epoch": 0.48, + "learning_rate": 1.1113248222451018e-05, + "loss": 0.9443, + "step": 15027 + }, + { + "epoch": 0.48, + "learning_rate": 1.1112218851250657e-05, + "loss": 0.9077, + "step": 15028 + }, + { + "epoch": 0.48, + "learning_rate": 1.1111189468117423e-05, + "loss": 0.9561, + "step": 15029 + }, + { + "epoch": 0.48, + "learning_rate": 1.1110160073062353e-05, + "loss": 1.0591, + "step": 15030 + }, + { + "epoch": 0.48, + "learning_rate": 1.1109130666096497e-05, + "loss": 0.8979, + "step": 15031 + }, + { + "epoch": 0.48, + "learning_rate": 1.1108101247230893e-05, + "loss": 0.9199, + "step": 15032 + }, + { + "epoch": 0.48, + "learning_rate": 1.110707181647659e-05, + "loss": 0.8994, + "step": 15033 + }, + { + "epoch": 0.48, + "learning_rate": 1.1106042373844634e-05, + "loss": 1.0278, + "step": 15034 + }, + { + "epoch": 0.48, + "learning_rate": 1.1105012919346065e-05, + "loss": 0.998, + "step": 15035 + }, + { + "epoch": 0.48, + "learning_rate": 1.110398345299193e-05, + "loss": 1.0552, + "step": 15036 + }, + { + "epoch": 0.48, + "learning_rate": 1.1102953974793276e-05, + "loss": 0.8701, + "step": 15037 + }, + { + "epoch": 0.48, + "learning_rate": 1.1101924484761147e-05, + "loss": 1.084, + "step": 15038 + }, + { + "epoch": 0.48, + "learning_rate": 1.1100894982906584e-05, + "loss": 0.9399, + "step": 15039 + }, + { + "epoch": 0.48, + "learning_rate": 1.109986546924064e-05, + "loss": 0.9985, + "step": 15040 + }, + { + "epoch": 0.48, + "learning_rate": 1.1098835943774356e-05, + "loss": 0.9351, + "step": 15041 + }, + { + "epoch": 0.48, + "learning_rate": 1.109780640651878e-05, + "loss": 1.0796, + "step": 15042 + }, + { + "epoch": 0.48, + "learning_rate": 1.1096776857484952e-05, + "loss": 1.0386, + "step": 15043 + }, + { + "epoch": 0.48, + "learning_rate": 1.1095747296683925e-05, + "loss": 0.9473, + "step": 15044 + }, + { + "epoch": 0.48, + "learning_rate": 1.1094717724126741e-05, + "loss": 0.9351, + "step": 15045 + }, + { + "epoch": 0.48, + "learning_rate": 1.109368813982445e-05, + "loss": 0.9639, + "step": 15046 + }, + { + "epoch": 0.48, + "learning_rate": 1.1092658543788093e-05, + "loss": 0.9126, + "step": 15047 + }, + { + "epoch": 0.48, + "learning_rate": 1.109162893602872e-05, + "loss": 1.0347, + "step": 15048 + }, + { + "epoch": 0.48, + "learning_rate": 1.1090599316557375e-05, + "loss": 1.0005, + "step": 15049 + }, + { + "epoch": 0.48, + "learning_rate": 1.1089569685385112e-05, + "loss": 0.9961, + "step": 15050 + }, + { + "epoch": 0.48, + "learning_rate": 1.1088540042522967e-05, + "loss": 0.9312, + "step": 15051 + }, + { + "epoch": 0.48, + "learning_rate": 1.1087510387981992e-05, + "loss": 1.1226, + "step": 15052 + }, + { + "epoch": 0.48, + "learning_rate": 1.1086480721773236e-05, + "loss": 0.9595, + "step": 15053 + }, + { + "epoch": 0.48, + "learning_rate": 1.1085451043907745e-05, + "loss": 0.9766, + "step": 15054 + }, + { + "epoch": 0.48, + "learning_rate": 1.1084421354396562e-05, + "loss": 0.9517, + "step": 15055 + }, + { + "epoch": 0.48, + "learning_rate": 1.1083391653250742e-05, + "loss": 0.9502, + "step": 15056 + }, + { + "epoch": 0.48, + "learning_rate": 1.1082361940481328e-05, + "loss": 0.9995, + "step": 15057 + }, + { + "epoch": 0.48, + "learning_rate": 1.1081332216099365e-05, + "loss": 0.959, + "step": 15058 + }, + { + "epoch": 0.48, + "learning_rate": 1.1080302480115908e-05, + "loss": 0.9565, + "step": 15059 + }, + { + "epoch": 0.48, + "learning_rate": 1.1079272732541998e-05, + "loss": 1.0063, + "step": 15060 + }, + { + "epoch": 0.48, + "learning_rate": 1.1078242973388689e-05, + "loss": 1.0933, + "step": 15061 + }, + { + "epoch": 0.48, + "learning_rate": 1.1077213202667024e-05, + "loss": 0.9971, + "step": 15062 + }, + { + "epoch": 0.48, + "learning_rate": 1.1076183420388057e-05, + "loss": 0.9692, + "step": 15063 + }, + { + "epoch": 0.48, + "learning_rate": 1.107515362656283e-05, + "loss": 0.9575, + "step": 15064 + }, + { + "epoch": 0.48, + "learning_rate": 1.1074123821202392e-05, + "loss": 1.0527, + "step": 15065 + }, + { + "epoch": 0.48, + "learning_rate": 1.1073094004317799e-05, + "loss": 1.0054, + "step": 15066 + }, + { + "epoch": 0.48, + "learning_rate": 1.1072064175920092e-05, + "loss": 1.0674, + "step": 15067 + }, + { + "epoch": 0.48, + "learning_rate": 1.1071034336020326e-05, + "loss": 0.9697, + "step": 15068 + }, + { + "epoch": 0.48, + "learning_rate": 1.1070004484629543e-05, + "loss": 1.123, + "step": 15069 + }, + { + "epoch": 0.48, + "learning_rate": 1.1068974621758796e-05, + "loss": 1.0288, + "step": 15070 + }, + { + "epoch": 0.48, + "learning_rate": 1.1067944747419137e-05, + "loss": 0.9722, + "step": 15071 + }, + { + "epoch": 0.48, + "learning_rate": 1.1066914861621613e-05, + "loss": 0.5247, + "step": 15072 + }, + { + "epoch": 0.48, + "learning_rate": 1.1065884964377271e-05, + "loss": 0.8813, + "step": 15073 + }, + { + "epoch": 0.48, + "learning_rate": 1.1064855055697166e-05, + "loss": 1.228, + "step": 15074 + }, + { + "epoch": 0.48, + "learning_rate": 1.106382513559234e-05, + "loss": 0.9624, + "step": 15075 + }, + { + "epoch": 0.48, + "learning_rate": 1.1062795204073852e-05, + "loss": 1.0527, + "step": 15076 + }, + { + "epoch": 0.48, + "learning_rate": 1.1061765261152749e-05, + "loss": 0.9688, + "step": 15077 + }, + { + "epoch": 0.48, + "learning_rate": 1.1060735306840075e-05, + "loss": 1.0049, + "step": 15078 + }, + { + "epoch": 0.48, + "learning_rate": 1.1059705341146887e-05, + "loss": 0.9976, + "step": 15079 + }, + { + "epoch": 0.48, + "learning_rate": 1.1058675364084235e-05, + "loss": 0.9707, + "step": 15080 + }, + { + "epoch": 0.48, + "learning_rate": 1.1057645375663166e-05, + "loss": 1.0225, + "step": 15081 + }, + { + "epoch": 0.48, + "learning_rate": 1.1056615375894734e-05, + "loss": 0.9736, + "step": 15082 + }, + { + "epoch": 0.48, + "learning_rate": 1.1055585364789987e-05, + "loss": 0.9478, + "step": 15083 + }, + { + "epoch": 0.48, + "learning_rate": 1.105455534235998e-05, + "loss": 0.8979, + "step": 15084 + }, + { + "epoch": 0.48, + "learning_rate": 1.105352530861576e-05, + "loss": 0.8164, + "step": 15085 + }, + { + "epoch": 0.48, + "learning_rate": 1.1052495263568379e-05, + "loss": 1.0522, + "step": 15086 + }, + { + "epoch": 0.48, + "learning_rate": 1.1051465207228889e-05, + "loss": 1.001, + "step": 15087 + }, + { + "epoch": 0.48, + "learning_rate": 1.105043513960834e-05, + "loss": 0.4963, + "step": 15088 + }, + { + "epoch": 0.48, + "learning_rate": 1.1049405060717786e-05, + "loss": 1.04, + "step": 15089 + }, + { + "epoch": 0.48, + "learning_rate": 1.1048374970568276e-05, + "loss": 1.1465, + "step": 15090 + }, + { + "epoch": 0.48, + "learning_rate": 1.1047344869170864e-05, + "loss": 1.0566, + "step": 15091 + }, + { + "epoch": 0.48, + "learning_rate": 1.10463147565366e-05, + "loss": 0.8711, + "step": 15092 + }, + { + "epoch": 0.48, + "learning_rate": 1.1045284632676535e-05, + "loss": 0.8589, + "step": 15093 + }, + { + "epoch": 0.48, + "learning_rate": 1.1044254497601725e-05, + "loss": 1.0083, + "step": 15094 + }, + { + "epoch": 0.48, + "learning_rate": 1.104322435132322e-05, + "loss": 0.5527, + "step": 15095 + }, + { + "epoch": 0.48, + "learning_rate": 1.1042194193852071e-05, + "loss": 0.9834, + "step": 15096 + }, + { + "epoch": 0.48, + "learning_rate": 1.1041164025199332e-05, + "loss": 0.9092, + "step": 15097 + }, + { + "epoch": 0.48, + "learning_rate": 1.1040133845376056e-05, + "loss": 0.5469, + "step": 15098 + }, + { + "epoch": 0.48, + "learning_rate": 1.1039103654393295e-05, + "loss": 0.5459, + "step": 15099 + }, + { + "epoch": 0.48, + "learning_rate": 1.1038073452262102e-05, + "loss": 0.9307, + "step": 15100 + }, + { + "epoch": 0.48, + "learning_rate": 1.1037043238993528e-05, + "loss": 1.04, + "step": 15101 + }, + { + "epoch": 0.48, + "learning_rate": 1.103601301459863e-05, + "loss": 1.0479, + "step": 15102 + }, + { + "epoch": 0.48, + "learning_rate": 1.103498277908846e-05, + "loss": 0.9771, + "step": 15103 + }, + { + "epoch": 0.48, + "learning_rate": 1.1033952532474066e-05, + "loss": 0.9316, + "step": 15104 + }, + { + "epoch": 0.48, + "learning_rate": 1.1032922274766508e-05, + "loss": 1.0439, + "step": 15105 + }, + { + "epoch": 0.48, + "learning_rate": 1.1031892005976839e-05, + "loss": 1.0142, + "step": 15106 + }, + { + "epoch": 0.48, + "learning_rate": 1.1030861726116107e-05, + "loss": 1.0542, + "step": 15107 + }, + { + "epoch": 0.48, + "learning_rate": 1.1029831435195373e-05, + "loss": 1.0576, + "step": 15108 + }, + { + "epoch": 0.48, + "learning_rate": 1.1028801133225687e-05, + "loss": 0.9971, + "step": 15109 + }, + { + "epoch": 0.48, + "learning_rate": 1.1027770820218103e-05, + "loss": 1.0269, + "step": 15110 + }, + { + "epoch": 0.48, + "learning_rate": 1.1026740496183674e-05, + "loss": 0.9507, + "step": 15111 + }, + { + "epoch": 0.48, + "learning_rate": 1.1025710161133457e-05, + "loss": 0.9995, + "step": 15112 + }, + { + "epoch": 0.48, + "learning_rate": 1.1024679815078503e-05, + "loss": 0.9868, + "step": 15113 + }, + { + "epoch": 0.48, + "learning_rate": 1.1023649458029873e-05, + "loss": 0.9355, + "step": 15114 + }, + { + "epoch": 0.48, + "learning_rate": 1.1022619089998615e-05, + "loss": 0.9419, + "step": 15115 + }, + { + "epoch": 0.48, + "learning_rate": 1.1021588710995783e-05, + "loss": 0.9409, + "step": 15116 + }, + { + "epoch": 0.48, + "learning_rate": 1.1020558321032437e-05, + "loss": 0.9165, + "step": 15117 + }, + { + "epoch": 0.48, + "learning_rate": 1.1019527920119633e-05, + "loss": 0.9321, + "step": 15118 + }, + { + "epoch": 0.48, + "learning_rate": 1.101849750826842e-05, + "loss": 1.0234, + "step": 15119 + }, + { + "epoch": 0.48, + "learning_rate": 1.1017467085489855e-05, + "loss": 0.9287, + "step": 15120 + }, + { + "epoch": 0.48, + "learning_rate": 1.1016436651794997e-05, + "loss": 0.9902, + "step": 15121 + }, + { + "epoch": 0.48, + "learning_rate": 1.1015406207194897e-05, + "loss": 1.1172, + "step": 15122 + }, + { + "epoch": 0.48, + "learning_rate": 1.101437575170061e-05, + "loss": 1.0464, + "step": 15123 + }, + { + "epoch": 0.48, + "learning_rate": 1.1013345285323196e-05, + "loss": 0.9922, + "step": 15124 + }, + { + "epoch": 0.48, + "learning_rate": 1.101231480807371e-05, + "loss": 1.0972, + "step": 15125 + }, + { + "epoch": 0.48, + "learning_rate": 1.1011284319963205e-05, + "loss": 0.9678, + "step": 15126 + }, + { + "epoch": 0.48, + "learning_rate": 1.1010253821002737e-05, + "loss": 1.1416, + "step": 15127 + }, + { + "epoch": 0.48, + "learning_rate": 1.1009223311203363e-05, + "loss": 0.9922, + "step": 15128 + }, + { + "epoch": 0.48, + "learning_rate": 1.1008192790576144e-05, + "loss": 0.9023, + "step": 15129 + }, + { + "epoch": 0.48, + "learning_rate": 1.1007162259132128e-05, + "loss": 1.0322, + "step": 15130 + }, + { + "epoch": 0.48, + "learning_rate": 1.100613171688238e-05, + "loss": 1.0015, + "step": 15131 + }, + { + "epoch": 0.48, + "learning_rate": 1.1005101163837948e-05, + "loss": 0.9897, + "step": 15132 + }, + { + "epoch": 0.48, + "learning_rate": 1.1004070600009898e-05, + "loss": 0.9146, + "step": 15133 + }, + { + "epoch": 0.48, + "learning_rate": 1.1003040025409278e-05, + "loss": 1.0132, + "step": 15134 + }, + { + "epoch": 0.48, + "learning_rate": 1.1002009440047147e-05, + "loss": 0.9614, + "step": 15135 + }, + { + "epoch": 0.48, + "learning_rate": 1.1000978843934569e-05, + "loss": 1.0645, + "step": 15136 + }, + { + "epoch": 0.48, + "learning_rate": 1.0999948237082592e-05, + "loss": 1.0596, + "step": 15137 + }, + { + "epoch": 0.48, + "learning_rate": 1.0998917619502276e-05, + "loss": 1.0718, + "step": 15138 + }, + { + "epoch": 0.48, + "learning_rate": 1.099788699120468e-05, + "loss": 0.894, + "step": 15139 + }, + { + "epoch": 0.48, + "learning_rate": 1.0996856352200864e-05, + "loss": 0.8872, + "step": 15140 + }, + { + "epoch": 0.48, + "learning_rate": 1.0995825702501882e-05, + "loss": 0.8887, + "step": 15141 + }, + { + "epoch": 0.48, + "learning_rate": 1.099479504211879e-05, + "loss": 1.0981, + "step": 15142 + }, + { + "epoch": 0.48, + "learning_rate": 1.0993764371062647e-05, + "loss": 1.0688, + "step": 15143 + }, + { + "epoch": 0.48, + "learning_rate": 1.0992733689344517e-05, + "loss": 0.9375, + "step": 15144 + }, + { + "epoch": 0.48, + "learning_rate": 1.099170299697545e-05, + "loss": 1.0117, + "step": 15145 + }, + { + "epoch": 0.48, + "learning_rate": 1.099067229396651e-05, + "loss": 1.0249, + "step": 15146 + }, + { + "epoch": 0.48, + "learning_rate": 1.0989641580328752e-05, + "loss": 1.041, + "step": 15147 + }, + { + "epoch": 0.48, + "learning_rate": 1.0988610856073232e-05, + "loss": 1.0747, + "step": 15148 + }, + { + "epoch": 0.48, + "learning_rate": 1.0987580121211014e-05, + "loss": 0.9766, + "step": 15149 + }, + { + "epoch": 0.48, + "learning_rate": 1.0986549375753154e-05, + "loss": 0.8989, + "step": 15150 + }, + { + "epoch": 0.48, + "learning_rate": 1.0985518619710711e-05, + "loss": 0.9834, + "step": 15151 + }, + { + "epoch": 0.48, + "learning_rate": 1.0984487853094744e-05, + "loss": 1.2202, + "step": 15152 + }, + { + "epoch": 0.48, + "learning_rate": 1.0983457075916314e-05, + "loss": 0.9224, + "step": 15153 + }, + { + "epoch": 0.48, + "learning_rate": 1.0982426288186475e-05, + "loss": 1.0083, + "step": 15154 + }, + { + "epoch": 0.48, + "learning_rate": 1.0981395489916292e-05, + "loss": 0.8823, + "step": 15155 + }, + { + "epoch": 0.48, + "learning_rate": 1.0980364681116817e-05, + "loss": 1.0166, + "step": 15156 + }, + { + "epoch": 0.48, + "learning_rate": 1.0979333861799117e-05, + "loss": 1.0405, + "step": 15157 + }, + { + "epoch": 0.48, + "learning_rate": 1.0978303031974246e-05, + "loss": 0.9268, + "step": 15158 + }, + { + "epoch": 0.48, + "learning_rate": 1.0977272191653272e-05, + "loss": 1.1714, + "step": 15159 + }, + { + "epoch": 0.48, + "learning_rate": 1.0976241340847244e-05, + "loss": 0.9077, + "step": 15160 + }, + { + "epoch": 0.48, + "learning_rate": 1.0975210479567226e-05, + "loss": 1.0913, + "step": 15161 + }, + { + "epoch": 0.48, + "learning_rate": 1.0974179607824282e-05, + "loss": 1.0493, + "step": 15162 + }, + { + "epoch": 0.48, + "learning_rate": 1.097314872562947e-05, + "loss": 1.0806, + "step": 15163 + }, + { + "epoch": 0.48, + "learning_rate": 1.0972117832993846e-05, + "loss": 1.0977, + "step": 15164 + }, + { + "epoch": 0.48, + "learning_rate": 1.0971086929928475e-05, + "loss": 0.7412, + "step": 15165 + }, + { + "epoch": 0.49, + "learning_rate": 1.0970056016444414e-05, + "loss": 0.958, + "step": 15166 + }, + { + "epoch": 0.49, + "learning_rate": 1.096902509255273e-05, + "loss": 1.1128, + "step": 15167 + }, + { + "epoch": 0.49, + "learning_rate": 1.0967994158264475e-05, + "loss": 0.9175, + "step": 15168 + }, + { + "epoch": 0.49, + "learning_rate": 1.0966963213590715e-05, + "loss": 1.0596, + "step": 15169 + }, + { + "epoch": 0.49, + "learning_rate": 1.0965932258542513e-05, + "loss": 0.9102, + "step": 15170 + }, + { + "epoch": 0.49, + "learning_rate": 1.0964901293130924e-05, + "loss": 1.0078, + "step": 15171 + }, + { + "epoch": 0.49, + "learning_rate": 1.0963870317367011e-05, + "loss": 1.0142, + "step": 15172 + }, + { + "epoch": 0.49, + "learning_rate": 1.0962839331261836e-05, + "loss": 1.0371, + "step": 15173 + }, + { + "epoch": 0.49, + "learning_rate": 1.0961808334826465e-05, + "loss": 0.9868, + "step": 15174 + }, + { + "epoch": 0.49, + "learning_rate": 1.0960777328071951e-05, + "loss": 0.5269, + "step": 15175 + }, + { + "epoch": 0.49, + "learning_rate": 1.095974631100936e-05, + "loss": 0.9302, + "step": 15176 + }, + { + "epoch": 0.49, + "learning_rate": 1.0958715283649755e-05, + "loss": 0.9595, + "step": 15177 + }, + { + "epoch": 0.49, + "learning_rate": 1.0957684246004197e-05, + "loss": 0.9351, + "step": 15178 + }, + { + "epoch": 0.49, + "learning_rate": 1.0956653198083745e-05, + "loss": 1.0039, + "step": 15179 + }, + { + "epoch": 0.49, + "learning_rate": 1.095562213989946e-05, + "loss": 1.0234, + "step": 15180 + }, + { + "epoch": 0.49, + "learning_rate": 1.095459107146241e-05, + "loss": 1.0093, + "step": 15181 + }, + { + "epoch": 0.49, + "learning_rate": 1.0953559992783656e-05, + "loss": 0.938, + "step": 15182 + }, + { + "epoch": 0.49, + "learning_rate": 1.0952528903874258e-05, + "loss": 0.9922, + "step": 15183 + }, + { + "epoch": 0.49, + "learning_rate": 1.0951497804745277e-05, + "loss": 0.9873, + "step": 15184 + }, + { + "epoch": 0.49, + "learning_rate": 1.095046669540778e-05, + "loss": 0.8984, + "step": 15185 + }, + { + "epoch": 0.49, + "learning_rate": 1.0949435575872825e-05, + "loss": 1.0356, + "step": 15186 + }, + { + "epoch": 0.49, + "learning_rate": 1.0948404446151477e-05, + "loss": 0.8936, + "step": 15187 + }, + { + "epoch": 0.49, + "learning_rate": 1.0947373306254802e-05, + "loss": 0.8247, + "step": 15188 + }, + { + "epoch": 0.49, + "learning_rate": 1.0946342156193858e-05, + "loss": 0.9634, + "step": 15189 + }, + { + "epoch": 0.49, + "learning_rate": 1.094531099597971e-05, + "loss": 1.1689, + "step": 15190 + }, + { + "epoch": 0.49, + "learning_rate": 1.0944279825623423e-05, + "loss": 1.0552, + "step": 15191 + }, + { + "epoch": 0.49, + "learning_rate": 1.0943248645136055e-05, + "loss": 1.0498, + "step": 15192 + }, + { + "epoch": 0.49, + "learning_rate": 1.0942217454528674e-05, + "loss": 1.0293, + "step": 15193 + }, + { + "epoch": 0.49, + "learning_rate": 1.0941186253812346e-05, + "loss": 1.0044, + "step": 15194 + }, + { + "epoch": 0.49, + "learning_rate": 1.0940155042998126e-05, + "loss": 0.9644, + "step": 15195 + }, + { + "epoch": 0.49, + "learning_rate": 1.0939123822097086e-05, + "loss": 0.5054, + "step": 15196 + }, + { + "epoch": 0.49, + "learning_rate": 1.0938092591120288e-05, + "loss": 0.9136, + "step": 15197 + }, + { + "epoch": 0.49, + "learning_rate": 1.0937061350078792e-05, + "loss": 0.9829, + "step": 15198 + }, + { + "epoch": 0.49, + "learning_rate": 1.0936030098983666e-05, + "loss": 0.8643, + "step": 15199 + }, + { + "epoch": 0.49, + "learning_rate": 1.0934998837845971e-05, + "loss": 0.854, + "step": 15200 + }, + { + "epoch": 0.49, + "learning_rate": 1.0933967566676775e-05, + "loss": 1.0439, + "step": 15201 + }, + { + "epoch": 0.49, + "learning_rate": 1.093293628548714e-05, + "loss": 1.0908, + "step": 15202 + }, + { + "epoch": 0.49, + "learning_rate": 1.0931904994288131e-05, + "loss": 1.1587, + "step": 15203 + }, + { + "epoch": 0.49, + "learning_rate": 1.0930873693090815e-05, + "loss": 1.0566, + "step": 15204 + }, + { + "epoch": 0.49, + "learning_rate": 1.0929842381906253e-05, + "loss": 0.9863, + "step": 15205 + }, + { + "epoch": 0.49, + "learning_rate": 1.0928811060745513e-05, + "loss": 0.9097, + "step": 15206 + }, + { + "epoch": 0.49, + "learning_rate": 1.0927779729619654e-05, + "loss": 1.0811, + "step": 15207 + }, + { + "epoch": 0.49, + "learning_rate": 1.092674838853975e-05, + "loss": 0.9883, + "step": 15208 + }, + { + "epoch": 0.49, + "learning_rate": 1.0925717037516859e-05, + "loss": 0.8799, + "step": 15209 + }, + { + "epoch": 0.49, + "learning_rate": 1.0924685676562049e-05, + "loss": 0.9712, + "step": 15210 + }, + { + "epoch": 0.49, + "learning_rate": 1.0923654305686384e-05, + "loss": 1.083, + "step": 15211 + }, + { + "epoch": 0.49, + "learning_rate": 1.0922622924900935e-05, + "loss": 1.0039, + "step": 15212 + }, + { + "epoch": 0.49, + "learning_rate": 1.0921591534216757e-05, + "loss": 0.9229, + "step": 15213 + }, + { + "epoch": 0.49, + "learning_rate": 1.0920560133644923e-05, + "loss": 1.0039, + "step": 15214 + }, + { + "epoch": 0.49, + "learning_rate": 1.09195287231965e-05, + "loss": 1.0049, + "step": 15215 + }, + { + "epoch": 0.49, + "learning_rate": 1.0918497302882554e-05, + "loss": 1.0527, + "step": 15216 + }, + { + "epoch": 0.49, + "learning_rate": 1.0917465872714144e-05, + "loss": 0.9023, + "step": 15217 + }, + { + "epoch": 0.49, + "learning_rate": 1.091643443270234e-05, + "loss": 0.9053, + "step": 15218 + }, + { + "epoch": 0.49, + "learning_rate": 1.0915402982858212e-05, + "loss": 1.1445, + "step": 15219 + }, + { + "epoch": 0.49, + "learning_rate": 1.0914371523192821e-05, + "loss": 1.0391, + "step": 15220 + }, + { + "epoch": 0.49, + "learning_rate": 1.0913340053717236e-05, + "loss": 1.0181, + "step": 15221 + }, + { + "epoch": 0.49, + "learning_rate": 1.0912308574442523e-05, + "loss": 0.9595, + "step": 15222 + }, + { + "epoch": 0.49, + "learning_rate": 1.091127708537975e-05, + "loss": 1.0366, + "step": 15223 + }, + { + "epoch": 0.49, + "learning_rate": 1.0910245586539983e-05, + "loss": 0.9482, + "step": 15224 + }, + { + "epoch": 0.49, + "learning_rate": 1.0909214077934286e-05, + "loss": 1.0669, + "step": 15225 + }, + { + "epoch": 0.49, + "learning_rate": 1.0908182559573728e-05, + "loss": 0.8608, + "step": 15226 + }, + { + "epoch": 0.49, + "learning_rate": 1.0907151031469379e-05, + "loss": 1.0703, + "step": 15227 + }, + { + "epoch": 0.49, + "learning_rate": 1.0906119493632298e-05, + "loss": 0.9873, + "step": 15228 + }, + { + "epoch": 0.49, + "learning_rate": 1.0905087946073563e-05, + "loss": 1.0137, + "step": 15229 + }, + { + "epoch": 0.49, + "learning_rate": 1.0904056388804236e-05, + "loss": 1.0439, + "step": 15230 + }, + { + "epoch": 0.49, + "learning_rate": 1.0903024821835382e-05, + "loss": 0.8618, + "step": 15231 + }, + { + "epoch": 0.49, + "learning_rate": 1.090199324517807e-05, + "loss": 0.9736, + "step": 15232 + }, + { + "epoch": 0.49, + "learning_rate": 1.0900961658843375e-05, + "loss": 1.0112, + "step": 15233 + }, + { + "epoch": 0.49, + "learning_rate": 1.0899930062842353e-05, + "loss": 1.0923, + "step": 15234 + }, + { + "epoch": 0.49, + "learning_rate": 1.0898898457186078e-05, + "loss": 1.0679, + "step": 15235 + }, + { + "epoch": 0.49, + "learning_rate": 1.0897866841885618e-05, + "loss": 1.0088, + "step": 15236 + }, + { + "epoch": 0.49, + "learning_rate": 1.0896835216952039e-05, + "loss": 0.9087, + "step": 15237 + }, + { + "epoch": 0.49, + "learning_rate": 1.0895803582396415e-05, + "loss": 0.8872, + "step": 15238 + }, + { + "epoch": 0.49, + "learning_rate": 1.0894771938229807e-05, + "loss": 1.0542, + "step": 15239 + }, + { + "epoch": 0.49, + "learning_rate": 1.0893740284463286e-05, + "loss": 0.9883, + "step": 15240 + }, + { + "epoch": 0.49, + "learning_rate": 1.0892708621107922e-05, + "loss": 0.9541, + "step": 15241 + }, + { + "epoch": 0.49, + "learning_rate": 1.0891676948174784e-05, + "loss": 1.0034, + "step": 15242 + }, + { + "epoch": 0.49, + "learning_rate": 1.0890645265674935e-05, + "loss": 1.0566, + "step": 15243 + }, + { + "epoch": 0.49, + "learning_rate": 1.0889613573619449e-05, + "loss": 1.02, + "step": 15244 + }, + { + "epoch": 0.49, + "learning_rate": 1.0888581872019395e-05, + "loss": 0.9263, + "step": 15245 + }, + { + "epoch": 0.49, + "learning_rate": 1.0887550160885842e-05, + "loss": 0.9756, + "step": 15246 + }, + { + "epoch": 0.49, + "learning_rate": 1.088651844022986e-05, + "loss": 0.9932, + "step": 15247 + }, + { + "epoch": 0.49, + "learning_rate": 1.088548671006251e-05, + "loss": 1.0317, + "step": 15248 + }, + { + "epoch": 0.49, + "learning_rate": 1.0884454970394871e-05, + "loss": 1.0088, + "step": 15249 + }, + { + "epoch": 0.49, + "learning_rate": 1.088342322123801e-05, + "loss": 1.0122, + "step": 15250 + }, + { + "epoch": 0.49, + "learning_rate": 1.0882391462602995e-05, + "loss": 1.0654, + "step": 15251 + }, + { + "epoch": 0.49, + "learning_rate": 1.0881359694500894e-05, + "loss": 0.9404, + "step": 15252 + }, + { + "epoch": 0.49, + "learning_rate": 1.0880327916942783e-05, + "loss": 1.001, + "step": 15253 + }, + { + "epoch": 0.49, + "learning_rate": 1.0879296129939725e-05, + "loss": 1.0366, + "step": 15254 + }, + { + "epoch": 0.49, + "learning_rate": 1.0878264333502793e-05, + "loss": 1.0391, + "step": 15255 + }, + { + "epoch": 0.49, + "learning_rate": 1.0877232527643056e-05, + "loss": 0.9351, + "step": 15256 + }, + { + "epoch": 0.49, + "learning_rate": 1.0876200712371588e-05, + "loss": 0.9495, + "step": 15257 + }, + { + "epoch": 0.49, + "learning_rate": 1.0875168887699453e-05, + "loss": 0.9507, + "step": 15258 + }, + { + "epoch": 0.49, + "learning_rate": 1.0874137053637724e-05, + "loss": 0.9673, + "step": 15259 + }, + { + "epoch": 0.49, + "learning_rate": 1.0873105210197472e-05, + "loss": 0.9536, + "step": 15260 + }, + { + "epoch": 0.49, + "learning_rate": 1.087207335738977e-05, + "loss": 1.0244, + "step": 15261 + }, + { + "epoch": 0.49, + "learning_rate": 1.0871041495225685e-05, + "loss": 1.1372, + "step": 15262 + }, + { + "epoch": 0.49, + "learning_rate": 1.0870009623716287e-05, + "loss": 0.9736, + "step": 15263 + }, + { + "epoch": 0.49, + "learning_rate": 1.0868977742872651e-05, + "loss": 0.5249, + "step": 15264 + }, + { + "epoch": 0.49, + "learning_rate": 1.0867945852705845e-05, + "loss": 1.0439, + "step": 15265 + }, + { + "epoch": 0.49, + "learning_rate": 1.0866913953226938e-05, + "loss": 0.9883, + "step": 15266 + }, + { + "epoch": 0.49, + "learning_rate": 1.0865882044447006e-05, + "loss": 0.9619, + "step": 15267 + }, + { + "epoch": 0.49, + "learning_rate": 1.0864850126377117e-05, + "loss": 1.0166, + "step": 15268 + }, + { + "epoch": 0.49, + "learning_rate": 1.0863818199028347e-05, + "loss": 0.9912, + "step": 15269 + }, + { + "epoch": 0.49, + "learning_rate": 1.086278626241176e-05, + "loss": 1.082, + "step": 15270 + }, + { + "epoch": 0.49, + "learning_rate": 1.086175431653843e-05, + "loss": 0.8379, + "step": 15271 + }, + { + "epoch": 0.49, + "learning_rate": 1.0860722361419436e-05, + "loss": 1.0493, + "step": 15272 + }, + { + "epoch": 0.49, + "learning_rate": 1.0859690397065838e-05, + "loss": 1.0166, + "step": 15273 + }, + { + "epoch": 0.49, + "learning_rate": 1.0858658423488715e-05, + "loss": 1.0024, + "step": 15274 + }, + { + "epoch": 0.49, + "learning_rate": 1.0857626440699139e-05, + "loss": 1.0278, + "step": 15275 + }, + { + "epoch": 0.49, + "learning_rate": 1.0856594448708178e-05, + "loss": 1.0464, + "step": 15276 + }, + { + "epoch": 0.49, + "learning_rate": 1.0855562447526911e-05, + "loss": 0.9961, + "step": 15277 + }, + { + "epoch": 0.49, + "learning_rate": 1.0854530437166399e-05, + "loss": 1.0659, + "step": 15278 + }, + { + "epoch": 0.49, + "learning_rate": 1.0853498417637726e-05, + "loss": 1.0483, + "step": 15279 + }, + { + "epoch": 0.49, + "learning_rate": 1.085246638895196e-05, + "loss": 1.0669, + "step": 15280 + }, + { + "epoch": 0.49, + "learning_rate": 1.0851434351120172e-05, + "loss": 1.0532, + "step": 15281 + }, + { + "epoch": 0.49, + "learning_rate": 1.0850402304153433e-05, + "loss": 1.0127, + "step": 15282 + }, + { + "epoch": 0.49, + "learning_rate": 1.084937024806282e-05, + "loss": 0.9985, + "step": 15283 + }, + { + "epoch": 0.49, + "learning_rate": 1.0848338182859405e-05, + "loss": 0.9863, + "step": 15284 + }, + { + "epoch": 0.49, + "learning_rate": 1.084730610855426e-05, + "loss": 0.9653, + "step": 15285 + }, + { + "epoch": 0.49, + "learning_rate": 1.084627402515846e-05, + "loss": 1.0308, + "step": 15286 + }, + { + "epoch": 0.49, + "learning_rate": 1.0845241932683075e-05, + "loss": 0.9985, + "step": 15287 + }, + { + "epoch": 0.49, + "learning_rate": 1.0844209831139179e-05, + "loss": 0.9551, + "step": 15288 + }, + { + "epoch": 0.49, + "learning_rate": 1.0843177720537845e-05, + "loss": 1.021, + "step": 15289 + }, + { + "epoch": 0.49, + "learning_rate": 1.0842145600890149e-05, + "loss": 0.9824, + "step": 15290 + }, + { + "epoch": 0.49, + "learning_rate": 1.084111347220716e-05, + "loss": 0.96, + "step": 15291 + }, + { + "epoch": 0.49, + "learning_rate": 1.0840081334499956e-05, + "loss": 1.0107, + "step": 15292 + }, + { + "epoch": 0.49, + "learning_rate": 1.0839049187779611e-05, + "loss": 0.896, + "step": 15293 + }, + { + "epoch": 0.49, + "learning_rate": 1.0838017032057194e-05, + "loss": 1.0713, + "step": 15294 + }, + { + "epoch": 0.49, + "learning_rate": 1.0836984867343785e-05, + "loss": 1.0381, + "step": 15295 + }, + { + "epoch": 0.49, + "learning_rate": 1.083595269365045e-05, + "loss": 1.0132, + "step": 15296 + }, + { + "epoch": 0.49, + "learning_rate": 1.0834920510988272e-05, + "loss": 0.9155, + "step": 15297 + }, + { + "epoch": 0.49, + "learning_rate": 1.0833888319368318e-05, + "loss": 0.897, + "step": 15298 + }, + { + "epoch": 0.49, + "learning_rate": 1.0832856118801669e-05, + "loss": 0.8765, + "step": 15299 + }, + { + "epoch": 0.49, + "learning_rate": 1.0831823909299391e-05, + "loss": 0.9487, + "step": 15300 + }, + { + "epoch": 0.49, + "learning_rate": 1.0830791690872564e-05, + "loss": 0.9834, + "step": 15301 + }, + { + "epoch": 0.49, + "learning_rate": 1.0829759463532265e-05, + "loss": 0.8662, + "step": 15302 + }, + { + "epoch": 0.49, + "learning_rate": 1.0828727227289564e-05, + "loss": 0.9292, + "step": 15303 + }, + { + "epoch": 0.49, + "learning_rate": 1.0827694982155534e-05, + "loss": 0.9653, + "step": 15304 + }, + { + "epoch": 0.49, + "learning_rate": 1.0826662728141255e-05, + "loss": 0.9219, + "step": 15305 + }, + { + "epoch": 0.49, + "learning_rate": 1.0825630465257802e-05, + "loss": 0.9561, + "step": 15306 + }, + { + "epoch": 0.49, + "learning_rate": 1.0824598193516245e-05, + "loss": 0.9717, + "step": 15307 + }, + { + "epoch": 0.49, + "learning_rate": 1.0823565912927662e-05, + "loss": 1.0112, + "step": 15308 + }, + { + "epoch": 0.49, + "learning_rate": 1.0822533623503128e-05, + "loss": 1.0283, + "step": 15309 + }, + { + "epoch": 0.49, + "learning_rate": 1.0821501325253722e-05, + "loss": 1.0425, + "step": 15310 + }, + { + "epoch": 0.49, + "learning_rate": 1.0820469018190513e-05, + "loss": 0.959, + "step": 15311 + }, + { + "epoch": 0.49, + "learning_rate": 1.081943670232458e-05, + "loss": 0.9272, + "step": 15312 + }, + { + "epoch": 0.49, + "learning_rate": 1.0818404377666996e-05, + "loss": 1.0962, + "step": 15313 + }, + { + "epoch": 0.49, + "learning_rate": 1.0817372044228843e-05, + "loss": 1.0581, + "step": 15314 + }, + { + "epoch": 0.49, + "learning_rate": 1.081633970202119e-05, + "loss": 1.0342, + "step": 15315 + }, + { + "epoch": 0.49, + "learning_rate": 1.0815307351055116e-05, + "loss": 1.1445, + "step": 15316 + }, + { + "epoch": 0.49, + "learning_rate": 1.0814274991341698e-05, + "loss": 0.9946, + "step": 15317 + }, + { + "epoch": 0.49, + "learning_rate": 1.0813242622892009e-05, + "loss": 0.9551, + "step": 15318 + }, + { + "epoch": 0.49, + "learning_rate": 1.0812210245717128e-05, + "loss": 1.1816, + "step": 15319 + }, + { + "epoch": 0.49, + "learning_rate": 1.0811177859828128e-05, + "loss": 1.0747, + "step": 15320 + }, + { + "epoch": 0.49, + "learning_rate": 1.0810145465236085e-05, + "loss": 0.9888, + "step": 15321 + }, + { + "epoch": 0.49, + "learning_rate": 1.0809113061952083e-05, + "loss": 0.8574, + "step": 15322 + }, + { + "epoch": 0.49, + "learning_rate": 1.0808080649987191e-05, + "loss": 1.0352, + "step": 15323 + }, + { + "epoch": 0.49, + "learning_rate": 1.0807048229352488e-05, + "loss": 1.0142, + "step": 15324 + }, + { + "epoch": 0.49, + "learning_rate": 1.0806015800059054e-05, + "loss": 0.9282, + "step": 15325 + }, + { + "epoch": 0.49, + "learning_rate": 1.0804983362117958e-05, + "loss": 0.9907, + "step": 15326 + }, + { + "epoch": 0.49, + "learning_rate": 1.0803950915540283e-05, + "loss": 1.0337, + "step": 15327 + }, + { + "epoch": 0.49, + "learning_rate": 1.0802918460337101e-05, + "loss": 1.0186, + "step": 15328 + }, + { + "epoch": 0.49, + "learning_rate": 1.0801885996519498e-05, + "loss": 0.8418, + "step": 15329 + }, + { + "epoch": 0.49, + "learning_rate": 1.0800853524098543e-05, + "loss": 0.9907, + "step": 15330 + }, + { + "epoch": 0.49, + "learning_rate": 1.0799821043085314e-05, + "loss": 0.8843, + "step": 15331 + }, + { + "epoch": 0.49, + "learning_rate": 1.0798788553490894e-05, + "loss": 0.9341, + "step": 15332 + }, + { + "epoch": 0.49, + "learning_rate": 1.0797756055326359e-05, + "loss": 1.0249, + "step": 15333 + }, + { + "epoch": 0.49, + "learning_rate": 1.0796723548602778e-05, + "loss": 0.9775, + "step": 15334 + }, + { + "epoch": 0.49, + "learning_rate": 1.0795691033331239e-05, + "loss": 0.9683, + "step": 15335 + }, + { + "epoch": 0.49, + "learning_rate": 1.0794658509522817e-05, + "loss": 0.9175, + "step": 15336 + }, + { + "epoch": 0.49, + "learning_rate": 1.0793625977188587e-05, + "loss": 0.959, + "step": 15337 + }, + { + "epoch": 0.49, + "learning_rate": 1.0792593436339627e-05, + "loss": 0.9814, + "step": 15338 + }, + { + "epoch": 0.49, + "learning_rate": 1.0791560886987016e-05, + "loss": 0.9761, + "step": 15339 + }, + { + "epoch": 0.49, + "learning_rate": 1.0790528329141839e-05, + "loss": 0.9883, + "step": 15340 + }, + { + "epoch": 0.49, + "learning_rate": 1.078949576281516e-05, + "loss": 0.9951, + "step": 15341 + }, + { + "epoch": 0.49, + "learning_rate": 1.078846318801807e-05, + "loss": 0.9219, + "step": 15342 + }, + { + "epoch": 0.49, + "learning_rate": 1.078743060476164e-05, + "loss": 0.9482, + "step": 15343 + }, + { + "epoch": 0.49, + "learning_rate": 1.0786398013056955e-05, + "loss": 0.9434, + "step": 15344 + }, + { + "epoch": 0.49, + "learning_rate": 1.0785365412915088e-05, + "loss": 1.0591, + "step": 15345 + }, + { + "epoch": 0.49, + "learning_rate": 1.0784332804347117e-05, + "loss": 0.8848, + "step": 15346 + }, + { + "epoch": 0.49, + "learning_rate": 1.0783300187364122e-05, + "loss": 0.8696, + "step": 15347 + }, + { + "epoch": 0.49, + "learning_rate": 1.0782267561977189e-05, + "loss": 1.0479, + "step": 15348 + }, + { + "epoch": 0.49, + "learning_rate": 1.0781234928197386e-05, + "loss": 0.9697, + "step": 15349 + }, + { + "epoch": 0.49, + "learning_rate": 1.0780202286035799e-05, + "loss": 1.0337, + "step": 15350 + }, + { + "epoch": 0.49, + "learning_rate": 1.0779169635503503e-05, + "loss": 1.0713, + "step": 15351 + }, + { + "epoch": 0.49, + "learning_rate": 1.0778136976611581e-05, + "loss": 0.915, + "step": 15352 + }, + { + "epoch": 0.49, + "learning_rate": 1.0777104309371111e-05, + "loss": 0.875, + "step": 15353 + }, + { + "epoch": 0.49, + "learning_rate": 1.0776071633793171e-05, + "loss": 0.9336, + "step": 15354 + }, + { + "epoch": 0.49, + "learning_rate": 1.077503894988884e-05, + "loss": 1.0049, + "step": 15355 + }, + { + "epoch": 0.49, + "learning_rate": 1.07740062576692e-05, + "loss": 0.9419, + "step": 15356 + }, + { + "epoch": 0.49, + "learning_rate": 1.077297355714533e-05, + "loss": 1.0908, + "step": 15357 + }, + { + "epoch": 0.49, + "learning_rate": 1.0771940848328308e-05, + "loss": 1.0435, + "step": 15358 + }, + { + "epoch": 0.49, + "learning_rate": 1.0770908131229216e-05, + "loss": 1.0098, + "step": 15359 + }, + { + "epoch": 0.49, + "learning_rate": 1.0769875405859132e-05, + "loss": 1.0386, + "step": 15360 + }, + { + "epoch": 0.49, + "learning_rate": 1.0768842672229138e-05, + "loss": 0.9702, + "step": 15361 + }, + { + "epoch": 0.49, + "learning_rate": 1.0767809930350313e-05, + "loss": 1.0151, + "step": 15362 + }, + { + "epoch": 0.49, + "learning_rate": 1.0766777180233736e-05, + "loss": 0.8853, + "step": 15363 + }, + { + "epoch": 0.49, + "learning_rate": 1.0765744421890492e-05, + "loss": 1.063, + "step": 15364 + }, + { + "epoch": 0.49, + "learning_rate": 1.0764711655331652e-05, + "loss": 1.1084, + "step": 15365 + }, + { + "epoch": 0.49, + "learning_rate": 1.0763678880568306e-05, + "loss": 0.9961, + "step": 15366 + }, + { + "epoch": 0.49, + "learning_rate": 1.0762646097611531e-05, + "loss": 1.062, + "step": 15367 + }, + { + "epoch": 0.49, + "learning_rate": 1.0761613306472407e-05, + "loss": 0.9756, + "step": 15368 + }, + { + "epoch": 0.49, + "learning_rate": 1.0760580507162012e-05, + "loss": 1.0566, + "step": 15369 + }, + { + "epoch": 0.49, + "learning_rate": 1.0759547699691435e-05, + "loss": 1.0273, + "step": 15370 + }, + { + "epoch": 0.49, + "learning_rate": 1.0758514884071749e-05, + "loss": 0.9692, + "step": 15371 + }, + { + "epoch": 0.49, + "learning_rate": 1.0757482060314037e-05, + "loss": 1.0161, + "step": 15372 + }, + { + "epoch": 0.49, + "learning_rate": 1.0756449228429382e-05, + "loss": 1.0181, + "step": 15373 + }, + { + "epoch": 0.49, + "learning_rate": 1.0755416388428865e-05, + "loss": 1.0327, + "step": 15374 + }, + { + "epoch": 0.49, + "learning_rate": 1.0754383540323566e-05, + "loss": 0.9849, + "step": 15375 + }, + { + "epoch": 0.49, + "learning_rate": 1.0753350684124563e-05, + "loss": 1.0142, + "step": 15376 + }, + { + "epoch": 0.49, + "learning_rate": 1.0752317819842946e-05, + "loss": 0.876, + "step": 15377 + }, + { + "epoch": 0.49, + "learning_rate": 1.0751284947489789e-05, + "loss": 1.0259, + "step": 15378 + }, + { + "epoch": 0.49, + "learning_rate": 1.0750252067076176e-05, + "loss": 1.0474, + "step": 15379 + }, + { + "epoch": 0.49, + "learning_rate": 1.0749219178613188e-05, + "loss": 1.1758, + "step": 15380 + }, + { + "epoch": 0.49, + "learning_rate": 1.0748186282111907e-05, + "loss": 0.9531, + "step": 15381 + }, + { + "epoch": 0.49, + "learning_rate": 1.074715337758342e-05, + "loss": 0.9912, + "step": 15382 + }, + { + "epoch": 0.49, + "learning_rate": 1.0746120465038798e-05, + "loss": 1.0635, + "step": 15383 + }, + { + "epoch": 0.49, + "learning_rate": 1.0745087544489132e-05, + "loss": 0.9614, + "step": 15384 + }, + { + "epoch": 0.49, + "learning_rate": 1.07440546159455e-05, + "loss": 1.0068, + "step": 15385 + }, + { + "epoch": 0.49, + "learning_rate": 1.0743021679418987e-05, + "loss": 0.9824, + "step": 15386 + }, + { + "epoch": 0.49, + "learning_rate": 1.0741988734920675e-05, + "loss": 1.0088, + "step": 15387 + }, + { + "epoch": 0.49, + "learning_rate": 1.0740955782461641e-05, + "loss": 0.8633, + "step": 15388 + }, + { + "epoch": 0.49, + "learning_rate": 1.0739922822052975e-05, + "loss": 0.9463, + "step": 15389 + }, + { + "epoch": 0.49, + "learning_rate": 1.0738889853705754e-05, + "loss": 1.021, + "step": 15390 + }, + { + "epoch": 0.49, + "learning_rate": 1.0737856877431063e-05, + "loss": 1.0068, + "step": 15391 + }, + { + "epoch": 0.49, + "learning_rate": 1.0736823893239985e-05, + "loss": 1.1353, + "step": 15392 + }, + { + "epoch": 0.49, + "learning_rate": 1.0735790901143603e-05, + "loss": 1.0352, + "step": 15393 + }, + { + "epoch": 0.49, + "learning_rate": 1.0734757901152996e-05, + "loss": 1.0029, + "step": 15394 + }, + { + "epoch": 0.49, + "learning_rate": 1.0733724893279251e-05, + "loss": 0.8936, + "step": 15395 + }, + { + "epoch": 0.49, + "learning_rate": 1.073269187753345e-05, + "loss": 0.957, + "step": 15396 + }, + { + "epoch": 0.49, + "learning_rate": 1.0731658853926678e-05, + "loss": 0.9326, + "step": 15397 + }, + { + "epoch": 0.49, + "learning_rate": 1.0730625822470015e-05, + "loss": 0.9595, + "step": 15398 + }, + { + "epoch": 0.49, + "learning_rate": 1.0729592783174545e-05, + "loss": 1.0127, + "step": 15399 + }, + { + "epoch": 0.49, + "learning_rate": 1.0728559736051354e-05, + "loss": 1.0352, + "step": 15400 + }, + { + "epoch": 0.49, + "learning_rate": 1.0727526681111522e-05, + "loss": 0.9043, + "step": 15401 + }, + { + "epoch": 0.49, + "learning_rate": 1.0726493618366131e-05, + "loss": 1.0044, + "step": 15402 + }, + { + "epoch": 0.49, + "learning_rate": 1.0725460547826275e-05, + "loss": 0.8882, + "step": 15403 + }, + { + "epoch": 0.49, + "learning_rate": 1.0724427469503023e-05, + "loss": 0.9336, + "step": 15404 + }, + { + "epoch": 0.49, + "learning_rate": 1.0723394383407471e-05, + "loss": 0.9497, + "step": 15405 + }, + { + "epoch": 0.49, + "learning_rate": 1.0722361289550694e-05, + "loss": 1.1758, + "step": 15406 + }, + { + "epoch": 0.49, + "learning_rate": 1.072132818794378e-05, + "loss": 0.9834, + "step": 15407 + }, + { + "epoch": 0.49, + "learning_rate": 1.0720295078597816e-05, + "loss": 0.9688, + "step": 15408 + }, + { + "epoch": 0.49, + "learning_rate": 1.071926196152388e-05, + "loss": 1.104, + "step": 15409 + }, + { + "epoch": 0.49, + "learning_rate": 1.071822883673306e-05, + "loss": 0.9438, + "step": 15410 + }, + { + "epoch": 0.49, + "learning_rate": 1.0717195704236438e-05, + "loss": 0.9668, + "step": 15411 + }, + { + "epoch": 0.49, + "learning_rate": 1.0716162564045103e-05, + "loss": 1.0405, + "step": 15412 + }, + { + "epoch": 0.49, + "learning_rate": 1.0715129416170136e-05, + "loss": 1.0601, + "step": 15413 + }, + { + "epoch": 0.49, + "learning_rate": 1.071409626062262e-05, + "loss": 1.0879, + "step": 15414 + }, + { + "epoch": 0.49, + "learning_rate": 1.071306309741364e-05, + "loss": 1.041, + "step": 15415 + }, + { + "epoch": 0.49, + "learning_rate": 1.0712029926554285e-05, + "loss": 0.9541, + "step": 15416 + }, + { + "epoch": 0.49, + "learning_rate": 1.0710996748055638e-05, + "loss": 1.0703, + "step": 15417 + }, + { + "epoch": 0.49, + "learning_rate": 1.0709963561928779e-05, + "loss": 0.9961, + "step": 15418 + }, + { + "epoch": 0.49, + "learning_rate": 1.07089303681848e-05, + "loss": 0.9561, + "step": 15419 + }, + { + "epoch": 0.49, + "learning_rate": 1.070789716683478e-05, + "loss": 1.0088, + "step": 15420 + }, + { + "epoch": 0.49, + "learning_rate": 1.0706863957889808e-05, + "loss": 1.042, + "step": 15421 + }, + { + "epoch": 0.49, + "learning_rate": 1.0705830741360968e-05, + "loss": 0.98, + "step": 15422 + }, + { + "epoch": 0.49, + "learning_rate": 1.0704797517259346e-05, + "loss": 1.1113, + "step": 15423 + }, + { + "epoch": 0.49, + "learning_rate": 1.0703764285596027e-05, + "loss": 0.9473, + "step": 15424 + }, + { + "epoch": 0.49, + "learning_rate": 1.0702731046382096e-05, + "loss": 0.9575, + "step": 15425 + }, + { + "epoch": 0.49, + "learning_rate": 1.0701697799628636e-05, + "loss": 1.1152, + "step": 15426 + }, + { + "epoch": 0.49, + "learning_rate": 1.070066454534674e-05, + "loss": 1.0068, + "step": 15427 + }, + { + "epoch": 0.49, + "learning_rate": 1.0699631283547484e-05, + "loss": 1.1294, + "step": 15428 + }, + { + "epoch": 0.49, + "learning_rate": 1.069859801424196e-05, + "loss": 0.9902, + "step": 15429 + }, + { + "epoch": 0.49, + "learning_rate": 1.0697564737441254e-05, + "loss": 0.9448, + "step": 15430 + }, + { + "epoch": 0.49, + "learning_rate": 1.0696531453156452e-05, + "loss": 0.9292, + "step": 15431 + }, + { + "epoch": 0.49, + "learning_rate": 1.0695498161398634e-05, + "loss": 0.8823, + "step": 15432 + }, + { + "epoch": 0.49, + "learning_rate": 1.0694464862178893e-05, + "loss": 1.0562, + "step": 15433 + }, + { + "epoch": 0.49, + "learning_rate": 1.069343155550831e-05, + "loss": 0.9854, + "step": 15434 + }, + { + "epoch": 0.49, + "learning_rate": 1.0692398241397979e-05, + "loss": 1.0239, + "step": 15435 + }, + { + "epoch": 0.49, + "learning_rate": 1.0691364919858975e-05, + "loss": 1.0161, + "step": 15436 + }, + { + "epoch": 0.49, + "learning_rate": 1.0690331590902395e-05, + "loss": 1.002, + "step": 15437 + }, + { + "epoch": 0.49, + "learning_rate": 1.068929825453932e-05, + "loss": 1.0488, + "step": 15438 + }, + { + "epoch": 0.49, + "learning_rate": 1.0688264910780837e-05, + "loss": 0.8643, + "step": 15439 + }, + { + "epoch": 0.49, + "learning_rate": 1.0687231559638034e-05, + "loss": 1.0269, + "step": 15440 + }, + { + "epoch": 0.49, + "learning_rate": 1.0686198201121996e-05, + "loss": 1.0625, + "step": 15441 + }, + { + "epoch": 0.49, + "learning_rate": 1.0685164835243813e-05, + "loss": 0.895, + "step": 15442 + }, + { + "epoch": 0.49, + "learning_rate": 1.0684131462014566e-05, + "loss": 0.9258, + "step": 15443 + }, + { + "epoch": 0.49, + "learning_rate": 1.0683098081445347e-05, + "loss": 0.9941, + "step": 15444 + }, + { + "epoch": 0.49, + "learning_rate": 1.0682064693547245e-05, + "loss": 0.9985, + "step": 15445 + }, + { + "epoch": 0.49, + "learning_rate": 1.068103129833134e-05, + "loss": 1.0015, + "step": 15446 + }, + { + "epoch": 0.49, + "learning_rate": 1.0679997895808724e-05, + "loss": 0.9507, + "step": 15447 + }, + { + "epoch": 0.49, + "learning_rate": 1.0678964485990483e-05, + "loss": 1.1191, + "step": 15448 + }, + { + "epoch": 0.49, + "learning_rate": 1.0677931068887704e-05, + "loss": 0.9565, + "step": 15449 + }, + { + "epoch": 0.49, + "learning_rate": 1.0676897644511477e-05, + "loss": 1.0654, + "step": 15450 + }, + { + "epoch": 0.49, + "learning_rate": 1.0675864212872883e-05, + "loss": 1.0854, + "step": 15451 + }, + { + "epoch": 0.49, + "learning_rate": 1.0674830773983019e-05, + "loss": 0.9331, + "step": 15452 + }, + { + "epoch": 0.49, + "learning_rate": 1.0673797327852967e-05, + "loss": 0.8838, + "step": 15453 + }, + { + "epoch": 0.49, + "learning_rate": 1.0672763874493812e-05, + "loss": 1.0234, + "step": 15454 + }, + { + "epoch": 0.49, + "learning_rate": 1.0671730413916648e-05, + "loss": 1.0537, + "step": 15455 + }, + { + "epoch": 0.49, + "learning_rate": 1.0670696946132556e-05, + "loss": 1.0918, + "step": 15456 + }, + { + "epoch": 0.49, + "learning_rate": 1.0669663471152633e-05, + "loss": 1.0703, + "step": 15457 + }, + { + "epoch": 0.49, + "learning_rate": 1.066862998898796e-05, + "loss": 0.8547, + "step": 15458 + }, + { + "epoch": 0.49, + "learning_rate": 1.0667596499649628e-05, + "loss": 0.9868, + "step": 15459 + }, + { + "epoch": 0.49, + "learning_rate": 1.0666563003148722e-05, + "loss": 0.9971, + "step": 15460 + }, + { + "epoch": 0.49, + "learning_rate": 1.0665529499496336e-05, + "loss": 1.0859, + "step": 15461 + }, + { + "epoch": 0.49, + "learning_rate": 1.0664495988703554e-05, + "loss": 1.0234, + "step": 15462 + }, + { + "epoch": 0.49, + "learning_rate": 1.0663462470781464e-05, + "loss": 0.915, + "step": 15463 + }, + { + "epoch": 0.49, + "learning_rate": 1.0662428945741157e-05, + "loss": 0.9395, + "step": 15464 + }, + { + "epoch": 0.49, + "learning_rate": 1.0661395413593722e-05, + "loss": 0.9453, + "step": 15465 + }, + { + "epoch": 0.49, + "learning_rate": 1.0660361874350245e-05, + "loss": 0.9648, + "step": 15466 + }, + { + "epoch": 0.49, + "learning_rate": 1.0659328328021814e-05, + "loss": 1.0063, + "step": 15467 + }, + { + "epoch": 0.49, + "learning_rate": 1.0658294774619522e-05, + "loss": 1.0366, + "step": 15468 + }, + { + "epoch": 0.49, + "learning_rate": 1.0657261214154454e-05, + "loss": 0.8955, + "step": 15469 + }, + { + "epoch": 0.49, + "learning_rate": 1.0656227646637702e-05, + "loss": 0.9814, + "step": 15470 + }, + { + "epoch": 0.49, + "learning_rate": 1.0655194072080355e-05, + "loss": 0.9629, + "step": 15471 + }, + { + "epoch": 0.49, + "learning_rate": 1.06541604904935e-05, + "loss": 1.0483, + "step": 15472 + }, + { + "epoch": 0.49, + "learning_rate": 1.0653126901888225e-05, + "loss": 1.0, + "step": 15473 + }, + { + "epoch": 0.49, + "learning_rate": 1.0652093306275621e-05, + "loss": 1.0117, + "step": 15474 + }, + { + "epoch": 0.49, + "learning_rate": 1.065105970366678e-05, + "loss": 0.7612, + "step": 15475 + }, + { + "epoch": 0.49, + "learning_rate": 1.0650026094072788e-05, + "loss": 0.8086, + "step": 15476 + }, + { + "epoch": 0.49, + "learning_rate": 1.0648992477504733e-05, + "loss": 1.0073, + "step": 15477 + }, + { + "epoch": 0.49, + "learning_rate": 1.064795885397371e-05, + "loss": 0.9912, + "step": 15478 + }, + { + "epoch": 0.5, + "learning_rate": 1.0646925223490805e-05, + "loss": 1.0469, + "step": 15479 + }, + { + "epoch": 0.5, + "learning_rate": 1.064589158606711e-05, + "loss": 0.9604, + "step": 15480 + }, + { + "epoch": 0.5, + "learning_rate": 1.0644857941713708e-05, + "loss": 1.1318, + "step": 15481 + }, + { + "epoch": 0.5, + "learning_rate": 1.0643824290441697e-05, + "loss": 1.0278, + "step": 15482 + }, + { + "epoch": 0.5, + "learning_rate": 1.0642790632262165e-05, + "loss": 1.041, + "step": 15483 + }, + { + "epoch": 0.5, + "learning_rate": 1.06417569671862e-05, + "loss": 0.9849, + "step": 15484 + }, + { + "epoch": 0.5, + "learning_rate": 1.0640723295224893e-05, + "loss": 1.085, + "step": 15485 + }, + { + "epoch": 0.5, + "learning_rate": 1.0639689616389337e-05, + "loss": 0.8975, + "step": 15486 + }, + { + "epoch": 0.5, + "learning_rate": 1.0638655930690616e-05, + "loss": 0.9458, + "step": 15487 + }, + { + "epoch": 0.5, + "learning_rate": 1.0637622238139821e-05, + "loss": 0.9834, + "step": 15488 + }, + { + "epoch": 0.5, + "learning_rate": 1.0636588538748052e-05, + "loss": 0.877, + "step": 15489 + }, + { + "epoch": 0.5, + "learning_rate": 1.063555483252639e-05, + "loss": 1.0195, + "step": 15490 + }, + { + "epoch": 0.5, + "learning_rate": 1.0634521119485924e-05, + "loss": 0.9941, + "step": 15491 + }, + { + "epoch": 0.5, + "learning_rate": 1.0633487399637754e-05, + "loss": 0.9951, + "step": 15492 + }, + { + "epoch": 0.5, + "learning_rate": 1.0632453672992962e-05, + "loss": 1.0166, + "step": 15493 + }, + { + "epoch": 0.5, + "learning_rate": 1.0631419939562642e-05, + "loss": 0.9521, + "step": 15494 + }, + { + "epoch": 0.5, + "learning_rate": 1.0630386199357887e-05, + "loss": 0.9756, + "step": 15495 + }, + { + "epoch": 0.5, + "learning_rate": 1.0629352452389784e-05, + "loss": 0.9561, + "step": 15496 + }, + { + "epoch": 0.5, + "learning_rate": 1.0628318698669426e-05, + "loss": 0.8872, + "step": 15497 + }, + { + "epoch": 0.5, + "learning_rate": 1.0627284938207903e-05, + "loss": 0.9517, + "step": 15498 + }, + { + "epoch": 0.5, + "learning_rate": 1.0626251171016309e-05, + "loss": 1.0986, + "step": 15499 + }, + { + "epoch": 0.5, + "learning_rate": 1.0625217397105732e-05, + "loss": 1.0, + "step": 15500 + }, + { + "epoch": 0.5, + "learning_rate": 1.0624183616487263e-05, + "loss": 1.0415, + "step": 15501 + }, + { + "epoch": 0.5, + "learning_rate": 1.0623149829171994e-05, + "loss": 0.958, + "step": 15502 + }, + { + "epoch": 0.5, + "learning_rate": 1.0622116035171022e-05, + "loss": 0.9253, + "step": 15503 + }, + { + "epoch": 0.5, + "learning_rate": 1.062108223449543e-05, + "loss": 0.999, + "step": 15504 + }, + { + "epoch": 0.5, + "learning_rate": 1.0620048427156311e-05, + "loss": 1.0493, + "step": 15505 + }, + { + "epoch": 0.5, + "learning_rate": 1.0619014613164765e-05, + "loss": 0.9902, + "step": 15506 + }, + { + "epoch": 0.5, + "learning_rate": 1.061798079253187e-05, + "loss": 1.1064, + "step": 15507 + }, + { + "epoch": 0.5, + "learning_rate": 1.0616946965268729e-05, + "loss": 1.0889, + "step": 15508 + }, + { + "epoch": 0.5, + "learning_rate": 1.0615913131386428e-05, + "loss": 0.9453, + "step": 15509 + }, + { + "epoch": 0.5, + "learning_rate": 1.0614879290896062e-05, + "loss": 0.9082, + "step": 15510 + }, + { + "epoch": 0.5, + "learning_rate": 1.0613845443808721e-05, + "loss": 1.0591, + "step": 15511 + }, + { + "epoch": 0.5, + "learning_rate": 1.06128115901355e-05, + "loss": 0.8823, + "step": 15512 + }, + { + "epoch": 0.5, + "learning_rate": 1.0611777729887485e-05, + "loss": 1.1011, + "step": 15513 + }, + { + "epoch": 0.5, + "learning_rate": 1.0610743863075776e-05, + "loss": 0.9414, + "step": 15514 + }, + { + "epoch": 0.5, + "learning_rate": 1.0609709989711459e-05, + "loss": 1.0454, + "step": 15515 + }, + { + "epoch": 0.5, + "learning_rate": 1.060867610980563e-05, + "loss": 1.064, + "step": 15516 + }, + { + "epoch": 0.5, + "learning_rate": 1.0607642223369376e-05, + "loss": 1.042, + "step": 15517 + }, + { + "epoch": 0.5, + "learning_rate": 1.0606608330413798e-05, + "loss": 0.915, + "step": 15518 + }, + { + "epoch": 0.5, + "learning_rate": 1.0605574430949983e-05, + "loss": 1.0742, + "step": 15519 + }, + { + "epoch": 0.5, + "learning_rate": 1.0604540524989024e-05, + "loss": 0.8892, + "step": 15520 + }, + { + "epoch": 0.5, + "learning_rate": 1.0603506612542016e-05, + "loss": 0.9517, + "step": 15521 + }, + { + "epoch": 0.5, + "learning_rate": 1.0602472693620048e-05, + "loss": 0.9399, + "step": 15522 + }, + { + "epoch": 0.5, + "learning_rate": 1.0601438768234213e-05, + "loss": 1.0474, + "step": 15523 + }, + { + "epoch": 0.5, + "learning_rate": 1.0600404836395608e-05, + "loss": 1.0303, + "step": 15524 + }, + { + "epoch": 0.5, + "learning_rate": 1.0599370898115324e-05, + "loss": 0.9287, + "step": 15525 + }, + { + "epoch": 0.5, + "learning_rate": 1.0598336953404453e-05, + "loss": 0.958, + "step": 15526 + }, + { + "epoch": 0.5, + "learning_rate": 1.0597303002274088e-05, + "loss": 1.0356, + "step": 15527 + }, + { + "epoch": 0.5, + "learning_rate": 1.0596269044735326e-05, + "loss": 1.0889, + "step": 15528 + }, + { + "epoch": 0.5, + "learning_rate": 1.0595235080799253e-05, + "loss": 0.9268, + "step": 15529 + }, + { + "epoch": 0.5, + "learning_rate": 1.0594201110476968e-05, + "loss": 1.0117, + "step": 15530 + }, + { + "epoch": 0.5, + "learning_rate": 1.0593167133779562e-05, + "loss": 1.0117, + "step": 15531 + }, + { + "epoch": 0.5, + "learning_rate": 1.059213315071813e-05, + "loss": 0.9551, + "step": 15532 + }, + { + "epoch": 0.5, + "learning_rate": 1.0591099161303764e-05, + "loss": 1.0454, + "step": 15533 + }, + { + "epoch": 0.5, + "learning_rate": 1.0590065165547561e-05, + "loss": 1.0117, + "step": 15534 + }, + { + "epoch": 0.5, + "learning_rate": 1.0589031163460607e-05, + "loss": 0.9028, + "step": 15535 + }, + { + "epoch": 0.5, + "learning_rate": 1.0587997155054007e-05, + "loss": 0.9746, + "step": 15536 + }, + { + "epoch": 0.5, + "learning_rate": 1.0586963140338842e-05, + "loss": 1.0259, + "step": 15537 + }, + { + "epoch": 0.5, + "learning_rate": 1.0585929119326214e-05, + "loss": 0.9658, + "step": 15538 + }, + { + "epoch": 0.5, + "learning_rate": 1.0584895092027215e-05, + "loss": 1.0693, + "step": 15539 + }, + { + "epoch": 0.5, + "learning_rate": 1.058386105845294e-05, + "loss": 1.0151, + "step": 15540 + }, + { + "epoch": 0.5, + "learning_rate": 1.0582827018614481e-05, + "loss": 0.9946, + "step": 15541 + }, + { + "epoch": 0.5, + "learning_rate": 1.0581792972522932e-05, + "loss": 1.1328, + "step": 15542 + }, + { + "epoch": 0.5, + "learning_rate": 1.0580758920189391e-05, + "loss": 0.8931, + "step": 15543 + }, + { + "epoch": 0.5, + "learning_rate": 1.0579724861624949e-05, + "loss": 1.0068, + "step": 15544 + }, + { + "epoch": 0.5, + "learning_rate": 1.05786907968407e-05, + "loss": 0.9771, + "step": 15545 + }, + { + "epoch": 0.5, + "learning_rate": 1.057765672584774e-05, + "loss": 0.9648, + "step": 15546 + }, + { + "epoch": 0.5, + "learning_rate": 1.057662264865716e-05, + "loss": 1.02, + "step": 15547 + }, + { + "epoch": 0.5, + "learning_rate": 1.0575588565280061e-05, + "loss": 0.7415, + "step": 15548 + }, + { + "epoch": 0.5, + "learning_rate": 1.0574554475727531e-05, + "loss": 0.9224, + "step": 15549 + }, + { + "epoch": 0.5, + "learning_rate": 1.0573520380010669e-05, + "loss": 1.1357, + "step": 15550 + }, + { + "epoch": 0.5, + "learning_rate": 1.0572486278140566e-05, + "loss": 0.991, + "step": 15551 + }, + { + "epoch": 0.5, + "learning_rate": 1.0571452170128322e-05, + "loss": 1.2036, + "step": 15552 + }, + { + "epoch": 0.5, + "learning_rate": 1.0570418055985026e-05, + "loss": 0.9097, + "step": 15553 + }, + { + "epoch": 0.5, + "learning_rate": 1.0569383935721775e-05, + "loss": 1.0371, + "step": 15554 + }, + { + "epoch": 0.5, + "learning_rate": 1.056834980934967e-05, + "loss": 0.8496, + "step": 15555 + }, + { + "epoch": 0.5, + "learning_rate": 1.0567315676879795e-05, + "loss": 0.9487, + "step": 15556 + }, + { + "epoch": 0.5, + "learning_rate": 1.056628153832325e-05, + "loss": 0.9858, + "step": 15557 + }, + { + "epoch": 0.5, + "learning_rate": 1.0565247393691134e-05, + "loss": 0.8896, + "step": 15558 + }, + { + "epoch": 0.5, + "learning_rate": 1.0564213242994539e-05, + "loss": 0.9248, + "step": 15559 + }, + { + "epoch": 0.5, + "learning_rate": 1.056317908624456e-05, + "loss": 1.082, + "step": 15560 + }, + { + "epoch": 0.5, + "learning_rate": 1.056214492345229e-05, + "loss": 0.9814, + "step": 15561 + }, + { + "epoch": 0.5, + "learning_rate": 1.0561110754628829e-05, + "loss": 0.9629, + "step": 15562 + }, + { + "epoch": 0.5, + "learning_rate": 1.0560076579785272e-05, + "loss": 0.9956, + "step": 15563 + }, + { + "epoch": 0.5, + "learning_rate": 1.0559042398932713e-05, + "loss": 1.0303, + "step": 15564 + }, + { + "epoch": 0.5, + "learning_rate": 1.0558008212082245e-05, + "loss": 0.9756, + "step": 15565 + }, + { + "epoch": 0.5, + "learning_rate": 1.0556974019244968e-05, + "loss": 1.0293, + "step": 15566 + }, + { + "epoch": 0.5, + "learning_rate": 1.0555939820431978e-05, + "loss": 0.939, + "step": 15567 + }, + { + "epoch": 0.5, + "learning_rate": 1.0554905615654368e-05, + "loss": 0.5381, + "step": 15568 + }, + { + "epoch": 0.5, + "learning_rate": 1.0553871404923233e-05, + "loss": 0.9541, + "step": 15569 + }, + { + "epoch": 0.5, + "learning_rate": 1.0552837188249674e-05, + "loss": 0.936, + "step": 15570 + }, + { + "epoch": 0.5, + "learning_rate": 1.055180296564478e-05, + "loss": 1.0928, + "step": 15571 + }, + { + "epoch": 0.5, + "learning_rate": 1.0550768737119651e-05, + "loss": 1.0591, + "step": 15572 + }, + { + "epoch": 0.5, + "learning_rate": 1.0549734502685389e-05, + "loss": 0.9453, + "step": 15573 + }, + { + "epoch": 0.5, + "learning_rate": 1.054870026235308e-05, + "loss": 0.998, + "step": 15574 + }, + { + "epoch": 0.5, + "learning_rate": 1.0547666016133824e-05, + "loss": 1.0483, + "step": 15575 + }, + { + "epoch": 0.5, + "learning_rate": 1.0546631764038719e-05, + "loss": 0.9409, + "step": 15576 + }, + { + "epoch": 0.5, + "learning_rate": 1.0545597506078857e-05, + "loss": 0.9673, + "step": 15577 + }, + { + "epoch": 0.5, + "learning_rate": 1.0544563242265342e-05, + "loss": 0.9033, + "step": 15578 + }, + { + "epoch": 0.5, + "learning_rate": 1.0543528972609264e-05, + "loss": 1.104, + "step": 15579 + }, + { + "epoch": 0.5, + "learning_rate": 1.0542494697121722e-05, + "loss": 1.0005, + "step": 15580 + }, + { + "epoch": 0.5, + "learning_rate": 1.0541460415813811e-05, + "loss": 1.0488, + "step": 15581 + }, + { + "epoch": 0.5, + "learning_rate": 1.0540426128696632e-05, + "loss": 0.9014, + "step": 15582 + }, + { + "epoch": 0.5, + "learning_rate": 1.0539391835781276e-05, + "loss": 0.8833, + "step": 15583 + }, + { + "epoch": 0.5, + "learning_rate": 1.0538357537078843e-05, + "loss": 1.0283, + "step": 15584 + }, + { + "epoch": 0.5, + "learning_rate": 1.0537323232600429e-05, + "loss": 0.8833, + "step": 15585 + }, + { + "epoch": 0.5, + "learning_rate": 1.0536288922357134e-05, + "loss": 0.9629, + "step": 15586 + }, + { + "epoch": 0.5, + "learning_rate": 1.053525460636005e-05, + "loss": 1.062, + "step": 15587 + }, + { + "epoch": 0.5, + "learning_rate": 1.0534220284620278e-05, + "loss": 0.9438, + "step": 15588 + }, + { + "epoch": 0.5, + "learning_rate": 1.0533185957148913e-05, + "loss": 0.5317, + "step": 15589 + }, + { + "epoch": 0.5, + "learning_rate": 1.0532151623957052e-05, + "loss": 1.022, + "step": 15590 + }, + { + "epoch": 0.5, + "learning_rate": 1.0531117285055794e-05, + "loss": 0.9639, + "step": 15591 + }, + { + "epoch": 0.5, + "learning_rate": 1.0530082940456235e-05, + "loss": 0.9033, + "step": 15592 + }, + { + "epoch": 0.5, + "learning_rate": 1.0529048590169472e-05, + "loss": 1.0352, + "step": 15593 + }, + { + "epoch": 0.5, + "learning_rate": 1.0528014234206605e-05, + "loss": 1.0625, + "step": 15594 + }, + { + "epoch": 0.5, + "learning_rate": 1.0526979872578726e-05, + "loss": 1.0, + "step": 15595 + }, + { + "epoch": 0.5, + "learning_rate": 1.0525945505296939e-05, + "loss": 1.0508, + "step": 15596 + }, + { + "epoch": 0.5, + "learning_rate": 1.052491113237234e-05, + "loss": 0.9946, + "step": 15597 + }, + { + "epoch": 0.5, + "learning_rate": 1.0523876753816023e-05, + "loss": 1.0664, + "step": 15598 + }, + { + "epoch": 0.5, + "learning_rate": 1.0522842369639088e-05, + "loss": 1.0439, + "step": 15599 + }, + { + "epoch": 0.5, + "learning_rate": 1.0521807979852634e-05, + "loss": 1.0234, + "step": 15600 + }, + { + "epoch": 0.5, + "learning_rate": 1.052077358446776e-05, + "loss": 1.0249, + "step": 15601 + }, + { + "epoch": 0.5, + "learning_rate": 1.0519739183495558e-05, + "loss": 1.1392, + "step": 15602 + }, + { + "epoch": 0.5, + "learning_rate": 1.0518704776947133e-05, + "loss": 1.1011, + "step": 15603 + }, + { + "epoch": 0.5, + "learning_rate": 1.051767036483358e-05, + "loss": 1.0449, + "step": 15604 + }, + { + "epoch": 0.5, + "learning_rate": 1.0516635947165993e-05, + "loss": 0.8701, + "step": 15605 + }, + { + "epoch": 0.5, + "learning_rate": 1.0515601523955476e-05, + "loss": 1.0024, + "step": 15606 + }, + { + "epoch": 0.5, + "learning_rate": 1.0514567095213126e-05, + "loss": 0.5073, + "step": 15607 + }, + { + "epoch": 0.5, + "learning_rate": 1.051353266095004e-05, + "loss": 0.9614, + "step": 15608 + }, + { + "epoch": 0.5, + "learning_rate": 1.0512498221177319e-05, + "loss": 0.9644, + "step": 15609 + }, + { + "epoch": 0.5, + "learning_rate": 1.0511463775906057e-05, + "loss": 0.9766, + "step": 15610 + }, + { + "epoch": 0.5, + "learning_rate": 1.0510429325147354e-05, + "loss": 0.9932, + "step": 15611 + }, + { + "epoch": 0.5, + "learning_rate": 1.0509394868912313e-05, + "loss": 0.918, + "step": 15612 + }, + { + "epoch": 0.5, + "learning_rate": 1.0508360407212024e-05, + "loss": 0.9541, + "step": 15613 + }, + { + "epoch": 0.5, + "learning_rate": 1.0507325940057596e-05, + "loss": 1.0693, + "step": 15614 + }, + { + "epoch": 0.5, + "learning_rate": 1.0506291467460116e-05, + "loss": 0.8384, + "step": 15615 + }, + { + "epoch": 0.5, + "learning_rate": 1.0505256989430692e-05, + "loss": 0.8774, + "step": 15616 + }, + { + "epoch": 0.5, + "learning_rate": 1.0504222505980422e-05, + "loss": 0.9688, + "step": 15617 + }, + { + "epoch": 0.5, + "learning_rate": 1.0503188017120398e-05, + "loss": 1.0317, + "step": 15618 + }, + { + "epoch": 0.5, + "learning_rate": 1.0502153522861726e-05, + "loss": 1.0625, + "step": 15619 + }, + { + "epoch": 0.5, + "learning_rate": 1.0501119023215503e-05, + "loss": 1.0237, + "step": 15620 + }, + { + "epoch": 0.5, + "learning_rate": 1.0500084518192825e-05, + "loss": 1.0801, + "step": 15621 + }, + { + "epoch": 0.5, + "learning_rate": 1.0499050007804796e-05, + "loss": 1.0195, + "step": 15622 + }, + { + "epoch": 0.5, + "learning_rate": 1.0498015492062514e-05, + "loss": 0.8555, + "step": 15623 + }, + { + "epoch": 0.5, + "learning_rate": 1.0496980970977073e-05, + "loss": 1.061, + "step": 15624 + }, + { + "epoch": 0.5, + "learning_rate": 1.0495946444559577e-05, + "loss": 1.0312, + "step": 15625 + }, + { + "epoch": 0.5, + "learning_rate": 1.0494911912821126e-05, + "loss": 0.9531, + "step": 15626 + }, + { + "epoch": 0.5, + "learning_rate": 1.0493877375772818e-05, + "loss": 1.022, + "step": 15627 + }, + { + "epoch": 0.5, + "learning_rate": 1.0492842833425751e-05, + "loss": 0.9858, + "step": 15628 + }, + { + "epoch": 0.5, + "learning_rate": 1.0491808285791028e-05, + "loss": 0.9468, + "step": 15629 + }, + { + "epoch": 0.5, + "learning_rate": 1.0490773732879744e-05, + "loss": 0.9277, + "step": 15630 + }, + { + "epoch": 0.5, + "learning_rate": 1.0489739174703002e-05, + "loss": 1.0293, + "step": 15631 + }, + { + "epoch": 0.5, + "learning_rate": 1.04887046112719e-05, + "loss": 1.0259, + "step": 15632 + }, + { + "epoch": 0.5, + "learning_rate": 1.048767004259754e-05, + "loss": 0.9155, + "step": 15633 + }, + { + "epoch": 0.5, + "learning_rate": 1.0486635468691017e-05, + "loss": 1.002, + "step": 15634 + }, + { + "epoch": 0.5, + "learning_rate": 1.0485600889563439e-05, + "loss": 0.9473, + "step": 15635 + }, + { + "epoch": 0.5, + "learning_rate": 1.0484566305225898e-05, + "loss": 0.9478, + "step": 15636 + }, + { + "epoch": 0.5, + "learning_rate": 1.0483531715689496e-05, + "loss": 1.0615, + "step": 15637 + }, + { + "epoch": 0.5, + "learning_rate": 1.0482497120965336e-05, + "loss": 1.0352, + "step": 15638 + }, + { + "epoch": 0.5, + "learning_rate": 1.0481462521064514e-05, + "loss": 0.8975, + "step": 15639 + }, + { + "epoch": 0.5, + "learning_rate": 1.0480427915998132e-05, + "loss": 0.9341, + "step": 15640 + }, + { + "epoch": 0.5, + "learning_rate": 1.047939330577729e-05, + "loss": 0.9282, + "step": 15641 + }, + { + "epoch": 0.5, + "learning_rate": 1.0478358690413092e-05, + "loss": 0.9761, + "step": 15642 + }, + { + "epoch": 0.5, + "learning_rate": 1.047732406991663e-05, + "loss": 0.9546, + "step": 15643 + }, + { + "epoch": 0.5, + "learning_rate": 1.047628944429901e-05, + "loss": 1.0728, + "step": 15644 + }, + { + "epoch": 0.5, + "learning_rate": 1.0475254813571335e-05, + "loss": 0.9634, + "step": 15645 + }, + { + "epoch": 0.5, + "learning_rate": 1.0474220177744699e-05, + "loss": 1.0386, + "step": 15646 + }, + { + "epoch": 0.5, + "learning_rate": 1.0473185536830206e-05, + "loss": 0.9829, + "step": 15647 + }, + { + "epoch": 0.5, + "learning_rate": 1.0472150890838955e-05, + "loss": 1.0205, + "step": 15648 + }, + { + "epoch": 0.5, + "learning_rate": 1.0471116239782047e-05, + "loss": 1.0806, + "step": 15649 + }, + { + "epoch": 0.5, + "learning_rate": 1.0470081583670583e-05, + "loss": 1.0039, + "step": 15650 + }, + { + "epoch": 0.5, + "learning_rate": 1.0469046922515667e-05, + "loss": 0.9561, + "step": 15651 + }, + { + "epoch": 0.5, + "learning_rate": 1.0468012256328392e-05, + "loss": 0.9897, + "step": 15652 + }, + { + "epoch": 0.5, + "learning_rate": 1.0466977585119869e-05, + "loss": 0.9585, + "step": 15653 + }, + { + "epoch": 0.5, + "learning_rate": 1.046594290890119e-05, + "loss": 1.0107, + "step": 15654 + }, + { + "epoch": 0.5, + "learning_rate": 1.046490822768346e-05, + "loss": 0.9526, + "step": 15655 + }, + { + "epoch": 0.5, + "learning_rate": 1.0463873541477778e-05, + "loss": 1.0972, + "step": 15656 + }, + { + "epoch": 0.5, + "learning_rate": 1.0462838850295247e-05, + "loss": 0.9819, + "step": 15657 + }, + { + "epoch": 0.5, + "learning_rate": 1.0461804154146969e-05, + "loss": 1.0923, + "step": 15658 + }, + { + "epoch": 0.5, + "learning_rate": 1.0460769453044042e-05, + "loss": 0.54, + "step": 15659 + }, + { + "epoch": 0.5, + "learning_rate": 1.0459734746997567e-05, + "loss": 1.0518, + "step": 15660 + }, + { + "epoch": 0.5, + "learning_rate": 1.045870003601865e-05, + "loss": 1.0269, + "step": 15661 + }, + { + "epoch": 0.5, + "learning_rate": 1.0457665320118388e-05, + "loss": 1.0449, + "step": 15662 + }, + { + "epoch": 0.5, + "learning_rate": 1.0456630599307883e-05, + "loss": 0.939, + "step": 15663 + }, + { + "epoch": 0.5, + "learning_rate": 1.0455595873598238e-05, + "loss": 0.9648, + "step": 15664 + }, + { + "epoch": 0.5, + "learning_rate": 1.0454561143000554e-05, + "loss": 1.0176, + "step": 15665 + }, + { + "epoch": 0.5, + "learning_rate": 1.0453526407525932e-05, + "loss": 1.0088, + "step": 15666 + }, + { + "epoch": 0.5, + "learning_rate": 1.0452491667185471e-05, + "loss": 0.9956, + "step": 15667 + }, + { + "epoch": 0.5, + "learning_rate": 1.0451456921990277e-05, + "loss": 1.0508, + "step": 15668 + }, + { + "epoch": 0.5, + "learning_rate": 1.0450422171951453e-05, + "loss": 1.1196, + "step": 15669 + }, + { + "epoch": 0.5, + "learning_rate": 1.0449387417080092e-05, + "loss": 0.9561, + "step": 15670 + }, + { + "epoch": 0.5, + "learning_rate": 1.0448352657387304e-05, + "loss": 0.9414, + "step": 15671 + }, + { + "epoch": 0.5, + "learning_rate": 1.044731789288419e-05, + "loss": 1.0078, + "step": 15672 + }, + { + "epoch": 0.5, + "learning_rate": 1.0446283123581848e-05, + "loss": 0.9897, + "step": 15673 + }, + { + "epoch": 0.5, + "learning_rate": 1.0445248349491382e-05, + "loss": 1.0303, + "step": 15674 + }, + { + "epoch": 0.5, + "learning_rate": 1.0444213570623894e-05, + "loss": 1.0078, + "step": 15675 + }, + { + "epoch": 0.5, + "learning_rate": 1.0443178786990487e-05, + "loss": 1.0728, + "step": 15676 + }, + { + "epoch": 0.5, + "learning_rate": 1.0442143998602262e-05, + "loss": 0.978, + "step": 15677 + }, + { + "epoch": 0.5, + "learning_rate": 1.0441109205470319e-05, + "loss": 0.8853, + "step": 15678 + }, + { + "epoch": 0.5, + "learning_rate": 1.0440074407605762e-05, + "loss": 0.9419, + "step": 15679 + }, + { + "epoch": 0.5, + "learning_rate": 1.0439039605019698e-05, + "loss": 0.8975, + "step": 15680 + }, + { + "epoch": 0.5, + "learning_rate": 1.0438004797723222e-05, + "loss": 1.085, + "step": 15681 + }, + { + "epoch": 0.5, + "learning_rate": 1.0436969985727439e-05, + "loss": 0.9775, + "step": 15682 + }, + { + "epoch": 0.5, + "learning_rate": 1.0435935169043451e-05, + "loss": 0.9771, + "step": 15683 + }, + { + "epoch": 0.5, + "learning_rate": 1.0434900347682363e-05, + "loss": 1.0728, + "step": 15684 + }, + { + "epoch": 0.5, + "learning_rate": 1.0433865521655276e-05, + "loss": 1.0605, + "step": 15685 + }, + { + "epoch": 0.5, + "learning_rate": 1.043283069097329e-05, + "loss": 1.0425, + "step": 15686 + }, + { + "epoch": 0.5, + "learning_rate": 1.0431795855647513e-05, + "loss": 0.9419, + "step": 15687 + }, + { + "epoch": 0.5, + "learning_rate": 1.043076101568904e-05, + "loss": 0.8364, + "step": 15688 + }, + { + "epoch": 0.5, + "learning_rate": 1.0429726171108982e-05, + "loss": 1.1011, + "step": 15689 + }, + { + "epoch": 0.5, + "learning_rate": 1.0428691321918435e-05, + "loss": 0.5098, + "step": 15690 + }, + { + "epoch": 0.5, + "learning_rate": 1.0427656468128508e-05, + "loss": 0.8848, + "step": 15691 + }, + { + "epoch": 0.5, + "learning_rate": 1.0426621609750298e-05, + "loss": 1.0444, + "step": 15692 + }, + { + "epoch": 0.5, + "learning_rate": 1.042558674679491e-05, + "loss": 0.9272, + "step": 15693 + }, + { + "epoch": 0.5, + "learning_rate": 1.0424551879273448e-05, + "loss": 1.0, + "step": 15694 + }, + { + "epoch": 0.5, + "learning_rate": 1.0423517007197015e-05, + "loss": 0.9082, + "step": 15695 + }, + { + "epoch": 0.5, + "learning_rate": 1.042248213057671e-05, + "loss": 0.9688, + "step": 15696 + }, + { + "epoch": 0.5, + "learning_rate": 1.0421447249423642e-05, + "loss": 0.9531, + "step": 15697 + }, + { + "epoch": 0.5, + "learning_rate": 1.0420412363748913e-05, + "loss": 1.0474, + "step": 15698 + }, + { + "epoch": 0.5, + "learning_rate": 1.0419377473563621e-05, + "loss": 1.0508, + "step": 15699 + }, + { + "epoch": 0.5, + "learning_rate": 1.0418342578878875e-05, + "loss": 0.877, + "step": 15700 + }, + { + "epoch": 0.5, + "learning_rate": 1.0417307679705776e-05, + "loss": 0.9575, + "step": 15701 + }, + { + "epoch": 0.5, + "learning_rate": 1.0416272776055427e-05, + "loss": 0.8789, + "step": 15702 + }, + { + "epoch": 0.5, + "learning_rate": 1.0415237867938933e-05, + "loss": 1.0859, + "step": 15703 + }, + { + "epoch": 0.5, + "learning_rate": 1.0414202955367394e-05, + "loss": 1.0293, + "step": 15704 + }, + { + "epoch": 0.5, + "learning_rate": 1.0413168038351917e-05, + "loss": 0.9214, + "step": 15705 + }, + { + "epoch": 0.5, + "learning_rate": 1.0412133116903605e-05, + "loss": 0.9204, + "step": 15706 + }, + { + "epoch": 0.5, + "learning_rate": 1.041109819103356e-05, + "loss": 0.9937, + "step": 15707 + }, + { + "epoch": 0.5, + "learning_rate": 1.0410063260752885e-05, + "loss": 1.0156, + "step": 15708 + }, + { + "epoch": 0.5, + "learning_rate": 1.0409028326072685e-05, + "loss": 1.1538, + "step": 15709 + }, + { + "epoch": 0.5, + "learning_rate": 1.0407993387004067e-05, + "loss": 0.9741, + "step": 15710 + }, + { + "epoch": 0.5, + "learning_rate": 1.0406958443558128e-05, + "loss": 1.0117, + "step": 15711 + }, + { + "epoch": 0.5, + "learning_rate": 1.0405923495745978e-05, + "loss": 0.979, + "step": 15712 + }, + { + "epoch": 0.5, + "learning_rate": 1.0404888543578715e-05, + "loss": 1.1543, + "step": 15713 + }, + { + "epoch": 0.5, + "learning_rate": 1.040385358706745e-05, + "loss": 1.0718, + "step": 15714 + }, + { + "epoch": 0.5, + "learning_rate": 1.0402818626223278e-05, + "loss": 0.9258, + "step": 15715 + }, + { + "epoch": 0.5, + "learning_rate": 1.0401783661057312e-05, + "loss": 1.061, + "step": 15716 + }, + { + "epoch": 0.5, + "learning_rate": 1.0400748691580651e-05, + "loss": 0.9595, + "step": 15717 + }, + { + "epoch": 0.5, + "learning_rate": 1.03997137178044e-05, + "loss": 1.0298, + "step": 15718 + }, + { + "epoch": 0.5, + "learning_rate": 1.0398678739739662e-05, + "loss": 0.9463, + "step": 15719 + }, + { + "epoch": 0.5, + "learning_rate": 1.0397643757397543e-05, + "loss": 0.9023, + "step": 15720 + }, + { + "epoch": 0.5, + "learning_rate": 1.039660877078915e-05, + "loss": 1.0806, + "step": 15721 + }, + { + "epoch": 0.5, + "learning_rate": 1.039557377992558e-05, + "loss": 0.9268, + "step": 15722 + }, + { + "epoch": 0.5, + "learning_rate": 1.0394538784817941e-05, + "loss": 0.9746, + "step": 15723 + }, + { + "epoch": 0.5, + "learning_rate": 1.039350378547734e-05, + "loss": 1.1689, + "step": 15724 + }, + { + "epoch": 0.5, + "learning_rate": 1.0392468781914877e-05, + "loss": 1.0322, + "step": 15725 + }, + { + "epoch": 0.5, + "learning_rate": 1.0391433774141658e-05, + "loss": 1.0254, + "step": 15726 + }, + { + "epoch": 0.5, + "learning_rate": 1.0390398762168787e-05, + "loss": 0.8462, + "step": 15727 + }, + { + "epoch": 0.5, + "learning_rate": 1.038936374600737e-05, + "loss": 1.1206, + "step": 15728 + }, + { + "epoch": 0.5, + "learning_rate": 1.0388328725668515e-05, + "loss": 1.1191, + "step": 15729 + }, + { + "epoch": 0.5, + "learning_rate": 1.0387293701163319e-05, + "loss": 0.9478, + "step": 15730 + }, + { + "epoch": 0.5, + "learning_rate": 1.0386258672502889e-05, + "loss": 1.0093, + "step": 15731 + }, + { + "epoch": 0.5, + "learning_rate": 1.0385223639698331e-05, + "loss": 0.9194, + "step": 15732 + }, + { + "epoch": 0.5, + "learning_rate": 1.0384188602760755e-05, + "loss": 1.041, + "step": 15733 + }, + { + "epoch": 0.5, + "learning_rate": 1.0383153561701256e-05, + "loss": 0.8237, + "step": 15734 + }, + { + "epoch": 0.5, + "learning_rate": 1.0382118516530943e-05, + "loss": 0.9897, + "step": 15735 + }, + { + "epoch": 0.5, + "learning_rate": 1.0381083467260922e-05, + "loss": 1.0835, + "step": 15736 + }, + { + "epoch": 0.5, + "learning_rate": 1.03800484139023e-05, + "loss": 0.5691, + "step": 15737 + }, + { + "epoch": 0.5, + "learning_rate": 1.0379013356466175e-05, + "loss": 0.9155, + "step": 15738 + }, + { + "epoch": 0.5, + "learning_rate": 1.0377978294963658e-05, + "loss": 1.083, + "step": 15739 + }, + { + "epoch": 0.5, + "learning_rate": 1.0376943229405851e-05, + "loss": 1.1123, + "step": 15740 + }, + { + "epoch": 0.5, + "learning_rate": 1.0375908159803862e-05, + "loss": 1.085, + "step": 15741 + }, + { + "epoch": 0.5, + "learning_rate": 1.0374873086168793e-05, + "loss": 0.9624, + "step": 15742 + }, + { + "epoch": 0.5, + "learning_rate": 1.0373838008511752e-05, + "loss": 0.8809, + "step": 15743 + }, + { + "epoch": 0.5, + "learning_rate": 1.0372802926843843e-05, + "loss": 1.0278, + "step": 15744 + }, + { + "epoch": 0.5, + "learning_rate": 1.0371767841176173e-05, + "loss": 0.9878, + "step": 15745 + }, + { + "epoch": 0.5, + "learning_rate": 1.0370732751519841e-05, + "loss": 0.9829, + "step": 15746 + }, + { + "epoch": 0.5, + "learning_rate": 1.0369697657885957e-05, + "loss": 1.0352, + "step": 15747 + }, + { + "epoch": 0.5, + "learning_rate": 1.0368662560285632e-05, + "loss": 0.8975, + "step": 15748 + }, + { + "epoch": 0.5, + "learning_rate": 1.036762745872996e-05, + "loss": 1.0259, + "step": 15749 + }, + { + "epoch": 0.5, + "learning_rate": 1.0366592353230054e-05, + "loss": 1.0908, + "step": 15750 + }, + { + "epoch": 0.5, + "learning_rate": 1.0365557243797018e-05, + "loss": 0.9834, + "step": 15751 + }, + { + "epoch": 0.5, + "learning_rate": 1.0364522130441958e-05, + "loss": 0.9507, + "step": 15752 + }, + { + "epoch": 0.5, + "learning_rate": 1.036348701317598e-05, + "loss": 0.915, + "step": 15753 + }, + { + "epoch": 0.5, + "learning_rate": 1.0362451892010184e-05, + "loss": 1.0542, + "step": 15754 + }, + { + "epoch": 0.5, + "learning_rate": 1.0361416766955685e-05, + "loss": 0.8184, + "step": 15755 + }, + { + "epoch": 0.5, + "learning_rate": 1.0360381638023583e-05, + "loss": 1.0044, + "step": 15756 + }, + { + "epoch": 0.5, + "learning_rate": 1.0359346505224983e-05, + "loss": 1.0586, + "step": 15757 + }, + { + "epoch": 0.5, + "learning_rate": 1.0358311368570992e-05, + "loss": 1.0278, + "step": 15758 + }, + { + "epoch": 0.5, + "learning_rate": 1.035727622807272e-05, + "loss": 0.9106, + "step": 15759 + }, + { + "epoch": 0.5, + "learning_rate": 1.035624108374127e-05, + "loss": 1.0283, + "step": 15760 + }, + { + "epoch": 0.5, + "learning_rate": 1.0355205935587744e-05, + "loss": 0.9653, + "step": 15761 + }, + { + "epoch": 0.5, + "learning_rate": 1.0354170783623252e-05, + "loss": 0.978, + "step": 15762 + }, + { + "epoch": 0.5, + "learning_rate": 1.0353135627858901e-05, + "loss": 1.1826, + "step": 15763 + }, + { + "epoch": 0.5, + "learning_rate": 1.0352100468305794e-05, + "loss": 1.0747, + "step": 15764 + }, + { + "epoch": 0.5, + "learning_rate": 1.035106530497504e-05, + "loss": 1.0752, + "step": 15765 + }, + { + "epoch": 0.5, + "learning_rate": 1.0350030137877743e-05, + "loss": 0.9902, + "step": 15766 + }, + { + "epoch": 0.5, + "learning_rate": 1.0348994967025012e-05, + "loss": 0.9497, + "step": 15767 + }, + { + "epoch": 0.5, + "learning_rate": 1.0347959792427948e-05, + "loss": 1.0396, + "step": 15768 + }, + { + "epoch": 0.5, + "learning_rate": 1.0346924614097662e-05, + "loss": 0.8345, + "step": 15769 + }, + { + "epoch": 0.5, + "learning_rate": 1.0345889432045262e-05, + "loss": 1.0684, + "step": 15770 + }, + { + "epoch": 0.5, + "learning_rate": 1.0344854246281845e-05, + "loss": 1.0693, + "step": 15771 + }, + { + "epoch": 0.5, + "learning_rate": 1.0343819056818527e-05, + "loss": 1.0791, + "step": 15772 + }, + { + "epoch": 0.5, + "learning_rate": 1.034278386366641e-05, + "loss": 1.0864, + "step": 15773 + }, + { + "epoch": 0.5, + "learning_rate": 1.0341748666836604e-05, + "loss": 0.5542, + "step": 15774 + }, + { + "epoch": 0.5, + "learning_rate": 1.0340713466340211e-05, + "loss": 0.8408, + "step": 15775 + }, + { + "epoch": 0.5, + "learning_rate": 1.0339678262188339e-05, + "loss": 0.9951, + "step": 15776 + }, + { + "epoch": 0.5, + "learning_rate": 1.0338643054392098e-05, + "loss": 0.8604, + "step": 15777 + }, + { + "epoch": 0.5, + "learning_rate": 1.033760784296259e-05, + "loss": 1.0327, + "step": 15778 + }, + { + "epoch": 0.5, + "learning_rate": 1.0336572627910923e-05, + "loss": 0.9614, + "step": 15779 + }, + { + "epoch": 0.5, + "learning_rate": 1.0335537409248204e-05, + "loss": 1.0122, + "step": 15780 + }, + { + "epoch": 0.5, + "learning_rate": 1.033450218698554e-05, + "loss": 0.9268, + "step": 15781 + }, + { + "epoch": 0.5, + "learning_rate": 1.033346696113404e-05, + "loss": 1.0518, + "step": 15782 + }, + { + "epoch": 0.5, + "learning_rate": 1.0332431731704806e-05, + "loss": 1.1538, + "step": 15783 + }, + { + "epoch": 0.5, + "learning_rate": 1.033139649870895e-05, + "loss": 1.0005, + "step": 15784 + }, + { + "epoch": 0.5, + "learning_rate": 1.0330361262157575e-05, + "loss": 0.8989, + "step": 15785 + }, + { + "epoch": 0.5, + "learning_rate": 1.0329326022061787e-05, + "loss": 0.9653, + "step": 15786 + }, + { + "epoch": 0.5, + "learning_rate": 1.03282907784327e-05, + "loss": 0.9409, + "step": 15787 + }, + { + "epoch": 0.5, + "learning_rate": 1.0327255531281413e-05, + "loss": 0.9492, + "step": 15788 + }, + { + "epoch": 0.5, + "learning_rate": 1.0326220280619036e-05, + "loss": 0.855, + "step": 15789 + }, + { + "epoch": 0.5, + "learning_rate": 1.0325185026456679e-05, + "loss": 1.0542, + "step": 15790 + }, + { + "epoch": 0.51, + "learning_rate": 1.0324149768805445e-05, + "loss": 0.9541, + "step": 15791 + }, + { + "epoch": 0.51, + "learning_rate": 1.0323114507676442e-05, + "loss": 1.0532, + "step": 15792 + }, + { + "epoch": 0.51, + "learning_rate": 1.032207924308078e-05, + "loss": 0.5532, + "step": 15793 + }, + { + "epoch": 0.51, + "learning_rate": 1.032104397502956e-05, + "loss": 1.0522, + "step": 15794 + }, + { + "epoch": 0.51, + "learning_rate": 1.0320008703533897e-05, + "loss": 0.9883, + "step": 15795 + }, + { + "epoch": 0.51, + "learning_rate": 1.0318973428604895e-05, + "loss": 1.0991, + "step": 15796 + }, + { + "epoch": 0.51, + "learning_rate": 1.031793815025366e-05, + "loss": 0.9912, + "step": 15797 + }, + { + "epoch": 0.51, + "learning_rate": 1.03169028684913e-05, + "loss": 0.9932, + "step": 15798 + }, + { + "epoch": 0.51, + "learning_rate": 1.0315867583328925e-05, + "loss": 1.0322, + "step": 15799 + }, + { + "epoch": 0.51, + "learning_rate": 1.0314832294777638e-05, + "loss": 0.874, + "step": 15800 + }, + { + "epoch": 0.51, + "learning_rate": 1.0313797002848551e-05, + "loss": 1.0015, + "step": 15801 + }, + { + "epoch": 0.51, + "learning_rate": 1.0312761707552768e-05, + "loss": 1.1001, + "step": 15802 + }, + { + "epoch": 0.51, + "learning_rate": 1.0311726408901397e-05, + "loss": 0.9858, + "step": 15803 + }, + { + "epoch": 0.51, + "learning_rate": 1.0310691106905551e-05, + "loss": 1.0635, + "step": 15804 + }, + { + "epoch": 0.51, + "learning_rate": 1.0309655801576329e-05, + "loss": 0.8975, + "step": 15805 + }, + { + "epoch": 0.51, + "learning_rate": 1.0308620492924843e-05, + "loss": 0.8462, + "step": 15806 + }, + { + "epoch": 0.51, + "learning_rate": 1.0307585180962201e-05, + "loss": 0.9102, + "step": 15807 + }, + { + "epoch": 0.51, + "learning_rate": 1.0306549865699513e-05, + "loss": 1.0815, + "step": 15808 + }, + { + "epoch": 0.51, + "learning_rate": 1.0305514547147883e-05, + "loss": 1.0132, + "step": 15809 + }, + { + "epoch": 0.51, + "learning_rate": 1.0304479225318418e-05, + "loss": 0.959, + "step": 15810 + }, + { + "epoch": 0.51, + "learning_rate": 1.030344390022223e-05, + "loss": 0.9585, + "step": 15811 + }, + { + "epoch": 0.51, + "learning_rate": 1.0302408571870425e-05, + "loss": 0.8955, + "step": 15812 + }, + { + "epoch": 0.51, + "learning_rate": 1.0301373240274111e-05, + "loss": 0.9668, + "step": 15813 + }, + { + "epoch": 0.51, + "learning_rate": 1.0300337905444393e-05, + "loss": 1.0342, + "step": 15814 + }, + { + "epoch": 0.51, + "learning_rate": 1.0299302567392382e-05, + "loss": 0.5847, + "step": 15815 + }, + { + "epoch": 0.51, + "learning_rate": 1.0298267226129188e-05, + "loss": 0.9941, + "step": 15816 + }, + { + "epoch": 0.51, + "learning_rate": 1.0297231881665917e-05, + "loss": 1.0107, + "step": 15817 + }, + { + "epoch": 0.51, + "learning_rate": 1.0296196534013674e-05, + "loss": 1.0103, + "step": 15818 + }, + { + "epoch": 0.51, + "learning_rate": 1.0295161183183571e-05, + "loss": 1.1016, + "step": 15819 + }, + { + "epoch": 0.51, + "learning_rate": 1.0294125829186719e-05, + "loss": 1.0415, + "step": 15820 + }, + { + "epoch": 0.51, + "learning_rate": 1.0293090472034217e-05, + "loss": 1.0234, + "step": 15821 + }, + { + "epoch": 0.51, + "learning_rate": 1.029205511173718e-05, + "loss": 1.0225, + "step": 15822 + }, + { + "epoch": 0.51, + "learning_rate": 1.029101974830672e-05, + "loss": 1.022, + "step": 15823 + }, + { + "epoch": 0.51, + "learning_rate": 1.0289984381753935e-05, + "loss": 0.9312, + "step": 15824 + }, + { + "epoch": 0.51, + "learning_rate": 1.0288949012089936e-05, + "loss": 0.9561, + "step": 15825 + }, + { + "epoch": 0.51, + "learning_rate": 1.0287913639325841e-05, + "loss": 0.9668, + "step": 15826 + }, + { + "epoch": 0.51, + "learning_rate": 1.0286878263472745e-05, + "loss": 1.0576, + "step": 15827 + }, + { + "epoch": 0.51, + "learning_rate": 1.0285842884541766e-05, + "loss": 0.8867, + "step": 15828 + }, + { + "epoch": 0.51, + "learning_rate": 1.028480750254401e-05, + "loss": 1.063, + "step": 15829 + }, + { + "epoch": 0.51, + "learning_rate": 1.0283772117490583e-05, + "loss": 0.9199, + "step": 15830 + }, + { + "epoch": 0.51, + "learning_rate": 1.0282736729392598e-05, + "loss": 1.0352, + "step": 15831 + }, + { + "epoch": 0.51, + "learning_rate": 1.0281701338261157e-05, + "loss": 0.9072, + "step": 15832 + }, + { + "epoch": 0.51, + "learning_rate": 1.0280665944107372e-05, + "loss": 1.0259, + "step": 15833 + }, + { + "epoch": 0.51, + "learning_rate": 1.0279630546942353e-05, + "loss": 0.9067, + "step": 15834 + }, + { + "epoch": 0.51, + "learning_rate": 1.0278595146777213e-05, + "loss": 0.8887, + "step": 15835 + }, + { + "epoch": 0.51, + "learning_rate": 1.027755974362305e-05, + "loss": 0.531, + "step": 15836 + }, + { + "epoch": 0.51, + "learning_rate": 1.0276524337490979e-05, + "loss": 0.9517, + "step": 15837 + }, + { + "epoch": 0.51, + "learning_rate": 1.0275488928392107e-05, + "loss": 0.8022, + "step": 15838 + }, + { + "epoch": 0.51, + "learning_rate": 1.0274453516337545e-05, + "loss": 0.9263, + "step": 15839 + }, + { + "epoch": 0.51, + "learning_rate": 1.02734181013384e-05, + "loss": 0.8862, + "step": 15840 + }, + { + "epoch": 0.51, + "learning_rate": 1.0272382683405781e-05, + "loss": 1.2168, + "step": 15841 + }, + { + "epoch": 0.51, + "learning_rate": 1.02713472625508e-05, + "loss": 1.041, + "step": 15842 + }, + { + "epoch": 0.51, + "learning_rate": 1.027031183878456e-05, + "loss": 0.9956, + "step": 15843 + }, + { + "epoch": 0.51, + "learning_rate": 1.0269276412118174e-05, + "loss": 1.0015, + "step": 15844 + }, + { + "epoch": 0.51, + "learning_rate": 1.026824098256275e-05, + "loss": 0.9956, + "step": 15845 + }, + { + "epoch": 0.51, + "learning_rate": 1.0267205550129398e-05, + "loss": 1.0791, + "step": 15846 + }, + { + "epoch": 0.51, + "learning_rate": 1.0266170114829226e-05, + "loss": 1.02, + "step": 15847 + }, + { + "epoch": 0.51, + "learning_rate": 1.0265134676673341e-05, + "loss": 1.0337, + "step": 15848 + }, + { + "epoch": 0.51, + "learning_rate": 1.0264099235672855e-05, + "loss": 0.9092, + "step": 15849 + }, + { + "epoch": 0.51, + "learning_rate": 1.0263063791838882e-05, + "loss": 0.897, + "step": 15850 + }, + { + "epoch": 0.51, + "learning_rate": 1.0262028345182517e-05, + "loss": 1.021, + "step": 15851 + }, + { + "epoch": 0.51, + "learning_rate": 1.0260992895714884e-05, + "loss": 0.9956, + "step": 15852 + }, + { + "epoch": 0.51, + "learning_rate": 1.0259957443447083e-05, + "loss": 0.8916, + "step": 15853 + }, + { + "epoch": 0.51, + "learning_rate": 1.0258921988390229e-05, + "loss": 1.0605, + "step": 15854 + }, + { + "epoch": 0.51, + "learning_rate": 1.0257886530555428e-05, + "loss": 0.9595, + "step": 15855 + }, + { + "epoch": 0.51, + "learning_rate": 1.0256851069953789e-05, + "loss": 1.0127, + "step": 15856 + }, + { + "epoch": 0.51, + "learning_rate": 1.0255815606596422e-05, + "loss": 0.9478, + "step": 15857 + }, + { + "epoch": 0.51, + "learning_rate": 1.0254780140494435e-05, + "loss": 0.979, + "step": 15858 + }, + { + "epoch": 0.51, + "learning_rate": 1.0253744671658943e-05, + "loss": 0.9424, + "step": 15859 + }, + { + "epoch": 0.51, + "learning_rate": 1.0252709200101048e-05, + "loss": 0.9414, + "step": 15860 + }, + { + "epoch": 0.51, + "learning_rate": 1.0251673725831868e-05, + "loss": 1.0938, + "step": 15861 + }, + { + "epoch": 0.51, + "learning_rate": 1.0250638248862504e-05, + "loss": 0.9067, + "step": 15862 + }, + { + "epoch": 0.51, + "learning_rate": 1.0249602769204068e-05, + "loss": 1.0239, + "step": 15863 + }, + { + "epoch": 0.51, + "learning_rate": 1.024856728686767e-05, + "loss": 1.0078, + "step": 15864 + }, + { + "epoch": 0.51, + "learning_rate": 1.0247531801864424e-05, + "loss": 1.0137, + "step": 15865 + }, + { + "epoch": 0.51, + "learning_rate": 1.0246496314205432e-05, + "loss": 0.96, + "step": 15866 + }, + { + "epoch": 0.51, + "learning_rate": 1.024546082390181e-05, + "loss": 0.9648, + "step": 15867 + }, + { + "epoch": 0.51, + "learning_rate": 1.0244425330964665e-05, + "loss": 1.0732, + "step": 15868 + }, + { + "epoch": 0.51, + "learning_rate": 1.0243389835405105e-05, + "loss": 1.0083, + "step": 15869 + }, + { + "epoch": 0.51, + "learning_rate": 1.024235433723424e-05, + "loss": 0.8755, + "step": 15870 + }, + { + "epoch": 0.51, + "learning_rate": 1.0241318836463186e-05, + "loss": 1.0215, + "step": 15871 + }, + { + "epoch": 0.51, + "learning_rate": 1.0240283333103043e-05, + "loss": 0.9868, + "step": 15872 + }, + { + "epoch": 0.51, + "learning_rate": 1.0239247827164931e-05, + "loss": 1.0806, + "step": 15873 + }, + { + "epoch": 0.51, + "learning_rate": 1.023821231865995e-05, + "loss": 1.0098, + "step": 15874 + }, + { + "epoch": 0.51, + "learning_rate": 1.0237176807599217e-05, + "loss": 0.9951, + "step": 15875 + }, + { + "epoch": 0.51, + "learning_rate": 1.0236141293993837e-05, + "loss": 1.0488, + "step": 15876 + }, + { + "epoch": 0.51, + "learning_rate": 1.0235105777854923e-05, + "loss": 1.0273, + "step": 15877 + }, + { + "epoch": 0.51, + "learning_rate": 1.0234070259193585e-05, + "loss": 0.9395, + "step": 15878 + }, + { + "epoch": 0.51, + "learning_rate": 1.0233034738020933e-05, + "loss": 0.8916, + "step": 15879 + }, + { + "epoch": 0.51, + "learning_rate": 1.0231999214348076e-05, + "loss": 1.0278, + "step": 15880 + }, + { + "epoch": 0.51, + "learning_rate": 1.0230963688186122e-05, + "loss": 0.9194, + "step": 15881 + }, + { + "epoch": 0.51, + "learning_rate": 1.0229928159546186e-05, + "loss": 1.0063, + "step": 15882 + }, + { + "epoch": 0.51, + "learning_rate": 1.022889262843937e-05, + "loss": 1.0342, + "step": 15883 + }, + { + "epoch": 0.51, + "learning_rate": 1.0227857094876796e-05, + "loss": 1.0903, + "step": 15884 + }, + { + "epoch": 0.51, + "learning_rate": 1.0226821558869565e-05, + "loss": 0.9614, + "step": 15885 + }, + { + "epoch": 0.51, + "learning_rate": 1.022578602042879e-05, + "loss": 1.0107, + "step": 15886 + }, + { + "epoch": 0.51, + "learning_rate": 1.0224750479565581e-05, + "loss": 1.043, + "step": 15887 + }, + { + "epoch": 0.51, + "learning_rate": 1.0223714936291046e-05, + "loss": 0.9023, + "step": 15888 + }, + { + "epoch": 0.51, + "learning_rate": 1.02226793906163e-05, + "loss": 1.063, + "step": 15889 + }, + { + "epoch": 0.51, + "learning_rate": 1.022164384255245e-05, + "loss": 1.0117, + "step": 15890 + }, + { + "epoch": 0.51, + "learning_rate": 1.0220608292110605e-05, + "loss": 0.9414, + "step": 15891 + }, + { + "epoch": 0.51, + "learning_rate": 1.0219572739301879e-05, + "loss": 0.9658, + "step": 15892 + }, + { + "epoch": 0.51, + "learning_rate": 1.021853718413738e-05, + "loss": 0.9878, + "step": 15893 + }, + { + "epoch": 0.51, + "learning_rate": 1.0217501626628216e-05, + "loss": 0.8843, + "step": 15894 + }, + { + "epoch": 0.51, + "learning_rate": 1.0216466066785507e-05, + "loss": 0.9893, + "step": 15895 + }, + { + "epoch": 0.51, + "learning_rate": 1.0215430504620352e-05, + "loss": 1.0479, + "step": 15896 + }, + { + "epoch": 0.51, + "learning_rate": 1.0214394940143866e-05, + "loss": 0.9634, + "step": 15897 + }, + { + "epoch": 0.51, + "learning_rate": 1.021335937336716e-05, + "loss": 0.9399, + "step": 15898 + }, + { + "epoch": 0.51, + "learning_rate": 1.0212323804301348e-05, + "loss": 0.9961, + "step": 15899 + }, + { + "epoch": 0.51, + "learning_rate": 1.021128823295753e-05, + "loss": 0.9131, + "step": 15900 + }, + { + "epoch": 0.51, + "learning_rate": 1.0210252659346826e-05, + "loss": 1.0029, + "step": 15901 + }, + { + "epoch": 0.51, + "learning_rate": 1.0209217083480345e-05, + "loss": 0.8662, + "step": 15902 + }, + { + "epoch": 0.51, + "learning_rate": 1.0208181505369197e-05, + "loss": 0.8413, + "step": 15903 + }, + { + "epoch": 0.51, + "learning_rate": 1.0207145925024489e-05, + "loss": 0.9941, + "step": 15904 + }, + { + "epoch": 0.51, + "learning_rate": 1.0206110342457334e-05, + "loss": 1.0088, + "step": 15905 + }, + { + "epoch": 0.51, + "learning_rate": 1.0205074757678846e-05, + "loss": 0.8118, + "step": 15906 + }, + { + "epoch": 0.51, + "learning_rate": 1.020403917070013e-05, + "loss": 1.0684, + "step": 15907 + }, + { + "epoch": 0.51, + "learning_rate": 1.0203003581532303e-05, + "loss": 0.9937, + "step": 15908 + }, + { + "epoch": 0.51, + "learning_rate": 1.0201967990186468e-05, + "loss": 1.0811, + "step": 15909 + }, + { + "epoch": 0.51, + "learning_rate": 1.0200932396673744e-05, + "loss": 0.978, + "step": 15910 + }, + { + "epoch": 0.51, + "learning_rate": 1.0199896801005236e-05, + "loss": 0.9521, + "step": 15911 + }, + { + "epoch": 0.51, + "learning_rate": 1.0198861203192059e-05, + "loss": 0.938, + "step": 15912 + }, + { + "epoch": 0.51, + "learning_rate": 1.0197825603245317e-05, + "loss": 1.0059, + "step": 15913 + }, + { + "epoch": 0.51, + "learning_rate": 1.0196790001176127e-05, + "loss": 1.0303, + "step": 15914 + }, + { + "epoch": 0.51, + "learning_rate": 1.01957543969956e-05, + "loss": 0.9072, + "step": 15915 + }, + { + "epoch": 0.51, + "learning_rate": 1.0194718790714841e-05, + "loss": 0.9897, + "step": 15916 + }, + { + "epoch": 0.51, + "learning_rate": 1.0193683182344967e-05, + "loss": 0.9531, + "step": 15917 + }, + { + "epoch": 0.51, + "learning_rate": 1.019264757189709e-05, + "loss": 0.9209, + "step": 15918 + }, + { + "epoch": 0.51, + "learning_rate": 1.0191611959382312e-05, + "loss": 0.8369, + "step": 15919 + }, + { + "epoch": 0.51, + "learning_rate": 1.0190576344811754e-05, + "loss": 1.0352, + "step": 15920 + }, + { + "epoch": 0.51, + "learning_rate": 1.0189540728196521e-05, + "loss": 0.9893, + "step": 15921 + }, + { + "epoch": 0.51, + "learning_rate": 1.0188505109547726e-05, + "loss": 0.9121, + "step": 15922 + }, + { + "epoch": 0.51, + "learning_rate": 1.018746948887648e-05, + "loss": 0.9946, + "step": 15923 + }, + { + "epoch": 0.51, + "learning_rate": 1.0186433866193893e-05, + "loss": 0.791, + "step": 15924 + }, + { + "epoch": 0.51, + "learning_rate": 1.0185398241511077e-05, + "loss": 0.8999, + "step": 15925 + }, + { + "epoch": 0.51, + "learning_rate": 1.0184362614839146e-05, + "loss": 1.0107, + "step": 15926 + }, + { + "epoch": 0.51, + "learning_rate": 1.0183326986189204e-05, + "loss": 0.9561, + "step": 15927 + }, + { + "epoch": 0.51, + "learning_rate": 1.0182291355572367e-05, + "loss": 1.0459, + "step": 15928 + }, + { + "epoch": 0.51, + "learning_rate": 1.0181255722999748e-05, + "loss": 0.9839, + "step": 15929 + }, + { + "epoch": 0.51, + "learning_rate": 1.0180220088482453e-05, + "loss": 0.8906, + "step": 15930 + }, + { + "epoch": 0.51, + "learning_rate": 1.01791844520316e-05, + "loss": 0.9229, + "step": 15931 + }, + { + "epoch": 0.51, + "learning_rate": 1.017814881365829e-05, + "loss": 1.0181, + "step": 15932 + }, + { + "epoch": 0.51, + "learning_rate": 1.0177113173373645e-05, + "loss": 0.8784, + "step": 15933 + }, + { + "epoch": 0.51, + "learning_rate": 1.017607753118877e-05, + "loss": 0.9839, + "step": 15934 + }, + { + "epoch": 0.51, + "learning_rate": 1.0175041887114779e-05, + "loss": 0.9688, + "step": 15935 + }, + { + "epoch": 0.51, + "learning_rate": 1.0174006241162782e-05, + "loss": 1.0723, + "step": 15936 + }, + { + "epoch": 0.51, + "learning_rate": 1.017297059334389e-05, + "loss": 1.0181, + "step": 15937 + }, + { + "epoch": 0.51, + "learning_rate": 1.0171934943669215e-05, + "loss": 0.9185, + "step": 15938 + }, + { + "epoch": 0.51, + "learning_rate": 1.0170899292149868e-05, + "loss": 0.7952, + "step": 15939 + }, + { + "epoch": 0.51, + "learning_rate": 1.0169863638796964e-05, + "loss": 0.8838, + "step": 15940 + }, + { + "epoch": 0.51, + "learning_rate": 1.0168827983621608e-05, + "loss": 0.9712, + "step": 15941 + }, + { + "epoch": 0.51, + "learning_rate": 1.0167792326634913e-05, + "loss": 0.9868, + "step": 15942 + }, + { + "epoch": 0.51, + "learning_rate": 1.0166756667847995e-05, + "loss": 0.8276, + "step": 15943 + }, + { + "epoch": 0.51, + "learning_rate": 1.0165721007271963e-05, + "loss": 1.125, + "step": 15944 + }, + { + "epoch": 0.51, + "learning_rate": 1.0164685344917923e-05, + "loss": 0.9595, + "step": 15945 + }, + { + "epoch": 0.51, + "learning_rate": 1.0163649680796996e-05, + "loss": 1.0298, + "step": 15946 + }, + { + "epoch": 0.51, + "learning_rate": 1.0162614014920289e-05, + "loss": 0.9028, + "step": 15947 + }, + { + "epoch": 0.51, + "learning_rate": 1.0161578347298913e-05, + "loss": 0.9443, + "step": 15948 + }, + { + "epoch": 0.51, + "learning_rate": 1.016054267794398e-05, + "loss": 0.9438, + "step": 15949 + }, + { + "epoch": 0.51, + "learning_rate": 1.01595070068666e-05, + "loss": 1.0513, + "step": 15950 + }, + { + "epoch": 0.51, + "learning_rate": 1.0158471334077889e-05, + "loss": 0.8838, + "step": 15951 + }, + { + "epoch": 0.51, + "learning_rate": 1.0157435659588957e-05, + "loss": 0.9556, + "step": 15952 + }, + { + "epoch": 0.51, + "learning_rate": 1.0156399983410912e-05, + "loss": 0.5422, + "step": 15953 + }, + { + "epoch": 0.51, + "learning_rate": 1.0155364305554868e-05, + "loss": 0.8662, + "step": 15954 + }, + { + "epoch": 0.51, + "learning_rate": 1.015432862603194e-05, + "loss": 0.8833, + "step": 15955 + }, + { + "epoch": 0.51, + "learning_rate": 1.0153292944853233e-05, + "loss": 1.0425, + "step": 15956 + }, + { + "epoch": 0.51, + "learning_rate": 1.0152257262029866e-05, + "loss": 1.0542, + "step": 15957 + }, + { + "epoch": 0.51, + "learning_rate": 1.0151221577572943e-05, + "loss": 0.8999, + "step": 15958 + }, + { + "epoch": 0.51, + "learning_rate": 1.0150185891493582e-05, + "loss": 0.9922, + "step": 15959 + }, + { + "epoch": 0.51, + "learning_rate": 1.0149150203802894e-05, + "loss": 0.9229, + "step": 15960 + }, + { + "epoch": 0.51, + "learning_rate": 1.0148114514511987e-05, + "loss": 1.0669, + "step": 15961 + }, + { + "epoch": 0.51, + "learning_rate": 1.0147078823631974e-05, + "loss": 0.9697, + "step": 15962 + }, + { + "epoch": 0.51, + "learning_rate": 1.0146043131173972e-05, + "loss": 0.9985, + "step": 15963 + }, + { + "epoch": 0.51, + "learning_rate": 1.0145007437149085e-05, + "loss": 1.1079, + "step": 15964 + }, + { + "epoch": 0.51, + "learning_rate": 1.014397174156843e-05, + "loss": 1.0234, + "step": 15965 + }, + { + "epoch": 0.51, + "learning_rate": 1.0142936044443118e-05, + "loss": 0.5085, + "step": 15966 + }, + { + "epoch": 0.51, + "learning_rate": 1.0141900345784264e-05, + "loss": 1.0493, + "step": 15967 + }, + { + "epoch": 0.51, + "learning_rate": 1.014086464560297e-05, + "loss": 0.9487, + "step": 15968 + }, + { + "epoch": 0.51, + "learning_rate": 1.0139828943910358e-05, + "loss": 0.8647, + "step": 15969 + }, + { + "epoch": 0.51, + "learning_rate": 1.0138793240717535e-05, + "loss": 0.9868, + "step": 15970 + }, + { + "epoch": 0.51, + "learning_rate": 1.0137757536035616e-05, + "loss": 1.0264, + "step": 15971 + }, + { + "epoch": 0.51, + "learning_rate": 1.0136721829875709e-05, + "loss": 1.0542, + "step": 15972 + }, + { + "epoch": 0.51, + "learning_rate": 1.0135686122248927e-05, + "loss": 0.9448, + "step": 15973 + }, + { + "epoch": 0.51, + "learning_rate": 1.0134650413166386e-05, + "loss": 0.9619, + "step": 15974 + }, + { + "epoch": 0.51, + "learning_rate": 1.0133614702639194e-05, + "loss": 1.1182, + "step": 15975 + }, + { + "epoch": 0.51, + "learning_rate": 1.0132578990678463e-05, + "loss": 1.208, + "step": 15976 + }, + { + "epoch": 0.51, + "learning_rate": 1.0131543277295306e-05, + "loss": 1.0742, + "step": 15977 + }, + { + "epoch": 0.51, + "learning_rate": 1.0130507562500838e-05, + "loss": 0.9155, + "step": 15978 + }, + { + "epoch": 0.51, + "learning_rate": 1.0129471846306165e-05, + "loss": 0.9131, + "step": 15979 + }, + { + "epoch": 0.51, + "learning_rate": 1.0128436128722402e-05, + "loss": 0.8774, + "step": 15980 + }, + { + "epoch": 0.51, + "learning_rate": 1.0127400409760664e-05, + "loss": 1.0356, + "step": 15981 + }, + { + "epoch": 0.51, + "learning_rate": 1.0126364689432062e-05, + "loss": 0.8921, + "step": 15982 + }, + { + "epoch": 0.51, + "learning_rate": 1.0125328967747705e-05, + "loss": 1.0381, + "step": 15983 + }, + { + "epoch": 0.51, + "learning_rate": 1.0124293244718706e-05, + "loss": 0.9834, + "step": 15984 + }, + { + "epoch": 0.51, + "learning_rate": 1.0123257520356178e-05, + "loss": 0.8682, + "step": 15985 + }, + { + "epoch": 0.51, + "learning_rate": 1.0122221794671234e-05, + "loss": 0.9478, + "step": 15986 + }, + { + "epoch": 0.51, + "learning_rate": 1.0121186067674985e-05, + "loss": 0.8892, + "step": 15987 + }, + { + "epoch": 0.51, + "learning_rate": 1.0120150339378543e-05, + "loss": 1.0327, + "step": 15988 + }, + { + "epoch": 0.51, + "learning_rate": 1.0119114609793023e-05, + "loss": 1.0122, + "step": 15989 + }, + { + "epoch": 0.51, + "learning_rate": 1.0118078878929534e-05, + "loss": 0.9683, + "step": 15990 + }, + { + "epoch": 0.51, + "learning_rate": 1.0117043146799188e-05, + "loss": 1.0459, + "step": 15991 + }, + { + "epoch": 0.51, + "learning_rate": 1.01160074134131e-05, + "loss": 1.0654, + "step": 15992 + }, + { + "epoch": 0.51, + "learning_rate": 1.0114971678782381e-05, + "loss": 1.0142, + "step": 15993 + }, + { + "epoch": 0.51, + "learning_rate": 1.0113935942918143e-05, + "loss": 1.0649, + "step": 15994 + }, + { + "epoch": 0.51, + "learning_rate": 1.0112900205831498e-05, + "loss": 1.0586, + "step": 15995 + }, + { + "epoch": 0.51, + "learning_rate": 1.011186446753356e-05, + "loss": 0.9028, + "step": 15996 + }, + { + "epoch": 0.51, + "learning_rate": 1.0110828728035438e-05, + "loss": 0.8604, + "step": 15997 + }, + { + "epoch": 0.51, + "learning_rate": 1.0109792987348245e-05, + "loss": 0.9976, + "step": 15998 + }, + { + "epoch": 0.51, + "learning_rate": 1.01087572454831e-05, + "loss": 0.8455, + "step": 15999 + }, + { + "epoch": 0.51, + "learning_rate": 1.0107721502451107e-05, + "loss": 0.9849, + "step": 16000 + }, + { + "epoch": 0.51, + "learning_rate": 1.0106685758263384e-05, + "loss": 0.958, + "step": 16001 + }, + { + "epoch": 0.51, + "learning_rate": 1.0105650012931036e-05, + "loss": 1.0967, + "step": 16002 + }, + { + "epoch": 0.51, + "learning_rate": 1.0104614266465183e-05, + "loss": 1.0044, + "step": 16003 + }, + { + "epoch": 0.51, + "learning_rate": 1.0103578518876936e-05, + "loss": 0.9033, + "step": 16004 + }, + { + "epoch": 0.51, + "learning_rate": 1.0102542770177404e-05, + "loss": 1.0488, + "step": 16005 + }, + { + "epoch": 0.51, + "learning_rate": 1.01015070203777e-05, + "loss": 0.5671, + "step": 16006 + }, + { + "epoch": 0.51, + "learning_rate": 1.0100471269488941e-05, + "loss": 1.0322, + "step": 16007 + }, + { + "epoch": 0.51, + "learning_rate": 1.0099435517522238e-05, + "loss": 1.0815, + "step": 16008 + }, + { + "epoch": 0.51, + "learning_rate": 1.00983997644887e-05, + "loss": 1.0044, + "step": 16009 + }, + { + "epoch": 0.51, + "learning_rate": 1.009736401039944e-05, + "loss": 0.9214, + "step": 16010 + }, + { + "epoch": 0.51, + "learning_rate": 1.0096328255265573e-05, + "loss": 1.0488, + "step": 16011 + }, + { + "epoch": 0.51, + "learning_rate": 1.009529249909821e-05, + "loss": 0.9961, + "step": 16012 + }, + { + "epoch": 0.51, + "learning_rate": 1.0094256741908466e-05, + "loss": 1.0469, + "step": 16013 + }, + { + "epoch": 0.51, + "learning_rate": 1.0093220983707448e-05, + "loss": 1.0254, + "step": 16014 + }, + { + "epoch": 0.51, + "learning_rate": 1.0092185224506274e-05, + "loss": 1.0249, + "step": 16015 + }, + { + "epoch": 0.51, + "learning_rate": 1.0091149464316055e-05, + "loss": 1.1021, + "step": 16016 + }, + { + "epoch": 0.51, + "learning_rate": 1.0090113703147903e-05, + "loss": 1.1216, + "step": 16017 + }, + { + "epoch": 0.51, + "learning_rate": 1.008907794101293e-05, + "loss": 0.9082, + "step": 16018 + }, + { + "epoch": 0.51, + "learning_rate": 1.0088042177922248e-05, + "loss": 0.9424, + "step": 16019 + }, + { + "epoch": 0.51, + "learning_rate": 1.0087006413886974e-05, + "loss": 1.0215, + "step": 16020 + }, + { + "epoch": 0.51, + "learning_rate": 1.0085970648918215e-05, + "loss": 1.0024, + "step": 16021 + }, + { + "epoch": 0.51, + "learning_rate": 1.0084934883027086e-05, + "loss": 0.8945, + "step": 16022 + }, + { + "epoch": 0.51, + "learning_rate": 1.0083899116224702e-05, + "loss": 1.0825, + "step": 16023 + }, + { + "epoch": 0.51, + "learning_rate": 1.0082863348522174e-05, + "loss": 1.0298, + "step": 16024 + }, + { + "epoch": 0.51, + "learning_rate": 1.008182757993061e-05, + "loss": 0.9292, + "step": 16025 + }, + { + "epoch": 0.51, + "learning_rate": 1.0080791810461128e-05, + "loss": 0.8979, + "step": 16026 + }, + { + "epoch": 0.51, + "learning_rate": 1.0079756040124841e-05, + "loss": 0.9619, + "step": 16027 + }, + { + "epoch": 0.51, + "learning_rate": 1.007872026893286e-05, + "loss": 0.9756, + "step": 16028 + }, + { + "epoch": 0.51, + "learning_rate": 1.0077684496896296e-05, + "loss": 1.0288, + "step": 16029 + }, + { + "epoch": 0.51, + "learning_rate": 1.0076648724026263e-05, + "loss": 0.9204, + "step": 16030 + }, + { + "epoch": 0.51, + "learning_rate": 1.0075612950333875e-05, + "loss": 0.9194, + "step": 16031 + }, + { + "epoch": 0.51, + "learning_rate": 1.0074577175830244e-05, + "loss": 0.9551, + "step": 16032 + }, + { + "epoch": 0.51, + "learning_rate": 1.0073541400526481e-05, + "loss": 0.9639, + "step": 16033 + }, + { + "epoch": 0.51, + "learning_rate": 1.0072505624433701e-05, + "loss": 1.0776, + "step": 16034 + }, + { + "epoch": 0.51, + "learning_rate": 1.0071469847563017e-05, + "loss": 1.0195, + "step": 16035 + }, + { + "epoch": 0.51, + "learning_rate": 1.007043406992554e-05, + "loss": 1.001, + "step": 16036 + }, + { + "epoch": 0.51, + "learning_rate": 1.0069398291532383e-05, + "loss": 0.98, + "step": 16037 + }, + { + "epoch": 0.51, + "learning_rate": 1.0068362512394661e-05, + "loss": 0.8394, + "step": 16038 + }, + { + "epoch": 0.51, + "learning_rate": 1.0067326732523482e-05, + "loss": 1.0044, + "step": 16039 + }, + { + "epoch": 0.51, + "learning_rate": 1.0066290951929965e-05, + "loss": 0.9287, + "step": 16040 + }, + { + "epoch": 0.51, + "learning_rate": 1.0065255170625216e-05, + "loss": 0.9619, + "step": 16041 + }, + { + "epoch": 0.51, + "learning_rate": 1.006421938862035e-05, + "loss": 0.9282, + "step": 16042 + }, + { + "epoch": 0.51, + "learning_rate": 1.0063183605926488e-05, + "loss": 0.9927, + "step": 16043 + }, + { + "epoch": 0.51, + "learning_rate": 1.006214782255473e-05, + "loss": 0.9565, + "step": 16044 + }, + { + "epoch": 0.51, + "learning_rate": 1.0061112038516196e-05, + "loss": 1.0156, + "step": 16045 + }, + { + "epoch": 0.51, + "learning_rate": 1.0060076253821998e-05, + "loss": 1.0015, + "step": 16046 + }, + { + "epoch": 0.51, + "learning_rate": 1.0059040468483248e-05, + "loss": 1.0156, + "step": 16047 + }, + { + "epoch": 0.51, + "learning_rate": 1.0058004682511058e-05, + "loss": 0.877, + "step": 16048 + }, + { + "epoch": 0.51, + "learning_rate": 1.0056968895916541e-05, + "loss": 0.8315, + "step": 16049 + }, + { + "epoch": 0.51, + "learning_rate": 1.0055933108710816e-05, + "loss": 0.9019, + "step": 16050 + }, + { + "epoch": 0.51, + "learning_rate": 1.0054897320904986e-05, + "loss": 0.9253, + "step": 16051 + }, + { + "epoch": 0.51, + "learning_rate": 1.005386153251017e-05, + "loss": 1.1079, + "step": 16052 + }, + { + "epoch": 0.51, + "learning_rate": 1.0052825743537478e-05, + "loss": 1.0039, + "step": 16053 + }, + { + "epoch": 0.51, + "learning_rate": 1.0051789953998029e-05, + "loss": 0.9102, + "step": 16054 + }, + { + "epoch": 0.51, + "learning_rate": 1.0050754163902925e-05, + "loss": 1.0132, + "step": 16055 + }, + { + "epoch": 0.51, + "learning_rate": 1.0049718373263287e-05, + "loss": 0.9858, + "step": 16056 + }, + { + "epoch": 0.51, + "learning_rate": 1.0048682582090228e-05, + "loss": 0.9604, + "step": 16057 + }, + { + "epoch": 0.51, + "learning_rate": 1.0047646790394855e-05, + "loss": 0.8984, + "step": 16058 + }, + { + "epoch": 0.51, + "learning_rate": 1.004661099818829e-05, + "loss": 1.0049, + "step": 16059 + }, + { + "epoch": 0.51, + "learning_rate": 1.0045575205481635e-05, + "loss": 1.0288, + "step": 16060 + }, + { + "epoch": 0.51, + "learning_rate": 1.0044539412286011e-05, + "loss": 1.0073, + "step": 16061 + }, + { + "epoch": 0.51, + "learning_rate": 1.0043503618612528e-05, + "loss": 0.9712, + "step": 16062 + }, + { + "epoch": 0.51, + "learning_rate": 1.00424678244723e-05, + "loss": 1.0176, + "step": 16063 + }, + { + "epoch": 0.51, + "learning_rate": 1.0041432029876439e-05, + "loss": 1.0332, + "step": 16064 + }, + { + "epoch": 0.51, + "learning_rate": 1.0040396234836058e-05, + "loss": 0.9453, + "step": 16065 + }, + { + "epoch": 0.51, + "learning_rate": 1.0039360439362271e-05, + "loss": 1.0093, + "step": 16066 + }, + { + "epoch": 0.51, + "learning_rate": 1.0038324643466189e-05, + "loss": 1.0098, + "step": 16067 + }, + { + "epoch": 0.51, + "learning_rate": 1.0037288847158927e-05, + "loss": 0.9551, + "step": 16068 + }, + { + "epoch": 0.51, + "learning_rate": 1.0036253050451597e-05, + "loss": 0.9351, + "step": 16069 + }, + { + "epoch": 0.51, + "learning_rate": 1.003521725335531e-05, + "loss": 1.0088, + "step": 16070 + }, + { + "epoch": 0.51, + "learning_rate": 1.003418145588118e-05, + "loss": 0.9482, + "step": 16071 + }, + { + "epoch": 0.51, + "learning_rate": 1.0033145658040324e-05, + "loss": 0.9438, + "step": 16072 + }, + { + "epoch": 0.51, + "learning_rate": 1.0032109859843851e-05, + "loss": 0.9771, + "step": 16073 + }, + { + "epoch": 0.51, + "learning_rate": 1.0031074061302873e-05, + "loss": 0.8867, + "step": 16074 + }, + { + "epoch": 0.51, + "learning_rate": 1.0030038262428506e-05, + "loss": 1.0654, + "step": 16075 + }, + { + "epoch": 0.51, + "learning_rate": 1.0029002463231864e-05, + "loss": 0.9746, + "step": 16076 + }, + { + "epoch": 0.51, + "learning_rate": 1.0027966663724053e-05, + "loss": 1.0776, + "step": 16077 + }, + { + "epoch": 0.51, + "learning_rate": 1.0026930863916194e-05, + "loss": 0.9893, + "step": 16078 + }, + { + "epoch": 0.51, + "learning_rate": 1.0025895063819394e-05, + "loss": 1.064, + "step": 16079 + }, + { + "epoch": 0.51, + "learning_rate": 1.002485926344477e-05, + "loss": 1.1045, + "step": 16080 + }, + { + "epoch": 0.51, + "learning_rate": 1.0023823462803436e-05, + "loss": 0.8501, + "step": 16081 + }, + { + "epoch": 0.51, + "learning_rate": 1.0022787661906501e-05, + "loss": 0.9385, + "step": 16082 + }, + { + "epoch": 0.51, + "learning_rate": 1.0021751860765078e-05, + "loss": 0.8696, + "step": 16083 + }, + { + "epoch": 0.51, + "learning_rate": 1.0020716059390284e-05, + "loss": 0.9614, + "step": 16084 + }, + { + "epoch": 0.51, + "learning_rate": 1.0019680257793228e-05, + "loss": 0.8765, + "step": 16085 + }, + { + "epoch": 0.51, + "learning_rate": 1.0018644455985025e-05, + "loss": 0.7529, + "step": 16086 + }, + { + "epoch": 0.51, + "learning_rate": 1.0017608653976787e-05, + "loss": 0.9727, + "step": 16087 + }, + { + "epoch": 0.51, + "learning_rate": 1.0016572851779628e-05, + "loss": 1.127, + "step": 16088 + }, + { + "epoch": 0.51, + "learning_rate": 1.001553704940466e-05, + "loss": 0.8647, + "step": 16089 + }, + { + "epoch": 0.51, + "learning_rate": 1.0014501246862997e-05, + "loss": 0.9722, + "step": 16090 + }, + { + "epoch": 0.51, + "learning_rate": 1.0013465444165755e-05, + "loss": 0.9761, + "step": 16091 + }, + { + "epoch": 0.51, + "learning_rate": 1.001242964132404e-05, + "loss": 1.0352, + "step": 16092 + }, + { + "epoch": 0.51, + "learning_rate": 1.0011393838348969e-05, + "loss": 1.04, + "step": 16093 + }, + { + "epoch": 0.51, + "learning_rate": 1.0010358035251655e-05, + "loss": 0.9189, + "step": 16094 + }, + { + "epoch": 0.51, + "learning_rate": 1.000932223204321e-05, + "loss": 0.8696, + "step": 16095 + }, + { + "epoch": 0.51, + "learning_rate": 1.0008286428734751e-05, + "loss": 1.0166, + "step": 16096 + }, + { + "epoch": 0.51, + "learning_rate": 1.0007250625337384e-05, + "loss": 0.8765, + "step": 16097 + }, + { + "epoch": 0.51, + "learning_rate": 1.0006214821862226e-05, + "loss": 0.9741, + "step": 16098 + }, + { + "epoch": 0.51, + "learning_rate": 1.0005179018320392e-05, + "loss": 0.8716, + "step": 16099 + }, + { + "epoch": 0.51, + "learning_rate": 1.0004143214722992e-05, + "loss": 1.0527, + "step": 16100 + }, + { + "epoch": 0.51, + "learning_rate": 1.0003107411081139e-05, + "loss": 1.0283, + "step": 16101 + }, + { + "epoch": 0.51, + "learning_rate": 1.0002071607405947e-05, + "loss": 0.8354, + "step": 16102 + }, + { + "epoch": 0.51, + "learning_rate": 1.0001035803708532e-05, + "loss": 0.9624, + "step": 16103 + }, + { + "epoch": 0.52, + "learning_rate": 1e-05, + "loss": 1.0, + "step": 16104 + }, + { + "epoch": 0.52, + "learning_rate": 9.998964196291471e-06, + "loss": 0.9199, + "step": 16105 + }, + { + "epoch": 0.52, + "learning_rate": 9.997928392594053e-06, + "loss": 0.8765, + "step": 16106 + }, + { + "epoch": 0.52, + "learning_rate": 9.996892588918861e-06, + "loss": 0.9673, + "step": 16107 + }, + { + "epoch": 0.52, + "learning_rate": 9.995856785277013e-06, + "loss": 0.9146, + "step": 16108 + }, + { + "epoch": 0.52, + "learning_rate": 9.994820981679611e-06, + "loss": 0.781, + "step": 16109 + }, + { + "epoch": 0.52, + "learning_rate": 9.993785178137778e-06, + "loss": 0.8989, + "step": 16110 + }, + { + "epoch": 0.52, + "learning_rate": 9.99274937466262e-06, + "loss": 1.0234, + "step": 16111 + }, + { + "epoch": 0.52, + "learning_rate": 9.991713571265254e-06, + "loss": 1.1055, + "step": 16112 + }, + { + "epoch": 0.52, + "learning_rate": 9.990677767956792e-06, + "loss": 1.0303, + "step": 16113 + }, + { + "epoch": 0.52, + "learning_rate": 9.989641964748347e-06, + "loss": 0.9258, + "step": 16114 + }, + { + "epoch": 0.52, + "learning_rate": 9.988606161651031e-06, + "loss": 1.0562, + "step": 16115 + }, + { + "epoch": 0.52, + "learning_rate": 9.987570358675965e-06, + "loss": 1.0532, + "step": 16116 + }, + { + "epoch": 0.52, + "learning_rate": 9.986534555834251e-06, + "loss": 0.9834, + "step": 16117 + }, + { + "epoch": 0.52, + "learning_rate": 9.985498753137005e-06, + "loss": 0.8848, + "step": 16118 + }, + { + "epoch": 0.52, + "learning_rate": 9.984462950595341e-06, + "loss": 1.0444, + "step": 16119 + }, + { + "epoch": 0.52, + "learning_rate": 9.983427148220373e-06, + "loss": 0.9849, + "step": 16120 + }, + { + "epoch": 0.52, + "learning_rate": 9.982391346023216e-06, + "loss": 0.9302, + "step": 16121 + }, + { + "epoch": 0.52, + "learning_rate": 9.981355544014978e-06, + "loss": 1.1021, + "step": 16122 + }, + { + "epoch": 0.52, + "learning_rate": 9.980319742206777e-06, + "loss": 0.8201, + "step": 16123 + }, + { + "epoch": 0.52, + "learning_rate": 9.979283940609722e-06, + "loss": 0.9238, + "step": 16124 + }, + { + "epoch": 0.52, + "learning_rate": 9.978248139234925e-06, + "loss": 1.0635, + "step": 16125 + }, + { + "epoch": 0.52, + "learning_rate": 9.977212338093502e-06, + "loss": 0.9561, + "step": 16126 + }, + { + "epoch": 0.52, + "learning_rate": 9.976176537196568e-06, + "loss": 1.0503, + "step": 16127 + }, + { + "epoch": 0.52, + "learning_rate": 9.975140736555231e-06, + "loss": 1.0308, + "step": 16128 + }, + { + "epoch": 0.52, + "learning_rate": 9.974104936180608e-06, + "loss": 0.9834, + "step": 16129 + }, + { + "epoch": 0.52, + "learning_rate": 9.973069136083808e-06, + "loss": 0.9258, + "step": 16130 + }, + { + "epoch": 0.52, + "learning_rate": 9.97203333627595e-06, + "loss": 1.0127, + "step": 16131 + }, + { + "epoch": 0.52, + "learning_rate": 9.970997536768141e-06, + "loss": 0.9521, + "step": 16132 + }, + { + "epoch": 0.52, + "learning_rate": 9.969961737571497e-06, + "loss": 1.0596, + "step": 16133 + }, + { + "epoch": 0.52, + "learning_rate": 9.968925938697129e-06, + "loss": 0.9941, + "step": 16134 + }, + { + "epoch": 0.52, + "learning_rate": 9.967890140156152e-06, + "loss": 0.7944, + "step": 16135 + }, + { + "epoch": 0.52, + "learning_rate": 9.966854341959678e-06, + "loss": 0.5308, + "step": 16136 + }, + { + "epoch": 0.52, + "learning_rate": 9.96581854411882e-06, + "loss": 0.9644, + "step": 16137 + }, + { + "epoch": 0.52, + "learning_rate": 9.964782746644695e-06, + "loss": 0.8462, + "step": 16138 + }, + { + "epoch": 0.52, + "learning_rate": 9.96374694954841e-06, + "loss": 1.0391, + "step": 16139 + }, + { + "epoch": 0.52, + "learning_rate": 9.962711152841078e-06, + "loss": 1.0913, + "step": 16140 + }, + { + "epoch": 0.52, + "learning_rate": 9.961675356533814e-06, + "loss": 1.1006, + "step": 16141 + }, + { + "epoch": 0.52, + "learning_rate": 9.960639560637732e-06, + "loss": 0.9941, + "step": 16142 + }, + { + "epoch": 0.52, + "learning_rate": 9.959603765163944e-06, + "loss": 0.9058, + "step": 16143 + }, + { + "epoch": 0.52, + "learning_rate": 9.958567970123563e-06, + "loss": 1.0625, + "step": 16144 + }, + { + "epoch": 0.52, + "learning_rate": 9.957532175527702e-06, + "loss": 1.0244, + "step": 16145 + }, + { + "epoch": 0.52, + "learning_rate": 9.956496381387477e-06, + "loss": 1.0015, + "step": 16146 + }, + { + "epoch": 0.52, + "learning_rate": 9.955460587713994e-06, + "loss": 1.063, + "step": 16147 + }, + { + "epoch": 0.52, + "learning_rate": 9.954424794518368e-06, + "loss": 1.0293, + "step": 16148 + }, + { + "epoch": 0.52, + "learning_rate": 9.953389001811716e-06, + "loss": 0.9111, + "step": 16149 + }, + { + "epoch": 0.52, + "learning_rate": 9.952353209605146e-06, + "loss": 0.9902, + "step": 16150 + }, + { + "epoch": 0.52, + "learning_rate": 9.951317417909775e-06, + "loss": 0.8291, + "step": 16151 + }, + { + "epoch": 0.52, + "learning_rate": 9.950281626736713e-06, + "loss": 1.0845, + "step": 16152 + }, + { + "epoch": 0.52, + "learning_rate": 9.949245836097075e-06, + "loss": 0.9409, + "step": 16153 + }, + { + "epoch": 0.52, + "learning_rate": 9.948210046001978e-06, + "loss": 1.0684, + "step": 16154 + }, + { + "epoch": 0.52, + "learning_rate": 9.947174256462526e-06, + "loss": 1.0454, + "step": 16155 + }, + { + "epoch": 0.52, + "learning_rate": 9.946138467489833e-06, + "loss": 0.9404, + "step": 16156 + }, + { + "epoch": 0.52, + "learning_rate": 9.945102679095016e-06, + "loss": 0.9656, + "step": 16157 + }, + { + "epoch": 0.52, + "learning_rate": 9.944066891289188e-06, + "loss": 0.9922, + "step": 16158 + }, + { + "epoch": 0.52, + "learning_rate": 9.943031104083459e-06, + "loss": 0.8894, + "step": 16159 + }, + { + "epoch": 0.52, + "learning_rate": 9.941995317488944e-06, + "loss": 0.9199, + "step": 16160 + }, + { + "epoch": 0.52, + "learning_rate": 9.940959531516757e-06, + "loss": 0.9673, + "step": 16161 + }, + { + "epoch": 0.52, + "learning_rate": 9.939923746178006e-06, + "loss": 0.9961, + "step": 16162 + }, + { + "epoch": 0.52, + "learning_rate": 9.938887961483807e-06, + "loss": 1.0195, + "step": 16163 + }, + { + "epoch": 0.52, + "learning_rate": 9.937852177445273e-06, + "loss": 0.8945, + "step": 16164 + }, + { + "epoch": 0.52, + "learning_rate": 9.936816394073515e-06, + "loss": 0.9849, + "step": 16165 + }, + { + "epoch": 0.52, + "learning_rate": 9.935780611379651e-06, + "loss": 0.9556, + "step": 16166 + }, + { + "epoch": 0.52, + "learning_rate": 9.934744829374787e-06, + "loss": 0.8911, + "step": 16167 + }, + { + "epoch": 0.52, + "learning_rate": 9.933709048070039e-06, + "loss": 1.1382, + "step": 16168 + }, + { + "epoch": 0.52, + "learning_rate": 9.932673267476522e-06, + "loss": 0.8818, + "step": 16169 + }, + { + "epoch": 0.52, + "learning_rate": 9.931637487605342e-06, + "loss": 0.9326, + "step": 16170 + }, + { + "epoch": 0.52, + "learning_rate": 9.93060170846762e-06, + "loss": 1.0654, + "step": 16171 + }, + { + "epoch": 0.52, + "learning_rate": 9.929565930074463e-06, + "loss": 0.9893, + "step": 16172 + }, + { + "epoch": 0.52, + "learning_rate": 9.928530152436984e-06, + "loss": 1.1074, + "step": 16173 + }, + { + "epoch": 0.52, + "learning_rate": 9.9274943755663e-06, + "loss": 0.8862, + "step": 16174 + }, + { + "epoch": 0.52, + "learning_rate": 9.92645859947352e-06, + "loss": 1.0078, + "step": 16175 + }, + { + "epoch": 0.52, + "learning_rate": 9.92542282416976e-06, + "loss": 0.938, + "step": 16176 + }, + { + "epoch": 0.52, + "learning_rate": 9.924387049666128e-06, + "loss": 0.9321, + "step": 16177 + }, + { + "epoch": 0.52, + "learning_rate": 9.92335127597374e-06, + "loss": 0.5244, + "step": 16178 + }, + { + "epoch": 0.52, + "learning_rate": 9.922315503103707e-06, + "loss": 0.9043, + "step": 16179 + }, + { + "epoch": 0.52, + "learning_rate": 9.921279731067144e-06, + "loss": 0.9258, + "step": 16180 + }, + { + "epoch": 0.52, + "learning_rate": 9.92024395987516e-06, + "loss": 0.9668, + "step": 16181 + }, + { + "epoch": 0.52, + "learning_rate": 9.919208189538872e-06, + "loss": 0.5505, + "step": 16182 + }, + { + "epoch": 0.52, + "learning_rate": 9.918172420069391e-06, + "loss": 0.5325, + "step": 16183 + }, + { + "epoch": 0.52, + "learning_rate": 9.917136651477833e-06, + "loss": 1.0186, + "step": 16184 + }, + { + "epoch": 0.52, + "learning_rate": 9.916100883775302e-06, + "loss": 1.0381, + "step": 16185 + }, + { + "epoch": 0.52, + "learning_rate": 9.915065116972916e-06, + "loss": 1.0845, + "step": 16186 + }, + { + "epoch": 0.52, + "learning_rate": 9.914029351081788e-06, + "loss": 1.0293, + "step": 16187 + }, + { + "epoch": 0.52, + "learning_rate": 9.912993586113029e-06, + "loss": 0.9453, + "step": 16188 + }, + { + "epoch": 0.52, + "learning_rate": 9.911957822077752e-06, + "loss": 0.9761, + "step": 16189 + }, + { + "epoch": 0.52, + "learning_rate": 9.910922058987072e-06, + "loss": 1.0063, + "step": 16190 + }, + { + "epoch": 0.52, + "learning_rate": 9.909886296852103e-06, + "loss": 0.8735, + "step": 16191 + }, + { + "epoch": 0.52, + "learning_rate": 9.908850535683949e-06, + "loss": 1.0244, + "step": 16192 + }, + { + "epoch": 0.52, + "learning_rate": 9.907814775493729e-06, + "loss": 0.9949, + "step": 16193 + }, + { + "epoch": 0.52, + "learning_rate": 9.906779016292554e-06, + "loss": 0.9658, + "step": 16194 + }, + { + "epoch": 0.52, + "learning_rate": 9.905743258091539e-06, + "loss": 1.0718, + "step": 16195 + }, + { + "epoch": 0.52, + "learning_rate": 9.904707500901791e-06, + "loss": 0.9341, + "step": 16196 + }, + { + "epoch": 0.52, + "learning_rate": 9.903671744734429e-06, + "loss": 1.1055, + "step": 16197 + }, + { + "epoch": 0.52, + "learning_rate": 9.90263598960056e-06, + "loss": 0.9375, + "step": 16198 + }, + { + "epoch": 0.52, + "learning_rate": 9.901600235511306e-06, + "loss": 1.061, + "step": 16199 + }, + { + "epoch": 0.52, + "learning_rate": 9.900564482477767e-06, + "loss": 1.0776, + "step": 16200 + }, + { + "epoch": 0.52, + "learning_rate": 9.89952873051106e-06, + "loss": 1.0107, + "step": 16201 + }, + { + "epoch": 0.52, + "learning_rate": 9.898492979622301e-06, + "loss": 0.9487, + "step": 16202 + }, + { + "epoch": 0.52, + "learning_rate": 9.897457229822597e-06, + "loss": 1.0669, + "step": 16203 + }, + { + "epoch": 0.52, + "learning_rate": 9.896421481123066e-06, + "loss": 1.0093, + "step": 16204 + }, + { + "epoch": 0.52, + "learning_rate": 9.895385733534819e-06, + "loss": 1.0425, + "step": 16205 + }, + { + "epoch": 0.52, + "learning_rate": 9.89434998706897e-06, + "loss": 1.1011, + "step": 16206 + }, + { + "epoch": 0.52, + "learning_rate": 9.893314241736622e-06, + "loss": 1.1289, + "step": 16207 + }, + { + "epoch": 0.52, + "learning_rate": 9.892278497548896e-06, + "loss": 1.0098, + "step": 16208 + }, + { + "epoch": 0.52, + "learning_rate": 9.891242754516903e-06, + "loss": 0.9502, + "step": 16209 + }, + { + "epoch": 0.52, + "learning_rate": 9.890207012651756e-06, + "loss": 1.0161, + "step": 16210 + }, + { + "epoch": 0.52, + "learning_rate": 9.889171271964565e-06, + "loss": 1.0063, + "step": 16211 + }, + { + "epoch": 0.52, + "learning_rate": 9.888135532466442e-06, + "loss": 0.5593, + "step": 16212 + }, + { + "epoch": 0.52, + "learning_rate": 9.887099794168502e-06, + "loss": 0.9658, + "step": 16213 + }, + { + "epoch": 0.52, + "learning_rate": 9.88606405708186e-06, + "loss": 0.853, + "step": 16214 + }, + { + "epoch": 0.52, + "learning_rate": 9.88502832121762e-06, + "loss": 0.9756, + "step": 16215 + }, + { + "epoch": 0.52, + "learning_rate": 9.883992586586903e-06, + "loss": 0.9805, + "step": 16216 + }, + { + "epoch": 0.52, + "learning_rate": 9.882956853200814e-06, + "loss": 1.0601, + "step": 16217 + }, + { + "epoch": 0.52, + "learning_rate": 9.88192112107047e-06, + "loss": 1.0273, + "step": 16218 + }, + { + "epoch": 0.52, + "learning_rate": 9.880885390206979e-06, + "loss": 0.9766, + "step": 16219 + }, + { + "epoch": 0.52, + "learning_rate": 9.879849660621457e-06, + "loss": 0.8423, + "step": 16220 + }, + { + "epoch": 0.52, + "learning_rate": 9.878813932325016e-06, + "loss": 0.9697, + "step": 16221 + }, + { + "epoch": 0.52, + "learning_rate": 9.877778205328771e-06, + "loss": 1.0469, + "step": 16222 + }, + { + "epoch": 0.52, + "learning_rate": 9.876742479643826e-06, + "loss": 0.9136, + "step": 16223 + }, + { + "epoch": 0.52, + "learning_rate": 9.875706755281297e-06, + "loss": 0.873, + "step": 16224 + }, + { + "epoch": 0.52, + "learning_rate": 9.874671032252299e-06, + "loss": 0.9932, + "step": 16225 + }, + { + "epoch": 0.52, + "learning_rate": 9.873635310567941e-06, + "loss": 0.9722, + "step": 16226 + }, + { + "epoch": 0.52, + "learning_rate": 9.872599590239337e-06, + "loss": 1.0537, + "step": 16227 + }, + { + "epoch": 0.52, + "learning_rate": 9.871563871277598e-06, + "loss": 1.0806, + "step": 16228 + }, + { + "epoch": 0.52, + "learning_rate": 9.87052815369384e-06, + "loss": 0.9956, + "step": 16229 + }, + { + "epoch": 0.52, + "learning_rate": 9.869492437499167e-06, + "loss": 1.0342, + "step": 16230 + }, + { + "epoch": 0.52, + "learning_rate": 9.868456722704697e-06, + "loss": 0.9922, + "step": 16231 + }, + { + "epoch": 0.52, + "learning_rate": 9.86742100932154e-06, + "loss": 1.0625, + "step": 16232 + }, + { + "epoch": 0.52, + "learning_rate": 9.86638529736081e-06, + "loss": 1.0449, + "step": 16233 + }, + { + "epoch": 0.52, + "learning_rate": 9.865349586833617e-06, + "loss": 0.9863, + "step": 16234 + }, + { + "epoch": 0.52, + "learning_rate": 9.864313877751073e-06, + "loss": 0.979, + "step": 16235 + }, + { + "epoch": 0.52, + "learning_rate": 9.863278170124293e-06, + "loss": 1.0103, + "step": 16236 + }, + { + "epoch": 0.52, + "learning_rate": 9.862242463964389e-06, + "loss": 1.0278, + "step": 16237 + }, + { + "epoch": 0.52, + "learning_rate": 9.86120675928247e-06, + "loss": 1.0376, + "step": 16238 + }, + { + "epoch": 0.52, + "learning_rate": 9.860171056089646e-06, + "loss": 1.0493, + "step": 16239 + }, + { + "epoch": 0.52, + "learning_rate": 9.859135354397031e-06, + "loss": 0.9927, + "step": 16240 + }, + { + "epoch": 0.52, + "learning_rate": 9.85809965421574e-06, + "loss": 0.9668, + "step": 16241 + }, + { + "epoch": 0.52, + "learning_rate": 9.857063955556882e-06, + "loss": 0.9775, + "step": 16242 + }, + { + "epoch": 0.52, + "learning_rate": 9.85602825843157e-06, + "loss": 0.999, + "step": 16243 + }, + { + "epoch": 0.52, + "learning_rate": 9.85499256285092e-06, + "loss": 0.9546, + "step": 16244 + }, + { + "epoch": 0.52, + "learning_rate": 9.853956868826033e-06, + "loss": 1.0498, + "step": 16245 + }, + { + "epoch": 0.52, + "learning_rate": 9.852921176368027e-06, + "loss": 0.9478, + "step": 16246 + }, + { + "epoch": 0.52, + "learning_rate": 9.851885485488017e-06, + "loss": 0.875, + "step": 16247 + }, + { + "epoch": 0.52, + "learning_rate": 9.85084979619711e-06, + "loss": 0.9668, + "step": 16248 + }, + { + "epoch": 0.52, + "learning_rate": 9.849814108506422e-06, + "loss": 1.0381, + "step": 16249 + }, + { + "epoch": 0.52, + "learning_rate": 9.84877842242706e-06, + "loss": 1.0542, + "step": 16250 + }, + { + "epoch": 0.52, + "learning_rate": 9.847742737970139e-06, + "loss": 0.9438, + "step": 16251 + }, + { + "epoch": 0.52, + "learning_rate": 9.84670705514677e-06, + "loss": 1.1147, + "step": 16252 + }, + { + "epoch": 0.52, + "learning_rate": 9.845671373968064e-06, + "loss": 0.8462, + "step": 16253 + }, + { + "epoch": 0.52, + "learning_rate": 9.844635694445133e-06, + "loss": 0.8701, + "step": 16254 + }, + { + "epoch": 0.52, + "learning_rate": 9.843600016589093e-06, + "loss": 0.8457, + "step": 16255 + }, + { + "epoch": 0.52, + "learning_rate": 9.842564340411047e-06, + "loss": 0.9819, + "step": 16256 + }, + { + "epoch": 0.52, + "learning_rate": 9.841528665922113e-06, + "loss": 1.0283, + "step": 16257 + }, + { + "epoch": 0.52, + "learning_rate": 9.8404929931334e-06, + "loss": 1.1226, + "step": 16258 + }, + { + "epoch": 0.52, + "learning_rate": 9.839457322056023e-06, + "loss": 1.0835, + "step": 16259 + }, + { + "epoch": 0.52, + "learning_rate": 9.838421652701092e-06, + "loss": 0.9722, + "step": 16260 + }, + { + "epoch": 0.52, + "learning_rate": 9.837385985079715e-06, + "loss": 0.9683, + "step": 16261 + }, + { + "epoch": 0.52, + "learning_rate": 9.836350319203007e-06, + "loss": 0.8955, + "step": 16262 + }, + { + "epoch": 0.52, + "learning_rate": 9.835314655082078e-06, + "loss": 0.9985, + "step": 16263 + }, + { + "epoch": 0.52, + "learning_rate": 9.834278992728042e-06, + "loss": 1.0088, + "step": 16264 + }, + { + "epoch": 0.52, + "learning_rate": 9.833243332152007e-06, + "loss": 0.9268, + "step": 16265 + }, + { + "epoch": 0.52, + "learning_rate": 9.832207673365087e-06, + "loss": 1.0625, + "step": 16266 + }, + { + "epoch": 0.52, + "learning_rate": 9.831172016378399e-06, + "loss": 0.9966, + "step": 16267 + }, + { + "epoch": 0.52, + "learning_rate": 9.830136361203042e-06, + "loss": 0.9512, + "step": 16268 + }, + { + "epoch": 0.52, + "learning_rate": 9.829100707850134e-06, + "loss": 0.9819, + "step": 16269 + }, + { + "epoch": 0.52, + "learning_rate": 9.828065056330787e-06, + "loss": 1.0396, + "step": 16270 + }, + { + "epoch": 0.52, + "learning_rate": 9.827029406656112e-06, + "loss": 0.9854, + "step": 16271 + }, + { + "epoch": 0.52, + "learning_rate": 9.82599375883722e-06, + "loss": 0.9668, + "step": 16272 + }, + { + "epoch": 0.52, + "learning_rate": 9.824958112885223e-06, + "loss": 0.8672, + "step": 16273 + }, + { + "epoch": 0.52, + "learning_rate": 9.823922468811235e-06, + "loss": 1.0435, + "step": 16274 + }, + { + "epoch": 0.52, + "learning_rate": 9.822886826626359e-06, + "loss": 1.0049, + "step": 16275 + }, + { + "epoch": 0.52, + "learning_rate": 9.821851186341711e-06, + "loss": 0.5164, + "step": 16276 + }, + { + "epoch": 0.52, + "learning_rate": 9.820815547968406e-06, + "loss": 1.0933, + "step": 16277 + }, + { + "epoch": 0.52, + "learning_rate": 9.819779911517548e-06, + "loss": 1.002, + "step": 16278 + }, + { + "epoch": 0.52, + "learning_rate": 9.818744277000255e-06, + "loss": 1.0312, + "step": 16279 + }, + { + "epoch": 0.52, + "learning_rate": 9.817708644427634e-06, + "loss": 0.8916, + "step": 16280 + }, + { + "epoch": 0.52, + "learning_rate": 9.816673013810798e-06, + "loss": 0.981, + "step": 16281 + }, + { + "epoch": 0.52, + "learning_rate": 9.815637385160861e-06, + "loss": 1.0483, + "step": 16282 + }, + { + "epoch": 0.52, + "learning_rate": 9.814601758488926e-06, + "loss": 1.0459, + "step": 16283 + }, + { + "epoch": 0.52, + "learning_rate": 9.81356613380611e-06, + "loss": 1.104, + "step": 16284 + }, + { + "epoch": 0.52, + "learning_rate": 9.812530511123524e-06, + "loss": 0.9736, + "step": 16285 + }, + { + "epoch": 0.52, + "learning_rate": 9.811494890452277e-06, + "loss": 0.7852, + "step": 16286 + }, + { + "epoch": 0.52, + "learning_rate": 9.81045927180348e-06, + "loss": 1.0405, + "step": 16287 + }, + { + "epoch": 0.52, + "learning_rate": 9.809423655188248e-06, + "loss": 0.9883, + "step": 16288 + }, + { + "epoch": 0.52, + "learning_rate": 9.808388040617691e-06, + "loss": 0.959, + "step": 16289 + }, + { + "epoch": 0.52, + "learning_rate": 9.807352428102915e-06, + "loss": 0.9048, + "step": 16290 + }, + { + "epoch": 0.52, + "learning_rate": 9.806316817655034e-06, + "loss": 0.9844, + "step": 16291 + }, + { + "epoch": 0.52, + "learning_rate": 9.80528120928516e-06, + "loss": 1.0278, + "step": 16292 + }, + { + "epoch": 0.52, + "learning_rate": 9.804245603004403e-06, + "loss": 0.9717, + "step": 16293 + }, + { + "epoch": 0.52, + "learning_rate": 9.803209998823876e-06, + "loss": 0.9272, + "step": 16294 + }, + { + "epoch": 0.52, + "learning_rate": 9.802174396754686e-06, + "loss": 0.8789, + "step": 16295 + }, + { + "epoch": 0.52, + "learning_rate": 9.801138796807945e-06, + "loss": 0.9863, + "step": 16296 + }, + { + "epoch": 0.52, + "learning_rate": 9.800103198994768e-06, + "loss": 1.0571, + "step": 16297 + }, + { + "epoch": 0.52, + "learning_rate": 9.79906760332626e-06, + "loss": 1.0024, + "step": 16298 + }, + { + "epoch": 0.52, + "learning_rate": 9.798032009813533e-06, + "loss": 1.0142, + "step": 16299 + }, + { + "epoch": 0.52, + "learning_rate": 9.796996418467702e-06, + "loss": 0.8472, + "step": 16300 + }, + { + "epoch": 0.52, + "learning_rate": 9.795960829299872e-06, + "loss": 0.8896, + "step": 16301 + }, + { + "epoch": 0.52, + "learning_rate": 9.794925242321156e-06, + "loss": 0.9058, + "step": 16302 + }, + { + "epoch": 0.52, + "learning_rate": 9.793889657542667e-06, + "loss": 0.9067, + "step": 16303 + }, + { + "epoch": 0.52, + "learning_rate": 9.792854074975513e-06, + "loss": 0.9087, + "step": 16304 + }, + { + "epoch": 0.52, + "learning_rate": 9.79181849463081e-06, + "loss": 0.9648, + "step": 16305 + }, + { + "epoch": 0.52, + "learning_rate": 9.79078291651966e-06, + "loss": 0.9771, + "step": 16306 + }, + { + "epoch": 0.52, + "learning_rate": 9.789747340653176e-06, + "loss": 0.9458, + "step": 16307 + }, + { + "epoch": 0.52, + "learning_rate": 9.788711767042471e-06, + "loss": 0.9854, + "step": 16308 + }, + { + "epoch": 0.52, + "learning_rate": 9.787676195698657e-06, + "loss": 0.9531, + "step": 16309 + }, + { + "epoch": 0.52, + "learning_rate": 9.786640626632841e-06, + "loss": 0.9897, + "step": 16310 + }, + { + "epoch": 0.52, + "learning_rate": 9.785605059856134e-06, + "loss": 1.1226, + "step": 16311 + }, + { + "epoch": 0.52, + "learning_rate": 9.784569495379653e-06, + "loss": 1.0693, + "step": 16312 + }, + { + "epoch": 0.52, + "learning_rate": 9.783533933214498e-06, + "loss": 0.9775, + "step": 16313 + }, + { + "epoch": 0.52, + "learning_rate": 9.782498373371785e-06, + "loss": 0.9771, + "step": 16314 + }, + { + "epoch": 0.52, + "learning_rate": 9.781462815862623e-06, + "loss": 0.9595, + "step": 16315 + }, + { + "epoch": 0.52, + "learning_rate": 9.780427260698124e-06, + "loss": 0.9224, + "step": 16316 + }, + { + "epoch": 0.52, + "learning_rate": 9.779391707889397e-06, + "loss": 0.9653, + "step": 16317 + }, + { + "epoch": 0.52, + "learning_rate": 9.778356157447552e-06, + "loss": 0.9702, + "step": 16318 + }, + { + "epoch": 0.52, + "learning_rate": 9.777320609383702e-06, + "loss": 1.0151, + "step": 16319 + }, + { + "epoch": 0.52, + "learning_rate": 9.776285063708959e-06, + "loss": 0.9727, + "step": 16320 + }, + { + "epoch": 0.52, + "learning_rate": 9.775249520434424e-06, + "loss": 0.9829, + "step": 16321 + }, + { + "epoch": 0.52, + "learning_rate": 9.774213979571214e-06, + "loss": 1.0605, + "step": 16322 + }, + { + "epoch": 0.52, + "learning_rate": 9.773178441130437e-06, + "loss": 1.0571, + "step": 16323 + }, + { + "epoch": 0.52, + "learning_rate": 9.772142905123206e-06, + "loss": 0.9697, + "step": 16324 + }, + { + "epoch": 0.52, + "learning_rate": 9.77110737156063e-06, + "loss": 1.1221, + "step": 16325 + }, + { + "epoch": 0.52, + "learning_rate": 9.770071840453816e-06, + "loss": 1.062, + "step": 16326 + }, + { + "epoch": 0.52, + "learning_rate": 9.769036311813883e-06, + "loss": 1.0815, + "step": 16327 + }, + { + "epoch": 0.52, + "learning_rate": 9.76800078565193e-06, + "loss": 1.0562, + "step": 16328 + }, + { + "epoch": 0.52, + "learning_rate": 9.766965261979072e-06, + "loss": 1.0288, + "step": 16329 + }, + { + "epoch": 0.52, + "learning_rate": 9.765929740806417e-06, + "loss": 0.9756, + "step": 16330 + }, + { + "epoch": 0.52, + "learning_rate": 9.764894222145079e-06, + "loss": 0.8901, + "step": 16331 + }, + { + "epoch": 0.52, + "learning_rate": 9.763858706006164e-06, + "loss": 0.9512, + "step": 16332 + }, + { + "epoch": 0.52, + "learning_rate": 9.762823192400788e-06, + "loss": 0.9746, + "step": 16333 + }, + { + "epoch": 0.52, + "learning_rate": 9.761787681340053e-06, + "loss": 0.8877, + "step": 16334 + }, + { + "epoch": 0.52, + "learning_rate": 9.760752172835076e-06, + "loss": 0.9268, + "step": 16335 + }, + { + "epoch": 0.52, + "learning_rate": 9.759716666896959e-06, + "loss": 0.9678, + "step": 16336 + }, + { + "epoch": 0.52, + "learning_rate": 9.758681163536819e-06, + "loss": 0.8154, + "step": 16337 + }, + { + "epoch": 0.52, + "learning_rate": 9.757645662765762e-06, + "loss": 0.8569, + "step": 16338 + }, + { + "epoch": 0.52, + "learning_rate": 9.7566101645949e-06, + "loss": 0.9556, + "step": 16339 + }, + { + "epoch": 0.52, + "learning_rate": 9.755574669035339e-06, + "loss": 1.0347, + "step": 16340 + }, + { + "epoch": 0.52, + "learning_rate": 9.754539176098192e-06, + "loss": 0.8726, + "step": 16341 + }, + { + "epoch": 0.52, + "learning_rate": 9.753503685794572e-06, + "loss": 0.9951, + "step": 16342 + }, + { + "epoch": 0.52, + "learning_rate": 9.75246819813558e-06, + "loss": 1.0972, + "step": 16343 + }, + { + "epoch": 0.52, + "learning_rate": 9.751432713132333e-06, + "loss": 0.9829, + "step": 16344 + }, + { + "epoch": 0.52, + "learning_rate": 9.750397230795937e-06, + "loss": 0.9795, + "step": 16345 + }, + { + "epoch": 0.52, + "learning_rate": 9.7493617511375e-06, + "loss": 1.0205, + "step": 16346 + }, + { + "epoch": 0.52, + "learning_rate": 9.748326274168137e-06, + "loss": 0.9771, + "step": 16347 + }, + { + "epoch": 0.52, + "learning_rate": 9.747290799898952e-06, + "loss": 0.5032, + "step": 16348 + }, + { + "epoch": 0.52, + "learning_rate": 9.746255328341057e-06, + "loss": 0.9561, + "step": 16349 + }, + { + "epoch": 0.52, + "learning_rate": 9.745219859505568e-06, + "loss": 0.9644, + "step": 16350 + }, + { + "epoch": 0.52, + "learning_rate": 9.744184393403584e-06, + "loss": 0.9692, + "step": 16351 + }, + { + "epoch": 0.52, + "learning_rate": 9.743148930046216e-06, + "loss": 0.8813, + "step": 16352 + }, + { + "epoch": 0.52, + "learning_rate": 9.742113469444577e-06, + "loss": 0.9688, + "step": 16353 + }, + { + "epoch": 0.52, + "learning_rate": 9.741078011609773e-06, + "loss": 0.9619, + "step": 16354 + }, + { + "epoch": 0.52, + "learning_rate": 9.740042556552918e-06, + "loss": 0.9878, + "step": 16355 + }, + { + "epoch": 0.52, + "learning_rate": 9.739007104285118e-06, + "loss": 0.9702, + "step": 16356 + }, + { + "epoch": 0.52, + "learning_rate": 9.737971654817486e-06, + "loss": 1.032, + "step": 16357 + }, + { + "epoch": 0.52, + "learning_rate": 9.736936208161125e-06, + "loss": 1.1045, + "step": 16358 + }, + { + "epoch": 0.52, + "learning_rate": 9.735900764327147e-06, + "loss": 0.8354, + "step": 16359 + }, + { + "epoch": 0.52, + "learning_rate": 9.734865323326662e-06, + "loss": 1.0674, + "step": 16360 + }, + { + "epoch": 0.52, + "learning_rate": 9.733829885170778e-06, + "loss": 0.9946, + "step": 16361 + }, + { + "epoch": 0.52, + "learning_rate": 9.732794449870606e-06, + "loss": 0.957, + "step": 16362 + }, + { + "epoch": 0.52, + "learning_rate": 9.731759017437252e-06, + "loss": 0.9736, + "step": 16363 + }, + { + "epoch": 0.52, + "learning_rate": 9.730723587881827e-06, + "loss": 1.0396, + "step": 16364 + }, + { + "epoch": 0.52, + "learning_rate": 9.729688161215445e-06, + "loss": 0.9326, + "step": 16365 + }, + { + "epoch": 0.52, + "learning_rate": 9.728652737449206e-06, + "loss": 0.9985, + "step": 16366 + }, + { + "epoch": 0.52, + "learning_rate": 9.727617316594222e-06, + "loss": 0.8999, + "step": 16367 + }, + { + "epoch": 0.52, + "learning_rate": 9.726581898661603e-06, + "loss": 0.9333, + "step": 16368 + }, + { + "epoch": 0.52, + "learning_rate": 9.725546483662458e-06, + "loss": 0.999, + "step": 16369 + }, + { + "epoch": 0.52, + "learning_rate": 9.724511071607894e-06, + "loss": 1.0723, + "step": 16370 + }, + { + "epoch": 0.52, + "learning_rate": 9.723475662509023e-06, + "loss": 0.9727, + "step": 16371 + }, + { + "epoch": 0.52, + "learning_rate": 9.722440256376956e-06, + "loss": 1.0366, + "step": 16372 + }, + { + "epoch": 0.52, + "learning_rate": 9.721404853222794e-06, + "loss": 0.9341, + "step": 16373 + }, + { + "epoch": 0.52, + "learning_rate": 9.720369453057648e-06, + "loss": 0.5752, + "step": 16374 + }, + { + "epoch": 0.52, + "learning_rate": 9.71933405589263e-06, + "loss": 0.9458, + "step": 16375 + }, + { + "epoch": 0.52, + "learning_rate": 9.718298661738845e-06, + "loss": 0.5505, + "step": 16376 + }, + { + "epoch": 0.52, + "learning_rate": 9.717263270607406e-06, + "loss": 1.0337, + "step": 16377 + }, + { + "epoch": 0.52, + "learning_rate": 9.716227882509419e-06, + "loss": 0.8721, + "step": 16378 + }, + { + "epoch": 0.52, + "learning_rate": 9.715192497455992e-06, + "loss": 0.8613, + "step": 16379 + }, + { + "epoch": 0.52, + "learning_rate": 9.714157115458236e-06, + "loss": 0.9023, + "step": 16380 + }, + { + "epoch": 0.52, + "learning_rate": 9.713121736527256e-06, + "loss": 1.0312, + "step": 16381 + }, + { + "epoch": 0.52, + "learning_rate": 9.712086360674162e-06, + "loss": 0.9824, + "step": 16382 + }, + { + "epoch": 0.52, + "learning_rate": 9.711050987910065e-06, + "loss": 1.0767, + "step": 16383 + }, + { + "epoch": 0.52, + "learning_rate": 9.71001561824607e-06, + "loss": 0.9209, + "step": 16384 + }, + { + "epoch": 0.52, + "learning_rate": 9.708980251693284e-06, + "loss": 1.0649, + "step": 16385 + }, + { + "epoch": 0.52, + "learning_rate": 9.707944888262819e-06, + "loss": 1.0405, + "step": 16386 + }, + { + "epoch": 0.52, + "learning_rate": 9.706909527965783e-06, + "loss": 0.9805, + "step": 16387 + }, + { + "epoch": 0.52, + "learning_rate": 9.705874170813286e-06, + "loss": 0.5452, + "step": 16388 + }, + { + "epoch": 0.52, + "learning_rate": 9.70483881681643e-06, + "loss": 1.0278, + "step": 16389 + }, + { + "epoch": 0.52, + "learning_rate": 9.703803465986328e-06, + "loss": 0.9253, + "step": 16390 + }, + { + "epoch": 0.52, + "learning_rate": 9.702768118334086e-06, + "loss": 0.9927, + "step": 16391 + }, + { + "epoch": 0.52, + "learning_rate": 9.701732773870813e-06, + "loss": 1.0176, + "step": 16392 + }, + { + "epoch": 0.52, + "learning_rate": 9.700697432607618e-06, + "loss": 0.9165, + "step": 16393 + }, + { + "epoch": 0.52, + "learning_rate": 9.699662094555608e-06, + "loss": 1.1968, + "step": 16394 + }, + { + "epoch": 0.52, + "learning_rate": 9.698626759725895e-06, + "loss": 1.0317, + "step": 16395 + }, + { + "epoch": 0.52, + "learning_rate": 9.69759142812958e-06, + "loss": 0.9236, + "step": 16396 + }, + { + "epoch": 0.52, + "learning_rate": 9.696556099777773e-06, + "loss": 1.0029, + "step": 16397 + }, + { + "epoch": 0.52, + "learning_rate": 9.695520774681584e-06, + "loss": 0.9414, + "step": 16398 + }, + { + "epoch": 0.52, + "learning_rate": 9.69448545285212e-06, + "loss": 0.9917, + "step": 16399 + }, + { + "epoch": 0.52, + "learning_rate": 9.69345013430049e-06, + "loss": 0.9214, + "step": 16400 + }, + { + "epoch": 0.52, + "learning_rate": 9.692414819037799e-06, + "loss": 0.8901, + "step": 16401 + }, + { + "epoch": 0.52, + "learning_rate": 9.691379507075157e-06, + "loss": 0.9966, + "step": 16402 + }, + { + "epoch": 0.52, + "learning_rate": 9.690344198423676e-06, + "loss": 0.8936, + "step": 16403 + }, + { + "epoch": 0.52, + "learning_rate": 9.689308893094456e-06, + "loss": 0.9531, + "step": 16404 + }, + { + "epoch": 0.52, + "learning_rate": 9.688273591098606e-06, + "loss": 1.0645, + "step": 16405 + }, + { + "epoch": 0.52, + "learning_rate": 9.687238292447236e-06, + "loss": 0.9663, + "step": 16406 + }, + { + "epoch": 0.52, + "learning_rate": 9.686202997151452e-06, + "loss": 0.8882, + "step": 16407 + }, + { + "epoch": 0.52, + "learning_rate": 9.685167705222364e-06, + "loss": 0.959, + "step": 16408 + }, + { + "epoch": 0.52, + "learning_rate": 9.684132416671077e-06, + "loss": 1.0137, + "step": 16409 + }, + { + "epoch": 0.52, + "learning_rate": 9.683097131508705e-06, + "loss": 0.9814, + "step": 16410 + }, + { + "epoch": 0.52, + "learning_rate": 9.682061849746344e-06, + "loss": 0.8862, + "step": 16411 + }, + { + "epoch": 0.52, + "learning_rate": 9.68102657139511e-06, + "loss": 0.936, + "step": 16412 + }, + { + "epoch": 0.52, + "learning_rate": 9.679991296466106e-06, + "loss": 0.9116, + "step": 16413 + }, + { + "epoch": 0.52, + "learning_rate": 9.678956024970441e-06, + "loss": 1.0151, + "step": 16414 + }, + { + "epoch": 0.52, + "learning_rate": 9.677920756919223e-06, + "loss": 0.9688, + "step": 16415 + }, + { + "epoch": 0.52, + "learning_rate": 9.676885492323558e-06, + "loss": 0.9482, + "step": 16416 + }, + { + "epoch": 0.53, + "learning_rate": 9.675850231194558e-06, + "loss": 0.9199, + "step": 16417 + }, + { + "epoch": 0.53, + "learning_rate": 9.674814973543326e-06, + "loss": 1.0186, + "step": 16418 + }, + { + "epoch": 0.53, + "learning_rate": 9.673779719380967e-06, + "loss": 0.7939, + "step": 16419 + }, + { + "epoch": 0.53, + "learning_rate": 9.67274446871859e-06, + "loss": 1.061, + "step": 16420 + }, + { + "epoch": 0.53, + "learning_rate": 9.671709221567302e-06, + "loss": 1.0171, + "step": 16421 + }, + { + "epoch": 0.53, + "learning_rate": 9.670673977938215e-06, + "loss": 1.002, + "step": 16422 + }, + { + "epoch": 0.53, + "learning_rate": 9.669638737842428e-06, + "loss": 0.8989, + "step": 16423 + }, + { + "epoch": 0.53, + "learning_rate": 9.668603501291051e-06, + "loss": 0.8535, + "step": 16424 + }, + { + "epoch": 0.53, + "learning_rate": 9.667568268295197e-06, + "loss": 0.9331, + "step": 16425 + }, + { + "epoch": 0.53, + "learning_rate": 9.666533038865963e-06, + "loss": 0.8901, + "step": 16426 + }, + { + "epoch": 0.53, + "learning_rate": 9.665497813014464e-06, + "loss": 0.9717, + "step": 16427 + }, + { + "epoch": 0.53, + "learning_rate": 9.664462590751799e-06, + "loss": 1.0391, + "step": 16428 + }, + { + "epoch": 0.53, + "learning_rate": 9.663427372089079e-06, + "loss": 0.8848, + "step": 16429 + }, + { + "epoch": 0.53, + "learning_rate": 9.662392157037412e-06, + "loss": 1.1094, + "step": 16430 + }, + { + "epoch": 0.53, + "learning_rate": 9.661356945607904e-06, + "loss": 0.9863, + "step": 16431 + }, + { + "epoch": 0.53, + "learning_rate": 9.660321737811661e-06, + "loss": 0.9663, + "step": 16432 + }, + { + "epoch": 0.53, + "learning_rate": 9.659286533659792e-06, + "loss": 0.9668, + "step": 16433 + }, + { + "epoch": 0.53, + "learning_rate": 9.6582513331634e-06, + "loss": 0.9067, + "step": 16434 + }, + { + "epoch": 0.53, + "learning_rate": 9.657216136333592e-06, + "loss": 1.0, + "step": 16435 + }, + { + "epoch": 0.53, + "learning_rate": 9.656180943181476e-06, + "loss": 1.0488, + "step": 16436 + }, + { + "epoch": 0.53, + "learning_rate": 9.655145753718156e-06, + "loss": 0.979, + "step": 16437 + }, + { + "epoch": 0.53, + "learning_rate": 9.654110567954741e-06, + "loss": 1.0039, + "step": 16438 + }, + { + "epoch": 0.53, + "learning_rate": 9.653075385902338e-06, + "loss": 1.0972, + "step": 16439 + }, + { + "epoch": 0.53, + "learning_rate": 9.652040207572057e-06, + "loss": 1.1143, + "step": 16440 + }, + { + "epoch": 0.53, + "learning_rate": 9.651005032974994e-06, + "loss": 0.9492, + "step": 16441 + }, + { + "epoch": 0.53, + "learning_rate": 9.64996986212226e-06, + "loss": 0.9375, + "step": 16442 + }, + { + "epoch": 0.53, + "learning_rate": 9.648934695024963e-06, + "loss": 1.0322, + "step": 16443 + }, + { + "epoch": 0.53, + "learning_rate": 9.647899531694208e-06, + "loss": 0.9712, + "step": 16444 + }, + { + "epoch": 0.53, + "learning_rate": 9.646864372141102e-06, + "loss": 1.0898, + "step": 16445 + }, + { + "epoch": 0.53, + "learning_rate": 9.64582921637675e-06, + "loss": 0.9624, + "step": 16446 + }, + { + "epoch": 0.53, + "learning_rate": 9.644794064412258e-06, + "loss": 1.0005, + "step": 16447 + }, + { + "epoch": 0.53, + "learning_rate": 9.643758916258737e-06, + "loss": 0.8813, + "step": 16448 + }, + { + "epoch": 0.53, + "learning_rate": 9.642723771927284e-06, + "loss": 0.9722, + "step": 16449 + }, + { + "epoch": 0.53, + "learning_rate": 9.64168863142901e-06, + "loss": 0.9902, + "step": 16450 + }, + { + "epoch": 0.53, + "learning_rate": 9.64065349477502e-06, + "loss": 0.9424, + "step": 16451 + }, + { + "epoch": 0.53, + "learning_rate": 9.63961836197642e-06, + "loss": 0.9536, + "step": 16452 + }, + { + "epoch": 0.53, + "learning_rate": 9.638583233044317e-06, + "loss": 0.8877, + "step": 16453 + }, + { + "epoch": 0.53, + "learning_rate": 9.637548107989816e-06, + "loss": 0.5083, + "step": 16454 + }, + { + "epoch": 0.53, + "learning_rate": 9.636512986824022e-06, + "loss": 0.9409, + "step": 16455 + }, + { + "epoch": 0.53, + "learning_rate": 9.635477869558047e-06, + "loss": 1.0234, + "step": 16456 + }, + { + "epoch": 0.53, + "learning_rate": 9.634442756202986e-06, + "loss": 0.9766, + "step": 16457 + }, + { + "epoch": 0.53, + "learning_rate": 9.63340764676995e-06, + "loss": 0.9766, + "step": 16458 + }, + { + "epoch": 0.53, + "learning_rate": 9.632372541270041e-06, + "loss": 0.8892, + "step": 16459 + }, + { + "epoch": 0.53, + "learning_rate": 9.631337439714371e-06, + "loss": 0.9258, + "step": 16460 + }, + { + "epoch": 0.53, + "learning_rate": 9.630302342114045e-06, + "loss": 0.9155, + "step": 16461 + }, + { + "epoch": 0.53, + "learning_rate": 9.629267248480162e-06, + "loss": 0.9766, + "step": 16462 + }, + { + "epoch": 0.53, + "learning_rate": 9.628232158823834e-06, + "loss": 0.9736, + "step": 16463 + }, + { + "epoch": 0.53, + "learning_rate": 9.62719707315616e-06, + "loss": 1.1118, + "step": 16464 + }, + { + "epoch": 0.53, + "learning_rate": 9.62616199148825e-06, + "loss": 0.9814, + "step": 16465 + }, + { + "epoch": 0.53, + "learning_rate": 9.62512691383121e-06, + "loss": 1.0493, + "step": 16466 + }, + { + "epoch": 0.53, + "learning_rate": 9.624091840196141e-06, + "loss": 0.9204, + "step": 16467 + }, + { + "epoch": 0.53, + "learning_rate": 9.623056770594152e-06, + "loss": 0.8496, + "step": 16468 + }, + { + "epoch": 0.53, + "learning_rate": 9.622021705036343e-06, + "loss": 0.9478, + "step": 16469 + }, + { + "epoch": 0.53, + "learning_rate": 9.620986643533825e-06, + "loss": 0.9712, + "step": 16470 + }, + { + "epoch": 0.53, + "learning_rate": 9.619951586097704e-06, + "loss": 1.041, + "step": 16471 + }, + { + "epoch": 0.53, + "learning_rate": 9.618916532739081e-06, + "loss": 0.9365, + "step": 16472 + }, + { + "epoch": 0.53, + "learning_rate": 9.61788148346906e-06, + "loss": 0.9263, + "step": 16473 + }, + { + "epoch": 0.53, + "learning_rate": 9.616846438298749e-06, + "loss": 0.9751, + "step": 16474 + }, + { + "epoch": 0.53, + "learning_rate": 9.615811397239249e-06, + "loss": 0.8921, + "step": 16475 + }, + { + "epoch": 0.53, + "learning_rate": 9.614776360301669e-06, + "loss": 0.9688, + "step": 16476 + }, + { + "epoch": 0.53, + "learning_rate": 9.613741327497111e-06, + "loss": 0.9507, + "step": 16477 + }, + { + "epoch": 0.53, + "learning_rate": 9.612706298836687e-06, + "loss": 0.9355, + "step": 16478 + }, + { + "epoch": 0.53, + "learning_rate": 9.61167127433149e-06, + "loss": 1.1074, + "step": 16479 + }, + { + "epoch": 0.53, + "learning_rate": 9.610636253992632e-06, + "loss": 0.9819, + "step": 16480 + }, + { + "epoch": 0.53, + "learning_rate": 9.609601237831215e-06, + "loss": 1.0498, + "step": 16481 + }, + { + "epoch": 0.53, + "learning_rate": 9.608566225858344e-06, + "loss": 0.9614, + "step": 16482 + }, + { + "epoch": 0.53, + "learning_rate": 9.607531218085126e-06, + "loss": 0.5403, + "step": 16483 + }, + { + "epoch": 0.53, + "learning_rate": 9.606496214522662e-06, + "loss": 1.0542, + "step": 16484 + }, + { + "epoch": 0.53, + "learning_rate": 9.605461215182059e-06, + "loss": 0.9761, + "step": 16485 + }, + { + "epoch": 0.53, + "learning_rate": 9.604426220074425e-06, + "loss": 0.8296, + "step": 16486 + }, + { + "epoch": 0.53, + "learning_rate": 9.603391229210855e-06, + "loss": 1.0596, + "step": 16487 + }, + { + "epoch": 0.53, + "learning_rate": 9.602356242602458e-06, + "loss": 1.0884, + "step": 16488 + }, + { + "epoch": 0.53, + "learning_rate": 9.60132126026034e-06, + "loss": 1.0859, + "step": 16489 + }, + { + "epoch": 0.53, + "learning_rate": 9.600286282195603e-06, + "loss": 1.0371, + "step": 16490 + }, + { + "epoch": 0.53, + "learning_rate": 9.59925130841935e-06, + "loss": 1.0933, + "step": 16491 + }, + { + "epoch": 0.53, + "learning_rate": 9.598216338942688e-06, + "loss": 1.0728, + "step": 16492 + }, + { + "epoch": 0.53, + "learning_rate": 9.597181373776725e-06, + "loss": 1.0137, + "step": 16493 + }, + { + "epoch": 0.53, + "learning_rate": 9.596146412932555e-06, + "loss": 0.9785, + "step": 16494 + }, + { + "epoch": 0.53, + "learning_rate": 9.595111456421287e-06, + "loss": 1.0293, + "step": 16495 + }, + { + "epoch": 0.53, + "learning_rate": 9.594076504254026e-06, + "loss": 0.9785, + "step": 16496 + }, + { + "epoch": 0.53, + "learning_rate": 9.593041556441874e-06, + "loss": 0.9937, + "step": 16497 + }, + { + "epoch": 0.53, + "learning_rate": 9.592006612995937e-06, + "loss": 1.0537, + "step": 16498 + }, + { + "epoch": 0.53, + "learning_rate": 9.590971673927315e-06, + "loss": 0.9985, + "step": 16499 + }, + { + "epoch": 0.53, + "learning_rate": 9.589936739247118e-06, + "loss": 1.0068, + "step": 16500 + }, + { + "epoch": 0.53, + "learning_rate": 9.588901808966446e-06, + "loss": 0.9883, + "step": 16501 + }, + { + "epoch": 0.53, + "learning_rate": 9.5878668830964e-06, + "loss": 1.0649, + "step": 16502 + }, + { + "epoch": 0.53, + "learning_rate": 9.586831961648087e-06, + "loss": 1.0513, + "step": 16503 + }, + { + "epoch": 0.53, + "learning_rate": 9.58579704463261e-06, + "loss": 0.998, + "step": 16504 + }, + { + "epoch": 0.53, + "learning_rate": 9.584762132061069e-06, + "loss": 0.9272, + "step": 16505 + }, + { + "epoch": 0.53, + "learning_rate": 9.583727223944576e-06, + "loss": 0.9307, + "step": 16506 + }, + { + "epoch": 0.53, + "learning_rate": 9.582692320294227e-06, + "loss": 0.96, + "step": 16507 + }, + { + "epoch": 0.53, + "learning_rate": 9.58165742112113e-06, + "loss": 1.1001, + "step": 16508 + }, + { + "epoch": 0.53, + "learning_rate": 9.580622526436382e-06, + "loss": 1.0278, + "step": 16509 + }, + { + "epoch": 0.53, + "learning_rate": 9.57958763625109e-06, + "loss": 1.0024, + "step": 16510 + }, + { + "epoch": 0.53, + "learning_rate": 9.578552750576361e-06, + "loss": 0.9614, + "step": 16511 + }, + { + "epoch": 0.53, + "learning_rate": 9.577517869423292e-06, + "loss": 1.0059, + "step": 16512 + }, + { + "epoch": 0.53, + "learning_rate": 9.576482992802989e-06, + "loss": 0.5383, + "step": 16513 + }, + { + "epoch": 0.53, + "learning_rate": 9.575448120726553e-06, + "loss": 0.8999, + "step": 16514 + }, + { + "epoch": 0.53, + "learning_rate": 9.57441325320509e-06, + "loss": 1.0176, + "step": 16515 + }, + { + "epoch": 0.53, + "learning_rate": 9.573378390249707e-06, + "loss": 0.9536, + "step": 16516 + }, + { + "epoch": 0.53, + "learning_rate": 9.572343531871496e-06, + "loss": 0.9971, + "step": 16517 + }, + { + "epoch": 0.53, + "learning_rate": 9.571308678081566e-06, + "loss": 0.9766, + "step": 16518 + }, + { + "epoch": 0.53, + "learning_rate": 9.57027382889102e-06, + "loss": 1.0601, + "step": 16519 + }, + { + "epoch": 0.53, + "learning_rate": 9.56923898431096e-06, + "loss": 0.8623, + "step": 16520 + }, + { + "epoch": 0.53, + "learning_rate": 9.568204144352488e-06, + "loss": 0.8672, + "step": 16521 + }, + { + "epoch": 0.53, + "learning_rate": 9.567169309026709e-06, + "loss": 0.998, + "step": 16522 + }, + { + "epoch": 0.53, + "learning_rate": 9.566134478344728e-06, + "loss": 0.9707, + "step": 16523 + }, + { + "epoch": 0.53, + "learning_rate": 9.56509965231764e-06, + "loss": 0.895, + "step": 16524 + }, + { + "epoch": 0.53, + "learning_rate": 9.56406483095655e-06, + "loss": 0.9424, + "step": 16525 + }, + { + "epoch": 0.53, + "learning_rate": 9.563030014272565e-06, + "loss": 0.9438, + "step": 16526 + }, + { + "epoch": 0.53, + "learning_rate": 9.561995202276781e-06, + "loss": 0.8682, + "step": 16527 + }, + { + "epoch": 0.53, + "learning_rate": 9.560960394980306e-06, + "loss": 1.0605, + "step": 16528 + }, + { + "epoch": 0.53, + "learning_rate": 9.559925592394238e-06, + "loss": 1.0117, + "step": 16529 + }, + { + "epoch": 0.53, + "learning_rate": 9.558890794529683e-06, + "loss": 1.0029, + "step": 16530 + }, + { + "epoch": 0.53, + "learning_rate": 9.557856001397745e-06, + "loss": 0.9805, + "step": 16531 + }, + { + "epoch": 0.53, + "learning_rate": 9.556821213009518e-06, + "loss": 1.0181, + "step": 16532 + }, + { + "epoch": 0.53, + "learning_rate": 9.555786429376111e-06, + "loss": 0.9893, + "step": 16533 + }, + { + "epoch": 0.53, + "learning_rate": 9.554751650508621e-06, + "loss": 0.9707, + "step": 16534 + }, + { + "epoch": 0.53, + "learning_rate": 9.553716876418155e-06, + "loss": 1.0093, + "step": 16535 + }, + { + "epoch": 0.53, + "learning_rate": 9.552682107115812e-06, + "loss": 0.9312, + "step": 16536 + }, + { + "epoch": 0.53, + "learning_rate": 9.551647342612696e-06, + "loss": 0.9507, + "step": 16537 + }, + { + "epoch": 0.53, + "learning_rate": 9.550612582919906e-06, + "loss": 0.9429, + "step": 16538 + }, + { + "epoch": 0.53, + "learning_rate": 9.549577828048553e-06, + "loss": 1.0181, + "step": 16539 + }, + { + "epoch": 0.53, + "learning_rate": 9.548543078009725e-06, + "loss": 0.854, + "step": 16540 + }, + { + "epoch": 0.53, + "learning_rate": 9.54750833281453e-06, + "loss": 0.9468, + "step": 16541 + }, + { + "epoch": 0.53, + "learning_rate": 9.54647359247407e-06, + "loss": 1.0542, + "step": 16542 + }, + { + "epoch": 0.53, + "learning_rate": 9.545438856999448e-06, + "loss": 0.9619, + "step": 16543 + }, + { + "epoch": 0.53, + "learning_rate": 9.544404126401762e-06, + "loss": 1.1064, + "step": 16544 + }, + { + "epoch": 0.53, + "learning_rate": 9.543369400692118e-06, + "loss": 0.9331, + "step": 16545 + }, + { + "epoch": 0.53, + "learning_rate": 9.542334679881617e-06, + "loss": 0.9668, + "step": 16546 + }, + { + "epoch": 0.53, + "learning_rate": 9.541299963981353e-06, + "loss": 0.9629, + "step": 16547 + }, + { + "epoch": 0.53, + "learning_rate": 9.540265253002434e-06, + "loss": 0.916, + "step": 16548 + }, + { + "epoch": 0.53, + "learning_rate": 9.539230546955961e-06, + "loss": 0.9946, + "step": 16549 + }, + { + "epoch": 0.53, + "learning_rate": 9.538195845853035e-06, + "loss": 0.9526, + "step": 16550 + }, + { + "epoch": 0.53, + "learning_rate": 9.537161149704756e-06, + "loss": 1.0425, + "step": 16551 + }, + { + "epoch": 0.53, + "learning_rate": 9.536126458522224e-06, + "loss": 1.0098, + "step": 16552 + }, + { + "epoch": 0.53, + "learning_rate": 9.535091772316542e-06, + "loss": 1.0142, + "step": 16553 + }, + { + "epoch": 0.53, + "learning_rate": 9.534057091098813e-06, + "loss": 0.9136, + "step": 16554 + }, + { + "epoch": 0.53, + "learning_rate": 9.533022414880136e-06, + "loss": 1.0186, + "step": 16555 + }, + { + "epoch": 0.53, + "learning_rate": 9.53198774367161e-06, + "loss": 0.9971, + "step": 16556 + }, + { + "epoch": 0.53, + "learning_rate": 9.530953077484337e-06, + "loss": 0.9736, + "step": 16557 + }, + { + "epoch": 0.53, + "learning_rate": 9.529918416329419e-06, + "loss": 0.9932, + "step": 16558 + }, + { + "epoch": 0.53, + "learning_rate": 9.528883760217953e-06, + "loss": 1.0049, + "step": 16559 + }, + { + "epoch": 0.53, + "learning_rate": 9.527849109161047e-06, + "loss": 0.9062, + "step": 16560 + }, + { + "epoch": 0.53, + "learning_rate": 9.5268144631698e-06, + "loss": 0.938, + "step": 16561 + }, + { + "epoch": 0.53, + "learning_rate": 9.525779822255306e-06, + "loss": 0.9565, + "step": 16562 + }, + { + "epoch": 0.53, + "learning_rate": 9.52474518642867e-06, + "loss": 1.1855, + "step": 16563 + }, + { + "epoch": 0.53, + "learning_rate": 9.523710555700992e-06, + "loss": 0.894, + "step": 16564 + }, + { + "epoch": 0.53, + "learning_rate": 9.522675930083372e-06, + "loss": 1.0498, + "step": 16565 + }, + { + "epoch": 0.53, + "learning_rate": 9.521641309586911e-06, + "loss": 1.0371, + "step": 16566 + }, + { + "epoch": 0.53, + "learning_rate": 9.52060669422271e-06, + "loss": 1.0261, + "step": 16567 + }, + { + "epoch": 0.53, + "learning_rate": 9.519572084001868e-06, + "loss": 0.9595, + "step": 16568 + }, + { + "epoch": 0.53, + "learning_rate": 9.51853747893549e-06, + "loss": 0.918, + "step": 16569 + }, + { + "epoch": 0.53, + "learning_rate": 9.517502879034669e-06, + "loss": 0.9868, + "step": 16570 + }, + { + "epoch": 0.53, + "learning_rate": 9.516468284310507e-06, + "loss": 0.9805, + "step": 16571 + }, + { + "epoch": 0.53, + "learning_rate": 9.515433694774105e-06, + "loss": 1.0464, + "step": 16572 + }, + { + "epoch": 0.53, + "learning_rate": 9.514399110436564e-06, + "loss": 1.0132, + "step": 16573 + }, + { + "epoch": 0.53, + "learning_rate": 9.513364531308983e-06, + "loss": 1.0283, + "step": 16574 + }, + { + "epoch": 0.53, + "learning_rate": 9.512329957402461e-06, + "loss": 0.9849, + "step": 16575 + }, + { + "epoch": 0.53, + "learning_rate": 9.511295388728105e-06, + "loss": 1.0078, + "step": 16576 + }, + { + "epoch": 0.53, + "learning_rate": 9.510260825297003e-06, + "loss": 1.1113, + "step": 16577 + }, + { + "epoch": 0.53, + "learning_rate": 9.50922626712026e-06, + "loss": 0.959, + "step": 16578 + }, + { + "epoch": 0.53, + "learning_rate": 9.508191714208976e-06, + "loss": 0.9468, + "step": 16579 + }, + { + "epoch": 0.53, + "learning_rate": 9.507157166574252e-06, + "loss": 1.064, + "step": 16580 + }, + { + "epoch": 0.53, + "learning_rate": 9.506122624227184e-06, + "loss": 1.0488, + "step": 16581 + }, + { + "epoch": 0.53, + "learning_rate": 9.505088087178876e-06, + "loss": 0.9058, + "step": 16582 + }, + { + "epoch": 0.53, + "learning_rate": 9.504053555440423e-06, + "loss": 0.874, + "step": 16583 + }, + { + "epoch": 0.53, + "learning_rate": 9.503019029022932e-06, + "loss": 0.8735, + "step": 16584 + }, + { + "epoch": 0.53, + "learning_rate": 9.501984507937492e-06, + "loss": 0.9497, + "step": 16585 + }, + { + "epoch": 0.53, + "learning_rate": 9.500949992195207e-06, + "loss": 1.0098, + "step": 16586 + }, + { + "epoch": 0.53, + "learning_rate": 9.499915481807177e-06, + "loss": 0.916, + "step": 16587 + }, + { + "epoch": 0.53, + "learning_rate": 9.498880976784498e-06, + "loss": 0.9263, + "step": 16588 + }, + { + "epoch": 0.53, + "learning_rate": 9.497846477138277e-06, + "loss": 0.939, + "step": 16589 + }, + { + "epoch": 0.53, + "learning_rate": 9.496811982879604e-06, + "loss": 1.0054, + "step": 16590 + }, + { + "epoch": 0.53, + "learning_rate": 9.495777494019583e-06, + "loss": 1.0747, + "step": 16591 + }, + { + "epoch": 0.53, + "learning_rate": 9.494743010569311e-06, + "loss": 0.9438, + "step": 16592 + }, + { + "epoch": 0.53, + "learning_rate": 9.493708532539886e-06, + "loss": 0.9062, + "step": 16593 + }, + { + "epoch": 0.53, + "learning_rate": 9.492674059942408e-06, + "loss": 0.9175, + "step": 16594 + }, + { + "epoch": 0.53, + "learning_rate": 9.491639592787977e-06, + "loss": 0.9004, + "step": 16595 + }, + { + "epoch": 0.53, + "learning_rate": 9.490605131087692e-06, + "loss": 0.9399, + "step": 16596 + }, + { + "epoch": 0.53, + "learning_rate": 9.489570674852648e-06, + "loss": 1.0361, + "step": 16597 + }, + { + "epoch": 0.53, + "learning_rate": 9.488536224093944e-06, + "loss": 0.9814, + "step": 16598 + }, + { + "epoch": 0.53, + "learning_rate": 9.487501778822685e-06, + "loss": 0.9966, + "step": 16599 + }, + { + "epoch": 0.53, + "learning_rate": 9.486467339049963e-06, + "loss": 0.835, + "step": 16600 + }, + { + "epoch": 0.53, + "learning_rate": 9.485432904786877e-06, + "loss": 1.0132, + "step": 16601 + }, + { + "epoch": 0.53, + "learning_rate": 9.484398476044527e-06, + "loss": 0.8848, + "step": 16602 + }, + { + "epoch": 0.53, + "learning_rate": 9.483364052834008e-06, + "loss": 0.9512, + "step": 16603 + }, + { + "epoch": 0.53, + "learning_rate": 9.482329635166424e-06, + "loss": 1.002, + "step": 16604 + }, + { + "epoch": 0.53, + "learning_rate": 9.481295223052869e-06, + "loss": 0.8662, + "step": 16605 + }, + { + "epoch": 0.53, + "learning_rate": 9.480260816504447e-06, + "loss": 0.9385, + "step": 16606 + }, + { + "epoch": 0.53, + "learning_rate": 9.479226415532246e-06, + "loss": 1.1392, + "step": 16607 + }, + { + "epoch": 0.53, + "learning_rate": 9.478192020147368e-06, + "loss": 0.9336, + "step": 16608 + }, + { + "epoch": 0.53, + "learning_rate": 9.477157630360915e-06, + "loss": 1.0283, + "step": 16609 + }, + { + "epoch": 0.53, + "learning_rate": 9.47612324618398e-06, + "loss": 1.0566, + "step": 16610 + }, + { + "epoch": 0.53, + "learning_rate": 9.475088867627663e-06, + "loss": 1.0288, + "step": 16611 + }, + { + "epoch": 0.53, + "learning_rate": 9.474054494703063e-06, + "loss": 0.9243, + "step": 16612 + }, + { + "epoch": 0.53, + "learning_rate": 9.473020127421274e-06, + "loss": 1.0352, + "step": 16613 + }, + { + "epoch": 0.53, + "learning_rate": 9.471985765793401e-06, + "loss": 0.8799, + "step": 16614 + }, + { + "epoch": 0.53, + "learning_rate": 9.470951409830533e-06, + "loss": 0.9272, + "step": 16615 + }, + { + "epoch": 0.53, + "learning_rate": 9.469917059543769e-06, + "loss": 0.8838, + "step": 16616 + }, + { + "epoch": 0.53, + "learning_rate": 9.46888271494421e-06, + "loss": 1.0884, + "step": 16617 + }, + { + "epoch": 0.53, + "learning_rate": 9.467848376042951e-06, + "loss": 1.061, + "step": 16618 + }, + { + "epoch": 0.53, + "learning_rate": 9.466814042851089e-06, + "loss": 0.9712, + "step": 16619 + }, + { + "epoch": 0.53, + "learning_rate": 9.465779715379723e-06, + "loss": 1.0532, + "step": 16620 + }, + { + "epoch": 0.53, + "learning_rate": 9.46474539363995e-06, + "loss": 0.874, + "step": 16621 + }, + { + "epoch": 0.53, + "learning_rate": 9.463711077642871e-06, + "loss": 0.9087, + "step": 16622 + }, + { + "epoch": 0.53, + "learning_rate": 9.462676767399573e-06, + "loss": 1.0093, + "step": 16623 + }, + { + "epoch": 0.53, + "learning_rate": 9.46164246292116e-06, + "loss": 0.9321, + "step": 16624 + }, + { + "epoch": 0.53, + "learning_rate": 9.460608164218727e-06, + "loss": 0.8838, + "step": 16625 + }, + { + "epoch": 0.53, + "learning_rate": 9.459573871303371e-06, + "loss": 0.939, + "step": 16626 + }, + { + "epoch": 0.53, + "learning_rate": 9.458539584186189e-06, + "loss": 1.0137, + "step": 16627 + }, + { + "epoch": 0.53, + "learning_rate": 9.457505302878281e-06, + "loss": 1.0049, + "step": 16628 + }, + { + "epoch": 0.53, + "learning_rate": 9.45647102739074e-06, + "loss": 1.0, + "step": 16629 + }, + { + "epoch": 0.53, + "learning_rate": 9.455436757734663e-06, + "loss": 1.0532, + "step": 16630 + }, + { + "epoch": 0.53, + "learning_rate": 9.454402493921145e-06, + "loss": 0.9424, + "step": 16631 + }, + { + "epoch": 0.53, + "learning_rate": 9.453368235961284e-06, + "loss": 0.9077, + "step": 16632 + }, + { + "epoch": 0.53, + "learning_rate": 9.452333983866178e-06, + "loss": 0.9922, + "step": 16633 + }, + { + "epoch": 0.53, + "learning_rate": 9.451299737646924e-06, + "loss": 0.5015, + "step": 16634 + }, + { + "epoch": 0.53, + "learning_rate": 9.450265497314615e-06, + "loss": 1.0273, + "step": 16635 + }, + { + "epoch": 0.53, + "learning_rate": 9.449231262880347e-06, + "loss": 1.0322, + "step": 16636 + }, + { + "epoch": 0.53, + "learning_rate": 9.448197034355222e-06, + "loss": 1.0469, + "step": 16637 + }, + { + "epoch": 0.53, + "learning_rate": 9.44716281175033e-06, + "loss": 1.0542, + "step": 16638 + }, + { + "epoch": 0.53, + "learning_rate": 9.44612859507677e-06, + "loss": 1.0352, + "step": 16639 + }, + { + "epoch": 0.53, + "learning_rate": 9.445094384345637e-06, + "loss": 0.9346, + "step": 16640 + }, + { + "epoch": 0.53, + "learning_rate": 9.444060179568026e-06, + "loss": 1.0396, + "step": 16641 + }, + { + "epoch": 0.53, + "learning_rate": 9.443025980755032e-06, + "loss": 0.9468, + "step": 16642 + }, + { + "epoch": 0.53, + "learning_rate": 9.441991787917756e-06, + "loss": 1.0527, + "step": 16643 + }, + { + "epoch": 0.53, + "learning_rate": 9.440957601067294e-06, + "loss": 1.0342, + "step": 16644 + }, + { + "epoch": 0.53, + "learning_rate": 9.439923420214733e-06, + "loss": 0.9756, + "step": 16645 + }, + { + "epoch": 0.53, + "learning_rate": 9.438889245371173e-06, + "loss": 1.0156, + "step": 16646 + }, + { + "epoch": 0.53, + "learning_rate": 9.437855076547713e-06, + "loss": 1.0156, + "step": 16647 + }, + { + "epoch": 0.53, + "learning_rate": 9.436820913755444e-06, + "loss": 1.0127, + "step": 16648 + }, + { + "epoch": 0.53, + "learning_rate": 9.435786757005465e-06, + "loss": 0.9556, + "step": 16649 + }, + { + "epoch": 0.53, + "learning_rate": 9.434752606308867e-06, + "loss": 0.9277, + "step": 16650 + }, + { + "epoch": 0.53, + "learning_rate": 9.43371846167675e-06, + "loss": 1.0073, + "step": 16651 + }, + { + "epoch": 0.53, + "learning_rate": 9.43268432312021e-06, + "loss": 1.0117, + "step": 16652 + }, + { + "epoch": 0.53, + "learning_rate": 9.431650190650336e-06, + "loss": 1.0449, + "step": 16653 + }, + { + "epoch": 0.53, + "learning_rate": 9.430616064278226e-06, + "loss": 1.0742, + "step": 16654 + }, + { + "epoch": 0.53, + "learning_rate": 9.429581944014976e-06, + "loss": 1.0161, + "step": 16655 + }, + { + "epoch": 0.53, + "learning_rate": 9.42854782987168e-06, + "loss": 0.9604, + "step": 16656 + }, + { + "epoch": 0.53, + "learning_rate": 9.427513721859434e-06, + "loss": 0.8291, + "step": 16657 + }, + { + "epoch": 0.53, + "learning_rate": 9.426479619989333e-06, + "loss": 1.1899, + "step": 16658 + }, + { + "epoch": 0.53, + "learning_rate": 9.425445524272474e-06, + "loss": 1.0522, + "step": 16659 + }, + { + "epoch": 0.53, + "learning_rate": 9.424411434719944e-06, + "loss": 0.9634, + "step": 16660 + }, + { + "epoch": 0.53, + "learning_rate": 9.423377351342842e-06, + "loss": 0.9902, + "step": 16661 + }, + { + "epoch": 0.53, + "learning_rate": 9.422343274152263e-06, + "loss": 1.0815, + "step": 16662 + }, + { + "epoch": 0.53, + "learning_rate": 9.421309203159303e-06, + "loss": 1.0444, + "step": 16663 + }, + { + "epoch": 0.53, + "learning_rate": 9.420275138375053e-06, + "loss": 1.0532, + "step": 16664 + }, + { + "epoch": 0.53, + "learning_rate": 9.419241079810609e-06, + "loss": 1.0239, + "step": 16665 + }, + { + "epoch": 0.53, + "learning_rate": 9.418207027477066e-06, + "loss": 1.0493, + "step": 16666 + }, + { + "epoch": 0.53, + "learning_rate": 9.417172981385522e-06, + "loss": 1.0396, + "step": 16667 + }, + { + "epoch": 0.53, + "learning_rate": 9.416138941547064e-06, + "loss": 0.9521, + "step": 16668 + }, + { + "epoch": 0.53, + "learning_rate": 9.415104907972787e-06, + "loss": 0.9775, + "step": 16669 + }, + { + "epoch": 0.53, + "learning_rate": 9.414070880673789e-06, + "loss": 0.8862, + "step": 16670 + }, + { + "epoch": 0.53, + "learning_rate": 9.41303685966116e-06, + "loss": 0.9185, + "step": 16671 + }, + { + "epoch": 0.53, + "learning_rate": 9.412002844945996e-06, + "loss": 0.791, + "step": 16672 + }, + { + "epoch": 0.53, + "learning_rate": 9.410968836539394e-06, + "loss": 0.9834, + "step": 16673 + }, + { + "epoch": 0.53, + "learning_rate": 9.409934834452444e-06, + "loss": 1.02, + "step": 16674 + }, + { + "epoch": 0.53, + "learning_rate": 9.40890083869624e-06, + "loss": 0.9219, + "step": 16675 + }, + { + "epoch": 0.53, + "learning_rate": 9.407866849281873e-06, + "loss": 0.958, + "step": 16676 + }, + { + "epoch": 0.53, + "learning_rate": 9.40683286622044e-06, + "loss": 0.9893, + "step": 16677 + }, + { + "epoch": 0.53, + "learning_rate": 9.405798889523036e-06, + "loss": 0.9844, + "step": 16678 + }, + { + "epoch": 0.53, + "learning_rate": 9.404764919200749e-06, + "loss": 0.9795, + "step": 16679 + }, + { + "epoch": 0.53, + "learning_rate": 9.403730955264677e-06, + "loss": 0.9678, + "step": 16680 + }, + { + "epoch": 0.53, + "learning_rate": 9.402696997725914e-06, + "loss": 1.0015, + "step": 16681 + }, + { + "epoch": 0.53, + "learning_rate": 9.40166304659555e-06, + "loss": 1.1191, + "step": 16682 + }, + { + "epoch": 0.53, + "learning_rate": 9.400629101884681e-06, + "loss": 0.9609, + "step": 16683 + }, + { + "epoch": 0.53, + "learning_rate": 9.399595163604395e-06, + "loss": 0.8965, + "step": 16684 + }, + { + "epoch": 0.53, + "learning_rate": 9.398561231765789e-06, + "loss": 1.0312, + "step": 16685 + }, + { + "epoch": 0.53, + "learning_rate": 9.397527306379955e-06, + "loss": 1.1484, + "step": 16686 + }, + { + "epoch": 0.53, + "learning_rate": 9.396493387457988e-06, + "loss": 0.5112, + "step": 16687 + }, + { + "epoch": 0.53, + "learning_rate": 9.395459475010977e-06, + "loss": 0.9121, + "step": 16688 + }, + { + "epoch": 0.53, + "learning_rate": 9.394425569050018e-06, + "loss": 1.0093, + "step": 16689 + }, + { + "epoch": 0.53, + "learning_rate": 9.393391669586205e-06, + "loss": 1.0024, + "step": 16690 + }, + { + "epoch": 0.53, + "learning_rate": 9.392357776630626e-06, + "loss": 1.0752, + "step": 16691 + }, + { + "epoch": 0.53, + "learning_rate": 9.391323890194374e-06, + "loss": 0.894, + "step": 16692 + }, + { + "epoch": 0.53, + "learning_rate": 9.390290010288545e-06, + "loss": 1.0312, + "step": 16693 + }, + { + "epoch": 0.53, + "learning_rate": 9.389256136924226e-06, + "loss": 0.9824, + "step": 16694 + }, + { + "epoch": 0.53, + "learning_rate": 9.388222270112515e-06, + "loss": 1.02, + "step": 16695 + }, + { + "epoch": 0.53, + "learning_rate": 9.387188409864502e-06, + "loss": 1.0298, + "step": 16696 + }, + { + "epoch": 0.53, + "learning_rate": 9.386154556191282e-06, + "loss": 0.9224, + "step": 16697 + }, + { + "epoch": 0.53, + "learning_rate": 9.385120709103943e-06, + "loss": 0.9639, + "step": 16698 + }, + { + "epoch": 0.53, + "learning_rate": 9.384086868613574e-06, + "loss": 0.9961, + "step": 16699 + }, + { + "epoch": 0.53, + "learning_rate": 9.383053034731275e-06, + "loss": 0.5029, + "step": 16700 + }, + { + "epoch": 0.53, + "learning_rate": 9.382019207468131e-06, + "loss": 0.562, + "step": 16701 + }, + { + "epoch": 0.53, + "learning_rate": 9.38098538683524e-06, + "loss": 0.9385, + "step": 16702 + }, + { + "epoch": 0.53, + "learning_rate": 9.379951572843689e-06, + "loss": 0.9468, + "step": 16703 + }, + { + "epoch": 0.53, + "learning_rate": 9.378917765504572e-06, + "loss": 0.5396, + "step": 16704 + }, + { + "epoch": 0.53, + "learning_rate": 9.377883964828983e-06, + "loss": 0.8691, + "step": 16705 + }, + { + "epoch": 0.53, + "learning_rate": 9.376850170828007e-06, + "loss": 0.9155, + "step": 16706 + }, + { + "epoch": 0.53, + "learning_rate": 9.375816383512739e-06, + "loss": 0.835, + "step": 16707 + }, + { + "epoch": 0.53, + "learning_rate": 9.374782602894272e-06, + "loss": 0.9243, + "step": 16708 + }, + { + "epoch": 0.53, + "learning_rate": 9.373748828983693e-06, + "loss": 1.1025, + "step": 16709 + }, + { + "epoch": 0.53, + "learning_rate": 9.372715061792097e-06, + "loss": 0.9736, + "step": 16710 + }, + { + "epoch": 0.53, + "learning_rate": 9.371681301330575e-06, + "loss": 0.9248, + "step": 16711 + }, + { + "epoch": 0.53, + "learning_rate": 9.37064754761022e-06, + "loss": 0.9775, + "step": 16712 + }, + { + "epoch": 0.53, + "learning_rate": 9.369613800642118e-06, + "loss": 0.9951, + "step": 16713 + }, + { + "epoch": 0.53, + "learning_rate": 9.368580060437362e-06, + "loss": 0.9517, + "step": 16714 + }, + { + "epoch": 0.53, + "learning_rate": 9.367546327007041e-06, + "loss": 1.0454, + "step": 16715 + }, + { + "epoch": 0.53, + "learning_rate": 9.36651260036225e-06, + "loss": 0.8296, + "step": 16716 + }, + { + "epoch": 0.53, + "learning_rate": 9.365478880514078e-06, + "loss": 1.0059, + "step": 16717 + }, + { + "epoch": 0.53, + "learning_rate": 9.364445167473614e-06, + "loss": 1.0225, + "step": 16718 + }, + { + "epoch": 0.53, + "learning_rate": 9.363411461251952e-06, + "loss": 0.9727, + "step": 16719 + }, + { + "epoch": 0.53, + "learning_rate": 9.36237776186018e-06, + "loss": 0.9189, + "step": 16720 + }, + { + "epoch": 0.53, + "learning_rate": 9.361344069309387e-06, + "loss": 1.0308, + "step": 16721 + }, + { + "epoch": 0.53, + "learning_rate": 9.360310383610668e-06, + "loss": 1.0273, + "step": 16722 + }, + { + "epoch": 0.53, + "learning_rate": 9.359276704775109e-06, + "loss": 0.9346, + "step": 16723 + }, + { + "epoch": 0.53, + "learning_rate": 9.358243032813801e-06, + "loss": 1.0581, + "step": 16724 + }, + { + "epoch": 0.53, + "learning_rate": 9.357209367737837e-06, + "loss": 0.9932, + "step": 16725 + }, + { + "epoch": 0.53, + "learning_rate": 9.356175709558302e-06, + "loss": 0.9512, + "step": 16726 + }, + { + "epoch": 0.53, + "learning_rate": 9.355142058286293e-06, + "loss": 0.8311, + "step": 16727 + }, + { + "epoch": 0.53, + "learning_rate": 9.354108413932897e-06, + "loss": 1.0967, + "step": 16728 + }, + { + "epoch": 0.54, + "learning_rate": 9.353074776509198e-06, + "loss": 0.9243, + "step": 16729 + }, + { + "epoch": 0.54, + "learning_rate": 9.352041146026293e-06, + "loss": 0.9634, + "step": 16730 + }, + { + "epoch": 0.54, + "learning_rate": 9.351007522495269e-06, + "loss": 0.9902, + "step": 16731 + }, + { + "epoch": 0.54, + "learning_rate": 9.349973905927214e-06, + "loss": 1.0874, + "step": 16732 + }, + { + "epoch": 0.54, + "learning_rate": 9.348940296333222e-06, + "loss": 0.8662, + "step": 16733 + }, + { + "epoch": 0.54, + "learning_rate": 9.347906693724379e-06, + "loss": 1.0166, + "step": 16734 + }, + { + "epoch": 0.54, + "learning_rate": 9.34687309811178e-06, + "loss": 1.0039, + "step": 16735 + }, + { + "epoch": 0.54, + "learning_rate": 9.345839509506506e-06, + "loss": 0.938, + "step": 16736 + }, + { + "epoch": 0.54, + "learning_rate": 9.34480592791965e-06, + "loss": 0.9531, + "step": 16737 + }, + { + "epoch": 0.54, + "learning_rate": 9.3437723533623e-06, + "loss": 0.9722, + "step": 16738 + }, + { + "epoch": 0.54, + "learning_rate": 9.342738785845547e-06, + "loss": 0.981, + "step": 16739 + }, + { + "epoch": 0.54, + "learning_rate": 9.34170522538048e-06, + "loss": 0.8975, + "step": 16740 + }, + { + "epoch": 0.54, + "learning_rate": 9.340671671978187e-06, + "loss": 0.9141, + "step": 16741 + }, + { + "epoch": 0.54, + "learning_rate": 9.33963812564976e-06, + "loss": 0.9243, + "step": 16742 + }, + { + "epoch": 0.54, + "learning_rate": 9.338604586406283e-06, + "loss": 0.9551, + "step": 16743 + }, + { + "epoch": 0.54, + "learning_rate": 9.337571054258846e-06, + "loss": 0.9736, + "step": 16744 + }, + { + "epoch": 0.54, + "learning_rate": 9.336537529218539e-06, + "loss": 0.9321, + "step": 16745 + }, + { + "epoch": 0.54, + "learning_rate": 9.33550401129645e-06, + "loss": 0.8469, + "step": 16746 + }, + { + "epoch": 0.54, + "learning_rate": 9.334470500503666e-06, + "loss": 1.0088, + "step": 16747 + }, + { + "epoch": 0.54, + "learning_rate": 9.333436996851278e-06, + "loss": 0.9521, + "step": 16748 + }, + { + "epoch": 0.54, + "learning_rate": 9.332403500350374e-06, + "loss": 1.0337, + "step": 16749 + }, + { + "epoch": 0.54, + "learning_rate": 9.331370011012045e-06, + "loss": 1.0278, + "step": 16750 + }, + { + "epoch": 0.54, + "learning_rate": 9.330336528847372e-06, + "loss": 1.0, + "step": 16751 + }, + { + "epoch": 0.54, + "learning_rate": 9.329303053867445e-06, + "loss": 0.979, + "step": 16752 + }, + { + "epoch": 0.54, + "learning_rate": 9.328269586083355e-06, + "loss": 0.9565, + "step": 16753 + }, + { + "epoch": 0.54, + "learning_rate": 9.327236125506191e-06, + "loss": 0.999, + "step": 16754 + }, + { + "epoch": 0.54, + "learning_rate": 9.326202672147037e-06, + "loss": 0.9795, + "step": 16755 + }, + { + "epoch": 0.54, + "learning_rate": 9.325169226016985e-06, + "loss": 0.9878, + "step": 16756 + }, + { + "epoch": 0.54, + "learning_rate": 9.32413578712712e-06, + "loss": 0.9658, + "step": 16757 + }, + { + "epoch": 0.54, + "learning_rate": 9.323102355488528e-06, + "loss": 1.022, + "step": 16758 + }, + { + "epoch": 0.54, + "learning_rate": 9.3220689311123e-06, + "loss": 1.0171, + "step": 16759 + }, + { + "epoch": 0.54, + "learning_rate": 9.32103551400952e-06, + "loss": 0.469, + "step": 16760 + }, + { + "epoch": 0.54, + "learning_rate": 9.320002104191278e-06, + "loss": 0.8433, + "step": 16761 + }, + { + "epoch": 0.54, + "learning_rate": 9.318968701668664e-06, + "loss": 0.9409, + "step": 16762 + }, + { + "epoch": 0.54, + "learning_rate": 9.31793530645276e-06, + "loss": 1.0146, + "step": 16763 + }, + { + "epoch": 0.54, + "learning_rate": 9.316901918554652e-06, + "loss": 0.9346, + "step": 16764 + }, + { + "epoch": 0.54, + "learning_rate": 9.315868537985437e-06, + "loss": 0.8896, + "step": 16765 + }, + { + "epoch": 0.54, + "learning_rate": 9.31483516475619e-06, + "loss": 1.0146, + "step": 16766 + }, + { + "epoch": 0.54, + "learning_rate": 9.313801798878008e-06, + "loss": 1.0396, + "step": 16767 + }, + { + "epoch": 0.54, + "learning_rate": 9.31276844036197e-06, + "loss": 0.9619, + "step": 16768 + }, + { + "epoch": 0.54, + "learning_rate": 9.311735089219165e-06, + "loss": 1.0835, + "step": 16769 + }, + { + "epoch": 0.54, + "learning_rate": 9.310701745460681e-06, + "loss": 0.9287, + "step": 16770 + }, + { + "epoch": 0.54, + "learning_rate": 9.309668409097607e-06, + "loss": 0.9365, + "step": 16771 + }, + { + "epoch": 0.54, + "learning_rate": 9.308635080141025e-06, + "loss": 0.8328, + "step": 16772 + }, + { + "epoch": 0.54, + "learning_rate": 9.307601758602028e-06, + "loss": 0.5452, + "step": 16773 + }, + { + "epoch": 0.54, + "learning_rate": 9.306568444491692e-06, + "loss": 0.9004, + "step": 16774 + }, + { + "epoch": 0.54, + "learning_rate": 9.305535137821112e-06, + "loss": 0.9307, + "step": 16775 + }, + { + "epoch": 0.54, + "learning_rate": 9.30450183860137e-06, + "loss": 0.8621, + "step": 16776 + }, + { + "epoch": 0.54, + "learning_rate": 9.303468546843553e-06, + "loss": 0.9746, + "step": 16777 + }, + { + "epoch": 0.54, + "learning_rate": 9.302435262558748e-06, + "loss": 1.0137, + "step": 16778 + }, + { + "epoch": 0.54, + "learning_rate": 9.30140198575804e-06, + "loss": 0.9883, + "step": 16779 + }, + { + "epoch": 0.54, + "learning_rate": 9.300368716452521e-06, + "loss": 1.0249, + "step": 16780 + }, + { + "epoch": 0.54, + "learning_rate": 9.299335454653266e-06, + "loss": 0.9917, + "step": 16781 + }, + { + "epoch": 0.54, + "learning_rate": 9.298302200371367e-06, + "loss": 0.7676, + "step": 16782 + }, + { + "epoch": 0.54, + "learning_rate": 9.297268953617909e-06, + "loss": 1.0269, + "step": 16783 + }, + { + "epoch": 0.54, + "learning_rate": 9.296235714403976e-06, + "loss": 1.0962, + "step": 16784 + }, + { + "epoch": 0.54, + "learning_rate": 9.295202482740656e-06, + "loss": 0.9922, + "step": 16785 + }, + { + "epoch": 0.54, + "learning_rate": 9.294169258639032e-06, + "loss": 1.0342, + "step": 16786 + }, + { + "epoch": 0.54, + "learning_rate": 9.293136042110193e-06, + "loss": 1.0566, + "step": 16787 + }, + { + "epoch": 0.54, + "learning_rate": 9.292102833165225e-06, + "loss": 0.9048, + "step": 16788 + }, + { + "epoch": 0.54, + "learning_rate": 9.291069631815204e-06, + "loss": 1.0806, + "step": 16789 + }, + { + "epoch": 0.54, + "learning_rate": 9.290036438071223e-06, + "loss": 0.9473, + "step": 16790 + }, + { + "epoch": 0.54, + "learning_rate": 9.289003251944367e-06, + "loss": 0.9399, + "step": 16791 + }, + { + "epoch": 0.54, + "learning_rate": 9.287970073445716e-06, + "loss": 1.083, + "step": 16792 + }, + { + "epoch": 0.54, + "learning_rate": 9.286936902586361e-06, + "loss": 0.8774, + "step": 16793 + }, + { + "epoch": 0.54, + "learning_rate": 9.285903739377382e-06, + "loss": 0.917, + "step": 16794 + }, + { + "epoch": 0.54, + "learning_rate": 9.28487058382987e-06, + "loss": 0.937, + "step": 16795 + }, + { + "epoch": 0.54, + "learning_rate": 9.2838374359549e-06, + "loss": 0.9829, + "step": 16796 + }, + { + "epoch": 0.54, + "learning_rate": 9.282804295763564e-06, + "loss": 0.9888, + "step": 16797 + }, + { + "epoch": 0.54, + "learning_rate": 9.281771163266944e-06, + "loss": 1.0547, + "step": 16798 + }, + { + "epoch": 0.54, + "learning_rate": 9.280738038476122e-06, + "loss": 1.0254, + "step": 16799 + }, + { + "epoch": 0.54, + "learning_rate": 9.279704921402187e-06, + "loss": 1.1445, + "step": 16800 + }, + { + "epoch": 0.54, + "learning_rate": 9.278671812056223e-06, + "loss": 0.9609, + "step": 16801 + }, + { + "epoch": 0.54, + "learning_rate": 9.27763871044931e-06, + "loss": 0.9751, + "step": 16802 + }, + { + "epoch": 0.54, + "learning_rate": 9.276605616592535e-06, + "loss": 0.9619, + "step": 16803 + }, + { + "epoch": 0.54, + "learning_rate": 9.27557253049698e-06, + "loss": 0.8965, + "step": 16804 + }, + { + "epoch": 0.54, + "learning_rate": 9.27453945217373e-06, + "loss": 0.9492, + "step": 16805 + }, + { + "epoch": 0.54, + "learning_rate": 9.27350638163387e-06, + "loss": 0.8267, + "step": 16806 + }, + { + "epoch": 0.54, + "learning_rate": 9.272473318888481e-06, + "loss": 0.9395, + "step": 16807 + }, + { + "epoch": 0.54, + "learning_rate": 9.27144026394865e-06, + "loss": 1.0244, + "step": 16808 + }, + { + "epoch": 0.54, + "learning_rate": 9.270407216825455e-06, + "loss": 0.8677, + "step": 16809 + }, + { + "epoch": 0.54, + "learning_rate": 9.269374177529988e-06, + "loss": 1.0483, + "step": 16810 + }, + { + "epoch": 0.54, + "learning_rate": 9.268341146073325e-06, + "loss": 0.9072, + "step": 16811 + }, + { + "epoch": 0.54, + "learning_rate": 9.267308122466551e-06, + "loss": 1.0288, + "step": 16812 + }, + { + "epoch": 0.54, + "learning_rate": 9.266275106720752e-06, + "loss": 1.0952, + "step": 16813 + }, + { + "epoch": 0.54, + "learning_rate": 9.265242098847006e-06, + "loss": 0.8379, + "step": 16814 + }, + { + "epoch": 0.54, + "learning_rate": 9.2642090988564e-06, + "loss": 1.0142, + "step": 16815 + }, + { + "epoch": 0.54, + "learning_rate": 9.263176106760016e-06, + "loss": 0.8975, + "step": 16816 + }, + { + "epoch": 0.54, + "learning_rate": 9.262143122568937e-06, + "loss": 1.0625, + "step": 16817 + }, + { + "epoch": 0.54, + "learning_rate": 9.261110146294251e-06, + "loss": 1.0181, + "step": 16818 + }, + { + "epoch": 0.54, + "learning_rate": 9.26007717794703e-06, + "loss": 1.0361, + "step": 16819 + }, + { + "epoch": 0.54, + "learning_rate": 9.259044217538362e-06, + "loss": 1.0557, + "step": 16820 + }, + { + "epoch": 0.54, + "learning_rate": 9.25801126507933e-06, + "loss": 1.1094, + "step": 16821 + }, + { + "epoch": 0.54, + "learning_rate": 9.256978320581014e-06, + "loss": 1.0249, + "step": 16822 + }, + { + "epoch": 0.54, + "learning_rate": 9.255945384054502e-06, + "loss": 0.9956, + "step": 16823 + }, + { + "epoch": 0.54, + "learning_rate": 9.25491245551087e-06, + "loss": 0.999, + "step": 16824 + }, + { + "epoch": 0.54, + "learning_rate": 9.253879534961207e-06, + "loss": 1.1084, + "step": 16825 + }, + { + "epoch": 0.54, + "learning_rate": 9.252846622416587e-06, + "loss": 0.9307, + "step": 16826 + }, + { + "epoch": 0.54, + "learning_rate": 9.251813717888094e-06, + "loss": 1.1426, + "step": 16827 + }, + { + "epoch": 0.54, + "learning_rate": 9.250780821386816e-06, + "loss": 1.0679, + "step": 16828 + }, + { + "epoch": 0.54, + "learning_rate": 9.249747932923827e-06, + "loss": 1.0298, + "step": 16829 + }, + { + "epoch": 0.54, + "learning_rate": 9.248715052510213e-06, + "loss": 0.897, + "step": 16830 + }, + { + "epoch": 0.54, + "learning_rate": 9.247682180157056e-06, + "loss": 0.8955, + "step": 16831 + }, + { + "epoch": 0.54, + "learning_rate": 9.246649315875435e-06, + "loss": 0.9775, + "step": 16832 + }, + { + "epoch": 0.54, + "learning_rate": 9.24561645967644e-06, + "loss": 0.9604, + "step": 16833 + }, + { + "epoch": 0.54, + "learning_rate": 9.244583611571139e-06, + "loss": 1.0015, + "step": 16834 + }, + { + "epoch": 0.54, + "learning_rate": 9.24355077157062e-06, + "loss": 0.9424, + "step": 16835 + }, + { + "epoch": 0.54, + "learning_rate": 9.242517939685965e-06, + "loss": 0.8931, + "step": 16836 + }, + { + "epoch": 0.54, + "learning_rate": 9.241485115928253e-06, + "loss": 0.8809, + "step": 16837 + }, + { + "epoch": 0.54, + "learning_rate": 9.240452300308567e-06, + "loss": 1.0361, + "step": 16838 + }, + { + "epoch": 0.54, + "learning_rate": 9.239419492837986e-06, + "loss": 0.9683, + "step": 16839 + }, + { + "epoch": 0.54, + "learning_rate": 9.238386693527598e-06, + "loss": 0.9341, + "step": 16840 + }, + { + "epoch": 0.54, + "learning_rate": 9.237353902388474e-06, + "loss": 1.0928, + "step": 16841 + }, + { + "epoch": 0.54, + "learning_rate": 9.236321119431697e-06, + "loss": 0.9458, + "step": 16842 + }, + { + "epoch": 0.54, + "learning_rate": 9.23528834466835e-06, + "loss": 1.0513, + "step": 16843 + }, + { + "epoch": 0.54, + "learning_rate": 9.234255578109511e-06, + "loss": 0.9302, + "step": 16844 + }, + { + "epoch": 0.54, + "learning_rate": 9.233222819766267e-06, + "loss": 0.5608, + "step": 16845 + }, + { + "epoch": 0.54, + "learning_rate": 9.23219006964969e-06, + "loss": 1.147, + "step": 16846 + }, + { + "epoch": 0.54, + "learning_rate": 9.231157327770864e-06, + "loss": 0.9443, + "step": 16847 + }, + { + "epoch": 0.54, + "learning_rate": 9.230124594140872e-06, + "loss": 0.9541, + "step": 16848 + }, + { + "epoch": 0.54, + "learning_rate": 9.229091868770787e-06, + "loss": 0.8926, + "step": 16849 + }, + { + "epoch": 0.54, + "learning_rate": 9.228059151671693e-06, + "loss": 1.002, + "step": 16850 + }, + { + "epoch": 0.54, + "learning_rate": 9.227026442854674e-06, + "loss": 0.999, + "step": 16851 + }, + { + "epoch": 0.54, + "learning_rate": 9.225993742330802e-06, + "loss": 0.9404, + "step": 16852 + }, + { + "epoch": 0.54, + "learning_rate": 9.224961050111163e-06, + "loss": 0.8604, + "step": 16853 + }, + { + "epoch": 0.54, + "learning_rate": 9.22392836620683e-06, + "loss": 1.0308, + "step": 16854 + }, + { + "epoch": 0.54, + "learning_rate": 9.22289569062889e-06, + "loss": 0.9551, + "step": 16855 + }, + { + "epoch": 0.54, + "learning_rate": 9.221863023388424e-06, + "loss": 1.0469, + "step": 16856 + }, + { + "epoch": 0.54, + "learning_rate": 9.220830364496499e-06, + "loss": 0.9927, + "step": 16857 + }, + { + "epoch": 0.54, + "learning_rate": 9.219797713964205e-06, + "loss": 0.9448, + "step": 16858 + }, + { + "epoch": 0.54, + "learning_rate": 9.218765071802615e-06, + "loss": 1.0566, + "step": 16859 + }, + { + "epoch": 0.54, + "learning_rate": 9.217732438022813e-06, + "loss": 1.02, + "step": 16860 + }, + { + "epoch": 0.54, + "learning_rate": 9.216699812635876e-06, + "loss": 0.9604, + "step": 16861 + }, + { + "epoch": 0.54, + "learning_rate": 9.215667195652885e-06, + "loss": 0.9272, + "step": 16862 + }, + { + "epoch": 0.54, + "learning_rate": 9.214634587084918e-06, + "loss": 0.937, + "step": 16863 + }, + { + "epoch": 0.54, + "learning_rate": 9.21360198694305e-06, + "loss": 0.9053, + "step": 16864 + }, + { + "epoch": 0.54, + "learning_rate": 9.212569395238362e-06, + "loss": 0.9648, + "step": 16865 + }, + { + "epoch": 0.54, + "learning_rate": 9.211536811981934e-06, + "loss": 0.998, + "step": 16866 + }, + { + "epoch": 0.54, + "learning_rate": 9.210504237184841e-06, + "loss": 0.8311, + "step": 16867 + }, + { + "epoch": 0.54, + "learning_rate": 9.209471670858166e-06, + "loss": 1.0518, + "step": 16868 + }, + { + "epoch": 0.54, + "learning_rate": 9.208439113012984e-06, + "loss": 1.064, + "step": 16869 + }, + { + "epoch": 0.54, + "learning_rate": 9.207406563660374e-06, + "loss": 0.9717, + "step": 16870 + }, + { + "epoch": 0.54, + "learning_rate": 9.20637402281142e-06, + "loss": 0.9697, + "step": 16871 + }, + { + "epoch": 0.54, + "learning_rate": 9.205341490477188e-06, + "loss": 1.0903, + "step": 16872 + }, + { + "epoch": 0.54, + "learning_rate": 9.204308966668765e-06, + "loss": 1.0664, + "step": 16873 + }, + { + "epoch": 0.54, + "learning_rate": 9.203276451397224e-06, + "loss": 0.9722, + "step": 16874 + }, + { + "epoch": 0.54, + "learning_rate": 9.202243944673646e-06, + "loss": 1.0537, + "step": 16875 + }, + { + "epoch": 0.54, + "learning_rate": 9.201211446509106e-06, + "loss": 0.9268, + "step": 16876 + }, + { + "epoch": 0.54, + "learning_rate": 9.200178956914685e-06, + "loss": 0.8521, + "step": 16877 + }, + { + "epoch": 0.54, + "learning_rate": 9.199146475901462e-06, + "loss": 0.9888, + "step": 16878 + }, + { + "epoch": 0.54, + "learning_rate": 9.198114003480507e-06, + "loss": 0.9541, + "step": 16879 + }, + { + "epoch": 0.54, + "learning_rate": 9.1970815396629e-06, + "loss": 0.9951, + "step": 16880 + }, + { + "epoch": 0.54, + "learning_rate": 9.196049084459722e-06, + "loss": 0.7913, + "step": 16881 + }, + { + "epoch": 0.54, + "learning_rate": 9.195016637882045e-06, + "loss": 0.936, + "step": 16882 + }, + { + "epoch": 0.54, + "learning_rate": 9.193984199940951e-06, + "loss": 0.9224, + "step": 16883 + }, + { + "epoch": 0.54, + "learning_rate": 9.192951770647515e-06, + "loss": 0.9663, + "step": 16884 + }, + { + "epoch": 0.54, + "learning_rate": 9.19191935001281e-06, + "loss": 1.0425, + "step": 16885 + }, + { + "epoch": 0.54, + "learning_rate": 9.190886938047922e-06, + "loss": 1.0933, + "step": 16886 + }, + { + "epoch": 0.54, + "learning_rate": 9.189854534763916e-06, + "loss": 0.9053, + "step": 16887 + }, + { + "epoch": 0.54, + "learning_rate": 9.188822140171875e-06, + "loss": 1.019, + "step": 16888 + }, + { + "epoch": 0.54, + "learning_rate": 9.187789754282875e-06, + "loss": 1.043, + "step": 16889 + }, + { + "epoch": 0.54, + "learning_rate": 9.186757377107995e-06, + "loss": 0.9365, + "step": 16890 + }, + { + "epoch": 0.54, + "learning_rate": 9.185725008658306e-06, + "loss": 0.9243, + "step": 16891 + }, + { + "epoch": 0.54, + "learning_rate": 9.184692648944886e-06, + "loss": 1.0396, + "step": 16892 + }, + { + "epoch": 0.54, + "learning_rate": 9.183660297978812e-06, + "loss": 0.9424, + "step": 16893 + }, + { + "epoch": 0.54, + "learning_rate": 9.182627955771159e-06, + "loss": 1.0737, + "step": 16894 + }, + { + "epoch": 0.54, + "learning_rate": 9.181595622333006e-06, + "loss": 1.0391, + "step": 16895 + }, + { + "epoch": 0.54, + "learning_rate": 9.180563297675423e-06, + "loss": 1.0229, + "step": 16896 + }, + { + "epoch": 0.54, + "learning_rate": 9.17953098180949e-06, + "loss": 0.9741, + "step": 16897 + }, + { + "epoch": 0.54, + "learning_rate": 9.178498674746281e-06, + "loss": 0.9961, + "step": 16898 + }, + { + "epoch": 0.54, + "learning_rate": 9.177466376496872e-06, + "loss": 1.084, + "step": 16899 + }, + { + "epoch": 0.54, + "learning_rate": 9.17643408707234e-06, + "loss": 0.9082, + "step": 16900 + }, + { + "epoch": 0.54, + "learning_rate": 9.17540180648376e-06, + "loss": 0.8984, + "step": 16901 + }, + { + "epoch": 0.54, + "learning_rate": 9.174369534742203e-06, + "loss": 1.0483, + "step": 16902 + }, + { + "epoch": 0.54, + "learning_rate": 9.173337271858748e-06, + "loss": 0.9248, + "step": 16903 + }, + { + "epoch": 0.54, + "learning_rate": 9.172305017844469e-06, + "loss": 0.8862, + "step": 16904 + }, + { + "epoch": 0.54, + "learning_rate": 9.17127277271044e-06, + "loss": 1.1025, + "step": 16905 + }, + { + "epoch": 0.54, + "learning_rate": 9.170240536467738e-06, + "loss": 0.9966, + "step": 16906 + }, + { + "epoch": 0.54, + "learning_rate": 9.169208309127436e-06, + "loss": 0.9238, + "step": 16907 + }, + { + "epoch": 0.54, + "learning_rate": 9.168176090700614e-06, + "loss": 0.9067, + "step": 16908 + }, + { + "epoch": 0.54, + "learning_rate": 9.167143881198336e-06, + "loss": 0.853, + "step": 16909 + }, + { + "epoch": 0.54, + "learning_rate": 9.166111680631685e-06, + "loss": 0.9863, + "step": 16910 + }, + { + "epoch": 0.54, + "learning_rate": 9.165079489011732e-06, + "loss": 0.8252, + "step": 16911 + }, + { + "epoch": 0.54, + "learning_rate": 9.164047306349551e-06, + "loss": 0.9619, + "step": 16912 + }, + { + "epoch": 0.54, + "learning_rate": 9.163015132656218e-06, + "loss": 0.9883, + "step": 16913 + }, + { + "epoch": 0.54, + "learning_rate": 9.161982967942806e-06, + "loss": 1.0215, + "step": 16914 + }, + { + "epoch": 0.54, + "learning_rate": 9.16095081222039e-06, + "loss": 1.0215, + "step": 16915 + }, + { + "epoch": 0.54, + "learning_rate": 9.159918665500047e-06, + "loss": 1.0474, + "step": 16916 + }, + { + "epoch": 0.54, + "learning_rate": 9.158886527792843e-06, + "loss": 0.9795, + "step": 16917 + }, + { + "epoch": 0.54, + "learning_rate": 9.157854399109856e-06, + "loss": 0.9238, + "step": 16918 + }, + { + "epoch": 0.54, + "learning_rate": 9.156822279462158e-06, + "loss": 0.9893, + "step": 16919 + }, + { + "epoch": 0.54, + "learning_rate": 9.155790168860825e-06, + "loss": 0.9893, + "step": 16920 + }, + { + "epoch": 0.54, + "learning_rate": 9.154758067316928e-06, + "loss": 1.0625, + "step": 16921 + }, + { + "epoch": 0.54, + "learning_rate": 9.153725974841542e-06, + "loss": 1.0127, + "step": 16922 + }, + { + "epoch": 0.54, + "learning_rate": 9.15269389144574e-06, + "loss": 0.8447, + "step": 16923 + }, + { + "epoch": 0.54, + "learning_rate": 9.151661817140598e-06, + "loss": 0.917, + "step": 16924 + }, + { + "epoch": 0.54, + "learning_rate": 9.150629751937183e-06, + "loss": 0.9595, + "step": 16925 + }, + { + "epoch": 0.54, + "learning_rate": 9.14959769584657e-06, + "loss": 0.8633, + "step": 16926 + }, + { + "epoch": 0.54, + "learning_rate": 9.148565648879832e-06, + "loss": 1.0156, + "step": 16927 + }, + { + "epoch": 0.54, + "learning_rate": 9.147533611048042e-06, + "loss": 0.9224, + "step": 16928 + }, + { + "epoch": 0.54, + "learning_rate": 9.146501582362277e-06, + "loss": 1.127, + "step": 16929 + }, + { + "epoch": 0.54, + "learning_rate": 9.145469562833603e-06, + "loss": 0.9077, + "step": 16930 + }, + { + "epoch": 0.54, + "learning_rate": 9.144437552473096e-06, + "loss": 0.9922, + "step": 16931 + }, + { + "epoch": 0.54, + "learning_rate": 9.143405551291824e-06, + "loss": 1.0137, + "step": 16932 + }, + { + "epoch": 0.54, + "learning_rate": 9.142373559300865e-06, + "loss": 1.0796, + "step": 16933 + }, + { + "epoch": 0.54, + "learning_rate": 9.141341576511289e-06, + "loss": 1.0044, + "step": 16934 + }, + { + "epoch": 0.54, + "learning_rate": 9.140309602934165e-06, + "loss": 1.0688, + "step": 16935 + }, + { + "epoch": 0.54, + "learning_rate": 9.139277638580568e-06, + "loss": 1.1328, + "step": 16936 + }, + { + "epoch": 0.54, + "learning_rate": 9.13824568346157e-06, + "loss": 0.9692, + "step": 16937 + }, + { + "epoch": 0.54, + "learning_rate": 9.13721373758824e-06, + "loss": 1.0093, + "step": 16938 + }, + { + "epoch": 0.54, + "learning_rate": 9.136181800971657e-06, + "loss": 0.9834, + "step": 16939 + }, + { + "epoch": 0.54, + "learning_rate": 9.135149873622884e-06, + "loss": 0.9814, + "step": 16940 + }, + { + "epoch": 0.54, + "learning_rate": 9.134117955552997e-06, + "loss": 0.9736, + "step": 16941 + }, + { + "epoch": 0.54, + "learning_rate": 9.133086046773064e-06, + "loss": 1.0088, + "step": 16942 + }, + { + "epoch": 0.54, + "learning_rate": 9.132054147294158e-06, + "loss": 0.9487, + "step": 16943 + }, + { + "epoch": 0.54, + "learning_rate": 9.13102225712735e-06, + "loss": 0.9204, + "step": 16944 + }, + { + "epoch": 0.54, + "learning_rate": 9.129990376283714e-06, + "loss": 0.9482, + "step": 16945 + }, + { + "epoch": 0.54, + "learning_rate": 9.128958504774322e-06, + "loss": 0.9573, + "step": 16946 + }, + { + "epoch": 0.54, + "learning_rate": 9.127926642610235e-06, + "loss": 0.9849, + "step": 16947 + }, + { + "epoch": 0.54, + "learning_rate": 9.12689478980253e-06, + "loss": 1.0098, + "step": 16948 + }, + { + "epoch": 0.54, + "learning_rate": 9.12586294636228e-06, + "loss": 1.0288, + "step": 16949 + }, + { + "epoch": 0.54, + "learning_rate": 9.12483111230055e-06, + "loss": 1.0088, + "step": 16950 + }, + { + "epoch": 0.54, + "learning_rate": 9.123799287628416e-06, + "loss": 0.8965, + "step": 16951 + }, + { + "epoch": 0.54, + "learning_rate": 9.122767472356945e-06, + "loss": 0.9653, + "step": 16952 + }, + { + "epoch": 0.54, + "learning_rate": 9.121735666497207e-06, + "loss": 1.0327, + "step": 16953 + }, + { + "epoch": 0.54, + "learning_rate": 9.12070387006028e-06, + "loss": 0.9819, + "step": 16954 + }, + { + "epoch": 0.54, + "learning_rate": 9.119672083057222e-06, + "loss": 0.9399, + "step": 16955 + }, + { + "epoch": 0.54, + "learning_rate": 9.118640305499108e-06, + "loss": 0.9341, + "step": 16956 + }, + { + "epoch": 0.54, + "learning_rate": 9.117608537397008e-06, + "loss": 0.5046, + "step": 16957 + }, + { + "epoch": 0.54, + "learning_rate": 9.116576778761993e-06, + "loss": 1.0264, + "step": 16958 + }, + { + "epoch": 0.54, + "learning_rate": 9.115545029605129e-06, + "loss": 0.9829, + "step": 16959 + }, + { + "epoch": 0.54, + "learning_rate": 9.11451328993749e-06, + "loss": 0.9761, + "step": 16960 + }, + { + "epoch": 0.54, + "learning_rate": 9.113481559770148e-06, + "loss": 0.9038, + "step": 16961 + }, + { + "epoch": 0.54, + "learning_rate": 9.112449839114162e-06, + "loss": 0.8931, + "step": 16962 + }, + { + "epoch": 0.54, + "learning_rate": 9.111418127980608e-06, + "loss": 1.0439, + "step": 16963 + }, + { + "epoch": 0.54, + "learning_rate": 9.110386426380553e-06, + "loss": 1.0239, + "step": 16964 + }, + { + "epoch": 0.54, + "learning_rate": 9.109354734325068e-06, + "loss": 1.1265, + "step": 16965 + }, + { + "epoch": 0.54, + "learning_rate": 9.108323051825221e-06, + "loss": 1.0562, + "step": 16966 + }, + { + "epoch": 0.54, + "learning_rate": 9.10729137889208e-06, + "loss": 0.5857, + "step": 16967 + }, + { + "epoch": 0.54, + "learning_rate": 9.106259715536717e-06, + "loss": 0.9404, + "step": 16968 + }, + { + "epoch": 0.54, + "learning_rate": 9.105228061770198e-06, + "loss": 0.98, + "step": 16969 + }, + { + "epoch": 0.54, + "learning_rate": 9.10419641760359e-06, + "loss": 0.9136, + "step": 16970 + }, + { + "epoch": 0.54, + "learning_rate": 9.103164783047963e-06, + "loss": 1.002, + "step": 16971 + }, + { + "epoch": 0.54, + "learning_rate": 9.102133158114384e-06, + "loss": 0.9741, + "step": 16972 + }, + { + "epoch": 0.54, + "learning_rate": 9.101101542813927e-06, + "loss": 0.9761, + "step": 16973 + }, + { + "epoch": 0.54, + "learning_rate": 9.10006993715765e-06, + "loss": 0.9951, + "step": 16974 + }, + { + "epoch": 0.54, + "learning_rate": 9.09903834115663e-06, + "loss": 1.0483, + "step": 16975 + }, + { + "epoch": 0.54, + "learning_rate": 9.098006754821932e-06, + "loss": 1.0396, + "step": 16976 + }, + { + "epoch": 0.54, + "learning_rate": 9.096975178164622e-06, + "loss": 0.5308, + "step": 16977 + }, + { + "epoch": 0.54, + "learning_rate": 9.095943611195767e-06, + "loss": 0.8843, + "step": 16978 + }, + { + "epoch": 0.54, + "learning_rate": 9.09491205392644e-06, + "loss": 0.8862, + "step": 16979 + }, + { + "epoch": 0.54, + "learning_rate": 9.093880506367703e-06, + "loss": 0.9741, + "step": 16980 + }, + { + "epoch": 0.54, + "learning_rate": 9.092848968530625e-06, + "loss": 0.9819, + "step": 16981 + }, + { + "epoch": 0.54, + "learning_rate": 9.091817440426273e-06, + "loss": 0.9463, + "step": 16982 + }, + { + "epoch": 0.54, + "learning_rate": 9.090785922065716e-06, + "loss": 0.9585, + "step": 16983 + }, + { + "epoch": 0.54, + "learning_rate": 9.089754413460024e-06, + "loss": 0.5149, + "step": 16984 + }, + { + "epoch": 0.54, + "learning_rate": 9.088722914620254e-06, + "loss": 0.8457, + "step": 16985 + }, + { + "epoch": 0.54, + "learning_rate": 9.08769142555748e-06, + "loss": 1.062, + "step": 16986 + }, + { + "epoch": 0.54, + "learning_rate": 9.086659946282766e-06, + "loss": 0.9326, + "step": 16987 + }, + { + "epoch": 0.54, + "learning_rate": 9.085628476807182e-06, + "loss": 0.981, + "step": 16988 + }, + { + "epoch": 0.54, + "learning_rate": 9.08459701714179e-06, + "loss": 0.5232, + "step": 16989 + }, + { + "epoch": 0.54, + "learning_rate": 9.08356556729766e-06, + "loss": 1.0386, + "step": 16990 + }, + { + "epoch": 0.54, + "learning_rate": 9.082534127285861e-06, + "loss": 1.0059, + "step": 16991 + }, + { + "epoch": 0.54, + "learning_rate": 9.081502697117453e-06, + "loss": 0.873, + "step": 16992 + }, + { + "epoch": 0.54, + "learning_rate": 9.080471276803502e-06, + "loss": 1.0918, + "step": 16993 + }, + { + "epoch": 0.54, + "learning_rate": 9.079439866355078e-06, + "loss": 0.9438, + "step": 16994 + }, + { + "epoch": 0.54, + "learning_rate": 9.078408465783244e-06, + "loss": 1.0522, + "step": 16995 + }, + { + "epoch": 0.54, + "learning_rate": 9.07737707509907e-06, + "loss": 1.019, + "step": 16996 + }, + { + "epoch": 0.54, + "learning_rate": 9.076345694313618e-06, + "loss": 0.998, + "step": 16997 + }, + { + "epoch": 0.54, + "learning_rate": 9.075314323437953e-06, + "loss": 1.0181, + "step": 16998 + }, + { + "epoch": 0.54, + "learning_rate": 9.074282962483146e-06, + "loss": 0.978, + "step": 16999 + }, + { + "epoch": 0.54, + "learning_rate": 9.073251611460255e-06, + "loss": 0.9238, + "step": 17000 + }, + { + "epoch": 0.54, + "learning_rate": 9.07222027038035e-06, + "loss": 0.9604, + "step": 17001 + }, + { + "epoch": 0.54, + "learning_rate": 9.071188939254492e-06, + "loss": 1.0093, + "step": 17002 + }, + { + "epoch": 0.54, + "learning_rate": 9.070157618093749e-06, + "loss": 0.9878, + "step": 17003 + }, + { + "epoch": 0.54, + "learning_rate": 9.069126306909187e-06, + "loss": 1.041, + "step": 17004 + }, + { + "epoch": 0.54, + "learning_rate": 9.068095005711869e-06, + "loss": 0.9526, + "step": 17005 + }, + { + "epoch": 0.54, + "learning_rate": 9.06706371451286e-06, + "loss": 0.9326, + "step": 17006 + }, + { + "epoch": 0.54, + "learning_rate": 9.066032433323228e-06, + "loss": 0.9326, + "step": 17007 + }, + { + "epoch": 0.54, + "learning_rate": 9.065001162154032e-06, + "loss": 0.9678, + "step": 17008 + }, + { + "epoch": 0.54, + "learning_rate": 9.06396990101634e-06, + "loss": 1.0488, + "step": 17009 + }, + { + "epoch": 0.54, + "learning_rate": 9.062938649921211e-06, + "loss": 1.0093, + "step": 17010 + }, + { + "epoch": 0.54, + "learning_rate": 9.061907408879715e-06, + "loss": 0.9995, + "step": 17011 + }, + { + "epoch": 0.54, + "learning_rate": 9.060876177902915e-06, + "loss": 0.8936, + "step": 17012 + }, + { + "epoch": 0.54, + "learning_rate": 9.059844957001876e-06, + "loss": 1.0547, + "step": 17013 + }, + { + "epoch": 0.54, + "learning_rate": 9.05881374618766e-06, + "loss": 0.9341, + "step": 17014 + }, + { + "epoch": 0.54, + "learning_rate": 9.05778254547133e-06, + "loss": 0.8906, + "step": 17015 + }, + { + "epoch": 0.54, + "learning_rate": 9.056751354863947e-06, + "loss": 0.9023, + "step": 17016 + }, + { + "epoch": 0.54, + "learning_rate": 9.05572017437658e-06, + "loss": 1.0166, + "step": 17017 + }, + { + "epoch": 0.54, + "learning_rate": 9.054689004020294e-06, + "loss": 0.9302, + "step": 17018 + }, + { + "epoch": 0.54, + "learning_rate": 9.053657843806145e-06, + "loss": 0.9253, + "step": 17019 + }, + { + "epoch": 0.54, + "learning_rate": 9.0526266937452e-06, + "loss": 0.9233, + "step": 17020 + }, + { + "epoch": 0.54, + "learning_rate": 9.051595553848523e-06, + "loss": 0.9604, + "step": 17021 + }, + { + "epoch": 0.54, + "learning_rate": 9.050564424127177e-06, + "loss": 0.8799, + "step": 17022 + }, + { + "epoch": 0.54, + "learning_rate": 9.049533304592225e-06, + "loss": 1.0171, + "step": 17023 + }, + { + "epoch": 0.54, + "learning_rate": 9.048502195254727e-06, + "loss": 0.9502, + "step": 17024 + }, + { + "epoch": 0.54, + "learning_rate": 9.047471096125746e-06, + "loss": 0.8633, + "step": 17025 + }, + { + "epoch": 0.54, + "learning_rate": 9.046440007216346e-06, + "loss": 0.9741, + "step": 17026 + }, + { + "epoch": 0.54, + "learning_rate": 9.04540892853759e-06, + "loss": 0.8564, + "step": 17027 + }, + { + "epoch": 0.54, + "learning_rate": 9.04437786010054e-06, + "loss": 0.9746, + "step": 17028 + }, + { + "epoch": 0.54, + "learning_rate": 9.043346801916262e-06, + "loss": 1.0586, + "step": 17029 + }, + { + "epoch": 0.54, + "learning_rate": 9.04231575399581e-06, + "loss": 1.0166, + "step": 17030 + }, + { + "epoch": 0.54, + "learning_rate": 9.04128471635025e-06, + "loss": 1.1187, + "step": 17031 + }, + { + "epoch": 0.54, + "learning_rate": 9.040253688990642e-06, + "loss": 1.0269, + "step": 17032 + }, + { + "epoch": 0.54, + "learning_rate": 9.039222671928052e-06, + "loss": 0.9336, + "step": 17033 + }, + { + "epoch": 0.54, + "learning_rate": 9.038191665173538e-06, + "loss": 1.0205, + "step": 17034 + }, + { + "epoch": 0.54, + "learning_rate": 9.037160668738164e-06, + "loss": 0.9805, + "step": 17035 + }, + { + "epoch": 0.54, + "learning_rate": 9.03612968263299e-06, + "loss": 0.9092, + "step": 17036 + }, + { + "epoch": 0.54, + "learning_rate": 9.035098706869082e-06, + "loss": 1.1787, + "step": 17037 + }, + { + "epoch": 0.54, + "learning_rate": 9.034067741457492e-06, + "loss": 1.0439, + "step": 17038 + }, + { + "epoch": 0.54, + "learning_rate": 9.033036786409288e-06, + "loss": 1.0098, + "step": 17039 + }, + { + "epoch": 0.54, + "learning_rate": 9.032005841735527e-06, + "loss": 1.0166, + "step": 17040 + }, + { + "epoch": 0.54, + "learning_rate": 9.030974907447273e-06, + "loss": 1.0854, + "step": 17041 + }, + { + "epoch": 0.55, + "learning_rate": 9.029943983555586e-06, + "loss": 1.043, + "step": 17042 + }, + { + "epoch": 0.55, + "learning_rate": 9.028913070071527e-06, + "loss": 0.9438, + "step": 17043 + }, + { + "epoch": 0.55, + "learning_rate": 9.027882167006159e-06, + "loss": 0.897, + "step": 17044 + }, + { + "epoch": 0.55, + "learning_rate": 9.026851274370535e-06, + "loss": 1.0088, + "step": 17045 + }, + { + "epoch": 0.55, + "learning_rate": 9.025820392175722e-06, + "loss": 1.0239, + "step": 17046 + }, + { + "epoch": 0.55, + "learning_rate": 9.024789520432776e-06, + "loss": 0.8838, + "step": 17047 + }, + { + "epoch": 0.55, + "learning_rate": 9.023758659152759e-06, + "loss": 0.9189, + "step": 17048 + }, + { + "epoch": 0.55, + "learning_rate": 9.022727808346731e-06, + "loss": 0.9204, + "step": 17049 + }, + { + "epoch": 0.55, + "learning_rate": 9.021696968025754e-06, + "loss": 1.0596, + "step": 17050 + }, + { + "epoch": 0.55, + "learning_rate": 9.020666138200885e-06, + "loss": 1.0317, + "step": 17051 + }, + { + "epoch": 0.55, + "learning_rate": 9.019635318883188e-06, + "loss": 1.0498, + "step": 17052 + }, + { + "epoch": 0.55, + "learning_rate": 9.018604510083715e-06, + "loss": 1.1626, + "step": 17053 + }, + { + "epoch": 0.55, + "learning_rate": 9.017573711813528e-06, + "loss": 0.9136, + "step": 17054 + }, + { + "epoch": 0.55, + "learning_rate": 9.01654292408369e-06, + "loss": 0.937, + "step": 17055 + }, + { + "epoch": 0.55, + "learning_rate": 9.015512146905257e-06, + "loss": 1.0303, + "step": 17056 + }, + { + "epoch": 0.55, + "learning_rate": 9.014481380289292e-06, + "loss": 0.8896, + "step": 17057 + }, + { + "epoch": 0.55, + "learning_rate": 9.013450624246847e-06, + "loss": 0.9888, + "step": 17058 + }, + { + "epoch": 0.55, + "learning_rate": 9.012419878788989e-06, + "loss": 0.8999, + "step": 17059 + }, + { + "epoch": 0.55, + "learning_rate": 9.011389143926771e-06, + "loss": 0.8975, + "step": 17060 + }, + { + "epoch": 0.55, + "learning_rate": 9.010358419671252e-06, + "loss": 0.9829, + "step": 17061 + }, + { + "epoch": 0.55, + "learning_rate": 9.009327706033496e-06, + "loss": 1.0396, + "step": 17062 + }, + { + "epoch": 0.55, + "learning_rate": 9.008297003024551e-06, + "loss": 0.9175, + "step": 17063 + }, + { + "epoch": 0.55, + "learning_rate": 9.007266310655486e-06, + "loss": 0.9541, + "step": 17064 + }, + { + "epoch": 0.55, + "learning_rate": 9.006235628937352e-06, + "loss": 1.0566, + "step": 17065 + }, + { + "epoch": 0.55, + "learning_rate": 9.00520495788121e-06, + "loss": 1.0781, + "step": 17066 + }, + { + "epoch": 0.55, + "learning_rate": 9.004174297498122e-06, + "loss": 0.9268, + "step": 17067 + }, + { + "epoch": 0.55, + "learning_rate": 9.00314364779914e-06, + "loss": 1.0493, + "step": 17068 + }, + { + "epoch": 0.55, + "learning_rate": 9.002113008795323e-06, + "loss": 0.9722, + "step": 17069 + }, + { + "epoch": 0.55, + "learning_rate": 9.001082380497727e-06, + "loss": 0.9126, + "step": 17070 + }, + { + "epoch": 0.55, + "learning_rate": 9.000051762917412e-06, + "loss": 0.8477, + "step": 17071 + }, + { + "epoch": 0.55, + "learning_rate": 8.999021156065435e-06, + "loss": 1.0967, + "step": 17072 + }, + { + "epoch": 0.55, + "learning_rate": 8.997990559952855e-06, + "loss": 1.0166, + "step": 17073 + }, + { + "epoch": 0.55, + "learning_rate": 8.996959974590729e-06, + "loss": 0.9419, + "step": 17074 + }, + { + "epoch": 0.55, + "learning_rate": 8.995929399990109e-06, + "loss": 0.5125, + "step": 17075 + }, + { + "epoch": 0.55, + "learning_rate": 8.994898836162056e-06, + "loss": 1.0234, + "step": 17076 + }, + { + "epoch": 0.55, + "learning_rate": 8.993868283117623e-06, + "loss": 1.0352, + "step": 17077 + }, + { + "epoch": 0.55, + "learning_rate": 8.992837740867873e-06, + "loss": 1.0107, + "step": 17078 + }, + { + "epoch": 0.55, + "learning_rate": 8.991807209423858e-06, + "loss": 0.8145, + "step": 17079 + }, + { + "epoch": 0.55, + "learning_rate": 8.990776688796637e-06, + "loss": 0.9253, + "step": 17080 + }, + { + "epoch": 0.55, + "learning_rate": 8.989746178997265e-06, + "loss": 0.9175, + "step": 17081 + }, + { + "epoch": 0.55, + "learning_rate": 8.988715680036802e-06, + "loss": 1.0386, + "step": 17082 + }, + { + "epoch": 0.55, + "learning_rate": 8.987685191926296e-06, + "loss": 1.0171, + "step": 17083 + }, + { + "epoch": 0.55, + "learning_rate": 8.986654714676808e-06, + "loss": 0.9692, + "step": 17084 + }, + { + "epoch": 0.55, + "learning_rate": 8.985624248299392e-06, + "loss": 1.1201, + "step": 17085 + }, + { + "epoch": 0.55, + "learning_rate": 8.984593792805108e-06, + "loss": 1.1152, + "step": 17086 + }, + { + "epoch": 0.55, + "learning_rate": 8.983563348205006e-06, + "loss": 1.0269, + "step": 17087 + }, + { + "epoch": 0.55, + "learning_rate": 8.982532914510146e-06, + "loss": 0.9043, + "step": 17088 + }, + { + "epoch": 0.55, + "learning_rate": 8.981502491731582e-06, + "loss": 0.9761, + "step": 17089 + }, + { + "epoch": 0.55, + "learning_rate": 8.980472079880372e-06, + "loss": 1.0728, + "step": 17090 + }, + { + "epoch": 0.55, + "learning_rate": 8.979441678967565e-06, + "loss": 0.9556, + "step": 17091 + }, + { + "epoch": 0.55, + "learning_rate": 8.97841128900422e-06, + "loss": 0.9116, + "step": 17092 + }, + { + "epoch": 0.55, + "learning_rate": 8.97738091000139e-06, + "loss": 0.9604, + "step": 17093 + }, + { + "epoch": 0.55, + "learning_rate": 8.97635054197013e-06, + "loss": 0.9341, + "step": 17094 + }, + { + "epoch": 0.55, + "learning_rate": 8.975320184921497e-06, + "loss": 0.9746, + "step": 17095 + }, + { + "epoch": 0.55, + "learning_rate": 8.974289838866546e-06, + "loss": 0.9746, + "step": 17096 + }, + { + "epoch": 0.55, + "learning_rate": 8.973259503816333e-06, + "loss": 1.0464, + "step": 17097 + }, + { + "epoch": 0.55, + "learning_rate": 8.972229179781902e-06, + "loss": 1.0259, + "step": 17098 + }, + { + "epoch": 0.55, + "learning_rate": 8.971198866774318e-06, + "loss": 0.9106, + "step": 17099 + }, + { + "epoch": 0.55, + "learning_rate": 8.97016856480463e-06, + "loss": 0.8706, + "step": 17100 + }, + { + "epoch": 0.55, + "learning_rate": 8.969138273883896e-06, + "loss": 0.8311, + "step": 17101 + }, + { + "epoch": 0.55, + "learning_rate": 8.968107994023166e-06, + "loss": 1.0693, + "step": 17102 + }, + { + "epoch": 0.55, + "learning_rate": 8.967077725233494e-06, + "loss": 1.0527, + "step": 17103 + }, + { + "epoch": 0.55, + "learning_rate": 8.966047467525934e-06, + "loss": 0.4834, + "step": 17104 + }, + { + "epoch": 0.55, + "learning_rate": 8.965017220911545e-06, + "loss": 1.0703, + "step": 17105 + }, + { + "epoch": 0.55, + "learning_rate": 8.963986985401372e-06, + "loss": 1.0801, + "step": 17106 + }, + { + "epoch": 0.55, + "learning_rate": 8.962956761006475e-06, + "loss": 1.0977, + "step": 17107 + }, + { + "epoch": 0.55, + "learning_rate": 8.961926547737902e-06, + "loss": 0.9712, + "step": 17108 + }, + { + "epoch": 0.55, + "learning_rate": 8.960896345606708e-06, + "loss": 1.0488, + "step": 17109 + }, + { + "epoch": 0.55, + "learning_rate": 8.959866154623946e-06, + "loss": 0.9946, + "step": 17110 + }, + { + "epoch": 0.55, + "learning_rate": 8.958835974800668e-06, + "loss": 0.8369, + "step": 17111 + }, + { + "epoch": 0.55, + "learning_rate": 8.957805806147934e-06, + "loss": 1.0249, + "step": 17112 + }, + { + "epoch": 0.55, + "learning_rate": 8.956775648676786e-06, + "loss": 0.9321, + "step": 17113 + }, + { + "epoch": 0.55, + "learning_rate": 8.955745502398278e-06, + "loss": 0.9243, + "step": 17114 + }, + { + "epoch": 0.55, + "learning_rate": 8.954715367323468e-06, + "loss": 0.9756, + "step": 17115 + }, + { + "epoch": 0.55, + "learning_rate": 8.953685243463404e-06, + "loss": 1.0088, + "step": 17116 + }, + { + "epoch": 0.55, + "learning_rate": 8.952655130829138e-06, + "loss": 1.0234, + "step": 17117 + }, + { + "epoch": 0.55, + "learning_rate": 8.951625029431726e-06, + "loss": 0.9858, + "step": 17118 + }, + { + "epoch": 0.55, + "learning_rate": 8.950594939282215e-06, + "loss": 0.9258, + "step": 17119 + }, + { + "epoch": 0.55, + "learning_rate": 8.949564860391664e-06, + "loss": 0.9805, + "step": 17120 + }, + { + "epoch": 0.55, + "learning_rate": 8.948534792771116e-06, + "loss": 0.9429, + "step": 17121 + }, + { + "epoch": 0.55, + "learning_rate": 8.947504736431625e-06, + "loss": 0.9429, + "step": 17122 + }, + { + "epoch": 0.55, + "learning_rate": 8.946474691384243e-06, + "loss": 0.9336, + "step": 17123 + }, + { + "epoch": 0.55, + "learning_rate": 8.945444657640022e-06, + "loss": 1.0337, + "step": 17124 + }, + { + "epoch": 0.55, + "learning_rate": 8.944414635210013e-06, + "loss": 0.4873, + "step": 17125 + }, + { + "epoch": 0.55, + "learning_rate": 8.943384624105267e-06, + "loss": 0.9927, + "step": 17126 + }, + { + "epoch": 0.55, + "learning_rate": 8.942354624336837e-06, + "loss": 0.8713, + "step": 17127 + }, + { + "epoch": 0.55, + "learning_rate": 8.94132463591577e-06, + "loss": 1.0522, + "step": 17128 + }, + { + "epoch": 0.55, + "learning_rate": 8.940294658853116e-06, + "loss": 0.9937, + "step": 17129 + }, + { + "epoch": 0.55, + "learning_rate": 8.939264693159926e-06, + "loss": 0.9951, + "step": 17130 + }, + { + "epoch": 0.55, + "learning_rate": 8.938234738847256e-06, + "loss": 1.0088, + "step": 17131 + }, + { + "epoch": 0.55, + "learning_rate": 8.93720479592615e-06, + "loss": 1.0293, + "step": 17132 + }, + { + "epoch": 0.55, + "learning_rate": 8.93617486440766e-06, + "loss": 0.9692, + "step": 17133 + }, + { + "epoch": 0.55, + "learning_rate": 8.935144944302835e-06, + "loss": 0.9956, + "step": 17134 + }, + { + "epoch": 0.55, + "learning_rate": 8.934115035622732e-06, + "loss": 1.0625, + "step": 17135 + }, + { + "epoch": 0.55, + "learning_rate": 8.933085138378392e-06, + "loss": 0.9756, + "step": 17136 + }, + { + "epoch": 0.55, + "learning_rate": 8.932055252580865e-06, + "loss": 0.9785, + "step": 17137 + }, + { + "epoch": 0.55, + "learning_rate": 8.931025378241205e-06, + "loss": 0.9668, + "step": 17138 + }, + { + "epoch": 0.55, + "learning_rate": 8.92999551537046e-06, + "loss": 0.9678, + "step": 17139 + }, + { + "epoch": 0.55, + "learning_rate": 8.928965663979677e-06, + "loss": 1.0454, + "step": 17140 + }, + { + "epoch": 0.55, + "learning_rate": 8.92793582407991e-06, + "loss": 1.0098, + "step": 17141 + }, + { + "epoch": 0.55, + "learning_rate": 8.926905995682206e-06, + "loss": 0.9966, + "step": 17142 + }, + { + "epoch": 0.55, + "learning_rate": 8.92587617879761e-06, + "loss": 1.0469, + "step": 17143 + }, + { + "epoch": 0.55, + "learning_rate": 8.924846373437174e-06, + "loss": 1.0308, + "step": 17144 + }, + { + "epoch": 0.55, + "learning_rate": 8.923816579611947e-06, + "loss": 0.9902, + "step": 17145 + }, + { + "epoch": 0.55, + "learning_rate": 8.92278679733298e-06, + "loss": 0.9678, + "step": 17146 + }, + { + "epoch": 0.55, + "learning_rate": 8.921757026611314e-06, + "loss": 1.063, + "step": 17147 + }, + { + "epoch": 0.55, + "learning_rate": 8.920727267458003e-06, + "loss": 1.1094, + "step": 17148 + }, + { + "epoch": 0.55, + "learning_rate": 8.919697519884094e-06, + "loss": 1.043, + "step": 17149 + }, + { + "epoch": 0.55, + "learning_rate": 8.918667783900637e-06, + "loss": 0.9995, + "step": 17150 + }, + { + "epoch": 0.55, + "learning_rate": 8.917638059518677e-06, + "loss": 1.0303, + "step": 17151 + }, + { + "epoch": 0.55, + "learning_rate": 8.916608346749263e-06, + "loss": 0.8389, + "step": 17152 + }, + { + "epoch": 0.55, + "learning_rate": 8.91557864560344e-06, + "loss": 0.854, + "step": 17153 + }, + { + "epoch": 0.55, + "learning_rate": 8.914548956092258e-06, + "loss": 0.9346, + "step": 17154 + }, + { + "epoch": 0.55, + "learning_rate": 8.913519278226766e-06, + "loss": 0.9658, + "step": 17155 + }, + { + "epoch": 0.55, + "learning_rate": 8.91248961201801e-06, + "loss": 0.9971, + "step": 17156 + }, + { + "epoch": 0.55, + "learning_rate": 8.911459957477035e-06, + "loss": 0.9648, + "step": 17157 + }, + { + "epoch": 0.55, + "learning_rate": 8.910430314614893e-06, + "loss": 1.0576, + "step": 17158 + }, + { + "epoch": 0.55, + "learning_rate": 8.909400683442628e-06, + "loss": 1.0732, + "step": 17159 + }, + { + "epoch": 0.55, + "learning_rate": 8.908371063971282e-06, + "loss": 1.0269, + "step": 17160 + }, + { + "epoch": 0.55, + "learning_rate": 8.90734145621191e-06, + "loss": 0.8271, + "step": 17161 + }, + { + "epoch": 0.55, + "learning_rate": 8.906311860175553e-06, + "loss": 0.9741, + "step": 17162 + }, + { + "epoch": 0.55, + "learning_rate": 8.905282275873259e-06, + "loss": 0.9702, + "step": 17163 + }, + { + "epoch": 0.55, + "learning_rate": 8.904252703316077e-06, + "loss": 0.9282, + "step": 17164 + }, + { + "epoch": 0.55, + "learning_rate": 8.903223142515051e-06, + "loss": 1.0376, + "step": 17165 + }, + { + "epoch": 0.55, + "learning_rate": 8.902193593481226e-06, + "loss": 0.9292, + "step": 17166 + }, + { + "epoch": 0.55, + "learning_rate": 8.901164056225648e-06, + "loss": 0.9326, + "step": 17167 + }, + { + "epoch": 0.55, + "learning_rate": 8.900134530759364e-06, + "loss": 0.8789, + "step": 17168 + }, + { + "epoch": 0.55, + "learning_rate": 8.899105017093418e-06, + "loss": 0.9502, + "step": 17169 + }, + { + "epoch": 0.55, + "learning_rate": 8.898075515238856e-06, + "loss": 0.9087, + "step": 17170 + }, + { + "epoch": 0.55, + "learning_rate": 8.897046025206725e-06, + "loss": 0.9634, + "step": 17171 + }, + { + "epoch": 0.55, + "learning_rate": 8.896016547008071e-06, + "loss": 0.9292, + "step": 17172 + }, + { + "epoch": 0.55, + "learning_rate": 8.89498708065394e-06, + "loss": 1.0283, + "step": 17173 + }, + { + "epoch": 0.55, + "learning_rate": 8.893957626155371e-06, + "loss": 0.9961, + "step": 17174 + }, + { + "epoch": 0.55, + "learning_rate": 8.892928183523412e-06, + "loss": 0.9976, + "step": 17175 + }, + { + "epoch": 0.55, + "learning_rate": 8.89189875276911e-06, + "loss": 0.9531, + "step": 17176 + }, + { + "epoch": 0.55, + "learning_rate": 8.890869333903507e-06, + "loss": 0.9795, + "step": 17177 + }, + { + "epoch": 0.55, + "learning_rate": 8.889839926937647e-06, + "loss": 0.9399, + "step": 17178 + }, + { + "epoch": 0.55, + "learning_rate": 8.888810531882578e-06, + "loss": 0.918, + "step": 17179 + }, + { + "epoch": 0.55, + "learning_rate": 8.887781148749346e-06, + "loss": 0.8503, + "step": 17180 + }, + { + "epoch": 0.55, + "learning_rate": 8.886751777548988e-06, + "loss": 0.9902, + "step": 17181 + }, + { + "epoch": 0.55, + "learning_rate": 8.88572241829255e-06, + "loss": 0.9946, + "step": 17182 + }, + { + "epoch": 0.55, + "learning_rate": 8.884693070991077e-06, + "loss": 0.8379, + "step": 17183 + }, + { + "epoch": 0.55, + "learning_rate": 8.883663735655612e-06, + "loss": 0.9609, + "step": 17184 + }, + { + "epoch": 0.55, + "learning_rate": 8.882634412297205e-06, + "loss": 0.9546, + "step": 17185 + }, + { + "epoch": 0.55, + "learning_rate": 8.88160510092689e-06, + "loss": 0.9844, + "step": 17186 + }, + { + "epoch": 0.55, + "learning_rate": 8.880575801555715e-06, + "loss": 1.0884, + "step": 17187 + }, + { + "epoch": 0.55, + "learning_rate": 8.879546514194725e-06, + "loss": 1.043, + "step": 17188 + }, + { + "epoch": 0.55, + "learning_rate": 8.878517238854958e-06, + "loss": 0.9409, + "step": 17189 + }, + { + "epoch": 0.55, + "learning_rate": 8.877487975547461e-06, + "loss": 0.874, + "step": 17190 + }, + { + "epoch": 0.55, + "learning_rate": 8.876458724283276e-06, + "loss": 0.9429, + "step": 17191 + }, + { + "epoch": 0.55, + "learning_rate": 8.875429485073444e-06, + "loss": 0.853, + "step": 17192 + }, + { + "epoch": 0.55, + "learning_rate": 8.87440025792901e-06, + "loss": 0.9175, + "step": 17193 + }, + { + "epoch": 0.55, + "learning_rate": 8.873371042861013e-06, + "loss": 1.0322, + "step": 17194 + }, + { + "epoch": 0.55, + "learning_rate": 8.872341839880503e-06, + "loss": 0.8921, + "step": 17195 + }, + { + "epoch": 0.55, + "learning_rate": 8.871312648998517e-06, + "loss": 0.906, + "step": 17196 + }, + { + "epoch": 0.55, + "learning_rate": 8.870283470226093e-06, + "loss": 0.9995, + "step": 17197 + }, + { + "epoch": 0.55, + "learning_rate": 8.869254303574278e-06, + "loss": 1.0083, + "step": 17198 + }, + { + "epoch": 0.55, + "learning_rate": 8.868225149054112e-06, + "loss": 0.8271, + "step": 17199 + }, + { + "epoch": 0.55, + "learning_rate": 8.867196006676637e-06, + "loss": 1.043, + "step": 17200 + }, + { + "epoch": 0.55, + "learning_rate": 8.866166876452897e-06, + "loss": 0.9243, + "step": 17201 + }, + { + "epoch": 0.55, + "learning_rate": 8.86513775839393e-06, + "loss": 1.0342, + "step": 17202 + }, + { + "epoch": 0.55, + "learning_rate": 8.864108652510785e-06, + "loss": 1.0322, + "step": 17203 + }, + { + "epoch": 0.55, + "learning_rate": 8.863079558814492e-06, + "loss": 0.7979, + "step": 17204 + }, + { + "epoch": 0.55, + "learning_rate": 8.862050477316096e-06, + "loss": 1.0044, + "step": 17205 + }, + { + "epoch": 0.55, + "learning_rate": 8.86102140802664e-06, + "loss": 1.0874, + "step": 17206 + }, + { + "epoch": 0.55, + "learning_rate": 8.859992350957163e-06, + "loss": 1.0396, + "step": 17207 + }, + { + "epoch": 0.55, + "learning_rate": 8.858963306118708e-06, + "loss": 0.9976, + "step": 17208 + }, + { + "epoch": 0.55, + "learning_rate": 8.857934273522312e-06, + "loss": 0.9912, + "step": 17209 + }, + { + "epoch": 0.55, + "learning_rate": 8.856905253179022e-06, + "loss": 1.0186, + "step": 17210 + }, + { + "epoch": 0.55, + "learning_rate": 8.85587624509987e-06, + "loss": 0.9888, + "step": 17211 + }, + { + "epoch": 0.55, + "learning_rate": 8.8548472492959e-06, + "loss": 0.9622, + "step": 17212 + }, + { + "epoch": 0.55, + "learning_rate": 8.853818265778148e-06, + "loss": 1.1099, + "step": 17213 + }, + { + "epoch": 0.55, + "learning_rate": 8.852789294557662e-06, + "loss": 0.8538, + "step": 17214 + }, + { + "epoch": 0.55, + "learning_rate": 8.851760335645475e-06, + "loss": 1.0259, + "step": 17215 + }, + { + "epoch": 0.55, + "learning_rate": 8.850731389052629e-06, + "loss": 1.0229, + "step": 17216 + }, + { + "epoch": 0.55, + "learning_rate": 8.849702454790163e-06, + "loss": 1.0459, + "step": 17217 + }, + { + "epoch": 0.55, + "learning_rate": 8.848673532869122e-06, + "loss": 0.9355, + "step": 17218 + }, + { + "epoch": 0.55, + "learning_rate": 8.847644623300534e-06, + "loss": 0.9355, + "step": 17219 + }, + { + "epoch": 0.55, + "learning_rate": 8.846615726095445e-06, + "loss": 0.9521, + "step": 17220 + }, + { + "epoch": 0.55, + "learning_rate": 8.845586841264892e-06, + "loss": 0.5239, + "step": 17221 + }, + { + "epoch": 0.55, + "learning_rate": 8.844557968819916e-06, + "loss": 0.9351, + "step": 17222 + }, + { + "epoch": 0.55, + "learning_rate": 8.843529108771553e-06, + "loss": 1.02, + "step": 17223 + }, + { + "epoch": 0.55, + "learning_rate": 8.842500261130846e-06, + "loss": 0.895, + "step": 17224 + }, + { + "epoch": 0.55, + "learning_rate": 8.841471425908829e-06, + "loss": 1.0112, + "step": 17225 + }, + { + "epoch": 0.55, + "learning_rate": 8.84044260311654e-06, + "loss": 0.9707, + "step": 17226 + }, + { + "epoch": 0.55, + "learning_rate": 8.839413792765018e-06, + "loss": 0.939, + "step": 17227 + }, + { + "epoch": 0.55, + "learning_rate": 8.8383849948653e-06, + "loss": 0.9902, + "step": 17228 + }, + { + "epoch": 0.55, + "learning_rate": 8.837356209428428e-06, + "loss": 1.0264, + "step": 17229 + }, + { + "epoch": 0.55, + "learning_rate": 8.836327436465437e-06, + "loss": 1.0879, + "step": 17230 + }, + { + "epoch": 0.55, + "learning_rate": 8.835298675987364e-06, + "loss": 0.9761, + "step": 17231 + }, + { + "epoch": 0.55, + "learning_rate": 8.834269928005246e-06, + "loss": 0.8447, + "step": 17232 + }, + { + "epoch": 0.55, + "learning_rate": 8.833241192530126e-06, + "loss": 1.0195, + "step": 17233 + }, + { + "epoch": 0.55, + "learning_rate": 8.832212469573032e-06, + "loss": 0.9282, + "step": 17234 + }, + { + "epoch": 0.55, + "learning_rate": 8.83118375914501e-06, + "loss": 0.96, + "step": 17235 + }, + { + "epoch": 0.55, + "learning_rate": 8.83015506125709e-06, + "loss": 1.0142, + "step": 17236 + }, + { + "epoch": 0.55, + "learning_rate": 8.829126375920311e-06, + "loss": 1.042, + "step": 17237 + }, + { + "epoch": 0.55, + "learning_rate": 8.82809770314571e-06, + "loss": 1.0703, + "step": 17238 + }, + { + "epoch": 0.55, + "learning_rate": 8.827069042944322e-06, + "loss": 0.9609, + "step": 17239 + }, + { + "epoch": 0.55, + "learning_rate": 8.826040395327188e-06, + "loss": 0.9424, + "step": 17240 + }, + { + "epoch": 0.55, + "learning_rate": 8.825011760305345e-06, + "loss": 0.9819, + "step": 17241 + }, + { + "epoch": 0.55, + "learning_rate": 8.82398313788982e-06, + "loss": 0.9077, + "step": 17242 + }, + { + "epoch": 0.55, + "learning_rate": 8.822954528091655e-06, + "loss": 0.9736, + "step": 17243 + }, + { + "epoch": 0.55, + "learning_rate": 8.821925930921885e-06, + "loss": 0.96, + "step": 17244 + }, + { + "epoch": 0.55, + "learning_rate": 8.820897346391546e-06, + "loss": 1.0498, + "step": 17245 + }, + { + "epoch": 0.55, + "learning_rate": 8.819868774511674e-06, + "loss": 0.9858, + "step": 17246 + }, + { + "epoch": 0.55, + "learning_rate": 8.818840215293304e-06, + "loss": 0.9204, + "step": 17247 + }, + { + "epoch": 0.55, + "learning_rate": 8.817811668747474e-06, + "loss": 0.9663, + "step": 17248 + }, + { + "epoch": 0.55, + "learning_rate": 8.816783134885215e-06, + "loss": 1.0044, + "step": 17249 + }, + { + "epoch": 0.55, + "learning_rate": 8.815754613717561e-06, + "loss": 0.9194, + "step": 17250 + }, + { + "epoch": 0.55, + "learning_rate": 8.81472610525555e-06, + "loss": 0.9961, + "step": 17251 + }, + { + "epoch": 0.55, + "learning_rate": 8.813697609510216e-06, + "loss": 1.04, + "step": 17252 + }, + { + "epoch": 0.55, + "learning_rate": 8.812669126492593e-06, + "loss": 1.0195, + "step": 17253 + }, + { + "epoch": 0.55, + "learning_rate": 8.811640656213717e-06, + "loss": 1.0869, + "step": 17254 + }, + { + "epoch": 0.55, + "learning_rate": 8.810612198684622e-06, + "loss": 1.0117, + "step": 17255 + }, + { + "epoch": 0.55, + "learning_rate": 8.809583753916345e-06, + "loss": 0.9727, + "step": 17256 + }, + { + "epoch": 0.55, + "learning_rate": 8.808555321919914e-06, + "loss": 0.5037, + "step": 17257 + }, + { + "epoch": 0.55, + "learning_rate": 8.807526902706364e-06, + "loss": 0.9629, + "step": 17258 + }, + { + "epoch": 0.55, + "learning_rate": 8.806498496286732e-06, + "loss": 0.9438, + "step": 17259 + }, + { + "epoch": 0.55, + "learning_rate": 8.805470102672049e-06, + "loss": 1.0176, + "step": 17260 + }, + { + "epoch": 0.55, + "learning_rate": 8.804441721873352e-06, + "loss": 1.0273, + "step": 17261 + }, + { + "epoch": 0.55, + "learning_rate": 8.80341335390167e-06, + "loss": 0.9595, + "step": 17262 + }, + { + "epoch": 0.55, + "learning_rate": 8.802384998768043e-06, + "loss": 0.9312, + "step": 17263 + }, + { + "epoch": 0.55, + "learning_rate": 8.801356656483495e-06, + "loss": 1.0142, + "step": 17264 + }, + { + "epoch": 0.55, + "learning_rate": 8.800328327059065e-06, + "loss": 1.0439, + "step": 17265 + }, + { + "epoch": 0.55, + "learning_rate": 8.799300010505782e-06, + "loss": 1.1162, + "step": 17266 + }, + { + "epoch": 0.55, + "learning_rate": 8.798271706834684e-06, + "loss": 0.9619, + "step": 17267 + }, + { + "epoch": 0.55, + "learning_rate": 8.797243416056797e-06, + "loss": 1.0483, + "step": 17268 + }, + { + "epoch": 0.55, + "learning_rate": 8.796215138183161e-06, + "loss": 0.5127, + "step": 17269 + }, + { + "epoch": 0.55, + "learning_rate": 8.795186873224804e-06, + "loss": 1.0562, + "step": 17270 + }, + { + "epoch": 0.55, + "learning_rate": 8.794158621192757e-06, + "loss": 0.9775, + "step": 17271 + }, + { + "epoch": 0.55, + "learning_rate": 8.793130382098054e-06, + "loss": 1.1191, + "step": 17272 + }, + { + "epoch": 0.55, + "learning_rate": 8.792102155951726e-06, + "loss": 1.0718, + "step": 17273 + }, + { + "epoch": 0.55, + "learning_rate": 8.791073942764806e-06, + "loss": 0.9062, + "step": 17274 + }, + { + "epoch": 0.55, + "learning_rate": 8.790045742548323e-06, + "loss": 1.0571, + "step": 17275 + }, + { + "epoch": 0.55, + "learning_rate": 8.789017555313311e-06, + "loss": 1.0361, + "step": 17276 + }, + { + "epoch": 0.55, + "learning_rate": 8.787989381070798e-06, + "loss": 1.0254, + "step": 17277 + }, + { + "epoch": 0.55, + "learning_rate": 8.786961219831822e-06, + "loss": 0.9751, + "step": 17278 + }, + { + "epoch": 0.55, + "learning_rate": 8.785933071607408e-06, + "loss": 1.0093, + "step": 17279 + }, + { + "epoch": 0.55, + "learning_rate": 8.784904936408586e-06, + "loss": 0.9634, + "step": 17280 + }, + { + "epoch": 0.55, + "learning_rate": 8.78387681424639e-06, + "loss": 1.0127, + "step": 17281 + }, + { + "epoch": 0.55, + "learning_rate": 8.782848705131852e-06, + "loss": 1.0513, + "step": 17282 + }, + { + "epoch": 0.55, + "learning_rate": 8.781820609075998e-06, + "loss": 1.0264, + "step": 17283 + }, + { + "epoch": 0.55, + "learning_rate": 8.780792526089862e-06, + "loss": 1.0303, + "step": 17284 + }, + { + "epoch": 0.55, + "learning_rate": 8.779764456184471e-06, + "loss": 1.0151, + "step": 17285 + }, + { + "epoch": 0.55, + "learning_rate": 8.778736399370862e-06, + "loss": 1.0747, + "step": 17286 + }, + { + "epoch": 0.55, + "learning_rate": 8.777708355660057e-06, + "loss": 0.9878, + "step": 17287 + }, + { + "epoch": 0.55, + "learning_rate": 8.776680325063087e-06, + "loss": 0.9253, + "step": 17288 + }, + { + "epoch": 0.55, + "learning_rate": 8.775652307590982e-06, + "loss": 0.9917, + "step": 17289 + }, + { + "epoch": 0.55, + "learning_rate": 8.774624303254774e-06, + "loss": 0.9883, + "step": 17290 + }, + { + "epoch": 0.55, + "learning_rate": 8.773596312065492e-06, + "loss": 0.9268, + "step": 17291 + }, + { + "epoch": 0.55, + "learning_rate": 8.772568334034164e-06, + "loss": 0.9194, + "step": 17292 + }, + { + "epoch": 0.55, + "learning_rate": 8.771540369171822e-06, + "loss": 0.8706, + "step": 17293 + }, + { + "epoch": 0.55, + "learning_rate": 8.77051241748949e-06, + "loss": 0.8809, + "step": 17294 + }, + { + "epoch": 0.55, + "learning_rate": 8.769484478998198e-06, + "loss": 1.0586, + "step": 17295 + }, + { + "epoch": 0.55, + "learning_rate": 8.768456553708975e-06, + "loss": 1.083, + "step": 17296 + }, + { + "epoch": 0.55, + "learning_rate": 8.767428641632851e-06, + "loss": 0.9326, + "step": 17297 + }, + { + "epoch": 0.55, + "learning_rate": 8.766400742780853e-06, + "loss": 1.1494, + "step": 17298 + }, + { + "epoch": 0.55, + "learning_rate": 8.765372857164011e-06, + "loss": 0.9556, + "step": 17299 + }, + { + "epoch": 0.55, + "learning_rate": 8.764344984793351e-06, + "loss": 1.0293, + "step": 17300 + }, + { + "epoch": 0.55, + "learning_rate": 8.763317125679907e-06, + "loss": 0.9761, + "step": 17301 + }, + { + "epoch": 0.55, + "learning_rate": 8.762289279834696e-06, + "loss": 0.4763, + "step": 17302 + }, + { + "epoch": 0.55, + "learning_rate": 8.761261447268752e-06, + "loss": 0.7856, + "step": 17303 + }, + { + "epoch": 0.55, + "learning_rate": 8.7602336279931e-06, + "loss": 0.9292, + "step": 17304 + }, + { + "epoch": 0.55, + "learning_rate": 8.759205822018771e-06, + "loss": 1.0088, + "step": 17305 + }, + { + "epoch": 0.55, + "learning_rate": 8.75817802935679e-06, + "loss": 0.928, + "step": 17306 + }, + { + "epoch": 0.55, + "learning_rate": 8.757150250018183e-06, + "loss": 0.8506, + "step": 17307 + }, + { + "epoch": 0.55, + "learning_rate": 8.756122484013983e-06, + "loss": 0.957, + "step": 17308 + }, + { + "epoch": 0.55, + "learning_rate": 8.75509473135521e-06, + "loss": 0.8984, + "step": 17309 + }, + { + "epoch": 0.55, + "learning_rate": 8.754066992052889e-06, + "loss": 0.9697, + "step": 17310 + }, + { + "epoch": 0.55, + "learning_rate": 8.753039266118052e-06, + "loss": 1.0186, + "step": 17311 + }, + { + "epoch": 0.55, + "learning_rate": 8.752011553561725e-06, + "loss": 0.9795, + "step": 17312 + }, + { + "epoch": 0.55, + "learning_rate": 8.750983854394934e-06, + "loss": 1.0547, + "step": 17313 + }, + { + "epoch": 0.55, + "learning_rate": 8.749956168628702e-06, + "loss": 0.8599, + "step": 17314 + }, + { + "epoch": 0.55, + "learning_rate": 8.748928496274056e-06, + "loss": 1.0264, + "step": 17315 + }, + { + "epoch": 0.55, + "learning_rate": 8.747900837342026e-06, + "loss": 1.0542, + "step": 17316 + }, + { + "epoch": 0.55, + "learning_rate": 8.74687319184363e-06, + "loss": 1.0029, + "step": 17317 + }, + { + "epoch": 0.55, + "learning_rate": 8.745845559789902e-06, + "loss": 0.9722, + "step": 17318 + }, + { + "epoch": 0.55, + "learning_rate": 8.744817941191862e-06, + "loss": 1.0088, + "step": 17319 + }, + { + "epoch": 0.55, + "learning_rate": 8.743790336060534e-06, + "loss": 1.0591, + "step": 17320 + }, + { + "epoch": 0.55, + "learning_rate": 8.742762744406945e-06, + "loss": 1.0835, + "step": 17321 + }, + { + "epoch": 0.55, + "learning_rate": 8.74173516624212e-06, + "loss": 0.9292, + "step": 17322 + }, + { + "epoch": 0.55, + "learning_rate": 8.740707601577087e-06, + "loss": 0.5259, + "step": 17323 + }, + { + "epoch": 0.55, + "learning_rate": 8.739680050422869e-06, + "loss": 0.9795, + "step": 17324 + }, + { + "epoch": 0.55, + "learning_rate": 8.738652512790487e-06, + "loss": 0.8984, + "step": 17325 + }, + { + "epoch": 0.55, + "learning_rate": 8.737624988690968e-06, + "loss": 1.1426, + "step": 17326 + }, + { + "epoch": 0.55, + "learning_rate": 8.736597478135334e-06, + "loss": 0.9531, + "step": 17327 + }, + { + "epoch": 0.55, + "learning_rate": 8.73556998113461e-06, + "loss": 0.9565, + "step": 17328 + }, + { + "epoch": 0.55, + "learning_rate": 8.734542497699823e-06, + "loss": 1.0195, + "step": 17329 + }, + { + "epoch": 0.55, + "learning_rate": 8.733515027841993e-06, + "loss": 1.0693, + "step": 17330 + }, + { + "epoch": 0.55, + "learning_rate": 8.73248757157215e-06, + "loss": 0.9858, + "step": 17331 + }, + { + "epoch": 0.55, + "learning_rate": 8.731460128901308e-06, + "loss": 1.0664, + "step": 17332 + }, + { + "epoch": 0.55, + "learning_rate": 8.730432699840496e-06, + "loss": 1.0005, + "step": 17333 + }, + { + "epoch": 0.55, + "learning_rate": 8.729405284400735e-06, + "loss": 1.0029, + "step": 17334 + }, + { + "epoch": 0.55, + "learning_rate": 8.728377882593049e-06, + "loss": 0.9619, + "step": 17335 + }, + { + "epoch": 0.55, + "learning_rate": 8.72735049442846e-06, + "loss": 0.9736, + "step": 17336 + }, + { + "epoch": 0.55, + "learning_rate": 8.726323119917993e-06, + "loss": 0.5146, + "step": 17337 + }, + { + "epoch": 0.55, + "learning_rate": 8.72529575907267e-06, + "loss": 0.9854, + "step": 17338 + }, + { + "epoch": 0.55, + "learning_rate": 8.724268411903516e-06, + "loss": 0.9702, + "step": 17339 + }, + { + "epoch": 0.55, + "learning_rate": 8.723241078421544e-06, + "loss": 0.9189, + "step": 17340 + }, + { + "epoch": 0.55, + "learning_rate": 8.722213758637785e-06, + "loss": 1.0508, + "step": 17341 + }, + { + "epoch": 0.55, + "learning_rate": 8.721186452563257e-06, + "loss": 0.9321, + "step": 17342 + }, + { + "epoch": 0.55, + "learning_rate": 8.720159160208983e-06, + "loss": 0.9922, + "step": 17343 + }, + { + "epoch": 0.55, + "learning_rate": 8.719131881585987e-06, + "loss": 0.9844, + "step": 17344 + }, + { + "epoch": 0.55, + "learning_rate": 8.718104616705285e-06, + "loss": 0.9634, + "step": 17345 + }, + { + "epoch": 0.55, + "learning_rate": 8.717077365577908e-06, + "loss": 0.9927, + "step": 17346 + }, + { + "epoch": 0.55, + "learning_rate": 8.716050128214865e-06, + "loss": 1.0112, + "step": 17347 + }, + { + "epoch": 0.55, + "learning_rate": 8.715022904627185e-06, + "loss": 0.9932, + "step": 17348 + }, + { + "epoch": 0.55, + "learning_rate": 8.713995694825887e-06, + "loss": 1.0786, + "step": 17349 + }, + { + "epoch": 0.55, + "learning_rate": 8.71296849882199e-06, + "loss": 0.8701, + "step": 17350 + }, + { + "epoch": 0.55, + "learning_rate": 8.711941316626517e-06, + "loss": 0.582, + "step": 17351 + }, + { + "epoch": 0.55, + "learning_rate": 8.710914148250493e-06, + "loss": 0.9956, + "step": 17352 + }, + { + "epoch": 0.55, + "learning_rate": 8.709886993704928e-06, + "loss": 0.9575, + "step": 17353 + }, + { + "epoch": 0.55, + "learning_rate": 8.708859853000852e-06, + "loss": 0.8843, + "step": 17354 + }, + { + "epoch": 0.56, + "learning_rate": 8.707832726149276e-06, + "loss": 1.0913, + "step": 17355 + }, + { + "epoch": 0.56, + "learning_rate": 8.706805613161228e-06, + "loss": 0.9443, + "step": 17356 + }, + { + "epoch": 0.56, + "learning_rate": 8.705778514047723e-06, + "loss": 0.9946, + "step": 17357 + }, + { + "epoch": 0.56, + "learning_rate": 8.704751428819784e-06, + "loss": 0.9321, + "step": 17358 + }, + { + "epoch": 0.56, + "learning_rate": 8.703724357488428e-06, + "loss": 0.9526, + "step": 17359 + }, + { + "epoch": 0.56, + "learning_rate": 8.702697300064675e-06, + "loss": 0.8503, + "step": 17360 + }, + { + "epoch": 0.56, + "learning_rate": 8.701670256559544e-06, + "loss": 0.9463, + "step": 17361 + }, + { + "epoch": 0.56, + "learning_rate": 8.700643226984054e-06, + "loss": 1.061, + "step": 17362 + }, + { + "epoch": 0.56, + "learning_rate": 8.699616211349226e-06, + "loss": 1.0615, + "step": 17363 + }, + { + "epoch": 0.56, + "learning_rate": 8.698589209666074e-06, + "loss": 0.8501, + "step": 17364 + }, + { + "epoch": 0.56, + "learning_rate": 8.69756222194562e-06, + "loss": 0.897, + "step": 17365 + }, + { + "epoch": 0.56, + "learning_rate": 8.696535248198881e-06, + "loss": 1.0317, + "step": 17366 + }, + { + "epoch": 0.56, + "learning_rate": 8.695508288436877e-06, + "loss": 0.8091, + "step": 17367 + }, + { + "epoch": 0.56, + "learning_rate": 8.694481342670625e-06, + "loss": 1.0479, + "step": 17368 + }, + { + "epoch": 0.56, + "learning_rate": 8.693454410911148e-06, + "loss": 0.9487, + "step": 17369 + }, + { + "epoch": 0.56, + "learning_rate": 8.692427493169455e-06, + "loss": 0.9517, + "step": 17370 + }, + { + "epoch": 0.56, + "learning_rate": 8.691400589456568e-06, + "loss": 0.8677, + "step": 17371 + }, + { + "epoch": 0.56, + "learning_rate": 8.690373699783505e-06, + "loss": 0.5576, + "step": 17372 + }, + { + "epoch": 0.56, + "learning_rate": 8.68934682416128e-06, + "loss": 1.0171, + "step": 17373 + }, + { + "epoch": 0.56, + "learning_rate": 8.688319962600914e-06, + "loss": 1.0068, + "step": 17374 + }, + { + "epoch": 0.56, + "learning_rate": 8.687293115113426e-06, + "loss": 1.0596, + "step": 17375 + }, + { + "epoch": 0.56, + "learning_rate": 8.686266281709832e-06, + "loss": 1.0605, + "step": 17376 + }, + { + "epoch": 0.56, + "learning_rate": 8.685239462401143e-06, + "loss": 1.0117, + "step": 17377 + }, + { + "epoch": 0.56, + "learning_rate": 8.684212657198378e-06, + "loss": 0.8584, + "step": 17378 + }, + { + "epoch": 0.56, + "learning_rate": 8.683185866112557e-06, + "loss": 0.9883, + "step": 17379 + }, + { + "epoch": 0.56, + "learning_rate": 8.682159089154694e-06, + "loss": 0.8481, + "step": 17380 + }, + { + "epoch": 0.56, + "learning_rate": 8.681132326335804e-06, + "loss": 0.9019, + "step": 17381 + }, + { + "epoch": 0.56, + "learning_rate": 8.680105577666906e-06, + "loss": 1.0864, + "step": 17382 + }, + { + "epoch": 0.56, + "learning_rate": 8.679078843159013e-06, + "loss": 0.9336, + "step": 17383 + }, + { + "epoch": 0.56, + "learning_rate": 8.678052122823148e-06, + "loss": 0.9736, + "step": 17384 + }, + { + "epoch": 0.56, + "learning_rate": 8.677025416670315e-06, + "loss": 0.8569, + "step": 17385 + }, + { + "epoch": 0.56, + "learning_rate": 8.675998724711534e-06, + "loss": 1.061, + "step": 17386 + }, + { + "epoch": 0.56, + "learning_rate": 8.674972046957824e-06, + "loss": 1.001, + "step": 17387 + }, + { + "epoch": 0.56, + "learning_rate": 8.673945383420197e-06, + "loss": 1.042, + "step": 17388 + }, + { + "epoch": 0.56, + "learning_rate": 8.672918734109668e-06, + "loss": 1.0132, + "step": 17389 + }, + { + "epoch": 0.56, + "learning_rate": 8.671892099037251e-06, + "loss": 0.8384, + "step": 17390 + }, + { + "epoch": 0.56, + "learning_rate": 8.670865478213966e-06, + "loss": 1.0312, + "step": 17391 + }, + { + "epoch": 0.56, + "learning_rate": 8.669838871650822e-06, + "loss": 1.0645, + "step": 17392 + }, + { + "epoch": 0.56, + "learning_rate": 8.668812279358834e-06, + "loss": 1.1035, + "step": 17393 + }, + { + "epoch": 0.56, + "learning_rate": 8.667785701349016e-06, + "loss": 0.9702, + "step": 17394 + }, + { + "epoch": 0.56, + "learning_rate": 8.666759137632382e-06, + "loss": 1.0259, + "step": 17395 + }, + { + "epoch": 0.56, + "learning_rate": 8.665732588219948e-06, + "loss": 1.0278, + "step": 17396 + }, + { + "epoch": 0.56, + "learning_rate": 8.664706053122728e-06, + "loss": 0.9507, + "step": 17397 + }, + { + "epoch": 0.56, + "learning_rate": 8.663679532351734e-06, + "loss": 0.9678, + "step": 17398 + }, + { + "epoch": 0.56, + "learning_rate": 8.66265302591798e-06, + "loss": 0.9258, + "step": 17399 + }, + { + "epoch": 0.56, + "learning_rate": 8.661626533832477e-06, + "loss": 1.0288, + "step": 17400 + }, + { + "epoch": 0.56, + "learning_rate": 8.660600056106242e-06, + "loss": 1.0854, + "step": 17401 + }, + { + "epoch": 0.56, + "learning_rate": 8.659573592750287e-06, + "loss": 1.002, + "step": 17402 + }, + { + "epoch": 0.56, + "learning_rate": 8.65854714377562e-06, + "loss": 0.9463, + "step": 17403 + }, + { + "epoch": 0.56, + "learning_rate": 8.65752070919326e-06, + "loss": 1.0454, + "step": 17404 + }, + { + "epoch": 0.56, + "learning_rate": 8.656494289014216e-06, + "loss": 0.9834, + "step": 17405 + }, + { + "epoch": 0.56, + "learning_rate": 8.655467883249501e-06, + "loss": 0.9907, + "step": 17406 + }, + { + "epoch": 0.56, + "learning_rate": 8.654441491910132e-06, + "loss": 1.0396, + "step": 17407 + }, + { + "epoch": 0.56, + "learning_rate": 8.653415115007112e-06, + "loss": 0.9067, + "step": 17408 + }, + { + "epoch": 0.56, + "learning_rate": 8.652388752551458e-06, + "loss": 0.9854, + "step": 17409 + }, + { + "epoch": 0.56, + "learning_rate": 8.651362404554182e-06, + "loss": 1.0898, + "step": 17410 + }, + { + "epoch": 0.56, + "learning_rate": 8.650336071026294e-06, + "loss": 1.0308, + "step": 17411 + }, + { + "epoch": 0.56, + "learning_rate": 8.649309751978805e-06, + "loss": 0.9087, + "step": 17412 + }, + { + "epoch": 0.56, + "learning_rate": 8.648283447422728e-06, + "loss": 1.0181, + "step": 17413 + }, + { + "epoch": 0.56, + "learning_rate": 8.647257157369079e-06, + "loss": 0.9795, + "step": 17414 + }, + { + "epoch": 0.56, + "learning_rate": 8.646230881828858e-06, + "loss": 0.9399, + "step": 17415 + }, + { + "epoch": 0.56, + "learning_rate": 8.645204620813084e-06, + "loss": 0.9673, + "step": 17416 + }, + { + "epoch": 0.56, + "learning_rate": 8.644178374332763e-06, + "loss": 0.9321, + "step": 17417 + }, + { + "epoch": 0.56, + "learning_rate": 8.643152142398908e-06, + "loss": 0.9697, + "step": 17418 + }, + { + "epoch": 0.56, + "learning_rate": 8.642125925022528e-06, + "loss": 0.542, + "step": 17419 + }, + { + "epoch": 0.56, + "learning_rate": 8.641099722214635e-06, + "loss": 0.8652, + "step": 17420 + }, + { + "epoch": 0.56, + "learning_rate": 8.640073533986236e-06, + "loss": 1.0415, + "step": 17421 + }, + { + "epoch": 0.56, + "learning_rate": 8.63904736034835e-06, + "loss": 0.959, + "step": 17422 + }, + { + "epoch": 0.56, + "learning_rate": 8.638021201311973e-06, + "loss": 0.9497, + "step": 17423 + }, + { + "epoch": 0.56, + "learning_rate": 8.636995056888122e-06, + "loss": 1.0356, + "step": 17424 + }, + { + "epoch": 0.56, + "learning_rate": 8.635968927087806e-06, + "loss": 0.9688, + "step": 17425 + }, + { + "epoch": 0.56, + "learning_rate": 8.634942811922033e-06, + "loss": 1.0181, + "step": 17426 + }, + { + "epoch": 0.56, + "learning_rate": 8.633916711401813e-06, + "loss": 0.9077, + "step": 17427 + }, + { + "epoch": 0.56, + "learning_rate": 8.632890625538156e-06, + "loss": 0.9375, + "step": 17428 + }, + { + "epoch": 0.56, + "learning_rate": 8.631864554342074e-06, + "loss": 0.8877, + "step": 17429 + }, + { + "epoch": 0.56, + "learning_rate": 8.630838497824564e-06, + "loss": 1.0186, + "step": 17430 + }, + { + "epoch": 0.56, + "learning_rate": 8.629812455996643e-06, + "loss": 0.5366, + "step": 17431 + }, + { + "epoch": 0.56, + "learning_rate": 8.62878642886932e-06, + "loss": 0.9771, + "step": 17432 + }, + { + "epoch": 0.56, + "learning_rate": 8.627760416453599e-06, + "loss": 1.0635, + "step": 17433 + }, + { + "epoch": 0.56, + "learning_rate": 8.62673441876049e-06, + "loss": 0.9731, + "step": 17434 + }, + { + "epoch": 0.56, + "learning_rate": 8.625708435801e-06, + "loss": 1.0181, + "step": 17435 + }, + { + "epoch": 0.56, + "learning_rate": 8.624682467586142e-06, + "loss": 0.8789, + "step": 17436 + }, + { + "epoch": 0.56, + "learning_rate": 8.623656514126917e-06, + "loss": 1.0293, + "step": 17437 + }, + { + "epoch": 0.56, + "learning_rate": 8.622630575434333e-06, + "loss": 0.9355, + "step": 17438 + }, + { + "epoch": 0.56, + "learning_rate": 8.621604651519398e-06, + "loss": 0.8584, + "step": 17439 + }, + { + "epoch": 0.56, + "learning_rate": 8.620578742393121e-06, + "loss": 1.0474, + "step": 17440 + }, + { + "epoch": 0.56, + "learning_rate": 8.619552848066509e-06, + "loss": 0.9341, + "step": 17441 + }, + { + "epoch": 0.56, + "learning_rate": 8.618526968550564e-06, + "loss": 1.0479, + "step": 17442 + }, + { + "epoch": 0.56, + "learning_rate": 8.617501103856296e-06, + "loss": 0.8833, + "step": 17443 + }, + { + "epoch": 0.56, + "learning_rate": 8.616475253994714e-06, + "loss": 0.8906, + "step": 17444 + }, + { + "epoch": 0.56, + "learning_rate": 8.615449418976819e-06, + "loss": 0.9595, + "step": 17445 + }, + { + "epoch": 0.56, + "learning_rate": 8.61442359881362e-06, + "loss": 1.019, + "step": 17446 + }, + { + "epoch": 0.56, + "learning_rate": 8.613397793516123e-06, + "loss": 0.9595, + "step": 17447 + }, + { + "epoch": 0.56, + "learning_rate": 8.61237200309533e-06, + "loss": 0.9268, + "step": 17448 + }, + { + "epoch": 0.56, + "learning_rate": 8.611346227562251e-06, + "loss": 0.9302, + "step": 17449 + }, + { + "epoch": 0.56, + "learning_rate": 8.610320466927892e-06, + "loss": 0.9995, + "step": 17450 + }, + { + "epoch": 0.56, + "learning_rate": 8.609294721203255e-06, + "loss": 0.9526, + "step": 17451 + }, + { + "epoch": 0.56, + "learning_rate": 8.60826899039935e-06, + "loss": 0.9648, + "step": 17452 + }, + { + "epoch": 0.56, + "learning_rate": 8.607243274527173e-06, + "loss": 0.9419, + "step": 17453 + }, + { + "epoch": 0.56, + "learning_rate": 8.606217573597738e-06, + "loss": 1.0186, + "step": 17454 + }, + { + "epoch": 0.56, + "learning_rate": 8.605191887622042e-06, + "loss": 0.9229, + "step": 17455 + }, + { + "epoch": 0.56, + "learning_rate": 8.604166216611095e-06, + "loss": 0.9922, + "step": 17456 + }, + { + "epoch": 0.56, + "learning_rate": 8.6031405605759e-06, + "loss": 1.0171, + "step": 17457 + }, + { + "epoch": 0.56, + "learning_rate": 8.60211491952746e-06, + "loss": 1.083, + "step": 17458 + }, + { + "epoch": 0.56, + "learning_rate": 8.601089293476785e-06, + "loss": 1.0176, + "step": 17459 + }, + { + "epoch": 0.56, + "learning_rate": 8.600063682434869e-06, + "loss": 0.9595, + "step": 17460 + }, + { + "epoch": 0.56, + "learning_rate": 8.59903808641272e-06, + "loss": 0.9341, + "step": 17461 + }, + { + "epoch": 0.56, + "learning_rate": 8.598012505421342e-06, + "loss": 1.02, + "step": 17462 + }, + { + "epoch": 0.56, + "learning_rate": 8.596986939471739e-06, + "loss": 0.9854, + "step": 17463 + }, + { + "epoch": 0.56, + "learning_rate": 8.595961388574911e-06, + "loss": 1.0083, + "step": 17464 + }, + { + "epoch": 0.56, + "learning_rate": 8.594935852741865e-06, + "loss": 0.9175, + "step": 17465 + }, + { + "epoch": 0.56, + "learning_rate": 8.593910331983603e-06, + "loss": 1.0156, + "step": 17466 + }, + { + "epoch": 0.56, + "learning_rate": 8.592884826311131e-06, + "loss": 0.9146, + "step": 17467 + }, + { + "epoch": 0.56, + "learning_rate": 8.591859335735442e-06, + "loss": 0.9971, + "step": 17468 + }, + { + "epoch": 0.56, + "learning_rate": 8.590833860267545e-06, + "loss": 1.0415, + "step": 17469 + }, + { + "epoch": 0.56, + "learning_rate": 8.58980839991844e-06, + "loss": 0.9658, + "step": 17470 + }, + { + "epoch": 0.56, + "learning_rate": 8.588782954699131e-06, + "loss": 1.0005, + "step": 17471 + }, + { + "epoch": 0.56, + "learning_rate": 8.587757524620618e-06, + "loss": 1.0566, + "step": 17472 + }, + { + "epoch": 0.56, + "learning_rate": 8.586732109693905e-06, + "loss": 1.0557, + "step": 17473 + }, + { + "epoch": 0.56, + "learning_rate": 8.585706709929993e-06, + "loss": 0.9014, + "step": 17474 + }, + { + "epoch": 0.56, + "learning_rate": 8.584681325339886e-06, + "loss": 1.0044, + "step": 17475 + }, + { + "epoch": 0.56, + "learning_rate": 8.583655955934578e-06, + "loss": 0.9761, + "step": 17476 + }, + { + "epoch": 0.56, + "learning_rate": 8.582630601725072e-06, + "loss": 0.9839, + "step": 17477 + }, + { + "epoch": 0.56, + "learning_rate": 8.581605262722374e-06, + "loss": 1.0044, + "step": 17478 + }, + { + "epoch": 0.56, + "learning_rate": 8.58057993893748e-06, + "loss": 1.0513, + "step": 17479 + }, + { + "epoch": 0.56, + "learning_rate": 8.579554630381395e-06, + "loss": 0.9907, + "step": 17480 + }, + { + "epoch": 0.56, + "learning_rate": 8.578529337065115e-06, + "loss": 0.5303, + "step": 17481 + }, + { + "epoch": 0.56, + "learning_rate": 8.577504058999644e-06, + "loss": 1.0879, + "step": 17482 + }, + { + "epoch": 0.56, + "learning_rate": 8.576478796195978e-06, + "loss": 0.957, + "step": 17483 + }, + { + "epoch": 0.56, + "learning_rate": 8.575453548665118e-06, + "loss": 0.8882, + "step": 17484 + }, + { + "epoch": 0.56, + "learning_rate": 8.574428316418065e-06, + "loss": 0.9897, + "step": 17485 + }, + { + "epoch": 0.56, + "learning_rate": 8.573403099465822e-06, + "loss": 1.0479, + "step": 17486 + }, + { + "epoch": 0.56, + "learning_rate": 8.57237789781938e-06, + "loss": 0.936, + "step": 17487 + }, + { + "epoch": 0.56, + "learning_rate": 8.571352711489745e-06, + "loss": 0.9473, + "step": 17488 + }, + { + "epoch": 0.56, + "learning_rate": 8.570327540487914e-06, + "loss": 0.9238, + "step": 17489 + }, + { + "epoch": 0.56, + "learning_rate": 8.56930238482489e-06, + "loss": 0.8579, + "step": 17490 + }, + { + "epoch": 0.56, + "learning_rate": 8.568277244511665e-06, + "loss": 0.938, + "step": 17491 + }, + { + "epoch": 0.56, + "learning_rate": 8.567252119559241e-06, + "loss": 0.9312, + "step": 17492 + }, + { + "epoch": 0.56, + "learning_rate": 8.566227009978616e-06, + "loss": 1.0122, + "step": 17493 + }, + { + "epoch": 0.56, + "learning_rate": 8.565201915780786e-06, + "loss": 1.0215, + "step": 17494 + }, + { + "epoch": 0.56, + "learning_rate": 8.564176836976753e-06, + "loss": 0.937, + "step": 17495 + }, + { + "epoch": 0.56, + "learning_rate": 8.563151773577514e-06, + "loss": 1.0327, + "step": 17496 + }, + { + "epoch": 0.56, + "learning_rate": 8.562126725594069e-06, + "loss": 0.958, + "step": 17497 + }, + { + "epoch": 0.56, + "learning_rate": 8.56110169303741e-06, + "loss": 0.9731, + "step": 17498 + }, + { + "epoch": 0.56, + "learning_rate": 8.560076675918537e-06, + "loss": 1.0371, + "step": 17499 + }, + { + "epoch": 0.56, + "learning_rate": 8.559051674248446e-06, + "loss": 0.9995, + "step": 17500 + }, + { + "epoch": 0.56, + "learning_rate": 8.558026688038139e-06, + "loss": 1.0254, + "step": 17501 + }, + { + "epoch": 0.56, + "learning_rate": 8.557001717298607e-06, + "loss": 0.9414, + "step": 17502 + }, + { + "epoch": 0.56, + "learning_rate": 8.55597676204085e-06, + "loss": 0.9937, + "step": 17503 + }, + { + "epoch": 0.56, + "learning_rate": 8.554951822275866e-06, + "loss": 0.9678, + "step": 17504 + }, + { + "epoch": 0.56, + "learning_rate": 8.553926898014651e-06, + "loss": 1.022, + "step": 17505 + }, + { + "epoch": 0.56, + "learning_rate": 8.552901989268199e-06, + "loss": 0.9932, + "step": 17506 + }, + { + "epoch": 0.56, + "learning_rate": 8.551877096047504e-06, + "loss": 1.0474, + "step": 17507 + }, + { + "epoch": 0.56, + "learning_rate": 8.550852218363568e-06, + "loss": 0.9209, + "step": 17508 + }, + { + "epoch": 0.56, + "learning_rate": 8.54982735622738e-06, + "loss": 1.0718, + "step": 17509 + }, + { + "epoch": 0.56, + "learning_rate": 8.548802509649944e-06, + "loss": 0.9922, + "step": 17510 + }, + { + "epoch": 0.56, + "learning_rate": 8.547777678642247e-06, + "loss": 0.9326, + "step": 17511 + }, + { + "epoch": 0.56, + "learning_rate": 8.546752863215294e-06, + "loss": 0.897, + "step": 17512 + }, + { + "epoch": 0.56, + "learning_rate": 8.54572806338007e-06, + "loss": 1.0439, + "step": 17513 + }, + { + "epoch": 0.56, + "learning_rate": 8.544703279147576e-06, + "loss": 0.96, + "step": 17514 + }, + { + "epoch": 0.56, + "learning_rate": 8.543678510528803e-06, + "loss": 1.0332, + "step": 17515 + }, + { + "epoch": 0.56, + "learning_rate": 8.542653757534748e-06, + "loss": 1.0303, + "step": 17516 + }, + { + "epoch": 0.56, + "learning_rate": 8.541629020176407e-06, + "loss": 0.9819, + "step": 17517 + }, + { + "epoch": 0.56, + "learning_rate": 8.540604298464771e-06, + "loss": 0.5149, + "step": 17518 + }, + { + "epoch": 0.56, + "learning_rate": 8.53957959241084e-06, + "loss": 0.9312, + "step": 17519 + }, + { + "epoch": 0.56, + "learning_rate": 8.538554902025602e-06, + "loss": 1.0571, + "step": 17520 + }, + { + "epoch": 0.56, + "learning_rate": 8.53753022732005e-06, + "loss": 0.9927, + "step": 17521 + }, + { + "epoch": 0.56, + "learning_rate": 8.53650556830518e-06, + "loss": 1.0098, + "step": 17522 + }, + { + "epoch": 0.56, + "learning_rate": 8.535480924991986e-06, + "loss": 0.9819, + "step": 17523 + }, + { + "epoch": 0.56, + "learning_rate": 8.534456297391461e-06, + "loss": 1.0366, + "step": 17524 + }, + { + "epoch": 0.56, + "learning_rate": 8.533431685514601e-06, + "loss": 0.9819, + "step": 17525 + }, + { + "epoch": 0.56, + "learning_rate": 8.532407089372392e-06, + "loss": 1.0933, + "step": 17526 + }, + { + "epoch": 0.56, + "learning_rate": 8.531382508975833e-06, + "loss": 0.9639, + "step": 17527 + }, + { + "epoch": 0.56, + "learning_rate": 8.530357944335913e-06, + "loss": 0.936, + "step": 17528 + }, + { + "epoch": 0.56, + "learning_rate": 8.529333395463625e-06, + "loss": 0.9531, + "step": 17529 + }, + { + "epoch": 0.56, + "learning_rate": 8.528308862369966e-06, + "loss": 0.9761, + "step": 17530 + }, + { + "epoch": 0.56, + "learning_rate": 8.527284345065919e-06, + "loss": 1.0088, + "step": 17531 + }, + { + "epoch": 0.56, + "learning_rate": 8.526259843562483e-06, + "loss": 0.9111, + "step": 17532 + }, + { + "epoch": 0.56, + "learning_rate": 8.525235357870648e-06, + "loss": 0.876, + "step": 17533 + }, + { + "epoch": 0.56, + "learning_rate": 8.524210888001405e-06, + "loss": 1.1499, + "step": 17534 + }, + { + "epoch": 0.56, + "learning_rate": 8.523186433965748e-06, + "loss": 0.9722, + "step": 17535 + }, + { + "epoch": 0.56, + "learning_rate": 8.522161995774665e-06, + "loss": 1.1777, + "step": 17536 + }, + { + "epoch": 0.56, + "learning_rate": 8.521137573439147e-06, + "loss": 0.8906, + "step": 17537 + }, + { + "epoch": 0.56, + "learning_rate": 8.520113166970186e-06, + "loss": 0.8896, + "step": 17538 + }, + { + "epoch": 0.56, + "learning_rate": 8.519088776378773e-06, + "loss": 1.0464, + "step": 17539 + }, + { + "epoch": 0.56, + "learning_rate": 8.518064401675899e-06, + "loss": 0.9253, + "step": 17540 + }, + { + "epoch": 0.56, + "learning_rate": 8.517040042872553e-06, + "loss": 0.9888, + "step": 17541 + }, + { + "epoch": 0.56, + "learning_rate": 8.51601569997973e-06, + "loss": 0.9585, + "step": 17542 + }, + { + "epoch": 0.56, + "learning_rate": 8.514991373008412e-06, + "loss": 0.9661, + "step": 17543 + }, + { + "epoch": 0.56, + "learning_rate": 8.513967061969594e-06, + "loss": 0.9443, + "step": 17544 + }, + { + "epoch": 0.56, + "learning_rate": 8.512942766874265e-06, + "loss": 1.0269, + "step": 17545 + }, + { + "epoch": 0.56, + "learning_rate": 8.511918487733412e-06, + "loss": 1.0586, + "step": 17546 + }, + { + "epoch": 0.56, + "learning_rate": 8.51089422455803e-06, + "loss": 1.0962, + "step": 17547 + }, + { + "epoch": 0.56, + "learning_rate": 8.509869977359102e-06, + "loss": 0.9375, + "step": 17548 + }, + { + "epoch": 0.56, + "learning_rate": 8.508845746147622e-06, + "loss": 1.0098, + "step": 17549 + }, + { + "epoch": 0.56, + "learning_rate": 8.50782153093458e-06, + "loss": 1.063, + "step": 17550 + }, + { + "epoch": 0.56, + "learning_rate": 8.506797331730957e-06, + "loss": 0.9512, + "step": 17551 + }, + { + "epoch": 0.56, + "learning_rate": 8.505773148547746e-06, + "loss": 1.0117, + "step": 17552 + }, + { + "epoch": 0.56, + "learning_rate": 8.504748981395937e-06, + "loss": 0.8931, + "step": 17553 + }, + { + "epoch": 0.56, + "learning_rate": 8.503724830286515e-06, + "loss": 1.0967, + "step": 17554 + }, + { + "epoch": 0.56, + "learning_rate": 8.50270069523047e-06, + "loss": 1.019, + "step": 17555 + }, + { + "epoch": 0.56, + "learning_rate": 8.50167657623879e-06, + "loss": 0.8896, + "step": 17556 + }, + { + "epoch": 0.56, + "learning_rate": 8.500652473322462e-06, + "loss": 1.0103, + "step": 17557 + }, + { + "epoch": 0.56, + "learning_rate": 8.499628386492476e-06, + "loss": 0.9917, + "step": 17558 + }, + { + "epoch": 0.56, + "learning_rate": 8.498604315759815e-06, + "loss": 1.0103, + "step": 17559 + }, + { + "epoch": 0.56, + "learning_rate": 8.497580261135467e-06, + "loss": 0.9316, + "step": 17560 + }, + { + "epoch": 0.56, + "learning_rate": 8.49655622263042e-06, + "loss": 1.0479, + "step": 17561 + }, + { + "epoch": 0.56, + "learning_rate": 8.495532200255661e-06, + "loss": 0.9639, + "step": 17562 + }, + { + "epoch": 0.56, + "learning_rate": 8.494508194022175e-06, + "loss": 0.9561, + "step": 17563 + }, + { + "epoch": 0.56, + "learning_rate": 8.493484203940953e-06, + "loss": 0.8628, + "step": 17564 + }, + { + "epoch": 0.56, + "learning_rate": 8.49246023002298e-06, + "loss": 0.9399, + "step": 17565 + }, + { + "epoch": 0.56, + "learning_rate": 8.491436272279237e-06, + "loss": 1.1133, + "step": 17566 + }, + { + "epoch": 0.56, + "learning_rate": 8.490412330720711e-06, + "loss": 1.0044, + "step": 17567 + }, + { + "epoch": 0.56, + "learning_rate": 8.489388405358392e-06, + "loss": 0.9346, + "step": 17568 + }, + { + "epoch": 0.56, + "learning_rate": 8.488364496203266e-06, + "loss": 0.9189, + "step": 17569 + }, + { + "epoch": 0.56, + "learning_rate": 8.487340603266313e-06, + "loss": 0.9287, + "step": 17570 + }, + { + "epoch": 0.56, + "learning_rate": 8.486316726558522e-06, + "loss": 0.9126, + "step": 17571 + }, + { + "epoch": 0.56, + "learning_rate": 8.485292866090877e-06, + "loss": 0.9775, + "step": 17572 + }, + { + "epoch": 0.56, + "learning_rate": 8.484269021874365e-06, + "loss": 0.9946, + "step": 17573 + }, + { + "epoch": 0.56, + "learning_rate": 8.483245193919965e-06, + "loss": 1.0059, + "step": 17574 + }, + { + "epoch": 0.56, + "learning_rate": 8.48222138223867e-06, + "loss": 0.9883, + "step": 17575 + }, + { + "epoch": 0.56, + "learning_rate": 8.481197586841456e-06, + "loss": 0.9629, + "step": 17576 + }, + { + "epoch": 0.56, + "learning_rate": 8.480173807739312e-06, + "loss": 0.897, + "step": 17577 + }, + { + "epoch": 0.56, + "learning_rate": 8.479150044943223e-06, + "loss": 0.9805, + "step": 17578 + }, + { + "epoch": 0.56, + "learning_rate": 8.478126298464167e-06, + "loss": 0.958, + "step": 17579 + }, + { + "epoch": 0.56, + "learning_rate": 8.477102568313138e-06, + "loss": 0.9434, + "step": 17580 + }, + { + "epoch": 0.56, + "learning_rate": 8.476078854501108e-06, + "loss": 0.5127, + "step": 17581 + }, + { + "epoch": 0.56, + "learning_rate": 8.475055157039065e-06, + "loss": 0.9893, + "step": 17582 + }, + { + "epoch": 0.56, + "learning_rate": 8.474031475937995e-06, + "loss": 0.9502, + "step": 17583 + }, + { + "epoch": 0.56, + "learning_rate": 8.473007811208874e-06, + "loss": 1.0088, + "step": 17584 + }, + { + "epoch": 0.56, + "learning_rate": 8.471984162862692e-06, + "loss": 0.9546, + "step": 17585 + }, + { + "epoch": 0.56, + "learning_rate": 8.470960530910429e-06, + "loss": 0.9253, + "step": 17586 + }, + { + "epoch": 0.56, + "learning_rate": 8.469936915363066e-06, + "loss": 0.5134, + "step": 17587 + }, + { + "epoch": 0.56, + "learning_rate": 8.46891331623159e-06, + "loss": 1.0278, + "step": 17588 + }, + { + "epoch": 0.56, + "learning_rate": 8.467889733526977e-06, + "loss": 1.0928, + "step": 17589 + }, + { + "epoch": 0.56, + "learning_rate": 8.466866167260212e-06, + "loss": 0.9824, + "step": 17590 + }, + { + "epoch": 0.56, + "learning_rate": 8.465842617442274e-06, + "loss": 0.9263, + "step": 17591 + }, + { + "epoch": 0.56, + "learning_rate": 8.464819084084149e-06, + "loss": 0.9302, + "step": 17592 + }, + { + "epoch": 0.56, + "learning_rate": 8.463795567196816e-06, + "loss": 1.0205, + "step": 17593 + }, + { + "epoch": 0.56, + "learning_rate": 8.462772066791256e-06, + "loss": 0.9575, + "step": 17594 + }, + { + "epoch": 0.56, + "learning_rate": 8.461748582878454e-06, + "loss": 1.0405, + "step": 17595 + }, + { + "epoch": 0.56, + "learning_rate": 8.460725115469383e-06, + "loss": 0.8877, + "step": 17596 + }, + { + "epoch": 0.56, + "learning_rate": 8.45970166457503e-06, + "loss": 1.0562, + "step": 17597 + }, + { + "epoch": 0.56, + "learning_rate": 8.45867823020637e-06, + "loss": 1.0752, + "step": 17598 + }, + { + "epoch": 0.56, + "learning_rate": 8.457654812374388e-06, + "loss": 0.9849, + "step": 17599 + }, + { + "epoch": 0.56, + "learning_rate": 8.456631411090064e-06, + "loss": 0.9561, + "step": 17600 + }, + { + "epoch": 0.56, + "learning_rate": 8.455608026364376e-06, + "loss": 0.9492, + "step": 17601 + }, + { + "epoch": 0.56, + "learning_rate": 8.454584658208305e-06, + "loss": 0.9453, + "step": 17602 + }, + { + "epoch": 0.56, + "learning_rate": 8.453561306632833e-06, + "loss": 0.9771, + "step": 17603 + }, + { + "epoch": 0.56, + "learning_rate": 8.452537971648935e-06, + "loss": 1.0742, + "step": 17604 + }, + { + "epoch": 0.56, + "learning_rate": 8.451514653267591e-06, + "loss": 0.8262, + "step": 17605 + }, + { + "epoch": 0.56, + "learning_rate": 8.45049135149978e-06, + "loss": 0.9932, + "step": 17606 + }, + { + "epoch": 0.56, + "learning_rate": 8.449468066356483e-06, + "loss": 1.002, + "step": 17607 + }, + { + "epoch": 0.56, + "learning_rate": 8.448444797848679e-06, + "loss": 0.8394, + "step": 17608 + }, + { + "epoch": 0.56, + "learning_rate": 8.447421545987343e-06, + "loss": 0.8994, + "step": 17609 + }, + { + "epoch": 0.56, + "learning_rate": 8.446398310783458e-06, + "loss": 1.0068, + "step": 17610 + }, + { + "epoch": 0.56, + "learning_rate": 8.445375092247999e-06, + "loss": 0.9873, + "step": 17611 + }, + { + "epoch": 0.56, + "learning_rate": 8.444351890391941e-06, + "loss": 0.9941, + "step": 17612 + }, + { + "epoch": 0.56, + "learning_rate": 8.443328705226268e-06, + "loss": 1.0171, + "step": 17613 + }, + { + "epoch": 0.56, + "learning_rate": 8.442305536761957e-06, + "loss": 0.9824, + "step": 17614 + }, + { + "epoch": 0.56, + "learning_rate": 8.44128238500998e-06, + "loss": 1.0483, + "step": 17615 + }, + { + "epoch": 0.56, + "learning_rate": 8.440259249981321e-06, + "loss": 1.0415, + "step": 17616 + }, + { + "epoch": 0.56, + "learning_rate": 8.439236131686953e-06, + "loss": 0.4937, + "step": 17617 + }, + { + "epoch": 0.56, + "learning_rate": 8.438213030137855e-06, + "loss": 1.0635, + "step": 17618 + }, + { + "epoch": 0.56, + "learning_rate": 8.437189945345003e-06, + "loss": 1.0396, + "step": 17619 + }, + { + "epoch": 0.56, + "learning_rate": 8.436166877319372e-06, + "loss": 1.0547, + "step": 17620 + }, + { + "epoch": 0.56, + "learning_rate": 8.435143826071939e-06, + "loss": 0.9893, + "step": 17621 + }, + { + "epoch": 0.56, + "learning_rate": 8.43412079161368e-06, + "loss": 0.9409, + "step": 17622 + }, + { + "epoch": 0.56, + "learning_rate": 8.433097773955572e-06, + "loss": 0.9058, + "step": 17623 + }, + { + "epoch": 0.56, + "learning_rate": 8.432074773108591e-06, + "loss": 1.0029, + "step": 17624 + }, + { + "epoch": 0.56, + "learning_rate": 8.431051789083712e-06, + "loss": 0.9302, + "step": 17625 + }, + { + "epoch": 0.56, + "learning_rate": 8.430028821891917e-06, + "loss": 0.9941, + "step": 17626 + }, + { + "epoch": 0.56, + "learning_rate": 8.42900587154417e-06, + "loss": 0.8823, + "step": 17627 + }, + { + "epoch": 0.56, + "learning_rate": 8.42798293805145e-06, + "loss": 0.9883, + "step": 17628 + }, + { + "epoch": 0.56, + "learning_rate": 8.426960021424734e-06, + "loss": 0.8765, + "step": 17629 + }, + { + "epoch": 0.56, + "learning_rate": 8.425937121674997e-06, + "loss": 0.9922, + "step": 17630 + }, + { + "epoch": 0.56, + "learning_rate": 8.424914238813212e-06, + "loss": 1.0151, + "step": 17631 + }, + { + "epoch": 0.56, + "learning_rate": 8.423891372850355e-06, + "loss": 0.8784, + "step": 17632 + }, + { + "epoch": 0.56, + "learning_rate": 8.422868523797402e-06, + "loss": 0.9502, + "step": 17633 + }, + { + "epoch": 0.56, + "learning_rate": 8.42184569166532e-06, + "loss": 0.9619, + "step": 17634 + }, + { + "epoch": 0.56, + "learning_rate": 8.420822876465088e-06, + "loss": 0.9482, + "step": 17635 + }, + { + "epoch": 0.56, + "learning_rate": 8.419800078207679e-06, + "loss": 0.9434, + "step": 17636 + }, + { + "epoch": 0.56, + "learning_rate": 8.418777296904064e-06, + "loss": 0.9048, + "step": 17637 + }, + { + "epoch": 0.56, + "learning_rate": 8.417754532565222e-06, + "loss": 1.0801, + "step": 17638 + }, + { + "epoch": 0.56, + "learning_rate": 8.416731785202119e-06, + "loss": 1.0713, + "step": 17639 + }, + { + "epoch": 0.56, + "learning_rate": 8.415709054825733e-06, + "loss": 0.9287, + "step": 17640 + }, + { + "epoch": 0.56, + "learning_rate": 8.41468634144704e-06, + "loss": 1.0127, + "step": 17641 + }, + { + "epoch": 0.56, + "learning_rate": 8.413663645077005e-06, + "loss": 1.1035, + "step": 17642 + }, + { + "epoch": 0.56, + "learning_rate": 8.412640965726603e-06, + "loss": 1.0933, + "step": 17643 + }, + { + "epoch": 0.56, + "learning_rate": 8.411618303406805e-06, + "loss": 1.0464, + "step": 17644 + }, + { + "epoch": 0.56, + "learning_rate": 8.410595658128586e-06, + "loss": 1.0649, + "step": 17645 + }, + { + "epoch": 0.56, + "learning_rate": 8.409573029902917e-06, + "loss": 0.9858, + "step": 17646 + }, + { + "epoch": 0.56, + "learning_rate": 8.40855041874077e-06, + "loss": 0.9546, + "step": 17647 + }, + { + "epoch": 0.56, + "learning_rate": 8.407527824653116e-06, + "loss": 0.894, + "step": 17648 + }, + { + "epoch": 0.56, + "learning_rate": 8.406505247650926e-06, + "loss": 0.8154, + "step": 17649 + }, + { + "epoch": 0.56, + "learning_rate": 8.405482687745168e-06, + "loss": 0.9165, + "step": 17650 + }, + { + "epoch": 0.56, + "learning_rate": 8.404460144946819e-06, + "loss": 0.9175, + "step": 17651 + }, + { + "epoch": 0.56, + "learning_rate": 8.403437619266843e-06, + "loss": 0.9873, + "step": 17652 + }, + { + "epoch": 0.56, + "learning_rate": 8.40241511071622e-06, + "loss": 1.1187, + "step": 17653 + }, + { + "epoch": 0.56, + "learning_rate": 8.40139261930591e-06, + "loss": 1.0659, + "step": 17654 + }, + { + "epoch": 0.56, + "learning_rate": 8.40037014504689e-06, + "loss": 1.0537, + "step": 17655 + }, + { + "epoch": 0.56, + "learning_rate": 8.399347687950129e-06, + "loss": 1.0493, + "step": 17656 + }, + { + "epoch": 0.56, + "learning_rate": 8.398325248026592e-06, + "loss": 0.9961, + "step": 17657 + }, + { + "epoch": 0.56, + "learning_rate": 8.397302825287257e-06, + "loss": 0.9805, + "step": 17658 + }, + { + "epoch": 0.56, + "learning_rate": 8.396280419743087e-06, + "loss": 0.9595, + "step": 17659 + }, + { + "epoch": 0.56, + "learning_rate": 8.395258031405052e-06, + "loss": 0.9873, + "step": 17660 + }, + { + "epoch": 0.56, + "learning_rate": 8.394235660284123e-06, + "loss": 0.9917, + "step": 17661 + }, + { + "epoch": 0.56, + "learning_rate": 8.393213306391268e-06, + "loss": 0.9824, + "step": 17662 + }, + { + "epoch": 0.56, + "learning_rate": 8.392190969737459e-06, + "loss": 0.9785, + "step": 17663 + }, + { + "epoch": 0.56, + "learning_rate": 8.39116865033366e-06, + "loss": 1.0278, + "step": 17664 + }, + { + "epoch": 0.56, + "learning_rate": 8.390146348190839e-06, + "loss": 0.9385, + "step": 17665 + }, + { + "epoch": 0.56, + "learning_rate": 8.389124063319966e-06, + "loss": 0.9131, + "step": 17666 + }, + { + "epoch": 0.57, + "learning_rate": 8.388101795732008e-06, + "loss": 0.542, + "step": 17667 + }, + { + "epoch": 0.57, + "learning_rate": 8.387079545437934e-06, + "loss": 0.5095, + "step": 17668 + }, + { + "epoch": 0.57, + "learning_rate": 8.386057312448713e-06, + "loss": 1.0093, + "step": 17669 + }, + { + "epoch": 0.57, + "learning_rate": 8.385035096775311e-06, + "loss": 1.0879, + "step": 17670 + }, + { + "epoch": 0.57, + "learning_rate": 8.384012898428698e-06, + "loss": 0.8716, + "step": 17671 + }, + { + "epoch": 0.57, + "learning_rate": 8.382990717419834e-06, + "loss": 0.531, + "step": 17672 + }, + { + "epoch": 0.57, + "learning_rate": 8.38196855375969e-06, + "loss": 1.0205, + "step": 17673 + }, + { + "epoch": 0.57, + "learning_rate": 8.380946407459233e-06, + "loss": 0.96, + "step": 17674 + }, + { + "epoch": 0.57, + "learning_rate": 8.379924278529429e-06, + "loss": 0.8965, + "step": 17675 + }, + { + "epoch": 0.57, + "learning_rate": 8.378902166981244e-06, + "loss": 0.8945, + "step": 17676 + }, + { + "epoch": 0.57, + "learning_rate": 8.377880072825647e-06, + "loss": 0.9868, + "step": 17677 + }, + { + "epoch": 0.57, + "learning_rate": 8.376857996073602e-06, + "loss": 1.0166, + "step": 17678 + }, + { + "epoch": 0.57, + "learning_rate": 8.375835936736072e-06, + "loss": 1.0938, + "step": 17679 + }, + { + "epoch": 0.57, + "learning_rate": 8.374813894824025e-06, + "loss": 1.0039, + "step": 17680 + }, + { + "epoch": 0.57, + "learning_rate": 8.373791870348427e-06, + "loss": 1.104, + "step": 17681 + }, + { + "epoch": 0.57, + "learning_rate": 8.37276986332024e-06, + "loss": 1.0718, + "step": 17682 + }, + { + "epoch": 0.57, + "learning_rate": 8.371747873750433e-06, + "loss": 0.9014, + "step": 17683 + }, + { + "epoch": 0.57, + "learning_rate": 8.37072590164997e-06, + "loss": 0.9224, + "step": 17684 + }, + { + "epoch": 0.57, + "learning_rate": 8.369703947029814e-06, + "loss": 0.979, + "step": 17685 + }, + { + "epoch": 0.57, + "learning_rate": 8.368682009900935e-06, + "loss": 0.9048, + "step": 17686 + }, + { + "epoch": 0.57, + "learning_rate": 8.367660090274288e-06, + "loss": 0.9785, + "step": 17687 + }, + { + "epoch": 0.57, + "learning_rate": 8.36663818816084e-06, + "loss": 0.9185, + "step": 17688 + }, + { + "epoch": 0.57, + "learning_rate": 8.365616303571559e-06, + "loss": 0.9912, + "step": 17689 + }, + { + "epoch": 0.57, + "learning_rate": 8.364594436517404e-06, + "loss": 1.0771, + "step": 17690 + }, + { + "epoch": 0.57, + "learning_rate": 8.363572587009342e-06, + "loss": 0.9336, + "step": 17691 + }, + { + "epoch": 0.57, + "learning_rate": 8.362550755058339e-06, + "loss": 0.9028, + "step": 17692 + }, + { + "epoch": 0.57, + "learning_rate": 8.361528940675351e-06, + "loss": 0.9829, + "step": 17693 + }, + { + "epoch": 0.57, + "learning_rate": 8.360507143871344e-06, + "loss": 0.957, + "step": 17694 + }, + { + "epoch": 0.57, + "learning_rate": 8.35948536465728e-06, + "loss": 0.5881, + "step": 17695 + }, + { + "epoch": 0.57, + "learning_rate": 8.358463603044123e-06, + "loss": 0.9009, + "step": 17696 + }, + { + "epoch": 0.57, + "learning_rate": 8.357441859042837e-06, + "loss": 0.9844, + "step": 17697 + }, + { + "epoch": 0.57, + "learning_rate": 8.356420132664379e-06, + "loss": 0.9307, + "step": 17698 + }, + { + "epoch": 0.57, + "learning_rate": 8.355398423919715e-06, + "loss": 0.896, + "step": 17699 + }, + { + "epoch": 0.57, + "learning_rate": 8.354376732819804e-06, + "loss": 0.9478, + "step": 17700 + }, + { + "epoch": 0.57, + "learning_rate": 8.353355059375613e-06, + "loss": 1.0347, + "step": 17701 + }, + { + "epoch": 0.57, + "learning_rate": 8.352333403598096e-06, + "loss": 1.0093, + "step": 17702 + }, + { + "epoch": 0.57, + "learning_rate": 8.351311765498222e-06, + "loss": 0.9272, + "step": 17703 + }, + { + "epoch": 0.57, + "learning_rate": 8.350290145086945e-06, + "loss": 0.999, + "step": 17704 + }, + { + "epoch": 0.57, + "learning_rate": 8.349268542375229e-06, + "loss": 0.9985, + "step": 17705 + }, + { + "epoch": 0.57, + "learning_rate": 8.348246957374035e-06, + "loss": 0.9463, + "step": 17706 + }, + { + "epoch": 0.57, + "learning_rate": 8.347225390094324e-06, + "loss": 0.9785, + "step": 17707 + }, + { + "epoch": 0.57, + "learning_rate": 8.346203840547053e-06, + "loss": 0.9429, + "step": 17708 + }, + { + "epoch": 0.57, + "learning_rate": 8.345182308743191e-06, + "loss": 1.02, + "step": 17709 + }, + { + "epoch": 0.57, + "learning_rate": 8.344160794693685e-06, + "loss": 1.0659, + "step": 17710 + }, + { + "epoch": 0.57, + "learning_rate": 8.343139298409502e-06, + "loss": 1.0137, + "step": 17711 + }, + { + "epoch": 0.57, + "learning_rate": 8.342117819901602e-06, + "loss": 0.9692, + "step": 17712 + }, + { + "epoch": 0.57, + "learning_rate": 8.341096359180941e-06, + "loss": 0.9204, + "step": 17713 + }, + { + "epoch": 0.57, + "learning_rate": 8.340074916258482e-06, + "loss": 0.9326, + "step": 17714 + }, + { + "epoch": 0.57, + "learning_rate": 8.339053491145181e-06, + "loss": 0.8994, + "step": 17715 + }, + { + "epoch": 0.57, + "learning_rate": 8.338032083852001e-06, + "loss": 1.0054, + "step": 17716 + }, + { + "epoch": 0.57, + "learning_rate": 8.337010694389894e-06, + "loss": 1.0581, + "step": 17717 + }, + { + "epoch": 0.57, + "learning_rate": 8.335989322769822e-06, + "loss": 0.9172, + "step": 17718 + }, + { + "epoch": 0.57, + "learning_rate": 8.334967969002743e-06, + "loss": 1.022, + "step": 17719 + }, + { + "epoch": 0.57, + "learning_rate": 8.333946633099614e-06, + "loss": 1.0229, + "step": 17720 + }, + { + "epoch": 0.57, + "learning_rate": 8.332925315071394e-06, + "loss": 1.002, + "step": 17721 + }, + { + "epoch": 0.57, + "learning_rate": 8.33190401492904e-06, + "loss": 0.895, + "step": 17722 + }, + { + "epoch": 0.57, + "learning_rate": 8.330882732683512e-06, + "loss": 0.8042, + "step": 17723 + }, + { + "epoch": 0.57, + "learning_rate": 8.329861468345768e-06, + "loss": 1.0093, + "step": 17724 + }, + { + "epoch": 0.57, + "learning_rate": 8.328840221926757e-06, + "loss": 0.9697, + "step": 17725 + }, + { + "epoch": 0.57, + "learning_rate": 8.327818993437442e-06, + "loss": 1.0127, + "step": 17726 + }, + { + "epoch": 0.57, + "learning_rate": 8.326797782888777e-06, + "loss": 1.0024, + "step": 17727 + }, + { + "epoch": 0.57, + "learning_rate": 8.325776590291722e-06, + "loss": 0.9443, + "step": 17728 + }, + { + "epoch": 0.57, + "learning_rate": 8.324755415657228e-06, + "loss": 0.9619, + "step": 17729 + }, + { + "epoch": 0.57, + "learning_rate": 8.323734258996257e-06, + "loss": 0.9668, + "step": 17730 + }, + { + "epoch": 0.57, + "learning_rate": 8.322713120319766e-06, + "loss": 0.9351, + "step": 17731 + }, + { + "epoch": 0.57, + "learning_rate": 8.321691999638703e-06, + "loss": 1.0391, + "step": 17732 + }, + { + "epoch": 0.57, + "learning_rate": 8.320670896964028e-06, + "loss": 1.0029, + "step": 17733 + }, + { + "epoch": 0.57, + "learning_rate": 8.319649812306693e-06, + "loss": 0.959, + "step": 17734 + }, + { + "epoch": 0.57, + "learning_rate": 8.318628745677658e-06, + "loss": 0.9722, + "step": 17735 + }, + { + "epoch": 0.57, + "learning_rate": 8.317607697087877e-06, + "loss": 0.9219, + "step": 17736 + }, + { + "epoch": 0.57, + "learning_rate": 8.316586666548301e-06, + "loss": 0.9966, + "step": 17737 + }, + { + "epoch": 0.57, + "learning_rate": 8.315565654069888e-06, + "loss": 0.9189, + "step": 17738 + }, + { + "epoch": 0.57, + "learning_rate": 8.314544659663593e-06, + "loss": 1.0884, + "step": 17739 + }, + { + "epoch": 0.57, + "learning_rate": 8.313523683340365e-06, + "loss": 1.0278, + "step": 17740 + }, + { + "epoch": 0.57, + "learning_rate": 8.312502725111163e-06, + "loss": 0.8696, + "step": 17741 + }, + { + "epoch": 0.57, + "learning_rate": 8.311481784986942e-06, + "loss": 0.936, + "step": 17742 + }, + { + "epoch": 0.57, + "learning_rate": 8.31046086297865e-06, + "loss": 0.8789, + "step": 17743 + }, + { + "epoch": 0.57, + "learning_rate": 8.309439959097241e-06, + "loss": 0.9438, + "step": 17744 + }, + { + "epoch": 0.57, + "learning_rate": 8.308419073353672e-06, + "loss": 0.8823, + "step": 17745 + }, + { + "epoch": 0.57, + "learning_rate": 8.307398205758895e-06, + "loss": 0.8906, + "step": 17746 + }, + { + "epoch": 0.57, + "learning_rate": 8.306377356323863e-06, + "loss": 0.939, + "step": 17747 + }, + { + "epoch": 0.57, + "learning_rate": 8.305356525059525e-06, + "loss": 1.1025, + "step": 17748 + }, + { + "epoch": 0.57, + "learning_rate": 8.304335711976837e-06, + "loss": 0.854, + "step": 17749 + }, + { + "epoch": 0.57, + "learning_rate": 8.30331491708675e-06, + "loss": 1.0059, + "step": 17750 + }, + { + "epoch": 0.57, + "learning_rate": 8.302294140400216e-06, + "loss": 0.874, + "step": 17751 + }, + { + "epoch": 0.57, + "learning_rate": 8.301273381928185e-06, + "loss": 0.9561, + "step": 17752 + }, + { + "epoch": 0.57, + "learning_rate": 8.300252641681613e-06, + "loss": 0.9966, + "step": 17753 + }, + { + "epoch": 0.57, + "learning_rate": 8.299231919671452e-06, + "loss": 1.0396, + "step": 17754 + }, + { + "epoch": 0.57, + "learning_rate": 8.298211215908647e-06, + "loss": 1.0225, + "step": 17755 + }, + { + "epoch": 0.57, + "learning_rate": 8.297190530404151e-06, + "loss": 0.876, + "step": 17756 + }, + { + "epoch": 0.57, + "learning_rate": 8.296169863168917e-06, + "loss": 0.9775, + "step": 17757 + }, + { + "epoch": 0.57, + "learning_rate": 8.295149214213894e-06, + "loss": 0.9756, + "step": 17758 + }, + { + "epoch": 0.57, + "learning_rate": 8.294128583550033e-06, + "loss": 1.0288, + "step": 17759 + }, + { + "epoch": 0.57, + "learning_rate": 8.293107971188285e-06, + "loss": 0.9458, + "step": 17760 + }, + { + "epoch": 0.57, + "learning_rate": 8.292087377139603e-06, + "loss": 1.0918, + "step": 17761 + }, + { + "epoch": 0.57, + "learning_rate": 8.291066801414929e-06, + "loss": 0.9673, + "step": 17762 + }, + { + "epoch": 0.57, + "learning_rate": 8.290046244025218e-06, + "loss": 1.0298, + "step": 17763 + }, + { + "epoch": 0.57, + "learning_rate": 8.289025704981417e-06, + "loss": 0.9683, + "step": 17764 + }, + { + "epoch": 0.57, + "learning_rate": 8.288005184294477e-06, + "loss": 1.041, + "step": 17765 + }, + { + "epoch": 0.57, + "learning_rate": 8.286984681975347e-06, + "loss": 1.0215, + "step": 17766 + }, + { + "epoch": 0.57, + "learning_rate": 8.285964198034973e-06, + "loss": 0.9912, + "step": 17767 + }, + { + "epoch": 0.57, + "learning_rate": 8.28494373248431e-06, + "loss": 1.0762, + "step": 17768 + }, + { + "epoch": 0.57, + "learning_rate": 8.283923285334304e-06, + "loss": 1.0171, + "step": 17769 + }, + { + "epoch": 0.57, + "learning_rate": 8.282902856595899e-06, + "loss": 0.9805, + "step": 17770 + }, + { + "epoch": 0.57, + "learning_rate": 8.281882446280046e-06, + "loss": 0.8892, + "step": 17771 + }, + { + "epoch": 0.57, + "learning_rate": 8.280862054397693e-06, + "loss": 1.002, + "step": 17772 + }, + { + "epoch": 0.57, + "learning_rate": 8.279841680959785e-06, + "loss": 0.9609, + "step": 17773 + }, + { + "epoch": 0.57, + "learning_rate": 8.278821325977276e-06, + "loss": 0.9038, + "step": 17774 + }, + { + "epoch": 0.57, + "learning_rate": 8.277800989461106e-06, + "loss": 1.0171, + "step": 17775 + }, + { + "epoch": 0.57, + "learning_rate": 8.276780671422231e-06, + "loss": 0.9995, + "step": 17776 + }, + { + "epoch": 0.57, + "learning_rate": 8.27576037187159e-06, + "loss": 0.8413, + "step": 17777 + }, + { + "epoch": 0.57, + "learning_rate": 8.274740090820128e-06, + "loss": 0.9795, + "step": 17778 + }, + { + "epoch": 0.57, + "learning_rate": 8.273719828278796e-06, + "loss": 1.0068, + "step": 17779 + }, + { + "epoch": 0.57, + "learning_rate": 8.272699584258543e-06, + "loss": 0.9653, + "step": 17780 + }, + { + "epoch": 0.57, + "learning_rate": 8.27167935877031e-06, + "loss": 0.9775, + "step": 17781 + }, + { + "epoch": 0.57, + "learning_rate": 8.270659151825045e-06, + "loss": 0.9482, + "step": 17782 + }, + { + "epoch": 0.57, + "learning_rate": 8.269638963433692e-06, + "loss": 0.9023, + "step": 17783 + }, + { + "epoch": 0.57, + "learning_rate": 8.268618793607201e-06, + "loss": 0.8833, + "step": 17784 + }, + { + "epoch": 0.57, + "learning_rate": 8.267598642356512e-06, + "loss": 0.936, + "step": 17785 + }, + { + "epoch": 0.57, + "learning_rate": 8.266578509692574e-06, + "loss": 0.9819, + "step": 17786 + }, + { + "epoch": 0.57, + "learning_rate": 8.26555839562633e-06, + "loss": 1.0015, + "step": 17787 + }, + { + "epoch": 0.57, + "learning_rate": 8.264538300168721e-06, + "loss": 0.9922, + "step": 17788 + }, + { + "epoch": 0.57, + "learning_rate": 8.263518223330698e-06, + "loss": 0.9048, + "step": 17789 + }, + { + "epoch": 0.57, + "learning_rate": 8.262498165123202e-06, + "loss": 0.8843, + "step": 17790 + }, + { + "epoch": 0.57, + "learning_rate": 8.261478125557176e-06, + "loss": 1.0405, + "step": 17791 + }, + { + "epoch": 0.57, + "learning_rate": 8.260458104643574e-06, + "loss": 0.9824, + "step": 17792 + }, + { + "epoch": 0.57, + "learning_rate": 8.259438102393322e-06, + "loss": 1.2144, + "step": 17793 + }, + { + "epoch": 0.57, + "learning_rate": 8.258418118817377e-06, + "loss": 0.8989, + "step": 17794 + }, + { + "epoch": 0.57, + "learning_rate": 8.257398153926676e-06, + "loss": 1.0117, + "step": 17795 + }, + { + "epoch": 0.57, + "learning_rate": 8.256378207732165e-06, + "loss": 1.0088, + "step": 17796 + }, + { + "epoch": 0.57, + "learning_rate": 8.255358280244785e-06, + "loss": 0.9434, + "step": 17797 + }, + { + "epoch": 0.57, + "learning_rate": 8.25433837147548e-06, + "loss": 1.0205, + "step": 17798 + }, + { + "epoch": 0.57, + "learning_rate": 8.253318481435198e-06, + "loss": 0.5154, + "step": 17799 + }, + { + "epoch": 0.57, + "learning_rate": 8.252298610134871e-06, + "loss": 0.9746, + "step": 17800 + }, + { + "epoch": 0.57, + "learning_rate": 8.251278757585443e-06, + "loss": 0.9087, + "step": 17801 + }, + { + "epoch": 0.57, + "learning_rate": 8.25025892379786e-06, + "loss": 0.8689, + "step": 17802 + }, + { + "epoch": 0.57, + "learning_rate": 8.249239108783063e-06, + "loss": 0.5193, + "step": 17803 + }, + { + "epoch": 0.57, + "learning_rate": 8.248219312551993e-06, + "loss": 0.9126, + "step": 17804 + }, + { + "epoch": 0.57, + "learning_rate": 8.24719953511559e-06, + "loss": 0.738, + "step": 17805 + }, + { + "epoch": 0.57, + "learning_rate": 8.246179776484798e-06, + "loss": 1.0605, + "step": 17806 + }, + { + "epoch": 0.57, + "learning_rate": 8.245160036670558e-06, + "loss": 0.8213, + "step": 17807 + }, + { + "epoch": 0.57, + "learning_rate": 8.244140315683805e-06, + "loss": 0.9995, + "step": 17808 + }, + { + "epoch": 0.57, + "learning_rate": 8.243120613535483e-06, + "loss": 0.916, + "step": 17809 + }, + { + "epoch": 0.57, + "learning_rate": 8.242100930236533e-06, + "loss": 0.8945, + "step": 17810 + }, + { + "epoch": 0.57, + "learning_rate": 8.241081265797893e-06, + "loss": 1.1191, + "step": 17811 + }, + { + "epoch": 0.57, + "learning_rate": 8.240061620230507e-06, + "loss": 0.9785, + "step": 17812 + }, + { + "epoch": 0.57, + "learning_rate": 8.23904199354531e-06, + "loss": 0.9268, + "step": 17813 + }, + { + "epoch": 0.57, + "learning_rate": 8.238022385753248e-06, + "loss": 0.9399, + "step": 17814 + }, + { + "epoch": 0.57, + "learning_rate": 8.23700279686525e-06, + "loss": 0.9087, + "step": 17815 + }, + { + "epoch": 0.57, + "learning_rate": 8.235983226892263e-06, + "loss": 0.9946, + "step": 17816 + }, + { + "epoch": 0.57, + "learning_rate": 8.234963675845224e-06, + "loss": 1.1318, + "step": 17817 + }, + { + "epoch": 0.57, + "learning_rate": 8.23394414373507e-06, + "loss": 0.9053, + "step": 17818 + }, + { + "epoch": 0.57, + "learning_rate": 8.232924630572742e-06, + "loss": 0.9004, + "step": 17819 + }, + { + "epoch": 0.57, + "learning_rate": 8.231905136369178e-06, + "loss": 1.0889, + "step": 17820 + }, + { + "epoch": 0.57, + "learning_rate": 8.230885661135312e-06, + "loss": 1.0312, + "step": 17821 + }, + { + "epoch": 0.57, + "learning_rate": 8.22986620488209e-06, + "loss": 0.9463, + "step": 17822 + }, + { + "epoch": 0.57, + "learning_rate": 8.22884676762044e-06, + "loss": 1.0796, + "step": 17823 + }, + { + "epoch": 0.57, + "learning_rate": 8.227827349361306e-06, + "loss": 0.9648, + "step": 17824 + }, + { + "epoch": 0.57, + "learning_rate": 8.226807950115623e-06, + "loss": 1.041, + "step": 17825 + }, + { + "epoch": 0.57, + "learning_rate": 8.225788569894327e-06, + "loss": 0.9019, + "step": 17826 + }, + { + "epoch": 0.57, + "learning_rate": 8.224769208708356e-06, + "loss": 0.9937, + "step": 17827 + }, + { + "epoch": 0.57, + "learning_rate": 8.223749866568647e-06, + "loss": 1.063, + "step": 17828 + }, + { + "epoch": 0.57, + "learning_rate": 8.222730543486138e-06, + "loss": 1.0737, + "step": 17829 + }, + { + "epoch": 0.57, + "learning_rate": 8.221711239471762e-06, + "loss": 0.8384, + "step": 17830 + }, + { + "epoch": 0.57, + "learning_rate": 8.220691954536457e-06, + "loss": 1.001, + "step": 17831 + }, + { + "epoch": 0.57, + "learning_rate": 8.219672688691156e-06, + "loss": 0.9688, + "step": 17832 + }, + { + "epoch": 0.57, + "learning_rate": 8.218653441946797e-06, + "loss": 1.0073, + "step": 17833 + }, + { + "epoch": 0.57, + "learning_rate": 8.217634214314317e-06, + "loss": 0.8418, + "step": 17834 + }, + { + "epoch": 0.57, + "learning_rate": 8.216615005804647e-06, + "loss": 1.0596, + "step": 17835 + }, + { + "epoch": 0.57, + "learning_rate": 8.215595816428725e-06, + "loss": 0.9458, + "step": 17836 + }, + { + "epoch": 0.57, + "learning_rate": 8.214576646197489e-06, + "loss": 0.9438, + "step": 17837 + }, + { + "epoch": 0.57, + "learning_rate": 8.213557495121864e-06, + "loss": 0.9292, + "step": 17838 + }, + { + "epoch": 0.57, + "learning_rate": 8.212538363212792e-06, + "loss": 0.9204, + "step": 17839 + }, + { + "epoch": 0.57, + "learning_rate": 8.211519250481205e-06, + "loss": 0.8191, + "step": 17840 + }, + { + "epoch": 0.57, + "learning_rate": 8.210500156938036e-06, + "loss": 1.022, + "step": 17841 + }, + { + "epoch": 0.57, + "learning_rate": 8.20948108259422e-06, + "loss": 1.0269, + "step": 17842 + }, + { + "epoch": 0.57, + "learning_rate": 8.20846202746069e-06, + "loss": 0.9756, + "step": 17843 + }, + { + "epoch": 0.57, + "learning_rate": 8.207442991548386e-06, + "loss": 0.9897, + "step": 17844 + }, + { + "epoch": 0.57, + "learning_rate": 8.206423974868228e-06, + "loss": 0.9731, + "step": 17845 + }, + { + "epoch": 0.57, + "learning_rate": 8.205404977431158e-06, + "loss": 0.9414, + "step": 17846 + }, + { + "epoch": 0.57, + "learning_rate": 8.204385999248105e-06, + "loss": 0.9976, + "step": 17847 + }, + { + "epoch": 0.57, + "learning_rate": 8.203367040330003e-06, + "loss": 1.1304, + "step": 17848 + }, + { + "epoch": 0.57, + "learning_rate": 8.202348100687783e-06, + "loss": 0.9854, + "step": 17849 + }, + { + "epoch": 0.57, + "learning_rate": 8.201329180332381e-06, + "loss": 1.0103, + "step": 17850 + }, + { + "epoch": 0.57, + "learning_rate": 8.200310279274724e-06, + "loss": 0.7998, + "step": 17851 + }, + { + "epoch": 0.57, + "learning_rate": 8.199291397525752e-06, + "loss": 0.9082, + "step": 17852 + }, + { + "epoch": 0.57, + "learning_rate": 8.198272535096386e-06, + "loss": 0.9565, + "step": 17853 + }, + { + "epoch": 0.57, + "learning_rate": 8.19725369199756e-06, + "loss": 0.8892, + "step": 17854 + }, + { + "epoch": 0.57, + "learning_rate": 8.196234868240207e-06, + "loss": 0.9414, + "step": 17855 + }, + { + "epoch": 0.57, + "learning_rate": 8.195216063835259e-06, + "loss": 1.1255, + "step": 17856 + }, + { + "epoch": 0.57, + "learning_rate": 8.194197278793644e-06, + "loss": 0.9058, + "step": 17857 + }, + { + "epoch": 0.57, + "learning_rate": 8.193178513126294e-06, + "loss": 1.0571, + "step": 17858 + }, + { + "epoch": 0.57, + "learning_rate": 8.19215976684414e-06, + "loss": 0.8638, + "step": 17859 + }, + { + "epoch": 0.57, + "learning_rate": 8.191141039958113e-06, + "loss": 0.8979, + "step": 17860 + }, + { + "epoch": 0.57, + "learning_rate": 8.190122332479138e-06, + "loss": 0.9648, + "step": 17861 + }, + { + "epoch": 0.57, + "learning_rate": 8.189103644418146e-06, + "loss": 0.9688, + "step": 17862 + }, + { + "epoch": 0.57, + "learning_rate": 8.188084975786068e-06, + "loss": 1.0371, + "step": 17863 + }, + { + "epoch": 0.57, + "learning_rate": 8.187066326593836e-06, + "loss": 1.0498, + "step": 17864 + }, + { + "epoch": 0.57, + "learning_rate": 8.186047696852372e-06, + "loss": 0.876, + "step": 17865 + }, + { + "epoch": 0.57, + "learning_rate": 8.18502908657261e-06, + "loss": 0.9902, + "step": 17866 + }, + { + "epoch": 0.57, + "learning_rate": 8.18401049576548e-06, + "loss": 1.0366, + "step": 17867 + }, + { + "epoch": 0.57, + "learning_rate": 8.182991924441904e-06, + "loss": 0.9873, + "step": 17868 + }, + { + "epoch": 0.57, + "learning_rate": 8.181973372612814e-06, + "loss": 1.0518, + "step": 17869 + }, + { + "epoch": 0.57, + "learning_rate": 8.18095484028914e-06, + "loss": 0.9619, + "step": 17870 + }, + { + "epoch": 0.57, + "learning_rate": 8.179936327481805e-06, + "loss": 1.0073, + "step": 17871 + }, + { + "epoch": 0.57, + "learning_rate": 8.17891783420174e-06, + "loss": 0.9048, + "step": 17872 + }, + { + "epoch": 0.57, + "learning_rate": 8.17789936045987e-06, + "loss": 1.0737, + "step": 17873 + }, + { + "epoch": 0.57, + "learning_rate": 8.176880906267123e-06, + "loss": 0.9556, + "step": 17874 + }, + { + "epoch": 0.57, + "learning_rate": 8.17586247163443e-06, + "loss": 0.8857, + "step": 17875 + }, + { + "epoch": 0.57, + "learning_rate": 8.174844056572711e-06, + "loss": 0.9463, + "step": 17876 + }, + { + "epoch": 0.57, + "learning_rate": 8.173825661092896e-06, + "loss": 0.8271, + "step": 17877 + }, + { + "epoch": 0.57, + "learning_rate": 8.172807285205909e-06, + "loss": 0.9746, + "step": 17878 + }, + { + "epoch": 0.57, + "learning_rate": 8.171788928922677e-06, + "loss": 1.0527, + "step": 17879 + }, + { + "epoch": 0.57, + "learning_rate": 8.170770592254127e-06, + "loss": 0.9614, + "step": 17880 + }, + { + "epoch": 0.57, + "learning_rate": 8.169752275211184e-06, + "loss": 0.9604, + "step": 17881 + }, + { + "epoch": 0.57, + "learning_rate": 8.168733977804779e-06, + "loss": 1.0356, + "step": 17882 + }, + { + "epoch": 0.57, + "learning_rate": 8.167715700045827e-06, + "loss": 1.0054, + "step": 17883 + }, + { + "epoch": 0.57, + "learning_rate": 8.166697441945257e-06, + "loss": 0.9917, + "step": 17884 + }, + { + "epoch": 0.57, + "learning_rate": 8.165679203513994e-06, + "loss": 0.9595, + "step": 17885 + }, + { + "epoch": 0.57, + "learning_rate": 8.164660984762965e-06, + "loss": 0.9565, + "step": 17886 + }, + { + "epoch": 0.57, + "learning_rate": 8.16364278570309e-06, + "loss": 0.9551, + "step": 17887 + }, + { + "epoch": 0.57, + "learning_rate": 8.162624606345296e-06, + "loss": 1.0586, + "step": 17888 + }, + { + "epoch": 0.57, + "learning_rate": 8.161606446700507e-06, + "loss": 0.8794, + "step": 17889 + }, + { + "epoch": 0.57, + "learning_rate": 8.160588306779651e-06, + "loss": 1.0112, + "step": 17890 + }, + { + "epoch": 0.57, + "learning_rate": 8.159570186593642e-06, + "loss": 1.0039, + "step": 17891 + }, + { + "epoch": 0.57, + "learning_rate": 8.158552086153407e-06, + "loss": 1.1143, + "step": 17892 + }, + { + "epoch": 0.57, + "learning_rate": 8.157534005469872e-06, + "loss": 0.9521, + "step": 17893 + }, + { + "epoch": 0.57, + "learning_rate": 8.156515944553956e-06, + "loss": 0.9326, + "step": 17894 + }, + { + "epoch": 0.57, + "learning_rate": 8.155497903416585e-06, + "loss": 0.8931, + "step": 17895 + }, + { + "epoch": 0.57, + "learning_rate": 8.15447988206868e-06, + "loss": 0.9819, + "step": 17896 + }, + { + "epoch": 0.57, + "learning_rate": 8.153461880521168e-06, + "loss": 1.0396, + "step": 17897 + }, + { + "epoch": 0.57, + "learning_rate": 8.15244389878496e-06, + "loss": 0.9702, + "step": 17898 + }, + { + "epoch": 0.57, + "learning_rate": 8.151425936870988e-06, + "loss": 0.9868, + "step": 17899 + }, + { + "epoch": 0.57, + "learning_rate": 8.150407994790166e-06, + "loss": 0.8955, + "step": 17900 + }, + { + "epoch": 0.57, + "learning_rate": 8.14939007255342e-06, + "loss": 0.9795, + "step": 17901 + }, + { + "epoch": 0.57, + "learning_rate": 8.148372170171672e-06, + "loss": 0.9487, + "step": 17902 + }, + { + "epoch": 0.57, + "learning_rate": 8.14735428765584e-06, + "loss": 0.9136, + "step": 17903 + }, + { + "epoch": 0.57, + "learning_rate": 8.146336425016849e-06, + "loss": 0.9971, + "step": 17904 + }, + { + "epoch": 0.57, + "learning_rate": 8.145318582265616e-06, + "loss": 1.0259, + "step": 17905 + }, + { + "epoch": 0.57, + "learning_rate": 8.14430075941306e-06, + "loss": 0.9976, + "step": 17906 + }, + { + "epoch": 0.57, + "learning_rate": 8.143282956470102e-06, + "loss": 0.939, + "step": 17907 + }, + { + "epoch": 0.57, + "learning_rate": 8.142265173447664e-06, + "loss": 1.0459, + "step": 17908 + }, + { + "epoch": 0.57, + "learning_rate": 8.141247410356668e-06, + "loss": 1.042, + "step": 17909 + }, + { + "epoch": 0.57, + "learning_rate": 8.140229667208027e-06, + "loss": 0.9702, + "step": 17910 + }, + { + "epoch": 0.57, + "learning_rate": 8.139211944012663e-06, + "loss": 0.9863, + "step": 17911 + }, + { + "epoch": 0.57, + "learning_rate": 8.138194240781497e-06, + "loss": 0.9932, + "step": 17912 + }, + { + "epoch": 0.57, + "learning_rate": 8.137176557525446e-06, + "loss": 0.938, + "step": 17913 + }, + { + "epoch": 0.57, + "learning_rate": 8.136158894255429e-06, + "loss": 1.0659, + "step": 17914 + }, + { + "epoch": 0.57, + "learning_rate": 8.135141250982364e-06, + "loss": 1.083, + "step": 17915 + }, + { + "epoch": 0.57, + "learning_rate": 8.134123627717168e-06, + "loss": 0.9058, + "step": 17916 + }, + { + "epoch": 0.57, + "learning_rate": 8.13310602447076e-06, + "loss": 0.9502, + "step": 17917 + }, + { + "epoch": 0.57, + "learning_rate": 8.132088441254059e-06, + "loss": 0.9648, + "step": 17918 + }, + { + "epoch": 0.57, + "learning_rate": 8.131070878077982e-06, + "loss": 1.0107, + "step": 17919 + }, + { + "epoch": 0.57, + "learning_rate": 8.13005333495345e-06, + "loss": 0.8843, + "step": 17920 + }, + { + "epoch": 0.57, + "learning_rate": 8.129035811891373e-06, + "loss": 0.5254, + "step": 17921 + }, + { + "epoch": 0.57, + "learning_rate": 8.12801830890267e-06, + "loss": 0.9824, + "step": 17922 + }, + { + "epoch": 0.57, + "learning_rate": 8.127000825998258e-06, + "loss": 1.0127, + "step": 17923 + }, + { + "epoch": 0.57, + "learning_rate": 8.125983363189056e-06, + "loss": 0.9438, + "step": 17924 + }, + { + "epoch": 0.57, + "learning_rate": 8.124965920485976e-06, + "loss": 0.9829, + "step": 17925 + }, + { + "epoch": 0.57, + "learning_rate": 8.123948497899939e-06, + "loss": 1.0039, + "step": 17926 + }, + { + "epoch": 0.57, + "learning_rate": 8.122931095441861e-06, + "loss": 0.5364, + "step": 17927 + }, + { + "epoch": 0.57, + "learning_rate": 8.12191371312265e-06, + "loss": 1.0786, + "step": 17928 + }, + { + "epoch": 0.57, + "learning_rate": 8.120896350953229e-06, + "loss": 0.9619, + "step": 17929 + }, + { + "epoch": 0.57, + "learning_rate": 8.11987900894451e-06, + "loss": 0.8477, + "step": 17930 + }, + { + "epoch": 0.57, + "learning_rate": 8.118861687107405e-06, + "loss": 0.9673, + "step": 17931 + }, + { + "epoch": 0.57, + "learning_rate": 8.117844385452835e-06, + "loss": 0.9458, + "step": 17932 + }, + { + "epoch": 0.57, + "learning_rate": 8.116827103991711e-06, + "loss": 0.8647, + "step": 17933 + }, + { + "epoch": 0.57, + "learning_rate": 8.115809842734949e-06, + "loss": 0.9756, + "step": 17934 + }, + { + "epoch": 0.57, + "learning_rate": 8.114792601693465e-06, + "loss": 0.8728, + "step": 17935 + }, + { + "epoch": 0.57, + "learning_rate": 8.113775380878167e-06, + "loss": 0.9409, + "step": 17936 + }, + { + "epoch": 0.57, + "learning_rate": 8.11275818029997e-06, + "loss": 1.0225, + "step": 17937 + }, + { + "epoch": 0.57, + "learning_rate": 8.111740999969789e-06, + "loss": 1.0288, + "step": 17938 + }, + { + "epoch": 0.57, + "learning_rate": 8.11072383989854e-06, + "loss": 1.062, + "step": 17939 + }, + { + "epoch": 0.57, + "learning_rate": 8.10970670009713e-06, + "loss": 1.0366, + "step": 17940 + }, + { + "epoch": 0.57, + "learning_rate": 8.108689580576476e-06, + "loss": 0.8726, + "step": 17941 + }, + { + "epoch": 0.57, + "learning_rate": 8.107672481347491e-06, + "loss": 0.9365, + "step": 17942 + }, + { + "epoch": 0.57, + "learning_rate": 8.10665540242109e-06, + "loss": 0.9795, + "step": 17943 + }, + { + "epoch": 0.57, + "learning_rate": 8.105638343808174e-06, + "loss": 0.9414, + "step": 17944 + }, + { + "epoch": 0.57, + "learning_rate": 8.104621305519664e-06, + "loss": 1.0298, + "step": 17945 + }, + { + "epoch": 0.57, + "learning_rate": 8.10360428756647e-06, + "loss": 0.8911, + "step": 17946 + }, + { + "epoch": 0.57, + "learning_rate": 8.102587289959502e-06, + "loss": 0.9536, + "step": 17947 + }, + { + "epoch": 0.57, + "learning_rate": 8.101570312709677e-06, + "loss": 0.9771, + "step": 17948 + }, + { + "epoch": 0.57, + "learning_rate": 8.100553355827897e-06, + "loss": 0.978, + "step": 17949 + }, + { + "epoch": 0.57, + "learning_rate": 8.09953641932508e-06, + "loss": 0.9575, + "step": 17950 + }, + { + "epoch": 0.57, + "learning_rate": 8.098519503212132e-06, + "loss": 1.019, + "step": 17951 + }, + { + "epoch": 0.57, + "learning_rate": 8.097502607499965e-06, + "loss": 1.0317, + "step": 17952 + }, + { + "epoch": 0.57, + "learning_rate": 8.096485732199491e-06, + "loss": 0.9546, + "step": 17953 + }, + { + "epoch": 0.57, + "learning_rate": 8.095468877321617e-06, + "loss": 0.8721, + "step": 17954 + }, + { + "epoch": 0.57, + "learning_rate": 8.094452042877254e-06, + "loss": 0.9399, + "step": 17955 + }, + { + "epoch": 0.57, + "learning_rate": 8.093435228877312e-06, + "loss": 0.9004, + "step": 17956 + }, + { + "epoch": 0.57, + "learning_rate": 8.0924184353327e-06, + "loss": 0.8652, + "step": 17957 + }, + { + "epoch": 0.57, + "learning_rate": 8.091401662254329e-06, + "loss": 0.9844, + "step": 17958 + }, + { + "epoch": 0.57, + "learning_rate": 8.090384909653105e-06, + "loss": 0.979, + "step": 17959 + }, + { + "epoch": 0.57, + "learning_rate": 8.089368177539935e-06, + "loss": 0.8975, + "step": 17960 + }, + { + "epoch": 0.57, + "learning_rate": 8.088351465925732e-06, + "loss": 1.0063, + "step": 17961 + }, + { + "epoch": 0.57, + "learning_rate": 8.0873347748214e-06, + "loss": 0.895, + "step": 17962 + }, + { + "epoch": 0.57, + "learning_rate": 8.086318104237852e-06, + "loss": 0.5244, + "step": 17963 + }, + { + "epoch": 0.57, + "learning_rate": 8.08530145418599e-06, + "loss": 0.9385, + "step": 17964 + }, + { + "epoch": 0.57, + "learning_rate": 8.084284824676728e-06, + "loss": 1.1411, + "step": 17965 + }, + { + "epoch": 0.57, + "learning_rate": 8.083268215720967e-06, + "loss": 0.9849, + "step": 17966 + }, + { + "epoch": 0.57, + "learning_rate": 8.082251627329617e-06, + "loss": 0.8755, + "step": 17967 + }, + { + "epoch": 0.57, + "learning_rate": 8.081235059513584e-06, + "loss": 1.0718, + "step": 17968 + }, + { + "epoch": 0.57, + "learning_rate": 8.080218512283775e-06, + "loss": 0.936, + "step": 17969 + }, + { + "epoch": 0.57, + "learning_rate": 8.079201985651097e-06, + "loss": 0.9146, + "step": 17970 + }, + { + "epoch": 0.57, + "learning_rate": 8.078185479626454e-06, + "loss": 0.9971, + "step": 17971 + }, + { + "epoch": 0.57, + "learning_rate": 8.077168994220757e-06, + "loss": 0.9316, + "step": 17972 + }, + { + "epoch": 0.57, + "learning_rate": 8.07615252944491e-06, + "loss": 0.9468, + "step": 17973 + }, + { + "epoch": 0.57, + "learning_rate": 8.075136085309814e-06, + "loss": 0.9282, + "step": 17974 + }, + { + "epoch": 0.57, + "learning_rate": 8.074119661826378e-06, + "loss": 0.9482, + "step": 17975 + }, + { + "epoch": 0.57, + "learning_rate": 8.073103259005506e-06, + "loss": 1.0132, + "step": 17976 + }, + { + "epoch": 0.57, + "learning_rate": 8.072086876858106e-06, + "loss": 0.9202, + "step": 17977 + }, + { + "epoch": 0.57, + "learning_rate": 8.071070515395077e-06, + "loss": 0.9541, + "step": 17978 + }, + { + "epoch": 0.57, + "learning_rate": 8.070054174627326e-06, + "loss": 0.874, + "step": 17979 + }, + { + "epoch": 0.58, + "learning_rate": 8.069037854565765e-06, + "loss": 0.9585, + "step": 17980 + }, + { + "epoch": 0.58, + "learning_rate": 8.068021555221283e-06, + "loss": 0.9795, + "step": 17981 + }, + { + "epoch": 0.58, + "learning_rate": 8.067005276604795e-06, + "loss": 0.9878, + "step": 17982 + }, + { + "epoch": 0.58, + "learning_rate": 8.065989018727199e-06, + "loss": 1.0303, + "step": 17983 + }, + { + "epoch": 0.58, + "learning_rate": 8.064972781599401e-06, + "loss": 1.1133, + "step": 17984 + }, + { + "epoch": 0.58, + "learning_rate": 8.063956565232303e-06, + "loss": 1.022, + "step": 17985 + }, + { + "epoch": 0.58, + "learning_rate": 8.062940369636808e-06, + "loss": 1.0322, + "step": 17986 + }, + { + "epoch": 0.58, + "learning_rate": 8.061924194823823e-06, + "loss": 0.957, + "step": 17987 + }, + { + "epoch": 0.58, + "learning_rate": 8.060908040804245e-06, + "loss": 0.9912, + "step": 17988 + }, + { + "epoch": 0.58, + "learning_rate": 8.059891907588975e-06, + "loss": 0.9712, + "step": 17989 + }, + { + "epoch": 0.58, + "learning_rate": 8.058875795188918e-06, + "loss": 1.0381, + "step": 17990 + }, + { + "epoch": 0.58, + "learning_rate": 8.057859703614976e-06, + "loss": 0.9854, + "step": 17991 + }, + { + "epoch": 0.58, + "learning_rate": 8.056843632878049e-06, + "loss": 1.0208, + "step": 17992 + }, + { + "epoch": 0.58, + "learning_rate": 8.055827582989043e-06, + "loss": 0.9224, + "step": 17993 + }, + { + "epoch": 0.58, + "learning_rate": 8.054811553958853e-06, + "loss": 0.981, + "step": 17994 + }, + { + "epoch": 0.58, + "learning_rate": 8.053795545798382e-06, + "loss": 0.8721, + "step": 17995 + }, + { + "epoch": 0.58, + "learning_rate": 8.05277955851853e-06, + "loss": 0.998, + "step": 17996 + }, + { + "epoch": 0.58, + "learning_rate": 8.0517635921302e-06, + "loss": 1.0156, + "step": 17997 + }, + { + "epoch": 0.58, + "learning_rate": 8.05074764664429e-06, + "loss": 1.0347, + "step": 17998 + }, + { + "epoch": 0.58, + "learning_rate": 8.0497317220717e-06, + "loss": 0.9834, + "step": 17999 + }, + { + "epoch": 0.58, + "learning_rate": 8.04871581842333e-06, + "loss": 1.0029, + "step": 18000 + }, + { + "epoch": 0.58, + "learning_rate": 8.047699935710078e-06, + "loss": 0.832, + "step": 18001 + }, + { + "epoch": 0.58, + "learning_rate": 8.046684073942846e-06, + "loss": 1.0103, + "step": 18002 + }, + { + "epoch": 0.58, + "learning_rate": 8.045668233132538e-06, + "loss": 0.5398, + "step": 18003 + }, + { + "epoch": 0.58, + "learning_rate": 8.044652413290041e-06, + "loss": 0.8801, + "step": 18004 + }, + { + "epoch": 0.58, + "learning_rate": 8.043636614426259e-06, + "loss": 0.9868, + "step": 18005 + }, + { + "epoch": 0.58, + "learning_rate": 8.042620836552091e-06, + "loss": 1.063, + "step": 18006 + }, + { + "epoch": 0.58, + "learning_rate": 8.041605079678436e-06, + "loss": 0.9688, + "step": 18007 + }, + { + "epoch": 0.58, + "learning_rate": 8.040589343816191e-06, + "loss": 0.9485, + "step": 18008 + }, + { + "epoch": 0.58, + "learning_rate": 8.039573628976254e-06, + "loss": 0.9541, + "step": 18009 + }, + { + "epoch": 0.58, + "learning_rate": 8.038557935169523e-06, + "loss": 1.0049, + "step": 18010 + }, + { + "epoch": 0.58, + "learning_rate": 8.037542262406896e-06, + "loss": 0.8098, + "step": 18011 + }, + { + "epoch": 0.58, + "learning_rate": 8.036526610699267e-06, + "loss": 0.9512, + "step": 18012 + }, + { + "epoch": 0.58, + "learning_rate": 8.035510980057532e-06, + "loss": 1.0027, + "step": 18013 + }, + { + "epoch": 0.58, + "learning_rate": 8.034495370492593e-06, + "loss": 0.9419, + "step": 18014 + }, + { + "epoch": 0.58, + "learning_rate": 8.033479782015341e-06, + "loss": 1.1011, + "step": 18015 + }, + { + "epoch": 0.58, + "learning_rate": 8.032464214636675e-06, + "loss": 0.9668, + "step": 18016 + }, + { + "epoch": 0.58, + "learning_rate": 8.03144866836749e-06, + "loss": 0.9546, + "step": 18017 + }, + { + "epoch": 0.58, + "learning_rate": 8.030433143218688e-06, + "loss": 0.8647, + "step": 18018 + }, + { + "epoch": 0.58, + "learning_rate": 8.029417639201153e-06, + "loss": 0.9756, + "step": 18019 + }, + { + "epoch": 0.58, + "learning_rate": 8.028402156325787e-06, + "loss": 0.9971, + "step": 18020 + }, + { + "epoch": 0.58, + "learning_rate": 8.027386694603482e-06, + "loss": 0.9551, + "step": 18021 + }, + { + "epoch": 0.58, + "learning_rate": 8.026371254045136e-06, + "loss": 1.0718, + "step": 18022 + }, + { + "epoch": 0.58, + "learning_rate": 8.025355834661641e-06, + "loss": 1.0845, + "step": 18023 + }, + { + "epoch": 0.58, + "learning_rate": 8.024340436463894e-06, + "loss": 0.9177, + "step": 18024 + }, + { + "epoch": 0.58, + "learning_rate": 8.023325059462785e-06, + "loss": 0.9526, + "step": 18025 + }, + { + "epoch": 0.58, + "learning_rate": 8.022309703669218e-06, + "loss": 0.8423, + "step": 18026 + }, + { + "epoch": 0.58, + "learning_rate": 8.021294369094072e-06, + "loss": 1.0308, + "step": 18027 + }, + { + "epoch": 0.58, + "learning_rate": 8.020279055748249e-06, + "loss": 1.0298, + "step": 18028 + }, + { + "epoch": 0.58, + "learning_rate": 8.01926376364264e-06, + "loss": 1.061, + "step": 18029 + }, + { + "epoch": 0.58, + "learning_rate": 8.01824849278814e-06, + "loss": 0.9419, + "step": 18030 + }, + { + "epoch": 0.58, + "learning_rate": 8.01723324319564e-06, + "loss": 0.9702, + "step": 18031 + }, + { + "epoch": 0.58, + "learning_rate": 8.016218014876034e-06, + "loss": 0.9629, + "step": 18032 + }, + { + "epoch": 0.58, + "learning_rate": 8.015202807840214e-06, + "loss": 1.0244, + "step": 18033 + }, + { + "epoch": 0.58, + "learning_rate": 8.014187622099067e-06, + "loss": 0.9609, + "step": 18034 + }, + { + "epoch": 0.58, + "learning_rate": 8.013172457663492e-06, + "loss": 0.9976, + "step": 18035 + }, + { + "epoch": 0.58, + "learning_rate": 8.012157314544375e-06, + "loss": 0.9956, + "step": 18036 + }, + { + "epoch": 0.58, + "learning_rate": 8.011142192752615e-06, + "loss": 0.9814, + "step": 18037 + }, + { + "epoch": 0.58, + "learning_rate": 8.010127092299093e-06, + "loss": 1.0342, + "step": 18038 + }, + { + "epoch": 0.58, + "learning_rate": 8.009112013194707e-06, + "loss": 0.9932, + "step": 18039 + }, + { + "epoch": 0.58, + "learning_rate": 8.008096955450345e-06, + "loss": 1.0742, + "step": 18040 + }, + { + "epoch": 0.58, + "learning_rate": 8.0070819190769e-06, + "loss": 0.9399, + "step": 18041 + }, + { + "epoch": 0.58, + "learning_rate": 8.006066904085261e-06, + "loss": 1.083, + "step": 18042 + }, + { + "epoch": 0.58, + "learning_rate": 8.005051910486315e-06, + "loss": 0.9678, + "step": 18043 + }, + { + "epoch": 0.58, + "learning_rate": 8.004036938290954e-06, + "loss": 0.9404, + "step": 18044 + }, + { + "epoch": 0.58, + "learning_rate": 8.003021987510069e-06, + "loss": 0.9951, + "step": 18045 + }, + { + "epoch": 0.58, + "learning_rate": 8.002007058154544e-06, + "loss": 1.0498, + "step": 18046 + }, + { + "epoch": 0.58, + "learning_rate": 8.000992150235276e-06, + "loss": 0.9561, + "step": 18047 + }, + { + "epoch": 0.58, + "learning_rate": 7.999977263763153e-06, + "loss": 0.9854, + "step": 18048 + }, + { + "epoch": 0.58, + "learning_rate": 7.998962398749054e-06, + "loss": 0.8848, + "step": 18049 + }, + { + "epoch": 0.58, + "learning_rate": 7.997947555203878e-06, + "loss": 0.9575, + "step": 18050 + }, + { + "epoch": 0.58, + "learning_rate": 7.996932733138506e-06, + "loss": 0.917, + "step": 18051 + }, + { + "epoch": 0.58, + "learning_rate": 7.99591793256383e-06, + "loss": 0.9951, + "step": 18052 + }, + { + "epoch": 0.58, + "learning_rate": 7.994903153490736e-06, + "loss": 0.9878, + "step": 18053 + }, + { + "epoch": 0.58, + "learning_rate": 7.993888395930112e-06, + "loss": 1.0356, + "step": 18054 + }, + { + "epoch": 0.58, + "learning_rate": 7.992873659892844e-06, + "loss": 0.9448, + "step": 18055 + }, + { + "epoch": 0.58, + "learning_rate": 7.991858945389828e-06, + "loss": 1.0024, + "step": 18056 + }, + { + "epoch": 0.58, + "learning_rate": 7.990844252431936e-06, + "loss": 0.939, + "step": 18057 + }, + { + "epoch": 0.58, + "learning_rate": 7.989829581030062e-06, + "loss": 0.9395, + "step": 18058 + }, + { + "epoch": 0.58, + "learning_rate": 7.988814931195092e-06, + "loss": 1.0186, + "step": 18059 + }, + { + "epoch": 0.58, + "learning_rate": 7.987800302937912e-06, + "loss": 0.9419, + "step": 18060 + }, + { + "epoch": 0.58, + "learning_rate": 7.986785696269407e-06, + "loss": 1.0112, + "step": 18061 + }, + { + "epoch": 0.58, + "learning_rate": 7.985771111200464e-06, + "loss": 0.9565, + "step": 18062 + }, + { + "epoch": 0.58, + "learning_rate": 7.984756547741973e-06, + "loss": 1.1079, + "step": 18063 + }, + { + "epoch": 0.58, + "learning_rate": 7.983742005904808e-06, + "loss": 0.8535, + "step": 18064 + }, + { + "epoch": 0.58, + "learning_rate": 7.982727485699862e-06, + "loss": 0.9829, + "step": 18065 + }, + { + "epoch": 0.58, + "learning_rate": 7.981712987138015e-06, + "loss": 0.9302, + "step": 18066 + }, + { + "epoch": 0.58, + "learning_rate": 7.980698510230156e-06, + "loss": 0.9941, + "step": 18067 + }, + { + "epoch": 0.58, + "learning_rate": 7.979684054987165e-06, + "loss": 1.0439, + "step": 18068 + }, + { + "epoch": 0.58, + "learning_rate": 7.97866962141993e-06, + "loss": 1.0093, + "step": 18069 + }, + { + "epoch": 0.58, + "learning_rate": 7.977655209539332e-06, + "loss": 0.8184, + "step": 18070 + }, + { + "epoch": 0.58, + "learning_rate": 7.97664081935626e-06, + "loss": 0.8784, + "step": 18071 + }, + { + "epoch": 0.58, + "learning_rate": 7.975626450881587e-06, + "loss": 1.0293, + "step": 18072 + }, + { + "epoch": 0.58, + "learning_rate": 7.974612104126203e-06, + "loss": 1.0391, + "step": 18073 + }, + { + "epoch": 0.58, + "learning_rate": 7.97359777910099e-06, + "loss": 1.0591, + "step": 18074 + }, + { + "epoch": 0.58, + "learning_rate": 7.972583475816829e-06, + "loss": 1.0376, + "step": 18075 + }, + { + "epoch": 0.58, + "learning_rate": 7.971569194284606e-06, + "loss": 0.9678, + "step": 18076 + }, + { + "epoch": 0.58, + "learning_rate": 7.970554934515198e-06, + "loss": 0.8516, + "step": 18077 + }, + { + "epoch": 0.58, + "learning_rate": 7.969540696519494e-06, + "loss": 0.9668, + "step": 18078 + }, + { + "epoch": 0.58, + "learning_rate": 7.968526480308367e-06, + "loss": 0.7651, + "step": 18079 + }, + { + "epoch": 0.58, + "learning_rate": 7.9675122858927e-06, + "loss": 0.939, + "step": 18080 + }, + { + "epoch": 0.58, + "learning_rate": 7.966498113283381e-06, + "loss": 0.9805, + "step": 18081 + }, + { + "epoch": 0.58, + "learning_rate": 7.965483962491287e-06, + "loss": 1.0337, + "step": 18082 + }, + { + "epoch": 0.58, + "learning_rate": 7.964469833527297e-06, + "loss": 0.8569, + "step": 18083 + }, + { + "epoch": 0.58, + "learning_rate": 7.963455726402292e-06, + "loss": 0.9316, + "step": 18084 + }, + { + "epoch": 0.58, + "learning_rate": 7.962441641127153e-06, + "loss": 0.9106, + "step": 18085 + }, + { + "epoch": 0.58, + "learning_rate": 7.961427577712764e-06, + "loss": 0.9365, + "step": 18086 + }, + { + "epoch": 0.58, + "learning_rate": 7.960413536169999e-06, + "loss": 0.9766, + "step": 18087 + }, + { + "epoch": 0.58, + "learning_rate": 7.959399516509738e-06, + "loss": 0.9116, + "step": 18088 + }, + { + "epoch": 0.58, + "learning_rate": 7.958385518742862e-06, + "loss": 0.9531, + "step": 18089 + }, + { + "epoch": 0.58, + "learning_rate": 7.957371542880248e-06, + "loss": 0.8623, + "step": 18090 + }, + { + "epoch": 0.58, + "learning_rate": 7.956357588932779e-06, + "loss": 0.5833, + "step": 18091 + }, + { + "epoch": 0.58, + "learning_rate": 7.955343656911331e-06, + "loss": 1.001, + "step": 18092 + }, + { + "epoch": 0.58, + "learning_rate": 7.954329746826782e-06, + "loss": 0.9873, + "step": 18093 + }, + { + "epoch": 0.58, + "learning_rate": 7.953315858690015e-06, + "loss": 1.0234, + "step": 18094 + }, + { + "epoch": 0.58, + "learning_rate": 7.952301992511899e-06, + "loss": 0.9287, + "step": 18095 + }, + { + "epoch": 0.58, + "learning_rate": 7.951288148303318e-06, + "loss": 0.9355, + "step": 18096 + }, + { + "epoch": 0.58, + "learning_rate": 7.950274326075146e-06, + "loss": 1.1104, + "step": 18097 + }, + { + "epoch": 0.58, + "learning_rate": 7.949260525838263e-06, + "loss": 0.8845, + "step": 18098 + }, + { + "epoch": 0.58, + "learning_rate": 7.948246747603545e-06, + "loss": 1.0078, + "step": 18099 + }, + { + "epoch": 0.58, + "learning_rate": 7.947232991381868e-06, + "loss": 1.084, + "step": 18100 + }, + { + "epoch": 0.58, + "learning_rate": 7.946219257184113e-06, + "loss": 0.9795, + "step": 18101 + }, + { + "epoch": 0.58, + "learning_rate": 7.945205545021148e-06, + "loss": 0.9961, + "step": 18102 + }, + { + "epoch": 0.58, + "learning_rate": 7.944191854903855e-06, + "loss": 0.9375, + "step": 18103 + }, + { + "epoch": 0.58, + "learning_rate": 7.943178186843109e-06, + "loss": 1.022, + "step": 18104 + }, + { + "epoch": 0.58, + "learning_rate": 7.942164540849781e-06, + "loss": 1.0239, + "step": 18105 + }, + { + "epoch": 0.58, + "learning_rate": 7.941150916934752e-06, + "loss": 0.876, + "step": 18106 + }, + { + "epoch": 0.58, + "learning_rate": 7.940137315108896e-06, + "loss": 1.0059, + "step": 18107 + }, + { + "epoch": 0.58, + "learning_rate": 7.939123735383086e-06, + "loss": 0.9053, + "step": 18108 + }, + { + "epoch": 0.58, + "learning_rate": 7.9381101777682e-06, + "loss": 0.9653, + "step": 18109 + }, + { + "epoch": 0.58, + "learning_rate": 7.937096642275107e-06, + "loss": 0.8569, + "step": 18110 + }, + { + "epoch": 0.58, + "learning_rate": 7.936083128914685e-06, + "loss": 0.5366, + "step": 18111 + }, + { + "epoch": 0.58, + "learning_rate": 7.935069637697806e-06, + "loss": 0.9536, + "step": 18112 + }, + { + "epoch": 0.58, + "learning_rate": 7.934056168635343e-06, + "loss": 0.8418, + "step": 18113 + }, + { + "epoch": 0.58, + "learning_rate": 7.933042721738172e-06, + "loss": 0.8149, + "step": 18114 + }, + { + "epoch": 0.58, + "learning_rate": 7.932029297017167e-06, + "loss": 1.0972, + "step": 18115 + }, + { + "epoch": 0.58, + "learning_rate": 7.9310158944832e-06, + "loss": 0.917, + "step": 18116 + }, + { + "epoch": 0.58, + "learning_rate": 7.93000251414714e-06, + "loss": 1.0093, + "step": 18117 + }, + { + "epoch": 0.58, + "learning_rate": 7.92898915601986e-06, + "loss": 0.9668, + "step": 18118 + }, + { + "epoch": 0.58, + "learning_rate": 7.927975820112237e-06, + "loss": 1.0708, + "step": 18119 + }, + { + "epoch": 0.58, + "learning_rate": 7.92696250643514e-06, + "loss": 1.0254, + "step": 18120 + }, + { + "epoch": 0.58, + "learning_rate": 7.925949214999441e-06, + "loss": 0.8667, + "step": 18121 + }, + { + "epoch": 0.58, + "learning_rate": 7.92493594581601e-06, + "loss": 0.9443, + "step": 18122 + }, + { + "epoch": 0.58, + "learning_rate": 7.92392269889572e-06, + "loss": 0.5464, + "step": 18123 + }, + { + "epoch": 0.58, + "learning_rate": 7.922909474249446e-06, + "loss": 1.0303, + "step": 18124 + }, + { + "epoch": 0.58, + "learning_rate": 7.92189627188805e-06, + "loss": 0.9585, + "step": 18125 + }, + { + "epoch": 0.58, + "learning_rate": 7.92088309182241e-06, + "loss": 0.957, + "step": 18126 + }, + { + "epoch": 0.58, + "learning_rate": 7.919869934063391e-06, + "loss": 1.0669, + "step": 18127 + }, + { + "epoch": 0.58, + "learning_rate": 7.918856798621864e-06, + "loss": 1.0552, + "step": 18128 + }, + { + "epoch": 0.58, + "learning_rate": 7.917843685508702e-06, + "loss": 0.938, + "step": 18129 + }, + { + "epoch": 0.58, + "learning_rate": 7.916830594734774e-06, + "loss": 1.001, + "step": 18130 + }, + { + "epoch": 0.58, + "learning_rate": 7.91581752631095e-06, + "loss": 1.0386, + "step": 18131 + }, + { + "epoch": 0.58, + "learning_rate": 7.914804480248094e-06, + "loss": 0.9907, + "step": 18132 + }, + { + "epoch": 0.58, + "learning_rate": 7.913791456557078e-06, + "loss": 1.0156, + "step": 18133 + }, + { + "epoch": 0.58, + "learning_rate": 7.912778455248772e-06, + "loss": 1.0513, + "step": 18134 + }, + { + "epoch": 0.58, + "learning_rate": 7.911765476334042e-06, + "loss": 0.9399, + "step": 18135 + }, + { + "epoch": 0.58, + "learning_rate": 7.910752519823756e-06, + "loss": 1.0205, + "step": 18136 + }, + { + "epoch": 0.58, + "learning_rate": 7.909739585728787e-06, + "loss": 1.0093, + "step": 18137 + }, + { + "epoch": 0.58, + "learning_rate": 7.908726674059996e-06, + "loss": 1.0044, + "step": 18138 + }, + { + "epoch": 0.58, + "learning_rate": 7.907713784828256e-06, + "loss": 1.0, + "step": 18139 + }, + { + "epoch": 0.58, + "learning_rate": 7.90670091804443e-06, + "loss": 0.9727, + "step": 18140 + }, + { + "epoch": 0.58, + "learning_rate": 7.905688073719387e-06, + "loss": 1.0234, + "step": 18141 + }, + { + "epoch": 0.58, + "learning_rate": 7.90467525186399e-06, + "loss": 0.9268, + "step": 18142 + }, + { + "epoch": 0.58, + "learning_rate": 7.903662452489112e-06, + "loss": 1.0918, + "step": 18143 + }, + { + "epoch": 0.58, + "learning_rate": 7.902649675605612e-06, + "loss": 1.1079, + "step": 18144 + }, + { + "epoch": 0.58, + "learning_rate": 7.901636921224362e-06, + "loss": 1.001, + "step": 18145 + }, + { + "epoch": 0.58, + "learning_rate": 7.90062418935623e-06, + "loss": 1.0098, + "step": 18146 + }, + { + "epoch": 0.58, + "learning_rate": 7.899611480012073e-06, + "loss": 1.0127, + "step": 18147 + }, + { + "epoch": 0.58, + "learning_rate": 7.898598793202759e-06, + "loss": 1.0386, + "step": 18148 + }, + { + "epoch": 0.58, + "learning_rate": 7.897586128939154e-06, + "loss": 0.8799, + "step": 18149 + }, + { + "epoch": 0.58, + "learning_rate": 7.896573487232124e-06, + "loss": 0.939, + "step": 18150 + }, + { + "epoch": 0.58, + "learning_rate": 7.895560868092533e-06, + "loss": 0.9131, + "step": 18151 + }, + { + "epoch": 0.58, + "learning_rate": 7.894548271531245e-06, + "loss": 0.9609, + "step": 18152 + }, + { + "epoch": 0.58, + "learning_rate": 7.893535697559122e-06, + "loss": 0.9009, + "step": 18153 + }, + { + "epoch": 0.58, + "learning_rate": 7.892523146187035e-06, + "loss": 0.8608, + "step": 18154 + }, + { + "epoch": 0.58, + "learning_rate": 7.891510617425837e-06, + "loss": 0.9834, + "step": 18155 + }, + { + "epoch": 0.58, + "learning_rate": 7.890498111286398e-06, + "loss": 0.8682, + "step": 18156 + }, + { + "epoch": 0.58, + "learning_rate": 7.889485627779579e-06, + "loss": 0.9648, + "step": 18157 + }, + { + "epoch": 0.58, + "learning_rate": 7.888473166916242e-06, + "loss": 1.0195, + "step": 18158 + }, + { + "epoch": 0.58, + "learning_rate": 7.887460728707253e-06, + "loss": 1.0659, + "step": 18159 + }, + { + "epoch": 0.58, + "learning_rate": 7.886448313163472e-06, + "loss": 0.9575, + "step": 18160 + }, + { + "epoch": 0.58, + "learning_rate": 7.885435920295764e-06, + "loss": 1.0098, + "step": 18161 + }, + { + "epoch": 0.58, + "learning_rate": 7.884423550114985e-06, + "loss": 0.8174, + "step": 18162 + }, + { + "epoch": 0.58, + "learning_rate": 7.883411202632e-06, + "loss": 1.0273, + "step": 18163 + }, + { + "epoch": 0.58, + "learning_rate": 7.882398877857671e-06, + "loss": 1.0073, + "step": 18164 + }, + { + "epoch": 0.58, + "learning_rate": 7.88138657580286e-06, + "loss": 0.8721, + "step": 18165 + }, + { + "epoch": 0.58, + "learning_rate": 7.880374296478425e-06, + "loss": 0.8887, + "step": 18166 + }, + { + "epoch": 0.58, + "learning_rate": 7.879362039895227e-06, + "loss": 0.9985, + "step": 18167 + }, + { + "epoch": 0.58, + "learning_rate": 7.878349806064128e-06, + "loss": 1.0806, + "step": 18168 + }, + { + "epoch": 0.58, + "learning_rate": 7.877337594995988e-06, + "loss": 0.9756, + "step": 18169 + }, + { + "epoch": 0.58, + "learning_rate": 7.876325406701668e-06, + "loss": 1.042, + "step": 18170 + }, + { + "epoch": 0.58, + "learning_rate": 7.875313241192023e-06, + "loss": 0.8945, + "step": 18171 + }, + { + "epoch": 0.58, + "learning_rate": 7.874301098477915e-06, + "loss": 0.9756, + "step": 18172 + }, + { + "epoch": 0.58, + "learning_rate": 7.873288978570205e-06, + "loss": 0.9312, + "step": 18173 + }, + { + "epoch": 0.58, + "learning_rate": 7.87227688147975e-06, + "loss": 0.9126, + "step": 18174 + }, + { + "epoch": 0.58, + "learning_rate": 7.871264807217408e-06, + "loss": 0.8794, + "step": 18175 + }, + { + "epoch": 0.58, + "learning_rate": 7.87025275579404e-06, + "loss": 0.937, + "step": 18176 + }, + { + "epoch": 0.58, + "learning_rate": 7.869240727220506e-06, + "loss": 0.9956, + "step": 18177 + }, + { + "epoch": 0.58, + "learning_rate": 7.868228721507659e-06, + "loss": 0.8989, + "step": 18178 + }, + { + "epoch": 0.58, + "learning_rate": 7.867216738666356e-06, + "loss": 1.1143, + "step": 18179 + }, + { + "epoch": 0.58, + "learning_rate": 7.866204778707458e-06, + "loss": 0.9497, + "step": 18180 + }, + { + "epoch": 0.58, + "learning_rate": 7.86519284164182e-06, + "loss": 0.9805, + "step": 18181 + }, + { + "epoch": 0.58, + "learning_rate": 7.864180927480302e-06, + "loss": 1.0254, + "step": 18182 + }, + { + "epoch": 0.58, + "learning_rate": 7.863169036233757e-06, + "loss": 0.9038, + "step": 18183 + }, + { + "epoch": 0.58, + "learning_rate": 7.86215716791305e-06, + "loss": 0.8281, + "step": 18184 + }, + { + "epoch": 0.58, + "learning_rate": 7.861145322529023e-06, + "loss": 1.1157, + "step": 18185 + }, + { + "epoch": 0.58, + "learning_rate": 7.860133500092541e-06, + "loss": 0.9697, + "step": 18186 + }, + { + "epoch": 0.58, + "learning_rate": 7.859121700614459e-06, + "loss": 1.0396, + "step": 18187 + }, + { + "epoch": 0.58, + "learning_rate": 7.858109924105632e-06, + "loss": 0.79, + "step": 18188 + }, + { + "epoch": 0.58, + "learning_rate": 7.857098170576914e-06, + "loss": 0.9888, + "step": 18189 + }, + { + "epoch": 0.58, + "learning_rate": 7.856086440039162e-06, + "loss": 0.8784, + "step": 18190 + }, + { + "epoch": 0.58, + "learning_rate": 7.855074732503228e-06, + "loss": 0.9497, + "step": 18191 + }, + { + "epoch": 0.58, + "learning_rate": 7.854063047979973e-06, + "loss": 0.8857, + "step": 18192 + }, + { + "epoch": 0.58, + "learning_rate": 7.853051386480243e-06, + "loss": 1.0093, + "step": 18193 + }, + { + "epoch": 0.58, + "learning_rate": 7.852039748014894e-06, + "loss": 0.8569, + "step": 18194 + }, + { + "epoch": 0.58, + "learning_rate": 7.851028132594783e-06, + "loss": 0.9907, + "step": 18195 + }, + { + "epoch": 0.58, + "learning_rate": 7.850016540230761e-06, + "loss": 1.0083, + "step": 18196 + }, + { + "epoch": 0.58, + "learning_rate": 7.849004970933682e-06, + "loss": 0.9888, + "step": 18197 + }, + { + "epoch": 0.58, + "learning_rate": 7.8479934247144e-06, + "loss": 1.0039, + "step": 18198 + }, + { + "epoch": 0.58, + "learning_rate": 7.84698190158377e-06, + "loss": 0.8745, + "step": 18199 + }, + { + "epoch": 0.58, + "learning_rate": 7.845970401552639e-06, + "loss": 0.9473, + "step": 18200 + }, + { + "epoch": 0.58, + "learning_rate": 7.844958924631859e-06, + "loss": 0.8027, + "step": 18201 + }, + { + "epoch": 0.58, + "learning_rate": 7.843947470832284e-06, + "loss": 1.0186, + "step": 18202 + }, + { + "epoch": 0.58, + "learning_rate": 7.84293604016477e-06, + "loss": 0.9478, + "step": 18203 + }, + { + "epoch": 0.58, + "learning_rate": 7.841924632640166e-06, + "loss": 0.8665, + "step": 18204 + }, + { + "epoch": 0.58, + "learning_rate": 7.840913248269318e-06, + "loss": 0.9473, + "step": 18205 + }, + { + "epoch": 0.58, + "learning_rate": 7.839901887063083e-06, + "loss": 0.9512, + "step": 18206 + }, + { + "epoch": 0.58, + "learning_rate": 7.838890549032313e-06, + "loss": 0.9282, + "step": 18207 + }, + { + "epoch": 0.58, + "learning_rate": 7.837879234187851e-06, + "loss": 0.9653, + "step": 18208 + }, + { + "epoch": 0.58, + "learning_rate": 7.836867942540553e-06, + "loss": 0.9414, + "step": 18209 + }, + { + "epoch": 0.58, + "learning_rate": 7.835856674101271e-06, + "loss": 0.9761, + "step": 18210 + }, + { + "epoch": 0.58, + "learning_rate": 7.834845428880848e-06, + "loss": 0.9385, + "step": 18211 + }, + { + "epoch": 0.58, + "learning_rate": 7.83383420689014e-06, + "loss": 0.9707, + "step": 18212 + }, + { + "epoch": 0.58, + "learning_rate": 7.832823008139992e-06, + "loss": 0.8057, + "step": 18213 + }, + { + "epoch": 0.58, + "learning_rate": 7.831811832641255e-06, + "loss": 0.917, + "step": 18214 + }, + { + "epoch": 0.58, + "learning_rate": 7.83080068040478e-06, + "loss": 0.9424, + "step": 18215 + }, + { + "epoch": 0.58, + "learning_rate": 7.829789551441409e-06, + "loss": 1.0083, + "step": 18216 + }, + { + "epoch": 0.58, + "learning_rate": 7.828778445761995e-06, + "loss": 1.019, + "step": 18217 + }, + { + "epoch": 0.58, + "learning_rate": 7.827767363377385e-06, + "loss": 1.0132, + "step": 18218 + }, + { + "epoch": 0.58, + "learning_rate": 7.826756304298428e-06, + "loss": 0.8252, + "step": 18219 + }, + { + "epoch": 0.58, + "learning_rate": 7.825745268535968e-06, + "loss": 0.9453, + "step": 18220 + }, + { + "epoch": 0.58, + "learning_rate": 7.824734256100858e-06, + "loss": 0.8999, + "step": 18221 + }, + { + "epoch": 0.58, + "learning_rate": 7.823723267003947e-06, + "loss": 0.9634, + "step": 18222 + }, + { + "epoch": 0.58, + "learning_rate": 7.82271230125607e-06, + "loss": 0.8745, + "step": 18223 + }, + { + "epoch": 0.58, + "learning_rate": 7.821701358868081e-06, + "loss": 1.0435, + "step": 18224 + }, + { + "epoch": 0.58, + "learning_rate": 7.820690439850824e-06, + "loss": 0.9204, + "step": 18225 + }, + { + "epoch": 0.58, + "learning_rate": 7.819679544215148e-06, + "loss": 1.0464, + "step": 18226 + }, + { + "epoch": 0.58, + "learning_rate": 7.818668671971899e-06, + "loss": 0.979, + "step": 18227 + }, + { + "epoch": 0.58, + "learning_rate": 7.81765782313192e-06, + "loss": 0.8667, + "step": 18228 + }, + { + "epoch": 0.58, + "learning_rate": 7.816646997706061e-06, + "loss": 0.9897, + "step": 18229 + }, + { + "epoch": 0.58, + "learning_rate": 7.815636195705159e-06, + "loss": 0.9487, + "step": 18230 + }, + { + "epoch": 0.58, + "learning_rate": 7.814625417140062e-06, + "loss": 0.8689, + "step": 18231 + }, + { + "epoch": 0.58, + "learning_rate": 7.813614662021618e-06, + "loss": 0.9692, + "step": 18232 + }, + { + "epoch": 0.58, + "learning_rate": 7.812603930360668e-06, + "loss": 1.0435, + "step": 18233 + }, + { + "epoch": 0.58, + "learning_rate": 7.811593222168057e-06, + "loss": 1.0273, + "step": 18234 + }, + { + "epoch": 0.58, + "learning_rate": 7.810582537454628e-06, + "loss": 1.0942, + "step": 18235 + }, + { + "epoch": 0.58, + "learning_rate": 7.809571876231226e-06, + "loss": 1.0024, + "step": 18236 + }, + { + "epoch": 0.58, + "learning_rate": 7.808561238508696e-06, + "loss": 0.9546, + "step": 18237 + }, + { + "epoch": 0.58, + "learning_rate": 7.807550624297878e-06, + "loss": 1.0425, + "step": 18238 + }, + { + "epoch": 0.58, + "learning_rate": 7.806540033609612e-06, + "loss": 1.0913, + "step": 18239 + }, + { + "epoch": 0.58, + "learning_rate": 7.805529466454745e-06, + "loss": 0.9932, + "step": 18240 + }, + { + "epoch": 0.58, + "learning_rate": 7.804518922844118e-06, + "loss": 0.9902, + "step": 18241 + }, + { + "epoch": 0.58, + "learning_rate": 7.803508402788572e-06, + "loss": 0.9985, + "step": 18242 + }, + { + "epoch": 0.58, + "learning_rate": 7.802497906298954e-06, + "loss": 0.8901, + "step": 18243 + }, + { + "epoch": 0.58, + "learning_rate": 7.801487433386097e-06, + "loss": 0.9878, + "step": 18244 + }, + { + "epoch": 0.58, + "learning_rate": 7.800476984060852e-06, + "loss": 1.0186, + "step": 18245 + }, + { + "epoch": 0.58, + "learning_rate": 7.79946655833405e-06, + "loss": 0.9199, + "step": 18246 + }, + { + "epoch": 0.58, + "learning_rate": 7.798456156216538e-06, + "loss": 0.9092, + "step": 18247 + }, + { + "epoch": 0.58, + "learning_rate": 7.797445777719153e-06, + "loss": 0.9014, + "step": 18248 + }, + { + "epoch": 0.58, + "learning_rate": 7.79643542285274e-06, + "loss": 0.9619, + "step": 18249 + }, + { + "epoch": 0.58, + "learning_rate": 7.795425091628135e-06, + "loss": 1.0137, + "step": 18250 + }, + { + "epoch": 0.58, + "learning_rate": 7.794414784056177e-06, + "loss": 1.0273, + "step": 18251 + }, + { + "epoch": 0.58, + "learning_rate": 7.79340450014771e-06, + "loss": 0.9336, + "step": 18252 + }, + { + "epoch": 0.58, + "learning_rate": 7.79239423991357e-06, + "loss": 0.877, + "step": 18253 + }, + { + "epoch": 0.58, + "learning_rate": 7.791384003364599e-06, + "loss": 1.0093, + "step": 18254 + }, + { + "epoch": 0.58, + "learning_rate": 7.790373790511628e-06, + "loss": 1.0625, + "step": 18255 + }, + { + "epoch": 0.58, + "learning_rate": 7.789363601365502e-06, + "loss": 0.8442, + "step": 18256 + }, + { + "epoch": 0.58, + "learning_rate": 7.78835343593706e-06, + "loss": 0.9165, + "step": 18257 + }, + { + "epoch": 0.58, + "learning_rate": 7.787343294237137e-06, + "loss": 0.9507, + "step": 18258 + }, + { + "epoch": 0.58, + "learning_rate": 7.786333176276571e-06, + "loss": 0.9673, + "step": 18259 + }, + { + "epoch": 0.58, + "learning_rate": 7.785323082066204e-06, + "loss": 0.5374, + "step": 18260 + }, + { + "epoch": 0.58, + "learning_rate": 7.784313011616865e-06, + "loss": 0.9395, + "step": 18261 + }, + { + "epoch": 0.58, + "learning_rate": 7.783302964939396e-06, + "loss": 0.9229, + "step": 18262 + }, + { + "epoch": 0.58, + "learning_rate": 7.782292942044631e-06, + "loss": 0.9619, + "step": 18263 + }, + { + "epoch": 0.58, + "learning_rate": 7.781282942943411e-06, + "loss": 0.9683, + "step": 18264 + }, + { + "epoch": 0.58, + "learning_rate": 7.780272967646567e-06, + "loss": 0.9702, + "step": 18265 + }, + { + "epoch": 0.58, + "learning_rate": 7.77926301616494e-06, + "loss": 0.9526, + "step": 18266 + }, + { + "epoch": 0.58, + "learning_rate": 7.778253088509366e-06, + "loss": 0.8784, + "step": 18267 + }, + { + "epoch": 0.58, + "learning_rate": 7.777243184690672e-06, + "loss": 1.0513, + "step": 18268 + }, + { + "epoch": 0.58, + "learning_rate": 7.7762333047197e-06, + "loss": 0.8774, + "step": 18269 + }, + { + "epoch": 0.58, + "learning_rate": 7.775223448607283e-06, + "loss": 0.9556, + "step": 18270 + }, + { + "epoch": 0.58, + "learning_rate": 7.774213616364258e-06, + "loss": 0.9224, + "step": 18271 + }, + { + "epoch": 0.58, + "learning_rate": 7.773203808001454e-06, + "loss": 0.8599, + "step": 18272 + }, + { + "epoch": 0.58, + "learning_rate": 7.772194023529711e-06, + "loss": 0.9443, + "step": 18273 + }, + { + "epoch": 0.58, + "learning_rate": 7.77118426295986e-06, + "loss": 1.0107, + "step": 18274 + }, + { + "epoch": 0.58, + "learning_rate": 7.770174526302739e-06, + "loss": 1.1289, + "step": 18275 + }, + { + "epoch": 0.58, + "learning_rate": 7.769164813569173e-06, + "loss": 0.9414, + "step": 18276 + }, + { + "epoch": 0.58, + "learning_rate": 7.768155124770001e-06, + "loss": 1.0063, + "step": 18277 + }, + { + "epoch": 0.58, + "learning_rate": 7.767145459916053e-06, + "loss": 0.9927, + "step": 18278 + }, + { + "epoch": 0.58, + "learning_rate": 7.766135819018163e-06, + "loss": 0.9185, + "step": 18279 + }, + { + "epoch": 0.58, + "learning_rate": 7.765126202087164e-06, + "loss": 0.9292, + "step": 18280 + }, + { + "epoch": 0.58, + "learning_rate": 7.764116609133887e-06, + "loss": 0.9155, + "step": 18281 + }, + { + "epoch": 0.58, + "learning_rate": 7.76310704016917e-06, + "loss": 1.0112, + "step": 18282 + }, + { + "epoch": 0.58, + "learning_rate": 7.762097495203833e-06, + "loss": 1.0078, + "step": 18283 + }, + { + "epoch": 0.58, + "learning_rate": 7.761087974248712e-06, + "loss": 0.9526, + "step": 18284 + }, + { + "epoch": 0.58, + "learning_rate": 7.760078477314642e-06, + "loss": 0.98, + "step": 18285 + }, + { + "epoch": 0.58, + "learning_rate": 7.759069004412449e-06, + "loss": 0.9517, + "step": 18286 + }, + { + "epoch": 0.58, + "learning_rate": 7.758059555552966e-06, + "loss": 0.9419, + "step": 18287 + }, + { + "epoch": 0.58, + "learning_rate": 7.757050130747025e-06, + "loss": 0.9648, + "step": 18288 + }, + { + "epoch": 0.58, + "learning_rate": 7.756040730005452e-06, + "loss": 0.9507, + "step": 18289 + }, + { + "epoch": 0.58, + "learning_rate": 7.75503135333908e-06, + "loss": 1.0371, + "step": 18290 + }, + { + "epoch": 0.58, + "learning_rate": 7.754022000758734e-06, + "loss": 0.9932, + "step": 18291 + }, + { + "epoch": 0.58, + "learning_rate": 7.753012672275247e-06, + "loss": 1.0791, + "step": 18292 + }, + { + "epoch": 0.59, + "learning_rate": 7.75200336789945e-06, + "loss": 0.9946, + "step": 18293 + }, + { + "epoch": 0.59, + "learning_rate": 7.750994087642166e-06, + "loss": 0.9316, + "step": 18294 + }, + { + "epoch": 0.59, + "learning_rate": 7.749984831514226e-06, + "loss": 0.9604, + "step": 18295 + }, + { + "epoch": 0.59, + "learning_rate": 7.748975599526459e-06, + "loss": 0.8613, + "step": 18296 + }, + { + "epoch": 0.59, + "learning_rate": 7.747966391689695e-06, + "loss": 0.5222, + "step": 18297 + }, + { + "epoch": 0.59, + "learning_rate": 7.746957208014757e-06, + "loss": 1.1348, + "step": 18298 + }, + { + "epoch": 0.59, + "learning_rate": 7.745948048512477e-06, + "loss": 1.0308, + "step": 18299 + }, + { + "epoch": 0.59, + "learning_rate": 7.744938913193678e-06, + "loss": 0.9961, + "step": 18300 + }, + { + "epoch": 0.59, + "learning_rate": 7.743929802069187e-06, + "loss": 1.0137, + "step": 18301 + }, + { + "epoch": 0.59, + "learning_rate": 7.742920715149832e-06, + "loss": 1.0396, + "step": 18302 + }, + { + "epoch": 0.59, + "learning_rate": 7.741911652446441e-06, + "loss": 0.9067, + "step": 18303 + }, + { + "epoch": 0.59, + "learning_rate": 7.740902613969839e-06, + "loss": 0.979, + "step": 18304 + }, + { + "epoch": 0.59, + "learning_rate": 7.739893599730856e-06, + "loss": 0.9478, + "step": 18305 + }, + { + "epoch": 0.59, + "learning_rate": 7.738884609740309e-06, + "loss": 1.0591, + "step": 18306 + }, + { + "epoch": 0.59, + "learning_rate": 7.737875644009027e-06, + "loss": 0.979, + "step": 18307 + }, + { + "epoch": 0.59, + "learning_rate": 7.736866702547836e-06, + "loss": 1.063, + "step": 18308 + }, + { + "epoch": 0.59, + "learning_rate": 7.73585778536756e-06, + "loss": 0.9668, + "step": 18309 + }, + { + "epoch": 0.59, + "learning_rate": 7.734848892479025e-06, + "loss": 0.8813, + "step": 18310 + }, + { + "epoch": 0.59, + "learning_rate": 7.733840023893054e-06, + "loss": 0.9717, + "step": 18311 + }, + { + "epoch": 0.59, + "learning_rate": 7.732831179620477e-06, + "loss": 0.9824, + "step": 18312 + }, + { + "epoch": 0.59, + "learning_rate": 7.731822359672107e-06, + "loss": 1.0039, + "step": 18313 + }, + { + "epoch": 0.59, + "learning_rate": 7.730813564058772e-06, + "loss": 0.981, + "step": 18314 + }, + { + "epoch": 0.59, + "learning_rate": 7.729804792791298e-06, + "loss": 0.9487, + "step": 18315 + }, + { + "epoch": 0.59, + "learning_rate": 7.728796045880504e-06, + "loss": 0.8535, + "step": 18316 + }, + { + "epoch": 0.59, + "learning_rate": 7.727787323337216e-06, + "loss": 0.9775, + "step": 18317 + }, + { + "epoch": 0.59, + "learning_rate": 7.726778625172255e-06, + "loss": 0.8862, + "step": 18318 + }, + { + "epoch": 0.59, + "learning_rate": 7.725769951396444e-06, + "loss": 0.9697, + "step": 18319 + }, + { + "epoch": 0.59, + "learning_rate": 7.724761302020606e-06, + "loss": 0.488, + "step": 18320 + }, + { + "epoch": 0.59, + "learning_rate": 7.72375267705556e-06, + "loss": 1.0229, + "step": 18321 + }, + { + "epoch": 0.59, + "learning_rate": 7.722744076512127e-06, + "loss": 0.9478, + "step": 18322 + }, + { + "epoch": 0.59, + "learning_rate": 7.72173550040113e-06, + "loss": 1.0225, + "step": 18323 + }, + { + "epoch": 0.59, + "learning_rate": 7.720726948733391e-06, + "loss": 0.998, + "step": 18324 + }, + { + "epoch": 0.59, + "learning_rate": 7.719718421519728e-06, + "loss": 0.8877, + "step": 18325 + }, + { + "epoch": 0.59, + "learning_rate": 7.718709918770964e-06, + "loss": 0.939, + "step": 18326 + }, + { + "epoch": 0.59, + "learning_rate": 7.717701440497917e-06, + "loss": 1.0264, + "step": 18327 + }, + { + "epoch": 0.59, + "learning_rate": 7.71669298671141e-06, + "loss": 0.895, + "step": 18328 + }, + { + "epoch": 0.59, + "learning_rate": 7.715684557422256e-06, + "loss": 0.9683, + "step": 18329 + }, + { + "epoch": 0.59, + "learning_rate": 7.71467615264128e-06, + "loss": 0.9512, + "step": 18330 + }, + { + "epoch": 0.59, + "learning_rate": 7.7136677723793e-06, + "loss": 0.9551, + "step": 18331 + }, + { + "epoch": 0.59, + "learning_rate": 7.712659416647137e-06, + "loss": 0.9512, + "step": 18332 + }, + { + "epoch": 0.59, + "learning_rate": 7.711651085455604e-06, + "loss": 1.0151, + "step": 18333 + }, + { + "epoch": 0.59, + "learning_rate": 7.710642778815523e-06, + "loss": 0.8472, + "step": 18334 + }, + { + "epoch": 0.59, + "learning_rate": 7.709634496737711e-06, + "loss": 0.8877, + "step": 18335 + }, + { + "epoch": 0.59, + "learning_rate": 7.708626239232986e-06, + "loss": 0.894, + "step": 18336 + }, + { + "epoch": 0.59, + "learning_rate": 7.707618006312164e-06, + "loss": 0.8672, + "step": 18337 + }, + { + "epoch": 0.59, + "learning_rate": 7.706609797986067e-06, + "loss": 0.896, + "step": 18338 + }, + { + "epoch": 0.59, + "learning_rate": 7.705601614265507e-06, + "loss": 0.9834, + "step": 18339 + }, + { + "epoch": 0.59, + "learning_rate": 7.704593455161301e-06, + "loss": 0.7476, + "step": 18340 + }, + { + "epoch": 0.59, + "learning_rate": 7.703585320684267e-06, + "loss": 0.96, + "step": 18341 + }, + { + "epoch": 0.59, + "learning_rate": 7.702577210845222e-06, + "loss": 0.9756, + "step": 18342 + }, + { + "epoch": 0.59, + "learning_rate": 7.701569125654984e-06, + "loss": 0.9839, + "step": 18343 + }, + { + "epoch": 0.59, + "learning_rate": 7.700561065124362e-06, + "loss": 0.9863, + "step": 18344 + }, + { + "epoch": 0.59, + "learning_rate": 7.699553029264175e-06, + "loss": 0.9414, + "step": 18345 + }, + { + "epoch": 0.59, + "learning_rate": 7.698545018085237e-06, + "loss": 1.0137, + "step": 18346 + }, + { + "epoch": 0.59, + "learning_rate": 7.697537031598364e-06, + "loss": 1.0469, + "step": 18347 + }, + { + "epoch": 0.59, + "learning_rate": 7.69652906981437e-06, + "loss": 1.0317, + "step": 18348 + }, + { + "epoch": 0.59, + "learning_rate": 7.695521132744073e-06, + "loss": 1.0132, + "step": 18349 + }, + { + "epoch": 0.59, + "learning_rate": 7.694513220398286e-06, + "loss": 0.8926, + "step": 18350 + }, + { + "epoch": 0.59, + "learning_rate": 7.693505332787814e-06, + "loss": 0.9844, + "step": 18351 + }, + { + "epoch": 0.59, + "learning_rate": 7.692497469923479e-06, + "loss": 0.5496, + "step": 18352 + }, + { + "epoch": 0.59, + "learning_rate": 7.691489631816093e-06, + "loss": 0.8447, + "step": 18353 + }, + { + "epoch": 0.59, + "learning_rate": 7.690481818476468e-06, + "loss": 0.9424, + "step": 18354 + }, + { + "epoch": 0.59, + "learning_rate": 7.689474029915415e-06, + "loss": 1.0288, + "step": 18355 + }, + { + "epoch": 0.59, + "learning_rate": 7.688466266143752e-06, + "loss": 0.8599, + "step": 18356 + }, + { + "epoch": 0.59, + "learning_rate": 7.687458527172285e-06, + "loss": 0.897, + "step": 18357 + }, + { + "epoch": 0.59, + "learning_rate": 7.686450813011835e-06, + "loss": 0.9805, + "step": 18358 + }, + { + "epoch": 0.59, + "learning_rate": 7.6854431236732e-06, + "loss": 1.1582, + "step": 18359 + }, + { + "epoch": 0.59, + "learning_rate": 7.6844354591672e-06, + "loss": 0.9434, + "step": 18360 + }, + { + "epoch": 0.59, + "learning_rate": 7.683427819504646e-06, + "loss": 0.9976, + "step": 18361 + }, + { + "epoch": 0.59, + "learning_rate": 7.682420204696346e-06, + "loss": 0.9019, + "step": 18362 + }, + { + "epoch": 0.59, + "learning_rate": 7.681412614753115e-06, + "loss": 1.0654, + "step": 18363 + }, + { + "epoch": 0.59, + "learning_rate": 7.680405049685757e-06, + "loss": 0.9102, + "step": 18364 + }, + { + "epoch": 0.59, + "learning_rate": 7.679397509505092e-06, + "loss": 0.9785, + "step": 18365 + }, + { + "epoch": 0.59, + "learning_rate": 7.67838999422192e-06, + "loss": 1.0386, + "step": 18366 + }, + { + "epoch": 0.59, + "learning_rate": 7.67738250384705e-06, + "loss": 1.0107, + "step": 18367 + }, + { + "epoch": 0.59, + "learning_rate": 7.676375038391299e-06, + "loss": 0.8345, + "step": 18368 + }, + { + "epoch": 0.59, + "learning_rate": 7.67536759786547e-06, + "loss": 1.0044, + "step": 18369 + }, + { + "epoch": 0.59, + "learning_rate": 7.674360182280377e-06, + "loss": 1.0493, + "step": 18370 + }, + { + "epoch": 0.59, + "learning_rate": 7.673352791646824e-06, + "loss": 0.9932, + "step": 18371 + }, + { + "epoch": 0.59, + "learning_rate": 7.67234542597562e-06, + "loss": 0.7461, + "step": 18372 + }, + { + "epoch": 0.59, + "learning_rate": 7.671338085277576e-06, + "loss": 0.8921, + "step": 18373 + }, + { + "epoch": 0.59, + "learning_rate": 7.670330769563494e-06, + "loss": 0.9346, + "step": 18374 + }, + { + "epoch": 0.59, + "learning_rate": 7.669323478844184e-06, + "loss": 0.9976, + "step": 18375 + }, + { + "epoch": 0.59, + "learning_rate": 7.668316213130455e-06, + "loss": 0.9282, + "step": 18376 + }, + { + "epoch": 0.59, + "learning_rate": 7.667308972433113e-06, + "loss": 0.9517, + "step": 18377 + }, + { + "epoch": 0.59, + "learning_rate": 7.666301756762963e-06, + "loss": 0.9854, + "step": 18378 + }, + { + "epoch": 0.59, + "learning_rate": 7.665294566130812e-06, + "loss": 1.1353, + "step": 18379 + }, + { + "epoch": 0.59, + "learning_rate": 7.66428740054747e-06, + "loss": 1.0005, + "step": 18380 + }, + { + "epoch": 0.59, + "learning_rate": 7.663280260023735e-06, + "loss": 0.9702, + "step": 18381 + }, + { + "epoch": 0.59, + "learning_rate": 7.662273144570419e-06, + "loss": 0.98, + "step": 18382 + }, + { + "epoch": 0.59, + "learning_rate": 7.661266054198324e-06, + "loss": 0.8926, + "step": 18383 + }, + { + "epoch": 0.59, + "learning_rate": 7.660258988918254e-06, + "loss": 0.8103, + "step": 18384 + }, + { + "epoch": 0.59, + "learning_rate": 7.659251948741019e-06, + "loss": 0.915, + "step": 18385 + }, + { + "epoch": 0.59, + "learning_rate": 7.658244933677417e-06, + "loss": 0.9756, + "step": 18386 + }, + { + "epoch": 0.59, + "learning_rate": 7.657237943738255e-06, + "loss": 1.0562, + "step": 18387 + }, + { + "epoch": 0.59, + "learning_rate": 7.656230978934343e-06, + "loss": 1.0942, + "step": 18388 + }, + { + "epoch": 0.59, + "learning_rate": 7.655224039276472e-06, + "loss": 0.9521, + "step": 18389 + }, + { + "epoch": 0.59, + "learning_rate": 7.654217124775453e-06, + "loss": 1.0767, + "step": 18390 + }, + { + "epoch": 0.59, + "learning_rate": 7.65321023544209e-06, + "loss": 0.4885, + "step": 18391 + }, + { + "epoch": 0.59, + "learning_rate": 7.652203371287181e-06, + "loss": 0.9771, + "step": 18392 + }, + { + "epoch": 0.59, + "learning_rate": 7.651196532321535e-06, + "loss": 1.0361, + "step": 18393 + }, + { + "epoch": 0.59, + "learning_rate": 7.650189718555947e-06, + "loss": 0.9858, + "step": 18394 + }, + { + "epoch": 0.59, + "learning_rate": 7.649182930001229e-06, + "loss": 1.0273, + "step": 18395 + }, + { + "epoch": 0.59, + "learning_rate": 7.648176166668171e-06, + "loss": 0.9531, + "step": 18396 + }, + { + "epoch": 0.59, + "learning_rate": 7.64716942856758e-06, + "loss": 0.8921, + "step": 18397 + }, + { + "epoch": 0.59, + "learning_rate": 7.646162715710258e-06, + "loss": 0.9907, + "step": 18398 + }, + { + "epoch": 0.59, + "learning_rate": 7.645156028107005e-06, + "loss": 1.0425, + "step": 18399 + }, + { + "epoch": 0.59, + "learning_rate": 7.644149365768621e-06, + "loss": 0.9712, + "step": 18400 + }, + { + "epoch": 0.59, + "learning_rate": 7.643142728705907e-06, + "loss": 0.8774, + "step": 18401 + }, + { + "epoch": 0.59, + "learning_rate": 7.642136116929662e-06, + "loss": 0.8848, + "step": 18402 + }, + { + "epoch": 0.59, + "learning_rate": 7.641129530450693e-06, + "loss": 0.9331, + "step": 18403 + }, + { + "epoch": 0.59, + "learning_rate": 7.640122969279788e-06, + "loss": 0.9648, + "step": 18404 + }, + { + "epoch": 0.59, + "learning_rate": 7.639116433427753e-06, + "loss": 0.9805, + "step": 18405 + }, + { + "epoch": 0.59, + "learning_rate": 7.638109922905385e-06, + "loss": 0.887, + "step": 18406 + }, + { + "epoch": 0.59, + "learning_rate": 7.637103437723483e-06, + "loss": 0.978, + "step": 18407 + }, + { + "epoch": 0.59, + "learning_rate": 7.636096977892848e-06, + "loss": 0.8706, + "step": 18408 + }, + { + "epoch": 0.59, + "learning_rate": 7.635090543424274e-06, + "loss": 0.9785, + "step": 18409 + }, + { + "epoch": 0.59, + "learning_rate": 7.634084134328561e-06, + "loss": 0.9497, + "step": 18410 + }, + { + "epoch": 0.59, + "learning_rate": 7.633077750616512e-06, + "loss": 0.9937, + "step": 18411 + }, + { + "epoch": 0.59, + "learning_rate": 7.632071392298915e-06, + "loss": 1.0107, + "step": 18412 + }, + { + "epoch": 0.59, + "learning_rate": 7.63106505938657e-06, + "loss": 0.9312, + "step": 18413 + }, + { + "epoch": 0.59, + "learning_rate": 7.630058751890277e-06, + "loss": 1.0474, + "step": 18414 + }, + { + "epoch": 0.59, + "learning_rate": 7.62905246982083e-06, + "loss": 0.9214, + "step": 18415 + }, + { + "epoch": 0.59, + "learning_rate": 7.628046213189028e-06, + "loss": 0.8647, + "step": 18416 + }, + { + "epoch": 0.59, + "learning_rate": 7.627039982005663e-06, + "loss": 0.9873, + "step": 18417 + }, + { + "epoch": 0.59, + "learning_rate": 7.626033776281534e-06, + "loss": 1.0146, + "step": 18418 + }, + { + "epoch": 0.59, + "learning_rate": 7.625027596027434e-06, + "loss": 0.9629, + "step": 18419 + }, + { + "epoch": 0.59, + "learning_rate": 7.624021441254159e-06, + "loss": 0.9585, + "step": 18420 + }, + { + "epoch": 0.59, + "learning_rate": 7.623015311972507e-06, + "loss": 0.9531, + "step": 18421 + }, + { + "epoch": 0.59, + "learning_rate": 7.622009208193267e-06, + "loss": 1.0308, + "step": 18422 + }, + { + "epoch": 0.59, + "learning_rate": 7.621003129927237e-06, + "loss": 0.98, + "step": 18423 + }, + { + "epoch": 0.59, + "learning_rate": 7.619997077185211e-06, + "loss": 0.8208, + "step": 18424 + }, + { + "epoch": 0.59, + "learning_rate": 7.61899104997798e-06, + "loss": 1.0181, + "step": 18425 + }, + { + "epoch": 0.59, + "learning_rate": 7.6179850483163445e-06, + "loss": 0.9751, + "step": 18426 + }, + { + "epoch": 0.59, + "learning_rate": 7.616979072211093e-06, + "loss": 1.0264, + "step": 18427 + }, + { + "epoch": 0.59, + "learning_rate": 7.615973121673015e-06, + "loss": 1.002, + "step": 18428 + }, + { + "epoch": 0.59, + "learning_rate": 7.614967196712908e-06, + "loss": 0.9946, + "step": 18429 + }, + { + "epoch": 0.59, + "learning_rate": 7.613961297341563e-06, + "loss": 0.8545, + "step": 18430 + }, + { + "epoch": 0.59, + "learning_rate": 7.612955423569772e-06, + "loss": 1.0381, + "step": 18431 + }, + { + "epoch": 0.59, + "learning_rate": 7.611949575408328e-06, + "loss": 0.4561, + "step": 18432 + }, + { + "epoch": 0.59, + "learning_rate": 7.610943752868026e-06, + "loss": 0.9521, + "step": 18433 + }, + { + "epoch": 0.59, + "learning_rate": 7.60993795595965e-06, + "loss": 0.9014, + "step": 18434 + }, + { + "epoch": 0.59, + "learning_rate": 7.6089321846939935e-06, + "loss": 1.0103, + "step": 18435 + }, + { + "epoch": 0.59, + "learning_rate": 7.60792643908185e-06, + "loss": 0.9893, + "step": 18436 + }, + { + "epoch": 0.59, + "learning_rate": 7.606920719134007e-06, + "loss": 0.9775, + "step": 18437 + }, + { + "epoch": 0.59, + "learning_rate": 7.605915024861257e-06, + "loss": 0.958, + "step": 18438 + }, + { + "epoch": 0.59, + "learning_rate": 7.604909356274388e-06, + "loss": 1.0479, + "step": 18439 + }, + { + "epoch": 0.59, + "learning_rate": 7.603903713384191e-06, + "loss": 1.0371, + "step": 18440 + }, + { + "epoch": 0.59, + "learning_rate": 7.6028980962014595e-06, + "loss": 0.9639, + "step": 18441 + }, + { + "epoch": 0.59, + "learning_rate": 7.601892504736975e-06, + "loss": 0.8237, + "step": 18442 + }, + { + "epoch": 0.59, + "learning_rate": 7.60088693900153e-06, + "loss": 1.0205, + "step": 18443 + }, + { + "epoch": 0.59, + "learning_rate": 7.599881399005913e-06, + "loss": 0.9292, + "step": 18444 + }, + { + "epoch": 0.59, + "learning_rate": 7.598875884760911e-06, + "loss": 0.9033, + "step": 18445 + }, + { + "epoch": 0.59, + "learning_rate": 7.597870396277314e-06, + "loss": 1.0088, + "step": 18446 + }, + { + "epoch": 0.59, + "learning_rate": 7.596864933565909e-06, + "loss": 0.8916, + "step": 18447 + }, + { + "epoch": 0.59, + "learning_rate": 7.595859496637489e-06, + "loss": 0.9722, + "step": 18448 + }, + { + "epoch": 0.59, + "learning_rate": 7.594854085502829e-06, + "loss": 0.9639, + "step": 18449 + }, + { + "epoch": 0.59, + "learning_rate": 7.593848700172725e-06, + "loss": 0.9219, + "step": 18450 + }, + { + "epoch": 0.59, + "learning_rate": 7.592843340657961e-06, + "loss": 1.0278, + "step": 18451 + }, + { + "epoch": 0.59, + "learning_rate": 7.591838006969324e-06, + "loss": 0.9526, + "step": 18452 + }, + { + "epoch": 0.59, + "learning_rate": 7.5908326991176e-06, + "loss": 0.917, + "step": 18453 + }, + { + "epoch": 0.59, + "learning_rate": 7.589827417113573e-06, + "loss": 0.8721, + "step": 18454 + }, + { + "epoch": 0.59, + "learning_rate": 7.588822160968035e-06, + "loss": 1.0376, + "step": 18455 + }, + { + "epoch": 0.59, + "learning_rate": 7.587816930691765e-06, + "loss": 1.0176, + "step": 18456 + }, + { + "epoch": 0.59, + "learning_rate": 7.586811726295549e-06, + "loss": 1.0449, + "step": 18457 + }, + { + "epoch": 0.59, + "learning_rate": 7.585806547790171e-06, + "loss": 1.0273, + "step": 18458 + }, + { + "epoch": 0.59, + "learning_rate": 7.584801395186417e-06, + "loss": 0.9849, + "step": 18459 + }, + { + "epoch": 0.59, + "learning_rate": 7.5837962684950735e-06, + "loss": 0.9175, + "step": 18460 + }, + { + "epoch": 0.59, + "learning_rate": 7.582791167726921e-06, + "loss": 1.064, + "step": 18461 + }, + { + "epoch": 0.59, + "learning_rate": 7.581786092892742e-06, + "loss": 1.0122, + "step": 18462 + }, + { + "epoch": 0.59, + "learning_rate": 7.580781044003324e-06, + "loss": 0.918, + "step": 18463 + }, + { + "epoch": 0.59, + "learning_rate": 7.579776021069447e-06, + "loss": 0.9028, + "step": 18464 + }, + { + "epoch": 0.59, + "learning_rate": 7.578771024101896e-06, + "loss": 0.9438, + "step": 18465 + }, + { + "epoch": 0.59, + "learning_rate": 7.577766053111452e-06, + "loss": 0.9688, + "step": 18466 + }, + { + "epoch": 0.59, + "learning_rate": 7.576761108108896e-06, + "loss": 0.5095, + "step": 18467 + }, + { + "epoch": 0.59, + "learning_rate": 7.5757561891050115e-06, + "loss": 0.813, + "step": 18468 + }, + { + "epoch": 0.59, + "learning_rate": 7.5747512961105795e-06, + "loss": 0.8677, + "step": 18469 + }, + { + "epoch": 0.59, + "learning_rate": 7.573746429136383e-06, + "loss": 1.1235, + "step": 18470 + }, + { + "epoch": 0.59, + "learning_rate": 7.572741588193206e-06, + "loss": 0.8921, + "step": 18471 + }, + { + "epoch": 0.59, + "learning_rate": 7.57173677329182e-06, + "loss": 0.9937, + "step": 18472 + }, + { + "epoch": 0.59, + "learning_rate": 7.570731984443013e-06, + "loss": 1.0337, + "step": 18473 + }, + { + "epoch": 0.59, + "learning_rate": 7.569727221657561e-06, + "loss": 1.0381, + "step": 18474 + }, + { + "epoch": 0.59, + "learning_rate": 7.568722484946248e-06, + "loss": 0.9111, + "step": 18475 + }, + { + "epoch": 0.59, + "learning_rate": 7.567717774319851e-06, + "loss": 0.9385, + "step": 18476 + }, + { + "epoch": 0.59, + "learning_rate": 7.56671308978915e-06, + "loss": 1.064, + "step": 18477 + }, + { + "epoch": 0.59, + "learning_rate": 7.565708431364926e-06, + "loss": 1.0269, + "step": 18478 + }, + { + "epoch": 0.59, + "learning_rate": 7.56470379905796e-06, + "loss": 0.8638, + "step": 18479 + }, + { + "epoch": 0.59, + "learning_rate": 7.5636991928790226e-06, + "loss": 0.8828, + "step": 18480 + }, + { + "epoch": 0.59, + "learning_rate": 7.562694612838896e-06, + "loss": 0.9507, + "step": 18481 + }, + { + "epoch": 0.59, + "learning_rate": 7.56169005894836e-06, + "loss": 0.8647, + "step": 18482 + }, + { + "epoch": 0.59, + "learning_rate": 7.56068553121819e-06, + "loss": 0.9644, + "step": 18483 + }, + { + "epoch": 0.59, + "learning_rate": 7.559681029659165e-06, + "loss": 0.8579, + "step": 18484 + }, + { + "epoch": 0.59, + "learning_rate": 7.558676554282061e-06, + "loss": 0.9678, + "step": 18485 + }, + { + "epoch": 0.59, + "learning_rate": 7.557672105097661e-06, + "loss": 0.9072, + "step": 18486 + }, + { + "epoch": 0.59, + "learning_rate": 7.556667682116732e-06, + "loss": 1.0137, + "step": 18487 + }, + { + "epoch": 0.59, + "learning_rate": 7.555663285350053e-06, + "loss": 1.0669, + "step": 18488 + }, + { + "epoch": 0.59, + "learning_rate": 7.554658914808404e-06, + "loss": 0.4995, + "step": 18489 + }, + { + "epoch": 0.59, + "learning_rate": 7.553654570502557e-06, + "loss": 1.1279, + "step": 18490 + }, + { + "epoch": 0.59, + "learning_rate": 7.552650252443289e-06, + "loss": 0.8467, + "step": 18491 + }, + { + "epoch": 0.59, + "learning_rate": 7.551645960641375e-06, + "loss": 1.0576, + "step": 18492 + }, + { + "epoch": 0.59, + "learning_rate": 7.5506416951075905e-06, + "loss": 0.8584, + "step": 18493 + }, + { + "epoch": 0.59, + "learning_rate": 7.549637455852714e-06, + "loss": 0.9722, + "step": 18494 + }, + { + "epoch": 0.59, + "learning_rate": 7.548633242887512e-06, + "loss": 1.0605, + "step": 18495 + }, + { + "epoch": 0.59, + "learning_rate": 7.547629056222762e-06, + "loss": 1.1411, + "step": 18496 + }, + { + "epoch": 0.59, + "learning_rate": 7.546624895869238e-06, + "loss": 0.9302, + "step": 18497 + }, + { + "epoch": 0.59, + "learning_rate": 7.545620761837713e-06, + "loss": 0.9712, + "step": 18498 + }, + { + "epoch": 0.59, + "learning_rate": 7.544616654138963e-06, + "loss": 1.0513, + "step": 18499 + }, + { + "epoch": 0.59, + "learning_rate": 7.543612572783757e-06, + "loss": 0.8901, + "step": 18500 + }, + { + "epoch": 0.59, + "learning_rate": 7.542608517782873e-06, + "loss": 0.9277, + "step": 18501 + }, + { + "epoch": 0.59, + "learning_rate": 7.541604489147077e-06, + "loss": 0.9814, + "step": 18502 + }, + { + "epoch": 0.59, + "learning_rate": 7.540600486887143e-06, + "loss": 0.9468, + "step": 18503 + }, + { + "epoch": 0.59, + "learning_rate": 7.539596511013843e-06, + "loss": 0.9854, + "step": 18504 + }, + { + "epoch": 0.59, + "learning_rate": 7.538592561537954e-06, + "loss": 0.8623, + "step": 18505 + }, + { + "epoch": 0.59, + "learning_rate": 7.537588638470239e-06, + "loss": 0.9888, + "step": 18506 + }, + { + "epoch": 0.59, + "learning_rate": 7.536584741821473e-06, + "loss": 0.8535, + "step": 18507 + }, + { + "epoch": 0.59, + "learning_rate": 7.535580871602427e-06, + "loss": 0.9048, + "step": 18508 + }, + { + "epoch": 0.59, + "learning_rate": 7.534577027823873e-06, + "loss": 1.0913, + "step": 18509 + }, + { + "epoch": 0.59, + "learning_rate": 7.5335732104965775e-06, + "loss": 0.9434, + "step": 18510 + }, + { + "epoch": 0.59, + "learning_rate": 7.53256941963131e-06, + "loss": 0.9927, + "step": 18511 + }, + { + "epoch": 0.59, + "learning_rate": 7.531565655238843e-06, + "loss": 1.0005, + "step": 18512 + }, + { + "epoch": 0.59, + "learning_rate": 7.530561917329944e-06, + "loss": 0.9312, + "step": 18513 + }, + { + "epoch": 0.59, + "learning_rate": 7.5295582059153815e-06, + "loss": 0.9551, + "step": 18514 + }, + { + "epoch": 0.59, + "learning_rate": 7.528554521005927e-06, + "loss": 0.9946, + "step": 18515 + }, + { + "epoch": 0.59, + "learning_rate": 7.52755086261235e-06, + "loss": 0.8945, + "step": 18516 + }, + { + "epoch": 0.59, + "learning_rate": 7.526547230745412e-06, + "loss": 0.9961, + "step": 18517 + }, + { + "epoch": 0.59, + "learning_rate": 7.525543625415885e-06, + "loss": 0.8376, + "step": 18518 + }, + { + "epoch": 0.59, + "learning_rate": 7.5245400466345364e-06, + "loss": 1.0005, + "step": 18519 + }, + { + "epoch": 0.59, + "learning_rate": 7.523536494412134e-06, + "loss": 1.1113, + "step": 18520 + }, + { + "epoch": 0.59, + "learning_rate": 7.5225329687594435e-06, + "loss": 0.9814, + "step": 18521 + }, + { + "epoch": 0.59, + "learning_rate": 7.521529469687232e-06, + "loss": 1.0054, + "step": 18522 + }, + { + "epoch": 0.59, + "learning_rate": 7.520525997206267e-06, + "loss": 0.9658, + "step": 18523 + }, + { + "epoch": 0.59, + "learning_rate": 7.519522551327317e-06, + "loss": 0.9888, + "step": 18524 + }, + { + "epoch": 0.59, + "learning_rate": 7.518519132061142e-06, + "loss": 0.9521, + "step": 18525 + }, + { + "epoch": 0.59, + "learning_rate": 7.51751573941851e-06, + "loss": 0.9976, + "step": 18526 + }, + { + "epoch": 0.59, + "learning_rate": 7.516512373410185e-06, + "loss": 1.0103, + "step": 18527 + }, + { + "epoch": 0.59, + "learning_rate": 7.515509034046935e-06, + "loss": 0.939, + "step": 18528 + }, + { + "epoch": 0.59, + "learning_rate": 7.514505721339524e-06, + "loss": 0.8887, + "step": 18529 + }, + { + "epoch": 0.59, + "learning_rate": 7.513502435298715e-06, + "loss": 0.9673, + "step": 18530 + }, + { + "epoch": 0.59, + "learning_rate": 7.512499175935277e-06, + "loss": 1.0068, + "step": 18531 + }, + { + "epoch": 0.59, + "learning_rate": 7.511495943259966e-06, + "loss": 1.0186, + "step": 18532 + }, + { + "epoch": 0.59, + "learning_rate": 7.510492737283549e-06, + "loss": 1.1367, + "step": 18533 + }, + { + "epoch": 0.59, + "learning_rate": 7.50948955801679e-06, + "loss": 0.9648, + "step": 18534 + }, + { + "epoch": 0.59, + "learning_rate": 7.508486405470452e-06, + "loss": 0.917, + "step": 18535 + }, + { + "epoch": 0.59, + "learning_rate": 7.507483279655297e-06, + "loss": 0.9487, + "step": 18536 + }, + { + "epoch": 0.59, + "learning_rate": 7.5064801805820875e-06, + "loss": 0.981, + "step": 18537 + }, + { + "epoch": 0.59, + "learning_rate": 7.505477108261587e-06, + "loss": 1.0547, + "step": 18538 + }, + { + "epoch": 0.59, + "learning_rate": 7.504474062704559e-06, + "loss": 1.0522, + "step": 18539 + }, + { + "epoch": 0.59, + "learning_rate": 7.503471043921759e-06, + "loss": 0.8535, + "step": 18540 + }, + { + "epoch": 0.59, + "learning_rate": 7.502468051923952e-06, + "loss": 0.96, + "step": 18541 + }, + { + "epoch": 0.59, + "learning_rate": 7.501465086721897e-06, + "loss": 0.9707, + "step": 18542 + }, + { + "epoch": 0.59, + "learning_rate": 7.500462148326358e-06, + "loss": 0.9512, + "step": 18543 + }, + { + "epoch": 0.59, + "learning_rate": 7.499459236748097e-06, + "loss": 0.958, + "step": 18544 + }, + { + "epoch": 0.59, + "learning_rate": 7.498456351997867e-06, + "loss": 0.8765, + "step": 18545 + }, + { + "epoch": 0.59, + "learning_rate": 7.4974534940864345e-06, + "loss": 0.5203, + "step": 18546 + }, + { + "epoch": 0.59, + "learning_rate": 7.496450663024553e-06, + "loss": 0.9917, + "step": 18547 + }, + { + "epoch": 0.59, + "learning_rate": 7.4954478588229865e-06, + "loss": 0.9453, + "step": 18548 + }, + { + "epoch": 0.59, + "learning_rate": 7.494445081492494e-06, + "loss": 0.9937, + "step": 18549 + }, + { + "epoch": 0.59, + "learning_rate": 7.493442331043832e-06, + "loss": 0.978, + "step": 18550 + }, + { + "epoch": 0.59, + "learning_rate": 7.492439607487758e-06, + "loss": 1.0786, + "step": 18551 + }, + { + "epoch": 0.59, + "learning_rate": 7.491436910835032e-06, + "loss": 0.8882, + "step": 18552 + }, + { + "epoch": 0.59, + "learning_rate": 7.490434241096414e-06, + "loss": 0.7964, + "step": 18553 + }, + { + "epoch": 0.59, + "learning_rate": 7.489431598282659e-06, + "loss": 0.8818, + "step": 18554 + }, + { + "epoch": 0.59, + "learning_rate": 7.488428982404524e-06, + "loss": 1.0073, + "step": 18555 + }, + { + "epoch": 0.59, + "learning_rate": 7.4874263934727655e-06, + "loss": 0.8984, + "step": 18556 + }, + { + "epoch": 0.59, + "learning_rate": 7.48642383149814e-06, + "loss": 0.9878, + "step": 18557 + }, + { + "epoch": 0.59, + "learning_rate": 7.485421296491405e-06, + "loss": 1.0767, + "step": 18558 + }, + { + "epoch": 0.59, + "learning_rate": 7.484418788463317e-06, + "loss": 1.0703, + "step": 18559 + }, + { + "epoch": 0.59, + "learning_rate": 7.483416307424631e-06, + "loss": 0.9941, + "step": 18560 + }, + { + "epoch": 0.59, + "learning_rate": 7.482413853386104e-06, + "loss": 1.0728, + "step": 18561 + }, + { + "epoch": 0.59, + "learning_rate": 7.481411426358492e-06, + "loss": 0.8706, + "step": 18562 + }, + { + "epoch": 0.59, + "learning_rate": 7.480409026352545e-06, + "loss": 1.1172, + "step": 18563 + }, + { + "epoch": 0.59, + "learning_rate": 7.47940665337902e-06, + "loss": 0.9697, + "step": 18564 + }, + { + "epoch": 0.59, + "learning_rate": 7.478404307448672e-06, + "loss": 0.9614, + "step": 18565 + }, + { + "epoch": 0.59, + "learning_rate": 7.477401988572257e-06, + "loss": 0.9463, + "step": 18566 + }, + { + "epoch": 0.59, + "learning_rate": 7.476399696760525e-06, + "loss": 1.0093, + "step": 18567 + }, + { + "epoch": 0.59, + "learning_rate": 7.4753974320242315e-06, + "loss": 1.0571, + "step": 18568 + }, + { + "epoch": 0.59, + "learning_rate": 7.474395194374134e-06, + "loss": 1.0205, + "step": 18569 + }, + { + "epoch": 0.59, + "learning_rate": 7.4733929838209765e-06, + "loss": 0.8896, + "step": 18570 + }, + { + "epoch": 0.59, + "learning_rate": 7.4723908003755154e-06, + "loss": 0.8438, + "step": 18571 + }, + { + "epoch": 0.59, + "learning_rate": 7.471388644048504e-06, + "loss": 0.9678, + "step": 18572 + }, + { + "epoch": 0.59, + "learning_rate": 7.4703865148506935e-06, + "loss": 0.9111, + "step": 18573 + }, + { + "epoch": 0.59, + "learning_rate": 7.469384412792837e-06, + "loss": 0.9888, + "step": 18574 + }, + { + "epoch": 0.59, + "learning_rate": 7.468382337885685e-06, + "loss": 0.9399, + "step": 18575 + }, + { + "epoch": 0.59, + "learning_rate": 7.467380290139988e-06, + "loss": 0.959, + "step": 18576 + }, + { + "epoch": 0.59, + "learning_rate": 7.466378269566501e-06, + "loss": 0.9814, + "step": 18577 + }, + { + "epoch": 0.59, + "learning_rate": 7.465376276175968e-06, + "loss": 0.998, + "step": 18578 + }, + { + "epoch": 0.59, + "learning_rate": 7.464374309979143e-06, + "loss": 1.0103, + "step": 18579 + }, + { + "epoch": 0.59, + "learning_rate": 7.463372370986775e-06, + "loss": 0.9531, + "step": 18580 + }, + { + "epoch": 0.59, + "learning_rate": 7.462370459209614e-06, + "loss": 0.9326, + "step": 18581 + }, + { + "epoch": 0.59, + "learning_rate": 7.461368574658407e-06, + "loss": 1.0278, + "step": 18582 + }, + { + "epoch": 0.59, + "learning_rate": 7.46036671734391e-06, + "loss": 0.9634, + "step": 18583 + }, + { + "epoch": 0.59, + "learning_rate": 7.4593648872768674e-06, + "loss": 1.0518, + "step": 18584 + }, + { + "epoch": 0.59, + "learning_rate": 7.4583630844680255e-06, + "loss": 0.9141, + "step": 18585 + }, + { + "epoch": 0.59, + "learning_rate": 7.457361308928135e-06, + "loss": 0.813, + "step": 18586 + }, + { + "epoch": 0.59, + "learning_rate": 7.456359560667943e-06, + "loss": 0.8511, + "step": 18587 + }, + { + "epoch": 0.59, + "learning_rate": 7.455357839698201e-06, + "loss": 0.8975, + "step": 18588 + }, + { + "epoch": 0.59, + "learning_rate": 7.454356146029649e-06, + "loss": 0.999, + "step": 18589 + }, + { + "epoch": 0.59, + "learning_rate": 7.45335447967304e-06, + "loss": 0.9312, + "step": 18590 + }, + { + "epoch": 0.59, + "learning_rate": 7.452352840639118e-06, + "loss": 0.9692, + "step": 18591 + }, + { + "epoch": 0.59, + "learning_rate": 7.451351228938633e-06, + "loss": 0.9424, + "step": 18592 + }, + { + "epoch": 0.59, + "learning_rate": 7.450349644582326e-06, + "loss": 1.0244, + "step": 18593 + }, + { + "epoch": 0.59, + "learning_rate": 7.449348087580947e-06, + "loss": 0.9189, + "step": 18594 + }, + { + "epoch": 0.59, + "learning_rate": 7.4483465579452385e-06, + "loss": 1.0327, + "step": 18595 + }, + { + "epoch": 0.59, + "learning_rate": 7.447345055685948e-06, + "loss": 0.998, + "step": 18596 + }, + { + "epoch": 0.59, + "learning_rate": 7.446343580813819e-06, + "loss": 0.4836, + "step": 18597 + }, + { + "epoch": 0.59, + "learning_rate": 7.445342133339599e-06, + "loss": 1.0225, + "step": 18598 + }, + { + "epoch": 0.59, + "learning_rate": 7.4443407132740344e-06, + "loss": 1.0781, + "step": 18599 + }, + { + "epoch": 0.59, + "learning_rate": 7.4433393206278605e-06, + "loss": 0.9087, + "step": 18600 + }, + { + "epoch": 0.59, + "learning_rate": 7.4423379554118245e-06, + "loss": 1.0889, + "step": 18601 + }, + { + "epoch": 0.59, + "learning_rate": 7.441336617636674e-06, + "loss": 0.8359, + "step": 18602 + }, + { + "epoch": 0.59, + "learning_rate": 7.440335307313149e-06, + "loss": 0.8701, + "step": 18603 + }, + { + "epoch": 0.59, + "learning_rate": 7.439334024451993e-06, + "loss": 0.9702, + "step": 18604 + }, + { + "epoch": 0.59, + "learning_rate": 7.4383327690639495e-06, + "loss": 0.916, + "step": 18605 + }, + { + "epoch": 0.6, + "learning_rate": 7.437331541159759e-06, + "loss": 0.9014, + "step": 18606 + }, + { + "epoch": 0.6, + "learning_rate": 7.436330340750171e-06, + "loss": 0.958, + "step": 18607 + }, + { + "epoch": 0.6, + "learning_rate": 7.435329167845915e-06, + "loss": 0.9932, + "step": 18608 + }, + { + "epoch": 0.6, + "learning_rate": 7.4343280224577395e-06, + "loss": 0.9854, + "step": 18609 + }, + { + "epoch": 0.6, + "learning_rate": 7.4333269045963855e-06, + "loss": 0.9917, + "step": 18610 + }, + { + "epoch": 0.6, + "learning_rate": 7.432325814272592e-06, + "loss": 0.8552, + "step": 18611 + }, + { + "epoch": 0.6, + "learning_rate": 7.4313247514971e-06, + "loss": 0.8711, + "step": 18612 + }, + { + "epoch": 0.6, + "learning_rate": 7.4303237162806515e-06, + "loss": 0.7637, + "step": 18613 + }, + { + "epoch": 0.6, + "learning_rate": 7.4293227086339906e-06, + "loss": 1.0933, + "step": 18614 + }, + { + "epoch": 0.6, + "learning_rate": 7.428321728567846e-06, + "loss": 0.8887, + "step": 18615 + }, + { + "epoch": 0.6, + "learning_rate": 7.427320776092965e-06, + "loss": 0.9062, + "step": 18616 + }, + { + "epoch": 0.6, + "learning_rate": 7.4263198512200825e-06, + "loss": 0.9263, + "step": 18617 + }, + { + "epoch": 0.6, + "learning_rate": 7.425318953959943e-06, + "loss": 0.9517, + "step": 18618 + }, + { + "epoch": 0.6, + "learning_rate": 7.424318084323279e-06, + "loss": 0.9824, + "step": 18619 + }, + { + "epoch": 0.6, + "learning_rate": 7.423317242320831e-06, + "loss": 0.9224, + "step": 18620 + }, + { + "epoch": 0.6, + "learning_rate": 7.422316427963339e-06, + "loss": 0.9429, + "step": 18621 + }, + { + "epoch": 0.6, + "learning_rate": 7.421315641261542e-06, + "loss": 0.9756, + "step": 18622 + }, + { + "epoch": 0.6, + "learning_rate": 7.420314882226169e-06, + "loss": 1.0732, + "step": 18623 + }, + { + "epoch": 0.6, + "learning_rate": 7.419314150867964e-06, + "loss": 0.9639, + "step": 18624 + }, + { + "epoch": 0.6, + "learning_rate": 7.418313447197663e-06, + "loss": 0.9731, + "step": 18625 + }, + { + "epoch": 0.6, + "learning_rate": 7.417312771225999e-06, + "loss": 0.9175, + "step": 18626 + }, + { + "epoch": 0.6, + "learning_rate": 7.416312122963711e-06, + "loss": 0.8801, + "step": 18627 + }, + { + "epoch": 0.6, + "learning_rate": 7.415311502421538e-06, + "loss": 0.9517, + "step": 18628 + }, + { + "epoch": 0.6, + "learning_rate": 7.414310909610212e-06, + "loss": 0.8779, + "step": 18629 + }, + { + "epoch": 0.6, + "learning_rate": 7.413310344540464e-06, + "loss": 0.8599, + "step": 18630 + }, + { + "epoch": 0.6, + "learning_rate": 7.412309807223035e-06, + "loss": 0.9736, + "step": 18631 + }, + { + "epoch": 0.6, + "learning_rate": 7.411309297668656e-06, + "loss": 1.063, + "step": 18632 + }, + { + "epoch": 0.6, + "learning_rate": 7.410308815888068e-06, + "loss": 0.9033, + "step": 18633 + }, + { + "epoch": 0.6, + "learning_rate": 7.409308361891996e-06, + "loss": 0.8208, + "step": 18634 + }, + { + "epoch": 0.6, + "learning_rate": 7.408307935691179e-06, + "loss": 1.002, + "step": 18635 + }, + { + "epoch": 0.6, + "learning_rate": 7.407307537296349e-06, + "loss": 1.0522, + "step": 18636 + }, + { + "epoch": 0.6, + "learning_rate": 7.406307166718243e-06, + "loss": 0.4993, + "step": 18637 + }, + { + "epoch": 0.6, + "learning_rate": 7.4053068239675885e-06, + "loss": 0.9043, + "step": 18638 + }, + { + "epoch": 0.6, + "learning_rate": 7.404306509055119e-06, + "loss": 1.1128, + "step": 18639 + }, + { + "epoch": 0.6, + "learning_rate": 7.4033062219915685e-06, + "loss": 1.0181, + "step": 18640 + }, + { + "epoch": 0.6, + "learning_rate": 7.402305962787668e-06, + "loss": 0.8867, + "step": 18641 + }, + { + "epoch": 0.6, + "learning_rate": 7.401305731454149e-06, + "loss": 1.1562, + "step": 18642 + }, + { + "epoch": 0.6, + "learning_rate": 7.400305528001745e-06, + "loss": 0.9448, + "step": 18643 + }, + { + "epoch": 0.6, + "learning_rate": 7.399305352441184e-06, + "loss": 1.0122, + "step": 18644 + }, + { + "epoch": 0.6, + "learning_rate": 7.398305204783203e-06, + "loss": 1.0508, + "step": 18645 + }, + { + "epoch": 0.6, + "learning_rate": 7.397305085038522e-06, + "loss": 0.9287, + "step": 18646 + }, + { + "epoch": 0.6, + "learning_rate": 7.3963049932178785e-06, + "loss": 0.9707, + "step": 18647 + }, + { + "epoch": 0.6, + "learning_rate": 7.395304929332001e-06, + "loss": 0.9502, + "step": 18648 + }, + { + "epoch": 0.6, + "learning_rate": 7.394304893391618e-06, + "loss": 0.9468, + "step": 18649 + }, + { + "epoch": 0.6, + "learning_rate": 7.39330488540746e-06, + "loss": 0.998, + "step": 18650 + }, + { + "epoch": 0.6, + "learning_rate": 7.392304905390256e-06, + "loss": 0.8979, + "step": 18651 + }, + { + "epoch": 0.6, + "learning_rate": 7.391304953350739e-06, + "loss": 0.9507, + "step": 18652 + }, + { + "epoch": 0.6, + "learning_rate": 7.390305029299627e-06, + "loss": 1.0581, + "step": 18653 + }, + { + "epoch": 0.6, + "learning_rate": 7.389305133247655e-06, + "loss": 0.9839, + "step": 18654 + }, + { + "epoch": 0.6, + "learning_rate": 7.388305265205549e-06, + "loss": 1.0029, + "step": 18655 + }, + { + "epoch": 0.6, + "learning_rate": 7.387305425184037e-06, + "loss": 0.5037, + "step": 18656 + }, + { + "epoch": 0.6, + "learning_rate": 7.386305613193846e-06, + "loss": 1.0337, + "step": 18657 + }, + { + "epoch": 0.6, + "learning_rate": 7.385305829245703e-06, + "loss": 0.9136, + "step": 18658 + }, + { + "epoch": 0.6, + "learning_rate": 7.384306073350335e-06, + "loss": 0.9292, + "step": 18659 + }, + { + "epoch": 0.6, + "learning_rate": 7.383306345518473e-06, + "loss": 1.0371, + "step": 18660 + }, + { + "epoch": 0.6, + "learning_rate": 7.382306645760832e-06, + "loss": 0.8486, + "step": 18661 + }, + { + "epoch": 0.6, + "learning_rate": 7.381306974088145e-06, + "loss": 0.9014, + "step": 18662 + }, + { + "epoch": 0.6, + "learning_rate": 7.380307330511134e-06, + "loss": 1.0093, + "step": 18663 + }, + { + "epoch": 0.6, + "learning_rate": 7.379307715040528e-06, + "loss": 1.1123, + "step": 18664 + }, + { + "epoch": 0.6, + "learning_rate": 7.378308127687048e-06, + "loss": 0.894, + "step": 18665 + }, + { + "epoch": 0.6, + "learning_rate": 7.377308568461422e-06, + "loss": 0.8867, + "step": 18666 + }, + { + "epoch": 0.6, + "learning_rate": 7.376309037374376e-06, + "loss": 0.8965, + "step": 18667 + }, + { + "epoch": 0.6, + "learning_rate": 7.3753095344366256e-06, + "loss": 0.8555, + "step": 18668 + }, + { + "epoch": 0.6, + "learning_rate": 7.3743100596589e-06, + "loss": 0.9478, + "step": 18669 + }, + { + "epoch": 0.6, + "learning_rate": 7.37331061305192e-06, + "loss": 0.8564, + "step": 18670 + }, + { + "epoch": 0.6, + "learning_rate": 7.372311194626412e-06, + "loss": 0.9741, + "step": 18671 + }, + { + "epoch": 0.6, + "learning_rate": 7.371311804393098e-06, + "loss": 0.9751, + "step": 18672 + }, + { + "epoch": 0.6, + "learning_rate": 7.370312442362697e-06, + "loss": 1.0269, + "step": 18673 + }, + { + "epoch": 0.6, + "learning_rate": 7.369313108545933e-06, + "loss": 1.0278, + "step": 18674 + }, + { + "epoch": 0.6, + "learning_rate": 7.368313802953528e-06, + "loss": 0.9429, + "step": 18675 + }, + { + "epoch": 0.6, + "learning_rate": 7.367314525596204e-06, + "loss": 0.9805, + "step": 18676 + }, + { + "epoch": 0.6, + "learning_rate": 7.3663152764846824e-06, + "loss": 1.0244, + "step": 18677 + }, + { + "epoch": 0.6, + "learning_rate": 7.365316055629681e-06, + "loss": 1.0674, + "step": 18678 + }, + { + "epoch": 0.6, + "learning_rate": 7.364316863041923e-06, + "loss": 1.0278, + "step": 18679 + }, + { + "epoch": 0.6, + "learning_rate": 7.363317698732126e-06, + "loss": 0.875, + "step": 18680 + }, + { + "epoch": 0.6, + "learning_rate": 7.362318562711015e-06, + "loss": 1.1348, + "step": 18681 + }, + { + "epoch": 0.6, + "learning_rate": 7.361319454989306e-06, + "loss": 0.9971, + "step": 18682 + }, + { + "epoch": 0.6, + "learning_rate": 7.360320375577721e-06, + "loss": 0.9741, + "step": 18683 + }, + { + "epoch": 0.6, + "learning_rate": 7.359321324486974e-06, + "loss": 0.9463, + "step": 18684 + }, + { + "epoch": 0.6, + "learning_rate": 7.358322301727786e-06, + "loss": 1.0425, + "step": 18685 + }, + { + "epoch": 0.6, + "learning_rate": 7.357323307310878e-06, + "loss": 0.9565, + "step": 18686 + }, + { + "epoch": 0.6, + "learning_rate": 7.356324341246963e-06, + "loss": 0.8896, + "step": 18687 + }, + { + "epoch": 0.6, + "learning_rate": 7.355325403546764e-06, + "loss": 0.9551, + "step": 18688 + }, + { + "epoch": 0.6, + "learning_rate": 7.354326494220996e-06, + "loss": 0.9639, + "step": 18689 + }, + { + "epoch": 0.6, + "learning_rate": 7.353327613280381e-06, + "loss": 1.0537, + "step": 18690 + }, + { + "epoch": 0.6, + "learning_rate": 7.352328760735628e-06, + "loss": 1.0527, + "step": 18691 + }, + { + "epoch": 0.6, + "learning_rate": 7.351329936597456e-06, + "loss": 0.8882, + "step": 18692 + }, + { + "epoch": 0.6, + "learning_rate": 7.350331140876582e-06, + "loss": 0.5063, + "step": 18693 + }, + { + "epoch": 0.6, + "learning_rate": 7.349332373583723e-06, + "loss": 0.9604, + "step": 18694 + }, + { + "epoch": 0.6, + "learning_rate": 7.348333634729594e-06, + "loss": 0.9058, + "step": 18695 + }, + { + "epoch": 0.6, + "learning_rate": 7.34733492432491e-06, + "loss": 0.8557, + "step": 18696 + }, + { + "epoch": 0.6, + "learning_rate": 7.346336242380392e-06, + "loss": 1.0371, + "step": 18697 + }, + { + "epoch": 0.6, + "learning_rate": 7.345337588906744e-06, + "loss": 0.937, + "step": 18698 + }, + { + "epoch": 0.6, + "learning_rate": 7.344338963914686e-06, + "loss": 0.9692, + "step": 18699 + }, + { + "epoch": 0.6, + "learning_rate": 7.343340367414931e-06, + "loss": 0.9414, + "step": 18700 + }, + { + "epoch": 0.6, + "learning_rate": 7.342341799418193e-06, + "loss": 0.8735, + "step": 18701 + }, + { + "epoch": 0.6, + "learning_rate": 7.341343259935188e-06, + "loss": 0.9927, + "step": 18702 + }, + { + "epoch": 0.6, + "learning_rate": 7.340344748976626e-06, + "loss": 0.8721, + "step": 18703 + }, + { + "epoch": 0.6, + "learning_rate": 7.339346266553222e-06, + "loss": 1.0127, + "step": 18704 + }, + { + "epoch": 0.6, + "learning_rate": 7.3383478126756925e-06, + "loss": 0.9385, + "step": 18705 + }, + { + "epoch": 0.6, + "learning_rate": 7.33734938735474e-06, + "loss": 0.7603, + "step": 18706 + }, + { + "epoch": 0.6, + "learning_rate": 7.336350990601082e-06, + "loss": 1.0649, + "step": 18707 + }, + { + "epoch": 0.6, + "learning_rate": 7.335352622425431e-06, + "loss": 0.9419, + "step": 18708 + }, + { + "epoch": 0.6, + "learning_rate": 7.334354282838496e-06, + "loss": 0.936, + "step": 18709 + }, + { + "epoch": 0.6, + "learning_rate": 7.33335597185099e-06, + "loss": 1.0039, + "step": 18710 + }, + { + "epoch": 0.6, + "learning_rate": 7.332357689473624e-06, + "loss": 0.9106, + "step": 18711 + }, + { + "epoch": 0.6, + "learning_rate": 7.331359435717107e-06, + "loss": 0.9126, + "step": 18712 + }, + { + "epoch": 0.6, + "learning_rate": 7.33036121059215e-06, + "loss": 1.0244, + "step": 18713 + }, + { + "epoch": 0.6, + "learning_rate": 7.329363014109463e-06, + "loss": 0.8682, + "step": 18714 + }, + { + "epoch": 0.6, + "learning_rate": 7.328364846279754e-06, + "loss": 0.9429, + "step": 18715 + }, + { + "epoch": 0.6, + "learning_rate": 7.3273667071137345e-06, + "loss": 0.8789, + "step": 18716 + }, + { + "epoch": 0.6, + "learning_rate": 7.326368596622111e-06, + "loss": 0.9253, + "step": 18717 + }, + { + "epoch": 0.6, + "learning_rate": 7.325370514815594e-06, + "loss": 0.9644, + "step": 18718 + }, + { + "epoch": 0.6, + "learning_rate": 7.32437246170489e-06, + "loss": 1.0229, + "step": 18719 + }, + { + "epoch": 0.6, + "learning_rate": 7.323374437300712e-06, + "loss": 1.0029, + "step": 18720 + }, + { + "epoch": 0.6, + "learning_rate": 7.322376441613761e-06, + "loss": 0.8936, + "step": 18721 + }, + { + "epoch": 0.6, + "learning_rate": 7.321378474654749e-06, + "loss": 0.9067, + "step": 18722 + }, + { + "epoch": 0.6, + "learning_rate": 7.32038053643438e-06, + "loss": 0.9819, + "step": 18723 + }, + { + "epoch": 0.6, + "learning_rate": 7.319382626963362e-06, + "loss": 0.9512, + "step": 18724 + }, + { + "epoch": 0.6, + "learning_rate": 7.3183847462524e-06, + "loss": 0.9688, + "step": 18725 + }, + { + "epoch": 0.6, + "learning_rate": 7.317386894312204e-06, + "loss": 1.0337, + "step": 18726 + }, + { + "epoch": 0.6, + "learning_rate": 7.316389071153478e-06, + "loss": 0.9136, + "step": 18727 + }, + { + "epoch": 0.6, + "learning_rate": 7.31539127678693e-06, + "loss": 0.9741, + "step": 18728 + }, + { + "epoch": 0.6, + "learning_rate": 7.314393511223259e-06, + "loss": 0.9302, + "step": 18729 + }, + { + "epoch": 0.6, + "learning_rate": 7.313395774473173e-06, + "loss": 0.855, + "step": 18730 + }, + { + "epoch": 0.6, + "learning_rate": 7.312398066547377e-06, + "loss": 1.0493, + "step": 18731 + }, + { + "epoch": 0.6, + "learning_rate": 7.311400387456575e-06, + "loss": 1.0044, + "step": 18732 + }, + { + "epoch": 0.6, + "learning_rate": 7.310402737211471e-06, + "loss": 0.938, + "step": 18733 + }, + { + "epoch": 0.6, + "learning_rate": 7.3094051158227696e-06, + "loss": 1.0103, + "step": 18734 + }, + { + "epoch": 0.6, + "learning_rate": 7.3084075233011774e-06, + "loss": 0.9438, + "step": 18735 + }, + { + "epoch": 0.6, + "learning_rate": 7.307409959657389e-06, + "loss": 1.0586, + "step": 18736 + }, + { + "epoch": 0.6, + "learning_rate": 7.3064124249021115e-06, + "loss": 0.9795, + "step": 18737 + }, + { + "epoch": 0.6, + "learning_rate": 7.305414919046049e-06, + "loss": 0.9551, + "step": 18738 + }, + { + "epoch": 0.6, + "learning_rate": 7.304417442099901e-06, + "loss": 0.8521, + "step": 18739 + }, + { + "epoch": 0.6, + "learning_rate": 7.303419994074371e-06, + "loss": 0.9922, + "step": 18740 + }, + { + "epoch": 0.6, + "learning_rate": 7.30242257498016e-06, + "loss": 0.9409, + "step": 18741 + }, + { + "epoch": 0.6, + "learning_rate": 7.30142518482797e-06, + "loss": 1.0176, + "step": 18742 + }, + { + "epoch": 0.6, + "learning_rate": 7.300427823628505e-06, + "loss": 1.0371, + "step": 18743 + }, + { + "epoch": 0.6, + "learning_rate": 7.299430491392456e-06, + "loss": 0.9067, + "step": 18744 + }, + { + "epoch": 0.6, + "learning_rate": 7.2984331881305294e-06, + "loss": 0.9922, + "step": 18745 + }, + { + "epoch": 0.6, + "learning_rate": 7.2974359138534254e-06, + "loss": 0.9795, + "step": 18746 + }, + { + "epoch": 0.6, + "learning_rate": 7.296438668571843e-06, + "loss": 0.9858, + "step": 18747 + }, + { + "epoch": 0.6, + "learning_rate": 7.295441452296482e-06, + "loss": 0.9902, + "step": 18748 + }, + { + "epoch": 0.6, + "learning_rate": 7.294444265038041e-06, + "loss": 1.0156, + "step": 18749 + }, + { + "epoch": 0.6, + "learning_rate": 7.293447106807222e-06, + "loss": 1.0713, + "step": 18750 + }, + { + "epoch": 0.6, + "learning_rate": 7.2924499776147166e-06, + "loss": 0.936, + "step": 18751 + }, + { + "epoch": 0.6, + "learning_rate": 7.291452877471226e-06, + "loss": 1.0098, + "step": 18752 + }, + { + "epoch": 0.6, + "learning_rate": 7.290455806387447e-06, + "loss": 0.8862, + "step": 18753 + }, + { + "epoch": 0.6, + "learning_rate": 7.2894587643740796e-06, + "loss": 0.958, + "step": 18754 + }, + { + "epoch": 0.6, + "learning_rate": 7.288461751441821e-06, + "loss": 0.9756, + "step": 18755 + }, + { + "epoch": 0.6, + "learning_rate": 7.287464767601368e-06, + "loss": 0.8604, + "step": 18756 + }, + { + "epoch": 0.6, + "learning_rate": 7.286467812863414e-06, + "loss": 0.9341, + "step": 18757 + }, + { + "epoch": 0.6, + "learning_rate": 7.285470887238659e-06, + "loss": 1.0376, + "step": 18758 + }, + { + "epoch": 0.6, + "learning_rate": 7.284473990737795e-06, + "loss": 0.9595, + "step": 18759 + }, + { + "epoch": 0.6, + "learning_rate": 7.283477123371519e-06, + "loss": 1.0234, + "step": 18760 + }, + { + "epoch": 0.6, + "learning_rate": 7.28248028515053e-06, + "loss": 0.9429, + "step": 18761 + }, + { + "epoch": 0.6, + "learning_rate": 7.2814834760855176e-06, + "loss": 0.9741, + "step": 18762 + }, + { + "epoch": 0.6, + "learning_rate": 7.280486696187179e-06, + "loss": 1.0107, + "step": 18763 + }, + { + "epoch": 0.6, + "learning_rate": 7.2794899454662076e-06, + "loss": 0.9722, + "step": 18764 + }, + { + "epoch": 0.6, + "learning_rate": 7.2784932239333e-06, + "loss": 0.9648, + "step": 18765 + }, + { + "epoch": 0.6, + "learning_rate": 7.27749653159915e-06, + "loss": 0.9482, + "step": 18766 + }, + { + "epoch": 0.6, + "learning_rate": 7.276499868474446e-06, + "loss": 0.8804, + "step": 18767 + }, + { + "epoch": 0.6, + "learning_rate": 7.275503234569884e-06, + "loss": 0.5378, + "step": 18768 + }, + { + "epoch": 0.6, + "learning_rate": 7.274506629896156e-06, + "loss": 0.8643, + "step": 18769 + }, + { + "epoch": 0.6, + "learning_rate": 7.273510054463957e-06, + "loss": 0.9478, + "step": 18770 + }, + { + "epoch": 0.6, + "learning_rate": 7.2725135082839785e-06, + "loss": 1.0054, + "step": 18771 + }, + { + "epoch": 0.6, + "learning_rate": 7.271516991366909e-06, + "loss": 0.8662, + "step": 18772 + }, + { + "epoch": 0.6, + "learning_rate": 7.270520503723448e-06, + "loss": 0.9458, + "step": 18773 + }, + { + "epoch": 0.6, + "learning_rate": 7.269524045364275e-06, + "loss": 0.8613, + "step": 18774 + }, + { + "epoch": 0.6, + "learning_rate": 7.2685276163000896e-06, + "loss": 1.0234, + "step": 18775 + }, + { + "epoch": 0.6, + "learning_rate": 7.267531216541579e-06, + "loss": 1.0396, + "step": 18776 + }, + { + "epoch": 0.6, + "learning_rate": 7.266534846099434e-06, + "loss": 0.9858, + "step": 18777 + }, + { + "epoch": 0.6, + "learning_rate": 7.2655385049843455e-06, + "loss": 0.9604, + "step": 18778 + }, + { + "epoch": 0.6, + "learning_rate": 7.264542193207002e-06, + "loss": 0.9478, + "step": 18779 + }, + { + "epoch": 0.6, + "learning_rate": 7.263545910778098e-06, + "loss": 0.8186, + "step": 18780 + }, + { + "epoch": 0.6, + "learning_rate": 7.262549657708313e-06, + "loss": 0.9834, + "step": 18781 + }, + { + "epoch": 0.6, + "learning_rate": 7.261553434008339e-06, + "loss": 1.0107, + "step": 18782 + }, + { + "epoch": 0.6, + "learning_rate": 7.260557239688868e-06, + "loss": 0.9448, + "step": 18783 + }, + { + "epoch": 0.6, + "learning_rate": 7.259561074760586e-06, + "loss": 0.8672, + "step": 18784 + }, + { + "epoch": 0.6, + "learning_rate": 7.25856493923418e-06, + "loss": 0.9766, + "step": 18785 + }, + { + "epoch": 0.6, + "learning_rate": 7.2575688331203385e-06, + "loss": 1.0859, + "step": 18786 + }, + { + "epoch": 0.6, + "learning_rate": 7.256572756429748e-06, + "loss": 0.8306, + "step": 18787 + }, + { + "epoch": 0.6, + "learning_rate": 7.2555767091731e-06, + "loss": 0.8418, + "step": 18788 + }, + { + "epoch": 0.6, + "learning_rate": 7.254580691361073e-06, + "loss": 0.9863, + "step": 18789 + }, + { + "epoch": 0.6, + "learning_rate": 7.253584703004355e-06, + "loss": 1.0552, + "step": 18790 + }, + { + "epoch": 0.6, + "learning_rate": 7.252588744113634e-06, + "loss": 1.0312, + "step": 18791 + }, + { + "epoch": 0.6, + "learning_rate": 7.251592814699596e-06, + "loss": 0.9956, + "step": 18792 + }, + { + "epoch": 0.6, + "learning_rate": 7.250596914772925e-06, + "loss": 1.0381, + "step": 18793 + }, + { + "epoch": 0.6, + "learning_rate": 7.249601044344305e-06, + "loss": 0.9526, + "step": 18794 + }, + { + "epoch": 0.6, + "learning_rate": 7.248605203424426e-06, + "loss": 0.835, + "step": 18795 + }, + { + "epoch": 0.6, + "learning_rate": 7.247609392023967e-06, + "loss": 0.9487, + "step": 18796 + }, + { + "epoch": 0.6, + "learning_rate": 7.246613610153611e-06, + "loss": 0.9395, + "step": 18797 + }, + { + "epoch": 0.6, + "learning_rate": 7.245617857824043e-06, + "loss": 0.9243, + "step": 18798 + }, + { + "epoch": 0.6, + "learning_rate": 7.2446221350459486e-06, + "loss": 0.8975, + "step": 18799 + }, + { + "epoch": 0.6, + "learning_rate": 7.243626441830009e-06, + "loss": 1.002, + "step": 18800 + }, + { + "epoch": 0.6, + "learning_rate": 7.2426307781869075e-06, + "loss": 1.0225, + "step": 18801 + }, + { + "epoch": 0.6, + "learning_rate": 7.241635144127325e-06, + "loss": 0.9082, + "step": 18802 + }, + { + "epoch": 0.6, + "learning_rate": 7.240639539661947e-06, + "loss": 1.02, + "step": 18803 + }, + { + "epoch": 0.6, + "learning_rate": 7.23964396480145e-06, + "loss": 0.7817, + "step": 18804 + }, + { + "epoch": 0.6, + "learning_rate": 7.23864841955652e-06, + "loss": 0.98, + "step": 18805 + }, + { + "epoch": 0.6, + "learning_rate": 7.237652903937835e-06, + "loss": 0.9395, + "step": 18806 + }, + { + "epoch": 0.6, + "learning_rate": 7.236657417956076e-06, + "loss": 0.8584, + "step": 18807 + }, + { + "epoch": 0.6, + "learning_rate": 7.235661961621926e-06, + "loss": 0.9648, + "step": 18808 + }, + { + "epoch": 0.6, + "learning_rate": 7.234666534946063e-06, + "loss": 0.8794, + "step": 18809 + }, + { + "epoch": 0.6, + "learning_rate": 7.233671137939169e-06, + "loss": 0.9819, + "step": 18810 + }, + { + "epoch": 0.6, + "learning_rate": 7.232675770611925e-06, + "loss": 0.8687, + "step": 18811 + }, + { + "epoch": 0.6, + "learning_rate": 7.2316804329750034e-06, + "loss": 0.8843, + "step": 18812 + }, + { + "epoch": 0.6, + "learning_rate": 7.230685125039085e-06, + "loss": 0.9883, + "step": 18813 + }, + { + "epoch": 0.6, + "learning_rate": 7.2296898468148535e-06, + "loss": 1.0396, + "step": 18814 + }, + { + "epoch": 0.6, + "learning_rate": 7.228694598312982e-06, + "loss": 0.9375, + "step": 18815 + }, + { + "epoch": 0.6, + "learning_rate": 7.227699379544151e-06, + "loss": 1.0273, + "step": 18816 + }, + { + "epoch": 0.6, + "learning_rate": 7.226704190519036e-06, + "loss": 1.0063, + "step": 18817 + }, + { + "epoch": 0.6, + "learning_rate": 7.2257090312483225e-06, + "loss": 0.9468, + "step": 18818 + }, + { + "epoch": 0.6, + "learning_rate": 7.224713901742675e-06, + "loss": 0.9216, + "step": 18819 + }, + { + "epoch": 0.6, + "learning_rate": 7.223718802012776e-06, + "loss": 0.9634, + "step": 18820 + }, + { + "epoch": 0.6, + "learning_rate": 7.222723732069301e-06, + "loss": 0.8979, + "step": 18821 + }, + { + "epoch": 0.6, + "learning_rate": 7.221728691922927e-06, + "loss": 0.9751, + "step": 18822 + }, + { + "epoch": 0.6, + "learning_rate": 7.220733681584329e-06, + "loss": 0.9014, + "step": 18823 + }, + { + "epoch": 0.6, + "learning_rate": 7.219738701064182e-06, + "loss": 0.8506, + "step": 18824 + }, + { + "epoch": 0.6, + "learning_rate": 7.218743750373164e-06, + "loss": 0.9355, + "step": 18825 + }, + { + "epoch": 0.6, + "learning_rate": 7.21774882952195e-06, + "loss": 0.9443, + "step": 18826 + }, + { + "epoch": 0.6, + "learning_rate": 7.216753938521208e-06, + "loss": 1.1182, + "step": 18827 + }, + { + "epoch": 0.6, + "learning_rate": 7.215759077381616e-06, + "loss": 0.9458, + "step": 18828 + }, + { + "epoch": 0.6, + "learning_rate": 7.214764246113848e-06, + "loss": 0.8931, + "step": 18829 + }, + { + "epoch": 0.6, + "learning_rate": 7.213769444728575e-06, + "loss": 1.001, + "step": 18830 + }, + { + "epoch": 0.6, + "learning_rate": 7.212774673236474e-06, + "loss": 0.8584, + "step": 18831 + }, + { + "epoch": 0.6, + "learning_rate": 7.211779931648218e-06, + "loss": 0.9639, + "step": 18832 + }, + { + "epoch": 0.6, + "learning_rate": 7.210785219974479e-06, + "loss": 0.8979, + "step": 18833 + }, + { + "epoch": 0.6, + "learning_rate": 7.2097905382259225e-06, + "loss": 0.8022, + "step": 18834 + }, + { + "epoch": 0.6, + "learning_rate": 7.208795886413227e-06, + "loss": 0.8994, + "step": 18835 + }, + { + "epoch": 0.6, + "learning_rate": 7.207801264547063e-06, + "loss": 0.9858, + "step": 18836 + }, + { + "epoch": 0.6, + "learning_rate": 7.2068066726381e-06, + "loss": 0.9902, + "step": 18837 + }, + { + "epoch": 0.6, + "learning_rate": 7.2058121106970105e-06, + "loss": 1.0068, + "step": 18838 + }, + { + "epoch": 0.6, + "learning_rate": 7.2048175787344666e-06, + "loss": 1.0571, + "step": 18839 + }, + { + "epoch": 0.6, + "learning_rate": 7.203823076761136e-06, + "loss": 0.8701, + "step": 18840 + }, + { + "epoch": 0.6, + "learning_rate": 7.20282860478769e-06, + "loss": 0.9971, + "step": 18841 + }, + { + "epoch": 0.6, + "learning_rate": 7.201834162824794e-06, + "loss": 0.8037, + "step": 18842 + }, + { + "epoch": 0.6, + "learning_rate": 7.200839750883122e-06, + "loss": 0.8369, + "step": 18843 + }, + { + "epoch": 0.6, + "learning_rate": 7.199845368973342e-06, + "loss": 0.9878, + "step": 18844 + }, + { + "epoch": 0.6, + "learning_rate": 7.198851017106123e-06, + "loss": 0.9297, + "step": 18845 + }, + { + "epoch": 0.6, + "learning_rate": 7.19785669529213e-06, + "loss": 1.0049, + "step": 18846 + }, + { + "epoch": 0.6, + "learning_rate": 7.196862403542034e-06, + "loss": 1.0103, + "step": 18847 + }, + { + "epoch": 0.6, + "learning_rate": 7.195868141866504e-06, + "loss": 1.0376, + "step": 18848 + }, + { + "epoch": 0.6, + "learning_rate": 7.194873910276205e-06, + "loss": 0.9839, + "step": 18849 + }, + { + "epoch": 0.6, + "learning_rate": 7.193879708781805e-06, + "loss": 1.0044, + "step": 18850 + }, + { + "epoch": 0.6, + "learning_rate": 7.1928855373939675e-06, + "loss": 0.9678, + "step": 18851 + }, + { + "epoch": 0.6, + "learning_rate": 7.191891396123361e-06, + "loss": 0.917, + "step": 18852 + }, + { + "epoch": 0.6, + "learning_rate": 7.190897284980653e-06, + "loss": 1.0654, + "step": 18853 + }, + { + "epoch": 0.6, + "learning_rate": 7.189903203976509e-06, + "loss": 1.0049, + "step": 18854 + }, + { + "epoch": 0.6, + "learning_rate": 7.188909153121593e-06, + "loss": 1.0859, + "step": 18855 + }, + { + "epoch": 0.6, + "learning_rate": 7.187915132426574e-06, + "loss": 1.019, + "step": 18856 + }, + { + "epoch": 0.6, + "learning_rate": 7.1869211419021105e-06, + "loss": 1.0376, + "step": 18857 + }, + { + "epoch": 0.6, + "learning_rate": 7.1859271815588684e-06, + "loss": 0.873, + "step": 18858 + }, + { + "epoch": 0.6, + "learning_rate": 7.1849332514075135e-06, + "loss": 0.551, + "step": 18859 + }, + { + "epoch": 0.6, + "learning_rate": 7.183939351458708e-06, + "loss": 1.0146, + "step": 18860 + }, + { + "epoch": 0.6, + "learning_rate": 7.182945481723119e-06, + "loss": 1.04, + "step": 18861 + }, + { + "epoch": 0.6, + "learning_rate": 7.1819516422114064e-06, + "loss": 1.0229, + "step": 18862 + }, + { + "epoch": 0.6, + "learning_rate": 7.180957832934239e-06, + "loss": 1.1475, + "step": 18863 + }, + { + "epoch": 0.6, + "learning_rate": 7.179964053902269e-06, + "loss": 0.9761, + "step": 18864 + }, + { + "epoch": 0.6, + "learning_rate": 7.1789703051261626e-06, + "loss": 0.9077, + "step": 18865 + }, + { + "epoch": 0.6, + "learning_rate": 7.1779765866165825e-06, + "loss": 0.8105, + "step": 18866 + }, + { + "epoch": 0.6, + "learning_rate": 7.176982898384193e-06, + "loss": 0.9995, + "step": 18867 + }, + { + "epoch": 0.6, + "learning_rate": 7.175989240439649e-06, + "loss": 1.0273, + "step": 18868 + }, + { + "epoch": 0.6, + "learning_rate": 7.174995612793618e-06, + "loss": 0.9507, + "step": 18869 + }, + { + "epoch": 0.6, + "learning_rate": 7.174002015456755e-06, + "loss": 0.9321, + "step": 18870 + }, + { + "epoch": 0.6, + "learning_rate": 7.17300844843973e-06, + "loss": 0.9961, + "step": 18871 + }, + { + "epoch": 0.6, + "learning_rate": 7.172014911753189e-06, + "loss": 1.0342, + "step": 18872 + }, + { + "epoch": 0.6, + "learning_rate": 7.1710214054078e-06, + "loss": 0.9282, + "step": 18873 + }, + { + "epoch": 0.6, + "learning_rate": 7.1700279294142185e-06, + "loss": 0.9219, + "step": 18874 + }, + { + "epoch": 0.6, + "learning_rate": 7.169034483783106e-06, + "loss": 1.1592, + "step": 18875 + }, + { + "epoch": 0.6, + "learning_rate": 7.16804106852512e-06, + "loss": 0.9512, + "step": 18876 + }, + { + "epoch": 0.6, + "learning_rate": 7.167047683650921e-06, + "loss": 0.9009, + "step": 18877 + }, + { + "epoch": 0.6, + "learning_rate": 7.166054329171165e-06, + "loss": 0.7915, + "step": 18878 + }, + { + "epoch": 0.6, + "learning_rate": 7.165061005096509e-06, + "loss": 1.0068, + "step": 18879 + }, + { + "epoch": 0.6, + "learning_rate": 7.16406771143761e-06, + "loss": 0.8823, + "step": 18880 + }, + { + "epoch": 0.6, + "learning_rate": 7.163074448205127e-06, + "loss": 0.9185, + "step": 18881 + }, + { + "epoch": 0.6, + "learning_rate": 7.162081215409713e-06, + "loss": 0.9795, + "step": 18882 + }, + { + "epoch": 0.6, + "learning_rate": 7.161088013062029e-06, + "loss": 0.9556, + "step": 18883 + }, + { + "epoch": 0.6, + "learning_rate": 7.160094841172729e-06, + "loss": 0.96, + "step": 18884 + }, + { + "epoch": 0.6, + "learning_rate": 7.159101699752466e-06, + "loss": 0.9971, + "step": 18885 + }, + { + "epoch": 0.6, + "learning_rate": 7.158108588811902e-06, + "loss": 0.9253, + "step": 18886 + }, + { + "epoch": 0.6, + "learning_rate": 7.157115508361683e-06, + "loss": 0.8423, + "step": 18887 + }, + { + "epoch": 0.6, + "learning_rate": 7.15612245841247e-06, + "loss": 1.0869, + "step": 18888 + }, + { + "epoch": 0.6, + "learning_rate": 7.155129438974917e-06, + "loss": 0.8623, + "step": 18889 + }, + { + "epoch": 0.6, + "learning_rate": 7.154136450059674e-06, + "loss": 0.915, + "step": 18890 + }, + { + "epoch": 0.6, + "learning_rate": 7.153143491677398e-06, + "loss": 1.1006, + "step": 18891 + }, + { + "epoch": 0.6, + "learning_rate": 7.152150563838742e-06, + "loss": 0.8496, + "step": 18892 + }, + { + "epoch": 0.6, + "learning_rate": 7.151157666554357e-06, + "loss": 1.0146, + "step": 18893 + }, + { + "epoch": 0.6, + "learning_rate": 7.150164799834902e-06, + "loss": 0.9492, + "step": 18894 + }, + { + "epoch": 0.6, + "learning_rate": 7.1491719636910215e-06, + "loss": 0.9917, + "step": 18895 + }, + { + "epoch": 0.6, + "learning_rate": 7.148179158133369e-06, + "loss": 0.9048, + "step": 18896 + }, + { + "epoch": 0.6, + "learning_rate": 7.147186383172599e-06, + "loss": 0.9814, + "step": 18897 + }, + { + "epoch": 0.6, + "learning_rate": 7.146193638819362e-06, + "loss": 0.8965, + "step": 18898 + }, + { + "epoch": 0.6, + "learning_rate": 7.145200925084308e-06, + "loss": 0.9375, + "step": 18899 + }, + { + "epoch": 0.6, + "learning_rate": 7.14420824197809e-06, + "loss": 1.063, + "step": 18900 + }, + { + "epoch": 0.6, + "learning_rate": 7.14321558951136e-06, + "loss": 1.104, + "step": 18901 + }, + { + "epoch": 0.6, + "learning_rate": 7.142222967694761e-06, + "loss": 0.9438, + "step": 18902 + }, + { + "epoch": 0.6, + "learning_rate": 7.1412303765389455e-06, + "loss": 1.0273, + "step": 18903 + }, + { + "epoch": 0.6, + "learning_rate": 7.140237816054566e-06, + "loss": 0.9556, + "step": 18904 + }, + { + "epoch": 0.6, + "learning_rate": 7.139245286252269e-06, + "loss": 1.0312, + "step": 18905 + }, + { + "epoch": 0.6, + "learning_rate": 7.138252787142703e-06, + "loss": 0.9429, + "step": 18906 + }, + { + "epoch": 0.6, + "learning_rate": 7.1372603187365185e-06, + "loss": 0.9688, + "step": 18907 + }, + { + "epoch": 0.6, + "learning_rate": 7.136267881044363e-06, + "loss": 0.8374, + "step": 18908 + }, + { + "epoch": 0.6, + "learning_rate": 7.135275474076887e-06, + "loss": 0.9922, + "step": 18909 + }, + { + "epoch": 0.6, + "learning_rate": 7.134283097844731e-06, + "loss": 0.8486, + "step": 18910 + }, + { + "epoch": 0.6, + "learning_rate": 7.133290752358546e-06, + "loss": 0.9702, + "step": 18911 + }, + { + "epoch": 0.6, + "learning_rate": 7.1322984376289775e-06, + "loss": 0.9941, + "step": 18912 + }, + { + "epoch": 0.6, + "learning_rate": 7.131306153666673e-06, + "loss": 0.9561, + "step": 18913 + }, + { + "epoch": 0.6, + "learning_rate": 7.130313900482279e-06, + "loss": 0.9199, + "step": 18914 + }, + { + "epoch": 0.6, + "learning_rate": 7.12932167808644e-06, + "loss": 0.8999, + "step": 18915 + }, + { + "epoch": 0.6, + "learning_rate": 7.128329486489807e-06, + "loss": 0.8936, + "step": 18916 + }, + { + "epoch": 0.6, + "learning_rate": 7.127337325703016e-06, + "loss": 0.9907, + "step": 18917 + }, + { + "epoch": 0.61, + "learning_rate": 7.126345195736717e-06, + "loss": 0.9365, + "step": 18918 + }, + { + "epoch": 0.61, + "learning_rate": 7.125353096601554e-06, + "loss": 0.9165, + "step": 18919 + }, + { + "epoch": 0.61, + "learning_rate": 7.12436102830817e-06, + "loss": 0.9521, + "step": 18920 + }, + { + "epoch": 0.61, + "learning_rate": 7.12336899086721e-06, + "loss": 0.9468, + "step": 18921 + }, + { + "epoch": 0.61, + "learning_rate": 7.122376984289316e-06, + "loss": 0.9976, + "step": 18922 + }, + { + "epoch": 0.61, + "learning_rate": 7.121385008585135e-06, + "loss": 0.8809, + "step": 18923 + }, + { + "epoch": 0.61, + "learning_rate": 7.120393063765307e-06, + "loss": 1.0488, + "step": 18924 + }, + { + "epoch": 0.61, + "learning_rate": 7.1194011498404725e-06, + "loss": 0.9624, + "step": 18925 + }, + { + "epoch": 0.61, + "learning_rate": 7.118409266821275e-06, + "loss": 0.8867, + "step": 18926 + }, + { + "epoch": 0.61, + "learning_rate": 7.117417414718357e-06, + "loss": 0.9565, + "step": 18927 + }, + { + "epoch": 0.61, + "learning_rate": 7.116425593542362e-06, + "loss": 0.876, + "step": 18928 + }, + { + "epoch": 0.61, + "learning_rate": 7.115433803303927e-06, + "loss": 0.8286, + "step": 18929 + }, + { + "epoch": 0.61, + "learning_rate": 7.1144420440136945e-06, + "loss": 0.959, + "step": 18930 + }, + { + "epoch": 0.61, + "learning_rate": 7.113450315682309e-06, + "loss": 0.895, + "step": 18931 + }, + { + "epoch": 0.61, + "learning_rate": 7.112458618320403e-06, + "loss": 1.0273, + "step": 18932 + }, + { + "epoch": 0.61, + "learning_rate": 7.111466951938622e-06, + "loss": 1.0596, + "step": 18933 + }, + { + "epoch": 0.61, + "learning_rate": 7.110475316547602e-06, + "loss": 1.0093, + "step": 18934 + }, + { + "epoch": 0.61, + "learning_rate": 7.109483712157985e-06, + "loss": 0.8579, + "step": 18935 + }, + { + "epoch": 0.61, + "learning_rate": 7.108492138780409e-06, + "loss": 0.9561, + "step": 18936 + }, + { + "epoch": 0.61, + "learning_rate": 7.1075005964255106e-06, + "loss": 1.0054, + "step": 18937 + }, + { + "epoch": 0.61, + "learning_rate": 7.1065090851039295e-06, + "loss": 0.8535, + "step": 18938 + }, + { + "epoch": 0.61, + "learning_rate": 7.1055176048263085e-06, + "loss": 0.8916, + "step": 18939 + }, + { + "epoch": 0.61, + "learning_rate": 7.104526155603278e-06, + "loss": 1.0234, + "step": 18940 + }, + { + "epoch": 0.61, + "learning_rate": 7.103534737445474e-06, + "loss": 0.9111, + "step": 18941 + }, + { + "epoch": 0.61, + "learning_rate": 7.102543350363537e-06, + "loss": 0.897, + "step": 18942 + }, + { + "epoch": 0.61, + "learning_rate": 7.101551994368105e-06, + "loss": 1.0654, + "step": 18943 + }, + { + "epoch": 0.61, + "learning_rate": 7.100560669469812e-06, + "loss": 0.9121, + "step": 18944 + }, + { + "epoch": 0.61, + "learning_rate": 7.099569375679292e-06, + "loss": 0.9395, + "step": 18945 + }, + { + "epoch": 0.61, + "learning_rate": 7.098578113007184e-06, + "loss": 0.9365, + "step": 18946 + }, + { + "epoch": 0.61, + "learning_rate": 7.0975868814641255e-06, + "loss": 0.4897, + "step": 18947 + }, + { + "epoch": 0.61, + "learning_rate": 7.096595681060744e-06, + "loss": 0.9668, + "step": 18948 + }, + { + "epoch": 0.61, + "learning_rate": 7.095604511807677e-06, + "loss": 0.9194, + "step": 18949 + }, + { + "epoch": 0.61, + "learning_rate": 7.09461337371556e-06, + "loss": 0.9785, + "step": 18950 + }, + { + "epoch": 0.61, + "learning_rate": 7.093622266795025e-06, + "loss": 0.8911, + "step": 18951 + }, + { + "epoch": 0.61, + "learning_rate": 7.092631191056707e-06, + "loss": 0.9761, + "step": 18952 + }, + { + "epoch": 0.61, + "learning_rate": 7.091640146511238e-06, + "loss": 1.001, + "step": 18953 + }, + { + "epoch": 0.61, + "learning_rate": 7.090649133169256e-06, + "loss": 0.9307, + "step": 18954 + }, + { + "epoch": 0.61, + "learning_rate": 7.089658151041384e-06, + "loss": 0.9771, + "step": 18955 + }, + { + "epoch": 0.61, + "learning_rate": 7.0886672001382595e-06, + "loss": 1.0596, + "step": 18956 + }, + { + "epoch": 0.61, + "learning_rate": 7.087676280470514e-06, + "loss": 0.9814, + "step": 18957 + }, + { + "epoch": 0.61, + "learning_rate": 7.0866853920487796e-06, + "loss": 1.0171, + "step": 18958 + }, + { + "epoch": 0.61, + "learning_rate": 7.0856945348836845e-06, + "loss": 1.0107, + "step": 18959 + }, + { + "epoch": 0.61, + "learning_rate": 7.084703708985864e-06, + "loss": 0.9346, + "step": 18960 + }, + { + "epoch": 0.61, + "learning_rate": 7.083712914365946e-06, + "loss": 0.9971, + "step": 18961 + }, + { + "epoch": 0.61, + "learning_rate": 7.0827221510345636e-06, + "loss": 0.8926, + "step": 18962 + }, + { + "epoch": 0.61, + "learning_rate": 7.08173141900234e-06, + "loss": 0.9048, + "step": 18963 + }, + { + "epoch": 0.61, + "learning_rate": 7.080740718279908e-06, + "loss": 0.9761, + "step": 18964 + }, + { + "epoch": 0.61, + "learning_rate": 7.079750048877899e-06, + "loss": 0.8882, + "step": 18965 + }, + { + "epoch": 0.61, + "learning_rate": 7.078759410806939e-06, + "loss": 1.0508, + "step": 18966 + }, + { + "epoch": 0.61, + "learning_rate": 7.0777688040776595e-06, + "loss": 0.9297, + "step": 18967 + }, + { + "epoch": 0.61, + "learning_rate": 7.076778228700685e-06, + "loss": 1.0322, + "step": 18968 + }, + { + "epoch": 0.61, + "learning_rate": 7.0757876846866454e-06, + "loss": 0.9165, + "step": 18969 + }, + { + "epoch": 0.61, + "learning_rate": 7.0747971720461674e-06, + "loss": 0.8877, + "step": 18970 + }, + { + "epoch": 0.61, + "learning_rate": 7.073806690789877e-06, + "loss": 1.0439, + "step": 18971 + }, + { + "epoch": 0.61, + "learning_rate": 7.072816240928403e-06, + "loss": 0.9058, + "step": 18972 + }, + { + "epoch": 0.61, + "learning_rate": 7.071825822472374e-06, + "loss": 0.9131, + "step": 18973 + }, + { + "epoch": 0.61, + "learning_rate": 7.070835435432409e-06, + "loss": 1.0547, + "step": 18974 + }, + { + "epoch": 0.61, + "learning_rate": 7.069845079819139e-06, + "loss": 0.9131, + "step": 18975 + }, + { + "epoch": 0.61, + "learning_rate": 7.068854755643188e-06, + "loss": 0.9922, + "step": 18976 + }, + { + "epoch": 0.61, + "learning_rate": 7.067864462915184e-06, + "loss": 1.0884, + "step": 18977 + }, + { + "epoch": 0.61, + "learning_rate": 7.066874201645748e-06, + "loss": 0.9492, + "step": 18978 + }, + { + "epoch": 0.61, + "learning_rate": 7.065883971845505e-06, + "loss": 1.0571, + "step": 18979 + }, + { + "epoch": 0.61, + "learning_rate": 7.064893773525079e-06, + "loss": 0.9907, + "step": 18980 + }, + { + "epoch": 0.61, + "learning_rate": 7.063903606695095e-06, + "loss": 1.0225, + "step": 18981 + }, + { + "epoch": 0.61, + "learning_rate": 7.062913471366175e-06, + "loss": 0.9109, + "step": 18982 + }, + { + "epoch": 0.61, + "learning_rate": 7.061923367548944e-06, + "loss": 1.0869, + "step": 18983 + }, + { + "epoch": 0.61, + "learning_rate": 7.060933295254027e-06, + "loss": 1.0479, + "step": 18984 + }, + { + "epoch": 0.61, + "learning_rate": 7.059943254492038e-06, + "loss": 0.9282, + "step": 18985 + }, + { + "epoch": 0.61, + "learning_rate": 7.058953245273605e-06, + "loss": 1.0293, + "step": 18986 + }, + { + "epoch": 0.61, + "learning_rate": 7.057963267609348e-06, + "loss": 1.0547, + "step": 18987 + }, + { + "epoch": 0.61, + "learning_rate": 7.05697332150989e-06, + "loss": 0.9429, + "step": 18988 + }, + { + "epoch": 0.61, + "learning_rate": 7.05598340698585e-06, + "loss": 0.5293, + "step": 18989 + }, + { + "epoch": 0.61, + "learning_rate": 7.054993524047851e-06, + "loss": 0.9609, + "step": 18990 + }, + { + "epoch": 0.61, + "learning_rate": 7.0540036727065105e-06, + "loss": 0.9678, + "step": 18991 + }, + { + "epoch": 0.61, + "learning_rate": 7.053013852972455e-06, + "loss": 0.8813, + "step": 18992 + }, + { + "epoch": 0.61, + "learning_rate": 7.052024064856296e-06, + "loss": 0.9277, + "step": 18993 + }, + { + "epoch": 0.61, + "learning_rate": 7.051034308368655e-06, + "loss": 0.9404, + "step": 18994 + }, + { + "epoch": 0.61, + "learning_rate": 7.050044583520153e-06, + "loss": 0.9648, + "step": 18995 + }, + { + "epoch": 0.61, + "learning_rate": 7.049054890321407e-06, + "loss": 1.043, + "step": 18996 + }, + { + "epoch": 0.61, + "learning_rate": 7.048065228783036e-06, + "loss": 0.792, + "step": 18997 + }, + { + "epoch": 0.61, + "learning_rate": 7.047075598915658e-06, + "loss": 0.9463, + "step": 18998 + }, + { + "epoch": 0.61, + "learning_rate": 7.046086000729896e-06, + "loss": 0.9331, + "step": 18999 + }, + { + "epoch": 0.61, + "learning_rate": 7.045096434236357e-06, + "loss": 1.0728, + "step": 19000 + }, + { + "epoch": 0.61, + "learning_rate": 7.0441068994456644e-06, + "loss": 0.9961, + "step": 19001 + }, + { + "epoch": 0.61, + "learning_rate": 7.043117396368431e-06, + "loss": 0.9126, + "step": 19002 + }, + { + "epoch": 0.61, + "learning_rate": 7.0421279250152765e-06, + "loss": 0.8979, + "step": 19003 + }, + { + "epoch": 0.61, + "learning_rate": 7.041138485396815e-06, + "loss": 0.9595, + "step": 19004 + }, + { + "epoch": 0.61, + "learning_rate": 7.0401490775236635e-06, + "loss": 1.0269, + "step": 19005 + }, + { + "epoch": 0.61, + "learning_rate": 7.039159701406439e-06, + "loss": 0.9858, + "step": 19006 + }, + { + "epoch": 0.61, + "learning_rate": 7.038170357055754e-06, + "loss": 0.9717, + "step": 19007 + }, + { + "epoch": 0.61, + "learning_rate": 7.0371810444822205e-06, + "loss": 1.0913, + "step": 19008 + }, + { + "epoch": 0.61, + "learning_rate": 7.036191763696456e-06, + "loss": 0.8694, + "step": 19009 + }, + { + "epoch": 0.61, + "learning_rate": 7.035202514709073e-06, + "loss": 1.0708, + "step": 19010 + }, + { + "epoch": 0.61, + "learning_rate": 7.034213297530685e-06, + "loss": 0.8818, + "step": 19011 + }, + { + "epoch": 0.61, + "learning_rate": 7.0332241121719105e-06, + "loss": 0.9438, + "step": 19012 + }, + { + "epoch": 0.61, + "learning_rate": 7.032234958643354e-06, + "loss": 0.9492, + "step": 19013 + }, + { + "epoch": 0.61, + "learning_rate": 7.031245836955634e-06, + "loss": 0.8257, + "step": 19014 + }, + { + "epoch": 0.61, + "learning_rate": 7.030256747119358e-06, + "loss": 1.0293, + "step": 19015 + }, + { + "epoch": 0.61, + "learning_rate": 7.0292676891451406e-06, + "loss": 0.9902, + "step": 19016 + }, + { + "epoch": 0.61, + "learning_rate": 7.028278663043596e-06, + "loss": 0.9619, + "step": 19017 + }, + { + "epoch": 0.61, + "learning_rate": 7.02728966882533e-06, + "loss": 0.918, + "step": 19018 + }, + { + "epoch": 0.61, + "learning_rate": 7.0263007065009545e-06, + "loss": 0.9058, + "step": 19019 + }, + { + "epoch": 0.61, + "learning_rate": 7.025311776081083e-06, + "loss": 0.9907, + "step": 19020 + }, + { + "epoch": 0.61, + "learning_rate": 7.024322877576324e-06, + "loss": 1.0054, + "step": 19021 + }, + { + "epoch": 0.61, + "learning_rate": 7.02333401099729e-06, + "loss": 1.0303, + "step": 19022 + }, + { + "epoch": 0.61, + "learning_rate": 7.022345176354585e-06, + "loss": 0.9622, + "step": 19023 + }, + { + "epoch": 0.61, + "learning_rate": 7.021356373658818e-06, + "loss": 0.9189, + "step": 19024 + }, + { + "epoch": 0.61, + "learning_rate": 7.0203676029206034e-06, + "loss": 1.0532, + "step": 19025 + }, + { + "epoch": 0.61, + "learning_rate": 7.019378864150545e-06, + "loss": 0.9624, + "step": 19026 + }, + { + "epoch": 0.61, + "learning_rate": 7.018390157359254e-06, + "loss": 0.8862, + "step": 19027 + }, + { + "epoch": 0.61, + "learning_rate": 7.017401482557337e-06, + "loss": 0.9688, + "step": 19028 + }, + { + "epoch": 0.61, + "learning_rate": 7.0164128397554e-06, + "loss": 1.0229, + "step": 19029 + }, + { + "epoch": 0.61, + "learning_rate": 7.0154242289640565e-06, + "loss": 0.9155, + "step": 19030 + }, + { + "epoch": 0.61, + "learning_rate": 7.014435650193902e-06, + "loss": 1.0088, + "step": 19031 + }, + { + "epoch": 0.61, + "learning_rate": 7.013447103455549e-06, + "loss": 0.8745, + "step": 19032 + }, + { + "epoch": 0.61, + "learning_rate": 7.012458588759603e-06, + "loss": 0.8823, + "step": 19033 + }, + { + "epoch": 0.61, + "learning_rate": 7.0114701061166715e-06, + "loss": 1.0024, + "step": 19034 + }, + { + "epoch": 0.61, + "learning_rate": 7.010481655537358e-06, + "loss": 0.9604, + "step": 19035 + }, + { + "epoch": 0.61, + "learning_rate": 7.0094932370322655e-06, + "loss": 1.0698, + "step": 19036 + }, + { + "epoch": 0.61, + "learning_rate": 7.008504850612008e-06, + "loss": 1.019, + "step": 19037 + }, + { + "epoch": 0.61, + "learning_rate": 7.007516496287176e-06, + "loss": 1.02, + "step": 19038 + }, + { + "epoch": 0.61, + "learning_rate": 7.006528174068381e-06, + "loss": 1.0083, + "step": 19039 + }, + { + "epoch": 0.61, + "learning_rate": 7.005539883966226e-06, + "loss": 0.9507, + "step": 19040 + }, + { + "epoch": 0.61, + "learning_rate": 7.004551625991314e-06, + "loss": 0.9053, + "step": 19041 + }, + { + "epoch": 0.61, + "learning_rate": 7.003563400154248e-06, + "loss": 1.0186, + "step": 19042 + }, + { + "epoch": 0.61, + "learning_rate": 7.002575206465629e-06, + "loss": 0.9194, + "step": 19043 + }, + { + "epoch": 0.61, + "learning_rate": 7.001587044936061e-06, + "loss": 1.063, + "step": 19044 + }, + { + "epoch": 0.61, + "learning_rate": 7.000598915576151e-06, + "loss": 0.8799, + "step": 19045 + }, + { + "epoch": 0.61, + "learning_rate": 6.999610818396489e-06, + "loss": 0.9438, + "step": 19046 + }, + { + "epoch": 0.61, + "learning_rate": 6.998622753407685e-06, + "loss": 0.9482, + "step": 19047 + }, + { + "epoch": 0.61, + "learning_rate": 6.997634720620334e-06, + "loss": 1.0356, + "step": 19048 + }, + { + "epoch": 0.61, + "learning_rate": 6.996646720045042e-06, + "loss": 1.022, + "step": 19049 + }, + { + "epoch": 0.61, + "learning_rate": 6.995658751692406e-06, + "loss": 0.9854, + "step": 19050 + }, + { + "epoch": 0.61, + "learning_rate": 6.994670815573028e-06, + "loss": 0.8955, + "step": 19051 + }, + { + "epoch": 0.61, + "learning_rate": 6.993682911697507e-06, + "loss": 0.9404, + "step": 19052 + }, + { + "epoch": 0.61, + "learning_rate": 6.9926950400764386e-06, + "loss": 0.8662, + "step": 19053 + }, + { + "epoch": 0.61, + "learning_rate": 6.991707200720424e-06, + "loss": 0.9751, + "step": 19054 + }, + { + "epoch": 0.61, + "learning_rate": 6.990719393640062e-06, + "loss": 0.9595, + "step": 19055 + }, + { + "epoch": 0.61, + "learning_rate": 6.989731618845952e-06, + "loss": 0.9395, + "step": 19056 + }, + { + "epoch": 0.61, + "learning_rate": 6.988743876348688e-06, + "loss": 1.0425, + "step": 19057 + }, + { + "epoch": 0.61, + "learning_rate": 6.98775616615887e-06, + "loss": 0.9561, + "step": 19058 + }, + { + "epoch": 0.61, + "learning_rate": 6.986768488287095e-06, + "loss": 0.9077, + "step": 19059 + }, + { + "epoch": 0.61, + "learning_rate": 6.98578084274396e-06, + "loss": 0.9873, + "step": 19060 + }, + { + "epoch": 0.61, + "learning_rate": 6.984793229540059e-06, + "loss": 0.9961, + "step": 19061 + }, + { + "epoch": 0.61, + "learning_rate": 6.983805648685993e-06, + "loss": 0.894, + "step": 19062 + }, + { + "epoch": 0.61, + "learning_rate": 6.9828181001923505e-06, + "loss": 0.9565, + "step": 19063 + }, + { + "epoch": 0.61, + "learning_rate": 6.981830584069731e-06, + "loss": 0.7935, + "step": 19064 + }, + { + "epoch": 0.61, + "learning_rate": 6.9808431003287295e-06, + "loss": 0.9399, + "step": 19065 + }, + { + "epoch": 0.61, + "learning_rate": 6.979855648979938e-06, + "loss": 1.0444, + "step": 19066 + }, + { + "epoch": 0.61, + "learning_rate": 6.97886823003396e-06, + "loss": 0.9805, + "step": 19067 + }, + { + "epoch": 0.61, + "learning_rate": 6.977880843501376e-06, + "loss": 1.0513, + "step": 19068 + }, + { + "epoch": 0.61, + "learning_rate": 6.976893489392787e-06, + "loss": 1.0908, + "step": 19069 + }, + { + "epoch": 0.61, + "learning_rate": 6.975906167718784e-06, + "loss": 0.9326, + "step": 19070 + }, + { + "epoch": 0.61, + "learning_rate": 6.974918878489963e-06, + "loss": 0.8579, + "step": 19071 + }, + { + "epoch": 0.61, + "learning_rate": 6.973931621716911e-06, + "loss": 0.9653, + "step": 19072 + }, + { + "epoch": 0.61, + "learning_rate": 6.972944397410226e-06, + "loss": 0.895, + "step": 19073 + }, + { + "epoch": 0.61, + "learning_rate": 6.971957205580497e-06, + "loss": 1.0156, + "step": 19074 + }, + { + "epoch": 0.61, + "learning_rate": 6.970970046238321e-06, + "loss": 1.0288, + "step": 19075 + }, + { + "epoch": 0.61, + "learning_rate": 6.969982919394279e-06, + "loss": 0.9058, + "step": 19076 + }, + { + "epoch": 0.61, + "learning_rate": 6.968995825058966e-06, + "loss": 0.8906, + "step": 19077 + }, + { + "epoch": 0.61, + "learning_rate": 6.9680087632429745e-06, + "loss": 1.0659, + "step": 19078 + }, + { + "epoch": 0.61, + "learning_rate": 6.9670217339568935e-06, + "loss": 1.0332, + "step": 19079 + }, + { + "epoch": 0.61, + "learning_rate": 6.966034737211311e-06, + "loss": 0.918, + "step": 19080 + }, + { + "epoch": 0.61, + "learning_rate": 6.96504777301682e-06, + "loss": 1.0239, + "step": 19081 + }, + { + "epoch": 0.61, + "learning_rate": 6.9640608413840105e-06, + "loss": 0.9194, + "step": 19082 + }, + { + "epoch": 0.61, + "learning_rate": 6.963073942323464e-06, + "loss": 1.0063, + "step": 19083 + }, + { + "epoch": 0.61, + "learning_rate": 6.962087075845772e-06, + "loss": 0.8442, + "step": 19084 + }, + { + "epoch": 0.61, + "learning_rate": 6.961100241961525e-06, + "loss": 0.8735, + "step": 19085 + }, + { + "epoch": 0.61, + "learning_rate": 6.9601134406813085e-06, + "loss": 0.918, + "step": 19086 + }, + { + "epoch": 0.61, + "learning_rate": 6.95912667201571e-06, + "loss": 1.0054, + "step": 19087 + }, + { + "epoch": 0.61, + "learning_rate": 6.958139935975319e-06, + "loss": 0.9233, + "step": 19088 + }, + { + "epoch": 0.61, + "learning_rate": 6.957153232570717e-06, + "loss": 0.7871, + "step": 19089 + }, + { + "epoch": 0.61, + "learning_rate": 6.956166561812498e-06, + "loss": 0.9453, + "step": 19090 + }, + { + "epoch": 0.61, + "learning_rate": 6.955179923711241e-06, + "loss": 0.8696, + "step": 19091 + }, + { + "epoch": 0.61, + "learning_rate": 6.954193318277531e-06, + "loss": 0.9487, + "step": 19092 + }, + { + "epoch": 0.61, + "learning_rate": 6.953206745521958e-06, + "loss": 0.8071, + "step": 19093 + }, + { + "epoch": 0.61, + "learning_rate": 6.952220205455103e-06, + "loss": 1.0459, + "step": 19094 + }, + { + "epoch": 0.61, + "learning_rate": 6.951233698087554e-06, + "loss": 1.0151, + "step": 19095 + }, + { + "epoch": 0.61, + "learning_rate": 6.950247223429892e-06, + "loss": 0.9375, + "step": 19096 + }, + { + "epoch": 0.61, + "learning_rate": 6.9492607814927034e-06, + "loss": 0.9722, + "step": 19097 + }, + { + "epoch": 0.61, + "learning_rate": 6.94827437228657e-06, + "loss": 0.939, + "step": 19098 + }, + { + "epoch": 0.61, + "learning_rate": 6.947287995822073e-06, + "loss": 0.9766, + "step": 19099 + }, + { + "epoch": 0.61, + "learning_rate": 6.946301652109798e-06, + "loss": 1.0576, + "step": 19100 + }, + { + "epoch": 0.61, + "learning_rate": 6.945315341160328e-06, + "loss": 0.9727, + "step": 19101 + }, + { + "epoch": 0.61, + "learning_rate": 6.944329062984243e-06, + "loss": 1.0542, + "step": 19102 + }, + { + "epoch": 0.61, + "learning_rate": 6.943342817592125e-06, + "loss": 0.8887, + "step": 19103 + }, + { + "epoch": 0.61, + "learning_rate": 6.942356604994553e-06, + "loss": 1.0059, + "step": 19104 + }, + { + "epoch": 0.61, + "learning_rate": 6.941370425202116e-06, + "loss": 0.9307, + "step": 19105 + }, + { + "epoch": 0.61, + "learning_rate": 6.9403842782253874e-06, + "loss": 0.8525, + "step": 19106 + }, + { + "epoch": 0.61, + "learning_rate": 6.9393981640749486e-06, + "loss": 1.0117, + "step": 19107 + }, + { + "epoch": 0.61, + "learning_rate": 6.938412082761379e-06, + "loss": 0.9165, + "step": 19108 + }, + { + "epoch": 0.61, + "learning_rate": 6.93742603429526e-06, + "loss": 1.0557, + "step": 19109 + }, + { + "epoch": 0.61, + "learning_rate": 6.936440018687169e-06, + "loss": 1.0298, + "step": 19110 + }, + { + "epoch": 0.61, + "learning_rate": 6.935454035947688e-06, + "loss": 0.9292, + "step": 19111 + }, + { + "epoch": 0.61, + "learning_rate": 6.934468086087392e-06, + "loss": 0.9761, + "step": 19112 + }, + { + "epoch": 0.61, + "learning_rate": 6.933482169116865e-06, + "loss": 1.0317, + "step": 19113 + }, + { + "epoch": 0.61, + "learning_rate": 6.932496285046677e-06, + "loss": 1.0132, + "step": 19114 + }, + { + "epoch": 0.61, + "learning_rate": 6.931510433887407e-06, + "loss": 0.9839, + "step": 19115 + }, + { + "epoch": 0.61, + "learning_rate": 6.930524615649636e-06, + "loss": 0.8916, + "step": 19116 + }, + { + "epoch": 0.61, + "learning_rate": 6.9295388303439385e-06, + "loss": 0.958, + "step": 19117 + }, + { + "epoch": 0.61, + "learning_rate": 6.92855307798089e-06, + "loss": 0.9248, + "step": 19118 + }, + { + "epoch": 0.61, + "learning_rate": 6.92756735857107e-06, + "loss": 0.9917, + "step": 19119 + }, + { + "epoch": 0.61, + "learning_rate": 6.9265816721250546e-06, + "loss": 1.063, + "step": 19120 + }, + { + "epoch": 0.61, + "learning_rate": 6.925596018653411e-06, + "loss": 0.9375, + "step": 19121 + }, + { + "epoch": 0.61, + "learning_rate": 6.924610398166721e-06, + "loss": 0.981, + "step": 19122 + }, + { + "epoch": 0.61, + "learning_rate": 6.923624810675558e-06, + "loss": 0.9399, + "step": 19123 + }, + { + "epoch": 0.61, + "learning_rate": 6.922639256190496e-06, + "loss": 0.9214, + "step": 19124 + }, + { + "epoch": 0.61, + "learning_rate": 6.921653734722107e-06, + "loss": 0.9658, + "step": 19125 + }, + { + "epoch": 0.61, + "learning_rate": 6.920668246280968e-06, + "loss": 0.9019, + "step": 19126 + }, + { + "epoch": 0.61, + "learning_rate": 6.919682790877651e-06, + "loss": 0.9902, + "step": 19127 + }, + { + "epoch": 0.61, + "learning_rate": 6.918697368522731e-06, + "loss": 0.8564, + "step": 19128 + }, + { + "epoch": 0.61, + "learning_rate": 6.9177119792267755e-06, + "loss": 0.9756, + "step": 19129 + }, + { + "epoch": 0.61, + "learning_rate": 6.916726623000358e-06, + "loss": 1.0024, + "step": 19130 + }, + { + "epoch": 0.61, + "learning_rate": 6.915741299854053e-06, + "loss": 1.0073, + "step": 19131 + }, + { + "epoch": 0.61, + "learning_rate": 6.914756009798429e-06, + "loss": 0.9487, + "step": 19132 + }, + { + "epoch": 0.61, + "learning_rate": 6.91377075284406e-06, + "loss": 0.9961, + "step": 19133 + }, + { + "epoch": 0.61, + "learning_rate": 6.912785529001516e-06, + "loss": 0.8271, + "step": 19134 + }, + { + "epoch": 0.61, + "learning_rate": 6.911800338281367e-06, + "loss": 0.9805, + "step": 19135 + }, + { + "epoch": 0.61, + "learning_rate": 6.9108151806941795e-06, + "loss": 0.9688, + "step": 19136 + }, + { + "epoch": 0.61, + "learning_rate": 6.909830056250527e-06, + "loss": 1.0024, + "step": 19137 + }, + { + "epoch": 0.61, + "learning_rate": 6.908844964960978e-06, + "loss": 0.9702, + "step": 19138 + }, + { + "epoch": 0.61, + "learning_rate": 6.907859906836101e-06, + "loss": 0.8862, + "step": 19139 + }, + { + "epoch": 0.61, + "learning_rate": 6.906874881886467e-06, + "loss": 0.9971, + "step": 19140 + }, + { + "epoch": 0.61, + "learning_rate": 6.905889890122641e-06, + "loss": 0.9668, + "step": 19141 + }, + { + "epoch": 0.61, + "learning_rate": 6.904904931555192e-06, + "loss": 0.958, + "step": 19142 + }, + { + "epoch": 0.61, + "learning_rate": 6.903920006194689e-06, + "loss": 1.0322, + "step": 19143 + }, + { + "epoch": 0.61, + "learning_rate": 6.902935114051695e-06, + "loss": 0.939, + "step": 19144 + }, + { + "epoch": 0.61, + "learning_rate": 6.901950255136782e-06, + "loss": 1.0981, + "step": 19145 + }, + { + "epoch": 0.61, + "learning_rate": 6.9009654294605135e-06, + "loss": 0.9585, + "step": 19146 + }, + { + "epoch": 0.61, + "learning_rate": 6.899980637033454e-06, + "loss": 0.9243, + "step": 19147 + }, + { + "epoch": 0.61, + "learning_rate": 6.898995877866172e-06, + "loss": 0.9546, + "step": 19148 + }, + { + "epoch": 0.61, + "learning_rate": 6.898011151969233e-06, + "loss": 0.9443, + "step": 19149 + }, + { + "epoch": 0.61, + "learning_rate": 6.897026459353206e-06, + "loss": 0.9717, + "step": 19150 + }, + { + "epoch": 0.61, + "learning_rate": 6.8960418000286455e-06, + "loss": 1.0229, + "step": 19151 + }, + { + "epoch": 0.61, + "learning_rate": 6.895057174006121e-06, + "loss": 1.0498, + "step": 19152 + }, + { + "epoch": 0.61, + "learning_rate": 6.8940725812961966e-06, + "loss": 0.8867, + "step": 19153 + }, + { + "epoch": 0.61, + "learning_rate": 6.893088021909436e-06, + "loss": 0.978, + "step": 19154 + }, + { + "epoch": 0.61, + "learning_rate": 6.892103495856404e-06, + "loss": 0.958, + "step": 19155 + }, + { + "epoch": 0.61, + "learning_rate": 6.891119003147659e-06, + "loss": 1.0386, + "step": 19156 + }, + { + "epoch": 0.61, + "learning_rate": 6.89013454379377e-06, + "loss": 0.8784, + "step": 19157 + }, + { + "epoch": 0.61, + "learning_rate": 6.8891501178052965e-06, + "loss": 0.9966, + "step": 19158 + }, + { + "epoch": 0.61, + "learning_rate": 6.888165725192797e-06, + "loss": 0.9355, + "step": 19159 + }, + { + "epoch": 0.61, + "learning_rate": 6.887181365966833e-06, + "loss": 0.896, + "step": 19160 + }, + { + "epoch": 0.61, + "learning_rate": 6.8861970401379695e-06, + "loss": 0.8633, + "step": 19161 + }, + { + "epoch": 0.61, + "learning_rate": 6.885212747716766e-06, + "loss": 1.0566, + "step": 19162 + }, + { + "epoch": 0.61, + "learning_rate": 6.884228488713782e-06, + "loss": 0.8389, + "step": 19163 + }, + { + "epoch": 0.61, + "learning_rate": 6.883244263139578e-06, + "loss": 0.9167, + "step": 19164 + }, + { + "epoch": 0.61, + "learning_rate": 6.882260071004717e-06, + "loss": 0.8228, + "step": 19165 + }, + { + "epoch": 0.61, + "learning_rate": 6.881275912319751e-06, + "loss": 0.9761, + "step": 19166 + }, + { + "epoch": 0.61, + "learning_rate": 6.880291787095244e-06, + "loss": 0.9653, + "step": 19167 + }, + { + "epoch": 0.61, + "learning_rate": 6.879307695341752e-06, + "loss": 1.0732, + "step": 19168 + }, + { + "epoch": 0.61, + "learning_rate": 6.878323637069834e-06, + "loss": 1.0874, + "step": 19169 + }, + { + "epoch": 0.61, + "learning_rate": 6.877339612290051e-06, + "loss": 1.0405, + "step": 19170 + }, + { + "epoch": 0.61, + "learning_rate": 6.876355621012957e-06, + "loss": 0.9033, + "step": 19171 + }, + { + "epoch": 0.61, + "learning_rate": 6.8753716632491095e-06, + "loss": 1.0225, + "step": 19172 + }, + { + "epoch": 0.61, + "learning_rate": 6.8743877390090695e-06, + "loss": 0.8965, + "step": 19173 + }, + { + "epoch": 0.61, + "learning_rate": 6.873403848303387e-06, + "loss": 0.96, + "step": 19174 + }, + { + "epoch": 0.61, + "learning_rate": 6.872419991142621e-06, + "loss": 1.0347, + "step": 19175 + }, + { + "epoch": 0.61, + "learning_rate": 6.871436167537327e-06, + "loss": 0.9639, + "step": 19176 + }, + { + "epoch": 0.61, + "learning_rate": 6.870452377498058e-06, + "loss": 0.9404, + "step": 19177 + }, + { + "epoch": 0.61, + "learning_rate": 6.869468621035373e-06, + "loss": 0.9395, + "step": 19178 + }, + { + "epoch": 0.61, + "learning_rate": 6.868484898159829e-06, + "loss": 0.856, + "step": 19179 + }, + { + "epoch": 0.61, + "learning_rate": 6.8675012088819706e-06, + "loss": 1.0654, + "step": 19180 + }, + { + "epoch": 0.61, + "learning_rate": 6.866517553212362e-06, + "loss": 1.0234, + "step": 19181 + }, + { + "epoch": 0.61, + "learning_rate": 6.865533931161549e-06, + "loss": 0.9468, + "step": 19182 + }, + { + "epoch": 0.61, + "learning_rate": 6.864550342740088e-06, + "loss": 0.9805, + "step": 19183 + }, + { + "epoch": 0.61, + "learning_rate": 6.8635667879585335e-06, + "loss": 0.9922, + "step": 19184 + }, + { + "epoch": 0.61, + "learning_rate": 6.862583266827434e-06, + "loss": 0.9282, + "step": 19185 + }, + { + "epoch": 0.61, + "learning_rate": 6.861599779357344e-06, + "loss": 1.123, + "step": 19186 + }, + { + "epoch": 0.61, + "learning_rate": 6.860616325558815e-06, + "loss": 0.9546, + "step": 19187 + }, + { + "epoch": 0.61, + "learning_rate": 6.859632905442402e-06, + "loss": 0.9888, + "step": 19188 + }, + { + "epoch": 0.61, + "learning_rate": 6.858649519018648e-06, + "loss": 0.979, + "step": 19189 + }, + { + "epoch": 0.61, + "learning_rate": 6.8576661662981114e-06, + "loss": 0.9097, + "step": 19190 + }, + { + "epoch": 0.61, + "learning_rate": 6.856682847291336e-06, + "loss": 0.8643, + "step": 19191 + }, + { + "epoch": 0.61, + "learning_rate": 6.855699562008875e-06, + "loss": 0.9624, + "step": 19192 + }, + { + "epoch": 0.61, + "learning_rate": 6.854716310461278e-06, + "loss": 0.9106, + "step": 19193 + }, + { + "epoch": 0.61, + "learning_rate": 6.853733092659094e-06, + "loss": 1.0386, + "step": 19194 + }, + { + "epoch": 0.61, + "learning_rate": 6.852749908612871e-06, + "loss": 1.0928, + "step": 19195 + }, + { + "epoch": 0.61, + "learning_rate": 6.8517667583331636e-06, + "loss": 0.9072, + "step": 19196 + }, + { + "epoch": 0.61, + "learning_rate": 6.85078364183051e-06, + "loss": 0.8521, + "step": 19197 + }, + { + "epoch": 0.61, + "learning_rate": 6.8498005591154625e-06, + "loss": 1.0278, + "step": 19198 + }, + { + "epoch": 0.61, + "learning_rate": 6.848817510198569e-06, + "loss": 0.8525, + "step": 19199 + }, + { + "epoch": 0.61, + "learning_rate": 6.847834495090375e-06, + "loss": 0.9678, + "step": 19200 + }, + { + "epoch": 0.61, + "learning_rate": 6.846851513801429e-06, + "loss": 0.9165, + "step": 19201 + }, + { + "epoch": 0.61, + "learning_rate": 6.845868566342276e-06, + "loss": 0.8906, + "step": 19202 + }, + { + "epoch": 0.61, + "learning_rate": 6.844885652723467e-06, + "loss": 0.8296, + "step": 19203 + }, + { + "epoch": 0.61, + "learning_rate": 6.84390277295554e-06, + "loss": 1.0308, + "step": 19204 + }, + { + "epoch": 0.61, + "learning_rate": 6.842919927049041e-06, + "loss": 0.9077, + "step": 19205 + }, + { + "epoch": 0.61, + "learning_rate": 6.841937115014519e-06, + "loss": 0.9731, + "step": 19206 + }, + { + "epoch": 0.61, + "learning_rate": 6.840954336862516e-06, + "loss": 0.968, + "step": 19207 + }, + { + "epoch": 0.61, + "learning_rate": 6.839971592603576e-06, + "loss": 0.9497, + "step": 19208 + }, + { + "epoch": 0.61, + "learning_rate": 6.838988882248243e-06, + "loss": 0.9639, + "step": 19209 + }, + { + "epoch": 0.61, + "learning_rate": 6.8380062058070614e-06, + "loss": 0.9219, + "step": 19210 + }, + { + "epoch": 0.61, + "learning_rate": 6.837023563290579e-06, + "loss": 1.0479, + "step": 19211 + }, + { + "epoch": 0.61, + "learning_rate": 6.836040954709329e-06, + "loss": 1.0337, + "step": 19212 + }, + { + "epoch": 0.61, + "learning_rate": 6.835058380073856e-06, + "loss": 0.9282, + "step": 19213 + }, + { + "epoch": 0.61, + "learning_rate": 6.834075839394705e-06, + "loss": 0.9941, + "step": 19214 + }, + { + "epoch": 0.61, + "learning_rate": 6.833093332682417e-06, + "loss": 1.125, + "step": 19215 + }, + { + "epoch": 0.61, + "learning_rate": 6.832110859947532e-06, + "loss": 1.0225, + "step": 19216 + }, + { + "epoch": 0.61, + "learning_rate": 6.831128421200591e-06, + "loss": 0.874, + "step": 19217 + }, + { + "epoch": 0.61, + "learning_rate": 6.8301460164521395e-06, + "loss": 0.9199, + "step": 19218 + }, + { + "epoch": 0.61, + "learning_rate": 6.829163645712709e-06, + "loss": 0.9648, + "step": 19219 + }, + { + "epoch": 0.61, + "learning_rate": 6.8281813089928425e-06, + "loss": 0.9951, + "step": 19220 + }, + { + "epoch": 0.61, + "learning_rate": 6.82719900630308e-06, + "loss": 1.019, + "step": 19221 + }, + { + "epoch": 0.61, + "learning_rate": 6.826216737653962e-06, + "loss": 0.9551, + "step": 19222 + }, + { + "epoch": 0.61, + "learning_rate": 6.825234503056028e-06, + "loss": 1.1274, + "step": 19223 + }, + { + "epoch": 0.61, + "learning_rate": 6.8242523025198114e-06, + "loss": 0.9575, + "step": 19224 + }, + { + "epoch": 0.61, + "learning_rate": 6.823270136055853e-06, + "loss": 0.8945, + "step": 19225 + }, + { + "epoch": 0.61, + "learning_rate": 6.822288003674691e-06, + "loss": 1.0127, + "step": 19226 + }, + { + "epoch": 0.61, + "learning_rate": 6.821305905386861e-06, + "loss": 0.9839, + "step": 19227 + }, + { + "epoch": 0.61, + "learning_rate": 6.8203238412029e-06, + "loss": 1.0195, + "step": 19228 + }, + { + "epoch": 0.61, + "learning_rate": 6.819341811133349e-06, + "loss": 0.5359, + "step": 19229 + }, + { + "epoch": 0.61, + "learning_rate": 6.818359815188735e-06, + "loss": 0.9775, + "step": 19230 + }, + { + "epoch": 0.62, + "learning_rate": 6.8173778533796005e-06, + "loss": 1.0034, + "step": 19231 + }, + { + "epoch": 0.62, + "learning_rate": 6.816395925716481e-06, + "loss": 0.9634, + "step": 19232 + }, + { + "epoch": 0.62, + "learning_rate": 6.815414032209909e-06, + "loss": 1.0239, + "step": 19233 + }, + { + "epoch": 0.62, + "learning_rate": 6.814432172870422e-06, + "loss": 0.9702, + "step": 19234 + }, + { + "epoch": 0.62, + "learning_rate": 6.813450347708549e-06, + "loss": 0.9629, + "step": 19235 + }, + { + "epoch": 0.62, + "learning_rate": 6.812468556734827e-06, + "loss": 1.0259, + "step": 19236 + }, + { + "epoch": 0.62, + "learning_rate": 6.81148679995979e-06, + "loss": 0.9951, + "step": 19237 + }, + { + "epoch": 0.62, + "learning_rate": 6.810505077393971e-06, + "loss": 0.7378, + "step": 19238 + }, + { + "epoch": 0.62, + "learning_rate": 6.809523389047902e-06, + "loss": 0.9409, + "step": 19239 + }, + { + "epoch": 0.62, + "learning_rate": 6.808541734932117e-06, + "loss": 0.9639, + "step": 19240 + }, + { + "epoch": 0.62, + "learning_rate": 6.80756011505715e-06, + "loss": 0.9219, + "step": 19241 + }, + { + "epoch": 0.62, + "learning_rate": 6.8065785294335256e-06, + "loss": 1.0176, + "step": 19242 + }, + { + "epoch": 0.62, + "learning_rate": 6.8055969780717805e-06, + "loss": 1.0122, + "step": 19243 + }, + { + "epoch": 0.62, + "learning_rate": 6.804615460982442e-06, + "loss": 0.9639, + "step": 19244 + }, + { + "epoch": 0.62, + "learning_rate": 6.803633978176045e-06, + "loss": 1.1602, + "step": 19245 + }, + { + "epoch": 0.62, + "learning_rate": 6.8026525296631185e-06, + "loss": 0.9995, + "step": 19246 + }, + { + "epoch": 0.62, + "learning_rate": 6.801671115454192e-06, + "loss": 1.0557, + "step": 19247 + }, + { + "epoch": 0.62, + "learning_rate": 6.800689735559795e-06, + "loss": 0.9011, + "step": 19248 + }, + { + "epoch": 0.62, + "learning_rate": 6.799708389990456e-06, + "loss": 1.0371, + "step": 19249 + }, + { + "epoch": 0.62, + "learning_rate": 6.798727078756704e-06, + "loss": 0.9922, + "step": 19250 + }, + { + "epoch": 0.62, + "learning_rate": 6.797745801869066e-06, + "loss": 1.0054, + "step": 19251 + }, + { + "epoch": 0.62, + "learning_rate": 6.796764559338072e-06, + "loss": 0.9775, + "step": 19252 + }, + { + "epoch": 0.62, + "learning_rate": 6.795783351174248e-06, + "loss": 0.9414, + "step": 19253 + }, + { + "epoch": 0.62, + "learning_rate": 6.7948021773881235e-06, + "loss": 1.0303, + "step": 19254 + }, + { + "epoch": 0.62, + "learning_rate": 6.793821037990225e-06, + "loss": 0.873, + "step": 19255 + }, + { + "epoch": 0.62, + "learning_rate": 6.792839932991081e-06, + "loss": 0.9736, + "step": 19256 + }, + { + "epoch": 0.62, + "learning_rate": 6.791858862401212e-06, + "loss": 0.9717, + "step": 19257 + }, + { + "epoch": 0.62, + "learning_rate": 6.790877826231146e-06, + "loss": 0.9639, + "step": 19258 + }, + { + "epoch": 0.62, + "learning_rate": 6.789896824491409e-06, + "loss": 0.9551, + "step": 19259 + }, + { + "epoch": 0.62, + "learning_rate": 6.788915857192527e-06, + "loss": 1.0063, + "step": 19260 + }, + { + "epoch": 0.62, + "learning_rate": 6.7879349243450235e-06, + "loss": 0.9629, + "step": 19261 + }, + { + "epoch": 0.62, + "learning_rate": 6.786954025959425e-06, + "loss": 0.9043, + "step": 19262 + }, + { + "epoch": 0.62, + "learning_rate": 6.785973162046253e-06, + "loss": 0.8652, + "step": 19263 + }, + { + "epoch": 0.62, + "learning_rate": 6.7849923326160335e-06, + "loss": 0.9868, + "step": 19264 + }, + { + "epoch": 0.62, + "learning_rate": 6.784011537679287e-06, + "loss": 0.8662, + "step": 19265 + }, + { + "epoch": 0.62, + "learning_rate": 6.783030777246535e-06, + "loss": 0.9849, + "step": 19266 + }, + { + "epoch": 0.62, + "learning_rate": 6.782050051328304e-06, + "loss": 0.9512, + "step": 19267 + }, + { + "epoch": 0.62, + "learning_rate": 6.781069359935117e-06, + "loss": 0.8838, + "step": 19268 + }, + { + "epoch": 0.62, + "learning_rate": 6.78008870307749e-06, + "loss": 0.8223, + "step": 19269 + }, + { + "epoch": 0.62, + "learning_rate": 6.779108080765948e-06, + "loss": 0.9023, + "step": 19270 + }, + { + "epoch": 0.62, + "learning_rate": 6.778127493011014e-06, + "loss": 1.0229, + "step": 19271 + }, + { + "epoch": 0.62, + "learning_rate": 6.777146939823203e-06, + "loss": 0.9302, + "step": 19272 + }, + { + "epoch": 0.62, + "learning_rate": 6.776166421213042e-06, + "loss": 1.0439, + "step": 19273 + }, + { + "epoch": 0.62, + "learning_rate": 6.775185937191044e-06, + "loss": 0.9331, + "step": 19274 + }, + { + "epoch": 0.62, + "learning_rate": 6.774205487767733e-06, + "loss": 0.8687, + "step": 19275 + }, + { + "epoch": 0.62, + "learning_rate": 6.773225072953626e-06, + "loss": 0.9336, + "step": 19276 + }, + { + "epoch": 0.62, + "learning_rate": 6.772244692759245e-06, + "loss": 0.9932, + "step": 19277 + }, + { + "epoch": 0.62, + "learning_rate": 6.771264347195103e-06, + "loss": 1.0811, + "step": 19278 + }, + { + "epoch": 0.62, + "learning_rate": 6.770284036271727e-06, + "loss": 0.8984, + "step": 19279 + }, + { + "epoch": 0.62, + "learning_rate": 6.769303759999624e-06, + "loss": 0.9019, + "step": 19280 + }, + { + "epoch": 0.62, + "learning_rate": 6.768323518389317e-06, + "loss": 0.9141, + "step": 19281 + }, + { + "epoch": 0.62, + "learning_rate": 6.767343311451321e-06, + "loss": 0.8779, + "step": 19282 + }, + { + "epoch": 0.62, + "learning_rate": 6.7663631391961546e-06, + "loss": 1.0186, + "step": 19283 + }, + { + "epoch": 0.62, + "learning_rate": 6.765383001634331e-06, + "loss": 0.5156, + "step": 19284 + }, + { + "epoch": 0.62, + "learning_rate": 6.76440289877637e-06, + "loss": 0.9717, + "step": 19285 + }, + { + "epoch": 0.62, + "learning_rate": 6.763422830632789e-06, + "loss": 0.9697, + "step": 19286 + }, + { + "epoch": 0.62, + "learning_rate": 6.762442797214093e-06, + "loss": 0.9971, + "step": 19287 + }, + { + "epoch": 0.62, + "learning_rate": 6.7614627985308045e-06, + "loss": 0.8511, + "step": 19288 + }, + { + "epoch": 0.62, + "learning_rate": 6.760482834593435e-06, + "loss": 1.1636, + "step": 19289 + }, + { + "epoch": 0.62, + "learning_rate": 6.759502905412498e-06, + "loss": 1.0205, + "step": 19290 + }, + { + "epoch": 0.62, + "learning_rate": 6.758523010998511e-06, + "loss": 0.8843, + "step": 19291 + }, + { + "epoch": 0.62, + "learning_rate": 6.7575431513619825e-06, + "loss": 0.8486, + "step": 19292 + }, + { + "epoch": 0.62, + "learning_rate": 6.75656332651343e-06, + "loss": 0.9751, + "step": 19293 + }, + { + "epoch": 0.62, + "learning_rate": 6.755583536463366e-06, + "loss": 1.0308, + "step": 19294 + }, + { + "epoch": 0.62, + "learning_rate": 6.754603781222295e-06, + "loss": 0.98, + "step": 19295 + }, + { + "epoch": 0.62, + "learning_rate": 6.753624060800735e-06, + "loss": 1.0068, + "step": 19296 + }, + { + "epoch": 0.62, + "learning_rate": 6.752644375209194e-06, + "loss": 0.9961, + "step": 19297 + }, + { + "epoch": 0.62, + "learning_rate": 6.751664724458187e-06, + "loss": 0.9907, + "step": 19298 + }, + { + "epoch": 0.62, + "learning_rate": 6.750685108558221e-06, + "loss": 0.8877, + "step": 19299 + }, + { + "epoch": 0.62, + "learning_rate": 6.749705527519807e-06, + "loss": 1.0117, + "step": 19300 + }, + { + "epoch": 0.62, + "learning_rate": 6.748725981353461e-06, + "loss": 0.853, + "step": 19301 + }, + { + "epoch": 0.62, + "learning_rate": 6.747746470069683e-06, + "loss": 0.9204, + "step": 19302 + }, + { + "epoch": 0.62, + "learning_rate": 6.746766993678986e-06, + "loss": 0.9263, + "step": 19303 + }, + { + "epoch": 0.62, + "learning_rate": 6.745787552191878e-06, + "loss": 1.1221, + "step": 19304 + }, + { + "epoch": 0.62, + "learning_rate": 6.744808145618869e-06, + "loss": 0.8672, + "step": 19305 + }, + { + "epoch": 0.62, + "learning_rate": 6.743828773970464e-06, + "loss": 1.0034, + "step": 19306 + }, + { + "epoch": 0.62, + "learning_rate": 6.742849437257175e-06, + "loss": 0.8813, + "step": 19307 + }, + { + "epoch": 0.62, + "learning_rate": 6.741870135489506e-06, + "loss": 0.9795, + "step": 19308 + }, + { + "epoch": 0.62, + "learning_rate": 6.740890868677965e-06, + "loss": 1.0664, + "step": 19309 + }, + { + "epoch": 0.62, + "learning_rate": 6.739911636833056e-06, + "loss": 0.9731, + "step": 19310 + }, + { + "epoch": 0.62, + "learning_rate": 6.7389324399652885e-06, + "loss": 0.895, + "step": 19311 + }, + { + "epoch": 0.62, + "learning_rate": 6.737953278085167e-06, + "loss": 1.1099, + "step": 19312 + }, + { + "epoch": 0.62, + "learning_rate": 6.736974151203197e-06, + "loss": 1.0747, + "step": 19313 + }, + { + "epoch": 0.62, + "learning_rate": 6.735995059329881e-06, + "loss": 0.9775, + "step": 19314 + }, + { + "epoch": 0.62, + "learning_rate": 6.735016002475727e-06, + "loss": 1.0713, + "step": 19315 + }, + { + "epoch": 0.62, + "learning_rate": 6.734036980651239e-06, + "loss": 0.8765, + "step": 19316 + }, + { + "epoch": 0.62, + "learning_rate": 6.733057993866917e-06, + "loss": 0.9326, + "step": 19317 + }, + { + "epoch": 0.62, + "learning_rate": 6.732079042133271e-06, + "loss": 0.8804, + "step": 19318 + }, + { + "epoch": 0.62, + "learning_rate": 6.731100125460795e-06, + "loss": 0.9951, + "step": 19319 + }, + { + "epoch": 0.62, + "learning_rate": 6.7301212438599996e-06, + "loss": 0.9272, + "step": 19320 + }, + { + "epoch": 0.62, + "learning_rate": 6.729142397341383e-06, + "loss": 1.0474, + "step": 19321 + }, + { + "epoch": 0.62, + "learning_rate": 6.728163585915448e-06, + "loss": 0.9429, + "step": 19322 + }, + { + "epoch": 0.62, + "learning_rate": 6.727184809592697e-06, + "loss": 0.9189, + "step": 19323 + }, + { + "epoch": 0.62, + "learning_rate": 6.726206068383636e-06, + "loss": 0.8745, + "step": 19324 + }, + { + "epoch": 0.62, + "learning_rate": 6.7252273622987565e-06, + "loss": 0.9854, + "step": 19325 + }, + { + "epoch": 0.62, + "learning_rate": 6.724248691348563e-06, + "loss": 1.0015, + "step": 19326 + }, + { + "epoch": 0.62, + "learning_rate": 6.723270055543554e-06, + "loss": 0.8975, + "step": 19327 + }, + { + "epoch": 0.62, + "learning_rate": 6.722291454894232e-06, + "loss": 0.5493, + "step": 19328 + }, + { + "epoch": 0.62, + "learning_rate": 6.721312889411095e-06, + "loss": 0.9307, + "step": 19329 + }, + { + "epoch": 0.62, + "learning_rate": 6.720334359104641e-06, + "loss": 0.8901, + "step": 19330 + }, + { + "epoch": 0.62, + "learning_rate": 6.719355863985375e-06, + "loss": 0.9653, + "step": 19331 + }, + { + "epoch": 0.62, + "learning_rate": 6.718377404063785e-06, + "loss": 0.9067, + "step": 19332 + }, + { + "epoch": 0.62, + "learning_rate": 6.7173989793503735e-06, + "loss": 1.0488, + "step": 19333 + }, + { + "epoch": 0.62, + "learning_rate": 6.716420589855639e-06, + "loss": 1.0078, + "step": 19334 + }, + { + "epoch": 0.62, + "learning_rate": 6.715442235590077e-06, + "loss": 0.9204, + "step": 19335 + }, + { + "epoch": 0.62, + "learning_rate": 6.714463916564185e-06, + "loss": 0.9976, + "step": 19336 + }, + { + "epoch": 0.62, + "learning_rate": 6.713485632788457e-06, + "loss": 1.0771, + "step": 19337 + }, + { + "epoch": 0.62, + "learning_rate": 6.712507384273392e-06, + "loss": 0.8823, + "step": 19338 + }, + { + "epoch": 0.62, + "learning_rate": 6.7115291710294895e-06, + "loss": 1.0054, + "step": 19339 + }, + { + "epoch": 0.62, + "learning_rate": 6.710550993067234e-06, + "loss": 0.9644, + "step": 19340 + }, + { + "epoch": 0.62, + "learning_rate": 6.709572850397126e-06, + "loss": 1.0332, + "step": 19341 + }, + { + "epoch": 0.62, + "learning_rate": 6.70859474302966e-06, + "loss": 1.0396, + "step": 19342 + }, + { + "epoch": 0.62, + "learning_rate": 6.70761667097533e-06, + "loss": 0.936, + "step": 19343 + }, + { + "epoch": 0.62, + "learning_rate": 6.706638634244629e-06, + "loss": 0.4998, + "step": 19344 + }, + { + "epoch": 0.62, + "learning_rate": 6.70566063284805e-06, + "loss": 1.0464, + "step": 19345 + }, + { + "epoch": 0.62, + "learning_rate": 6.70468266679609e-06, + "loss": 0.9487, + "step": 19346 + }, + { + "epoch": 0.62, + "learning_rate": 6.703704736099236e-06, + "loss": 0.9614, + "step": 19347 + }, + { + "epoch": 0.62, + "learning_rate": 6.70272684076798e-06, + "loss": 1.0264, + "step": 19348 + }, + { + "epoch": 0.62, + "learning_rate": 6.701748980812817e-06, + "loss": 0.8196, + "step": 19349 + }, + { + "epoch": 0.62, + "learning_rate": 6.700771156244237e-06, + "loss": 1.0146, + "step": 19350 + }, + { + "epoch": 0.62, + "learning_rate": 6.699793367072732e-06, + "loss": 1.0225, + "step": 19351 + }, + { + "epoch": 0.62, + "learning_rate": 6.698815613308791e-06, + "loss": 0.9668, + "step": 19352 + }, + { + "epoch": 0.62, + "learning_rate": 6.697837894962904e-06, + "loss": 1.0522, + "step": 19353 + }, + { + "epoch": 0.62, + "learning_rate": 6.696860212045564e-06, + "loss": 1.0186, + "step": 19354 + }, + { + "epoch": 0.62, + "learning_rate": 6.695882564567255e-06, + "loss": 1.0884, + "step": 19355 + }, + { + "epoch": 0.62, + "learning_rate": 6.694904952538473e-06, + "loss": 0.9482, + "step": 19356 + }, + { + "epoch": 0.62, + "learning_rate": 6.693927375969701e-06, + "loss": 0.519, + "step": 19357 + }, + { + "epoch": 0.62, + "learning_rate": 6.692949834871429e-06, + "loss": 0.8936, + "step": 19358 + }, + { + "epoch": 0.62, + "learning_rate": 6.691972329254144e-06, + "loss": 1.0225, + "step": 19359 + }, + { + "epoch": 0.62, + "learning_rate": 6.6909948591283366e-06, + "loss": 0.8906, + "step": 19360 + }, + { + "epoch": 0.62, + "learning_rate": 6.690017424504491e-06, + "loss": 0.9541, + "step": 19361 + }, + { + "epoch": 0.62, + "learning_rate": 6.689040025393098e-06, + "loss": 1.0649, + "step": 19362 + }, + { + "epoch": 0.62, + "learning_rate": 6.688062661804639e-06, + "loss": 0.9033, + "step": 19363 + }, + { + "epoch": 0.62, + "learning_rate": 6.687085333749603e-06, + "loss": 0.9512, + "step": 19364 + }, + { + "epoch": 0.62, + "learning_rate": 6.686108041238473e-06, + "loss": 0.9692, + "step": 19365 + }, + { + "epoch": 0.62, + "learning_rate": 6.685130784281737e-06, + "loss": 0.9546, + "step": 19366 + }, + { + "epoch": 0.62, + "learning_rate": 6.684153562889878e-06, + "loss": 0.9883, + "step": 19367 + }, + { + "epoch": 0.62, + "learning_rate": 6.683176377073382e-06, + "loss": 1.0518, + "step": 19368 + }, + { + "epoch": 0.62, + "learning_rate": 6.682199226842736e-06, + "loss": 0.8159, + "step": 19369 + }, + { + "epoch": 0.62, + "learning_rate": 6.681222112208418e-06, + "loss": 1.0278, + "step": 19370 + }, + { + "epoch": 0.62, + "learning_rate": 6.680245033180912e-06, + "loss": 0.8823, + "step": 19371 + }, + { + "epoch": 0.62, + "learning_rate": 6.679267989770704e-06, + "loss": 0.8906, + "step": 19372 + }, + { + "epoch": 0.62, + "learning_rate": 6.6782909819882734e-06, + "loss": 0.9868, + "step": 19373 + }, + { + "epoch": 0.62, + "learning_rate": 6.6773140098441055e-06, + "loss": 0.9658, + "step": 19374 + }, + { + "epoch": 0.62, + "learning_rate": 6.676337073348682e-06, + "loss": 0.9033, + "step": 19375 + }, + { + "epoch": 0.62, + "learning_rate": 6.675360172512482e-06, + "loss": 1.0093, + "step": 19376 + }, + { + "epoch": 0.62, + "learning_rate": 6.674383307345991e-06, + "loss": 0.9932, + "step": 19377 + }, + { + "epoch": 0.62, + "learning_rate": 6.673406477859684e-06, + "loss": 0.9194, + "step": 19378 + }, + { + "epoch": 0.62, + "learning_rate": 6.672429684064044e-06, + "loss": 1.0708, + "step": 19379 + }, + { + "epoch": 0.62, + "learning_rate": 6.671452925969549e-06, + "loss": 0.8125, + "step": 19380 + }, + { + "epoch": 0.62, + "learning_rate": 6.670476203586681e-06, + "loss": 1.0317, + "step": 19381 + }, + { + "epoch": 0.62, + "learning_rate": 6.6694995169259194e-06, + "loss": 0.9692, + "step": 19382 + }, + { + "epoch": 0.62, + "learning_rate": 6.66852286599774e-06, + "loss": 0.9536, + "step": 19383 + }, + { + "epoch": 0.62, + "learning_rate": 6.667546250812629e-06, + "loss": 0.9287, + "step": 19384 + }, + { + "epoch": 0.62, + "learning_rate": 6.6665696713810535e-06, + "loss": 1.0181, + "step": 19385 + }, + { + "epoch": 0.62, + "learning_rate": 6.665593127713497e-06, + "loss": 1.0205, + "step": 19386 + }, + { + "epoch": 0.62, + "learning_rate": 6.664616619820435e-06, + "loss": 0.7832, + "step": 19387 + }, + { + "epoch": 0.62, + "learning_rate": 6.6636401477123455e-06, + "loss": 0.8916, + "step": 19388 + }, + { + "epoch": 0.62, + "learning_rate": 6.662663711399705e-06, + "loss": 1.144, + "step": 19389 + }, + { + "epoch": 0.62, + "learning_rate": 6.6616873108929904e-06, + "loss": 0.8149, + "step": 19390 + }, + { + "epoch": 0.62, + "learning_rate": 6.6607109462026765e-06, + "loss": 1.0532, + "step": 19391 + }, + { + "epoch": 0.62, + "learning_rate": 6.659734617339239e-06, + "loss": 0.9595, + "step": 19392 + }, + { + "epoch": 0.62, + "learning_rate": 6.65875832431315e-06, + "loss": 0.9951, + "step": 19393 + }, + { + "epoch": 0.62, + "learning_rate": 6.657782067134886e-06, + "loss": 0.9062, + "step": 19394 + }, + { + "epoch": 0.62, + "learning_rate": 6.656805845814922e-06, + "loss": 0.9736, + "step": 19395 + }, + { + "epoch": 0.62, + "learning_rate": 6.655829660363733e-06, + "loss": 0.8979, + "step": 19396 + }, + { + "epoch": 0.62, + "learning_rate": 6.65485351079179e-06, + "loss": 0.8328, + "step": 19397 + }, + { + "epoch": 0.62, + "learning_rate": 6.6538773971095646e-06, + "loss": 1.0137, + "step": 19398 + }, + { + "epoch": 0.62, + "learning_rate": 6.652901319327534e-06, + "loss": 0.9036, + "step": 19399 + }, + { + "epoch": 0.62, + "learning_rate": 6.651925277456166e-06, + "loss": 1.0728, + "step": 19400 + }, + { + "epoch": 0.62, + "learning_rate": 6.650949271505937e-06, + "loss": 0.9844, + "step": 19401 + }, + { + "epoch": 0.62, + "learning_rate": 6.649973301487314e-06, + "loss": 1.0464, + "step": 19402 + }, + { + "epoch": 0.62, + "learning_rate": 6.6489973674107674e-06, + "loss": 0.8965, + "step": 19403 + }, + { + "epoch": 0.62, + "learning_rate": 6.648021469286772e-06, + "loss": 0.9419, + "step": 19404 + }, + { + "epoch": 0.62, + "learning_rate": 6.647045607125797e-06, + "loss": 1.021, + "step": 19405 + }, + { + "epoch": 0.62, + "learning_rate": 6.646069780938312e-06, + "loss": 0.9331, + "step": 19406 + }, + { + "epoch": 0.62, + "learning_rate": 6.645093990734788e-06, + "loss": 1.0122, + "step": 19407 + }, + { + "epoch": 0.62, + "learning_rate": 6.64411823652569e-06, + "loss": 0.9336, + "step": 19408 + }, + { + "epoch": 0.62, + "learning_rate": 6.643142518321489e-06, + "loss": 0.9941, + "step": 19409 + }, + { + "epoch": 0.62, + "learning_rate": 6.642166836132653e-06, + "loss": 0.8472, + "step": 19410 + }, + { + "epoch": 0.62, + "learning_rate": 6.641191189969651e-06, + "loss": 0.9878, + "step": 19411 + }, + { + "epoch": 0.62, + "learning_rate": 6.640215579842951e-06, + "loss": 0.9067, + "step": 19412 + }, + { + "epoch": 0.62, + "learning_rate": 6.639240005763019e-06, + "loss": 0.9526, + "step": 19413 + }, + { + "epoch": 0.62, + "learning_rate": 6.638264467740323e-06, + "loss": 1.0908, + "step": 19414 + }, + { + "epoch": 0.62, + "learning_rate": 6.637288965785331e-06, + "loss": 0.9834, + "step": 19415 + }, + { + "epoch": 0.62, + "learning_rate": 6.636313499908504e-06, + "loss": 1.0156, + "step": 19416 + }, + { + "epoch": 0.62, + "learning_rate": 6.635338070120309e-06, + "loss": 0.9624, + "step": 19417 + }, + { + "epoch": 0.62, + "learning_rate": 6.634362676431214e-06, + "loss": 0.9224, + "step": 19418 + }, + { + "epoch": 0.62, + "learning_rate": 6.633387318851683e-06, + "loss": 0.5032, + "step": 19419 + }, + { + "epoch": 0.62, + "learning_rate": 6.632411997392178e-06, + "loss": 0.9023, + "step": 19420 + }, + { + "epoch": 0.62, + "learning_rate": 6.631436712063167e-06, + "loss": 0.9106, + "step": 19421 + }, + { + "epoch": 0.62, + "learning_rate": 6.630461462875115e-06, + "loss": 0.8984, + "step": 19422 + }, + { + "epoch": 0.62, + "learning_rate": 6.629486249838479e-06, + "loss": 0.9932, + "step": 19423 + }, + { + "epoch": 0.62, + "learning_rate": 6.628511072963726e-06, + "loss": 0.9209, + "step": 19424 + }, + { + "epoch": 0.62, + "learning_rate": 6.6275359322613156e-06, + "loss": 1.1279, + "step": 19425 + }, + { + "epoch": 0.62, + "learning_rate": 6.626560827741713e-06, + "loss": 0.8752, + "step": 19426 + }, + { + "epoch": 0.62, + "learning_rate": 6.62558575941538e-06, + "loss": 1.022, + "step": 19427 + }, + { + "epoch": 0.62, + "learning_rate": 6.6246107272927775e-06, + "loss": 0.938, + "step": 19428 + }, + { + "epoch": 0.62, + "learning_rate": 6.623635731384365e-06, + "loss": 0.9863, + "step": 19429 + }, + { + "epoch": 0.62, + "learning_rate": 6.622660771700608e-06, + "loss": 1.0522, + "step": 19430 + }, + { + "epoch": 0.62, + "learning_rate": 6.62168584825196e-06, + "loss": 0.9458, + "step": 19431 + }, + { + "epoch": 0.62, + "learning_rate": 6.620710961048884e-06, + "loss": 0.9829, + "step": 19432 + }, + { + "epoch": 0.62, + "learning_rate": 6.619736110101839e-06, + "loss": 0.9829, + "step": 19433 + }, + { + "epoch": 0.62, + "learning_rate": 6.6187612954212845e-06, + "loss": 1.042, + "step": 19434 + }, + { + "epoch": 0.62, + "learning_rate": 6.6177865170176815e-06, + "loss": 1.0098, + "step": 19435 + }, + { + "epoch": 0.62, + "learning_rate": 6.616811774901484e-06, + "loss": 0.8828, + "step": 19436 + }, + { + "epoch": 0.62, + "learning_rate": 6.615837069083153e-06, + "loss": 0.9575, + "step": 19437 + }, + { + "epoch": 0.62, + "learning_rate": 6.614862399573143e-06, + "loss": 0.7905, + "step": 19438 + }, + { + "epoch": 0.62, + "learning_rate": 6.613887766381912e-06, + "loss": 0.9292, + "step": 19439 + }, + { + "epoch": 0.62, + "learning_rate": 6.612913169519922e-06, + "loss": 0.9038, + "step": 19440 + }, + { + "epoch": 0.62, + "learning_rate": 6.6119386089976215e-06, + "loss": 1.1035, + "step": 19441 + }, + { + "epoch": 0.62, + "learning_rate": 6.6109640848254705e-06, + "loss": 0.9414, + "step": 19442 + }, + { + "epoch": 0.62, + "learning_rate": 6.609989597013924e-06, + "loss": 0.9956, + "step": 19443 + }, + { + "epoch": 0.62, + "learning_rate": 6.609015145573438e-06, + "loss": 0.8594, + "step": 19444 + }, + { + "epoch": 0.62, + "learning_rate": 6.608040730514468e-06, + "loss": 0.8833, + "step": 19445 + }, + { + "epoch": 0.62, + "learning_rate": 6.607066351847467e-06, + "loss": 0.9678, + "step": 19446 + }, + { + "epoch": 0.62, + "learning_rate": 6.606092009582886e-06, + "loss": 0.9917, + "step": 19447 + }, + { + "epoch": 0.62, + "learning_rate": 6.605117703731182e-06, + "loss": 0.8906, + "step": 19448 + }, + { + "epoch": 0.62, + "learning_rate": 6.604143434302807e-06, + "loss": 1.0576, + "step": 19449 + }, + { + "epoch": 0.62, + "learning_rate": 6.603169201308217e-06, + "loss": 1.0181, + "step": 19450 + }, + { + "epoch": 0.62, + "learning_rate": 6.602195004757861e-06, + "loss": 0.9165, + "step": 19451 + }, + { + "epoch": 0.62, + "learning_rate": 6.601220844662195e-06, + "loss": 0.9082, + "step": 19452 + }, + { + "epoch": 0.62, + "learning_rate": 6.600246721031664e-06, + "loss": 1.0342, + "step": 19453 + }, + { + "epoch": 0.62, + "learning_rate": 6.599272633876725e-06, + "loss": 1.0039, + "step": 19454 + }, + { + "epoch": 0.62, + "learning_rate": 6.598298583207825e-06, + "loss": 0.9795, + "step": 19455 + }, + { + "epoch": 0.62, + "learning_rate": 6.597324569035417e-06, + "loss": 0.9688, + "step": 19456 + }, + { + "epoch": 0.62, + "learning_rate": 6.5963505913699485e-06, + "loss": 0.9761, + "step": 19457 + }, + { + "epoch": 0.62, + "learning_rate": 6.595376650221874e-06, + "loss": 0.9429, + "step": 19458 + }, + { + "epoch": 0.62, + "learning_rate": 6.594402745601639e-06, + "loss": 1.0127, + "step": 19459 + }, + { + "epoch": 0.62, + "learning_rate": 6.593428877519698e-06, + "loss": 1.0596, + "step": 19460 + }, + { + "epoch": 0.62, + "learning_rate": 6.59245504598649e-06, + "loss": 1.0469, + "step": 19461 + }, + { + "epoch": 0.62, + "learning_rate": 6.5914812510124685e-06, + "loss": 0.9194, + "step": 19462 + }, + { + "epoch": 0.62, + "learning_rate": 6.59050749260808e-06, + "loss": 0.9272, + "step": 19463 + }, + { + "epoch": 0.62, + "learning_rate": 6.589533770783774e-06, + "loss": 1.0874, + "step": 19464 + }, + { + "epoch": 0.62, + "learning_rate": 6.588560085549995e-06, + "loss": 0.9033, + "step": 19465 + }, + { + "epoch": 0.62, + "learning_rate": 6.587586436917191e-06, + "loss": 0.8992, + "step": 19466 + }, + { + "epoch": 0.62, + "learning_rate": 6.586612824895812e-06, + "loss": 0.9204, + "step": 19467 + }, + { + "epoch": 0.62, + "learning_rate": 6.585639249496297e-06, + "loss": 1.0239, + "step": 19468 + }, + { + "epoch": 0.62, + "learning_rate": 6.584665710729092e-06, + "loss": 1.041, + "step": 19469 + }, + { + "epoch": 0.62, + "learning_rate": 6.583692208604644e-06, + "loss": 0.9937, + "step": 19470 + }, + { + "epoch": 0.62, + "learning_rate": 6.582718743133398e-06, + "loss": 0.9111, + "step": 19471 + }, + { + "epoch": 0.62, + "learning_rate": 6.581745314325798e-06, + "loss": 0.7607, + "step": 19472 + }, + { + "epoch": 0.62, + "learning_rate": 6.5807719221922885e-06, + "loss": 1.1006, + "step": 19473 + }, + { + "epoch": 0.62, + "learning_rate": 6.579798566743314e-06, + "loss": 0.9941, + "step": 19474 + }, + { + "epoch": 0.62, + "learning_rate": 6.578825247989314e-06, + "loss": 1.0312, + "step": 19475 + }, + { + "epoch": 0.62, + "learning_rate": 6.577851965940731e-06, + "loss": 0.9932, + "step": 19476 + }, + { + "epoch": 0.62, + "learning_rate": 6.576878720608011e-06, + "loss": 1.0688, + "step": 19477 + }, + { + "epoch": 0.62, + "learning_rate": 6.575905512001592e-06, + "loss": 0.96, + "step": 19478 + }, + { + "epoch": 0.62, + "learning_rate": 6.574932340131917e-06, + "loss": 0.9736, + "step": 19479 + }, + { + "epoch": 0.62, + "learning_rate": 6.57395920500943e-06, + "loss": 0.9927, + "step": 19480 + }, + { + "epoch": 0.62, + "learning_rate": 6.572986106644567e-06, + "loss": 1.0801, + "step": 19481 + }, + { + "epoch": 0.62, + "learning_rate": 6.572013045047774e-06, + "loss": 0.8623, + "step": 19482 + }, + { + "epoch": 0.62, + "learning_rate": 6.571040020229483e-06, + "loss": 0.9858, + "step": 19483 + }, + { + "epoch": 0.62, + "learning_rate": 6.570067032200137e-06, + "loss": 0.8486, + "step": 19484 + }, + { + "epoch": 0.62, + "learning_rate": 6.569094080970179e-06, + "loss": 1.0254, + "step": 19485 + }, + { + "epoch": 0.62, + "learning_rate": 6.568121166550041e-06, + "loss": 0.9258, + "step": 19486 + }, + { + "epoch": 0.62, + "learning_rate": 6.567148288950167e-06, + "loss": 0.8936, + "step": 19487 + }, + { + "epoch": 0.62, + "learning_rate": 6.566175448180991e-06, + "loss": 1.0703, + "step": 19488 + }, + { + "epoch": 0.62, + "learning_rate": 6.565202644252953e-06, + "loss": 0.9111, + "step": 19489 + }, + { + "epoch": 0.62, + "learning_rate": 6.564229877176491e-06, + "loss": 0.9253, + "step": 19490 + }, + { + "epoch": 0.62, + "learning_rate": 6.5632571469620385e-06, + "loss": 0.8906, + "step": 19491 + }, + { + "epoch": 0.62, + "learning_rate": 6.562284453620032e-06, + "loss": 0.9399, + "step": 19492 + }, + { + "epoch": 0.62, + "learning_rate": 6.561311797160907e-06, + "loss": 1.0117, + "step": 19493 + }, + { + "epoch": 0.62, + "learning_rate": 6.560339177595103e-06, + "loss": 0.9219, + "step": 19494 + }, + { + "epoch": 0.62, + "learning_rate": 6.559366594933052e-06, + "loss": 0.9448, + "step": 19495 + }, + { + "epoch": 0.62, + "learning_rate": 6.558394049185191e-06, + "loss": 0.9287, + "step": 19496 + }, + { + "epoch": 0.62, + "learning_rate": 6.5574215403619515e-06, + "loss": 0.8691, + "step": 19497 + }, + { + "epoch": 0.62, + "learning_rate": 6.556449068473773e-06, + "loss": 0.9868, + "step": 19498 + }, + { + "epoch": 0.62, + "learning_rate": 6.555476633531081e-06, + "loss": 0.9453, + "step": 19499 + }, + { + "epoch": 0.62, + "learning_rate": 6.554504235544313e-06, + "loss": 0.9268, + "step": 19500 + }, + { + "epoch": 0.62, + "learning_rate": 6.5535318745238994e-06, + "loss": 1.0596, + "step": 19501 + }, + { + "epoch": 0.62, + "learning_rate": 6.552559550480275e-06, + "loss": 1.0508, + "step": 19502 + }, + { + "epoch": 0.62, + "learning_rate": 6.551587263423872e-06, + "loss": 1.0361, + "step": 19503 + }, + { + "epoch": 0.62, + "learning_rate": 6.55061501336512e-06, + "loss": 0.8818, + "step": 19504 + }, + { + "epoch": 0.62, + "learning_rate": 6.549642800314458e-06, + "loss": 0.9814, + "step": 19505 + }, + { + "epoch": 0.62, + "learning_rate": 6.548670624282303e-06, + "loss": 0.9463, + "step": 19506 + }, + { + "epoch": 0.62, + "learning_rate": 6.547698485279095e-06, + "loss": 1.0024, + "step": 19507 + }, + { + "epoch": 0.62, + "learning_rate": 6.54672638331526e-06, + "loss": 0.939, + "step": 19508 + }, + { + "epoch": 0.62, + "learning_rate": 6.545754318401228e-06, + "loss": 1.0376, + "step": 19509 + }, + { + "epoch": 0.62, + "learning_rate": 6.544782290547431e-06, + "loss": 0.9292, + "step": 19510 + }, + { + "epoch": 0.62, + "learning_rate": 6.543810299764295e-06, + "loss": 1.0806, + "step": 19511 + }, + { + "epoch": 0.62, + "learning_rate": 6.54283834606225e-06, + "loss": 1.0122, + "step": 19512 + }, + { + "epoch": 0.62, + "learning_rate": 6.541866429451728e-06, + "loss": 1.0312, + "step": 19513 + }, + { + "epoch": 0.62, + "learning_rate": 6.5408945499431466e-06, + "loss": 0.8345, + "step": 19514 + }, + { + "epoch": 0.62, + "learning_rate": 6.539922707546939e-06, + "loss": 1.0137, + "step": 19515 + }, + { + "epoch": 0.62, + "learning_rate": 6.538950902273531e-06, + "loss": 1.0039, + "step": 19516 + }, + { + "epoch": 0.62, + "learning_rate": 6.537979134133349e-06, + "loss": 0.9077, + "step": 19517 + }, + { + "epoch": 0.62, + "learning_rate": 6.537007403136819e-06, + "loss": 0.5671, + "step": 19518 + }, + { + "epoch": 0.62, + "learning_rate": 6.536035709294371e-06, + "loss": 0.8423, + "step": 19519 + }, + { + "epoch": 0.62, + "learning_rate": 6.535064052616425e-06, + "loss": 1.0776, + "step": 19520 + }, + { + "epoch": 0.62, + "learning_rate": 6.534092433113404e-06, + "loss": 0.999, + "step": 19521 + }, + { + "epoch": 0.62, + "learning_rate": 6.533120850795735e-06, + "loss": 1.0054, + "step": 19522 + }, + { + "epoch": 0.62, + "learning_rate": 6.532149305673844e-06, + "loss": 0.8506, + "step": 19523 + }, + { + "epoch": 0.62, + "learning_rate": 6.531177797758155e-06, + "loss": 1.0278, + "step": 19524 + }, + { + "epoch": 0.62, + "learning_rate": 6.530206327059085e-06, + "loss": 1.0068, + "step": 19525 + }, + { + "epoch": 0.62, + "learning_rate": 6.529234893587063e-06, + "loss": 0.915, + "step": 19526 + }, + { + "epoch": 0.62, + "learning_rate": 6.528263497352509e-06, + "loss": 1.1016, + "step": 19527 + }, + { + "epoch": 0.62, + "learning_rate": 6.5272921383658465e-06, + "loss": 0.8486, + "step": 19528 + }, + { + "epoch": 0.62, + "learning_rate": 6.526320816637497e-06, + "loss": 1.0132, + "step": 19529 + }, + { + "epoch": 0.62, + "learning_rate": 6.525349532177878e-06, + "loss": 1.0742, + "step": 19530 + }, + { + "epoch": 0.62, + "learning_rate": 6.524378284997413e-06, + "loss": 0.959, + "step": 19531 + }, + { + "epoch": 0.62, + "learning_rate": 6.5234070751065225e-06, + "loss": 1.019, + "step": 19532 + }, + { + "epoch": 0.62, + "learning_rate": 6.522435902515626e-06, + "loss": 0.9521, + "step": 19533 + }, + { + "epoch": 0.62, + "learning_rate": 6.521464767235144e-06, + "loss": 1.0332, + "step": 19534 + }, + { + "epoch": 0.62, + "learning_rate": 6.520493669275499e-06, + "loss": 1.022, + "step": 19535 + }, + { + "epoch": 0.62, + "learning_rate": 6.519522608647102e-06, + "loss": 1.0571, + "step": 19536 + }, + { + "epoch": 0.62, + "learning_rate": 6.518551585360375e-06, + "loss": 0.9814, + "step": 19537 + }, + { + "epoch": 0.62, + "learning_rate": 6.517580599425737e-06, + "loss": 0.8843, + "step": 19538 + }, + { + "epoch": 0.62, + "learning_rate": 6.516609650853605e-06, + "loss": 0.9761, + "step": 19539 + }, + { + "epoch": 0.62, + "learning_rate": 6.515638739654396e-06, + "loss": 0.8779, + "step": 19540 + }, + { + "epoch": 0.62, + "learning_rate": 6.5146678658385265e-06, + "loss": 1.0112, + "step": 19541 + }, + { + "epoch": 0.62, + "learning_rate": 6.513697029416415e-06, + "loss": 0.854, + "step": 19542 + }, + { + "epoch": 0.62, + "learning_rate": 6.512726230398478e-06, + "loss": 1.0, + "step": 19543 + }, + { + "epoch": 0.63, + "learning_rate": 6.511755468795127e-06, + "loss": 0.8838, + "step": 19544 + }, + { + "epoch": 0.63, + "learning_rate": 6.510784744616777e-06, + "loss": 0.9243, + "step": 19545 + }, + { + "epoch": 0.63, + "learning_rate": 6.509814057873847e-06, + "loss": 0.9438, + "step": 19546 + }, + { + "epoch": 0.63, + "learning_rate": 6.508843408576748e-06, + "loss": 0.9766, + "step": 19547 + }, + { + "epoch": 0.63, + "learning_rate": 6.507872796735897e-06, + "loss": 0.8711, + "step": 19548 + }, + { + "epoch": 0.63, + "learning_rate": 6.506902222361704e-06, + "loss": 0.8652, + "step": 19549 + }, + { + "epoch": 0.63, + "learning_rate": 6.505931685464589e-06, + "loss": 0.5085, + "step": 19550 + }, + { + "epoch": 0.63, + "learning_rate": 6.504961186054956e-06, + "loss": 0.9229, + "step": 19551 + }, + { + "epoch": 0.63, + "learning_rate": 6.503990724143222e-06, + "loss": 0.8149, + "step": 19552 + }, + { + "epoch": 0.63, + "learning_rate": 6.503020299739797e-06, + "loss": 0.9536, + "step": 19553 + }, + { + "epoch": 0.63, + "learning_rate": 6.502049912855095e-06, + "loss": 1.0146, + "step": 19554 + }, + { + "epoch": 0.63, + "learning_rate": 6.501079563499526e-06, + "loss": 0.5007, + "step": 19555 + }, + { + "epoch": 0.63, + "learning_rate": 6.500109251683502e-06, + "loss": 0.9492, + "step": 19556 + }, + { + "epoch": 0.63, + "learning_rate": 6.499138977417429e-06, + "loss": 1.124, + "step": 19557 + }, + { + "epoch": 0.63, + "learning_rate": 6.4981687407117275e-06, + "loss": 0.978, + "step": 19558 + }, + { + "epoch": 0.63, + "learning_rate": 6.497198541576795e-06, + "loss": 1.019, + "step": 19559 + }, + { + "epoch": 0.63, + "learning_rate": 6.496228380023044e-06, + "loss": 0.9336, + "step": 19560 + }, + { + "epoch": 0.63, + "learning_rate": 6.495258256060886e-06, + "loss": 1.0073, + "step": 19561 + }, + { + "epoch": 0.63, + "learning_rate": 6.494288169700728e-06, + "loss": 1.0601, + "step": 19562 + }, + { + "epoch": 0.63, + "learning_rate": 6.49331812095298e-06, + "loss": 0.8594, + "step": 19563 + }, + { + "epoch": 0.63, + "learning_rate": 6.492348109828044e-06, + "loss": 0.9609, + "step": 19564 + }, + { + "epoch": 0.63, + "learning_rate": 6.491378136336333e-06, + "loss": 0.9478, + "step": 19565 + }, + { + "epoch": 0.63, + "learning_rate": 6.490408200488252e-06, + "loss": 0.9031, + "step": 19566 + }, + { + "epoch": 0.63, + "learning_rate": 6.489438302294205e-06, + "loss": 0.9282, + "step": 19567 + }, + { + "epoch": 0.63, + "learning_rate": 6.488468441764601e-06, + "loss": 0.9033, + "step": 19568 + }, + { + "epoch": 0.63, + "learning_rate": 6.487498618909845e-06, + "loss": 0.8984, + "step": 19569 + }, + { + "epoch": 0.63, + "learning_rate": 6.4865288337403385e-06, + "loss": 0.9971, + "step": 19570 + }, + { + "epoch": 0.63, + "learning_rate": 6.48555908626649e-06, + "loss": 0.8472, + "step": 19571 + }, + { + "epoch": 0.63, + "learning_rate": 6.4845893764987035e-06, + "loss": 0.9922, + "step": 19572 + }, + { + "epoch": 0.63, + "learning_rate": 6.483619704447384e-06, + "loss": 0.772, + "step": 19573 + }, + { + "epoch": 0.63, + "learning_rate": 6.482650070122933e-06, + "loss": 0.9941, + "step": 19574 + }, + { + "epoch": 0.63, + "learning_rate": 6.481680473535752e-06, + "loss": 0.9795, + "step": 19575 + }, + { + "epoch": 0.63, + "learning_rate": 6.480710914696247e-06, + "loss": 0.9243, + "step": 19576 + }, + { + "epoch": 0.63, + "learning_rate": 6.479741393614816e-06, + "loss": 0.9531, + "step": 19577 + }, + { + "epoch": 0.63, + "learning_rate": 6.478771910301866e-06, + "loss": 1.0278, + "step": 19578 + }, + { + "epoch": 0.63, + "learning_rate": 6.477802464767795e-06, + "loss": 0.9487, + "step": 19579 + }, + { + "epoch": 0.63, + "learning_rate": 6.476833057023007e-06, + "loss": 0.9194, + "step": 19580 + }, + { + "epoch": 0.63, + "learning_rate": 6.4758636870779035e-06, + "loss": 0.8589, + "step": 19581 + }, + { + "epoch": 0.63, + "learning_rate": 6.47489435494288e-06, + "loss": 0.9873, + "step": 19582 + }, + { + "epoch": 0.63, + "learning_rate": 6.473925060628337e-06, + "loss": 0.8408, + "step": 19583 + }, + { + "epoch": 0.63, + "learning_rate": 6.472955804144676e-06, + "loss": 0.7795, + "step": 19584 + }, + { + "epoch": 0.63, + "learning_rate": 6.4719865855022964e-06, + "loss": 0.9648, + "step": 19585 + }, + { + "epoch": 0.63, + "learning_rate": 6.471017404711597e-06, + "loss": 1.0986, + "step": 19586 + }, + { + "epoch": 0.63, + "learning_rate": 6.470048261782973e-06, + "loss": 0.9995, + "step": 19587 + }, + { + "epoch": 0.63, + "learning_rate": 6.4690791567268295e-06, + "loss": 0.8572, + "step": 19588 + }, + { + "epoch": 0.63, + "learning_rate": 6.468110089553554e-06, + "loss": 0.9878, + "step": 19589 + }, + { + "epoch": 0.63, + "learning_rate": 6.46714106027355e-06, + "loss": 0.8945, + "step": 19590 + }, + { + "epoch": 0.63, + "learning_rate": 6.466172068897212e-06, + "loss": 0.8911, + "step": 19591 + }, + { + "epoch": 0.63, + "learning_rate": 6.465203115434935e-06, + "loss": 0.9146, + "step": 19592 + }, + { + "epoch": 0.63, + "learning_rate": 6.464234199897117e-06, + "loss": 0.957, + "step": 19593 + }, + { + "epoch": 0.63, + "learning_rate": 6.463265322294154e-06, + "loss": 0.9966, + "step": 19594 + }, + { + "epoch": 0.63, + "learning_rate": 6.462296482636439e-06, + "loss": 1.0654, + "step": 19595 + }, + { + "epoch": 0.63, + "learning_rate": 6.461327680934373e-06, + "loss": 0.8364, + "step": 19596 + }, + { + "epoch": 0.63, + "learning_rate": 6.460358917198339e-06, + "loss": 0.9673, + "step": 19597 + }, + { + "epoch": 0.63, + "learning_rate": 6.459390191438736e-06, + "loss": 0.9746, + "step": 19598 + }, + { + "epoch": 0.63, + "learning_rate": 6.458421503665958e-06, + "loss": 0.9878, + "step": 19599 + }, + { + "epoch": 0.63, + "learning_rate": 6.457452853890398e-06, + "loss": 1.0322, + "step": 19600 + }, + { + "epoch": 0.63, + "learning_rate": 6.456484242122447e-06, + "loss": 0.8965, + "step": 19601 + }, + { + "epoch": 0.63, + "learning_rate": 6.455515668372503e-06, + "loss": 1.04, + "step": 19602 + }, + { + "epoch": 0.63, + "learning_rate": 6.45454713265095e-06, + "loss": 0.9766, + "step": 19603 + }, + { + "epoch": 0.63, + "learning_rate": 6.453578634968183e-06, + "loss": 0.9702, + "step": 19604 + }, + { + "epoch": 0.63, + "learning_rate": 6.452610175334591e-06, + "loss": 0.9834, + "step": 19605 + }, + { + "epoch": 0.63, + "learning_rate": 6.451641753760565e-06, + "loss": 0.9868, + "step": 19606 + }, + { + "epoch": 0.63, + "learning_rate": 6.450673370256497e-06, + "loss": 1.0229, + "step": 19607 + }, + { + "epoch": 0.63, + "learning_rate": 6.449705024832777e-06, + "loss": 0.9609, + "step": 19608 + }, + { + "epoch": 0.63, + "learning_rate": 6.448736717499791e-06, + "loss": 0.8232, + "step": 19609 + }, + { + "epoch": 0.63, + "learning_rate": 6.4477684482679295e-06, + "loss": 0.9736, + "step": 19610 + }, + { + "epoch": 0.63, + "learning_rate": 6.446800217147584e-06, + "loss": 0.9209, + "step": 19611 + }, + { + "epoch": 0.63, + "learning_rate": 6.4458320241491366e-06, + "loss": 1.0562, + "step": 19612 + }, + { + "epoch": 0.63, + "learning_rate": 6.44486386928298e-06, + "loss": 0.9497, + "step": 19613 + }, + { + "epoch": 0.63, + "learning_rate": 6.443895752559498e-06, + "loss": 0.8979, + "step": 19614 + }, + { + "epoch": 0.63, + "learning_rate": 6.442927673989079e-06, + "loss": 0.936, + "step": 19615 + }, + { + "epoch": 0.63, + "learning_rate": 6.441959633582109e-06, + "loss": 0.8853, + "step": 19616 + }, + { + "epoch": 0.63, + "learning_rate": 6.440991631348974e-06, + "loss": 0.9375, + "step": 19617 + }, + { + "epoch": 0.63, + "learning_rate": 6.440023667300064e-06, + "loss": 0.8413, + "step": 19618 + }, + { + "epoch": 0.63, + "learning_rate": 6.439055741445756e-06, + "loss": 0.8936, + "step": 19619 + }, + { + "epoch": 0.63, + "learning_rate": 6.43808785379644e-06, + "loss": 0.9844, + "step": 19620 + }, + { + "epoch": 0.63, + "learning_rate": 6.437120004362498e-06, + "loss": 1.0063, + "step": 19621 + }, + { + "epoch": 0.63, + "learning_rate": 6.436152193154315e-06, + "loss": 0.9453, + "step": 19622 + }, + { + "epoch": 0.63, + "learning_rate": 6.435184420182274e-06, + "loss": 1.0107, + "step": 19623 + }, + { + "epoch": 0.63, + "learning_rate": 6.434216685456761e-06, + "loss": 0.9473, + "step": 19624 + }, + { + "epoch": 0.63, + "learning_rate": 6.433248988988155e-06, + "loss": 0.9116, + "step": 19625 + }, + { + "epoch": 0.63, + "learning_rate": 6.4322813307868445e-06, + "loss": 1.0483, + "step": 19626 + }, + { + "epoch": 0.63, + "learning_rate": 6.4313137108632025e-06, + "loss": 0.8403, + "step": 19627 + }, + { + "epoch": 0.63, + "learning_rate": 6.4303461292276156e-06, + "loss": 0.9263, + "step": 19628 + }, + { + "epoch": 0.63, + "learning_rate": 6.429378585890463e-06, + "loss": 0.9111, + "step": 19629 + }, + { + "epoch": 0.63, + "learning_rate": 6.428411080862127e-06, + "loss": 0.9663, + "step": 19630 + }, + { + "epoch": 0.63, + "learning_rate": 6.427443614152987e-06, + "loss": 0.8936, + "step": 19631 + }, + { + "epoch": 0.63, + "learning_rate": 6.426476185773423e-06, + "loss": 1.0186, + "step": 19632 + }, + { + "epoch": 0.63, + "learning_rate": 6.425508795733819e-06, + "loss": 0.918, + "step": 19633 + }, + { + "epoch": 0.63, + "learning_rate": 6.4245414440445455e-06, + "loss": 1.0312, + "step": 19634 + }, + { + "epoch": 0.63, + "learning_rate": 6.423574130715984e-06, + "loss": 0.9893, + "step": 19635 + }, + { + "epoch": 0.63, + "learning_rate": 6.422606855758516e-06, + "loss": 0.8408, + "step": 19636 + }, + { + "epoch": 0.63, + "learning_rate": 6.421639619182516e-06, + "loss": 1.0186, + "step": 19637 + }, + { + "epoch": 0.63, + "learning_rate": 6.420672420998362e-06, + "loss": 0.9653, + "step": 19638 + }, + { + "epoch": 0.63, + "learning_rate": 6.419705261216431e-06, + "loss": 1.1333, + "step": 19639 + }, + { + "epoch": 0.63, + "learning_rate": 6.418738139847102e-06, + "loss": 1.0454, + "step": 19640 + }, + { + "epoch": 0.63, + "learning_rate": 6.4177710569007524e-06, + "loss": 0.855, + "step": 19641 + }, + { + "epoch": 0.63, + "learning_rate": 6.41680401238775e-06, + "loss": 0.939, + "step": 19642 + }, + { + "epoch": 0.63, + "learning_rate": 6.415837006318476e-06, + "loss": 1.0171, + "step": 19643 + }, + { + "epoch": 0.63, + "learning_rate": 6.4148700387033035e-06, + "loss": 0.8862, + "step": 19644 + }, + { + "epoch": 0.63, + "learning_rate": 6.413903109552608e-06, + "loss": 0.8525, + "step": 19645 + }, + { + "epoch": 0.63, + "learning_rate": 6.412936218876763e-06, + "loss": 1.0713, + "step": 19646 + }, + { + "epoch": 0.63, + "learning_rate": 6.411969366686144e-06, + "loss": 1.019, + "step": 19647 + }, + { + "epoch": 0.63, + "learning_rate": 6.411002552991121e-06, + "loss": 0.9331, + "step": 19648 + }, + { + "epoch": 0.63, + "learning_rate": 6.410035777802072e-06, + "loss": 0.8896, + "step": 19649 + }, + { + "epoch": 0.63, + "learning_rate": 6.409069041129362e-06, + "loss": 0.9849, + "step": 19650 + }, + { + "epoch": 0.63, + "learning_rate": 6.408102342983368e-06, + "loss": 0.854, + "step": 19651 + }, + { + "epoch": 0.63, + "learning_rate": 6.407135683374461e-06, + "loss": 0.9287, + "step": 19652 + }, + { + "epoch": 0.63, + "learning_rate": 6.406169062313011e-06, + "loss": 0.9009, + "step": 19653 + }, + { + "epoch": 0.63, + "learning_rate": 6.405202479809389e-06, + "loss": 0.9004, + "step": 19654 + }, + { + "epoch": 0.63, + "learning_rate": 6.404235935873967e-06, + "loss": 0.9204, + "step": 19655 + }, + { + "epoch": 0.63, + "learning_rate": 6.403269430517115e-06, + "loss": 1.0464, + "step": 19656 + }, + { + "epoch": 0.63, + "learning_rate": 6.402302963749201e-06, + "loss": 1.0742, + "step": 19657 + }, + { + "epoch": 0.63, + "learning_rate": 6.401336535580594e-06, + "loss": 1.1357, + "step": 19658 + }, + { + "epoch": 0.63, + "learning_rate": 6.400370146021662e-06, + "loss": 0.9458, + "step": 19659 + }, + { + "epoch": 0.63, + "learning_rate": 6.399403795082773e-06, + "loss": 1.022, + "step": 19660 + }, + { + "epoch": 0.63, + "learning_rate": 6.398437482774298e-06, + "loss": 0.9575, + "step": 19661 + }, + { + "epoch": 0.63, + "learning_rate": 6.397471209106603e-06, + "loss": 0.9604, + "step": 19662 + }, + { + "epoch": 0.63, + "learning_rate": 6.396504974090054e-06, + "loss": 1.0063, + "step": 19663 + }, + { + "epoch": 0.63, + "learning_rate": 6.395538777735023e-06, + "loss": 0.958, + "step": 19664 + }, + { + "epoch": 0.63, + "learning_rate": 6.394572620051868e-06, + "loss": 0.9102, + "step": 19665 + }, + { + "epoch": 0.63, + "learning_rate": 6.393606501050957e-06, + "loss": 0.5078, + "step": 19666 + }, + { + "epoch": 0.63, + "learning_rate": 6.39264042074266e-06, + "loss": 1.0024, + "step": 19667 + }, + { + "epoch": 0.63, + "learning_rate": 6.391674379137336e-06, + "loss": 0.9937, + "step": 19668 + }, + { + "epoch": 0.63, + "learning_rate": 6.390708376245354e-06, + "loss": 0.979, + "step": 19669 + }, + { + "epoch": 0.63, + "learning_rate": 6.389742412077074e-06, + "loss": 0.9678, + "step": 19670 + }, + { + "epoch": 0.63, + "learning_rate": 6.388776486642869e-06, + "loss": 1.0098, + "step": 19671 + }, + { + "epoch": 0.63, + "learning_rate": 6.38781059995309e-06, + "loss": 0.957, + "step": 19672 + }, + { + "epoch": 0.63, + "learning_rate": 6.386844752018106e-06, + "loss": 0.897, + "step": 19673 + }, + { + "epoch": 0.63, + "learning_rate": 6.38587894284828e-06, + "loss": 1.0107, + "step": 19674 + }, + { + "epoch": 0.63, + "learning_rate": 6.384913172453973e-06, + "loss": 0.9888, + "step": 19675 + }, + { + "epoch": 0.63, + "learning_rate": 6.383947440845545e-06, + "loss": 0.9697, + "step": 19676 + }, + { + "epoch": 0.63, + "learning_rate": 6.382981748033359e-06, + "loss": 0.9795, + "step": 19677 + }, + { + "epoch": 0.63, + "learning_rate": 6.382016094027776e-06, + "loss": 1.0239, + "step": 19678 + }, + { + "epoch": 0.63, + "learning_rate": 6.381050478839161e-06, + "loss": 0.9517, + "step": 19679 + }, + { + "epoch": 0.63, + "learning_rate": 6.380084902477865e-06, + "loss": 0.9155, + "step": 19680 + }, + { + "epoch": 0.63, + "learning_rate": 6.379119364954251e-06, + "loss": 1.0117, + "step": 19681 + }, + { + "epoch": 0.63, + "learning_rate": 6.37815386627868e-06, + "loss": 0.5117, + "step": 19682 + }, + { + "epoch": 0.63, + "learning_rate": 6.377188406461509e-06, + "loss": 0.9541, + "step": 19683 + }, + { + "epoch": 0.63, + "learning_rate": 6.376222985513096e-06, + "loss": 0.9629, + "step": 19684 + }, + { + "epoch": 0.63, + "learning_rate": 6.375257603443801e-06, + "loss": 0.8228, + "step": 19685 + }, + { + "epoch": 0.63, + "learning_rate": 6.374292260263983e-06, + "loss": 0.8696, + "step": 19686 + }, + { + "epoch": 0.63, + "learning_rate": 6.373326955983994e-06, + "loss": 1.0698, + "step": 19687 + }, + { + "epoch": 0.63, + "learning_rate": 6.372361690614192e-06, + "loss": 0.5205, + "step": 19688 + }, + { + "epoch": 0.63, + "learning_rate": 6.371396464164934e-06, + "loss": 0.9634, + "step": 19689 + }, + { + "epoch": 0.63, + "learning_rate": 6.370431276646576e-06, + "loss": 1.0059, + "step": 19690 + }, + { + "epoch": 0.63, + "learning_rate": 6.369466128069475e-06, + "loss": 0.9199, + "step": 19691 + }, + { + "epoch": 0.63, + "learning_rate": 6.3685010184439835e-06, + "loss": 0.8916, + "step": 19692 + }, + { + "epoch": 0.63, + "learning_rate": 6.367535947780457e-06, + "loss": 0.9756, + "step": 19693 + }, + { + "epoch": 0.63, + "learning_rate": 6.3665709160892494e-06, + "loss": 0.9941, + "step": 19694 + }, + { + "epoch": 0.63, + "learning_rate": 6.365605923380714e-06, + "loss": 0.9683, + "step": 19695 + }, + { + "epoch": 0.63, + "learning_rate": 6.364640969665205e-06, + "loss": 0.9443, + "step": 19696 + }, + { + "epoch": 0.63, + "learning_rate": 6.363676054953076e-06, + "loss": 0.9414, + "step": 19697 + }, + { + "epoch": 0.63, + "learning_rate": 6.362711179254678e-06, + "loss": 0.854, + "step": 19698 + }, + { + "epoch": 0.63, + "learning_rate": 6.361746342580363e-06, + "loss": 0.9668, + "step": 19699 + }, + { + "epoch": 0.63, + "learning_rate": 6.360781544940482e-06, + "loss": 0.4805, + "step": 19700 + }, + { + "epoch": 0.63, + "learning_rate": 6.359816786345391e-06, + "loss": 1.0073, + "step": 19701 + }, + { + "epoch": 0.63, + "learning_rate": 6.358852066805437e-06, + "loss": 0.9736, + "step": 19702 + }, + { + "epoch": 0.63, + "learning_rate": 6.3578873863309675e-06, + "loss": 1.0049, + "step": 19703 + }, + { + "epoch": 0.63, + "learning_rate": 6.356922744932335e-06, + "loss": 0.9775, + "step": 19704 + }, + { + "epoch": 0.63, + "learning_rate": 6.35595814261989e-06, + "loss": 1.0195, + "step": 19705 + }, + { + "epoch": 0.63, + "learning_rate": 6.354993579403981e-06, + "loss": 0.8735, + "step": 19706 + }, + { + "epoch": 0.63, + "learning_rate": 6.354029055294957e-06, + "loss": 1.0776, + "step": 19707 + }, + { + "epoch": 0.63, + "learning_rate": 6.353064570303167e-06, + "loss": 0.9043, + "step": 19708 + }, + { + "epoch": 0.63, + "learning_rate": 6.352100124438961e-06, + "loss": 0.998, + "step": 19709 + }, + { + "epoch": 0.63, + "learning_rate": 6.35113571771268e-06, + "loss": 1.1489, + "step": 19710 + }, + { + "epoch": 0.63, + "learning_rate": 6.3501713501346726e-06, + "loss": 1.0249, + "step": 19711 + }, + { + "epoch": 0.63, + "learning_rate": 6.349207021715289e-06, + "loss": 1.0566, + "step": 19712 + }, + { + "epoch": 0.63, + "learning_rate": 6.348242732464873e-06, + "loss": 1.0498, + "step": 19713 + }, + { + "epoch": 0.63, + "learning_rate": 6.34727848239377e-06, + "loss": 0.8994, + "step": 19714 + }, + { + "epoch": 0.63, + "learning_rate": 6.346314271512328e-06, + "loss": 0.9673, + "step": 19715 + }, + { + "epoch": 0.63, + "learning_rate": 6.345350099830893e-06, + "loss": 0.9932, + "step": 19716 + }, + { + "epoch": 0.63, + "learning_rate": 6.344385967359804e-06, + "loss": 1.0742, + "step": 19717 + }, + { + "epoch": 0.63, + "learning_rate": 6.343421874109407e-06, + "loss": 0.8662, + "step": 19718 + }, + { + "epoch": 0.63, + "learning_rate": 6.342457820090045e-06, + "loss": 0.9741, + "step": 19719 + }, + { + "epoch": 0.63, + "learning_rate": 6.341493805312064e-06, + "loss": 0.8599, + "step": 19720 + }, + { + "epoch": 0.63, + "learning_rate": 6.340529829785806e-06, + "loss": 1.0127, + "step": 19721 + }, + { + "epoch": 0.63, + "learning_rate": 6.339565893521611e-06, + "loss": 0.958, + "step": 19722 + }, + { + "epoch": 0.63, + "learning_rate": 6.338601996529825e-06, + "loss": 0.9321, + "step": 19723 + }, + { + "epoch": 0.63, + "learning_rate": 6.3376381388207895e-06, + "loss": 0.8467, + "step": 19724 + }, + { + "epoch": 0.63, + "learning_rate": 6.336674320404841e-06, + "loss": 0.9243, + "step": 19725 + }, + { + "epoch": 0.63, + "learning_rate": 6.335710541292321e-06, + "loss": 0.9922, + "step": 19726 + }, + { + "epoch": 0.63, + "learning_rate": 6.334746801493572e-06, + "loss": 0.9175, + "step": 19727 + }, + { + "epoch": 0.63, + "learning_rate": 6.3337831010189335e-06, + "loss": 1.0029, + "step": 19728 + }, + { + "epoch": 0.63, + "learning_rate": 6.3328194398787455e-06, + "loss": 1.0586, + "step": 19729 + }, + { + "epoch": 0.63, + "learning_rate": 6.331855818083348e-06, + "loss": 1.0127, + "step": 19730 + }, + { + "epoch": 0.63, + "learning_rate": 6.330892235643076e-06, + "loss": 0.9238, + "step": 19731 + }, + { + "epoch": 0.63, + "learning_rate": 6.32992869256827e-06, + "loss": 0.8584, + "step": 19732 + }, + { + "epoch": 0.63, + "learning_rate": 6.328965188869267e-06, + "loss": 0.9995, + "step": 19733 + }, + { + "epoch": 0.63, + "learning_rate": 6.3280017245564025e-06, + "loss": 0.8091, + "step": 19734 + }, + { + "epoch": 0.63, + "learning_rate": 6.327038299640018e-06, + "loss": 1.0137, + "step": 19735 + }, + { + "epoch": 0.63, + "learning_rate": 6.326074914130449e-06, + "loss": 1.0117, + "step": 19736 + }, + { + "epoch": 0.63, + "learning_rate": 6.325111568038027e-06, + "loss": 0.9531, + "step": 19737 + }, + { + "epoch": 0.63, + "learning_rate": 6.324148261373091e-06, + "loss": 0.9263, + "step": 19738 + }, + { + "epoch": 0.63, + "learning_rate": 6.323184994145979e-06, + "loss": 1.0571, + "step": 19739 + }, + { + "epoch": 0.63, + "learning_rate": 6.322221766367019e-06, + "loss": 0.96, + "step": 19740 + }, + { + "epoch": 0.63, + "learning_rate": 6.3212585780465515e-06, + "loss": 0.9556, + "step": 19741 + }, + { + "epoch": 0.63, + "learning_rate": 6.320295429194907e-06, + "loss": 0.9731, + "step": 19742 + }, + { + "epoch": 0.63, + "learning_rate": 6.31933231982242e-06, + "loss": 0.9609, + "step": 19743 + }, + { + "epoch": 0.63, + "learning_rate": 6.318369249939423e-06, + "loss": 1.0117, + "step": 19744 + }, + { + "epoch": 0.63, + "learning_rate": 6.3174062195562504e-06, + "loss": 0.9087, + "step": 19745 + }, + { + "epoch": 0.63, + "learning_rate": 6.316443228683231e-06, + "loss": 1.0771, + "step": 19746 + }, + { + "epoch": 0.63, + "learning_rate": 6.315480277330706e-06, + "loss": 0.9805, + "step": 19747 + }, + { + "epoch": 0.63, + "learning_rate": 6.3145173655089944e-06, + "loss": 0.8984, + "step": 19748 + }, + { + "epoch": 0.63, + "learning_rate": 6.313554493228431e-06, + "loss": 0.9512, + "step": 19749 + }, + { + "epoch": 0.63, + "learning_rate": 6.31259166049935e-06, + "loss": 0.9375, + "step": 19750 + }, + { + "epoch": 0.63, + "learning_rate": 6.311628867332079e-06, + "loss": 0.9116, + "step": 19751 + }, + { + "epoch": 0.63, + "learning_rate": 6.310666113736947e-06, + "loss": 0.8564, + "step": 19752 + }, + { + "epoch": 0.63, + "learning_rate": 6.309703399724288e-06, + "loss": 1.019, + "step": 19753 + }, + { + "epoch": 0.63, + "learning_rate": 6.3087407253044275e-06, + "loss": 0.9678, + "step": 19754 + }, + { + "epoch": 0.63, + "learning_rate": 6.307778090487691e-06, + "loss": 0.5615, + "step": 19755 + }, + { + "epoch": 0.63, + "learning_rate": 6.3068154952844086e-06, + "loss": 0.9316, + "step": 19756 + }, + { + "epoch": 0.63, + "learning_rate": 6.305852939704911e-06, + "loss": 1.0605, + "step": 19757 + }, + { + "epoch": 0.63, + "learning_rate": 6.30489042375952e-06, + "loss": 0.8916, + "step": 19758 + }, + { + "epoch": 0.63, + "learning_rate": 6.303927947458566e-06, + "loss": 0.9761, + "step": 19759 + }, + { + "epoch": 0.63, + "learning_rate": 6.302965510812376e-06, + "loss": 0.8647, + "step": 19760 + }, + { + "epoch": 0.63, + "learning_rate": 6.302003113831272e-06, + "loss": 0.9766, + "step": 19761 + }, + { + "epoch": 0.63, + "learning_rate": 6.301040756525587e-06, + "loss": 0.8984, + "step": 19762 + }, + { + "epoch": 0.63, + "learning_rate": 6.300078438905637e-06, + "loss": 1.0322, + "step": 19763 + }, + { + "epoch": 0.63, + "learning_rate": 6.299116160981751e-06, + "loss": 0.8955, + "step": 19764 + }, + { + "epoch": 0.63, + "learning_rate": 6.298153922764252e-06, + "loss": 0.9312, + "step": 19765 + }, + { + "epoch": 0.63, + "learning_rate": 6.297191724263464e-06, + "loss": 1.0225, + "step": 19766 + }, + { + "epoch": 0.63, + "learning_rate": 6.296229565489711e-06, + "loss": 1.02, + "step": 19767 + }, + { + "epoch": 0.63, + "learning_rate": 6.2952674464533155e-06, + "loss": 0.9106, + "step": 19768 + }, + { + "epoch": 0.63, + "learning_rate": 6.2943053671646045e-06, + "loss": 0.8667, + "step": 19769 + }, + { + "epoch": 0.63, + "learning_rate": 6.293343327633892e-06, + "loss": 0.9844, + "step": 19770 + }, + { + "epoch": 0.63, + "learning_rate": 6.292381327871503e-06, + "loss": 0.9341, + "step": 19771 + }, + { + "epoch": 0.63, + "learning_rate": 6.291419367887759e-06, + "loss": 0.8408, + "step": 19772 + }, + { + "epoch": 0.63, + "learning_rate": 6.29045744769298e-06, + "loss": 1.1548, + "step": 19773 + }, + { + "epoch": 0.63, + "learning_rate": 6.2894955672974875e-06, + "loss": 0.9194, + "step": 19774 + }, + { + "epoch": 0.63, + "learning_rate": 6.288533726711603e-06, + "loss": 1.0215, + "step": 19775 + }, + { + "epoch": 0.63, + "learning_rate": 6.287571925945642e-06, + "loss": 0.9229, + "step": 19776 + }, + { + "epoch": 0.63, + "learning_rate": 6.286610165009928e-06, + "loss": 1.0259, + "step": 19777 + }, + { + "epoch": 0.63, + "learning_rate": 6.2856484439147745e-06, + "loss": 0.9082, + "step": 19778 + }, + { + "epoch": 0.63, + "learning_rate": 6.284686762670503e-06, + "loss": 1.0532, + "step": 19779 + }, + { + "epoch": 0.63, + "learning_rate": 6.283725121287432e-06, + "loss": 0.9692, + "step": 19780 + }, + { + "epoch": 0.63, + "learning_rate": 6.282763519775876e-06, + "loss": 0.9229, + "step": 19781 + }, + { + "epoch": 0.63, + "learning_rate": 6.281801958146153e-06, + "loss": 0.9658, + "step": 19782 + }, + { + "epoch": 0.63, + "learning_rate": 6.28084043640858e-06, + "loss": 0.8457, + "step": 19783 + }, + { + "epoch": 0.63, + "learning_rate": 6.279878954573475e-06, + "loss": 0.9956, + "step": 19784 + }, + { + "epoch": 0.63, + "learning_rate": 6.278917512651152e-06, + "loss": 0.9243, + "step": 19785 + }, + { + "epoch": 0.63, + "learning_rate": 6.277956110651924e-06, + "loss": 0.9863, + "step": 19786 + }, + { + "epoch": 0.63, + "learning_rate": 6.276994748586106e-06, + "loss": 1.022, + "step": 19787 + }, + { + "epoch": 0.63, + "learning_rate": 6.2760334264640144e-06, + "loss": 0.9365, + "step": 19788 + }, + { + "epoch": 0.63, + "learning_rate": 6.2750721442959635e-06, + "loss": 0.8706, + "step": 19789 + }, + { + "epoch": 0.63, + "learning_rate": 6.274110902092265e-06, + "loss": 1.0122, + "step": 19790 + }, + { + "epoch": 0.63, + "learning_rate": 6.273149699863234e-06, + "loss": 1.0503, + "step": 19791 + }, + { + "epoch": 0.63, + "learning_rate": 6.272188537619186e-06, + "loss": 0.9424, + "step": 19792 + }, + { + "epoch": 0.63, + "learning_rate": 6.271227415370425e-06, + "loss": 0.9409, + "step": 19793 + }, + { + "epoch": 0.63, + "learning_rate": 6.270266333127266e-06, + "loss": 0.9102, + "step": 19794 + }, + { + "epoch": 0.63, + "learning_rate": 6.269305290900023e-06, + "loss": 1.0024, + "step": 19795 + }, + { + "epoch": 0.63, + "learning_rate": 6.268344288699004e-06, + "loss": 0.9814, + "step": 19796 + }, + { + "epoch": 0.63, + "learning_rate": 6.26738332653452e-06, + "loss": 0.9165, + "step": 19797 + }, + { + "epoch": 0.63, + "learning_rate": 6.266422404416884e-06, + "loss": 0.9277, + "step": 19798 + }, + { + "epoch": 0.63, + "learning_rate": 6.265461522356402e-06, + "loss": 0.9004, + "step": 19799 + }, + { + "epoch": 0.63, + "learning_rate": 6.2645006803633905e-06, + "loss": 0.9409, + "step": 19800 + }, + { + "epoch": 0.63, + "learning_rate": 6.263539878448146e-06, + "loss": 0.9839, + "step": 19801 + }, + { + "epoch": 0.63, + "learning_rate": 6.262579116620986e-06, + "loss": 0.8579, + "step": 19802 + }, + { + "epoch": 0.63, + "learning_rate": 6.261618394892213e-06, + "loss": 0.9824, + "step": 19803 + }, + { + "epoch": 0.63, + "learning_rate": 6.260657713272138e-06, + "loss": 0.8926, + "step": 19804 + }, + { + "epoch": 0.63, + "learning_rate": 6.259697071771069e-06, + "loss": 0.875, + "step": 19805 + }, + { + "epoch": 0.63, + "learning_rate": 6.258736470399309e-06, + "loss": 0.8413, + "step": 19806 + }, + { + "epoch": 0.63, + "learning_rate": 6.257775909167171e-06, + "loss": 0.8853, + "step": 19807 + }, + { + "epoch": 0.63, + "learning_rate": 6.256815388084952e-06, + "loss": 0.9883, + "step": 19808 + }, + { + "epoch": 0.63, + "learning_rate": 6.2558549071629595e-06, + "loss": 1.0005, + "step": 19809 + }, + { + "epoch": 0.63, + "learning_rate": 6.2548944664115005e-06, + "loss": 0.488, + "step": 19810 + }, + { + "epoch": 0.63, + "learning_rate": 6.25393406584088e-06, + "loss": 0.835, + "step": 19811 + }, + { + "epoch": 0.63, + "learning_rate": 6.252973705461401e-06, + "loss": 1.0903, + "step": 19812 + }, + { + "epoch": 0.63, + "learning_rate": 6.252013385283367e-06, + "loss": 1.0542, + "step": 19813 + }, + { + "epoch": 0.63, + "learning_rate": 6.251053105317082e-06, + "loss": 0.8281, + "step": 19814 + }, + { + "epoch": 0.63, + "learning_rate": 6.25009286557285e-06, + "loss": 0.4839, + "step": 19815 + }, + { + "epoch": 0.63, + "learning_rate": 6.249132666060968e-06, + "loss": 0.916, + "step": 19816 + }, + { + "epoch": 0.63, + "learning_rate": 6.248172506791741e-06, + "loss": 1.0396, + "step": 19817 + }, + { + "epoch": 0.63, + "learning_rate": 6.247212387775472e-06, + "loss": 0.9102, + "step": 19818 + }, + { + "epoch": 0.63, + "learning_rate": 6.246252309022463e-06, + "loss": 1.0293, + "step": 19819 + }, + { + "epoch": 0.63, + "learning_rate": 6.2452922705430085e-06, + "loss": 0.9512, + "step": 19820 + }, + { + "epoch": 0.63, + "learning_rate": 6.244332272347413e-06, + "loss": 1.1538, + "step": 19821 + }, + { + "epoch": 0.63, + "learning_rate": 6.243372314445978e-06, + "loss": 0.9805, + "step": 19822 + }, + { + "epoch": 0.63, + "learning_rate": 6.242412396848998e-06, + "loss": 0.9961, + "step": 19823 + }, + { + "epoch": 0.63, + "learning_rate": 6.241452519566775e-06, + "loss": 1.0469, + "step": 19824 + }, + { + "epoch": 0.63, + "learning_rate": 6.240492682609608e-06, + "loss": 0.9819, + "step": 19825 + }, + { + "epoch": 0.63, + "learning_rate": 6.239532885987793e-06, + "loss": 0.8745, + "step": 19826 + }, + { + "epoch": 0.63, + "learning_rate": 6.238573129711627e-06, + "loss": 0.936, + "step": 19827 + }, + { + "epoch": 0.63, + "learning_rate": 6.237613413791408e-06, + "loss": 1.002, + "step": 19828 + }, + { + "epoch": 0.63, + "learning_rate": 6.236653738237434e-06, + "loss": 0.9253, + "step": 19829 + }, + { + "epoch": 0.63, + "learning_rate": 6.2356941030600036e-06, + "loss": 0.937, + "step": 19830 + }, + { + "epoch": 0.63, + "learning_rate": 6.2347345082694065e-06, + "loss": 0.8389, + "step": 19831 + }, + { + "epoch": 0.63, + "learning_rate": 6.233774953875939e-06, + "loss": 1.0083, + "step": 19832 + }, + { + "epoch": 0.63, + "learning_rate": 6.232815439889901e-06, + "loss": 0.8525, + "step": 19833 + }, + { + "epoch": 0.63, + "learning_rate": 6.231855966321581e-06, + "loss": 0.8452, + "step": 19834 + }, + { + "epoch": 0.63, + "learning_rate": 6.230896533181278e-06, + "loss": 1.0425, + "step": 19835 + }, + { + "epoch": 0.63, + "learning_rate": 6.229937140479283e-06, + "loss": 0.9507, + "step": 19836 + }, + { + "epoch": 0.63, + "learning_rate": 6.228977788225894e-06, + "loss": 0.9077, + "step": 19837 + }, + { + "epoch": 0.63, + "learning_rate": 6.2280184764313964e-06, + "loss": 1.0029, + "step": 19838 + }, + { + "epoch": 0.63, + "learning_rate": 6.227059205106085e-06, + "loss": 0.9277, + "step": 19839 + }, + { + "epoch": 0.63, + "learning_rate": 6.226099974260253e-06, + "loss": 0.8965, + "step": 19840 + }, + { + "epoch": 0.63, + "learning_rate": 6.225140783904191e-06, + "loss": 0.8979, + "step": 19841 + }, + { + "epoch": 0.63, + "learning_rate": 6.224181634048192e-06, + "loss": 0.8779, + "step": 19842 + }, + { + "epoch": 0.63, + "learning_rate": 6.2232225247025435e-06, + "loss": 0.8716, + "step": 19843 + }, + { + "epoch": 0.63, + "learning_rate": 6.2222634558775395e-06, + "loss": 1.0044, + "step": 19844 + }, + { + "epoch": 0.63, + "learning_rate": 6.22130442758347e-06, + "loss": 0.937, + "step": 19845 + }, + { + "epoch": 0.63, + "learning_rate": 6.220345439830618e-06, + "loss": 0.9614, + "step": 19846 + }, + { + "epoch": 0.63, + "learning_rate": 6.219386492629277e-06, + "loss": 0.4978, + "step": 19847 + }, + { + "epoch": 0.63, + "learning_rate": 6.218427585989734e-06, + "loss": 0.9204, + "step": 19848 + }, + { + "epoch": 0.63, + "learning_rate": 6.217468719922278e-06, + "loss": 0.8467, + "step": 19849 + }, + { + "epoch": 0.63, + "learning_rate": 6.216509894437197e-06, + "loss": 0.854, + "step": 19850 + }, + { + "epoch": 0.63, + "learning_rate": 6.2155511095447775e-06, + "loss": 0.9536, + "step": 19851 + }, + { + "epoch": 0.63, + "learning_rate": 6.21459236525531e-06, + "loss": 1.0176, + "step": 19852 + }, + { + "epoch": 0.63, + "learning_rate": 6.213633661579072e-06, + "loss": 0.9219, + "step": 19853 + }, + { + "epoch": 0.63, + "learning_rate": 6.212674998526355e-06, + "loss": 0.8828, + "step": 19854 + }, + { + "epoch": 0.63, + "learning_rate": 6.211716376107443e-06, + "loss": 0.7925, + "step": 19855 + }, + { + "epoch": 0.64, + "learning_rate": 6.210757794332622e-06, + "loss": 0.9404, + "step": 19856 + }, + { + "epoch": 0.64, + "learning_rate": 6.209799253212177e-06, + "loss": 0.894, + "step": 19857 + }, + { + "epoch": 0.64, + "learning_rate": 6.208840752756391e-06, + "loss": 1.0278, + "step": 19858 + }, + { + "epoch": 0.64, + "learning_rate": 6.207882292975549e-06, + "loss": 1.0518, + "step": 19859 + }, + { + "epoch": 0.64, + "learning_rate": 6.206923873879933e-06, + "loss": 0.9736, + "step": 19860 + }, + { + "epoch": 0.64, + "learning_rate": 6.205965495479824e-06, + "loss": 0.8853, + "step": 19861 + }, + { + "epoch": 0.64, + "learning_rate": 6.2050071577855065e-06, + "loss": 1.0088, + "step": 19862 + }, + { + "epoch": 0.64, + "learning_rate": 6.204048860807261e-06, + "loss": 1.041, + "step": 19863 + }, + { + "epoch": 0.64, + "learning_rate": 6.203090604555375e-06, + "loss": 0.9502, + "step": 19864 + }, + { + "epoch": 0.64, + "learning_rate": 6.2021323890401205e-06, + "loss": 0.9512, + "step": 19865 + }, + { + "epoch": 0.64, + "learning_rate": 6.201174214271782e-06, + "loss": 0.9863, + "step": 19866 + }, + { + "epoch": 0.64, + "learning_rate": 6.2002160802606425e-06, + "loss": 1.082, + "step": 19867 + }, + { + "epoch": 0.64, + "learning_rate": 6.1992579870169765e-06, + "loss": 0.979, + "step": 19868 + }, + { + "epoch": 0.64, + "learning_rate": 6.198299934551068e-06, + "loss": 1.0703, + "step": 19869 + }, + { + "epoch": 0.64, + "learning_rate": 6.197341922873192e-06, + "loss": 1.0283, + "step": 19870 + }, + { + "epoch": 0.64, + "learning_rate": 6.196383951993629e-06, + "loss": 0.9053, + "step": 19871 + }, + { + "epoch": 0.64, + "learning_rate": 6.195426021922657e-06, + "loss": 0.5134, + "step": 19872 + }, + { + "epoch": 0.64, + "learning_rate": 6.194468132670552e-06, + "loss": 0.9624, + "step": 19873 + }, + { + "epoch": 0.64, + "learning_rate": 6.1935102842475925e-06, + "loss": 0.9458, + "step": 19874 + }, + { + "epoch": 0.64, + "learning_rate": 6.1925524766640575e-06, + "loss": 0.9507, + "step": 19875 + }, + { + "epoch": 0.64, + "learning_rate": 6.191594709930218e-06, + "loss": 1.1572, + "step": 19876 + }, + { + "epoch": 0.64, + "learning_rate": 6.190636984056353e-06, + "loss": 0.894, + "step": 19877 + }, + { + "epoch": 0.64, + "learning_rate": 6.189679299052735e-06, + "loss": 0.8872, + "step": 19878 + }, + { + "epoch": 0.64, + "learning_rate": 6.188721654929644e-06, + "loss": 0.8926, + "step": 19879 + }, + { + "epoch": 0.64, + "learning_rate": 6.187764051697349e-06, + "loss": 1.0088, + "step": 19880 + }, + { + "epoch": 0.64, + "learning_rate": 6.186806489366127e-06, + "loss": 0.9961, + "step": 19881 + }, + { + "epoch": 0.64, + "learning_rate": 6.185848967946252e-06, + "loss": 0.9839, + "step": 19882 + }, + { + "epoch": 0.64, + "learning_rate": 6.184891487447999e-06, + "loss": 0.8799, + "step": 19883 + }, + { + "epoch": 0.64, + "learning_rate": 6.183934047881636e-06, + "loss": 0.8809, + "step": 19884 + }, + { + "epoch": 0.64, + "learning_rate": 6.182976649257435e-06, + "loss": 1.0635, + "step": 19885 + }, + { + "epoch": 0.64, + "learning_rate": 6.1820192915856725e-06, + "loss": 0.877, + "step": 19886 + }, + { + "epoch": 0.64, + "learning_rate": 6.181061974876615e-06, + "loss": 1.0273, + "step": 19887 + }, + { + "epoch": 0.64, + "learning_rate": 6.180104699140537e-06, + "loss": 1.0, + "step": 19888 + }, + { + "epoch": 0.64, + "learning_rate": 6.179147464387709e-06, + "loss": 0.8877, + "step": 19889 + }, + { + "epoch": 0.64, + "learning_rate": 6.178190270628403e-06, + "loss": 0.937, + "step": 19890 + }, + { + "epoch": 0.64, + "learning_rate": 6.177233117872881e-06, + "loss": 0.916, + "step": 19891 + }, + { + "epoch": 0.64, + "learning_rate": 6.176276006131417e-06, + "loss": 1.0332, + "step": 19892 + }, + { + "epoch": 0.64, + "learning_rate": 6.175318935414281e-06, + "loss": 0.8589, + "step": 19893 + }, + { + "epoch": 0.64, + "learning_rate": 6.174361905731738e-06, + "loss": 1.0396, + "step": 19894 + }, + { + "epoch": 0.64, + "learning_rate": 6.173404917094059e-06, + "loss": 0.9585, + "step": 19895 + }, + { + "epoch": 0.64, + "learning_rate": 6.17244796951151e-06, + "loss": 0.8691, + "step": 19896 + }, + { + "epoch": 0.64, + "learning_rate": 6.171491062994357e-06, + "loss": 0.9868, + "step": 19897 + }, + { + "epoch": 0.64, + "learning_rate": 6.1705341975528735e-06, + "loss": 0.9351, + "step": 19898 + }, + { + "epoch": 0.64, + "learning_rate": 6.169577373197315e-06, + "loss": 0.9751, + "step": 19899 + }, + { + "epoch": 0.64, + "learning_rate": 6.1686205899379515e-06, + "loss": 0.8311, + "step": 19900 + }, + { + "epoch": 0.64, + "learning_rate": 6.167663847785049e-06, + "loss": 0.8843, + "step": 19901 + }, + { + "epoch": 0.64, + "learning_rate": 6.166707146748871e-06, + "loss": 0.9048, + "step": 19902 + }, + { + "epoch": 0.64, + "learning_rate": 6.165750486839688e-06, + "loss": 1.0127, + "step": 19903 + }, + { + "epoch": 0.64, + "learning_rate": 6.164793868067753e-06, + "loss": 1.0, + "step": 19904 + }, + { + "epoch": 0.64, + "learning_rate": 6.163837290443339e-06, + "loss": 0.7944, + "step": 19905 + }, + { + "epoch": 0.64, + "learning_rate": 6.162880753976704e-06, + "loss": 0.8398, + "step": 19906 + }, + { + "epoch": 0.64, + "learning_rate": 6.161924258678112e-06, + "loss": 0.915, + "step": 19907 + }, + { + "epoch": 0.64, + "learning_rate": 6.160967804557825e-06, + "loss": 0.9292, + "step": 19908 + }, + { + "epoch": 0.64, + "learning_rate": 6.160011391626104e-06, + "loss": 0.8904, + "step": 19909 + }, + { + "epoch": 0.64, + "learning_rate": 6.159055019893209e-06, + "loss": 0.9678, + "step": 19910 + }, + { + "epoch": 0.64, + "learning_rate": 6.158098689369405e-06, + "loss": 1.1519, + "step": 19911 + }, + { + "epoch": 0.64, + "learning_rate": 6.157142400064948e-06, + "loss": 1.0054, + "step": 19912 + }, + { + "epoch": 0.64, + "learning_rate": 6.156186151990104e-06, + "loss": 0.9155, + "step": 19913 + }, + { + "epoch": 0.64, + "learning_rate": 6.155229945155128e-06, + "loss": 0.9185, + "step": 19914 + }, + { + "epoch": 0.64, + "learning_rate": 6.154273779570276e-06, + "loss": 0.9751, + "step": 19915 + }, + { + "epoch": 0.64, + "learning_rate": 6.153317655245812e-06, + "loss": 0.9414, + "step": 19916 + }, + { + "epoch": 0.64, + "learning_rate": 6.1523615721919916e-06, + "loss": 0.9756, + "step": 19917 + }, + { + "epoch": 0.64, + "learning_rate": 6.1514055304190725e-06, + "loss": 1.0166, + "step": 19918 + }, + { + "epoch": 0.64, + "learning_rate": 6.150449529937313e-06, + "loss": 0.5454, + "step": 19919 + }, + { + "epoch": 0.64, + "learning_rate": 6.149493570756975e-06, + "loss": 0.8857, + "step": 19920 + }, + { + "epoch": 0.64, + "learning_rate": 6.148537652888303e-06, + "loss": 0.9521, + "step": 19921 + }, + { + "epoch": 0.64, + "learning_rate": 6.147581776341561e-06, + "loss": 0.9692, + "step": 19922 + }, + { + "epoch": 0.64, + "learning_rate": 6.146625941127002e-06, + "loss": 0.9756, + "step": 19923 + }, + { + "epoch": 0.64, + "learning_rate": 6.145670147254883e-06, + "loss": 0.896, + "step": 19924 + }, + { + "epoch": 0.64, + "learning_rate": 6.144714394735457e-06, + "loss": 0.9839, + "step": 19925 + }, + { + "epoch": 0.64, + "learning_rate": 6.143758683578979e-06, + "loss": 0.9585, + "step": 19926 + }, + { + "epoch": 0.64, + "learning_rate": 6.142803013795702e-06, + "loss": 0.9614, + "step": 19927 + }, + { + "epoch": 0.64, + "learning_rate": 6.141847385395883e-06, + "loss": 1.043, + "step": 19928 + }, + { + "epoch": 0.64, + "learning_rate": 6.14089179838977e-06, + "loss": 0.8794, + "step": 19929 + }, + { + "epoch": 0.64, + "learning_rate": 6.139936252787615e-06, + "loss": 0.9492, + "step": 19930 + }, + { + "epoch": 0.64, + "learning_rate": 6.138980748599672e-06, + "loss": 0.792, + "step": 19931 + }, + { + "epoch": 0.64, + "learning_rate": 6.138025285836193e-06, + "loss": 0.8784, + "step": 19932 + }, + { + "epoch": 0.64, + "learning_rate": 6.137069864507428e-06, + "loss": 0.9492, + "step": 19933 + }, + { + "epoch": 0.64, + "learning_rate": 6.136114484623627e-06, + "loss": 0.936, + "step": 19934 + }, + { + "epoch": 0.64, + "learning_rate": 6.135159146195048e-06, + "loss": 0.8994, + "step": 19935 + }, + { + "epoch": 0.64, + "learning_rate": 6.134203849231928e-06, + "loss": 0.8662, + "step": 19936 + }, + { + "epoch": 0.64, + "learning_rate": 6.133248593744523e-06, + "loss": 0.9717, + "step": 19937 + }, + { + "epoch": 0.64, + "learning_rate": 6.13229337974308e-06, + "loss": 0.9634, + "step": 19938 + }, + { + "epoch": 0.64, + "learning_rate": 6.131338207237849e-06, + "loss": 0.9136, + "step": 19939 + }, + { + "epoch": 0.64, + "learning_rate": 6.130383076239079e-06, + "loss": 0.896, + "step": 19940 + }, + { + "epoch": 0.64, + "learning_rate": 6.1294279867570135e-06, + "loss": 0.9272, + "step": 19941 + }, + { + "epoch": 0.64, + "learning_rate": 6.128472938801906e-06, + "loss": 0.9048, + "step": 19942 + }, + { + "epoch": 0.64, + "learning_rate": 6.127517932383997e-06, + "loss": 0.9448, + "step": 19943 + }, + { + "epoch": 0.64, + "learning_rate": 6.126562967513534e-06, + "loss": 0.874, + "step": 19944 + }, + { + "epoch": 0.64, + "learning_rate": 6.125608044200764e-06, + "loss": 0.9126, + "step": 19945 + }, + { + "epoch": 0.64, + "learning_rate": 6.124653162455932e-06, + "loss": 0.9609, + "step": 19946 + }, + { + "epoch": 0.64, + "learning_rate": 6.123698322289282e-06, + "loss": 0.9795, + "step": 19947 + }, + { + "epoch": 0.64, + "learning_rate": 6.122743523711059e-06, + "loss": 0.9116, + "step": 19948 + }, + { + "epoch": 0.64, + "learning_rate": 6.121788766731507e-06, + "loss": 0.9414, + "step": 19949 + }, + { + "epoch": 0.64, + "learning_rate": 6.12083405136087e-06, + "loss": 0.8252, + "step": 19950 + }, + { + "epoch": 0.64, + "learning_rate": 6.119879377609389e-06, + "loss": 1.0049, + "step": 19951 + }, + { + "epoch": 0.64, + "learning_rate": 6.118924745487309e-06, + "loss": 0.959, + "step": 19952 + }, + { + "epoch": 0.64, + "learning_rate": 6.117970155004871e-06, + "loss": 1.0093, + "step": 19953 + }, + { + "epoch": 0.64, + "learning_rate": 6.117015606172316e-06, + "loss": 0.8013, + "step": 19954 + }, + { + "epoch": 0.64, + "learning_rate": 6.116061098999886e-06, + "loss": 0.9072, + "step": 19955 + }, + { + "epoch": 0.64, + "learning_rate": 6.115106633497822e-06, + "loss": 0.9863, + "step": 19956 + }, + { + "epoch": 0.64, + "learning_rate": 6.114152209676365e-06, + "loss": 0.9653, + "step": 19957 + }, + { + "epoch": 0.64, + "learning_rate": 6.113197827545757e-06, + "loss": 0.8403, + "step": 19958 + }, + { + "epoch": 0.64, + "learning_rate": 6.112243487116232e-06, + "loss": 0.9424, + "step": 19959 + }, + { + "epoch": 0.64, + "learning_rate": 6.11128918839803e-06, + "loss": 0.9883, + "step": 19960 + }, + { + "epoch": 0.64, + "learning_rate": 6.1103349314013915e-06, + "loss": 0.8872, + "step": 19961 + }, + { + "epoch": 0.64, + "learning_rate": 6.109380716136555e-06, + "loss": 1.0029, + "step": 19962 + }, + { + "epoch": 0.64, + "learning_rate": 6.1084265426137564e-06, + "loss": 0.981, + "step": 19963 + }, + { + "epoch": 0.64, + "learning_rate": 6.107472410843235e-06, + "loss": 0.9683, + "step": 19964 + }, + { + "epoch": 0.64, + "learning_rate": 6.1065183208352264e-06, + "loss": 0.9106, + "step": 19965 + }, + { + "epoch": 0.64, + "learning_rate": 6.105564272599971e-06, + "loss": 0.979, + "step": 19966 + }, + { + "epoch": 0.64, + "learning_rate": 6.104610266147697e-06, + "loss": 0.978, + "step": 19967 + }, + { + "epoch": 0.64, + "learning_rate": 6.103656301488643e-06, + "loss": 0.9858, + "step": 19968 + }, + { + "epoch": 0.64, + "learning_rate": 6.102702378633045e-06, + "loss": 1.0557, + "step": 19969 + }, + { + "epoch": 0.64, + "learning_rate": 6.101748497591138e-06, + "loss": 1.0513, + "step": 19970 + }, + { + "epoch": 0.64, + "learning_rate": 6.100794658373154e-06, + "loss": 0.9395, + "step": 19971 + }, + { + "epoch": 0.64, + "learning_rate": 6.099840860989328e-06, + "loss": 0.8867, + "step": 19972 + }, + { + "epoch": 0.64, + "learning_rate": 6.098887105449898e-06, + "loss": 0.9463, + "step": 19973 + }, + { + "epoch": 0.64, + "learning_rate": 6.097933391765087e-06, + "loss": 1.0083, + "step": 19974 + }, + { + "epoch": 0.64, + "learning_rate": 6.096979719945132e-06, + "loss": 0.9331, + "step": 19975 + }, + { + "epoch": 0.64, + "learning_rate": 6.096026090000265e-06, + "loss": 0.8804, + "step": 19976 + }, + { + "epoch": 0.64, + "learning_rate": 6.095072501940718e-06, + "loss": 0.8916, + "step": 19977 + }, + { + "epoch": 0.64, + "learning_rate": 6.094118955776721e-06, + "loss": 0.9346, + "step": 19978 + }, + { + "epoch": 0.64, + "learning_rate": 6.093165451518506e-06, + "loss": 0.9746, + "step": 19979 + }, + { + "epoch": 0.64, + "learning_rate": 6.0922119891763e-06, + "loss": 1.0366, + "step": 19980 + }, + { + "epoch": 0.64, + "learning_rate": 6.091258568760338e-06, + "loss": 1.0049, + "step": 19981 + }, + { + "epoch": 0.64, + "learning_rate": 6.090305190280843e-06, + "loss": 1.0254, + "step": 19982 + }, + { + "epoch": 0.64, + "learning_rate": 6.089351853748046e-06, + "loss": 0.9614, + "step": 19983 + }, + { + "epoch": 0.64, + "learning_rate": 6.088398559172175e-06, + "loss": 1.0454, + "step": 19984 + }, + { + "epoch": 0.64, + "learning_rate": 6.08744530656346e-06, + "loss": 0.7769, + "step": 19985 + }, + { + "epoch": 0.64, + "learning_rate": 6.086492095932126e-06, + "loss": 1.0234, + "step": 19986 + }, + { + "epoch": 0.64, + "learning_rate": 6.085538927288401e-06, + "loss": 0.9893, + "step": 19987 + }, + { + "epoch": 0.64, + "learning_rate": 6.084585800642512e-06, + "loss": 0.7693, + "step": 19988 + }, + { + "epoch": 0.64, + "learning_rate": 6.083632716004682e-06, + "loss": 1.0293, + "step": 19989 + }, + { + "epoch": 0.64, + "learning_rate": 6.082679673385138e-06, + "loss": 1.083, + "step": 19990 + }, + { + "epoch": 0.64, + "learning_rate": 6.081726672794106e-06, + "loss": 0.9502, + "step": 19991 + }, + { + "epoch": 0.64, + "learning_rate": 6.080773714241812e-06, + "loss": 0.9297, + "step": 19992 + }, + { + "epoch": 0.64, + "learning_rate": 6.079820797738477e-06, + "loss": 1.0215, + "step": 19993 + }, + { + "epoch": 0.64, + "learning_rate": 6.078867923294327e-06, + "loss": 0.9106, + "step": 19994 + }, + { + "epoch": 0.64, + "learning_rate": 6.077915090919582e-06, + "loss": 0.9771, + "step": 19995 + }, + { + "epoch": 0.64, + "learning_rate": 6.076962300624471e-06, + "loss": 0.9077, + "step": 19996 + }, + { + "epoch": 0.64, + "learning_rate": 6.076009552419213e-06, + "loss": 1.0312, + "step": 19997 + }, + { + "epoch": 0.64, + "learning_rate": 6.075056846314027e-06, + "loss": 0.5193, + "step": 19998 + }, + { + "epoch": 0.64, + "learning_rate": 6.0741041823191384e-06, + "loss": 0.9927, + "step": 19999 + }, + { + "epoch": 0.64, + "learning_rate": 6.073151560444765e-06, + "loss": 0.9204, + "step": 20000 + }, + { + "epoch": 0.64, + "learning_rate": 6.072198980701131e-06, + "loss": 0.9512, + "step": 20001 + }, + { + "epoch": 0.64, + "learning_rate": 6.071246443098453e-06, + "loss": 0.9907, + "step": 20002 + }, + { + "epoch": 0.64, + "learning_rate": 6.070293947646959e-06, + "loss": 0.9136, + "step": 20003 + }, + { + "epoch": 0.64, + "learning_rate": 6.0693414943568554e-06, + "loss": 0.9517, + "step": 20004 + }, + { + "epoch": 0.64, + "learning_rate": 6.068389083238368e-06, + "loss": 0.8457, + "step": 20005 + }, + { + "epoch": 0.64, + "learning_rate": 6.067436714301715e-06, + "loss": 0.8779, + "step": 20006 + }, + { + "epoch": 0.64, + "learning_rate": 6.066484387557114e-06, + "loss": 0.873, + "step": 20007 + }, + { + "epoch": 0.64, + "learning_rate": 6.0655321030147814e-06, + "loss": 1.1035, + "step": 20008 + }, + { + "epoch": 0.64, + "learning_rate": 6.064579860684935e-06, + "loss": 0.9731, + "step": 20009 + }, + { + "epoch": 0.64, + "learning_rate": 6.0636276605777915e-06, + "loss": 1.0435, + "step": 20010 + }, + { + "epoch": 0.64, + "learning_rate": 6.06267550270357e-06, + "loss": 0.9209, + "step": 20011 + }, + { + "epoch": 0.64, + "learning_rate": 6.06172338707248e-06, + "loss": 1.0381, + "step": 20012 + }, + { + "epoch": 0.64, + "learning_rate": 6.060771313694737e-06, + "loss": 0.8931, + "step": 20013 + }, + { + "epoch": 0.64, + "learning_rate": 6.059819282580561e-06, + "loss": 0.9187, + "step": 20014 + }, + { + "epoch": 0.64, + "learning_rate": 6.0588672937401615e-06, + "loss": 0.9868, + "step": 20015 + }, + { + "epoch": 0.64, + "learning_rate": 6.0579153471837536e-06, + "loss": 1.0142, + "step": 20016 + }, + { + "epoch": 0.64, + "learning_rate": 6.056963442921553e-06, + "loss": 0.9922, + "step": 20017 + }, + { + "epoch": 0.64, + "learning_rate": 6.056011580963775e-06, + "loss": 1.0693, + "step": 20018 + }, + { + "epoch": 0.64, + "learning_rate": 6.0550597613206205e-06, + "loss": 0.9521, + "step": 20019 + }, + { + "epoch": 0.64, + "learning_rate": 6.054107984002311e-06, + "loss": 1.0015, + "step": 20020 + }, + { + "epoch": 0.64, + "learning_rate": 6.0531562490190565e-06, + "loss": 0.9395, + "step": 20021 + }, + { + "epoch": 0.64, + "learning_rate": 6.052204556381066e-06, + "loss": 0.9849, + "step": 20022 + }, + { + "epoch": 0.64, + "learning_rate": 6.0512529060985515e-06, + "loss": 1.0542, + "step": 20023 + }, + { + "epoch": 0.64, + "learning_rate": 6.050301298181723e-06, + "loss": 0.9375, + "step": 20024 + }, + { + "epoch": 0.64, + "learning_rate": 6.04934973264079e-06, + "loss": 0.9565, + "step": 20025 + }, + { + "epoch": 0.64, + "learning_rate": 6.0483982094859675e-06, + "loss": 1.0391, + "step": 20026 + }, + { + "epoch": 0.64, + "learning_rate": 6.047446728727453e-06, + "loss": 1.043, + "step": 20027 + }, + { + "epoch": 0.64, + "learning_rate": 6.046495290375463e-06, + "loss": 0.939, + "step": 20028 + }, + { + "epoch": 0.64, + "learning_rate": 6.0455438944402e-06, + "loss": 1.0537, + "step": 20029 + }, + { + "epoch": 0.64, + "learning_rate": 6.044592540931877e-06, + "loss": 0.8911, + "step": 20030 + }, + { + "epoch": 0.64, + "learning_rate": 6.043641229860699e-06, + "loss": 1.0933, + "step": 20031 + }, + { + "epoch": 0.64, + "learning_rate": 6.0426899612368695e-06, + "loss": 0.9146, + "step": 20032 + }, + { + "epoch": 0.64, + "learning_rate": 6.041738735070599e-06, + "loss": 1.0557, + "step": 20033 + }, + { + "epoch": 0.64, + "learning_rate": 6.0407875513720915e-06, + "loss": 0.7065, + "step": 20034 + }, + { + "epoch": 0.64, + "learning_rate": 6.039836410151549e-06, + "loss": 0.8999, + "step": 20035 + }, + { + "epoch": 0.64, + "learning_rate": 6.038885311419183e-06, + "loss": 0.8203, + "step": 20036 + }, + { + "epoch": 0.64, + "learning_rate": 6.037934255185191e-06, + "loss": 1.0674, + "step": 20037 + }, + { + "epoch": 0.64, + "learning_rate": 6.0369832414597784e-06, + "loss": 0.9805, + "step": 20038 + }, + { + "epoch": 0.64, + "learning_rate": 6.03603227025315e-06, + "loss": 0.9917, + "step": 20039 + }, + { + "epoch": 0.64, + "learning_rate": 6.035081341575508e-06, + "loss": 1.0127, + "step": 20040 + }, + { + "epoch": 0.64, + "learning_rate": 6.034130455437058e-06, + "loss": 0.9463, + "step": 20041 + }, + { + "epoch": 0.64, + "learning_rate": 6.033179611847999e-06, + "loss": 0.9668, + "step": 20042 + }, + { + "epoch": 0.64, + "learning_rate": 6.03222881081853e-06, + "loss": 1.0532, + "step": 20043 + }, + { + "epoch": 0.64, + "learning_rate": 6.031278052358855e-06, + "loss": 0.8989, + "step": 20044 + }, + { + "epoch": 0.64, + "learning_rate": 6.030327336479174e-06, + "loss": 1.0488, + "step": 20045 + }, + { + "epoch": 0.64, + "learning_rate": 6.029376663189685e-06, + "loss": 0.9307, + "step": 20046 + }, + { + "epoch": 0.64, + "learning_rate": 6.028426032500594e-06, + "loss": 0.9048, + "step": 20047 + }, + { + "epoch": 0.64, + "learning_rate": 6.027475444422093e-06, + "loss": 0.885, + "step": 20048 + }, + { + "epoch": 0.64, + "learning_rate": 6.0265248989643895e-06, + "loss": 0.981, + "step": 20049 + }, + { + "epoch": 0.64, + "learning_rate": 6.025574396137672e-06, + "loss": 0.9849, + "step": 20050 + }, + { + "epoch": 0.64, + "learning_rate": 6.024623935952142e-06, + "loss": 1.062, + "step": 20051 + }, + { + "epoch": 0.64, + "learning_rate": 6.0236735184179975e-06, + "loss": 0.896, + "step": 20052 + }, + { + "epoch": 0.64, + "learning_rate": 6.022723143545435e-06, + "loss": 0.8442, + "step": 20053 + }, + { + "epoch": 0.64, + "learning_rate": 6.021772811344653e-06, + "loss": 0.9497, + "step": 20054 + }, + { + "epoch": 0.64, + "learning_rate": 6.020822521825844e-06, + "loss": 1.0146, + "step": 20055 + }, + { + "epoch": 0.64, + "learning_rate": 6.01987227499921e-06, + "loss": 1.0195, + "step": 20056 + }, + { + "epoch": 0.64, + "learning_rate": 6.0189220708749375e-06, + "loss": 0.8545, + "step": 20057 + }, + { + "epoch": 0.64, + "learning_rate": 6.017971909463225e-06, + "loss": 0.8735, + "step": 20058 + }, + { + "epoch": 0.64, + "learning_rate": 6.0170217907742675e-06, + "loss": 0.9683, + "step": 20059 + }, + { + "epoch": 0.64, + "learning_rate": 6.016071714818258e-06, + "loss": 0.854, + "step": 20060 + }, + { + "epoch": 0.64, + "learning_rate": 6.015121681605389e-06, + "loss": 0.9468, + "step": 20061 + }, + { + "epoch": 0.64, + "learning_rate": 6.014171691145855e-06, + "loss": 0.9375, + "step": 20062 + }, + { + "epoch": 0.64, + "learning_rate": 6.0132217434498466e-06, + "loss": 0.9956, + "step": 20063 + }, + { + "epoch": 0.64, + "learning_rate": 6.0122718385275615e-06, + "loss": 1.0889, + "step": 20064 + }, + { + "epoch": 0.64, + "learning_rate": 6.011321976389182e-06, + "loss": 0.918, + "step": 20065 + }, + { + "epoch": 0.64, + "learning_rate": 6.010372157044903e-06, + "loss": 0.9897, + "step": 20066 + }, + { + "epoch": 0.64, + "learning_rate": 6.0094223805049145e-06, + "loss": 0.9399, + "step": 20067 + }, + { + "epoch": 0.64, + "learning_rate": 6.008472646779409e-06, + "loss": 0.9834, + "step": 20068 + }, + { + "epoch": 0.64, + "learning_rate": 6.007522955878573e-06, + "loss": 0.9292, + "step": 20069 + }, + { + "epoch": 0.64, + "learning_rate": 6.006573307812599e-06, + "loss": 0.9692, + "step": 20070 + }, + { + "epoch": 0.64, + "learning_rate": 6.005623702591674e-06, + "loss": 0.9736, + "step": 20071 + }, + { + "epoch": 0.64, + "learning_rate": 6.004674140225983e-06, + "loss": 0.8413, + "step": 20072 + }, + { + "epoch": 0.64, + "learning_rate": 6.00372462072572e-06, + "loss": 1.0073, + "step": 20073 + }, + { + "epoch": 0.64, + "learning_rate": 6.002775144101065e-06, + "loss": 0.9824, + "step": 20074 + }, + { + "epoch": 0.64, + "learning_rate": 6.001825710362213e-06, + "loss": 0.9771, + "step": 20075 + }, + { + "epoch": 0.64, + "learning_rate": 6.000876319519343e-06, + "loss": 0.9678, + "step": 20076 + }, + { + "epoch": 0.64, + "learning_rate": 5.999926971582645e-06, + "loss": 0.9468, + "step": 20077 + }, + { + "epoch": 0.64, + "learning_rate": 5.9989776665623026e-06, + "loss": 0.9229, + "step": 20078 + }, + { + "epoch": 0.64, + "learning_rate": 5.998028404468506e-06, + "loss": 0.9897, + "step": 20079 + }, + { + "epoch": 0.64, + "learning_rate": 5.997079185311431e-06, + "loss": 1.0269, + "step": 20080 + }, + { + "epoch": 0.64, + "learning_rate": 5.996130009101269e-06, + "loss": 0.9893, + "step": 20081 + }, + { + "epoch": 0.64, + "learning_rate": 5.9951808758481984e-06, + "loss": 0.9409, + "step": 20082 + }, + { + "epoch": 0.64, + "learning_rate": 5.994231785562406e-06, + "loss": 0.9639, + "step": 20083 + }, + { + "epoch": 0.64, + "learning_rate": 5.9932827382540715e-06, + "loss": 0.8381, + "step": 20084 + }, + { + "epoch": 0.64, + "learning_rate": 5.992333733933378e-06, + "loss": 0.8608, + "step": 20085 + }, + { + "epoch": 0.64, + "learning_rate": 5.991384772610514e-06, + "loss": 0.9707, + "step": 20086 + }, + { + "epoch": 0.64, + "learning_rate": 5.990435854295651e-06, + "loss": 1.0376, + "step": 20087 + }, + { + "epoch": 0.64, + "learning_rate": 5.989486978998973e-06, + "loss": 0.9712, + "step": 20088 + }, + { + "epoch": 0.64, + "learning_rate": 5.98853814673066e-06, + "loss": 1.0498, + "step": 20089 + }, + { + "epoch": 0.64, + "learning_rate": 5.987589357500896e-06, + "loss": 0.5366, + "step": 20090 + }, + { + "epoch": 0.64, + "learning_rate": 5.986640611319855e-06, + "loss": 0.9883, + "step": 20091 + }, + { + "epoch": 0.64, + "learning_rate": 5.9856919081977195e-06, + "loss": 0.9331, + "step": 20092 + }, + { + "epoch": 0.64, + "learning_rate": 5.984743248144666e-06, + "loss": 0.8838, + "step": 20093 + }, + { + "epoch": 0.64, + "learning_rate": 5.983794631170879e-06, + "loss": 1.0625, + "step": 20094 + }, + { + "epoch": 0.64, + "learning_rate": 5.982846057286525e-06, + "loss": 0.8901, + "step": 20095 + }, + { + "epoch": 0.64, + "learning_rate": 5.981897526501787e-06, + "loss": 0.9849, + "step": 20096 + }, + { + "epoch": 0.64, + "learning_rate": 5.980949038826843e-06, + "loss": 0.8293, + "step": 20097 + }, + { + "epoch": 0.64, + "learning_rate": 5.980000594271868e-06, + "loss": 0.9736, + "step": 20098 + }, + { + "epoch": 0.64, + "learning_rate": 5.979052192847036e-06, + "loss": 0.9971, + "step": 20099 + }, + { + "epoch": 0.64, + "learning_rate": 5.9781038345625245e-06, + "loss": 0.8062, + "step": 20100 + }, + { + "epoch": 0.64, + "learning_rate": 5.977155519428512e-06, + "loss": 0.998, + "step": 20101 + }, + { + "epoch": 0.64, + "learning_rate": 5.9762072474551635e-06, + "loss": 1.0205, + "step": 20102 + }, + { + "epoch": 0.64, + "learning_rate": 5.975259018652658e-06, + "loss": 1.0312, + "step": 20103 + }, + { + "epoch": 0.64, + "learning_rate": 5.9743108330311695e-06, + "loss": 0.8354, + "step": 20104 + }, + { + "epoch": 0.64, + "learning_rate": 5.97336269060087e-06, + "loss": 0.5134, + "step": 20105 + }, + { + "epoch": 0.64, + "learning_rate": 5.972414591371934e-06, + "loss": 0.8638, + "step": 20106 + }, + { + "epoch": 0.64, + "learning_rate": 5.971466535354529e-06, + "loss": 0.9297, + "step": 20107 + }, + { + "epoch": 0.64, + "learning_rate": 5.970518522558832e-06, + "loss": 0.8066, + "step": 20108 + }, + { + "epoch": 0.64, + "learning_rate": 5.969570552995014e-06, + "loss": 1.0386, + "step": 20109 + }, + { + "epoch": 0.64, + "learning_rate": 5.968622626673239e-06, + "loss": 0.9097, + "step": 20110 + }, + { + "epoch": 0.64, + "learning_rate": 5.967674743603683e-06, + "loss": 0.8608, + "step": 20111 + }, + { + "epoch": 0.64, + "learning_rate": 5.966726903796514e-06, + "loss": 0.9624, + "step": 20112 + }, + { + "epoch": 0.64, + "learning_rate": 5.965779107261901e-06, + "loss": 1.04, + "step": 20113 + }, + { + "epoch": 0.64, + "learning_rate": 5.964831354010014e-06, + "loss": 0.9604, + "step": 20114 + }, + { + "epoch": 0.64, + "learning_rate": 5.963883644051022e-06, + "loss": 0.9619, + "step": 20115 + }, + { + "epoch": 0.64, + "learning_rate": 5.96293597739509e-06, + "loss": 0.9409, + "step": 20116 + }, + { + "epoch": 0.64, + "learning_rate": 5.961988354052388e-06, + "loss": 0.8882, + "step": 20117 + }, + { + "epoch": 0.64, + "learning_rate": 5.96104077403308e-06, + "loss": 0.9644, + "step": 20118 + }, + { + "epoch": 0.64, + "learning_rate": 5.960093237347335e-06, + "loss": 1.0552, + "step": 20119 + }, + { + "epoch": 0.64, + "learning_rate": 5.959145744005321e-06, + "loss": 0.8535, + "step": 20120 + }, + { + "epoch": 0.64, + "learning_rate": 5.958198294017199e-06, + "loss": 0.9282, + "step": 20121 + }, + { + "epoch": 0.64, + "learning_rate": 5.957250887393136e-06, + "loss": 0.9116, + "step": 20122 + }, + { + "epoch": 0.64, + "learning_rate": 5.956303524143294e-06, + "loss": 0.9849, + "step": 20123 + }, + { + "epoch": 0.64, + "learning_rate": 5.955356204277844e-06, + "loss": 0.9194, + "step": 20124 + }, + { + "epoch": 0.64, + "learning_rate": 5.954408927806947e-06, + "loss": 1.0688, + "step": 20125 + }, + { + "epoch": 0.64, + "learning_rate": 5.953461694740761e-06, + "loss": 0.979, + "step": 20126 + }, + { + "epoch": 0.64, + "learning_rate": 5.952514505089453e-06, + "loss": 1.0469, + "step": 20127 + }, + { + "epoch": 0.64, + "learning_rate": 5.951567358863184e-06, + "loss": 0.9082, + "step": 20128 + }, + { + "epoch": 0.64, + "learning_rate": 5.950620256072117e-06, + "loss": 0.9521, + "step": 20129 + }, + { + "epoch": 0.64, + "learning_rate": 5.9496731967264135e-06, + "loss": 1.0156, + "step": 20130 + }, + { + "epoch": 0.64, + "learning_rate": 5.948726180836234e-06, + "loss": 0.8682, + "step": 20131 + }, + { + "epoch": 0.64, + "learning_rate": 5.947779208411743e-06, + "loss": 0.9351, + "step": 20132 + }, + { + "epoch": 0.64, + "learning_rate": 5.946832279463092e-06, + "loss": 0.9355, + "step": 20133 + }, + { + "epoch": 0.64, + "learning_rate": 5.945885394000443e-06, + "loss": 0.9082, + "step": 20134 + }, + { + "epoch": 0.64, + "learning_rate": 5.94493855203396e-06, + "loss": 0.9521, + "step": 20135 + }, + { + "epoch": 0.64, + "learning_rate": 5.9439917535737964e-06, + "loss": 0.9453, + "step": 20136 + }, + { + "epoch": 0.64, + "learning_rate": 5.943044998630113e-06, + "loss": 0.8071, + "step": 20137 + }, + { + "epoch": 0.64, + "learning_rate": 5.942098287213067e-06, + "loss": 0.7979, + "step": 20138 + }, + { + "epoch": 0.64, + "learning_rate": 5.941151619332819e-06, + "loss": 0.8042, + "step": 20139 + }, + { + "epoch": 0.64, + "learning_rate": 5.940204994999518e-06, + "loss": 0.8433, + "step": 20140 + }, + { + "epoch": 0.64, + "learning_rate": 5.939258414223324e-06, + "loss": 0.9805, + "step": 20141 + }, + { + "epoch": 0.64, + "learning_rate": 5.938311877014393e-06, + "loss": 1.0342, + "step": 20142 + }, + { + "epoch": 0.64, + "learning_rate": 5.9373653833828805e-06, + "loss": 1.0093, + "step": 20143 + }, + { + "epoch": 0.64, + "learning_rate": 5.936418933338941e-06, + "loss": 0.9419, + "step": 20144 + }, + { + "epoch": 0.64, + "learning_rate": 5.935472526892729e-06, + "loss": 0.9604, + "step": 20145 + }, + { + "epoch": 0.64, + "learning_rate": 5.934526164054399e-06, + "loss": 0.9722, + "step": 20146 + }, + { + "epoch": 0.64, + "learning_rate": 5.933579844834108e-06, + "loss": 0.9478, + "step": 20147 + }, + { + "epoch": 0.64, + "learning_rate": 5.932633569242e-06, + "loss": 0.9287, + "step": 20148 + }, + { + "epoch": 0.64, + "learning_rate": 5.931687337288232e-06, + "loss": 1.002, + "step": 20149 + }, + { + "epoch": 0.64, + "learning_rate": 5.930741148982957e-06, + "loss": 0.8179, + "step": 20150 + }, + { + "epoch": 0.64, + "learning_rate": 5.929795004336325e-06, + "loss": 0.978, + "step": 20151 + }, + { + "epoch": 0.64, + "learning_rate": 5.928848903358488e-06, + "loss": 1.0264, + "step": 20152 + }, + { + "epoch": 0.64, + "learning_rate": 5.927902846059598e-06, + "loss": 0.9385, + "step": 20153 + }, + { + "epoch": 0.64, + "learning_rate": 5.926956832449806e-06, + "loss": 0.9985, + "step": 20154 + }, + { + "epoch": 0.64, + "learning_rate": 5.926010862539256e-06, + "loss": 1.0688, + "step": 20155 + }, + { + "epoch": 0.64, + "learning_rate": 5.925064936338101e-06, + "loss": 1.0244, + "step": 20156 + }, + { + "epoch": 0.64, + "learning_rate": 5.924119053856489e-06, + "loss": 0.8877, + "step": 20157 + }, + { + "epoch": 0.64, + "learning_rate": 5.92317321510457e-06, + "loss": 0.957, + "step": 20158 + }, + { + "epoch": 0.64, + "learning_rate": 5.92222742009249e-06, + "loss": 0.8628, + "step": 20159 + }, + { + "epoch": 0.64, + "learning_rate": 5.921281668830395e-06, + "loss": 0.9316, + "step": 20160 + }, + { + "epoch": 0.64, + "learning_rate": 5.920335961328434e-06, + "loss": 0.979, + "step": 20161 + }, + { + "epoch": 0.64, + "learning_rate": 5.919390297596756e-06, + "loss": 0.8833, + "step": 20162 + }, + { + "epoch": 0.64, + "learning_rate": 5.918444677645501e-06, + "loss": 0.9829, + "step": 20163 + }, + { + "epoch": 0.64, + "learning_rate": 5.917499101484819e-06, + "loss": 0.9277, + "step": 20164 + }, + { + "epoch": 0.64, + "learning_rate": 5.9165535691248525e-06, + "loss": 1.022, + "step": 20165 + }, + { + "epoch": 0.64, + "learning_rate": 5.9156080805757455e-06, + "loss": 0.9741, + "step": 20166 + }, + { + "epoch": 0.64, + "learning_rate": 5.914662635847646e-06, + "loss": 0.8564, + "step": 20167 + }, + { + "epoch": 0.64, + "learning_rate": 5.913717234950692e-06, + "loss": 1.0186, + "step": 20168 + }, + { + "epoch": 0.65, + "learning_rate": 5.9127718778950325e-06, + "loss": 1.0732, + "step": 20169 + }, + { + "epoch": 0.65, + "learning_rate": 5.911826564690808e-06, + "loss": 1.064, + "step": 20170 + }, + { + "epoch": 0.65, + "learning_rate": 5.910881295348157e-06, + "loss": 0.9907, + "step": 20171 + }, + { + "epoch": 0.65, + "learning_rate": 5.909936069877225e-06, + "loss": 0.9668, + "step": 20172 + }, + { + "epoch": 0.65, + "learning_rate": 5.908990888288151e-06, + "loss": 1.0327, + "step": 20173 + }, + { + "epoch": 0.65, + "learning_rate": 5.908045750591078e-06, + "loss": 0.7485, + "step": 20174 + }, + { + "epoch": 0.65, + "learning_rate": 5.907100656796145e-06, + "loss": 0.8779, + "step": 20175 + }, + { + "epoch": 0.65, + "learning_rate": 5.906155606913493e-06, + "loss": 0.8735, + "step": 20176 + }, + { + "epoch": 0.65, + "learning_rate": 5.905210600953264e-06, + "loss": 0.8286, + "step": 20177 + }, + { + "epoch": 0.65, + "learning_rate": 5.904265638925588e-06, + "loss": 1.0439, + "step": 20178 + }, + { + "epoch": 0.65, + "learning_rate": 5.90332072084061e-06, + "loss": 0.9941, + "step": 20179 + }, + { + "epoch": 0.65, + "learning_rate": 5.902375846708468e-06, + "loss": 0.936, + "step": 20180 + }, + { + "epoch": 0.65, + "learning_rate": 5.901431016539296e-06, + "loss": 0.9092, + "step": 20181 + }, + { + "epoch": 0.65, + "learning_rate": 5.900486230343235e-06, + "loss": 1.0454, + "step": 20182 + }, + { + "epoch": 0.65, + "learning_rate": 5.899541488130418e-06, + "loss": 1.0171, + "step": 20183 + }, + { + "epoch": 0.65, + "learning_rate": 5.898596789910986e-06, + "loss": 0.9009, + "step": 20184 + }, + { + "epoch": 0.65, + "learning_rate": 5.897652135695069e-06, + "loss": 0.9199, + "step": 20185 + }, + { + "epoch": 0.65, + "learning_rate": 5.8967075254928045e-06, + "loss": 0.9346, + "step": 20186 + }, + { + "epoch": 0.65, + "learning_rate": 5.895762959314326e-06, + "loss": 1.0605, + "step": 20187 + }, + { + "epoch": 0.65, + "learning_rate": 5.894818437169768e-06, + "loss": 0.8848, + "step": 20188 + }, + { + "epoch": 0.65, + "learning_rate": 5.893873959069265e-06, + "loss": 0.8823, + "step": 20189 + }, + { + "epoch": 0.65, + "learning_rate": 5.892929525022951e-06, + "loss": 0.9021, + "step": 20190 + }, + { + "epoch": 0.65, + "learning_rate": 5.891985135040959e-06, + "loss": 1.0298, + "step": 20191 + }, + { + "epoch": 0.65, + "learning_rate": 5.891040789133421e-06, + "loss": 0.918, + "step": 20192 + }, + { + "epoch": 0.65, + "learning_rate": 5.890096487310465e-06, + "loss": 1.0283, + "step": 20193 + }, + { + "epoch": 0.65, + "learning_rate": 5.889152229582225e-06, + "loss": 0.9126, + "step": 20194 + }, + { + "epoch": 0.65, + "learning_rate": 5.888208015958832e-06, + "loss": 0.8965, + "step": 20195 + }, + { + "epoch": 0.65, + "learning_rate": 5.887263846450417e-06, + "loss": 0.9819, + "step": 20196 + }, + { + "epoch": 0.65, + "learning_rate": 5.8863197210671085e-06, + "loss": 0.9502, + "step": 20197 + }, + { + "epoch": 0.65, + "learning_rate": 5.8853756398190375e-06, + "loss": 0.96, + "step": 20198 + }, + { + "epoch": 0.65, + "learning_rate": 5.8844316027163315e-06, + "loss": 1.0854, + "step": 20199 + }, + { + "epoch": 0.65, + "learning_rate": 5.883487609769121e-06, + "loss": 0.9487, + "step": 20200 + }, + { + "epoch": 0.65, + "learning_rate": 5.882543660987531e-06, + "loss": 1.0156, + "step": 20201 + }, + { + "epoch": 0.65, + "learning_rate": 5.881599756381691e-06, + "loss": 0.9399, + "step": 20202 + }, + { + "epoch": 0.65, + "learning_rate": 5.880655895961728e-06, + "loss": 0.9204, + "step": 20203 + }, + { + "epoch": 0.65, + "learning_rate": 5.8797120797377675e-06, + "loss": 0.9917, + "step": 20204 + }, + { + "epoch": 0.65, + "learning_rate": 5.8787683077199355e-06, + "loss": 1.0112, + "step": 20205 + }, + { + "epoch": 0.65, + "learning_rate": 5.877824579918358e-06, + "loss": 0.9854, + "step": 20206 + }, + { + "epoch": 0.65, + "learning_rate": 5.876880896343163e-06, + "loss": 0.9961, + "step": 20207 + }, + { + "epoch": 0.65, + "learning_rate": 5.875937257004471e-06, + "loss": 0.8682, + "step": 20208 + }, + { + "epoch": 0.65, + "learning_rate": 5.874993661912411e-06, + "loss": 0.9097, + "step": 20209 + }, + { + "epoch": 0.65, + "learning_rate": 5.8740501110771e-06, + "loss": 0.9238, + "step": 20210 + }, + { + "epoch": 0.65, + "learning_rate": 5.873106604508665e-06, + "loss": 0.8262, + "step": 20211 + }, + { + "epoch": 0.65, + "learning_rate": 5.87216314221723e-06, + "loss": 1.0151, + "step": 20212 + }, + { + "epoch": 0.65, + "learning_rate": 5.871219724212915e-06, + "loss": 1.0098, + "step": 20213 + }, + { + "epoch": 0.65, + "learning_rate": 5.870276350505843e-06, + "loss": 1.1694, + "step": 20214 + }, + { + "epoch": 0.65, + "learning_rate": 5.869333021106139e-06, + "loss": 0.8784, + "step": 20215 + }, + { + "epoch": 0.65, + "learning_rate": 5.8683897360239164e-06, + "loss": 0.9922, + "step": 20216 + }, + { + "epoch": 0.65, + "learning_rate": 5.8674464952693e-06, + "loss": 0.9468, + "step": 20217 + }, + { + "epoch": 0.65, + "learning_rate": 5.866503298852409e-06, + "loss": 0.8584, + "step": 20218 + }, + { + "epoch": 0.65, + "learning_rate": 5.865560146783361e-06, + "loss": 1.0015, + "step": 20219 + }, + { + "epoch": 0.65, + "learning_rate": 5.8646170390722775e-06, + "loss": 1.0254, + "step": 20220 + }, + { + "epoch": 0.65, + "learning_rate": 5.863673975729278e-06, + "loss": 0.9438, + "step": 20221 + }, + { + "epoch": 0.65, + "learning_rate": 5.862730956764482e-06, + "loss": 0.8945, + "step": 20222 + }, + { + "epoch": 0.65, + "learning_rate": 5.861787982187998e-06, + "loss": 0.8652, + "step": 20223 + }, + { + "epoch": 0.65, + "learning_rate": 5.860845052009951e-06, + "loss": 0.939, + "step": 20224 + }, + { + "epoch": 0.65, + "learning_rate": 5.859902166240457e-06, + "loss": 0.9458, + "step": 20225 + }, + { + "epoch": 0.65, + "learning_rate": 5.858959324889628e-06, + "loss": 0.9336, + "step": 20226 + }, + { + "epoch": 0.65, + "learning_rate": 5.858016527967583e-06, + "loss": 0.9106, + "step": 20227 + }, + { + "epoch": 0.65, + "learning_rate": 5.857073775484436e-06, + "loss": 1.1172, + "step": 20228 + }, + { + "epoch": 0.65, + "learning_rate": 5.856131067450305e-06, + "loss": 1.0469, + "step": 20229 + }, + { + "epoch": 0.65, + "learning_rate": 5.8551884038753025e-06, + "loss": 0.9229, + "step": 20230 + }, + { + "epoch": 0.65, + "learning_rate": 5.854245784769539e-06, + "loss": 0.9814, + "step": 20231 + }, + { + "epoch": 0.65, + "learning_rate": 5.853303210143129e-06, + "loss": 0.9082, + "step": 20232 + }, + { + "epoch": 0.65, + "learning_rate": 5.8523606800061865e-06, + "loss": 1.0425, + "step": 20233 + }, + { + "epoch": 0.65, + "learning_rate": 5.851418194368823e-06, + "loss": 1.0293, + "step": 20234 + }, + { + "epoch": 0.65, + "learning_rate": 5.850475753241152e-06, + "loss": 0.5388, + "step": 20235 + }, + { + "epoch": 0.65, + "learning_rate": 5.849533356633283e-06, + "loss": 1.0244, + "step": 20236 + }, + { + "epoch": 0.65, + "learning_rate": 5.848591004555331e-06, + "loss": 0.8726, + "step": 20237 + }, + { + "epoch": 0.65, + "learning_rate": 5.847648697017399e-06, + "loss": 0.9424, + "step": 20238 + }, + { + "epoch": 0.65, + "learning_rate": 5.846706434029602e-06, + "loss": 0.9175, + "step": 20239 + }, + { + "epoch": 0.65, + "learning_rate": 5.845764215602048e-06, + "loss": 0.9478, + "step": 20240 + }, + { + "epoch": 0.65, + "learning_rate": 5.844822041744846e-06, + "loss": 1.0464, + "step": 20241 + }, + { + "epoch": 0.65, + "learning_rate": 5.843879912468104e-06, + "loss": 0.9683, + "step": 20242 + }, + { + "epoch": 0.65, + "learning_rate": 5.842937827781934e-06, + "loss": 1.019, + "step": 20243 + }, + { + "epoch": 0.65, + "learning_rate": 5.841995787696438e-06, + "loss": 0.8091, + "step": 20244 + }, + { + "epoch": 0.65, + "learning_rate": 5.841053792221727e-06, + "loss": 0.9346, + "step": 20245 + }, + { + "epoch": 0.65, + "learning_rate": 5.840111841367905e-06, + "loss": 0.9854, + "step": 20246 + }, + { + "epoch": 0.65, + "learning_rate": 5.839169935145078e-06, + "loss": 0.8667, + "step": 20247 + }, + { + "epoch": 0.65, + "learning_rate": 5.838228073563355e-06, + "loss": 1.0073, + "step": 20248 + }, + { + "epoch": 0.65, + "learning_rate": 5.8372862566328345e-06, + "loss": 0.9961, + "step": 20249 + }, + { + "epoch": 0.65, + "learning_rate": 5.83634448436363e-06, + "loss": 0.9258, + "step": 20250 + }, + { + "epoch": 0.65, + "learning_rate": 5.835402756765836e-06, + "loss": 0.981, + "step": 20251 + }, + { + "epoch": 0.65, + "learning_rate": 5.8344610738495685e-06, + "loss": 0.873, + "step": 20252 + }, + { + "epoch": 0.65, + "learning_rate": 5.833519435624917e-06, + "loss": 1.0386, + "step": 20253 + }, + { + "epoch": 0.65, + "learning_rate": 5.832577842101995e-06, + "loss": 0.9692, + "step": 20254 + }, + { + "epoch": 0.65, + "learning_rate": 5.831636293290896e-06, + "loss": 0.9517, + "step": 20255 + }, + { + "epoch": 0.65, + "learning_rate": 5.83069478920173e-06, + "loss": 0.9238, + "step": 20256 + }, + { + "epoch": 0.65, + "learning_rate": 5.829753329844592e-06, + "loss": 0.8628, + "step": 20257 + }, + { + "epoch": 0.65, + "learning_rate": 5.8288119152295876e-06, + "loss": 0.8711, + "step": 20258 + }, + { + "epoch": 0.65, + "learning_rate": 5.827870545366815e-06, + "loss": 0.9917, + "step": 20259 + }, + { + "epoch": 0.65, + "learning_rate": 5.8269292202663755e-06, + "loss": 1.0586, + "step": 20260 + }, + { + "epoch": 0.65, + "learning_rate": 5.82598793993836e-06, + "loss": 0.9785, + "step": 20261 + }, + { + "epoch": 0.65, + "learning_rate": 5.82504670439288e-06, + "loss": 0.9497, + "step": 20262 + }, + { + "epoch": 0.65, + "learning_rate": 5.8241055136400256e-06, + "loss": 0.9751, + "step": 20263 + }, + { + "epoch": 0.65, + "learning_rate": 5.8231643676898995e-06, + "loss": 1.0015, + "step": 20264 + }, + { + "epoch": 0.65, + "learning_rate": 5.822223266552598e-06, + "loss": 1.0698, + "step": 20265 + }, + { + "epoch": 0.65, + "learning_rate": 5.821282210238213e-06, + "loss": 0.9912, + "step": 20266 + }, + { + "epoch": 0.65, + "learning_rate": 5.820341198756849e-06, + "loss": 0.981, + "step": 20267 + }, + { + "epoch": 0.65, + "learning_rate": 5.819400232118598e-06, + "loss": 0.9526, + "step": 20268 + }, + { + "epoch": 0.65, + "learning_rate": 5.8184593103335525e-06, + "loss": 0.8643, + "step": 20269 + }, + { + "epoch": 0.65, + "learning_rate": 5.817518433411813e-06, + "loss": 0.9268, + "step": 20270 + }, + { + "epoch": 0.65, + "learning_rate": 5.816577601363474e-06, + "loss": 0.9766, + "step": 20271 + }, + { + "epoch": 0.65, + "learning_rate": 5.815636814198622e-06, + "loss": 0.9839, + "step": 20272 + }, + { + "epoch": 0.65, + "learning_rate": 5.814696071927359e-06, + "loss": 0.9653, + "step": 20273 + }, + { + "epoch": 0.65, + "learning_rate": 5.813755374559773e-06, + "loss": 0.9673, + "step": 20274 + }, + { + "epoch": 0.65, + "learning_rate": 5.812814722105962e-06, + "loss": 1.0508, + "step": 20275 + }, + { + "epoch": 0.65, + "learning_rate": 5.811874114576015e-06, + "loss": 0.9697, + "step": 20276 + }, + { + "epoch": 0.65, + "learning_rate": 5.810933551980022e-06, + "loss": 0.9683, + "step": 20277 + }, + { + "epoch": 0.65, + "learning_rate": 5.809993034328074e-06, + "loss": 1.0762, + "step": 20278 + }, + { + "epoch": 0.65, + "learning_rate": 5.809052561630266e-06, + "loss": 0.9189, + "step": 20279 + }, + { + "epoch": 0.65, + "learning_rate": 5.808112133896682e-06, + "loss": 0.8955, + "step": 20280 + }, + { + "epoch": 0.65, + "learning_rate": 5.807171751137421e-06, + "loss": 0.8872, + "step": 20281 + }, + { + "epoch": 0.65, + "learning_rate": 5.8062314133625605e-06, + "loss": 1.0352, + "step": 20282 + }, + { + "epoch": 0.65, + "learning_rate": 5.805291120582204e-06, + "loss": 1.0259, + "step": 20283 + }, + { + "epoch": 0.65, + "learning_rate": 5.804350872806423e-06, + "loss": 0.9854, + "step": 20284 + }, + { + "epoch": 0.65, + "learning_rate": 5.80341067004532e-06, + "loss": 0.9736, + "step": 20285 + }, + { + "epoch": 0.65, + "learning_rate": 5.802470512308969e-06, + "loss": 0.8662, + "step": 20286 + }, + { + "epoch": 0.65, + "learning_rate": 5.80153039960747e-06, + "loss": 0.958, + "step": 20287 + }, + { + "epoch": 0.65, + "learning_rate": 5.800590331950898e-06, + "loss": 0.9443, + "step": 20288 + }, + { + "epoch": 0.65, + "learning_rate": 5.799650309349348e-06, + "loss": 0.8628, + "step": 20289 + }, + { + "epoch": 0.65, + "learning_rate": 5.798710331812902e-06, + "loss": 1.0283, + "step": 20290 + }, + { + "epoch": 0.65, + "learning_rate": 5.7977703993516445e-06, + "loss": 0.8916, + "step": 20291 + }, + { + "epoch": 0.65, + "learning_rate": 5.796830511975655e-06, + "loss": 0.96, + "step": 20292 + }, + { + "epoch": 0.65, + "learning_rate": 5.795890669695028e-06, + "loss": 0.9121, + "step": 20293 + }, + { + "epoch": 0.65, + "learning_rate": 5.794950872519835e-06, + "loss": 0.9175, + "step": 20294 + }, + { + "epoch": 0.65, + "learning_rate": 5.79401112046017e-06, + "loss": 0.8779, + "step": 20295 + }, + { + "epoch": 0.65, + "learning_rate": 5.793071413526108e-06, + "loss": 0.7415, + "step": 20296 + }, + { + "epoch": 0.65, + "learning_rate": 5.7921317517277365e-06, + "loss": 1.019, + "step": 20297 + }, + { + "epoch": 0.65, + "learning_rate": 5.791192135075134e-06, + "loss": 0.9092, + "step": 20298 + }, + { + "epoch": 0.65, + "learning_rate": 5.790252563578382e-06, + "loss": 1.0537, + "step": 20299 + }, + { + "epoch": 0.65, + "learning_rate": 5.789313037247557e-06, + "loss": 0.9209, + "step": 20300 + }, + { + "epoch": 0.65, + "learning_rate": 5.788373556092748e-06, + "loss": 0.9346, + "step": 20301 + }, + { + "epoch": 0.65, + "learning_rate": 5.787434120124025e-06, + "loss": 0.917, + "step": 20302 + }, + { + "epoch": 0.65, + "learning_rate": 5.786494729351475e-06, + "loss": 1.0288, + "step": 20303 + }, + { + "epoch": 0.65, + "learning_rate": 5.785555383785173e-06, + "loss": 0.8535, + "step": 20304 + }, + { + "epoch": 0.65, + "learning_rate": 5.784616083435198e-06, + "loss": 0.9282, + "step": 20305 + }, + { + "epoch": 0.65, + "learning_rate": 5.783676828311623e-06, + "loss": 0.8862, + "step": 20306 + }, + { + "epoch": 0.65, + "learning_rate": 5.782737618424533e-06, + "loss": 0.7939, + "step": 20307 + }, + { + "epoch": 0.65, + "learning_rate": 5.781798453783998e-06, + "loss": 0.9878, + "step": 20308 + }, + { + "epoch": 0.65, + "learning_rate": 5.7808593344000995e-06, + "loss": 0.9668, + "step": 20309 + }, + { + "epoch": 0.65, + "learning_rate": 5.779920260282912e-06, + "loss": 0.8325, + "step": 20310 + }, + { + "epoch": 0.65, + "learning_rate": 5.7789812314425055e-06, + "loss": 0.8462, + "step": 20311 + }, + { + "epoch": 0.65, + "learning_rate": 5.778042247888962e-06, + "loss": 0.9453, + "step": 20312 + }, + { + "epoch": 0.65, + "learning_rate": 5.777103309632353e-06, + "loss": 0.9419, + "step": 20313 + }, + { + "epoch": 0.65, + "learning_rate": 5.776164416682748e-06, + "loss": 0.9756, + "step": 20314 + }, + { + "epoch": 0.65, + "learning_rate": 5.775225569050228e-06, + "loss": 0.8848, + "step": 20315 + }, + { + "epoch": 0.65, + "learning_rate": 5.774286766744862e-06, + "loss": 0.9165, + "step": 20316 + }, + { + "epoch": 0.65, + "learning_rate": 5.7733480097767185e-06, + "loss": 0.9932, + "step": 20317 + }, + { + "epoch": 0.65, + "learning_rate": 5.7724092981558766e-06, + "loss": 0.939, + "step": 20318 + }, + { + "epoch": 0.65, + "learning_rate": 5.7714706318924e-06, + "loss": 0.8979, + "step": 20319 + }, + { + "epoch": 0.65, + "learning_rate": 5.7705320109963716e-06, + "loss": 0.9648, + "step": 20320 + }, + { + "epoch": 0.65, + "learning_rate": 5.769593435477845e-06, + "loss": 0.9834, + "step": 20321 + }, + { + "epoch": 0.65, + "learning_rate": 5.768654905346905e-06, + "loss": 0.7783, + "step": 20322 + }, + { + "epoch": 0.65, + "learning_rate": 5.767716420613609e-06, + "loss": 0.9336, + "step": 20323 + }, + { + "epoch": 0.65, + "learning_rate": 5.766777981288036e-06, + "loss": 0.9248, + "step": 20324 + }, + { + "epoch": 0.65, + "learning_rate": 5.765839587380246e-06, + "loss": 0.7852, + "step": 20325 + }, + { + "epoch": 0.65, + "learning_rate": 5.764901238900316e-06, + "loss": 0.8691, + "step": 20326 + }, + { + "epoch": 0.65, + "learning_rate": 5.7639629358583045e-06, + "loss": 1.0, + "step": 20327 + }, + { + "epoch": 0.65, + "learning_rate": 5.763024678264289e-06, + "loss": 0.8711, + "step": 20328 + }, + { + "epoch": 0.65, + "learning_rate": 5.762086466128323e-06, + "loss": 0.9863, + "step": 20329 + }, + { + "epoch": 0.65, + "learning_rate": 5.7611482994604824e-06, + "loss": 0.9941, + "step": 20330 + }, + { + "epoch": 0.65, + "learning_rate": 5.760210178270827e-06, + "loss": 0.9805, + "step": 20331 + }, + { + "epoch": 0.65, + "learning_rate": 5.7592721025694246e-06, + "loss": 0.9258, + "step": 20332 + }, + { + "epoch": 0.65, + "learning_rate": 5.758334072366337e-06, + "loss": 0.958, + "step": 20333 + }, + { + "epoch": 0.65, + "learning_rate": 5.757396087671634e-06, + "loss": 1.0703, + "step": 20334 + }, + { + "epoch": 0.65, + "learning_rate": 5.756458148495376e-06, + "loss": 0.9043, + "step": 20335 + }, + { + "epoch": 0.65, + "learning_rate": 5.755520254847625e-06, + "loss": 1.0874, + "step": 20336 + }, + { + "epoch": 0.65, + "learning_rate": 5.754582406738441e-06, + "loss": 0.9849, + "step": 20337 + }, + { + "epoch": 0.65, + "learning_rate": 5.753644604177893e-06, + "loss": 0.9746, + "step": 20338 + }, + { + "epoch": 0.65, + "learning_rate": 5.752706847176033e-06, + "loss": 1.0293, + "step": 20339 + }, + { + "epoch": 0.65, + "learning_rate": 5.751769135742933e-06, + "loss": 0.8921, + "step": 20340 + }, + { + "epoch": 0.65, + "learning_rate": 5.750831469888645e-06, + "loss": 0.9468, + "step": 20341 + }, + { + "epoch": 0.65, + "learning_rate": 5.749893849623236e-06, + "loss": 0.9492, + "step": 20342 + }, + { + "epoch": 0.65, + "learning_rate": 5.748956274956763e-06, + "loss": 0.9385, + "step": 20343 + }, + { + "epoch": 0.65, + "learning_rate": 5.748018745899284e-06, + "loss": 1.0649, + "step": 20344 + }, + { + "epoch": 0.65, + "learning_rate": 5.747081262460853e-06, + "loss": 1.0, + "step": 20345 + }, + { + "epoch": 0.65, + "learning_rate": 5.746143824651537e-06, + "loss": 1.0327, + "step": 20346 + }, + { + "epoch": 0.65, + "learning_rate": 5.7452064324813876e-06, + "loss": 0.8857, + "step": 20347 + }, + { + "epoch": 0.65, + "learning_rate": 5.744269085960467e-06, + "loss": 0.9604, + "step": 20348 + }, + { + "epoch": 0.65, + "learning_rate": 5.74333178509883e-06, + "loss": 0.8682, + "step": 20349 + }, + { + "epoch": 0.65, + "learning_rate": 5.742394529906528e-06, + "loss": 1.0029, + "step": 20350 + }, + { + "epoch": 0.65, + "learning_rate": 5.741457320393623e-06, + "loss": 0.936, + "step": 20351 + }, + { + "epoch": 0.65, + "learning_rate": 5.7405201565701686e-06, + "loss": 0.9473, + "step": 20352 + }, + { + "epoch": 0.65, + "learning_rate": 5.739583038446215e-06, + "loss": 0.9502, + "step": 20353 + }, + { + "epoch": 0.65, + "learning_rate": 5.7386459660318235e-06, + "loss": 0.9521, + "step": 20354 + }, + { + "epoch": 0.65, + "learning_rate": 5.737708939337046e-06, + "loss": 0.9722, + "step": 20355 + }, + { + "epoch": 0.65, + "learning_rate": 5.73677195837193e-06, + "loss": 0.9614, + "step": 20356 + }, + { + "epoch": 0.65, + "learning_rate": 5.735835023146536e-06, + "loss": 0.9053, + "step": 20357 + }, + { + "epoch": 0.65, + "learning_rate": 5.734898133670912e-06, + "loss": 1.0059, + "step": 20358 + }, + { + "epoch": 0.65, + "learning_rate": 5.733961289955112e-06, + "loss": 0.9185, + "step": 20359 + }, + { + "epoch": 0.65, + "learning_rate": 5.733024492009181e-06, + "loss": 0.9463, + "step": 20360 + }, + { + "epoch": 0.65, + "learning_rate": 5.732087739843179e-06, + "loss": 0.9634, + "step": 20361 + }, + { + "epoch": 0.65, + "learning_rate": 5.73115103346715e-06, + "loss": 0.9375, + "step": 20362 + }, + { + "epoch": 0.65, + "learning_rate": 5.730214372891147e-06, + "loss": 0.9561, + "step": 20363 + }, + { + "epoch": 0.65, + "learning_rate": 5.7292777581252155e-06, + "loss": 0.8623, + "step": 20364 + }, + { + "epoch": 0.65, + "learning_rate": 5.728341189179412e-06, + "loss": 1.1221, + "step": 20365 + }, + { + "epoch": 0.65, + "learning_rate": 5.727404666063778e-06, + "loss": 1.0381, + "step": 20366 + }, + { + "epoch": 0.65, + "learning_rate": 5.7264681887883644e-06, + "loss": 1.106, + "step": 20367 + }, + { + "epoch": 0.65, + "learning_rate": 5.725531757363213e-06, + "loss": 0.8818, + "step": 20368 + }, + { + "epoch": 0.65, + "learning_rate": 5.724595371798379e-06, + "loss": 1.0664, + "step": 20369 + }, + { + "epoch": 0.65, + "learning_rate": 5.723659032103901e-06, + "loss": 0.7817, + "step": 20370 + }, + { + "epoch": 0.65, + "learning_rate": 5.722722738289833e-06, + "loss": 0.978, + "step": 20371 + }, + { + "epoch": 0.65, + "learning_rate": 5.721786490366212e-06, + "loss": 0.978, + "step": 20372 + }, + { + "epoch": 0.65, + "learning_rate": 5.7208502883430964e-06, + "loss": 0.9829, + "step": 20373 + }, + { + "epoch": 0.65, + "learning_rate": 5.719914132230511e-06, + "loss": 0.895, + "step": 20374 + }, + { + "epoch": 0.65, + "learning_rate": 5.718978022038515e-06, + "loss": 0.9141, + "step": 20375 + }, + { + "epoch": 0.65, + "learning_rate": 5.718041957777143e-06, + "loss": 1.0474, + "step": 20376 + }, + { + "epoch": 0.65, + "learning_rate": 5.717105939456445e-06, + "loss": 1.0742, + "step": 20377 + }, + { + "epoch": 0.65, + "learning_rate": 5.716169967086458e-06, + "loss": 0.9946, + "step": 20378 + }, + { + "epoch": 0.65, + "learning_rate": 5.715234040677229e-06, + "loss": 0.9673, + "step": 20379 + }, + { + "epoch": 0.65, + "learning_rate": 5.714298160238791e-06, + "loss": 0.8706, + "step": 20380 + }, + { + "epoch": 0.65, + "learning_rate": 5.713362325781201e-06, + "loss": 0.9512, + "step": 20381 + }, + { + "epoch": 0.65, + "learning_rate": 5.712426537314479e-06, + "loss": 0.9448, + "step": 20382 + }, + { + "epoch": 0.65, + "learning_rate": 5.711490794848681e-06, + "loss": 0.9424, + "step": 20383 + }, + { + "epoch": 0.65, + "learning_rate": 5.710555098393835e-06, + "loss": 1.0439, + "step": 20384 + }, + { + "epoch": 0.65, + "learning_rate": 5.709619447959991e-06, + "loss": 1.001, + "step": 20385 + }, + { + "epoch": 0.65, + "learning_rate": 5.708683843557177e-06, + "loss": 0.9839, + "step": 20386 + }, + { + "epoch": 0.65, + "learning_rate": 5.70774828519544e-06, + "loss": 0.8232, + "step": 20387 + }, + { + "epoch": 0.65, + "learning_rate": 5.7068127728848135e-06, + "loss": 1.0415, + "step": 20388 + }, + { + "epoch": 0.65, + "learning_rate": 5.705877306635334e-06, + "loss": 0.9409, + "step": 20389 + }, + { + "epoch": 0.65, + "learning_rate": 5.704941886457035e-06, + "loss": 0.9165, + "step": 20390 + }, + { + "epoch": 0.65, + "learning_rate": 5.704006512359961e-06, + "loss": 0.9463, + "step": 20391 + }, + { + "epoch": 0.65, + "learning_rate": 5.703071184354136e-06, + "loss": 0.896, + "step": 20392 + }, + { + "epoch": 0.65, + "learning_rate": 5.7021359024496085e-06, + "loss": 0.9541, + "step": 20393 + }, + { + "epoch": 0.65, + "learning_rate": 5.701200666656403e-06, + "loss": 1.0137, + "step": 20394 + }, + { + "epoch": 0.65, + "learning_rate": 5.7002654769845555e-06, + "loss": 0.9395, + "step": 20395 + }, + { + "epoch": 0.65, + "learning_rate": 5.699330333444104e-06, + "loss": 0.9316, + "step": 20396 + }, + { + "epoch": 0.65, + "learning_rate": 5.698395236045078e-06, + "loss": 1.0156, + "step": 20397 + }, + { + "epoch": 0.65, + "learning_rate": 5.69746018479751e-06, + "loss": 1.0273, + "step": 20398 + }, + { + "epoch": 0.65, + "learning_rate": 5.6965251797114284e-06, + "loss": 0.9407, + "step": 20399 + }, + { + "epoch": 0.65, + "learning_rate": 5.695590220796872e-06, + "loss": 1.0938, + "step": 20400 + }, + { + "epoch": 0.65, + "learning_rate": 5.694655308063867e-06, + "loss": 0.9824, + "step": 20401 + }, + { + "epoch": 0.65, + "learning_rate": 5.6937204415224475e-06, + "loss": 0.9507, + "step": 20402 + }, + { + "epoch": 0.65, + "learning_rate": 5.6927856211826425e-06, + "loss": 1.1416, + "step": 20403 + }, + { + "epoch": 0.65, + "learning_rate": 5.691850847054481e-06, + "loss": 1.0488, + "step": 20404 + }, + { + "epoch": 0.65, + "learning_rate": 5.690916119147987e-06, + "loss": 1.0298, + "step": 20405 + }, + { + "epoch": 0.65, + "learning_rate": 5.6899814374732e-06, + "loss": 1.0366, + "step": 20406 + }, + { + "epoch": 0.65, + "learning_rate": 5.689046802040137e-06, + "loss": 0.9263, + "step": 20407 + }, + { + "epoch": 0.65, + "learning_rate": 5.688112212858834e-06, + "loss": 0.9517, + "step": 20408 + }, + { + "epoch": 0.65, + "learning_rate": 5.687177669939311e-06, + "loss": 0.9561, + "step": 20409 + }, + { + "epoch": 0.65, + "learning_rate": 5.6862431732916035e-06, + "loss": 1.0796, + "step": 20410 + }, + { + "epoch": 0.65, + "learning_rate": 5.685308722925731e-06, + "loss": 0.9365, + "step": 20411 + }, + { + "epoch": 0.65, + "learning_rate": 5.684374318851721e-06, + "loss": 0.9541, + "step": 20412 + }, + { + "epoch": 0.65, + "learning_rate": 5.683439961079596e-06, + "loss": 1.0024, + "step": 20413 + }, + { + "epoch": 0.65, + "learning_rate": 5.682505649619384e-06, + "loss": 0.4817, + "step": 20414 + }, + { + "epoch": 0.65, + "learning_rate": 5.681571384481106e-06, + "loss": 0.9111, + "step": 20415 + }, + { + "epoch": 0.65, + "learning_rate": 5.680637165674791e-06, + "loss": 1.0366, + "step": 20416 + }, + { + "epoch": 0.65, + "learning_rate": 5.679702993210454e-06, + "loss": 0.856, + "step": 20417 + }, + { + "epoch": 0.65, + "learning_rate": 5.6787688670981326e-06, + "loss": 0.939, + "step": 20418 + }, + { + "epoch": 0.65, + "learning_rate": 5.677834787347829e-06, + "loss": 1.0317, + "step": 20419 + }, + { + "epoch": 0.65, + "learning_rate": 5.676900753969579e-06, + "loss": 0.8916, + "step": 20420 + }, + { + "epoch": 0.65, + "learning_rate": 5.675966766973395e-06, + "loss": 0.8906, + "step": 20421 + }, + { + "epoch": 0.65, + "learning_rate": 5.6750328263693065e-06, + "loss": 1.0034, + "step": 20422 + }, + { + "epoch": 0.65, + "learning_rate": 5.674098932167325e-06, + "loss": 0.9448, + "step": 20423 + }, + { + "epoch": 0.65, + "learning_rate": 5.673165084377479e-06, + "loss": 0.8882, + "step": 20424 + }, + { + "epoch": 0.65, + "learning_rate": 5.672231283009777e-06, + "loss": 0.9634, + "step": 20425 + }, + { + "epoch": 0.65, + "learning_rate": 5.671297528074253e-06, + "loss": 0.9482, + "step": 20426 + }, + { + "epoch": 0.65, + "learning_rate": 5.670363819580909e-06, + "loss": 0.7783, + "step": 20427 + }, + { + "epoch": 0.65, + "learning_rate": 5.66943015753977e-06, + "loss": 0.9463, + "step": 20428 + }, + { + "epoch": 0.65, + "learning_rate": 5.668496541960851e-06, + "loss": 0.9805, + "step": 20429 + }, + { + "epoch": 0.65, + "learning_rate": 5.667562972854175e-06, + "loss": 0.9331, + "step": 20430 + }, + { + "epoch": 0.65, + "learning_rate": 5.666629450229748e-06, + "loss": 1.0327, + "step": 20431 + }, + { + "epoch": 0.65, + "learning_rate": 5.665695974097595e-06, + "loss": 0.936, + "step": 20432 + }, + { + "epoch": 0.65, + "learning_rate": 5.664762544467728e-06, + "loss": 1.0752, + "step": 20433 + }, + { + "epoch": 0.65, + "learning_rate": 5.663829161350161e-06, + "loss": 0.9121, + "step": 20434 + }, + { + "epoch": 0.65, + "learning_rate": 5.662895824754902e-06, + "loss": 0.8545, + "step": 20435 + }, + { + "epoch": 0.65, + "learning_rate": 5.661962534691976e-06, + "loss": 0.9448, + "step": 20436 + }, + { + "epoch": 0.65, + "learning_rate": 5.661029291171391e-06, + "loss": 0.8784, + "step": 20437 + }, + { + "epoch": 0.65, + "learning_rate": 5.660096094203156e-06, + "loss": 1.0723, + "step": 20438 + }, + { + "epoch": 0.65, + "learning_rate": 5.659162943797291e-06, + "loss": 0.8591, + "step": 20439 + }, + { + "epoch": 0.65, + "learning_rate": 5.658229839963798e-06, + "loss": 0.9272, + "step": 20440 + }, + { + "epoch": 0.65, + "learning_rate": 5.657296782712699e-06, + "loss": 0.9331, + "step": 20441 + }, + { + "epoch": 0.65, + "learning_rate": 5.656363772053998e-06, + "loss": 0.8115, + "step": 20442 + }, + { + "epoch": 0.65, + "learning_rate": 5.655430807997708e-06, + "loss": 0.9517, + "step": 20443 + }, + { + "epoch": 0.65, + "learning_rate": 5.654497890553832e-06, + "loss": 1.0308, + "step": 20444 + }, + { + "epoch": 0.65, + "learning_rate": 5.653565019732388e-06, + "loss": 0.8618, + "step": 20445 + }, + { + "epoch": 0.65, + "learning_rate": 5.6526321955433775e-06, + "loss": 0.8638, + "step": 20446 + }, + { + "epoch": 0.65, + "learning_rate": 5.651699417996817e-06, + "loss": 0.9399, + "step": 20447 + }, + { + "epoch": 0.65, + "learning_rate": 5.650766687102704e-06, + "loss": 0.9414, + "step": 20448 + }, + { + "epoch": 0.65, + "learning_rate": 5.649834002871058e-06, + "loss": 1.062, + "step": 20449 + }, + { + "epoch": 0.65, + "learning_rate": 5.648901365311872e-06, + "loss": 0.9404, + "step": 20450 + }, + { + "epoch": 0.65, + "learning_rate": 5.647968774435163e-06, + "loss": 0.958, + "step": 20451 + }, + { + "epoch": 0.65, + "learning_rate": 5.647036230250928e-06, + "loss": 1.0156, + "step": 20452 + }, + { + "epoch": 0.65, + "learning_rate": 5.646103732769182e-06, + "loss": 0.9521, + "step": 20453 + }, + { + "epoch": 0.65, + "learning_rate": 5.645171281999918e-06, + "loss": 0.8491, + "step": 20454 + }, + { + "epoch": 0.65, + "learning_rate": 5.644238877953152e-06, + "loss": 0.9766, + "step": 20455 + }, + { + "epoch": 0.65, + "learning_rate": 5.64330652063888e-06, + "loss": 0.9199, + "step": 20456 + }, + { + "epoch": 0.65, + "learning_rate": 5.642374210067109e-06, + "loss": 1.1758, + "step": 20457 + }, + { + "epoch": 0.65, + "learning_rate": 5.641441946247836e-06, + "loss": 0.9082, + "step": 20458 + }, + { + "epoch": 0.65, + "learning_rate": 5.6405097291910725e-06, + "loss": 0.9956, + "step": 20459 + }, + { + "epoch": 0.65, + "learning_rate": 5.639577558906808e-06, + "loss": 0.9019, + "step": 20460 + }, + { + "epoch": 0.65, + "learning_rate": 5.6386454354050565e-06, + "loss": 0.9429, + "step": 20461 + }, + { + "epoch": 0.65, + "learning_rate": 5.637713358695808e-06, + "loss": 0.98, + "step": 20462 + }, + { + "epoch": 0.65, + "learning_rate": 5.6367813287890705e-06, + "loss": 0.916, + "step": 20463 + }, + { + "epoch": 0.65, + "learning_rate": 5.635849345694841e-06, + "loss": 0.9126, + "step": 20464 + }, + { + "epoch": 0.65, + "learning_rate": 5.6349174094231175e-06, + "loss": 0.9536, + "step": 20465 + }, + { + "epoch": 0.65, + "learning_rate": 5.633985519983895e-06, + "loss": 1.0576, + "step": 20466 + }, + { + "epoch": 0.65, + "learning_rate": 5.633053677387181e-06, + "loss": 0.9053, + "step": 20467 + }, + { + "epoch": 0.65, + "learning_rate": 5.632121881642962e-06, + "loss": 0.9736, + "step": 20468 + }, + { + "epoch": 0.65, + "learning_rate": 5.631190132761247e-06, + "loss": 0.9248, + "step": 20469 + }, + { + "epoch": 0.65, + "learning_rate": 5.630258430752022e-06, + "loss": 0.96, + "step": 20470 + }, + { + "epoch": 0.65, + "learning_rate": 5.6293267756252965e-06, + "loss": 0.9492, + "step": 20471 + }, + { + "epoch": 0.65, + "learning_rate": 5.628395167391049e-06, + "loss": 0.8735, + "step": 20472 + }, + { + "epoch": 0.65, + "learning_rate": 5.627463606059287e-06, + "loss": 1.0342, + "step": 20473 + }, + { + "epoch": 0.65, + "learning_rate": 5.626532091639997e-06, + "loss": 0.9937, + "step": 20474 + }, + { + "epoch": 0.65, + "learning_rate": 5.625600624143181e-06, + "loss": 0.9858, + "step": 20475 + }, + { + "epoch": 0.65, + "learning_rate": 5.62466920357883e-06, + "loss": 0.9868, + "step": 20476 + }, + { + "epoch": 0.65, + "learning_rate": 5.623737829956931e-06, + "loss": 0.9463, + "step": 20477 + }, + { + "epoch": 0.65, + "learning_rate": 5.622806503287486e-06, + "loss": 0.8945, + "step": 20478 + }, + { + "epoch": 0.65, + "learning_rate": 5.621875223580484e-06, + "loss": 0.9678, + "step": 20479 + }, + { + "epoch": 0.65, + "learning_rate": 5.620943990845909e-06, + "loss": 0.9502, + "step": 20480 + }, + { + "epoch": 0.65, + "learning_rate": 5.620012805093764e-06, + "loss": 0.5254, + "step": 20481 + }, + { + "epoch": 0.66, + "learning_rate": 5.6190816663340345e-06, + "loss": 0.9189, + "step": 20482 + }, + { + "epoch": 0.66, + "learning_rate": 5.618150574576706e-06, + "loss": 0.9614, + "step": 20483 + }, + { + "epoch": 0.66, + "learning_rate": 5.6172195298317755e-06, + "loss": 0.9497, + "step": 20484 + }, + { + "epoch": 0.66, + "learning_rate": 5.616288532109225e-06, + "loss": 0.8848, + "step": 20485 + }, + { + "epoch": 0.66, + "learning_rate": 5.6153575814190516e-06, + "loss": 1.0122, + "step": 20486 + }, + { + "epoch": 0.66, + "learning_rate": 5.614426677771239e-06, + "loss": 1.002, + "step": 20487 + }, + { + "epoch": 0.66, + "learning_rate": 5.613495821175772e-06, + "loss": 1.0161, + "step": 20488 + }, + { + "epoch": 0.66, + "learning_rate": 5.612565011642638e-06, + "loss": 0.9521, + "step": 20489 + }, + { + "epoch": 0.66, + "learning_rate": 5.61163424918183e-06, + "loss": 0.9844, + "step": 20490 + }, + { + "epoch": 0.66, + "learning_rate": 5.610703533803325e-06, + "loss": 0.9761, + "step": 20491 + }, + { + "epoch": 0.66, + "learning_rate": 5.609772865517117e-06, + "loss": 0.9199, + "step": 20492 + }, + { + "epoch": 0.66, + "learning_rate": 5.608842244333184e-06, + "loss": 0.959, + "step": 20493 + }, + { + "epoch": 0.66, + "learning_rate": 5.607911670261521e-06, + "loss": 0.8914, + "step": 20494 + }, + { + "epoch": 0.66, + "learning_rate": 5.606981143312097e-06, + "loss": 0.9707, + "step": 20495 + }, + { + "epoch": 0.66, + "learning_rate": 5.606050663494908e-06, + "loss": 0.9912, + "step": 20496 + }, + { + "epoch": 0.66, + "learning_rate": 5.605120230819927e-06, + "loss": 0.9424, + "step": 20497 + }, + { + "epoch": 0.66, + "learning_rate": 5.604189845297148e-06, + "loss": 0.7852, + "step": 20498 + }, + { + "epoch": 0.66, + "learning_rate": 5.603259506936541e-06, + "loss": 1.0562, + "step": 20499 + }, + { + "epoch": 0.66, + "learning_rate": 5.602329215748099e-06, + "loss": 0.9727, + "step": 20500 + }, + { + "epoch": 0.66, + "learning_rate": 5.601398971741792e-06, + "loss": 0.8721, + "step": 20501 + }, + { + "epoch": 0.66, + "learning_rate": 5.600468774927614e-06, + "loss": 0.8789, + "step": 20502 + }, + { + "epoch": 0.66, + "learning_rate": 5.59953862531553e-06, + "loss": 1.0269, + "step": 20503 + }, + { + "epoch": 0.66, + "learning_rate": 5.598608522915529e-06, + "loss": 0.9307, + "step": 20504 + }, + { + "epoch": 0.66, + "learning_rate": 5.597678467737585e-06, + "loss": 0.9272, + "step": 20505 + }, + { + "epoch": 0.66, + "learning_rate": 5.596748459791682e-06, + "loss": 1.0513, + "step": 20506 + }, + { + "epoch": 0.66, + "learning_rate": 5.595818499087791e-06, + "loss": 0.9976, + "step": 20507 + }, + { + "epoch": 0.66, + "learning_rate": 5.594888585635897e-06, + "loss": 0.9863, + "step": 20508 + }, + { + "epoch": 0.66, + "learning_rate": 5.593958719445973e-06, + "loss": 1.043, + "step": 20509 + }, + { + "epoch": 0.66, + "learning_rate": 5.5930289005279945e-06, + "loss": 0.4993, + "step": 20510 + }, + { + "epoch": 0.66, + "learning_rate": 5.592099128891937e-06, + "loss": 0.926, + "step": 20511 + }, + { + "epoch": 0.66, + "learning_rate": 5.591169404547781e-06, + "loss": 0.9258, + "step": 20512 + }, + { + "epoch": 0.66, + "learning_rate": 5.590239727505491e-06, + "loss": 0.9199, + "step": 20513 + }, + { + "epoch": 0.66, + "learning_rate": 5.589310097775055e-06, + "loss": 1.0205, + "step": 20514 + }, + { + "epoch": 0.66, + "learning_rate": 5.588380515366435e-06, + "loss": 0.9302, + "step": 20515 + }, + { + "epoch": 0.66, + "learning_rate": 5.587450980289614e-06, + "loss": 0.9761, + "step": 20516 + }, + { + "epoch": 0.66, + "learning_rate": 5.5865214925545595e-06, + "loss": 0.9868, + "step": 20517 + }, + { + "epoch": 0.66, + "learning_rate": 5.585592052171246e-06, + "loss": 1.001, + "step": 20518 + }, + { + "epoch": 0.66, + "learning_rate": 5.58466265914964e-06, + "loss": 1.0117, + "step": 20519 + }, + { + "epoch": 0.66, + "learning_rate": 5.583733313499719e-06, + "loss": 0.9956, + "step": 20520 + }, + { + "epoch": 0.66, + "learning_rate": 5.582804015231455e-06, + "loss": 0.9478, + "step": 20521 + }, + { + "epoch": 0.66, + "learning_rate": 5.581874764354809e-06, + "loss": 1.0806, + "step": 20522 + }, + { + "epoch": 0.66, + "learning_rate": 5.580945560879763e-06, + "loss": 0.9482, + "step": 20523 + }, + { + "epoch": 0.66, + "learning_rate": 5.580016404816279e-06, + "loss": 0.9453, + "step": 20524 + }, + { + "epoch": 0.66, + "learning_rate": 5.579087296174323e-06, + "loss": 0.9126, + "step": 20525 + }, + { + "epoch": 0.66, + "learning_rate": 5.578158234963872e-06, + "loss": 0.9819, + "step": 20526 + }, + { + "epoch": 0.66, + "learning_rate": 5.5772292211948885e-06, + "loss": 0.96, + "step": 20527 + }, + { + "epoch": 0.66, + "learning_rate": 5.576300254877338e-06, + "loss": 0.7998, + "step": 20528 + }, + { + "epoch": 0.66, + "learning_rate": 5.575371336021192e-06, + "loss": 0.9287, + "step": 20529 + }, + { + "epoch": 0.66, + "learning_rate": 5.574442464636412e-06, + "loss": 0.8901, + "step": 20530 + }, + { + "epoch": 0.66, + "learning_rate": 5.57351364073297e-06, + "loss": 0.957, + "step": 20531 + }, + { + "epoch": 0.66, + "learning_rate": 5.572584864320828e-06, + "loss": 0.9536, + "step": 20532 + }, + { + "epoch": 0.66, + "learning_rate": 5.57165613540995e-06, + "loss": 1.0034, + "step": 20533 + }, + { + "epoch": 0.66, + "learning_rate": 5.570727454010296e-06, + "loss": 0.9644, + "step": 20534 + }, + { + "epoch": 0.66, + "learning_rate": 5.56979882013184e-06, + "loss": 0.9595, + "step": 20535 + }, + { + "epoch": 0.66, + "learning_rate": 5.568870233784534e-06, + "loss": 0.9316, + "step": 20536 + }, + { + "epoch": 0.66, + "learning_rate": 5.567941694978351e-06, + "loss": 1.0127, + "step": 20537 + }, + { + "epoch": 0.66, + "learning_rate": 5.567013203723244e-06, + "loss": 0.8857, + "step": 20538 + }, + { + "epoch": 0.66, + "learning_rate": 5.566084760029188e-06, + "loss": 1.0356, + "step": 20539 + }, + { + "epoch": 0.66, + "learning_rate": 5.565156363906127e-06, + "loss": 0.9243, + "step": 20540 + }, + { + "epoch": 0.66, + "learning_rate": 5.564228015364032e-06, + "loss": 0.9834, + "step": 20541 + }, + { + "epoch": 0.66, + "learning_rate": 5.56329971441286e-06, + "loss": 1.0132, + "step": 20542 + }, + { + "epoch": 0.66, + "learning_rate": 5.5623714610625746e-06, + "loss": 0.8936, + "step": 20543 + }, + { + "epoch": 0.66, + "learning_rate": 5.561443255323129e-06, + "loss": 0.9375, + "step": 20544 + }, + { + "epoch": 0.66, + "learning_rate": 5.560515097204488e-06, + "loss": 0.9102, + "step": 20545 + }, + { + "epoch": 0.66, + "learning_rate": 5.559586986716602e-06, + "loss": 0.9614, + "step": 20546 + }, + { + "epoch": 0.66, + "learning_rate": 5.558658923869442e-06, + "loss": 0.8921, + "step": 20547 + }, + { + "epoch": 0.66, + "learning_rate": 5.557730908672948e-06, + "loss": 1.0405, + "step": 20548 + }, + { + "epoch": 0.66, + "learning_rate": 5.556802941137089e-06, + "loss": 0.5251, + "step": 20549 + }, + { + "epoch": 0.66, + "learning_rate": 5.555875021271812e-06, + "loss": 1.0371, + "step": 20550 + }, + { + "epoch": 0.66, + "learning_rate": 5.554947149087084e-06, + "loss": 0.9902, + "step": 20551 + }, + { + "epoch": 0.66, + "learning_rate": 5.5540193245928476e-06, + "loss": 1.002, + "step": 20552 + }, + { + "epoch": 0.66, + "learning_rate": 5.553091547799067e-06, + "loss": 0.9893, + "step": 20553 + }, + { + "epoch": 0.66, + "learning_rate": 5.552163818715694e-06, + "loss": 0.8711, + "step": 20554 + }, + { + "epoch": 0.66, + "learning_rate": 5.55123613735268e-06, + "loss": 0.9976, + "step": 20555 + }, + { + "epoch": 0.66, + "learning_rate": 5.550308503719975e-06, + "loss": 0.9155, + "step": 20556 + }, + { + "epoch": 0.66, + "learning_rate": 5.549380917827538e-06, + "loss": 0.8882, + "step": 20557 + }, + { + "epoch": 0.66, + "learning_rate": 5.548453379685315e-06, + "loss": 1.0112, + "step": 20558 + }, + { + "epoch": 0.66, + "learning_rate": 5.547525889303265e-06, + "loss": 1.0508, + "step": 20559 + }, + { + "epoch": 0.66, + "learning_rate": 5.546598446691334e-06, + "loss": 1.0425, + "step": 20560 + }, + { + "epoch": 0.66, + "learning_rate": 5.545671051859467e-06, + "loss": 0.9932, + "step": 20561 + }, + { + "epoch": 0.66, + "learning_rate": 5.544743704817625e-06, + "loss": 0.8955, + "step": 20562 + }, + { + "epoch": 0.66, + "learning_rate": 5.543816405575752e-06, + "loss": 1.0083, + "step": 20563 + }, + { + "epoch": 0.66, + "learning_rate": 5.542889154143794e-06, + "loss": 1.0679, + "step": 20564 + }, + { + "epoch": 0.66, + "learning_rate": 5.541961950531707e-06, + "loss": 0.9131, + "step": 20565 + }, + { + "epoch": 0.66, + "learning_rate": 5.541034794749433e-06, + "loss": 1.1646, + "step": 20566 + }, + { + "epoch": 0.66, + "learning_rate": 5.540107686806917e-06, + "loss": 0.9248, + "step": 20567 + }, + { + "epoch": 0.66, + "learning_rate": 5.539180626714113e-06, + "loss": 0.939, + "step": 20568 + }, + { + "epoch": 0.66, + "learning_rate": 5.538253614480966e-06, + "loss": 0.9375, + "step": 20569 + }, + { + "epoch": 0.66, + "learning_rate": 5.537326650117414e-06, + "loss": 1.0024, + "step": 20570 + }, + { + "epoch": 0.66, + "learning_rate": 5.536399733633413e-06, + "loss": 0.9756, + "step": 20571 + }, + { + "epoch": 0.66, + "learning_rate": 5.5354728650389024e-06, + "loss": 1.0044, + "step": 20572 + }, + { + "epoch": 0.66, + "learning_rate": 5.5345460443438215e-06, + "loss": 1.0181, + "step": 20573 + }, + { + "epoch": 0.66, + "learning_rate": 5.5336192715581235e-06, + "loss": 0.939, + "step": 20574 + }, + { + "epoch": 0.66, + "learning_rate": 5.532692546691746e-06, + "loss": 1.105, + "step": 20575 + }, + { + "epoch": 0.66, + "learning_rate": 5.531765869754635e-06, + "loss": 0.8379, + "step": 20576 + }, + { + "epoch": 0.66, + "learning_rate": 5.530839240756732e-06, + "loss": 0.9756, + "step": 20577 + }, + { + "epoch": 0.66, + "learning_rate": 5.5299126597079786e-06, + "loss": 1.0444, + "step": 20578 + }, + { + "epoch": 0.66, + "learning_rate": 5.52898612661831e-06, + "loss": 0.9419, + "step": 20579 + }, + { + "epoch": 0.66, + "learning_rate": 5.528059641497676e-06, + "loss": 1.0283, + "step": 20580 + }, + { + "epoch": 0.66, + "learning_rate": 5.527133204356008e-06, + "loss": 0.8896, + "step": 20581 + }, + { + "epoch": 0.66, + "learning_rate": 5.526206815203257e-06, + "loss": 1.0908, + "step": 20582 + }, + { + "epoch": 0.66, + "learning_rate": 5.5252804740493485e-06, + "loss": 0.9902, + "step": 20583 + }, + { + "epoch": 0.66, + "learning_rate": 5.524354180904233e-06, + "loss": 1.0425, + "step": 20584 + }, + { + "epoch": 0.66, + "learning_rate": 5.523427935777845e-06, + "loss": 0.8784, + "step": 20585 + }, + { + "epoch": 0.66, + "learning_rate": 5.522501738680119e-06, + "loss": 0.769, + "step": 20586 + }, + { + "epoch": 0.66, + "learning_rate": 5.521575589620991e-06, + "loss": 0.9595, + "step": 20587 + }, + { + "epoch": 0.66, + "learning_rate": 5.520649488610406e-06, + "loss": 0.9272, + "step": 20588 + }, + { + "epoch": 0.66, + "learning_rate": 5.519723435658289e-06, + "loss": 0.9897, + "step": 20589 + }, + { + "epoch": 0.66, + "learning_rate": 5.518797430774583e-06, + "loss": 0.9258, + "step": 20590 + }, + { + "epoch": 0.66, + "learning_rate": 5.517871473969219e-06, + "loss": 0.9194, + "step": 20591 + }, + { + "epoch": 0.66, + "learning_rate": 5.516945565252141e-06, + "loss": 1.1157, + "step": 20592 + }, + { + "epoch": 0.66, + "learning_rate": 5.516019704633269e-06, + "loss": 0.9941, + "step": 20593 + }, + { + "epoch": 0.66, + "learning_rate": 5.515093892122545e-06, + "loss": 0.9551, + "step": 20594 + }, + { + "epoch": 0.66, + "learning_rate": 5.514168127729897e-06, + "loss": 1.0029, + "step": 20595 + }, + { + "epoch": 0.66, + "learning_rate": 5.513242411465264e-06, + "loss": 0.8525, + "step": 20596 + }, + { + "epoch": 0.66, + "learning_rate": 5.51231674333857e-06, + "loss": 1.0361, + "step": 20597 + }, + { + "epoch": 0.66, + "learning_rate": 5.511391123359755e-06, + "loss": 0.9097, + "step": 20598 + }, + { + "epoch": 0.66, + "learning_rate": 5.510465551538745e-06, + "loss": 1.0239, + "step": 20599 + }, + { + "epoch": 0.66, + "learning_rate": 5.509540027885471e-06, + "loss": 1.0703, + "step": 20600 + }, + { + "epoch": 0.66, + "learning_rate": 5.508614552409859e-06, + "loss": 0.8311, + "step": 20601 + }, + { + "epoch": 0.66, + "learning_rate": 5.507689125121847e-06, + "loss": 0.8687, + "step": 20602 + }, + { + "epoch": 0.66, + "learning_rate": 5.506763746031354e-06, + "loss": 0.9434, + "step": 20603 + }, + { + "epoch": 0.66, + "learning_rate": 5.505838415148317e-06, + "loss": 0.939, + "step": 20604 + }, + { + "epoch": 0.66, + "learning_rate": 5.504913132482659e-06, + "loss": 1.0244, + "step": 20605 + }, + { + "epoch": 0.66, + "learning_rate": 5.503987898044305e-06, + "loss": 0.9844, + "step": 20606 + }, + { + "epoch": 0.66, + "learning_rate": 5.503062711843189e-06, + "loss": 0.8079, + "step": 20607 + }, + { + "epoch": 0.66, + "learning_rate": 5.502137573889232e-06, + "loss": 1.043, + "step": 20608 + }, + { + "epoch": 0.66, + "learning_rate": 5.5012124841923586e-06, + "loss": 0.9609, + "step": 20609 + }, + { + "epoch": 0.66, + "learning_rate": 5.5002874427624995e-06, + "loss": 0.9458, + "step": 20610 + }, + { + "epoch": 0.66, + "learning_rate": 5.4993624496095764e-06, + "loss": 0.9434, + "step": 20611 + }, + { + "epoch": 0.66, + "learning_rate": 5.498437504743509e-06, + "loss": 0.8708, + "step": 20612 + }, + { + "epoch": 0.66, + "learning_rate": 5.497512608174229e-06, + "loss": 0.9175, + "step": 20613 + }, + { + "epoch": 0.66, + "learning_rate": 5.496587759911652e-06, + "loss": 0.8647, + "step": 20614 + }, + { + "epoch": 0.66, + "learning_rate": 5.4956629599657105e-06, + "loss": 0.8628, + "step": 20615 + }, + { + "epoch": 0.66, + "learning_rate": 5.494738208346314e-06, + "loss": 0.9673, + "step": 20616 + }, + { + "epoch": 0.66, + "learning_rate": 5.4938135050633945e-06, + "loss": 0.8833, + "step": 20617 + }, + { + "epoch": 0.66, + "learning_rate": 5.492888850126864e-06, + "loss": 1.0938, + "step": 20618 + }, + { + "epoch": 0.66, + "learning_rate": 5.491964243546652e-06, + "loss": 1.02, + "step": 20619 + }, + { + "epoch": 0.66, + "learning_rate": 5.491039685332671e-06, + "loss": 0.9756, + "step": 20620 + }, + { + "epoch": 0.66, + "learning_rate": 5.4901151754948475e-06, + "loss": 0.9414, + "step": 20621 + }, + { + "epoch": 0.66, + "learning_rate": 5.489190714043098e-06, + "loss": 0.8916, + "step": 20622 + }, + { + "epoch": 0.66, + "learning_rate": 5.488266300987337e-06, + "loss": 0.9985, + "step": 20623 + }, + { + "epoch": 0.66, + "learning_rate": 5.487341936337484e-06, + "loss": 0.9033, + "step": 20624 + }, + { + "epoch": 0.66, + "learning_rate": 5.4864176201034605e-06, + "loss": 1.147, + "step": 20625 + }, + { + "epoch": 0.66, + "learning_rate": 5.485493352295177e-06, + "loss": 0.9741, + "step": 20626 + }, + { + "epoch": 0.66, + "learning_rate": 5.484569132922557e-06, + "loss": 1.02, + "step": 20627 + }, + { + "epoch": 0.66, + "learning_rate": 5.483644961995509e-06, + "loss": 0.8535, + "step": 20628 + }, + { + "epoch": 0.66, + "learning_rate": 5.482720839523956e-06, + "loss": 0.8086, + "step": 20629 + }, + { + "epoch": 0.66, + "learning_rate": 5.481796765517808e-06, + "loss": 0.9604, + "step": 20630 + }, + { + "epoch": 0.66, + "learning_rate": 5.480872739986982e-06, + "loss": 0.4844, + "step": 20631 + }, + { + "epoch": 0.66, + "learning_rate": 5.479948762941385e-06, + "loss": 0.9854, + "step": 20632 + }, + { + "epoch": 0.66, + "learning_rate": 5.47902483439094e-06, + "loss": 0.8896, + "step": 20633 + }, + { + "epoch": 0.66, + "learning_rate": 5.47810095434555e-06, + "loss": 0.9609, + "step": 20634 + }, + { + "epoch": 0.66, + "learning_rate": 5.477177122815136e-06, + "loss": 1.0586, + "step": 20635 + }, + { + "epoch": 0.66, + "learning_rate": 5.476253339809603e-06, + "loss": 0.9141, + "step": 20636 + }, + { + "epoch": 0.66, + "learning_rate": 5.475329605338873e-06, + "loss": 0.522, + "step": 20637 + }, + { + "epoch": 0.66, + "learning_rate": 5.474405919412839e-06, + "loss": 1.0273, + "step": 20638 + }, + { + "epoch": 0.66, + "learning_rate": 5.473482282041426e-06, + "loss": 0.9404, + "step": 20639 + }, + { + "epoch": 0.66, + "learning_rate": 5.472558693234534e-06, + "loss": 0.9795, + "step": 20640 + }, + { + "epoch": 0.66, + "learning_rate": 5.4716351530020805e-06, + "loss": 0.9529, + "step": 20641 + }, + { + "epoch": 0.66, + "learning_rate": 5.4707116613539664e-06, + "loss": 1.0337, + "step": 20642 + }, + { + "epoch": 0.66, + "learning_rate": 5.469788218300107e-06, + "loss": 0.9385, + "step": 20643 + }, + { + "epoch": 0.66, + "learning_rate": 5.468864823850407e-06, + "loss": 0.8799, + "step": 20644 + }, + { + "epoch": 0.66, + "learning_rate": 5.467941478014772e-06, + "loss": 0.8633, + "step": 20645 + }, + { + "epoch": 0.66, + "learning_rate": 5.467018180803106e-06, + "loss": 0.9585, + "step": 20646 + }, + { + "epoch": 0.66, + "learning_rate": 5.466094932225321e-06, + "loss": 0.9434, + "step": 20647 + }, + { + "epoch": 0.66, + "learning_rate": 5.465171732291317e-06, + "loss": 1.02, + "step": 20648 + }, + { + "epoch": 0.66, + "learning_rate": 5.464248581011002e-06, + "loss": 0.9448, + "step": 20649 + }, + { + "epoch": 0.66, + "learning_rate": 5.463325478394284e-06, + "loss": 1.0137, + "step": 20650 + }, + { + "epoch": 0.66, + "learning_rate": 5.462402424451057e-06, + "loss": 0.7991, + "step": 20651 + }, + { + "epoch": 0.66, + "learning_rate": 5.461479419191233e-06, + "loss": 0.9692, + "step": 20652 + }, + { + "epoch": 0.66, + "learning_rate": 5.460556462624714e-06, + "loss": 0.7915, + "step": 20653 + }, + { + "epoch": 0.66, + "learning_rate": 5.459633554761397e-06, + "loss": 1.0332, + "step": 20654 + }, + { + "epoch": 0.66, + "learning_rate": 5.4587106956111845e-06, + "loss": 0.8569, + "step": 20655 + }, + { + "epoch": 0.66, + "learning_rate": 5.457787885183985e-06, + "loss": 1.0039, + "step": 20656 + }, + { + "epoch": 0.66, + "learning_rate": 5.456865123489689e-06, + "loss": 0.9854, + "step": 20657 + }, + { + "epoch": 0.66, + "learning_rate": 5.455942410538207e-06, + "loss": 1.0034, + "step": 20658 + }, + { + "epoch": 0.66, + "learning_rate": 5.4550197463394295e-06, + "loss": 1.0151, + "step": 20659 + }, + { + "epoch": 0.66, + "learning_rate": 5.454097130903269e-06, + "loss": 1.0127, + "step": 20660 + }, + { + "epoch": 0.66, + "learning_rate": 5.453174564239606e-06, + "loss": 1.0, + "step": 20661 + }, + { + "epoch": 0.66, + "learning_rate": 5.452252046358353e-06, + "loss": 0.8721, + "step": 20662 + }, + { + "epoch": 0.66, + "learning_rate": 5.451329577269398e-06, + "loss": 0.9282, + "step": 20663 + }, + { + "epoch": 0.66, + "learning_rate": 5.450407156982644e-06, + "loss": 0.9375, + "step": 20664 + }, + { + "epoch": 0.66, + "learning_rate": 5.4494847855079846e-06, + "loss": 0.769, + "step": 20665 + }, + { + "epoch": 0.66, + "learning_rate": 5.4485624628553205e-06, + "loss": 1.0259, + "step": 20666 + }, + { + "epoch": 0.66, + "learning_rate": 5.44764018903454e-06, + "loss": 0.8931, + "step": 20667 + }, + { + "epoch": 0.66, + "learning_rate": 5.44671796405555e-06, + "loss": 1.0088, + "step": 20668 + }, + { + "epoch": 0.66, + "learning_rate": 5.445795787928228e-06, + "loss": 0.9233, + "step": 20669 + }, + { + "epoch": 0.66, + "learning_rate": 5.444873660662482e-06, + "loss": 0.98, + "step": 20670 + }, + { + "epoch": 0.66, + "learning_rate": 5.443951582268197e-06, + "loss": 0.936, + "step": 20671 + }, + { + "epoch": 0.66, + "learning_rate": 5.44302955275527e-06, + "loss": 0.9302, + "step": 20672 + }, + { + "epoch": 0.66, + "learning_rate": 5.442107572133591e-06, + "loss": 0.9854, + "step": 20673 + }, + { + "epoch": 0.66, + "learning_rate": 5.4411856404130566e-06, + "loss": 0.9688, + "step": 20674 + }, + { + "epoch": 0.66, + "learning_rate": 5.440263757603556e-06, + "loss": 1.0449, + "step": 20675 + }, + { + "epoch": 0.66, + "learning_rate": 5.439341923714977e-06, + "loss": 0.9854, + "step": 20676 + }, + { + "epoch": 0.66, + "learning_rate": 5.438420138757209e-06, + "loss": 0.9175, + "step": 20677 + }, + { + "epoch": 0.66, + "learning_rate": 5.4374984027401475e-06, + "loss": 0.8906, + "step": 20678 + }, + { + "epoch": 0.66, + "learning_rate": 5.436576715673676e-06, + "loss": 0.9277, + "step": 20679 + }, + { + "epoch": 0.66, + "learning_rate": 5.435655077567687e-06, + "loss": 1.0527, + "step": 20680 + }, + { + "epoch": 0.66, + "learning_rate": 5.434733488432067e-06, + "loss": 1.0776, + "step": 20681 + }, + { + "epoch": 0.66, + "learning_rate": 5.433811948276706e-06, + "loss": 0.9023, + "step": 20682 + }, + { + "epoch": 0.66, + "learning_rate": 5.4328904571114896e-06, + "loss": 0.9839, + "step": 20683 + }, + { + "epoch": 0.66, + "learning_rate": 5.431969014946303e-06, + "loss": 1.0239, + "step": 20684 + }, + { + "epoch": 0.66, + "learning_rate": 5.4310476217910305e-06, + "loss": 0.8027, + "step": 20685 + }, + { + "epoch": 0.66, + "learning_rate": 5.430126277655564e-06, + "loss": 0.9751, + "step": 20686 + }, + { + "epoch": 0.66, + "learning_rate": 5.42920498254978e-06, + "loss": 0.9077, + "step": 20687 + }, + { + "epoch": 0.66, + "learning_rate": 5.428283736483574e-06, + "loss": 1.0796, + "step": 20688 + }, + { + "epoch": 0.66, + "learning_rate": 5.427362539466822e-06, + "loss": 0.9282, + "step": 20689 + }, + { + "epoch": 0.66, + "learning_rate": 5.426441391509409e-06, + "loss": 0.8794, + "step": 20690 + }, + { + "epoch": 0.66, + "learning_rate": 5.425520292621216e-06, + "loss": 0.9893, + "step": 20691 + }, + { + "epoch": 0.66, + "learning_rate": 5.424599242812129e-06, + "loss": 1.0005, + "step": 20692 + }, + { + "epoch": 0.66, + "learning_rate": 5.42367824209203e-06, + "loss": 1.0049, + "step": 20693 + }, + { + "epoch": 0.66, + "learning_rate": 5.422757290470795e-06, + "loss": 0.9302, + "step": 20694 + }, + { + "epoch": 0.66, + "learning_rate": 5.421836387958311e-06, + "loss": 0.9966, + "step": 20695 + }, + { + "epoch": 0.66, + "learning_rate": 5.4209155345644525e-06, + "loss": 0.9888, + "step": 20696 + }, + { + "epoch": 0.66, + "learning_rate": 5.419994730299106e-06, + "loss": 0.9614, + "step": 20697 + }, + { + "epoch": 0.66, + "learning_rate": 5.419073975172148e-06, + "loss": 1.0278, + "step": 20698 + }, + { + "epoch": 0.66, + "learning_rate": 5.418153269193455e-06, + "loss": 0.9609, + "step": 20699 + }, + { + "epoch": 0.66, + "learning_rate": 5.417232612372903e-06, + "loss": 0.5432, + "step": 20700 + }, + { + "epoch": 0.66, + "learning_rate": 5.416312004720379e-06, + "loss": 1.0586, + "step": 20701 + }, + { + "epoch": 0.66, + "learning_rate": 5.415391446245748e-06, + "loss": 0.9478, + "step": 20702 + }, + { + "epoch": 0.66, + "learning_rate": 5.414470936958898e-06, + "loss": 1.0234, + "step": 20703 + }, + { + "epoch": 0.66, + "learning_rate": 5.413550476869693e-06, + "loss": 0.9536, + "step": 20704 + }, + { + "epoch": 0.66, + "learning_rate": 5.412630065988027e-06, + "loss": 0.9854, + "step": 20705 + }, + { + "epoch": 0.66, + "learning_rate": 5.411709704323754e-06, + "loss": 0.8887, + "step": 20706 + }, + { + "epoch": 0.66, + "learning_rate": 5.4107893918867614e-06, + "loss": 0.856, + "step": 20707 + }, + { + "epoch": 0.66, + "learning_rate": 5.409869128686915e-06, + "loss": 0.866, + "step": 20708 + }, + { + "epoch": 0.66, + "learning_rate": 5.408948914734098e-06, + "loss": 0.9507, + "step": 20709 + }, + { + "epoch": 0.66, + "learning_rate": 5.408028750038174e-06, + "loss": 0.9219, + "step": 20710 + }, + { + "epoch": 0.66, + "learning_rate": 5.407108634609023e-06, + "loss": 0.9731, + "step": 20711 + }, + { + "epoch": 0.66, + "learning_rate": 5.406188568456509e-06, + "loss": 0.9385, + "step": 20712 + }, + { + "epoch": 0.66, + "learning_rate": 5.4052685515905155e-06, + "loss": 1.0664, + "step": 20713 + }, + { + "epoch": 0.66, + "learning_rate": 5.404348584020897e-06, + "loss": 0.9204, + "step": 20714 + }, + { + "epoch": 0.66, + "learning_rate": 5.403428665757537e-06, + "loss": 0.9146, + "step": 20715 + }, + { + "epoch": 0.66, + "learning_rate": 5.402508796810295e-06, + "loss": 0.9844, + "step": 20716 + }, + { + "epoch": 0.66, + "learning_rate": 5.4015889771890495e-06, + "loss": 0.8345, + "step": 20717 + }, + { + "epoch": 0.66, + "learning_rate": 5.4006692069036615e-06, + "loss": 0.5007, + "step": 20718 + }, + { + "epoch": 0.66, + "learning_rate": 5.399749485964007e-06, + "loss": 0.8945, + "step": 20719 + }, + { + "epoch": 0.66, + "learning_rate": 5.398829814379949e-06, + "loss": 0.9556, + "step": 20720 + }, + { + "epoch": 0.66, + "learning_rate": 5.397910192161354e-06, + "loss": 1.0161, + "step": 20721 + }, + { + "epoch": 0.66, + "learning_rate": 5.3969906193180856e-06, + "loss": 0.8877, + "step": 20722 + }, + { + "epoch": 0.66, + "learning_rate": 5.3960710958600186e-06, + "loss": 0.9712, + "step": 20723 + }, + { + "epoch": 0.66, + "learning_rate": 5.395151621797008e-06, + "loss": 0.9316, + "step": 20724 + }, + { + "epoch": 0.66, + "learning_rate": 5.394232197138929e-06, + "loss": 0.8765, + "step": 20725 + }, + { + "epoch": 0.66, + "learning_rate": 5.393312821895635e-06, + "loss": 1.0557, + "step": 20726 + }, + { + "epoch": 0.66, + "learning_rate": 5.392393496077003e-06, + "loss": 0.8491, + "step": 20727 + }, + { + "epoch": 0.66, + "learning_rate": 5.391474219692887e-06, + "loss": 0.9814, + "step": 20728 + }, + { + "epoch": 0.66, + "learning_rate": 5.3905549927531545e-06, + "loss": 0.9575, + "step": 20729 + }, + { + "epoch": 0.66, + "learning_rate": 5.38963581526766e-06, + "loss": 0.5122, + "step": 20730 + }, + { + "epoch": 0.66, + "learning_rate": 5.388716687246274e-06, + "loss": 0.8147, + "step": 20731 + }, + { + "epoch": 0.66, + "learning_rate": 5.3877976086988514e-06, + "loss": 0.9053, + "step": 20732 + }, + { + "epoch": 0.66, + "learning_rate": 5.38687857963526e-06, + "loss": 1.0, + "step": 20733 + }, + { + "epoch": 0.66, + "learning_rate": 5.385959600065356e-06, + "loss": 1.0952, + "step": 20734 + }, + { + "epoch": 0.66, + "learning_rate": 5.385040669998994e-06, + "loss": 0.9492, + "step": 20735 + }, + { + "epoch": 0.66, + "learning_rate": 5.384121789446043e-06, + "loss": 0.9668, + "step": 20736 + }, + { + "epoch": 0.66, + "learning_rate": 5.383202958416356e-06, + "loss": 0.937, + "step": 20737 + }, + { + "epoch": 0.66, + "learning_rate": 5.382284176919792e-06, + "loss": 0.9839, + "step": 20738 + }, + { + "epoch": 0.66, + "learning_rate": 5.381365444966205e-06, + "loss": 1.0005, + "step": 20739 + }, + { + "epoch": 0.66, + "learning_rate": 5.3804467625654565e-06, + "loss": 0.8774, + "step": 20740 + }, + { + "epoch": 0.66, + "learning_rate": 5.3795281297274e-06, + "loss": 0.9233, + "step": 20741 + }, + { + "epoch": 0.66, + "learning_rate": 5.378609546461896e-06, + "loss": 0.896, + "step": 20742 + }, + { + "epoch": 0.66, + "learning_rate": 5.377691012778796e-06, + "loss": 0.9707, + "step": 20743 + }, + { + "epoch": 0.66, + "learning_rate": 5.376772528687958e-06, + "loss": 0.9517, + "step": 20744 + }, + { + "epoch": 0.66, + "learning_rate": 5.375854094199229e-06, + "loss": 0.8496, + "step": 20745 + }, + { + "epoch": 0.66, + "learning_rate": 5.374935709322471e-06, + "loss": 0.894, + "step": 20746 + }, + { + "epoch": 0.66, + "learning_rate": 5.374017374067531e-06, + "loss": 0.9468, + "step": 20747 + }, + { + "epoch": 0.66, + "learning_rate": 5.373099088444269e-06, + "loss": 0.9561, + "step": 20748 + }, + { + "epoch": 0.66, + "learning_rate": 5.372180852462527e-06, + "loss": 1.0059, + "step": 20749 + }, + { + "epoch": 0.66, + "learning_rate": 5.371262666132169e-06, + "loss": 0.8594, + "step": 20750 + }, + { + "epoch": 0.66, + "learning_rate": 5.370344529463038e-06, + "loss": 0.5315, + "step": 20751 + }, + { + "epoch": 0.66, + "learning_rate": 5.369426442464988e-06, + "loss": 0.8018, + "step": 20752 + }, + { + "epoch": 0.66, + "learning_rate": 5.368508405147863e-06, + "loss": 1.0352, + "step": 20753 + }, + { + "epoch": 0.66, + "learning_rate": 5.36759041752152e-06, + "loss": 0.978, + "step": 20754 + }, + { + "epoch": 0.66, + "learning_rate": 5.366672479595802e-06, + "loss": 1.0576, + "step": 20755 + }, + { + "epoch": 0.66, + "learning_rate": 5.365754591380563e-06, + "loss": 0.9365, + "step": 20756 + }, + { + "epoch": 0.66, + "learning_rate": 5.364836752885645e-06, + "loss": 0.854, + "step": 20757 + }, + { + "epoch": 0.66, + "learning_rate": 5.3639189641209065e-06, + "loss": 0.9346, + "step": 20758 + }, + { + "epoch": 0.66, + "learning_rate": 5.36300122509618e-06, + "loss": 0.9473, + "step": 20759 + }, + { + "epoch": 0.66, + "learning_rate": 5.36208353582132e-06, + "loss": 0.8213, + "step": 20760 + }, + { + "epoch": 0.66, + "learning_rate": 5.361165896306169e-06, + "loss": 0.9399, + "step": 20761 + }, + { + "epoch": 0.66, + "learning_rate": 5.360248306560577e-06, + "loss": 1.0083, + "step": 20762 + }, + { + "epoch": 0.66, + "learning_rate": 5.359330766594381e-06, + "loss": 0.9946, + "step": 20763 + }, + { + "epoch": 0.66, + "learning_rate": 5.3584132764174345e-06, + "loss": 0.9785, + "step": 20764 + }, + { + "epoch": 0.66, + "learning_rate": 5.357495836039572e-06, + "loss": 0.8613, + "step": 20765 + }, + { + "epoch": 0.66, + "learning_rate": 5.356578445470651e-06, + "loss": 1.0156, + "step": 20766 + }, + { + "epoch": 0.66, + "learning_rate": 5.355661104720493e-06, + "loss": 0.9995, + "step": 20767 + }, + { + "epoch": 0.66, + "learning_rate": 5.354743813798957e-06, + "loss": 0.9639, + "step": 20768 + }, + { + "epoch": 0.66, + "learning_rate": 5.353826572715875e-06, + "loss": 1.0283, + "step": 20769 + }, + { + "epoch": 0.66, + "learning_rate": 5.352909381481095e-06, + "loss": 1.0527, + "step": 20770 + }, + { + "epoch": 0.66, + "learning_rate": 5.351992240104451e-06, + "loss": 1.0264, + "step": 20771 + }, + { + "epoch": 0.66, + "learning_rate": 5.35107514859579e-06, + "loss": 0.8779, + "step": 20772 + }, + { + "epoch": 0.66, + "learning_rate": 5.350158106964947e-06, + "loss": 0.8389, + "step": 20773 + }, + { + "epoch": 0.66, + "learning_rate": 5.349241115221761e-06, + "loss": 0.9443, + "step": 20774 + }, + { + "epoch": 0.66, + "learning_rate": 5.348324173376067e-06, + "loss": 0.9614, + "step": 20775 + }, + { + "epoch": 0.66, + "learning_rate": 5.347407281437711e-06, + "loss": 0.969, + "step": 20776 + }, + { + "epoch": 0.66, + "learning_rate": 5.346490439416525e-06, + "loss": 0.9404, + "step": 20777 + }, + { + "epoch": 0.66, + "learning_rate": 5.3455736473223425e-06, + "loss": 0.8809, + "step": 20778 + }, + { + "epoch": 0.66, + "learning_rate": 5.344656905165008e-06, + "loss": 0.9907, + "step": 20779 + }, + { + "epoch": 0.66, + "learning_rate": 5.343740212954348e-06, + "loss": 0.9351, + "step": 20780 + }, + { + "epoch": 0.66, + "learning_rate": 5.342823570700206e-06, + "loss": 0.8608, + "step": 20781 + }, + { + "epoch": 0.66, + "learning_rate": 5.341906978412414e-06, + "loss": 0.9912, + "step": 20782 + }, + { + "epoch": 0.66, + "learning_rate": 5.3409904361008035e-06, + "loss": 0.9331, + "step": 20783 + }, + { + "epoch": 0.66, + "learning_rate": 5.340073943775206e-06, + "loss": 1.0547, + "step": 20784 + }, + { + "epoch": 0.66, + "learning_rate": 5.339157501445461e-06, + "loss": 0.8154, + "step": 20785 + }, + { + "epoch": 0.66, + "learning_rate": 5.338241109121393e-06, + "loss": 1.0239, + "step": 20786 + }, + { + "epoch": 0.66, + "learning_rate": 5.337324766812845e-06, + "loss": 0.9385, + "step": 20787 + }, + { + "epoch": 0.66, + "learning_rate": 5.33640847452964e-06, + "loss": 1.0259, + "step": 20788 + }, + { + "epoch": 0.66, + "learning_rate": 5.335492232281612e-06, + "loss": 0.9194, + "step": 20789 + }, + { + "epoch": 0.66, + "learning_rate": 5.334576040078585e-06, + "loss": 0.9634, + "step": 20790 + }, + { + "epoch": 0.66, + "learning_rate": 5.333659897930396e-06, + "loss": 0.7493, + "step": 20791 + }, + { + "epoch": 0.66, + "learning_rate": 5.332743805846871e-06, + "loss": 0.9473, + "step": 20792 + }, + { + "epoch": 0.66, + "learning_rate": 5.331827763837842e-06, + "loss": 1.0605, + "step": 20793 + }, + { + "epoch": 0.67, + "learning_rate": 5.33091177191313e-06, + "loss": 0.937, + "step": 20794 + }, + { + "epoch": 0.67, + "learning_rate": 5.329995830082573e-06, + "loss": 0.873, + "step": 20795 + }, + { + "epoch": 0.67, + "learning_rate": 5.3290799383559895e-06, + "loss": 1.0415, + "step": 20796 + }, + { + "epoch": 0.67, + "learning_rate": 5.3281640967432116e-06, + "loss": 0.915, + "step": 20797 + }, + { + "epoch": 0.67, + "learning_rate": 5.327248305254057e-06, + "loss": 0.9692, + "step": 20798 + }, + { + "epoch": 0.67, + "learning_rate": 5.326332563898361e-06, + "loss": 0.978, + "step": 20799 + }, + { + "epoch": 0.67, + "learning_rate": 5.32541687268594e-06, + "loss": 0.9351, + "step": 20800 + }, + { + "epoch": 0.67, + "learning_rate": 5.324501231626625e-06, + "loss": 1.0049, + "step": 20801 + }, + { + "epoch": 0.67, + "learning_rate": 5.323585640730235e-06, + "loss": 0.9648, + "step": 20802 + }, + { + "epoch": 0.67, + "learning_rate": 5.322670100006603e-06, + "loss": 0.9873, + "step": 20803 + }, + { + "epoch": 0.67, + "learning_rate": 5.321754609465536e-06, + "loss": 0.9761, + "step": 20804 + }, + { + "epoch": 0.67, + "learning_rate": 5.3208391691168686e-06, + "loss": 0.9473, + "step": 20805 + }, + { + "epoch": 0.67, + "learning_rate": 5.319923778970413e-06, + "loss": 0.9302, + "step": 20806 + }, + { + "epoch": 0.67, + "learning_rate": 5.319008439036002e-06, + "loss": 0.876, + "step": 20807 + }, + { + "epoch": 0.67, + "learning_rate": 5.318093149323444e-06, + "loss": 1.0552, + "step": 20808 + }, + { + "epoch": 0.67, + "learning_rate": 5.317177909842569e-06, + "loss": 1.0024, + "step": 20809 + }, + { + "epoch": 0.67, + "learning_rate": 5.316262720603189e-06, + "loss": 0.8608, + "step": 20810 + }, + { + "epoch": 0.67, + "learning_rate": 5.315347581615133e-06, + "loss": 0.9326, + "step": 20811 + }, + { + "epoch": 0.67, + "learning_rate": 5.314432492888206e-06, + "loss": 1.0342, + "step": 20812 + }, + { + "epoch": 0.67, + "learning_rate": 5.313517454432237e-06, + "loss": 0.9326, + "step": 20813 + }, + { + "epoch": 0.67, + "learning_rate": 5.312602466257034e-06, + "loss": 0.9321, + "step": 20814 + }, + { + "epoch": 0.67, + "learning_rate": 5.3116875283724226e-06, + "loss": 0.9102, + "step": 20815 + }, + { + "epoch": 0.67, + "learning_rate": 5.310772640788216e-06, + "loss": 0.9185, + "step": 20816 + }, + { + "epoch": 0.67, + "learning_rate": 5.3098578035142245e-06, + "loss": 0.9424, + "step": 20817 + }, + { + "epoch": 0.67, + "learning_rate": 5.308943016560273e-06, + "loss": 0.9194, + "step": 20818 + }, + { + "epoch": 0.67, + "learning_rate": 5.30802827993617e-06, + "loss": 0.8989, + "step": 20819 + }, + { + "epoch": 0.67, + "learning_rate": 5.307113593651727e-06, + "loss": 1.0459, + "step": 20820 + }, + { + "epoch": 0.67, + "learning_rate": 5.306198957716766e-06, + "loss": 0.8818, + "step": 20821 + }, + { + "epoch": 0.67, + "learning_rate": 5.305284372141095e-06, + "loss": 1.0039, + "step": 20822 + }, + { + "epoch": 0.67, + "learning_rate": 5.304369836934523e-06, + "loss": 1.0098, + "step": 20823 + }, + { + "epoch": 0.67, + "learning_rate": 5.30345535210687e-06, + "loss": 0.9917, + "step": 20824 + }, + { + "epoch": 0.67, + "learning_rate": 5.302540917667938e-06, + "loss": 1.0342, + "step": 20825 + }, + { + "epoch": 0.67, + "learning_rate": 5.301626533627549e-06, + "loss": 0.8628, + "step": 20826 + }, + { + "epoch": 0.67, + "learning_rate": 5.300712199995506e-06, + "loss": 0.9341, + "step": 20827 + }, + { + "epoch": 0.67, + "learning_rate": 5.299797916781619e-06, + "loss": 0.6841, + "step": 20828 + }, + { + "epoch": 0.67, + "learning_rate": 5.298883683995697e-06, + "loss": 0.8613, + "step": 20829 + }, + { + "epoch": 0.67, + "learning_rate": 5.297969501647552e-06, + "loss": 1.0283, + "step": 20830 + }, + { + "epoch": 0.67, + "learning_rate": 5.297055369746987e-06, + "loss": 0.7874, + "step": 20831 + }, + { + "epoch": 0.67, + "learning_rate": 5.296141288303818e-06, + "loss": 1.0474, + "step": 20832 + }, + { + "epoch": 0.67, + "learning_rate": 5.295227257327842e-06, + "loss": 0.4973, + "step": 20833 + }, + { + "epoch": 0.67, + "learning_rate": 5.294313276828877e-06, + "loss": 0.9141, + "step": 20834 + }, + { + "epoch": 0.67, + "learning_rate": 5.293399346816717e-06, + "loss": 1.022, + "step": 20835 + }, + { + "epoch": 0.67, + "learning_rate": 5.292485467301176e-06, + "loss": 0.9712, + "step": 20836 + }, + { + "epoch": 0.67, + "learning_rate": 5.291571638292052e-06, + "loss": 0.96, + "step": 20837 + }, + { + "epoch": 0.67, + "learning_rate": 5.2906578597991575e-06, + "loss": 0.9897, + "step": 20838 + }, + { + "epoch": 0.67, + "learning_rate": 5.289744131832288e-06, + "loss": 0.8838, + "step": 20839 + }, + { + "epoch": 0.67, + "learning_rate": 5.288830454401256e-06, + "loss": 1.0864, + "step": 20840 + }, + { + "epoch": 0.67, + "learning_rate": 5.2879168275158574e-06, + "loss": 0.897, + "step": 20841 + }, + { + "epoch": 0.67, + "learning_rate": 5.287003251185897e-06, + "loss": 0.9478, + "step": 20842 + }, + { + "epoch": 0.67, + "learning_rate": 5.286089725421172e-06, + "loss": 0.9419, + "step": 20843 + }, + { + "epoch": 0.67, + "learning_rate": 5.2851762502314906e-06, + "loss": 0.9448, + "step": 20844 + }, + { + "epoch": 0.67, + "learning_rate": 5.284262825626646e-06, + "loss": 0.8809, + "step": 20845 + }, + { + "epoch": 0.67, + "learning_rate": 5.283349451616447e-06, + "loss": 1.0552, + "step": 20846 + }, + { + "epoch": 0.67, + "learning_rate": 5.282436128210682e-06, + "loss": 0.896, + "step": 20847 + }, + { + "epoch": 0.67, + "learning_rate": 5.2815228554191625e-06, + "loss": 0.918, + "step": 20848 + }, + { + "epoch": 0.67, + "learning_rate": 5.2806096332516785e-06, + "loss": 1.0776, + "step": 20849 + }, + { + "epoch": 0.67, + "learning_rate": 5.279696461718031e-06, + "loss": 0.9653, + "step": 20850 + }, + { + "epoch": 0.67, + "learning_rate": 5.278783340828013e-06, + "loss": 0.9551, + "step": 20851 + }, + { + "epoch": 0.67, + "learning_rate": 5.277870270591429e-06, + "loss": 0.8496, + "step": 20852 + }, + { + "epoch": 0.67, + "learning_rate": 5.276957251018065e-06, + "loss": 0.5093, + "step": 20853 + }, + { + "epoch": 0.67, + "learning_rate": 5.276044282117727e-06, + "loss": 0.9297, + "step": 20854 + }, + { + "epoch": 0.67, + "learning_rate": 5.275131363900206e-06, + "loss": 0.9302, + "step": 20855 + }, + { + "epoch": 0.67, + "learning_rate": 5.274218496375296e-06, + "loss": 0.9658, + "step": 20856 + }, + { + "epoch": 0.67, + "learning_rate": 5.273305679552788e-06, + "loss": 0.9619, + "step": 20857 + }, + { + "epoch": 0.67, + "learning_rate": 5.272392913442484e-06, + "loss": 0.8188, + "step": 20858 + }, + { + "epoch": 0.67, + "learning_rate": 5.271480198054164e-06, + "loss": 0.8896, + "step": 20859 + }, + { + "epoch": 0.67, + "learning_rate": 5.270567533397634e-06, + "loss": 1.0293, + "step": 20860 + }, + { + "epoch": 0.67, + "learning_rate": 5.269654919482682e-06, + "loss": 0.8599, + "step": 20861 + }, + { + "epoch": 0.67, + "learning_rate": 5.268742356319092e-06, + "loss": 0.9902, + "step": 20862 + }, + { + "epoch": 0.67, + "learning_rate": 5.267829843916663e-06, + "loss": 0.9692, + "step": 20863 + }, + { + "epoch": 0.67, + "learning_rate": 5.266917382285183e-06, + "loss": 0.9824, + "step": 20864 + }, + { + "epoch": 0.67, + "learning_rate": 5.2660049714344385e-06, + "loss": 1.0068, + "step": 20865 + }, + { + "epoch": 0.67, + "learning_rate": 5.265092611374225e-06, + "loss": 1.0381, + "step": 20866 + }, + { + "epoch": 0.67, + "learning_rate": 5.264180302114327e-06, + "loss": 0.9106, + "step": 20867 + }, + { + "epoch": 0.67, + "learning_rate": 5.263268043664529e-06, + "loss": 1.0166, + "step": 20868 + }, + { + "epoch": 0.67, + "learning_rate": 5.262355836034626e-06, + "loss": 0.8281, + "step": 20869 + }, + { + "epoch": 0.67, + "learning_rate": 5.2614436792343985e-06, + "loss": 0.9731, + "step": 20870 + }, + { + "epoch": 0.67, + "learning_rate": 5.260531573273644e-06, + "loss": 0.894, + "step": 20871 + }, + { + "epoch": 0.67, + "learning_rate": 5.259619518162131e-06, + "loss": 0.8506, + "step": 20872 + }, + { + "epoch": 0.67, + "learning_rate": 5.258707513909659e-06, + "loss": 0.8667, + "step": 20873 + }, + { + "epoch": 0.67, + "learning_rate": 5.257795560526005e-06, + "loss": 1.0229, + "step": 20874 + }, + { + "epoch": 0.67, + "learning_rate": 5.256883658020958e-06, + "loss": 1.0195, + "step": 20875 + }, + { + "epoch": 0.67, + "learning_rate": 5.255971806404298e-06, + "loss": 0.9761, + "step": 20876 + }, + { + "epoch": 0.67, + "learning_rate": 5.2550600056858125e-06, + "loss": 1.0566, + "step": 20877 + }, + { + "epoch": 0.67, + "learning_rate": 5.2541482558752786e-06, + "loss": 1.0029, + "step": 20878 + }, + { + "epoch": 0.67, + "learning_rate": 5.253236556982489e-06, + "loss": 0.9858, + "step": 20879 + }, + { + "epoch": 0.67, + "learning_rate": 5.25232490901721e-06, + "loss": 0.8364, + "step": 20880 + }, + { + "epoch": 0.67, + "learning_rate": 5.2514133119892345e-06, + "loss": 0.9404, + "step": 20881 + }, + { + "epoch": 0.67, + "learning_rate": 5.250501765908334e-06, + "loss": 0.9648, + "step": 20882 + }, + { + "epoch": 0.67, + "learning_rate": 5.249590270784297e-06, + "loss": 1.0352, + "step": 20883 + }, + { + "epoch": 0.67, + "learning_rate": 5.248678826626896e-06, + "loss": 0.9199, + "step": 20884 + }, + { + "epoch": 0.67, + "learning_rate": 5.247767433445916e-06, + "loss": 0.8271, + "step": 20885 + }, + { + "epoch": 0.67, + "learning_rate": 5.246856091251132e-06, + "loss": 0.8994, + "step": 20886 + }, + { + "epoch": 0.67, + "learning_rate": 5.245944800052322e-06, + "loss": 1.002, + "step": 20887 + }, + { + "epoch": 0.67, + "learning_rate": 5.245033559859258e-06, + "loss": 0.8809, + "step": 20888 + }, + { + "epoch": 0.67, + "learning_rate": 5.244122370681726e-06, + "loss": 1.0273, + "step": 20889 + }, + { + "epoch": 0.67, + "learning_rate": 5.243211232529492e-06, + "loss": 0.8823, + "step": 20890 + }, + { + "epoch": 0.67, + "learning_rate": 5.242300145412343e-06, + "loss": 0.8916, + "step": 20891 + }, + { + "epoch": 0.67, + "learning_rate": 5.241389109340043e-06, + "loss": 0.9097, + "step": 20892 + }, + { + "epoch": 0.67, + "learning_rate": 5.2404781243223756e-06, + "loss": 0.8979, + "step": 20893 + }, + { + "epoch": 0.67, + "learning_rate": 5.23956719036911e-06, + "loss": 1.1597, + "step": 20894 + }, + { + "epoch": 0.67, + "learning_rate": 5.23865630749002e-06, + "loss": 0.9233, + "step": 20895 + }, + { + "epoch": 0.67, + "learning_rate": 5.237745475694873e-06, + "loss": 0.9448, + "step": 20896 + }, + { + "epoch": 0.67, + "learning_rate": 5.236834694993452e-06, + "loss": 0.9854, + "step": 20897 + }, + { + "epoch": 0.67, + "learning_rate": 5.235923965395519e-06, + "loss": 0.8755, + "step": 20898 + }, + { + "epoch": 0.67, + "learning_rate": 5.235013286910853e-06, + "loss": 1.0132, + "step": 20899 + }, + { + "epoch": 0.67, + "learning_rate": 5.23410265954922e-06, + "loss": 0.9241, + "step": 20900 + }, + { + "epoch": 0.67, + "learning_rate": 5.233192083320389e-06, + "loss": 0.9419, + "step": 20901 + }, + { + "epoch": 0.67, + "learning_rate": 5.2322815582341336e-06, + "loss": 0.98, + "step": 20902 + }, + { + "epoch": 0.67, + "learning_rate": 5.23137108430022e-06, + "loss": 0.9106, + "step": 20903 + }, + { + "epoch": 0.67, + "learning_rate": 5.2304606615284135e-06, + "loss": 1.0234, + "step": 20904 + }, + { + "epoch": 0.67, + "learning_rate": 5.229550289928489e-06, + "loss": 0.9287, + "step": 20905 + }, + { + "epoch": 0.67, + "learning_rate": 5.228639969510212e-06, + "loss": 1.0024, + "step": 20906 + }, + { + "epoch": 0.67, + "learning_rate": 5.2277297002833415e-06, + "loss": 0.9927, + "step": 20907 + }, + { + "epoch": 0.67, + "learning_rate": 5.2268194822576545e-06, + "loss": 1.0186, + "step": 20908 + }, + { + "epoch": 0.67, + "learning_rate": 5.225909315442912e-06, + "loss": 0.9077, + "step": 20909 + }, + { + "epoch": 0.67, + "learning_rate": 5.224999199848878e-06, + "loss": 0.9092, + "step": 20910 + }, + { + "epoch": 0.67, + "learning_rate": 5.224089135485315e-06, + "loss": 0.8726, + "step": 20911 + }, + { + "epoch": 0.67, + "learning_rate": 5.223179122361992e-06, + "loss": 1.0718, + "step": 20912 + }, + { + "epoch": 0.67, + "learning_rate": 5.2222691604886684e-06, + "loss": 0.906, + "step": 20913 + }, + { + "epoch": 0.67, + "learning_rate": 5.221359249875112e-06, + "loss": 0.8354, + "step": 20914 + }, + { + "epoch": 0.67, + "learning_rate": 5.2204493905310785e-06, + "loss": 0.8823, + "step": 20915 + }, + { + "epoch": 0.67, + "learning_rate": 5.219539582466338e-06, + "loss": 0.8984, + "step": 20916 + }, + { + "epoch": 0.67, + "learning_rate": 5.218629825690644e-06, + "loss": 1.0078, + "step": 20917 + }, + { + "epoch": 0.67, + "learning_rate": 5.2177201202137625e-06, + "loss": 0.936, + "step": 20918 + }, + { + "epoch": 0.67, + "learning_rate": 5.216810466045448e-06, + "loss": 1.0352, + "step": 20919 + }, + { + "epoch": 0.67, + "learning_rate": 5.215900863195467e-06, + "loss": 1.0088, + "step": 20920 + }, + { + "epoch": 0.67, + "learning_rate": 5.2149913116735695e-06, + "loss": 1.021, + "step": 20921 + }, + { + "epoch": 0.67, + "learning_rate": 5.214081811489524e-06, + "loss": 1.0059, + "step": 20922 + }, + { + "epoch": 0.67, + "learning_rate": 5.213172362653081e-06, + "loss": 0.8945, + "step": 20923 + }, + { + "epoch": 0.67, + "learning_rate": 5.212262965174009e-06, + "loss": 0.8242, + "step": 20924 + }, + { + "epoch": 0.67, + "learning_rate": 5.211353619062047e-06, + "loss": 1.0176, + "step": 20925 + }, + { + "epoch": 0.67, + "learning_rate": 5.2104443243269644e-06, + "loss": 1.0234, + "step": 20926 + }, + { + "epoch": 0.67, + "learning_rate": 5.209535080978512e-06, + "loss": 1.043, + "step": 20927 + }, + { + "epoch": 0.67, + "learning_rate": 5.208625889026448e-06, + "loss": 1.0186, + "step": 20928 + }, + { + "epoch": 0.67, + "learning_rate": 5.207716748480522e-06, + "loss": 0.9395, + "step": 20929 + }, + { + "epoch": 0.67, + "learning_rate": 5.206807659350496e-06, + "loss": 0.936, + "step": 20930 + }, + { + "epoch": 0.67, + "learning_rate": 5.205898621646115e-06, + "loss": 0.9775, + "step": 20931 + }, + { + "epoch": 0.67, + "learning_rate": 5.204989635377142e-06, + "loss": 1.0039, + "step": 20932 + }, + { + "epoch": 0.67, + "learning_rate": 5.204080700553318e-06, + "loss": 0.981, + "step": 20933 + }, + { + "epoch": 0.67, + "learning_rate": 5.203171817184403e-06, + "loss": 0.7793, + "step": 20934 + }, + { + "epoch": 0.67, + "learning_rate": 5.202262985280142e-06, + "loss": 1.021, + "step": 20935 + }, + { + "epoch": 0.67, + "learning_rate": 5.201354204850293e-06, + "loss": 0.8691, + "step": 20936 + }, + { + "epoch": 0.67, + "learning_rate": 5.200445475904598e-06, + "loss": 1.1162, + "step": 20937 + }, + { + "epoch": 0.67, + "learning_rate": 5.199536798452815e-06, + "loss": 0.939, + "step": 20938 + }, + { + "epoch": 0.67, + "learning_rate": 5.198628172504691e-06, + "loss": 0.8867, + "step": 20939 + }, + { + "epoch": 0.67, + "learning_rate": 5.197719598069971e-06, + "loss": 0.938, + "step": 20940 + }, + { + "epoch": 0.67, + "learning_rate": 5.196811075158402e-06, + "loss": 1.021, + "step": 20941 + }, + { + "epoch": 0.67, + "learning_rate": 5.195902603779738e-06, + "loss": 0.8813, + "step": 20942 + }, + { + "epoch": 0.67, + "learning_rate": 5.1949941839437175e-06, + "loss": 0.9136, + "step": 20943 + }, + { + "epoch": 0.67, + "learning_rate": 5.194085815660096e-06, + "loss": 1.0527, + "step": 20944 + }, + { + "epoch": 0.67, + "learning_rate": 5.193177498938615e-06, + "loss": 0.958, + "step": 20945 + }, + { + "epoch": 0.67, + "learning_rate": 5.192269233789015e-06, + "loss": 0.9409, + "step": 20946 + }, + { + "epoch": 0.67, + "learning_rate": 5.19136102022105e-06, + "loss": 0.9521, + "step": 20947 + }, + { + "epoch": 0.67, + "learning_rate": 5.190452858244458e-06, + "loss": 0.9043, + "step": 20948 + }, + { + "epoch": 0.67, + "learning_rate": 5.189544747868981e-06, + "loss": 0.8035, + "step": 20949 + }, + { + "epoch": 0.67, + "learning_rate": 5.188636689104368e-06, + "loss": 0.9473, + "step": 20950 + }, + { + "epoch": 0.67, + "learning_rate": 5.187728681960359e-06, + "loss": 0.8936, + "step": 20951 + }, + { + "epoch": 0.67, + "learning_rate": 5.1868207264466905e-06, + "loss": 0.8735, + "step": 20952 + }, + { + "epoch": 0.67, + "learning_rate": 5.185912822573113e-06, + "loss": 0.8545, + "step": 20953 + }, + { + "epoch": 0.67, + "learning_rate": 5.185004970349364e-06, + "loss": 1.0859, + "step": 20954 + }, + { + "epoch": 0.67, + "learning_rate": 5.184097169785183e-06, + "loss": 0.9424, + "step": 20955 + }, + { + "epoch": 0.67, + "learning_rate": 5.183189420890304e-06, + "loss": 0.8799, + "step": 20956 + }, + { + "epoch": 0.67, + "learning_rate": 5.182281723674476e-06, + "loss": 1.0195, + "step": 20957 + }, + { + "epoch": 0.67, + "learning_rate": 5.181374078147429e-06, + "loss": 0.9155, + "step": 20958 + }, + { + "epoch": 0.67, + "learning_rate": 5.180466484318909e-06, + "loss": 0.9124, + "step": 20959 + }, + { + "epoch": 0.67, + "learning_rate": 5.179558942198646e-06, + "loss": 0.8711, + "step": 20960 + }, + { + "epoch": 0.67, + "learning_rate": 5.178651451796384e-06, + "loss": 0.832, + "step": 20961 + }, + { + "epoch": 0.67, + "learning_rate": 5.177744013121856e-06, + "loss": 1.064, + "step": 20962 + }, + { + "epoch": 0.67, + "learning_rate": 5.176836626184798e-06, + "loss": 1.0605, + "step": 20963 + }, + { + "epoch": 0.67, + "learning_rate": 5.175929290994941e-06, + "loss": 0.8398, + "step": 20964 + }, + { + "epoch": 0.67, + "learning_rate": 5.175022007562027e-06, + "loss": 1.0493, + "step": 20965 + }, + { + "epoch": 0.67, + "learning_rate": 5.174114775895784e-06, + "loss": 1.0312, + "step": 20966 + }, + { + "epoch": 0.67, + "learning_rate": 5.173207596005951e-06, + "loss": 0.9937, + "step": 20967 + }, + { + "epoch": 0.67, + "learning_rate": 5.172300467902256e-06, + "loss": 0.8745, + "step": 20968 + }, + { + "epoch": 0.67, + "learning_rate": 5.171393391594437e-06, + "loss": 0.9414, + "step": 20969 + }, + { + "epoch": 0.67, + "learning_rate": 5.170486367092224e-06, + "loss": 0.9482, + "step": 20970 + }, + { + "epoch": 0.67, + "learning_rate": 5.169579394405346e-06, + "loss": 1.0068, + "step": 20971 + }, + { + "epoch": 0.67, + "learning_rate": 5.168672473543532e-06, + "loss": 0.937, + "step": 20972 + }, + { + "epoch": 0.67, + "learning_rate": 5.167765604516519e-06, + "loss": 1.0073, + "step": 20973 + }, + { + "epoch": 0.67, + "learning_rate": 5.1668587873340284e-06, + "loss": 0.9932, + "step": 20974 + }, + { + "epoch": 0.67, + "learning_rate": 5.165952022005798e-06, + "loss": 1.0239, + "step": 20975 + }, + { + "epoch": 0.67, + "learning_rate": 5.16504530854155e-06, + "loss": 0.8726, + "step": 20976 + }, + { + "epoch": 0.67, + "learning_rate": 5.16413864695102e-06, + "loss": 0.9932, + "step": 20977 + }, + { + "epoch": 0.67, + "learning_rate": 5.1632320372439235e-06, + "loss": 0.9092, + "step": 20978 + }, + { + "epoch": 0.67, + "learning_rate": 5.162325479429999e-06, + "loss": 0.958, + "step": 20979 + }, + { + "epoch": 0.67, + "learning_rate": 5.161418973518964e-06, + "loss": 1.022, + "step": 20980 + }, + { + "epoch": 0.67, + "learning_rate": 5.160512519520551e-06, + "loss": 1.0732, + "step": 20981 + }, + { + "epoch": 0.67, + "learning_rate": 5.15960611744448e-06, + "loss": 0.8843, + "step": 20982 + }, + { + "epoch": 0.67, + "learning_rate": 5.1586997673004815e-06, + "loss": 1.0181, + "step": 20983 + }, + { + "epoch": 0.67, + "learning_rate": 5.157793469098277e-06, + "loss": 0.9888, + "step": 20984 + }, + { + "epoch": 0.67, + "learning_rate": 5.156887222847589e-06, + "loss": 0.8481, + "step": 20985 + }, + { + "epoch": 0.67, + "learning_rate": 5.155981028558137e-06, + "loss": 0.916, + "step": 20986 + }, + { + "epoch": 0.67, + "learning_rate": 5.155074886239651e-06, + "loss": 1.0264, + "step": 20987 + }, + { + "epoch": 0.67, + "learning_rate": 5.154168795901846e-06, + "loss": 0.9541, + "step": 20988 + }, + { + "epoch": 0.67, + "learning_rate": 5.1532627575544514e-06, + "loss": 0.9956, + "step": 20989 + }, + { + "epoch": 0.67, + "learning_rate": 5.152356771207182e-06, + "loss": 0.8953, + "step": 20990 + }, + { + "epoch": 0.67, + "learning_rate": 5.151450836869756e-06, + "loss": 0.9062, + "step": 20991 + }, + { + "epoch": 0.67, + "learning_rate": 5.150544954551899e-06, + "loss": 0.8413, + "step": 20992 + }, + { + "epoch": 0.67, + "learning_rate": 5.1496391242633305e-06, + "loss": 0.9834, + "step": 20993 + }, + { + "epoch": 0.67, + "learning_rate": 5.148733346013763e-06, + "loss": 0.7881, + "step": 20994 + }, + { + "epoch": 0.67, + "learning_rate": 5.147827619812915e-06, + "loss": 1.0811, + "step": 20995 + }, + { + "epoch": 0.67, + "learning_rate": 5.146921945670511e-06, + "loss": 0.9321, + "step": 20996 + }, + { + "epoch": 0.67, + "learning_rate": 5.146016323596259e-06, + "loss": 0.8257, + "step": 20997 + }, + { + "epoch": 0.67, + "learning_rate": 5.145110753599884e-06, + "loss": 0.8506, + "step": 20998 + }, + { + "epoch": 0.67, + "learning_rate": 5.144205235691094e-06, + "loss": 0.8901, + "step": 20999 + }, + { + "epoch": 0.67, + "learning_rate": 5.143299769879613e-06, + "loss": 0.9019, + "step": 21000 + }, + { + "epoch": 0.67, + "learning_rate": 5.142394356175145e-06, + "loss": 0.9595, + "step": 21001 + }, + { + "epoch": 0.67, + "learning_rate": 5.141488994587412e-06, + "loss": 0.8794, + "step": 21002 + }, + { + "epoch": 0.67, + "learning_rate": 5.140583685126122e-06, + "loss": 0.9263, + "step": 21003 + }, + { + "epoch": 0.67, + "learning_rate": 5.139678427800994e-06, + "loss": 0.8989, + "step": 21004 + }, + { + "epoch": 0.67, + "learning_rate": 5.1387732226217335e-06, + "loss": 0.9663, + "step": 21005 + }, + { + "epoch": 0.67, + "learning_rate": 5.137868069598061e-06, + "loss": 0.8804, + "step": 21006 + }, + { + "epoch": 0.67, + "learning_rate": 5.136962968739683e-06, + "loss": 0.9985, + "step": 21007 + }, + { + "epoch": 0.67, + "learning_rate": 5.136057920056309e-06, + "loss": 0.854, + "step": 21008 + }, + { + "epoch": 0.67, + "learning_rate": 5.135152923557647e-06, + "loss": 0.9985, + "step": 21009 + }, + { + "epoch": 0.67, + "learning_rate": 5.134247979253414e-06, + "loss": 0.9834, + "step": 21010 + }, + { + "epoch": 0.67, + "learning_rate": 5.133343087153311e-06, + "loss": 0.8306, + "step": 21011 + }, + { + "epoch": 0.67, + "learning_rate": 5.132438247267054e-06, + "loss": 0.9512, + "step": 21012 + }, + { + "epoch": 0.67, + "learning_rate": 5.1315334596043456e-06, + "loss": 0.9229, + "step": 21013 + }, + { + "epoch": 0.67, + "learning_rate": 5.130628724174896e-06, + "loss": 0.9688, + "step": 21014 + }, + { + "epoch": 0.67, + "learning_rate": 5.129724040988413e-06, + "loss": 0.9072, + "step": 21015 + }, + { + "epoch": 0.67, + "learning_rate": 5.1288194100545996e-06, + "loss": 0.9839, + "step": 21016 + }, + { + "epoch": 0.67, + "learning_rate": 5.12791483138316e-06, + "loss": 0.813, + "step": 21017 + }, + { + "epoch": 0.67, + "learning_rate": 5.127010304983806e-06, + "loss": 1.0137, + "step": 21018 + }, + { + "epoch": 0.67, + "learning_rate": 5.126105830866234e-06, + "loss": 0.895, + "step": 21019 + }, + { + "epoch": 0.67, + "learning_rate": 5.125201409040156e-06, + "loss": 0.9453, + "step": 21020 + }, + { + "epoch": 0.67, + "learning_rate": 5.124297039515267e-06, + "loss": 0.9287, + "step": 21021 + }, + { + "epoch": 0.67, + "learning_rate": 5.123392722301283e-06, + "loss": 0.9414, + "step": 21022 + }, + { + "epoch": 0.67, + "learning_rate": 5.122488457407892e-06, + "loss": 1.0186, + "step": 21023 + }, + { + "epoch": 0.67, + "learning_rate": 5.121584244844804e-06, + "loss": 0.4998, + "step": 21024 + }, + { + "epoch": 0.67, + "learning_rate": 5.120680084621714e-06, + "loss": 0.9624, + "step": 21025 + }, + { + "epoch": 0.67, + "learning_rate": 5.1197759767483305e-06, + "loss": 0.918, + "step": 21026 + }, + { + "epoch": 0.67, + "learning_rate": 5.1188719212343456e-06, + "loss": 1.0483, + "step": 21027 + }, + { + "epoch": 0.67, + "learning_rate": 5.117967918089466e-06, + "loss": 1.0029, + "step": 21028 + }, + { + "epoch": 0.67, + "learning_rate": 5.117063967323389e-06, + "loss": 0.9775, + "step": 21029 + }, + { + "epoch": 0.67, + "learning_rate": 5.11616006894581e-06, + "loss": 1.061, + "step": 21030 + }, + { + "epoch": 0.67, + "learning_rate": 5.115256222966426e-06, + "loss": 0.9492, + "step": 21031 + }, + { + "epoch": 0.67, + "learning_rate": 5.11435242939494e-06, + "loss": 0.9487, + "step": 21032 + }, + { + "epoch": 0.67, + "learning_rate": 5.113448688241045e-06, + "loss": 1.0415, + "step": 21033 + }, + { + "epoch": 0.67, + "learning_rate": 5.112544999514433e-06, + "loss": 0.9341, + "step": 21034 + }, + { + "epoch": 0.67, + "learning_rate": 5.111641363224808e-06, + "loss": 0.8726, + "step": 21035 + }, + { + "epoch": 0.67, + "learning_rate": 5.110737779381858e-06, + "loss": 0.981, + "step": 21036 + }, + { + "epoch": 0.67, + "learning_rate": 5.109834247995283e-06, + "loss": 0.9326, + "step": 21037 + }, + { + "epoch": 0.67, + "learning_rate": 5.108930769074776e-06, + "loss": 0.937, + "step": 21038 + }, + { + "epoch": 0.67, + "learning_rate": 5.108027342630027e-06, + "loss": 0.8779, + "step": 21039 + }, + { + "epoch": 0.67, + "learning_rate": 5.107123968670728e-06, + "loss": 0.9307, + "step": 21040 + }, + { + "epoch": 0.67, + "learning_rate": 5.1062206472065765e-06, + "loss": 0.5432, + "step": 21041 + }, + { + "epoch": 0.67, + "learning_rate": 5.105317378247258e-06, + "loss": 0.8813, + "step": 21042 + }, + { + "epoch": 0.67, + "learning_rate": 5.10441416180247e-06, + "loss": 1.0767, + "step": 21043 + }, + { + "epoch": 0.67, + "learning_rate": 5.103510997881896e-06, + "loss": 0.9023, + "step": 21044 + }, + { + "epoch": 0.67, + "learning_rate": 5.102607886495239e-06, + "loss": 0.9053, + "step": 21045 + }, + { + "epoch": 0.67, + "learning_rate": 5.10170482765217e-06, + "loss": 0.9409, + "step": 21046 + }, + { + "epoch": 0.67, + "learning_rate": 5.100801821362391e-06, + "loss": 0.9619, + "step": 21047 + }, + { + "epoch": 0.67, + "learning_rate": 5.099898867635584e-06, + "loss": 0.8906, + "step": 21048 + }, + { + "epoch": 0.67, + "learning_rate": 5.0989959664814414e-06, + "loss": 0.9155, + "step": 21049 + }, + { + "epoch": 0.67, + "learning_rate": 5.098093117909644e-06, + "loss": 0.9907, + "step": 21050 + }, + { + "epoch": 0.67, + "learning_rate": 5.097190321929887e-06, + "loss": 1.0195, + "step": 21051 + }, + { + "epoch": 0.67, + "learning_rate": 5.096287578551847e-06, + "loss": 1.043, + "step": 21052 + }, + { + "epoch": 0.67, + "learning_rate": 5.095384887785222e-06, + "loss": 0.9053, + "step": 21053 + }, + { + "epoch": 0.67, + "learning_rate": 5.094482249639683e-06, + "loss": 0.9824, + "step": 21054 + }, + { + "epoch": 0.67, + "learning_rate": 5.093579664124921e-06, + "loss": 1.2188, + "step": 21055 + }, + { + "epoch": 0.67, + "learning_rate": 5.092677131250617e-06, + "loss": 1.0386, + "step": 21056 + }, + { + "epoch": 0.67, + "learning_rate": 5.09177465102646e-06, + "loss": 0.9639, + "step": 21057 + }, + { + "epoch": 0.67, + "learning_rate": 5.0908722234621255e-06, + "loss": 0.9751, + "step": 21058 + }, + { + "epoch": 0.67, + "learning_rate": 5.089969848567302e-06, + "loss": 1.0327, + "step": 21059 + }, + { + "epoch": 0.67, + "learning_rate": 5.089067526351667e-06, + "loss": 0.5144, + "step": 21060 + }, + { + "epoch": 0.67, + "learning_rate": 5.088165256824904e-06, + "loss": 1.0186, + "step": 21061 + }, + { + "epoch": 0.67, + "learning_rate": 5.087263039996688e-06, + "loss": 0.9175, + "step": 21062 + }, + { + "epoch": 0.67, + "learning_rate": 5.086360875876704e-06, + "loss": 0.8638, + "step": 21063 + }, + { + "epoch": 0.67, + "learning_rate": 5.085458764474628e-06, + "loss": 0.9863, + "step": 21064 + }, + { + "epoch": 0.67, + "learning_rate": 5.084556705800143e-06, + "loss": 1.0264, + "step": 21065 + }, + { + "epoch": 0.67, + "learning_rate": 5.083654699862921e-06, + "loss": 0.957, + "step": 21066 + }, + { + "epoch": 0.67, + "learning_rate": 5.082752746672647e-06, + "loss": 0.9883, + "step": 21067 + }, + { + "epoch": 0.67, + "learning_rate": 5.081850846238994e-06, + "loss": 1.0073, + "step": 21068 + }, + { + "epoch": 0.67, + "learning_rate": 5.080948998571638e-06, + "loss": 0.9268, + "step": 21069 + }, + { + "epoch": 0.67, + "learning_rate": 5.08004720368025e-06, + "loss": 0.9268, + "step": 21070 + }, + { + "epoch": 0.67, + "learning_rate": 5.079145461574516e-06, + "loss": 0.8403, + "step": 21071 + }, + { + "epoch": 0.67, + "learning_rate": 5.078243772264104e-06, + "loss": 0.8901, + "step": 21072 + }, + { + "epoch": 0.67, + "learning_rate": 5.077342135758686e-06, + "loss": 1.0112, + "step": 21073 + }, + { + "epoch": 0.67, + "learning_rate": 5.076440552067943e-06, + "loss": 0.8755, + "step": 21074 + }, + { + "epoch": 0.67, + "learning_rate": 5.075539021201543e-06, + "loss": 0.8604, + "step": 21075 + }, + { + "epoch": 0.67, + "learning_rate": 5.074637543169155e-06, + "loss": 1.0078, + "step": 21076 + }, + { + "epoch": 0.67, + "learning_rate": 5.073736117980459e-06, + "loss": 1.0854, + "step": 21077 + }, + { + "epoch": 0.67, + "learning_rate": 5.072834745645123e-06, + "loss": 1.0854, + "step": 21078 + }, + { + "epoch": 0.67, + "learning_rate": 5.071933426172813e-06, + "loss": 0.9414, + "step": 21079 + }, + { + "epoch": 0.67, + "learning_rate": 5.071032159573207e-06, + "loss": 1.0698, + "step": 21080 + }, + { + "epoch": 0.67, + "learning_rate": 5.0701309458559665e-06, + "loss": 0.916, + "step": 21081 + }, + { + "epoch": 0.67, + "learning_rate": 5.069229785030769e-06, + "loss": 0.9595, + "step": 21082 + }, + { + "epoch": 0.67, + "learning_rate": 5.068328677107279e-06, + "loss": 0.9478, + "step": 21083 + }, + { + "epoch": 0.67, + "learning_rate": 5.067427622095165e-06, + "loss": 1.0566, + "step": 21084 + }, + { + "epoch": 0.67, + "learning_rate": 5.066526620004089e-06, + "loss": 0.9204, + "step": 21085 + }, + { + "epoch": 0.67, + "learning_rate": 5.065625670843726e-06, + "loss": 0.8315, + "step": 21086 + }, + { + "epoch": 0.67, + "learning_rate": 5.0647247746237325e-06, + "loss": 0.9048, + "step": 21087 + }, + { + "epoch": 0.67, + "learning_rate": 5.063823931353786e-06, + "loss": 0.8921, + "step": 21088 + }, + { + "epoch": 0.67, + "learning_rate": 5.062923141043543e-06, + "loss": 0.9297, + "step": 21089 + }, + { + "epoch": 0.67, + "learning_rate": 5.0620224037026766e-06, + "loss": 0.8154, + "step": 21090 + }, + { + "epoch": 0.67, + "learning_rate": 5.061121719340838e-06, + "loss": 0.8872, + "step": 21091 + }, + { + "epoch": 0.67, + "learning_rate": 5.060221087967701e-06, + "loss": 0.8716, + "step": 21092 + }, + { + "epoch": 0.67, + "learning_rate": 5.05932050959292e-06, + "loss": 0.9966, + "step": 21093 + }, + { + "epoch": 0.67, + "learning_rate": 5.0584199842261665e-06, + "loss": 0.9365, + "step": 21094 + }, + { + "epoch": 0.67, + "learning_rate": 5.057519511877091e-06, + "loss": 0.8628, + "step": 21095 + }, + { + "epoch": 0.67, + "learning_rate": 5.056619092555368e-06, + "loss": 1.043, + "step": 21096 + }, + { + "epoch": 0.67, + "learning_rate": 5.055718726270647e-06, + "loss": 1.0605, + "step": 21097 + }, + { + "epoch": 0.67, + "learning_rate": 5.054818413032598e-06, + "loss": 0.959, + "step": 21098 + }, + { + "epoch": 0.67, + "learning_rate": 5.053918152850868e-06, + "loss": 0.813, + "step": 21099 + }, + { + "epoch": 0.67, + "learning_rate": 5.053017945735125e-06, + "loss": 0.8711, + "step": 21100 + }, + { + "epoch": 0.67, + "learning_rate": 5.052117791695023e-06, + "loss": 0.896, + "step": 21101 + }, + { + "epoch": 0.67, + "learning_rate": 5.0512176907402225e-06, + "loss": 0.8271, + "step": 21102 + }, + { + "epoch": 0.67, + "learning_rate": 5.0503176428803756e-06, + "loss": 0.8643, + "step": 21103 + }, + { + "epoch": 0.67, + "learning_rate": 5.049417648125147e-06, + "loss": 0.8745, + "step": 21104 + }, + { + "epoch": 0.67, + "learning_rate": 5.048517706484188e-06, + "loss": 0.9043, + "step": 21105 + }, + { + "epoch": 0.67, + "learning_rate": 5.047617817967153e-06, + "loss": 0.9795, + "step": 21106 + }, + { + "epoch": 0.68, + "learning_rate": 5.046717982583695e-06, + "loss": 0.8887, + "step": 21107 + }, + { + "epoch": 0.68, + "learning_rate": 5.045818200343474e-06, + "loss": 0.8359, + "step": 21108 + }, + { + "epoch": 0.68, + "learning_rate": 5.044918471256137e-06, + "loss": 0.9341, + "step": 21109 + }, + { + "epoch": 0.68, + "learning_rate": 5.044018795331345e-06, + "loss": 0.8809, + "step": 21110 + }, + { + "epoch": 0.68, + "learning_rate": 5.043119172578747e-06, + "loss": 0.9702, + "step": 21111 + }, + { + "epoch": 0.68, + "learning_rate": 5.04221960300799e-06, + "loss": 0.7705, + "step": 21112 + }, + { + "epoch": 0.68, + "learning_rate": 5.041320086628731e-06, + "loss": 0.9844, + "step": 21113 + }, + { + "epoch": 0.68, + "learning_rate": 5.040420623450622e-06, + "loss": 0.9565, + "step": 21114 + }, + { + "epoch": 0.68, + "learning_rate": 5.039521213483307e-06, + "loss": 0.8879, + "step": 21115 + }, + { + "epoch": 0.68, + "learning_rate": 5.0386218567364435e-06, + "loss": 1.0269, + "step": 21116 + }, + { + "epoch": 0.68, + "learning_rate": 5.037722553219675e-06, + "loss": 1.0483, + "step": 21117 + }, + { + "epoch": 0.68, + "learning_rate": 5.036823302942648e-06, + "loss": 0.9727, + "step": 21118 + }, + { + "epoch": 0.68, + "learning_rate": 5.035924105915018e-06, + "loss": 0.896, + "step": 21119 + }, + { + "epoch": 0.68, + "learning_rate": 5.035024962146429e-06, + "loss": 0.874, + "step": 21120 + }, + { + "epoch": 0.68, + "learning_rate": 5.0341258716465235e-06, + "loss": 0.9434, + "step": 21121 + }, + { + "epoch": 0.68, + "learning_rate": 5.033226834424955e-06, + "loss": 0.9629, + "step": 21122 + }, + { + "epoch": 0.68, + "learning_rate": 5.032327850491366e-06, + "loss": 0.8672, + "step": 21123 + }, + { + "epoch": 0.68, + "learning_rate": 5.031428919855396e-06, + "loss": 0.8945, + "step": 21124 + }, + { + "epoch": 0.68, + "learning_rate": 5.030530042526699e-06, + "loss": 0.8521, + "step": 21125 + }, + { + "epoch": 0.68, + "learning_rate": 5.029631218514911e-06, + "loss": 0.9106, + "step": 21126 + }, + { + "epoch": 0.68, + "learning_rate": 5.028732447829683e-06, + "loss": 0.8833, + "step": 21127 + }, + { + "epoch": 0.68, + "learning_rate": 5.027833730480654e-06, + "loss": 0.8696, + "step": 21128 + }, + { + "epoch": 0.68, + "learning_rate": 5.026935066477467e-06, + "loss": 0.856, + "step": 21129 + }, + { + "epoch": 0.68, + "learning_rate": 5.026036455829757e-06, + "loss": 0.9863, + "step": 21130 + }, + { + "epoch": 0.68, + "learning_rate": 5.025137898547177e-06, + "loss": 0.8657, + "step": 21131 + }, + { + "epoch": 0.68, + "learning_rate": 5.024239394639356e-06, + "loss": 0.9771, + "step": 21132 + }, + { + "epoch": 0.68, + "learning_rate": 5.0233409441159445e-06, + "loss": 0.9102, + "step": 21133 + }, + { + "epoch": 0.68, + "learning_rate": 5.022442546986571e-06, + "loss": 0.9653, + "step": 21134 + }, + { + "epoch": 0.68, + "learning_rate": 5.021544203260885e-06, + "loss": 0.8442, + "step": 21135 + }, + { + "epoch": 0.68, + "learning_rate": 5.0206459129485206e-06, + "loss": 0.9121, + "step": 21136 + }, + { + "epoch": 0.68, + "learning_rate": 5.0197476760591146e-06, + "loss": 0.9209, + "step": 21137 + }, + { + "epoch": 0.68, + "learning_rate": 5.018849492602299e-06, + "loss": 0.9038, + "step": 21138 + }, + { + "epoch": 0.68, + "learning_rate": 5.017951362587721e-06, + "loss": 1.0361, + "step": 21139 + }, + { + "epoch": 0.68, + "learning_rate": 5.017053286025005e-06, + "loss": 0.9307, + "step": 21140 + }, + { + "epoch": 0.68, + "learning_rate": 5.016155262923798e-06, + "loss": 1.0371, + "step": 21141 + }, + { + "epoch": 0.68, + "learning_rate": 5.015257293293725e-06, + "loss": 1.0488, + "step": 21142 + }, + { + "epoch": 0.68, + "learning_rate": 5.014359377144431e-06, + "loss": 0.9346, + "step": 21143 + }, + { + "epoch": 0.68, + "learning_rate": 5.013461514485536e-06, + "loss": 1.0181, + "step": 21144 + }, + { + "epoch": 0.68, + "learning_rate": 5.012563705326685e-06, + "loss": 0.9199, + "step": 21145 + }, + { + "epoch": 0.68, + "learning_rate": 5.011665949677501e-06, + "loss": 1.0078, + "step": 21146 + }, + { + "epoch": 0.68, + "learning_rate": 5.010768247547624e-06, + "loss": 1.0332, + "step": 21147 + }, + { + "epoch": 0.68, + "learning_rate": 5.0098705989466774e-06, + "loss": 0.9609, + "step": 21148 + }, + { + "epoch": 0.68, + "learning_rate": 5.0089730038843035e-06, + "loss": 0.8616, + "step": 21149 + }, + { + "epoch": 0.68, + "learning_rate": 5.008075462370119e-06, + "loss": 0.9624, + "step": 21150 + }, + { + "epoch": 0.68, + "learning_rate": 5.007177974413768e-06, + "loss": 0.9585, + "step": 21151 + }, + { + "epoch": 0.68, + "learning_rate": 5.006280540024864e-06, + "loss": 0.9922, + "step": 21152 + }, + { + "epoch": 0.68, + "learning_rate": 5.0053831592130465e-06, + "loss": 0.9341, + "step": 21153 + }, + { + "epoch": 0.68, + "learning_rate": 5.004485831987938e-06, + "loss": 0.8804, + "step": 21154 + }, + { + "epoch": 0.68, + "learning_rate": 5.003588558359169e-06, + "loss": 0.9272, + "step": 21155 + }, + { + "epoch": 0.68, + "learning_rate": 5.002691338336365e-06, + "loss": 0.9106, + "step": 21156 + }, + { + "epoch": 0.68, + "learning_rate": 5.0017941719291505e-06, + "loss": 0.9727, + "step": 21157 + }, + { + "epoch": 0.68, + "learning_rate": 5.000897059147156e-06, + "loss": 0.9023, + "step": 21158 + }, + { + "epoch": 0.68, + "learning_rate": 5.000000000000003e-06, + "loss": 0.8774, + "step": 21159 + }, + { + "epoch": 0.68, + "learning_rate": 4.9991029944973114e-06, + "loss": 0.8535, + "step": 21160 + }, + { + "epoch": 0.68, + "learning_rate": 4.998206042648716e-06, + "loss": 0.8979, + "step": 21161 + }, + { + "epoch": 0.68, + "learning_rate": 4.997309144463832e-06, + "loss": 0.9937, + "step": 21162 + }, + { + "epoch": 0.68, + "learning_rate": 4.996412299952281e-06, + "loss": 0.9551, + "step": 21163 + }, + { + "epoch": 0.68, + "learning_rate": 4.995515509123693e-06, + "loss": 0.9182, + "step": 21164 + }, + { + "epoch": 0.68, + "learning_rate": 4.994618771987679e-06, + "loss": 1.0752, + "step": 21165 + }, + { + "epoch": 0.68, + "learning_rate": 4.993722088553872e-06, + "loss": 0.9531, + "step": 21166 + }, + { + "epoch": 0.68, + "learning_rate": 4.992825458831885e-06, + "loss": 0.9604, + "step": 21167 + }, + { + "epoch": 0.68, + "learning_rate": 4.991928882831339e-06, + "loss": 1.0552, + "step": 21168 + }, + { + "epoch": 0.68, + "learning_rate": 4.99103236056185e-06, + "loss": 0.875, + "step": 21169 + }, + { + "epoch": 0.68, + "learning_rate": 4.990135892033045e-06, + "loss": 0.9663, + "step": 21170 + }, + { + "epoch": 0.68, + "learning_rate": 4.989239477254533e-06, + "loss": 1.0171, + "step": 21171 + }, + { + "epoch": 0.68, + "learning_rate": 4.988343116235939e-06, + "loss": 0.9951, + "step": 21172 + }, + { + "epoch": 0.68, + "learning_rate": 4.987446808986877e-06, + "loss": 0.9727, + "step": 21173 + }, + { + "epoch": 0.68, + "learning_rate": 4.986550555516962e-06, + "loss": 0.8218, + "step": 21174 + }, + { + "epoch": 0.68, + "learning_rate": 4.985654355835808e-06, + "loss": 1.0166, + "step": 21175 + }, + { + "epoch": 0.68, + "learning_rate": 4.984758209953038e-06, + "loss": 0.9331, + "step": 21176 + }, + { + "epoch": 0.68, + "learning_rate": 4.983862117878257e-06, + "loss": 0.9341, + "step": 21177 + }, + { + "epoch": 0.68, + "learning_rate": 4.982966079621088e-06, + "loss": 0.9326, + "step": 21178 + }, + { + "epoch": 0.68, + "learning_rate": 4.982070095191136e-06, + "loss": 1.0674, + "step": 21179 + }, + { + "epoch": 0.68, + "learning_rate": 4.981174164598023e-06, + "loss": 1.0396, + "step": 21180 + }, + { + "epoch": 0.68, + "learning_rate": 4.9802782878513555e-06, + "loss": 0.9453, + "step": 21181 + }, + { + "epoch": 0.68, + "learning_rate": 4.979382464960747e-06, + "loss": 0.9907, + "step": 21182 + }, + { + "epoch": 0.68, + "learning_rate": 4.978486695935805e-06, + "loss": 0.8496, + "step": 21183 + }, + { + "epoch": 0.68, + "learning_rate": 4.977590980786145e-06, + "loss": 0.9111, + "step": 21184 + }, + { + "epoch": 0.68, + "learning_rate": 4.976695319521373e-06, + "loss": 0.9268, + "step": 21185 + }, + { + "epoch": 0.68, + "learning_rate": 4.975799712151104e-06, + "loss": 0.8501, + "step": 21186 + }, + { + "epoch": 0.68, + "learning_rate": 4.974904158684941e-06, + "loss": 0.9458, + "step": 21187 + }, + { + "epoch": 0.68, + "learning_rate": 4.974008659132501e-06, + "loss": 0.5317, + "step": 21188 + }, + { + "epoch": 0.68, + "learning_rate": 4.973113213503379e-06, + "loss": 0.9185, + "step": 21189 + }, + { + "epoch": 0.68, + "learning_rate": 4.9722178218071925e-06, + "loss": 0.9092, + "step": 21190 + }, + { + "epoch": 0.68, + "learning_rate": 4.97132248405354e-06, + "loss": 1.0464, + "step": 21191 + }, + { + "epoch": 0.68, + "learning_rate": 4.970427200252035e-06, + "loss": 0.9785, + "step": 21192 + }, + { + "epoch": 0.68, + "learning_rate": 4.9695319704122765e-06, + "loss": 0.8574, + "step": 21193 + }, + { + "epoch": 0.68, + "learning_rate": 4.968636794543876e-06, + "loss": 0.9595, + "step": 21194 + }, + { + "epoch": 0.68, + "learning_rate": 4.9677416726564356e-06, + "loss": 0.8862, + "step": 21195 + }, + { + "epoch": 0.68, + "learning_rate": 4.966846604759556e-06, + "loss": 0.8872, + "step": 21196 + }, + { + "epoch": 0.68, + "learning_rate": 4.96595159086284e-06, + "loss": 0.9155, + "step": 21197 + }, + { + "epoch": 0.68, + "learning_rate": 4.965056630975894e-06, + "loss": 0.9653, + "step": 21198 + }, + { + "epoch": 0.68, + "learning_rate": 4.964161725108314e-06, + "loss": 0.8489, + "step": 21199 + }, + { + "epoch": 0.68, + "learning_rate": 4.9632668732697105e-06, + "loss": 0.9414, + "step": 21200 + }, + { + "epoch": 0.68, + "learning_rate": 4.962372075469678e-06, + "loss": 0.9131, + "step": 21201 + }, + { + "epoch": 0.68, + "learning_rate": 4.9614773317178146e-06, + "loss": 0.9165, + "step": 21202 + }, + { + "epoch": 0.68, + "learning_rate": 4.960582642023727e-06, + "loss": 0.9688, + "step": 21203 + }, + { + "epoch": 0.68, + "learning_rate": 4.9596880063970095e-06, + "loss": 0.8779, + "step": 21204 + }, + { + "epoch": 0.68, + "learning_rate": 4.958793424847258e-06, + "loss": 0.8101, + "step": 21205 + }, + { + "epoch": 0.68, + "learning_rate": 4.957898897384078e-06, + "loss": 0.9995, + "step": 21206 + }, + { + "epoch": 0.68, + "learning_rate": 4.957004424017061e-06, + "loss": 0.9761, + "step": 21207 + }, + { + "epoch": 0.68, + "learning_rate": 4.956110004755803e-06, + "loss": 0.9907, + "step": 21208 + }, + { + "epoch": 0.68, + "learning_rate": 4.955215639609906e-06, + "loss": 0.8638, + "step": 21209 + }, + { + "epoch": 0.68, + "learning_rate": 4.954321328588957e-06, + "loss": 0.9434, + "step": 21210 + }, + { + "epoch": 0.68, + "learning_rate": 4.953427071702564e-06, + "loss": 0.9849, + "step": 21211 + }, + { + "epoch": 0.68, + "learning_rate": 4.952532868960306e-06, + "loss": 0.9722, + "step": 21212 + }, + { + "epoch": 0.68, + "learning_rate": 4.951638720371787e-06, + "loss": 0.8779, + "step": 21213 + }, + { + "epoch": 0.68, + "learning_rate": 4.950744625946595e-06, + "loss": 0.9551, + "step": 21214 + }, + { + "epoch": 0.68, + "learning_rate": 4.949850585694328e-06, + "loss": 1.0361, + "step": 21215 + }, + { + "epoch": 0.68, + "learning_rate": 4.948956599624572e-06, + "loss": 1.0283, + "step": 21216 + }, + { + "epoch": 0.68, + "learning_rate": 4.948062667746923e-06, + "loss": 1.0034, + "step": 21217 + }, + { + "epoch": 0.68, + "learning_rate": 4.947168790070969e-06, + "loss": 1.0112, + "step": 21218 + }, + { + "epoch": 0.68, + "learning_rate": 4.946274966606307e-06, + "loss": 0.9531, + "step": 21219 + }, + { + "epoch": 0.68, + "learning_rate": 4.945381197362515e-06, + "loss": 1.0049, + "step": 21220 + }, + { + "epoch": 0.68, + "learning_rate": 4.944487482349193e-06, + "loss": 0.8779, + "step": 21221 + }, + { + "epoch": 0.68, + "learning_rate": 4.9435938215759205e-06, + "loss": 0.897, + "step": 21222 + }, + { + "epoch": 0.68, + "learning_rate": 4.942700215052294e-06, + "loss": 0.9141, + "step": 21223 + }, + { + "epoch": 0.68, + "learning_rate": 4.941806662787894e-06, + "loss": 0.8633, + "step": 21224 + }, + { + "epoch": 0.68, + "learning_rate": 4.940913164792312e-06, + "loss": 0.9736, + "step": 21225 + }, + { + "epoch": 0.68, + "learning_rate": 4.940019721075135e-06, + "loss": 0.9507, + "step": 21226 + }, + { + "epoch": 0.68, + "learning_rate": 4.939126331645946e-06, + "loss": 0.9155, + "step": 21227 + }, + { + "epoch": 0.68, + "learning_rate": 4.938232996514326e-06, + "loss": 0.9785, + "step": 21228 + }, + { + "epoch": 0.68, + "learning_rate": 4.937339715689867e-06, + "loss": 0.9888, + "step": 21229 + }, + { + "epoch": 0.68, + "learning_rate": 4.936446489182147e-06, + "loss": 0.8774, + "step": 21230 + }, + { + "epoch": 0.68, + "learning_rate": 4.935553317000756e-06, + "loss": 0.9116, + "step": 21231 + }, + { + "epoch": 0.68, + "learning_rate": 4.934660199155269e-06, + "loss": 0.9609, + "step": 21232 + }, + { + "epoch": 0.68, + "learning_rate": 4.933767135655275e-06, + "loss": 0.8599, + "step": 21233 + }, + { + "epoch": 0.68, + "learning_rate": 4.932874126510353e-06, + "loss": 0.8975, + "step": 21234 + }, + { + "epoch": 0.68, + "learning_rate": 4.931981171730084e-06, + "loss": 0.9775, + "step": 21235 + }, + { + "epoch": 0.68, + "learning_rate": 4.931088271324044e-06, + "loss": 0.8491, + "step": 21236 + }, + { + "epoch": 0.68, + "learning_rate": 4.93019542530182e-06, + "loss": 0.9326, + "step": 21237 + }, + { + "epoch": 0.68, + "learning_rate": 4.9293026336729856e-06, + "loss": 0.8276, + "step": 21238 + }, + { + "epoch": 0.68, + "learning_rate": 4.9284098964471246e-06, + "loss": 0.9609, + "step": 21239 + }, + { + "epoch": 0.68, + "learning_rate": 4.927517213633812e-06, + "loss": 1.0391, + "step": 21240 + }, + { + "epoch": 0.68, + "learning_rate": 4.926624585242628e-06, + "loss": 0.8809, + "step": 21241 + }, + { + "epoch": 0.68, + "learning_rate": 4.925732011283142e-06, + "loss": 0.9258, + "step": 21242 + }, + { + "epoch": 0.68, + "learning_rate": 4.9248394917649386e-06, + "loss": 0.8877, + "step": 21243 + }, + { + "epoch": 0.68, + "learning_rate": 4.923947026697589e-06, + "loss": 0.8872, + "step": 21244 + }, + { + "epoch": 0.68, + "learning_rate": 4.9230546160906715e-06, + "loss": 0.9341, + "step": 21245 + }, + { + "epoch": 0.68, + "learning_rate": 4.92216225995376e-06, + "loss": 0.9214, + "step": 21246 + }, + { + "epoch": 0.68, + "learning_rate": 4.921269958296424e-06, + "loss": 0.9565, + "step": 21247 + }, + { + "epoch": 0.68, + "learning_rate": 4.920377711128244e-06, + "loss": 0.9448, + "step": 21248 + }, + { + "epoch": 0.68, + "learning_rate": 4.919485518458789e-06, + "loss": 0.9629, + "step": 21249 + }, + { + "epoch": 0.68, + "learning_rate": 4.918593380297631e-06, + "loss": 0.9248, + "step": 21250 + }, + { + "epoch": 0.68, + "learning_rate": 4.9177012966543406e-06, + "loss": 1.0425, + "step": 21251 + }, + { + "epoch": 0.68, + "learning_rate": 4.916809267538493e-06, + "loss": 0.8477, + "step": 21252 + }, + { + "epoch": 0.68, + "learning_rate": 4.915917292959652e-06, + "loss": 1.0601, + "step": 21253 + }, + { + "epoch": 0.68, + "learning_rate": 4.915025372927396e-06, + "loss": 0.8916, + "step": 21254 + }, + { + "epoch": 0.68, + "learning_rate": 4.914133507451285e-06, + "loss": 0.8833, + "step": 21255 + }, + { + "epoch": 0.68, + "learning_rate": 4.913241696540903e-06, + "loss": 0.9487, + "step": 21256 + }, + { + "epoch": 0.68, + "learning_rate": 4.9123499402057975e-06, + "loss": 0.8389, + "step": 21257 + }, + { + "epoch": 0.68, + "learning_rate": 4.911458238455551e-06, + "loss": 0.9775, + "step": 21258 + }, + { + "epoch": 0.68, + "learning_rate": 4.910566591299722e-06, + "loss": 1.0005, + "step": 21259 + }, + { + "epoch": 0.68, + "learning_rate": 4.909674998747885e-06, + "loss": 0.9712, + "step": 21260 + }, + { + "epoch": 0.68, + "learning_rate": 4.9087834608095975e-06, + "loss": 0.9507, + "step": 21261 + }, + { + "epoch": 0.68, + "learning_rate": 4.9078919774944335e-06, + "loss": 0.9111, + "step": 21262 + }, + { + "epoch": 0.68, + "learning_rate": 4.907000548811948e-06, + "loss": 0.9966, + "step": 21263 + }, + { + "epoch": 0.68, + "learning_rate": 4.906109174771718e-06, + "loss": 1.0132, + "step": 21264 + }, + { + "epoch": 0.68, + "learning_rate": 4.905217855383291e-06, + "loss": 0.7971, + "step": 21265 + }, + { + "epoch": 0.68, + "learning_rate": 4.904326590656242e-06, + "loss": 0.8687, + "step": 21266 + }, + { + "epoch": 0.68, + "learning_rate": 4.903435380600126e-06, + "loss": 1.0229, + "step": 21267 + }, + { + "epoch": 0.68, + "learning_rate": 4.90254422522451e-06, + "loss": 0.9541, + "step": 21268 + }, + { + "epoch": 0.68, + "learning_rate": 4.90165312453895e-06, + "loss": 1.0757, + "step": 21269 + }, + { + "epoch": 0.68, + "learning_rate": 4.900762078553013e-06, + "loss": 0.9087, + "step": 21270 + }, + { + "epoch": 0.68, + "learning_rate": 4.899871087276255e-06, + "loss": 0.9775, + "step": 21271 + }, + { + "epoch": 0.68, + "learning_rate": 4.898980150718236e-06, + "loss": 0.8916, + "step": 21272 + }, + { + "epoch": 0.68, + "learning_rate": 4.898089268888511e-06, + "loss": 1.001, + "step": 21273 + }, + { + "epoch": 0.68, + "learning_rate": 4.897198441796645e-06, + "loss": 0.9829, + "step": 21274 + }, + { + "epoch": 0.68, + "learning_rate": 4.896307669452188e-06, + "loss": 0.8984, + "step": 21275 + }, + { + "epoch": 0.68, + "learning_rate": 4.895416951864705e-06, + "loss": 0.9751, + "step": 21276 + }, + { + "epoch": 0.68, + "learning_rate": 4.894526289043746e-06, + "loss": 0.5151, + "step": 21277 + }, + { + "epoch": 0.68, + "learning_rate": 4.893635680998873e-06, + "loss": 0.9043, + "step": 21278 + }, + { + "epoch": 0.68, + "learning_rate": 4.8927451277396365e-06, + "loss": 0.9824, + "step": 21279 + }, + { + "epoch": 0.68, + "learning_rate": 4.891854629275593e-06, + "loss": 0.9585, + "step": 21280 + }, + { + "epoch": 0.68, + "learning_rate": 4.890964185616293e-06, + "loss": 0.8813, + "step": 21281 + }, + { + "epoch": 0.68, + "learning_rate": 4.890073796771295e-06, + "loss": 0.7964, + "step": 21282 + }, + { + "epoch": 0.68, + "learning_rate": 4.889183462750147e-06, + "loss": 1.0093, + "step": 21283 + }, + { + "epoch": 0.68, + "learning_rate": 4.8882931835624085e-06, + "loss": 1.0024, + "step": 21284 + }, + { + "epoch": 0.68, + "learning_rate": 4.887402959217625e-06, + "loss": 0.4771, + "step": 21285 + }, + { + "epoch": 0.68, + "learning_rate": 4.886512789725348e-06, + "loss": 0.9434, + "step": 21286 + }, + { + "epoch": 0.68, + "learning_rate": 4.885622675095131e-06, + "loss": 0.9502, + "step": 21287 + }, + { + "epoch": 0.68, + "learning_rate": 4.884732615336524e-06, + "loss": 0.9976, + "step": 21288 + }, + { + "epoch": 0.68, + "learning_rate": 4.883842610459073e-06, + "loss": 0.9375, + "step": 21289 + }, + { + "epoch": 0.68, + "learning_rate": 4.882952660472326e-06, + "loss": 0.8462, + "step": 21290 + }, + { + "epoch": 0.68, + "learning_rate": 4.882062765385838e-06, + "loss": 0.8555, + "step": 21291 + }, + { + "epoch": 0.68, + "learning_rate": 4.881172925209147e-06, + "loss": 0.9946, + "step": 21292 + }, + { + "epoch": 0.68, + "learning_rate": 4.8802831399518105e-06, + "loss": 0.9536, + "step": 21293 + }, + { + "epoch": 0.68, + "learning_rate": 4.879393409623368e-06, + "loss": 0.9048, + "step": 21294 + }, + { + "epoch": 0.68, + "learning_rate": 4.878503734233369e-06, + "loss": 0.8652, + "step": 21295 + }, + { + "epoch": 0.68, + "learning_rate": 4.877614113791351e-06, + "loss": 0.9468, + "step": 21296 + }, + { + "epoch": 0.68, + "learning_rate": 4.876724548306869e-06, + "loss": 0.8818, + "step": 21297 + }, + { + "epoch": 0.68, + "learning_rate": 4.875835037789459e-06, + "loss": 1.0278, + "step": 21298 + }, + { + "epoch": 0.68, + "learning_rate": 4.874945582248671e-06, + "loss": 0.9185, + "step": 21299 + }, + { + "epoch": 0.68, + "learning_rate": 4.8740561816940425e-06, + "loss": 0.9907, + "step": 21300 + }, + { + "epoch": 0.68, + "learning_rate": 4.87316683613512e-06, + "loss": 0.9282, + "step": 21301 + }, + { + "epoch": 0.68, + "learning_rate": 4.872277545581444e-06, + "loss": 0.7888, + "step": 21302 + }, + { + "epoch": 0.68, + "learning_rate": 4.871388310042556e-06, + "loss": 0.4661, + "step": 21303 + }, + { + "epoch": 0.68, + "learning_rate": 4.8704991295279894e-06, + "loss": 1.0269, + "step": 21304 + }, + { + "epoch": 0.68, + "learning_rate": 4.869610004047295e-06, + "loss": 0.9277, + "step": 21305 + }, + { + "epoch": 0.68, + "learning_rate": 4.868720933610003e-06, + "loss": 0.9985, + "step": 21306 + }, + { + "epoch": 0.68, + "learning_rate": 4.867831918225661e-06, + "loss": 0.9165, + "step": 21307 + }, + { + "epoch": 0.68, + "learning_rate": 4.866942957903798e-06, + "loss": 1.0342, + "step": 21308 + }, + { + "epoch": 0.68, + "learning_rate": 4.866054052653963e-06, + "loss": 0.8691, + "step": 21309 + }, + { + "epoch": 0.68, + "learning_rate": 4.86516520248568e-06, + "loss": 0.8809, + "step": 21310 + }, + { + "epoch": 0.68, + "learning_rate": 4.864276407408495e-06, + "loss": 0.9556, + "step": 21311 + }, + { + "epoch": 0.68, + "learning_rate": 4.863387667431936e-06, + "loss": 1.002, + "step": 21312 + }, + { + "epoch": 0.68, + "learning_rate": 4.862498982565546e-06, + "loss": 1.1177, + "step": 21313 + }, + { + "epoch": 0.68, + "learning_rate": 4.861610352818852e-06, + "loss": 0.9609, + "step": 21314 + }, + { + "epoch": 0.68, + "learning_rate": 4.860721778201397e-06, + "loss": 0.9194, + "step": 21315 + }, + { + "epoch": 0.68, + "learning_rate": 4.859833258722705e-06, + "loss": 0.8843, + "step": 21316 + }, + { + "epoch": 0.68, + "learning_rate": 4.858944794392321e-06, + "loss": 0.8823, + "step": 21317 + }, + { + "epoch": 0.68, + "learning_rate": 4.858056385219761e-06, + "loss": 0.895, + "step": 21318 + }, + { + "epoch": 0.68, + "learning_rate": 4.85716803121457e-06, + "loss": 0.875, + "step": 21319 + }, + { + "epoch": 0.68, + "learning_rate": 4.85627973238627e-06, + "loss": 0.9824, + "step": 21320 + }, + { + "epoch": 0.68, + "learning_rate": 4.8553914887443986e-06, + "loss": 0.9902, + "step": 21321 + }, + { + "epoch": 0.68, + "learning_rate": 4.854503300298481e-06, + "loss": 1.1431, + "step": 21322 + }, + { + "epoch": 0.68, + "learning_rate": 4.85361516705805e-06, + "loss": 0.9189, + "step": 21323 + }, + { + "epoch": 0.68, + "learning_rate": 4.852727089032634e-06, + "loss": 1.0151, + "step": 21324 + }, + { + "epoch": 0.68, + "learning_rate": 4.851839066231759e-06, + "loss": 0.9614, + "step": 21325 + }, + { + "epoch": 0.68, + "learning_rate": 4.850951098664949e-06, + "loss": 0.9683, + "step": 21326 + }, + { + "epoch": 0.68, + "learning_rate": 4.850063186341739e-06, + "loss": 0.7981, + "step": 21327 + }, + { + "epoch": 0.68, + "learning_rate": 4.849175329271652e-06, + "loss": 0.8975, + "step": 21328 + }, + { + "epoch": 0.68, + "learning_rate": 4.848287527464208e-06, + "loss": 0.8638, + "step": 21329 + }, + { + "epoch": 0.68, + "learning_rate": 4.847399780928943e-06, + "loss": 0.9312, + "step": 21330 + }, + { + "epoch": 0.68, + "learning_rate": 4.84651208967537e-06, + "loss": 0.8628, + "step": 21331 + }, + { + "epoch": 0.68, + "learning_rate": 4.845624453713023e-06, + "loss": 0.8633, + "step": 21332 + }, + { + "epoch": 0.68, + "learning_rate": 4.844736873051422e-06, + "loss": 0.8606, + "step": 21333 + }, + { + "epoch": 0.68, + "learning_rate": 4.8438493477000885e-06, + "loss": 0.4924, + "step": 21334 + }, + { + "epoch": 0.68, + "learning_rate": 4.842961877668541e-06, + "loss": 0.9751, + "step": 21335 + }, + { + "epoch": 0.68, + "learning_rate": 4.8420744629663076e-06, + "loss": 0.9326, + "step": 21336 + }, + { + "epoch": 0.68, + "learning_rate": 4.8411871036029035e-06, + "loss": 0.9526, + "step": 21337 + }, + { + "epoch": 0.68, + "learning_rate": 4.840299799587857e-06, + "loss": 0.9307, + "step": 21338 + }, + { + "epoch": 0.68, + "learning_rate": 4.839412550930682e-06, + "loss": 1.0508, + "step": 21339 + }, + { + "epoch": 0.68, + "learning_rate": 4.8385253576409e-06, + "loss": 1.0342, + "step": 21340 + }, + { + "epoch": 0.68, + "learning_rate": 4.837638219728024e-06, + "loss": 0.9424, + "step": 21341 + }, + { + "epoch": 0.68, + "learning_rate": 4.83675113720158e-06, + "loss": 1.0464, + "step": 21342 + }, + { + "epoch": 0.68, + "learning_rate": 4.835864110071079e-06, + "loss": 1.0171, + "step": 21343 + }, + { + "epoch": 0.68, + "learning_rate": 4.8349771383460445e-06, + "loss": 0.9233, + "step": 21344 + }, + { + "epoch": 0.68, + "learning_rate": 4.834090222035984e-06, + "loss": 0.8887, + "step": 21345 + }, + { + "epoch": 0.68, + "learning_rate": 4.8332033611504224e-06, + "loss": 1.0273, + "step": 21346 + }, + { + "epoch": 0.68, + "learning_rate": 4.83231655569887e-06, + "loss": 0.9829, + "step": 21347 + }, + { + "epoch": 0.68, + "learning_rate": 4.831429805690842e-06, + "loss": 0.8303, + "step": 21348 + }, + { + "epoch": 0.68, + "learning_rate": 4.830543111135849e-06, + "loss": 0.8955, + "step": 21349 + }, + { + "epoch": 0.68, + "learning_rate": 4.829656472043411e-06, + "loss": 1.0146, + "step": 21350 + }, + { + "epoch": 0.68, + "learning_rate": 4.828769888423031e-06, + "loss": 0.4929, + "step": 21351 + }, + { + "epoch": 0.68, + "learning_rate": 4.827883360284233e-06, + "loss": 0.9863, + "step": 21352 + }, + { + "epoch": 0.68, + "learning_rate": 4.826996887636517e-06, + "loss": 0.8872, + "step": 21353 + }, + { + "epoch": 0.68, + "learning_rate": 4.826110470489403e-06, + "loss": 0.9658, + "step": 21354 + }, + { + "epoch": 0.68, + "learning_rate": 4.825224108852399e-06, + "loss": 0.9648, + "step": 21355 + }, + { + "epoch": 0.68, + "learning_rate": 4.824337802735012e-06, + "loss": 1.043, + "step": 21356 + }, + { + "epoch": 0.68, + "learning_rate": 4.8234515521467485e-06, + "loss": 0.8726, + "step": 21357 + }, + { + "epoch": 0.68, + "learning_rate": 4.822565357097124e-06, + "loss": 0.9985, + "step": 21358 + }, + { + "epoch": 0.68, + "learning_rate": 4.821679217595641e-06, + "loss": 0.8877, + "step": 21359 + }, + { + "epoch": 0.68, + "learning_rate": 4.820793133651811e-06, + "loss": 0.998, + "step": 21360 + }, + { + "epoch": 0.68, + "learning_rate": 4.819907105275136e-06, + "loss": 0.834, + "step": 21361 + }, + { + "epoch": 0.68, + "learning_rate": 4.819021132475131e-06, + "loss": 1.0444, + "step": 21362 + }, + { + "epoch": 0.68, + "learning_rate": 4.8181352152612885e-06, + "loss": 0.9995, + "step": 21363 + }, + { + "epoch": 0.68, + "learning_rate": 4.817249353643123e-06, + "loss": 0.8794, + "step": 21364 + }, + { + "epoch": 0.68, + "learning_rate": 4.8163635476301326e-06, + "loss": 0.8589, + "step": 21365 + }, + { + "epoch": 0.68, + "learning_rate": 4.815477797231829e-06, + "loss": 0.8906, + "step": 21366 + }, + { + "epoch": 0.68, + "learning_rate": 4.814592102457707e-06, + "loss": 0.9922, + "step": 21367 + }, + { + "epoch": 0.68, + "learning_rate": 4.813706463317276e-06, + "loss": 1.0063, + "step": 21368 + }, + { + "epoch": 0.68, + "learning_rate": 4.812820879820034e-06, + "loss": 0.9204, + "step": 21369 + }, + { + "epoch": 0.68, + "learning_rate": 4.811935351975484e-06, + "loss": 1.0708, + "step": 21370 + }, + { + "epoch": 0.68, + "learning_rate": 4.811049879793121e-06, + "loss": 0.9395, + "step": 21371 + }, + { + "epoch": 0.68, + "learning_rate": 4.8101644632824555e-06, + "loss": 0.8521, + "step": 21372 + }, + { + "epoch": 0.68, + "learning_rate": 4.80927910245298e-06, + "loss": 0.8813, + "step": 21373 + }, + { + "epoch": 0.68, + "learning_rate": 4.808393797314192e-06, + "loss": 0.9048, + "step": 21374 + }, + { + "epoch": 0.68, + "learning_rate": 4.8075085478755965e-06, + "loss": 0.9746, + "step": 21375 + }, + { + "epoch": 0.68, + "learning_rate": 4.806623354146684e-06, + "loss": 0.7998, + "step": 21376 + }, + { + "epoch": 0.68, + "learning_rate": 4.805738216136958e-06, + "loss": 0.8032, + "step": 21377 + }, + { + "epoch": 0.68, + "learning_rate": 4.804853133855912e-06, + "loss": 0.9634, + "step": 21378 + }, + { + "epoch": 0.68, + "learning_rate": 4.803968107313043e-06, + "loss": 0.9766, + "step": 21379 + }, + { + "epoch": 0.68, + "learning_rate": 4.803083136517842e-06, + "loss": 0.8979, + "step": 21380 + }, + { + "epoch": 0.68, + "learning_rate": 4.80219822147981e-06, + "loss": 0.9834, + "step": 21381 + }, + { + "epoch": 0.68, + "learning_rate": 4.801313362208435e-06, + "loss": 0.9419, + "step": 21382 + }, + { + "epoch": 0.68, + "learning_rate": 4.800428558713218e-06, + "loss": 0.8975, + "step": 21383 + }, + { + "epoch": 0.68, + "learning_rate": 4.799543811003645e-06, + "loss": 0.917, + "step": 21384 + }, + { + "epoch": 0.68, + "learning_rate": 4.798659119089219e-06, + "loss": 0.8003, + "step": 21385 + }, + { + "epoch": 0.68, + "learning_rate": 4.7977744829794145e-06, + "loss": 0.9067, + "step": 21386 + }, + { + "epoch": 0.68, + "learning_rate": 4.7968899026837375e-06, + "loss": 0.874, + "step": 21387 + }, + { + "epoch": 0.68, + "learning_rate": 4.796005378211669e-06, + "loss": 0.9155, + "step": 21388 + }, + { + "epoch": 0.68, + "learning_rate": 4.795120909572707e-06, + "loss": 0.9634, + "step": 21389 + }, + { + "epoch": 0.68, + "learning_rate": 4.7942364967763345e-06, + "loss": 0.9482, + "step": 21390 + }, + { + "epoch": 0.68, + "learning_rate": 4.793352139832046e-06, + "loss": 0.936, + "step": 21391 + }, + { + "epoch": 0.68, + "learning_rate": 4.792467838749326e-06, + "loss": 0.9263, + "step": 21392 + }, + { + "epoch": 0.68, + "learning_rate": 4.791583593537663e-06, + "loss": 0.8003, + "step": 21393 + }, + { + "epoch": 0.68, + "learning_rate": 4.79069940420654e-06, + "loss": 1.0796, + "step": 21394 + }, + { + "epoch": 0.68, + "learning_rate": 4.789815270765451e-06, + "loss": 0.8652, + "step": 21395 + }, + { + "epoch": 0.68, + "learning_rate": 4.788931193223875e-06, + "loss": 0.8525, + "step": 21396 + }, + { + "epoch": 0.68, + "learning_rate": 4.7880471715913025e-06, + "loss": 0.835, + "step": 21397 + }, + { + "epoch": 0.68, + "learning_rate": 4.787163205877211e-06, + "loss": 0.9307, + "step": 21398 + }, + { + "epoch": 0.68, + "learning_rate": 4.786279296091094e-06, + "loss": 1.0347, + "step": 21399 + }, + { + "epoch": 0.68, + "learning_rate": 4.7853954422424295e-06, + "loss": 0.9312, + "step": 21400 + }, + { + "epoch": 0.68, + "learning_rate": 4.784511644340701e-06, + "loss": 1.0171, + "step": 21401 + }, + { + "epoch": 0.68, + "learning_rate": 4.783627902395386e-06, + "loss": 0.8833, + "step": 21402 + }, + { + "epoch": 0.68, + "learning_rate": 4.782744216415974e-06, + "loss": 0.9585, + "step": 21403 + }, + { + "epoch": 0.68, + "learning_rate": 4.781860586411939e-06, + "loss": 0.9663, + "step": 21404 + }, + { + "epoch": 0.68, + "learning_rate": 4.780977012392769e-06, + "loss": 0.9062, + "step": 21405 + }, + { + "epoch": 0.68, + "learning_rate": 4.7800934943679355e-06, + "loss": 1.0073, + "step": 21406 + }, + { + "epoch": 0.68, + "learning_rate": 4.779210032346929e-06, + "loss": 0.9316, + "step": 21407 + }, + { + "epoch": 0.68, + "learning_rate": 4.778326626339213e-06, + "loss": 0.9873, + "step": 21408 + }, + { + "epoch": 0.68, + "learning_rate": 4.777443276354279e-06, + "loss": 1.0776, + "step": 21409 + }, + { + "epoch": 0.68, + "learning_rate": 4.776559982401593e-06, + "loss": 0.4788, + "step": 21410 + }, + { + "epoch": 0.68, + "learning_rate": 4.7756767444906415e-06, + "loss": 0.8408, + "step": 21411 + }, + { + "epoch": 0.68, + "learning_rate": 4.774793562630896e-06, + "loss": 1.0093, + "step": 21412 + }, + { + "epoch": 0.68, + "learning_rate": 4.773910436831829e-06, + "loss": 0.9668, + "step": 21413 + }, + { + "epoch": 0.68, + "learning_rate": 4.7730273671029235e-06, + "loss": 1.0298, + "step": 21414 + }, + { + "epoch": 0.68, + "learning_rate": 4.772144353453648e-06, + "loss": 0.9707, + "step": 21415 + }, + { + "epoch": 0.68, + "learning_rate": 4.771261395893476e-06, + "loss": 0.9863, + "step": 21416 + }, + { + "epoch": 0.68, + "learning_rate": 4.7703784944318845e-06, + "loss": 0.959, + "step": 21417 + }, + { + "epoch": 0.68, + "learning_rate": 4.769495649078345e-06, + "loss": 0.9775, + "step": 21418 + }, + { + "epoch": 0.68, + "learning_rate": 4.768612859842323e-06, + "loss": 0.9624, + "step": 21419 + }, + { + "epoch": 0.69, + "learning_rate": 4.767730126733301e-06, + "loss": 0.9663, + "step": 21420 + }, + { + "epoch": 0.69, + "learning_rate": 4.766847449760738e-06, + "loss": 0.9985, + "step": 21421 + }, + { + "epoch": 0.69, + "learning_rate": 4.765964828934115e-06, + "loss": 0.9951, + "step": 21422 + }, + { + "epoch": 0.69, + "learning_rate": 4.765082264262898e-06, + "loss": 1.0415, + "step": 21423 + }, + { + "epoch": 0.69, + "learning_rate": 4.764199755756552e-06, + "loss": 0.9106, + "step": 21424 + }, + { + "epoch": 0.69, + "learning_rate": 4.763317303424546e-06, + "loss": 0.8579, + "step": 21425 + }, + { + "epoch": 0.69, + "learning_rate": 4.762434907276353e-06, + "loss": 0.9272, + "step": 21426 + }, + { + "epoch": 0.69, + "learning_rate": 4.761552567321433e-06, + "loss": 0.8384, + "step": 21427 + }, + { + "epoch": 0.69, + "learning_rate": 4.760670283569259e-06, + "loss": 0.9238, + "step": 21428 + }, + { + "epoch": 0.69, + "learning_rate": 4.759788056029292e-06, + "loss": 0.9658, + "step": 21429 + }, + { + "epoch": 0.69, + "learning_rate": 4.758905884711007e-06, + "loss": 0.9482, + "step": 21430 + }, + { + "epoch": 0.69, + "learning_rate": 4.758023769623853e-06, + "loss": 0.9258, + "step": 21431 + }, + { + "epoch": 0.69, + "learning_rate": 4.7571417107773065e-06, + "loss": 0.9248, + "step": 21432 + }, + { + "epoch": 0.69, + "learning_rate": 4.756259708180823e-06, + "loss": 0.9995, + "step": 21433 + }, + { + "epoch": 0.69, + "learning_rate": 4.755377761843874e-06, + "loss": 0.8218, + "step": 21434 + }, + { + "epoch": 0.69, + "learning_rate": 4.754495871775913e-06, + "loss": 0.9502, + "step": 21435 + }, + { + "epoch": 0.69, + "learning_rate": 4.7536140379864085e-06, + "loss": 0.8975, + "step": 21436 + }, + { + "epoch": 0.69, + "learning_rate": 4.752732260484816e-06, + "loss": 0.9839, + "step": 21437 + }, + { + "epoch": 0.69, + "learning_rate": 4.751850539280607e-06, + "loss": 0.9766, + "step": 21438 + }, + { + "epoch": 0.69, + "learning_rate": 4.750968874383225e-06, + "loss": 1.0137, + "step": 21439 + }, + { + "epoch": 0.69, + "learning_rate": 4.750087265802141e-06, + "loss": 1.0107, + "step": 21440 + }, + { + "epoch": 0.69, + "learning_rate": 4.749205713546808e-06, + "loss": 0.9194, + "step": 21441 + }, + { + "epoch": 0.69, + "learning_rate": 4.74832421762669e-06, + "loss": 1.0083, + "step": 21442 + }, + { + "epoch": 0.69, + "learning_rate": 4.747442778051236e-06, + "loss": 1.0615, + "step": 21443 + }, + { + "epoch": 0.69, + "learning_rate": 4.746561394829912e-06, + "loss": 0.9966, + "step": 21444 + }, + { + "epoch": 0.69, + "learning_rate": 4.74568006797217e-06, + "loss": 0.9141, + "step": 21445 + }, + { + "epoch": 0.69, + "learning_rate": 4.744798797487465e-06, + "loss": 0.9434, + "step": 21446 + }, + { + "epoch": 0.69, + "learning_rate": 4.7439175833852504e-06, + "loss": 1.0063, + "step": 21447 + }, + { + "epoch": 0.69, + "learning_rate": 4.743036425674986e-06, + "loss": 0.9121, + "step": 21448 + }, + { + "epoch": 0.69, + "learning_rate": 4.742155324366119e-06, + "loss": 0.5342, + "step": 21449 + }, + { + "epoch": 0.69, + "learning_rate": 4.74127427946811e-06, + "loss": 0.9268, + "step": 21450 + }, + { + "epoch": 0.69, + "learning_rate": 4.740393290990408e-06, + "loss": 0.9814, + "step": 21451 + }, + { + "epoch": 0.69, + "learning_rate": 4.739512358942461e-06, + "loss": 0.9458, + "step": 21452 + }, + { + "epoch": 0.69, + "learning_rate": 4.738631483333729e-06, + "loss": 0.98, + "step": 21453 + }, + { + "epoch": 0.69, + "learning_rate": 4.737750664173657e-06, + "loss": 0.9604, + "step": 21454 + }, + { + "epoch": 0.69, + "learning_rate": 4.736869901471693e-06, + "loss": 0.9956, + "step": 21455 + }, + { + "epoch": 0.69, + "learning_rate": 4.735989195237294e-06, + "loss": 0.9302, + "step": 21456 + }, + { + "epoch": 0.69, + "learning_rate": 4.735108545479904e-06, + "loss": 1.0347, + "step": 21457 + }, + { + "epoch": 0.69, + "learning_rate": 4.73422795220897e-06, + "loss": 0.8027, + "step": 21458 + }, + { + "epoch": 0.69, + "learning_rate": 4.733347415433946e-06, + "loss": 0.9658, + "step": 21459 + }, + { + "epoch": 0.69, + "learning_rate": 4.732466935164274e-06, + "loss": 0.8955, + "step": 21460 + }, + { + "epoch": 0.69, + "learning_rate": 4.7315865114093994e-06, + "loss": 0.8208, + "step": 21461 + }, + { + "epoch": 0.69, + "learning_rate": 4.730706144178774e-06, + "loss": 0.8237, + "step": 21462 + }, + { + "epoch": 0.69, + "learning_rate": 4.729825833481841e-06, + "loss": 0.9766, + "step": 21463 + }, + { + "epoch": 0.69, + "learning_rate": 4.728945579328038e-06, + "loss": 0.9995, + "step": 21464 + }, + { + "epoch": 0.69, + "learning_rate": 4.7280653817268205e-06, + "loss": 0.9951, + "step": 21465 + }, + { + "epoch": 0.69, + "learning_rate": 4.727185240687623e-06, + "loss": 0.8726, + "step": 21466 + }, + { + "epoch": 0.69, + "learning_rate": 4.726305156219896e-06, + "loss": 0.9126, + "step": 21467 + }, + { + "epoch": 0.69, + "learning_rate": 4.725425128333077e-06, + "loss": 1.0273, + "step": 21468 + }, + { + "epoch": 0.69, + "learning_rate": 4.724545157036611e-06, + "loss": 1.0049, + "step": 21469 + }, + { + "epoch": 0.69, + "learning_rate": 4.723665242339931e-06, + "loss": 0.7869, + "step": 21470 + }, + { + "epoch": 0.69, + "learning_rate": 4.722785384252488e-06, + "loss": 0.8945, + "step": 21471 + }, + { + "epoch": 0.69, + "learning_rate": 4.721905582783714e-06, + "loss": 0.9492, + "step": 21472 + }, + { + "epoch": 0.69, + "learning_rate": 4.721025837943054e-06, + "loss": 0.9224, + "step": 21473 + }, + { + "epoch": 0.69, + "learning_rate": 4.720146149739941e-06, + "loss": 1.0264, + "step": 21474 + }, + { + "epoch": 0.69, + "learning_rate": 4.7192665181838245e-06, + "loss": 1.0405, + "step": 21475 + }, + { + "epoch": 0.69, + "learning_rate": 4.7183869432841255e-06, + "loss": 0.8882, + "step": 21476 + }, + { + "epoch": 0.69, + "learning_rate": 4.717507425050294e-06, + "loss": 0.9351, + "step": 21477 + }, + { + "epoch": 0.69, + "learning_rate": 4.7166279634917575e-06, + "loss": 0.8569, + "step": 21478 + }, + { + "epoch": 0.69, + "learning_rate": 4.715748558617959e-06, + "loss": 0.939, + "step": 21479 + }, + { + "epoch": 0.69, + "learning_rate": 4.714869210438326e-06, + "loss": 0.9092, + "step": 21480 + }, + { + "epoch": 0.69, + "learning_rate": 4.713989918962302e-06, + "loss": 0.978, + "step": 21481 + }, + { + "epoch": 0.69, + "learning_rate": 4.713110684199311e-06, + "loss": 0.9346, + "step": 21482 + }, + { + "epoch": 0.69, + "learning_rate": 4.7122315061587985e-06, + "loss": 0.9365, + "step": 21483 + }, + { + "epoch": 0.69, + "learning_rate": 4.711352384850183e-06, + "loss": 0.9844, + "step": 21484 + }, + { + "epoch": 0.69, + "learning_rate": 4.710473320282907e-06, + "loss": 0.9756, + "step": 21485 + }, + { + "epoch": 0.69, + "learning_rate": 4.709594312466394e-06, + "loss": 0.854, + "step": 21486 + }, + { + "epoch": 0.69, + "learning_rate": 4.708715361410082e-06, + "loss": 0.8735, + "step": 21487 + }, + { + "epoch": 0.69, + "learning_rate": 4.707836467123395e-06, + "loss": 0.9062, + "step": 21488 + }, + { + "epoch": 0.69, + "learning_rate": 4.7069576296157704e-06, + "loss": 1.0356, + "step": 21489 + }, + { + "epoch": 0.69, + "learning_rate": 4.706078848896631e-06, + "loss": 0.9849, + "step": 21490 + }, + { + "epoch": 0.69, + "learning_rate": 4.705200124975406e-06, + "loss": 1.0688, + "step": 21491 + }, + { + "epoch": 0.69, + "learning_rate": 4.704321457861522e-06, + "loss": 0.8486, + "step": 21492 + }, + { + "epoch": 0.69, + "learning_rate": 4.70344284756441e-06, + "loss": 0.9204, + "step": 21493 + }, + { + "epoch": 0.69, + "learning_rate": 4.702564294093491e-06, + "loss": 0.9556, + "step": 21494 + }, + { + "epoch": 0.69, + "learning_rate": 4.701685797458198e-06, + "loss": 1.0562, + "step": 21495 + }, + { + "epoch": 0.69, + "learning_rate": 4.700807357667953e-06, + "loss": 0.897, + "step": 21496 + }, + { + "epoch": 0.69, + "learning_rate": 4.699928974732175e-06, + "loss": 0.9106, + "step": 21497 + }, + { + "epoch": 0.69, + "learning_rate": 4.699050648660298e-06, + "loss": 0.9058, + "step": 21498 + }, + { + "epoch": 0.69, + "learning_rate": 4.6981723794617406e-06, + "loss": 0.9067, + "step": 21499 + }, + { + "epoch": 0.69, + "learning_rate": 4.697294167145923e-06, + "loss": 0.5127, + "step": 21500 + }, + { + "epoch": 0.69, + "learning_rate": 4.696416011722272e-06, + "loss": 1.0732, + "step": 21501 + }, + { + "epoch": 0.69, + "learning_rate": 4.695537913200209e-06, + "loss": 0.834, + "step": 21502 + }, + { + "epoch": 0.69, + "learning_rate": 4.694659871589149e-06, + "loss": 0.8472, + "step": 21503 + }, + { + "epoch": 0.69, + "learning_rate": 4.693781886898521e-06, + "loss": 0.9399, + "step": 21504 + }, + { + "epoch": 0.69, + "learning_rate": 4.692903959137741e-06, + "loss": 1.0542, + "step": 21505 + }, + { + "epoch": 0.69, + "learning_rate": 4.692026088316227e-06, + "loss": 1.0493, + "step": 21506 + }, + { + "epoch": 0.69, + "learning_rate": 4.691148274443396e-06, + "loss": 0.9399, + "step": 21507 + }, + { + "epoch": 0.69, + "learning_rate": 4.690270517528671e-06, + "loss": 1.0254, + "step": 21508 + }, + { + "epoch": 0.69, + "learning_rate": 4.6893928175814644e-06, + "loss": 0.8267, + "step": 21509 + }, + { + "epoch": 0.69, + "learning_rate": 4.688515174611198e-06, + "loss": 0.8975, + "step": 21510 + }, + { + "epoch": 0.69, + "learning_rate": 4.6876375886272815e-06, + "loss": 0.9448, + "step": 21511 + }, + { + "epoch": 0.69, + "learning_rate": 4.686760059639139e-06, + "loss": 0.9121, + "step": 21512 + }, + { + "epoch": 0.69, + "learning_rate": 4.685882587656179e-06, + "loss": 0.9731, + "step": 21513 + }, + { + "epoch": 0.69, + "learning_rate": 4.68500517268782e-06, + "loss": 0.8877, + "step": 21514 + }, + { + "epoch": 0.69, + "learning_rate": 4.684127814743468e-06, + "loss": 0.8818, + "step": 21515 + }, + { + "epoch": 0.69, + "learning_rate": 4.6832505138325445e-06, + "loss": 0.8867, + "step": 21516 + }, + { + "epoch": 0.69, + "learning_rate": 4.682373269964455e-06, + "loss": 0.9746, + "step": 21517 + }, + { + "epoch": 0.69, + "learning_rate": 4.681496083148619e-06, + "loss": 0.9185, + "step": 21518 + }, + { + "epoch": 0.69, + "learning_rate": 4.6806189533944405e-06, + "loss": 1.0142, + "step": 21519 + }, + { + "epoch": 0.69, + "learning_rate": 4.679741880711337e-06, + "loss": 0.9653, + "step": 21520 + }, + { + "epoch": 0.69, + "learning_rate": 4.678864865108715e-06, + "loss": 0.9272, + "step": 21521 + }, + { + "epoch": 0.69, + "learning_rate": 4.677987906595983e-06, + "loss": 0.959, + "step": 21522 + }, + { + "epoch": 0.69, + "learning_rate": 4.677111005182549e-06, + "loss": 0.9644, + "step": 21523 + }, + { + "epoch": 0.69, + "learning_rate": 4.676234160877825e-06, + "loss": 1.0034, + "step": 21524 + }, + { + "epoch": 0.69, + "learning_rate": 4.675357373691213e-06, + "loss": 0.9341, + "step": 21525 + }, + { + "epoch": 0.69, + "learning_rate": 4.674480643632126e-06, + "loss": 0.8738, + "step": 21526 + }, + { + "epoch": 0.69, + "learning_rate": 4.6736039707099655e-06, + "loss": 1.0298, + "step": 21527 + }, + { + "epoch": 0.69, + "learning_rate": 4.672727354934145e-06, + "loss": 1.0024, + "step": 21528 + }, + { + "epoch": 0.69, + "learning_rate": 4.671850796314059e-06, + "loss": 0.896, + "step": 21529 + }, + { + "epoch": 0.69, + "learning_rate": 4.67097429485912e-06, + "loss": 0.7808, + "step": 21530 + }, + { + "epoch": 0.69, + "learning_rate": 4.670097850578724e-06, + "loss": 1.0718, + "step": 21531 + }, + { + "epoch": 0.69, + "learning_rate": 4.669221463482284e-06, + "loss": 0.8113, + "step": 21532 + }, + { + "epoch": 0.69, + "learning_rate": 4.668345133579194e-06, + "loss": 0.9248, + "step": 21533 + }, + { + "epoch": 0.69, + "learning_rate": 4.667468860878864e-06, + "loss": 0.8418, + "step": 21534 + }, + { + "epoch": 0.69, + "learning_rate": 4.666592645390692e-06, + "loss": 0.9268, + "step": 21535 + }, + { + "epoch": 0.69, + "learning_rate": 4.665716487124077e-06, + "loss": 0.9912, + "step": 21536 + }, + { + "epoch": 0.69, + "learning_rate": 4.664840386088416e-06, + "loss": 0.5098, + "step": 21537 + }, + { + "epoch": 0.69, + "learning_rate": 4.663964342293118e-06, + "loss": 0.9209, + "step": 21538 + }, + { + "epoch": 0.69, + "learning_rate": 4.663088355747572e-06, + "loss": 0.4758, + "step": 21539 + }, + { + "epoch": 0.69, + "learning_rate": 4.6622124264611866e-06, + "loss": 1.0298, + "step": 21540 + }, + { + "epoch": 0.69, + "learning_rate": 4.661336554443354e-06, + "loss": 0.9761, + "step": 21541 + }, + { + "epoch": 0.69, + "learning_rate": 4.660460739703467e-06, + "loss": 1.02, + "step": 21542 + }, + { + "epoch": 0.69, + "learning_rate": 4.659584982250932e-06, + "loss": 0.9775, + "step": 21543 + }, + { + "epoch": 0.69, + "learning_rate": 4.658709282095138e-06, + "loss": 0.9175, + "step": 21544 + }, + { + "epoch": 0.69, + "learning_rate": 4.657833639245482e-06, + "loss": 0.8936, + "step": 21545 + }, + { + "epoch": 0.69, + "learning_rate": 4.656958053711357e-06, + "loss": 0.9746, + "step": 21546 + }, + { + "epoch": 0.69, + "learning_rate": 4.65608252550216e-06, + "loss": 1.0576, + "step": 21547 + }, + { + "epoch": 0.69, + "learning_rate": 4.655207054627281e-06, + "loss": 0.9175, + "step": 21548 + }, + { + "epoch": 0.69, + "learning_rate": 4.654331641096118e-06, + "loss": 0.999, + "step": 21549 + }, + { + "epoch": 0.69, + "learning_rate": 4.6534562849180565e-06, + "loss": 1.0645, + "step": 21550 + }, + { + "epoch": 0.69, + "learning_rate": 4.6525809861025e-06, + "loss": 0.9507, + "step": 21551 + }, + { + "epoch": 0.69, + "learning_rate": 4.651705744658822e-06, + "loss": 0.9761, + "step": 21552 + }, + { + "epoch": 0.69, + "learning_rate": 4.6508305605964276e-06, + "loss": 0.8662, + "step": 21553 + }, + { + "epoch": 0.69, + "learning_rate": 4.6499554339246965e-06, + "loss": 1.0366, + "step": 21554 + }, + { + "epoch": 0.69, + "learning_rate": 4.649080364653027e-06, + "loss": 0.9023, + "step": 21555 + }, + { + "epoch": 0.69, + "learning_rate": 4.648205352790799e-06, + "loss": 0.877, + "step": 21556 + }, + { + "epoch": 0.69, + "learning_rate": 4.647330398347408e-06, + "loss": 0.9521, + "step": 21557 + }, + { + "epoch": 0.69, + "learning_rate": 4.646455501332237e-06, + "loss": 1.0059, + "step": 21558 + }, + { + "epoch": 0.69, + "learning_rate": 4.645580661754675e-06, + "loss": 0.8413, + "step": 21559 + }, + { + "epoch": 0.69, + "learning_rate": 4.644705879624102e-06, + "loss": 0.8926, + "step": 21560 + }, + { + "epoch": 0.69, + "learning_rate": 4.6438311549499115e-06, + "loss": 0.8613, + "step": 21561 + }, + { + "epoch": 0.69, + "learning_rate": 4.642956487741482e-06, + "loss": 0.8857, + "step": 21562 + }, + { + "epoch": 0.69, + "learning_rate": 4.642081878008203e-06, + "loss": 0.9165, + "step": 21563 + }, + { + "epoch": 0.69, + "learning_rate": 4.641207325759454e-06, + "loss": 0.8276, + "step": 21564 + }, + { + "epoch": 0.69, + "learning_rate": 4.640332831004621e-06, + "loss": 0.8872, + "step": 21565 + }, + { + "epoch": 0.69, + "learning_rate": 4.6394583937530855e-06, + "loss": 0.8726, + "step": 21566 + }, + { + "epoch": 0.69, + "learning_rate": 4.638584014014229e-06, + "loss": 0.874, + "step": 21567 + }, + { + "epoch": 0.69, + "learning_rate": 4.637709691797428e-06, + "loss": 0.9927, + "step": 21568 + }, + { + "epoch": 0.69, + "learning_rate": 4.636835427112072e-06, + "loss": 0.9062, + "step": 21569 + }, + { + "epoch": 0.69, + "learning_rate": 4.635961219967532e-06, + "loss": 0.5498, + "step": 21570 + }, + { + "epoch": 0.69, + "learning_rate": 4.635087070373196e-06, + "loss": 1.0034, + "step": 21571 + }, + { + "epoch": 0.69, + "learning_rate": 4.634212978338433e-06, + "loss": 0.9326, + "step": 21572 + }, + { + "epoch": 0.69, + "learning_rate": 4.633338943872634e-06, + "loss": 0.9644, + "step": 21573 + }, + { + "epoch": 0.69, + "learning_rate": 4.6324649669851614e-06, + "loss": 0.9512, + "step": 21574 + }, + { + "epoch": 0.69, + "learning_rate": 4.631591047685403e-06, + "loss": 0.9419, + "step": 21575 + }, + { + "epoch": 0.69, + "learning_rate": 4.630717185982728e-06, + "loss": 0.9424, + "step": 21576 + }, + { + "epoch": 0.69, + "learning_rate": 4.629843381886519e-06, + "loss": 0.8618, + "step": 21577 + }, + { + "epoch": 0.69, + "learning_rate": 4.628969635406142e-06, + "loss": 0.9775, + "step": 21578 + }, + { + "epoch": 0.69, + "learning_rate": 4.6280959465509815e-06, + "loss": 0.9639, + "step": 21579 + }, + { + "epoch": 0.69, + "learning_rate": 4.627222315330406e-06, + "loss": 0.9502, + "step": 21580 + }, + { + "epoch": 0.69, + "learning_rate": 4.626348741753788e-06, + "loss": 0.9863, + "step": 21581 + }, + { + "epoch": 0.69, + "learning_rate": 4.6254752258304966e-06, + "loss": 0.9473, + "step": 21582 + }, + { + "epoch": 0.69, + "learning_rate": 4.6246017675699125e-06, + "loss": 0.9102, + "step": 21583 + }, + { + "epoch": 0.69, + "learning_rate": 4.623728366981398e-06, + "loss": 0.9141, + "step": 21584 + }, + { + "epoch": 0.69, + "learning_rate": 4.622855024074332e-06, + "loss": 0.8813, + "step": 21585 + }, + { + "epoch": 0.69, + "learning_rate": 4.621981738858081e-06, + "loss": 0.9365, + "step": 21586 + }, + { + "epoch": 0.69, + "learning_rate": 4.62110851134201e-06, + "loss": 0.9951, + "step": 21587 + }, + { + "epoch": 0.69, + "learning_rate": 4.620235341535494e-06, + "loss": 1.0659, + "step": 21588 + }, + { + "epoch": 0.69, + "learning_rate": 4.6193622294479e-06, + "loss": 0.8071, + "step": 21589 + }, + { + "epoch": 0.69, + "learning_rate": 4.618489175088594e-06, + "loss": 1.0659, + "step": 21590 + }, + { + "epoch": 0.69, + "learning_rate": 4.617616178466939e-06, + "loss": 0.9565, + "step": 21591 + }, + { + "epoch": 0.69, + "learning_rate": 4.61674323959231e-06, + "loss": 0.9082, + "step": 21592 + }, + { + "epoch": 0.69, + "learning_rate": 4.6158703584740636e-06, + "loss": 0.4985, + "step": 21593 + }, + { + "epoch": 0.69, + "learning_rate": 4.614997535121574e-06, + "loss": 0.8911, + "step": 21594 + }, + { + "epoch": 0.69, + "learning_rate": 4.6141247695441975e-06, + "loss": 0.9243, + "step": 21595 + }, + { + "epoch": 0.69, + "learning_rate": 4.613252061751308e-06, + "loss": 0.9541, + "step": 21596 + }, + { + "epoch": 0.69, + "learning_rate": 4.612379411752255e-06, + "loss": 0.9941, + "step": 21597 + }, + { + "epoch": 0.69, + "learning_rate": 4.611506819556412e-06, + "loss": 0.8108, + "step": 21598 + }, + { + "epoch": 0.69, + "learning_rate": 4.610634285173132e-06, + "loss": 1.0234, + "step": 21599 + }, + { + "epoch": 0.69, + "learning_rate": 4.609761808611787e-06, + "loss": 1.0737, + "step": 21600 + }, + { + "epoch": 0.69, + "learning_rate": 4.608889389881728e-06, + "loss": 0.4861, + "step": 21601 + }, + { + "epoch": 0.69, + "learning_rate": 4.608017028992322e-06, + "loss": 0.9482, + "step": 21602 + }, + { + "epoch": 0.69, + "learning_rate": 4.607144725952924e-06, + "loss": 0.8745, + "step": 21603 + }, + { + "epoch": 0.69, + "learning_rate": 4.6062724807729e-06, + "loss": 0.9097, + "step": 21604 + }, + { + "epoch": 0.69, + "learning_rate": 4.605400293461596e-06, + "loss": 0.8892, + "step": 21605 + }, + { + "epoch": 0.69, + "learning_rate": 4.604528164028381e-06, + "loss": 0.8784, + "step": 21606 + }, + { + "epoch": 0.69, + "learning_rate": 4.603656092482602e-06, + "loss": 1.0142, + "step": 21607 + }, + { + "epoch": 0.69, + "learning_rate": 4.602784078833626e-06, + "loss": 0.7688, + "step": 21608 + }, + { + "epoch": 0.69, + "learning_rate": 4.601912123090799e-06, + "loss": 0.894, + "step": 21609 + }, + { + "epoch": 0.69, + "learning_rate": 4.601040225263486e-06, + "loss": 1.0288, + "step": 21610 + }, + { + "epoch": 0.69, + "learning_rate": 4.600168385361034e-06, + "loss": 0.9253, + "step": 21611 + }, + { + "epoch": 0.69, + "learning_rate": 4.5992966033928e-06, + "loss": 0.9468, + "step": 21612 + }, + { + "epoch": 0.69, + "learning_rate": 4.5984248793681316e-06, + "loss": 0.9146, + "step": 21613 + }, + { + "epoch": 0.69, + "learning_rate": 4.597553213296391e-06, + "loss": 0.9746, + "step": 21614 + }, + { + "epoch": 0.69, + "learning_rate": 4.596681605186921e-06, + "loss": 0.8604, + "step": 21615 + }, + { + "epoch": 0.69, + "learning_rate": 4.59581005504908e-06, + "loss": 0.9038, + "step": 21616 + }, + { + "epoch": 0.69, + "learning_rate": 4.5949385628922134e-06, + "loss": 0.999, + "step": 21617 + }, + { + "epoch": 0.69, + "learning_rate": 4.594067128725678e-06, + "loss": 0.915, + "step": 21618 + }, + { + "epoch": 0.69, + "learning_rate": 4.593195752558819e-06, + "loss": 0.9468, + "step": 21619 + }, + { + "epoch": 0.69, + "learning_rate": 4.592324434400986e-06, + "loss": 0.8984, + "step": 21620 + }, + { + "epoch": 0.69, + "learning_rate": 4.591453174261522e-06, + "loss": 0.9946, + "step": 21621 + }, + { + "epoch": 0.69, + "learning_rate": 4.590581972149785e-06, + "loss": 0.5234, + "step": 21622 + }, + { + "epoch": 0.69, + "learning_rate": 4.589710828075112e-06, + "loss": 0.9424, + "step": 21623 + }, + { + "epoch": 0.69, + "learning_rate": 4.5888397420468566e-06, + "loss": 0.8667, + "step": 21624 + }, + { + "epoch": 0.69, + "learning_rate": 4.587968714074363e-06, + "loss": 0.9468, + "step": 21625 + }, + { + "epoch": 0.69, + "learning_rate": 4.587097744166976e-06, + "loss": 0.9287, + "step": 21626 + }, + { + "epoch": 0.69, + "learning_rate": 4.586226832334035e-06, + "loss": 0.958, + "step": 21627 + }, + { + "epoch": 0.69, + "learning_rate": 4.585355978584892e-06, + "loss": 0.9614, + "step": 21628 + }, + { + "epoch": 0.69, + "learning_rate": 4.584485182928887e-06, + "loss": 0.9272, + "step": 21629 + }, + { + "epoch": 0.69, + "learning_rate": 4.5836144453753595e-06, + "loss": 0.978, + "step": 21630 + }, + { + "epoch": 0.69, + "learning_rate": 4.582743765933657e-06, + "loss": 0.9248, + "step": 21631 + }, + { + "epoch": 0.69, + "learning_rate": 4.581873144613115e-06, + "loss": 0.9502, + "step": 21632 + }, + { + "epoch": 0.69, + "learning_rate": 4.581002581423081e-06, + "loss": 0.8823, + "step": 21633 + }, + { + "epoch": 0.69, + "learning_rate": 4.580132076372893e-06, + "loss": 1.0283, + "step": 21634 + }, + { + "epoch": 0.69, + "learning_rate": 4.579261629471888e-06, + "loss": 0.9961, + "step": 21635 + }, + { + "epoch": 0.69, + "learning_rate": 4.578391240729403e-06, + "loss": 1.0005, + "step": 21636 + }, + { + "epoch": 0.69, + "learning_rate": 4.577520910154784e-06, + "loss": 0.9497, + "step": 21637 + }, + { + "epoch": 0.69, + "learning_rate": 4.57665063775736e-06, + "loss": 0.8931, + "step": 21638 + }, + { + "epoch": 0.69, + "learning_rate": 4.575780423546476e-06, + "loss": 0.9873, + "step": 21639 + }, + { + "epoch": 0.69, + "learning_rate": 4.574910267531462e-06, + "loss": 0.9907, + "step": 21640 + }, + { + "epoch": 0.69, + "learning_rate": 4.574040169721664e-06, + "loss": 1.0259, + "step": 21641 + }, + { + "epoch": 0.69, + "learning_rate": 4.573170130126402e-06, + "loss": 0.958, + "step": 21642 + }, + { + "epoch": 0.69, + "learning_rate": 4.572300148755021e-06, + "loss": 0.8643, + "step": 21643 + }, + { + "epoch": 0.69, + "learning_rate": 4.57143022561685e-06, + "loss": 1.0522, + "step": 21644 + }, + { + "epoch": 0.69, + "learning_rate": 4.5705603607212275e-06, + "loss": 1.0327, + "step": 21645 + }, + { + "epoch": 0.69, + "learning_rate": 4.5696905540774805e-06, + "loss": 0.9917, + "step": 21646 + }, + { + "epoch": 0.69, + "learning_rate": 4.5688208056949475e-06, + "loss": 1.0894, + "step": 21647 + }, + { + "epoch": 0.69, + "learning_rate": 4.567951115582953e-06, + "loss": 0.9053, + "step": 21648 + }, + { + "epoch": 0.69, + "learning_rate": 4.567081483750837e-06, + "loss": 0.9536, + "step": 21649 + }, + { + "epoch": 0.69, + "learning_rate": 4.566211910207918e-06, + "loss": 0.8545, + "step": 21650 + }, + { + "epoch": 0.69, + "learning_rate": 4.565342394963533e-06, + "loss": 0.8486, + "step": 21651 + }, + { + "epoch": 0.69, + "learning_rate": 4.5644729380270085e-06, + "loss": 0.8696, + "step": 21652 + }, + { + "epoch": 0.69, + "learning_rate": 4.563603539407675e-06, + "loss": 0.8633, + "step": 21653 + }, + { + "epoch": 0.69, + "learning_rate": 4.5627341991148556e-06, + "loss": 1.0283, + "step": 21654 + }, + { + "epoch": 0.69, + "learning_rate": 4.561864917157885e-06, + "loss": 0.9463, + "step": 21655 + }, + { + "epoch": 0.69, + "learning_rate": 4.560995693546085e-06, + "loss": 0.9463, + "step": 21656 + }, + { + "epoch": 0.69, + "learning_rate": 4.560126528288781e-06, + "loss": 0.8848, + "step": 21657 + }, + { + "epoch": 0.69, + "learning_rate": 4.559257421395295e-06, + "loss": 0.9492, + "step": 21658 + }, + { + "epoch": 0.69, + "learning_rate": 4.55838837287496e-06, + "loss": 0.9541, + "step": 21659 + }, + { + "epoch": 0.69, + "learning_rate": 4.557519382737091e-06, + "loss": 0.9819, + "step": 21660 + }, + { + "epoch": 0.69, + "learning_rate": 4.5566504509910205e-06, + "loss": 0.8887, + "step": 21661 + }, + { + "epoch": 0.69, + "learning_rate": 4.555781577646061e-06, + "loss": 0.9731, + "step": 21662 + }, + { + "epoch": 0.69, + "learning_rate": 4.554912762711544e-06, + "loss": 0.979, + "step": 21663 + }, + { + "epoch": 0.69, + "learning_rate": 4.5540440061967875e-06, + "loss": 0.917, + "step": 21664 + }, + { + "epoch": 0.69, + "learning_rate": 4.553175308111109e-06, + "loss": 1.0923, + "step": 21665 + }, + { + "epoch": 0.69, + "learning_rate": 4.552306668463831e-06, + "loss": 0.8223, + "step": 21666 + }, + { + "epoch": 0.69, + "learning_rate": 4.551438087264275e-06, + "loss": 0.8794, + "step": 21667 + }, + { + "epoch": 0.69, + "learning_rate": 4.550569564521758e-06, + "loss": 0.9282, + "step": 21668 + }, + { + "epoch": 0.69, + "learning_rate": 4.549701100245595e-06, + "loss": 1.0107, + "step": 21669 + }, + { + "epoch": 0.69, + "learning_rate": 4.548832694445111e-06, + "loss": 0.9331, + "step": 21670 + }, + { + "epoch": 0.69, + "learning_rate": 4.547964347129615e-06, + "loss": 0.9678, + "step": 21671 + }, + { + "epoch": 0.69, + "learning_rate": 4.547096058308433e-06, + "loss": 0.8574, + "step": 21672 + }, + { + "epoch": 0.69, + "learning_rate": 4.5462278279908736e-06, + "loss": 0.936, + "step": 21673 + }, + { + "epoch": 0.69, + "learning_rate": 4.545359656186253e-06, + "loss": 0.9541, + "step": 21674 + }, + { + "epoch": 0.69, + "learning_rate": 4.5444915429038836e-06, + "loss": 0.9023, + "step": 21675 + }, + { + "epoch": 0.69, + "learning_rate": 4.5436234881530865e-06, + "loss": 1.0059, + "step": 21676 + }, + { + "epoch": 0.69, + "learning_rate": 4.542755491943166e-06, + "loss": 0.9932, + "step": 21677 + }, + { + "epoch": 0.69, + "learning_rate": 4.541887554283443e-06, + "loss": 0.9424, + "step": 21678 + }, + { + "epoch": 0.69, + "learning_rate": 4.541019675183226e-06, + "loss": 0.9463, + "step": 21679 + }, + { + "epoch": 0.69, + "learning_rate": 4.540151854651827e-06, + "loss": 0.5093, + "step": 21680 + }, + { + "epoch": 0.69, + "learning_rate": 4.539284092698551e-06, + "loss": 0.8296, + "step": 21681 + }, + { + "epoch": 0.69, + "learning_rate": 4.538416389332717e-06, + "loss": 0.978, + "step": 21682 + }, + { + "epoch": 0.69, + "learning_rate": 4.537548744563627e-06, + "loss": 0.8647, + "step": 21683 + }, + { + "epoch": 0.69, + "learning_rate": 4.536681158400598e-06, + "loss": 0.959, + "step": 21684 + }, + { + "epoch": 0.69, + "learning_rate": 4.535813630852929e-06, + "loss": 0.9224, + "step": 21685 + }, + { + "epoch": 0.69, + "learning_rate": 4.534946161929935e-06, + "loss": 1.0981, + "step": 21686 + }, + { + "epoch": 0.69, + "learning_rate": 4.534078751640921e-06, + "loss": 1.001, + "step": 21687 + }, + { + "epoch": 0.69, + "learning_rate": 4.533211399995195e-06, + "loss": 0.9575, + "step": 21688 + }, + { + "epoch": 0.69, + "learning_rate": 4.532344107002054e-06, + "loss": 1.0054, + "step": 21689 + }, + { + "epoch": 0.69, + "learning_rate": 4.531476872670813e-06, + "loss": 0.9648, + "step": 21690 + }, + { + "epoch": 0.69, + "learning_rate": 4.53060969701077e-06, + "loss": 0.9683, + "step": 21691 + }, + { + "epoch": 0.69, + "learning_rate": 4.529742580031236e-06, + "loss": 1.0264, + "step": 21692 + }, + { + "epoch": 0.69, + "learning_rate": 4.528875521741505e-06, + "loss": 0.9224, + "step": 21693 + }, + { + "epoch": 0.69, + "learning_rate": 4.528008522150893e-06, + "loss": 0.9458, + "step": 21694 + }, + { + "epoch": 0.69, + "learning_rate": 4.527141581268686e-06, + "loss": 0.9297, + "step": 21695 + }, + { + "epoch": 0.69, + "learning_rate": 4.526274699104195e-06, + "loss": 0.9941, + "step": 21696 + }, + { + "epoch": 0.69, + "learning_rate": 4.525407875666716e-06, + "loss": 0.9941, + "step": 21697 + }, + { + "epoch": 0.69, + "learning_rate": 4.524541110965556e-06, + "loss": 0.5085, + "step": 21698 + }, + { + "epoch": 0.69, + "learning_rate": 4.523674405010005e-06, + "loss": 1.001, + "step": 21699 + }, + { + "epoch": 0.69, + "learning_rate": 4.52280775780937e-06, + "loss": 0.9146, + "step": 21700 + }, + { + "epoch": 0.69, + "learning_rate": 4.521941169372943e-06, + "loss": 0.8994, + "step": 21701 + }, + { + "epoch": 0.69, + "learning_rate": 4.521074639710033e-06, + "loss": 0.8906, + "step": 21702 + }, + { + "epoch": 0.69, + "learning_rate": 4.520208168829919e-06, + "loss": 0.8687, + "step": 21703 + }, + { + "epoch": 0.69, + "learning_rate": 4.519341756741911e-06, + "loss": 0.9946, + "step": 21704 + }, + { + "epoch": 0.69, + "learning_rate": 4.518475403455297e-06, + "loss": 0.9355, + "step": 21705 + }, + { + "epoch": 0.69, + "learning_rate": 4.517609108979379e-06, + "loss": 0.8774, + "step": 21706 + }, + { + "epoch": 0.69, + "learning_rate": 4.516742873323447e-06, + "loss": 1.0737, + "step": 21707 + }, + { + "epoch": 0.69, + "learning_rate": 4.515876696496791e-06, + "loss": 0.9263, + "step": 21708 + }, + { + "epoch": 0.69, + "learning_rate": 4.515010578508713e-06, + "loss": 0.8789, + "step": 21709 + }, + { + "epoch": 0.69, + "learning_rate": 4.514144519368502e-06, + "loss": 0.8887, + "step": 21710 + }, + { + "epoch": 0.69, + "learning_rate": 4.513278519085444e-06, + "loss": 1.0493, + "step": 21711 + }, + { + "epoch": 0.69, + "learning_rate": 4.512412577668839e-06, + "loss": 0.9907, + "step": 21712 + }, + { + "epoch": 0.69, + "learning_rate": 4.511546695127973e-06, + "loss": 0.9287, + "step": 21713 + }, + { + "epoch": 0.69, + "learning_rate": 4.510680871472134e-06, + "loss": 0.9824, + "step": 21714 + }, + { + "epoch": 0.69, + "learning_rate": 4.509815106710617e-06, + "loss": 1.0903, + "step": 21715 + }, + { + "epoch": 0.69, + "learning_rate": 4.5089494008527045e-06, + "loss": 0.9209, + "step": 21716 + }, + { + "epoch": 0.69, + "learning_rate": 4.5080837539076914e-06, + "loss": 0.96, + "step": 21717 + }, + { + "epoch": 0.69, + "learning_rate": 4.507218165884861e-06, + "loss": 0.7949, + "step": 21718 + }, + { + "epoch": 0.69, + "learning_rate": 4.506352636793502e-06, + "loss": 0.8184, + "step": 21719 + }, + { + "epoch": 0.69, + "learning_rate": 4.5054871666428935e-06, + "loss": 0.9351, + "step": 21720 + }, + { + "epoch": 0.69, + "learning_rate": 4.504621755442332e-06, + "loss": 0.9521, + "step": 21721 + }, + { + "epoch": 0.69, + "learning_rate": 4.503756403201093e-06, + "loss": 0.9785, + "step": 21722 + }, + { + "epoch": 0.69, + "learning_rate": 4.502891109928468e-06, + "loss": 0.8413, + "step": 21723 + }, + { + "epoch": 0.69, + "learning_rate": 4.5020258756337376e-06, + "loss": 0.9175, + "step": 21724 + }, + { + "epoch": 0.69, + "learning_rate": 4.501160700326185e-06, + "loss": 0.8618, + "step": 21725 + }, + { + "epoch": 0.69, + "learning_rate": 4.500295584015087e-06, + "loss": 0.9531, + "step": 21726 + }, + { + "epoch": 0.69, + "learning_rate": 4.499430526709736e-06, + "loss": 0.9165, + "step": 21727 + }, + { + "epoch": 0.69, + "learning_rate": 4.498565528419404e-06, + "loss": 0.9214, + "step": 21728 + }, + { + "epoch": 0.69, + "learning_rate": 4.497700589153379e-06, + "loss": 1.0029, + "step": 21729 + }, + { + "epoch": 0.69, + "learning_rate": 4.4968357089209326e-06, + "loss": 1.0415, + "step": 21730 + }, + { + "epoch": 0.69, + "learning_rate": 4.495970887731353e-06, + "loss": 0.9048, + "step": 21731 + }, + { + "epoch": 0.7, + "learning_rate": 4.495106125593913e-06, + "loss": 0.9932, + "step": 21732 + }, + { + "epoch": 0.7, + "learning_rate": 4.494241422517892e-06, + "loss": 0.8301, + "step": 21733 + }, + { + "epoch": 0.7, + "learning_rate": 4.493376778512564e-06, + "loss": 0.9976, + "step": 21734 + }, + { + "epoch": 0.7, + "learning_rate": 4.492512193587212e-06, + "loss": 0.981, + "step": 21735 + }, + { + "epoch": 0.7, + "learning_rate": 4.491647667751105e-06, + "loss": 0.9619, + "step": 21736 + }, + { + "epoch": 0.7, + "learning_rate": 4.490783201013527e-06, + "loss": 0.9282, + "step": 21737 + }, + { + "epoch": 0.7, + "learning_rate": 4.489918793383743e-06, + "loss": 0.978, + "step": 21738 + }, + { + "epoch": 0.7, + "learning_rate": 4.489054444871041e-06, + "loss": 0.9341, + "step": 21739 + }, + { + "epoch": 0.7, + "learning_rate": 4.488190155484676e-06, + "loss": 0.8999, + "step": 21740 + }, + { + "epoch": 0.7, + "learning_rate": 4.487325925233935e-06, + "loss": 0.8784, + "step": 21741 + }, + { + "epoch": 0.7, + "learning_rate": 4.486461754128083e-06, + "loss": 1.0771, + "step": 21742 + }, + { + "epoch": 0.7, + "learning_rate": 4.485597642176397e-06, + "loss": 0.9785, + "step": 21743 + }, + { + "epoch": 0.7, + "learning_rate": 4.484733589388142e-06, + "loss": 1.0415, + "step": 21744 + }, + { + "epoch": 0.7, + "learning_rate": 4.483869595772595e-06, + "loss": 0.9492, + "step": 21745 + }, + { + "epoch": 0.7, + "learning_rate": 4.483005661339023e-06, + "loss": 0.894, + "step": 21746 + }, + { + "epoch": 0.7, + "learning_rate": 4.4821417860966935e-06, + "loss": 0.9668, + "step": 21747 + }, + { + "epoch": 0.7, + "learning_rate": 4.481277970054872e-06, + "loss": 1.0601, + "step": 21748 + }, + { + "epoch": 0.7, + "learning_rate": 4.480414213222834e-06, + "loss": 0.9468, + "step": 21749 + }, + { + "epoch": 0.7, + "learning_rate": 4.47955051560984e-06, + "loss": 0.9639, + "step": 21750 + }, + { + "epoch": 0.7, + "learning_rate": 4.478686877225161e-06, + "loss": 0.8818, + "step": 21751 + }, + { + "epoch": 0.7, + "learning_rate": 4.477823298078062e-06, + "loss": 1.1069, + "step": 21752 + }, + { + "epoch": 0.7, + "learning_rate": 4.476959778177804e-06, + "loss": 0.876, + "step": 21753 + }, + { + "epoch": 0.7, + "learning_rate": 4.476096317533659e-06, + "loss": 0.8662, + "step": 21754 + }, + { + "epoch": 0.7, + "learning_rate": 4.475232916154885e-06, + "loss": 0.8975, + "step": 21755 + }, + { + "epoch": 0.7, + "learning_rate": 4.4743695740507464e-06, + "loss": 0.9517, + "step": 21756 + }, + { + "epoch": 0.7, + "learning_rate": 4.4735062912305095e-06, + "loss": 0.9868, + "step": 21757 + }, + { + "epoch": 0.7, + "learning_rate": 4.472643067703433e-06, + "loss": 0.8518, + "step": 21758 + }, + { + "epoch": 0.7, + "learning_rate": 4.471779903478776e-06, + "loss": 0.856, + "step": 21759 + }, + { + "epoch": 0.7, + "learning_rate": 4.470916798565805e-06, + "loss": 0.8643, + "step": 21760 + }, + { + "epoch": 0.7, + "learning_rate": 4.470053752973774e-06, + "loss": 0.9092, + "step": 21761 + }, + { + "epoch": 0.7, + "learning_rate": 4.469190766711955e-06, + "loss": 1.0044, + "step": 21762 + }, + { + "epoch": 0.7, + "learning_rate": 4.468327839789589e-06, + "loss": 0.9624, + "step": 21763 + }, + { + "epoch": 0.7, + "learning_rate": 4.467464972215947e-06, + "loss": 0.4958, + "step": 21764 + }, + { + "epoch": 0.7, + "learning_rate": 4.466602164000279e-06, + "loss": 0.9014, + "step": 21765 + }, + { + "epoch": 0.7, + "learning_rate": 4.46573941515185e-06, + "loss": 0.9575, + "step": 21766 + }, + { + "epoch": 0.7, + "learning_rate": 4.464876725679908e-06, + "loss": 0.854, + "step": 21767 + }, + { + "epoch": 0.7, + "learning_rate": 4.464014095593716e-06, + "loss": 1.0332, + "step": 21768 + }, + { + "epoch": 0.7, + "learning_rate": 4.463151524902523e-06, + "loss": 0.833, + "step": 21769 + }, + { + "epoch": 0.7, + "learning_rate": 4.4622890136155925e-06, + "loss": 0.9556, + "step": 21770 + }, + { + "epoch": 0.7, + "learning_rate": 4.461426561742166e-06, + "loss": 0.8921, + "step": 21771 + }, + { + "epoch": 0.7, + "learning_rate": 4.460564169291506e-06, + "loss": 0.8882, + "step": 21772 + }, + { + "epoch": 0.7, + "learning_rate": 4.459701836272857e-06, + "loss": 0.877, + "step": 21773 + }, + { + "epoch": 0.7, + "learning_rate": 4.458839562695481e-06, + "loss": 0.9429, + "step": 21774 + }, + { + "epoch": 0.7, + "learning_rate": 4.457977348568618e-06, + "loss": 0.8193, + "step": 21775 + }, + { + "epoch": 0.7, + "learning_rate": 4.45711519390153e-06, + "loss": 0.9285, + "step": 21776 + }, + { + "epoch": 0.7, + "learning_rate": 4.45625309870346e-06, + "loss": 0.9395, + "step": 21777 + }, + { + "epoch": 0.7, + "learning_rate": 4.455391062983659e-06, + "loss": 1.0435, + "step": 21778 + }, + { + "epoch": 0.7, + "learning_rate": 4.454529086751373e-06, + "loss": 0.7617, + "step": 21779 + }, + { + "epoch": 0.7, + "learning_rate": 4.453667170015855e-06, + "loss": 1.0317, + "step": 21780 + }, + { + "epoch": 0.7, + "learning_rate": 4.452805312786347e-06, + "loss": 0.9858, + "step": 21781 + }, + { + "epoch": 0.7, + "learning_rate": 4.451943515072102e-06, + "loss": 0.9219, + "step": 21782 + }, + { + "epoch": 0.7, + "learning_rate": 4.451081776882359e-06, + "loss": 0.9624, + "step": 21783 + }, + { + "epoch": 0.7, + "learning_rate": 4.450220098226371e-06, + "loss": 0.8657, + "step": 21784 + }, + { + "epoch": 0.7, + "learning_rate": 4.4493584791133796e-06, + "loss": 0.7798, + "step": 21785 + }, + { + "epoch": 0.7, + "learning_rate": 4.448496919552628e-06, + "loss": 0.9453, + "step": 21786 + }, + { + "epoch": 0.7, + "learning_rate": 4.447635419553357e-06, + "loss": 0.9424, + "step": 21787 + }, + { + "epoch": 0.7, + "learning_rate": 4.446773979124816e-06, + "loss": 0.4685, + "step": 21788 + }, + { + "epoch": 0.7, + "learning_rate": 4.4459125982762406e-06, + "loss": 0.8223, + "step": 21789 + }, + { + "epoch": 0.7, + "learning_rate": 4.445051277016881e-06, + "loss": 0.9316, + "step": 21790 + }, + { + "epoch": 0.7, + "learning_rate": 4.444190015355971e-06, + "loss": 1.0298, + "step": 21791 + }, + { + "epoch": 0.7, + "learning_rate": 4.4433288133027555e-06, + "loss": 0.9399, + "step": 21792 + }, + { + "epoch": 0.7, + "learning_rate": 4.442467670866467e-06, + "loss": 0.9277, + "step": 21793 + }, + { + "epoch": 0.7, + "learning_rate": 4.4416065880563545e-06, + "loss": 0.8862, + "step": 21794 + }, + { + "epoch": 0.7, + "learning_rate": 4.440745564881646e-06, + "loss": 0.9429, + "step": 21795 + }, + { + "epoch": 0.7, + "learning_rate": 4.439884601351591e-06, + "loss": 0.8896, + "step": 21796 + }, + { + "epoch": 0.7, + "learning_rate": 4.4390236974754185e-06, + "loss": 0.8843, + "step": 21797 + }, + { + "epoch": 0.7, + "learning_rate": 4.438162853262364e-06, + "loss": 0.8965, + "step": 21798 + }, + { + "epoch": 0.7, + "learning_rate": 4.437302068721671e-06, + "loss": 0.8311, + "step": 21799 + }, + { + "epoch": 0.7, + "learning_rate": 4.436441343862569e-06, + "loss": 0.9116, + "step": 21800 + }, + { + "epoch": 0.7, + "learning_rate": 4.4355806786942925e-06, + "loss": 0.8088, + "step": 21801 + }, + { + "epoch": 0.7, + "learning_rate": 4.4347200732260785e-06, + "loss": 0.918, + "step": 21802 + }, + { + "epoch": 0.7, + "learning_rate": 4.43385952746716e-06, + "loss": 0.9937, + "step": 21803 + }, + { + "epoch": 0.7, + "learning_rate": 4.432999041426764e-06, + "loss": 0.9438, + "step": 21804 + }, + { + "epoch": 0.7, + "learning_rate": 4.432138615114131e-06, + "loss": 1.0508, + "step": 21805 + }, + { + "epoch": 0.7, + "learning_rate": 4.431278248538485e-06, + "loss": 1.04, + "step": 21806 + }, + { + "epoch": 0.7, + "learning_rate": 4.430417941709068e-06, + "loss": 0.9824, + "step": 21807 + }, + { + "epoch": 0.7, + "learning_rate": 4.429557694635093e-06, + "loss": 0.9556, + "step": 21808 + }, + { + "epoch": 0.7, + "learning_rate": 4.4286975073258045e-06, + "loss": 0.9287, + "step": 21809 + }, + { + "epoch": 0.7, + "learning_rate": 4.427837379790422e-06, + "loss": 0.9604, + "step": 21810 + }, + { + "epoch": 0.7, + "learning_rate": 4.42697731203818e-06, + "loss": 0.9814, + "step": 21811 + }, + { + "epoch": 0.7, + "learning_rate": 4.426117304078301e-06, + "loss": 0.8511, + "step": 21812 + }, + { + "epoch": 0.7, + "learning_rate": 4.425257355920018e-06, + "loss": 1.0532, + "step": 21813 + }, + { + "epoch": 0.7, + "learning_rate": 4.424397467572549e-06, + "loss": 1.0186, + "step": 21814 + }, + { + "epoch": 0.7, + "learning_rate": 4.423537639045132e-06, + "loss": 0.9756, + "step": 21815 + }, + { + "epoch": 0.7, + "learning_rate": 4.422677870346977e-06, + "loss": 1.061, + "step": 21816 + }, + { + "epoch": 0.7, + "learning_rate": 4.421818161487322e-06, + "loss": 0.8936, + "step": 21817 + }, + { + "epoch": 0.7, + "learning_rate": 4.420958512475378e-06, + "loss": 0.7744, + "step": 21818 + }, + { + "epoch": 0.7, + "learning_rate": 4.420098923320378e-06, + "loss": 0.9502, + "step": 21819 + }, + { + "epoch": 0.7, + "learning_rate": 4.41923939403154e-06, + "loss": 0.9404, + "step": 21820 + }, + { + "epoch": 0.7, + "learning_rate": 4.418379924618088e-06, + "loss": 0.9092, + "step": 21821 + }, + { + "epoch": 0.7, + "learning_rate": 4.417520515089239e-06, + "loss": 1.0444, + "step": 21822 + }, + { + "epoch": 0.7, + "learning_rate": 4.416661165454225e-06, + "loss": 0.9487, + "step": 21823 + }, + { + "epoch": 0.7, + "learning_rate": 4.4158018757222475e-06, + "loss": 1.0835, + "step": 21824 + }, + { + "epoch": 0.7, + "learning_rate": 4.414942645902541e-06, + "loss": 0.9507, + "step": 21825 + }, + { + "epoch": 0.7, + "learning_rate": 4.414083476004315e-06, + "loss": 1.0342, + "step": 21826 + }, + { + "epoch": 0.7, + "learning_rate": 4.413224366036794e-06, + "loss": 1.0342, + "step": 21827 + }, + { + "epoch": 0.7, + "learning_rate": 4.412365316009191e-06, + "loss": 0.8643, + "step": 21828 + }, + { + "epoch": 0.7, + "learning_rate": 4.411506325930726e-06, + "loss": 0.98, + "step": 21829 + }, + { + "epoch": 0.7, + "learning_rate": 4.410647395810612e-06, + "loss": 0.8462, + "step": 21830 + }, + { + "epoch": 0.7, + "learning_rate": 4.409788525658068e-06, + "loss": 0.8379, + "step": 21831 + }, + { + "epoch": 0.7, + "learning_rate": 4.4089297154822994e-06, + "loss": 0.9609, + "step": 21832 + }, + { + "epoch": 0.7, + "learning_rate": 4.408070965292534e-06, + "loss": 0.873, + "step": 21833 + }, + { + "epoch": 0.7, + "learning_rate": 4.407212275097972e-06, + "loss": 0.7573, + "step": 21834 + }, + { + "epoch": 0.7, + "learning_rate": 4.406353644907837e-06, + "loss": 1.0039, + "step": 21835 + }, + { + "epoch": 0.7, + "learning_rate": 4.405495074731338e-06, + "loss": 0.9849, + "step": 21836 + }, + { + "epoch": 0.7, + "learning_rate": 4.40463656457768e-06, + "loss": 0.9453, + "step": 21837 + }, + { + "epoch": 0.7, + "learning_rate": 4.403778114456083e-06, + "loss": 1.0054, + "step": 21838 + }, + { + "epoch": 0.7, + "learning_rate": 4.402919724375753e-06, + "loss": 0.8975, + "step": 21839 + }, + { + "epoch": 0.7, + "learning_rate": 4.402061394345897e-06, + "loss": 0.9434, + "step": 21840 + }, + { + "epoch": 0.7, + "learning_rate": 4.401203124375729e-06, + "loss": 0.9839, + "step": 21841 + }, + { + "epoch": 0.7, + "learning_rate": 4.400344914474456e-06, + "loss": 1.0381, + "step": 21842 + }, + { + "epoch": 0.7, + "learning_rate": 4.399486764651282e-06, + "loss": 0.978, + "step": 21843 + }, + { + "epoch": 0.7, + "learning_rate": 4.398628674915418e-06, + "loss": 0.9053, + "step": 21844 + }, + { + "epoch": 0.7, + "learning_rate": 4.397770645276071e-06, + "loss": 0.9878, + "step": 21845 + }, + { + "epoch": 0.7, + "learning_rate": 4.396912675742444e-06, + "loss": 0.8994, + "step": 21846 + }, + { + "epoch": 0.7, + "learning_rate": 4.39605476632374e-06, + "loss": 0.9429, + "step": 21847 + }, + { + "epoch": 0.7, + "learning_rate": 4.395196917029169e-06, + "loss": 1.0752, + "step": 21848 + }, + { + "epoch": 0.7, + "learning_rate": 4.394339127867927e-06, + "loss": 1.0913, + "step": 21849 + }, + { + "epoch": 0.7, + "learning_rate": 4.393481398849228e-06, + "loss": 0.9707, + "step": 21850 + }, + { + "epoch": 0.7, + "learning_rate": 4.3926237299822635e-06, + "loss": 0.9165, + "step": 21851 + }, + { + "epoch": 0.7, + "learning_rate": 4.391766121276245e-06, + "loss": 1.0459, + "step": 21852 + }, + { + "epoch": 0.7, + "learning_rate": 4.3909085727403686e-06, + "loss": 0.8652, + "step": 21853 + }, + { + "epoch": 0.7, + "learning_rate": 4.390051084383836e-06, + "loss": 0.9229, + "step": 21854 + }, + { + "epoch": 0.7, + "learning_rate": 4.389193656215842e-06, + "loss": 0.8398, + "step": 21855 + }, + { + "epoch": 0.7, + "learning_rate": 4.388336288245595e-06, + "loss": 0.9009, + "step": 21856 + }, + { + "epoch": 0.7, + "learning_rate": 4.387478980482285e-06, + "loss": 0.9551, + "step": 21857 + }, + { + "epoch": 0.7, + "learning_rate": 4.386621732935117e-06, + "loss": 1.0054, + "step": 21858 + }, + { + "epoch": 0.7, + "learning_rate": 4.3857645456132815e-06, + "loss": 0.9302, + "step": 21859 + }, + { + "epoch": 0.7, + "learning_rate": 4.384907418525987e-06, + "loss": 0.9819, + "step": 21860 + }, + { + "epoch": 0.7, + "learning_rate": 4.384050351682413e-06, + "loss": 0.4597, + "step": 21861 + }, + { + "epoch": 0.7, + "learning_rate": 4.383193345091767e-06, + "loss": 0.9624, + "step": 21862 + }, + { + "epoch": 0.7, + "learning_rate": 4.382336398763238e-06, + "loss": 0.8921, + "step": 21863 + }, + { + "epoch": 0.7, + "learning_rate": 4.381479512706025e-06, + "loss": 0.9873, + "step": 21864 + }, + { + "epoch": 0.7, + "learning_rate": 4.380622686929314e-06, + "loss": 0.9814, + "step": 21865 + }, + { + "epoch": 0.7, + "learning_rate": 4.379765921442307e-06, + "loss": 0.9795, + "step": 21866 + }, + { + "epoch": 0.7, + "learning_rate": 4.378909216254188e-06, + "loss": 0.9248, + "step": 21867 + }, + { + "epoch": 0.7, + "learning_rate": 4.378052571374159e-06, + "loss": 1.0195, + "step": 21868 + }, + { + "epoch": 0.7, + "learning_rate": 4.3771959868113956e-06, + "loss": 0.9932, + "step": 21869 + }, + { + "epoch": 0.7, + "learning_rate": 4.376339462575101e-06, + "loss": 0.9888, + "step": 21870 + }, + { + "epoch": 0.7, + "learning_rate": 4.375482998674456e-06, + "loss": 0.7861, + "step": 21871 + }, + { + "epoch": 0.7, + "learning_rate": 4.374626595118657e-06, + "loss": 1.0894, + "step": 21872 + }, + { + "epoch": 0.7, + "learning_rate": 4.373770251916885e-06, + "loss": 0.9595, + "step": 21873 + }, + { + "epoch": 0.7, + "learning_rate": 4.372913969078336e-06, + "loss": 0.9028, + "step": 21874 + }, + { + "epoch": 0.7, + "learning_rate": 4.372057746612191e-06, + "loss": 0.9321, + "step": 21875 + }, + { + "epoch": 0.7, + "learning_rate": 4.371201584527639e-06, + "loss": 0.9048, + "step": 21876 + }, + { + "epoch": 0.7, + "learning_rate": 4.37034548283386e-06, + "loss": 0.9243, + "step": 21877 + }, + { + "epoch": 0.7, + "learning_rate": 4.369489441540047e-06, + "loss": 0.9072, + "step": 21878 + }, + { + "epoch": 0.7, + "learning_rate": 4.368633460655377e-06, + "loss": 0.897, + "step": 21879 + }, + { + "epoch": 0.7, + "learning_rate": 4.367777540189042e-06, + "loss": 0.7158, + "step": 21880 + }, + { + "epoch": 0.7, + "learning_rate": 4.36692168015022e-06, + "loss": 0.9819, + "step": 21881 + }, + { + "epoch": 0.7, + "learning_rate": 4.366065880548091e-06, + "loss": 0.8916, + "step": 21882 + }, + { + "epoch": 0.7, + "learning_rate": 4.365210141391841e-06, + "loss": 0.8267, + "step": 21883 + }, + { + "epoch": 0.7, + "learning_rate": 4.364354462690652e-06, + "loss": 0.9561, + "step": 21884 + }, + { + "epoch": 0.7, + "learning_rate": 4.363498844453702e-06, + "loss": 0.9111, + "step": 21885 + }, + { + "epoch": 0.7, + "learning_rate": 4.362643286690168e-06, + "loss": 0.9717, + "step": 21886 + }, + { + "epoch": 0.7, + "learning_rate": 4.3617877894092355e-06, + "loss": 0.9517, + "step": 21887 + }, + { + "epoch": 0.7, + "learning_rate": 4.360932352620076e-06, + "loss": 0.9395, + "step": 21888 + }, + { + "epoch": 0.7, + "learning_rate": 4.360076976331876e-06, + "loss": 0.9487, + "step": 21889 + }, + { + "epoch": 0.7, + "learning_rate": 4.359221660553807e-06, + "loss": 1.0381, + "step": 21890 + }, + { + "epoch": 0.7, + "learning_rate": 4.358366405295048e-06, + "loss": 1.0098, + "step": 21891 + }, + { + "epoch": 0.7, + "learning_rate": 4.3575112105647685e-06, + "loss": 0.9888, + "step": 21892 + }, + { + "epoch": 0.7, + "learning_rate": 4.356656076372153e-06, + "loss": 0.9272, + "step": 21893 + }, + { + "epoch": 0.7, + "learning_rate": 4.355801002726369e-06, + "loss": 0.9575, + "step": 21894 + }, + { + "epoch": 0.7, + "learning_rate": 4.354945989636596e-06, + "loss": 0.8511, + "step": 21895 + }, + { + "epoch": 0.7, + "learning_rate": 4.354091037112002e-06, + "loss": 1.0264, + "step": 21896 + }, + { + "epoch": 0.7, + "learning_rate": 4.3532361451617674e-06, + "loss": 0.9116, + "step": 21897 + }, + { + "epoch": 0.7, + "learning_rate": 4.3523813137950585e-06, + "loss": 0.9888, + "step": 21898 + }, + { + "epoch": 0.7, + "learning_rate": 4.351526543021047e-06, + "loss": 0.9258, + "step": 21899 + }, + { + "epoch": 0.7, + "learning_rate": 4.3506718328489015e-06, + "loss": 0.8809, + "step": 21900 + }, + { + "epoch": 0.7, + "learning_rate": 4.349817183287798e-06, + "loss": 0.9165, + "step": 21901 + }, + { + "epoch": 0.7, + "learning_rate": 4.3489625943469e-06, + "loss": 0.917, + "step": 21902 + }, + { + "epoch": 0.7, + "learning_rate": 4.348108066035382e-06, + "loss": 1.0273, + "step": 21903 + }, + { + "epoch": 0.7, + "learning_rate": 4.347253598362405e-06, + "loss": 0.939, + "step": 21904 + }, + { + "epoch": 0.7, + "learning_rate": 4.346399191337145e-06, + "loss": 0.8125, + "step": 21905 + }, + { + "epoch": 0.7, + "learning_rate": 4.345544844968764e-06, + "loss": 0.9937, + "step": 21906 + }, + { + "epoch": 0.7, + "learning_rate": 4.34469055926643e-06, + "loss": 0.8525, + "step": 21907 + }, + { + "epoch": 0.7, + "learning_rate": 4.3438363342393034e-06, + "loss": 0.9692, + "step": 21908 + }, + { + "epoch": 0.7, + "learning_rate": 4.342982169896555e-06, + "loss": 0.9814, + "step": 21909 + }, + { + "epoch": 0.7, + "learning_rate": 4.342128066247345e-06, + "loss": 0.9302, + "step": 21910 + }, + { + "epoch": 0.7, + "learning_rate": 4.341274023300843e-06, + "loss": 0.9746, + "step": 21911 + }, + { + "epoch": 0.7, + "learning_rate": 4.340420041066203e-06, + "loss": 0.9033, + "step": 21912 + }, + { + "epoch": 0.7, + "learning_rate": 4.3395661195526004e-06, + "loss": 0.8745, + "step": 21913 + }, + { + "epoch": 0.7, + "learning_rate": 4.338712258769182e-06, + "loss": 0.8892, + "step": 21914 + }, + { + "epoch": 0.7, + "learning_rate": 4.337858458725118e-06, + "loss": 0.9194, + "step": 21915 + }, + { + "epoch": 0.7, + "learning_rate": 4.337004719429563e-06, + "loss": 0.8198, + "step": 21916 + }, + { + "epoch": 0.7, + "learning_rate": 4.336151040891683e-06, + "loss": 1.0122, + "step": 21917 + }, + { + "epoch": 0.7, + "learning_rate": 4.335297423120631e-06, + "loss": 0.8716, + "step": 21918 + }, + { + "epoch": 0.7, + "learning_rate": 4.334443866125572e-06, + "loss": 1.0762, + "step": 21919 + }, + { + "epoch": 0.7, + "learning_rate": 4.333590369915659e-06, + "loss": 0.8931, + "step": 21920 + }, + { + "epoch": 0.7, + "learning_rate": 4.3327369345000505e-06, + "loss": 0.8308, + "step": 21921 + }, + { + "epoch": 0.7, + "learning_rate": 4.331883559887899e-06, + "loss": 1.0225, + "step": 21922 + }, + { + "epoch": 0.7, + "learning_rate": 4.331030246088367e-06, + "loss": 0.9907, + "step": 21923 + }, + { + "epoch": 0.7, + "learning_rate": 4.330176993110608e-06, + "loss": 1.0713, + "step": 21924 + }, + { + "epoch": 0.7, + "learning_rate": 4.329323800963769e-06, + "loss": 0.7334, + "step": 21925 + }, + { + "epoch": 0.7, + "learning_rate": 4.328470669657014e-06, + "loss": 1.0356, + "step": 21926 + }, + { + "epoch": 0.7, + "learning_rate": 4.327617599199489e-06, + "loss": 0.9678, + "step": 21927 + }, + { + "epoch": 0.7, + "learning_rate": 4.3267645896003515e-06, + "loss": 0.9385, + "step": 21928 + }, + { + "epoch": 0.7, + "learning_rate": 4.325911640868752e-06, + "loss": 0.9854, + "step": 21929 + }, + { + "epoch": 0.7, + "learning_rate": 4.32505875301384e-06, + "loss": 0.9031, + "step": 21930 + }, + { + "epoch": 0.7, + "learning_rate": 4.3242059260447646e-06, + "loss": 0.5044, + "step": 21931 + }, + { + "epoch": 0.7, + "learning_rate": 4.32335315997068e-06, + "loss": 0.9541, + "step": 21932 + }, + { + "epoch": 0.7, + "learning_rate": 4.322500454800731e-06, + "loss": 0.9072, + "step": 21933 + }, + { + "epoch": 0.7, + "learning_rate": 4.321647810544072e-06, + "loss": 0.8931, + "step": 21934 + }, + { + "epoch": 0.7, + "learning_rate": 4.320795227209843e-06, + "loss": 0.9204, + "step": 21935 + }, + { + "epoch": 0.7, + "learning_rate": 4.319942704807204e-06, + "loss": 0.9604, + "step": 21936 + }, + { + "epoch": 0.7, + "learning_rate": 4.3190902433452855e-06, + "loss": 0.8398, + "step": 21937 + }, + { + "epoch": 0.7, + "learning_rate": 4.318237842833246e-06, + "loss": 0.9868, + "step": 21938 + }, + { + "epoch": 0.7, + "learning_rate": 4.317385503280221e-06, + "loss": 0.498, + "step": 21939 + }, + { + "epoch": 0.7, + "learning_rate": 4.316533224695365e-06, + "loss": 0.7803, + "step": 21940 + }, + { + "epoch": 0.7, + "learning_rate": 4.315681007087814e-06, + "loss": 0.9448, + "step": 21941 + }, + { + "epoch": 0.7, + "learning_rate": 4.314828850466718e-06, + "loss": 0.9702, + "step": 21942 + }, + { + "epoch": 0.7, + "learning_rate": 4.313976754841216e-06, + "loss": 0.9219, + "step": 21943 + }, + { + "epoch": 0.7, + "learning_rate": 4.313124720220451e-06, + "loss": 0.8916, + "step": 21944 + }, + { + "epoch": 0.7, + "learning_rate": 4.3122727466135596e-06, + "loss": 1.0259, + "step": 21945 + }, + { + "epoch": 0.7, + "learning_rate": 4.311420834029692e-06, + "loss": 0.9941, + "step": 21946 + }, + { + "epoch": 0.7, + "learning_rate": 4.3105689824779775e-06, + "loss": 0.9194, + "step": 21947 + }, + { + "epoch": 0.7, + "learning_rate": 4.3097171919675655e-06, + "loss": 0.8853, + "step": 21948 + }, + { + "epoch": 0.7, + "learning_rate": 4.308865462507587e-06, + "loss": 0.978, + "step": 21949 + }, + { + "epoch": 0.7, + "learning_rate": 4.308013794107185e-06, + "loss": 0.8848, + "step": 21950 + }, + { + "epoch": 0.7, + "learning_rate": 4.3071621867754975e-06, + "loss": 0.8845, + "step": 21951 + }, + { + "epoch": 0.7, + "learning_rate": 4.306310640521659e-06, + "loss": 0.8857, + "step": 21952 + }, + { + "epoch": 0.7, + "learning_rate": 4.305459155354802e-06, + "loss": 1.1162, + "step": 21953 + }, + { + "epoch": 0.7, + "learning_rate": 4.30460773128407e-06, + "loss": 0.9746, + "step": 21954 + }, + { + "epoch": 0.7, + "learning_rate": 4.303756368318589e-06, + "loss": 0.8826, + "step": 21955 + }, + { + "epoch": 0.7, + "learning_rate": 4.302905066467502e-06, + "loss": 0.7554, + "step": 21956 + }, + { + "epoch": 0.7, + "learning_rate": 4.3020538257399345e-06, + "loss": 1.0103, + "step": 21957 + }, + { + "epoch": 0.7, + "learning_rate": 4.301202646145032e-06, + "loss": 0.9932, + "step": 21958 + }, + { + "epoch": 0.7, + "learning_rate": 4.300351527691909e-06, + "loss": 0.9292, + "step": 21959 + }, + { + "epoch": 0.7, + "learning_rate": 4.29950047038971e-06, + "loss": 1.002, + "step": 21960 + }, + { + "epoch": 0.7, + "learning_rate": 4.298649474247559e-06, + "loss": 0.8042, + "step": 21961 + }, + { + "epoch": 0.7, + "learning_rate": 4.2977985392745925e-06, + "loss": 0.938, + "step": 21962 + }, + { + "epoch": 0.7, + "learning_rate": 4.296947665479937e-06, + "loss": 0.9194, + "step": 21963 + }, + { + "epoch": 0.7, + "learning_rate": 4.296096852872716e-06, + "loss": 0.9404, + "step": 21964 + }, + { + "epoch": 0.7, + "learning_rate": 4.295246101462069e-06, + "loss": 0.9209, + "step": 21965 + }, + { + "epoch": 0.7, + "learning_rate": 4.294395411257116e-06, + "loss": 0.9209, + "step": 21966 + }, + { + "epoch": 0.7, + "learning_rate": 4.293544782266983e-06, + "loss": 0.8721, + "step": 21967 + }, + { + "epoch": 0.7, + "learning_rate": 4.292694214500802e-06, + "loss": 1.0518, + "step": 21968 + }, + { + "epoch": 0.7, + "learning_rate": 4.291843707967696e-06, + "loss": 0.8315, + "step": 21969 + }, + { + "epoch": 0.7, + "learning_rate": 4.290993262676785e-06, + "loss": 0.9614, + "step": 21970 + }, + { + "epoch": 0.7, + "learning_rate": 4.290142878637202e-06, + "loss": 0.9438, + "step": 21971 + }, + { + "epoch": 0.7, + "learning_rate": 4.289292555858063e-06, + "loss": 0.9712, + "step": 21972 + }, + { + "epoch": 0.7, + "learning_rate": 4.288442294348498e-06, + "loss": 0.9644, + "step": 21973 + }, + { + "epoch": 0.7, + "learning_rate": 4.287592094117626e-06, + "loss": 0.8892, + "step": 21974 + }, + { + "epoch": 0.7, + "learning_rate": 4.286741955174569e-06, + "loss": 1.0093, + "step": 21975 + }, + { + "epoch": 0.7, + "learning_rate": 4.285891877528444e-06, + "loss": 0.8491, + "step": 21976 + }, + { + "epoch": 0.7, + "learning_rate": 4.285041861188378e-06, + "loss": 0.9268, + "step": 21977 + }, + { + "epoch": 0.7, + "learning_rate": 4.2841919061634855e-06, + "loss": 0.9219, + "step": 21978 + }, + { + "epoch": 0.7, + "learning_rate": 4.283342012462891e-06, + "loss": 0.938, + "step": 21979 + }, + { + "epoch": 0.7, + "learning_rate": 4.2824921800957055e-06, + "loss": 0.8909, + "step": 21980 + }, + { + "epoch": 0.7, + "learning_rate": 4.281642409071058e-06, + "loss": 0.9614, + "step": 21981 + }, + { + "epoch": 0.7, + "learning_rate": 4.2807926993980534e-06, + "loss": 0.9492, + "step": 21982 + }, + { + "epoch": 0.7, + "learning_rate": 4.279943051085815e-06, + "loss": 0.9604, + "step": 21983 + }, + { + "epoch": 0.7, + "learning_rate": 4.279093464143455e-06, + "loss": 0.7532, + "step": 21984 + }, + { + "epoch": 0.7, + "learning_rate": 4.278243938580093e-06, + "loss": 0.9795, + "step": 21985 + }, + { + "epoch": 0.7, + "learning_rate": 4.277394474404838e-06, + "loss": 0.8701, + "step": 21986 + }, + { + "epoch": 0.7, + "learning_rate": 4.2765450716268105e-06, + "loss": 0.9673, + "step": 21987 + }, + { + "epoch": 0.7, + "learning_rate": 4.275695730255116e-06, + "loss": 0.832, + "step": 21988 + }, + { + "epoch": 0.7, + "learning_rate": 4.274846450298879e-06, + "loss": 0.8555, + "step": 21989 + }, + { + "epoch": 0.7, + "learning_rate": 4.273997231767195e-06, + "loss": 1.0498, + "step": 21990 + }, + { + "epoch": 0.7, + "learning_rate": 4.273148074669188e-06, + "loss": 0.939, + "step": 21991 + }, + { + "epoch": 0.7, + "learning_rate": 4.2722989790139595e-06, + "loss": 1.0596, + "step": 21992 + }, + { + "epoch": 0.7, + "learning_rate": 4.271449944810627e-06, + "loss": 0.9214, + "step": 21993 + }, + { + "epoch": 0.7, + "learning_rate": 4.270600972068294e-06, + "loss": 0.9307, + "step": 21994 + }, + { + "epoch": 0.7, + "learning_rate": 4.269752060796075e-06, + "loss": 0.5215, + "step": 21995 + }, + { + "epoch": 0.7, + "learning_rate": 4.2689032110030745e-06, + "loss": 0.9512, + "step": 21996 + }, + { + "epoch": 0.7, + "learning_rate": 4.2680544226984e-06, + "loss": 0.9292, + "step": 21997 + }, + { + "epoch": 0.7, + "learning_rate": 4.267205695891154e-06, + "loss": 0.9204, + "step": 21998 + }, + { + "epoch": 0.7, + "learning_rate": 4.266357030590449e-06, + "loss": 0.9653, + "step": 21999 + }, + { + "epoch": 0.7, + "learning_rate": 4.265508426805385e-06, + "loss": 0.8867, + "step": 22000 + }, + { + "epoch": 0.7, + "learning_rate": 4.264659884545071e-06, + "loss": 0.9312, + "step": 22001 + }, + { + "epoch": 0.7, + "learning_rate": 4.263811403818606e-06, + "loss": 0.9346, + "step": 22002 + }, + { + "epoch": 0.7, + "learning_rate": 4.2629629846351e-06, + "loss": 0.9785, + "step": 22003 + }, + { + "epoch": 0.7, + "learning_rate": 4.262114627003652e-06, + "loss": 0.9648, + "step": 22004 + }, + { + "epoch": 0.7, + "learning_rate": 4.261266330933363e-06, + "loss": 0.9058, + "step": 22005 + }, + { + "epoch": 0.7, + "learning_rate": 4.260418096433332e-06, + "loss": 0.9966, + "step": 22006 + }, + { + "epoch": 0.7, + "learning_rate": 4.259569923512665e-06, + "loss": 0.9844, + "step": 22007 + }, + { + "epoch": 0.7, + "learning_rate": 4.258721812180461e-06, + "loss": 0.9258, + "step": 22008 + }, + { + "epoch": 0.7, + "learning_rate": 4.257873762445814e-06, + "loss": 0.8994, + "step": 22009 + }, + { + "epoch": 0.7, + "learning_rate": 4.2570257743178285e-06, + "loss": 0.887, + "step": 22010 + }, + { + "epoch": 0.7, + "learning_rate": 4.256177847805603e-06, + "loss": 0.9253, + "step": 22011 + }, + { + "epoch": 0.7, + "learning_rate": 4.255329982918226e-06, + "loss": 1.0801, + "step": 22012 + }, + { + "epoch": 0.7, + "learning_rate": 4.254482179664805e-06, + "loss": 1.0225, + "step": 22013 + }, + { + "epoch": 0.7, + "learning_rate": 4.253634438054433e-06, + "loss": 1.0176, + "step": 22014 + }, + { + "epoch": 0.7, + "learning_rate": 4.252786758096198e-06, + "loss": 0.4082, + "step": 22015 + }, + { + "epoch": 0.7, + "learning_rate": 4.251939139799205e-06, + "loss": 0.9854, + "step": 22016 + }, + { + "epoch": 0.7, + "learning_rate": 4.251091583172538e-06, + "loss": 0.9863, + "step": 22017 + }, + { + "epoch": 0.7, + "learning_rate": 4.250244088225302e-06, + "loss": 0.9434, + "step": 22018 + }, + { + "epoch": 0.7, + "learning_rate": 4.249396654966582e-06, + "loss": 1.0088, + "step": 22019 + }, + { + "epoch": 0.7, + "learning_rate": 4.248549283405472e-06, + "loss": 0.8271, + "step": 22020 + }, + { + "epoch": 0.7, + "learning_rate": 4.247701973551058e-06, + "loss": 0.9072, + "step": 22021 + }, + { + "epoch": 0.7, + "learning_rate": 4.246854725412441e-06, + "loss": 0.8325, + "step": 22022 + }, + { + "epoch": 0.7, + "learning_rate": 4.2460075389987e-06, + "loss": 0.8101, + "step": 22023 + }, + { + "epoch": 0.7, + "learning_rate": 4.245160414318935e-06, + "loss": 0.9976, + "step": 22024 + }, + { + "epoch": 0.7, + "learning_rate": 4.244313351382225e-06, + "loss": 0.9463, + "step": 22025 + }, + { + "epoch": 0.7, + "learning_rate": 4.243466350197671e-06, + "loss": 0.9648, + "step": 22026 + }, + { + "epoch": 0.7, + "learning_rate": 4.242619410774344e-06, + "loss": 0.9551, + "step": 22027 + }, + { + "epoch": 0.7, + "learning_rate": 4.2417725331213425e-06, + "loss": 0.8965, + "step": 22028 + }, + { + "epoch": 0.7, + "learning_rate": 4.240925717247745e-06, + "loss": 1.0049, + "step": 22029 + }, + { + "epoch": 0.7, + "learning_rate": 4.240078963162644e-06, + "loss": 0.4771, + "step": 22030 + }, + { + "epoch": 0.7, + "learning_rate": 4.2392322708751176e-06, + "loss": 0.5168, + "step": 22031 + }, + { + "epoch": 0.7, + "learning_rate": 4.238385640394258e-06, + "loss": 1.0059, + "step": 22032 + }, + { + "epoch": 0.7, + "learning_rate": 4.237539071729138e-06, + "loss": 0.9204, + "step": 22033 + }, + { + "epoch": 0.7, + "learning_rate": 4.236692564888855e-06, + "loss": 0.9263, + "step": 22034 + }, + { + "epoch": 0.7, + "learning_rate": 4.235846119882475e-06, + "loss": 0.8779, + "step": 22035 + }, + { + "epoch": 0.7, + "learning_rate": 4.234999736719091e-06, + "loss": 1.0513, + "step": 22036 + }, + { + "epoch": 0.7, + "learning_rate": 4.2341534154077745e-06, + "loss": 0.9341, + "step": 22037 + }, + { + "epoch": 0.7, + "learning_rate": 4.233307155957614e-06, + "loss": 0.9507, + "step": 22038 + }, + { + "epoch": 0.7, + "learning_rate": 4.232460958377683e-06, + "loss": 0.9521, + "step": 22039 + }, + { + "epoch": 0.7, + "learning_rate": 4.231614822677066e-06, + "loss": 0.9478, + "step": 22040 + }, + { + "epoch": 0.7, + "learning_rate": 4.2307687488648375e-06, + "loss": 0.8687, + "step": 22041 + }, + { + "epoch": 0.7, + "learning_rate": 4.229922736950075e-06, + "loss": 0.9521, + "step": 22042 + }, + { + "epoch": 0.7, + "learning_rate": 4.2290767869418525e-06, + "loss": 0.873, + "step": 22043 + }, + { + "epoch": 0.7, + "learning_rate": 4.228230898849253e-06, + "loss": 0.9771, + "step": 22044 + }, + { + "epoch": 0.71, + "learning_rate": 4.227385072681344e-06, + "loss": 0.9253, + "step": 22045 + }, + { + "epoch": 0.71, + "learning_rate": 4.226539308447207e-06, + "loss": 0.856, + "step": 22046 + }, + { + "epoch": 0.71, + "learning_rate": 4.225693606155915e-06, + "loss": 0.8359, + "step": 22047 + }, + { + "epoch": 0.71, + "learning_rate": 4.2248479658165355e-06, + "loss": 1.0049, + "step": 22048 + }, + { + "epoch": 0.71, + "learning_rate": 4.22400238743815e-06, + "loss": 0.9556, + "step": 22049 + }, + { + "epoch": 0.71, + "learning_rate": 4.223156871029825e-06, + "loss": 0.9336, + "step": 22050 + }, + { + "epoch": 0.71, + "learning_rate": 4.22231141660063e-06, + "loss": 0.8193, + "step": 22051 + }, + { + "epoch": 0.71, + "learning_rate": 4.221466024159644e-06, + "loss": 0.9595, + "step": 22052 + }, + { + "epoch": 0.71, + "learning_rate": 4.220620693715931e-06, + "loss": 0.9819, + "step": 22053 + }, + { + "epoch": 0.71, + "learning_rate": 4.2197754252785586e-06, + "loss": 1.1016, + "step": 22054 + }, + { + "epoch": 0.71, + "learning_rate": 4.218930218856602e-06, + "loss": 0.874, + "step": 22055 + }, + { + "epoch": 0.71, + "learning_rate": 4.2180850744591225e-06, + "loss": 0.9937, + "step": 22056 + }, + { + "epoch": 0.71, + "learning_rate": 4.217239992095195e-06, + "loss": 0.9565, + "step": 22057 + }, + { + "epoch": 0.71, + "learning_rate": 4.216394971773882e-06, + "loss": 0.915, + "step": 22058 + }, + { + "epoch": 0.71, + "learning_rate": 4.2155500135042495e-06, + "loss": 0.9395, + "step": 22059 + }, + { + "epoch": 0.71, + "learning_rate": 4.21470511729536e-06, + "loss": 1.0127, + "step": 22060 + }, + { + "epoch": 0.71, + "learning_rate": 4.213860283156286e-06, + "loss": 0.9004, + "step": 22061 + }, + { + "epoch": 0.71, + "learning_rate": 4.213015511096083e-06, + "loss": 0.9746, + "step": 22062 + }, + { + "epoch": 0.71, + "learning_rate": 4.212170801123824e-06, + "loss": 0.8711, + "step": 22063 + }, + { + "epoch": 0.71, + "learning_rate": 4.211326153248565e-06, + "loss": 0.938, + "step": 22064 + }, + { + "epoch": 0.71, + "learning_rate": 4.210481567479371e-06, + "loss": 0.9189, + "step": 22065 + }, + { + "epoch": 0.71, + "learning_rate": 4.209637043825297e-06, + "loss": 0.8955, + "step": 22066 + }, + { + "epoch": 0.71, + "learning_rate": 4.2087925822954136e-06, + "loss": 0.8877, + "step": 22067 + }, + { + "epoch": 0.71, + "learning_rate": 4.207948182898774e-06, + "loss": 0.9634, + "step": 22068 + }, + { + "epoch": 0.71, + "learning_rate": 4.2071038456444415e-06, + "loss": 0.9531, + "step": 22069 + }, + { + "epoch": 0.71, + "learning_rate": 4.206259570541471e-06, + "loss": 0.9966, + "step": 22070 + }, + { + "epoch": 0.71, + "learning_rate": 4.205415357598926e-06, + "loss": 0.9136, + "step": 22071 + }, + { + "epoch": 0.71, + "learning_rate": 4.2045712068258625e-06, + "loss": 0.9761, + "step": 22072 + }, + { + "epoch": 0.71, + "learning_rate": 4.2037271182313344e-06, + "loss": 0.8989, + "step": 22073 + }, + { + "epoch": 0.71, + "learning_rate": 4.202883091824397e-06, + "loss": 0.9048, + "step": 22074 + }, + { + "epoch": 0.71, + "learning_rate": 4.202039127614111e-06, + "loss": 1.0845, + "step": 22075 + }, + { + "epoch": 0.71, + "learning_rate": 4.2011952256095245e-06, + "loss": 0.9272, + "step": 22076 + }, + { + "epoch": 0.71, + "learning_rate": 4.2003513858197e-06, + "loss": 0.9722, + "step": 22077 + }, + { + "epoch": 0.71, + "learning_rate": 4.199507608253681e-06, + "loss": 0.9829, + "step": 22078 + }, + { + "epoch": 0.71, + "learning_rate": 4.198663892920535e-06, + "loss": 0.8447, + "step": 22079 + }, + { + "epoch": 0.71, + "learning_rate": 4.197820239829295e-06, + "loss": 0.8809, + "step": 22080 + }, + { + "epoch": 0.71, + "learning_rate": 4.1969766489890285e-06, + "loss": 1.0605, + "step": 22081 + }, + { + "epoch": 0.71, + "learning_rate": 4.196133120408775e-06, + "loss": 0.5386, + "step": 22082 + }, + { + "epoch": 0.71, + "learning_rate": 4.195289654097594e-06, + "loss": 1.0605, + "step": 22083 + }, + { + "epoch": 0.71, + "learning_rate": 4.194446250064528e-06, + "loss": 0.9917, + "step": 22084 + }, + { + "epoch": 0.71, + "learning_rate": 4.19360290831863e-06, + "loss": 0.7126, + "step": 22085 + }, + { + "epoch": 0.71, + "learning_rate": 4.192759628868948e-06, + "loss": 0.9653, + "step": 22086 + }, + { + "epoch": 0.71, + "learning_rate": 4.191916411724527e-06, + "loss": 0.9434, + "step": 22087 + }, + { + "epoch": 0.71, + "learning_rate": 4.1910732568944125e-06, + "loss": 0.4724, + "step": 22088 + }, + { + "epoch": 0.71, + "learning_rate": 4.1902301643876555e-06, + "loss": 1.0249, + "step": 22089 + }, + { + "epoch": 0.71, + "learning_rate": 4.189387134213297e-06, + "loss": 0.9849, + "step": 22090 + }, + { + "epoch": 0.71, + "learning_rate": 4.188544166380385e-06, + "loss": 0.8838, + "step": 22091 + }, + { + "epoch": 0.71, + "learning_rate": 4.187701260897963e-06, + "loss": 0.8115, + "step": 22092 + }, + { + "epoch": 0.71, + "learning_rate": 4.186858417775071e-06, + "loss": 0.873, + "step": 22093 + }, + { + "epoch": 0.71, + "learning_rate": 4.1860156370207575e-06, + "loss": 1.0, + "step": 22094 + }, + { + "epoch": 0.71, + "learning_rate": 4.185172918644061e-06, + "loss": 0.9575, + "step": 22095 + }, + { + "epoch": 0.71, + "learning_rate": 4.18433026265402e-06, + "loss": 0.8906, + "step": 22096 + }, + { + "epoch": 0.71, + "learning_rate": 4.1834876690596835e-06, + "loss": 0.8481, + "step": 22097 + }, + { + "epoch": 0.71, + "learning_rate": 4.182645137870086e-06, + "loss": 0.938, + "step": 22098 + }, + { + "epoch": 0.71, + "learning_rate": 4.181802669094265e-06, + "loss": 1.063, + "step": 22099 + }, + { + "epoch": 0.71, + "learning_rate": 4.180960262741266e-06, + "loss": 1.0142, + "step": 22100 + }, + { + "epoch": 0.71, + "learning_rate": 4.180117918820118e-06, + "loss": 0.9502, + "step": 22101 + }, + { + "epoch": 0.71, + "learning_rate": 4.179275637339872e-06, + "loss": 0.9824, + "step": 22102 + }, + { + "epoch": 0.71, + "learning_rate": 4.178433418309549e-06, + "loss": 0.8267, + "step": 22103 + }, + { + "epoch": 0.71, + "learning_rate": 4.177591261738196e-06, + "loss": 0.9312, + "step": 22104 + }, + { + "epoch": 0.71, + "learning_rate": 4.176749167634841e-06, + "loss": 1.0649, + "step": 22105 + }, + { + "epoch": 0.71, + "learning_rate": 4.175907136008527e-06, + "loss": 0.9458, + "step": 22106 + }, + { + "epoch": 0.71, + "learning_rate": 4.175065166868278e-06, + "loss": 0.915, + "step": 22107 + }, + { + "epoch": 0.71, + "learning_rate": 4.174223260223139e-06, + "loss": 0.8521, + "step": 22108 + }, + { + "epoch": 0.71, + "learning_rate": 4.173381416082136e-06, + "loss": 0.9941, + "step": 22109 + }, + { + "epoch": 0.71, + "learning_rate": 4.172539634454301e-06, + "loss": 0.8647, + "step": 22110 + }, + { + "epoch": 0.71, + "learning_rate": 4.171697915348664e-06, + "loss": 0.9341, + "step": 22111 + }, + { + "epoch": 0.71, + "learning_rate": 4.170856258774261e-06, + "loss": 0.8877, + "step": 22112 + }, + { + "epoch": 0.71, + "learning_rate": 4.170014664740114e-06, + "loss": 0.9814, + "step": 22113 + }, + { + "epoch": 0.71, + "learning_rate": 4.169173133255262e-06, + "loss": 0.9341, + "step": 22114 + }, + { + "epoch": 0.71, + "learning_rate": 4.168331664328727e-06, + "loss": 1.1597, + "step": 22115 + }, + { + "epoch": 0.71, + "learning_rate": 4.167490257969539e-06, + "loss": 0.9404, + "step": 22116 + }, + { + "epoch": 0.71, + "learning_rate": 4.166648914186729e-06, + "loss": 0.9497, + "step": 22117 + }, + { + "epoch": 0.71, + "learning_rate": 4.165807632989318e-06, + "loss": 0.4661, + "step": 22118 + }, + { + "epoch": 0.71, + "learning_rate": 4.164966414386332e-06, + "loss": 0.897, + "step": 22119 + }, + { + "epoch": 0.71, + "learning_rate": 4.1641252583868006e-06, + "loss": 0.9346, + "step": 22120 + }, + { + "epoch": 0.71, + "learning_rate": 4.163284164999744e-06, + "loss": 0.9141, + "step": 22121 + }, + { + "epoch": 0.71, + "learning_rate": 4.162443134234192e-06, + "loss": 1.0376, + "step": 22122 + }, + { + "epoch": 0.71, + "learning_rate": 4.16160216609916e-06, + "loss": 0.9126, + "step": 22123 + }, + { + "epoch": 0.71, + "learning_rate": 4.160761260603683e-06, + "loss": 0.8477, + "step": 22124 + }, + { + "epoch": 0.71, + "learning_rate": 4.159920417756768e-06, + "loss": 0.895, + "step": 22125 + }, + { + "epoch": 0.71, + "learning_rate": 4.159079637567447e-06, + "loss": 0.9595, + "step": 22126 + }, + { + "epoch": 0.71, + "learning_rate": 4.158238920044733e-06, + "loss": 0.916, + "step": 22127 + }, + { + "epoch": 0.71, + "learning_rate": 4.1573982651976545e-06, + "loss": 0.9834, + "step": 22128 + }, + { + "epoch": 0.71, + "learning_rate": 4.1565576730352216e-06, + "loss": 0.8145, + "step": 22129 + }, + { + "epoch": 0.71, + "learning_rate": 4.155717143566461e-06, + "loss": 0.957, + "step": 22130 + }, + { + "epoch": 0.71, + "learning_rate": 4.154876676800389e-06, + "loss": 0.9741, + "step": 22131 + }, + { + "epoch": 0.71, + "learning_rate": 4.154036272746021e-06, + "loss": 0.9761, + "step": 22132 + }, + { + "epoch": 0.71, + "learning_rate": 4.153195931412368e-06, + "loss": 0.7168, + "step": 22133 + }, + { + "epoch": 0.71, + "learning_rate": 4.152355652808457e-06, + "loss": 0.9722, + "step": 22134 + }, + { + "epoch": 0.71, + "learning_rate": 4.151515436943295e-06, + "loss": 0.8081, + "step": 22135 + }, + { + "epoch": 0.71, + "learning_rate": 4.150675283825902e-06, + "loss": 0.9229, + "step": 22136 + }, + { + "epoch": 0.71, + "learning_rate": 4.14983519346529e-06, + "loss": 0.9561, + "step": 22137 + }, + { + "epoch": 0.71, + "learning_rate": 4.148995165870468e-06, + "loss": 1.0137, + "step": 22138 + }, + { + "epoch": 0.71, + "learning_rate": 4.148155201050457e-06, + "loss": 0.9214, + "step": 22139 + }, + { + "epoch": 0.71, + "learning_rate": 4.147315299014263e-06, + "loss": 0.9092, + "step": 22140 + }, + { + "epoch": 0.71, + "learning_rate": 4.1464754597708974e-06, + "loss": 0.981, + "step": 22141 + }, + { + "epoch": 0.71, + "learning_rate": 4.14563568332937e-06, + "loss": 0.9541, + "step": 22142 + }, + { + "epoch": 0.71, + "learning_rate": 4.144795969698695e-06, + "loss": 0.8887, + "step": 22143 + }, + { + "epoch": 0.71, + "learning_rate": 4.143956318887876e-06, + "loss": 0.9429, + "step": 22144 + }, + { + "epoch": 0.71, + "learning_rate": 4.1431167309059285e-06, + "loss": 0.9424, + "step": 22145 + }, + { + "epoch": 0.71, + "learning_rate": 4.142277205761852e-06, + "loss": 0.9009, + "step": 22146 + }, + { + "epoch": 0.71, + "learning_rate": 4.141437743464665e-06, + "loss": 1.0215, + "step": 22147 + }, + { + "epoch": 0.71, + "learning_rate": 4.140598344023361e-06, + "loss": 1.0522, + "step": 22148 + }, + { + "epoch": 0.71, + "learning_rate": 4.139759007446955e-06, + "loss": 1.0068, + "step": 22149 + }, + { + "epoch": 0.71, + "learning_rate": 4.138919733744445e-06, + "loss": 1.0269, + "step": 22150 + }, + { + "epoch": 0.71, + "learning_rate": 4.138080522924844e-06, + "loss": 1.0728, + "step": 22151 + }, + { + "epoch": 0.71, + "learning_rate": 4.137241374997147e-06, + "loss": 0.9946, + "step": 22152 + }, + { + "epoch": 0.71, + "learning_rate": 4.136402289970365e-06, + "loss": 0.9014, + "step": 22153 + }, + { + "epoch": 0.71, + "learning_rate": 4.135563267853494e-06, + "loss": 0.8726, + "step": 22154 + }, + { + "epoch": 0.71, + "learning_rate": 4.134724308655545e-06, + "loss": 0.9858, + "step": 22155 + }, + { + "epoch": 0.71, + "learning_rate": 4.133885412385506e-06, + "loss": 0.8896, + "step": 22156 + }, + { + "epoch": 0.71, + "learning_rate": 4.133046579052389e-06, + "loss": 1.0425, + "step": 22157 + }, + { + "epoch": 0.71, + "learning_rate": 4.132207808665184e-06, + "loss": 1.0059, + "step": 22158 + }, + { + "epoch": 0.71, + "learning_rate": 4.1313691012329e-06, + "loss": 1.0566, + "step": 22159 + }, + { + "epoch": 0.71, + "learning_rate": 4.130530456764524e-06, + "loss": 0.937, + "step": 22160 + }, + { + "epoch": 0.71, + "learning_rate": 4.129691875269066e-06, + "loss": 0.9971, + "step": 22161 + }, + { + "epoch": 0.71, + "learning_rate": 4.128853356755518e-06, + "loss": 0.8638, + "step": 22162 + }, + { + "epoch": 0.71, + "learning_rate": 4.128014901232874e-06, + "loss": 0.9136, + "step": 22163 + }, + { + "epoch": 0.71, + "learning_rate": 4.127176508710128e-06, + "loss": 0.9756, + "step": 22164 + }, + { + "epoch": 0.71, + "learning_rate": 4.126338179196282e-06, + "loss": 0.8806, + "step": 22165 + }, + { + "epoch": 0.71, + "learning_rate": 4.125499912700324e-06, + "loss": 1.1128, + "step": 22166 + }, + { + "epoch": 0.71, + "learning_rate": 4.124661709231252e-06, + "loss": 0.8223, + "step": 22167 + }, + { + "epoch": 0.71, + "learning_rate": 4.123823568798055e-06, + "loss": 1.04, + "step": 22168 + }, + { + "epoch": 0.71, + "learning_rate": 4.122985491409732e-06, + "loss": 1.0938, + "step": 22169 + }, + { + "epoch": 0.71, + "learning_rate": 4.12214747707527e-06, + "loss": 0.8604, + "step": 22170 + }, + { + "epoch": 0.71, + "learning_rate": 4.121309525803659e-06, + "loss": 1.0132, + "step": 22171 + }, + { + "epoch": 0.71, + "learning_rate": 4.120471637603889e-06, + "loss": 0.9214, + "step": 22172 + }, + { + "epoch": 0.71, + "learning_rate": 4.119633812484953e-06, + "loss": 0.8428, + "step": 22173 + }, + { + "epoch": 0.71, + "learning_rate": 4.118796050455835e-06, + "loss": 0.9316, + "step": 22174 + }, + { + "epoch": 0.71, + "learning_rate": 4.11795835152553e-06, + "loss": 1.0518, + "step": 22175 + }, + { + "epoch": 0.71, + "learning_rate": 4.117120715703023e-06, + "loss": 0.9072, + "step": 22176 + }, + { + "epoch": 0.71, + "learning_rate": 4.116283142997298e-06, + "loss": 1.0117, + "step": 22177 + }, + { + "epoch": 0.71, + "learning_rate": 4.115445633417341e-06, + "loss": 0.9302, + "step": 22178 + }, + { + "epoch": 0.71, + "learning_rate": 4.114608186972143e-06, + "loss": 1.0049, + "step": 22179 + }, + { + "epoch": 0.71, + "learning_rate": 4.1137708036706845e-06, + "loss": 0.918, + "step": 22180 + }, + { + "epoch": 0.71, + "learning_rate": 4.112933483521948e-06, + "loss": 0.8296, + "step": 22181 + }, + { + "epoch": 0.71, + "learning_rate": 4.112096226534924e-06, + "loss": 1.0093, + "step": 22182 + }, + { + "epoch": 0.71, + "learning_rate": 4.111259032718587e-06, + "loss": 0.9966, + "step": 22183 + }, + { + "epoch": 0.71, + "learning_rate": 4.1104219020819256e-06, + "loss": 0.9927, + "step": 22184 + }, + { + "epoch": 0.71, + "learning_rate": 4.1095848346339206e-06, + "loss": 1.0405, + "step": 22185 + }, + { + "epoch": 0.71, + "learning_rate": 4.10874783038355e-06, + "loss": 0.9551, + "step": 22186 + }, + { + "epoch": 0.71, + "learning_rate": 4.1079108893397924e-06, + "loss": 0.9517, + "step": 22187 + }, + { + "epoch": 0.71, + "learning_rate": 4.107074011511632e-06, + "loss": 0.8726, + "step": 22188 + }, + { + "epoch": 0.71, + "learning_rate": 4.106237196908044e-06, + "loss": 0.8125, + "step": 22189 + }, + { + "epoch": 0.71, + "learning_rate": 4.10540044553801e-06, + "loss": 1.0146, + "step": 22190 + }, + { + "epoch": 0.71, + "learning_rate": 4.104563757410502e-06, + "loss": 0.5212, + "step": 22191 + }, + { + "epoch": 0.71, + "learning_rate": 4.103727132534507e-06, + "loss": 0.8955, + "step": 22192 + }, + { + "epoch": 0.71, + "learning_rate": 4.102890570918988e-06, + "loss": 0.8491, + "step": 22193 + }, + { + "epoch": 0.71, + "learning_rate": 4.10205407257293e-06, + "loss": 0.9165, + "step": 22194 + }, + { + "epoch": 0.71, + "learning_rate": 4.1012176375053e-06, + "loss": 1.043, + "step": 22195 + }, + { + "epoch": 0.71, + "learning_rate": 4.10038126572508e-06, + "loss": 0.9009, + "step": 22196 + }, + { + "epoch": 0.71, + "learning_rate": 4.099544957241237e-06, + "loss": 0.9365, + "step": 22197 + }, + { + "epoch": 0.71, + "learning_rate": 4.09870871206275e-06, + "loss": 0.9092, + "step": 22198 + }, + { + "epoch": 0.71, + "learning_rate": 4.097872530198582e-06, + "loss": 0.9521, + "step": 22199 + }, + { + "epoch": 0.71, + "learning_rate": 4.0970364116577185e-06, + "loss": 1.0869, + "step": 22200 + }, + { + "epoch": 0.71, + "learning_rate": 4.096200356449114e-06, + "loss": 0.999, + "step": 22201 + }, + { + "epoch": 0.71, + "learning_rate": 4.095364364581748e-06, + "loss": 1.0425, + "step": 22202 + }, + { + "epoch": 0.71, + "learning_rate": 4.094528436064584e-06, + "loss": 0.8442, + "step": 22203 + }, + { + "epoch": 0.71, + "learning_rate": 4.093692570906599e-06, + "loss": 0.9097, + "step": 22204 + }, + { + "epoch": 0.71, + "learning_rate": 4.092856769116751e-06, + "loss": 0.9727, + "step": 22205 + }, + { + "epoch": 0.71, + "learning_rate": 4.092021030704017e-06, + "loss": 0.897, + "step": 22206 + }, + { + "epoch": 0.71, + "learning_rate": 4.091185355677357e-06, + "loss": 0.4895, + "step": 22207 + }, + { + "epoch": 0.71, + "learning_rate": 4.090349744045739e-06, + "loss": 0.9287, + "step": 22208 + }, + { + "epoch": 0.71, + "learning_rate": 4.089514195818125e-06, + "loss": 0.8677, + "step": 22209 + }, + { + "epoch": 0.71, + "learning_rate": 4.088678711003485e-06, + "loss": 0.9883, + "step": 22210 + }, + { + "epoch": 0.71, + "learning_rate": 4.0878432896107775e-06, + "loss": 0.998, + "step": 22211 + }, + { + "epoch": 0.71, + "learning_rate": 4.087007931648972e-06, + "loss": 0.9482, + "step": 22212 + }, + { + "epoch": 0.71, + "learning_rate": 4.0861726371270224e-06, + "loss": 0.9194, + "step": 22213 + }, + { + "epoch": 0.71, + "learning_rate": 4.0853374060539e-06, + "loss": 0.9043, + "step": 22214 + }, + { + "epoch": 0.71, + "learning_rate": 4.08450223843856e-06, + "loss": 0.9995, + "step": 22215 + }, + { + "epoch": 0.71, + "learning_rate": 4.083667134289965e-06, + "loss": 0.9404, + "step": 22216 + }, + { + "epoch": 0.71, + "learning_rate": 4.08283209361707e-06, + "loss": 0.8203, + "step": 22217 + }, + { + "epoch": 0.71, + "learning_rate": 4.081997116428842e-06, + "loss": 1.0854, + "step": 22218 + }, + { + "epoch": 0.71, + "learning_rate": 4.08116220273423e-06, + "loss": 0.9385, + "step": 22219 + }, + { + "epoch": 0.71, + "learning_rate": 4.080327352542202e-06, + "loss": 1.0127, + "step": 22220 + }, + { + "epoch": 0.71, + "learning_rate": 4.079492565861709e-06, + "loss": 0.9648, + "step": 22221 + }, + { + "epoch": 0.71, + "learning_rate": 4.0786578427017056e-06, + "loss": 0.8677, + "step": 22222 + }, + { + "epoch": 0.71, + "learning_rate": 4.077823183071153e-06, + "loss": 0.8809, + "step": 22223 + }, + { + "epoch": 0.71, + "learning_rate": 4.076988586979004e-06, + "loss": 0.9556, + "step": 22224 + }, + { + "epoch": 0.71, + "learning_rate": 4.0761540544342116e-06, + "loss": 1.02, + "step": 22225 + }, + { + "epoch": 0.71, + "learning_rate": 4.0753195854457265e-06, + "loss": 1.0278, + "step": 22226 + }, + { + "epoch": 0.71, + "learning_rate": 4.074485180022508e-06, + "loss": 1.1504, + "step": 22227 + }, + { + "epoch": 0.71, + "learning_rate": 4.073650838173503e-06, + "loss": 0.9512, + "step": 22228 + }, + { + "epoch": 0.71, + "learning_rate": 4.0728165599076685e-06, + "loss": 0.9922, + "step": 22229 + }, + { + "epoch": 0.71, + "learning_rate": 4.071982345233954e-06, + "loss": 0.9717, + "step": 22230 + }, + { + "epoch": 0.71, + "learning_rate": 4.071148194161306e-06, + "loss": 0.978, + "step": 22231 + }, + { + "epoch": 0.71, + "learning_rate": 4.070314106698674e-06, + "loss": 0.9946, + "step": 22232 + }, + { + "epoch": 0.71, + "learning_rate": 4.069480082855012e-06, + "loss": 0.8706, + "step": 22233 + }, + { + "epoch": 0.71, + "learning_rate": 4.068646122639261e-06, + "loss": 0.9272, + "step": 22234 + }, + { + "epoch": 0.71, + "learning_rate": 4.067812226060377e-06, + "loss": 0.9282, + "step": 22235 + }, + { + "epoch": 0.71, + "learning_rate": 4.066978393127299e-06, + "loss": 0.7886, + "step": 22236 + }, + { + "epoch": 0.71, + "learning_rate": 4.066144623848979e-06, + "loss": 0.8579, + "step": 22237 + }, + { + "epoch": 0.71, + "learning_rate": 4.065310918234361e-06, + "loss": 0.9927, + "step": 22238 + }, + { + "epoch": 0.71, + "learning_rate": 4.064477276292388e-06, + "loss": 0.9688, + "step": 22239 + }, + { + "epoch": 0.71, + "learning_rate": 4.063643698032001e-06, + "loss": 0.8862, + "step": 22240 + }, + { + "epoch": 0.71, + "learning_rate": 4.062810183462151e-06, + "loss": 0.7993, + "step": 22241 + }, + { + "epoch": 0.71, + "learning_rate": 4.061976732591774e-06, + "loss": 0.6985, + "step": 22242 + }, + { + "epoch": 0.71, + "learning_rate": 4.061143345429817e-06, + "loss": 0.8838, + "step": 22243 + }, + { + "epoch": 0.71, + "learning_rate": 4.060310021985217e-06, + "loss": 0.9053, + "step": 22244 + }, + { + "epoch": 0.71, + "learning_rate": 4.059476762266922e-06, + "loss": 0.938, + "step": 22245 + }, + { + "epoch": 0.71, + "learning_rate": 4.0586435662838605e-06, + "loss": 0.9111, + "step": 22246 + }, + { + "epoch": 0.71, + "learning_rate": 4.057810434044982e-06, + "loss": 0.9927, + "step": 22247 + }, + { + "epoch": 0.71, + "learning_rate": 4.056977365559217e-06, + "loss": 0.9282, + "step": 22248 + }, + { + "epoch": 0.71, + "learning_rate": 4.056144360835511e-06, + "loss": 1.0176, + "step": 22249 + }, + { + "epoch": 0.71, + "learning_rate": 4.055311419882794e-06, + "loss": 0.9414, + "step": 22250 + }, + { + "epoch": 0.71, + "learning_rate": 4.0544785427100095e-06, + "loss": 0.957, + "step": 22251 + }, + { + "epoch": 0.71, + "learning_rate": 4.053645729326088e-06, + "loss": 0.8755, + "step": 22252 + }, + { + "epoch": 0.71, + "learning_rate": 4.052812979739973e-06, + "loss": 0.8501, + "step": 22253 + }, + { + "epoch": 0.71, + "learning_rate": 4.051980293960584e-06, + "loss": 0.9609, + "step": 22254 + }, + { + "epoch": 0.71, + "learning_rate": 4.051147671996869e-06, + "loss": 1.0947, + "step": 22255 + }, + { + "epoch": 0.71, + "learning_rate": 4.050315113857751e-06, + "loss": 0.998, + "step": 22256 + }, + { + "epoch": 0.71, + "learning_rate": 4.049482619552172e-06, + "loss": 0.9966, + "step": 22257 + }, + { + "epoch": 0.71, + "learning_rate": 4.048650189089054e-06, + "loss": 0.9038, + "step": 22258 + }, + { + "epoch": 0.71, + "learning_rate": 4.047817822477337e-06, + "loss": 0.9043, + "step": 22259 + }, + { + "epoch": 0.71, + "learning_rate": 4.046985519725946e-06, + "loss": 0.8467, + "step": 22260 + }, + { + "epoch": 0.71, + "learning_rate": 4.0461532808438145e-06, + "loss": 0.9658, + "step": 22261 + }, + { + "epoch": 0.71, + "learning_rate": 4.0453211058398635e-06, + "loss": 1.0405, + "step": 22262 + }, + { + "epoch": 0.71, + "learning_rate": 4.04448899472303e-06, + "loss": 0.8711, + "step": 22263 + }, + { + "epoch": 0.71, + "learning_rate": 4.0436569475022394e-06, + "loss": 0.7637, + "step": 22264 + }, + { + "epoch": 0.71, + "learning_rate": 4.0428249641864125e-06, + "loss": 1.0166, + "step": 22265 + }, + { + "epoch": 0.71, + "learning_rate": 4.041993044784486e-06, + "loss": 0.9219, + "step": 22266 + }, + { + "epoch": 0.71, + "learning_rate": 4.041161189305376e-06, + "loss": 0.9883, + "step": 22267 + }, + { + "epoch": 0.71, + "learning_rate": 4.040329397758014e-06, + "loss": 1.0439, + "step": 22268 + }, + { + "epoch": 0.71, + "learning_rate": 4.0394976701513235e-06, + "loss": 0.9971, + "step": 22269 + }, + { + "epoch": 0.71, + "learning_rate": 4.038666006494225e-06, + "loss": 0.9053, + "step": 22270 + }, + { + "epoch": 0.71, + "learning_rate": 4.0378344067956386e-06, + "loss": 0.9375, + "step": 22271 + }, + { + "epoch": 0.71, + "learning_rate": 4.037002871064495e-06, + "loss": 0.9741, + "step": 22272 + }, + { + "epoch": 0.71, + "learning_rate": 4.036171399309707e-06, + "loss": 1.0288, + "step": 22273 + }, + { + "epoch": 0.71, + "learning_rate": 4.035339991540204e-06, + "loss": 0.8989, + "step": 22274 + }, + { + "epoch": 0.71, + "learning_rate": 4.034508647764901e-06, + "loss": 1.0, + "step": 22275 + }, + { + "epoch": 0.71, + "learning_rate": 4.033677367992717e-06, + "loss": 0.916, + "step": 22276 + }, + { + "epoch": 0.71, + "learning_rate": 4.032846152232569e-06, + "loss": 1.0117, + "step": 22277 + }, + { + "epoch": 0.71, + "learning_rate": 4.032015000493381e-06, + "loss": 0.9829, + "step": 22278 + }, + { + "epoch": 0.71, + "learning_rate": 4.031183912784063e-06, + "loss": 0.9756, + "step": 22279 + }, + { + "epoch": 0.71, + "learning_rate": 4.03035288911354e-06, + "loss": 0.8184, + "step": 22280 + }, + { + "epoch": 0.71, + "learning_rate": 4.02952192949072e-06, + "loss": 0.9233, + "step": 22281 + }, + { + "epoch": 0.71, + "learning_rate": 4.0286910339245255e-06, + "loss": 0.958, + "step": 22282 + }, + { + "epoch": 0.71, + "learning_rate": 4.0278602024238666e-06, + "loss": 0.9419, + "step": 22283 + }, + { + "epoch": 0.71, + "learning_rate": 4.027029434997659e-06, + "loss": 0.854, + "step": 22284 + }, + { + "epoch": 0.71, + "learning_rate": 4.026198731654811e-06, + "loss": 1.0103, + "step": 22285 + }, + { + "epoch": 0.71, + "learning_rate": 4.0253680924042426e-06, + "loss": 1.0181, + "step": 22286 + }, + { + "epoch": 0.71, + "learning_rate": 4.024537517254859e-06, + "loss": 0.9614, + "step": 22287 + }, + { + "epoch": 0.71, + "learning_rate": 4.023707006215578e-06, + "loss": 0.8179, + "step": 22288 + }, + { + "epoch": 0.71, + "learning_rate": 4.022876559295303e-06, + "loss": 0.9043, + "step": 22289 + }, + { + "epoch": 0.71, + "learning_rate": 4.022046176502951e-06, + "loss": 0.8892, + "step": 22290 + }, + { + "epoch": 0.71, + "learning_rate": 4.021215857847427e-06, + "loss": 1.0483, + "step": 22291 + }, + { + "epoch": 0.71, + "learning_rate": 4.020385603337641e-06, + "loss": 0.9463, + "step": 22292 + }, + { + "epoch": 0.71, + "learning_rate": 4.019555412982494e-06, + "loss": 1.0176, + "step": 22293 + }, + { + "epoch": 0.71, + "learning_rate": 4.018725286790904e-06, + "loss": 0.9604, + "step": 22294 + }, + { + "epoch": 0.71, + "learning_rate": 4.017895224771767e-06, + "loss": 1.0308, + "step": 22295 + }, + { + "epoch": 0.71, + "learning_rate": 4.0170652269339986e-06, + "loss": 0.8252, + "step": 22296 + }, + { + "epoch": 0.71, + "learning_rate": 4.016235293286494e-06, + "loss": 0.9624, + "step": 22297 + }, + { + "epoch": 0.71, + "learning_rate": 4.01540542383817e-06, + "loss": 0.9004, + "step": 22298 + }, + { + "epoch": 0.71, + "learning_rate": 4.0145756185979146e-06, + "loss": 0.9014, + "step": 22299 + }, + { + "epoch": 0.71, + "learning_rate": 4.013745877574643e-06, + "loss": 0.8516, + "step": 22300 + }, + { + "epoch": 0.71, + "learning_rate": 4.012916200777248e-06, + "loss": 1.0288, + "step": 22301 + }, + { + "epoch": 0.71, + "learning_rate": 4.012086588214641e-06, + "loss": 0.915, + "step": 22302 + }, + { + "epoch": 0.71, + "learning_rate": 4.011257039895718e-06, + "loss": 0.8804, + "step": 22303 + }, + { + "epoch": 0.71, + "learning_rate": 4.010427555829374e-06, + "loss": 0.9927, + "step": 22304 + }, + { + "epoch": 0.71, + "learning_rate": 4.009598136024517e-06, + "loss": 0.9072, + "step": 22305 + }, + { + "epoch": 0.71, + "learning_rate": 4.008768780490042e-06, + "loss": 0.9409, + "step": 22306 + }, + { + "epoch": 0.71, + "learning_rate": 4.007939489234845e-06, + "loss": 0.834, + "step": 22307 + }, + { + "epoch": 0.71, + "learning_rate": 4.007110262267828e-06, + "loss": 0.9453, + "step": 22308 + }, + { + "epoch": 0.71, + "learning_rate": 4.006281099597886e-06, + "loss": 0.9946, + "step": 22309 + }, + { + "epoch": 0.71, + "learning_rate": 4.0054520012339106e-06, + "loss": 0.9243, + "step": 22310 + }, + { + "epoch": 0.71, + "learning_rate": 4.004622967184804e-06, + "loss": 0.9473, + "step": 22311 + }, + { + "epoch": 0.71, + "learning_rate": 4.003793997459454e-06, + "loss": 0.8311, + "step": 22312 + }, + { + "epoch": 0.71, + "learning_rate": 4.002965092066762e-06, + "loss": 0.9126, + "step": 22313 + }, + { + "epoch": 0.71, + "learning_rate": 4.002136251015617e-06, + "loss": 0.9126, + "step": 22314 + }, + { + "epoch": 0.71, + "learning_rate": 4.001307474314912e-06, + "loss": 0.9316, + "step": 22315 + }, + { + "epoch": 0.71, + "learning_rate": 4.000478761973536e-06, + "loss": 0.8481, + "step": 22316 + }, + { + "epoch": 0.71, + "learning_rate": 3.999650114000386e-06, + "loss": 0.8765, + "step": 22317 + }, + { + "epoch": 0.71, + "learning_rate": 3.998821530404345e-06, + "loss": 0.8657, + "step": 22318 + }, + { + "epoch": 0.71, + "learning_rate": 3.997993011194313e-06, + "loss": 0.8623, + "step": 22319 + }, + { + "epoch": 0.71, + "learning_rate": 3.997164556379167e-06, + "loss": 0.8799, + "step": 22320 + }, + { + "epoch": 0.71, + "learning_rate": 3.996336165967811e-06, + "loss": 0.936, + "step": 22321 + }, + { + "epoch": 0.71, + "learning_rate": 3.995507839969115e-06, + "loss": 0.9648, + "step": 22322 + }, + { + "epoch": 0.71, + "learning_rate": 3.994679578391979e-06, + "loss": 0.8682, + "step": 22323 + }, + { + "epoch": 0.71, + "learning_rate": 3.993851381245281e-06, + "loss": 1.0166, + "step": 22324 + }, + { + "epoch": 0.71, + "learning_rate": 3.993023248537913e-06, + "loss": 0.9521, + "step": 22325 + }, + { + "epoch": 0.71, + "learning_rate": 3.992195180278754e-06, + "loss": 1.0166, + "step": 22326 + }, + { + "epoch": 0.71, + "learning_rate": 3.991367176476696e-06, + "loss": 0.5447, + "step": 22327 + }, + { + "epoch": 0.71, + "learning_rate": 3.990539237140617e-06, + "loss": 0.9839, + "step": 22328 + }, + { + "epoch": 0.71, + "learning_rate": 3.989711362279403e-06, + "loss": 0.8633, + "step": 22329 + }, + { + "epoch": 0.71, + "learning_rate": 3.9888835519019285e-06, + "loss": 0.9204, + "step": 22330 + }, + { + "epoch": 0.71, + "learning_rate": 3.988055806017085e-06, + "loss": 0.9404, + "step": 22331 + }, + { + "epoch": 0.71, + "learning_rate": 3.987228124633746e-06, + "loss": 0.9497, + "step": 22332 + }, + { + "epoch": 0.71, + "learning_rate": 3.986400507760798e-06, + "loss": 0.8667, + "step": 22333 + }, + { + "epoch": 0.71, + "learning_rate": 3.9855729554071135e-06, + "loss": 0.957, + "step": 22334 + }, + { + "epoch": 0.71, + "learning_rate": 3.984745467581578e-06, + "loss": 0.9087, + "step": 22335 + }, + { + "epoch": 0.71, + "learning_rate": 3.983918044293066e-06, + "loss": 0.8218, + "step": 22336 + }, + { + "epoch": 0.71, + "learning_rate": 3.9830906855504545e-06, + "loss": 0.7754, + "step": 22337 + }, + { + "epoch": 0.71, + "learning_rate": 3.982263391362619e-06, + "loss": 0.9238, + "step": 22338 + }, + { + "epoch": 0.71, + "learning_rate": 3.98143616173844e-06, + "loss": 1.0073, + "step": 22339 + }, + { + "epoch": 0.71, + "learning_rate": 3.980608996686787e-06, + "loss": 1.0557, + "step": 22340 + }, + { + "epoch": 0.71, + "learning_rate": 3.9797818962165406e-06, + "loss": 0.9121, + "step": 22341 + }, + { + "epoch": 0.71, + "learning_rate": 3.97895486033657e-06, + "loss": 0.9092, + "step": 22342 + }, + { + "epoch": 0.71, + "learning_rate": 3.978127889055752e-06, + "loss": 0.9229, + "step": 22343 + }, + { + "epoch": 0.71, + "learning_rate": 3.977300982382953e-06, + "loss": 1.0215, + "step": 22344 + }, + { + "epoch": 0.71, + "learning_rate": 3.976474140327053e-06, + "loss": 0.8594, + "step": 22345 + }, + { + "epoch": 0.71, + "learning_rate": 3.975647362896914e-06, + "loss": 0.9983, + "step": 22346 + }, + { + "epoch": 0.71, + "learning_rate": 3.974820650101417e-06, + "loss": 0.9443, + "step": 22347 + }, + { + "epoch": 0.71, + "learning_rate": 3.973994001949424e-06, + "loss": 0.8999, + "step": 22348 + }, + { + "epoch": 0.71, + "learning_rate": 3.973167418449803e-06, + "loss": 0.9312, + "step": 22349 + }, + { + "epoch": 0.71, + "learning_rate": 3.97234089961143e-06, + "loss": 0.939, + "step": 22350 + }, + { + "epoch": 0.71, + "learning_rate": 3.9715144454431666e-06, + "loss": 0.9478, + "step": 22351 + }, + { + "epoch": 0.71, + "learning_rate": 3.970688055953879e-06, + "loss": 0.9634, + "step": 22352 + }, + { + "epoch": 0.71, + "learning_rate": 3.969861731152438e-06, + "loss": 0.8613, + "step": 22353 + }, + { + "epoch": 0.71, + "learning_rate": 3.9690354710477075e-06, + "loss": 1.1255, + "step": 22354 + }, + { + "epoch": 0.71, + "learning_rate": 3.968209275648548e-06, + "loss": 0.9355, + "step": 22355 + }, + { + "epoch": 0.71, + "learning_rate": 3.967383144963831e-06, + "loss": 0.9888, + "step": 22356 + }, + { + "epoch": 0.71, + "learning_rate": 3.966557079002413e-06, + "loss": 0.9126, + "step": 22357 + }, + { + "epoch": 0.72, + "learning_rate": 3.965731077773167e-06, + "loss": 0.9453, + "step": 22358 + }, + { + "epoch": 0.72, + "learning_rate": 3.96490514128494e-06, + "loss": 0.9004, + "step": 22359 + }, + { + "epoch": 0.72, + "learning_rate": 3.964079269546606e-06, + "loss": 1.0591, + "step": 22360 + }, + { + "epoch": 0.72, + "learning_rate": 3.963253462567018e-06, + "loss": 0.9768, + "step": 22361 + }, + { + "epoch": 0.72, + "learning_rate": 3.9624277203550425e-06, + "loss": 0.8477, + "step": 22362 + }, + { + "epoch": 0.72, + "learning_rate": 3.961602042919532e-06, + "loss": 0.8491, + "step": 22363 + }, + { + "epoch": 0.72, + "learning_rate": 3.960776430269352e-06, + "loss": 0.9888, + "step": 22364 + }, + { + "epoch": 0.72, + "learning_rate": 3.9599508824133535e-06, + "loss": 0.9233, + "step": 22365 + }, + { + "epoch": 0.72, + "learning_rate": 3.959125399360404e-06, + "loss": 1.0356, + "step": 22366 + }, + { + "epoch": 0.72, + "learning_rate": 3.958299981119347e-06, + "loss": 0.958, + "step": 22367 + }, + { + "epoch": 0.72, + "learning_rate": 3.957474627699047e-06, + "loss": 0.5173, + "step": 22368 + }, + { + "epoch": 0.72, + "learning_rate": 3.956649339108354e-06, + "loss": 0.9971, + "step": 22369 + }, + { + "epoch": 0.72, + "learning_rate": 3.955824115356128e-06, + "loss": 1.0063, + "step": 22370 + }, + { + "epoch": 0.72, + "learning_rate": 3.954998956451217e-06, + "loss": 1.0591, + "step": 22371 + }, + { + "epoch": 0.72, + "learning_rate": 3.954173862402481e-06, + "loss": 0.8691, + "step": 22372 + }, + { + "epoch": 0.72, + "learning_rate": 3.953348833218763e-06, + "loss": 0.8887, + "step": 22373 + }, + { + "epoch": 0.72, + "learning_rate": 3.952523868908927e-06, + "loss": 0.9526, + "step": 22374 + }, + { + "epoch": 0.72, + "learning_rate": 3.951698969481812e-06, + "loss": 0.9106, + "step": 22375 + }, + { + "epoch": 0.72, + "learning_rate": 3.950874134946276e-06, + "loss": 0.8701, + "step": 22376 + }, + { + "epoch": 0.72, + "learning_rate": 3.950049365311161e-06, + "loss": 0.9678, + "step": 22377 + }, + { + "epoch": 0.72, + "learning_rate": 3.949224660585325e-06, + "loss": 0.7832, + "step": 22378 + }, + { + "epoch": 0.72, + "learning_rate": 3.948400020777607e-06, + "loss": 0.8999, + "step": 22379 + }, + { + "epoch": 0.72, + "learning_rate": 3.947575445896864e-06, + "loss": 1.0288, + "step": 22380 + }, + { + "epoch": 0.72, + "learning_rate": 3.946750935951938e-06, + "loss": 0.9111, + "step": 22381 + }, + { + "epoch": 0.72, + "learning_rate": 3.945926490951674e-06, + "loss": 0.9365, + "step": 22382 + }, + { + "epoch": 0.72, + "learning_rate": 3.945102110904915e-06, + "loss": 0.9424, + "step": 22383 + }, + { + "epoch": 0.72, + "learning_rate": 3.944277795820514e-06, + "loss": 0.8521, + "step": 22384 + }, + { + "epoch": 0.72, + "learning_rate": 3.9434535457073055e-06, + "loss": 0.875, + "step": 22385 + }, + { + "epoch": 0.72, + "learning_rate": 3.942629360574141e-06, + "loss": 0.9014, + "step": 22386 + }, + { + "epoch": 0.72, + "learning_rate": 3.941805240429859e-06, + "loss": 0.9668, + "step": 22387 + }, + { + "epoch": 0.72, + "learning_rate": 3.940981185283299e-06, + "loss": 0.9126, + "step": 22388 + }, + { + "epoch": 0.72, + "learning_rate": 3.940157195143309e-06, + "loss": 1.1504, + "step": 22389 + }, + { + "epoch": 0.72, + "learning_rate": 3.939333270018725e-06, + "loss": 1.0151, + "step": 22390 + }, + { + "epoch": 0.72, + "learning_rate": 3.938509409918384e-06, + "loss": 0.8657, + "step": 22391 + }, + { + "epoch": 0.72, + "learning_rate": 3.937685614851132e-06, + "loss": 0.8691, + "step": 22392 + }, + { + "epoch": 0.72, + "learning_rate": 3.936861884825804e-06, + "loss": 1.0063, + "step": 22393 + }, + { + "epoch": 0.72, + "learning_rate": 3.936038219851235e-06, + "loss": 0.8662, + "step": 22394 + }, + { + "epoch": 0.72, + "learning_rate": 3.935214619936268e-06, + "loss": 0.8179, + "step": 22395 + }, + { + "epoch": 0.72, + "learning_rate": 3.934391085089735e-06, + "loss": 0.9185, + "step": 22396 + }, + { + "epoch": 0.72, + "learning_rate": 3.933567615320473e-06, + "loss": 0.9038, + "step": 22397 + }, + { + "epoch": 0.72, + "learning_rate": 3.9327442106373115e-06, + "loss": 0.9263, + "step": 22398 + }, + { + "epoch": 0.72, + "learning_rate": 3.931920871049095e-06, + "loss": 0.8813, + "step": 22399 + }, + { + "epoch": 0.72, + "learning_rate": 3.931097596564648e-06, + "loss": 1.0649, + "step": 22400 + }, + { + "epoch": 0.72, + "learning_rate": 3.930274387192811e-06, + "loss": 0.8633, + "step": 22401 + }, + { + "epoch": 0.72, + "learning_rate": 3.929451242942407e-06, + "loss": 0.7852, + "step": 22402 + }, + { + "epoch": 0.72, + "learning_rate": 3.928628163822278e-06, + "loss": 1.0889, + "step": 22403 + }, + { + "epoch": 0.72, + "learning_rate": 3.9278051498412475e-06, + "loss": 0.8813, + "step": 22404 + }, + { + "epoch": 0.72, + "learning_rate": 3.9269822010081484e-06, + "loss": 0.9155, + "step": 22405 + }, + { + "epoch": 0.72, + "learning_rate": 3.926159317331806e-06, + "loss": 0.9048, + "step": 22406 + }, + { + "epoch": 0.72, + "learning_rate": 3.925336498821055e-06, + "loss": 1.0889, + "step": 22407 + }, + { + "epoch": 0.72, + "learning_rate": 3.924513745484716e-06, + "loss": 0.9863, + "step": 22408 + }, + { + "epoch": 0.72, + "learning_rate": 3.923691057331625e-06, + "loss": 0.9443, + "step": 22409 + }, + { + "epoch": 0.72, + "learning_rate": 3.9228684343706004e-06, + "loss": 0.9336, + "step": 22410 + }, + { + "epoch": 0.72, + "learning_rate": 3.9220458766104785e-06, + "loss": 0.9917, + "step": 22411 + }, + { + "epoch": 0.72, + "learning_rate": 3.92122338406007e-06, + "loss": 0.8984, + "step": 22412 + }, + { + "epoch": 0.72, + "learning_rate": 3.920400956728211e-06, + "loss": 0.9775, + "step": 22413 + }, + { + "epoch": 0.72, + "learning_rate": 3.9195785946237176e-06, + "loss": 0.9761, + "step": 22414 + }, + { + "epoch": 0.72, + "learning_rate": 3.91875629775542e-06, + "loss": 0.9756, + "step": 22415 + }, + { + "epoch": 0.72, + "learning_rate": 3.917934066132133e-06, + "loss": 0.917, + "step": 22416 + }, + { + "epoch": 0.72, + "learning_rate": 3.9171118997626865e-06, + "loss": 0.9326, + "step": 22417 + }, + { + "epoch": 0.72, + "learning_rate": 3.916289798655894e-06, + "loss": 0.8579, + "step": 22418 + }, + { + "epoch": 0.72, + "learning_rate": 3.915467762820584e-06, + "loss": 1.0317, + "step": 22419 + }, + { + "epoch": 0.72, + "learning_rate": 3.914645792265564e-06, + "loss": 0.8667, + "step": 22420 + }, + { + "epoch": 0.72, + "learning_rate": 3.913823886999665e-06, + "loss": 0.9336, + "step": 22421 + }, + { + "epoch": 0.72, + "learning_rate": 3.913002047031695e-06, + "loss": 0.8774, + "step": 22422 + }, + { + "epoch": 0.72, + "learning_rate": 3.91218027237048e-06, + "loss": 0.8193, + "step": 22423 + }, + { + "epoch": 0.72, + "learning_rate": 3.911358563024831e-06, + "loss": 0.8057, + "step": 22424 + }, + { + "epoch": 0.72, + "learning_rate": 3.910536919003568e-06, + "loss": 1.0034, + "step": 22425 + }, + { + "epoch": 0.72, + "learning_rate": 3.909715340315504e-06, + "loss": 0.9609, + "step": 22426 + }, + { + "epoch": 0.72, + "learning_rate": 3.908893826969456e-06, + "loss": 0.9756, + "step": 22427 + }, + { + "epoch": 0.72, + "learning_rate": 3.9080723789742305e-06, + "loss": 0.9902, + "step": 22428 + }, + { + "epoch": 0.72, + "learning_rate": 3.907250996338649e-06, + "loss": 1.0562, + "step": 22429 + }, + { + "epoch": 0.72, + "learning_rate": 3.906429679071519e-06, + "loss": 0.8862, + "step": 22430 + }, + { + "epoch": 0.72, + "learning_rate": 3.905608427181659e-06, + "loss": 0.9756, + "step": 22431 + }, + { + "epoch": 0.72, + "learning_rate": 3.904787240677874e-06, + "loss": 0.8857, + "step": 22432 + }, + { + "epoch": 0.72, + "learning_rate": 3.903966119568974e-06, + "loss": 1.0117, + "step": 22433 + }, + { + "epoch": 0.72, + "learning_rate": 3.9031450638637725e-06, + "loss": 0.9521, + "step": 22434 + }, + { + "epoch": 0.72, + "learning_rate": 3.9023240735710775e-06, + "loss": 1.0249, + "step": 22435 + }, + { + "epoch": 0.72, + "learning_rate": 3.901503148699693e-06, + "loss": 1.0752, + "step": 22436 + }, + { + "epoch": 0.72, + "learning_rate": 3.900682289258435e-06, + "loss": 0.979, + "step": 22437 + }, + { + "epoch": 0.72, + "learning_rate": 3.899861495256105e-06, + "loss": 0.9453, + "step": 22438 + }, + { + "epoch": 0.72, + "learning_rate": 3.899040766701506e-06, + "loss": 0.8774, + "step": 22439 + }, + { + "epoch": 0.72, + "learning_rate": 3.898220103603451e-06, + "loss": 0.9927, + "step": 22440 + }, + { + "epoch": 0.72, + "learning_rate": 3.897399505970742e-06, + "loss": 0.9209, + "step": 22441 + }, + { + "epoch": 0.72, + "learning_rate": 3.896578973812181e-06, + "loss": 0.8589, + "step": 22442 + }, + { + "epoch": 0.72, + "learning_rate": 3.895758507136569e-06, + "loss": 0.9985, + "step": 22443 + }, + { + "epoch": 0.72, + "learning_rate": 3.894938105952717e-06, + "loss": 0.8047, + "step": 22444 + }, + { + "epoch": 0.72, + "learning_rate": 3.894117770269418e-06, + "loss": 0.8496, + "step": 22445 + }, + { + "epoch": 0.72, + "learning_rate": 3.893297500095481e-06, + "loss": 0.9072, + "step": 22446 + }, + { + "epoch": 0.72, + "learning_rate": 3.892477295439701e-06, + "loss": 0.9688, + "step": 22447 + }, + { + "epoch": 0.72, + "learning_rate": 3.891657156310883e-06, + "loss": 1.1191, + "step": 22448 + }, + { + "epoch": 0.72, + "learning_rate": 3.890837082717822e-06, + "loss": 0.9155, + "step": 22449 + }, + { + "epoch": 0.72, + "learning_rate": 3.89001707466932e-06, + "loss": 0.9365, + "step": 22450 + }, + { + "epoch": 0.72, + "learning_rate": 3.889197132174169e-06, + "loss": 1.0425, + "step": 22451 + }, + { + "epoch": 0.72, + "learning_rate": 3.8883772552411715e-06, + "loss": 0.9165, + "step": 22452 + }, + { + "epoch": 0.72, + "learning_rate": 3.887557443879118e-06, + "loss": 0.9922, + "step": 22453 + }, + { + "epoch": 0.72, + "learning_rate": 3.886737698096814e-06, + "loss": 0.9272, + "step": 22454 + }, + { + "epoch": 0.72, + "learning_rate": 3.885918017903045e-06, + "loss": 0.9551, + "step": 22455 + }, + { + "epoch": 0.72, + "learning_rate": 3.885098403306611e-06, + "loss": 0.8994, + "step": 22456 + }, + { + "epoch": 0.72, + "learning_rate": 3.884278854316305e-06, + "loss": 1.0229, + "step": 22457 + }, + { + "epoch": 0.72, + "learning_rate": 3.883459370940916e-06, + "loss": 0.9385, + "step": 22458 + }, + { + "epoch": 0.72, + "learning_rate": 3.882639953189238e-06, + "loss": 0.9014, + "step": 22459 + }, + { + "epoch": 0.72, + "learning_rate": 3.881820601070064e-06, + "loss": 0.8767, + "step": 22460 + }, + { + "epoch": 0.72, + "learning_rate": 3.881001314592181e-06, + "loss": 0.9883, + "step": 22461 + }, + { + "epoch": 0.72, + "learning_rate": 3.880182093764386e-06, + "loss": 0.9824, + "step": 22462 + }, + { + "epoch": 0.72, + "learning_rate": 3.8793629385954605e-06, + "loss": 0.9536, + "step": 22463 + }, + { + "epoch": 0.72, + "learning_rate": 3.878543849094203e-06, + "loss": 0.8481, + "step": 22464 + }, + { + "epoch": 0.72, + "learning_rate": 3.877724825269388e-06, + "loss": 0.9458, + "step": 22465 + }, + { + "epoch": 0.72, + "learning_rate": 3.876905867129813e-06, + "loss": 0.897, + "step": 22466 + }, + { + "epoch": 0.72, + "learning_rate": 3.876086974684259e-06, + "loss": 0.9067, + "step": 22467 + }, + { + "epoch": 0.72, + "learning_rate": 3.875268147941517e-06, + "loss": 1.0059, + "step": 22468 + }, + { + "epoch": 0.72, + "learning_rate": 3.874449386910365e-06, + "loss": 0.9609, + "step": 22469 + }, + { + "epoch": 0.72, + "learning_rate": 3.873630691599596e-06, + "loss": 0.9141, + "step": 22470 + }, + { + "epoch": 0.72, + "learning_rate": 3.87281206201799e-06, + "loss": 0.9585, + "step": 22471 + }, + { + "epoch": 0.72, + "learning_rate": 3.871993498174328e-06, + "loss": 0.937, + "step": 22472 + }, + { + "epoch": 0.72, + "learning_rate": 3.87117500007739e-06, + "loss": 0.9893, + "step": 22473 + }, + { + "epoch": 0.72, + "learning_rate": 3.8703565677359665e-06, + "loss": 0.8647, + "step": 22474 + }, + { + "epoch": 0.72, + "learning_rate": 3.869538201158829e-06, + "loss": 0.978, + "step": 22475 + }, + { + "epoch": 0.72, + "learning_rate": 3.868719900354765e-06, + "loss": 0.9507, + "step": 22476 + }, + { + "epoch": 0.72, + "learning_rate": 3.86790166533255e-06, + "loss": 0.9214, + "step": 22477 + }, + { + "epoch": 0.72, + "learning_rate": 3.8670834961009605e-06, + "loss": 0.8965, + "step": 22478 + }, + { + "epoch": 0.72, + "learning_rate": 3.866265392668781e-06, + "loss": 0.9331, + "step": 22479 + }, + { + "epoch": 0.72, + "learning_rate": 3.8654473550447846e-06, + "loss": 0.877, + "step": 22480 + }, + { + "epoch": 0.72, + "learning_rate": 3.86462938323775e-06, + "loss": 0.9688, + "step": 22481 + }, + { + "epoch": 0.72, + "learning_rate": 3.863811477256447e-06, + "loss": 0.9575, + "step": 22482 + }, + { + "epoch": 0.72, + "learning_rate": 3.862993637109661e-06, + "loss": 1.0288, + "step": 22483 + }, + { + "epoch": 0.72, + "learning_rate": 3.862175862806156e-06, + "loss": 1.0093, + "step": 22484 + }, + { + "epoch": 0.72, + "learning_rate": 3.861358154354715e-06, + "loss": 0.96, + "step": 22485 + }, + { + "epoch": 0.72, + "learning_rate": 3.860540511764104e-06, + "loss": 0.9966, + "step": 22486 + }, + { + "epoch": 0.72, + "learning_rate": 3.859722935043105e-06, + "loss": 0.9053, + "step": 22487 + }, + { + "epoch": 0.72, + "learning_rate": 3.858905424200476e-06, + "loss": 0.9531, + "step": 22488 + }, + { + "epoch": 0.72, + "learning_rate": 3.8580879792449984e-06, + "loss": 0.8916, + "step": 22489 + }, + { + "epoch": 0.72, + "learning_rate": 3.857270600185438e-06, + "loss": 0.8862, + "step": 22490 + }, + { + "epoch": 0.72, + "learning_rate": 3.856453287030567e-06, + "loss": 0.9355, + "step": 22491 + }, + { + "epoch": 0.72, + "learning_rate": 3.855636039789149e-06, + "loss": 1.0176, + "step": 22492 + }, + { + "epoch": 0.72, + "learning_rate": 3.8548188584699605e-06, + "loss": 0.9146, + "step": 22493 + }, + { + "epoch": 0.72, + "learning_rate": 3.854001743081764e-06, + "loss": 0.9844, + "step": 22494 + }, + { + "epoch": 0.72, + "learning_rate": 3.853184693633328e-06, + "loss": 0.9102, + "step": 22495 + }, + { + "epoch": 0.72, + "learning_rate": 3.852367710133413e-06, + "loss": 1.0029, + "step": 22496 + }, + { + "epoch": 0.72, + "learning_rate": 3.851550792590791e-06, + "loss": 0.9014, + "step": 22497 + }, + { + "epoch": 0.72, + "learning_rate": 3.850733941014223e-06, + "loss": 0.9858, + "step": 22498 + }, + { + "epoch": 0.72, + "learning_rate": 3.849917155412476e-06, + "loss": 0.9805, + "step": 22499 + }, + { + "epoch": 0.72, + "learning_rate": 3.849100435794309e-06, + "loss": 0.9131, + "step": 22500 + }, + { + "epoch": 0.72, + "learning_rate": 3.848283782168489e-06, + "loss": 1.0024, + "step": 22501 + }, + { + "epoch": 0.72, + "learning_rate": 3.847467194543776e-06, + "loss": 0.9897, + "step": 22502 + }, + { + "epoch": 0.72, + "learning_rate": 3.846650672928931e-06, + "loss": 1.0298, + "step": 22503 + }, + { + "epoch": 0.72, + "learning_rate": 3.84583421733271e-06, + "loss": 0.9575, + "step": 22504 + }, + { + "epoch": 0.72, + "learning_rate": 3.8450178277638814e-06, + "loss": 0.9087, + "step": 22505 + }, + { + "epoch": 0.72, + "learning_rate": 3.844201504231195e-06, + "loss": 0.8477, + "step": 22506 + }, + { + "epoch": 0.72, + "learning_rate": 3.8433852467434175e-06, + "loss": 0.916, + "step": 22507 + }, + { + "epoch": 0.72, + "learning_rate": 3.8425690553092995e-06, + "loss": 1.0215, + "step": 22508 + }, + { + "epoch": 0.72, + "learning_rate": 3.841752929937606e-06, + "loss": 0.8931, + "step": 22509 + }, + { + "epoch": 0.72, + "learning_rate": 3.840936870637083e-06, + "loss": 1.0435, + "step": 22510 + }, + { + "epoch": 0.72, + "learning_rate": 3.840120877416495e-06, + "loss": 0.9331, + "step": 22511 + }, + { + "epoch": 0.72, + "learning_rate": 3.839304950284587e-06, + "loss": 0.9336, + "step": 22512 + }, + { + "epoch": 0.72, + "learning_rate": 3.838489089250122e-06, + "loss": 0.9609, + "step": 22513 + }, + { + "epoch": 0.72, + "learning_rate": 3.8376732943218475e-06, + "loss": 0.9902, + "step": 22514 + }, + { + "epoch": 0.72, + "learning_rate": 3.836857565508522e-06, + "loss": 0.8726, + "step": 22515 + }, + { + "epoch": 0.72, + "learning_rate": 3.836041902818892e-06, + "loss": 0.8118, + "step": 22516 + }, + { + "epoch": 0.72, + "learning_rate": 3.8352263062617125e-06, + "loss": 1.0527, + "step": 22517 + }, + { + "epoch": 0.72, + "learning_rate": 3.8344107758457276e-06, + "loss": 0.8452, + "step": 22518 + }, + { + "epoch": 0.72, + "learning_rate": 3.8335953115796954e-06, + "loss": 1.043, + "step": 22519 + }, + { + "epoch": 0.72, + "learning_rate": 3.83277991347236e-06, + "loss": 0.9253, + "step": 22520 + }, + { + "epoch": 0.72, + "learning_rate": 3.831964581532467e-06, + "loss": 0.9219, + "step": 22521 + }, + { + "epoch": 0.72, + "learning_rate": 3.831149315768772e-06, + "loss": 0.8628, + "step": 22522 + }, + { + "epoch": 0.72, + "learning_rate": 3.830334116190013e-06, + "loss": 1.0571, + "step": 22523 + }, + { + "epoch": 0.72, + "learning_rate": 3.829518982804945e-06, + "loss": 0.9023, + "step": 22524 + }, + { + "epoch": 0.72, + "learning_rate": 3.828703915622309e-06, + "loss": 0.9409, + "step": 22525 + }, + { + "epoch": 0.72, + "learning_rate": 3.82788891465085e-06, + "loss": 0.958, + "step": 22526 + }, + { + "epoch": 0.72, + "learning_rate": 3.827073979899308e-06, + "loss": 0.9058, + "step": 22527 + }, + { + "epoch": 0.72, + "learning_rate": 3.826259111376435e-06, + "loss": 0.9473, + "step": 22528 + }, + { + "epoch": 0.72, + "learning_rate": 3.825444309090964e-06, + "loss": 0.9019, + "step": 22529 + }, + { + "epoch": 0.72, + "learning_rate": 3.8246295730516455e-06, + "loss": 0.9736, + "step": 22530 + }, + { + "epoch": 0.72, + "learning_rate": 3.823814903267214e-06, + "loss": 1.0083, + "step": 22531 + }, + { + "epoch": 0.72, + "learning_rate": 3.823000299746421e-06, + "loss": 0.9106, + "step": 22532 + }, + { + "epoch": 0.72, + "learning_rate": 3.82218576249799e-06, + "loss": 0.8779, + "step": 22533 + }, + { + "epoch": 0.72, + "learning_rate": 3.8213712915306734e-06, + "loss": 0.874, + "step": 22534 + }, + { + "epoch": 0.72, + "learning_rate": 3.8205568868532e-06, + "loss": 0.9087, + "step": 22535 + }, + { + "epoch": 0.72, + "learning_rate": 3.819742548474317e-06, + "loss": 0.96, + "step": 22536 + }, + { + "epoch": 0.72, + "learning_rate": 3.818928276402751e-06, + "loss": 0.9497, + "step": 22537 + }, + { + "epoch": 0.72, + "learning_rate": 3.818114070647249e-06, + "loss": 1.1016, + "step": 22538 + }, + { + "epoch": 0.72, + "learning_rate": 3.817299931216537e-06, + "loss": 0.9336, + "step": 22539 + }, + { + "epoch": 0.72, + "learning_rate": 3.816485858119361e-06, + "loss": 0.9565, + "step": 22540 + }, + { + "epoch": 0.72, + "learning_rate": 3.815671851364442e-06, + "loss": 0.9243, + "step": 22541 + }, + { + "epoch": 0.72, + "learning_rate": 3.814857910960522e-06, + "loss": 0.9478, + "step": 22542 + }, + { + "epoch": 0.72, + "learning_rate": 3.8140440369163278e-06, + "loss": 1.0444, + "step": 22543 + }, + { + "epoch": 0.72, + "learning_rate": 3.813230229240599e-06, + "loss": 0.979, + "step": 22544 + }, + { + "epoch": 0.72, + "learning_rate": 3.812416487942059e-06, + "loss": 0.9585, + "step": 22545 + }, + { + "epoch": 0.72, + "learning_rate": 3.8116028130294446e-06, + "loss": 0.8921, + "step": 22546 + }, + { + "epoch": 0.72, + "learning_rate": 3.8107892045114836e-06, + "loss": 0.9448, + "step": 22547 + }, + { + "epoch": 0.72, + "learning_rate": 3.809975662396905e-06, + "loss": 0.9487, + "step": 22548 + }, + { + "epoch": 0.72, + "learning_rate": 3.809162186694433e-06, + "loss": 0.8091, + "step": 22549 + }, + { + "epoch": 0.72, + "learning_rate": 3.8083487774128026e-06, + "loss": 0.9888, + "step": 22550 + }, + { + "epoch": 0.72, + "learning_rate": 3.807535434560734e-06, + "loss": 0.9434, + "step": 22551 + }, + { + "epoch": 0.72, + "learning_rate": 3.8067221581469594e-06, + "loss": 0.9424, + "step": 22552 + }, + { + "epoch": 0.72, + "learning_rate": 3.805908948180199e-06, + "loss": 0.8555, + "step": 22553 + }, + { + "epoch": 0.72, + "learning_rate": 3.8050958046691834e-06, + "loss": 0.873, + "step": 22554 + }, + { + "epoch": 0.72, + "learning_rate": 3.8042827276226334e-06, + "loss": 0.8311, + "step": 22555 + }, + { + "epoch": 0.72, + "learning_rate": 3.803469717049273e-06, + "loss": 1.0098, + "step": 22556 + }, + { + "epoch": 0.72, + "learning_rate": 3.8026567729578214e-06, + "loss": 0.8965, + "step": 22557 + }, + { + "epoch": 0.72, + "learning_rate": 3.801843895357006e-06, + "loss": 0.8374, + "step": 22558 + }, + { + "epoch": 0.72, + "learning_rate": 3.8010310842555462e-06, + "loss": 0.9482, + "step": 22559 + }, + { + "epoch": 0.72, + "learning_rate": 3.8002183396621593e-06, + "loss": 0.9238, + "step": 22560 + }, + { + "epoch": 0.72, + "learning_rate": 3.7994056615855713e-06, + "loss": 0.7976, + "step": 22561 + }, + { + "epoch": 0.72, + "learning_rate": 3.798593050034498e-06, + "loss": 0.9424, + "step": 22562 + }, + { + "epoch": 0.72, + "learning_rate": 3.7977805050176553e-06, + "loss": 1.0654, + "step": 22563 + }, + { + "epoch": 0.72, + "learning_rate": 3.796968026543766e-06, + "loss": 0.8979, + "step": 22564 + }, + { + "epoch": 0.72, + "learning_rate": 3.796155614621545e-06, + "loss": 0.929, + "step": 22565 + }, + { + "epoch": 0.72, + "learning_rate": 3.7953432692597046e-06, + "loss": 1.0352, + "step": 22566 + }, + { + "epoch": 0.72, + "learning_rate": 3.794530990466967e-06, + "loss": 1.0171, + "step": 22567 + }, + { + "epoch": 0.72, + "learning_rate": 3.7937187782520414e-06, + "loss": 0.9434, + "step": 22568 + }, + { + "epoch": 0.72, + "learning_rate": 3.7929066326236475e-06, + "loss": 0.9146, + "step": 22569 + }, + { + "epoch": 0.72, + "learning_rate": 3.792094553590495e-06, + "loss": 0.8667, + "step": 22570 + }, + { + "epoch": 0.72, + "learning_rate": 3.7912825411612985e-06, + "loss": 0.9624, + "step": 22571 + }, + { + "epoch": 0.72, + "learning_rate": 3.790470595344765e-06, + "loss": 1.001, + "step": 22572 + }, + { + "epoch": 0.72, + "learning_rate": 3.7896587161496136e-06, + "loss": 0.8823, + "step": 22573 + }, + { + "epoch": 0.72, + "learning_rate": 3.7888469035845464e-06, + "loss": 0.9009, + "step": 22574 + }, + { + "epoch": 0.72, + "learning_rate": 3.7880351576582818e-06, + "loss": 1.0308, + "step": 22575 + }, + { + "epoch": 0.72, + "learning_rate": 3.7872234783795213e-06, + "loss": 0.9653, + "step": 22576 + }, + { + "epoch": 0.72, + "learning_rate": 3.7864118657569836e-06, + "loss": 0.9663, + "step": 22577 + }, + { + "epoch": 0.72, + "learning_rate": 3.785600319799364e-06, + "loss": 1.0283, + "step": 22578 + }, + { + "epoch": 0.72, + "learning_rate": 3.7847888405153775e-06, + "loss": 0.8184, + "step": 22579 + }, + { + "epoch": 0.72, + "learning_rate": 3.7839774279137254e-06, + "loss": 0.9502, + "step": 22580 + }, + { + "epoch": 0.72, + "learning_rate": 3.783166082003118e-06, + "loss": 0.876, + "step": 22581 + }, + { + "epoch": 0.72, + "learning_rate": 3.7823548027922563e-06, + "loss": 0.9756, + "step": 22582 + }, + { + "epoch": 0.72, + "learning_rate": 3.781543590289849e-06, + "loss": 1.0127, + "step": 22583 + }, + { + "epoch": 0.72, + "learning_rate": 3.7807324445045924e-06, + "loss": 0.9561, + "step": 22584 + }, + { + "epoch": 0.72, + "learning_rate": 3.7799213654452017e-06, + "loss": 0.9409, + "step": 22585 + }, + { + "epoch": 0.72, + "learning_rate": 3.779110353120363e-06, + "loss": 0.9521, + "step": 22586 + }, + { + "epoch": 0.72, + "learning_rate": 3.778299407538789e-06, + "loss": 0.9092, + "step": 22587 + }, + { + "epoch": 0.72, + "learning_rate": 3.7774885287091734e-06, + "loss": 0.895, + "step": 22588 + }, + { + "epoch": 0.72, + "learning_rate": 3.7766777166402225e-06, + "loss": 0.9351, + "step": 22589 + }, + { + "epoch": 0.72, + "learning_rate": 3.7758669713406282e-06, + "loss": 0.979, + "step": 22590 + }, + { + "epoch": 0.72, + "learning_rate": 3.775056292819097e-06, + "loss": 1.0078, + "step": 22591 + }, + { + "epoch": 0.72, + "learning_rate": 3.774245681084322e-06, + "loss": 1.0181, + "step": 22592 + }, + { + "epoch": 0.72, + "learning_rate": 3.7734351361450007e-06, + "loss": 0.8667, + "step": 22593 + }, + { + "epoch": 0.72, + "learning_rate": 3.7726246580098247e-06, + "loss": 0.9619, + "step": 22594 + }, + { + "epoch": 0.72, + "learning_rate": 3.7718142466874984e-06, + "loss": 0.9194, + "step": 22595 + }, + { + "epoch": 0.72, + "learning_rate": 3.7710039021867085e-06, + "loss": 0.877, + "step": 22596 + }, + { + "epoch": 0.72, + "learning_rate": 3.770193624516156e-06, + "loss": 0.998, + "step": 22597 + }, + { + "epoch": 0.72, + "learning_rate": 3.769383413684532e-06, + "loss": 0.8242, + "step": 22598 + }, + { + "epoch": 0.72, + "learning_rate": 3.768573269700524e-06, + "loss": 0.8335, + "step": 22599 + }, + { + "epoch": 0.72, + "learning_rate": 3.767763192572832e-06, + "loss": 0.9092, + "step": 22600 + }, + { + "epoch": 0.72, + "learning_rate": 3.7669531823101435e-06, + "loss": 0.9072, + "step": 22601 + }, + { + "epoch": 0.72, + "learning_rate": 3.766143238921145e-06, + "loss": 1.0576, + "step": 22602 + }, + { + "epoch": 0.72, + "learning_rate": 3.765333362414534e-06, + "loss": 0.9946, + "step": 22603 + }, + { + "epoch": 0.72, + "learning_rate": 3.7645235527989956e-06, + "loss": 0.9878, + "step": 22604 + }, + { + "epoch": 0.72, + "learning_rate": 3.763713810083215e-06, + "loss": 0.8774, + "step": 22605 + }, + { + "epoch": 0.72, + "learning_rate": 3.7629041342758875e-06, + "loss": 0.7368, + "step": 22606 + }, + { + "epoch": 0.72, + "learning_rate": 3.7620945253856913e-06, + "loss": 0.896, + "step": 22607 + }, + { + "epoch": 0.72, + "learning_rate": 3.7612849834213207e-06, + "loss": 0.4729, + "step": 22608 + }, + { + "epoch": 0.72, + "learning_rate": 3.760475508391458e-06, + "loss": 0.9614, + "step": 22609 + }, + { + "epoch": 0.72, + "learning_rate": 3.7596661003047875e-06, + "loss": 0.9185, + "step": 22610 + }, + { + "epoch": 0.72, + "learning_rate": 3.75885675916999e-06, + "loss": 0.9648, + "step": 22611 + }, + { + "epoch": 0.72, + "learning_rate": 3.7580474849957547e-06, + "loss": 0.9595, + "step": 22612 + }, + { + "epoch": 0.72, + "learning_rate": 3.757238277790759e-06, + "loss": 0.9292, + "step": 22613 + }, + { + "epoch": 0.72, + "learning_rate": 3.756429137563691e-06, + "loss": 0.9189, + "step": 22614 + }, + { + "epoch": 0.72, + "learning_rate": 3.755620064323228e-06, + "loss": 0.7734, + "step": 22615 + }, + { + "epoch": 0.72, + "learning_rate": 3.7548110580780504e-06, + "loss": 0.8391, + "step": 22616 + }, + { + "epoch": 0.72, + "learning_rate": 3.754002118836835e-06, + "loss": 0.9775, + "step": 22617 + }, + { + "epoch": 0.72, + "learning_rate": 3.7531932466082677e-06, + "loss": 0.8584, + "step": 22618 + }, + { + "epoch": 0.72, + "learning_rate": 3.752384441401019e-06, + "loss": 0.9043, + "step": 22619 + }, + { + "epoch": 0.72, + "learning_rate": 3.7515757032237754e-06, + "loss": 0.9946, + "step": 22620 + }, + { + "epoch": 0.72, + "learning_rate": 3.750767032085205e-06, + "loss": 0.9336, + "step": 22621 + }, + { + "epoch": 0.72, + "learning_rate": 3.7499584279939925e-06, + "loss": 0.8892, + "step": 22622 + }, + { + "epoch": 0.72, + "learning_rate": 3.7491498909588076e-06, + "loss": 0.9741, + "step": 22623 + }, + { + "epoch": 0.72, + "learning_rate": 3.7483414209883274e-06, + "loss": 0.9092, + "step": 22624 + }, + { + "epoch": 0.72, + "learning_rate": 3.7475330180912206e-06, + "loss": 1.0093, + "step": 22625 + }, + { + "epoch": 0.72, + "learning_rate": 3.7467246822761683e-06, + "loss": 1.0269, + "step": 22626 + }, + { + "epoch": 0.72, + "learning_rate": 3.7459164135518358e-06, + "loss": 0.8364, + "step": 22627 + }, + { + "epoch": 0.72, + "learning_rate": 3.745108211926901e-06, + "loss": 0.8442, + "step": 22628 + }, + { + "epoch": 0.72, + "learning_rate": 3.74430007741003e-06, + "loss": 0.9976, + "step": 22629 + }, + { + "epoch": 0.72, + "learning_rate": 3.7434920100099025e-06, + "loss": 0.8813, + "step": 22630 + }, + { + "epoch": 0.72, + "learning_rate": 3.7426840097351747e-06, + "loss": 0.4783, + "step": 22631 + }, + { + "epoch": 0.72, + "learning_rate": 3.7418760765945262e-06, + "loss": 0.9097, + "step": 22632 + }, + { + "epoch": 0.72, + "learning_rate": 3.7410682105966167e-06, + "loss": 0.9731, + "step": 22633 + }, + { + "epoch": 0.72, + "learning_rate": 3.740260411750122e-06, + "loss": 0.8789, + "step": 22634 + }, + { + "epoch": 0.72, + "learning_rate": 3.739452680063702e-06, + "loss": 0.9136, + "step": 22635 + }, + { + "epoch": 0.72, + "learning_rate": 3.7386450155460296e-06, + "loss": 0.9849, + "step": 22636 + }, + { + "epoch": 0.72, + "learning_rate": 3.737837418205762e-06, + "loss": 0.8843, + "step": 22637 + }, + { + "epoch": 0.72, + "learning_rate": 3.7370298880515766e-06, + "loss": 0.9844, + "step": 22638 + }, + { + "epoch": 0.72, + "learning_rate": 3.736222425092121e-06, + "loss": 0.9243, + "step": 22639 + }, + { + "epoch": 0.72, + "learning_rate": 3.73541502933607e-06, + "loss": 0.8926, + "step": 22640 + }, + { + "epoch": 0.72, + "learning_rate": 3.734607700792079e-06, + "loss": 1.0044, + "step": 22641 + }, + { + "epoch": 0.72, + "learning_rate": 3.7338004394688164e-06, + "loss": 0.958, + "step": 22642 + }, + { + "epoch": 0.72, + "learning_rate": 3.732993245374941e-06, + "loss": 0.9639, + "step": 22643 + }, + { + "epoch": 0.72, + "learning_rate": 3.732186118519108e-06, + "loss": 0.9575, + "step": 22644 + }, + { + "epoch": 0.72, + "learning_rate": 3.731379058909984e-06, + "loss": 0.9873, + "step": 22645 + }, + { + "epoch": 0.72, + "learning_rate": 3.730572066556225e-06, + "loss": 0.8706, + "step": 22646 + }, + { + "epoch": 0.72, + "learning_rate": 3.7297651414664872e-06, + "loss": 0.8403, + "step": 22647 + }, + { + "epoch": 0.72, + "learning_rate": 3.728958283649432e-06, + "loss": 0.4897, + "step": 22648 + }, + { + "epoch": 0.72, + "learning_rate": 3.728151493113714e-06, + "loss": 0.9438, + "step": 22649 + }, + { + "epoch": 0.72, + "learning_rate": 3.7273447698679855e-06, + "loss": 0.9541, + "step": 22650 + }, + { + "epoch": 0.72, + "learning_rate": 3.726538113920909e-06, + "loss": 0.9487, + "step": 22651 + }, + { + "epoch": 0.72, + "learning_rate": 3.725731525281132e-06, + "loss": 1.0366, + "step": 22652 + }, + { + "epoch": 0.72, + "learning_rate": 3.724925003957316e-06, + "loss": 0.8838, + "step": 22653 + }, + { + "epoch": 0.72, + "learning_rate": 3.724118549958109e-06, + "loss": 0.9131, + "step": 22654 + }, + { + "epoch": 0.72, + "learning_rate": 3.7233121632921643e-06, + "loss": 0.8789, + "step": 22655 + }, + { + "epoch": 0.72, + "learning_rate": 3.7225058439681295e-06, + "loss": 0.9199, + "step": 22656 + }, + { + "epoch": 0.72, + "learning_rate": 3.721699591994663e-06, + "loss": 0.9053, + "step": 22657 + }, + { + "epoch": 0.72, + "learning_rate": 3.720893407380408e-06, + "loss": 0.9741, + "step": 22658 + }, + { + "epoch": 0.72, + "learning_rate": 3.7200872901340214e-06, + "loss": 1.1123, + "step": 22659 + }, + { + "epoch": 0.72, + "learning_rate": 3.7192812402641475e-06, + "loss": 0.7925, + "step": 22660 + }, + { + "epoch": 0.72, + "learning_rate": 3.7184752577794346e-06, + "loss": 0.9355, + "step": 22661 + }, + { + "epoch": 0.72, + "learning_rate": 3.7176693426885268e-06, + "loss": 1.0034, + "step": 22662 + }, + { + "epoch": 0.72, + "learning_rate": 3.716863495000078e-06, + "loss": 0.9492, + "step": 22663 + }, + { + "epoch": 0.72, + "learning_rate": 3.7160577147227262e-06, + "loss": 0.8743, + "step": 22664 + }, + { + "epoch": 0.72, + "learning_rate": 3.715252001865124e-06, + "loss": 0.9141, + "step": 22665 + }, + { + "epoch": 0.72, + "learning_rate": 3.7144463564359077e-06, + "loss": 0.9937, + "step": 22666 + }, + { + "epoch": 0.72, + "learning_rate": 3.71364077844373e-06, + "loss": 0.9907, + "step": 22667 + }, + { + "epoch": 0.72, + "learning_rate": 3.7128352678972302e-06, + "loss": 0.9624, + "step": 22668 + }, + { + "epoch": 0.72, + "learning_rate": 3.712029824805048e-06, + "loss": 1.0, + "step": 22669 + }, + { + "epoch": 0.72, + "learning_rate": 3.7112244491758243e-06, + "loss": 0.9375, + "step": 22670 + }, + { + "epoch": 0.73, + "learning_rate": 3.7104191410182044e-06, + "loss": 1.0459, + "step": 22671 + }, + { + "epoch": 0.73, + "learning_rate": 3.7096139003408237e-06, + "loss": 1.0093, + "step": 22672 + }, + { + "epoch": 0.73, + "learning_rate": 3.708808727152328e-06, + "loss": 0.9634, + "step": 22673 + }, + { + "epoch": 0.73, + "learning_rate": 3.708003621461347e-06, + "loss": 0.8896, + "step": 22674 + }, + { + "epoch": 0.73, + "learning_rate": 3.707198583276531e-06, + "loss": 1.0312, + "step": 22675 + }, + { + "epoch": 0.73, + "learning_rate": 3.7063936126065026e-06, + "loss": 0.9438, + "step": 22676 + }, + { + "epoch": 0.73, + "learning_rate": 3.7055887094599086e-06, + "loss": 0.9678, + "step": 22677 + }, + { + "epoch": 0.73, + "learning_rate": 3.704783873845379e-06, + "loss": 0.98, + "step": 22678 + }, + { + "epoch": 0.73, + "learning_rate": 3.7039791057715534e-06, + "loss": 0.9023, + "step": 22679 + }, + { + "epoch": 0.73, + "learning_rate": 3.703174405247062e-06, + "loss": 0.9053, + "step": 22680 + }, + { + "epoch": 0.73, + "learning_rate": 3.702369772280542e-06, + "loss": 0.9341, + "step": 22681 + }, + { + "epoch": 0.73, + "learning_rate": 3.701565206880625e-06, + "loss": 0.8877, + "step": 22682 + }, + { + "epoch": 0.73, + "learning_rate": 3.700760709055944e-06, + "loss": 1.0537, + "step": 22683 + }, + { + "epoch": 0.73, + "learning_rate": 3.6999562788151233e-06, + "loss": 0.9126, + "step": 22684 + }, + { + "epoch": 0.73, + "learning_rate": 3.6991519161668042e-06, + "loss": 0.9653, + "step": 22685 + }, + { + "epoch": 0.73, + "learning_rate": 3.698347621119608e-06, + "loss": 0.9731, + "step": 22686 + }, + { + "epoch": 0.73, + "learning_rate": 3.69754339368217e-06, + "loss": 0.9233, + "step": 22687 + }, + { + "epoch": 0.73, + "learning_rate": 3.696739233863117e-06, + "loss": 0.8188, + "step": 22688 + }, + { + "epoch": 0.73, + "learning_rate": 3.695935141671072e-06, + "loss": 0.9932, + "step": 22689 + }, + { + "epoch": 0.73, + "learning_rate": 3.6951311171146697e-06, + "loss": 0.9961, + "step": 22690 + }, + { + "epoch": 0.73, + "learning_rate": 3.694327160202532e-06, + "loss": 0.917, + "step": 22691 + }, + { + "epoch": 0.73, + "learning_rate": 3.6935232709432824e-06, + "loss": 0.8931, + "step": 22692 + }, + { + "epoch": 0.73, + "learning_rate": 3.692719449345552e-06, + "loss": 0.8179, + "step": 22693 + }, + { + "epoch": 0.73, + "learning_rate": 3.6919156954179612e-06, + "loss": 0.9077, + "step": 22694 + }, + { + "epoch": 0.73, + "learning_rate": 3.69111200916913e-06, + "loss": 1.0015, + "step": 22695 + }, + { + "epoch": 0.73, + "learning_rate": 3.6903083906076875e-06, + "loss": 1.0039, + "step": 22696 + }, + { + "epoch": 0.73, + "learning_rate": 3.6895048397422493e-06, + "loss": 0.9448, + "step": 22697 + }, + { + "epoch": 0.73, + "learning_rate": 3.6887013565814477e-06, + "loss": 0.918, + "step": 22698 + }, + { + "epoch": 0.73, + "learning_rate": 3.6878979411338887e-06, + "loss": 0.9507, + "step": 22699 + }, + { + "epoch": 0.73, + "learning_rate": 3.6870945934082025e-06, + "loss": 1.001, + "step": 22700 + }, + { + "epoch": 0.73, + "learning_rate": 3.686291313413001e-06, + "loss": 0.8975, + "step": 22701 + }, + { + "epoch": 0.73, + "learning_rate": 3.6854881011569087e-06, + "loss": 0.8809, + "step": 22702 + }, + { + "epoch": 0.73, + "learning_rate": 3.6846849566485377e-06, + "loss": 0.855, + "step": 22703 + }, + { + "epoch": 0.73, + "learning_rate": 3.6838818798965103e-06, + "loss": 0.8975, + "step": 22704 + }, + { + "epoch": 0.73, + "learning_rate": 3.6830788709094366e-06, + "loss": 1.0103, + "step": 22705 + }, + { + "epoch": 0.73, + "learning_rate": 3.682275929695943e-06, + "loss": 0.8574, + "step": 22706 + }, + { + "epoch": 0.73, + "learning_rate": 3.6814730562646295e-06, + "loss": 0.9497, + "step": 22707 + }, + { + "epoch": 0.73, + "learning_rate": 3.680670250624121e-06, + "loss": 1.0234, + "step": 22708 + }, + { + "epoch": 0.73, + "learning_rate": 3.679867512783023e-06, + "loss": 0.9087, + "step": 22709 + }, + { + "epoch": 0.73, + "learning_rate": 3.679064842749955e-06, + "loss": 0.8662, + "step": 22710 + }, + { + "epoch": 0.73, + "learning_rate": 3.678262240533522e-06, + "loss": 0.8228, + "step": 22711 + }, + { + "epoch": 0.73, + "learning_rate": 3.6774597061423436e-06, + "loss": 0.96, + "step": 22712 + }, + { + "epoch": 0.73, + "learning_rate": 3.676657239585023e-06, + "loss": 0.8086, + "step": 22713 + }, + { + "epoch": 0.73, + "learning_rate": 3.6758548408701734e-06, + "loss": 0.8984, + "step": 22714 + }, + { + "epoch": 0.73, + "learning_rate": 3.6750525100063973e-06, + "loss": 0.7812, + "step": 22715 + }, + { + "epoch": 0.73, + "learning_rate": 3.674250247002312e-06, + "loss": 0.9312, + "step": 22716 + }, + { + "epoch": 0.73, + "learning_rate": 3.673448051866517e-06, + "loss": 0.9844, + "step": 22717 + }, + { + "epoch": 0.73, + "learning_rate": 3.6726459246076262e-06, + "loss": 0.9233, + "step": 22718 + }, + { + "epoch": 0.73, + "learning_rate": 3.671843865234238e-06, + "loss": 0.9551, + "step": 22719 + }, + { + "epoch": 0.73, + "learning_rate": 3.6710418737549645e-06, + "loss": 0.8652, + "step": 22720 + }, + { + "epoch": 0.73, + "learning_rate": 3.670239950178408e-06, + "loss": 0.9678, + "step": 22721 + }, + { + "epoch": 0.73, + "learning_rate": 3.6694380945131714e-06, + "loss": 0.8848, + "step": 22722 + }, + { + "epoch": 0.73, + "learning_rate": 3.668636306767854e-06, + "loss": 0.9067, + "step": 22723 + }, + { + "epoch": 0.73, + "learning_rate": 3.6678345869510657e-06, + "loss": 0.9204, + "step": 22724 + }, + { + "epoch": 0.73, + "learning_rate": 3.6670329350713997e-06, + "loss": 0.5234, + "step": 22725 + }, + { + "epoch": 0.73, + "learning_rate": 3.666231351137466e-06, + "loss": 0.9268, + "step": 22726 + }, + { + "epoch": 0.73, + "learning_rate": 3.6654298351578586e-06, + "loss": 0.8853, + "step": 22727 + }, + { + "epoch": 0.73, + "learning_rate": 3.6646283871411803e-06, + "loss": 0.96, + "step": 22728 + }, + { + "epoch": 0.73, + "learning_rate": 3.6638270070960224e-06, + "loss": 0.9419, + "step": 22729 + }, + { + "epoch": 0.73, + "learning_rate": 3.663025695030993e-06, + "loss": 0.9946, + "step": 22730 + }, + { + "epoch": 0.73, + "learning_rate": 3.66222445095468e-06, + "loss": 0.96, + "step": 22731 + }, + { + "epoch": 0.73, + "learning_rate": 3.6614232748756882e-06, + "loss": 0.8955, + "step": 22732 + }, + { + "epoch": 0.73, + "learning_rate": 3.6606221668026098e-06, + "loss": 0.8799, + "step": 22733 + }, + { + "epoch": 0.73, + "learning_rate": 3.6598211267440353e-06, + "loss": 1.0381, + "step": 22734 + }, + { + "epoch": 0.73, + "learning_rate": 3.659020154708567e-06, + "loss": 0.9521, + "step": 22735 + }, + { + "epoch": 0.73, + "learning_rate": 3.6582192507047955e-06, + "loss": 0.9116, + "step": 22736 + }, + { + "epoch": 0.73, + "learning_rate": 3.657418414741312e-06, + "loss": 0.8613, + "step": 22737 + }, + { + "epoch": 0.73, + "learning_rate": 3.6566176468267047e-06, + "loss": 0.957, + "step": 22738 + }, + { + "epoch": 0.73, + "learning_rate": 3.6558169469695737e-06, + "loss": 0.896, + "step": 22739 + }, + { + "epoch": 0.73, + "learning_rate": 3.655016315178502e-06, + "loss": 0.7974, + "step": 22740 + }, + { + "epoch": 0.73, + "learning_rate": 3.6542157514620867e-06, + "loss": 0.8794, + "step": 22741 + }, + { + "epoch": 0.73, + "learning_rate": 3.6534152558289103e-06, + "loss": 1.0034, + "step": 22742 + }, + { + "epoch": 0.73, + "learning_rate": 3.6526148282875706e-06, + "loss": 1.0054, + "step": 22743 + }, + { + "epoch": 0.73, + "learning_rate": 3.6518144688466415e-06, + "loss": 0.9741, + "step": 22744 + }, + { + "epoch": 0.73, + "learning_rate": 3.651014177514721e-06, + "loss": 0.9111, + "step": 22745 + }, + { + "epoch": 0.73, + "learning_rate": 3.650213954300389e-06, + "loss": 0.9653, + "step": 22746 + }, + { + "epoch": 0.73, + "learning_rate": 3.649413799212236e-06, + "loss": 0.9268, + "step": 22747 + }, + { + "epoch": 0.73, + "learning_rate": 3.6486137122588417e-06, + "loss": 0.9126, + "step": 22748 + }, + { + "epoch": 0.73, + "learning_rate": 3.6478136934487962e-06, + "loss": 0.9644, + "step": 22749 + }, + { + "epoch": 0.73, + "learning_rate": 3.647013742790676e-06, + "loss": 0.9165, + "step": 22750 + }, + { + "epoch": 0.73, + "learning_rate": 3.646213860293074e-06, + "loss": 0.8999, + "step": 22751 + }, + { + "epoch": 0.73, + "learning_rate": 3.6454140459645582e-06, + "loss": 0.9541, + "step": 22752 + }, + { + "epoch": 0.73, + "learning_rate": 3.64461429981372e-06, + "loss": 0.7925, + "step": 22753 + }, + { + "epoch": 0.73, + "learning_rate": 3.643814621849134e-06, + "loss": 0.4946, + "step": 22754 + }, + { + "epoch": 0.73, + "learning_rate": 3.643015012079386e-06, + "loss": 1.0649, + "step": 22755 + }, + { + "epoch": 0.73, + "learning_rate": 3.6422154705130475e-06, + "loss": 0.8774, + "step": 22756 + }, + { + "epoch": 0.73, + "learning_rate": 3.641415997158704e-06, + "loss": 0.9478, + "step": 22757 + }, + { + "epoch": 0.73, + "learning_rate": 3.640616592024926e-06, + "loss": 0.9844, + "step": 22758 + }, + { + "epoch": 0.73, + "learning_rate": 3.6398172551203006e-06, + "loss": 0.8677, + "step": 22759 + }, + { + "epoch": 0.73, + "learning_rate": 3.6390179864533892e-06, + "loss": 0.9453, + "step": 22760 + }, + { + "epoch": 0.73, + "learning_rate": 3.638218786032779e-06, + "loss": 0.8213, + "step": 22761 + }, + { + "epoch": 0.73, + "learning_rate": 3.637419653867037e-06, + "loss": 0.7603, + "step": 22762 + }, + { + "epoch": 0.73, + "learning_rate": 3.6366205899647444e-06, + "loss": 1.0151, + "step": 22763 + }, + { + "epoch": 0.73, + "learning_rate": 3.6358215943344664e-06, + "loss": 0.917, + "step": 22764 + }, + { + "epoch": 0.73, + "learning_rate": 3.6350226669847823e-06, + "loss": 0.9224, + "step": 22765 + }, + { + "epoch": 0.73, + "learning_rate": 3.6342238079242607e-06, + "loss": 0.8872, + "step": 22766 + }, + { + "epoch": 0.73, + "learning_rate": 3.6334250171614727e-06, + "loss": 0.9351, + "step": 22767 + }, + { + "epoch": 0.73, + "learning_rate": 3.632626294704984e-06, + "loss": 0.8418, + "step": 22768 + }, + { + "epoch": 0.73, + "learning_rate": 3.6318276405633714e-06, + "loss": 0.8901, + "step": 22769 + }, + { + "epoch": 0.73, + "learning_rate": 3.6310290547451978e-06, + "loss": 0.8438, + "step": 22770 + }, + { + "epoch": 0.73, + "learning_rate": 3.6302305372590363e-06, + "loss": 0.8784, + "step": 22771 + }, + { + "epoch": 0.73, + "learning_rate": 3.6294320881134515e-06, + "loss": 1.0938, + "step": 22772 + }, + { + "epoch": 0.73, + "learning_rate": 3.628633707317006e-06, + "loss": 0.918, + "step": 22773 + }, + { + "epoch": 0.73, + "learning_rate": 3.627835394878274e-06, + "loss": 0.9111, + "step": 22774 + }, + { + "epoch": 0.73, + "learning_rate": 3.6270371508058155e-06, + "loss": 0.9204, + "step": 22775 + }, + { + "epoch": 0.73, + "learning_rate": 3.6262389751081938e-06, + "loss": 0.8301, + "step": 22776 + }, + { + "epoch": 0.73, + "learning_rate": 3.6254408677939724e-06, + "loss": 0.8633, + "step": 22777 + }, + { + "epoch": 0.73, + "learning_rate": 3.624642828871717e-06, + "loss": 0.843, + "step": 22778 + }, + { + "epoch": 0.73, + "learning_rate": 3.6238448583499854e-06, + "loss": 0.9155, + "step": 22779 + }, + { + "epoch": 0.73, + "learning_rate": 3.623046956237346e-06, + "loss": 0.812, + "step": 22780 + }, + { + "epoch": 0.73, + "learning_rate": 3.622249122542354e-06, + "loss": 0.895, + "step": 22781 + }, + { + "epoch": 0.73, + "learning_rate": 3.621451357273571e-06, + "loss": 0.9253, + "step": 22782 + }, + { + "epoch": 0.73, + "learning_rate": 3.6206536604395516e-06, + "loss": 0.8965, + "step": 22783 + }, + { + "epoch": 0.73, + "learning_rate": 3.619856032048862e-06, + "loss": 0.918, + "step": 22784 + }, + { + "epoch": 0.73, + "learning_rate": 3.619058472110052e-06, + "loss": 0.8994, + "step": 22785 + }, + { + "epoch": 0.73, + "learning_rate": 3.618260980631686e-06, + "loss": 0.8589, + "step": 22786 + }, + { + "epoch": 0.73, + "learning_rate": 3.6174635576223138e-06, + "loss": 0.9644, + "step": 22787 + }, + { + "epoch": 0.73, + "learning_rate": 3.6166662030904965e-06, + "loss": 0.9883, + "step": 22788 + }, + { + "epoch": 0.73, + "learning_rate": 3.615868917044787e-06, + "loss": 1.0273, + "step": 22789 + }, + { + "epoch": 0.73, + "learning_rate": 3.6150716994937375e-06, + "loss": 0.9106, + "step": 22790 + }, + { + "epoch": 0.73, + "learning_rate": 3.6142745504459e-06, + "loss": 0.9214, + "step": 22791 + }, + { + "epoch": 0.73, + "learning_rate": 3.6134774699098317e-06, + "loss": 0.834, + "step": 22792 + }, + { + "epoch": 0.73, + "learning_rate": 3.6126804578940787e-06, + "loss": 0.9683, + "step": 22793 + }, + { + "epoch": 0.73, + "learning_rate": 3.6118835144071986e-06, + "loss": 0.9585, + "step": 22794 + }, + { + "epoch": 0.73, + "learning_rate": 3.611086639457735e-06, + "loss": 0.5017, + "step": 22795 + }, + { + "epoch": 0.73, + "learning_rate": 3.610289833054247e-06, + "loss": 0.981, + "step": 22796 + }, + { + "epoch": 0.73, + "learning_rate": 3.6094930952052708e-06, + "loss": 0.8696, + "step": 22797 + }, + { + "epoch": 0.73, + "learning_rate": 3.6086964259193657e-06, + "loss": 0.9727, + "step": 22798 + }, + { + "epoch": 0.73, + "learning_rate": 3.60789982520507e-06, + "loss": 0.9766, + "step": 22799 + }, + { + "epoch": 0.73, + "learning_rate": 3.6071032930709383e-06, + "loss": 0.9277, + "step": 22800 + }, + { + "epoch": 0.73, + "learning_rate": 3.6063068295255088e-06, + "loss": 0.8613, + "step": 22801 + }, + { + "epoch": 0.73, + "learning_rate": 3.605510434577335e-06, + "loss": 0.8799, + "step": 22802 + }, + { + "epoch": 0.73, + "learning_rate": 3.6047141082349533e-06, + "loss": 0.8096, + "step": 22803 + }, + { + "epoch": 0.73, + "learning_rate": 3.6039178505069197e-06, + "loss": 0.7866, + "step": 22804 + }, + { + "epoch": 0.73, + "learning_rate": 3.6031216614017604e-06, + "loss": 0.8384, + "step": 22805 + }, + { + "epoch": 0.73, + "learning_rate": 3.602325540928029e-06, + "loss": 1.0249, + "step": 22806 + }, + { + "epoch": 0.73, + "learning_rate": 3.601529489094261e-06, + "loss": 0.917, + "step": 22807 + }, + { + "epoch": 0.73, + "learning_rate": 3.6007335059090044e-06, + "loss": 0.8735, + "step": 22808 + }, + { + "epoch": 0.73, + "learning_rate": 3.599937591380791e-06, + "loss": 0.9199, + "step": 22809 + }, + { + "epoch": 0.73, + "learning_rate": 3.5991417455181666e-06, + "loss": 0.9736, + "step": 22810 + }, + { + "epoch": 0.73, + "learning_rate": 3.5983459683296684e-06, + "loss": 0.9082, + "step": 22811 + }, + { + "epoch": 0.73, + "learning_rate": 3.597550259823832e-06, + "loss": 0.9292, + "step": 22812 + }, + { + "epoch": 0.73, + "learning_rate": 3.5967546200091917e-06, + "loss": 0.9487, + "step": 22813 + }, + { + "epoch": 0.73, + "learning_rate": 3.595959048894292e-06, + "loss": 1.0605, + "step": 22814 + }, + { + "epoch": 0.73, + "learning_rate": 3.595163546487663e-06, + "loss": 0.8169, + "step": 22815 + }, + { + "epoch": 0.73, + "learning_rate": 3.594368112797839e-06, + "loss": 0.9717, + "step": 22816 + }, + { + "epoch": 0.73, + "learning_rate": 3.593572747833358e-06, + "loss": 0.874, + "step": 22817 + }, + { + "epoch": 0.73, + "learning_rate": 3.592777451602748e-06, + "loss": 0.9106, + "step": 22818 + }, + { + "epoch": 0.73, + "learning_rate": 3.5919822241145476e-06, + "loss": 1.0464, + "step": 22819 + }, + { + "epoch": 0.73, + "learning_rate": 3.5911870653772863e-06, + "loss": 0.7925, + "step": 22820 + }, + { + "epoch": 0.73, + "learning_rate": 3.5903919753994944e-06, + "loss": 0.9355, + "step": 22821 + }, + { + "epoch": 0.73, + "learning_rate": 3.5895969541897003e-06, + "loss": 0.9448, + "step": 22822 + }, + { + "epoch": 0.73, + "learning_rate": 3.5888020017564396e-06, + "loss": 0.9536, + "step": 22823 + }, + { + "epoch": 0.73, + "learning_rate": 3.588007118108233e-06, + "loss": 0.9839, + "step": 22824 + }, + { + "epoch": 0.73, + "learning_rate": 3.5872123032536186e-06, + "loss": 0.9053, + "step": 22825 + }, + { + "epoch": 0.73, + "learning_rate": 3.5864175572011184e-06, + "loss": 0.8799, + "step": 22826 + }, + { + "epoch": 0.73, + "learning_rate": 3.5856228799592594e-06, + "loss": 1.0679, + "step": 22827 + }, + { + "epoch": 0.73, + "learning_rate": 3.5848282715365646e-06, + "loss": 0.8003, + "step": 22828 + }, + { + "epoch": 0.73, + "learning_rate": 3.584033731941565e-06, + "loss": 0.8252, + "step": 22829 + }, + { + "epoch": 0.73, + "learning_rate": 3.5832392611827802e-06, + "loss": 1.0078, + "step": 22830 + }, + { + "epoch": 0.73, + "learning_rate": 3.58244485926874e-06, + "loss": 1.0449, + "step": 22831 + }, + { + "epoch": 0.73, + "learning_rate": 3.581650526207959e-06, + "loss": 1.0264, + "step": 22832 + }, + { + "epoch": 0.73, + "learning_rate": 3.5808562620089695e-06, + "loss": 0.9624, + "step": 22833 + }, + { + "epoch": 0.73, + "learning_rate": 3.580062066680288e-06, + "loss": 0.9341, + "step": 22834 + }, + { + "epoch": 0.73, + "learning_rate": 3.5792679402304344e-06, + "loss": 0.9448, + "step": 22835 + }, + { + "epoch": 0.73, + "learning_rate": 3.5784738826679253e-06, + "loss": 0.9282, + "step": 22836 + }, + { + "epoch": 0.73, + "learning_rate": 3.577679894001289e-06, + "loss": 0.8774, + "step": 22837 + }, + { + "epoch": 0.73, + "learning_rate": 3.576885974239037e-06, + "loss": 0.8936, + "step": 22838 + }, + { + "epoch": 0.73, + "learning_rate": 3.5760921233896918e-06, + "loss": 0.9531, + "step": 22839 + }, + { + "epoch": 0.73, + "learning_rate": 3.575298341461766e-06, + "loss": 0.9731, + "step": 22840 + }, + { + "epoch": 0.73, + "learning_rate": 3.5745046284637807e-06, + "loss": 0.9907, + "step": 22841 + }, + { + "epoch": 0.73, + "learning_rate": 3.5737109844042505e-06, + "loss": 0.9336, + "step": 22842 + }, + { + "epoch": 0.73, + "learning_rate": 3.5729174092916886e-06, + "loss": 0.8506, + "step": 22843 + }, + { + "epoch": 0.73, + "learning_rate": 3.5721239031346067e-06, + "loss": 0.9722, + "step": 22844 + }, + { + "epoch": 0.73, + "learning_rate": 3.571330465941525e-06, + "loss": 0.8948, + "step": 22845 + }, + { + "epoch": 0.73, + "learning_rate": 3.5705370977209484e-06, + "loss": 0.8853, + "step": 22846 + }, + { + "epoch": 0.73, + "learning_rate": 3.5697437984813965e-06, + "loss": 1.041, + "step": 22847 + }, + { + "epoch": 0.73, + "learning_rate": 3.568950568231374e-06, + "loss": 0.9023, + "step": 22848 + }, + { + "epoch": 0.73, + "learning_rate": 3.568157406979401e-06, + "loss": 0.8843, + "step": 22849 + }, + { + "epoch": 0.73, + "learning_rate": 3.567364314733973e-06, + "loss": 0.979, + "step": 22850 + }, + { + "epoch": 0.73, + "learning_rate": 3.566571291503611e-06, + "loss": 0.8896, + "step": 22851 + }, + { + "epoch": 0.73, + "learning_rate": 3.565778337296816e-06, + "loss": 0.8105, + "step": 22852 + }, + { + "epoch": 0.73, + "learning_rate": 3.5649854521221006e-06, + "loss": 1.0625, + "step": 22853 + }, + { + "epoch": 0.73, + "learning_rate": 3.5641926359879663e-06, + "loss": 0.96, + "step": 22854 + }, + { + "epoch": 0.73, + "learning_rate": 3.5633998889029253e-06, + "loss": 0.8857, + "step": 22855 + }, + { + "epoch": 0.73, + "learning_rate": 3.5626072108754806e-06, + "loss": 0.771, + "step": 22856 + }, + { + "epoch": 0.73, + "learning_rate": 3.561814601914135e-06, + "loss": 0.8208, + "step": 22857 + }, + { + "epoch": 0.73, + "learning_rate": 3.5610220620273906e-06, + "loss": 0.9243, + "step": 22858 + }, + { + "epoch": 0.73, + "learning_rate": 3.560229591223756e-06, + "loss": 0.9722, + "step": 22859 + }, + { + "epoch": 0.73, + "learning_rate": 3.559437189511732e-06, + "loss": 0.8853, + "step": 22860 + }, + { + "epoch": 0.73, + "learning_rate": 3.5586448568998143e-06, + "loss": 0.9258, + "step": 22861 + }, + { + "epoch": 0.73, + "learning_rate": 3.5578525933965124e-06, + "loss": 0.8271, + "step": 22862 + }, + { + "epoch": 0.73, + "learning_rate": 3.557060399010318e-06, + "loss": 0.9922, + "step": 22863 + }, + { + "epoch": 0.73, + "learning_rate": 3.556268273749739e-06, + "loss": 0.875, + "step": 22864 + }, + { + "epoch": 0.73, + "learning_rate": 3.5554762176232694e-06, + "loss": 0.8984, + "step": 22865 + }, + { + "epoch": 0.73, + "learning_rate": 3.554684230639408e-06, + "loss": 0.9463, + "step": 22866 + }, + { + "epoch": 0.73, + "learning_rate": 3.553892312806648e-06, + "loss": 0.917, + "step": 22867 + }, + { + "epoch": 0.73, + "learning_rate": 3.553100464133493e-06, + "loss": 1.0166, + "step": 22868 + }, + { + "epoch": 0.73, + "learning_rate": 3.5523086846284303e-06, + "loss": 0.9751, + "step": 22869 + }, + { + "epoch": 0.73, + "learning_rate": 3.5515169742999646e-06, + "loss": 0.8813, + "step": 22870 + }, + { + "epoch": 0.73, + "learning_rate": 3.5507253331565804e-06, + "loss": 0.9844, + "step": 22871 + }, + { + "epoch": 0.73, + "learning_rate": 3.549933761206783e-06, + "loss": 0.8638, + "step": 22872 + }, + { + "epoch": 0.73, + "learning_rate": 3.5491422584590506e-06, + "loss": 0.51, + "step": 22873 + }, + { + "epoch": 0.73, + "learning_rate": 3.5483508249218858e-06, + "loss": 0.9482, + "step": 22874 + }, + { + "epoch": 0.73, + "learning_rate": 3.5475594606037723e-06, + "loss": 1.022, + "step": 22875 + }, + { + "epoch": 0.73, + "learning_rate": 3.5467681655132093e-06, + "loss": 0.9673, + "step": 22876 + }, + { + "epoch": 0.73, + "learning_rate": 3.5459769396586773e-06, + "loss": 0.8335, + "step": 22877 + }, + { + "epoch": 0.73, + "learning_rate": 3.5451857830486736e-06, + "loss": 0.77, + "step": 22878 + }, + { + "epoch": 0.73, + "learning_rate": 3.544394695691683e-06, + "loss": 0.9707, + "step": 22879 + }, + { + "epoch": 0.73, + "learning_rate": 3.5436036775961925e-06, + "loss": 0.9512, + "step": 22880 + }, + { + "epoch": 0.73, + "learning_rate": 3.542812728770685e-06, + "loss": 1.0068, + "step": 22881 + }, + { + "epoch": 0.73, + "learning_rate": 3.5420218492236536e-06, + "loss": 0.8955, + "step": 22882 + }, + { + "epoch": 0.73, + "learning_rate": 3.5412310389635773e-06, + "loss": 0.8662, + "step": 22883 + }, + { + "epoch": 0.73, + "learning_rate": 3.540440297998947e-06, + "loss": 0.8608, + "step": 22884 + }, + { + "epoch": 0.73, + "learning_rate": 3.5396496263382408e-06, + "loss": 0.9658, + "step": 22885 + }, + { + "epoch": 0.73, + "learning_rate": 3.5388590239899467e-06, + "loss": 0.8364, + "step": 22886 + }, + { + "epoch": 0.73, + "learning_rate": 3.5380684909625453e-06, + "loss": 1.0034, + "step": 22887 + }, + { + "epoch": 0.73, + "learning_rate": 3.537278027264517e-06, + "loss": 0.9585, + "step": 22888 + }, + { + "epoch": 0.73, + "learning_rate": 3.5364876329043385e-06, + "loss": 1.0181, + "step": 22889 + }, + { + "epoch": 0.73, + "learning_rate": 3.5356973078904976e-06, + "loss": 0.9316, + "step": 22890 + }, + { + "epoch": 0.73, + "learning_rate": 3.534907052231468e-06, + "loss": 0.9609, + "step": 22891 + }, + { + "epoch": 0.73, + "learning_rate": 3.534116865935734e-06, + "loss": 0.9668, + "step": 22892 + }, + { + "epoch": 0.73, + "learning_rate": 3.533326749011765e-06, + "loss": 0.8979, + "step": 22893 + }, + { + "epoch": 0.73, + "learning_rate": 3.532536701468051e-06, + "loss": 0.8975, + "step": 22894 + }, + { + "epoch": 0.73, + "learning_rate": 3.5317467233130542e-06, + "loss": 0.9175, + "step": 22895 + }, + { + "epoch": 0.73, + "learning_rate": 3.5309568145552585e-06, + "loss": 0.8853, + "step": 22896 + }, + { + "epoch": 0.73, + "learning_rate": 3.530166975203133e-06, + "loss": 0.9385, + "step": 22897 + }, + { + "epoch": 0.73, + "learning_rate": 3.5293772052651597e-06, + "loss": 0.877, + "step": 22898 + }, + { + "epoch": 0.73, + "learning_rate": 3.5285875047498075e-06, + "loss": 0.938, + "step": 22899 + }, + { + "epoch": 0.73, + "learning_rate": 3.527797873665546e-06, + "loss": 0.9995, + "step": 22900 + }, + { + "epoch": 0.73, + "learning_rate": 3.527008312020853e-06, + "loss": 0.9463, + "step": 22901 + }, + { + "epoch": 0.73, + "learning_rate": 3.5262188198241978e-06, + "loss": 1.0635, + "step": 22902 + }, + { + "epoch": 0.73, + "learning_rate": 3.5254293970840447e-06, + "loss": 0.8647, + "step": 22903 + }, + { + "epoch": 0.73, + "learning_rate": 3.5246400438088735e-06, + "loss": 0.8682, + "step": 22904 + }, + { + "epoch": 0.73, + "learning_rate": 3.5238507600071468e-06, + "loss": 0.9292, + "step": 22905 + }, + { + "epoch": 0.73, + "learning_rate": 3.523061545687332e-06, + "loss": 0.9512, + "step": 22906 + }, + { + "epoch": 0.73, + "learning_rate": 3.522272400857901e-06, + "loss": 0.8252, + "step": 22907 + }, + { + "epoch": 0.73, + "learning_rate": 3.521483325527315e-06, + "loss": 0.9829, + "step": 22908 + }, + { + "epoch": 0.73, + "learning_rate": 3.5206943197040456e-06, + "loss": 0.9849, + "step": 22909 + }, + { + "epoch": 0.73, + "learning_rate": 3.519905383396556e-06, + "loss": 0.9985, + "step": 22910 + }, + { + "epoch": 0.73, + "learning_rate": 3.519116516613309e-06, + "loss": 0.9106, + "step": 22911 + }, + { + "epoch": 0.73, + "learning_rate": 3.518327719362766e-06, + "loss": 0.8418, + "step": 22912 + }, + { + "epoch": 0.73, + "learning_rate": 3.5175389916533964e-06, + "loss": 1.0576, + "step": 22913 + }, + { + "epoch": 0.73, + "learning_rate": 3.516750333493656e-06, + "loss": 0.8188, + "step": 22914 + }, + { + "epoch": 0.73, + "learning_rate": 3.5159617448920125e-06, + "loss": 0.9502, + "step": 22915 + }, + { + "epoch": 0.73, + "learning_rate": 3.5151732258569195e-06, + "loss": 0.9468, + "step": 22916 + }, + { + "epoch": 0.73, + "learning_rate": 3.5143847763968477e-06, + "loss": 0.9102, + "step": 22917 + }, + { + "epoch": 0.73, + "learning_rate": 3.5135963965202423e-06, + "loss": 1.0327, + "step": 22918 + }, + { + "epoch": 0.73, + "learning_rate": 3.512808086235573e-06, + "loss": 0.9688, + "step": 22919 + }, + { + "epoch": 0.73, + "learning_rate": 3.5120198455512887e-06, + "loss": 0.9966, + "step": 22920 + }, + { + "epoch": 0.73, + "learning_rate": 3.5112316744758558e-06, + "loss": 0.998, + "step": 22921 + }, + { + "epoch": 0.73, + "learning_rate": 3.510443573017721e-06, + "loss": 0.8838, + "step": 22922 + }, + { + "epoch": 0.73, + "learning_rate": 3.509655541185348e-06, + "loss": 0.938, + "step": 22923 + }, + { + "epoch": 0.73, + "learning_rate": 3.508867578987185e-06, + "loss": 0.9741, + "step": 22924 + }, + { + "epoch": 0.73, + "learning_rate": 3.5080796864316947e-06, + "loss": 1.0688, + "step": 22925 + }, + { + "epoch": 0.73, + "learning_rate": 3.507291863527319e-06, + "loss": 0.8809, + "step": 22926 + }, + { + "epoch": 0.73, + "learning_rate": 3.506504110282518e-06, + "loss": 1.0513, + "step": 22927 + }, + { + "epoch": 0.73, + "learning_rate": 3.5057164267057388e-06, + "loss": 0.9312, + "step": 22928 + }, + { + "epoch": 0.73, + "learning_rate": 3.504928812805437e-06, + "loss": 0.9541, + "step": 22929 + }, + { + "epoch": 0.73, + "learning_rate": 3.504141268590059e-06, + "loss": 1.0532, + "step": 22930 + }, + { + "epoch": 0.73, + "learning_rate": 3.5033537940680575e-06, + "loss": 0.8994, + "step": 22931 + }, + { + "epoch": 0.73, + "learning_rate": 3.5025663892478813e-06, + "loss": 0.8936, + "step": 22932 + }, + { + "epoch": 0.73, + "learning_rate": 3.501779054137975e-06, + "loss": 0.8501, + "step": 22933 + }, + { + "epoch": 0.73, + "learning_rate": 3.500991788746786e-06, + "loss": 0.96, + "step": 22934 + }, + { + "epoch": 0.73, + "learning_rate": 3.5002045930827644e-06, + "loss": 0.8823, + "step": 22935 + }, + { + "epoch": 0.73, + "learning_rate": 3.4994174671543513e-06, + "loss": 0.917, + "step": 22936 + }, + { + "epoch": 0.73, + "learning_rate": 3.4986304109699965e-06, + "loss": 1.0391, + "step": 22937 + }, + { + "epoch": 0.73, + "learning_rate": 3.497843424538143e-06, + "loss": 0.877, + "step": 22938 + }, + { + "epoch": 0.73, + "learning_rate": 3.4970565078672303e-06, + "loss": 0.8516, + "step": 22939 + }, + { + "epoch": 0.73, + "learning_rate": 3.4962696609657064e-06, + "loss": 0.9922, + "step": 22940 + }, + { + "epoch": 0.73, + "learning_rate": 3.495482883842012e-06, + "loss": 0.9565, + "step": 22941 + }, + { + "epoch": 0.73, + "learning_rate": 3.4946961765045837e-06, + "loss": 0.5095, + "step": 22942 + }, + { + "epoch": 0.73, + "learning_rate": 3.4939095389618692e-06, + "loss": 0.9448, + "step": 22943 + }, + { + "epoch": 0.73, + "learning_rate": 3.4931229712223047e-06, + "loss": 0.8838, + "step": 22944 + }, + { + "epoch": 0.73, + "learning_rate": 3.4923364732943254e-06, + "loss": 0.8296, + "step": 22945 + }, + { + "epoch": 0.73, + "learning_rate": 3.4915500451863783e-06, + "loss": 0.9131, + "step": 22946 + }, + { + "epoch": 0.73, + "learning_rate": 3.490763686906895e-06, + "loss": 0.9604, + "step": 22947 + }, + { + "epoch": 0.73, + "learning_rate": 3.489977398464309e-06, + "loss": 0.8721, + "step": 22948 + }, + { + "epoch": 0.73, + "learning_rate": 3.489191179867065e-06, + "loss": 1.0459, + "step": 22949 + }, + { + "epoch": 0.73, + "learning_rate": 3.488405031123594e-06, + "loss": 0.8682, + "step": 22950 + }, + { + "epoch": 0.73, + "learning_rate": 3.4876189522423264e-06, + "loss": 0.9644, + "step": 22951 + }, + { + "epoch": 0.73, + "learning_rate": 3.4868329432317037e-06, + "loss": 1.022, + "step": 22952 + }, + { + "epoch": 0.73, + "learning_rate": 3.4860470041001515e-06, + "loss": 0.9805, + "step": 22953 + }, + { + "epoch": 0.73, + "learning_rate": 3.4852611348561095e-06, + "loss": 0.959, + "step": 22954 + }, + { + "epoch": 0.73, + "learning_rate": 3.4844753355080053e-06, + "loss": 1.0103, + "step": 22955 + }, + { + "epoch": 0.73, + "learning_rate": 3.483689606064269e-06, + "loss": 0.9863, + "step": 22956 + }, + { + "epoch": 0.73, + "learning_rate": 3.4829039465333282e-06, + "loss": 0.9795, + "step": 22957 + }, + { + "epoch": 0.73, + "learning_rate": 3.4821183569236185e-06, + "loss": 0.8735, + "step": 22958 + }, + { + "epoch": 0.73, + "learning_rate": 3.4813328372435627e-06, + "loss": 0.5095, + "step": 22959 + }, + { + "epoch": 0.73, + "learning_rate": 3.480547387501594e-06, + "loss": 0.8882, + "step": 22960 + }, + { + "epoch": 0.73, + "learning_rate": 3.4797620077061325e-06, + "loss": 0.9268, + "step": 22961 + }, + { + "epoch": 0.73, + "learning_rate": 3.4789766978656157e-06, + "loss": 0.9258, + "step": 22962 + }, + { + "epoch": 0.73, + "learning_rate": 3.4781914579884547e-06, + "loss": 0.8667, + "step": 22963 + }, + { + "epoch": 0.73, + "learning_rate": 3.477406288083085e-06, + "loss": 0.8916, + "step": 22964 + }, + { + "epoch": 0.73, + "learning_rate": 3.476621188157924e-06, + "loss": 0.8804, + "step": 22965 + }, + { + "epoch": 0.73, + "learning_rate": 3.4758361582214008e-06, + "loss": 0.8999, + "step": 22966 + }, + { + "epoch": 0.73, + "learning_rate": 3.4750511982819313e-06, + "loss": 0.8911, + "step": 22967 + }, + { + "epoch": 0.73, + "learning_rate": 3.4742663083479456e-06, + "loss": 0.9121, + "step": 22968 + }, + { + "epoch": 0.73, + "learning_rate": 3.4734814884278554e-06, + "loss": 1.0688, + "step": 22969 + }, + { + "epoch": 0.73, + "learning_rate": 3.4726967385300936e-06, + "loss": 0.8892, + "step": 22970 + }, + { + "epoch": 0.73, + "learning_rate": 3.471912058663064e-06, + "loss": 0.9409, + "step": 22971 + }, + { + "epoch": 0.73, + "learning_rate": 3.4711274488351967e-06, + "loss": 0.9194, + "step": 22972 + }, + { + "epoch": 0.73, + "learning_rate": 3.4703429090549023e-06, + "loss": 0.918, + "step": 22973 + }, + { + "epoch": 0.73, + "learning_rate": 3.4695584393306057e-06, + "loss": 0.8442, + "step": 22974 + }, + { + "epoch": 0.73, + "learning_rate": 3.468774039670717e-06, + "loss": 0.917, + "step": 22975 + }, + { + "epoch": 0.73, + "learning_rate": 3.467989710083657e-06, + "loss": 0.9766, + "step": 22976 + }, + { + "epoch": 0.73, + "learning_rate": 3.4672054505778375e-06, + "loss": 0.9858, + "step": 22977 + }, + { + "epoch": 0.73, + "learning_rate": 3.4664212611616744e-06, + "loss": 0.9536, + "step": 22978 + }, + { + "epoch": 0.73, + "learning_rate": 3.4656371418435765e-06, + "loss": 1.0103, + "step": 22979 + }, + { + "epoch": 0.73, + "learning_rate": 3.4648530926319634e-06, + "loss": 0.8706, + "step": 22980 + }, + { + "epoch": 0.73, + "learning_rate": 3.46406911353524e-06, + "loss": 0.9414, + "step": 22981 + }, + { + "epoch": 0.73, + "learning_rate": 3.463285204561826e-06, + "loss": 0.4204, + "step": 22982 + }, + { + "epoch": 0.74, + "learning_rate": 3.4625013657201255e-06, + "loss": 0.4692, + "step": 22983 + }, + { + "epoch": 0.74, + "learning_rate": 3.461717597018548e-06, + "loss": 0.9214, + "step": 22984 + }, + { + "epoch": 0.74, + "learning_rate": 3.4609338984655083e-06, + "loss": 0.957, + "step": 22985 + }, + { + "epoch": 0.74, + "learning_rate": 3.460150270069409e-06, + "loss": 0.9609, + "step": 22986 + }, + { + "epoch": 0.74, + "learning_rate": 3.4593667118386574e-06, + "loss": 0.9727, + "step": 22987 + }, + { + "epoch": 0.74, + "learning_rate": 3.4585832237816653e-06, + "loss": 0.7676, + "step": 22988 + }, + { + "epoch": 0.74, + "learning_rate": 3.4577998059068354e-06, + "loss": 1.0215, + "step": 22989 + }, + { + "epoch": 0.74, + "learning_rate": 3.4570164582225696e-06, + "loss": 0.9297, + "step": 22990 + }, + { + "epoch": 0.74, + "learning_rate": 3.4562331807372796e-06, + "loss": 1.0132, + "step": 22991 + }, + { + "epoch": 0.74, + "learning_rate": 3.4554499734593615e-06, + "loss": 0.8582, + "step": 22992 + }, + { + "epoch": 0.74, + "learning_rate": 3.454666836397228e-06, + "loss": 0.939, + "step": 22993 + }, + { + "epoch": 0.74, + "learning_rate": 3.4538837695592696e-06, + "loss": 0.8633, + "step": 22994 + }, + { + "epoch": 0.74, + "learning_rate": 3.4531007729538966e-06, + "loss": 0.9385, + "step": 22995 + }, + { + "epoch": 0.74, + "learning_rate": 3.4523178465895037e-06, + "loss": 0.9722, + "step": 22996 + }, + { + "epoch": 0.74, + "learning_rate": 3.4515349904744964e-06, + "loss": 0.9868, + "step": 22997 + }, + { + "epoch": 0.74, + "learning_rate": 3.450752204617268e-06, + "loss": 1.0908, + "step": 22998 + }, + { + "epoch": 0.74, + "learning_rate": 3.4499694890262235e-06, + "loss": 0.9805, + "step": 22999 + }, + { + "epoch": 0.74, + "learning_rate": 3.4491868437097575e-06, + "loss": 0.8, + "step": 23000 + }, + { + "epoch": 0.74, + "learning_rate": 3.4484042686762653e-06, + "loss": 1.0166, + "step": 23001 + }, + { + "epoch": 0.74, + "learning_rate": 3.4476217639341423e-06, + "loss": 0.9756, + "step": 23002 + }, + { + "epoch": 0.74, + "learning_rate": 3.446839329491789e-06, + "loss": 0.9097, + "step": 23003 + }, + { + "epoch": 0.74, + "learning_rate": 3.4460569653575936e-06, + "loss": 0.7979, + "step": 23004 + }, + { + "epoch": 0.74, + "learning_rate": 3.4452746715399564e-06, + "loss": 0.957, + "step": 23005 + }, + { + "epoch": 0.74, + "learning_rate": 3.4444924480472652e-06, + "loss": 0.9092, + "step": 23006 + }, + { + "epoch": 0.74, + "learning_rate": 3.4437102948879176e-06, + "loss": 0.9316, + "step": 23007 + }, + { + "epoch": 0.74, + "learning_rate": 3.4429282120703024e-06, + "loss": 1.0483, + "step": 23008 + }, + { + "epoch": 0.74, + "learning_rate": 3.4421461996028092e-06, + "loss": 0.8901, + "step": 23009 + }, + { + "epoch": 0.74, + "learning_rate": 3.441364257493828e-06, + "loss": 0.8652, + "step": 23010 + }, + { + "epoch": 0.74, + "learning_rate": 3.4405823857517516e-06, + "loss": 0.9429, + "step": 23011 + }, + { + "epoch": 0.74, + "learning_rate": 3.4398005843849634e-06, + "loss": 0.9126, + "step": 23012 + }, + { + "epoch": 0.74, + "learning_rate": 3.439018853401859e-06, + "loss": 0.8892, + "step": 23013 + }, + { + "epoch": 0.74, + "learning_rate": 3.438237192810816e-06, + "loss": 0.9551, + "step": 23014 + }, + { + "epoch": 0.74, + "learning_rate": 3.4374556026202333e-06, + "loss": 0.8911, + "step": 23015 + }, + { + "epoch": 0.74, + "learning_rate": 3.4366740828384824e-06, + "loss": 0.9087, + "step": 23016 + }, + { + "epoch": 0.74, + "learning_rate": 3.4358926334739584e-06, + "loss": 0.8896, + "step": 23017 + }, + { + "epoch": 0.74, + "learning_rate": 3.435111254535037e-06, + "loss": 0.9995, + "step": 23018 + }, + { + "epoch": 0.74, + "learning_rate": 3.4343299460301106e-06, + "loss": 0.9966, + "step": 23019 + }, + { + "epoch": 0.74, + "learning_rate": 3.4335487079675535e-06, + "loss": 0.9321, + "step": 23020 + }, + { + "epoch": 0.74, + "learning_rate": 3.4327675403557547e-06, + "loss": 0.9355, + "step": 23021 + }, + { + "epoch": 0.74, + "learning_rate": 3.4319864432030914e-06, + "loss": 0.9272, + "step": 23022 + }, + { + "epoch": 0.74, + "learning_rate": 3.4312054165179454e-06, + "loss": 0.9248, + "step": 23023 + }, + { + "epoch": 0.74, + "learning_rate": 3.4304244603086923e-06, + "loss": 0.9199, + "step": 23024 + }, + { + "epoch": 0.74, + "learning_rate": 3.4296435745837163e-06, + "loss": 0.9727, + "step": 23025 + }, + { + "epoch": 0.74, + "learning_rate": 3.42886275935139e-06, + "loss": 0.9692, + "step": 23026 + }, + { + "epoch": 0.74, + "learning_rate": 3.4280820146200968e-06, + "loss": 0.8618, + "step": 23027 + }, + { + "epoch": 0.74, + "learning_rate": 3.42730134039821e-06, + "loss": 0.8159, + "step": 23028 + }, + { + "epoch": 0.74, + "learning_rate": 3.4265207366941032e-06, + "loss": 0.9292, + "step": 23029 + }, + { + "epoch": 0.74, + "learning_rate": 3.425740203516156e-06, + "loss": 0.9722, + "step": 23030 + }, + { + "epoch": 0.74, + "learning_rate": 3.4249597408727407e-06, + "loss": 0.8994, + "step": 23031 + }, + { + "epoch": 0.74, + "learning_rate": 3.424179348772231e-06, + "loss": 0.8242, + "step": 23032 + }, + { + "epoch": 0.74, + "learning_rate": 3.4233990272229955e-06, + "loss": 0.7571, + "step": 23033 + }, + { + "epoch": 0.74, + "learning_rate": 3.422618776233413e-06, + "loss": 0.9897, + "step": 23034 + }, + { + "epoch": 0.74, + "learning_rate": 3.4218385958118484e-06, + "loss": 0.8909, + "step": 23035 + }, + { + "epoch": 0.74, + "learning_rate": 3.4210584859666783e-06, + "loss": 0.8584, + "step": 23036 + }, + { + "epoch": 0.74, + "learning_rate": 3.4202784467062667e-06, + "loss": 0.9214, + "step": 23037 + }, + { + "epoch": 0.74, + "learning_rate": 3.4194984780389904e-06, + "loss": 0.8789, + "step": 23038 + }, + { + "epoch": 0.74, + "learning_rate": 3.418718579973207e-06, + "loss": 0.9062, + "step": 23039 + }, + { + "epoch": 0.74, + "learning_rate": 3.4179387525172935e-06, + "loss": 0.9446, + "step": 23040 + }, + { + "epoch": 0.74, + "learning_rate": 3.417158995679608e-06, + "loss": 0.9458, + "step": 23041 + }, + { + "epoch": 0.74, + "learning_rate": 3.416379309468525e-06, + "loss": 1.0293, + "step": 23042 + }, + { + "epoch": 0.74, + "learning_rate": 3.4155996938924017e-06, + "loss": 0.9326, + "step": 23043 + }, + { + "epoch": 0.74, + "learning_rate": 3.41482014895961e-06, + "loss": 0.4939, + "step": 23044 + }, + { + "epoch": 0.74, + "learning_rate": 3.4140406746785104e-06, + "loss": 1.001, + "step": 23045 + }, + { + "epoch": 0.74, + "learning_rate": 3.4132612710574655e-06, + "loss": 0.7871, + "step": 23046 + }, + { + "epoch": 0.74, + "learning_rate": 3.4124819381048337e-06, + "loss": 0.937, + "step": 23047 + }, + { + "epoch": 0.74, + "learning_rate": 3.411702675828983e-06, + "loss": 0.9834, + "step": 23048 + }, + { + "epoch": 0.74, + "learning_rate": 3.4109234842382677e-06, + "loss": 0.8726, + "step": 23049 + }, + { + "epoch": 0.74, + "learning_rate": 3.4101443633410547e-06, + "loss": 0.8853, + "step": 23050 + }, + { + "epoch": 0.74, + "learning_rate": 3.409365313145695e-06, + "loss": 0.8994, + "step": 23051 + }, + { + "epoch": 0.74, + "learning_rate": 3.4085863336605563e-06, + "loss": 0.9077, + "step": 23052 + }, + { + "epoch": 0.74, + "learning_rate": 3.4078074248939895e-06, + "loss": 0.9346, + "step": 23053 + }, + { + "epoch": 0.74, + "learning_rate": 3.407028586854354e-06, + "loss": 0.8491, + "step": 23054 + }, + { + "epoch": 0.74, + "learning_rate": 3.4062498195500027e-06, + "loss": 0.9507, + "step": 23055 + }, + { + "epoch": 0.74, + "learning_rate": 3.4054711229892954e-06, + "loss": 1.0562, + "step": 23056 + }, + { + "epoch": 0.74, + "learning_rate": 3.404692497180582e-06, + "loss": 0.9443, + "step": 23057 + }, + { + "epoch": 0.74, + "learning_rate": 3.403913942132221e-06, + "loss": 0.9175, + "step": 23058 + }, + { + "epoch": 0.74, + "learning_rate": 3.40313545785256e-06, + "loss": 1.0298, + "step": 23059 + }, + { + "epoch": 0.74, + "learning_rate": 3.402357044349962e-06, + "loss": 0.9336, + "step": 23060 + }, + { + "epoch": 0.74, + "learning_rate": 3.401578701632764e-06, + "loss": 0.9839, + "step": 23061 + }, + { + "epoch": 0.74, + "learning_rate": 3.400800429709327e-06, + "loss": 0.9614, + "step": 23062 + }, + { + "epoch": 0.74, + "learning_rate": 3.4000222285879936e-06, + "loss": 1.0508, + "step": 23063 + }, + { + "epoch": 0.74, + "learning_rate": 3.3992440982771213e-06, + "loss": 0.958, + "step": 23064 + }, + { + "epoch": 0.74, + "learning_rate": 3.398466038785051e-06, + "loss": 0.8076, + "step": 23065 + }, + { + "epoch": 0.74, + "learning_rate": 3.3976880501201382e-06, + "loss": 0.7983, + "step": 23066 + }, + { + "epoch": 0.74, + "learning_rate": 3.3969101322907237e-06, + "loss": 0.8696, + "step": 23067 + }, + { + "epoch": 0.74, + "learning_rate": 3.396132285305157e-06, + "loss": 0.9302, + "step": 23068 + }, + { + "epoch": 0.74, + "learning_rate": 3.395354509171779e-06, + "loss": 0.916, + "step": 23069 + }, + { + "epoch": 0.74, + "learning_rate": 3.394576803898939e-06, + "loss": 0.9316, + "step": 23070 + }, + { + "epoch": 0.74, + "learning_rate": 3.393799169494978e-06, + "loss": 0.9072, + "step": 23071 + }, + { + "epoch": 0.74, + "learning_rate": 3.3930216059682432e-06, + "loss": 0.9224, + "step": 23072 + }, + { + "epoch": 0.74, + "learning_rate": 3.3922441133270734e-06, + "loss": 0.5002, + "step": 23073 + }, + { + "epoch": 0.74, + "learning_rate": 3.391466691579809e-06, + "loss": 0.8735, + "step": 23074 + }, + { + "epoch": 0.74, + "learning_rate": 3.3906893407347963e-06, + "loss": 1.0444, + "step": 23075 + }, + { + "epoch": 0.74, + "learning_rate": 3.3899120608003712e-06, + "loss": 0.8516, + "step": 23076 + }, + { + "epoch": 0.74, + "learning_rate": 3.3891348517848745e-06, + "loss": 0.8672, + "step": 23077 + }, + { + "epoch": 0.74, + "learning_rate": 3.3883577136966417e-06, + "loss": 0.8018, + "step": 23078 + }, + { + "epoch": 0.74, + "learning_rate": 3.3875806465440152e-06, + "loss": 0.8457, + "step": 23079 + }, + { + "epoch": 0.74, + "learning_rate": 3.3868036503353275e-06, + "loss": 0.9922, + "step": 23080 + }, + { + "epoch": 0.74, + "learning_rate": 3.3860267250789216e-06, + "loss": 0.8848, + "step": 23081 + }, + { + "epoch": 0.74, + "learning_rate": 3.385249870783124e-06, + "loss": 0.9414, + "step": 23082 + }, + { + "epoch": 0.74, + "learning_rate": 3.3844730874562825e-06, + "loss": 0.9746, + "step": 23083 + }, + { + "epoch": 0.74, + "learning_rate": 3.3836963751067155e-06, + "loss": 0.509, + "step": 23084 + }, + { + "epoch": 0.74, + "learning_rate": 3.3829197337427676e-06, + "loss": 1.0684, + "step": 23085 + }, + { + "epoch": 0.74, + "learning_rate": 3.382143163372764e-06, + "loss": 1.0, + "step": 23086 + }, + { + "epoch": 0.74, + "learning_rate": 3.3813666640050434e-06, + "loss": 0.874, + "step": 23087 + }, + { + "epoch": 0.74, + "learning_rate": 3.3805902356479304e-06, + "loss": 0.9121, + "step": 23088 + }, + { + "epoch": 0.74, + "learning_rate": 3.379813878309761e-06, + "loss": 0.9287, + "step": 23089 + }, + { + "epoch": 0.74, + "learning_rate": 3.3790375919988584e-06, + "loss": 0.9731, + "step": 23090 + }, + { + "epoch": 0.74, + "learning_rate": 3.3782613767235618e-06, + "loss": 0.8003, + "step": 23091 + }, + { + "epoch": 0.74, + "learning_rate": 3.377485232492186e-06, + "loss": 0.9111, + "step": 23092 + }, + { + "epoch": 0.74, + "learning_rate": 3.3767091593130664e-06, + "loss": 0.9595, + "step": 23093 + }, + { + "epoch": 0.74, + "learning_rate": 3.375933157194524e-06, + "loss": 0.8242, + "step": 23094 + }, + { + "epoch": 0.74, + "learning_rate": 3.3751572261448917e-06, + "loss": 0.98, + "step": 23095 + }, + { + "epoch": 0.74, + "learning_rate": 3.3743813661724866e-06, + "loss": 0.9146, + "step": 23096 + }, + { + "epoch": 0.74, + "learning_rate": 3.373605577285639e-06, + "loss": 0.9414, + "step": 23097 + }, + { + "epoch": 0.74, + "learning_rate": 3.3728298594926713e-06, + "loss": 0.9072, + "step": 23098 + }, + { + "epoch": 0.74, + "learning_rate": 3.372054212801904e-06, + "loss": 0.9766, + "step": 23099 + }, + { + "epoch": 0.74, + "learning_rate": 3.371278637221657e-06, + "loss": 0.9844, + "step": 23100 + }, + { + "epoch": 0.74, + "learning_rate": 3.3705031327602554e-06, + "loss": 1.0283, + "step": 23101 + }, + { + "epoch": 0.74, + "learning_rate": 3.3697276994260164e-06, + "loss": 0.9551, + "step": 23102 + }, + { + "epoch": 0.74, + "learning_rate": 3.3689523372272637e-06, + "loss": 0.9287, + "step": 23103 + }, + { + "epoch": 0.74, + "learning_rate": 3.368177046172311e-06, + "loss": 1.0068, + "step": 23104 + }, + { + "epoch": 0.74, + "learning_rate": 3.3674018262694808e-06, + "loss": 0.9785, + "step": 23105 + }, + { + "epoch": 0.74, + "learning_rate": 3.3666266775270893e-06, + "loss": 1.0303, + "step": 23106 + }, + { + "epoch": 0.74, + "learning_rate": 3.3658515999534526e-06, + "loss": 0.897, + "step": 23107 + }, + { + "epoch": 0.74, + "learning_rate": 3.3650765935568817e-06, + "loss": 0.9131, + "step": 23108 + }, + { + "epoch": 0.74, + "learning_rate": 3.3643016583456987e-06, + "loss": 1.0796, + "step": 23109 + }, + { + "epoch": 0.74, + "learning_rate": 3.363526794328212e-06, + "loss": 1.0068, + "step": 23110 + }, + { + "epoch": 0.74, + "learning_rate": 3.3627520015127426e-06, + "loss": 0.9209, + "step": 23111 + }, + { + "epoch": 0.74, + "learning_rate": 3.3619772799075967e-06, + "loss": 0.8896, + "step": 23112 + }, + { + "epoch": 0.74, + "learning_rate": 3.3612026295210885e-06, + "loss": 0.9819, + "step": 23113 + }, + { + "epoch": 0.74, + "learning_rate": 3.360428050361526e-06, + "loss": 0.9097, + "step": 23114 + }, + { + "epoch": 0.74, + "learning_rate": 3.3596535424372245e-06, + "loss": 0.8843, + "step": 23115 + }, + { + "epoch": 0.74, + "learning_rate": 3.358879105756492e-06, + "loss": 1.0239, + "step": 23116 + }, + { + "epoch": 0.74, + "learning_rate": 3.3581047403276335e-06, + "loss": 0.4812, + "step": 23117 + }, + { + "epoch": 0.74, + "learning_rate": 3.357330446158963e-06, + "loss": 0.8564, + "step": 23118 + }, + { + "epoch": 0.74, + "learning_rate": 3.3565562232587824e-06, + "loss": 0.8911, + "step": 23119 + }, + { + "epoch": 0.74, + "learning_rate": 3.3557820716354026e-06, + "loss": 1.0317, + "step": 23120 + }, + { + "epoch": 0.74, + "learning_rate": 3.35500799129713e-06, + "loss": 0.8899, + "step": 23121 + }, + { + "epoch": 0.74, + "learning_rate": 3.354233982252266e-06, + "loss": 0.8232, + "step": 23122 + }, + { + "epoch": 0.74, + "learning_rate": 3.3534600445091126e-06, + "loss": 0.9414, + "step": 23123 + }, + { + "epoch": 0.74, + "learning_rate": 3.352686178075981e-06, + "loss": 0.7764, + "step": 23124 + }, + { + "epoch": 0.74, + "learning_rate": 3.3519123829611667e-06, + "loss": 0.7856, + "step": 23125 + }, + { + "epoch": 0.74, + "learning_rate": 3.351138659172978e-06, + "loss": 0.9121, + "step": 23126 + }, + { + "epoch": 0.74, + "learning_rate": 3.350365006719709e-06, + "loss": 0.8457, + "step": 23127 + }, + { + "epoch": 0.74, + "learning_rate": 3.349591425609671e-06, + "loss": 0.8787, + "step": 23128 + }, + { + "epoch": 0.74, + "learning_rate": 3.348817915851149e-06, + "loss": 0.9902, + "step": 23129 + }, + { + "epoch": 0.74, + "learning_rate": 3.3480444774524543e-06, + "loss": 0.8174, + "step": 23130 + }, + { + "epoch": 0.74, + "learning_rate": 3.3472711104218768e-06, + "loss": 1.0044, + "step": 23131 + }, + { + "epoch": 0.74, + "learning_rate": 3.3464978147677206e-06, + "loss": 0.8892, + "step": 23132 + }, + { + "epoch": 0.74, + "learning_rate": 3.345724590498276e-06, + "loss": 0.8687, + "step": 23133 + }, + { + "epoch": 0.74, + "learning_rate": 3.344951437621845e-06, + "loss": 0.9312, + "step": 23134 + }, + { + "epoch": 0.74, + "learning_rate": 3.3441783561467158e-06, + "loss": 0.7891, + "step": 23135 + }, + { + "epoch": 0.74, + "learning_rate": 3.343405346081194e-06, + "loss": 0.9429, + "step": 23136 + }, + { + "epoch": 0.74, + "learning_rate": 3.3426324074335592e-06, + "loss": 0.9673, + "step": 23137 + }, + { + "epoch": 0.74, + "learning_rate": 3.3418595402121135e-06, + "loss": 0.9517, + "step": 23138 + }, + { + "epoch": 0.74, + "learning_rate": 3.3410867444251426e-06, + "loss": 1.0093, + "step": 23139 + }, + { + "epoch": 0.74, + "learning_rate": 3.3403140200809446e-06, + "loss": 0.7839, + "step": 23140 + }, + { + "epoch": 0.74, + "learning_rate": 3.339541367187803e-06, + "loss": 0.9756, + "step": 23141 + }, + { + "epoch": 0.74, + "learning_rate": 3.338768785754015e-06, + "loss": 0.4919, + "step": 23142 + }, + { + "epoch": 0.74, + "learning_rate": 3.3379962757878624e-06, + "loss": 1.0127, + "step": 23143 + }, + { + "epoch": 0.74, + "learning_rate": 3.3372238372976427e-06, + "loss": 0.8809, + "step": 23144 + }, + { + "epoch": 0.74, + "learning_rate": 3.3364514702916306e-06, + "loss": 0.5312, + "step": 23145 + }, + { + "epoch": 0.74, + "learning_rate": 3.3356791747781236e-06, + "loss": 0.9683, + "step": 23146 + }, + { + "epoch": 0.74, + "learning_rate": 3.3349069507653998e-06, + "loss": 1.0459, + "step": 23147 + }, + { + "epoch": 0.74, + "learning_rate": 3.33413479826175e-06, + "loss": 0.9482, + "step": 23148 + }, + { + "epoch": 0.74, + "learning_rate": 3.3333627172754546e-06, + "loss": 0.9839, + "step": 23149 + }, + { + "epoch": 0.74, + "learning_rate": 3.332590707814801e-06, + "loss": 0.8955, + "step": 23150 + }, + { + "epoch": 0.74, + "learning_rate": 3.331818769888071e-06, + "loss": 0.9336, + "step": 23151 + }, + { + "epoch": 0.74, + "learning_rate": 3.331046903503544e-06, + "loss": 0.8623, + "step": 23152 + }, + { + "epoch": 0.74, + "learning_rate": 3.3302751086695016e-06, + "loss": 0.9224, + "step": 23153 + }, + { + "epoch": 0.74, + "learning_rate": 3.329503385394228e-06, + "loss": 0.9204, + "step": 23154 + }, + { + "epoch": 0.74, + "learning_rate": 3.3287317336860004e-06, + "loss": 0.8135, + "step": 23155 + }, + { + "epoch": 0.74, + "learning_rate": 3.3279601535530946e-06, + "loss": 0.9424, + "step": 23156 + }, + { + "epoch": 0.74, + "learning_rate": 3.327188645003796e-06, + "loss": 1.0181, + "step": 23157 + }, + { + "epoch": 0.74, + "learning_rate": 3.326417208046374e-06, + "loss": 0.9331, + "step": 23158 + }, + { + "epoch": 0.74, + "learning_rate": 3.3256458426891134e-06, + "loss": 0.8989, + "step": 23159 + }, + { + "epoch": 0.74, + "learning_rate": 3.3248745489402855e-06, + "loss": 0.9971, + "step": 23160 + }, + { + "epoch": 0.74, + "learning_rate": 3.324103326808167e-06, + "loss": 0.4944, + "step": 23161 + }, + { + "epoch": 0.74, + "learning_rate": 3.3233321763010273e-06, + "loss": 0.9453, + "step": 23162 + }, + { + "epoch": 0.74, + "learning_rate": 3.3225610974271473e-06, + "loss": 1.04, + "step": 23163 + }, + { + "epoch": 0.74, + "learning_rate": 3.321790090194793e-06, + "loss": 0.47, + "step": 23164 + }, + { + "epoch": 0.74, + "learning_rate": 3.3210191546122437e-06, + "loss": 0.917, + "step": 23165 + }, + { + "epoch": 0.74, + "learning_rate": 3.3202482906877662e-06, + "loss": 0.9165, + "step": 23166 + }, + { + "epoch": 0.74, + "learning_rate": 3.319477498429633e-06, + "loss": 0.957, + "step": 23167 + }, + { + "epoch": 0.74, + "learning_rate": 3.318706777846108e-06, + "loss": 1.0522, + "step": 23168 + }, + { + "epoch": 0.74, + "learning_rate": 3.3179361289454694e-06, + "loss": 0.9546, + "step": 23169 + }, + { + "epoch": 0.74, + "learning_rate": 3.3171655517359758e-06, + "loss": 1.0176, + "step": 23170 + }, + { + "epoch": 0.74, + "learning_rate": 3.3163950462259043e-06, + "loss": 0.96, + "step": 23171 + }, + { + "epoch": 0.74, + "learning_rate": 3.3156246124235135e-06, + "loss": 0.8594, + "step": 23172 + }, + { + "epoch": 0.74, + "learning_rate": 3.314854250337075e-06, + "loss": 0.9634, + "step": 23173 + }, + { + "epoch": 0.74, + "learning_rate": 3.3140839599748533e-06, + "loss": 0.8574, + "step": 23174 + }, + { + "epoch": 0.74, + "learning_rate": 3.31331374134511e-06, + "loss": 1.0024, + "step": 23175 + }, + { + "epoch": 0.74, + "learning_rate": 3.3125435944561056e-06, + "loss": 1.022, + "step": 23176 + }, + { + "epoch": 0.74, + "learning_rate": 3.3117735193161114e-06, + "loss": 0.9214, + "step": 23177 + }, + { + "epoch": 0.74, + "learning_rate": 3.311003515933383e-06, + "loss": 0.8999, + "step": 23178 + }, + { + "epoch": 0.74, + "learning_rate": 3.310233584316186e-06, + "loss": 1.0894, + "step": 23179 + }, + { + "epoch": 0.74, + "learning_rate": 3.3094637244727756e-06, + "loss": 0.8433, + "step": 23180 + }, + { + "epoch": 0.74, + "learning_rate": 3.308693936411421e-06, + "loss": 0.9331, + "step": 23181 + }, + { + "epoch": 0.74, + "learning_rate": 3.3079242201403685e-06, + "loss": 0.8491, + "step": 23182 + }, + { + "epoch": 0.74, + "learning_rate": 3.307154575667886e-06, + "loss": 0.9009, + "step": 23183 + }, + { + "epoch": 0.74, + "learning_rate": 3.306385003002225e-06, + "loss": 0.835, + "step": 23184 + }, + { + "epoch": 0.74, + "learning_rate": 3.3056155021516467e-06, + "loss": 0.981, + "step": 23185 + }, + { + "epoch": 0.74, + "learning_rate": 3.3048460731244027e-06, + "loss": 1.0156, + "step": 23186 + }, + { + "epoch": 0.74, + "learning_rate": 3.3040767159287536e-06, + "loss": 0.9639, + "step": 23187 + }, + { + "epoch": 0.74, + "learning_rate": 3.3033074305729473e-06, + "loss": 1.0361, + "step": 23188 + }, + { + "epoch": 0.74, + "learning_rate": 3.3025382170652476e-06, + "loss": 0.8276, + "step": 23189 + }, + { + "epoch": 0.74, + "learning_rate": 3.3017690754138942e-06, + "loss": 0.8472, + "step": 23190 + }, + { + "epoch": 0.74, + "learning_rate": 3.3010000056271484e-06, + "loss": 1.0176, + "step": 23191 + }, + { + "epoch": 0.74, + "learning_rate": 3.300231007713255e-06, + "loss": 0.4895, + "step": 23192 + }, + { + "epoch": 0.74, + "learning_rate": 3.299462081680471e-06, + "loss": 0.998, + "step": 23193 + }, + { + "epoch": 0.74, + "learning_rate": 3.298693227537043e-06, + "loss": 0.8833, + "step": 23194 + }, + { + "epoch": 0.74, + "learning_rate": 3.2979244452912174e-06, + "loss": 0.8975, + "step": 23195 + }, + { + "epoch": 0.74, + "learning_rate": 3.2971557349512486e-06, + "loss": 0.8457, + "step": 23196 + }, + { + "epoch": 0.74, + "learning_rate": 3.2963870965253786e-06, + "loss": 1.0063, + "step": 23197 + }, + { + "epoch": 0.74, + "learning_rate": 3.2956185300218546e-06, + "loss": 0.981, + "step": 23198 + }, + { + "epoch": 0.74, + "learning_rate": 3.2948500354489255e-06, + "loss": 0.894, + "step": 23199 + }, + { + "epoch": 0.74, + "learning_rate": 3.2940816128148357e-06, + "loss": 1.0825, + "step": 23200 + }, + { + "epoch": 0.74, + "learning_rate": 3.293313262127824e-06, + "loss": 0.7188, + "step": 23201 + }, + { + "epoch": 0.74, + "learning_rate": 3.292544983396143e-06, + "loss": 0.9858, + "step": 23202 + }, + { + "epoch": 0.74, + "learning_rate": 3.2917767766280275e-06, + "loss": 0.8271, + "step": 23203 + }, + { + "epoch": 0.74, + "learning_rate": 3.2910086418317256e-06, + "loss": 1.0039, + "step": 23204 + }, + { + "epoch": 0.74, + "learning_rate": 3.2902405790154755e-06, + "loss": 0.9233, + "step": 23205 + }, + { + "epoch": 0.74, + "learning_rate": 3.289472588187519e-06, + "loss": 0.8613, + "step": 23206 + }, + { + "epoch": 0.74, + "learning_rate": 3.288704669356091e-06, + "loss": 0.8755, + "step": 23207 + }, + { + "epoch": 0.74, + "learning_rate": 3.2879368225294374e-06, + "loss": 0.916, + "step": 23208 + }, + { + "epoch": 0.74, + "learning_rate": 3.287169047715789e-06, + "loss": 0.9683, + "step": 23209 + }, + { + "epoch": 0.74, + "learning_rate": 3.2864013449233923e-06, + "loss": 0.8535, + "step": 23210 + }, + { + "epoch": 0.74, + "learning_rate": 3.2856337141604776e-06, + "loss": 0.9023, + "step": 23211 + }, + { + "epoch": 0.74, + "learning_rate": 3.2848661554352824e-06, + "loss": 1.0615, + "step": 23212 + }, + { + "epoch": 0.74, + "learning_rate": 3.284098668756038e-06, + "loss": 1.0132, + "step": 23213 + }, + { + "epoch": 0.74, + "learning_rate": 3.283331254130987e-06, + "loss": 0.8667, + "step": 23214 + }, + { + "epoch": 0.74, + "learning_rate": 3.282563911568353e-06, + "loss": 1.0117, + "step": 23215 + }, + { + "epoch": 0.74, + "learning_rate": 3.281796641076377e-06, + "loss": 0.9707, + "step": 23216 + }, + { + "epoch": 0.74, + "learning_rate": 3.281029442663285e-06, + "loss": 0.8926, + "step": 23217 + }, + { + "epoch": 0.74, + "learning_rate": 3.2802623163373138e-06, + "loss": 0.7686, + "step": 23218 + }, + { + "epoch": 0.74, + "learning_rate": 3.279495262106691e-06, + "loss": 0.4521, + "step": 23219 + }, + { + "epoch": 0.74, + "learning_rate": 3.2787282799796472e-06, + "loss": 1.082, + "step": 23220 + }, + { + "epoch": 0.74, + "learning_rate": 3.277961369964405e-06, + "loss": 0.9775, + "step": 23221 + }, + { + "epoch": 0.74, + "learning_rate": 3.2771945320692024e-06, + "loss": 0.8413, + "step": 23222 + }, + { + "epoch": 0.74, + "learning_rate": 3.276427766302258e-06, + "loss": 0.9756, + "step": 23223 + }, + { + "epoch": 0.74, + "learning_rate": 3.2756610726718065e-06, + "loss": 0.8745, + "step": 23224 + }, + { + "epoch": 0.74, + "learning_rate": 3.2748944511860657e-06, + "loss": 0.7903, + "step": 23225 + }, + { + "epoch": 0.74, + "learning_rate": 3.274127901853268e-06, + "loss": 0.832, + "step": 23226 + }, + { + "epoch": 0.74, + "learning_rate": 3.2733614246816338e-06, + "loss": 0.9678, + "step": 23227 + }, + { + "epoch": 0.74, + "learning_rate": 3.272595019679388e-06, + "loss": 0.8418, + "step": 23228 + }, + { + "epoch": 0.74, + "learning_rate": 3.2718286868547467e-06, + "loss": 0.8965, + "step": 23229 + }, + { + "epoch": 0.74, + "learning_rate": 3.271062426215942e-06, + "loss": 0.9287, + "step": 23230 + }, + { + "epoch": 0.74, + "learning_rate": 3.2702962377711867e-06, + "loss": 0.9141, + "step": 23231 + }, + { + "epoch": 0.74, + "learning_rate": 3.269530121528708e-06, + "loss": 0.9487, + "step": 23232 + }, + { + "epoch": 0.74, + "learning_rate": 3.268764077496722e-06, + "loss": 0.9492, + "step": 23233 + }, + { + "epoch": 0.74, + "learning_rate": 3.2679981056834477e-06, + "loss": 0.9019, + "step": 23234 + }, + { + "epoch": 0.74, + "learning_rate": 3.2672322060970996e-06, + "loss": 0.875, + "step": 23235 + }, + { + "epoch": 0.74, + "learning_rate": 3.2664663787459017e-06, + "loss": 0.8877, + "step": 23236 + }, + { + "epoch": 0.74, + "learning_rate": 3.265700623638065e-06, + "loss": 0.916, + "step": 23237 + }, + { + "epoch": 0.74, + "learning_rate": 3.2649349407818097e-06, + "loss": 0.9077, + "step": 23238 + }, + { + "epoch": 0.74, + "learning_rate": 3.2641693301853494e-06, + "loss": 0.9185, + "step": 23239 + }, + { + "epoch": 0.74, + "learning_rate": 3.2634037918568927e-06, + "loss": 0.8311, + "step": 23240 + }, + { + "epoch": 0.74, + "learning_rate": 3.262638325804662e-06, + "loss": 0.917, + "step": 23241 + }, + { + "epoch": 0.74, + "learning_rate": 3.2618729320368647e-06, + "loss": 0.8809, + "step": 23242 + }, + { + "epoch": 0.74, + "learning_rate": 3.261107610561711e-06, + "loss": 0.9263, + "step": 23243 + }, + { + "epoch": 0.74, + "learning_rate": 3.2603423613874174e-06, + "loss": 0.915, + "step": 23244 + }, + { + "epoch": 0.74, + "learning_rate": 3.2595771845221915e-06, + "loss": 0.915, + "step": 23245 + }, + { + "epoch": 0.74, + "learning_rate": 3.2588120799742386e-06, + "loss": 0.8669, + "step": 23246 + }, + { + "epoch": 0.74, + "learning_rate": 3.2580470477517744e-06, + "loss": 0.845, + "step": 23247 + }, + { + "epoch": 0.74, + "learning_rate": 3.2572820878630017e-06, + "loss": 0.9355, + "step": 23248 + }, + { + "epoch": 0.74, + "learning_rate": 3.256517200316135e-06, + "loss": 0.9731, + "step": 23249 + }, + { + "epoch": 0.74, + "learning_rate": 3.2557523851193696e-06, + "loss": 0.9551, + "step": 23250 + }, + { + "epoch": 0.74, + "learning_rate": 3.25498764228092e-06, + "loss": 0.9448, + "step": 23251 + }, + { + "epoch": 0.74, + "learning_rate": 3.2542229718089848e-06, + "loss": 0.5164, + "step": 23252 + }, + { + "epoch": 0.74, + "learning_rate": 3.253458373711774e-06, + "loss": 0.8911, + "step": 23253 + }, + { + "epoch": 0.74, + "learning_rate": 3.2526938479974855e-06, + "loss": 0.9336, + "step": 23254 + }, + { + "epoch": 0.74, + "learning_rate": 3.251929394674327e-06, + "loss": 0.8589, + "step": 23255 + }, + { + "epoch": 0.74, + "learning_rate": 3.2511650137504957e-06, + "loss": 0.9355, + "step": 23256 + }, + { + "epoch": 0.74, + "learning_rate": 3.2504007052342e-06, + "loss": 0.9092, + "step": 23257 + }, + { + "epoch": 0.74, + "learning_rate": 3.249636469133628e-06, + "loss": 0.8413, + "step": 23258 + }, + { + "epoch": 0.74, + "learning_rate": 3.2488723054569905e-06, + "loss": 0.8604, + "step": 23259 + }, + { + "epoch": 0.74, + "learning_rate": 3.2481082142124763e-06, + "loss": 1.0737, + "step": 23260 + }, + { + "epoch": 0.74, + "learning_rate": 3.247344195408293e-06, + "loss": 1.0361, + "step": 23261 + }, + { + "epoch": 0.74, + "learning_rate": 3.246580249052629e-06, + "loss": 0.9319, + "step": 23262 + }, + { + "epoch": 0.74, + "learning_rate": 3.2458163751536875e-06, + "loss": 0.9009, + "step": 23263 + }, + { + "epoch": 0.74, + "learning_rate": 3.2450525737196615e-06, + "loss": 1.0059, + "step": 23264 + }, + { + "epoch": 0.74, + "learning_rate": 3.244288844758746e-06, + "loss": 0.874, + "step": 23265 + }, + { + "epoch": 0.74, + "learning_rate": 3.2435251882791307e-06, + "loss": 1.0488, + "step": 23266 + }, + { + "epoch": 0.74, + "learning_rate": 3.2427616042890154e-06, + "loss": 1.0205, + "step": 23267 + }, + { + "epoch": 0.74, + "learning_rate": 3.2419980927965865e-06, + "loss": 0.9053, + "step": 23268 + }, + { + "epoch": 0.74, + "learning_rate": 3.2412346538100415e-06, + "loss": 0.9097, + "step": 23269 + }, + { + "epoch": 0.74, + "learning_rate": 3.2404712873375666e-06, + "loss": 0.8833, + "step": 23270 + }, + { + "epoch": 0.74, + "learning_rate": 3.2397079933873555e-06, + "loss": 0.853, + "step": 23271 + }, + { + "epoch": 0.74, + "learning_rate": 3.238944771967596e-06, + "loss": 0.8701, + "step": 23272 + }, + { + "epoch": 0.74, + "learning_rate": 3.238181623086477e-06, + "loss": 0.9756, + "step": 23273 + }, + { + "epoch": 0.74, + "learning_rate": 3.237418546752181e-06, + "loss": 1.0532, + "step": 23274 + }, + { + "epoch": 0.74, + "learning_rate": 3.236655542972904e-06, + "loss": 0.8662, + "step": 23275 + }, + { + "epoch": 0.74, + "learning_rate": 3.2358926117568246e-06, + "loss": 0.8865, + "step": 23276 + }, + { + "epoch": 0.74, + "learning_rate": 3.235129753112134e-06, + "loss": 0.8682, + "step": 23277 + }, + { + "epoch": 0.74, + "learning_rate": 3.2343669670470133e-06, + "loss": 0.9712, + "step": 23278 + }, + { + "epoch": 0.74, + "learning_rate": 3.2336042535696486e-06, + "loss": 0.9404, + "step": 23279 + }, + { + "epoch": 0.74, + "learning_rate": 3.232841612688218e-06, + "loss": 0.9434, + "step": 23280 + }, + { + "epoch": 0.74, + "learning_rate": 3.2320790444109107e-06, + "loss": 0.979, + "step": 23281 + }, + { + "epoch": 0.74, + "learning_rate": 3.231316548745902e-06, + "loss": 0.8848, + "step": 23282 + }, + { + "epoch": 0.74, + "learning_rate": 3.230554125701377e-06, + "loss": 0.9673, + "step": 23283 + }, + { + "epoch": 0.74, + "learning_rate": 3.2297917752855156e-06, + "loss": 0.894, + "step": 23284 + }, + { + "epoch": 0.74, + "learning_rate": 3.2290294975064928e-06, + "loss": 1.0132, + "step": 23285 + }, + { + "epoch": 0.74, + "learning_rate": 3.228267292372492e-06, + "loss": 0.9346, + "step": 23286 + }, + { + "epoch": 0.74, + "learning_rate": 3.2275051598916896e-06, + "loss": 0.7783, + "step": 23287 + }, + { + "epoch": 0.74, + "learning_rate": 3.226743100072257e-06, + "loss": 0.8682, + "step": 23288 + }, + { + "epoch": 0.74, + "learning_rate": 3.2259811129223784e-06, + "loss": 0.9307, + "step": 23289 + }, + { + "epoch": 0.74, + "learning_rate": 3.2252191984502256e-06, + "loss": 0.8945, + "step": 23290 + }, + { + "epoch": 0.74, + "learning_rate": 3.2244573566639693e-06, + "loss": 0.936, + "step": 23291 + }, + { + "epoch": 0.74, + "learning_rate": 3.22369558757179e-06, + "loss": 0.8828, + "step": 23292 + }, + { + "epoch": 0.74, + "learning_rate": 3.222933891181854e-06, + "loss": 0.9961, + "step": 23293 + }, + { + "epoch": 0.74, + "learning_rate": 3.2221722675023436e-06, + "loss": 1.0034, + "step": 23294 + }, + { + "epoch": 0.74, + "learning_rate": 3.2214107165414164e-06, + "loss": 0.8555, + "step": 23295 + }, + { + "epoch": 0.75, + "learning_rate": 3.2206492383072542e-06, + "loss": 1.0073, + "step": 23296 + }, + { + "epoch": 0.75, + "learning_rate": 3.2198878328080184e-06, + "loss": 0.7847, + "step": 23297 + }, + { + "epoch": 0.75, + "learning_rate": 3.219126500051887e-06, + "loss": 0.7725, + "step": 23298 + }, + { + "epoch": 0.75, + "learning_rate": 3.218365240047019e-06, + "loss": 0.8877, + "step": 23299 + }, + { + "epoch": 0.75, + "learning_rate": 3.2176040528015907e-06, + "loss": 0.9058, + "step": 23300 + }, + { + "epoch": 0.75, + "learning_rate": 3.2168429383237597e-06, + "loss": 1.0615, + "step": 23301 + }, + { + "epoch": 0.75, + "learning_rate": 3.2160818966217044e-06, + "loss": 0.9756, + "step": 23302 + }, + { + "epoch": 0.75, + "learning_rate": 3.215320927703576e-06, + "loss": 0.874, + "step": 23303 + }, + { + "epoch": 0.75, + "learning_rate": 3.214560031577548e-06, + "loss": 0.9189, + "step": 23304 + }, + { + "epoch": 0.75, + "learning_rate": 3.2137992082517777e-06, + "loss": 0.7129, + "step": 23305 + }, + { + "epoch": 0.75, + "learning_rate": 3.2130384577344344e-06, + "loss": 0.8662, + "step": 23306 + }, + { + "epoch": 0.75, + "learning_rate": 3.2122777800336747e-06, + "loss": 0.8823, + "step": 23307 + }, + { + "epoch": 0.75, + "learning_rate": 3.2115171751576644e-06, + "loss": 0.916, + "step": 23308 + }, + { + "epoch": 0.75, + "learning_rate": 3.210756643114559e-06, + "loss": 0.9941, + "step": 23309 + }, + { + "epoch": 0.75, + "learning_rate": 3.2099961839125272e-06, + "loss": 0.8931, + "step": 23310 + }, + { + "epoch": 0.75, + "learning_rate": 3.2092357975597146e-06, + "loss": 0.9556, + "step": 23311 + }, + { + "epoch": 0.75, + "learning_rate": 3.2084754840642905e-06, + "loss": 0.876, + "step": 23312 + }, + { + "epoch": 0.75, + "learning_rate": 3.207715243434404e-06, + "loss": 0.9434, + "step": 23313 + }, + { + "epoch": 0.75, + "learning_rate": 3.2069550756782186e-06, + "loss": 0.9912, + "step": 23314 + }, + { + "epoch": 0.75, + "learning_rate": 3.2061949808038837e-06, + "loss": 0.9072, + "step": 23315 + }, + { + "epoch": 0.75, + "learning_rate": 3.205434958819562e-06, + "loss": 0.9512, + "step": 23316 + }, + { + "epoch": 0.75, + "learning_rate": 3.204675009733402e-06, + "loss": 0.9995, + "step": 23317 + }, + { + "epoch": 0.75, + "learning_rate": 3.2039151335535603e-06, + "loss": 0.7983, + "step": 23318 + }, + { + "epoch": 0.75, + "learning_rate": 3.2031553302881835e-06, + "loss": 0.9282, + "step": 23319 + }, + { + "epoch": 0.75, + "learning_rate": 3.2023955999454305e-06, + "loss": 1.1479, + "step": 23320 + }, + { + "epoch": 0.75, + "learning_rate": 3.2016359425334467e-06, + "loss": 0.9734, + "step": 23321 + }, + { + "epoch": 0.75, + "learning_rate": 3.2008763580603872e-06, + "loss": 0.9365, + "step": 23322 + }, + { + "epoch": 0.75, + "learning_rate": 3.200116846534401e-06, + "loss": 0.9011, + "step": 23323 + }, + { + "epoch": 0.75, + "learning_rate": 3.1993574079636313e-06, + "loss": 1.021, + "step": 23324 + }, + { + "epoch": 0.75, + "learning_rate": 3.1985980423562333e-06, + "loss": 0.877, + "step": 23325 + }, + { + "epoch": 0.75, + "learning_rate": 3.197838749720351e-06, + "loss": 0.8535, + "step": 23326 + }, + { + "epoch": 0.75, + "learning_rate": 3.197079530064128e-06, + "loss": 0.9629, + "step": 23327 + }, + { + "epoch": 0.75, + "learning_rate": 3.1963203833957157e-06, + "loss": 0.8721, + "step": 23328 + }, + { + "epoch": 0.75, + "learning_rate": 3.1955613097232553e-06, + "loss": 1.0044, + "step": 23329 + }, + { + "epoch": 0.75, + "learning_rate": 3.1948023090548874e-06, + "loss": 0.8623, + "step": 23330 + }, + { + "epoch": 0.75, + "learning_rate": 3.1940433813987636e-06, + "loss": 0.8418, + "step": 23331 + }, + { + "epoch": 0.75, + "learning_rate": 3.193284526763021e-06, + "loss": 0.8589, + "step": 23332 + }, + { + "epoch": 0.75, + "learning_rate": 3.1925257451558024e-06, + "loss": 0.8027, + "step": 23333 + }, + { + "epoch": 0.75, + "learning_rate": 3.191767036585245e-06, + "loss": 0.8882, + "step": 23334 + }, + { + "epoch": 0.75, + "learning_rate": 3.1910084010594955e-06, + "loss": 1.0332, + "step": 23335 + }, + { + "epoch": 0.75, + "learning_rate": 3.1902498385866878e-06, + "loss": 0.9326, + "step": 23336 + }, + { + "epoch": 0.75, + "learning_rate": 3.189491349174965e-06, + "loss": 0.8945, + "step": 23337 + }, + { + "epoch": 0.75, + "learning_rate": 3.188732932832459e-06, + "loss": 0.9487, + "step": 23338 + }, + { + "epoch": 0.75, + "learning_rate": 3.187974589567314e-06, + "loss": 0.9741, + "step": 23339 + }, + { + "epoch": 0.75, + "learning_rate": 3.1872163193876626e-06, + "loss": 0.9551, + "step": 23340 + }, + { + "epoch": 0.75, + "learning_rate": 3.18645812230164e-06, + "loss": 0.8433, + "step": 23341 + }, + { + "epoch": 0.75, + "learning_rate": 3.1856999983173774e-06, + "loss": 0.8115, + "step": 23342 + }, + { + "epoch": 0.75, + "learning_rate": 3.1849419474430156e-06, + "loss": 0.9316, + "step": 23343 + }, + { + "epoch": 0.75, + "learning_rate": 3.1841839696866804e-06, + "loss": 0.9175, + "step": 23344 + }, + { + "epoch": 0.75, + "learning_rate": 3.183426065056512e-06, + "loss": 0.915, + "step": 23345 + }, + { + "epoch": 0.75, + "learning_rate": 3.182668233560634e-06, + "loss": 0.8228, + "step": 23346 + }, + { + "epoch": 0.75, + "learning_rate": 3.181910475207187e-06, + "loss": 0.9272, + "step": 23347 + }, + { + "epoch": 0.75, + "learning_rate": 3.1811527900042883e-06, + "loss": 0.9028, + "step": 23348 + }, + { + "epoch": 0.75, + "learning_rate": 3.1803951779600774e-06, + "loss": 1.0557, + "step": 23349 + }, + { + "epoch": 0.75, + "learning_rate": 3.1796376390826745e-06, + "loss": 0.876, + "step": 23350 + }, + { + "epoch": 0.75, + "learning_rate": 3.1788801733802154e-06, + "loss": 0.9575, + "step": 23351 + }, + { + "epoch": 0.75, + "learning_rate": 3.1781227808608186e-06, + "loss": 1.0518, + "step": 23352 + }, + { + "epoch": 0.75, + "learning_rate": 3.177365461532619e-06, + "loss": 0.9189, + "step": 23353 + }, + { + "epoch": 0.75, + "learning_rate": 3.1766082154037324e-06, + "loss": 1.1025, + "step": 23354 + }, + { + "epoch": 0.75, + "learning_rate": 3.1758510424822942e-06, + "loss": 0.8975, + "step": 23355 + }, + { + "epoch": 0.75, + "learning_rate": 3.175093942776416e-06, + "loss": 0.9771, + "step": 23356 + }, + { + "epoch": 0.75, + "learning_rate": 3.1743369162942295e-06, + "loss": 0.8701, + "step": 23357 + }, + { + "epoch": 0.75, + "learning_rate": 3.173579963043851e-06, + "loss": 0.9922, + "step": 23358 + }, + { + "epoch": 0.75, + "learning_rate": 3.1728230830334073e-06, + "loss": 0.9043, + "step": 23359 + }, + { + "epoch": 0.75, + "learning_rate": 3.1720662762710117e-06, + "loss": 0.9917, + "step": 23360 + }, + { + "epoch": 0.75, + "learning_rate": 3.171309542764792e-06, + "loss": 0.9668, + "step": 23361 + }, + { + "epoch": 0.75, + "learning_rate": 3.170552882522864e-06, + "loss": 0.8604, + "step": 23362 + }, + { + "epoch": 0.75, + "learning_rate": 3.1697962955533445e-06, + "loss": 0.8203, + "step": 23363 + }, + { + "epoch": 0.75, + "learning_rate": 3.1690397818643483e-06, + "loss": 0.9478, + "step": 23364 + }, + { + "epoch": 0.75, + "learning_rate": 3.168283341463998e-06, + "loss": 0.9531, + "step": 23365 + }, + { + "epoch": 0.75, + "learning_rate": 3.1675269743604044e-06, + "loss": 0.9448, + "step": 23366 + }, + { + "epoch": 0.75, + "learning_rate": 3.1667706805616873e-06, + "loss": 0.9331, + "step": 23367 + }, + { + "epoch": 0.75, + "learning_rate": 3.1660144600759566e-06, + "loss": 0.5032, + "step": 23368 + }, + { + "epoch": 0.75, + "learning_rate": 3.1652583129113257e-06, + "loss": 1.0684, + "step": 23369 + }, + { + "epoch": 0.75, + "learning_rate": 3.1645022390759116e-06, + "loss": 0.8936, + "step": 23370 + }, + { + "epoch": 0.75, + "learning_rate": 3.163746238577824e-06, + "loss": 1.1191, + "step": 23371 + }, + { + "epoch": 0.75, + "learning_rate": 3.1629903114251726e-06, + "loss": 0.9087, + "step": 23372 + }, + { + "epoch": 0.75, + "learning_rate": 3.1622344576260644e-06, + "loss": 0.9707, + "step": 23373 + }, + { + "epoch": 0.75, + "learning_rate": 3.1614786771886162e-06, + "loss": 0.9849, + "step": 23374 + }, + { + "epoch": 0.75, + "learning_rate": 3.1607229701209307e-06, + "loss": 0.8975, + "step": 23375 + }, + { + "epoch": 0.75, + "learning_rate": 3.1599673364311223e-06, + "loss": 1.0566, + "step": 23376 + }, + { + "epoch": 0.75, + "learning_rate": 3.15921177612729e-06, + "loss": 0.8706, + "step": 23377 + }, + { + "epoch": 0.75, + "learning_rate": 3.1584562892175507e-06, + "loss": 0.9771, + "step": 23378 + }, + { + "epoch": 0.75, + "learning_rate": 3.1577008757099967e-06, + "loss": 0.8906, + "step": 23379 + }, + { + "epoch": 0.75, + "learning_rate": 3.1569455356127444e-06, + "loss": 0.9312, + "step": 23380 + }, + { + "epoch": 0.75, + "learning_rate": 3.1561902689338888e-06, + "loss": 0.9243, + "step": 23381 + }, + { + "epoch": 0.75, + "learning_rate": 3.155435075681541e-06, + "loss": 0.8804, + "step": 23382 + }, + { + "epoch": 0.75, + "learning_rate": 3.154679955863795e-06, + "loss": 0.9004, + "step": 23383 + }, + { + "epoch": 0.75, + "learning_rate": 3.1539249094887613e-06, + "loss": 0.7795, + "step": 23384 + }, + { + "epoch": 0.75, + "learning_rate": 3.1531699365645372e-06, + "loss": 0.8647, + "step": 23385 + }, + { + "epoch": 0.75, + "learning_rate": 3.1524150370992213e-06, + "loss": 0.8433, + "step": 23386 + }, + { + "epoch": 0.75, + "learning_rate": 3.15166021110091e-06, + "loss": 0.8809, + "step": 23387 + }, + { + "epoch": 0.75, + "learning_rate": 3.150905458577709e-06, + "loss": 0.811, + "step": 23388 + }, + { + "epoch": 0.75, + "learning_rate": 3.1501507795377086e-06, + "loss": 1.0142, + "step": 23389 + }, + { + "epoch": 0.75, + "learning_rate": 3.149396173989012e-06, + "loss": 0.8701, + "step": 23390 + }, + { + "epoch": 0.75, + "learning_rate": 3.14864164193971e-06, + "loss": 0.9507, + "step": 23391 + }, + { + "epoch": 0.75, + "learning_rate": 3.147887183397904e-06, + "loss": 0.9097, + "step": 23392 + }, + { + "epoch": 0.75, + "learning_rate": 3.147132798371685e-06, + "loss": 1.0254, + "step": 23393 + }, + { + "epoch": 0.75, + "learning_rate": 3.146378486869146e-06, + "loss": 0.9229, + "step": 23394 + }, + { + "epoch": 0.75, + "learning_rate": 3.145624248898378e-06, + "loss": 0.8608, + "step": 23395 + }, + { + "epoch": 0.75, + "learning_rate": 3.1448700844674784e-06, + "loss": 1.04, + "step": 23396 + }, + { + "epoch": 0.75, + "learning_rate": 3.144115993584533e-06, + "loss": 0.873, + "step": 23397 + }, + { + "epoch": 0.75, + "learning_rate": 3.1433619762576386e-06, + "loss": 0.9424, + "step": 23398 + }, + { + "epoch": 0.75, + "learning_rate": 3.1426080324948783e-06, + "loss": 0.8467, + "step": 23399 + }, + { + "epoch": 0.75, + "learning_rate": 3.1418541623043507e-06, + "loss": 0.9517, + "step": 23400 + }, + { + "epoch": 0.75, + "learning_rate": 3.141100365694132e-06, + "loss": 0.9463, + "step": 23401 + }, + { + "epoch": 0.75, + "learning_rate": 3.1403466426723182e-06, + "loss": 0.9048, + "step": 23402 + }, + { + "epoch": 0.75, + "learning_rate": 3.139592993246989e-06, + "loss": 0.8457, + "step": 23403 + }, + { + "epoch": 0.75, + "learning_rate": 3.138839417426238e-06, + "loss": 0.9751, + "step": 23404 + }, + { + "epoch": 0.75, + "learning_rate": 3.1380859152181444e-06, + "loss": 0.9756, + "step": 23405 + }, + { + "epoch": 0.75, + "learning_rate": 3.1373324866307964e-06, + "loss": 1.0591, + "step": 23406 + }, + { + "epoch": 0.75, + "learning_rate": 3.136579131672276e-06, + "loss": 0.8652, + "step": 23407 + }, + { + "epoch": 0.75, + "learning_rate": 3.1358258503506654e-06, + "loss": 0.834, + "step": 23408 + }, + { + "epoch": 0.75, + "learning_rate": 3.135072642674043e-06, + "loss": 0.8955, + "step": 23409 + }, + { + "epoch": 0.75, + "learning_rate": 3.1343195086504974e-06, + "loss": 0.9912, + "step": 23410 + }, + { + "epoch": 0.75, + "learning_rate": 3.1335664482881045e-06, + "loss": 0.8228, + "step": 23411 + }, + { + "epoch": 0.75, + "learning_rate": 3.13281346159494e-06, + "loss": 0.8892, + "step": 23412 + }, + { + "epoch": 0.75, + "learning_rate": 3.132060548579091e-06, + "loss": 0.8843, + "step": 23413 + }, + { + "epoch": 0.75, + "learning_rate": 3.1313077092486286e-06, + "loss": 0.8721, + "step": 23414 + }, + { + "epoch": 0.75, + "learning_rate": 3.1305549436116356e-06, + "loss": 0.9829, + "step": 23415 + }, + { + "epoch": 0.75, + "learning_rate": 3.129802251676185e-06, + "loss": 1.1196, + "step": 23416 + }, + { + "epoch": 0.75, + "learning_rate": 3.129049633450353e-06, + "loss": 1.0708, + "step": 23417 + }, + { + "epoch": 0.75, + "learning_rate": 3.1282970889422104e-06, + "loss": 0.918, + "step": 23418 + }, + { + "epoch": 0.75, + "learning_rate": 3.127544618159838e-06, + "loss": 0.9277, + "step": 23419 + }, + { + "epoch": 0.75, + "learning_rate": 3.126792221111302e-06, + "loss": 0.9785, + "step": 23420 + }, + { + "epoch": 0.75, + "learning_rate": 3.1260398978046834e-06, + "loss": 0.937, + "step": 23421 + }, + { + "epoch": 0.75, + "learning_rate": 3.1252876482480443e-06, + "loss": 0.8979, + "step": 23422 + }, + { + "epoch": 0.75, + "learning_rate": 3.124535472449467e-06, + "loss": 0.8335, + "step": 23423 + }, + { + "epoch": 0.75, + "learning_rate": 3.1237833704170084e-06, + "loss": 1.0635, + "step": 23424 + }, + { + "epoch": 0.75, + "learning_rate": 3.1230313421587475e-06, + "loss": 1.0029, + "step": 23425 + }, + { + "epoch": 0.75, + "learning_rate": 3.1222793876827463e-06, + "loss": 0.8711, + "step": 23426 + }, + { + "epoch": 0.75, + "learning_rate": 3.1215275069970773e-06, + "loss": 0.8643, + "step": 23427 + }, + { + "epoch": 0.75, + "learning_rate": 3.1207757001098036e-06, + "loss": 0.8152, + "step": 23428 + }, + { + "epoch": 0.75, + "learning_rate": 3.1200239670289956e-06, + "loss": 0.9004, + "step": 23429 + }, + { + "epoch": 0.75, + "learning_rate": 3.1192723077627163e-06, + "loss": 1.0581, + "step": 23430 + }, + { + "epoch": 0.75, + "learning_rate": 3.1185207223190293e-06, + "loss": 0.8516, + "step": 23431 + }, + { + "epoch": 0.75, + "learning_rate": 3.1177692107059964e-06, + "loss": 0.9375, + "step": 23432 + }, + { + "epoch": 0.75, + "learning_rate": 3.1170177729316863e-06, + "loss": 1.0581, + "step": 23433 + }, + { + "epoch": 0.75, + "learning_rate": 3.116266409004154e-06, + "loss": 0.9849, + "step": 23434 + }, + { + "epoch": 0.75, + "learning_rate": 3.1155151189314682e-06, + "loss": 0.9175, + "step": 23435 + }, + { + "epoch": 0.75, + "learning_rate": 3.1147639027216826e-06, + "loss": 0.9429, + "step": 23436 + }, + { + "epoch": 0.75, + "learning_rate": 3.114012760382863e-06, + "loss": 0.9077, + "step": 23437 + }, + { + "epoch": 0.75, + "learning_rate": 3.1132616919230664e-06, + "loss": 0.9233, + "step": 23438 + }, + { + "epoch": 0.75, + "learning_rate": 3.112510697350348e-06, + "loss": 0.9258, + "step": 23439 + }, + { + "epoch": 0.75, + "learning_rate": 3.111759776672766e-06, + "loss": 0.9976, + "step": 23440 + }, + { + "epoch": 0.75, + "learning_rate": 3.11100892989838e-06, + "loss": 0.8442, + "step": 23441 + }, + { + "epoch": 0.75, + "learning_rate": 3.1102581570352406e-06, + "loss": 0.8555, + "step": 23442 + }, + { + "epoch": 0.75, + "learning_rate": 3.1095074580914107e-06, + "loss": 0.853, + "step": 23443 + }, + { + "epoch": 0.75, + "learning_rate": 3.108756833074935e-06, + "loss": 0.9878, + "step": 23444 + }, + { + "epoch": 0.75, + "learning_rate": 3.1080062819938783e-06, + "loss": 0.9268, + "step": 23445 + }, + { + "epoch": 0.75, + "learning_rate": 3.1072558048562805e-06, + "loss": 0.9165, + "step": 23446 + }, + { + "epoch": 0.75, + "learning_rate": 3.1065054016702024e-06, + "loss": 1.0161, + "step": 23447 + }, + { + "epoch": 0.75, + "learning_rate": 3.105755072443689e-06, + "loss": 1.0342, + "step": 23448 + }, + { + "epoch": 0.75, + "learning_rate": 3.1050048171847967e-06, + "loss": 0.981, + "step": 23449 + }, + { + "epoch": 0.75, + "learning_rate": 3.1042546359015712e-06, + "loss": 0.9893, + "step": 23450 + }, + { + "epoch": 0.75, + "learning_rate": 3.1035045286020595e-06, + "loss": 0.9634, + "step": 23451 + }, + { + "epoch": 0.75, + "learning_rate": 3.1027544952943134e-06, + "loss": 0.979, + "step": 23452 + }, + { + "epoch": 0.75, + "learning_rate": 3.1020045359863793e-06, + "loss": 0.7686, + "step": 23453 + }, + { + "epoch": 0.75, + "learning_rate": 3.1012546506862986e-06, + "loss": 0.9102, + "step": 23454 + }, + { + "epoch": 0.75, + "learning_rate": 3.100504839402123e-06, + "loss": 0.9717, + "step": 23455 + }, + { + "epoch": 0.75, + "learning_rate": 3.099755102141895e-06, + "loss": 0.7415, + "step": 23456 + }, + { + "epoch": 0.75, + "learning_rate": 3.0990054389136537e-06, + "loss": 1.0244, + "step": 23457 + }, + { + "epoch": 0.75, + "learning_rate": 3.098255849725451e-06, + "loss": 0.9595, + "step": 23458 + }, + { + "epoch": 0.75, + "learning_rate": 3.097506334585321e-06, + "loss": 0.9263, + "step": 23459 + }, + { + "epoch": 0.75, + "learning_rate": 3.0967568935013113e-06, + "loss": 0.9175, + "step": 23460 + }, + { + "epoch": 0.75, + "learning_rate": 3.0960075264814604e-06, + "loss": 0.999, + "step": 23461 + }, + { + "epoch": 0.75, + "learning_rate": 3.0952582335338076e-06, + "loss": 0.4644, + "step": 23462 + }, + { + "epoch": 0.75, + "learning_rate": 3.0945090146663893e-06, + "loss": 0.9253, + "step": 23463 + }, + { + "epoch": 0.75, + "learning_rate": 3.0937598698872504e-06, + "loss": 0.9897, + "step": 23464 + }, + { + "epoch": 0.75, + "learning_rate": 3.09301079920442e-06, + "loss": 0.9976, + "step": 23465 + }, + { + "epoch": 0.75, + "learning_rate": 3.0922618026259444e-06, + "loss": 0.9756, + "step": 23466 + }, + { + "epoch": 0.75, + "learning_rate": 3.0915128801598505e-06, + "loss": 1.0386, + "step": 23467 + }, + { + "epoch": 0.75, + "learning_rate": 3.090764031814183e-06, + "loss": 0.9238, + "step": 23468 + }, + { + "epoch": 0.75, + "learning_rate": 3.0900152575969653e-06, + "loss": 0.9648, + "step": 23469 + }, + { + "epoch": 0.75, + "learning_rate": 3.0892665575162394e-06, + "loss": 0.9209, + "step": 23470 + }, + { + "epoch": 0.75, + "learning_rate": 3.0885179315800317e-06, + "loss": 0.8838, + "step": 23471 + }, + { + "epoch": 0.75, + "learning_rate": 3.0877693797963803e-06, + "loss": 0.9507, + "step": 23472 + }, + { + "epoch": 0.75, + "learning_rate": 3.0870209021733098e-06, + "loss": 0.9858, + "step": 23473 + }, + { + "epoch": 0.75, + "learning_rate": 3.0862724987188586e-06, + "loss": 0.9414, + "step": 23474 + }, + { + "epoch": 0.75, + "learning_rate": 3.0855241694410478e-06, + "loss": 0.9502, + "step": 23475 + }, + { + "epoch": 0.75, + "learning_rate": 3.084775914347916e-06, + "loss": 0.8433, + "step": 23476 + }, + { + "epoch": 0.75, + "learning_rate": 3.0840277334474788e-06, + "loss": 1.0278, + "step": 23477 + }, + { + "epoch": 0.75, + "learning_rate": 3.083279626747774e-06, + "loss": 0.8447, + "step": 23478 + }, + { + "epoch": 0.75, + "learning_rate": 3.08253159425682e-06, + "loss": 0.9775, + "step": 23479 + }, + { + "epoch": 0.75, + "learning_rate": 3.081783635982649e-06, + "loss": 0.8145, + "step": 23480 + }, + { + "epoch": 0.75, + "learning_rate": 3.08103575193328e-06, + "loss": 0.9551, + "step": 23481 + }, + { + "epoch": 0.75, + "learning_rate": 3.0802879421167432e-06, + "loss": 0.8691, + "step": 23482 + }, + { + "epoch": 0.75, + "learning_rate": 3.0795402065410582e-06, + "loss": 0.8975, + "step": 23483 + }, + { + "epoch": 0.75, + "learning_rate": 3.0787925452142477e-06, + "loss": 0.9917, + "step": 23484 + }, + { + "epoch": 0.75, + "learning_rate": 3.0780449581443304e-06, + "loss": 0.8452, + "step": 23485 + }, + { + "epoch": 0.75, + "learning_rate": 3.077297445339332e-06, + "loss": 0.9458, + "step": 23486 + }, + { + "epoch": 0.75, + "learning_rate": 3.0765500068072686e-06, + "loss": 0.9434, + "step": 23487 + }, + { + "epoch": 0.75, + "learning_rate": 3.075802642556164e-06, + "loss": 0.7993, + "step": 23488 + }, + { + "epoch": 0.75, + "learning_rate": 3.0750553525940295e-06, + "loss": 0.9087, + "step": 23489 + }, + { + "epoch": 0.75, + "learning_rate": 3.074308136928891e-06, + "loss": 1.0205, + "step": 23490 + }, + { + "epoch": 0.75, + "learning_rate": 3.0735609955687608e-06, + "loss": 0.9004, + "step": 23491 + }, + { + "epoch": 0.75, + "learning_rate": 3.072813928521655e-06, + "loss": 0.9536, + "step": 23492 + }, + { + "epoch": 0.75, + "learning_rate": 3.072066935795587e-06, + "loss": 0.9619, + "step": 23493 + }, + { + "epoch": 0.75, + "learning_rate": 3.071320017398576e-06, + "loss": 0.936, + "step": 23494 + }, + { + "epoch": 0.75, + "learning_rate": 3.070573173338632e-06, + "loss": 0.9932, + "step": 23495 + }, + { + "epoch": 0.75, + "learning_rate": 3.069826403623767e-06, + "loss": 0.9248, + "step": 23496 + }, + { + "epoch": 0.75, + "learning_rate": 3.069079708261997e-06, + "loss": 0.542, + "step": 23497 + }, + { + "epoch": 0.75, + "learning_rate": 3.06833308726133e-06, + "loss": 0.8447, + "step": 23498 + }, + { + "epoch": 0.75, + "learning_rate": 3.0675865406297766e-06, + "loss": 1.0981, + "step": 23499 + }, + { + "epoch": 0.75, + "learning_rate": 3.0668400683753483e-06, + "loss": 1.0352, + "step": 23500 + }, + { + "epoch": 0.75, + "learning_rate": 3.066093670506054e-06, + "loss": 0.9648, + "step": 23501 + }, + { + "epoch": 0.75, + "learning_rate": 3.0653473470298977e-06, + "loss": 0.9146, + "step": 23502 + }, + { + "epoch": 0.75, + "learning_rate": 3.0646010979548923e-06, + "loss": 1.0029, + "step": 23503 + }, + { + "epoch": 0.75, + "learning_rate": 3.063854923289038e-06, + "loss": 0.9038, + "step": 23504 + }, + { + "epoch": 0.75, + "learning_rate": 3.063108823040346e-06, + "loss": 0.8887, + "step": 23505 + }, + { + "epoch": 0.75, + "learning_rate": 3.0623627972168203e-06, + "loss": 0.9678, + "step": 23506 + }, + { + "epoch": 0.75, + "learning_rate": 3.061616845826464e-06, + "loss": 0.7715, + "step": 23507 + }, + { + "epoch": 0.75, + "learning_rate": 3.0608709688772753e-06, + "loss": 0.9399, + "step": 23508 + }, + { + "epoch": 0.75, + "learning_rate": 3.0601251663772657e-06, + "loss": 0.8901, + "step": 23509 + }, + { + "epoch": 0.75, + "learning_rate": 3.0593794383344277e-06, + "loss": 0.939, + "step": 23510 + }, + { + "epoch": 0.75, + "learning_rate": 3.0586337847567714e-06, + "loss": 0.8896, + "step": 23511 + }, + { + "epoch": 0.75, + "learning_rate": 3.0578882056522885e-06, + "loss": 1.1128, + "step": 23512 + }, + { + "epoch": 0.75, + "learning_rate": 3.057142701028989e-06, + "loss": 0.9883, + "step": 23513 + }, + { + "epoch": 0.75, + "learning_rate": 3.0563972708948574e-06, + "loss": 1.0581, + "step": 23514 + }, + { + "epoch": 0.75, + "learning_rate": 3.055651915257901e-06, + "loss": 0.9229, + "step": 23515 + }, + { + "epoch": 0.75, + "learning_rate": 3.0549066341261112e-06, + "loss": 0.9033, + "step": 23516 + }, + { + "epoch": 0.75, + "learning_rate": 3.0541614275074903e-06, + "loss": 0.9727, + "step": 23517 + }, + { + "epoch": 0.75, + "learning_rate": 3.0534162954100264e-06, + "loss": 0.8555, + "step": 23518 + }, + { + "epoch": 0.75, + "learning_rate": 3.052671237841721e-06, + "loss": 0.9668, + "step": 23519 + }, + { + "epoch": 0.75, + "learning_rate": 3.051926254810561e-06, + "loss": 0.915, + "step": 23520 + }, + { + "epoch": 0.75, + "learning_rate": 3.0511813463245487e-06, + "loss": 0.9385, + "step": 23521 + }, + { + "epoch": 0.75, + "learning_rate": 3.0504365123916646e-06, + "loss": 0.9888, + "step": 23522 + }, + { + "epoch": 0.75, + "learning_rate": 3.0496917530199076e-06, + "loss": 0.855, + "step": 23523 + }, + { + "epoch": 0.75, + "learning_rate": 3.048947068217263e-06, + "loss": 0.9312, + "step": 23524 + }, + { + "epoch": 0.75, + "learning_rate": 3.048202457991726e-06, + "loss": 0.9595, + "step": 23525 + }, + { + "epoch": 0.75, + "learning_rate": 3.0474579223512803e-06, + "loss": 0.7859, + "step": 23526 + }, + { + "epoch": 0.75, + "learning_rate": 3.0467134613039184e-06, + "loss": 0.8394, + "step": 23527 + }, + { + "epoch": 0.75, + "learning_rate": 3.0459690748576263e-06, + "loss": 0.9883, + "step": 23528 + }, + { + "epoch": 0.75, + "learning_rate": 3.04522476302039e-06, + "loss": 0.917, + "step": 23529 + }, + { + "epoch": 0.75, + "learning_rate": 3.044480525800191e-06, + "loss": 1.0693, + "step": 23530 + }, + { + "epoch": 0.75, + "learning_rate": 3.0437363632050223e-06, + "loss": 0.9517, + "step": 23531 + }, + { + "epoch": 0.75, + "learning_rate": 3.0429922752428587e-06, + "loss": 1.0322, + "step": 23532 + }, + { + "epoch": 0.75, + "learning_rate": 3.0422482619216933e-06, + "loss": 0.9683, + "step": 23533 + }, + { + "epoch": 0.75, + "learning_rate": 3.041504323249502e-06, + "loss": 0.9512, + "step": 23534 + }, + { + "epoch": 0.75, + "learning_rate": 3.0407604592342664e-06, + "loss": 0.9077, + "step": 23535 + }, + { + "epoch": 0.75, + "learning_rate": 3.0400166698839707e-06, + "loss": 0.9409, + "step": 23536 + }, + { + "epoch": 0.75, + "learning_rate": 3.0392729552065936e-06, + "loss": 0.9414, + "step": 23537 + }, + { + "epoch": 0.75, + "learning_rate": 3.0385293152101115e-06, + "loss": 0.9937, + "step": 23538 + }, + { + "epoch": 0.75, + "learning_rate": 3.0377857499025075e-06, + "loss": 1.083, + "step": 23539 + }, + { + "epoch": 0.75, + "learning_rate": 3.037042259291758e-06, + "loss": 0.9644, + "step": 23540 + }, + { + "epoch": 0.75, + "learning_rate": 3.0362988433858356e-06, + "loss": 0.874, + "step": 23541 + }, + { + "epoch": 0.75, + "learning_rate": 3.035555502192722e-06, + "loss": 0.8867, + "step": 23542 + }, + { + "epoch": 0.75, + "learning_rate": 3.034812235720387e-06, + "loss": 1.0601, + "step": 23543 + }, + { + "epoch": 0.75, + "learning_rate": 3.034069043976812e-06, + "loss": 0.9883, + "step": 23544 + }, + { + "epoch": 0.75, + "learning_rate": 3.0333259269699654e-06, + "loss": 0.9702, + "step": 23545 + }, + { + "epoch": 0.75, + "learning_rate": 3.032582884707822e-06, + "loss": 0.9844, + "step": 23546 + }, + { + "epoch": 0.75, + "learning_rate": 3.031839917198349e-06, + "loss": 0.7537, + "step": 23547 + }, + { + "epoch": 0.75, + "learning_rate": 3.031097024449526e-06, + "loss": 0.7041, + "step": 23548 + }, + { + "epoch": 0.75, + "learning_rate": 3.030354206469316e-06, + "loss": 0.8833, + "step": 23549 + }, + { + "epoch": 0.75, + "learning_rate": 3.029611463265694e-06, + "loss": 0.9443, + "step": 23550 + }, + { + "epoch": 0.75, + "learning_rate": 3.028868794846627e-06, + "loss": 0.9321, + "step": 23551 + }, + { + "epoch": 0.75, + "learning_rate": 3.028126201220083e-06, + "loss": 0.8936, + "step": 23552 + }, + { + "epoch": 0.75, + "learning_rate": 3.0273836823940252e-06, + "loss": 0.9258, + "step": 23553 + }, + { + "epoch": 0.75, + "learning_rate": 3.0266412383764264e-06, + "loss": 0.9927, + "step": 23554 + }, + { + "epoch": 0.75, + "learning_rate": 3.0258988691752466e-06, + "loss": 0.9771, + "step": 23555 + }, + { + "epoch": 0.75, + "learning_rate": 3.025156574798457e-06, + "loss": 0.9443, + "step": 23556 + }, + { + "epoch": 0.75, + "learning_rate": 3.024414355254014e-06, + "loss": 0.8777, + "step": 23557 + }, + { + "epoch": 0.75, + "learning_rate": 3.0236722105498884e-06, + "loss": 0.9497, + "step": 23558 + }, + { + "epoch": 0.75, + "learning_rate": 3.0229301406940393e-06, + "loss": 0.9912, + "step": 23559 + }, + { + "epoch": 0.75, + "learning_rate": 3.022188145694427e-06, + "loss": 0.75, + "step": 23560 + }, + { + "epoch": 0.75, + "learning_rate": 3.0214462255590103e-06, + "loss": 0.9697, + "step": 23561 + }, + { + "epoch": 0.75, + "learning_rate": 3.020704380295756e-06, + "loss": 0.9116, + "step": 23562 + }, + { + "epoch": 0.75, + "learning_rate": 3.0199626099126154e-06, + "loss": 1.0068, + "step": 23563 + }, + { + "epoch": 0.75, + "learning_rate": 3.0192209144175543e-06, + "loss": 0.9048, + "step": 23564 + }, + { + "epoch": 0.75, + "learning_rate": 3.0184792938185237e-06, + "loss": 0.4563, + "step": 23565 + }, + { + "epoch": 0.75, + "learning_rate": 3.0177377481234903e-06, + "loss": 0.9282, + "step": 23566 + }, + { + "epoch": 0.75, + "learning_rate": 3.016996277340397e-06, + "loss": 0.9307, + "step": 23567 + }, + { + "epoch": 0.75, + "learning_rate": 3.0162548814772073e-06, + "loss": 0.8794, + "step": 23568 + }, + { + "epoch": 0.75, + "learning_rate": 3.0155135605418718e-06, + "loss": 1.1499, + "step": 23569 + }, + { + "epoch": 0.75, + "learning_rate": 3.014772314542348e-06, + "loss": 0.9155, + "step": 23570 + }, + { + "epoch": 0.75, + "learning_rate": 3.014031143486583e-06, + "loss": 0.9883, + "step": 23571 + }, + { + "epoch": 0.75, + "learning_rate": 3.0132900473825353e-06, + "loss": 0.9712, + "step": 23572 + }, + { + "epoch": 0.75, + "learning_rate": 3.012549026238154e-06, + "loss": 1.0513, + "step": 23573 + }, + { + "epoch": 0.75, + "learning_rate": 3.011808080061387e-06, + "loss": 0.885, + "step": 23574 + }, + { + "epoch": 0.75, + "learning_rate": 3.0110672088601835e-06, + "loss": 0.9399, + "step": 23575 + }, + { + "epoch": 0.75, + "learning_rate": 3.010326412642496e-06, + "loss": 0.8833, + "step": 23576 + }, + { + "epoch": 0.75, + "learning_rate": 3.009585691416267e-06, + "loss": 0.9043, + "step": 23577 + }, + { + "epoch": 0.75, + "learning_rate": 3.00884504518945e-06, + "loss": 0.937, + "step": 23578 + }, + { + "epoch": 0.75, + "learning_rate": 3.0081044739699883e-06, + "loss": 0.9097, + "step": 23579 + }, + { + "epoch": 0.75, + "learning_rate": 3.007363977765825e-06, + "loss": 0.834, + "step": 23580 + }, + { + "epoch": 0.75, + "learning_rate": 3.00662355658491e-06, + "loss": 0.8657, + "step": 23581 + }, + { + "epoch": 0.75, + "learning_rate": 3.005883210435184e-06, + "loss": 0.9834, + "step": 23582 + }, + { + "epoch": 0.75, + "learning_rate": 3.0051429393245867e-06, + "loss": 0.8586, + "step": 23583 + }, + { + "epoch": 0.75, + "learning_rate": 3.004402743261069e-06, + "loss": 0.8779, + "step": 23584 + }, + { + "epoch": 0.75, + "learning_rate": 3.0036626222525665e-06, + "loss": 0.9727, + "step": 23585 + }, + { + "epoch": 0.75, + "learning_rate": 3.002922576307017e-06, + "loss": 0.9639, + "step": 23586 + }, + { + "epoch": 0.75, + "learning_rate": 3.002182605432369e-06, + "loss": 0.8813, + "step": 23587 + }, + { + "epoch": 0.75, + "learning_rate": 3.001442709636553e-06, + "loss": 0.7974, + "step": 23588 + }, + { + "epoch": 0.75, + "learning_rate": 3.000702888927517e-06, + "loss": 0.7979, + "step": 23589 + }, + { + "epoch": 0.75, + "learning_rate": 2.999963143313186e-06, + "loss": 0.8975, + "step": 23590 + }, + { + "epoch": 0.75, + "learning_rate": 2.999223472801508e-06, + "loss": 0.8638, + "step": 23591 + }, + { + "epoch": 0.75, + "learning_rate": 2.9984838774004086e-06, + "loss": 0.9795, + "step": 23592 + }, + { + "epoch": 0.75, + "learning_rate": 2.9977443571178333e-06, + "loss": 0.9727, + "step": 23593 + }, + { + "epoch": 0.75, + "learning_rate": 2.9970049119617072e-06, + "loss": 0.9927, + "step": 23594 + }, + { + "epoch": 0.75, + "learning_rate": 2.996265541939972e-06, + "loss": 0.915, + "step": 23595 + }, + { + "epoch": 0.75, + "learning_rate": 2.995526247060555e-06, + "loss": 0.8926, + "step": 23596 + }, + { + "epoch": 0.75, + "learning_rate": 2.994787027331391e-06, + "loss": 0.9399, + "step": 23597 + }, + { + "epoch": 0.75, + "learning_rate": 2.9940478827604034e-06, + "loss": 0.8604, + "step": 23598 + }, + { + "epoch": 0.75, + "learning_rate": 2.993308813355533e-06, + "loss": 1.062, + "step": 23599 + }, + { + "epoch": 0.75, + "learning_rate": 2.992569819124702e-06, + "loss": 0.9258, + "step": 23600 + }, + { + "epoch": 0.75, + "learning_rate": 2.9918309000758426e-06, + "loss": 0.8716, + "step": 23601 + }, + { + "epoch": 0.75, + "learning_rate": 2.99109205621688e-06, + "loss": 0.8984, + "step": 23602 + }, + { + "epoch": 0.75, + "learning_rate": 2.990353287555745e-06, + "loss": 0.998, + "step": 23603 + }, + { + "epoch": 0.75, + "learning_rate": 2.9896145941003606e-06, + "loss": 0.9111, + "step": 23604 + }, + { + "epoch": 0.75, + "learning_rate": 2.988875975858654e-06, + "loss": 0.812, + "step": 23605 + }, + { + "epoch": 0.75, + "learning_rate": 2.988137432838545e-06, + "loss": 0.9624, + "step": 23606 + }, + { + "epoch": 0.75, + "learning_rate": 2.9873989650479653e-06, + "loss": 1.0513, + "step": 23607 + }, + { + "epoch": 0.75, + "learning_rate": 2.9866605724948284e-06, + "loss": 0.875, + "step": 23608 + }, + { + "epoch": 0.76, + "learning_rate": 2.985922255187066e-06, + "loss": 1.002, + "step": 23609 + }, + { + "epoch": 0.76, + "learning_rate": 2.9851840131325917e-06, + "loss": 0.8408, + "step": 23610 + }, + { + "epoch": 0.76, + "learning_rate": 2.9844458463393322e-06, + "loss": 0.998, + "step": 23611 + }, + { + "epoch": 0.76, + "learning_rate": 2.983707754815204e-06, + "loss": 1.021, + "step": 23612 + }, + { + "epoch": 0.76, + "learning_rate": 2.982969738568127e-06, + "loss": 0.8452, + "step": 23613 + }, + { + "epoch": 0.76, + "learning_rate": 2.9822317976060157e-06, + "loss": 0.8301, + "step": 23614 + }, + { + "epoch": 0.76, + "learning_rate": 2.981493931936793e-06, + "loss": 0.9346, + "step": 23615 + }, + { + "epoch": 0.76, + "learning_rate": 2.9807561415683696e-06, + "loss": 0.937, + "step": 23616 + }, + { + "epoch": 0.76, + "learning_rate": 2.9800184265086674e-06, + "loss": 0.9375, + "step": 23617 + }, + { + "epoch": 0.76, + "learning_rate": 2.9792807867655983e-06, + "loss": 0.7422, + "step": 23618 + }, + { + "epoch": 0.76, + "learning_rate": 2.978543222347076e-06, + "loss": 0.9199, + "step": 23619 + }, + { + "epoch": 0.76, + "learning_rate": 2.977805733261011e-06, + "loss": 0.8911, + "step": 23620 + }, + { + "epoch": 0.76, + "learning_rate": 2.9770683195153215e-06, + "loss": 0.9321, + "step": 23621 + }, + { + "epoch": 0.76, + "learning_rate": 2.976330981117913e-06, + "loss": 1.04, + "step": 23622 + }, + { + "epoch": 0.76, + "learning_rate": 2.9755937180767037e-06, + "loss": 0.4436, + "step": 23623 + }, + { + "epoch": 0.76, + "learning_rate": 2.974856530399599e-06, + "loss": 0.7871, + "step": 23624 + }, + { + "epoch": 0.76, + "learning_rate": 2.9741194180945055e-06, + "loss": 0.832, + "step": 23625 + }, + { + "epoch": 0.76, + "learning_rate": 2.9733823811693385e-06, + "loss": 0.9048, + "step": 23626 + }, + { + "epoch": 0.76, + "learning_rate": 2.972645419632002e-06, + "loss": 0.9282, + "step": 23627 + }, + { + "epoch": 0.76, + "learning_rate": 2.971908533490402e-06, + "loss": 0.8633, + "step": 23628 + }, + { + "epoch": 0.76, + "learning_rate": 2.9711717227524415e-06, + "loss": 0.897, + "step": 23629 + }, + { + "epoch": 0.76, + "learning_rate": 2.970434987426033e-06, + "loss": 0.9019, + "step": 23630 + }, + { + "epoch": 0.76, + "learning_rate": 2.969698327519075e-06, + "loss": 0.875, + "step": 23631 + }, + { + "epoch": 0.76, + "learning_rate": 2.968961743039475e-06, + "loss": 0.9619, + "step": 23632 + }, + { + "epoch": 0.76, + "learning_rate": 2.9682252339951313e-06, + "loss": 0.8193, + "step": 23633 + }, + { + "epoch": 0.76, + "learning_rate": 2.967488800393955e-06, + "loss": 1.0039, + "step": 23634 + }, + { + "epoch": 0.76, + "learning_rate": 2.966752442243833e-06, + "loss": 0.8198, + "step": 23635 + }, + { + "epoch": 0.76, + "learning_rate": 2.9660161595526783e-06, + "loss": 0.9185, + "step": 23636 + }, + { + "epoch": 0.76, + "learning_rate": 2.9652799523283815e-06, + "loss": 0.9834, + "step": 23637 + }, + { + "epoch": 0.76, + "learning_rate": 2.964543820578848e-06, + "loss": 0.946, + "step": 23638 + }, + { + "epoch": 0.76, + "learning_rate": 2.96380776431197e-06, + "loss": 1.0049, + "step": 23639 + }, + { + "epoch": 0.76, + "learning_rate": 2.9630717835356504e-06, + "loss": 0.8271, + "step": 23640 + }, + { + "epoch": 0.76, + "learning_rate": 2.962335878257779e-06, + "loss": 0.8564, + "step": 23641 + }, + { + "epoch": 0.76, + "learning_rate": 2.9616000484862616e-06, + "loss": 0.9395, + "step": 23642 + }, + { + "epoch": 0.76, + "learning_rate": 2.96086429422898e-06, + "loss": 0.9434, + "step": 23643 + }, + { + "epoch": 0.76, + "learning_rate": 2.960128615493836e-06, + "loss": 0.9712, + "step": 23644 + }, + { + "epoch": 0.76, + "learning_rate": 2.9593930122887184e-06, + "loss": 0.9644, + "step": 23645 + }, + { + "epoch": 0.76, + "learning_rate": 2.9586574846215245e-06, + "loss": 0.9824, + "step": 23646 + }, + { + "epoch": 0.76, + "learning_rate": 2.95792203250014e-06, + "loss": 0.8091, + "step": 23647 + }, + { + "epoch": 0.76, + "learning_rate": 2.9571866559324603e-06, + "loss": 0.7305, + "step": 23648 + }, + { + "epoch": 0.76, + "learning_rate": 2.956451354926374e-06, + "loss": 0.9214, + "step": 23649 + }, + { + "epoch": 0.76, + "learning_rate": 2.9557161294897696e-06, + "loss": 1.0005, + "step": 23650 + }, + { + "epoch": 0.76, + "learning_rate": 2.954980979630532e-06, + "loss": 0.9409, + "step": 23651 + }, + { + "epoch": 0.76, + "learning_rate": 2.954245905356553e-06, + "loss": 0.5237, + "step": 23652 + }, + { + "epoch": 0.76, + "learning_rate": 2.953510906675716e-06, + "loss": 0.9761, + "step": 23653 + }, + { + "epoch": 0.76, + "learning_rate": 2.9527759835959103e-06, + "loss": 0.8452, + "step": 23654 + }, + { + "epoch": 0.76, + "learning_rate": 2.9520411361250166e-06, + "loss": 0.8281, + "step": 23655 + }, + { + "epoch": 0.76, + "learning_rate": 2.9513063642709237e-06, + "loss": 0.8931, + "step": 23656 + }, + { + "epoch": 0.76, + "learning_rate": 2.950571668041513e-06, + "loss": 0.8276, + "step": 23657 + }, + { + "epoch": 0.76, + "learning_rate": 2.949837047444666e-06, + "loss": 0.9644, + "step": 23658 + }, + { + "epoch": 0.76, + "learning_rate": 2.9491025024882614e-06, + "loss": 1.0112, + "step": 23659 + }, + { + "epoch": 0.76, + "learning_rate": 2.9483680331801868e-06, + "loss": 0.9946, + "step": 23660 + }, + { + "epoch": 0.76, + "learning_rate": 2.9476336395283157e-06, + "loss": 1.0, + "step": 23661 + }, + { + "epoch": 0.76, + "learning_rate": 2.9468993215405337e-06, + "loss": 1.0059, + "step": 23662 + }, + { + "epoch": 0.76, + "learning_rate": 2.9461650792247165e-06, + "loss": 0.8945, + "step": 23663 + }, + { + "epoch": 0.76, + "learning_rate": 2.9454309125887405e-06, + "loss": 1.0112, + "step": 23664 + }, + { + "epoch": 0.76, + "learning_rate": 2.9446968216404803e-06, + "loss": 0.894, + "step": 23665 + }, + { + "epoch": 0.76, + "learning_rate": 2.943962806387818e-06, + "loss": 0.957, + "step": 23666 + }, + { + "epoch": 0.76, + "learning_rate": 2.9432288668386244e-06, + "loss": 0.9453, + "step": 23667 + }, + { + "epoch": 0.76, + "learning_rate": 2.9424950030007737e-06, + "loss": 0.9624, + "step": 23668 + }, + { + "epoch": 0.76, + "learning_rate": 2.941761214882143e-06, + "loss": 1.0454, + "step": 23669 + }, + { + "epoch": 0.76, + "learning_rate": 2.9410275024905997e-06, + "loss": 0.9658, + "step": 23670 + }, + { + "epoch": 0.76, + "learning_rate": 2.940293865834022e-06, + "loss": 0.8569, + "step": 23671 + }, + { + "epoch": 0.76, + "learning_rate": 2.939560304920277e-06, + "loss": 0.8535, + "step": 23672 + }, + { + "epoch": 0.76, + "learning_rate": 2.9388268197572366e-06, + "loss": 0.8564, + "step": 23673 + }, + { + "epoch": 0.76, + "learning_rate": 2.938093410352767e-06, + "loss": 0.9351, + "step": 23674 + }, + { + "epoch": 0.76, + "learning_rate": 2.9373600767147413e-06, + "loss": 0.939, + "step": 23675 + }, + { + "epoch": 0.76, + "learning_rate": 2.936626818851023e-06, + "loss": 0.9312, + "step": 23676 + }, + { + "epoch": 0.76, + "learning_rate": 2.9358936367694845e-06, + "loss": 0.9995, + "step": 23677 + }, + { + "epoch": 0.76, + "learning_rate": 2.9351605304779862e-06, + "loss": 0.9473, + "step": 23678 + }, + { + "epoch": 0.76, + "learning_rate": 2.9344274999844023e-06, + "loss": 1.0371, + "step": 23679 + }, + { + "epoch": 0.76, + "learning_rate": 2.933694545296586e-06, + "loss": 1.0186, + "step": 23680 + }, + { + "epoch": 0.76, + "learning_rate": 2.932961666422409e-06, + "loss": 0.981, + "step": 23681 + }, + { + "epoch": 0.76, + "learning_rate": 2.9322288633697293e-06, + "loss": 0.9805, + "step": 23682 + }, + { + "epoch": 0.76, + "learning_rate": 2.9314961361464145e-06, + "loss": 1.1123, + "step": 23683 + }, + { + "epoch": 0.76, + "learning_rate": 2.9307634847603205e-06, + "loss": 0.8232, + "step": 23684 + }, + { + "epoch": 0.76, + "learning_rate": 2.9300309092193126e-06, + "loss": 0.9536, + "step": 23685 + }, + { + "epoch": 0.76, + "learning_rate": 2.929298409531246e-06, + "loss": 0.7847, + "step": 23686 + }, + { + "epoch": 0.76, + "learning_rate": 2.928565985703988e-06, + "loss": 0.47, + "step": 23687 + }, + { + "epoch": 0.76, + "learning_rate": 2.927833637745384e-06, + "loss": 0.8813, + "step": 23688 + }, + { + "epoch": 0.76, + "learning_rate": 2.927101365663302e-06, + "loss": 1.043, + "step": 23689 + }, + { + "epoch": 0.76, + "learning_rate": 2.9263691694655915e-06, + "loss": 0.874, + "step": 23690 + }, + { + "epoch": 0.76, + "learning_rate": 2.925637049160114e-06, + "loss": 0.9653, + "step": 23691 + }, + { + "epoch": 0.76, + "learning_rate": 2.924905004754718e-06, + "loss": 0.8521, + "step": 23692 + }, + { + "epoch": 0.76, + "learning_rate": 2.924173036257264e-06, + "loss": 0.8862, + "step": 23693 + }, + { + "epoch": 0.76, + "learning_rate": 2.9234411436755982e-06, + "loss": 0.9351, + "step": 23694 + }, + { + "epoch": 0.76, + "learning_rate": 2.9227093270175844e-06, + "loss": 0.8472, + "step": 23695 + }, + { + "epoch": 0.76, + "learning_rate": 2.9219775862910604e-06, + "loss": 0.9077, + "step": 23696 + }, + { + "epoch": 0.76, + "learning_rate": 2.921245921503887e-06, + "loss": 0.895, + "step": 23697 + }, + { + "epoch": 0.76, + "learning_rate": 2.9205143326639063e-06, + "loss": 0.9604, + "step": 23698 + }, + { + "epoch": 0.76, + "learning_rate": 2.9197828197789756e-06, + "loss": 0.8472, + "step": 23699 + }, + { + "epoch": 0.76, + "learning_rate": 2.9190513828569354e-06, + "loss": 0.9873, + "step": 23700 + }, + { + "epoch": 0.76, + "learning_rate": 2.9183200219056406e-06, + "loss": 1.0117, + "step": 23701 + }, + { + "epoch": 0.76, + "learning_rate": 2.9175887369329336e-06, + "loss": 0.9868, + "step": 23702 + }, + { + "epoch": 0.76, + "learning_rate": 2.916857527946663e-06, + "loss": 1.002, + "step": 23703 + }, + { + "epoch": 0.76, + "learning_rate": 2.916126394954667e-06, + "loss": 0.8501, + "step": 23704 + }, + { + "epoch": 0.76, + "learning_rate": 2.915395337964798e-06, + "loss": 0.9321, + "step": 23705 + }, + { + "epoch": 0.76, + "learning_rate": 2.914664356984893e-06, + "loss": 0.9185, + "step": 23706 + }, + { + "epoch": 0.76, + "learning_rate": 2.9139334520228012e-06, + "loss": 0.9458, + "step": 23707 + }, + { + "epoch": 0.76, + "learning_rate": 2.9132026230863607e-06, + "loss": 1.0723, + "step": 23708 + }, + { + "epoch": 0.76, + "learning_rate": 2.912471870183411e-06, + "loss": 0.9287, + "step": 23709 + }, + { + "epoch": 0.76, + "learning_rate": 2.9117411933217953e-06, + "loss": 0.917, + "step": 23710 + }, + { + "epoch": 0.76, + "learning_rate": 2.9110105925093536e-06, + "loss": 0.9922, + "step": 23711 + }, + { + "epoch": 0.76, + "learning_rate": 2.9102800677539213e-06, + "loss": 0.9922, + "step": 23712 + }, + { + "epoch": 0.76, + "learning_rate": 2.9095496190633343e-06, + "loss": 0.9231, + "step": 23713 + }, + { + "epoch": 0.76, + "learning_rate": 2.9088192464454357e-06, + "loss": 1.021, + "step": 23714 + }, + { + "epoch": 0.76, + "learning_rate": 2.9080889499080555e-06, + "loss": 0.8208, + "step": 23715 + }, + { + "epoch": 0.76, + "learning_rate": 2.9073587294590343e-06, + "loss": 0.9214, + "step": 23716 + }, + { + "epoch": 0.76, + "learning_rate": 2.906628585106205e-06, + "loss": 0.9038, + "step": 23717 + }, + { + "epoch": 0.76, + "learning_rate": 2.9058985168574007e-06, + "loss": 1.0405, + "step": 23718 + }, + { + "epoch": 0.76, + "learning_rate": 2.905168524720451e-06, + "loss": 0.9907, + "step": 23719 + }, + { + "epoch": 0.76, + "learning_rate": 2.9044386087031927e-06, + "loss": 0.8799, + "step": 23720 + }, + { + "epoch": 0.76, + "learning_rate": 2.9037087688134526e-06, + "loss": 0.9688, + "step": 23721 + }, + { + "epoch": 0.76, + "learning_rate": 2.9029790050590665e-06, + "loss": 0.9146, + "step": 23722 + }, + { + "epoch": 0.76, + "learning_rate": 2.9022493174478593e-06, + "loss": 0.9209, + "step": 23723 + }, + { + "epoch": 0.76, + "learning_rate": 2.901519705987663e-06, + "loss": 0.9229, + "step": 23724 + }, + { + "epoch": 0.76, + "learning_rate": 2.900790170686305e-06, + "loss": 1.0835, + "step": 23725 + }, + { + "epoch": 0.76, + "learning_rate": 2.900060711551611e-06, + "loss": 0.9556, + "step": 23726 + }, + { + "epoch": 0.76, + "learning_rate": 2.899331328591405e-06, + "loss": 0.9783, + "step": 23727 + }, + { + "epoch": 0.76, + "learning_rate": 2.898602021813518e-06, + "loss": 0.968, + "step": 23728 + }, + { + "epoch": 0.76, + "learning_rate": 2.8978727912257698e-06, + "loss": 1.0479, + "step": 23729 + }, + { + "epoch": 0.76, + "learning_rate": 2.8971436368359886e-06, + "loss": 0.9692, + "step": 23730 + }, + { + "epoch": 0.76, + "learning_rate": 2.896414558651991e-06, + "loss": 1.0669, + "step": 23731 + }, + { + "epoch": 0.76, + "learning_rate": 2.895685556681611e-06, + "loss": 0.9116, + "step": 23732 + }, + { + "epoch": 0.76, + "learning_rate": 2.894956630932656e-06, + "loss": 0.9326, + "step": 23733 + }, + { + "epoch": 0.76, + "learning_rate": 2.894227781412955e-06, + "loss": 0.9185, + "step": 23734 + }, + { + "epoch": 0.76, + "learning_rate": 2.8934990081303225e-06, + "loss": 0.9399, + "step": 23735 + }, + { + "epoch": 0.76, + "learning_rate": 2.8927703110925854e-06, + "loss": 0.8862, + "step": 23736 + }, + { + "epoch": 0.76, + "learning_rate": 2.892041690307552e-06, + "loss": 1.0063, + "step": 23737 + }, + { + "epoch": 0.76, + "learning_rate": 2.891313145783049e-06, + "loss": 0.8584, + "step": 23738 + }, + { + "epoch": 0.76, + "learning_rate": 2.8905846775268843e-06, + "loss": 1.0015, + "step": 23739 + }, + { + "epoch": 0.76, + "learning_rate": 2.889856285546885e-06, + "loss": 0.896, + "step": 23740 + }, + { + "epoch": 0.76, + "learning_rate": 2.889127969850851e-06, + "loss": 0.9067, + "step": 23741 + }, + { + "epoch": 0.76, + "learning_rate": 2.888399730446608e-06, + "loss": 0.9541, + "step": 23742 + }, + { + "epoch": 0.76, + "learning_rate": 2.887671567341962e-06, + "loss": 0.9927, + "step": 23743 + }, + { + "epoch": 0.76, + "learning_rate": 2.8869434805447316e-06, + "loss": 1.103, + "step": 23744 + }, + { + "epoch": 0.76, + "learning_rate": 2.886215470062722e-06, + "loss": 0.9219, + "step": 23745 + }, + { + "epoch": 0.76, + "learning_rate": 2.88548753590375e-06, + "loss": 1.0005, + "step": 23746 + }, + { + "epoch": 0.76, + "learning_rate": 2.884759678075624e-06, + "loss": 0.9116, + "step": 23747 + }, + { + "epoch": 0.76, + "learning_rate": 2.8840318965861514e-06, + "loss": 0.8965, + "step": 23748 + }, + { + "epoch": 0.76, + "learning_rate": 2.883304191443137e-06, + "loss": 0.8843, + "step": 23749 + }, + { + "epoch": 0.76, + "learning_rate": 2.8825765626543965e-06, + "loss": 0.8945, + "step": 23750 + }, + { + "epoch": 0.76, + "learning_rate": 2.881849010227732e-06, + "loss": 1.041, + "step": 23751 + }, + { + "epoch": 0.76, + "learning_rate": 2.8811215341709463e-06, + "loss": 1.0259, + "step": 23752 + }, + { + "epoch": 0.76, + "learning_rate": 2.8803941344918505e-06, + "loss": 0.8149, + "step": 23753 + }, + { + "epoch": 0.76, + "learning_rate": 2.879666811198244e-06, + "loss": 0.9966, + "step": 23754 + }, + { + "epoch": 0.76, + "learning_rate": 2.8789395642979346e-06, + "loss": 0.9209, + "step": 23755 + }, + { + "epoch": 0.76, + "learning_rate": 2.8782123937987226e-06, + "loss": 1.0068, + "step": 23756 + }, + { + "epoch": 0.76, + "learning_rate": 2.877485299708409e-06, + "loss": 0.9551, + "step": 23757 + }, + { + "epoch": 0.76, + "learning_rate": 2.876758282034793e-06, + "loss": 0.8647, + "step": 23758 + }, + { + "epoch": 0.76, + "learning_rate": 2.87603134078568e-06, + "loss": 1.019, + "step": 23759 + }, + { + "epoch": 0.76, + "learning_rate": 2.875304475968863e-06, + "loss": 0.8179, + "step": 23760 + }, + { + "epoch": 0.76, + "learning_rate": 2.874577687592147e-06, + "loss": 0.8809, + "step": 23761 + }, + { + "epoch": 0.76, + "learning_rate": 2.873850975663327e-06, + "loss": 0.9839, + "step": 23762 + }, + { + "epoch": 0.76, + "learning_rate": 2.8731243401901977e-06, + "loss": 0.981, + "step": 23763 + }, + { + "epoch": 0.76, + "learning_rate": 2.872397781180555e-06, + "loss": 0.8804, + "step": 23764 + }, + { + "epoch": 0.76, + "learning_rate": 2.8716712986421978e-06, + "loss": 0.9893, + "step": 23765 + }, + { + "epoch": 0.76, + "learning_rate": 2.8709448925829155e-06, + "loss": 0.9858, + "step": 23766 + }, + { + "epoch": 0.76, + "learning_rate": 2.870218563010507e-06, + "loss": 1.0562, + "step": 23767 + }, + { + "epoch": 0.76, + "learning_rate": 2.869492309932761e-06, + "loss": 0.9429, + "step": 23768 + }, + { + "epoch": 0.76, + "learning_rate": 2.8687661333574723e-06, + "loss": 0.8237, + "step": 23769 + }, + { + "epoch": 0.76, + "learning_rate": 2.868040033292432e-06, + "loss": 0.9819, + "step": 23770 + }, + { + "epoch": 0.76, + "learning_rate": 2.867314009745428e-06, + "loss": 0.9253, + "step": 23771 + }, + { + "epoch": 0.76, + "learning_rate": 2.866588062724247e-06, + "loss": 0.9023, + "step": 23772 + }, + { + "epoch": 0.76, + "learning_rate": 2.8658621922366847e-06, + "loss": 0.9102, + "step": 23773 + }, + { + "epoch": 0.76, + "learning_rate": 2.865136398290522e-06, + "loss": 1.0244, + "step": 23774 + }, + { + "epoch": 0.76, + "learning_rate": 2.8644106808935535e-06, + "loss": 0.854, + "step": 23775 + }, + { + "epoch": 0.76, + "learning_rate": 2.8636850400535576e-06, + "loss": 0.8457, + "step": 23776 + }, + { + "epoch": 0.76, + "learning_rate": 2.8629594757783263e-06, + "loss": 0.98, + "step": 23777 + }, + { + "epoch": 0.76, + "learning_rate": 2.86223398807564e-06, + "loss": 0.9309, + "step": 23778 + }, + { + "epoch": 0.76, + "learning_rate": 2.8615085769532846e-06, + "loss": 0.8574, + "step": 23779 + }, + { + "epoch": 0.76, + "learning_rate": 2.8607832424190376e-06, + "loss": 0.9404, + "step": 23780 + }, + { + "epoch": 0.76, + "learning_rate": 2.8600579844806887e-06, + "loss": 1.0493, + "step": 23781 + }, + { + "epoch": 0.76, + "learning_rate": 2.859332803146012e-06, + "loss": 0.9961, + "step": 23782 + }, + { + "epoch": 0.76, + "learning_rate": 2.858607698422795e-06, + "loss": 0.8813, + "step": 23783 + }, + { + "epoch": 0.76, + "learning_rate": 2.85788267031881e-06, + "loss": 0.9077, + "step": 23784 + }, + { + "epoch": 0.76, + "learning_rate": 2.8571577188418466e-06, + "loss": 0.9253, + "step": 23785 + }, + { + "epoch": 0.76, + "learning_rate": 2.8564328439996692e-06, + "loss": 0.8862, + "step": 23786 + }, + { + "epoch": 0.76, + "learning_rate": 2.855708045800063e-06, + "loss": 0.9902, + "step": 23787 + }, + { + "epoch": 0.76, + "learning_rate": 2.854983324250801e-06, + "loss": 0.9282, + "step": 23788 + }, + { + "epoch": 0.76, + "learning_rate": 2.8542586793596625e-06, + "loss": 0.9937, + "step": 23789 + }, + { + "epoch": 0.76, + "learning_rate": 2.8535341111344208e-06, + "loss": 0.813, + "step": 23790 + }, + { + "epoch": 0.76, + "learning_rate": 2.852809619582845e-06, + "loss": 0.9634, + "step": 23791 + }, + { + "epoch": 0.76, + "learning_rate": 2.8520852047127146e-06, + "loss": 1.0522, + "step": 23792 + }, + { + "epoch": 0.76, + "learning_rate": 2.8513608665317995e-06, + "loss": 0.938, + "step": 23793 + }, + { + "epoch": 0.76, + "learning_rate": 2.8506366050478675e-06, + "loss": 0.9375, + "step": 23794 + }, + { + "epoch": 0.76, + "learning_rate": 2.8499124202686947e-06, + "loss": 0.9805, + "step": 23795 + }, + { + "epoch": 0.76, + "learning_rate": 2.8491883122020493e-06, + "loss": 0.979, + "step": 23796 + }, + { + "epoch": 0.76, + "learning_rate": 2.848464280855695e-06, + "loss": 0.9185, + "step": 23797 + }, + { + "epoch": 0.76, + "learning_rate": 2.8477403262374083e-06, + "loss": 0.9131, + "step": 23798 + }, + { + "epoch": 0.76, + "learning_rate": 2.847016448354948e-06, + "loss": 0.8315, + "step": 23799 + }, + { + "epoch": 0.76, + "learning_rate": 2.8462926472160877e-06, + "loss": 0.916, + "step": 23800 + }, + { + "epoch": 0.76, + "learning_rate": 2.845568922828591e-06, + "loss": 0.8616, + "step": 23801 + }, + { + "epoch": 0.76, + "learning_rate": 2.844845275200221e-06, + "loss": 0.9736, + "step": 23802 + }, + { + "epoch": 0.76, + "learning_rate": 2.844121704338739e-06, + "loss": 0.9233, + "step": 23803 + }, + { + "epoch": 0.76, + "learning_rate": 2.8433982102519155e-06, + "loss": 0.9287, + "step": 23804 + }, + { + "epoch": 0.76, + "learning_rate": 2.842674792947505e-06, + "loss": 0.8657, + "step": 23805 + }, + { + "epoch": 0.76, + "learning_rate": 2.8419514524332757e-06, + "loss": 0.9819, + "step": 23806 + }, + { + "epoch": 0.76, + "learning_rate": 2.8412281887169814e-06, + "loss": 0.9116, + "step": 23807 + }, + { + "epoch": 0.76, + "learning_rate": 2.8405050018063927e-06, + "loss": 0.9932, + "step": 23808 + }, + { + "epoch": 0.76, + "learning_rate": 2.839781891709256e-06, + "loss": 1.0044, + "step": 23809 + }, + { + "epoch": 0.76, + "learning_rate": 2.8390588584333367e-06, + "loss": 0.9341, + "step": 23810 + }, + { + "epoch": 0.76, + "learning_rate": 2.8383359019863878e-06, + "loss": 0.9648, + "step": 23811 + }, + { + "epoch": 0.76, + "learning_rate": 2.837613022376172e-06, + "loss": 0.9507, + "step": 23812 + }, + { + "epoch": 0.76, + "learning_rate": 2.8368902196104385e-06, + "loss": 0.8486, + "step": 23813 + }, + { + "epoch": 0.76, + "learning_rate": 2.8361674936969474e-06, + "loss": 0.9507, + "step": 23814 + }, + { + "epoch": 0.76, + "learning_rate": 2.835444844643451e-06, + "loss": 1.0723, + "step": 23815 + }, + { + "epoch": 0.76, + "learning_rate": 2.8347222724577028e-06, + "loss": 0.8315, + "step": 23816 + }, + { + "epoch": 0.76, + "learning_rate": 2.8339997771474513e-06, + "loss": 0.9326, + "step": 23817 + }, + { + "epoch": 0.76, + "learning_rate": 2.8332773587204543e-06, + "loss": 1.0015, + "step": 23818 + }, + { + "epoch": 0.76, + "learning_rate": 2.8325550171844562e-06, + "loss": 1.0742, + "step": 23819 + }, + { + "epoch": 0.76, + "learning_rate": 2.831832752547212e-06, + "loss": 0.9473, + "step": 23820 + }, + { + "epoch": 0.76, + "learning_rate": 2.8311105648164683e-06, + "loss": 0.9004, + "step": 23821 + }, + { + "epoch": 0.76, + "learning_rate": 2.830388453999976e-06, + "loss": 1.0264, + "step": 23822 + }, + { + "epoch": 0.76, + "learning_rate": 2.82966642010548e-06, + "loss": 0.8813, + "step": 23823 + }, + { + "epoch": 0.76, + "learning_rate": 2.828944463140729e-06, + "loss": 0.9321, + "step": 23824 + }, + { + "epoch": 0.76, + "learning_rate": 2.8282225831134636e-06, + "loss": 1.0122, + "step": 23825 + }, + { + "epoch": 0.76, + "learning_rate": 2.827500780031436e-06, + "loss": 0.9111, + "step": 23826 + }, + { + "epoch": 0.76, + "learning_rate": 2.826779053902383e-06, + "loss": 1.0283, + "step": 23827 + }, + { + "epoch": 0.76, + "learning_rate": 2.826057404734055e-06, + "loss": 0.939, + "step": 23828 + }, + { + "epoch": 0.76, + "learning_rate": 2.8253358325341917e-06, + "loss": 0.9634, + "step": 23829 + }, + { + "epoch": 0.76, + "learning_rate": 2.824614337310534e-06, + "loss": 0.9229, + "step": 23830 + }, + { + "epoch": 0.76, + "learning_rate": 2.823892919070821e-06, + "loss": 0.979, + "step": 23831 + }, + { + "epoch": 0.76, + "learning_rate": 2.8231715778227977e-06, + "loss": 0.9321, + "step": 23832 + }, + { + "epoch": 0.76, + "learning_rate": 2.8224503135741966e-06, + "loss": 0.9741, + "step": 23833 + }, + { + "epoch": 0.76, + "learning_rate": 2.821729126332763e-06, + "loss": 0.8716, + "step": 23834 + }, + { + "epoch": 0.76, + "learning_rate": 2.8210080161062324e-06, + "loss": 0.9351, + "step": 23835 + }, + { + "epoch": 0.76, + "learning_rate": 2.820286982902336e-06, + "loss": 0.5112, + "step": 23836 + }, + { + "epoch": 0.76, + "learning_rate": 2.8195660267288184e-06, + "loss": 0.9224, + "step": 23837 + }, + { + "epoch": 0.76, + "learning_rate": 2.8188451475934097e-06, + "loss": 0.9082, + "step": 23838 + }, + { + "epoch": 0.76, + "learning_rate": 2.818124345503842e-06, + "loss": 0.9702, + "step": 23839 + }, + { + "epoch": 0.76, + "learning_rate": 2.8174036204678556e-06, + "loss": 0.9316, + "step": 23840 + }, + { + "epoch": 0.76, + "learning_rate": 2.816682972493178e-06, + "loss": 1.0464, + "step": 23841 + }, + { + "epoch": 0.76, + "learning_rate": 2.8159624015875386e-06, + "loss": 0.8618, + "step": 23842 + }, + { + "epoch": 0.76, + "learning_rate": 2.815241907758676e-06, + "loss": 1.0371, + "step": 23843 + }, + { + "epoch": 0.76, + "learning_rate": 2.8145214910143128e-06, + "loss": 0.9077, + "step": 23844 + }, + { + "epoch": 0.76, + "learning_rate": 2.8138011513621834e-06, + "loss": 0.9497, + "step": 23845 + }, + { + "epoch": 0.76, + "learning_rate": 2.8130808888100146e-06, + "loss": 0.8604, + "step": 23846 + }, + { + "epoch": 0.76, + "learning_rate": 2.8123607033655343e-06, + "loss": 0.9526, + "step": 23847 + }, + { + "epoch": 0.76, + "learning_rate": 2.8116405950364655e-06, + "loss": 0.9863, + "step": 23848 + }, + { + "epoch": 0.76, + "learning_rate": 2.8109205638305414e-06, + "loss": 0.9341, + "step": 23849 + }, + { + "epoch": 0.76, + "learning_rate": 2.810200609755479e-06, + "loss": 0.9009, + "step": 23850 + }, + { + "epoch": 0.76, + "learning_rate": 2.80948073281901e-06, + "loss": 0.957, + "step": 23851 + }, + { + "epoch": 0.76, + "learning_rate": 2.8087609330288522e-06, + "loss": 0.4446, + "step": 23852 + }, + { + "epoch": 0.76, + "learning_rate": 2.8080412103927355e-06, + "loss": 0.9058, + "step": 23853 + }, + { + "epoch": 0.76, + "learning_rate": 2.8073215649183727e-06, + "loss": 0.896, + "step": 23854 + }, + { + "epoch": 0.76, + "learning_rate": 2.8066019966134907e-06, + "loss": 0.8394, + "step": 23855 + }, + { + "epoch": 0.76, + "learning_rate": 2.8058825054858052e-06, + "loss": 0.8628, + "step": 23856 + }, + { + "epoch": 0.76, + "learning_rate": 2.8051630915430416e-06, + "loss": 0.9321, + "step": 23857 + }, + { + "epoch": 0.76, + "learning_rate": 2.8044437547929125e-06, + "loss": 0.9624, + "step": 23858 + }, + { + "epoch": 0.76, + "learning_rate": 2.8037244952431408e-06, + "loss": 0.9746, + "step": 23859 + }, + { + "epoch": 0.76, + "learning_rate": 2.8030053129014378e-06, + "loss": 1.0117, + "step": 23860 + }, + { + "epoch": 0.76, + "learning_rate": 2.8022862077755284e-06, + "loss": 0.916, + "step": 23861 + }, + { + "epoch": 0.76, + "learning_rate": 2.801567179873117e-06, + "loss": 1.0273, + "step": 23862 + }, + { + "epoch": 0.76, + "learning_rate": 2.8008482292019257e-06, + "loss": 0.8828, + "step": 23863 + }, + { + "epoch": 0.76, + "learning_rate": 2.8001293557696607e-06, + "loss": 0.9736, + "step": 23864 + }, + { + "epoch": 0.76, + "learning_rate": 2.7994105595840437e-06, + "loss": 0.9185, + "step": 23865 + }, + { + "epoch": 0.76, + "learning_rate": 2.7986918406527787e-06, + "loss": 0.9121, + "step": 23866 + }, + { + "epoch": 0.76, + "learning_rate": 2.7979731989835832e-06, + "loss": 0.9707, + "step": 23867 + }, + { + "epoch": 0.76, + "learning_rate": 2.7972546345841654e-06, + "loss": 1.0791, + "step": 23868 + }, + { + "epoch": 0.76, + "learning_rate": 2.7965361474622333e-06, + "loss": 0.938, + "step": 23869 + }, + { + "epoch": 0.76, + "learning_rate": 2.7958177376254926e-06, + "loss": 0.8735, + "step": 23870 + }, + { + "epoch": 0.76, + "learning_rate": 2.795099405081657e-06, + "loss": 0.9668, + "step": 23871 + }, + { + "epoch": 0.76, + "learning_rate": 2.7943811498384275e-06, + "loss": 1.0742, + "step": 23872 + }, + { + "epoch": 0.76, + "learning_rate": 2.793662971903518e-06, + "loss": 0.9097, + "step": 23873 + }, + { + "epoch": 0.76, + "learning_rate": 2.7929448712846287e-06, + "loss": 1.001, + "step": 23874 + }, + { + "epoch": 0.76, + "learning_rate": 2.792226847989461e-06, + "loss": 1.019, + "step": 23875 + }, + { + "epoch": 0.76, + "learning_rate": 2.791508902025726e-06, + "loss": 0.9224, + "step": 23876 + }, + { + "epoch": 0.76, + "learning_rate": 2.7907910334011214e-06, + "loss": 0.9487, + "step": 23877 + }, + { + "epoch": 0.76, + "learning_rate": 2.790073242123348e-06, + "loss": 0.9678, + "step": 23878 + }, + { + "epoch": 0.76, + "learning_rate": 2.7893555282001115e-06, + "loss": 0.5186, + "step": 23879 + }, + { + "epoch": 0.76, + "learning_rate": 2.78863789163911e-06, + "loss": 0.9487, + "step": 23880 + }, + { + "epoch": 0.76, + "learning_rate": 2.78792033244804e-06, + "loss": 0.8843, + "step": 23881 + }, + { + "epoch": 0.76, + "learning_rate": 2.787202850634606e-06, + "loss": 0.9648, + "step": 23882 + }, + { + "epoch": 0.76, + "learning_rate": 2.7864854462065027e-06, + "loss": 0.9751, + "step": 23883 + }, + { + "epoch": 0.76, + "learning_rate": 2.785768119171427e-06, + "loss": 0.9785, + "step": 23884 + }, + { + "epoch": 0.76, + "learning_rate": 2.7850508695370716e-06, + "loss": 0.8037, + "step": 23885 + }, + { + "epoch": 0.76, + "learning_rate": 2.784333697311139e-06, + "loss": 0.876, + "step": 23886 + }, + { + "epoch": 0.76, + "learning_rate": 2.7836166025013167e-06, + "loss": 0.938, + "step": 23887 + }, + { + "epoch": 0.76, + "learning_rate": 2.782899585115304e-06, + "loss": 0.8833, + "step": 23888 + }, + { + "epoch": 0.76, + "learning_rate": 2.782182645160789e-06, + "loss": 0.8901, + "step": 23889 + }, + { + "epoch": 0.76, + "learning_rate": 2.7814657826454685e-06, + "loss": 0.9766, + "step": 23890 + }, + { + "epoch": 0.76, + "learning_rate": 2.7807489975770307e-06, + "loss": 0.8481, + "step": 23891 + }, + { + "epoch": 0.76, + "learning_rate": 2.7800322899631673e-06, + "loss": 0.9331, + "step": 23892 + }, + { + "epoch": 0.76, + "learning_rate": 2.7793156598115634e-06, + "loss": 0.9644, + "step": 23893 + }, + { + "epoch": 0.76, + "learning_rate": 2.7785991071299135e-06, + "loss": 0.9473, + "step": 23894 + }, + { + "epoch": 0.76, + "learning_rate": 2.7778826319259e-06, + "loss": 0.8594, + "step": 23895 + }, + { + "epoch": 0.76, + "learning_rate": 2.777166234207217e-06, + "loss": 0.8757, + "step": 23896 + }, + { + "epoch": 0.76, + "learning_rate": 2.7764499139815417e-06, + "loss": 0.9744, + "step": 23897 + }, + { + "epoch": 0.76, + "learning_rate": 2.775733671256572e-06, + "loss": 0.8599, + "step": 23898 + }, + { + "epoch": 0.76, + "learning_rate": 2.7750175060399777e-06, + "loss": 0.9526, + "step": 23899 + }, + { + "epoch": 0.76, + "learning_rate": 2.774301418339451e-06, + "loss": 0.9346, + "step": 23900 + }, + { + "epoch": 0.76, + "learning_rate": 2.7735854081626723e-06, + "loss": 0.9565, + "step": 23901 + }, + { + "epoch": 0.76, + "learning_rate": 2.772869475517326e-06, + "loss": 0.9062, + "step": 23902 + }, + { + "epoch": 0.76, + "learning_rate": 2.772153620411089e-06, + "loss": 0.9839, + "step": 23903 + }, + { + "epoch": 0.76, + "learning_rate": 2.7714378428516477e-06, + "loss": 0.9033, + "step": 23904 + }, + { + "epoch": 0.76, + "learning_rate": 2.7707221428466747e-06, + "loss": 0.8037, + "step": 23905 + }, + { + "epoch": 0.76, + "learning_rate": 2.7700065204038586e-06, + "loss": 0.9961, + "step": 23906 + }, + { + "epoch": 0.76, + "learning_rate": 2.769290975530864e-06, + "loss": 0.9883, + "step": 23907 + }, + { + "epoch": 0.76, + "learning_rate": 2.7685755082353783e-06, + "loss": 0.9897, + "step": 23908 + }, + { + "epoch": 0.76, + "learning_rate": 2.767860118525071e-06, + "loss": 1.0288, + "step": 23909 + }, + { + "epoch": 0.76, + "learning_rate": 2.7671448064076234e-06, + "loss": 0.8848, + "step": 23910 + }, + { + "epoch": 0.76, + "learning_rate": 2.766429571890704e-06, + "loss": 0.9741, + "step": 23911 + }, + { + "epoch": 0.76, + "learning_rate": 2.7657144149819924e-06, + "loss": 0.9297, + "step": 23912 + }, + { + "epoch": 0.76, + "learning_rate": 2.764999335689159e-06, + "loss": 0.9463, + "step": 23913 + }, + { + "epoch": 0.76, + "learning_rate": 2.7642843340198757e-06, + "loss": 0.9541, + "step": 23914 + }, + { + "epoch": 0.76, + "learning_rate": 2.76356940998181e-06, + "loss": 0.9321, + "step": 23915 + }, + { + "epoch": 0.76, + "learning_rate": 2.7628545635826387e-06, + "loss": 0.9888, + "step": 23916 + }, + { + "epoch": 0.76, + "learning_rate": 2.7621397948300244e-06, + "loss": 0.8643, + "step": 23917 + }, + { + "epoch": 0.76, + "learning_rate": 2.7614251037316442e-06, + "loss": 0.9341, + "step": 23918 + }, + { + "epoch": 0.76, + "learning_rate": 2.76071049029516e-06, + "loss": 0.96, + "step": 23919 + }, + { + "epoch": 0.76, + "learning_rate": 2.7599959545282375e-06, + "loss": 0.9912, + "step": 23920 + }, + { + "epoch": 0.77, + "learning_rate": 2.7592814964385494e-06, + "loss": 0.957, + "step": 23921 + }, + { + "epoch": 0.77, + "learning_rate": 2.7585671160337567e-06, + "loss": 0.9668, + "step": 23922 + }, + { + "epoch": 0.77, + "learning_rate": 2.757852813321521e-06, + "loss": 0.8945, + "step": 23923 + }, + { + "epoch": 0.77, + "learning_rate": 2.7571385883095127e-06, + "loss": 0.9766, + "step": 23924 + }, + { + "epoch": 0.77, + "learning_rate": 2.756424441005392e-06, + "loss": 1.0195, + "step": 23925 + }, + { + "epoch": 0.77, + "learning_rate": 2.7557103714168166e-06, + "loss": 0.9619, + "step": 23926 + }, + { + "epoch": 0.77, + "learning_rate": 2.7549963795514544e-06, + "loss": 0.9956, + "step": 23927 + }, + { + "epoch": 0.77, + "learning_rate": 2.75428246541696e-06, + "loss": 0.917, + "step": 23928 + }, + { + "epoch": 0.77, + "learning_rate": 2.7535686290210007e-06, + "loss": 0.8096, + "step": 23929 + }, + { + "epoch": 0.77, + "learning_rate": 2.7528548703712244e-06, + "loss": 0.9707, + "step": 23930 + }, + { + "epoch": 0.77, + "learning_rate": 2.7521411894752982e-06, + "loss": 0.8721, + "step": 23931 + }, + { + "epoch": 0.77, + "learning_rate": 2.7514275863408723e-06, + "loss": 0.8462, + "step": 23932 + }, + { + "epoch": 0.77, + "learning_rate": 2.7507140609756087e-06, + "loss": 0.9517, + "step": 23933 + }, + { + "epoch": 0.77, + "learning_rate": 2.750000613387157e-06, + "loss": 0.8687, + "step": 23934 + }, + { + "epoch": 0.77, + "learning_rate": 2.7492872435831785e-06, + "loss": 0.9214, + "step": 23935 + }, + { + "epoch": 0.77, + "learning_rate": 2.7485739515713228e-06, + "loss": 0.937, + "step": 23936 + }, + { + "epoch": 0.77, + "learning_rate": 2.747860737359244e-06, + "loss": 0.9321, + "step": 23937 + }, + { + "epoch": 0.77, + "learning_rate": 2.747147600954588e-06, + "loss": 0.9712, + "step": 23938 + }, + { + "epoch": 0.77, + "learning_rate": 2.7464345423650164e-06, + "loss": 1.062, + "step": 23939 + }, + { + "epoch": 0.77, + "learning_rate": 2.7457215615981704e-06, + "loss": 0.8428, + "step": 23940 + }, + { + "epoch": 0.77, + "learning_rate": 2.745008658661706e-06, + "loss": 0.8896, + "step": 23941 + }, + { + "epoch": 0.77, + "learning_rate": 2.7442958335632673e-06, + "loss": 0.9062, + "step": 23942 + }, + { + "epoch": 0.77, + "learning_rate": 2.7435830863105063e-06, + "loss": 1.0532, + "step": 23943 + }, + { + "epoch": 0.77, + "learning_rate": 2.7428704169110677e-06, + "loss": 0.9585, + "step": 23944 + }, + { + "epoch": 0.77, + "learning_rate": 2.7421578253725988e-06, + "loss": 0.7505, + "step": 23945 + }, + { + "epoch": 0.77, + "learning_rate": 2.7414453117027395e-06, + "loss": 0.9824, + "step": 23946 + }, + { + "epoch": 0.77, + "learning_rate": 2.7407328759091423e-06, + "loss": 0.8618, + "step": 23947 + }, + { + "epoch": 0.77, + "learning_rate": 2.7400205179994445e-06, + "loss": 0.9033, + "step": 23948 + }, + { + "epoch": 0.77, + "learning_rate": 2.739308237981294e-06, + "loss": 0.9443, + "step": 23949 + }, + { + "epoch": 0.77, + "learning_rate": 2.7385960358623274e-06, + "loss": 0.9922, + "step": 23950 + }, + { + "epoch": 0.77, + "learning_rate": 2.7378839116501955e-06, + "loss": 0.9536, + "step": 23951 + }, + { + "epoch": 0.77, + "learning_rate": 2.7371718653525247e-06, + "loss": 1.0137, + "step": 23952 + }, + { + "epoch": 0.77, + "learning_rate": 2.7364598969769663e-06, + "loss": 0.9258, + "step": 23953 + }, + { + "epoch": 0.77, + "learning_rate": 2.735748006531149e-06, + "loss": 0.8936, + "step": 23954 + }, + { + "epoch": 0.77, + "learning_rate": 2.735036194022721e-06, + "loss": 0.8428, + "step": 23955 + }, + { + "epoch": 0.77, + "learning_rate": 2.734324459459311e-06, + "loss": 0.9507, + "step": 23956 + }, + { + "epoch": 0.77, + "learning_rate": 2.733612802848561e-06, + "loss": 0.9565, + "step": 23957 + }, + { + "epoch": 0.77, + "learning_rate": 2.7329012241981023e-06, + "loss": 0.9233, + "step": 23958 + }, + { + "epoch": 0.77, + "learning_rate": 2.7321897235155724e-06, + "loss": 0.9883, + "step": 23959 + }, + { + "epoch": 0.77, + "learning_rate": 2.731478300808599e-06, + "loss": 0.9307, + "step": 23960 + }, + { + "epoch": 0.77, + "learning_rate": 2.7307669560848225e-06, + "loss": 0.8291, + "step": 23961 + }, + { + "epoch": 0.77, + "learning_rate": 2.730055689351868e-06, + "loss": 0.9146, + "step": 23962 + }, + { + "epoch": 0.77, + "learning_rate": 2.7293445006173734e-06, + "loss": 0.9487, + "step": 23963 + }, + { + "epoch": 0.77, + "learning_rate": 2.728633389888966e-06, + "loss": 0.9312, + "step": 23964 + }, + { + "epoch": 0.77, + "learning_rate": 2.727922357174272e-06, + "loss": 0.9253, + "step": 23965 + }, + { + "epoch": 0.77, + "learning_rate": 2.727211402480925e-06, + "loss": 0.8901, + "step": 23966 + }, + { + "epoch": 0.77, + "learning_rate": 2.72650052581655e-06, + "loss": 0.9355, + "step": 23967 + }, + { + "epoch": 0.77, + "learning_rate": 2.7257897271887748e-06, + "loss": 0.9424, + "step": 23968 + }, + { + "epoch": 0.77, + "learning_rate": 2.7250790066052222e-06, + "loss": 1.0151, + "step": 23969 + }, + { + "epoch": 0.77, + "learning_rate": 2.7243683640735243e-06, + "loss": 0.939, + "step": 23970 + }, + { + "epoch": 0.77, + "learning_rate": 2.723657799601297e-06, + "loss": 0.9243, + "step": 23971 + }, + { + "epoch": 0.77, + "learning_rate": 2.7229473131961726e-06, + "loss": 0.856, + "step": 23972 + }, + { + "epoch": 0.77, + "learning_rate": 2.722236904865766e-06, + "loss": 0.8867, + "step": 23973 + }, + { + "epoch": 0.77, + "learning_rate": 2.72152657461771e-06, + "loss": 0.8975, + "step": 23974 + }, + { + "epoch": 0.77, + "learning_rate": 2.7208163224596106e-06, + "loss": 1.0649, + "step": 23975 + }, + { + "epoch": 0.77, + "learning_rate": 2.720106148399101e-06, + "loss": 0.8989, + "step": 23976 + }, + { + "epoch": 0.77, + "learning_rate": 2.71939605244379e-06, + "loss": 0.9268, + "step": 23977 + }, + { + "epoch": 0.77, + "learning_rate": 2.718686034601307e-06, + "loss": 0.5227, + "step": 23978 + }, + { + "epoch": 0.77, + "learning_rate": 2.71797609487926e-06, + "loss": 0.9023, + "step": 23979 + }, + { + "epoch": 0.77, + "learning_rate": 2.7172662332852727e-06, + "loss": 0.9062, + "step": 23980 + }, + { + "epoch": 0.77, + "learning_rate": 2.7165564498269594e-06, + "loss": 0.9116, + "step": 23981 + }, + { + "epoch": 0.77, + "learning_rate": 2.715846744511934e-06, + "loss": 0.8394, + "step": 23982 + }, + { + "epoch": 0.77, + "learning_rate": 2.715137117347809e-06, + "loss": 0.9189, + "step": 23983 + }, + { + "epoch": 0.77, + "learning_rate": 2.714427568342203e-06, + "loss": 0.8623, + "step": 23984 + }, + { + "epoch": 0.77, + "learning_rate": 2.7137180975027224e-06, + "loss": 0.9209, + "step": 23985 + }, + { + "epoch": 0.77, + "learning_rate": 2.713008704836986e-06, + "loss": 0.8989, + "step": 23986 + }, + { + "epoch": 0.77, + "learning_rate": 2.7122993903525986e-06, + "loss": 0.978, + "step": 23987 + }, + { + "epoch": 0.77, + "learning_rate": 2.7115901540571752e-06, + "loss": 0.4636, + "step": 23988 + }, + { + "epoch": 0.77, + "learning_rate": 2.7108809959583237e-06, + "loss": 0.9561, + "step": 23989 + }, + { + "epoch": 0.77, + "learning_rate": 2.710171916063652e-06, + "loss": 0.9512, + "step": 23990 + }, + { + "epoch": 0.77, + "learning_rate": 2.709462914380765e-06, + "loss": 0.8374, + "step": 23991 + }, + { + "epoch": 0.77, + "learning_rate": 2.7087539909172744e-06, + "loss": 0.8755, + "step": 23992 + }, + { + "epoch": 0.77, + "learning_rate": 2.708045145680781e-06, + "loss": 0.8623, + "step": 23993 + }, + { + "epoch": 0.77, + "learning_rate": 2.7073363786788964e-06, + "loss": 0.9224, + "step": 23994 + }, + { + "epoch": 0.77, + "learning_rate": 2.706627689919218e-06, + "loss": 0.9189, + "step": 23995 + }, + { + "epoch": 0.77, + "learning_rate": 2.705919079409358e-06, + "loss": 0.9971, + "step": 23996 + }, + { + "epoch": 0.77, + "learning_rate": 2.705210547156908e-06, + "loss": 0.9888, + "step": 23997 + }, + { + "epoch": 0.77, + "learning_rate": 2.7045020931694786e-06, + "loss": 0.9023, + "step": 23998 + }, + { + "epoch": 0.77, + "learning_rate": 2.7037937174546647e-06, + "loss": 0.8604, + "step": 23999 + }, + { + "epoch": 0.77, + "learning_rate": 2.7030854200200718e-06, + "loss": 0.9741, + "step": 24000 + }, + { + "epoch": 0.77, + "learning_rate": 2.7023772008732927e-06, + "loss": 0.8955, + "step": 24001 + }, + { + "epoch": 0.77, + "learning_rate": 2.7016690600219342e-06, + "loss": 0.9287, + "step": 24002 + }, + { + "epoch": 0.77, + "learning_rate": 2.7009609974735883e-06, + "loss": 0.9526, + "step": 24003 + }, + { + "epoch": 0.77, + "learning_rate": 2.700253013235853e-06, + "loss": 0.875, + "step": 24004 + }, + { + "epoch": 0.77, + "learning_rate": 2.6995451073163215e-06, + "loss": 0.9189, + "step": 24005 + }, + { + "epoch": 0.77, + "learning_rate": 2.6988372797225938e-06, + "loss": 0.8984, + "step": 24006 + }, + { + "epoch": 0.77, + "learning_rate": 2.6981295304622613e-06, + "loss": 0.8359, + "step": 24007 + }, + { + "epoch": 0.77, + "learning_rate": 2.6974218595429147e-06, + "loss": 0.9766, + "step": 24008 + }, + { + "epoch": 0.77, + "learning_rate": 2.6967142669721522e-06, + "loss": 0.8647, + "step": 24009 + }, + { + "epoch": 0.77, + "learning_rate": 2.6960067527575605e-06, + "loss": 1.0366, + "step": 24010 + }, + { + "epoch": 0.77, + "learning_rate": 2.6952993169067355e-06, + "loss": 0.8926, + "step": 24011 + }, + { + "epoch": 0.77, + "learning_rate": 2.694591959427264e-06, + "loss": 0.9292, + "step": 24012 + }, + { + "epoch": 0.77, + "learning_rate": 2.693884680326735e-06, + "loss": 0.5117, + "step": 24013 + }, + { + "epoch": 0.77, + "learning_rate": 2.693177479612735e-06, + "loss": 0.9048, + "step": 24014 + }, + { + "epoch": 0.77, + "learning_rate": 2.692470357292857e-06, + "loss": 1.0054, + "step": 24015 + }, + { + "epoch": 0.77, + "learning_rate": 2.6917633133746812e-06, + "loss": 0.9604, + "step": 24016 + }, + { + "epoch": 0.77, + "learning_rate": 2.6910563478658003e-06, + "loss": 0.8735, + "step": 24017 + }, + { + "epoch": 0.77, + "learning_rate": 2.6903494607737923e-06, + "loss": 0.9707, + "step": 24018 + }, + { + "epoch": 0.77, + "learning_rate": 2.6896426521062503e-06, + "loss": 0.959, + "step": 24019 + }, + { + "epoch": 0.77, + "learning_rate": 2.688935921870747e-06, + "loss": 0.8677, + "step": 24020 + }, + { + "epoch": 0.77, + "learning_rate": 2.6882292700748714e-06, + "loss": 0.9263, + "step": 24021 + }, + { + "epoch": 0.77, + "learning_rate": 2.6875226967262014e-06, + "loss": 1.0122, + "step": 24022 + }, + { + "epoch": 0.77, + "learning_rate": 2.6868162018323217e-06, + "loss": 0.9751, + "step": 24023 + }, + { + "epoch": 0.77, + "learning_rate": 2.686109785400809e-06, + "loss": 0.9229, + "step": 24024 + }, + { + "epoch": 0.77, + "learning_rate": 2.6854034474392455e-06, + "loss": 1.021, + "step": 24025 + }, + { + "epoch": 0.77, + "learning_rate": 2.6846971879552054e-06, + "loss": 0.9043, + "step": 24026 + }, + { + "epoch": 0.77, + "learning_rate": 2.6839910069562738e-06, + "loss": 0.9043, + "step": 24027 + }, + { + "epoch": 0.77, + "learning_rate": 2.683284904450015e-06, + "loss": 0.8247, + "step": 24028 + }, + { + "epoch": 0.77, + "learning_rate": 2.6825788804440155e-06, + "loss": 0.9619, + "step": 24029 + }, + { + "epoch": 0.77, + "learning_rate": 2.6818729349458427e-06, + "loss": 0.9033, + "step": 24030 + }, + { + "epoch": 0.77, + "learning_rate": 2.6811670679630776e-06, + "loss": 0.9468, + "step": 24031 + }, + { + "epoch": 0.77, + "learning_rate": 2.6804612795032856e-06, + "loss": 0.9312, + "step": 24032 + }, + { + "epoch": 0.77, + "learning_rate": 2.679755569574047e-06, + "loss": 0.9497, + "step": 24033 + }, + { + "epoch": 0.77, + "learning_rate": 2.6790499381829292e-06, + "loss": 0.9053, + "step": 24034 + }, + { + "epoch": 0.77, + "learning_rate": 2.6783443853375035e-06, + "loss": 0.8271, + "step": 24035 + }, + { + "epoch": 0.77, + "learning_rate": 2.677638911045336e-06, + "loss": 0.9473, + "step": 24036 + }, + { + "epoch": 0.77, + "learning_rate": 2.6769335153140018e-06, + "loss": 0.8831, + "step": 24037 + }, + { + "epoch": 0.77, + "learning_rate": 2.676228198151064e-06, + "loss": 0.8706, + "step": 24038 + }, + { + "epoch": 0.77, + "learning_rate": 2.675522959564095e-06, + "loss": 0.8345, + "step": 24039 + }, + { + "epoch": 0.77, + "learning_rate": 2.674817799560655e-06, + "loss": 0.9775, + "step": 24040 + }, + { + "epoch": 0.77, + "learning_rate": 2.674112718148315e-06, + "loss": 0.4829, + "step": 24041 + }, + { + "epoch": 0.77, + "learning_rate": 2.6734077153346392e-06, + "loss": 1.0293, + "step": 24042 + }, + { + "epoch": 0.77, + "learning_rate": 2.6727027911271895e-06, + "loss": 1.0254, + "step": 24043 + }, + { + "epoch": 0.77, + "learning_rate": 2.6719979455335254e-06, + "loss": 0.877, + "step": 24044 + }, + { + "epoch": 0.77, + "learning_rate": 2.671293178561217e-06, + "loss": 0.8677, + "step": 24045 + }, + { + "epoch": 0.77, + "learning_rate": 2.6705884902178216e-06, + "loss": 0.8516, + "step": 24046 + }, + { + "epoch": 0.77, + "learning_rate": 2.6698838805108973e-06, + "loss": 0.9839, + "step": 24047 + }, + { + "epoch": 0.77, + "learning_rate": 2.669179349448009e-06, + "loss": 1.0146, + "step": 24048 + }, + { + "epoch": 0.77, + "learning_rate": 2.6684748970367124e-06, + "loss": 0.9414, + "step": 24049 + }, + { + "epoch": 0.77, + "learning_rate": 2.667770523284564e-06, + "loss": 0.8359, + "step": 24050 + }, + { + "epoch": 0.77, + "learning_rate": 2.6670662281991253e-06, + "loss": 0.8369, + "step": 24051 + }, + { + "epoch": 0.77, + "learning_rate": 2.6663620117879506e-06, + "loss": 0.9443, + "step": 24052 + }, + { + "epoch": 0.77, + "learning_rate": 2.6656578740585927e-06, + "loss": 0.9365, + "step": 24053 + }, + { + "epoch": 0.77, + "learning_rate": 2.6649538150186103e-06, + "loss": 0.8765, + "step": 24054 + }, + { + "epoch": 0.77, + "learning_rate": 2.664249834675553e-06, + "loss": 0.9336, + "step": 24055 + }, + { + "epoch": 0.77, + "learning_rate": 2.6635459330369794e-06, + "loss": 0.9678, + "step": 24056 + }, + { + "epoch": 0.77, + "learning_rate": 2.6628421101104385e-06, + "loss": 0.7976, + "step": 24057 + }, + { + "epoch": 0.77, + "learning_rate": 2.6621383659034805e-06, + "loss": 0.9604, + "step": 24058 + }, + { + "epoch": 0.77, + "learning_rate": 2.661434700423654e-06, + "loss": 0.9741, + "step": 24059 + }, + { + "epoch": 0.77, + "learning_rate": 2.660731113678515e-06, + "loss": 0.8989, + "step": 24060 + }, + { + "epoch": 0.77, + "learning_rate": 2.6600276056756045e-06, + "loss": 0.8511, + "step": 24061 + }, + { + "epoch": 0.77, + "learning_rate": 2.6593241764224777e-06, + "loss": 0.8965, + "step": 24062 + }, + { + "epoch": 0.77, + "learning_rate": 2.6586208259266755e-06, + "loss": 0.894, + "step": 24063 + }, + { + "epoch": 0.77, + "learning_rate": 2.6579175541957526e-06, + "loss": 0.8037, + "step": 24064 + }, + { + "epoch": 0.77, + "learning_rate": 2.6572143612372425e-06, + "loss": 0.957, + "step": 24065 + }, + { + "epoch": 0.77, + "learning_rate": 2.6565112470586983e-06, + "loss": 0.4487, + "step": 24066 + }, + { + "epoch": 0.77, + "learning_rate": 2.6558082116676588e-06, + "loss": 0.9089, + "step": 24067 + }, + { + "epoch": 0.77, + "learning_rate": 2.6551052550716716e-06, + "loss": 0.9209, + "step": 24068 + }, + { + "epoch": 0.77, + "learning_rate": 2.6544023772782736e-06, + "loss": 0.7769, + "step": 24069 + }, + { + "epoch": 0.77, + "learning_rate": 2.65369957829501e-06, + "loss": 0.9204, + "step": 24070 + }, + { + "epoch": 0.77, + "learning_rate": 2.6529968581294174e-06, + "loss": 0.9473, + "step": 24071 + }, + { + "epoch": 0.77, + "learning_rate": 2.652294216789044e-06, + "loss": 0.8574, + "step": 24072 + }, + { + "epoch": 0.77, + "learning_rate": 2.6515916542814146e-06, + "loss": 0.8677, + "step": 24073 + }, + { + "epoch": 0.77, + "learning_rate": 2.650889170614077e-06, + "loss": 0.9414, + "step": 24074 + }, + { + "epoch": 0.77, + "learning_rate": 2.6501867657945624e-06, + "loss": 0.9268, + "step": 24075 + }, + { + "epoch": 0.77, + "learning_rate": 2.6494844398304122e-06, + "loss": 0.9487, + "step": 24076 + }, + { + "epoch": 0.77, + "learning_rate": 2.6487821927291558e-06, + "loss": 0.9561, + "step": 24077 + }, + { + "epoch": 0.77, + "learning_rate": 2.648080024498334e-06, + "loss": 1.0361, + "step": 24078 + }, + { + "epoch": 0.77, + "learning_rate": 2.647377935145473e-06, + "loss": 0.9375, + "step": 24079 + }, + { + "epoch": 0.77, + "learning_rate": 2.646675924678116e-06, + "loss": 0.9424, + "step": 24080 + }, + { + "epoch": 0.77, + "learning_rate": 2.6459739931037808e-06, + "loss": 0.8091, + "step": 24081 + }, + { + "epoch": 0.77, + "learning_rate": 2.64527214043001e-06, + "loss": 0.9072, + "step": 24082 + }, + { + "epoch": 0.77, + "learning_rate": 2.644570366664326e-06, + "loss": 0.8374, + "step": 24083 + }, + { + "epoch": 0.77, + "learning_rate": 2.643868671814265e-06, + "loss": 0.7607, + "step": 24084 + }, + { + "epoch": 0.77, + "learning_rate": 2.64316705588735e-06, + "loss": 0.895, + "step": 24085 + }, + { + "epoch": 0.77, + "learning_rate": 2.6424655188911085e-06, + "loss": 0.9419, + "step": 24086 + }, + { + "epoch": 0.77, + "learning_rate": 2.641764060833073e-06, + "loss": 0.8862, + "step": 24087 + }, + { + "epoch": 0.77, + "learning_rate": 2.641062681720764e-06, + "loss": 0.8564, + "step": 24088 + }, + { + "epoch": 0.77, + "learning_rate": 2.640361381561706e-06, + "loss": 1.0015, + "step": 24089 + }, + { + "epoch": 0.77, + "learning_rate": 2.6396601603634265e-06, + "loss": 0.9653, + "step": 24090 + }, + { + "epoch": 0.77, + "learning_rate": 2.638959018133449e-06, + "loss": 0.9297, + "step": 24091 + }, + { + "epoch": 0.77, + "learning_rate": 2.638257954879291e-06, + "loss": 1.0205, + "step": 24092 + }, + { + "epoch": 0.77, + "learning_rate": 2.6375569706084804e-06, + "loss": 0.9824, + "step": 24093 + }, + { + "epoch": 0.77, + "learning_rate": 2.6368560653285326e-06, + "loss": 0.9019, + "step": 24094 + }, + { + "epoch": 0.77, + "learning_rate": 2.6361552390469724e-06, + "loss": 0.9736, + "step": 24095 + }, + { + "epoch": 0.77, + "learning_rate": 2.6354544917713165e-06, + "loss": 0.9292, + "step": 24096 + }, + { + "epoch": 0.77, + "learning_rate": 2.6347538235090842e-06, + "loss": 0.9556, + "step": 24097 + }, + { + "epoch": 0.77, + "learning_rate": 2.634053234267788e-06, + "loss": 0.9438, + "step": 24098 + }, + { + "epoch": 0.77, + "learning_rate": 2.6333527240549518e-06, + "loss": 0.833, + "step": 24099 + }, + { + "epoch": 0.77, + "learning_rate": 2.632652292878085e-06, + "loss": 0.979, + "step": 24100 + }, + { + "epoch": 0.77, + "learning_rate": 2.631951940744708e-06, + "loss": 1.0215, + "step": 24101 + }, + { + "epoch": 0.77, + "learning_rate": 2.631251667662332e-06, + "loss": 1.0581, + "step": 24102 + }, + { + "epoch": 0.77, + "learning_rate": 2.630551473638471e-06, + "loss": 0.8188, + "step": 24103 + }, + { + "epoch": 0.77, + "learning_rate": 2.629851358680633e-06, + "loss": 0.9351, + "step": 24104 + }, + { + "epoch": 0.77, + "learning_rate": 2.6291513227963363e-06, + "loss": 0.9734, + "step": 24105 + }, + { + "epoch": 0.77, + "learning_rate": 2.6284513659930856e-06, + "loss": 0.8716, + "step": 24106 + }, + { + "epoch": 0.77, + "learning_rate": 2.627751488278395e-06, + "loss": 0.8789, + "step": 24107 + }, + { + "epoch": 0.77, + "learning_rate": 2.62705168965977e-06, + "loss": 0.9785, + "step": 24108 + }, + { + "epoch": 0.77, + "learning_rate": 2.6263519701447228e-06, + "loss": 0.9731, + "step": 24109 + }, + { + "epoch": 0.77, + "learning_rate": 2.625652329740759e-06, + "loss": 0.8691, + "step": 24110 + }, + { + "epoch": 0.77, + "learning_rate": 2.624952768455383e-06, + "loss": 0.9326, + "step": 24111 + }, + { + "epoch": 0.77, + "learning_rate": 2.624253286296099e-06, + "loss": 0.9019, + "step": 24112 + }, + { + "epoch": 0.77, + "learning_rate": 2.6235538832704177e-06, + "loss": 0.9038, + "step": 24113 + }, + { + "epoch": 0.77, + "learning_rate": 2.6228545593858357e-06, + "loss": 0.9277, + "step": 24114 + }, + { + "epoch": 0.77, + "learning_rate": 2.6221553146498623e-06, + "loss": 0.874, + "step": 24115 + }, + { + "epoch": 0.77, + "learning_rate": 2.6214561490699953e-06, + "loss": 0.8364, + "step": 24116 + }, + { + "epoch": 0.77, + "learning_rate": 2.620757062653743e-06, + "loss": 0.9482, + "step": 24117 + }, + { + "epoch": 0.77, + "learning_rate": 2.620058055408594e-06, + "loss": 1.0083, + "step": 24118 + }, + { + "epoch": 0.77, + "learning_rate": 2.6193591273420572e-06, + "loss": 0.9634, + "step": 24119 + }, + { + "epoch": 0.77, + "learning_rate": 2.6186602784616267e-06, + "loss": 0.9307, + "step": 24120 + }, + { + "epoch": 0.77, + "learning_rate": 2.6179615087748046e-06, + "loss": 0.9277, + "step": 24121 + }, + { + "epoch": 0.77, + "learning_rate": 2.6172628182890825e-06, + "loss": 0.9229, + "step": 24122 + }, + { + "epoch": 0.77, + "learning_rate": 2.616564207011962e-06, + "loss": 0.8687, + "step": 24123 + }, + { + "epoch": 0.77, + "learning_rate": 2.615865674950934e-06, + "loss": 0.7891, + "step": 24124 + }, + { + "epoch": 0.77, + "learning_rate": 2.6151672221134993e-06, + "loss": 0.8237, + "step": 24125 + }, + { + "epoch": 0.77, + "learning_rate": 2.6144688485071425e-06, + "loss": 0.8589, + "step": 24126 + }, + { + "epoch": 0.77, + "learning_rate": 2.613770554139363e-06, + "loss": 0.7734, + "step": 24127 + }, + { + "epoch": 0.77, + "learning_rate": 2.6130723390176494e-06, + "loss": 0.9736, + "step": 24128 + }, + { + "epoch": 0.77, + "learning_rate": 2.6123742031494957e-06, + "loss": 0.9727, + "step": 24129 + }, + { + "epoch": 0.77, + "learning_rate": 2.611676146542391e-06, + "loss": 0.8638, + "step": 24130 + }, + { + "epoch": 0.77, + "learning_rate": 2.6109781692038215e-06, + "loss": 0.9331, + "step": 24131 + }, + { + "epoch": 0.77, + "learning_rate": 2.610280271141282e-06, + "loss": 0.9131, + "step": 24132 + }, + { + "epoch": 0.77, + "learning_rate": 2.609582452362256e-06, + "loss": 0.9927, + "step": 24133 + }, + { + "epoch": 0.77, + "learning_rate": 2.608884712874228e-06, + "loss": 0.9556, + "step": 24134 + }, + { + "epoch": 0.77, + "learning_rate": 2.6081870526846897e-06, + "loss": 0.9189, + "step": 24135 + }, + { + "epoch": 0.77, + "learning_rate": 2.6074894718011234e-06, + "loss": 0.9873, + "step": 24136 + }, + { + "epoch": 0.77, + "learning_rate": 2.6067919702310118e-06, + "loss": 0.9897, + "step": 24137 + }, + { + "epoch": 0.77, + "learning_rate": 2.606094547981841e-06, + "loss": 0.9165, + "step": 24138 + }, + { + "epoch": 0.77, + "learning_rate": 2.605397205061091e-06, + "loss": 1.0908, + "step": 24139 + }, + { + "epoch": 0.77, + "learning_rate": 2.6046999414762476e-06, + "loss": 0.9731, + "step": 24140 + }, + { + "epoch": 0.77, + "learning_rate": 2.604002757234789e-06, + "loss": 0.9229, + "step": 24141 + }, + { + "epoch": 0.77, + "learning_rate": 2.603305652344196e-06, + "loss": 0.8711, + "step": 24142 + }, + { + "epoch": 0.77, + "learning_rate": 2.6026086268119434e-06, + "loss": 0.9058, + "step": 24143 + }, + { + "epoch": 0.77, + "learning_rate": 2.601911680645516e-06, + "loss": 0.8677, + "step": 24144 + }, + { + "epoch": 0.77, + "learning_rate": 2.6012148138523854e-06, + "loss": 1.002, + "step": 24145 + }, + { + "epoch": 0.77, + "learning_rate": 2.6005180264400343e-06, + "loss": 0.8682, + "step": 24146 + }, + { + "epoch": 0.77, + "learning_rate": 2.599821318415936e-06, + "loss": 1.0278, + "step": 24147 + }, + { + "epoch": 0.77, + "learning_rate": 2.5991246897875644e-06, + "loss": 0.9287, + "step": 24148 + }, + { + "epoch": 0.77, + "learning_rate": 2.5984281405623913e-06, + "loss": 0.7954, + "step": 24149 + }, + { + "epoch": 0.77, + "learning_rate": 2.5977316707478962e-06, + "loss": 0.8145, + "step": 24150 + }, + { + "epoch": 0.77, + "learning_rate": 2.597035280351544e-06, + "loss": 0.9414, + "step": 24151 + }, + { + "epoch": 0.77, + "learning_rate": 2.5963389693808127e-06, + "loss": 1.0386, + "step": 24152 + }, + { + "epoch": 0.77, + "learning_rate": 2.595642737843168e-06, + "loss": 0.8906, + "step": 24153 + }, + { + "epoch": 0.77, + "learning_rate": 2.5949465857460842e-06, + "loss": 0.8755, + "step": 24154 + }, + { + "epoch": 0.77, + "learning_rate": 2.5942505130970296e-06, + "loss": 1.0024, + "step": 24155 + }, + { + "epoch": 0.77, + "learning_rate": 2.5935545199034694e-06, + "loss": 0.915, + "step": 24156 + }, + { + "epoch": 0.77, + "learning_rate": 2.592858606172869e-06, + "loss": 0.8721, + "step": 24157 + }, + { + "epoch": 0.77, + "learning_rate": 2.592162771912702e-06, + "loss": 0.9053, + "step": 24158 + }, + { + "epoch": 0.77, + "learning_rate": 2.591467017130426e-06, + "loss": 0.7837, + "step": 24159 + }, + { + "epoch": 0.77, + "learning_rate": 2.590771341833512e-06, + "loss": 0.9814, + "step": 24160 + }, + { + "epoch": 0.77, + "learning_rate": 2.5900757460294203e-06, + "loss": 0.9771, + "step": 24161 + }, + { + "epoch": 0.77, + "learning_rate": 2.5893802297256166e-06, + "loss": 0.8916, + "step": 24162 + }, + { + "epoch": 0.77, + "learning_rate": 2.5886847929295613e-06, + "loss": 0.9482, + "step": 24163 + }, + { + "epoch": 0.77, + "learning_rate": 2.5879894356487166e-06, + "loss": 0.9326, + "step": 24164 + }, + { + "epoch": 0.77, + "learning_rate": 2.5872941578905388e-06, + "loss": 0.6895, + "step": 24165 + }, + { + "epoch": 0.77, + "learning_rate": 2.586598959662493e-06, + "loss": 1.0215, + "step": 24166 + }, + { + "epoch": 0.77, + "learning_rate": 2.5859038409720326e-06, + "loss": 0.9402, + "step": 24167 + }, + { + "epoch": 0.77, + "learning_rate": 2.585208801826622e-06, + "loss": 1.0415, + "step": 24168 + }, + { + "epoch": 0.77, + "learning_rate": 2.5845138422337146e-06, + "loss": 0.8896, + "step": 24169 + }, + { + "epoch": 0.77, + "learning_rate": 2.583818962200767e-06, + "loss": 0.855, + "step": 24170 + }, + { + "epoch": 0.77, + "learning_rate": 2.583124161735231e-06, + "loss": 1.0498, + "step": 24171 + }, + { + "epoch": 0.77, + "learning_rate": 2.582429440844566e-06, + "loss": 0.9263, + "step": 24172 + }, + { + "epoch": 0.77, + "learning_rate": 2.5817347995362217e-06, + "loss": 0.8862, + "step": 24173 + }, + { + "epoch": 0.77, + "learning_rate": 2.581040237817656e-06, + "loss": 0.9487, + "step": 24174 + }, + { + "epoch": 0.77, + "learning_rate": 2.5803457556963176e-06, + "loss": 0.9556, + "step": 24175 + }, + { + "epoch": 0.77, + "learning_rate": 2.579651353179654e-06, + "loss": 0.8438, + "step": 24176 + }, + { + "epoch": 0.77, + "learning_rate": 2.578957030275122e-06, + "loss": 0.9795, + "step": 24177 + }, + { + "epoch": 0.77, + "learning_rate": 2.5782627869901687e-06, + "loss": 0.9424, + "step": 24178 + }, + { + "epoch": 0.77, + "learning_rate": 2.5775686233322384e-06, + "loss": 0.8862, + "step": 24179 + }, + { + "epoch": 0.77, + "learning_rate": 2.5768745393087845e-06, + "loss": 0.9604, + "step": 24180 + }, + { + "epoch": 0.77, + "learning_rate": 2.5761805349272515e-06, + "loss": 1.0024, + "step": 24181 + }, + { + "epoch": 0.77, + "learning_rate": 2.5754866101950828e-06, + "loss": 0.8789, + "step": 24182 + }, + { + "epoch": 0.77, + "learning_rate": 2.574792765119728e-06, + "loss": 1.0449, + "step": 24183 + }, + { + "epoch": 0.77, + "learning_rate": 2.5740989997086264e-06, + "loss": 0.8716, + "step": 24184 + }, + { + "epoch": 0.77, + "learning_rate": 2.5734053139692296e-06, + "loss": 0.9429, + "step": 24185 + }, + { + "epoch": 0.77, + "learning_rate": 2.572711707908968e-06, + "loss": 0.7485, + "step": 24186 + }, + { + "epoch": 0.77, + "learning_rate": 2.5720181815352942e-06, + "loss": 0.9751, + "step": 24187 + }, + { + "epoch": 0.77, + "learning_rate": 2.57132473485564e-06, + "loss": 1.0015, + "step": 24188 + }, + { + "epoch": 0.77, + "learning_rate": 2.5706313678774533e-06, + "loss": 0.8228, + "step": 24189 + }, + { + "epoch": 0.77, + "learning_rate": 2.5699380806081665e-06, + "loss": 0.9863, + "step": 24190 + }, + { + "epoch": 0.77, + "learning_rate": 2.569244873055223e-06, + "loss": 1.0044, + "step": 24191 + }, + { + "epoch": 0.77, + "learning_rate": 2.5685517452260566e-06, + "loss": 0.8247, + "step": 24192 + }, + { + "epoch": 0.77, + "learning_rate": 2.5678586971281103e-06, + "loss": 0.8701, + "step": 24193 + }, + { + "epoch": 0.77, + "learning_rate": 2.567165728768809e-06, + "loss": 0.7568, + "step": 24194 + }, + { + "epoch": 0.77, + "learning_rate": 2.566472840155596e-06, + "loss": 0.9697, + "step": 24195 + }, + { + "epoch": 0.77, + "learning_rate": 2.5657800312959e-06, + "loss": 0.9316, + "step": 24196 + }, + { + "epoch": 0.77, + "learning_rate": 2.565087302197159e-06, + "loss": 0.9507, + "step": 24197 + }, + { + "epoch": 0.77, + "learning_rate": 2.5643946528667994e-06, + "loss": 1.0093, + "step": 24198 + }, + { + "epoch": 0.77, + "learning_rate": 2.5637020833122593e-06, + "loss": 0.9292, + "step": 24199 + }, + { + "epoch": 0.77, + "learning_rate": 2.563009593540966e-06, + "loss": 0.7949, + "step": 24200 + }, + { + "epoch": 0.77, + "learning_rate": 2.562317183560349e-06, + "loss": 0.9673, + "step": 24201 + }, + { + "epoch": 0.77, + "learning_rate": 2.561624853377833e-06, + "loss": 0.8774, + "step": 24202 + }, + { + "epoch": 0.77, + "learning_rate": 2.560932603000854e-06, + "loss": 0.7754, + "step": 24203 + }, + { + "epoch": 0.77, + "learning_rate": 2.560240432436831e-06, + "loss": 0.8716, + "step": 24204 + }, + { + "epoch": 0.77, + "learning_rate": 2.5595483416931975e-06, + "loss": 0.9565, + "step": 24205 + }, + { + "epoch": 0.77, + "learning_rate": 2.5588563307773738e-06, + "loss": 0.9111, + "step": 24206 + }, + { + "epoch": 0.77, + "learning_rate": 2.5581643996967876e-06, + "loss": 0.9531, + "step": 24207 + }, + { + "epoch": 0.77, + "learning_rate": 2.5574725484588623e-06, + "loss": 0.9028, + "step": 24208 + }, + { + "epoch": 0.77, + "learning_rate": 2.5567807770710198e-06, + "loss": 0.8799, + "step": 24209 + }, + { + "epoch": 0.77, + "learning_rate": 2.5560890855406773e-06, + "loss": 0.874, + "step": 24210 + }, + { + "epoch": 0.77, + "learning_rate": 2.555397473875264e-06, + "loss": 0.8857, + "step": 24211 + }, + { + "epoch": 0.77, + "learning_rate": 2.554705942082194e-06, + "loss": 0.9917, + "step": 24212 + }, + { + "epoch": 0.77, + "learning_rate": 2.5540144901688923e-06, + "loss": 0.9443, + "step": 24213 + }, + { + "epoch": 0.77, + "learning_rate": 2.553323118142774e-06, + "loss": 0.9282, + "step": 24214 + }, + { + "epoch": 0.77, + "learning_rate": 2.552631826011257e-06, + "loss": 0.894, + "step": 24215 + }, + { + "epoch": 0.77, + "learning_rate": 2.551940613781755e-06, + "loss": 0.9741, + "step": 24216 + }, + { + "epoch": 0.77, + "learning_rate": 2.5512494814616906e-06, + "loss": 0.7983, + "step": 24217 + }, + { + "epoch": 0.77, + "learning_rate": 2.5505584290584716e-06, + "loss": 0.9351, + "step": 24218 + }, + { + "epoch": 0.77, + "learning_rate": 2.54986745657952e-06, + "loss": 0.9033, + "step": 24219 + }, + { + "epoch": 0.77, + "learning_rate": 2.549176564032245e-06, + "loss": 0.8936, + "step": 24220 + }, + { + "epoch": 0.77, + "learning_rate": 2.548485751424056e-06, + "loss": 0.8652, + "step": 24221 + }, + { + "epoch": 0.77, + "learning_rate": 2.547795018762371e-06, + "loss": 0.9609, + "step": 24222 + }, + { + "epoch": 0.77, + "learning_rate": 2.547104366054598e-06, + "loss": 0.918, + "step": 24223 + }, + { + "epoch": 0.77, + "learning_rate": 2.5464137933081466e-06, + "loss": 0.9326, + "step": 24224 + }, + { + "epoch": 0.77, + "learning_rate": 2.545723300530423e-06, + "loss": 0.8608, + "step": 24225 + }, + { + "epoch": 0.77, + "learning_rate": 2.545032887728841e-06, + "loss": 0.9004, + "step": 24226 + }, + { + "epoch": 0.77, + "learning_rate": 2.5443425549108026e-06, + "loss": 0.8555, + "step": 24227 + }, + { + "epoch": 0.77, + "learning_rate": 2.54365230208372e-06, + "loss": 1.0024, + "step": 24228 + }, + { + "epoch": 0.77, + "learning_rate": 2.542962129254993e-06, + "loss": 0.9556, + "step": 24229 + }, + { + "epoch": 0.77, + "learning_rate": 2.542272036432035e-06, + "loss": 0.9189, + "step": 24230 + }, + { + "epoch": 0.77, + "learning_rate": 2.541582023622239e-06, + "loss": 0.9966, + "step": 24231 + }, + { + "epoch": 0.77, + "learning_rate": 2.540892090833016e-06, + "loss": 0.5317, + "step": 24232 + }, + { + "epoch": 0.77, + "learning_rate": 2.5402022380717616e-06, + "loss": 0.9736, + "step": 24233 + }, + { + "epoch": 0.78, + "learning_rate": 2.5395124653458857e-06, + "loss": 0.9604, + "step": 24234 + }, + { + "epoch": 0.78, + "learning_rate": 2.53882277266278e-06, + "loss": 0.8618, + "step": 24235 + }, + { + "epoch": 0.78, + "learning_rate": 2.538133160029852e-06, + "loss": 0.9683, + "step": 24236 + }, + { + "epoch": 0.78, + "learning_rate": 2.5374436274544933e-06, + "loss": 0.8071, + "step": 24237 + }, + { + "epoch": 0.78, + "learning_rate": 2.5367541749441114e-06, + "loss": 0.9067, + "step": 24238 + }, + { + "epoch": 0.78, + "learning_rate": 2.5360648025060918e-06, + "loss": 0.8101, + "step": 24239 + }, + { + "epoch": 0.78, + "learning_rate": 2.5353755101478396e-06, + "loss": 0.9873, + "step": 24240 + }, + { + "epoch": 0.78, + "learning_rate": 2.534686297876743e-06, + "loss": 0.9751, + "step": 24241 + }, + { + "epoch": 0.78, + "learning_rate": 2.5339971657002027e-06, + "loss": 0.9111, + "step": 24242 + }, + { + "epoch": 0.78, + "learning_rate": 2.5333081136256076e-06, + "loss": 0.9048, + "step": 24243 + }, + { + "epoch": 0.78, + "learning_rate": 2.5326191416603553e-06, + "loss": 0.9375, + "step": 24244 + }, + { + "epoch": 0.78, + "learning_rate": 2.5319302498118326e-06, + "loss": 0.9302, + "step": 24245 + }, + { + "epoch": 0.78, + "learning_rate": 2.5312414380874386e-06, + "loss": 1.0542, + "step": 24246 + }, + { + "epoch": 0.78, + "learning_rate": 2.530552706494552e-06, + "loss": 0.7925, + "step": 24247 + }, + { + "epoch": 0.78, + "learning_rate": 2.5298640550405706e-06, + "loss": 0.8633, + "step": 24248 + }, + { + "epoch": 0.78, + "learning_rate": 2.5291754837328787e-06, + "loss": 0.9385, + "step": 24249 + }, + { + "epoch": 0.78, + "learning_rate": 2.528486992578868e-06, + "loss": 0.8359, + "step": 24250 + }, + { + "epoch": 0.78, + "learning_rate": 2.52779858158592e-06, + "loss": 0.9961, + "step": 24251 + }, + { + "epoch": 0.78, + "learning_rate": 2.5271102507614255e-06, + "loss": 0.835, + "step": 24252 + }, + { + "epoch": 0.78, + "learning_rate": 2.5264220001127683e-06, + "loss": 0.8059, + "step": 24253 + }, + { + "epoch": 0.78, + "learning_rate": 2.525733829647331e-06, + "loss": 1.0244, + "step": 24254 + }, + { + "epoch": 0.78, + "learning_rate": 2.5250457393724946e-06, + "loss": 0.8286, + "step": 24255 + }, + { + "epoch": 0.78, + "learning_rate": 2.5243577292956477e-06, + "loss": 1.0161, + "step": 24256 + }, + { + "epoch": 0.78, + "learning_rate": 2.5236697994241665e-06, + "loss": 0.9336, + "step": 24257 + }, + { + "epoch": 0.78, + "learning_rate": 2.522981949765436e-06, + "loss": 0.9854, + "step": 24258 + }, + { + "epoch": 0.78, + "learning_rate": 2.522294180326834e-06, + "loss": 0.8716, + "step": 24259 + }, + { + "epoch": 0.78, + "learning_rate": 2.521606491115738e-06, + "loss": 0.9409, + "step": 24260 + }, + { + "epoch": 0.78, + "learning_rate": 2.520918882139529e-06, + "loss": 0.8975, + "step": 24261 + }, + { + "epoch": 0.78, + "learning_rate": 2.5202313534055832e-06, + "loss": 0.9473, + "step": 24262 + }, + { + "epoch": 0.78, + "learning_rate": 2.5195439049212777e-06, + "loss": 0.9697, + "step": 24263 + }, + { + "epoch": 0.78, + "learning_rate": 2.518856536693983e-06, + "loss": 0.9409, + "step": 24264 + }, + { + "epoch": 0.78, + "learning_rate": 2.5181692487310827e-06, + "loss": 1.0137, + "step": 24265 + }, + { + "epoch": 0.78, + "learning_rate": 2.5174820410399414e-06, + "loss": 0.853, + "step": 24266 + }, + { + "epoch": 0.78, + "learning_rate": 2.51679491362794e-06, + "loss": 0.939, + "step": 24267 + }, + { + "epoch": 0.78, + "learning_rate": 2.5161078665024475e-06, + "loss": 0.9678, + "step": 24268 + }, + { + "epoch": 0.78, + "learning_rate": 2.515420899670834e-06, + "loss": 1.0122, + "step": 24269 + }, + { + "epoch": 0.78, + "learning_rate": 2.5147340131404676e-06, + "loss": 0.9443, + "step": 24270 + }, + { + "epoch": 0.78, + "learning_rate": 2.514047206918725e-06, + "loss": 0.8584, + "step": 24271 + }, + { + "epoch": 0.78, + "learning_rate": 2.5133604810129673e-06, + "loss": 0.9326, + "step": 24272 + }, + { + "epoch": 0.78, + "learning_rate": 2.512673835430569e-06, + "loss": 0.8115, + "step": 24273 + }, + { + "epoch": 0.78, + "learning_rate": 2.5119872701788904e-06, + "loss": 0.853, + "step": 24274 + }, + { + "epoch": 0.78, + "learning_rate": 2.511300785265305e-06, + "loss": 0.8892, + "step": 24275 + }, + { + "epoch": 0.78, + "learning_rate": 2.5106143806971727e-06, + "loss": 0.8076, + "step": 24276 + }, + { + "epoch": 0.78, + "learning_rate": 2.5099280564818605e-06, + "loss": 0.9536, + "step": 24277 + }, + { + "epoch": 0.78, + "learning_rate": 2.509241812626727e-06, + "loss": 0.8604, + "step": 24278 + }, + { + "epoch": 0.78, + "learning_rate": 2.5085556491391416e-06, + "loss": 0.8755, + "step": 24279 + }, + { + "epoch": 0.78, + "learning_rate": 2.5078695660264597e-06, + "loss": 0.9863, + "step": 24280 + }, + { + "epoch": 0.78, + "learning_rate": 2.5071835632960495e-06, + "loss": 0.9663, + "step": 24281 + }, + { + "epoch": 0.78, + "learning_rate": 2.5064976409552634e-06, + "loss": 0.8633, + "step": 24282 + }, + { + "epoch": 0.78, + "learning_rate": 2.5058117990114696e-06, + "loss": 0.8594, + "step": 24283 + }, + { + "epoch": 0.78, + "learning_rate": 2.5051260374720153e-06, + "loss": 0.9292, + "step": 24284 + }, + { + "epoch": 0.78, + "learning_rate": 2.5044403563442676e-06, + "loss": 0.9434, + "step": 24285 + }, + { + "epoch": 0.78, + "learning_rate": 2.5037547556355758e-06, + "loss": 0.9966, + "step": 24286 + }, + { + "epoch": 0.78, + "learning_rate": 2.5030692353533014e-06, + "loss": 0.9102, + "step": 24287 + }, + { + "epoch": 0.78, + "learning_rate": 2.5023837955047945e-06, + "loss": 0.9907, + "step": 24288 + }, + { + "epoch": 0.78, + "learning_rate": 2.501698436097415e-06, + "loss": 0.9492, + "step": 24289 + }, + { + "epoch": 0.78, + "learning_rate": 2.501013157138509e-06, + "loss": 0.9258, + "step": 24290 + }, + { + "epoch": 0.78, + "learning_rate": 2.5003279586354378e-06, + "loss": 0.9941, + "step": 24291 + }, + { + "epoch": 0.78, + "learning_rate": 2.499642840595542e-06, + "loss": 0.9805, + "step": 24292 + }, + { + "epoch": 0.78, + "learning_rate": 2.4989578030261806e-06, + "loss": 0.9922, + "step": 24293 + }, + { + "epoch": 0.78, + "learning_rate": 2.4982728459346974e-06, + "loss": 0.9287, + "step": 24294 + }, + { + "epoch": 0.78, + "learning_rate": 2.4975879693284467e-06, + "loss": 0.9668, + "step": 24295 + }, + { + "epoch": 0.78, + "learning_rate": 2.496903173214771e-06, + "loss": 0.9111, + "step": 24296 + }, + { + "epoch": 0.78, + "learning_rate": 2.496218457601023e-06, + "loss": 0.8574, + "step": 24297 + }, + { + "epoch": 0.78, + "learning_rate": 2.495533822494546e-06, + "loss": 0.9805, + "step": 24298 + }, + { + "epoch": 0.78, + "learning_rate": 2.494849267902686e-06, + "loss": 0.9819, + "step": 24299 + }, + { + "epoch": 0.78, + "learning_rate": 2.4941647938327827e-06, + "loss": 0.8569, + "step": 24300 + }, + { + "epoch": 0.78, + "learning_rate": 2.4934804002921887e-06, + "loss": 0.9922, + "step": 24301 + }, + { + "epoch": 0.78, + "learning_rate": 2.492796087288242e-06, + "loss": 0.8936, + "step": 24302 + }, + { + "epoch": 0.78, + "learning_rate": 2.492111854828282e-06, + "loss": 0.8555, + "step": 24303 + }, + { + "epoch": 0.78, + "learning_rate": 2.4914277029196545e-06, + "loss": 0.9673, + "step": 24304 + }, + { + "epoch": 0.78, + "learning_rate": 2.4907436315696954e-06, + "loss": 0.979, + "step": 24305 + }, + { + "epoch": 0.78, + "learning_rate": 2.490059640785749e-06, + "loss": 0.9331, + "step": 24306 + }, + { + "epoch": 0.78, + "learning_rate": 2.4893757305751507e-06, + "loss": 0.9404, + "step": 24307 + }, + { + "epoch": 0.78, + "learning_rate": 2.4886919009452395e-06, + "loss": 1.0469, + "step": 24308 + }, + { + "epoch": 0.78, + "learning_rate": 2.4880081519033473e-06, + "loss": 0.8994, + "step": 24309 + }, + { + "epoch": 0.78, + "learning_rate": 2.4873244834568166e-06, + "loss": 0.9312, + "step": 24310 + }, + { + "epoch": 0.78, + "learning_rate": 2.4866408956129773e-06, + "loss": 0.9971, + "step": 24311 + }, + { + "epoch": 0.78, + "learning_rate": 2.485957388379169e-06, + "loss": 0.9033, + "step": 24312 + }, + { + "epoch": 0.78, + "learning_rate": 2.4852739617627176e-06, + "loss": 0.9087, + "step": 24313 + }, + { + "epoch": 0.78, + "learning_rate": 2.4845906157709667e-06, + "loss": 1.0518, + "step": 24314 + }, + { + "epoch": 0.78, + "learning_rate": 2.4839073504112343e-06, + "loss": 0.9683, + "step": 24315 + }, + { + "epoch": 0.78, + "learning_rate": 2.4832241656908605e-06, + "loss": 1.0142, + "step": 24316 + }, + { + "epoch": 0.78, + "learning_rate": 2.4825410616171695e-06, + "loss": 0.9731, + "step": 24317 + }, + { + "epoch": 0.78, + "learning_rate": 2.4818580381974967e-06, + "loss": 0.8281, + "step": 24318 + }, + { + "epoch": 0.78, + "learning_rate": 2.481175095439162e-06, + "loss": 0.9194, + "step": 24319 + }, + { + "epoch": 0.78, + "learning_rate": 2.4804922333495018e-06, + "loss": 0.9268, + "step": 24320 + }, + { + "epoch": 0.78, + "learning_rate": 2.4798094519358363e-06, + "loss": 0.9683, + "step": 24321 + }, + { + "epoch": 0.78, + "learning_rate": 2.479126751205494e-06, + "loss": 0.8965, + "step": 24322 + }, + { + "epoch": 0.78, + "learning_rate": 2.4784441311657935e-06, + "loss": 0.9565, + "step": 24323 + }, + { + "epoch": 0.78, + "learning_rate": 2.477761591824066e-06, + "loss": 0.9585, + "step": 24324 + }, + { + "epoch": 0.78, + "learning_rate": 2.4770791331876296e-06, + "loss": 0.9058, + "step": 24325 + }, + { + "epoch": 0.78, + "learning_rate": 2.47639675526381e-06, + "loss": 0.9839, + "step": 24326 + }, + { + "epoch": 0.78, + "learning_rate": 2.4757144580599246e-06, + "loss": 0.9243, + "step": 24327 + }, + { + "epoch": 0.78, + "learning_rate": 2.475032241583298e-06, + "loss": 0.9634, + "step": 24328 + }, + { + "epoch": 0.78, + "learning_rate": 2.474350105841248e-06, + "loss": 0.9229, + "step": 24329 + }, + { + "epoch": 0.78, + "learning_rate": 2.4736680508410902e-06, + "loss": 1.0063, + "step": 24330 + }, + { + "epoch": 0.78, + "learning_rate": 2.4729860765901437e-06, + "loss": 0.9897, + "step": 24331 + }, + { + "epoch": 0.78, + "learning_rate": 2.4723041830957283e-06, + "loss": 0.9497, + "step": 24332 + }, + { + "epoch": 0.78, + "learning_rate": 2.4716223703651544e-06, + "loss": 1.001, + "step": 24333 + }, + { + "epoch": 0.78, + "learning_rate": 2.470940638405744e-06, + "loss": 1.0215, + "step": 24334 + }, + { + "epoch": 0.78, + "learning_rate": 2.470258987224805e-06, + "loss": 1.0093, + "step": 24335 + }, + { + "epoch": 0.78, + "learning_rate": 2.4695774168296593e-06, + "loss": 1.0542, + "step": 24336 + }, + { + "epoch": 0.78, + "learning_rate": 2.4688959272276083e-06, + "loss": 0.8354, + "step": 24337 + }, + { + "epoch": 0.78, + "learning_rate": 2.468214518425971e-06, + "loss": 1.0322, + "step": 24338 + }, + { + "epoch": 0.78, + "learning_rate": 2.4675331904320533e-06, + "loss": 0.9536, + "step": 24339 + }, + { + "epoch": 0.78, + "learning_rate": 2.4668519432531713e-06, + "loss": 0.9849, + "step": 24340 + }, + { + "epoch": 0.78, + "learning_rate": 2.466170776896627e-06, + "loss": 0.9111, + "step": 24341 + }, + { + "epoch": 0.78, + "learning_rate": 2.4654896913697357e-06, + "loss": 1.0542, + "step": 24342 + }, + { + "epoch": 0.78, + "learning_rate": 2.4648086866798016e-06, + "loss": 0.9512, + "step": 24343 + }, + { + "epoch": 0.78, + "learning_rate": 2.4641277628341296e-06, + "loss": 0.897, + "step": 24344 + }, + { + "epoch": 0.78, + "learning_rate": 2.4634469198400246e-06, + "loss": 0.9937, + "step": 24345 + }, + { + "epoch": 0.78, + "learning_rate": 2.4627661577047957e-06, + "loss": 0.9089, + "step": 24346 + }, + { + "epoch": 0.78, + "learning_rate": 2.4620854764357438e-06, + "loss": 0.9722, + "step": 24347 + }, + { + "epoch": 0.78, + "learning_rate": 2.46140487604017e-06, + "loss": 1.0283, + "step": 24348 + }, + { + "epoch": 0.78, + "learning_rate": 2.4607243565253813e-06, + "loss": 0.8862, + "step": 24349 + }, + { + "epoch": 0.78, + "learning_rate": 2.4600439178986724e-06, + "loss": 0.8247, + "step": 24350 + }, + { + "epoch": 0.78, + "learning_rate": 2.4593635601673518e-06, + "loss": 0.9424, + "step": 24351 + }, + { + "epoch": 0.78, + "learning_rate": 2.458683283338713e-06, + "loss": 0.8799, + "step": 24352 + }, + { + "epoch": 0.78, + "learning_rate": 2.458003087420058e-06, + "loss": 0.9082, + "step": 24353 + }, + { + "epoch": 0.78, + "learning_rate": 2.4573229724186787e-06, + "loss": 0.8672, + "step": 24354 + }, + { + "epoch": 0.78, + "learning_rate": 2.456642938341879e-06, + "loss": 0.9761, + "step": 24355 + }, + { + "epoch": 0.78, + "learning_rate": 2.4559629851969493e-06, + "loss": 0.8804, + "step": 24356 + }, + { + "epoch": 0.78, + "learning_rate": 2.45528311299119e-06, + "loss": 0.7739, + "step": 24357 + }, + { + "epoch": 0.78, + "learning_rate": 2.454603321731891e-06, + "loss": 0.9692, + "step": 24358 + }, + { + "epoch": 0.78, + "learning_rate": 2.453923611426352e-06, + "loss": 0.9897, + "step": 24359 + }, + { + "epoch": 0.78, + "learning_rate": 2.4532439820818566e-06, + "loss": 0.5288, + "step": 24360 + }, + { + "epoch": 0.78, + "learning_rate": 2.452564433705703e-06, + "loss": 1.0181, + "step": 24361 + }, + { + "epoch": 0.78, + "learning_rate": 2.4518849663051767e-06, + "loss": 0.915, + "step": 24362 + }, + { + "epoch": 0.78, + "learning_rate": 2.4512055798875744e-06, + "loss": 1.0166, + "step": 24363 + }, + { + "epoch": 0.78, + "learning_rate": 2.450526274460179e-06, + "loss": 0.8853, + "step": 24364 + }, + { + "epoch": 0.78, + "learning_rate": 2.449847050030284e-06, + "loss": 0.9214, + "step": 24365 + }, + { + "epoch": 0.78, + "learning_rate": 2.449167906605174e-06, + "loss": 0.9175, + "step": 24366 + }, + { + "epoch": 0.78, + "learning_rate": 2.448488844192135e-06, + "loss": 0.8784, + "step": 24367 + }, + { + "epoch": 0.78, + "learning_rate": 2.4478098627984515e-06, + "loss": 0.9873, + "step": 24368 + }, + { + "epoch": 0.78, + "learning_rate": 2.447130962431411e-06, + "loss": 0.9985, + "step": 24369 + }, + { + "epoch": 0.78, + "learning_rate": 2.4464521430982947e-06, + "loss": 0.9907, + "step": 24370 + }, + { + "epoch": 0.78, + "learning_rate": 2.4457734048063885e-06, + "loss": 0.8613, + "step": 24371 + }, + { + "epoch": 0.78, + "learning_rate": 2.4450947475629718e-06, + "loss": 0.8462, + "step": 24372 + }, + { + "epoch": 0.78, + "learning_rate": 2.444416171375329e-06, + "loss": 0.9312, + "step": 24373 + }, + { + "epoch": 0.78, + "learning_rate": 2.4437376762507394e-06, + "loss": 0.8721, + "step": 24374 + }, + { + "epoch": 0.78, + "learning_rate": 2.443059262196481e-06, + "loss": 0.9028, + "step": 24375 + }, + { + "epoch": 0.78, + "learning_rate": 2.44238092921983e-06, + "loss": 0.811, + "step": 24376 + }, + { + "epoch": 0.78, + "learning_rate": 2.4417026773280704e-06, + "loss": 0.8506, + "step": 24377 + }, + { + "epoch": 0.78, + "learning_rate": 2.4410245065284733e-06, + "loss": 0.9512, + "step": 24378 + }, + { + "epoch": 0.78, + "learning_rate": 2.44034641682832e-06, + "loss": 1.0708, + "step": 24379 + }, + { + "epoch": 0.78, + "learning_rate": 2.43966840823488e-06, + "loss": 1.0801, + "step": 24380 + }, + { + "epoch": 0.78, + "learning_rate": 2.4389904807554365e-06, + "loss": 0.9395, + "step": 24381 + }, + { + "epoch": 0.78, + "learning_rate": 2.438312634397252e-06, + "loss": 0.8154, + "step": 24382 + }, + { + "epoch": 0.78, + "learning_rate": 2.4376348691676066e-06, + "loss": 0.8882, + "step": 24383 + }, + { + "epoch": 0.78, + "learning_rate": 2.436957185073766e-06, + "loss": 1.022, + "step": 24384 + }, + { + "epoch": 0.78, + "learning_rate": 2.436279582123008e-06, + "loss": 0.8994, + "step": 24385 + }, + { + "epoch": 0.78, + "learning_rate": 2.4356020603225983e-06, + "loss": 0.9517, + "step": 24386 + }, + { + "epoch": 0.78, + "learning_rate": 2.434924619679804e-06, + "loss": 0.8496, + "step": 24387 + }, + { + "epoch": 0.78, + "learning_rate": 2.4342472602018986e-06, + "loss": 0.9443, + "step": 24388 + }, + { + "epoch": 0.78, + "learning_rate": 2.433569981896148e-06, + "loss": 0.9966, + "step": 24389 + }, + { + "epoch": 0.78, + "learning_rate": 2.4328927847698125e-06, + "loss": 0.8896, + "step": 24390 + }, + { + "epoch": 0.78, + "learning_rate": 2.4322156688301667e-06, + "loss": 0.8311, + "step": 24391 + }, + { + "epoch": 0.78, + "learning_rate": 2.4315386340844717e-06, + "loss": 0.8643, + "step": 24392 + }, + { + "epoch": 0.78, + "learning_rate": 2.4308616805399867e-06, + "loss": 0.9194, + "step": 24393 + }, + { + "epoch": 0.78, + "learning_rate": 2.430184808203981e-06, + "loss": 0.8608, + "step": 24394 + }, + { + "epoch": 0.78, + "learning_rate": 2.429508017083714e-06, + "loss": 0.9678, + "step": 24395 + }, + { + "epoch": 0.78, + "learning_rate": 2.428831307186449e-06, + "loss": 0.8335, + "step": 24396 + }, + { + "epoch": 0.78, + "learning_rate": 2.4281546785194442e-06, + "loss": 0.9146, + "step": 24397 + }, + { + "epoch": 0.78, + "learning_rate": 2.4274781310899596e-06, + "loss": 0.9214, + "step": 24398 + }, + { + "epoch": 0.78, + "learning_rate": 2.4268016649052515e-06, + "loss": 0.812, + "step": 24399 + }, + { + "epoch": 0.78, + "learning_rate": 2.426125279972583e-06, + "loss": 1.0615, + "step": 24400 + }, + { + "epoch": 0.78, + "learning_rate": 2.425448976299205e-06, + "loss": 0.9395, + "step": 24401 + }, + { + "epoch": 0.78, + "learning_rate": 2.4247727538923782e-06, + "loss": 0.8398, + "step": 24402 + }, + { + "epoch": 0.78, + "learning_rate": 2.4240966127593547e-06, + "loss": 0.9307, + "step": 24403 + }, + { + "epoch": 0.78, + "learning_rate": 2.423420552907394e-06, + "loss": 0.9209, + "step": 24404 + }, + { + "epoch": 0.78, + "learning_rate": 2.4227445743437406e-06, + "loss": 0.8345, + "step": 24405 + }, + { + "epoch": 0.78, + "learning_rate": 2.4220686770756543e-06, + "loss": 0.8652, + "step": 24406 + }, + { + "epoch": 0.78, + "learning_rate": 2.421392861110382e-06, + "loss": 0.936, + "step": 24407 + }, + { + "epoch": 0.78, + "learning_rate": 2.4207171264551786e-06, + "loss": 0.8657, + "step": 24408 + }, + { + "epoch": 0.78, + "learning_rate": 2.4200414731172905e-06, + "loss": 0.8691, + "step": 24409 + }, + { + "epoch": 0.78, + "learning_rate": 2.4193659011039707e-06, + "loss": 0.8672, + "step": 24410 + }, + { + "epoch": 0.78, + "learning_rate": 2.418690410422462e-06, + "loss": 0.9102, + "step": 24411 + }, + { + "epoch": 0.78, + "learning_rate": 2.4180150010800197e-06, + "loss": 0.4651, + "step": 24412 + }, + { + "epoch": 0.78, + "learning_rate": 2.4173396730838803e-06, + "loss": 0.8765, + "step": 24413 + }, + { + "epoch": 0.78, + "learning_rate": 2.4166644264412965e-06, + "loss": 0.9409, + "step": 24414 + }, + { + "epoch": 0.78, + "learning_rate": 2.4159892611595083e-06, + "loss": 0.9282, + "step": 24415 + }, + { + "epoch": 0.78, + "learning_rate": 2.4153141772457633e-06, + "loss": 0.9307, + "step": 24416 + }, + { + "epoch": 0.78, + "learning_rate": 2.414639174707302e-06, + "loss": 0.938, + "step": 24417 + }, + { + "epoch": 0.78, + "learning_rate": 2.4139642535513684e-06, + "loss": 0.856, + "step": 24418 + }, + { + "epoch": 0.78, + "learning_rate": 2.4132894137852036e-06, + "loss": 1.0283, + "step": 24419 + }, + { + "epoch": 0.78, + "learning_rate": 2.4126146554160455e-06, + "loss": 0.8271, + "step": 24420 + }, + { + "epoch": 0.78, + "learning_rate": 2.4119399784511333e-06, + "loss": 0.8882, + "step": 24421 + }, + { + "epoch": 0.78, + "learning_rate": 2.4112653828977093e-06, + "loss": 0.8435, + "step": 24422 + }, + { + "epoch": 0.78, + "learning_rate": 2.4105908687630066e-06, + "loss": 0.8691, + "step": 24423 + }, + { + "epoch": 0.78, + "learning_rate": 2.409916436054267e-06, + "loss": 0.8691, + "step": 24424 + }, + { + "epoch": 0.78, + "learning_rate": 2.4092420847787235e-06, + "loss": 0.9355, + "step": 24425 + }, + { + "epoch": 0.78, + "learning_rate": 2.4085678149436087e-06, + "loss": 0.873, + "step": 24426 + }, + { + "epoch": 0.78, + "learning_rate": 2.4078936265561625e-06, + "loss": 0.9231, + "step": 24427 + }, + { + "epoch": 0.78, + "learning_rate": 2.4072195196236146e-06, + "loss": 0.959, + "step": 24428 + }, + { + "epoch": 0.78, + "learning_rate": 2.4065454941531963e-06, + "loss": 0.8564, + "step": 24429 + }, + { + "epoch": 0.78, + "learning_rate": 2.4058715501521424e-06, + "loss": 0.814, + "step": 24430 + }, + { + "epoch": 0.78, + "learning_rate": 2.405197687627683e-06, + "loss": 0.8843, + "step": 24431 + }, + { + "epoch": 0.78, + "learning_rate": 2.404523906587044e-06, + "loss": 0.9795, + "step": 24432 + }, + { + "epoch": 0.78, + "learning_rate": 2.403850207037459e-06, + "loss": 0.9736, + "step": 24433 + }, + { + "epoch": 0.78, + "learning_rate": 2.4031765889861546e-06, + "loss": 0.9961, + "step": 24434 + }, + { + "epoch": 0.78, + "learning_rate": 2.402503052440356e-06, + "loss": 1.019, + "step": 24435 + }, + { + "epoch": 0.78, + "learning_rate": 2.4018295974072925e-06, + "loss": 1.02, + "step": 24436 + }, + { + "epoch": 0.78, + "learning_rate": 2.4011562238941886e-06, + "loss": 0.8022, + "step": 24437 + }, + { + "epoch": 0.78, + "learning_rate": 2.400482931908267e-06, + "loss": 0.9053, + "step": 24438 + }, + { + "epoch": 0.78, + "learning_rate": 2.399809721456754e-06, + "loss": 0.937, + "step": 24439 + }, + { + "epoch": 0.78, + "learning_rate": 2.3991365925468692e-06, + "loss": 0.9385, + "step": 24440 + }, + { + "epoch": 0.78, + "learning_rate": 2.3984635451858384e-06, + "loss": 1.0752, + "step": 24441 + }, + { + "epoch": 0.78, + "learning_rate": 2.397790579380881e-06, + "loss": 0.9966, + "step": 24442 + }, + { + "epoch": 0.78, + "learning_rate": 2.397117695139218e-06, + "loss": 0.9214, + "step": 24443 + }, + { + "epoch": 0.78, + "learning_rate": 2.396444892468064e-06, + "loss": 0.8799, + "step": 24444 + }, + { + "epoch": 0.78, + "learning_rate": 2.395772171374644e-06, + "loss": 0.9177, + "step": 24445 + }, + { + "epoch": 0.78, + "learning_rate": 2.3950995318661706e-06, + "loss": 1.0356, + "step": 24446 + }, + { + "epoch": 0.78, + "learning_rate": 2.394426973949865e-06, + "loss": 0.917, + "step": 24447 + }, + { + "epoch": 0.78, + "learning_rate": 2.3937544976329363e-06, + "loss": 0.9028, + "step": 24448 + }, + { + "epoch": 0.78, + "learning_rate": 2.3930821029226112e-06, + "loss": 0.9697, + "step": 24449 + }, + { + "epoch": 0.78, + "learning_rate": 2.39240978982609e-06, + "loss": 0.9761, + "step": 24450 + }, + { + "epoch": 0.78, + "learning_rate": 2.391737558350594e-06, + "loss": 0.8896, + "step": 24451 + }, + { + "epoch": 0.78, + "learning_rate": 2.3910654085033325e-06, + "loss": 0.8086, + "step": 24452 + }, + { + "epoch": 0.78, + "learning_rate": 2.390393340291519e-06, + "loss": 0.9819, + "step": 24453 + }, + { + "epoch": 0.78, + "learning_rate": 2.389721353722362e-06, + "loss": 0.9922, + "step": 24454 + }, + { + "epoch": 0.78, + "learning_rate": 2.3890494488030733e-06, + "loss": 1.0005, + "step": 24455 + }, + { + "epoch": 0.78, + "learning_rate": 2.388377625540859e-06, + "loss": 0.9492, + "step": 24456 + }, + { + "epoch": 0.78, + "learning_rate": 2.3877058839429347e-06, + "loss": 0.8074, + "step": 24457 + }, + { + "epoch": 0.78, + "learning_rate": 2.3870342240164945e-06, + "loss": 0.9668, + "step": 24458 + }, + { + "epoch": 0.78, + "learning_rate": 2.386362645768755e-06, + "loss": 0.8633, + "step": 24459 + }, + { + "epoch": 0.78, + "learning_rate": 2.3856911492069148e-06, + "loss": 1.0503, + "step": 24460 + }, + { + "epoch": 0.78, + "learning_rate": 2.385019734338184e-06, + "loss": 0.8652, + "step": 24461 + }, + { + "epoch": 0.78, + "learning_rate": 2.3843484011697614e-06, + "loss": 0.8555, + "step": 24462 + }, + { + "epoch": 0.78, + "learning_rate": 2.383677149708855e-06, + "loss": 0.8843, + "step": 24463 + }, + { + "epoch": 0.78, + "learning_rate": 2.3830059799626628e-06, + "loss": 0.8459, + "step": 24464 + }, + { + "epoch": 0.78, + "learning_rate": 2.382334891938386e-06, + "loss": 0.9331, + "step": 24465 + }, + { + "epoch": 0.78, + "learning_rate": 2.381663885643223e-06, + "loss": 0.8452, + "step": 24466 + }, + { + "epoch": 0.78, + "learning_rate": 2.3809929610843775e-06, + "loss": 0.9458, + "step": 24467 + }, + { + "epoch": 0.78, + "learning_rate": 2.380322118269043e-06, + "loss": 0.8047, + "step": 24468 + }, + { + "epoch": 0.78, + "learning_rate": 2.379651357204421e-06, + "loss": 1.0015, + "step": 24469 + }, + { + "epoch": 0.78, + "learning_rate": 2.3789806778977064e-06, + "loss": 0.5051, + "step": 24470 + }, + { + "epoch": 0.78, + "learning_rate": 2.378310080356092e-06, + "loss": 0.9546, + "step": 24471 + }, + { + "epoch": 0.78, + "learning_rate": 2.3776395645867777e-06, + "loss": 0.8887, + "step": 24472 + }, + { + "epoch": 0.78, + "learning_rate": 2.3769691305969544e-06, + "loss": 0.8628, + "step": 24473 + }, + { + "epoch": 0.78, + "learning_rate": 2.376298778393814e-06, + "loss": 0.9341, + "step": 24474 + }, + { + "epoch": 0.78, + "learning_rate": 2.375628507984552e-06, + "loss": 0.959, + "step": 24475 + }, + { + "epoch": 0.78, + "learning_rate": 2.374958319376358e-06, + "loss": 1.0566, + "step": 24476 + }, + { + "epoch": 0.78, + "learning_rate": 2.3742882125764198e-06, + "loss": 0.9258, + "step": 24477 + }, + { + "epoch": 0.78, + "learning_rate": 2.373618187591933e-06, + "loss": 0.9263, + "step": 24478 + }, + { + "epoch": 0.78, + "learning_rate": 2.3729482444300787e-06, + "loss": 0.9814, + "step": 24479 + }, + { + "epoch": 0.78, + "learning_rate": 2.3722783830980544e-06, + "loss": 0.9326, + "step": 24480 + }, + { + "epoch": 0.78, + "learning_rate": 2.3716086036030352e-06, + "loss": 0.8086, + "step": 24481 + }, + { + "epoch": 0.78, + "learning_rate": 2.370938905952216e-06, + "loss": 0.7878, + "step": 24482 + }, + { + "epoch": 0.78, + "learning_rate": 2.3702692901527757e-06, + "loss": 0.9395, + "step": 24483 + }, + { + "epoch": 0.78, + "learning_rate": 2.3695997562119057e-06, + "loss": 0.769, + "step": 24484 + }, + { + "epoch": 0.78, + "learning_rate": 2.3689303041367815e-06, + "loss": 0.8682, + "step": 24485 + }, + { + "epoch": 0.78, + "learning_rate": 2.368260933934593e-06, + "loss": 0.9897, + "step": 24486 + }, + { + "epoch": 0.78, + "learning_rate": 2.367591645612518e-06, + "loss": 0.8967, + "step": 24487 + }, + { + "epoch": 0.78, + "learning_rate": 2.3669224391777367e-06, + "loss": 0.9585, + "step": 24488 + }, + { + "epoch": 0.78, + "learning_rate": 2.366253314637428e-06, + "loss": 0.9277, + "step": 24489 + }, + { + "epoch": 0.78, + "learning_rate": 2.3655842719987756e-06, + "loss": 0.8735, + "step": 24490 + }, + { + "epoch": 0.78, + "learning_rate": 2.3649153112689518e-06, + "loss": 1.0215, + "step": 24491 + }, + { + "epoch": 0.78, + "learning_rate": 2.3642464324551384e-06, + "loss": 0.9043, + "step": 24492 + }, + { + "epoch": 0.78, + "learning_rate": 2.363577635564508e-06, + "loss": 0.9204, + "step": 24493 + }, + { + "epoch": 0.78, + "learning_rate": 2.36290892060424e-06, + "loss": 1.0522, + "step": 24494 + }, + { + "epoch": 0.78, + "learning_rate": 2.3622402875815077e-06, + "loss": 0.9058, + "step": 24495 + }, + { + "epoch": 0.78, + "learning_rate": 2.361571736503484e-06, + "loss": 1.0312, + "step": 24496 + }, + { + "epoch": 0.78, + "learning_rate": 2.360903267377339e-06, + "loss": 1.1167, + "step": 24497 + }, + { + "epoch": 0.78, + "learning_rate": 2.3602348802102494e-06, + "loss": 1.0767, + "step": 24498 + }, + { + "epoch": 0.78, + "learning_rate": 2.359566575009382e-06, + "loss": 0.9165, + "step": 24499 + }, + { + "epoch": 0.78, + "learning_rate": 2.358898351781912e-06, + "loss": 0.9756, + "step": 24500 + }, + { + "epoch": 0.78, + "learning_rate": 2.358230210535004e-06, + "loss": 0.9482, + "step": 24501 + }, + { + "epoch": 0.78, + "learning_rate": 2.357562151275833e-06, + "loss": 0.9761, + "step": 24502 + }, + { + "epoch": 0.78, + "learning_rate": 2.356894174011557e-06, + "loss": 0.9463, + "step": 24503 + }, + { + "epoch": 0.78, + "learning_rate": 2.3562262787493495e-06, + "loss": 1.0273, + "step": 24504 + }, + { + "epoch": 0.78, + "learning_rate": 2.3555584654963714e-06, + "loss": 1.0059, + "step": 24505 + }, + { + "epoch": 0.78, + "learning_rate": 2.3548907342597936e-06, + "loss": 0.9375, + "step": 24506 + }, + { + "epoch": 0.78, + "learning_rate": 2.354223085046775e-06, + "loss": 1.0791, + "step": 24507 + }, + { + "epoch": 0.78, + "learning_rate": 2.353555517864483e-06, + "loss": 0.9243, + "step": 24508 + }, + { + "epoch": 0.78, + "learning_rate": 2.3528880327200777e-06, + "loss": 0.9053, + "step": 24509 + }, + { + "epoch": 0.78, + "learning_rate": 2.3522206296207196e-06, + "loss": 0.9644, + "step": 24510 + }, + { + "epoch": 0.78, + "learning_rate": 2.351553308573569e-06, + "loss": 0.9717, + "step": 24511 + }, + { + "epoch": 0.78, + "learning_rate": 2.3508860695857872e-06, + "loss": 0.9663, + "step": 24512 + }, + { + "epoch": 0.78, + "learning_rate": 2.3502189126645303e-06, + "loss": 0.9463, + "step": 24513 + }, + { + "epoch": 0.78, + "learning_rate": 2.349551837816961e-06, + "loss": 0.9849, + "step": 24514 + }, + { + "epoch": 0.78, + "learning_rate": 2.348884845050234e-06, + "loss": 0.9741, + "step": 24515 + }, + { + "epoch": 0.78, + "learning_rate": 2.3482179343715017e-06, + "loss": 0.8677, + "step": 24516 + }, + { + "epoch": 0.78, + "learning_rate": 2.347551105787925e-06, + "loss": 0.9067, + "step": 24517 + }, + { + "epoch": 0.78, + "learning_rate": 2.346884359306656e-06, + "loss": 0.8975, + "step": 24518 + }, + { + "epoch": 0.78, + "learning_rate": 2.346217694934847e-06, + "loss": 0.8818, + "step": 24519 + }, + { + "epoch": 0.78, + "learning_rate": 2.3455511126796484e-06, + "loss": 0.9458, + "step": 24520 + }, + { + "epoch": 0.78, + "learning_rate": 2.3448846125482182e-06, + "loss": 0.9736, + "step": 24521 + }, + { + "epoch": 0.78, + "learning_rate": 2.3442181945477017e-06, + "loss": 1.0024, + "step": 24522 + }, + { + "epoch": 0.78, + "learning_rate": 2.3435518586852526e-06, + "loss": 0.9478, + "step": 24523 + }, + { + "epoch": 0.78, + "learning_rate": 2.342885604968016e-06, + "loss": 0.8452, + "step": 24524 + }, + { + "epoch": 0.78, + "learning_rate": 2.3422194334031477e-06, + "loss": 0.8521, + "step": 24525 + }, + { + "epoch": 0.78, + "learning_rate": 2.3415533439977845e-06, + "loss": 1.0176, + "step": 24526 + }, + { + "epoch": 0.78, + "learning_rate": 2.34088733675908e-06, + "loss": 0.9214, + "step": 24527 + }, + { + "epoch": 0.78, + "learning_rate": 2.3402214116941756e-06, + "loss": 0.9331, + "step": 24528 + }, + { + "epoch": 0.78, + "learning_rate": 2.339555568810221e-06, + "loss": 0.8682, + "step": 24529 + }, + { + "epoch": 0.78, + "learning_rate": 2.338889808114354e-06, + "loss": 0.9883, + "step": 24530 + }, + { + "epoch": 0.78, + "learning_rate": 2.3382241296137233e-06, + "loss": 0.9595, + "step": 24531 + }, + { + "epoch": 0.78, + "learning_rate": 2.337558533315468e-06, + "loss": 0.876, + "step": 24532 + }, + { + "epoch": 0.78, + "learning_rate": 2.3368930192267293e-06, + "loss": 0.8638, + "step": 24533 + }, + { + "epoch": 0.78, + "learning_rate": 2.3362275873546438e-06, + "loss": 0.8413, + "step": 24534 + }, + { + "epoch": 0.78, + "learning_rate": 2.335562237706358e-06, + "loss": 0.9092, + "step": 24535 + }, + { + "epoch": 0.78, + "learning_rate": 2.3348969702890044e-06, + "loss": 1.0034, + "step": 24536 + }, + { + "epoch": 0.78, + "learning_rate": 2.334231785109725e-06, + "loss": 0.9639, + "step": 24537 + }, + { + "epoch": 0.78, + "learning_rate": 2.333566682175652e-06, + "loss": 0.7217, + "step": 24538 + }, + { + "epoch": 0.78, + "learning_rate": 2.332901661493927e-06, + "loss": 0.9458, + "step": 24539 + }, + { + "epoch": 0.78, + "learning_rate": 2.3322367230716812e-06, + "loss": 0.9146, + "step": 24540 + }, + { + "epoch": 0.78, + "learning_rate": 2.3315718669160492e-06, + "loss": 0.9312, + "step": 24541 + }, + { + "epoch": 0.78, + "learning_rate": 2.33090709303416e-06, + "loss": 0.9551, + "step": 24542 + }, + { + "epoch": 0.78, + "learning_rate": 2.330242401433155e-06, + "loss": 0.8789, + "step": 24543 + }, + { + "epoch": 0.78, + "learning_rate": 2.3295777921201567e-06, + "loss": 0.9487, + "step": 24544 + }, + { + "epoch": 0.78, + "learning_rate": 2.3289132651023017e-06, + "loss": 0.9312, + "step": 24545 + }, + { + "epoch": 0.78, + "learning_rate": 2.3282488203867147e-06, + "loss": 0.9414, + "step": 24546 + }, + { + "epoch": 0.79, + "learning_rate": 2.3275844579805306e-06, + "loss": 0.9087, + "step": 24547 + }, + { + "epoch": 0.79, + "learning_rate": 2.3269201778908745e-06, + "loss": 0.9229, + "step": 24548 + }, + { + "epoch": 0.79, + "learning_rate": 2.3262559801248718e-06, + "loss": 0.9243, + "step": 24549 + }, + { + "epoch": 0.79, + "learning_rate": 2.325591864689647e-06, + "loss": 0.886, + "step": 24550 + }, + { + "epoch": 0.79, + "learning_rate": 2.324927831592331e-06, + "loss": 0.8081, + "step": 24551 + }, + { + "epoch": 0.79, + "learning_rate": 2.324263880840041e-06, + "loss": 0.875, + "step": 24552 + }, + { + "epoch": 0.79, + "learning_rate": 2.3236000124399084e-06, + "loss": 0.8911, + "step": 24553 + }, + { + "epoch": 0.79, + "learning_rate": 2.3229362263990518e-06, + "loss": 0.936, + "step": 24554 + }, + { + "epoch": 0.79, + "learning_rate": 2.3222725227245925e-06, + "loss": 0.9634, + "step": 24555 + }, + { + "epoch": 0.79, + "learning_rate": 2.3216089014236485e-06, + "loss": 1.0259, + "step": 24556 + }, + { + "epoch": 0.79, + "learning_rate": 2.3209453625033473e-06, + "loss": 0.9087, + "step": 24557 + }, + { + "epoch": 0.79, + "learning_rate": 2.3202819059707992e-06, + "loss": 0.9434, + "step": 24558 + }, + { + "epoch": 0.79, + "learning_rate": 2.319618531833131e-06, + "loss": 0.9253, + "step": 24559 + }, + { + "epoch": 0.79, + "learning_rate": 2.3189552400974545e-06, + "loss": 0.916, + "step": 24560 + }, + { + "epoch": 0.79, + "learning_rate": 2.3182920307708855e-06, + "loss": 0.9888, + "step": 24561 + }, + { + "epoch": 0.79, + "learning_rate": 2.317628903860544e-06, + "loss": 0.9604, + "step": 24562 + }, + { + "epoch": 0.79, + "learning_rate": 2.316965859373541e-06, + "loss": 0.9644, + "step": 24563 + }, + { + "epoch": 0.79, + "learning_rate": 2.316302897316992e-06, + "loss": 0.917, + "step": 24564 + }, + { + "epoch": 0.79, + "learning_rate": 2.3156400176980052e-06, + "loss": 0.9243, + "step": 24565 + }, + { + "epoch": 0.79, + "learning_rate": 2.3149772205237e-06, + "loss": 0.8643, + "step": 24566 + }, + { + "epoch": 0.79, + "learning_rate": 2.3143145058011805e-06, + "loss": 0.8892, + "step": 24567 + }, + { + "epoch": 0.79, + "learning_rate": 2.3136518735375646e-06, + "loss": 0.9341, + "step": 24568 + }, + { + "epoch": 0.79, + "learning_rate": 2.3129893237399524e-06, + "loss": 0.8047, + "step": 24569 + }, + { + "epoch": 0.79, + "learning_rate": 2.3123268564154645e-06, + "loss": 0.9336, + "step": 24570 + }, + { + "epoch": 0.79, + "learning_rate": 2.3116644715711943e-06, + "loss": 0.8267, + "step": 24571 + }, + { + "epoch": 0.79, + "learning_rate": 2.3110021692142592e-06, + "loss": 0.9326, + "step": 24572 + }, + { + "epoch": 0.79, + "learning_rate": 2.3103399493517576e-06, + "loss": 1.0454, + "step": 24573 + }, + { + "epoch": 0.79, + "learning_rate": 2.3096778119908005e-06, + "loss": 0.897, + "step": 24574 + }, + { + "epoch": 0.79, + "learning_rate": 2.3090157571384875e-06, + "loss": 0.8721, + "step": 24575 + }, + { + "epoch": 0.79, + "learning_rate": 2.3083537848019245e-06, + "loss": 0.8599, + "step": 24576 + }, + { + "epoch": 0.79, + "learning_rate": 2.307691894988211e-06, + "loss": 0.9688, + "step": 24577 + }, + { + "epoch": 0.79, + "learning_rate": 2.307030087704456e-06, + "loss": 1.0376, + "step": 24578 + }, + { + "epoch": 0.79, + "learning_rate": 2.306368362957747e-06, + "loss": 0.9292, + "step": 24579 + }, + { + "epoch": 0.79, + "learning_rate": 2.305706720755195e-06, + "loss": 1.0034, + "step": 24580 + }, + { + "epoch": 0.79, + "learning_rate": 2.3050451611038905e-06, + "loss": 0.8936, + "step": 24581 + }, + { + "epoch": 0.79, + "learning_rate": 2.3043836840109378e-06, + "loss": 1.105, + "step": 24582 + }, + { + "epoch": 0.79, + "learning_rate": 2.3037222894834287e-06, + "loss": 0.9434, + "step": 24583 + }, + { + "epoch": 0.79, + "learning_rate": 2.303060977528465e-06, + "loss": 0.8442, + "step": 24584 + }, + { + "epoch": 0.79, + "learning_rate": 2.3023997481531377e-06, + "loss": 0.8523, + "step": 24585 + }, + { + "epoch": 0.79, + "learning_rate": 2.3017386013645415e-06, + "loss": 0.9214, + "step": 24586 + }, + { + "epoch": 0.79, + "learning_rate": 2.301077537169768e-06, + "loss": 0.9268, + "step": 24587 + }, + { + "epoch": 0.79, + "learning_rate": 2.300416555575914e-06, + "loss": 0.8892, + "step": 24588 + }, + { + "epoch": 0.79, + "learning_rate": 2.299755656590067e-06, + "loss": 1.0142, + "step": 24589 + }, + { + "epoch": 0.79, + "learning_rate": 2.2990948402193215e-06, + "loss": 0.47, + "step": 24590 + }, + { + "epoch": 0.79, + "learning_rate": 2.298434106470763e-06, + "loss": 1.0073, + "step": 24591 + }, + { + "epoch": 0.79, + "learning_rate": 2.2977734553514854e-06, + "loss": 0.8511, + "step": 24592 + }, + { + "epoch": 0.79, + "learning_rate": 2.297112886868574e-06, + "loss": 0.8726, + "step": 24593 + }, + { + "epoch": 0.79, + "learning_rate": 2.2964524010291167e-06, + "loss": 0.8945, + "step": 24594 + }, + { + "epoch": 0.79, + "learning_rate": 2.2957919978401967e-06, + "loss": 0.9976, + "step": 24595 + }, + { + "epoch": 0.79, + "learning_rate": 2.295131677308904e-06, + "loss": 0.9395, + "step": 24596 + }, + { + "epoch": 0.79, + "learning_rate": 2.2944714394423186e-06, + "loss": 0.9087, + "step": 24597 + }, + { + "epoch": 0.79, + "learning_rate": 2.293811284247529e-06, + "loss": 0.9434, + "step": 24598 + }, + { + "epoch": 0.79, + "learning_rate": 2.2931512117316167e-06, + "loss": 0.8215, + "step": 24599 + }, + { + "epoch": 0.79, + "learning_rate": 2.2924912219016614e-06, + "loss": 0.8286, + "step": 24600 + }, + { + "epoch": 0.79, + "learning_rate": 2.291831314764743e-06, + "loss": 0.9048, + "step": 24601 + }, + { + "epoch": 0.79, + "learning_rate": 2.2911714903279456e-06, + "loss": 1.0869, + "step": 24602 + }, + { + "epoch": 0.79, + "learning_rate": 2.2905117485983464e-06, + "loss": 0.8882, + "step": 24603 + }, + { + "epoch": 0.79, + "learning_rate": 2.2898520895830213e-06, + "loss": 1.0303, + "step": 24604 + }, + { + "epoch": 0.79, + "learning_rate": 2.2891925132890524e-06, + "loss": 0.9634, + "step": 24605 + }, + { + "epoch": 0.79, + "learning_rate": 2.2885330197235125e-06, + "loss": 0.9458, + "step": 24606 + }, + { + "epoch": 0.79, + "learning_rate": 2.2878736088934806e-06, + "loss": 0.958, + "step": 24607 + }, + { + "epoch": 0.79, + "learning_rate": 2.2872142808060294e-06, + "loss": 0.9722, + "step": 24608 + }, + { + "epoch": 0.79, + "learning_rate": 2.286555035468233e-06, + "loss": 1.0742, + "step": 24609 + }, + { + "epoch": 0.79, + "learning_rate": 2.2858958728871606e-06, + "loss": 0.9873, + "step": 24610 + }, + { + "epoch": 0.79, + "learning_rate": 2.2852367930698913e-06, + "loss": 0.9268, + "step": 24611 + }, + { + "epoch": 0.79, + "learning_rate": 2.2845777960234904e-06, + "loss": 0.9297, + "step": 24612 + }, + { + "epoch": 0.79, + "learning_rate": 2.2839188817550338e-06, + "loss": 0.8916, + "step": 24613 + }, + { + "epoch": 0.79, + "learning_rate": 2.283260050271585e-06, + "loss": 1.0283, + "step": 24614 + }, + { + "epoch": 0.79, + "learning_rate": 2.2826013015802205e-06, + "loss": 0.9395, + "step": 24615 + }, + { + "epoch": 0.79, + "learning_rate": 2.281942635687998e-06, + "loss": 0.8789, + "step": 24616 + }, + { + "epoch": 0.79, + "learning_rate": 2.281284052601991e-06, + "loss": 0.876, + "step": 24617 + }, + { + "epoch": 0.79, + "learning_rate": 2.280625552329261e-06, + "loss": 0.9771, + "step": 24618 + }, + { + "epoch": 0.79, + "learning_rate": 2.2799671348768783e-06, + "loss": 0.9038, + "step": 24619 + }, + { + "epoch": 0.79, + "learning_rate": 2.2793088002519013e-06, + "loss": 0.8477, + "step": 24620 + }, + { + "epoch": 0.79, + "learning_rate": 2.2786505484613997e-06, + "loss": 1.0737, + "step": 24621 + }, + { + "epoch": 0.79, + "learning_rate": 2.2779923795124283e-06, + "loss": 0.8115, + "step": 24622 + }, + { + "epoch": 0.79, + "learning_rate": 2.277334293412058e-06, + "loss": 0.833, + "step": 24623 + }, + { + "epoch": 0.79, + "learning_rate": 2.2766762901673376e-06, + "loss": 0.9346, + "step": 24624 + }, + { + "epoch": 0.79, + "learning_rate": 2.2760183697853367e-06, + "loss": 0.8384, + "step": 24625 + }, + { + "epoch": 0.79, + "learning_rate": 2.275360532273108e-06, + "loss": 0.9253, + "step": 24626 + }, + { + "epoch": 0.79, + "learning_rate": 2.2747027776377127e-06, + "loss": 0.8789, + "step": 24627 + }, + { + "epoch": 0.79, + "learning_rate": 2.2740451058862046e-06, + "loss": 0.8521, + "step": 24628 + }, + { + "epoch": 0.79, + "learning_rate": 2.273387517025645e-06, + "loss": 0.8672, + "step": 24629 + }, + { + "epoch": 0.79, + "learning_rate": 2.2727300110630823e-06, + "loss": 1.0894, + "step": 24630 + }, + { + "epoch": 0.79, + "learning_rate": 2.2720725880055806e-06, + "loss": 0.8813, + "step": 24631 + }, + { + "epoch": 0.79, + "learning_rate": 2.2714152478601813e-06, + "loss": 0.8828, + "step": 24632 + }, + { + "epoch": 0.79, + "learning_rate": 2.270757990633946e-06, + "loss": 0.8916, + "step": 24633 + }, + { + "epoch": 0.79, + "learning_rate": 2.2701008163339212e-06, + "loss": 0.8579, + "step": 24634 + }, + { + "epoch": 0.79, + "learning_rate": 2.269443724967161e-06, + "loss": 1.0293, + "step": 24635 + }, + { + "epoch": 0.79, + "learning_rate": 2.268786716540712e-06, + "loss": 0.9189, + "step": 24636 + }, + { + "epoch": 0.79, + "learning_rate": 2.2681297910616273e-06, + "loss": 0.9805, + "step": 24637 + }, + { + "epoch": 0.79, + "learning_rate": 2.267472948536953e-06, + "loss": 0.9678, + "step": 24638 + }, + { + "epoch": 0.79, + "learning_rate": 2.266816188973735e-06, + "loss": 0.9116, + "step": 24639 + }, + { + "epoch": 0.79, + "learning_rate": 2.2661595123790194e-06, + "loss": 0.509, + "step": 24640 + }, + { + "epoch": 0.79, + "learning_rate": 2.265502918759854e-06, + "loss": 0.5, + "step": 24641 + }, + { + "epoch": 0.79, + "learning_rate": 2.2648464081232836e-06, + "loss": 0.8545, + "step": 24642 + }, + { + "epoch": 0.79, + "learning_rate": 2.2641899804763467e-06, + "loss": 1.0122, + "step": 24643 + }, + { + "epoch": 0.79, + "learning_rate": 2.2635336358260928e-06, + "loss": 0.7886, + "step": 24644 + }, + { + "epoch": 0.79, + "learning_rate": 2.2628773741795585e-06, + "loss": 0.8853, + "step": 24645 + }, + { + "epoch": 0.79, + "learning_rate": 2.262221195543789e-06, + "loss": 0.9268, + "step": 24646 + }, + { + "epoch": 0.79, + "learning_rate": 2.261565099925822e-06, + "loss": 0.7729, + "step": 24647 + }, + { + "epoch": 0.79, + "learning_rate": 2.260909087332698e-06, + "loss": 0.8809, + "step": 24648 + }, + { + "epoch": 0.79, + "learning_rate": 2.2602531577714513e-06, + "loss": 1.0483, + "step": 24649 + }, + { + "epoch": 0.79, + "learning_rate": 2.2595973112491243e-06, + "loss": 0.9941, + "step": 24650 + }, + { + "epoch": 0.79, + "learning_rate": 2.2589415477727496e-06, + "loss": 0.9814, + "step": 24651 + }, + { + "epoch": 0.79, + "learning_rate": 2.258285867349367e-06, + "loss": 0.9019, + "step": 24652 + }, + { + "epoch": 0.79, + "learning_rate": 2.2576302699860097e-06, + "loss": 0.8965, + "step": 24653 + }, + { + "epoch": 0.79, + "learning_rate": 2.2569747556897103e-06, + "loss": 0.8818, + "step": 24654 + }, + { + "epoch": 0.79, + "learning_rate": 2.2563193244674988e-06, + "loss": 0.8845, + "step": 24655 + }, + { + "epoch": 0.79, + "learning_rate": 2.2556639763264144e-06, + "loss": 0.5234, + "step": 24656 + }, + { + "epoch": 0.79, + "learning_rate": 2.255008711273481e-06, + "loss": 0.9438, + "step": 24657 + }, + { + "epoch": 0.79, + "learning_rate": 2.254353529315736e-06, + "loss": 0.8174, + "step": 24658 + }, + { + "epoch": 0.79, + "learning_rate": 2.2536984304602016e-06, + "loss": 0.9199, + "step": 24659 + }, + { + "epoch": 0.79, + "learning_rate": 2.2530434147139124e-06, + "loss": 0.8538, + "step": 24660 + }, + { + "epoch": 0.79, + "learning_rate": 2.2523884820838937e-06, + "loss": 0.9546, + "step": 24661 + }, + { + "epoch": 0.79, + "learning_rate": 2.251733632577172e-06, + "loss": 0.9111, + "step": 24662 + }, + { + "epoch": 0.79, + "learning_rate": 2.251078866200771e-06, + "loss": 0.8821, + "step": 24663 + }, + { + "epoch": 0.79, + "learning_rate": 2.2504241829617203e-06, + "loss": 1.0229, + "step": 24664 + }, + { + "epoch": 0.79, + "learning_rate": 2.2497695828670386e-06, + "loss": 0.9717, + "step": 24665 + }, + { + "epoch": 0.79, + "learning_rate": 2.249115065923754e-06, + "loss": 0.9951, + "step": 24666 + }, + { + "epoch": 0.79, + "learning_rate": 2.2484606321388834e-06, + "loss": 0.8398, + "step": 24667 + }, + { + "epoch": 0.79, + "learning_rate": 2.2478062815194566e-06, + "loss": 0.8076, + "step": 24668 + }, + { + "epoch": 0.79, + "learning_rate": 2.2471520140724845e-06, + "loss": 1.0322, + "step": 24669 + }, + { + "epoch": 0.79, + "learning_rate": 2.2464978298049935e-06, + "loss": 0.8555, + "step": 24670 + }, + { + "epoch": 0.79, + "learning_rate": 2.2458437287239955e-06, + "loss": 1.1055, + "step": 24671 + }, + { + "epoch": 0.79, + "learning_rate": 2.245189710836516e-06, + "loss": 0.9658, + "step": 24672 + }, + { + "epoch": 0.79, + "learning_rate": 2.2445357761495657e-06, + "loss": 0.8911, + "step": 24673 + }, + { + "epoch": 0.79, + "learning_rate": 2.243881924670166e-06, + "loss": 0.9561, + "step": 24674 + }, + { + "epoch": 0.79, + "learning_rate": 2.243228156405326e-06, + "loss": 1.0073, + "step": 24675 + }, + { + "epoch": 0.79, + "learning_rate": 2.242574471362069e-06, + "loss": 0.9292, + "step": 24676 + }, + { + "epoch": 0.79, + "learning_rate": 2.2419208695473958e-06, + "loss": 0.8667, + "step": 24677 + }, + { + "epoch": 0.79, + "learning_rate": 2.2412673509683293e-06, + "loss": 0.8413, + "step": 24678 + }, + { + "epoch": 0.79, + "learning_rate": 2.240613915631875e-06, + "loss": 1.0063, + "step": 24679 + }, + { + "epoch": 0.79, + "learning_rate": 2.2399605635450474e-06, + "loss": 0.8916, + "step": 24680 + }, + { + "epoch": 0.79, + "learning_rate": 2.2393072947148555e-06, + "loss": 0.9966, + "step": 24681 + }, + { + "epoch": 0.79, + "learning_rate": 2.2386541091483037e-06, + "loss": 0.981, + "step": 24682 + }, + { + "epoch": 0.79, + "learning_rate": 2.238001006852406e-06, + "loss": 0.8936, + "step": 24683 + }, + { + "epoch": 0.79, + "learning_rate": 2.2373479878341676e-06, + "loss": 0.9761, + "step": 24684 + }, + { + "epoch": 0.79, + "learning_rate": 2.23669505210059e-06, + "loss": 0.9321, + "step": 24685 + }, + { + "epoch": 0.79, + "learning_rate": 2.236042199658687e-06, + "loss": 0.8594, + "step": 24686 + }, + { + "epoch": 0.79, + "learning_rate": 2.235389430515458e-06, + "loss": 0.8394, + "step": 24687 + }, + { + "epoch": 0.79, + "learning_rate": 2.234736744677903e-06, + "loss": 0.9697, + "step": 24688 + }, + { + "epoch": 0.79, + "learning_rate": 2.2340841421530336e-06, + "loss": 0.9282, + "step": 24689 + }, + { + "epoch": 0.79, + "learning_rate": 2.2334316229478415e-06, + "loss": 0.4939, + "step": 24690 + }, + { + "epoch": 0.79, + "learning_rate": 2.232779187069336e-06, + "loss": 0.9282, + "step": 24691 + }, + { + "epoch": 0.79, + "learning_rate": 2.232126834524514e-06, + "loss": 0.8335, + "step": 24692 + }, + { + "epoch": 0.79, + "learning_rate": 2.231474565320374e-06, + "loss": 0.9438, + "step": 24693 + }, + { + "epoch": 0.79, + "learning_rate": 2.2308223794639105e-06, + "loss": 1.0088, + "step": 24694 + }, + { + "epoch": 0.79, + "learning_rate": 2.2301702769621282e-06, + "loss": 1.0059, + "step": 24695 + }, + { + "epoch": 0.79, + "learning_rate": 2.2295182578220155e-06, + "loss": 0.9243, + "step": 24696 + }, + { + "epoch": 0.79, + "learning_rate": 2.228866322050576e-06, + "loss": 0.9863, + "step": 24697 + }, + { + "epoch": 0.79, + "learning_rate": 2.228214469654797e-06, + "loss": 0.917, + "step": 24698 + }, + { + "epoch": 0.79, + "learning_rate": 2.22756270064168e-06, + "loss": 1.0234, + "step": 24699 + }, + { + "epoch": 0.79, + "learning_rate": 2.2269110150182083e-06, + "loss": 0.9316, + "step": 24700 + }, + { + "epoch": 0.79, + "learning_rate": 2.2262594127913815e-06, + "loss": 0.4639, + "step": 24701 + }, + { + "epoch": 0.79, + "learning_rate": 2.2256078939681846e-06, + "loss": 0.9272, + "step": 24702 + }, + { + "epoch": 0.79, + "learning_rate": 2.224956458555614e-06, + "loss": 0.96, + "step": 24703 + }, + { + "epoch": 0.79, + "learning_rate": 2.2243051065606525e-06, + "loss": 0.915, + "step": 24704 + }, + { + "epoch": 0.79, + "learning_rate": 2.223653837990295e-06, + "loss": 0.8369, + "step": 24705 + }, + { + "epoch": 0.79, + "learning_rate": 2.223002652851525e-06, + "loss": 1.0044, + "step": 24706 + }, + { + "epoch": 0.79, + "learning_rate": 2.222351551151328e-06, + "loss": 0.9697, + "step": 24707 + }, + { + "epoch": 0.79, + "learning_rate": 2.22170053289669e-06, + "loss": 0.9878, + "step": 24708 + }, + { + "epoch": 0.79, + "learning_rate": 2.221049598094598e-06, + "loss": 0.8452, + "step": 24709 + }, + { + "epoch": 0.79, + "learning_rate": 2.2203987467520325e-06, + "loss": 1.0483, + "step": 24710 + }, + { + "epoch": 0.79, + "learning_rate": 2.2197479788759814e-06, + "loss": 1.0396, + "step": 24711 + }, + { + "epoch": 0.79, + "learning_rate": 2.2190972944734213e-06, + "loss": 0.9045, + "step": 24712 + }, + { + "epoch": 0.79, + "learning_rate": 2.2184466935513372e-06, + "loss": 0.8594, + "step": 24713 + }, + { + "epoch": 0.79, + "learning_rate": 2.2177961761167087e-06, + "loss": 0.9504, + "step": 24714 + }, + { + "epoch": 0.79, + "learning_rate": 2.217145742176514e-06, + "loss": 0.9233, + "step": 24715 + }, + { + "epoch": 0.79, + "learning_rate": 2.2164953917377295e-06, + "loss": 0.876, + "step": 24716 + }, + { + "epoch": 0.79, + "learning_rate": 2.2158451248073374e-06, + "loss": 0.8425, + "step": 24717 + }, + { + "epoch": 0.79, + "learning_rate": 2.2151949413923102e-06, + "loss": 0.9028, + "step": 24718 + }, + { + "epoch": 0.79, + "learning_rate": 2.2145448414996275e-06, + "loss": 0.9424, + "step": 24719 + }, + { + "epoch": 0.79, + "learning_rate": 2.2138948251362625e-06, + "loss": 1.0303, + "step": 24720 + }, + { + "epoch": 0.79, + "learning_rate": 2.2132448923091874e-06, + "loss": 0.8901, + "step": 24721 + }, + { + "epoch": 0.79, + "learning_rate": 2.212595043025375e-06, + "loss": 0.9277, + "step": 24722 + }, + { + "epoch": 0.79, + "learning_rate": 2.2119452772918026e-06, + "loss": 0.9858, + "step": 24723 + }, + { + "epoch": 0.79, + "learning_rate": 2.2112955951154346e-06, + "loss": 0.9146, + "step": 24724 + }, + { + "epoch": 0.79, + "learning_rate": 2.210645996503247e-06, + "loss": 0.9199, + "step": 24725 + }, + { + "epoch": 0.79, + "learning_rate": 2.2099964814622075e-06, + "loss": 0.9487, + "step": 24726 + }, + { + "epoch": 0.79, + "learning_rate": 2.209347049999282e-06, + "loss": 0.8779, + "step": 24727 + }, + { + "epoch": 0.79, + "learning_rate": 2.2086977021214417e-06, + "loss": 0.8914, + "step": 24728 + }, + { + "epoch": 0.79, + "learning_rate": 2.2080484378356528e-06, + "loss": 0.9019, + "step": 24729 + }, + { + "epoch": 0.79, + "learning_rate": 2.207399257148878e-06, + "loss": 0.8623, + "step": 24730 + }, + { + "epoch": 0.79, + "learning_rate": 2.2067501600680873e-06, + "loss": 0.8394, + "step": 24731 + }, + { + "epoch": 0.79, + "learning_rate": 2.2061011466002423e-06, + "loss": 0.9868, + "step": 24732 + }, + { + "epoch": 0.79, + "learning_rate": 2.205452216752303e-06, + "loss": 0.9141, + "step": 24733 + }, + { + "epoch": 0.79, + "learning_rate": 2.204803370531239e-06, + "loss": 0.9648, + "step": 24734 + }, + { + "epoch": 0.79, + "learning_rate": 2.2041546079440036e-06, + "loss": 0.9961, + "step": 24735 + }, + { + "epoch": 0.79, + "learning_rate": 2.2035059289975656e-06, + "loss": 0.9492, + "step": 24736 + }, + { + "epoch": 0.79, + "learning_rate": 2.202857333698877e-06, + "loss": 0.8809, + "step": 24737 + }, + { + "epoch": 0.79, + "learning_rate": 2.2022088220549007e-06, + "loss": 0.8633, + "step": 24738 + }, + { + "epoch": 0.79, + "learning_rate": 2.2015603940725915e-06, + "loss": 0.9214, + "step": 24739 + }, + { + "epoch": 0.79, + "learning_rate": 2.2009120497589122e-06, + "loss": 0.8901, + "step": 24740 + }, + { + "epoch": 0.79, + "learning_rate": 2.2002637891208112e-06, + "loss": 0.9614, + "step": 24741 + }, + { + "epoch": 0.79, + "learning_rate": 2.19961561216525e-06, + "loss": 0.9106, + "step": 24742 + }, + { + "epoch": 0.79, + "learning_rate": 2.1989675188991777e-06, + "loss": 0.9165, + "step": 24743 + }, + { + "epoch": 0.79, + "learning_rate": 2.198319509329556e-06, + "loss": 0.8381, + "step": 24744 + }, + { + "epoch": 0.79, + "learning_rate": 2.1976715834633254e-06, + "loss": 0.8472, + "step": 24745 + }, + { + "epoch": 0.79, + "learning_rate": 2.197023741307447e-06, + "loss": 0.9478, + "step": 24746 + }, + { + "epoch": 0.79, + "learning_rate": 2.1963759828688645e-06, + "loss": 1.0273, + "step": 24747 + }, + { + "epoch": 0.79, + "learning_rate": 2.1957283081545346e-06, + "loss": 0.8037, + "step": 24748 + }, + { + "epoch": 0.79, + "learning_rate": 2.1950807171713994e-06, + "loss": 0.8887, + "step": 24749 + }, + { + "epoch": 0.79, + "learning_rate": 2.194433209926412e-06, + "loss": 0.813, + "step": 24750 + }, + { + "epoch": 0.79, + "learning_rate": 2.1937857864265187e-06, + "loss": 0.8535, + "step": 24751 + }, + { + "epoch": 0.79, + "learning_rate": 2.193138446678664e-06, + "loss": 0.9077, + "step": 24752 + }, + { + "epoch": 0.79, + "learning_rate": 2.1924911906897904e-06, + "loss": 0.9385, + "step": 24753 + }, + { + "epoch": 0.79, + "learning_rate": 2.1918440184668487e-06, + "loss": 0.979, + "step": 24754 + }, + { + "epoch": 0.79, + "learning_rate": 2.1911969300167767e-06, + "loss": 0.8159, + "step": 24755 + }, + { + "epoch": 0.79, + "learning_rate": 2.1905499253465214e-06, + "loss": 0.9253, + "step": 24756 + }, + { + "epoch": 0.79, + "learning_rate": 2.18990300446302e-06, + "loss": 0.9526, + "step": 24757 + }, + { + "epoch": 0.79, + "learning_rate": 2.189256167373217e-06, + "loss": 0.915, + "step": 24758 + }, + { + "epoch": 0.79, + "learning_rate": 2.188609414084052e-06, + "loss": 0.9482, + "step": 24759 + }, + { + "epoch": 0.79, + "learning_rate": 2.187962744602463e-06, + "loss": 0.8882, + "step": 24760 + }, + { + "epoch": 0.79, + "learning_rate": 2.187316158935384e-06, + "loss": 0.9033, + "step": 24761 + }, + { + "epoch": 0.79, + "learning_rate": 2.1866696570897593e-06, + "loss": 0.8672, + "step": 24762 + }, + { + "epoch": 0.79, + "learning_rate": 2.18602323907252e-06, + "loss": 0.8374, + "step": 24763 + }, + { + "epoch": 0.79, + "learning_rate": 2.1853769048906047e-06, + "loss": 0.9194, + "step": 24764 + }, + { + "epoch": 0.79, + "learning_rate": 2.184730654550947e-06, + "loss": 0.8877, + "step": 24765 + }, + { + "epoch": 0.79, + "learning_rate": 2.1840844880604795e-06, + "loss": 0.9214, + "step": 24766 + }, + { + "epoch": 0.79, + "learning_rate": 2.1834384054261325e-06, + "loss": 0.9331, + "step": 24767 + }, + { + "epoch": 0.79, + "learning_rate": 2.1827924066548436e-06, + "loss": 0.8843, + "step": 24768 + }, + { + "epoch": 0.79, + "learning_rate": 2.1821464917535375e-06, + "loss": 0.9121, + "step": 24769 + }, + { + "epoch": 0.79, + "learning_rate": 2.18150066072915e-06, + "loss": 0.9355, + "step": 24770 + }, + { + "epoch": 0.79, + "learning_rate": 2.1808549135886083e-06, + "loss": 0.9629, + "step": 24771 + }, + { + "epoch": 0.79, + "learning_rate": 2.180209250338835e-06, + "loss": 0.96, + "step": 24772 + }, + { + "epoch": 0.79, + "learning_rate": 2.1795636709867663e-06, + "loss": 0.9146, + "step": 24773 + }, + { + "epoch": 0.79, + "learning_rate": 2.178918175539323e-06, + "loss": 0.855, + "step": 24774 + }, + { + "epoch": 0.79, + "learning_rate": 2.1782727640034295e-06, + "loss": 0.959, + "step": 24775 + }, + { + "epoch": 0.79, + "learning_rate": 2.177627436386015e-06, + "loss": 1.0625, + "step": 24776 + }, + { + "epoch": 0.79, + "learning_rate": 2.1769821926940006e-06, + "loss": 0.8335, + "step": 24777 + }, + { + "epoch": 0.79, + "learning_rate": 2.176337032934307e-06, + "loss": 0.9141, + "step": 24778 + }, + { + "epoch": 0.79, + "learning_rate": 2.1756919571138613e-06, + "loss": 0.8594, + "step": 24779 + }, + { + "epoch": 0.79, + "learning_rate": 2.1750469652395777e-06, + "loss": 1.001, + "step": 24780 + }, + { + "epoch": 0.79, + "learning_rate": 2.174402057318383e-06, + "loss": 0.9551, + "step": 24781 + }, + { + "epoch": 0.79, + "learning_rate": 2.1737572333571934e-06, + "loss": 0.8599, + "step": 24782 + }, + { + "epoch": 0.79, + "learning_rate": 2.173112493362927e-06, + "loss": 0.9097, + "step": 24783 + }, + { + "epoch": 0.79, + "learning_rate": 2.172467837342499e-06, + "loss": 0.8096, + "step": 24784 + }, + { + "epoch": 0.79, + "learning_rate": 2.1718232653028292e-06, + "loss": 0.9624, + "step": 24785 + }, + { + "epoch": 0.79, + "learning_rate": 2.1711787772508307e-06, + "loss": 0.8135, + "step": 24786 + }, + { + "epoch": 0.79, + "learning_rate": 2.1705343731934215e-06, + "loss": 0.7969, + "step": 24787 + }, + { + "epoch": 0.79, + "learning_rate": 2.169890053137511e-06, + "loss": 0.916, + "step": 24788 + }, + { + "epoch": 0.79, + "learning_rate": 2.16924581709002e-06, + "loss": 1.0122, + "step": 24789 + }, + { + "epoch": 0.79, + "learning_rate": 2.168601665057848e-06, + "loss": 0.9419, + "step": 24790 + }, + { + "epoch": 0.79, + "learning_rate": 2.1679575970479163e-06, + "loss": 0.8459, + "step": 24791 + }, + { + "epoch": 0.79, + "learning_rate": 2.167313613067129e-06, + "loss": 0.916, + "step": 24792 + }, + { + "epoch": 0.79, + "learning_rate": 2.1666697131224e-06, + "loss": 0.9258, + "step": 24793 + }, + { + "epoch": 0.79, + "learning_rate": 2.1660258972206325e-06, + "loss": 0.9702, + "step": 24794 + }, + { + "epoch": 0.79, + "learning_rate": 2.16538216536874e-06, + "loss": 0.9575, + "step": 24795 + }, + { + "epoch": 0.79, + "learning_rate": 2.1647385175736235e-06, + "loss": 1.0132, + "step": 24796 + }, + { + "epoch": 0.79, + "learning_rate": 2.164094953842197e-06, + "loss": 0.8882, + "step": 24797 + }, + { + "epoch": 0.79, + "learning_rate": 2.1634514741813528e-06, + "loss": 0.958, + "step": 24798 + }, + { + "epoch": 0.79, + "learning_rate": 2.162808078598004e-06, + "loss": 0.9209, + "step": 24799 + }, + { + "epoch": 0.79, + "learning_rate": 2.1621647670990487e-06, + "loss": 0.9536, + "step": 24800 + }, + { + "epoch": 0.79, + "learning_rate": 2.1615215396913935e-06, + "loss": 1.1045, + "step": 24801 + }, + { + "epoch": 0.79, + "learning_rate": 2.1608783963819347e-06, + "loss": 0.9321, + "step": 24802 + }, + { + "epoch": 0.79, + "learning_rate": 2.1602353371775765e-06, + "loss": 0.8379, + "step": 24803 + }, + { + "epoch": 0.79, + "learning_rate": 2.1595923620852175e-06, + "loss": 0.8921, + "step": 24804 + }, + { + "epoch": 0.79, + "learning_rate": 2.1589494711117543e-06, + "loss": 0.9019, + "step": 24805 + }, + { + "epoch": 0.79, + "learning_rate": 2.1583066642640826e-06, + "loss": 0.8818, + "step": 24806 + }, + { + "epoch": 0.79, + "learning_rate": 2.1576639415491053e-06, + "loss": 0.833, + "step": 24807 + }, + { + "epoch": 0.79, + "learning_rate": 2.1570213029737108e-06, + "loss": 0.7705, + "step": 24808 + }, + { + "epoch": 0.79, + "learning_rate": 2.156378748544801e-06, + "loss": 0.9497, + "step": 24809 + }, + { + "epoch": 0.79, + "learning_rate": 2.1557362782692646e-06, + "loss": 1.0142, + "step": 24810 + }, + { + "epoch": 0.79, + "learning_rate": 2.1550938921539955e-06, + "loss": 0.8306, + "step": 24811 + }, + { + "epoch": 0.79, + "learning_rate": 2.154451590205888e-06, + "loss": 0.9819, + "step": 24812 + }, + { + "epoch": 0.79, + "learning_rate": 2.153809372431833e-06, + "loss": 1.0557, + "step": 24813 + }, + { + "epoch": 0.79, + "learning_rate": 2.153167238838716e-06, + "loss": 0.9165, + "step": 24814 + }, + { + "epoch": 0.79, + "learning_rate": 2.1525251894334343e-06, + "loss": 0.8999, + "step": 24815 + }, + { + "epoch": 0.79, + "learning_rate": 2.1518832242228703e-06, + "loss": 0.9854, + "step": 24816 + }, + { + "epoch": 0.79, + "learning_rate": 2.1512413432139122e-06, + "loss": 0.8298, + "step": 24817 + }, + { + "epoch": 0.79, + "learning_rate": 2.15059954641345e-06, + "loss": 0.7595, + "step": 24818 + }, + { + "epoch": 0.79, + "learning_rate": 2.1499578338283677e-06, + "loss": 0.8567, + "step": 24819 + }, + { + "epoch": 0.79, + "learning_rate": 2.1493162054655493e-06, + "loss": 0.9478, + "step": 24820 + }, + { + "epoch": 0.79, + "learning_rate": 2.1486746613318766e-06, + "loss": 0.8179, + "step": 24821 + }, + { + "epoch": 0.79, + "learning_rate": 2.1480332014342387e-06, + "loss": 0.9277, + "step": 24822 + }, + { + "epoch": 0.79, + "learning_rate": 2.147391825779511e-06, + "loss": 0.9526, + "step": 24823 + }, + { + "epoch": 0.79, + "learning_rate": 2.146750534374581e-06, + "loss": 0.8379, + "step": 24824 + }, + { + "epoch": 0.79, + "learning_rate": 2.1461093272263244e-06, + "loss": 0.4504, + "step": 24825 + }, + { + "epoch": 0.79, + "learning_rate": 2.1454682043416244e-06, + "loss": 0.7876, + "step": 24826 + }, + { + "epoch": 0.79, + "learning_rate": 2.1448271657273577e-06, + "loss": 1.0093, + "step": 24827 + }, + { + "epoch": 0.79, + "learning_rate": 2.1441862113904024e-06, + "loss": 0.9111, + "step": 24828 + }, + { + "epoch": 0.79, + "learning_rate": 2.1435453413376307e-06, + "loss": 0.8589, + "step": 24829 + }, + { + "epoch": 0.79, + "learning_rate": 2.1429045555759265e-06, + "loss": 0.9272, + "step": 24830 + }, + { + "epoch": 0.79, + "learning_rate": 2.1422638541121567e-06, + "loss": 0.9785, + "step": 24831 + }, + { + "epoch": 0.79, + "learning_rate": 2.141623236953203e-06, + "loss": 1.0171, + "step": 24832 + }, + { + "epoch": 0.79, + "learning_rate": 2.1409827041059306e-06, + "loss": 0.9438, + "step": 24833 + }, + { + "epoch": 0.79, + "learning_rate": 2.1403422555772226e-06, + "loss": 0.8657, + "step": 24834 + }, + { + "epoch": 0.79, + "learning_rate": 2.1397018913739377e-06, + "loss": 0.897, + "step": 24835 + }, + { + "epoch": 0.79, + "learning_rate": 2.1390616115029548e-06, + "loss": 0.9243, + "step": 24836 + }, + { + "epoch": 0.79, + "learning_rate": 2.138421415971138e-06, + "loss": 0.9126, + "step": 24837 + }, + { + "epoch": 0.79, + "learning_rate": 2.1377813047853622e-06, + "loss": 0.8955, + "step": 24838 + }, + { + "epoch": 0.79, + "learning_rate": 2.137141277952488e-06, + "loss": 0.8418, + "step": 24839 + }, + { + "epoch": 0.79, + "learning_rate": 2.1365013354793885e-06, + "loss": 0.9077, + "step": 24840 + }, + { + "epoch": 0.79, + "learning_rate": 2.1358614773729237e-06, + "loss": 0.9666, + "step": 24841 + }, + { + "epoch": 0.79, + "learning_rate": 2.135221703639968e-06, + "loss": 0.9653, + "step": 24842 + }, + { + "epoch": 0.79, + "learning_rate": 2.134582014287373e-06, + "loss": 0.9414, + "step": 24843 + }, + { + "epoch": 0.79, + "learning_rate": 2.1339424093220106e-06, + "loss": 0.895, + "step": 24844 + }, + { + "epoch": 0.79, + "learning_rate": 2.133302888750738e-06, + "loss": 0.9434, + "step": 24845 + }, + { + "epoch": 0.79, + "learning_rate": 2.132663452580421e-06, + "loss": 0.8408, + "step": 24846 + }, + { + "epoch": 0.79, + "learning_rate": 2.1320241008179165e-06, + "loss": 0.9624, + "step": 24847 + }, + { + "epoch": 0.79, + "learning_rate": 2.131384833470088e-06, + "loss": 0.8203, + "step": 24848 + }, + { + "epoch": 0.79, + "learning_rate": 2.1307456505437918e-06, + "loss": 0.8325, + "step": 24849 + }, + { + "epoch": 0.79, + "learning_rate": 2.1301065520458853e-06, + "loss": 0.9819, + "step": 24850 + }, + { + "epoch": 0.79, + "learning_rate": 2.129467537983222e-06, + "loss": 0.9121, + "step": 24851 + }, + { + "epoch": 0.79, + "learning_rate": 2.128828608362665e-06, + "loss": 0.9155, + "step": 24852 + }, + { + "epoch": 0.79, + "learning_rate": 2.128189763191062e-06, + "loss": 0.8364, + "step": 24853 + }, + { + "epoch": 0.79, + "learning_rate": 2.1275510024752745e-06, + "loss": 0.9741, + "step": 24854 + }, + { + "epoch": 0.79, + "learning_rate": 2.1269123262221513e-06, + "loss": 0.9126, + "step": 24855 + }, + { + "epoch": 0.79, + "learning_rate": 2.126273734438543e-06, + "loss": 0.8589, + "step": 24856 + }, + { + "epoch": 0.79, + "learning_rate": 2.1256352271313062e-06, + "loss": 0.9185, + "step": 24857 + }, + { + "epoch": 0.79, + "learning_rate": 2.1249968043072877e-06, + "loss": 0.9888, + "step": 24858 + }, + { + "epoch": 0.8, + "learning_rate": 2.1243584659733386e-06, + "loss": 1.0098, + "step": 24859 + }, + { + "epoch": 0.8, + "learning_rate": 2.1237202121363043e-06, + "loss": 0.9185, + "step": 24860 + }, + { + "epoch": 0.8, + "learning_rate": 2.1230820428030376e-06, + "loss": 1.0527, + "step": 24861 + }, + { + "epoch": 0.8, + "learning_rate": 2.122443957980379e-06, + "loss": 0.9766, + "step": 24862 + }, + { + "epoch": 0.8, + "learning_rate": 2.1218059576751827e-06, + "loss": 0.9287, + "step": 24863 + }, + { + "epoch": 0.8, + "learning_rate": 2.121168041894286e-06, + "loss": 0.8301, + "step": 24864 + }, + { + "epoch": 0.8, + "learning_rate": 2.1205302106445412e-06, + "loss": 0.9971, + "step": 24865 + }, + { + "epoch": 0.8, + "learning_rate": 2.119892463932781e-06, + "loss": 0.9092, + "step": 24866 + }, + { + "epoch": 0.8, + "learning_rate": 2.1192548017658576e-06, + "loss": 0.8989, + "step": 24867 + }, + { + "epoch": 0.8, + "learning_rate": 2.1186172241506043e-06, + "loss": 0.8853, + "step": 24868 + }, + { + "epoch": 0.8, + "learning_rate": 2.1179797310938688e-06, + "loss": 0.9243, + "step": 24869 + }, + { + "epoch": 0.8, + "learning_rate": 2.117342322602485e-06, + "loss": 0.8911, + "step": 24870 + }, + { + "epoch": 0.8, + "learning_rate": 2.1167049986832957e-06, + "loss": 0.8857, + "step": 24871 + }, + { + "epoch": 0.8, + "learning_rate": 2.116067759343137e-06, + "loss": 0.8467, + "step": 24872 + }, + { + "epoch": 0.8, + "learning_rate": 2.115430604588846e-06, + "loss": 0.9351, + "step": 24873 + }, + { + "epoch": 0.8, + "learning_rate": 2.1147935344272562e-06, + "loss": 0.833, + "step": 24874 + }, + { + "epoch": 0.8, + "learning_rate": 2.1141565488652073e-06, + "loss": 0.936, + "step": 24875 + }, + { + "epoch": 0.8, + "learning_rate": 2.113519647909528e-06, + "loss": 0.9355, + "step": 24876 + }, + { + "epoch": 0.8, + "learning_rate": 2.112882831567058e-06, + "loss": 0.8896, + "step": 24877 + }, + { + "epoch": 0.8, + "learning_rate": 2.1122460998446226e-06, + "loss": 0.8545, + "step": 24878 + }, + { + "epoch": 0.8, + "learning_rate": 2.1116094527490594e-06, + "loss": 0.9482, + "step": 24879 + }, + { + "epoch": 0.8, + "learning_rate": 2.1109728902871974e-06, + "loss": 1.0142, + "step": 24880 + }, + { + "epoch": 0.8, + "learning_rate": 2.110336412465864e-06, + "loss": 0.9253, + "step": 24881 + }, + { + "epoch": 0.8, + "learning_rate": 2.109700019291887e-06, + "loss": 0.939, + "step": 24882 + }, + { + "epoch": 0.8, + "learning_rate": 2.109063710772099e-06, + "loss": 0.8911, + "step": 24883 + }, + { + "epoch": 0.8, + "learning_rate": 2.1084274869133213e-06, + "loss": 0.9106, + "step": 24884 + }, + { + "epoch": 0.8, + "learning_rate": 2.1077913477223853e-06, + "loss": 1.0156, + "step": 24885 + }, + { + "epoch": 0.8, + "learning_rate": 2.1071552932061113e-06, + "loss": 0.9722, + "step": 24886 + }, + { + "epoch": 0.8, + "learning_rate": 2.10651932337133e-06, + "loss": 0.8965, + "step": 24887 + }, + { + "epoch": 0.8, + "learning_rate": 2.105883438224856e-06, + "loss": 0.9263, + "step": 24888 + }, + { + "epoch": 0.8, + "learning_rate": 2.1052476377735178e-06, + "loss": 1.0005, + "step": 24889 + }, + { + "epoch": 0.8, + "learning_rate": 2.1046119220241324e-06, + "loss": 0.8872, + "step": 24890 + }, + { + "epoch": 0.8, + "learning_rate": 2.103976290983526e-06, + "loss": 0.8848, + "step": 24891 + }, + { + "epoch": 0.8, + "learning_rate": 2.103340744658512e-06, + "loss": 0.98, + "step": 24892 + }, + { + "epoch": 0.8, + "learning_rate": 2.1027052830559148e-06, + "loss": 0.9419, + "step": 24893 + }, + { + "epoch": 0.8, + "learning_rate": 2.10206990618255e-06, + "loss": 0.9453, + "step": 24894 + }, + { + "epoch": 0.8, + "learning_rate": 2.101434614045235e-06, + "loss": 0.9355, + "step": 24895 + }, + { + "epoch": 0.8, + "learning_rate": 2.100799406650781e-06, + "loss": 0.9116, + "step": 24896 + }, + { + "epoch": 0.8, + "learning_rate": 2.10016428400601e-06, + "loss": 0.8638, + "step": 24897 + }, + { + "epoch": 0.8, + "learning_rate": 2.099529246117733e-06, + "loss": 0.9175, + "step": 24898 + }, + { + "epoch": 0.8, + "learning_rate": 2.0988942929927603e-06, + "loss": 0.9517, + "step": 24899 + }, + { + "epoch": 0.8, + "learning_rate": 2.0982594246379108e-06, + "loss": 0.895, + "step": 24900 + }, + { + "epoch": 0.8, + "learning_rate": 2.097624641059989e-06, + "loss": 0.8882, + "step": 24901 + }, + { + "epoch": 0.8, + "learning_rate": 2.0969899422658114e-06, + "loss": 0.9399, + "step": 24902 + }, + { + "epoch": 0.8, + "learning_rate": 2.096355328262186e-06, + "loss": 0.9868, + "step": 24903 + }, + { + "epoch": 0.8, + "learning_rate": 2.09572079905592e-06, + "loss": 0.9346, + "step": 24904 + }, + { + "epoch": 0.8, + "learning_rate": 2.0950863546538204e-06, + "loss": 0.9395, + "step": 24905 + }, + { + "epoch": 0.8, + "learning_rate": 2.094451995062696e-06, + "loss": 0.9712, + "step": 24906 + }, + { + "epoch": 0.8, + "learning_rate": 2.0938177202893516e-06, + "loss": 0.8789, + "step": 24907 + }, + { + "epoch": 0.8, + "learning_rate": 2.0931835303405944e-06, + "loss": 0.9653, + "step": 24908 + }, + { + "epoch": 0.8, + "learning_rate": 2.092549425223225e-06, + "loss": 0.9263, + "step": 24909 + }, + { + "epoch": 0.8, + "learning_rate": 2.0919154049440536e-06, + "loss": 0.9072, + "step": 24910 + }, + { + "epoch": 0.8, + "learning_rate": 2.0912814695098725e-06, + "loss": 0.8584, + "step": 24911 + }, + { + "epoch": 0.8, + "learning_rate": 2.0906476189274914e-06, + "loss": 0.9185, + "step": 24912 + }, + { + "epoch": 0.8, + "learning_rate": 2.090013853203704e-06, + "loss": 0.915, + "step": 24913 + }, + { + "epoch": 0.8, + "learning_rate": 2.089380172345318e-06, + "loss": 0.4482, + "step": 24914 + }, + { + "epoch": 0.8, + "learning_rate": 2.088746576359123e-06, + "loss": 0.8882, + "step": 24915 + }, + { + "epoch": 0.8, + "learning_rate": 2.0881130652519256e-06, + "loss": 0.9028, + "step": 24916 + }, + { + "epoch": 0.8, + "learning_rate": 2.0874796390305185e-06, + "loss": 0.9536, + "step": 24917 + }, + { + "epoch": 0.8, + "learning_rate": 2.0868462977016967e-06, + "loss": 0.9634, + "step": 24918 + }, + { + "epoch": 0.8, + "learning_rate": 2.086213041272255e-06, + "loss": 0.9678, + "step": 24919 + }, + { + "epoch": 0.8, + "learning_rate": 2.0855798697489903e-06, + "loss": 0.8687, + "step": 24920 + }, + { + "epoch": 0.8, + "learning_rate": 2.084946783138693e-06, + "loss": 0.9272, + "step": 24921 + }, + { + "epoch": 0.8, + "learning_rate": 2.0843137814481584e-06, + "loss": 0.8936, + "step": 24922 + }, + { + "epoch": 0.8, + "learning_rate": 2.083680864684173e-06, + "loss": 0.8936, + "step": 24923 + }, + { + "epoch": 0.8, + "learning_rate": 2.083048032853534e-06, + "loss": 0.9561, + "step": 24924 + }, + { + "epoch": 0.8, + "learning_rate": 2.0824152859630274e-06, + "loss": 0.8345, + "step": 24925 + }, + { + "epoch": 0.8, + "learning_rate": 2.0817826240194426e-06, + "loss": 0.9268, + "step": 24926 + }, + { + "epoch": 0.8, + "learning_rate": 2.0811500470295642e-06, + "loss": 0.7488, + "step": 24927 + }, + { + "epoch": 0.8, + "learning_rate": 2.0805175550001835e-06, + "loss": 0.5125, + "step": 24928 + }, + { + "epoch": 0.8, + "learning_rate": 2.079885147938082e-06, + "loss": 1.0903, + "step": 24929 + }, + { + "epoch": 0.8, + "learning_rate": 2.079252825850051e-06, + "loss": 0.834, + "step": 24930 + }, + { + "epoch": 0.8, + "learning_rate": 2.0786205887428666e-06, + "loss": 0.9932, + "step": 24931 + }, + { + "epoch": 0.8, + "learning_rate": 2.0779884366233204e-06, + "loss": 1.0449, + "step": 24932 + }, + { + "epoch": 0.8, + "learning_rate": 2.0773563694981903e-06, + "loss": 0.8423, + "step": 24933 + }, + { + "epoch": 0.8, + "learning_rate": 2.076724387374258e-06, + "loss": 0.8643, + "step": 24934 + }, + { + "epoch": 0.8, + "learning_rate": 2.0760924902583014e-06, + "loss": 0.9434, + "step": 24935 + }, + { + "epoch": 0.8, + "learning_rate": 2.0754606781571053e-06, + "loss": 0.7881, + "step": 24936 + }, + { + "epoch": 0.8, + "learning_rate": 2.0748289510774456e-06, + "loss": 0.98, + "step": 24937 + }, + { + "epoch": 0.8, + "learning_rate": 2.074197309026097e-06, + "loss": 0.832, + "step": 24938 + }, + { + "epoch": 0.8, + "learning_rate": 2.073565752009842e-06, + "loss": 0.8511, + "step": 24939 + }, + { + "epoch": 0.8, + "learning_rate": 2.0729342800354546e-06, + "loss": 1.0254, + "step": 24940 + }, + { + "epoch": 0.8, + "learning_rate": 2.0723028931097067e-06, + "loss": 1.0513, + "step": 24941 + }, + { + "epoch": 0.8, + "learning_rate": 2.071671591239377e-06, + "loss": 0.9282, + "step": 24942 + }, + { + "epoch": 0.8, + "learning_rate": 2.0710403744312358e-06, + "loss": 0.9023, + "step": 24943 + }, + { + "epoch": 0.8, + "learning_rate": 2.0704092426920543e-06, + "loss": 0.8008, + "step": 24944 + }, + { + "epoch": 0.8, + "learning_rate": 2.069778196028608e-06, + "loss": 0.8682, + "step": 24945 + }, + { + "epoch": 0.8, + "learning_rate": 2.0691472344476627e-06, + "loss": 0.4792, + "step": 24946 + }, + { + "epoch": 0.8, + "learning_rate": 2.068516357955992e-06, + "loss": 0.9536, + "step": 24947 + }, + { + "epoch": 0.8, + "learning_rate": 2.0678855665603624e-06, + "loss": 1.0371, + "step": 24948 + }, + { + "epoch": 0.8, + "learning_rate": 2.0672548602675437e-06, + "loss": 0.9321, + "step": 24949 + }, + { + "epoch": 0.8, + "learning_rate": 2.066624239084296e-06, + "loss": 0.9243, + "step": 24950 + }, + { + "epoch": 0.8, + "learning_rate": 2.0659937030173937e-06, + "loss": 0.9453, + "step": 24951 + }, + { + "epoch": 0.8, + "learning_rate": 2.0653632520735956e-06, + "loss": 0.9448, + "step": 24952 + }, + { + "epoch": 0.8, + "learning_rate": 2.0647328862596706e-06, + "loss": 0.7793, + "step": 24953 + }, + { + "epoch": 0.8, + "learning_rate": 2.0641026055823765e-06, + "loss": 0.9009, + "step": 24954 + }, + { + "epoch": 0.8, + "learning_rate": 2.0634724100484836e-06, + "loss": 0.978, + "step": 24955 + }, + { + "epoch": 0.8, + "learning_rate": 2.0628422996647436e-06, + "loss": 0.8169, + "step": 24956 + }, + { + "epoch": 0.8, + "learning_rate": 2.0622122744379226e-06, + "loss": 0.8843, + "step": 24957 + }, + { + "epoch": 0.8, + "learning_rate": 2.0615823343747777e-06, + "loss": 0.8511, + "step": 24958 + }, + { + "epoch": 0.8, + "learning_rate": 2.06095247948207e-06, + "loss": 0.9136, + "step": 24959 + }, + { + "epoch": 0.8, + "learning_rate": 2.0603227097665535e-06, + "loss": 0.9131, + "step": 24960 + }, + { + "epoch": 0.8, + "learning_rate": 2.05969302523499e-06, + "loss": 0.9648, + "step": 24961 + }, + { + "epoch": 0.8, + "learning_rate": 2.059063425894129e-06, + "loss": 0.5081, + "step": 24962 + }, + { + "epoch": 0.8, + "learning_rate": 2.0584339117507346e-06, + "loss": 0.9087, + "step": 24963 + }, + { + "epoch": 0.8, + "learning_rate": 2.0578044828115507e-06, + "loss": 0.8911, + "step": 24964 + }, + { + "epoch": 0.8, + "learning_rate": 2.0571751390833363e-06, + "loss": 0.8838, + "step": 24965 + }, + { + "epoch": 0.8, + "learning_rate": 2.0565458805728388e-06, + "loss": 0.9609, + "step": 24966 + }, + { + "epoch": 0.8, + "learning_rate": 2.0559167072868158e-06, + "loss": 0.917, + "step": 24967 + }, + { + "epoch": 0.8, + "learning_rate": 2.0552876192320126e-06, + "loss": 1.0127, + "step": 24968 + }, + { + "epoch": 0.8, + "learning_rate": 2.0546586164151827e-06, + "loss": 0.9292, + "step": 24969 + }, + { + "epoch": 0.8, + "learning_rate": 2.054029698843072e-06, + "loss": 0.8823, + "step": 24970 + }, + { + "epoch": 0.8, + "learning_rate": 2.053400866522428e-06, + "loss": 0.9438, + "step": 24971 + }, + { + "epoch": 0.8, + "learning_rate": 2.052772119459996e-06, + "loss": 0.9497, + "step": 24972 + }, + { + "epoch": 0.8, + "learning_rate": 2.0521434576625265e-06, + "loss": 0.8735, + "step": 24973 + }, + { + "epoch": 0.8, + "learning_rate": 2.051514881136758e-06, + "loss": 0.8301, + "step": 24974 + }, + { + "epoch": 0.8, + "learning_rate": 2.0508863898894416e-06, + "loss": 0.8496, + "step": 24975 + }, + { + "epoch": 0.8, + "learning_rate": 2.050257983927313e-06, + "loss": 0.9204, + "step": 24976 + }, + { + "epoch": 0.8, + "learning_rate": 2.049629663257121e-06, + "loss": 0.9937, + "step": 24977 + }, + { + "epoch": 0.8, + "learning_rate": 2.049001427885603e-06, + "loss": 0.9536, + "step": 24978 + }, + { + "epoch": 0.8, + "learning_rate": 2.048373277819501e-06, + "loss": 0.9736, + "step": 24979 + }, + { + "epoch": 0.8, + "learning_rate": 2.0477452130655497e-06, + "loss": 0.9287, + "step": 24980 + }, + { + "epoch": 0.8, + "learning_rate": 2.0471172336304945e-06, + "loss": 0.957, + "step": 24981 + }, + { + "epoch": 0.8, + "learning_rate": 2.0464893395210695e-06, + "loss": 0.9097, + "step": 24982 + }, + { + "epoch": 0.8, + "learning_rate": 2.0458615307440077e-06, + "loss": 0.9482, + "step": 24983 + }, + { + "epoch": 0.8, + "learning_rate": 2.045233807306052e-06, + "loss": 0.9302, + "step": 24984 + }, + { + "epoch": 0.8, + "learning_rate": 2.044606169213934e-06, + "loss": 0.875, + "step": 24985 + }, + { + "epoch": 0.8, + "learning_rate": 2.0439786164743835e-06, + "loss": 0.8511, + "step": 24986 + }, + { + "epoch": 0.8, + "learning_rate": 2.0433511490941404e-06, + "loss": 0.9023, + "step": 24987 + }, + { + "epoch": 0.8, + "learning_rate": 2.042723767079934e-06, + "loss": 0.9751, + "step": 24988 + }, + { + "epoch": 0.8, + "learning_rate": 2.0420964704384916e-06, + "loss": 1.1084, + "step": 24989 + }, + { + "epoch": 0.8, + "learning_rate": 2.0414692591765494e-06, + "loss": 0.9092, + "step": 24990 + }, + { + "epoch": 0.8, + "learning_rate": 2.0408421333008312e-06, + "loss": 0.9824, + "step": 24991 + }, + { + "epoch": 0.8, + "learning_rate": 2.040215092818071e-06, + "loss": 0.9629, + "step": 24992 + }, + { + "epoch": 0.8, + "learning_rate": 2.0395881377349933e-06, + "loss": 0.9238, + "step": 24993 + }, + { + "epoch": 0.8, + "learning_rate": 2.038961268058325e-06, + "loss": 0.877, + "step": 24994 + }, + { + "epoch": 0.8, + "learning_rate": 2.0383344837947895e-06, + "loss": 0.4478, + "step": 24995 + }, + { + "epoch": 0.8, + "learning_rate": 2.037707784951115e-06, + "loss": 0.8535, + "step": 24996 + }, + { + "epoch": 0.8, + "learning_rate": 2.037081171534021e-06, + "loss": 1.0366, + "step": 24997 + }, + { + "epoch": 0.8, + "learning_rate": 2.036454643550236e-06, + "loss": 0.8643, + "step": 24998 + }, + { + "epoch": 0.8, + "learning_rate": 2.0358282010064766e-06, + "loss": 1.0044, + "step": 24999 + }, + { + "epoch": 0.8, + "learning_rate": 2.035201843909471e-06, + "loss": 0.4775, + "step": 25000 + }, + { + "epoch": 0.8, + "learning_rate": 2.03457557226593e-06, + "loss": 0.813, + "step": 25001 + }, + { + "epoch": 0.8, + "learning_rate": 2.0339493860825797e-06, + "loss": 0.9482, + "step": 25002 + }, + { + "epoch": 0.8, + "learning_rate": 2.0333232853661344e-06, + "loss": 0.8618, + "step": 25003 + }, + { + "epoch": 0.8, + "learning_rate": 2.032697270123315e-06, + "loss": 0.9844, + "step": 25004 + }, + { + "epoch": 0.8, + "learning_rate": 2.0320713403608337e-06, + "loss": 1.0161, + "step": 25005 + }, + { + "epoch": 0.8, + "learning_rate": 2.0314454960854114e-06, + "loss": 0.9341, + "step": 25006 + }, + { + "epoch": 0.8, + "learning_rate": 2.030819737303756e-06, + "loss": 0.9482, + "step": 25007 + }, + { + "epoch": 0.8, + "learning_rate": 2.0301940640225913e-06, + "loss": 0.9121, + "step": 25008 + }, + { + "epoch": 0.8, + "learning_rate": 2.0295684762486177e-06, + "loss": 0.9341, + "step": 25009 + }, + { + "epoch": 0.8, + "learning_rate": 2.0289429739885557e-06, + "loss": 0.9551, + "step": 25010 + }, + { + "epoch": 0.8, + "learning_rate": 2.028317557249112e-06, + "loss": 0.9058, + "step": 25011 + }, + { + "epoch": 0.8, + "learning_rate": 2.027692226037e-06, + "loss": 0.9873, + "step": 25012 + }, + { + "epoch": 0.8, + "learning_rate": 2.027066980358925e-06, + "loss": 0.853, + "step": 25013 + }, + { + "epoch": 0.8, + "learning_rate": 2.0264418202215998e-06, + "loss": 0.9976, + "step": 25014 + }, + { + "epoch": 0.8, + "learning_rate": 2.0258167456317267e-06, + "loss": 0.9634, + "step": 25015 + }, + { + "epoch": 0.8, + "learning_rate": 2.02519175659602e-06, + "loss": 0.9912, + "step": 25016 + }, + { + "epoch": 0.8, + "learning_rate": 2.0245668531211736e-06, + "loss": 0.9009, + "step": 25017 + }, + { + "epoch": 0.8, + "learning_rate": 2.0239420352139016e-06, + "loss": 1.0034, + "step": 25018 + }, + { + "epoch": 0.8, + "learning_rate": 2.0233173028809008e-06, + "loss": 0.9126, + "step": 25019 + }, + { + "epoch": 0.8, + "learning_rate": 2.0226926561288796e-06, + "loss": 0.8496, + "step": 25020 + }, + { + "epoch": 0.8, + "learning_rate": 2.0220680949645378e-06, + "loss": 0.8096, + "step": 25021 + }, + { + "epoch": 0.8, + "learning_rate": 2.021443619394573e-06, + "loss": 0.8394, + "step": 25022 + }, + { + "epoch": 0.8, + "learning_rate": 2.020819229425691e-06, + "loss": 0.8535, + "step": 25023 + }, + { + "epoch": 0.8, + "learning_rate": 2.0201949250645882e-06, + "loss": 0.9766, + "step": 25024 + }, + { + "epoch": 0.8, + "learning_rate": 2.0195707063179583e-06, + "loss": 0.9736, + "step": 25025 + }, + { + "epoch": 0.8, + "learning_rate": 2.018946573192505e-06, + "loss": 0.8721, + "step": 25026 + }, + { + "epoch": 0.8, + "learning_rate": 2.0183225256949233e-06, + "loss": 0.9058, + "step": 25027 + }, + { + "epoch": 0.8, + "learning_rate": 2.0176985638319034e-06, + "loss": 0.9272, + "step": 25028 + }, + { + "epoch": 0.8, + "learning_rate": 2.0170746876101465e-06, + "loss": 0.9878, + "step": 25029 + }, + { + "epoch": 0.8, + "learning_rate": 2.0164508970363404e-06, + "loss": 0.4802, + "step": 25030 + }, + { + "epoch": 0.8, + "learning_rate": 2.0158271921171847e-06, + "loss": 0.9526, + "step": 25031 + }, + { + "epoch": 0.8, + "learning_rate": 2.0152035728593645e-06, + "loss": 0.832, + "step": 25032 + }, + { + "epoch": 0.8, + "learning_rate": 2.0145800392695736e-06, + "loss": 0.8926, + "step": 25033 + }, + { + "epoch": 0.8, + "learning_rate": 2.013956591354499e-06, + "loss": 0.9033, + "step": 25034 + }, + { + "epoch": 0.8, + "learning_rate": 2.013333229120834e-06, + "loss": 0.9458, + "step": 25035 + }, + { + "epoch": 0.8, + "learning_rate": 2.0127099525752615e-06, + "loss": 0.9976, + "step": 25036 + }, + { + "epoch": 0.8, + "learning_rate": 2.012086761724473e-06, + "loss": 0.5291, + "step": 25037 + }, + { + "epoch": 0.8, + "learning_rate": 2.0114636565751544e-06, + "loss": 0.7852, + "step": 25038 + }, + { + "epoch": 0.8, + "learning_rate": 2.0108406371339894e-06, + "loss": 1.0049, + "step": 25039 + }, + { + "epoch": 0.8, + "learning_rate": 2.0102177034076585e-06, + "loss": 0.96, + "step": 25040 + }, + { + "epoch": 0.8, + "learning_rate": 2.0095948554028534e-06, + "loss": 0.4939, + "step": 25041 + }, + { + "epoch": 0.8, + "learning_rate": 2.0089720931262478e-06, + "loss": 0.978, + "step": 25042 + }, + { + "epoch": 0.8, + "learning_rate": 2.008349416584532e-06, + "loss": 0.8911, + "step": 25043 + }, + { + "epoch": 0.8, + "learning_rate": 2.0077268257843784e-06, + "loss": 0.9551, + "step": 25044 + }, + { + "epoch": 0.8, + "learning_rate": 2.0071043207324724e-06, + "loss": 0.9395, + "step": 25045 + }, + { + "epoch": 0.8, + "learning_rate": 2.006481901435493e-06, + "loss": 0.9517, + "step": 25046 + }, + { + "epoch": 0.8, + "learning_rate": 2.005859567900115e-06, + "loss": 0.918, + "step": 25047 + }, + { + "epoch": 0.8, + "learning_rate": 2.005237320133013e-06, + "loss": 0.9644, + "step": 25048 + }, + { + "epoch": 0.8, + "learning_rate": 2.00461515814087e-06, + "loss": 0.8745, + "step": 25049 + }, + { + "epoch": 0.8, + "learning_rate": 2.0039930819303543e-06, + "loss": 1.0083, + "step": 25050 + }, + { + "epoch": 0.8, + "learning_rate": 2.003371091508146e-06, + "loss": 0.9302, + "step": 25051 + }, + { + "epoch": 0.8, + "learning_rate": 2.0027491868809137e-06, + "loss": 0.9805, + "step": 25052 + }, + { + "epoch": 0.8, + "learning_rate": 2.0021273680553367e-06, + "loss": 0.8247, + "step": 25053 + }, + { + "epoch": 0.8, + "learning_rate": 2.001505635038076e-06, + "loss": 0.8784, + "step": 25054 + }, + { + "epoch": 0.8, + "learning_rate": 2.00088398783581e-06, + "loss": 0.8955, + "step": 25055 + }, + { + "epoch": 0.8, + "learning_rate": 2.0002624264552027e-06, + "loss": 1.02, + "step": 25056 + }, + { + "epoch": 0.8, + "learning_rate": 1.999640950902929e-06, + "loss": 0.9131, + "step": 25057 + }, + { + "epoch": 0.8, + "learning_rate": 1.999019561185651e-06, + "loss": 1.0508, + "step": 25058 + }, + { + "epoch": 0.8, + "learning_rate": 1.9983982573100413e-06, + "loss": 0.9243, + "step": 25059 + }, + { + "epoch": 0.8, + "learning_rate": 1.997777039282762e-06, + "loss": 0.8525, + "step": 25060 + }, + { + "epoch": 0.8, + "learning_rate": 1.9971559071104785e-06, + "loss": 0.9629, + "step": 25061 + }, + { + "epoch": 0.8, + "learning_rate": 1.9965348607998525e-06, + "loss": 0.8955, + "step": 25062 + }, + { + "epoch": 0.8, + "learning_rate": 1.9959139003575523e-06, + "loss": 0.812, + "step": 25063 + }, + { + "epoch": 0.8, + "learning_rate": 1.995293025790235e-06, + "loss": 0.9287, + "step": 25064 + }, + { + "epoch": 0.8, + "learning_rate": 1.9946722371045668e-06, + "loss": 1.0054, + "step": 25065 + }, + { + "epoch": 0.8, + "learning_rate": 1.9940515343072064e-06, + "loss": 0.9434, + "step": 25066 + }, + { + "epoch": 0.8, + "learning_rate": 1.993430917404808e-06, + "loss": 0.9023, + "step": 25067 + }, + { + "epoch": 0.8, + "learning_rate": 1.9928103864040394e-06, + "loss": 0.8496, + "step": 25068 + }, + { + "epoch": 0.8, + "learning_rate": 1.992189941311552e-06, + "loss": 0.8926, + "step": 25069 + }, + { + "epoch": 0.8, + "learning_rate": 1.991569582134002e-06, + "loss": 0.8311, + "step": 25070 + }, + { + "epoch": 0.8, + "learning_rate": 1.9909493088780496e-06, + "loss": 0.9028, + "step": 25071 + }, + { + "epoch": 0.8, + "learning_rate": 1.990329121550346e-06, + "loss": 0.8555, + "step": 25072 + }, + { + "epoch": 0.8, + "learning_rate": 1.9897090201575455e-06, + "loss": 0.9966, + "step": 25073 + }, + { + "epoch": 0.8, + "learning_rate": 1.989089004706304e-06, + "loss": 0.9272, + "step": 25074 + }, + { + "epoch": 0.8, + "learning_rate": 1.9884690752032674e-06, + "loss": 0.9185, + "step": 25075 + }, + { + "epoch": 0.8, + "learning_rate": 1.987849231655098e-06, + "loss": 0.8203, + "step": 25076 + }, + { + "epoch": 0.8, + "learning_rate": 1.987229474068433e-06, + "loss": 0.9517, + "step": 25077 + }, + { + "epoch": 0.8, + "learning_rate": 1.986609802449929e-06, + "loss": 0.9072, + "step": 25078 + }, + { + "epoch": 0.8, + "learning_rate": 1.9859902168062316e-06, + "loss": 1.04, + "step": 25079 + }, + { + "epoch": 0.8, + "learning_rate": 1.9853707171439927e-06, + "loss": 0.833, + "step": 25080 + }, + { + "epoch": 0.8, + "learning_rate": 1.984751303469853e-06, + "loss": 0.9688, + "step": 25081 + }, + { + "epoch": 0.8, + "learning_rate": 1.9841319757904633e-06, + "loss": 0.8521, + "step": 25082 + }, + { + "epoch": 0.8, + "learning_rate": 1.983512734112466e-06, + "loss": 0.9189, + "step": 25083 + }, + { + "epoch": 0.8, + "learning_rate": 1.982893578442505e-06, + "loss": 0.8774, + "step": 25084 + }, + { + "epoch": 0.8, + "learning_rate": 1.982274508787221e-06, + "loss": 0.8989, + "step": 25085 + }, + { + "epoch": 0.8, + "learning_rate": 1.98165552515326e-06, + "loss": 0.8022, + "step": 25086 + }, + { + "epoch": 0.8, + "learning_rate": 1.981036627547258e-06, + "loss": 0.9219, + "step": 25087 + }, + { + "epoch": 0.8, + "learning_rate": 1.9804178159758604e-06, + "loss": 0.957, + "step": 25088 + }, + { + "epoch": 0.8, + "learning_rate": 1.979799090445702e-06, + "loss": 0.9019, + "step": 25089 + }, + { + "epoch": 0.8, + "learning_rate": 1.979180450963425e-06, + "loss": 0.8872, + "step": 25090 + }, + { + "epoch": 0.8, + "learning_rate": 1.978561897535666e-06, + "loss": 1.0967, + "step": 25091 + }, + { + "epoch": 0.8, + "learning_rate": 1.9779434301690592e-06, + "loss": 1.0078, + "step": 25092 + }, + { + "epoch": 0.8, + "learning_rate": 1.977325048870239e-06, + "loss": 0.9858, + "step": 25093 + }, + { + "epoch": 0.8, + "learning_rate": 1.976706753645844e-06, + "loss": 0.8843, + "step": 25094 + }, + { + "epoch": 0.8, + "learning_rate": 1.9760885445025035e-06, + "loss": 0.7639, + "step": 25095 + }, + { + "epoch": 0.8, + "learning_rate": 1.9754704214468545e-06, + "loss": 0.9116, + "step": 25096 + }, + { + "epoch": 0.8, + "learning_rate": 1.974852384485525e-06, + "loss": 0.9165, + "step": 25097 + }, + { + "epoch": 0.8, + "learning_rate": 1.9742344336251497e-06, + "loss": 0.9609, + "step": 25098 + }, + { + "epoch": 0.8, + "learning_rate": 1.9736165688723563e-06, + "loss": 0.9341, + "step": 25099 + }, + { + "epoch": 0.8, + "learning_rate": 1.9729987902337744e-06, + "loss": 0.9771, + "step": 25100 + }, + { + "epoch": 0.8, + "learning_rate": 1.9723810977160294e-06, + "loss": 0.8301, + "step": 25101 + }, + { + "epoch": 0.8, + "learning_rate": 1.971763491325752e-06, + "loss": 1.0278, + "step": 25102 + }, + { + "epoch": 0.8, + "learning_rate": 1.9711459710695656e-06, + "loss": 0.9956, + "step": 25103 + }, + { + "epoch": 0.8, + "learning_rate": 1.9705285369540994e-06, + "loss": 1.0015, + "step": 25104 + }, + { + "epoch": 0.8, + "learning_rate": 1.9699111889859755e-06, + "loss": 0.9727, + "step": 25105 + }, + { + "epoch": 0.8, + "learning_rate": 1.969293927171817e-06, + "loss": 1.0298, + "step": 25106 + }, + { + "epoch": 0.8, + "learning_rate": 1.9686767515182448e-06, + "loss": 0.8477, + "step": 25107 + }, + { + "epoch": 0.8, + "learning_rate": 1.9680596620318847e-06, + "loss": 0.8613, + "step": 25108 + }, + { + "epoch": 0.8, + "learning_rate": 1.967442658719352e-06, + "loss": 0.9624, + "step": 25109 + }, + { + "epoch": 0.8, + "learning_rate": 1.9668257415872726e-06, + "loss": 0.9619, + "step": 25110 + }, + { + "epoch": 0.8, + "learning_rate": 1.966208910642262e-06, + "loss": 1.0195, + "step": 25111 + }, + { + "epoch": 0.8, + "learning_rate": 1.965592165890936e-06, + "loss": 0.8901, + "step": 25112 + }, + { + "epoch": 0.8, + "learning_rate": 1.964975507339917e-06, + "loss": 0.8755, + "step": 25113 + }, + { + "epoch": 0.8, + "learning_rate": 1.964358934995818e-06, + "loss": 0.9961, + "step": 25114 + }, + { + "epoch": 0.8, + "learning_rate": 1.9637424488652537e-06, + "loss": 0.8049, + "step": 25115 + }, + { + "epoch": 0.8, + "learning_rate": 1.9631260489548365e-06, + "loss": 0.8174, + "step": 25116 + }, + { + "epoch": 0.8, + "learning_rate": 1.9625097352711853e-06, + "loss": 0.9268, + "step": 25117 + }, + { + "epoch": 0.8, + "learning_rate": 1.9618935078209054e-06, + "loss": 0.9053, + "step": 25118 + }, + { + "epoch": 0.8, + "learning_rate": 1.9612773666106155e-06, + "loss": 0.8755, + "step": 25119 + }, + { + "epoch": 0.8, + "learning_rate": 1.96066131164692e-06, + "loss": 0.9321, + "step": 25120 + }, + { + "epoch": 0.8, + "learning_rate": 1.9600453429364354e-06, + "loss": 0.8303, + "step": 25121 + }, + { + "epoch": 0.8, + "learning_rate": 1.959429460485761e-06, + "loss": 1.0112, + "step": 25122 + }, + { + "epoch": 0.8, + "learning_rate": 1.958813664301513e-06, + "loss": 0.9253, + "step": 25123 + }, + { + "epoch": 0.8, + "learning_rate": 1.9581979543902918e-06, + "loss": 1.0322, + "step": 25124 + }, + { + "epoch": 0.8, + "learning_rate": 1.9575823307587096e-06, + "loss": 1.0615, + "step": 25125 + }, + { + "epoch": 0.8, + "learning_rate": 1.956966793413365e-06, + "loss": 0.9346, + "step": 25126 + }, + { + "epoch": 0.8, + "learning_rate": 1.956351342360868e-06, + "loss": 1.0474, + "step": 25127 + }, + { + "epoch": 0.8, + "learning_rate": 1.9557359776078157e-06, + "loss": 1.0063, + "step": 25128 + }, + { + "epoch": 0.8, + "learning_rate": 1.9551206991608185e-06, + "loss": 0.9053, + "step": 25129 + }, + { + "epoch": 0.8, + "learning_rate": 1.9545055070264684e-06, + "loss": 0.8745, + "step": 25130 + }, + { + "epoch": 0.8, + "learning_rate": 1.9538904012113714e-06, + "loss": 0.8511, + "step": 25131 + }, + { + "epoch": 0.8, + "learning_rate": 1.9532753817221226e-06, + "loss": 0.8618, + "step": 25132 + }, + { + "epoch": 0.8, + "learning_rate": 1.9526604485653268e-06, + "loss": 0.9766, + "step": 25133 + }, + { + "epoch": 0.8, + "learning_rate": 1.9520456017475754e-06, + "loss": 0.9868, + "step": 25134 + }, + { + "epoch": 0.8, + "learning_rate": 1.95143084127547e-06, + "loss": 0.8574, + "step": 25135 + }, + { + "epoch": 0.8, + "learning_rate": 1.9508161671556037e-06, + "loss": 0.9507, + "step": 25136 + }, + { + "epoch": 0.8, + "learning_rate": 1.9502015793945715e-06, + "loss": 0.8374, + "step": 25137 + }, + { + "epoch": 0.8, + "learning_rate": 1.9495870779989656e-06, + "loss": 1.0967, + "step": 25138 + }, + { + "epoch": 0.8, + "learning_rate": 1.9489726629753826e-06, + "loss": 0.9258, + "step": 25139 + }, + { + "epoch": 0.8, + "learning_rate": 1.948358334330409e-06, + "loss": 0.9336, + "step": 25140 + }, + { + "epoch": 0.8, + "learning_rate": 1.9477440920706437e-06, + "loss": 0.8745, + "step": 25141 + }, + { + "epoch": 0.8, + "learning_rate": 1.9471299362026684e-06, + "loss": 0.9312, + "step": 25142 + }, + { + "epoch": 0.8, + "learning_rate": 1.9465158667330796e-06, + "loss": 0.9507, + "step": 25143 + }, + { + "epoch": 0.8, + "learning_rate": 1.9459018836684628e-06, + "loss": 1.0654, + "step": 25144 + }, + { + "epoch": 0.8, + "learning_rate": 1.945287987015404e-06, + "loss": 0.9492, + "step": 25145 + }, + { + "epoch": 0.8, + "learning_rate": 1.944674176780489e-06, + "loss": 0.896, + "step": 25146 + }, + { + "epoch": 0.8, + "learning_rate": 1.9440604529703066e-06, + "loss": 0.8433, + "step": 25147 + }, + { + "epoch": 0.8, + "learning_rate": 1.9434468155914378e-06, + "loss": 0.979, + "step": 25148 + }, + { + "epoch": 0.8, + "learning_rate": 1.94283326465047e-06, + "loss": 0.9751, + "step": 25149 + }, + { + "epoch": 0.8, + "learning_rate": 1.9422198001539847e-06, + "loss": 0.8516, + "step": 25150 + }, + { + "epoch": 0.8, + "learning_rate": 1.941606422108563e-06, + "loss": 0.9287, + "step": 25151 + }, + { + "epoch": 0.8, + "learning_rate": 1.940993130520783e-06, + "loss": 0.9282, + "step": 25152 + }, + { + "epoch": 0.8, + "learning_rate": 1.94037992539723e-06, + "loss": 0.9458, + "step": 25153 + }, + { + "epoch": 0.8, + "learning_rate": 1.9397668067444807e-06, + "loss": 0.9917, + "step": 25154 + }, + { + "epoch": 0.8, + "learning_rate": 1.9391537745691102e-06, + "loss": 0.9678, + "step": 25155 + }, + { + "epoch": 0.8, + "learning_rate": 1.9385408288777007e-06, + "loss": 0.9658, + "step": 25156 + }, + { + "epoch": 0.8, + "learning_rate": 1.937927969676824e-06, + "loss": 0.7939, + "step": 25157 + }, + { + "epoch": 0.8, + "learning_rate": 1.9373151969730597e-06, + "loss": 0.958, + "step": 25158 + }, + { + "epoch": 0.8, + "learning_rate": 1.93670251077298e-06, + "loss": 0.8857, + "step": 25159 + }, + { + "epoch": 0.8, + "learning_rate": 1.9360899110831586e-06, + "loss": 0.9175, + "step": 25160 + }, + { + "epoch": 0.8, + "learning_rate": 1.935477397910165e-06, + "loss": 0.8872, + "step": 25161 + }, + { + "epoch": 0.8, + "learning_rate": 1.934864971260576e-06, + "loss": 0.7974, + "step": 25162 + }, + { + "epoch": 0.8, + "learning_rate": 1.934252631140958e-06, + "loss": 0.9209, + "step": 25163 + }, + { + "epoch": 0.8, + "learning_rate": 1.9336403775578837e-06, + "loss": 0.9014, + "step": 25164 + }, + { + "epoch": 0.8, + "learning_rate": 1.93302821051792e-06, + "loss": 0.8589, + "step": 25165 + }, + { + "epoch": 0.8, + "learning_rate": 1.9324161300276368e-06, + "loss": 1.0034, + "step": 25166 + }, + { + "epoch": 0.8, + "learning_rate": 1.9318041360936e-06, + "loss": 1.0708, + "step": 25167 + }, + { + "epoch": 0.8, + "learning_rate": 1.931192228722375e-06, + "loss": 0.9805, + "step": 25168 + }, + { + "epoch": 0.8, + "learning_rate": 1.9305804079205257e-06, + "loss": 0.9272, + "step": 25169 + }, + { + "epoch": 0.8, + "learning_rate": 1.92996867369462e-06, + "loss": 0.8193, + "step": 25170 + }, + { + "epoch": 0.8, + "learning_rate": 1.929357026051216e-06, + "loss": 0.9116, + "step": 25171 + }, + { + "epoch": 0.81, + "learning_rate": 1.9287454649968816e-06, + "loss": 0.9121, + "step": 25172 + }, + { + "epoch": 0.81, + "learning_rate": 1.9281339905381735e-06, + "loss": 0.8843, + "step": 25173 + }, + { + "epoch": 0.81, + "learning_rate": 1.9275226026816585e-06, + "loss": 0.4878, + "step": 25174 + }, + { + "epoch": 0.81, + "learning_rate": 1.926911301433888e-06, + "loss": 0.8293, + "step": 25175 + }, + { + "epoch": 0.81, + "learning_rate": 1.926300086801426e-06, + "loss": 1.0205, + "step": 25176 + }, + { + "epoch": 0.81, + "learning_rate": 1.9256889587908255e-06, + "loss": 0.9858, + "step": 25177 + }, + { + "epoch": 0.81, + "learning_rate": 1.9250779174086496e-06, + "loss": 0.9297, + "step": 25178 + }, + { + "epoch": 0.81, + "learning_rate": 1.9244669626614478e-06, + "loss": 0.9761, + "step": 25179 + }, + { + "epoch": 0.81, + "learning_rate": 1.9238560945557816e-06, + "loss": 0.9941, + "step": 25180 + }, + { + "epoch": 0.81, + "learning_rate": 1.923245313098198e-06, + "loss": 1.1372, + "step": 25181 + }, + { + "epoch": 0.81, + "learning_rate": 1.922634618295257e-06, + "loss": 0.8555, + "step": 25182 + }, + { + "epoch": 0.81, + "learning_rate": 1.9220240101535036e-06, + "loss": 0.8867, + "step": 25183 + }, + { + "epoch": 0.81, + "learning_rate": 1.9214134886794942e-06, + "loss": 0.9736, + "step": 25184 + }, + { + "epoch": 0.81, + "learning_rate": 1.920803053879775e-06, + "loss": 0.9761, + "step": 25185 + }, + { + "epoch": 0.81, + "learning_rate": 1.9201927057608994e-06, + "loss": 1.0293, + "step": 25186 + }, + { + "epoch": 0.81, + "learning_rate": 1.919582444329412e-06, + "loss": 0.9141, + "step": 25187 + }, + { + "epoch": 0.81, + "learning_rate": 1.9189722695918633e-06, + "loss": 0.9092, + "step": 25188 + }, + { + "epoch": 0.81, + "learning_rate": 1.9183621815547994e-06, + "loss": 0.8711, + "step": 25189 + }, + { + "epoch": 0.81, + "learning_rate": 1.9177521802247655e-06, + "loss": 0.4807, + "step": 25190 + }, + { + "epoch": 0.81, + "learning_rate": 1.9171422656083015e-06, + "loss": 0.8511, + "step": 25191 + }, + { + "epoch": 0.81, + "learning_rate": 1.91653243771196e-06, + "loss": 0.9478, + "step": 25192 + }, + { + "epoch": 0.81, + "learning_rate": 1.915922696542275e-06, + "loss": 0.9473, + "step": 25193 + }, + { + "epoch": 0.81, + "learning_rate": 1.9153130421057955e-06, + "loss": 0.9126, + "step": 25194 + }, + { + "epoch": 0.81, + "learning_rate": 1.91470347440906e-06, + "loss": 0.9536, + "step": 25195 + }, + { + "epoch": 0.81, + "learning_rate": 1.9140939934586033e-06, + "loss": 0.8955, + "step": 25196 + }, + { + "epoch": 0.81, + "learning_rate": 1.913484599260973e-06, + "loss": 0.855, + "step": 25197 + }, + { + "epoch": 0.81, + "learning_rate": 1.9128752918227034e-06, + "loss": 0.9751, + "step": 25198 + }, + { + "epoch": 0.81, + "learning_rate": 1.9122660711503316e-06, + "loss": 0.9077, + "step": 25199 + }, + { + "epoch": 0.81, + "learning_rate": 1.911656937250391e-06, + "loss": 0.9199, + "step": 25200 + }, + { + "epoch": 0.81, + "learning_rate": 1.9110478901294214e-06, + "loss": 0.9424, + "step": 25201 + }, + { + "epoch": 0.81, + "learning_rate": 1.910438929793954e-06, + "loss": 0.939, + "step": 25202 + }, + { + "epoch": 0.81, + "learning_rate": 1.9098300562505266e-06, + "loss": 0.8784, + "step": 25203 + }, + { + "epoch": 0.81, + "learning_rate": 1.9092212695056688e-06, + "loss": 0.8555, + "step": 25204 + }, + { + "epoch": 0.81, + "learning_rate": 1.908612569565912e-06, + "loss": 0.918, + "step": 25205 + }, + { + "epoch": 0.81, + "learning_rate": 1.908003956437784e-06, + "loss": 0.9819, + "step": 25206 + }, + { + "epoch": 0.81, + "learning_rate": 1.907395430127821e-06, + "loss": 0.8704, + "step": 25207 + }, + { + "epoch": 0.81, + "learning_rate": 1.9067869906425462e-06, + "loss": 0.9268, + "step": 25208 + }, + { + "epoch": 0.81, + "learning_rate": 1.9061786379884927e-06, + "loss": 0.8086, + "step": 25209 + }, + { + "epoch": 0.81, + "learning_rate": 1.9055703721721807e-06, + "loss": 0.918, + "step": 25210 + }, + { + "epoch": 0.81, + "learning_rate": 1.9049621932001439e-06, + "loss": 0.938, + "step": 25211 + }, + { + "epoch": 0.81, + "learning_rate": 1.9043541010789034e-06, + "loss": 0.9331, + "step": 25212 + }, + { + "epoch": 0.81, + "learning_rate": 1.9037460958149835e-06, + "loss": 0.5271, + "step": 25213 + }, + { + "epoch": 0.81, + "learning_rate": 1.9031381774149048e-06, + "loss": 0.9844, + "step": 25214 + }, + { + "epoch": 0.81, + "learning_rate": 1.9025303458851952e-06, + "loss": 0.8472, + "step": 25215 + }, + { + "epoch": 0.81, + "learning_rate": 1.901922601232371e-06, + "loss": 0.8279, + "step": 25216 + }, + { + "epoch": 0.81, + "learning_rate": 1.9013149434629562e-06, + "loss": 0.894, + "step": 25217 + }, + { + "epoch": 0.81, + "learning_rate": 1.900707372583467e-06, + "loss": 0.9009, + "step": 25218 + }, + { + "epoch": 0.81, + "learning_rate": 1.9000998886004285e-06, + "loss": 0.9062, + "step": 25219 + }, + { + "epoch": 0.81, + "learning_rate": 1.8994924915203494e-06, + "loss": 1.1245, + "step": 25220 + }, + { + "epoch": 0.81, + "learning_rate": 1.8988851813497523e-06, + "loss": 0.9346, + "step": 25221 + }, + { + "epoch": 0.81, + "learning_rate": 1.8982779580951493e-06, + "loss": 0.7915, + "step": 25222 + }, + { + "epoch": 0.81, + "learning_rate": 1.8976708217630602e-06, + "loss": 0.9556, + "step": 25223 + }, + { + "epoch": 0.81, + "learning_rate": 1.8970637723599927e-06, + "loss": 0.897, + "step": 25224 + }, + { + "epoch": 0.81, + "learning_rate": 1.8964568098924651e-06, + "loss": 0.8682, + "step": 25225 + }, + { + "epoch": 0.81, + "learning_rate": 1.8958499343669857e-06, + "loss": 0.8779, + "step": 25226 + }, + { + "epoch": 0.81, + "learning_rate": 1.895243145790072e-06, + "loss": 0.8809, + "step": 25227 + }, + { + "epoch": 0.81, + "learning_rate": 1.8946364441682253e-06, + "loss": 0.9834, + "step": 25228 + }, + { + "epoch": 0.81, + "learning_rate": 1.8940298295079607e-06, + "loss": 0.9321, + "step": 25229 + }, + { + "epoch": 0.81, + "learning_rate": 1.8934233018157822e-06, + "loss": 0.8789, + "step": 25230 + }, + { + "epoch": 0.81, + "learning_rate": 1.8928168610982022e-06, + "loss": 0.9717, + "step": 25231 + }, + { + "epoch": 0.81, + "learning_rate": 1.8922105073617224e-06, + "loss": 1.0107, + "step": 25232 + }, + { + "epoch": 0.81, + "learning_rate": 1.8916042406128533e-06, + "loss": 0.8354, + "step": 25233 + }, + { + "epoch": 0.81, + "learning_rate": 1.8909980608580968e-06, + "loss": 1.0386, + "step": 25234 + }, + { + "epoch": 0.81, + "learning_rate": 1.8903919681039562e-06, + "loss": 0.9634, + "step": 25235 + }, + { + "epoch": 0.81, + "learning_rate": 1.8897859623569314e-06, + "loss": 0.9189, + "step": 25236 + }, + { + "epoch": 0.81, + "learning_rate": 1.8891800436235308e-06, + "loss": 1.062, + "step": 25237 + }, + { + "epoch": 0.81, + "learning_rate": 1.8885742119102512e-06, + "loss": 0.938, + "step": 25238 + }, + { + "epoch": 0.81, + "learning_rate": 1.887968467223591e-06, + "loss": 0.9287, + "step": 25239 + }, + { + "epoch": 0.81, + "learning_rate": 1.8873628095700536e-06, + "loss": 0.9248, + "step": 25240 + }, + { + "epoch": 0.81, + "learning_rate": 1.8867572389561318e-06, + "loss": 1.0474, + "step": 25241 + }, + { + "epoch": 0.81, + "learning_rate": 1.886151755388328e-06, + "loss": 1.0723, + "step": 25242 + }, + { + "epoch": 0.81, + "learning_rate": 1.885546358873136e-06, + "loss": 0.96, + "step": 25243 + }, + { + "epoch": 0.81, + "learning_rate": 1.8849410494170518e-06, + "loss": 0.8325, + "step": 25244 + }, + { + "epoch": 0.81, + "learning_rate": 1.8843358270265644e-06, + "loss": 0.8477, + "step": 25245 + }, + { + "epoch": 0.81, + "learning_rate": 1.8837306917081756e-06, + "loss": 0.9775, + "step": 25246 + }, + { + "epoch": 0.81, + "learning_rate": 1.8831256434683709e-06, + "loss": 0.8643, + "step": 25247 + }, + { + "epoch": 0.81, + "learning_rate": 1.8825206823136477e-06, + "loss": 0.9434, + "step": 25248 + }, + { + "epoch": 0.81, + "learning_rate": 1.8819158082504906e-06, + "loss": 0.8711, + "step": 25249 + }, + { + "epoch": 0.81, + "learning_rate": 1.881311021285397e-06, + "loss": 0.7949, + "step": 25250 + }, + { + "epoch": 0.81, + "learning_rate": 1.8807063214248467e-06, + "loss": 1.0088, + "step": 25251 + }, + { + "epoch": 0.81, + "learning_rate": 1.8801017086753337e-06, + "loss": 0.9561, + "step": 25252 + }, + { + "epoch": 0.81, + "learning_rate": 1.8794971830433395e-06, + "loss": 0.9336, + "step": 25253 + }, + { + "epoch": 0.81, + "learning_rate": 1.8788927445353567e-06, + "loss": 0.9536, + "step": 25254 + }, + { + "epoch": 0.81, + "learning_rate": 1.8782883931578632e-06, + "loss": 0.938, + "step": 25255 + }, + { + "epoch": 0.81, + "learning_rate": 1.8776841289173497e-06, + "loss": 0.915, + "step": 25256 + }, + { + "epoch": 0.81, + "learning_rate": 1.8770799518202964e-06, + "loss": 0.897, + "step": 25257 + }, + { + "epoch": 0.81, + "learning_rate": 1.8764758618731848e-06, + "loss": 0.9951, + "step": 25258 + }, + { + "epoch": 0.81, + "learning_rate": 1.8758718590824942e-06, + "loss": 0.8335, + "step": 25259 + }, + { + "epoch": 0.81, + "learning_rate": 1.8752679434547105e-06, + "loss": 0.9097, + "step": 25260 + }, + { + "epoch": 0.81, + "learning_rate": 1.8746641149963063e-06, + "loss": 0.5195, + "step": 25261 + }, + { + "epoch": 0.81, + "learning_rate": 1.8740603737137664e-06, + "loss": 0.9961, + "step": 25262 + }, + { + "epoch": 0.81, + "learning_rate": 1.8734567196135622e-06, + "loss": 1.0273, + "step": 25263 + }, + { + "epoch": 0.81, + "learning_rate": 1.8728531527021765e-06, + "loss": 0.9546, + "step": 25264 + }, + { + "epoch": 0.81, + "learning_rate": 1.8722496729860817e-06, + "loss": 0.9688, + "step": 25265 + }, + { + "epoch": 0.81, + "learning_rate": 1.8716462804717516e-06, + "loss": 0.9292, + "step": 25266 + }, + { + "epoch": 0.81, + "learning_rate": 1.87104297516566e-06, + "loss": 1.0273, + "step": 25267 + }, + { + "epoch": 0.81, + "learning_rate": 1.8704397570742816e-06, + "loss": 0.9648, + "step": 25268 + }, + { + "epoch": 0.81, + "learning_rate": 1.8698366262040858e-06, + "loss": 0.8125, + "step": 25269 + }, + { + "epoch": 0.81, + "learning_rate": 1.8692335825615471e-06, + "loss": 0.8438, + "step": 25270 + }, + { + "epoch": 0.81, + "learning_rate": 1.8686306261531306e-06, + "loss": 0.7886, + "step": 25271 + }, + { + "epoch": 0.81, + "learning_rate": 1.8680277569853134e-06, + "loss": 0.948, + "step": 25272 + }, + { + "epoch": 0.81, + "learning_rate": 1.8674249750645534e-06, + "loss": 0.8818, + "step": 25273 + }, + { + "epoch": 0.81, + "learning_rate": 1.8668222803973258e-06, + "loss": 0.9199, + "step": 25274 + }, + { + "epoch": 0.81, + "learning_rate": 1.8662196729900905e-06, + "loss": 1.0044, + "step": 25275 + }, + { + "epoch": 0.81, + "learning_rate": 1.8656171528493184e-06, + "loss": 1.0005, + "step": 25276 + }, + { + "epoch": 0.81, + "learning_rate": 1.8650147199814728e-06, + "loss": 1.0464, + "step": 25277 + }, + { + "epoch": 0.81, + "learning_rate": 1.864412374393012e-06, + "loss": 1.0879, + "step": 25278 + }, + { + "epoch": 0.81, + "learning_rate": 1.8638101160904054e-06, + "loss": 0.8198, + "step": 25279 + }, + { + "epoch": 0.81, + "learning_rate": 1.863207945080112e-06, + "loss": 0.4653, + "step": 25280 + }, + { + "epoch": 0.81, + "learning_rate": 1.8626058613685882e-06, + "loss": 0.8633, + "step": 25281 + }, + { + "epoch": 0.81, + "learning_rate": 1.8620038649623018e-06, + "loss": 1.0078, + "step": 25282 + }, + { + "epoch": 0.81, + "learning_rate": 1.8614019558677066e-06, + "loss": 0.9292, + "step": 25283 + }, + { + "epoch": 0.81, + "learning_rate": 1.8608001340912573e-06, + "loss": 0.9014, + "step": 25284 + }, + { + "epoch": 0.81, + "learning_rate": 1.860198399639418e-06, + "loss": 0.8877, + "step": 25285 + }, + { + "epoch": 0.81, + "learning_rate": 1.8595967525186387e-06, + "loss": 0.8228, + "step": 25286 + }, + { + "epoch": 0.81, + "learning_rate": 1.85899519273538e-06, + "loss": 1.0332, + "step": 25287 + }, + { + "epoch": 0.81, + "learning_rate": 1.8583937202960922e-06, + "loss": 0.9546, + "step": 25288 + }, + { + "epoch": 0.81, + "learning_rate": 1.8577923352072292e-06, + "loss": 0.9976, + "step": 25289 + }, + { + "epoch": 0.81, + "learning_rate": 1.85719103747524e-06, + "loss": 0.8667, + "step": 25290 + }, + { + "epoch": 0.81, + "learning_rate": 1.856589827106583e-06, + "loss": 0.9697, + "step": 25291 + }, + { + "epoch": 0.81, + "learning_rate": 1.8559887041077007e-06, + "loss": 0.8091, + "step": 25292 + }, + { + "epoch": 0.81, + "learning_rate": 1.8553876684850492e-06, + "loss": 0.8916, + "step": 25293 + }, + { + "epoch": 0.81, + "learning_rate": 1.8547867202450709e-06, + "loss": 0.9814, + "step": 25294 + }, + { + "epoch": 0.81, + "learning_rate": 1.8541858593942219e-06, + "loss": 0.8354, + "step": 25295 + }, + { + "epoch": 0.81, + "learning_rate": 1.853585085938938e-06, + "loss": 0.8267, + "step": 25296 + }, + { + "epoch": 0.81, + "learning_rate": 1.8529843998856735e-06, + "loss": 0.8823, + "step": 25297 + }, + { + "epoch": 0.81, + "learning_rate": 1.8523838012408668e-06, + "loss": 0.9917, + "step": 25298 + }, + { + "epoch": 0.81, + "learning_rate": 1.8517832900109667e-06, + "loss": 0.9912, + "step": 25299 + }, + { + "epoch": 0.81, + "learning_rate": 1.851182866202411e-06, + "loss": 1.0322, + "step": 25300 + }, + { + "epoch": 0.81, + "learning_rate": 1.8505825298216473e-06, + "loss": 0.8906, + "step": 25301 + }, + { + "epoch": 0.81, + "learning_rate": 1.8499822808751144e-06, + "loss": 0.7959, + "step": 25302 + }, + { + "epoch": 0.81, + "learning_rate": 1.8493821193692518e-06, + "loss": 1.0796, + "step": 25303 + }, + { + "epoch": 0.81, + "learning_rate": 1.8487820453104954e-06, + "loss": 1.0205, + "step": 25304 + }, + { + "epoch": 0.81, + "learning_rate": 1.848182058705288e-06, + "loss": 0.9326, + "step": 25305 + }, + { + "epoch": 0.81, + "learning_rate": 1.847582159560064e-06, + "loss": 1.0522, + "step": 25306 + }, + { + "epoch": 0.81, + "learning_rate": 1.846982347881262e-06, + "loss": 0.9062, + "step": 25307 + }, + { + "epoch": 0.81, + "learning_rate": 1.8463826236753146e-06, + "loss": 1.0137, + "step": 25308 + }, + { + "epoch": 0.81, + "learning_rate": 1.8457829869486598e-06, + "loss": 0.9541, + "step": 25309 + }, + { + "epoch": 0.81, + "learning_rate": 1.8451834377077294e-06, + "loss": 0.9214, + "step": 25310 + }, + { + "epoch": 0.81, + "learning_rate": 1.8445839759589545e-06, + "loss": 0.8887, + "step": 25311 + }, + { + "epoch": 0.81, + "learning_rate": 1.8439846017087658e-06, + "loss": 0.9209, + "step": 25312 + }, + { + "epoch": 0.81, + "learning_rate": 1.8433853149635973e-06, + "loss": 0.9272, + "step": 25313 + }, + { + "epoch": 0.81, + "learning_rate": 1.8427861157298755e-06, + "loss": 0.9995, + "step": 25314 + }, + { + "epoch": 0.81, + "learning_rate": 1.8421870040140332e-06, + "loss": 0.9795, + "step": 25315 + }, + { + "epoch": 0.81, + "learning_rate": 1.8415879798224957e-06, + "loss": 0.8823, + "step": 25316 + }, + { + "epoch": 0.81, + "learning_rate": 1.8409890431616895e-06, + "loss": 0.9492, + "step": 25317 + }, + { + "epoch": 0.81, + "learning_rate": 1.8403901940380385e-06, + "loss": 0.9194, + "step": 25318 + }, + { + "epoch": 0.81, + "learning_rate": 1.8397914324579725e-06, + "loss": 0.9224, + "step": 25319 + }, + { + "epoch": 0.81, + "learning_rate": 1.8391927584279101e-06, + "loss": 0.7654, + "step": 25320 + }, + { + "epoch": 0.81, + "learning_rate": 1.838594171954281e-06, + "loss": 0.8887, + "step": 25321 + }, + { + "epoch": 0.81, + "learning_rate": 1.8379956730435033e-06, + "loss": 1.0757, + "step": 25322 + }, + { + "epoch": 0.81, + "learning_rate": 1.8373972617019964e-06, + "loss": 0.9224, + "step": 25323 + }, + { + "epoch": 0.81, + "learning_rate": 1.8367989379361861e-06, + "loss": 0.9365, + "step": 25324 + }, + { + "epoch": 0.81, + "learning_rate": 1.8362007017524886e-06, + "loss": 0.7947, + "step": 25325 + }, + { + "epoch": 0.81, + "learning_rate": 1.8356025531573185e-06, + "loss": 0.9844, + "step": 25326 + }, + { + "epoch": 0.81, + "learning_rate": 1.8350044921571009e-06, + "loss": 0.9004, + "step": 25327 + }, + { + "epoch": 0.81, + "learning_rate": 1.8344065187582483e-06, + "loss": 0.9253, + "step": 25328 + }, + { + "epoch": 0.81, + "learning_rate": 1.8338086329671734e-06, + "loss": 0.9536, + "step": 25329 + }, + { + "epoch": 0.81, + "learning_rate": 1.8332108347902966e-06, + "loss": 0.9019, + "step": 25330 + }, + { + "epoch": 0.81, + "learning_rate": 1.8326131242340274e-06, + "loss": 0.8257, + "step": 25331 + }, + { + "epoch": 0.81, + "learning_rate": 1.8320155013047813e-06, + "loss": 0.8169, + "step": 25332 + }, + { + "epoch": 0.81, + "learning_rate": 1.8314179660089693e-06, + "loss": 0.8433, + "step": 25333 + }, + { + "epoch": 0.81, + "learning_rate": 1.8308205183530025e-06, + "loss": 0.8862, + "step": 25334 + }, + { + "epoch": 0.81, + "learning_rate": 1.8302231583432883e-06, + "loss": 0.9971, + "step": 25335 + }, + { + "epoch": 0.81, + "learning_rate": 1.8296258859862392e-06, + "loss": 0.8872, + "step": 25336 + }, + { + "epoch": 0.81, + "learning_rate": 1.8290287012882601e-06, + "loss": 0.8706, + "step": 25337 + }, + { + "epoch": 0.81, + "learning_rate": 1.8284316042557614e-06, + "loss": 0.9058, + "step": 25338 + }, + { + "epoch": 0.81, + "learning_rate": 1.827834594895146e-06, + "loss": 1.019, + "step": 25339 + }, + { + "epoch": 0.81, + "learning_rate": 1.8272376732128262e-06, + "loss": 0.9409, + "step": 25340 + }, + { + "epoch": 0.81, + "learning_rate": 1.8266408392151957e-06, + "loss": 0.9482, + "step": 25341 + }, + { + "epoch": 0.81, + "learning_rate": 1.8260440929086654e-06, + "loss": 0.7588, + "step": 25342 + }, + { + "epoch": 0.81, + "learning_rate": 1.825447434299633e-06, + "loss": 0.8774, + "step": 25343 + }, + { + "epoch": 0.81, + "learning_rate": 1.824850863394505e-06, + "loss": 0.9644, + "step": 25344 + }, + { + "epoch": 0.81, + "learning_rate": 1.824254380199677e-06, + "loss": 0.8901, + "step": 25345 + }, + { + "epoch": 0.81, + "learning_rate": 1.823657984721553e-06, + "loss": 0.9238, + "step": 25346 + }, + { + "epoch": 0.81, + "learning_rate": 1.8230616769665278e-06, + "loss": 0.9221, + "step": 25347 + }, + { + "epoch": 0.81, + "learning_rate": 1.8224654569410061e-06, + "loss": 0.8953, + "step": 25348 + }, + { + "epoch": 0.81, + "learning_rate": 1.8218693246513753e-06, + "loss": 0.9502, + "step": 25349 + }, + { + "epoch": 0.81, + "learning_rate": 1.8212732801040367e-06, + "loss": 1.0166, + "step": 25350 + }, + { + "epoch": 0.81, + "learning_rate": 1.820677323305382e-06, + "loss": 0.8555, + "step": 25351 + }, + { + "epoch": 0.81, + "learning_rate": 1.8200814542618094e-06, + "loss": 0.9121, + "step": 25352 + }, + { + "epoch": 0.81, + "learning_rate": 1.8194856729797084e-06, + "loss": 0.9707, + "step": 25353 + }, + { + "epoch": 0.81, + "learning_rate": 1.8188899794654734e-06, + "loss": 0.937, + "step": 25354 + }, + { + "epoch": 0.81, + "learning_rate": 1.8182943737254954e-06, + "loss": 0.9399, + "step": 25355 + }, + { + "epoch": 0.81, + "learning_rate": 1.8176988557661635e-06, + "loss": 0.8799, + "step": 25356 + }, + { + "epoch": 0.81, + "learning_rate": 1.8171034255938647e-06, + "loss": 0.8516, + "step": 25357 + }, + { + "epoch": 0.81, + "learning_rate": 1.8165080832149917e-06, + "loss": 0.9126, + "step": 25358 + }, + { + "epoch": 0.81, + "learning_rate": 1.8159128286359284e-06, + "loss": 0.9746, + "step": 25359 + }, + { + "epoch": 0.81, + "learning_rate": 1.8153176618630641e-06, + "loss": 0.522, + "step": 25360 + }, + { + "epoch": 0.81, + "learning_rate": 1.8147225829027837e-06, + "loss": 0.8145, + "step": 25361 + }, + { + "epoch": 0.81, + "learning_rate": 1.8141275917614686e-06, + "loss": 0.8875, + "step": 25362 + }, + { + "epoch": 0.81, + "learning_rate": 1.8135326884455074e-06, + "loss": 0.9331, + "step": 25363 + }, + { + "epoch": 0.81, + "learning_rate": 1.812937872961279e-06, + "loss": 0.8594, + "step": 25364 + }, + { + "epoch": 0.81, + "learning_rate": 1.8123431453151652e-06, + "loss": 0.8967, + "step": 25365 + }, + { + "epoch": 0.81, + "learning_rate": 1.8117485055135498e-06, + "loss": 0.9165, + "step": 25366 + }, + { + "epoch": 0.81, + "learning_rate": 1.8111539535628098e-06, + "loss": 0.5056, + "step": 25367 + }, + { + "epoch": 0.81, + "learning_rate": 1.8105594894693236e-06, + "loss": 1.0708, + "step": 25368 + }, + { + "epoch": 0.81, + "learning_rate": 1.8099651132394725e-06, + "loss": 0.9263, + "step": 25369 + }, + { + "epoch": 0.81, + "learning_rate": 1.8093708248796304e-06, + "loss": 0.9883, + "step": 25370 + }, + { + "epoch": 0.81, + "learning_rate": 1.8087766243961758e-06, + "loss": 0.9702, + "step": 25371 + }, + { + "epoch": 0.81, + "learning_rate": 1.8081825117954799e-06, + "loss": 0.9243, + "step": 25372 + }, + { + "epoch": 0.81, + "learning_rate": 1.8075884870839212e-06, + "loss": 0.9941, + "step": 25373 + }, + { + "epoch": 0.81, + "learning_rate": 1.8069945502678688e-06, + "loss": 0.9097, + "step": 25374 + }, + { + "epoch": 0.81, + "learning_rate": 1.8064007013536999e-06, + "loss": 0.9126, + "step": 25375 + }, + { + "epoch": 0.81, + "learning_rate": 1.8058069403477806e-06, + "loss": 0.9058, + "step": 25376 + }, + { + "epoch": 0.81, + "learning_rate": 1.8052132672564859e-06, + "loss": 0.9238, + "step": 25377 + }, + { + "epoch": 0.81, + "learning_rate": 1.804619682086185e-06, + "loss": 0.8772, + "step": 25378 + }, + { + "epoch": 0.81, + "learning_rate": 1.8040261848432428e-06, + "loss": 1.0083, + "step": 25379 + }, + { + "epoch": 0.81, + "learning_rate": 1.8034327755340275e-06, + "loss": 0.8833, + "step": 25380 + }, + { + "epoch": 0.81, + "learning_rate": 1.8028394541649098e-06, + "loss": 1.0117, + "step": 25381 + }, + { + "epoch": 0.81, + "learning_rate": 1.8022462207422487e-06, + "loss": 0.938, + "step": 25382 + }, + { + "epoch": 0.81, + "learning_rate": 1.801653075272416e-06, + "loss": 0.9175, + "step": 25383 + }, + { + "epoch": 0.81, + "learning_rate": 1.8010600177617699e-06, + "loss": 0.8379, + "step": 25384 + }, + { + "epoch": 0.81, + "learning_rate": 1.8004670482166808e-06, + "loss": 1.1245, + "step": 25385 + }, + { + "epoch": 0.81, + "learning_rate": 1.7998741666435005e-06, + "loss": 0.9165, + "step": 25386 + }, + { + "epoch": 0.81, + "learning_rate": 1.799281373048597e-06, + "loss": 0.9219, + "step": 25387 + }, + { + "epoch": 0.81, + "learning_rate": 1.7986886674383263e-06, + "loss": 0.8411, + "step": 25388 + }, + { + "epoch": 0.81, + "learning_rate": 1.7980960498190514e-06, + "loss": 0.9639, + "step": 25389 + }, + { + "epoch": 0.81, + "learning_rate": 1.797503520197127e-06, + "loss": 0.8921, + "step": 25390 + }, + { + "epoch": 0.81, + "learning_rate": 1.7969110785789134e-06, + "loss": 0.9287, + "step": 25391 + }, + { + "epoch": 0.81, + "learning_rate": 1.7963187249707624e-06, + "loss": 0.9414, + "step": 25392 + }, + { + "epoch": 0.81, + "learning_rate": 1.795726459379038e-06, + "loss": 0.8418, + "step": 25393 + }, + { + "epoch": 0.81, + "learning_rate": 1.7951342818100825e-06, + "loss": 0.8735, + "step": 25394 + }, + { + "epoch": 0.81, + "learning_rate": 1.7945421922702589e-06, + "loss": 0.9155, + "step": 25395 + }, + { + "epoch": 0.81, + "learning_rate": 1.7939501907659141e-06, + "loss": 1.0239, + "step": 25396 + }, + { + "epoch": 0.81, + "learning_rate": 1.793358277303403e-06, + "loss": 0.8657, + "step": 25397 + }, + { + "epoch": 0.81, + "learning_rate": 1.792766451889073e-06, + "loss": 0.8853, + "step": 25398 + }, + { + "epoch": 0.81, + "learning_rate": 1.7921747145292788e-06, + "loss": 0.96, + "step": 25399 + }, + { + "epoch": 0.81, + "learning_rate": 1.7915830652303646e-06, + "loss": 0.8291, + "step": 25400 + }, + { + "epoch": 0.81, + "learning_rate": 1.7909915039986803e-06, + "loss": 0.9639, + "step": 25401 + }, + { + "epoch": 0.81, + "learning_rate": 1.79040003084057e-06, + "loss": 0.8208, + "step": 25402 + }, + { + "epoch": 0.81, + "learning_rate": 1.7898086457623831e-06, + "loss": 0.8555, + "step": 25403 + }, + { + "epoch": 0.81, + "learning_rate": 1.78921734877046e-06, + "loss": 0.8408, + "step": 25404 + }, + { + "epoch": 0.81, + "learning_rate": 1.7886261398711503e-06, + "loss": 0.8906, + "step": 25405 + }, + { + "epoch": 0.81, + "learning_rate": 1.7880350190707952e-06, + "loss": 1.0039, + "step": 25406 + }, + { + "epoch": 0.81, + "learning_rate": 1.787443986375732e-06, + "loss": 0.8931, + "step": 25407 + }, + { + "epoch": 0.81, + "learning_rate": 1.7868530417923091e-06, + "loss": 0.8403, + "step": 25408 + }, + { + "epoch": 0.81, + "learning_rate": 1.7862621853268625e-06, + "loss": 0.8979, + "step": 25409 + }, + { + "epoch": 0.81, + "learning_rate": 1.7856714169857293e-06, + "loss": 0.9766, + "step": 25410 + }, + { + "epoch": 0.81, + "learning_rate": 1.7850807367752544e-06, + "loss": 0.9434, + "step": 25411 + }, + { + "epoch": 0.81, + "learning_rate": 1.7844901447017705e-06, + "loss": 0.9277, + "step": 25412 + }, + { + "epoch": 0.81, + "learning_rate": 1.7838996407716124e-06, + "loss": 0.9053, + "step": 25413 + }, + { + "epoch": 0.81, + "learning_rate": 1.78330922499112e-06, + "loss": 1.0122, + "step": 25414 + }, + { + "epoch": 0.81, + "learning_rate": 1.7827188973666244e-06, + "loss": 0.9985, + "step": 25415 + }, + { + "epoch": 0.81, + "learning_rate": 1.7821286579044649e-06, + "loss": 0.9292, + "step": 25416 + }, + { + "epoch": 0.81, + "learning_rate": 1.7815385066109659e-06, + "loss": 0.9282, + "step": 25417 + }, + { + "epoch": 0.81, + "learning_rate": 1.780948443492464e-06, + "loss": 0.9722, + "step": 25418 + }, + { + "epoch": 0.81, + "learning_rate": 1.7803584685552877e-06, + "loss": 0.9414, + "step": 25419 + }, + { + "epoch": 0.81, + "learning_rate": 1.7797685818057697e-06, + "loss": 0.8906, + "step": 25420 + }, + { + "epoch": 0.81, + "learning_rate": 1.779178783250234e-06, + "loss": 0.9219, + "step": 25421 + }, + { + "epoch": 0.81, + "learning_rate": 1.778589072895015e-06, + "loss": 0.9375, + "step": 25422 + }, + { + "epoch": 0.81, + "learning_rate": 1.777999450746436e-06, + "loss": 0.8218, + "step": 25423 + }, + { + "epoch": 0.81, + "learning_rate": 1.7774099168108239e-06, + "loss": 0.8628, + "step": 25424 + }, + { + "epoch": 0.81, + "learning_rate": 1.7768204710945002e-06, + "loss": 0.8408, + "step": 25425 + }, + { + "epoch": 0.81, + "learning_rate": 1.7762311136037936e-06, + "loss": 0.7295, + "step": 25426 + }, + { + "epoch": 0.81, + "learning_rate": 1.775641844345024e-06, + "loss": 0.8867, + "step": 25427 + }, + { + "epoch": 0.81, + "learning_rate": 1.7750526633245168e-06, + "loss": 0.9131, + "step": 25428 + }, + { + "epoch": 0.81, + "learning_rate": 1.77446357054859e-06, + "loss": 0.8882, + "step": 25429 + }, + { + "epoch": 0.81, + "learning_rate": 1.773874566023568e-06, + "loss": 0.9048, + "step": 25430 + }, + { + "epoch": 0.81, + "learning_rate": 1.773285649755767e-06, + "loss": 0.8379, + "step": 25431 + }, + { + "epoch": 0.81, + "learning_rate": 1.772696821751506e-06, + "loss": 0.9834, + "step": 25432 + }, + { + "epoch": 0.81, + "learning_rate": 1.772108082017101e-06, + "loss": 0.9478, + "step": 25433 + }, + { + "epoch": 0.81, + "learning_rate": 1.7715194305588711e-06, + "loss": 1.0264, + "step": 25434 + }, + { + "epoch": 0.81, + "learning_rate": 1.7709308673831305e-06, + "loss": 0.9692, + "step": 25435 + }, + { + "epoch": 0.81, + "learning_rate": 1.770342392496195e-06, + "loss": 0.8823, + "step": 25436 + }, + { + "epoch": 0.81, + "learning_rate": 1.769754005904375e-06, + "loss": 0.8892, + "step": 25437 + }, + { + "epoch": 0.81, + "learning_rate": 1.7691657076139923e-06, + "loss": 0.9438, + "step": 25438 + }, + { + "epoch": 0.81, + "learning_rate": 1.7685774976313452e-06, + "loss": 0.9131, + "step": 25439 + }, + { + "epoch": 0.81, + "learning_rate": 1.7679893759627554e-06, + "loss": 0.8108, + "step": 25440 + }, + { + "epoch": 0.81, + "learning_rate": 1.7674013426145254e-06, + "loss": 0.6809, + "step": 25441 + }, + { + "epoch": 0.81, + "learning_rate": 1.7668133975929713e-06, + "loss": 0.4609, + "step": 25442 + }, + { + "epoch": 0.81, + "learning_rate": 1.7662255409043938e-06, + "loss": 0.9077, + "step": 25443 + }, + { + "epoch": 0.81, + "learning_rate": 1.7656377725551065e-06, + "loss": 0.9517, + "step": 25444 + }, + { + "epoch": 0.81, + "learning_rate": 1.7650500925514124e-06, + "loss": 0.9365, + "step": 25445 + }, + { + "epoch": 0.81, + "learning_rate": 1.7644625008996163e-06, + "loss": 0.9165, + "step": 25446 + }, + { + "epoch": 0.81, + "learning_rate": 1.763874997606021e-06, + "loss": 0.7808, + "step": 25447 + }, + { + "epoch": 0.81, + "learning_rate": 1.7632875826769347e-06, + "loss": 0.9966, + "step": 25448 + }, + { + "epoch": 0.81, + "learning_rate": 1.7627002561186534e-06, + "loss": 0.9404, + "step": 25449 + }, + { + "epoch": 0.81, + "learning_rate": 1.7621130179374846e-06, + "loss": 0.9395, + "step": 25450 + }, + { + "epoch": 0.81, + "learning_rate": 1.761525868139725e-06, + "loss": 0.9473, + "step": 25451 + }, + { + "epoch": 0.81, + "learning_rate": 1.7609388067316735e-06, + "loss": 0.9194, + "step": 25452 + }, + { + "epoch": 0.81, + "learning_rate": 1.7603518337196324e-06, + "loss": 0.9136, + "step": 25453 + }, + { + "epoch": 0.81, + "learning_rate": 1.7597649491098967e-06, + "loss": 0.9155, + "step": 25454 + }, + { + "epoch": 0.81, + "learning_rate": 1.7591781529087636e-06, + "loss": 0.7998, + "step": 25455 + }, + { + "epoch": 0.81, + "learning_rate": 1.758591445122526e-06, + "loss": 0.9136, + "step": 25456 + }, + { + "epoch": 0.81, + "learning_rate": 1.7580048257574833e-06, + "loss": 0.9873, + "step": 25457 + }, + { + "epoch": 0.81, + "learning_rate": 1.7574182948199247e-06, + "loss": 0.9204, + "step": 25458 + }, + { + "epoch": 0.81, + "learning_rate": 1.7568318523161476e-06, + "loss": 0.9062, + "step": 25459 + }, + { + "epoch": 0.81, + "learning_rate": 1.756245498252439e-06, + "loss": 1.0562, + "step": 25460 + }, + { + "epoch": 0.81, + "learning_rate": 1.7556592326350974e-06, + "loss": 0.873, + "step": 25461 + }, + { + "epoch": 0.81, + "learning_rate": 1.7550730554704033e-06, + "loss": 0.9302, + "step": 25462 + }, + { + "epoch": 0.81, + "learning_rate": 1.7544869667646525e-06, + "loss": 0.876, + "step": 25463 + }, + { + "epoch": 0.81, + "learning_rate": 1.7539009665241291e-06, + "loss": 0.4482, + "step": 25464 + }, + { + "epoch": 0.81, + "learning_rate": 1.7533150547551237e-06, + "loss": 0.9658, + "step": 25465 + }, + { + "epoch": 0.81, + "learning_rate": 1.7527292314639176e-06, + "loss": 0.9854, + "step": 25466 + }, + { + "epoch": 0.81, + "learning_rate": 1.7521434966568028e-06, + "loss": 0.8306, + "step": 25467 + }, + { + "epoch": 0.81, + "learning_rate": 1.7515578503400599e-06, + "loss": 0.8623, + "step": 25468 + }, + { + "epoch": 0.81, + "learning_rate": 1.7509722925199723e-06, + "loss": 0.9038, + "step": 25469 + }, + { + "epoch": 0.81, + "learning_rate": 1.7503868232028199e-06, + "loss": 0.9775, + "step": 25470 + }, + { + "epoch": 0.81, + "learning_rate": 1.7498014423948894e-06, + "loss": 0.9609, + "step": 25471 + }, + { + "epoch": 0.81, + "learning_rate": 1.7492161501024553e-06, + "loss": 0.9302, + "step": 25472 + }, + { + "epoch": 0.81, + "learning_rate": 1.748630946331804e-06, + "loss": 0.8853, + "step": 25473 + }, + { + "epoch": 0.81, + "learning_rate": 1.7480458310892067e-06, + "loss": 0.9692, + "step": 25474 + }, + { + "epoch": 0.81, + "learning_rate": 1.7474608043809483e-06, + "loss": 0.8901, + "step": 25475 + }, + { + "epoch": 0.81, + "learning_rate": 1.7468758662133012e-06, + "loss": 0.8125, + "step": 25476 + }, + { + "epoch": 0.81, + "learning_rate": 1.7462910165925429e-06, + "loss": 0.9302, + "step": 25477 + }, + { + "epoch": 0.81, + "learning_rate": 1.7457062555249438e-06, + "loss": 0.801, + "step": 25478 + }, + { + "epoch": 0.81, + "learning_rate": 1.7451215830167845e-06, + "loss": 0.8352, + "step": 25479 + }, + { + "epoch": 0.81, + "learning_rate": 1.744536999074332e-06, + "loss": 0.8931, + "step": 25480 + }, + { + "epoch": 0.81, + "learning_rate": 1.743952503703863e-06, + "loss": 0.9087, + "step": 25481 + }, + { + "epoch": 0.81, + "learning_rate": 1.743368096911644e-06, + "loss": 0.9409, + "step": 25482 + }, + { + "epoch": 0.81, + "learning_rate": 1.7427837787039493e-06, + "loss": 0.9712, + "step": 25483 + }, + { + "epoch": 0.81, + "learning_rate": 1.742199549087047e-06, + "loss": 0.9165, + "step": 25484 + }, + { + "epoch": 0.82, + "learning_rate": 1.7416154080672031e-06, + "loss": 0.9512, + "step": 25485 + }, + { + "epoch": 0.82, + "learning_rate": 1.7410313556506842e-06, + "loss": 0.8721, + "step": 25486 + }, + { + "epoch": 0.82, + "learning_rate": 1.7404473918437604e-06, + "loss": 0.9404, + "step": 25487 + }, + { + "epoch": 0.82, + "learning_rate": 1.7398635166526923e-06, + "loss": 0.8604, + "step": 25488 + }, + { + "epoch": 0.82, + "learning_rate": 1.7392797300837493e-06, + "loss": 0.9438, + "step": 25489 + }, + { + "epoch": 0.82, + "learning_rate": 1.7386960321431922e-06, + "loss": 0.835, + "step": 25490 + }, + { + "epoch": 0.82, + "learning_rate": 1.7381124228372848e-06, + "loss": 0.9609, + "step": 25491 + }, + { + "epoch": 0.82, + "learning_rate": 1.737528902172283e-06, + "loss": 0.9775, + "step": 25492 + }, + { + "epoch": 0.82, + "learning_rate": 1.7369454701544552e-06, + "loss": 0.9639, + "step": 25493 + }, + { + "epoch": 0.82, + "learning_rate": 1.7363621267900578e-06, + "loss": 0.9844, + "step": 25494 + }, + { + "epoch": 0.82, + "learning_rate": 1.7357788720853453e-06, + "loss": 1.0117, + "step": 25495 + }, + { + "epoch": 0.82, + "learning_rate": 1.7351957060465817e-06, + "loss": 0.7935, + "step": 25496 + }, + { + "epoch": 0.82, + "learning_rate": 1.7346126286800202e-06, + "loss": 0.981, + "step": 25497 + }, + { + "epoch": 0.82, + "learning_rate": 1.7340296399919187e-06, + "loss": 0.7471, + "step": 25498 + }, + { + "epoch": 0.82, + "learning_rate": 1.7334467399885313e-06, + "loss": 0.9004, + "step": 25499 + }, + { + "epoch": 0.82, + "learning_rate": 1.7328639286761106e-06, + "loss": 1.0166, + "step": 25500 + }, + { + "epoch": 0.82, + "learning_rate": 1.7322812060609084e-06, + "loss": 0.8369, + "step": 25501 + }, + { + "epoch": 0.82, + "learning_rate": 1.731698572149182e-06, + "loss": 1.0161, + "step": 25502 + }, + { + "epoch": 0.82, + "learning_rate": 1.731116026947175e-06, + "loss": 0.7827, + "step": 25503 + }, + { + "epoch": 0.82, + "learning_rate": 1.7305335704611449e-06, + "loss": 0.9551, + "step": 25504 + }, + { + "epoch": 0.82, + "learning_rate": 1.7299512026973342e-06, + "loss": 0.9131, + "step": 25505 + }, + { + "epoch": 0.82, + "learning_rate": 1.7293689236620004e-06, + "loss": 0.9189, + "step": 25506 + }, + { + "epoch": 0.82, + "learning_rate": 1.7287867333613783e-06, + "loss": 0.8848, + "step": 25507 + }, + { + "epoch": 0.82, + "learning_rate": 1.7282046318017232e-06, + "loss": 0.9028, + "step": 25508 + }, + { + "epoch": 0.82, + "learning_rate": 1.7276226189892763e-06, + "loss": 0.9136, + "step": 25509 + }, + { + "epoch": 0.82, + "learning_rate": 1.7270406949302842e-06, + "loss": 0.9434, + "step": 25510 + }, + { + "epoch": 0.82, + "learning_rate": 1.7264588596309884e-06, + "loss": 1.0522, + "step": 25511 + }, + { + "epoch": 0.82, + "learning_rate": 1.7258771130976336e-06, + "loss": 0.4656, + "step": 25512 + }, + { + "epoch": 0.82, + "learning_rate": 1.7252954553364575e-06, + "loss": 0.937, + "step": 25513 + }, + { + "epoch": 0.82, + "learning_rate": 1.7247138863537095e-06, + "loss": 0.9102, + "step": 25514 + }, + { + "epoch": 0.82, + "learning_rate": 1.7241324061556164e-06, + "loss": 0.9395, + "step": 25515 + }, + { + "epoch": 0.82, + "learning_rate": 1.723551014748427e-06, + "loss": 0.8828, + "step": 25516 + }, + { + "epoch": 0.82, + "learning_rate": 1.7229697121383714e-06, + "loss": 0.9395, + "step": 25517 + }, + { + "epoch": 0.82, + "learning_rate": 1.722388498331694e-06, + "loss": 0.8638, + "step": 25518 + }, + { + "epoch": 0.82, + "learning_rate": 1.7218073733346241e-06, + "loss": 0.9546, + "step": 25519 + }, + { + "epoch": 0.82, + "learning_rate": 1.7212263371534011e-06, + "loss": 0.9199, + "step": 25520 + }, + { + "epoch": 0.82, + "learning_rate": 1.7206453897942576e-06, + "loss": 0.957, + "step": 25521 + }, + { + "epoch": 0.82, + "learning_rate": 1.7200645312634256e-06, + "loss": 0.9595, + "step": 25522 + }, + { + "epoch": 0.82, + "learning_rate": 1.7194837615671344e-06, + "loss": 0.9768, + "step": 25523 + }, + { + "epoch": 0.82, + "learning_rate": 1.7189030807116213e-06, + "loss": 0.8525, + "step": 25524 + }, + { + "epoch": 0.82, + "learning_rate": 1.7183224887031114e-06, + "loss": 0.9536, + "step": 25525 + }, + { + "epoch": 0.82, + "learning_rate": 1.7177419855478362e-06, + "loss": 0.9131, + "step": 25526 + }, + { + "epoch": 0.82, + "learning_rate": 1.7171615712520218e-06, + "loss": 0.9917, + "step": 25527 + }, + { + "epoch": 0.82, + "learning_rate": 1.716581245821899e-06, + "loss": 0.894, + "step": 25528 + }, + { + "epoch": 0.82, + "learning_rate": 1.7160010092636914e-06, + "loss": 0.9048, + "step": 25529 + }, + { + "epoch": 0.82, + "learning_rate": 1.7154208615836255e-06, + "loss": 0.9614, + "step": 25530 + }, + { + "epoch": 0.82, + "learning_rate": 1.7148408027879216e-06, + "loss": 0.9551, + "step": 25531 + }, + { + "epoch": 0.82, + "learning_rate": 1.7142608328828091e-06, + "loss": 0.9258, + "step": 25532 + }, + { + "epoch": 0.82, + "learning_rate": 1.713680951874508e-06, + "loss": 0.9336, + "step": 25533 + }, + { + "epoch": 0.82, + "learning_rate": 1.713101159769236e-06, + "loss": 0.8784, + "step": 25534 + }, + { + "epoch": 0.82, + "learning_rate": 1.7125214565732196e-06, + "loss": 0.814, + "step": 25535 + }, + { + "epoch": 0.82, + "learning_rate": 1.711941842292677e-06, + "loss": 0.8313, + "step": 25536 + }, + { + "epoch": 0.82, + "learning_rate": 1.7113623169338222e-06, + "loss": 0.5007, + "step": 25537 + }, + { + "epoch": 0.82, + "learning_rate": 1.7107828805028792e-06, + "loss": 0.8984, + "step": 25538 + }, + { + "epoch": 0.82, + "learning_rate": 1.7102035330060617e-06, + "loss": 1.0156, + "step": 25539 + }, + { + "epoch": 0.82, + "learning_rate": 1.709624274449584e-06, + "loss": 1.0439, + "step": 25540 + }, + { + "epoch": 0.82, + "learning_rate": 1.709045104839664e-06, + "loss": 0.8862, + "step": 25541 + }, + { + "epoch": 0.82, + "learning_rate": 1.7084660241825113e-06, + "loss": 0.8706, + "step": 25542 + }, + { + "epoch": 0.82, + "learning_rate": 1.7078870324843444e-06, + "loss": 0.8857, + "step": 25543 + }, + { + "epoch": 0.82, + "learning_rate": 1.7073081297513727e-06, + "loss": 0.8572, + "step": 25544 + }, + { + "epoch": 0.82, + "learning_rate": 1.7067293159898067e-06, + "loss": 0.8643, + "step": 25545 + }, + { + "epoch": 0.82, + "learning_rate": 1.706150591205854e-06, + "loss": 0.9995, + "step": 25546 + }, + { + "epoch": 0.82, + "learning_rate": 1.7055719554057292e-06, + "loss": 0.9453, + "step": 25547 + }, + { + "epoch": 0.82, + "learning_rate": 1.7049934085956344e-06, + "loss": 0.8774, + "step": 25548 + }, + { + "epoch": 0.82, + "learning_rate": 1.704414950781783e-06, + "loss": 0.9155, + "step": 25549 + }, + { + "epoch": 0.82, + "learning_rate": 1.703836581970375e-06, + "loss": 0.8394, + "step": 25550 + }, + { + "epoch": 0.82, + "learning_rate": 1.703258302167624e-06, + "loss": 1.041, + "step": 25551 + }, + { + "epoch": 0.82, + "learning_rate": 1.7026801113797232e-06, + "loss": 0.8989, + "step": 25552 + }, + { + "epoch": 0.82, + "learning_rate": 1.7021020096128859e-06, + "loss": 0.876, + "step": 25553 + }, + { + "epoch": 0.82, + "learning_rate": 1.7015239968733066e-06, + "loss": 0.9233, + "step": 25554 + }, + { + "epoch": 0.82, + "learning_rate": 1.7009460731671933e-06, + "loss": 0.9839, + "step": 25555 + }, + { + "epoch": 0.82, + "learning_rate": 1.7003682385007414e-06, + "loss": 0.916, + "step": 25556 + }, + { + "epoch": 0.82, + "learning_rate": 1.6997904928801556e-06, + "loss": 0.79, + "step": 25557 + }, + { + "epoch": 0.82, + "learning_rate": 1.6992128363116289e-06, + "loss": 0.9282, + "step": 25558 + }, + { + "epoch": 0.82, + "learning_rate": 1.6986352688013664e-06, + "loss": 0.9404, + "step": 25559 + }, + { + "epoch": 0.82, + "learning_rate": 1.698057790355554e-06, + "loss": 0.9194, + "step": 25560 + }, + { + "epoch": 0.82, + "learning_rate": 1.697480400980397e-06, + "loss": 0.9854, + "step": 25561 + }, + { + "epoch": 0.82, + "learning_rate": 1.6969031006820835e-06, + "loss": 0.896, + "step": 25562 + }, + { + "epoch": 0.82, + "learning_rate": 1.6963258894668134e-06, + "loss": 0.8857, + "step": 25563 + }, + { + "epoch": 0.82, + "learning_rate": 1.6957487673407735e-06, + "loss": 0.9526, + "step": 25564 + }, + { + "epoch": 0.82, + "learning_rate": 1.6951717343101615e-06, + "loss": 0.9937, + "step": 25565 + }, + { + "epoch": 0.82, + "learning_rate": 1.6945947903811633e-06, + "loss": 0.9448, + "step": 25566 + }, + { + "epoch": 0.82, + "learning_rate": 1.6940179355599751e-06, + "loss": 0.4343, + "step": 25567 + }, + { + "epoch": 0.82, + "learning_rate": 1.6934411698527774e-06, + "loss": 0.8384, + "step": 25568 + }, + { + "epoch": 0.82, + "learning_rate": 1.6928644932657657e-06, + "loss": 0.937, + "step": 25569 + }, + { + "epoch": 0.82, + "learning_rate": 1.69228790580512e-06, + "loss": 0.9253, + "step": 25570 + }, + { + "epoch": 0.82, + "learning_rate": 1.6917114074770346e-06, + "loss": 0.8369, + "step": 25571 + }, + { + "epoch": 0.82, + "learning_rate": 1.6911349982876902e-06, + "loss": 1.0059, + "step": 25572 + }, + { + "epoch": 0.82, + "learning_rate": 1.6905586782432693e-06, + "loss": 0.9585, + "step": 25573 + }, + { + "epoch": 0.82, + "learning_rate": 1.6899824473499594e-06, + "loss": 1.0518, + "step": 25574 + }, + { + "epoch": 0.82, + "learning_rate": 1.6894063056139398e-06, + "loss": 0.9761, + "step": 25575 + }, + { + "epoch": 0.82, + "learning_rate": 1.6888302530413913e-06, + "loss": 0.918, + "step": 25576 + }, + { + "epoch": 0.82, + "learning_rate": 1.688254289638499e-06, + "loss": 0.9399, + "step": 25577 + }, + { + "epoch": 0.82, + "learning_rate": 1.6876784154114379e-06, + "loss": 0.8682, + "step": 25578 + }, + { + "epoch": 0.82, + "learning_rate": 1.6871026303663861e-06, + "loss": 0.5015, + "step": 25579 + }, + { + "epoch": 0.82, + "learning_rate": 1.6865269345095258e-06, + "loss": 0.8403, + "step": 25580 + }, + { + "epoch": 0.82, + "learning_rate": 1.6859513278470274e-06, + "loss": 0.9219, + "step": 25581 + }, + { + "epoch": 0.82, + "learning_rate": 1.6853758103850715e-06, + "loss": 0.9233, + "step": 25582 + }, + { + "epoch": 0.82, + "learning_rate": 1.684800382129832e-06, + "loss": 0.9746, + "step": 25583 + }, + { + "epoch": 0.82, + "learning_rate": 1.6842250430874818e-06, + "loss": 0.9102, + "step": 25584 + }, + { + "epoch": 0.82, + "learning_rate": 1.6836497932641904e-06, + "loss": 0.8857, + "step": 25585 + }, + { + "epoch": 0.82, + "learning_rate": 1.6830746326661352e-06, + "loss": 0.9307, + "step": 25586 + }, + { + "epoch": 0.82, + "learning_rate": 1.682499561299482e-06, + "loss": 1.0244, + "step": 25587 + }, + { + "epoch": 0.82, + "learning_rate": 1.6819245791704053e-06, + "loss": 0.9673, + "step": 25588 + }, + { + "epoch": 0.82, + "learning_rate": 1.681349686285072e-06, + "loss": 0.8447, + "step": 25589 + }, + { + "epoch": 0.82, + "learning_rate": 1.6807748826496506e-06, + "loss": 0.8474, + "step": 25590 + }, + { + "epoch": 0.82, + "learning_rate": 1.6802001682703039e-06, + "loss": 0.9414, + "step": 25591 + }, + { + "epoch": 0.82, + "learning_rate": 1.679625543153205e-06, + "loss": 0.9609, + "step": 25592 + }, + { + "epoch": 0.82, + "learning_rate": 1.679051007304513e-06, + "loss": 0.9038, + "step": 25593 + }, + { + "epoch": 0.82, + "learning_rate": 1.6784765607303965e-06, + "loss": 0.9248, + "step": 25594 + }, + { + "epoch": 0.82, + "learning_rate": 1.677902203437014e-06, + "loss": 0.9561, + "step": 25595 + }, + { + "epoch": 0.82, + "learning_rate": 1.6773279354305327e-06, + "loss": 1.0093, + "step": 25596 + }, + { + "epoch": 0.82, + "learning_rate": 1.6767537567171122e-06, + "loss": 0.7993, + "step": 25597 + }, + { + "epoch": 0.82, + "learning_rate": 1.6761796673029119e-06, + "loss": 0.8506, + "step": 25598 + }, + { + "epoch": 0.82, + "learning_rate": 1.6756056671940902e-06, + "loss": 0.834, + "step": 25599 + }, + { + "epoch": 0.82, + "learning_rate": 1.6750317563968077e-06, + "loss": 0.9453, + "step": 25600 + }, + { + "epoch": 0.82, + "learning_rate": 1.6744579349172186e-06, + "loss": 0.8794, + "step": 25601 + }, + { + "epoch": 0.82, + "learning_rate": 1.6738842027614844e-06, + "loss": 0.9331, + "step": 25602 + }, + { + "epoch": 0.82, + "learning_rate": 1.6733105599357546e-06, + "loss": 1.0068, + "step": 25603 + }, + { + "epoch": 0.82, + "learning_rate": 1.6727370064461923e-06, + "loss": 0.8823, + "step": 25604 + }, + { + "epoch": 0.82, + "learning_rate": 1.6721635422989414e-06, + "loss": 0.9658, + "step": 25605 + }, + { + "epoch": 0.82, + "learning_rate": 1.6715901675001611e-06, + "loss": 0.9785, + "step": 25606 + }, + { + "epoch": 0.82, + "learning_rate": 1.671016882055998e-06, + "loss": 0.8804, + "step": 25607 + }, + { + "epoch": 0.82, + "learning_rate": 1.6704436859726093e-06, + "loss": 0.9385, + "step": 25608 + }, + { + "epoch": 0.82, + "learning_rate": 1.6698705792561377e-06, + "loss": 0.918, + "step": 25609 + }, + { + "epoch": 0.82, + "learning_rate": 1.669297561912737e-06, + "loss": 0.9614, + "step": 25610 + }, + { + "epoch": 0.82, + "learning_rate": 1.668724633948553e-06, + "loss": 0.9561, + "step": 25611 + }, + { + "epoch": 0.82, + "learning_rate": 1.6681517953697368e-06, + "loss": 0.8716, + "step": 25612 + }, + { + "epoch": 0.82, + "learning_rate": 1.6675790461824258e-06, + "loss": 0.9155, + "step": 25613 + }, + { + "epoch": 0.82, + "learning_rate": 1.667006386392772e-06, + "loss": 0.8818, + "step": 25614 + }, + { + "epoch": 0.82, + "learning_rate": 1.666433816006916e-06, + "loss": 0.9141, + "step": 25615 + }, + { + "epoch": 0.82, + "learning_rate": 1.6658613350310027e-06, + "loss": 1.0869, + "step": 25616 + }, + { + "epoch": 0.82, + "learning_rate": 1.665288943471175e-06, + "loss": 0.9565, + "step": 25617 + }, + { + "epoch": 0.82, + "learning_rate": 1.6647166413335692e-06, + "loss": 0.9014, + "step": 25618 + }, + { + "epoch": 0.82, + "learning_rate": 1.6641444286243314e-06, + "loss": 0.9619, + "step": 25619 + }, + { + "epoch": 0.82, + "learning_rate": 1.663572305349599e-06, + "loss": 0.9624, + "step": 25620 + }, + { + "epoch": 0.82, + "learning_rate": 1.6630002715155058e-06, + "loss": 0.9517, + "step": 25621 + }, + { + "epoch": 0.82, + "learning_rate": 1.662428327128196e-06, + "loss": 0.9033, + "step": 25622 + }, + { + "epoch": 0.82, + "learning_rate": 1.6618564721938024e-06, + "loss": 0.957, + "step": 25623 + }, + { + "epoch": 0.82, + "learning_rate": 1.6612847067184579e-06, + "loss": 0.9033, + "step": 25624 + }, + { + "epoch": 0.82, + "learning_rate": 1.660713030708303e-06, + "loss": 0.8735, + "step": 25625 + }, + { + "epoch": 0.82, + "learning_rate": 1.6601414441694642e-06, + "loss": 0.8472, + "step": 25626 + }, + { + "epoch": 0.82, + "learning_rate": 1.6595699471080807e-06, + "loss": 1.0483, + "step": 25627 + }, + { + "epoch": 0.82, + "learning_rate": 1.658998539530281e-06, + "loss": 0.8472, + "step": 25628 + }, + { + "epoch": 0.82, + "learning_rate": 1.658427221442196e-06, + "loss": 0.9438, + "step": 25629 + }, + { + "epoch": 0.82, + "learning_rate": 1.6578559928499517e-06, + "loss": 0.9287, + "step": 25630 + }, + { + "epoch": 0.82, + "learning_rate": 1.657284853759683e-06, + "loss": 0.8809, + "step": 25631 + }, + { + "epoch": 0.82, + "learning_rate": 1.656713804177511e-06, + "loss": 0.9399, + "step": 25632 + }, + { + "epoch": 0.82, + "learning_rate": 1.6561428441095695e-06, + "loss": 0.9302, + "step": 25633 + }, + { + "epoch": 0.82, + "learning_rate": 1.655571973561978e-06, + "loss": 1.0039, + "step": 25634 + }, + { + "epoch": 0.82, + "learning_rate": 1.6550011925408694e-06, + "loss": 0.9336, + "step": 25635 + }, + { + "epoch": 0.82, + "learning_rate": 1.6544305010523566e-06, + "loss": 1.0591, + "step": 25636 + }, + { + "epoch": 0.82, + "learning_rate": 1.653859899102571e-06, + "loss": 1.061, + "step": 25637 + }, + { + "epoch": 0.82, + "learning_rate": 1.6532893866976296e-06, + "loss": 0.9248, + "step": 25638 + }, + { + "epoch": 0.82, + "learning_rate": 1.652718963843657e-06, + "loss": 1.0083, + "step": 25639 + }, + { + "epoch": 0.82, + "learning_rate": 1.6521486305467693e-06, + "loss": 0.8691, + "step": 25640 + }, + { + "epoch": 0.82, + "learning_rate": 1.6515783868130908e-06, + "loss": 0.8984, + "step": 25641 + }, + { + "epoch": 0.82, + "learning_rate": 1.6510082326487363e-06, + "loss": 0.8911, + "step": 25642 + }, + { + "epoch": 0.82, + "learning_rate": 1.6504381680598237e-06, + "loss": 0.9502, + "step": 25643 + }, + { + "epoch": 0.82, + "learning_rate": 1.6498681930524652e-06, + "loss": 0.9658, + "step": 25644 + }, + { + "epoch": 0.82, + "learning_rate": 1.6492983076327829e-06, + "loss": 0.9966, + "step": 25645 + }, + { + "epoch": 0.82, + "learning_rate": 1.6487285118068852e-06, + "loss": 0.9756, + "step": 25646 + }, + { + "epoch": 0.82, + "learning_rate": 1.6481588055808894e-06, + "loss": 0.9766, + "step": 25647 + }, + { + "epoch": 0.82, + "learning_rate": 1.6475891889609052e-06, + "loss": 0.9951, + "step": 25648 + }, + { + "epoch": 0.82, + "learning_rate": 1.6470196619530465e-06, + "loss": 1.0215, + "step": 25649 + }, + { + "epoch": 0.82, + "learning_rate": 1.6464502245634216e-06, + "loss": 0.9946, + "step": 25650 + }, + { + "epoch": 0.82, + "learning_rate": 1.6458808767981416e-06, + "loss": 0.9355, + "step": 25651 + }, + { + "epoch": 0.82, + "learning_rate": 1.645311618663311e-06, + "loss": 0.8613, + "step": 25652 + }, + { + "epoch": 0.82, + "learning_rate": 1.6447424501650432e-06, + "loss": 0.9397, + "step": 25653 + }, + { + "epoch": 0.82, + "learning_rate": 1.6441733713094389e-06, + "loss": 0.9541, + "step": 25654 + }, + { + "epoch": 0.82, + "learning_rate": 1.6436043821026094e-06, + "loss": 0.9399, + "step": 25655 + }, + { + "epoch": 0.82, + "learning_rate": 1.6430354825506556e-06, + "loss": 0.9189, + "step": 25656 + }, + { + "epoch": 0.82, + "learning_rate": 1.6424666726596828e-06, + "loss": 0.8105, + "step": 25657 + }, + { + "epoch": 0.82, + "learning_rate": 1.6418979524357903e-06, + "loss": 0.8818, + "step": 25658 + }, + { + "epoch": 0.82, + "learning_rate": 1.6413293218850857e-06, + "loss": 0.8643, + "step": 25659 + }, + { + "epoch": 0.82, + "learning_rate": 1.6407607810136628e-06, + "loss": 0.8926, + "step": 25660 + }, + { + "epoch": 0.82, + "learning_rate": 1.6401923298276278e-06, + "loss": 0.9141, + "step": 25661 + }, + { + "epoch": 0.82, + "learning_rate": 1.6396239683330773e-06, + "loss": 0.9531, + "step": 25662 + }, + { + "epoch": 0.82, + "learning_rate": 1.6390556965361072e-06, + "loss": 0.897, + "step": 25663 + }, + { + "epoch": 0.82, + "learning_rate": 1.6384875144428168e-06, + "loss": 1.0166, + "step": 25664 + }, + { + "epoch": 0.82, + "learning_rate": 1.637919422059303e-06, + "loss": 0.8428, + "step": 25665 + }, + { + "epoch": 0.82, + "learning_rate": 1.6373514193916552e-06, + "loss": 0.9668, + "step": 25666 + }, + { + "epoch": 0.82, + "learning_rate": 1.6367835064459747e-06, + "loss": 0.9287, + "step": 25667 + }, + { + "epoch": 0.82, + "learning_rate": 1.6362156832283505e-06, + "loss": 0.5168, + "step": 25668 + }, + { + "epoch": 0.82, + "learning_rate": 1.635647949744874e-06, + "loss": 0.853, + "step": 25669 + }, + { + "epoch": 0.82, + "learning_rate": 1.6350803060016396e-06, + "loss": 0.8779, + "step": 25670 + }, + { + "epoch": 0.82, + "learning_rate": 1.6345127520047333e-06, + "loss": 0.9736, + "step": 25671 + }, + { + "epoch": 0.82, + "learning_rate": 1.6339452877602512e-06, + "loss": 0.9448, + "step": 25672 + }, + { + "epoch": 0.82, + "learning_rate": 1.6333779132742732e-06, + "loss": 0.9556, + "step": 25673 + }, + { + "epoch": 0.82, + "learning_rate": 1.6328106285528921e-06, + "loss": 0.9604, + "step": 25674 + }, + { + "epoch": 0.82, + "learning_rate": 1.6322434336021908e-06, + "loss": 0.9214, + "step": 25675 + }, + { + "epoch": 0.82, + "learning_rate": 1.6316763284282578e-06, + "loss": 0.916, + "step": 25676 + }, + { + "epoch": 0.82, + "learning_rate": 1.6311093130371747e-06, + "loss": 0.7793, + "step": 25677 + }, + { + "epoch": 0.82, + "learning_rate": 1.630542387435029e-06, + "loss": 0.8232, + "step": 25678 + }, + { + "epoch": 0.82, + "learning_rate": 1.629975551627897e-06, + "loss": 0.9517, + "step": 25679 + }, + { + "epoch": 0.82, + "learning_rate": 1.6294088056218705e-06, + "loss": 0.9536, + "step": 25680 + }, + { + "epoch": 0.82, + "learning_rate": 1.6288421494230166e-06, + "loss": 0.9189, + "step": 25681 + }, + { + "epoch": 0.82, + "learning_rate": 1.6282755830374253e-06, + "loss": 0.8931, + "step": 25682 + }, + { + "epoch": 0.82, + "learning_rate": 1.6277091064711692e-06, + "loss": 0.9004, + "step": 25683 + }, + { + "epoch": 0.82, + "learning_rate": 1.6271427197303301e-06, + "loss": 0.8574, + "step": 25684 + }, + { + "epoch": 0.82, + "learning_rate": 1.6265764228209812e-06, + "loss": 1.0254, + "step": 25685 + }, + { + "epoch": 0.82, + "learning_rate": 1.626010215749203e-06, + "loss": 0.9897, + "step": 25686 + }, + { + "epoch": 0.82, + "learning_rate": 1.6254440985210663e-06, + "loss": 0.9312, + "step": 25687 + }, + { + "epoch": 0.82, + "learning_rate": 1.624878071142646e-06, + "loss": 1.0112, + "step": 25688 + }, + { + "epoch": 0.82, + "learning_rate": 1.624312133620013e-06, + "loss": 0.4712, + "step": 25689 + }, + { + "epoch": 0.82, + "learning_rate": 1.6237462859592435e-06, + "loss": 0.9995, + "step": 25690 + }, + { + "epoch": 0.82, + "learning_rate": 1.6231805281664037e-06, + "loss": 0.7749, + "step": 25691 + }, + { + "epoch": 0.82, + "learning_rate": 1.6226148602475677e-06, + "loss": 0.8379, + "step": 25692 + }, + { + "epoch": 0.82, + "learning_rate": 1.6220492822087997e-06, + "loss": 0.9326, + "step": 25693 + }, + { + "epoch": 0.82, + "learning_rate": 1.6214837940561735e-06, + "loss": 0.981, + "step": 25694 + }, + { + "epoch": 0.82, + "learning_rate": 1.6209183957957541e-06, + "loss": 1.0132, + "step": 25695 + }, + { + "epoch": 0.82, + "learning_rate": 1.620353087433606e-06, + "loss": 0.8652, + "step": 25696 + }, + { + "epoch": 0.82, + "learning_rate": 1.619787868975793e-06, + "loss": 0.8931, + "step": 25697 + }, + { + "epoch": 0.82, + "learning_rate": 1.6192227404283833e-06, + "loss": 0.9214, + "step": 25698 + }, + { + "epoch": 0.82, + "learning_rate": 1.6186577017974369e-06, + "loss": 0.8696, + "step": 25699 + }, + { + "epoch": 0.82, + "learning_rate": 1.618092753089019e-06, + "loss": 0.9082, + "step": 25700 + }, + { + "epoch": 0.82, + "learning_rate": 1.6175278943091888e-06, + "loss": 0.9922, + "step": 25701 + }, + { + "epoch": 0.82, + "learning_rate": 1.6169631254640084e-06, + "loss": 0.9355, + "step": 25702 + }, + { + "epoch": 0.82, + "learning_rate": 1.6163984465595317e-06, + "loss": 0.9819, + "step": 25703 + }, + { + "epoch": 0.82, + "learning_rate": 1.6158338576018251e-06, + "loss": 0.8008, + "step": 25704 + }, + { + "epoch": 0.82, + "learning_rate": 1.615269358596938e-06, + "loss": 0.8389, + "step": 25705 + }, + { + "epoch": 0.82, + "learning_rate": 1.6147049495509348e-06, + "loss": 0.9644, + "step": 25706 + }, + { + "epoch": 0.82, + "learning_rate": 1.614140630469867e-06, + "loss": 0.9146, + "step": 25707 + }, + { + "epoch": 0.82, + "learning_rate": 1.6135764013597865e-06, + "loss": 1.0068, + "step": 25708 + }, + { + "epoch": 0.82, + "learning_rate": 1.6130122622267518e-06, + "loss": 0.7776, + "step": 25709 + }, + { + "epoch": 0.82, + "learning_rate": 1.6124482130768126e-06, + "loss": 0.8618, + "step": 25710 + }, + { + "epoch": 0.82, + "learning_rate": 1.6118842539160229e-06, + "loss": 0.9507, + "step": 25711 + }, + { + "epoch": 0.82, + "learning_rate": 1.6113203847504277e-06, + "loss": 0.915, + "step": 25712 + }, + { + "epoch": 0.82, + "learning_rate": 1.6107566055860834e-06, + "loss": 0.8672, + "step": 25713 + }, + { + "epoch": 0.82, + "learning_rate": 1.6101929164290342e-06, + "loss": 0.8467, + "step": 25714 + }, + { + "epoch": 0.82, + "learning_rate": 1.6096293172853317e-06, + "loss": 0.9575, + "step": 25715 + }, + { + "epoch": 0.82, + "learning_rate": 1.6090658081610178e-06, + "loss": 0.8098, + "step": 25716 + }, + { + "epoch": 0.82, + "learning_rate": 1.6085023890621443e-06, + "loss": 0.8555, + "step": 25717 + }, + { + "epoch": 0.82, + "learning_rate": 1.607939059994752e-06, + "loss": 0.958, + "step": 25718 + }, + { + "epoch": 0.82, + "learning_rate": 1.607375820964887e-06, + "loss": 0.834, + "step": 25719 + }, + { + "epoch": 0.82, + "learning_rate": 1.606812671978587e-06, + "loss": 1.0, + "step": 25720 + }, + { + "epoch": 0.82, + "learning_rate": 1.6062496130419015e-06, + "loss": 1.0625, + "step": 25721 + }, + { + "epoch": 0.82, + "learning_rate": 1.6056866441608653e-06, + "loss": 0.9614, + "step": 25722 + }, + { + "epoch": 0.82, + "learning_rate": 1.605123765341523e-06, + "loss": 0.8501, + "step": 25723 + }, + { + "epoch": 0.82, + "learning_rate": 1.6045609765899095e-06, + "loss": 0.9248, + "step": 25724 + }, + { + "epoch": 0.82, + "learning_rate": 1.60399827791207e-06, + "loss": 0.8335, + "step": 25725 + }, + { + "epoch": 0.82, + "learning_rate": 1.6034356693140318e-06, + "loss": 0.917, + "step": 25726 + }, + { + "epoch": 0.82, + "learning_rate": 1.6028731508018392e-06, + "loss": 0.8428, + "step": 25727 + }, + { + "epoch": 0.82, + "learning_rate": 1.6023107223815204e-06, + "loss": 0.9541, + "step": 25728 + }, + { + "epoch": 0.82, + "learning_rate": 1.6017483840591164e-06, + "loss": 0.9839, + "step": 25729 + }, + { + "epoch": 0.82, + "learning_rate": 1.6011861358406554e-06, + "loss": 0.9448, + "step": 25730 + }, + { + "epoch": 0.82, + "learning_rate": 1.6006239777321742e-06, + "loss": 0.9385, + "step": 25731 + }, + { + "epoch": 0.82, + "learning_rate": 1.6000619097396996e-06, + "loss": 0.9131, + "step": 25732 + }, + { + "epoch": 0.82, + "learning_rate": 1.5994999318692684e-06, + "loss": 0.9741, + "step": 25733 + }, + { + "epoch": 0.82, + "learning_rate": 1.598938044126901e-06, + "loss": 0.9111, + "step": 25734 + }, + { + "epoch": 0.82, + "learning_rate": 1.5983762465186337e-06, + "loss": 0.8921, + "step": 25735 + }, + { + "epoch": 0.82, + "learning_rate": 1.5978145390504873e-06, + "loss": 0.8823, + "step": 25736 + }, + { + "epoch": 0.82, + "learning_rate": 1.5972529217284949e-06, + "loss": 0.4893, + "step": 25737 + }, + { + "epoch": 0.82, + "learning_rate": 1.5966913945586758e-06, + "loss": 0.8848, + "step": 25738 + }, + { + "epoch": 0.82, + "learning_rate": 1.5961299575470613e-06, + "loss": 0.875, + "step": 25739 + }, + { + "epoch": 0.82, + "learning_rate": 1.5955686106996705e-06, + "loss": 0.9487, + "step": 25740 + }, + { + "epoch": 0.82, + "learning_rate": 1.5950073540225274e-06, + "loss": 0.8608, + "step": 25741 + }, + { + "epoch": 0.82, + "learning_rate": 1.59444618752165e-06, + "loss": 1.002, + "step": 25742 + }, + { + "epoch": 0.82, + "learning_rate": 1.5938851112030651e-06, + "loss": 0.9644, + "step": 25743 + }, + { + "epoch": 0.82, + "learning_rate": 1.5933241250727883e-06, + "loss": 0.8696, + "step": 25744 + }, + { + "epoch": 0.82, + "learning_rate": 1.59276322913684e-06, + "loss": 0.9292, + "step": 25745 + }, + { + "epoch": 0.82, + "learning_rate": 1.592202423401239e-06, + "loss": 0.8872, + "step": 25746 + }, + { + "epoch": 0.82, + "learning_rate": 1.5916417078719981e-06, + "loss": 0.8972, + "step": 25747 + }, + { + "epoch": 0.82, + "learning_rate": 1.591081082555137e-06, + "loss": 0.9512, + "step": 25748 + }, + { + "epoch": 0.82, + "learning_rate": 1.590520547456671e-06, + "loss": 1.0303, + "step": 25749 + }, + { + "epoch": 0.82, + "learning_rate": 1.5899601025826116e-06, + "loss": 0.916, + "step": 25750 + }, + { + "epoch": 0.82, + "learning_rate": 1.58939974793897e-06, + "loss": 1.0034, + "step": 25751 + }, + { + "epoch": 0.82, + "learning_rate": 1.5888394835317634e-06, + "loss": 0.9722, + "step": 25752 + }, + { + "epoch": 0.82, + "learning_rate": 1.588279309366998e-06, + "loss": 0.9229, + "step": 25753 + }, + { + "epoch": 0.82, + "learning_rate": 1.5877192254506878e-06, + "loss": 0.8044, + "step": 25754 + }, + { + "epoch": 0.82, + "learning_rate": 1.5871592317888395e-06, + "loss": 1.0371, + "step": 25755 + }, + { + "epoch": 0.82, + "learning_rate": 1.5865993283874626e-06, + "loss": 0.9175, + "step": 25756 + }, + { + "epoch": 0.82, + "learning_rate": 1.5860395152525609e-06, + "loss": 0.9673, + "step": 25757 + }, + { + "epoch": 0.82, + "learning_rate": 1.5854797923901456e-06, + "loss": 0.9951, + "step": 25758 + }, + { + "epoch": 0.82, + "learning_rate": 1.584920159806217e-06, + "loss": 0.8711, + "step": 25759 + }, + { + "epoch": 0.82, + "learning_rate": 1.5843606175067838e-06, + "loss": 0.8831, + "step": 25760 + }, + { + "epoch": 0.82, + "learning_rate": 1.5838011654978458e-06, + "loss": 0.9771, + "step": 25761 + }, + { + "epoch": 0.82, + "learning_rate": 1.583241803785408e-06, + "loss": 0.9248, + "step": 25762 + }, + { + "epoch": 0.82, + "learning_rate": 1.5826825323754703e-06, + "loss": 0.957, + "step": 25763 + }, + { + "epoch": 0.82, + "learning_rate": 1.5821233512740342e-06, + "loss": 0.9458, + "step": 25764 + }, + { + "epoch": 0.82, + "learning_rate": 1.5815642604870952e-06, + "loss": 0.8594, + "step": 25765 + }, + { + "epoch": 0.82, + "learning_rate": 1.5810052600206572e-06, + "loss": 0.8506, + "step": 25766 + }, + { + "epoch": 0.82, + "learning_rate": 1.580446349880712e-06, + "loss": 0.8989, + "step": 25767 + }, + { + "epoch": 0.82, + "learning_rate": 1.579887530073262e-06, + "loss": 0.9829, + "step": 25768 + }, + { + "epoch": 0.82, + "learning_rate": 1.5793288006042974e-06, + "loss": 0.8887, + "step": 25769 + }, + { + "epoch": 0.82, + "learning_rate": 1.5787701614798201e-06, + "loss": 0.9756, + "step": 25770 + }, + { + "epoch": 0.82, + "learning_rate": 1.5782116127058145e-06, + "loss": 0.8188, + "step": 25771 + }, + { + "epoch": 0.82, + "learning_rate": 1.57765315428828e-06, + "loss": 0.9375, + "step": 25772 + }, + { + "epoch": 0.82, + "learning_rate": 1.577094786233202e-06, + "loss": 0.9785, + "step": 25773 + }, + { + "epoch": 0.82, + "learning_rate": 1.5765365085465789e-06, + "loss": 0.8887, + "step": 25774 + }, + { + "epoch": 0.82, + "learning_rate": 1.5759783212343927e-06, + "loss": 0.8003, + "step": 25775 + }, + { + "epoch": 0.82, + "learning_rate": 1.5754202243026394e-06, + "loss": 0.8223, + "step": 25776 + }, + { + "epoch": 0.82, + "learning_rate": 1.5748622177573004e-06, + "loss": 0.9678, + "step": 25777 + }, + { + "epoch": 0.82, + "learning_rate": 1.5743043016043702e-06, + "loss": 0.9424, + "step": 25778 + }, + { + "epoch": 0.82, + "learning_rate": 1.5737464758498243e-06, + "loss": 0.9199, + "step": 25779 + }, + { + "epoch": 0.82, + "learning_rate": 1.573188740499656e-06, + "loss": 0.8994, + "step": 25780 + }, + { + "epoch": 0.82, + "learning_rate": 1.5726310955598435e-06, + "loss": 0.9412, + "step": 25781 + }, + { + "epoch": 0.82, + "learning_rate": 1.5720735410363752e-06, + "loss": 0.4646, + "step": 25782 + }, + { + "epoch": 0.82, + "learning_rate": 1.5715160769352277e-06, + "loss": 0.9282, + "step": 25783 + }, + { + "epoch": 0.82, + "learning_rate": 1.5709587032623863e-06, + "loss": 0.9204, + "step": 25784 + }, + { + "epoch": 0.82, + "learning_rate": 1.5704014200238305e-06, + "loss": 0.9497, + "step": 25785 + }, + { + "epoch": 0.82, + "learning_rate": 1.5698442272255365e-06, + "loss": 1.0098, + "step": 25786 + }, + { + "epoch": 0.82, + "learning_rate": 1.569287124873483e-06, + "loss": 0.8877, + "step": 25787 + }, + { + "epoch": 0.82, + "learning_rate": 1.5687301129736498e-06, + "loss": 0.8154, + "step": 25788 + }, + { + "epoch": 0.82, + "learning_rate": 1.568173191532011e-06, + "loss": 0.9375, + "step": 25789 + }, + { + "epoch": 0.82, + "learning_rate": 1.5676163605545392e-06, + "loss": 0.9189, + "step": 25790 + }, + { + "epoch": 0.82, + "learning_rate": 1.5670596200472143e-06, + "loss": 0.9546, + "step": 25791 + }, + { + "epoch": 0.82, + "learning_rate": 1.566502970016004e-06, + "loss": 0.9419, + "step": 25792 + }, + { + "epoch": 0.82, + "learning_rate": 1.5659464104668864e-06, + "loss": 0.9688, + "step": 25793 + }, + { + "epoch": 0.82, + "learning_rate": 1.5653899414058282e-06, + "loss": 0.8569, + "step": 25794 + }, + { + "epoch": 0.82, + "learning_rate": 1.564833562838801e-06, + "loss": 0.8271, + "step": 25795 + }, + { + "epoch": 0.82, + "learning_rate": 1.5642772747717727e-06, + "loss": 0.8594, + "step": 25796 + }, + { + "epoch": 0.83, + "learning_rate": 1.5637210772107148e-06, + "loss": 1.0303, + "step": 25797 + }, + { + "epoch": 0.83, + "learning_rate": 1.5631649701615903e-06, + "loss": 0.8354, + "step": 25798 + }, + { + "epoch": 0.83, + "learning_rate": 1.5626089536303713e-06, + "loss": 0.9663, + "step": 25799 + }, + { + "epoch": 0.83, + "learning_rate": 1.5620530276230173e-06, + "loss": 0.8569, + "step": 25800 + }, + { + "epoch": 0.83, + "learning_rate": 1.5614971921455001e-06, + "loss": 1.0122, + "step": 25801 + }, + { + "epoch": 0.83, + "learning_rate": 1.5609414472037754e-06, + "loss": 1.0571, + "step": 25802 + }, + { + "epoch": 0.83, + "learning_rate": 1.5603857928038102e-06, + "loss": 0.8325, + "step": 25803 + }, + { + "epoch": 0.83, + "learning_rate": 1.5598302289515632e-06, + "loss": 0.9351, + "step": 25804 + }, + { + "epoch": 0.83, + "learning_rate": 1.5592747556529985e-06, + "loss": 0.936, + "step": 25805 + }, + { + "epoch": 0.83, + "learning_rate": 1.5587193729140716e-06, + "loss": 0.8813, + "step": 25806 + }, + { + "epoch": 0.83, + "learning_rate": 1.558164080740746e-06, + "loss": 0.9551, + "step": 25807 + }, + { + "epoch": 0.83, + "learning_rate": 1.5576088791389776e-06, + "loss": 0.9736, + "step": 25808 + }, + { + "epoch": 0.83, + "learning_rate": 1.5570537681147214e-06, + "loss": 0.9312, + "step": 25809 + }, + { + "epoch": 0.83, + "learning_rate": 1.5564987476739323e-06, + "loss": 0.8525, + "step": 25810 + }, + { + "epoch": 0.83, + "learning_rate": 1.5559438178225683e-06, + "loss": 0.8799, + "step": 25811 + }, + { + "epoch": 0.83, + "learning_rate": 1.55538897856658e-06, + "loss": 1.0093, + "step": 25812 + }, + { + "epoch": 0.83, + "learning_rate": 1.5548342299119247e-06, + "loss": 1.0444, + "step": 25813 + }, + { + "epoch": 0.83, + "learning_rate": 1.5542795718645487e-06, + "loss": 0.9062, + "step": 25814 + }, + { + "epoch": 0.83, + "learning_rate": 1.5537250044304076e-06, + "loss": 0.9531, + "step": 25815 + }, + { + "epoch": 0.83, + "learning_rate": 1.5531705276154496e-06, + "loss": 0.873, + "step": 25816 + }, + { + "epoch": 0.83, + "learning_rate": 1.5526161414256236e-06, + "loss": 0.9331, + "step": 25817 + }, + { + "epoch": 0.83, + "learning_rate": 1.5520618458668746e-06, + "loss": 0.8472, + "step": 25818 + }, + { + "epoch": 0.83, + "learning_rate": 1.5515076409451546e-06, + "loss": 0.936, + "step": 25819 + }, + { + "epoch": 0.83, + "learning_rate": 1.5509535266664055e-06, + "loss": 0.7954, + "step": 25820 + }, + { + "epoch": 0.83, + "learning_rate": 1.550399503036576e-06, + "loss": 0.9058, + "step": 25821 + }, + { + "epoch": 0.83, + "learning_rate": 1.5498455700616067e-06, + "loss": 0.8896, + "step": 25822 + }, + { + "epoch": 0.83, + "learning_rate": 1.5492917277474461e-06, + "loss": 0.9282, + "step": 25823 + }, + { + "epoch": 0.83, + "learning_rate": 1.5487379761000276e-06, + "loss": 0.98, + "step": 25824 + }, + { + "epoch": 0.83, + "learning_rate": 1.5481843151253008e-06, + "loss": 0.9482, + "step": 25825 + }, + { + "epoch": 0.83, + "learning_rate": 1.5476307448291982e-06, + "loss": 0.9385, + "step": 25826 + }, + { + "epoch": 0.83, + "learning_rate": 1.547077265217667e-06, + "loss": 0.9082, + "step": 25827 + }, + { + "epoch": 0.83, + "learning_rate": 1.5465238762966383e-06, + "loss": 0.856, + "step": 25828 + }, + { + "epoch": 0.83, + "learning_rate": 1.5459705780720547e-06, + "loss": 0.8765, + "step": 25829 + }, + { + "epoch": 0.83, + "learning_rate": 1.545417370549851e-06, + "loss": 0.8716, + "step": 25830 + }, + { + "epoch": 0.83, + "learning_rate": 1.5448642537359616e-06, + "loss": 0.9614, + "step": 25831 + }, + { + "epoch": 0.83, + "learning_rate": 1.5443112276363193e-06, + "loss": 0.9277, + "step": 25832 + }, + { + "epoch": 0.83, + "learning_rate": 1.5437582922568607e-06, + "loss": 0.9614, + "step": 25833 + }, + { + "epoch": 0.83, + "learning_rate": 1.5432054476035174e-06, + "loss": 0.9238, + "step": 25834 + }, + { + "epoch": 0.83, + "learning_rate": 1.5426526936822172e-06, + "loss": 0.8853, + "step": 25835 + }, + { + "epoch": 0.83, + "learning_rate": 1.5421000304988953e-06, + "loss": 1.2729, + "step": 25836 + }, + { + "epoch": 0.83, + "learning_rate": 1.541547458059478e-06, + "loss": 0.9561, + "step": 25837 + }, + { + "epoch": 0.83, + "learning_rate": 1.5409949763698973e-06, + "loss": 0.8486, + "step": 25838 + }, + { + "epoch": 0.83, + "learning_rate": 1.5404425854360783e-06, + "loss": 0.8828, + "step": 25839 + }, + { + "epoch": 0.83, + "learning_rate": 1.5398902852639485e-06, + "loss": 0.874, + "step": 25840 + }, + { + "epoch": 0.83, + "learning_rate": 1.539338075859429e-06, + "loss": 0.9229, + "step": 25841 + }, + { + "epoch": 0.83, + "learning_rate": 1.5387859572284513e-06, + "loss": 1.019, + "step": 25842 + }, + { + "epoch": 0.83, + "learning_rate": 1.5382339293769344e-06, + "loss": 0.9038, + "step": 25843 + }, + { + "epoch": 0.83, + "learning_rate": 1.5376819923108033e-06, + "loss": 1.0654, + "step": 25844 + }, + { + "epoch": 0.83, + "learning_rate": 1.5371301460359777e-06, + "loss": 0.853, + "step": 25845 + }, + { + "epoch": 0.83, + "learning_rate": 1.5365783905583832e-06, + "loss": 0.8965, + "step": 25846 + }, + { + "epoch": 0.83, + "learning_rate": 1.5360267258839312e-06, + "loss": 0.8779, + "step": 25847 + }, + { + "epoch": 0.83, + "learning_rate": 1.5354751520185473e-06, + "loss": 0.8901, + "step": 25848 + }, + { + "epoch": 0.83, + "learning_rate": 1.5349236689681447e-06, + "loss": 0.8921, + "step": 25849 + }, + { + "epoch": 0.83, + "learning_rate": 1.5343722767386448e-06, + "loss": 0.9351, + "step": 25850 + }, + { + "epoch": 0.83, + "learning_rate": 1.5338209753359589e-06, + "loss": 0.8345, + "step": 25851 + }, + { + "epoch": 0.83, + "learning_rate": 1.5332697647660056e-06, + "loss": 0.875, + "step": 25852 + }, + { + "epoch": 0.83, + "learning_rate": 1.5327186450346986e-06, + "loss": 0.8745, + "step": 25853 + }, + { + "epoch": 0.83, + "learning_rate": 1.5321676161479482e-06, + "loss": 0.9233, + "step": 25854 + }, + { + "epoch": 0.83, + "learning_rate": 1.5316166781116659e-06, + "loss": 0.4722, + "step": 25855 + }, + { + "epoch": 0.83, + "learning_rate": 1.531065830931766e-06, + "loss": 0.9727, + "step": 25856 + }, + { + "epoch": 0.83, + "learning_rate": 1.530515074614155e-06, + "loss": 0.9771, + "step": 25857 + }, + { + "epoch": 0.83, + "learning_rate": 1.529964409164746e-06, + "loss": 0.8955, + "step": 25858 + }, + { + "epoch": 0.83, + "learning_rate": 1.5294138345894417e-06, + "loss": 0.9561, + "step": 25859 + }, + { + "epoch": 0.83, + "learning_rate": 1.5288633508941552e-06, + "loss": 0.9614, + "step": 25860 + }, + { + "epoch": 0.83, + "learning_rate": 1.52831295808479e-06, + "loss": 0.9077, + "step": 25861 + }, + { + "epoch": 0.83, + "learning_rate": 1.52776265616725e-06, + "loss": 0.7944, + "step": 25862 + }, + { + "epoch": 0.83, + "learning_rate": 1.5272124451474378e-06, + "loss": 0.8911, + "step": 25863 + }, + { + "epoch": 0.83, + "learning_rate": 1.5266623250312608e-06, + "loss": 0.9551, + "step": 25864 + }, + { + "epoch": 0.83, + "learning_rate": 1.5261122958246178e-06, + "loss": 0.8608, + "step": 25865 + }, + { + "epoch": 0.83, + "learning_rate": 1.5255623575334133e-06, + "loss": 0.8818, + "step": 25866 + }, + { + "epoch": 0.83, + "learning_rate": 1.5250125101635427e-06, + "loss": 0.9897, + "step": 25867 + }, + { + "epoch": 0.83, + "learning_rate": 1.5244627537209121e-06, + "loss": 0.9233, + "step": 25868 + }, + { + "epoch": 0.83, + "learning_rate": 1.523913088211415e-06, + "loss": 0.9219, + "step": 25869 + }, + { + "epoch": 0.83, + "learning_rate": 1.5233635136409496e-06, + "loss": 0.9209, + "step": 25870 + }, + { + "epoch": 0.83, + "learning_rate": 1.5228140300154094e-06, + "loss": 0.8628, + "step": 25871 + }, + { + "epoch": 0.83, + "learning_rate": 1.5222646373406957e-06, + "loss": 1.0049, + "step": 25872 + }, + { + "epoch": 0.83, + "learning_rate": 1.5217153356226988e-06, + "loss": 0.8447, + "step": 25873 + }, + { + "epoch": 0.83, + "learning_rate": 1.5211661248673114e-06, + "loss": 0.9316, + "step": 25874 + }, + { + "epoch": 0.83, + "learning_rate": 1.5206170050804303e-06, + "loss": 0.9385, + "step": 25875 + }, + { + "epoch": 0.83, + "learning_rate": 1.5200679762679438e-06, + "loss": 0.9023, + "step": 25876 + }, + { + "epoch": 0.83, + "learning_rate": 1.5195190384357405e-06, + "loss": 0.905, + "step": 25877 + }, + { + "epoch": 0.83, + "learning_rate": 1.5189701915897149e-06, + "loss": 1.0112, + "step": 25878 + }, + { + "epoch": 0.83, + "learning_rate": 1.5184214357357517e-06, + "loss": 1.0229, + "step": 25879 + }, + { + "epoch": 0.83, + "learning_rate": 1.517872770879738e-06, + "loss": 0.8091, + "step": 25880 + }, + { + "epoch": 0.83, + "learning_rate": 1.5173241970275642e-06, + "loss": 0.9961, + "step": 25881 + }, + { + "epoch": 0.83, + "learning_rate": 1.5167757141851115e-06, + "loss": 0.8716, + "step": 25882 + }, + { + "epoch": 0.83, + "learning_rate": 1.516227322358268e-06, + "loss": 0.9731, + "step": 25883 + }, + { + "epoch": 0.83, + "learning_rate": 1.515679021552917e-06, + "loss": 0.8662, + "step": 25884 + }, + { + "epoch": 0.83, + "learning_rate": 1.5151308117749396e-06, + "loss": 0.9756, + "step": 25885 + }, + { + "epoch": 0.83, + "learning_rate": 1.5145826930302154e-06, + "loss": 0.8877, + "step": 25886 + }, + { + "epoch": 0.83, + "learning_rate": 1.5140346653246308e-06, + "loss": 0.9585, + "step": 25887 + }, + { + "epoch": 0.83, + "learning_rate": 1.5134867286640598e-06, + "loss": 0.9873, + "step": 25888 + }, + { + "epoch": 0.83, + "learning_rate": 1.5129388830543857e-06, + "loss": 0.9897, + "step": 25889 + }, + { + "epoch": 0.83, + "learning_rate": 1.5123911285014826e-06, + "loss": 0.9961, + "step": 25890 + }, + { + "epoch": 0.83, + "learning_rate": 1.5118434650112334e-06, + "loss": 0.8662, + "step": 25891 + }, + { + "epoch": 0.83, + "learning_rate": 1.511295892589506e-06, + "loss": 0.9736, + "step": 25892 + }, + { + "epoch": 0.83, + "learning_rate": 1.51074841124218e-06, + "loss": 0.9023, + "step": 25893 + }, + { + "epoch": 0.83, + "learning_rate": 1.5102010209751262e-06, + "loss": 0.9346, + "step": 25894 + }, + { + "epoch": 0.83, + "learning_rate": 1.509653721794221e-06, + "loss": 0.9673, + "step": 25895 + }, + { + "epoch": 0.83, + "learning_rate": 1.509106513705333e-06, + "loss": 0.8726, + "step": 25896 + }, + { + "epoch": 0.83, + "learning_rate": 1.5085593967143375e-06, + "loss": 0.8223, + "step": 25897 + }, + { + "epoch": 0.83, + "learning_rate": 1.5080123708270989e-06, + "loss": 0.8442, + "step": 25898 + }, + { + "epoch": 0.83, + "learning_rate": 1.5074654360494934e-06, + "loss": 0.8643, + "step": 25899 + }, + { + "epoch": 0.83, + "learning_rate": 1.5069185923873808e-06, + "loss": 0.897, + "step": 25900 + }, + { + "epoch": 0.83, + "learning_rate": 1.5063718398466343e-06, + "loss": 0.8462, + "step": 25901 + }, + { + "epoch": 0.83, + "learning_rate": 1.5058251784331158e-06, + "loss": 0.9272, + "step": 25902 + }, + { + "epoch": 0.83, + "learning_rate": 1.5052786081526937e-06, + "loss": 0.8096, + "step": 25903 + }, + { + "epoch": 0.83, + "learning_rate": 1.5047321290112282e-06, + "loss": 0.8979, + "step": 25904 + }, + { + "epoch": 0.83, + "learning_rate": 1.5041857410145878e-06, + "loss": 0.8589, + "step": 25905 + }, + { + "epoch": 0.83, + "learning_rate": 1.503639444168632e-06, + "loss": 0.9053, + "step": 25906 + }, + { + "epoch": 0.83, + "learning_rate": 1.5030932384792206e-06, + "loss": 0.4609, + "step": 25907 + }, + { + "epoch": 0.83, + "learning_rate": 1.5025471239522128e-06, + "loss": 1.0273, + "step": 25908 + }, + { + "epoch": 0.83, + "learning_rate": 1.5020011005934732e-06, + "loss": 0.8232, + "step": 25909 + }, + { + "epoch": 0.83, + "learning_rate": 1.501455168408853e-06, + "loss": 0.918, + "step": 25910 + }, + { + "epoch": 0.83, + "learning_rate": 1.5009093274042164e-06, + "loss": 0.8286, + "step": 25911 + }, + { + "epoch": 0.83, + "learning_rate": 1.5003635775854154e-06, + "loss": 1.063, + "step": 25912 + }, + { + "epoch": 0.83, + "learning_rate": 1.4998179189583051e-06, + "loss": 0.9097, + "step": 25913 + }, + { + "epoch": 0.83, + "learning_rate": 1.4992723515287423e-06, + "loss": 0.709, + "step": 25914 + }, + { + "epoch": 0.83, + "learning_rate": 1.4987268753025796e-06, + "loss": 0.9424, + "step": 25915 + }, + { + "epoch": 0.83, + "learning_rate": 1.4981814902856662e-06, + "loss": 0.7974, + "step": 25916 + }, + { + "epoch": 0.83, + "learning_rate": 1.4976361964838592e-06, + "loss": 0.8638, + "step": 25917 + }, + { + "epoch": 0.83, + "learning_rate": 1.4970909939030054e-06, + "loss": 0.8306, + "step": 25918 + }, + { + "epoch": 0.83, + "learning_rate": 1.496545882548952e-06, + "loss": 0.9849, + "step": 25919 + }, + { + "epoch": 0.83, + "learning_rate": 1.4960008624275525e-06, + "loss": 0.895, + "step": 25920 + }, + { + "epoch": 0.83, + "learning_rate": 1.4954559335446518e-06, + "loss": 0.9683, + "step": 25921 + }, + { + "epoch": 0.83, + "learning_rate": 1.4949110959060953e-06, + "loss": 0.8872, + "step": 25922 + }, + { + "epoch": 0.83, + "learning_rate": 1.4943663495177307e-06, + "loss": 0.9722, + "step": 25923 + }, + { + "epoch": 0.83, + "learning_rate": 1.4938216943854022e-06, + "loss": 0.9126, + "step": 25924 + }, + { + "epoch": 0.83, + "learning_rate": 1.4932771305149507e-06, + "loss": 0.8687, + "step": 25925 + }, + { + "epoch": 0.83, + "learning_rate": 1.4927326579122226e-06, + "loss": 0.9438, + "step": 25926 + }, + { + "epoch": 0.83, + "learning_rate": 1.4921882765830565e-06, + "loss": 0.999, + "step": 25927 + }, + { + "epoch": 0.83, + "learning_rate": 1.4916439865332955e-06, + "loss": 0.999, + "step": 25928 + }, + { + "epoch": 0.83, + "learning_rate": 1.4910997877687783e-06, + "loss": 0.8682, + "step": 25929 + }, + { + "epoch": 0.83, + "learning_rate": 1.4905556802953425e-06, + "loss": 0.9702, + "step": 25930 + }, + { + "epoch": 0.83, + "learning_rate": 1.4900116641188244e-06, + "loss": 0.7808, + "step": 25931 + }, + { + "epoch": 0.83, + "learning_rate": 1.4894677392450662e-06, + "loss": 0.8901, + "step": 25932 + }, + { + "epoch": 0.83, + "learning_rate": 1.4889239056798965e-06, + "loss": 0.9116, + "step": 25933 + }, + { + "epoch": 0.83, + "learning_rate": 1.4883801634291572e-06, + "loss": 1.0181, + "step": 25934 + }, + { + "epoch": 0.83, + "learning_rate": 1.4878365124986748e-06, + "loss": 1.0151, + "step": 25935 + }, + { + "epoch": 0.83, + "learning_rate": 1.4872929528942914e-06, + "loss": 0.9629, + "step": 25936 + }, + { + "epoch": 0.83, + "learning_rate": 1.486749484621829e-06, + "loss": 0.9141, + "step": 25937 + }, + { + "epoch": 0.83, + "learning_rate": 1.4862061076871238e-06, + "loss": 0.751, + "step": 25938 + }, + { + "epoch": 0.83, + "learning_rate": 1.4856628220960034e-06, + "loss": 0.894, + "step": 25939 + }, + { + "epoch": 0.83, + "learning_rate": 1.485119627854299e-06, + "loss": 1.0137, + "step": 25940 + }, + { + "epoch": 0.83, + "learning_rate": 1.4845765249678356e-06, + "loss": 0.8186, + "step": 25941 + }, + { + "epoch": 0.83, + "learning_rate": 1.4840335134424432e-06, + "loss": 0.8867, + "step": 25942 + }, + { + "epoch": 0.83, + "learning_rate": 1.4834905932839438e-06, + "loss": 0.9756, + "step": 25943 + }, + { + "epoch": 0.83, + "learning_rate": 1.4829477644981704e-06, + "loss": 0.9448, + "step": 25944 + }, + { + "epoch": 0.83, + "learning_rate": 1.4824050270909363e-06, + "loss": 0.9609, + "step": 25945 + }, + { + "epoch": 0.83, + "learning_rate": 1.481862381068071e-06, + "loss": 1.064, + "step": 25946 + }, + { + "epoch": 0.83, + "learning_rate": 1.4813198264353934e-06, + "loss": 0.8125, + "step": 25947 + }, + { + "epoch": 0.83, + "learning_rate": 1.4807773631987288e-06, + "loss": 0.9673, + "step": 25948 + }, + { + "epoch": 0.83, + "learning_rate": 1.4802349913638913e-06, + "loss": 0.9893, + "step": 25949 + }, + { + "epoch": 0.83, + "learning_rate": 1.4796927109367054e-06, + "loss": 1.001, + "step": 25950 + }, + { + "epoch": 0.83, + "learning_rate": 1.4791505219229874e-06, + "loss": 0.9443, + "step": 25951 + }, + { + "epoch": 0.83, + "learning_rate": 1.4786084243285537e-06, + "loss": 0.9263, + "step": 25952 + }, + { + "epoch": 0.83, + "learning_rate": 1.4780664181592185e-06, + "loss": 0.8921, + "step": 25953 + }, + { + "epoch": 0.83, + "learning_rate": 1.4775245034208008e-06, + "loss": 0.9053, + "step": 25954 + }, + { + "epoch": 0.83, + "learning_rate": 1.4769826801191111e-06, + "loss": 0.873, + "step": 25955 + }, + { + "epoch": 0.83, + "learning_rate": 1.4764409482599661e-06, + "loss": 0.8616, + "step": 25956 + }, + { + "epoch": 0.83, + "learning_rate": 1.475899307849178e-06, + "loss": 0.9448, + "step": 25957 + }, + { + "epoch": 0.83, + "learning_rate": 1.4753577588925527e-06, + "loss": 0.9731, + "step": 25958 + }, + { + "epoch": 0.83, + "learning_rate": 1.474816301395906e-06, + "loss": 0.9351, + "step": 25959 + }, + { + "epoch": 0.83, + "learning_rate": 1.4742749353650454e-06, + "loss": 0.8457, + "step": 25960 + }, + { + "epoch": 0.83, + "learning_rate": 1.4737336608057772e-06, + "loss": 0.9482, + "step": 25961 + }, + { + "epoch": 0.83, + "learning_rate": 1.4731924777239125e-06, + "loss": 1.0337, + "step": 25962 + }, + { + "epoch": 0.83, + "learning_rate": 1.4726513861252556e-06, + "loss": 0.9863, + "step": 25963 + }, + { + "epoch": 0.83, + "learning_rate": 1.4721103860156094e-06, + "loss": 0.8516, + "step": 25964 + }, + { + "epoch": 0.83, + "learning_rate": 1.4715694774007827e-06, + "loss": 0.8833, + "step": 25965 + }, + { + "epoch": 0.83, + "learning_rate": 1.4710286602865743e-06, + "loss": 0.9136, + "step": 25966 + }, + { + "epoch": 0.83, + "learning_rate": 1.4704879346787938e-06, + "loss": 0.8496, + "step": 25967 + }, + { + "epoch": 0.83, + "learning_rate": 1.4699473005832332e-06, + "loss": 0.9243, + "step": 25968 + }, + { + "epoch": 0.83, + "learning_rate": 1.4694067580056993e-06, + "loss": 0.8403, + "step": 25969 + }, + { + "epoch": 0.83, + "learning_rate": 1.4688663069519882e-06, + "loss": 0.9194, + "step": 25970 + }, + { + "epoch": 0.83, + "learning_rate": 1.4683259474279011e-06, + "loss": 0.8877, + "step": 25971 + }, + { + "epoch": 0.83, + "learning_rate": 1.4677856794392332e-06, + "loss": 0.8579, + "step": 25972 + }, + { + "epoch": 0.83, + "learning_rate": 1.467245502991783e-06, + "loss": 0.8516, + "step": 25973 + }, + { + "epoch": 0.83, + "learning_rate": 1.4667054180913453e-06, + "loss": 1.0762, + "step": 25974 + }, + { + "epoch": 0.83, + "learning_rate": 1.4661654247437152e-06, + "loss": 1.0337, + "step": 25975 + }, + { + "epoch": 0.83, + "learning_rate": 1.4656255229546824e-06, + "loss": 0.856, + "step": 25976 + }, + { + "epoch": 0.83, + "learning_rate": 1.4650857127300444e-06, + "loss": 0.9971, + "step": 25977 + }, + { + "epoch": 0.83, + "learning_rate": 1.464545994075588e-06, + "loss": 0.9487, + "step": 25978 + }, + { + "epoch": 0.83, + "learning_rate": 1.4640063669971095e-06, + "loss": 1.0532, + "step": 25979 + }, + { + "epoch": 0.83, + "learning_rate": 1.463466831500393e-06, + "loss": 0.7959, + "step": 25980 + }, + { + "epoch": 0.83, + "learning_rate": 1.4629273875912319e-06, + "loss": 0.9556, + "step": 25981 + }, + { + "epoch": 0.83, + "learning_rate": 1.4623880352754116e-06, + "loss": 0.9009, + "step": 25982 + }, + { + "epoch": 0.83, + "learning_rate": 1.4618487745587195e-06, + "loss": 0.9297, + "step": 25983 + }, + { + "epoch": 0.83, + "learning_rate": 1.4613096054469379e-06, + "loss": 0.938, + "step": 25984 + }, + { + "epoch": 0.83, + "learning_rate": 1.460770527945855e-06, + "loss": 0.9556, + "step": 25985 + }, + { + "epoch": 0.83, + "learning_rate": 1.4602315420612522e-06, + "loss": 0.9321, + "step": 25986 + }, + { + "epoch": 0.83, + "learning_rate": 1.4596926477989158e-06, + "loss": 0.8086, + "step": 25987 + }, + { + "epoch": 0.83, + "learning_rate": 1.4591538451646237e-06, + "loss": 0.7385, + "step": 25988 + }, + { + "epoch": 0.83, + "learning_rate": 1.458615134164163e-06, + "loss": 0.9214, + "step": 25989 + }, + { + "epoch": 0.83, + "learning_rate": 1.4580765148033038e-06, + "loss": 0.9502, + "step": 25990 + }, + { + "epoch": 0.83, + "learning_rate": 1.4575379870878315e-06, + "loss": 0.9419, + "step": 25991 + }, + { + "epoch": 0.83, + "learning_rate": 1.4569995510235201e-06, + "loss": 0.9082, + "step": 25992 + }, + { + "epoch": 0.83, + "learning_rate": 1.4564612066161521e-06, + "loss": 0.8081, + "step": 25993 + }, + { + "epoch": 0.83, + "learning_rate": 1.455922953871496e-06, + "loss": 0.9497, + "step": 25994 + }, + { + "epoch": 0.83, + "learning_rate": 1.4553847927953334e-06, + "loss": 0.9419, + "step": 25995 + }, + { + "epoch": 0.83, + "learning_rate": 1.454846723393436e-06, + "loss": 0.9526, + "step": 25996 + }, + { + "epoch": 0.83, + "learning_rate": 1.4543087456715744e-06, + "loss": 0.917, + "step": 25997 + }, + { + "epoch": 0.83, + "learning_rate": 1.4537708596355216e-06, + "loss": 0.9072, + "step": 25998 + }, + { + "epoch": 0.83, + "learning_rate": 1.4532330652910497e-06, + "loss": 0.8765, + "step": 25999 + }, + { + "epoch": 0.83, + "learning_rate": 1.4526953626439267e-06, + "loss": 0.9761, + "step": 26000 + }, + { + "epoch": 0.83, + "learning_rate": 1.452157751699924e-06, + "loss": 0.9556, + "step": 26001 + }, + { + "epoch": 0.83, + "learning_rate": 1.4516202324648088e-06, + "loss": 0.9932, + "step": 26002 + }, + { + "epoch": 0.83, + "learning_rate": 1.4510828049443459e-06, + "loss": 0.8809, + "step": 26003 + }, + { + "epoch": 0.83, + "learning_rate": 1.4505454691443043e-06, + "loss": 1.0186, + "step": 26004 + }, + { + "epoch": 0.83, + "learning_rate": 1.4500082250704483e-06, + "loss": 1.0635, + "step": 26005 + }, + { + "epoch": 0.83, + "learning_rate": 1.449471072728542e-06, + "loss": 0.8423, + "step": 26006 + }, + { + "epoch": 0.83, + "learning_rate": 1.4489340121243446e-06, + "loss": 0.969, + "step": 26007 + }, + { + "epoch": 0.83, + "learning_rate": 1.4483970432636241e-06, + "loss": 0.8691, + "step": 26008 + }, + { + "epoch": 0.83, + "learning_rate": 1.4478601661521374e-06, + "loss": 0.9468, + "step": 26009 + }, + { + "epoch": 0.83, + "learning_rate": 1.4473233807956466e-06, + "loss": 0.96, + "step": 26010 + }, + { + "epoch": 0.83, + "learning_rate": 1.4467866871999093e-06, + "loss": 0.9258, + "step": 26011 + }, + { + "epoch": 0.83, + "learning_rate": 1.4462500853706884e-06, + "loss": 0.8589, + "step": 26012 + }, + { + "epoch": 0.83, + "learning_rate": 1.445713575313733e-06, + "loss": 0.9863, + "step": 26013 + }, + { + "epoch": 0.83, + "learning_rate": 1.445177157034806e-06, + "loss": 0.9722, + "step": 26014 + }, + { + "epoch": 0.83, + "learning_rate": 1.4446408305396587e-06, + "loss": 0.8113, + "step": 26015 + }, + { + "epoch": 0.83, + "learning_rate": 1.4441045958340472e-06, + "loss": 0.8628, + "step": 26016 + }, + { + "epoch": 0.83, + "learning_rate": 1.4435684529237226e-06, + "loss": 0.915, + "step": 26017 + }, + { + "epoch": 0.83, + "learning_rate": 1.4430324018144416e-06, + "loss": 0.9102, + "step": 26018 + }, + { + "epoch": 0.83, + "learning_rate": 1.4424964425119513e-06, + "loss": 1.0151, + "step": 26019 + }, + { + "epoch": 0.83, + "learning_rate": 1.4419605750220044e-06, + "loss": 0.8711, + "step": 26020 + }, + { + "epoch": 0.83, + "learning_rate": 1.441424799350346e-06, + "loss": 0.9199, + "step": 26021 + }, + { + "epoch": 0.83, + "learning_rate": 1.4408891155027304e-06, + "loss": 0.8198, + "step": 26022 + }, + { + "epoch": 0.83, + "learning_rate": 1.4403535234848987e-06, + "loss": 0.9302, + "step": 26023 + }, + { + "epoch": 0.83, + "learning_rate": 1.4398180233026027e-06, + "loss": 0.811, + "step": 26024 + }, + { + "epoch": 0.83, + "learning_rate": 1.4392826149615835e-06, + "loss": 0.9438, + "step": 26025 + }, + { + "epoch": 0.83, + "learning_rate": 1.4387472984675899e-06, + "loss": 1.0503, + "step": 26026 + }, + { + "epoch": 0.83, + "learning_rate": 1.4382120738263617e-06, + "loss": 0.8452, + "step": 26027 + }, + { + "epoch": 0.83, + "learning_rate": 1.437676941043643e-06, + "loss": 0.9575, + "step": 26028 + }, + { + "epoch": 0.83, + "learning_rate": 1.437141900125172e-06, + "loss": 0.8848, + "step": 26029 + }, + { + "epoch": 0.83, + "learning_rate": 1.4366069510766934e-06, + "loss": 0.8784, + "step": 26030 + }, + { + "epoch": 0.83, + "learning_rate": 1.436072093903943e-06, + "loss": 0.853, + "step": 26031 + }, + { + "epoch": 0.83, + "learning_rate": 1.435537328612663e-06, + "loss": 0.9292, + "step": 26032 + }, + { + "epoch": 0.83, + "learning_rate": 1.4350026552085861e-06, + "loss": 0.9287, + "step": 26033 + }, + { + "epoch": 0.83, + "learning_rate": 1.4344680736974536e-06, + "loss": 0.9238, + "step": 26034 + }, + { + "epoch": 0.83, + "learning_rate": 1.433933584084999e-06, + "loss": 0.813, + "step": 26035 + }, + { + "epoch": 0.83, + "learning_rate": 1.433399186376957e-06, + "loss": 0.9282, + "step": 26036 + }, + { + "epoch": 0.83, + "learning_rate": 1.432864880579058e-06, + "loss": 0.8862, + "step": 26037 + }, + { + "epoch": 0.83, + "learning_rate": 1.4323306666970405e-06, + "loss": 0.9375, + "step": 26038 + }, + { + "epoch": 0.83, + "learning_rate": 1.431796544736629e-06, + "loss": 1.084, + "step": 26039 + }, + { + "epoch": 0.83, + "learning_rate": 1.4312625147035607e-06, + "loss": 0.7229, + "step": 26040 + }, + { + "epoch": 0.83, + "learning_rate": 1.4307285766035616e-06, + "loss": 1.0166, + "step": 26041 + }, + { + "epoch": 0.83, + "learning_rate": 1.4301947304423624e-06, + "loss": 0.9761, + "step": 26042 + }, + { + "epoch": 0.83, + "learning_rate": 1.4296609762256853e-06, + "loss": 0.9893, + "step": 26043 + }, + { + "epoch": 0.83, + "learning_rate": 1.4291273139592643e-06, + "loss": 0.8567, + "step": 26044 + }, + { + "epoch": 0.83, + "learning_rate": 1.4285937436488184e-06, + "loss": 1.0825, + "step": 26045 + }, + { + "epoch": 0.83, + "learning_rate": 1.4280602653000774e-06, + "loss": 1.0322, + "step": 26046 + }, + { + "epoch": 0.83, + "learning_rate": 1.4275268789187635e-06, + "loss": 0.9058, + "step": 26047 + }, + { + "epoch": 0.83, + "learning_rate": 1.4269935845105954e-06, + "loss": 0.9697, + "step": 26048 + }, + { + "epoch": 0.83, + "learning_rate": 1.4264603820813006e-06, + "loss": 1.0234, + "step": 26049 + }, + { + "epoch": 0.83, + "learning_rate": 1.4259272716365979e-06, + "loss": 1.0322, + "step": 26050 + }, + { + "epoch": 0.83, + "learning_rate": 1.4253942531822052e-06, + "loss": 0.8862, + "step": 26051 + }, + { + "epoch": 0.83, + "learning_rate": 1.424861326723841e-06, + "loss": 0.9805, + "step": 26052 + }, + { + "epoch": 0.83, + "learning_rate": 1.4243284922672263e-06, + "loss": 0.9253, + "step": 26053 + }, + { + "epoch": 0.83, + "learning_rate": 1.4237957498180733e-06, + "loss": 0.8506, + "step": 26054 + }, + { + "epoch": 0.83, + "learning_rate": 1.4232630993821017e-06, + "loss": 0.9331, + "step": 26055 + }, + { + "epoch": 0.83, + "learning_rate": 1.4227305409650238e-06, + "loss": 0.8872, + "step": 26056 + }, + { + "epoch": 0.83, + "learning_rate": 1.4221980745725584e-06, + "loss": 0.8647, + "step": 26057 + }, + { + "epoch": 0.83, + "learning_rate": 1.4216657002104095e-06, + "loss": 0.8848, + "step": 26058 + }, + { + "epoch": 0.83, + "learning_rate": 1.421133417884296e-06, + "loss": 0.855, + "step": 26059 + }, + { + "epoch": 0.83, + "learning_rate": 1.4206012275999237e-06, + "loss": 0.96, + "step": 26060 + }, + { + "epoch": 0.83, + "learning_rate": 1.4200691293630086e-06, + "loss": 0.9219, + "step": 26061 + }, + { + "epoch": 0.83, + "learning_rate": 1.419537123179252e-06, + "loss": 0.9175, + "step": 26062 + }, + { + "epoch": 0.83, + "learning_rate": 1.4190052090543694e-06, + "loss": 0.8784, + "step": 26063 + }, + { + "epoch": 0.83, + "learning_rate": 1.4184733869940604e-06, + "loss": 0.9346, + "step": 26064 + }, + { + "epoch": 0.83, + "learning_rate": 1.4179416570040393e-06, + "loss": 0.8755, + "step": 26065 + }, + { + "epoch": 0.83, + "learning_rate": 1.417410019090002e-06, + "loss": 0.7788, + "step": 26066 + }, + { + "epoch": 0.83, + "learning_rate": 1.416878473257659e-06, + "loss": 0.8516, + "step": 26067 + }, + { + "epoch": 0.83, + "learning_rate": 1.4163470195127072e-06, + "loss": 0.8599, + "step": 26068 + }, + { + "epoch": 0.83, + "learning_rate": 1.415815657860855e-06, + "loss": 0.8892, + "step": 26069 + }, + { + "epoch": 0.83, + "learning_rate": 1.4152843883077982e-06, + "loss": 0.915, + "step": 26070 + }, + { + "epoch": 0.83, + "learning_rate": 1.414753210859241e-06, + "loss": 0.9307, + "step": 26071 + }, + { + "epoch": 0.83, + "learning_rate": 1.41422212552088e-06, + "loss": 0.8877, + "step": 26072 + }, + { + "epoch": 0.83, + "learning_rate": 1.4136911322984126e-06, + "loss": 0.9331, + "step": 26073 + }, + { + "epoch": 0.83, + "learning_rate": 1.4131602311975356e-06, + "loss": 0.8267, + "step": 26074 + }, + { + "epoch": 0.83, + "learning_rate": 1.4126294222239467e-06, + "loss": 0.9194, + "step": 26075 + }, + { + "epoch": 0.83, + "learning_rate": 1.412098705383339e-06, + "loss": 0.8555, + "step": 26076 + }, + { + "epoch": 0.83, + "learning_rate": 1.4115680806814092e-06, + "loss": 0.8643, + "step": 26077 + }, + { + "epoch": 0.83, + "learning_rate": 1.411037548123847e-06, + "loss": 0.8999, + "step": 26078 + }, + { + "epoch": 0.83, + "learning_rate": 1.410507107716348e-06, + "loss": 0.9092, + "step": 26079 + }, + { + "epoch": 0.83, + "learning_rate": 1.4099767594646009e-06, + "loss": 0.9331, + "step": 26080 + }, + { + "epoch": 0.83, + "learning_rate": 1.4094465033742965e-06, + "loss": 1.0122, + "step": 26081 + }, + { + "epoch": 0.83, + "learning_rate": 1.4089163394511218e-06, + "loss": 0.9595, + "step": 26082 + }, + { + "epoch": 0.83, + "learning_rate": 1.4083862677007688e-06, + "loss": 0.8882, + "step": 26083 + }, + { + "epoch": 0.83, + "learning_rate": 1.4078562881289192e-06, + "loss": 0.9741, + "step": 26084 + }, + { + "epoch": 0.83, + "learning_rate": 1.4073264007412657e-06, + "loss": 0.9819, + "step": 26085 + }, + { + "epoch": 0.83, + "learning_rate": 1.406796605543489e-06, + "loss": 1.0234, + "step": 26086 + }, + { + "epoch": 0.83, + "learning_rate": 1.4062669025412756e-06, + "loss": 0.9473, + "step": 26087 + }, + { + "epoch": 0.83, + "learning_rate": 1.4057372917403034e-06, + "loss": 0.8584, + "step": 26088 + }, + { + "epoch": 0.83, + "learning_rate": 1.405207773146262e-06, + "loss": 0.9844, + "step": 26089 + }, + { + "epoch": 0.83, + "learning_rate": 1.4046783467648285e-06, + "loss": 0.9014, + "step": 26090 + }, + { + "epoch": 0.83, + "learning_rate": 1.4041490126016811e-06, + "loss": 0.9106, + "step": 26091 + }, + { + "epoch": 0.83, + "learning_rate": 1.4036197706625044e-06, + "loss": 0.8833, + "step": 26092 + }, + { + "epoch": 0.83, + "learning_rate": 1.4030906209529705e-06, + "loss": 0.9155, + "step": 26093 + }, + { + "epoch": 0.83, + "learning_rate": 1.4025615634787616e-06, + "loss": 0.9141, + "step": 26094 + }, + { + "epoch": 0.83, + "learning_rate": 1.402032598245553e-06, + "loss": 0.9395, + "step": 26095 + }, + { + "epoch": 0.83, + "learning_rate": 1.401503725259018e-06, + "loss": 0.8433, + "step": 26096 + }, + { + "epoch": 0.83, + "learning_rate": 1.4009749445248288e-06, + "loss": 0.894, + "step": 26097 + }, + { + "epoch": 0.83, + "learning_rate": 1.4004462560486653e-06, + "loss": 0.8245, + "step": 26098 + }, + { + "epoch": 0.83, + "learning_rate": 1.399917659836193e-06, + "loss": 0.9556, + "step": 26099 + }, + { + "epoch": 0.83, + "learning_rate": 1.3993891558930883e-06, + "loss": 0.9316, + "step": 26100 + }, + { + "epoch": 0.83, + "learning_rate": 1.3988607442250168e-06, + "loss": 1.0234, + "step": 26101 + }, + { + "epoch": 0.83, + "learning_rate": 1.3983324248376528e-06, + "loss": 0.9819, + "step": 26102 + }, + { + "epoch": 0.83, + "learning_rate": 1.397804197736663e-06, + "loss": 0.9106, + "step": 26103 + }, + { + "epoch": 0.83, + "learning_rate": 1.3972760629277128e-06, + "loss": 0.8755, + "step": 26104 + }, + { + "epoch": 0.83, + "learning_rate": 1.3967480204164662e-06, + "loss": 0.8999, + "step": 26105 + }, + { + "epoch": 0.83, + "learning_rate": 1.3962200702085959e-06, + "loss": 0.833, + "step": 26106 + }, + { + "epoch": 0.83, + "learning_rate": 1.3956922123097582e-06, + "loss": 0.8574, + "step": 26107 + }, + { + "epoch": 0.83, + "learning_rate": 1.395164446725623e-06, + "loss": 0.8784, + "step": 26108 + }, + { + "epoch": 0.83, + "learning_rate": 1.3946367734618481e-06, + "loss": 0.8042, + "step": 26109 + }, + { + "epoch": 0.84, + "learning_rate": 1.3941091925241012e-06, + "loss": 1.0225, + "step": 26110 + }, + { + "epoch": 0.84, + "learning_rate": 1.3935817039180333e-06, + "loss": 0.9072, + "step": 26111 + }, + { + "epoch": 0.84, + "learning_rate": 1.393054307649311e-06, + "loss": 0.9404, + "step": 26112 + }, + { + "epoch": 0.84, + "learning_rate": 1.3925270037235883e-06, + "loss": 0.8896, + "step": 26113 + }, + { + "epoch": 0.84, + "learning_rate": 1.3919997921465255e-06, + "loss": 0.8877, + "step": 26114 + }, + { + "epoch": 0.84, + "learning_rate": 1.391472672923777e-06, + "loss": 0.8218, + "step": 26115 + }, + { + "epoch": 0.84, + "learning_rate": 1.3909456460610015e-06, + "loss": 0.9224, + "step": 26116 + }, + { + "epoch": 0.84, + "learning_rate": 1.390418711563849e-06, + "loss": 0.9429, + "step": 26117 + }, + { + "epoch": 0.84, + "learning_rate": 1.3898918694379792e-06, + "loss": 0.8765, + "step": 26118 + }, + { + "epoch": 0.84, + "learning_rate": 1.3893651196890357e-06, + "loss": 0.8638, + "step": 26119 + }, + { + "epoch": 0.84, + "learning_rate": 1.388838462322678e-06, + "loss": 0.9028, + "step": 26120 + }, + { + "epoch": 0.84, + "learning_rate": 1.3883118973445497e-06, + "loss": 0.915, + "step": 26121 + }, + { + "epoch": 0.84, + "learning_rate": 1.3877854247603073e-06, + "loss": 0.9731, + "step": 26122 + }, + { + "epoch": 0.84, + "learning_rate": 1.387259044575593e-06, + "loss": 0.894, + "step": 26123 + }, + { + "epoch": 0.84, + "learning_rate": 1.3867327567960587e-06, + "loss": 0.9854, + "step": 26124 + }, + { + "epoch": 0.84, + "learning_rate": 1.38620656142735e-06, + "loss": 0.4636, + "step": 26125 + }, + { + "epoch": 0.84, + "learning_rate": 1.3856804584751116e-06, + "loss": 0.9326, + "step": 26126 + }, + { + "epoch": 0.84, + "learning_rate": 1.3851544479449852e-06, + "loss": 0.8838, + "step": 26127 + }, + { + "epoch": 0.84, + "learning_rate": 1.3846285298426198e-06, + "loss": 0.8174, + "step": 26128 + }, + { + "epoch": 0.84, + "learning_rate": 1.3841027041736554e-06, + "loss": 0.9155, + "step": 26129 + }, + { + "epoch": 0.84, + "learning_rate": 1.3835769709437307e-06, + "loss": 1.0181, + "step": 26130 + }, + { + "epoch": 0.84, + "learning_rate": 1.3830513301584913e-06, + "loss": 0.9692, + "step": 26131 + }, + { + "epoch": 0.84, + "learning_rate": 1.3825257818235726e-06, + "loss": 0.9551, + "step": 26132 + }, + { + "epoch": 0.84, + "learning_rate": 1.3820003259446157e-06, + "loss": 0.9424, + "step": 26133 + }, + { + "epoch": 0.84, + "learning_rate": 1.3814749625272594e-06, + "loss": 0.9243, + "step": 26134 + }, + { + "epoch": 0.84, + "learning_rate": 1.380949691577137e-06, + "loss": 0.8521, + "step": 26135 + }, + { + "epoch": 0.84, + "learning_rate": 1.3804245130998828e-06, + "loss": 0.8027, + "step": 26136 + }, + { + "epoch": 0.84, + "learning_rate": 1.3798994271011368e-06, + "loss": 1.0166, + "step": 26137 + }, + { + "epoch": 0.84, + "learning_rate": 1.3793744335865277e-06, + "loss": 0.9395, + "step": 26138 + }, + { + "epoch": 0.84, + "learning_rate": 1.3788495325616912e-06, + "loss": 0.9263, + "step": 26139 + }, + { + "epoch": 0.84, + "learning_rate": 1.3783247240322596e-06, + "loss": 1.0117, + "step": 26140 + }, + { + "epoch": 0.84, + "learning_rate": 1.3778000080038601e-06, + "loss": 0.9175, + "step": 26141 + }, + { + "epoch": 0.84, + "learning_rate": 1.3772753844821218e-06, + "loss": 0.8481, + "step": 26142 + }, + { + "epoch": 0.84, + "learning_rate": 1.3767508534726791e-06, + "loss": 0.7793, + "step": 26143 + }, + { + "epoch": 0.84, + "learning_rate": 1.376226414981152e-06, + "loss": 0.918, + "step": 26144 + }, + { + "epoch": 0.84, + "learning_rate": 1.375702069013175e-06, + "loss": 0.9429, + "step": 26145 + }, + { + "epoch": 0.84, + "learning_rate": 1.3751778155743668e-06, + "loss": 0.7722, + "step": 26146 + }, + { + "epoch": 0.84, + "learning_rate": 1.374653654670357e-06, + "loss": 0.894, + "step": 26147 + }, + { + "epoch": 0.84, + "learning_rate": 1.3741295863067683e-06, + "loss": 0.9409, + "step": 26148 + }, + { + "epoch": 0.84, + "learning_rate": 1.3736056104892215e-06, + "loss": 0.9683, + "step": 26149 + }, + { + "epoch": 0.84, + "learning_rate": 1.3730817272233377e-06, + "loss": 0.9463, + "step": 26150 + }, + { + "epoch": 0.84, + "learning_rate": 1.3725579365147412e-06, + "loss": 0.9966, + "step": 26151 + }, + { + "epoch": 0.84, + "learning_rate": 1.3720342383690477e-06, + "loss": 0.9873, + "step": 26152 + }, + { + "epoch": 0.84, + "learning_rate": 1.3715106327918802e-06, + "loss": 0.9238, + "step": 26153 + }, + { + "epoch": 0.84, + "learning_rate": 1.3709871197888514e-06, + "loss": 0.9888, + "step": 26154 + }, + { + "epoch": 0.84, + "learning_rate": 1.3704636993655851e-06, + "loss": 0.9336, + "step": 26155 + }, + { + "epoch": 0.84, + "learning_rate": 1.3699403715276893e-06, + "loss": 0.9893, + "step": 26156 + }, + { + "epoch": 0.84, + "learning_rate": 1.3694171362807828e-06, + "loss": 1.0366, + "step": 26157 + }, + { + "epoch": 0.84, + "learning_rate": 1.3688939936304767e-06, + "loss": 0.9336, + "step": 26158 + }, + { + "epoch": 0.84, + "learning_rate": 1.3683709435823888e-06, + "loss": 1.0513, + "step": 26159 + }, + { + "epoch": 0.84, + "learning_rate": 1.3678479861421246e-06, + "loss": 0.8691, + "step": 26160 + }, + { + "epoch": 0.84, + "learning_rate": 1.3673251213152993e-06, + "loss": 0.98, + "step": 26161 + }, + { + "epoch": 0.84, + "learning_rate": 1.366802349107521e-06, + "loss": 0.8916, + "step": 26162 + }, + { + "epoch": 0.84, + "learning_rate": 1.366279669524402e-06, + "loss": 0.8906, + "step": 26163 + }, + { + "epoch": 0.84, + "learning_rate": 1.3657570825715428e-06, + "loss": 0.9858, + "step": 26164 + }, + { + "epoch": 0.84, + "learning_rate": 1.3652345882545558e-06, + "loss": 0.9487, + "step": 26165 + }, + { + "epoch": 0.84, + "learning_rate": 1.3647121865790436e-06, + "loss": 0.8667, + "step": 26166 + }, + { + "epoch": 0.84, + "learning_rate": 1.3641898775506146e-06, + "loss": 0.9546, + "step": 26167 + }, + { + "epoch": 0.84, + "learning_rate": 1.36366766117487e-06, + "loss": 0.8784, + "step": 26168 + }, + { + "epoch": 0.84, + "learning_rate": 1.363145537457411e-06, + "loss": 0.4729, + "step": 26169 + }, + { + "epoch": 0.84, + "learning_rate": 1.3626235064038452e-06, + "loss": 0.8711, + "step": 26170 + }, + { + "epoch": 0.84, + "learning_rate": 1.362101568019768e-06, + "loss": 0.9888, + "step": 26171 + }, + { + "epoch": 0.84, + "learning_rate": 1.3615797223107797e-06, + "loss": 0.9092, + "step": 26172 + }, + { + "epoch": 0.84, + "learning_rate": 1.3610579692824832e-06, + "loss": 0.9268, + "step": 26173 + }, + { + "epoch": 0.84, + "learning_rate": 1.3605363089404723e-06, + "loss": 0.8916, + "step": 26174 + }, + { + "epoch": 0.84, + "learning_rate": 1.3600147412903442e-06, + "loss": 0.8735, + "step": 26175 + }, + { + "epoch": 0.84, + "learning_rate": 1.359493266337697e-06, + "loss": 0.5137, + "step": 26176 + }, + { + "epoch": 0.84, + "learning_rate": 1.3589718840881227e-06, + "loss": 0.9521, + "step": 26177 + }, + { + "epoch": 0.84, + "learning_rate": 1.3584505945472181e-06, + "loss": 0.8545, + "step": 26178 + }, + { + "epoch": 0.84, + "learning_rate": 1.3579293977205753e-06, + "loss": 0.8643, + "step": 26179 + }, + { + "epoch": 0.84, + "learning_rate": 1.3574082936137845e-06, + "loss": 0.9443, + "step": 26180 + }, + { + "epoch": 0.84, + "learning_rate": 1.3568872822324363e-06, + "loss": 0.9897, + "step": 26181 + }, + { + "epoch": 0.84, + "learning_rate": 1.3563663635821246e-06, + "loss": 0.8931, + "step": 26182 + }, + { + "epoch": 0.84, + "learning_rate": 1.3558455376684322e-06, + "loss": 0.9644, + "step": 26183 + }, + { + "epoch": 0.84, + "learning_rate": 1.3553248044969525e-06, + "loss": 0.9722, + "step": 26184 + }, + { + "epoch": 0.84, + "learning_rate": 1.354804164073268e-06, + "loss": 1.0112, + "step": 26185 + }, + { + "epoch": 0.84, + "learning_rate": 1.3542836164029717e-06, + "loss": 0.9644, + "step": 26186 + }, + { + "epoch": 0.84, + "learning_rate": 1.353763161491639e-06, + "loss": 0.9888, + "step": 26187 + }, + { + "epoch": 0.84, + "learning_rate": 1.3532427993448605e-06, + "loss": 0.4414, + "step": 26188 + }, + { + "epoch": 0.84, + "learning_rate": 1.3527225299682144e-06, + "loss": 0.9912, + "step": 26189 + }, + { + "epoch": 0.84, + "learning_rate": 1.352202353367289e-06, + "loss": 1.0132, + "step": 26190 + }, + { + "epoch": 0.84, + "learning_rate": 1.3516822695476573e-06, + "loss": 0.9414, + "step": 26191 + }, + { + "epoch": 0.84, + "learning_rate": 1.351162278514907e-06, + "loss": 0.9775, + "step": 26192 + }, + { + "epoch": 0.84, + "learning_rate": 1.3506423802746128e-06, + "loss": 0.8247, + "step": 26193 + }, + { + "epoch": 0.84, + "learning_rate": 1.3501225748323532e-06, + "loss": 0.9678, + "step": 26194 + }, + { + "epoch": 0.84, + "learning_rate": 1.349602862193703e-06, + "loss": 0.8354, + "step": 26195 + }, + { + "epoch": 0.84, + "learning_rate": 1.3490832423642419e-06, + "loss": 0.9819, + "step": 26196 + }, + { + "epoch": 0.84, + "learning_rate": 1.3485637153495424e-06, + "loss": 1.0049, + "step": 26197 + }, + { + "epoch": 0.84, + "learning_rate": 1.3480442811551807e-06, + "loss": 0.9971, + "step": 26198 + }, + { + "epoch": 0.84, + "learning_rate": 1.3475249397867263e-06, + "loss": 0.4985, + "step": 26199 + }, + { + "epoch": 0.84, + "learning_rate": 1.3470056912497553e-06, + "loss": 0.9102, + "step": 26200 + }, + { + "epoch": 0.84, + "learning_rate": 1.3464865355498357e-06, + "loss": 0.9336, + "step": 26201 + }, + { + "epoch": 0.84, + "learning_rate": 1.3459674726925397e-06, + "loss": 0.873, + "step": 26202 + }, + { + "epoch": 0.84, + "learning_rate": 1.345448502683432e-06, + "loss": 0.9097, + "step": 26203 + }, + { + "epoch": 0.84, + "learning_rate": 1.3449296255280863e-06, + "loss": 0.8896, + "step": 26204 + }, + { + "epoch": 0.84, + "learning_rate": 1.3444108412320644e-06, + "loss": 0.9526, + "step": 26205 + }, + { + "epoch": 0.84, + "learning_rate": 1.343892149800936e-06, + "loss": 0.9058, + "step": 26206 + }, + { + "epoch": 0.84, + "learning_rate": 1.3433735512402658e-06, + "loss": 1.0659, + "step": 26207 + }, + { + "epoch": 0.84, + "learning_rate": 1.3428550455556155e-06, + "loss": 0.9253, + "step": 26208 + }, + { + "epoch": 0.84, + "learning_rate": 1.342336632752549e-06, + "loss": 0.9351, + "step": 26209 + }, + { + "epoch": 0.84, + "learning_rate": 1.3418183128366292e-06, + "loss": 0.9409, + "step": 26210 + }, + { + "epoch": 0.84, + "learning_rate": 1.341300085813415e-06, + "loss": 0.9004, + "step": 26211 + }, + { + "epoch": 0.84, + "learning_rate": 1.340781951688469e-06, + "loss": 0.9556, + "step": 26212 + }, + { + "epoch": 0.84, + "learning_rate": 1.340263910467351e-06, + "loss": 0.8945, + "step": 26213 + }, + { + "epoch": 0.84, + "learning_rate": 1.339745962155613e-06, + "loss": 0.8428, + "step": 26214 + }, + { + "epoch": 0.84, + "learning_rate": 1.3392281067588197e-06, + "loss": 0.8608, + "step": 26215 + }, + { + "epoch": 0.84, + "learning_rate": 1.3387103442825234e-06, + "loss": 0.936, + "step": 26216 + }, + { + "epoch": 0.84, + "learning_rate": 1.338192674732276e-06, + "loss": 0.9136, + "step": 26217 + }, + { + "epoch": 0.84, + "learning_rate": 1.3376750981136377e-06, + "loss": 0.9111, + "step": 26218 + }, + { + "epoch": 0.84, + "learning_rate": 1.3371576144321597e-06, + "loss": 0.9761, + "step": 26219 + }, + { + "epoch": 0.84, + "learning_rate": 1.3366402236933896e-06, + "loss": 0.9077, + "step": 26220 + }, + { + "epoch": 0.84, + "learning_rate": 1.336122925902884e-06, + "loss": 0.9414, + "step": 26221 + }, + { + "epoch": 0.84, + "learning_rate": 1.3356057210661877e-06, + "loss": 0.7041, + "step": 26222 + }, + { + "epoch": 0.84, + "learning_rate": 1.3350886091888583e-06, + "loss": 0.877, + "step": 26223 + }, + { + "epoch": 0.84, + "learning_rate": 1.3345715902764334e-06, + "loss": 1.0522, + "step": 26224 + }, + { + "epoch": 0.84, + "learning_rate": 1.3340546643344676e-06, + "loss": 1.0044, + "step": 26225 + }, + { + "epoch": 0.84, + "learning_rate": 1.333537831368502e-06, + "loss": 0.8716, + "step": 26226 + }, + { + "epoch": 0.84, + "learning_rate": 1.3330210913840858e-06, + "loss": 0.832, + "step": 26227 + }, + { + "epoch": 0.84, + "learning_rate": 1.3325044443867586e-06, + "loss": 0.9253, + "step": 26228 + }, + { + "epoch": 0.84, + "learning_rate": 1.3319878903820682e-06, + "loss": 1.0796, + "step": 26229 + }, + { + "epoch": 0.84, + "learning_rate": 1.3314714293755527e-06, + "loss": 0.856, + "step": 26230 + }, + { + "epoch": 0.84, + "learning_rate": 1.3309550613727595e-06, + "loss": 0.9224, + "step": 26231 + }, + { + "epoch": 0.84, + "learning_rate": 1.3304387863792201e-06, + "loss": 0.8823, + "step": 26232 + }, + { + "epoch": 0.84, + "learning_rate": 1.3299226044004787e-06, + "loss": 0.9722, + "step": 26233 + }, + { + "epoch": 0.84, + "learning_rate": 1.3294065154420698e-06, + "loss": 0.9644, + "step": 26234 + }, + { + "epoch": 0.84, + "learning_rate": 1.3288905195095358e-06, + "loss": 0.4714, + "step": 26235 + }, + { + "epoch": 0.84, + "learning_rate": 1.3283746166084078e-06, + "loss": 0.9849, + "step": 26236 + }, + { + "epoch": 0.84, + "learning_rate": 1.3278588067442244e-06, + "loss": 0.9253, + "step": 26237 + }, + { + "epoch": 0.84, + "learning_rate": 1.327343089922518e-06, + "loss": 0.9199, + "step": 26238 + }, + { + "epoch": 0.84, + "learning_rate": 1.3268274661488233e-06, + "loss": 0.9111, + "step": 26239 + }, + { + "epoch": 0.84, + "learning_rate": 1.326311935428667e-06, + "loss": 0.9712, + "step": 26240 + }, + { + "epoch": 0.84, + "learning_rate": 1.3257964977675874e-06, + "loss": 0.9065, + "step": 26241 + }, + { + "epoch": 0.84, + "learning_rate": 1.3252811531711085e-06, + "loss": 0.8906, + "step": 26242 + }, + { + "epoch": 0.84, + "learning_rate": 1.3247659016447645e-06, + "loss": 0.8789, + "step": 26243 + }, + { + "epoch": 0.84, + "learning_rate": 1.3242507431940787e-06, + "loss": 1.0259, + "step": 26244 + }, + { + "epoch": 0.84, + "learning_rate": 1.3237356778245824e-06, + "loss": 1.0342, + "step": 26245 + }, + { + "epoch": 0.84, + "learning_rate": 1.3232207055418012e-06, + "loss": 0.7554, + "step": 26246 + }, + { + "epoch": 0.84, + "learning_rate": 1.3227058263512572e-06, + "loss": 1.0464, + "step": 26247 + }, + { + "epoch": 0.84, + "learning_rate": 1.3221910402584748e-06, + "loss": 0.9292, + "step": 26248 + }, + { + "epoch": 0.84, + "learning_rate": 1.3216763472689809e-06, + "loss": 0.9541, + "step": 26249 + }, + { + "epoch": 0.84, + "learning_rate": 1.321161747388292e-06, + "loss": 0.8608, + "step": 26250 + }, + { + "epoch": 0.84, + "learning_rate": 1.3206472406219362e-06, + "loss": 0.9712, + "step": 26251 + }, + { + "epoch": 0.84, + "learning_rate": 1.3201328269754276e-06, + "loss": 0.9741, + "step": 26252 + }, + { + "epoch": 0.84, + "learning_rate": 1.3196185064542867e-06, + "loss": 0.9829, + "step": 26253 + }, + { + "epoch": 0.84, + "learning_rate": 1.319104279064034e-06, + "loss": 0.9165, + "step": 26254 + }, + { + "epoch": 0.84, + "learning_rate": 1.3185901448101845e-06, + "loss": 0.9712, + "step": 26255 + }, + { + "epoch": 0.84, + "learning_rate": 1.3180761036982526e-06, + "loss": 0.8696, + "step": 26256 + }, + { + "epoch": 0.84, + "learning_rate": 1.317562155733757e-06, + "loss": 0.9668, + "step": 26257 + }, + { + "epoch": 0.84, + "learning_rate": 1.3170483009222113e-06, + "loss": 0.916, + "step": 26258 + }, + { + "epoch": 0.84, + "learning_rate": 1.3165345392691253e-06, + "loss": 0.8945, + "step": 26259 + }, + { + "epoch": 0.84, + "learning_rate": 1.316020870780015e-06, + "loss": 0.968, + "step": 26260 + }, + { + "epoch": 0.84, + "learning_rate": 1.3155072954603898e-06, + "loss": 0.5098, + "step": 26261 + }, + { + "epoch": 0.84, + "learning_rate": 1.314993813315758e-06, + "loss": 0.9082, + "step": 26262 + }, + { + "epoch": 0.84, + "learning_rate": 1.3144804243516318e-06, + "loss": 0.8164, + "step": 26263 + }, + { + "epoch": 0.84, + "learning_rate": 1.313967128573519e-06, + "loss": 0.8389, + "step": 26264 + }, + { + "epoch": 0.84, + "learning_rate": 1.3134539259869228e-06, + "loss": 0.9346, + "step": 26265 + }, + { + "epoch": 0.84, + "learning_rate": 1.3129408165973545e-06, + "loss": 1.0059, + "step": 26266 + }, + { + "epoch": 0.84, + "learning_rate": 1.3124278004103152e-06, + "loss": 0.8435, + "step": 26267 + }, + { + "epoch": 0.84, + "learning_rate": 1.311914877431313e-06, + "loss": 0.9067, + "step": 26268 + }, + { + "epoch": 0.84, + "learning_rate": 1.3114020476658473e-06, + "loss": 0.8057, + "step": 26269 + }, + { + "epoch": 0.84, + "learning_rate": 1.310889311119422e-06, + "loss": 1.0254, + "step": 26270 + }, + { + "epoch": 0.84, + "learning_rate": 1.3103766677975372e-06, + "loss": 0.9961, + "step": 26271 + }, + { + "epoch": 0.84, + "learning_rate": 1.3098641177056938e-06, + "loss": 0.7651, + "step": 26272 + }, + { + "epoch": 0.84, + "learning_rate": 1.3093516608493895e-06, + "loss": 0.9282, + "step": 26273 + }, + { + "epoch": 0.84, + "learning_rate": 1.3088392972341256e-06, + "loss": 0.9385, + "step": 26274 + }, + { + "epoch": 0.84, + "learning_rate": 1.3083270268653947e-06, + "loss": 0.9126, + "step": 26275 + }, + { + "epoch": 0.84, + "learning_rate": 1.3078148497486998e-06, + "loss": 1.0293, + "step": 26276 + }, + { + "epoch": 0.84, + "learning_rate": 1.3073027658895266e-06, + "loss": 0.875, + "step": 26277 + }, + { + "epoch": 0.84, + "learning_rate": 1.3067907752933773e-06, + "loss": 0.7529, + "step": 26278 + }, + { + "epoch": 0.84, + "learning_rate": 1.3062788779657387e-06, + "loss": 0.8501, + "step": 26279 + }, + { + "epoch": 0.84, + "learning_rate": 1.3057670739121087e-06, + "loss": 0.9663, + "step": 26280 + }, + { + "epoch": 0.84, + "learning_rate": 1.3052553631379738e-06, + "loss": 0.9331, + "step": 26281 + }, + { + "epoch": 0.84, + "learning_rate": 1.3047437456488276e-06, + "loss": 0.9297, + "step": 26282 + }, + { + "epoch": 0.84, + "learning_rate": 1.3042322214501557e-06, + "loss": 0.8853, + "step": 26283 + }, + { + "epoch": 0.84, + "learning_rate": 1.3037207905474526e-06, + "loss": 0.9419, + "step": 26284 + }, + { + "epoch": 0.84, + "learning_rate": 1.3032094529461968e-06, + "loss": 0.9727, + "step": 26285 + }, + { + "epoch": 0.84, + "learning_rate": 1.3026982086518803e-06, + "loss": 0.9204, + "step": 26286 + }, + { + "epoch": 0.84, + "learning_rate": 1.3021870576699835e-06, + "loss": 0.8794, + "step": 26287 + }, + { + "epoch": 0.84, + "learning_rate": 1.3016760000059959e-06, + "loss": 0.9834, + "step": 26288 + }, + { + "epoch": 0.84, + "learning_rate": 1.3011650356653959e-06, + "loss": 0.9556, + "step": 26289 + }, + { + "epoch": 0.84, + "learning_rate": 1.3006541646536696e-06, + "loss": 0.791, + "step": 26290 + }, + { + "epoch": 0.84, + "learning_rate": 1.3001433869762968e-06, + "loss": 0.9741, + "step": 26291 + }, + { + "epoch": 0.84, + "learning_rate": 1.2996327026387567e-06, + "loss": 0.8831, + "step": 26292 + }, + { + "epoch": 0.84, + "learning_rate": 1.2991221116465259e-06, + "loss": 0.8413, + "step": 26293 + }, + { + "epoch": 0.84, + "learning_rate": 1.2986116140050874e-06, + "loss": 1.0771, + "step": 26294 + }, + { + "epoch": 0.84, + "learning_rate": 1.2981012097199142e-06, + "loss": 0.9634, + "step": 26295 + }, + { + "epoch": 0.84, + "learning_rate": 1.297590898796487e-06, + "loss": 0.9609, + "step": 26296 + }, + { + "epoch": 0.84, + "learning_rate": 1.297080681240277e-06, + "loss": 1.0, + "step": 26297 + }, + { + "epoch": 0.84, + "learning_rate": 1.2965705570567578e-06, + "loss": 0.9653, + "step": 26298 + }, + { + "epoch": 0.84, + "learning_rate": 1.2960605262514059e-06, + "loss": 0.895, + "step": 26299 + }, + { + "epoch": 0.84, + "learning_rate": 1.2955505888296915e-06, + "loss": 0.8877, + "step": 26300 + }, + { + "epoch": 0.84, + "learning_rate": 1.2950407447970837e-06, + "loss": 0.9282, + "step": 26301 + }, + { + "epoch": 0.84, + "learning_rate": 1.2945309941590567e-06, + "loss": 0.9346, + "step": 26302 + }, + { + "epoch": 0.84, + "learning_rate": 1.2940213369210775e-06, + "loss": 0.9883, + "step": 26303 + }, + { + "epoch": 0.84, + "learning_rate": 1.2935117730886126e-06, + "loss": 0.9072, + "step": 26304 + }, + { + "epoch": 0.84, + "learning_rate": 1.2930023026671312e-06, + "loss": 0.9268, + "step": 26305 + }, + { + "epoch": 0.84, + "learning_rate": 1.2924929256621e-06, + "loss": 0.9326, + "step": 26306 + }, + { + "epoch": 0.84, + "learning_rate": 1.2919836420789823e-06, + "loss": 0.8867, + "step": 26307 + }, + { + "epoch": 0.84, + "learning_rate": 1.2914744519232392e-06, + "loss": 0.8257, + "step": 26308 + }, + { + "epoch": 0.84, + "learning_rate": 1.2909653552003409e-06, + "loss": 0.9189, + "step": 26309 + }, + { + "epoch": 0.84, + "learning_rate": 1.290456351915742e-06, + "loss": 0.8525, + "step": 26310 + }, + { + "epoch": 0.84, + "learning_rate": 1.2899474420749104e-06, + "loss": 0.9292, + "step": 26311 + }, + { + "epoch": 0.84, + "learning_rate": 1.2894386256833014e-06, + "loss": 0.4646, + "step": 26312 + }, + { + "epoch": 0.84, + "learning_rate": 1.2889299027463763e-06, + "loss": 0.8403, + "step": 26313 + }, + { + "epoch": 0.84, + "learning_rate": 1.2884212732695932e-06, + "loss": 0.4753, + "step": 26314 + }, + { + "epoch": 0.84, + "learning_rate": 1.2879127372584076e-06, + "loss": 0.8921, + "step": 26315 + }, + { + "epoch": 0.84, + "learning_rate": 1.2874042947182751e-06, + "loss": 1.0215, + "step": 26316 + }, + { + "epoch": 0.84, + "learning_rate": 1.2868959456546537e-06, + "loss": 0.8511, + "step": 26317 + }, + { + "epoch": 0.84, + "learning_rate": 1.2863876900729932e-06, + "loss": 1.0024, + "step": 26318 + }, + { + "epoch": 0.84, + "learning_rate": 1.2858795279787517e-06, + "loss": 0.9253, + "step": 26319 + }, + { + "epoch": 0.84, + "learning_rate": 1.2853714593773758e-06, + "loss": 0.8496, + "step": 26320 + }, + { + "epoch": 0.84, + "learning_rate": 1.2848634842743246e-06, + "loss": 0.9683, + "step": 26321 + }, + { + "epoch": 0.84, + "learning_rate": 1.284355602675037e-06, + "loss": 0.9482, + "step": 26322 + }, + { + "epoch": 0.84, + "learning_rate": 1.2838478145849708e-06, + "loss": 0.8872, + "step": 26323 + }, + { + "epoch": 0.84, + "learning_rate": 1.2833401200095686e-06, + "loss": 1.0181, + "step": 26324 + }, + { + "epoch": 0.84, + "learning_rate": 1.2828325189542824e-06, + "loss": 0.8542, + "step": 26325 + }, + { + "epoch": 0.84, + "learning_rate": 1.2823250114245534e-06, + "loss": 0.8491, + "step": 26326 + }, + { + "epoch": 0.84, + "learning_rate": 1.2818175974258307e-06, + "loss": 0.9819, + "step": 26327 + }, + { + "epoch": 0.84, + "learning_rate": 1.2813102769635533e-06, + "loss": 0.9282, + "step": 26328 + }, + { + "epoch": 0.84, + "learning_rate": 1.2808030500431723e-06, + "loss": 0.915, + "step": 26329 + }, + { + "epoch": 0.84, + "learning_rate": 1.2802959166701213e-06, + "loss": 0.9421, + "step": 26330 + }, + { + "epoch": 0.84, + "learning_rate": 1.2797888768498456e-06, + "loss": 0.855, + "step": 26331 + }, + { + "epoch": 0.84, + "learning_rate": 1.2792819305877823e-06, + "loss": 0.9438, + "step": 26332 + }, + { + "epoch": 0.84, + "learning_rate": 1.278775077889375e-06, + "loss": 0.9697, + "step": 26333 + }, + { + "epoch": 0.84, + "learning_rate": 1.2782683187600553e-06, + "loss": 0.8828, + "step": 26334 + }, + { + "epoch": 0.84, + "learning_rate": 1.2777616532052672e-06, + "loss": 0.887, + "step": 26335 + }, + { + "epoch": 0.84, + "learning_rate": 1.277255081230443e-06, + "loss": 0.835, + "step": 26336 + }, + { + "epoch": 0.84, + "learning_rate": 1.276748602841018e-06, + "loss": 0.834, + "step": 26337 + }, + { + "epoch": 0.84, + "learning_rate": 1.2762422180424239e-06, + "loss": 0.8198, + "step": 26338 + }, + { + "epoch": 0.84, + "learning_rate": 1.275735926840098e-06, + "loss": 0.9839, + "step": 26339 + }, + { + "epoch": 0.84, + "learning_rate": 1.2752297292394677e-06, + "loss": 0.7603, + "step": 26340 + }, + { + "epoch": 0.84, + "learning_rate": 1.2747236252459683e-06, + "loss": 0.8613, + "step": 26341 + }, + { + "epoch": 0.84, + "learning_rate": 1.2742176148650288e-06, + "loss": 0.856, + "step": 26342 + }, + { + "epoch": 0.84, + "learning_rate": 1.2737116981020737e-06, + "loss": 0.4895, + "step": 26343 + }, + { + "epoch": 0.84, + "learning_rate": 1.2732058749625376e-06, + "loss": 0.959, + "step": 26344 + }, + { + "epoch": 0.84, + "learning_rate": 1.2727001454518428e-06, + "loss": 1.0137, + "step": 26345 + }, + { + "epoch": 0.84, + "learning_rate": 1.2721945095754174e-06, + "loss": 0.7959, + "step": 26346 + }, + { + "epoch": 0.84, + "learning_rate": 1.2716889673386835e-06, + "loss": 0.9434, + "step": 26347 + }, + { + "epoch": 0.84, + "learning_rate": 1.2711835187470677e-06, + "loss": 0.9316, + "step": 26348 + }, + { + "epoch": 0.84, + "learning_rate": 1.2706781638059917e-06, + "loss": 1.0073, + "step": 26349 + }, + { + "epoch": 0.84, + "learning_rate": 1.2701729025208798e-06, + "loss": 0.98, + "step": 26350 + }, + { + "epoch": 0.84, + "learning_rate": 1.2696677348971476e-06, + "loss": 0.834, + "step": 26351 + }, + { + "epoch": 0.84, + "learning_rate": 1.269162660940224e-06, + "loss": 0.9321, + "step": 26352 + }, + { + "epoch": 0.84, + "learning_rate": 1.2686576806555173e-06, + "loss": 0.9043, + "step": 26353 + }, + { + "epoch": 0.84, + "learning_rate": 1.2681527940484516e-06, + "loss": 0.9204, + "step": 26354 + }, + { + "epoch": 0.84, + "learning_rate": 1.267648001124442e-06, + "loss": 0.8716, + "step": 26355 + }, + { + "epoch": 0.84, + "learning_rate": 1.2671433018889058e-06, + "loss": 0.8237, + "step": 26356 + }, + { + "epoch": 0.84, + "learning_rate": 1.2666386963472554e-06, + "loss": 0.937, + "step": 26357 + }, + { + "epoch": 0.84, + "learning_rate": 1.266134184504908e-06, + "loss": 0.8496, + "step": 26358 + }, + { + "epoch": 0.84, + "learning_rate": 1.2656297663672756e-06, + "loss": 0.9644, + "step": 26359 + }, + { + "epoch": 0.84, + "learning_rate": 1.2651254419397674e-06, + "loss": 0.916, + "step": 26360 + }, + { + "epoch": 0.84, + "learning_rate": 1.2646212112277956e-06, + "loss": 0.8682, + "step": 26361 + }, + { + "epoch": 0.84, + "learning_rate": 1.2641170742367715e-06, + "loss": 0.9409, + "step": 26362 + }, + { + "epoch": 0.84, + "learning_rate": 1.2636130309721006e-06, + "loss": 0.8613, + "step": 26363 + }, + { + "epoch": 0.84, + "learning_rate": 1.2631090814391945e-06, + "loss": 0.9287, + "step": 26364 + }, + { + "epoch": 0.84, + "learning_rate": 1.2626052256434573e-06, + "loss": 0.8823, + "step": 26365 + }, + { + "epoch": 0.84, + "learning_rate": 1.2621014635902984e-06, + "loss": 0.8867, + "step": 26366 + }, + { + "epoch": 0.84, + "learning_rate": 1.2615977952851188e-06, + "loss": 0.9189, + "step": 26367 + }, + { + "epoch": 0.84, + "learning_rate": 1.2610942207333244e-06, + "loss": 0.8989, + "step": 26368 + }, + { + "epoch": 0.84, + "learning_rate": 1.260590739940315e-06, + "loss": 0.8311, + "step": 26369 + }, + { + "epoch": 0.84, + "learning_rate": 1.2600873529114977e-06, + "loss": 0.9365, + "step": 26370 + }, + { + "epoch": 0.84, + "learning_rate": 1.2595840596522669e-06, + "loss": 0.8833, + "step": 26371 + }, + { + "epoch": 0.84, + "learning_rate": 1.2590808601680282e-06, + "loss": 0.9707, + "step": 26372 + }, + { + "epoch": 0.84, + "learning_rate": 1.2585777544641765e-06, + "loss": 0.9731, + "step": 26373 + }, + { + "epoch": 0.84, + "learning_rate": 1.258074742546115e-06, + "loss": 0.9233, + "step": 26374 + }, + { + "epoch": 0.84, + "learning_rate": 1.2575718244192314e-06, + "loss": 0.9756, + "step": 26375 + }, + { + "epoch": 0.84, + "learning_rate": 1.2570690000889285e-06, + "loss": 0.958, + "step": 26376 + }, + { + "epoch": 0.84, + "learning_rate": 1.2565662695605974e-06, + "loss": 0.895, + "step": 26377 + }, + { + "epoch": 0.84, + "learning_rate": 1.2560636328396357e-06, + "loss": 1.1006, + "step": 26378 + }, + { + "epoch": 0.84, + "learning_rate": 1.2555610899314318e-06, + "loss": 1.0161, + "step": 26379 + }, + { + "epoch": 0.84, + "learning_rate": 1.2550586408413823e-06, + "loss": 0.9414, + "step": 26380 + }, + { + "epoch": 0.84, + "learning_rate": 1.2545562855748738e-06, + "loss": 0.8618, + "step": 26381 + }, + { + "epoch": 0.84, + "learning_rate": 1.2540540241372978e-06, + "loss": 0.8418, + "step": 26382 + }, + { + "epoch": 0.84, + "learning_rate": 1.253551856534041e-06, + "loss": 0.7964, + "step": 26383 + }, + { + "epoch": 0.84, + "learning_rate": 1.2530497827704947e-06, + "loss": 0.8711, + "step": 26384 + }, + { + "epoch": 0.84, + "learning_rate": 1.2525478028520432e-06, + "loss": 1.0181, + "step": 26385 + }, + { + "epoch": 0.84, + "learning_rate": 1.2520459167840704e-06, + "loss": 0.9248, + "step": 26386 + }, + { + "epoch": 0.84, + "learning_rate": 1.2515441245719662e-06, + "loss": 0.8701, + "step": 26387 + }, + { + "epoch": 0.84, + "learning_rate": 1.2510424262211086e-06, + "loss": 0.8809, + "step": 26388 + }, + { + "epoch": 0.84, + "learning_rate": 1.2505408217368842e-06, + "loss": 0.9497, + "step": 26389 + }, + { + "epoch": 0.84, + "learning_rate": 1.2500393111246744e-06, + "loss": 0.7935, + "step": 26390 + }, + { + "epoch": 0.84, + "learning_rate": 1.2495378943898584e-06, + "loss": 0.936, + "step": 26391 + }, + { + "epoch": 0.84, + "learning_rate": 1.2490365715378139e-06, + "loss": 0.8198, + "step": 26392 + }, + { + "epoch": 0.84, + "learning_rate": 1.2485353425739244e-06, + "loss": 0.9419, + "step": 26393 + }, + { + "epoch": 0.84, + "learning_rate": 1.2480342075035622e-06, + "loss": 0.9033, + "step": 26394 + }, + { + "epoch": 0.84, + "learning_rate": 1.24753316633211e-06, + "loss": 1.002, + "step": 26395 + }, + { + "epoch": 0.84, + "learning_rate": 1.2470322190649375e-06, + "loss": 0.8906, + "step": 26396 + }, + { + "epoch": 0.84, + "learning_rate": 1.2465313657074263e-06, + "loss": 0.8511, + "step": 26397 + }, + { + "epoch": 0.84, + "learning_rate": 1.2460306062649418e-06, + "loss": 0.811, + "step": 26398 + }, + { + "epoch": 0.84, + "learning_rate": 1.2455299407428622e-06, + "loss": 0.8408, + "step": 26399 + }, + { + "epoch": 0.84, + "learning_rate": 1.2450293691465554e-06, + "loss": 0.9077, + "step": 26400 + }, + { + "epoch": 0.84, + "learning_rate": 1.2445288914813957e-06, + "loss": 0.9575, + "step": 26401 + }, + { + "epoch": 0.84, + "learning_rate": 1.2440285077527491e-06, + "loss": 0.8364, + "step": 26402 + }, + { + "epoch": 0.84, + "learning_rate": 1.243528217965988e-06, + "loss": 0.8374, + "step": 26403 + }, + { + "epoch": 0.84, + "learning_rate": 1.243028022126478e-06, + "loss": 0.9204, + "step": 26404 + }, + { + "epoch": 0.84, + "learning_rate": 1.2425279202395846e-06, + "loss": 0.9077, + "step": 26405 + }, + { + "epoch": 0.84, + "learning_rate": 1.2420279123106739e-06, + "loss": 0.8052, + "step": 26406 + }, + { + "epoch": 0.84, + "learning_rate": 1.2415279983451124e-06, + "loss": 0.8633, + "step": 26407 + }, + { + "epoch": 0.84, + "learning_rate": 1.241028178348259e-06, + "loss": 0.8599, + "step": 26408 + }, + { + "epoch": 0.84, + "learning_rate": 1.2405284523254823e-06, + "loss": 0.9204, + "step": 26409 + }, + { + "epoch": 0.84, + "learning_rate": 1.2400288202821388e-06, + "loss": 0.9004, + "step": 26410 + }, + { + "epoch": 0.84, + "learning_rate": 1.2395292822235926e-06, + "loss": 1.0127, + "step": 26411 + }, + { + "epoch": 0.84, + "learning_rate": 1.2390298381552024e-06, + "loss": 0.8921, + "step": 26412 + }, + { + "epoch": 0.84, + "learning_rate": 1.2385304880823256e-06, + "loss": 0.918, + "step": 26413 + }, + { + "epoch": 0.84, + "learning_rate": 1.2380312320103171e-06, + "loss": 0.9414, + "step": 26414 + }, + { + "epoch": 0.84, + "learning_rate": 1.2375320699445391e-06, + "loss": 0.9019, + "step": 26415 + }, + { + "epoch": 0.84, + "learning_rate": 1.237033001890342e-06, + "loss": 0.8921, + "step": 26416 + }, + { + "epoch": 0.84, + "learning_rate": 1.2365340278530835e-06, + "loss": 0.9204, + "step": 26417 + }, + { + "epoch": 0.84, + "learning_rate": 1.236035147838115e-06, + "loss": 0.9546, + "step": 26418 + }, + { + "epoch": 0.84, + "learning_rate": 1.235536361850792e-06, + "loss": 0.9697, + "step": 26419 + }, + { + "epoch": 0.84, + "learning_rate": 1.2350376698964639e-06, + "loss": 1.0078, + "step": 26420 + }, + { + "epoch": 0.84, + "learning_rate": 1.2345390719804806e-06, + "loss": 0.9229, + "step": 26421 + }, + { + "epoch": 0.84, + "learning_rate": 1.2340405681081891e-06, + "loss": 0.9663, + "step": 26422 + }, + { + "epoch": 0.85, + "learning_rate": 1.2335421582849437e-06, + "loss": 0.7715, + "step": 26423 + }, + { + "epoch": 0.85, + "learning_rate": 1.2330438425160884e-06, + "loss": 0.8887, + "step": 26424 + }, + { + "epoch": 0.85, + "learning_rate": 1.2325456208069674e-06, + "loss": 0.9478, + "step": 26425 + }, + { + "epoch": 0.85, + "learning_rate": 1.232047493162931e-06, + "loss": 0.8394, + "step": 26426 + }, + { + "epoch": 0.85, + "learning_rate": 1.2315494595893218e-06, + "loss": 0.8379, + "step": 26427 + }, + { + "epoch": 0.85, + "learning_rate": 1.2310515200914796e-06, + "loss": 1.0625, + "step": 26428 + }, + { + "epoch": 0.85, + "learning_rate": 1.2305536746747515e-06, + "loss": 0.8613, + "step": 26429 + }, + { + "epoch": 0.85, + "learning_rate": 1.2300559233444775e-06, + "loss": 1.0298, + "step": 26430 + }, + { + "epoch": 0.85, + "learning_rate": 1.2295582661059957e-06, + "loss": 0.9438, + "step": 26431 + }, + { + "epoch": 0.85, + "learning_rate": 1.2290607029646483e-06, + "loss": 0.8259, + "step": 26432 + }, + { + "epoch": 0.85, + "learning_rate": 1.2285632339257703e-06, + "loss": 0.8853, + "step": 26433 + }, + { + "epoch": 0.85, + "learning_rate": 1.2280658589947036e-06, + "loss": 0.894, + "step": 26434 + }, + { + "epoch": 0.85, + "learning_rate": 1.227568578176781e-06, + "loss": 0.854, + "step": 26435 + }, + { + "epoch": 0.85, + "learning_rate": 1.2270713914773403e-06, + "loss": 0.897, + "step": 26436 + }, + { + "epoch": 0.85, + "learning_rate": 1.2265742989017104e-06, + "loss": 0.9316, + "step": 26437 + }, + { + "epoch": 0.85, + "learning_rate": 1.2260773004552317e-06, + "loss": 0.8345, + "step": 26438 + }, + { + "epoch": 0.85, + "learning_rate": 1.2255803961432299e-06, + "loss": 1.0713, + "step": 26439 + }, + { + "epoch": 0.85, + "learning_rate": 1.225083585971042e-06, + "loss": 0.8687, + "step": 26440 + }, + { + "epoch": 0.85, + "learning_rate": 1.2245868699439932e-06, + "loss": 0.9331, + "step": 26441 + }, + { + "epoch": 0.85, + "learning_rate": 1.2240902480674198e-06, + "loss": 1.0347, + "step": 26442 + }, + { + "epoch": 0.85, + "learning_rate": 1.223593720346642e-06, + "loss": 0.8535, + "step": 26443 + }, + { + "epoch": 0.85, + "learning_rate": 1.2230972867869905e-06, + "loss": 0.9175, + "step": 26444 + }, + { + "epoch": 0.85, + "learning_rate": 1.2226009473937906e-06, + "loss": 0.9297, + "step": 26445 + }, + { + "epoch": 0.85, + "learning_rate": 1.2221047021723698e-06, + "loss": 0.937, + "step": 26446 + }, + { + "epoch": 0.85, + "learning_rate": 1.2216085511280484e-06, + "loss": 0.9429, + "step": 26447 + }, + { + "epoch": 0.85, + "learning_rate": 1.2211124942661545e-06, + "loss": 0.8877, + "step": 26448 + }, + { + "epoch": 0.85, + "learning_rate": 1.2206165315920048e-06, + "loss": 0.9277, + "step": 26449 + }, + { + "epoch": 0.85, + "learning_rate": 1.2201206631109275e-06, + "loss": 0.9976, + "step": 26450 + }, + { + "epoch": 0.85, + "learning_rate": 1.2196248888282348e-06, + "loss": 0.8633, + "step": 26451 + }, + { + "epoch": 0.85, + "learning_rate": 1.2191292087492512e-06, + "loss": 0.8848, + "step": 26452 + }, + { + "epoch": 0.85, + "learning_rate": 1.2186336228792906e-06, + "loss": 0.8906, + "step": 26453 + }, + { + "epoch": 0.85, + "learning_rate": 1.2181381312236751e-06, + "loss": 0.7915, + "step": 26454 + }, + { + "epoch": 0.85, + "learning_rate": 1.2176427337877151e-06, + "loss": 0.8574, + "step": 26455 + }, + { + "epoch": 0.85, + "learning_rate": 1.2171474305767317e-06, + "loss": 0.896, + "step": 26456 + }, + { + "epoch": 0.85, + "learning_rate": 1.2166522215960352e-06, + "loss": 0.8271, + "step": 26457 + }, + { + "epoch": 0.85, + "learning_rate": 1.2161571068509402e-06, + "loss": 0.9922, + "step": 26458 + }, + { + "epoch": 0.85, + "learning_rate": 1.215662086346756e-06, + "loss": 1.0171, + "step": 26459 + }, + { + "epoch": 0.85, + "learning_rate": 1.2151671600887971e-06, + "loss": 0.8853, + "step": 26460 + }, + { + "epoch": 0.85, + "learning_rate": 1.2146723280823703e-06, + "loss": 1.0957, + "step": 26461 + }, + { + "epoch": 0.85, + "learning_rate": 1.214177590332788e-06, + "loss": 0.8086, + "step": 26462 + }, + { + "epoch": 0.85, + "learning_rate": 1.2136829468453549e-06, + "loss": 0.7944, + "step": 26463 + }, + { + "epoch": 0.85, + "learning_rate": 1.2131883976253812e-06, + "loss": 0.9214, + "step": 26464 + }, + { + "epoch": 0.85, + "learning_rate": 1.2126939426781715e-06, + "loss": 0.9492, + "step": 26465 + }, + { + "epoch": 0.85, + "learning_rate": 1.2121995820090304e-06, + "loss": 0.9443, + "step": 26466 + }, + { + "epoch": 0.85, + "learning_rate": 1.2117053156232605e-06, + "loss": 0.7671, + "step": 26467 + }, + { + "epoch": 0.85, + "learning_rate": 1.2112111435261676e-06, + "loss": 0.9717, + "step": 26468 + }, + { + "epoch": 0.85, + "learning_rate": 1.2107170657230515e-06, + "loss": 0.9087, + "step": 26469 + }, + { + "epoch": 0.85, + "learning_rate": 1.210223082219213e-06, + "loss": 0.9756, + "step": 26470 + }, + { + "epoch": 0.85, + "learning_rate": 1.209729193019954e-06, + "loss": 0.8682, + "step": 26471 + }, + { + "epoch": 0.85, + "learning_rate": 1.2092353981305728e-06, + "loss": 0.8843, + "step": 26472 + }, + { + "epoch": 0.85, + "learning_rate": 1.2087416975563638e-06, + "loss": 0.8137, + "step": 26473 + }, + { + "epoch": 0.85, + "learning_rate": 1.2082480913026296e-06, + "loss": 0.9258, + "step": 26474 + }, + { + "epoch": 0.85, + "learning_rate": 1.2077545793746626e-06, + "loss": 0.9902, + "step": 26475 + }, + { + "epoch": 0.85, + "learning_rate": 1.2072611617777563e-06, + "loss": 0.8262, + "step": 26476 + }, + { + "epoch": 0.85, + "learning_rate": 1.2067678385172077e-06, + "loss": 0.8687, + "step": 26477 + }, + { + "epoch": 0.85, + "learning_rate": 1.206274609598307e-06, + "loss": 0.8867, + "step": 26478 + }, + { + "epoch": 0.85, + "learning_rate": 1.2057814750263497e-06, + "loss": 0.9204, + "step": 26479 + }, + { + "epoch": 0.85, + "learning_rate": 1.205288434806624e-06, + "loss": 0.9194, + "step": 26480 + }, + { + "epoch": 0.85, + "learning_rate": 1.2047954889444191e-06, + "loss": 0.4971, + "step": 26481 + }, + { + "epoch": 0.85, + "learning_rate": 1.204302637445024e-06, + "loss": 0.8608, + "step": 26482 + }, + { + "epoch": 0.85, + "learning_rate": 1.2038098803137277e-06, + "loss": 0.8359, + "step": 26483 + }, + { + "epoch": 0.85, + "learning_rate": 1.203317217555815e-06, + "loss": 0.9014, + "step": 26484 + }, + { + "epoch": 0.85, + "learning_rate": 1.202824649176575e-06, + "loss": 0.8242, + "step": 26485 + }, + { + "epoch": 0.85, + "learning_rate": 1.2023321751812878e-06, + "loss": 1.1162, + "step": 26486 + }, + { + "epoch": 0.85, + "learning_rate": 1.2018397955752414e-06, + "loss": 0.9634, + "step": 26487 + }, + { + "epoch": 0.85, + "learning_rate": 1.2013475103637172e-06, + "loss": 0.978, + "step": 26488 + }, + { + "epoch": 0.85, + "learning_rate": 1.2008553195519957e-06, + "loss": 1.0044, + "step": 26489 + }, + { + "epoch": 0.85, + "learning_rate": 1.2003632231453567e-06, + "loss": 0.8975, + "step": 26490 + }, + { + "epoch": 0.85, + "learning_rate": 1.1998712211490838e-06, + "loss": 0.9248, + "step": 26491 + }, + { + "epoch": 0.85, + "learning_rate": 1.1993793135684495e-06, + "loss": 0.9644, + "step": 26492 + }, + { + "epoch": 0.85, + "learning_rate": 1.1988875004087385e-06, + "loss": 1.0181, + "step": 26493 + }, + { + "epoch": 0.85, + "learning_rate": 1.1983957816752224e-06, + "loss": 0.937, + "step": 26494 + }, + { + "epoch": 0.85, + "learning_rate": 1.197904157373182e-06, + "loss": 0.8555, + "step": 26495 + }, + { + "epoch": 0.85, + "learning_rate": 1.1974126275078834e-06, + "loss": 0.4409, + "step": 26496 + }, + { + "epoch": 0.85, + "learning_rate": 1.196921192084608e-06, + "loss": 0.9448, + "step": 26497 + }, + { + "epoch": 0.85, + "learning_rate": 1.1964298511086236e-06, + "loss": 0.8105, + "step": 26498 + }, + { + "epoch": 0.85, + "learning_rate": 1.195938604585205e-06, + "loss": 0.8711, + "step": 26499 + }, + { + "epoch": 0.85, + "learning_rate": 1.1954474525196202e-06, + "loss": 0.8921, + "step": 26500 + }, + { + "epoch": 0.85, + "learning_rate": 1.1949563949171406e-06, + "loss": 0.894, + "step": 26501 + }, + { + "epoch": 0.85, + "learning_rate": 1.1944654317830329e-06, + "loss": 0.9248, + "step": 26502 + }, + { + "epoch": 0.85, + "learning_rate": 1.1939745631225696e-06, + "loss": 0.877, + "step": 26503 + }, + { + "epoch": 0.85, + "learning_rate": 1.1934837889410089e-06, + "loss": 0.9082, + "step": 26504 + }, + { + "epoch": 0.85, + "learning_rate": 1.192993109243622e-06, + "loss": 0.8535, + "step": 26505 + }, + { + "epoch": 0.85, + "learning_rate": 1.1925025240356714e-06, + "loss": 0.8779, + "step": 26506 + }, + { + "epoch": 0.85, + "learning_rate": 1.1920120333224217e-06, + "loss": 0.9546, + "step": 26507 + }, + { + "epoch": 0.85, + "learning_rate": 1.1915216371091353e-06, + "loss": 0.981, + "step": 26508 + }, + { + "epoch": 0.85, + "learning_rate": 1.1910313354010705e-06, + "loss": 0.9741, + "step": 26509 + }, + { + "epoch": 0.85, + "learning_rate": 1.1905411282034929e-06, + "loss": 0.9336, + "step": 26510 + }, + { + "epoch": 0.85, + "learning_rate": 1.1900510155216583e-06, + "loss": 0.8188, + "step": 26511 + }, + { + "epoch": 0.85, + "learning_rate": 1.1895609973608247e-06, + "loss": 0.9141, + "step": 26512 + }, + { + "epoch": 0.85, + "learning_rate": 1.1890710737262524e-06, + "loss": 0.9424, + "step": 26513 + }, + { + "epoch": 0.85, + "learning_rate": 1.1885812446231959e-06, + "loss": 0.9126, + "step": 26514 + }, + { + "epoch": 0.85, + "learning_rate": 1.188091510056909e-06, + "loss": 0.9395, + "step": 26515 + }, + { + "epoch": 0.85, + "learning_rate": 1.1876018700326498e-06, + "loss": 0.6646, + "step": 26516 + }, + { + "epoch": 0.85, + "learning_rate": 1.187112324555667e-06, + "loss": 0.9814, + "step": 26517 + }, + { + "epoch": 0.85, + "learning_rate": 1.186622873631218e-06, + "loss": 0.8301, + "step": 26518 + }, + { + "epoch": 0.85, + "learning_rate": 1.1861335172645505e-06, + "loss": 0.9854, + "step": 26519 + }, + { + "epoch": 0.85, + "learning_rate": 1.185644255460916e-06, + "loss": 0.8997, + "step": 26520 + }, + { + "epoch": 0.85, + "learning_rate": 1.1851550882255613e-06, + "loss": 0.9668, + "step": 26521 + }, + { + "epoch": 0.85, + "learning_rate": 1.184666015563739e-06, + "loss": 0.8823, + "step": 26522 + }, + { + "epoch": 0.85, + "learning_rate": 1.1841770374806915e-06, + "loss": 1.0225, + "step": 26523 + }, + { + "epoch": 0.85, + "learning_rate": 1.1836881539816702e-06, + "loss": 0.9961, + "step": 26524 + }, + { + "epoch": 0.85, + "learning_rate": 1.1831993650719176e-06, + "loss": 0.8955, + "step": 26525 + }, + { + "epoch": 0.85, + "learning_rate": 1.182710670756677e-06, + "loss": 0.9224, + "step": 26526 + }, + { + "epoch": 0.85, + "learning_rate": 1.1822220710411914e-06, + "loss": 0.8877, + "step": 26527 + }, + { + "epoch": 0.85, + "learning_rate": 1.181733565930705e-06, + "loss": 0.917, + "step": 26528 + }, + { + "epoch": 0.85, + "learning_rate": 1.181245155430457e-06, + "loss": 0.9058, + "step": 26529 + }, + { + "epoch": 0.85, + "learning_rate": 1.1807568395456891e-06, + "loss": 0.9307, + "step": 26530 + }, + { + "epoch": 0.85, + "learning_rate": 1.180268618281638e-06, + "loss": 0.9556, + "step": 26531 + }, + { + "epoch": 0.85, + "learning_rate": 1.1797804916435462e-06, + "loss": 0.8638, + "step": 26532 + }, + { + "epoch": 0.85, + "learning_rate": 1.1792924596366473e-06, + "loss": 0.9814, + "step": 26533 + }, + { + "epoch": 0.85, + "learning_rate": 1.1788045222661781e-06, + "loss": 0.7964, + "step": 26534 + }, + { + "epoch": 0.85, + "learning_rate": 1.1783166795373724e-06, + "loss": 0.9233, + "step": 26535 + }, + { + "epoch": 0.85, + "learning_rate": 1.177828931455467e-06, + "loss": 0.9253, + "step": 26536 + }, + { + "epoch": 0.85, + "learning_rate": 1.1773412780256909e-06, + "loss": 0.8613, + "step": 26537 + }, + { + "epoch": 0.85, + "learning_rate": 1.1768537192532803e-06, + "loss": 0.8745, + "step": 26538 + }, + { + "epoch": 0.85, + "learning_rate": 1.1763662551434617e-06, + "loss": 0.8093, + "step": 26539 + }, + { + "epoch": 0.85, + "learning_rate": 1.1758788857014735e-06, + "loss": 0.9556, + "step": 26540 + }, + { + "epoch": 0.85, + "learning_rate": 1.1753916109325336e-06, + "loss": 0.9829, + "step": 26541 + }, + { + "epoch": 0.85, + "learning_rate": 1.1749044308418766e-06, + "loss": 0.8394, + "step": 26542 + }, + { + "epoch": 0.85, + "learning_rate": 1.1744173454347274e-06, + "loss": 0.4631, + "step": 26543 + }, + { + "epoch": 0.85, + "learning_rate": 1.1739303547163138e-06, + "loss": 0.9038, + "step": 26544 + }, + { + "epoch": 0.85, + "learning_rate": 1.1734434586918565e-06, + "loss": 0.9453, + "step": 26545 + }, + { + "epoch": 0.85, + "learning_rate": 1.172956657366585e-06, + "loss": 0.936, + "step": 26546 + }, + { + "epoch": 0.85, + "learning_rate": 1.1724699507457194e-06, + "loss": 0.8711, + "step": 26547 + }, + { + "epoch": 0.85, + "learning_rate": 1.1719833388344804e-06, + "loss": 0.8403, + "step": 26548 + }, + { + "epoch": 0.85, + "learning_rate": 1.1714968216380885e-06, + "loss": 0.4937, + "step": 26549 + }, + { + "epoch": 0.85, + "learning_rate": 1.1710103991617672e-06, + "loss": 0.8452, + "step": 26550 + }, + { + "epoch": 0.85, + "learning_rate": 1.1705240714107301e-06, + "loss": 1.0308, + "step": 26551 + }, + { + "epoch": 0.85, + "learning_rate": 1.1700378383902e-06, + "loss": 0.9009, + "step": 26552 + }, + { + "epoch": 0.85, + "learning_rate": 1.1695517001053913e-06, + "loss": 0.8804, + "step": 26553 + }, + { + "epoch": 0.85, + "learning_rate": 1.1690656565615188e-06, + "loss": 0.9272, + "step": 26554 + }, + { + "epoch": 0.85, + "learning_rate": 1.1685797077637983e-06, + "loss": 0.9473, + "step": 26555 + }, + { + "epoch": 0.85, + "learning_rate": 1.1680938537174447e-06, + "loss": 0.979, + "step": 26556 + }, + { + "epoch": 0.85, + "learning_rate": 1.167608094427668e-06, + "loss": 0.9297, + "step": 26557 + }, + { + "epoch": 0.85, + "learning_rate": 1.167122429899682e-06, + "loss": 0.968, + "step": 26558 + }, + { + "epoch": 0.85, + "learning_rate": 1.1666368601386968e-06, + "loss": 0.9287, + "step": 26559 + }, + { + "epoch": 0.85, + "learning_rate": 1.1661513851499196e-06, + "loss": 0.9434, + "step": 26560 + }, + { + "epoch": 0.85, + "learning_rate": 1.1656660049385638e-06, + "loss": 0.5237, + "step": 26561 + }, + { + "epoch": 0.85, + "learning_rate": 1.165180719509832e-06, + "loss": 0.9082, + "step": 26562 + }, + { + "epoch": 0.85, + "learning_rate": 1.1646955288689376e-06, + "loss": 0.9399, + "step": 26563 + }, + { + "epoch": 0.85, + "learning_rate": 1.164210433021078e-06, + "loss": 0.9585, + "step": 26564 + }, + { + "epoch": 0.85, + "learning_rate": 1.163725431971463e-06, + "loss": 1.0522, + "step": 26565 + }, + { + "epoch": 0.85, + "learning_rate": 1.1632405257252922e-06, + "loss": 0.9399, + "step": 26566 + }, + { + "epoch": 0.85, + "learning_rate": 1.1627557142877733e-06, + "loss": 0.9849, + "step": 26567 + }, + { + "epoch": 0.85, + "learning_rate": 1.1622709976641022e-06, + "loss": 0.9419, + "step": 26568 + }, + { + "epoch": 0.85, + "learning_rate": 1.161786375859485e-06, + "loss": 0.9092, + "step": 26569 + }, + { + "epoch": 0.85, + "learning_rate": 1.1613018488791162e-06, + "loss": 0.8569, + "step": 26570 + }, + { + "epoch": 0.85, + "learning_rate": 1.1608174167282004e-06, + "loss": 0.8433, + "step": 26571 + }, + { + "epoch": 0.85, + "learning_rate": 1.160333079411926e-06, + "loss": 0.8887, + "step": 26572 + }, + { + "epoch": 0.85, + "learning_rate": 1.1598488369354977e-06, + "loss": 0.9854, + "step": 26573 + }, + { + "epoch": 0.85, + "learning_rate": 1.1593646893041043e-06, + "loss": 0.9683, + "step": 26574 + }, + { + "epoch": 0.85, + "learning_rate": 1.1588806365229466e-06, + "loss": 0.8745, + "step": 26575 + }, + { + "epoch": 0.85, + "learning_rate": 1.1583966785972122e-06, + "loss": 0.8203, + "step": 26576 + }, + { + "epoch": 0.85, + "learning_rate": 1.1579128155320984e-06, + "loss": 0.8711, + "step": 26577 + }, + { + "epoch": 0.85, + "learning_rate": 1.1574290473327943e-06, + "loss": 1.0415, + "step": 26578 + }, + { + "epoch": 0.85, + "learning_rate": 1.15694537400449e-06, + "loss": 0.9399, + "step": 26579 + }, + { + "epoch": 0.85, + "learning_rate": 1.1564617955523716e-06, + "loss": 0.9351, + "step": 26580 + }, + { + "epoch": 0.85, + "learning_rate": 1.1559783119816337e-06, + "loss": 0.8364, + "step": 26581 + }, + { + "epoch": 0.85, + "learning_rate": 1.1554949232974588e-06, + "loss": 0.8921, + "step": 26582 + }, + { + "epoch": 0.85, + "learning_rate": 1.1550116295050363e-06, + "loss": 0.9175, + "step": 26583 + }, + { + "epoch": 0.85, + "learning_rate": 1.1545284306095473e-06, + "loss": 0.9766, + "step": 26584 + }, + { + "epoch": 0.85, + "learning_rate": 1.154045326616181e-06, + "loss": 1.0239, + "step": 26585 + }, + { + "epoch": 0.85, + "learning_rate": 1.153562317530118e-06, + "loss": 1.0225, + "step": 26586 + }, + { + "epoch": 0.85, + "learning_rate": 1.1530794033565417e-06, + "loss": 0.9287, + "step": 26587 + }, + { + "epoch": 0.85, + "learning_rate": 1.1525965841006293e-06, + "loss": 0.9561, + "step": 26588 + }, + { + "epoch": 0.85, + "learning_rate": 1.152113859767565e-06, + "loss": 0.9458, + "step": 26589 + }, + { + "epoch": 0.85, + "learning_rate": 1.1516312303625253e-06, + "loss": 0.8726, + "step": 26590 + }, + { + "epoch": 0.85, + "learning_rate": 1.1511486958906914e-06, + "loss": 1.0605, + "step": 26591 + }, + { + "epoch": 0.85, + "learning_rate": 1.1506662563572392e-06, + "loss": 0.8647, + "step": 26592 + }, + { + "epoch": 0.85, + "learning_rate": 1.150183911767343e-06, + "loss": 0.8999, + "step": 26593 + }, + { + "epoch": 0.85, + "learning_rate": 1.1497016621261781e-06, + "loss": 0.761, + "step": 26594 + }, + { + "epoch": 0.85, + "learning_rate": 1.1492195074389211e-06, + "loss": 1.0259, + "step": 26595 + }, + { + "epoch": 0.85, + "learning_rate": 1.1487374477107404e-06, + "loss": 0.8799, + "step": 26596 + }, + { + "epoch": 0.85, + "learning_rate": 1.148255482946814e-06, + "loss": 0.9331, + "step": 26597 + }, + { + "epoch": 0.85, + "learning_rate": 1.1477736131523087e-06, + "loss": 0.8384, + "step": 26598 + }, + { + "epoch": 0.85, + "learning_rate": 1.1472918383323938e-06, + "loss": 0.9106, + "step": 26599 + }, + { + "epoch": 0.85, + "learning_rate": 1.1468101584922409e-06, + "loss": 0.4578, + "step": 26600 + }, + { + "epoch": 0.85, + "learning_rate": 1.1463285736370178e-06, + "loss": 0.833, + "step": 26601 + }, + { + "epoch": 0.85, + "learning_rate": 1.1458470837718904e-06, + "loss": 0.8823, + "step": 26602 + }, + { + "epoch": 0.85, + "learning_rate": 1.1453656889020214e-06, + "loss": 0.8711, + "step": 26603 + }, + { + "epoch": 0.85, + "learning_rate": 1.1448843890325822e-06, + "loss": 0.9175, + "step": 26604 + }, + { + "epoch": 0.85, + "learning_rate": 1.1444031841687297e-06, + "loss": 0.8545, + "step": 26605 + }, + { + "epoch": 0.85, + "learning_rate": 1.1439220743156333e-06, + "loss": 0.8979, + "step": 26606 + }, + { + "epoch": 0.85, + "learning_rate": 1.1434410594784494e-06, + "loss": 1.0615, + "step": 26607 + }, + { + "epoch": 0.85, + "learning_rate": 1.1429601396623447e-06, + "loss": 0.895, + "step": 26608 + }, + { + "epoch": 0.85, + "learning_rate": 1.142479314872471e-06, + "loss": 0.9038, + "step": 26609 + }, + { + "epoch": 0.85, + "learning_rate": 1.1419985851139936e-06, + "loss": 0.9722, + "step": 26610 + }, + { + "epoch": 0.85, + "learning_rate": 1.1415179503920658e-06, + "loss": 0.8843, + "step": 26611 + }, + { + "epoch": 0.85, + "learning_rate": 1.1410374107118483e-06, + "loss": 0.9761, + "step": 26612 + }, + { + "epoch": 0.85, + "learning_rate": 1.1405569660784932e-06, + "loss": 0.8799, + "step": 26613 + }, + { + "epoch": 0.85, + "learning_rate": 1.1400766164971577e-06, + "loss": 0.9458, + "step": 26614 + }, + { + "epoch": 0.85, + "learning_rate": 1.1395963619729932e-06, + "loss": 0.9272, + "step": 26615 + }, + { + "epoch": 0.85, + "learning_rate": 1.1391162025111568e-06, + "loss": 0.8853, + "step": 26616 + }, + { + "epoch": 0.85, + "learning_rate": 1.1386361381167942e-06, + "loss": 0.8579, + "step": 26617 + }, + { + "epoch": 0.85, + "learning_rate": 1.1381561687950592e-06, + "loss": 0.8789, + "step": 26618 + }, + { + "epoch": 0.85, + "learning_rate": 1.1376762945510988e-06, + "loss": 0.9521, + "step": 26619 + }, + { + "epoch": 0.85, + "learning_rate": 1.1371965153900666e-06, + "loss": 0.8896, + "step": 26620 + }, + { + "epoch": 0.85, + "learning_rate": 1.136716831317104e-06, + "loss": 0.9438, + "step": 26621 + }, + { + "epoch": 0.85, + "learning_rate": 1.1362372423373624e-06, + "loss": 0.8989, + "step": 26622 + }, + { + "epoch": 0.85, + "learning_rate": 1.1357577484559856e-06, + "loss": 0.9878, + "step": 26623 + }, + { + "epoch": 0.85, + "learning_rate": 1.1352783496781173e-06, + "loss": 0.9297, + "step": 26624 + }, + { + "epoch": 0.85, + "learning_rate": 1.1347990460088998e-06, + "loss": 0.9771, + "step": 26625 + }, + { + "epoch": 0.85, + "learning_rate": 1.134319837453478e-06, + "loss": 0.7729, + "step": 26626 + }, + { + "epoch": 0.85, + "learning_rate": 1.1338407240169912e-06, + "loss": 0.9307, + "step": 26627 + }, + { + "epoch": 0.85, + "learning_rate": 1.133361705704583e-06, + "loss": 1.0874, + "step": 26628 + }, + { + "epoch": 0.85, + "learning_rate": 1.1328827825213884e-06, + "loss": 0.8794, + "step": 26629 + }, + { + "epoch": 0.85, + "learning_rate": 1.1324039544725496e-06, + "loss": 0.9023, + "step": 26630 + }, + { + "epoch": 0.85, + "learning_rate": 1.1319252215632026e-06, + "loss": 1.0605, + "step": 26631 + }, + { + "epoch": 0.85, + "learning_rate": 1.1314465837984833e-06, + "loss": 0.915, + "step": 26632 + }, + { + "epoch": 0.85, + "learning_rate": 1.1309680411835244e-06, + "loss": 0.8643, + "step": 26633 + }, + { + "epoch": 0.85, + "learning_rate": 1.130489593723465e-06, + "loss": 0.8628, + "step": 26634 + }, + { + "epoch": 0.85, + "learning_rate": 1.1300112414234332e-06, + "loss": 0.9883, + "step": 26635 + }, + { + "epoch": 0.85, + "learning_rate": 1.1295329842885671e-06, + "loss": 0.8433, + "step": 26636 + }, + { + "epoch": 0.85, + "learning_rate": 1.1290548223239938e-06, + "loss": 0.9126, + "step": 26637 + }, + { + "epoch": 0.85, + "learning_rate": 1.128576755534846e-06, + "loss": 0.9136, + "step": 26638 + }, + { + "epoch": 0.85, + "learning_rate": 1.128098783926248e-06, + "loss": 1.0078, + "step": 26639 + }, + { + "epoch": 0.85, + "learning_rate": 1.127620907503333e-06, + "loss": 0.9761, + "step": 26640 + }, + { + "epoch": 0.85, + "learning_rate": 1.1271431262712262e-06, + "loss": 0.8564, + "step": 26641 + }, + { + "epoch": 0.85, + "learning_rate": 1.1266654402350518e-06, + "loss": 0.9502, + "step": 26642 + }, + { + "epoch": 0.85, + "learning_rate": 1.1261878493999389e-06, + "loss": 1.0459, + "step": 26643 + }, + { + "epoch": 0.85, + "learning_rate": 1.1257103537710067e-06, + "loss": 1.002, + "step": 26644 + }, + { + "epoch": 0.85, + "learning_rate": 1.1252329533533835e-06, + "loss": 0.9131, + "step": 26645 + }, + { + "epoch": 0.85, + "learning_rate": 1.1247556481521892e-06, + "loss": 0.9258, + "step": 26646 + }, + { + "epoch": 0.85, + "learning_rate": 1.1242784381725424e-06, + "loss": 0.8604, + "step": 26647 + }, + { + "epoch": 0.85, + "learning_rate": 1.1238013234195644e-06, + "loss": 0.8896, + "step": 26648 + }, + { + "epoch": 0.85, + "learning_rate": 1.1233243038983755e-06, + "loss": 0.9722, + "step": 26649 + }, + { + "epoch": 0.85, + "learning_rate": 1.1228473796140916e-06, + "loss": 0.8906, + "step": 26650 + }, + { + "epoch": 0.85, + "learning_rate": 1.1223705505718319e-06, + "loss": 1.0483, + "step": 26651 + }, + { + "epoch": 0.85, + "learning_rate": 1.1218938167767091e-06, + "loss": 1.0059, + "step": 26652 + }, + { + "epoch": 0.85, + "learning_rate": 1.1214171782338413e-06, + "loss": 0.9175, + "step": 26653 + }, + { + "epoch": 0.85, + "learning_rate": 1.120940634948341e-06, + "loss": 0.9048, + "step": 26654 + }, + { + "epoch": 0.85, + "learning_rate": 1.1204641869253219e-06, + "loss": 1.0044, + "step": 26655 + }, + { + "epoch": 0.85, + "learning_rate": 1.119987834169891e-06, + "loss": 0.7661, + "step": 26656 + }, + { + "epoch": 0.85, + "learning_rate": 1.1195115766871668e-06, + "loss": 0.8823, + "step": 26657 + }, + { + "epoch": 0.85, + "learning_rate": 1.1190354144822514e-06, + "loss": 0.7676, + "step": 26658 + }, + { + "epoch": 0.85, + "learning_rate": 1.1185593475602596e-06, + "loss": 0.9614, + "step": 26659 + }, + { + "epoch": 0.85, + "learning_rate": 1.1180833759262944e-06, + "loss": 0.9463, + "step": 26660 + }, + { + "epoch": 0.85, + "learning_rate": 1.117607499585468e-06, + "loss": 1.062, + "step": 26661 + }, + { + "epoch": 0.85, + "learning_rate": 1.1171317185428788e-06, + "loss": 0.9873, + "step": 26662 + }, + { + "epoch": 0.85, + "learning_rate": 1.116656032803638e-06, + "loss": 0.8496, + "step": 26663 + }, + { + "epoch": 0.85, + "learning_rate": 1.116180442372844e-06, + "loss": 0.8013, + "step": 26664 + }, + { + "epoch": 0.85, + "learning_rate": 1.1157049472556036e-06, + "loss": 0.9829, + "step": 26665 + }, + { + "epoch": 0.85, + "learning_rate": 1.1152295474570142e-06, + "loss": 0.9429, + "step": 26666 + }, + { + "epoch": 0.85, + "learning_rate": 1.1147542429821812e-06, + "loss": 0.958, + "step": 26667 + }, + { + "epoch": 0.85, + "learning_rate": 1.1142790338361987e-06, + "loss": 0.9229, + "step": 26668 + }, + { + "epoch": 0.85, + "learning_rate": 1.1138039200241723e-06, + "loss": 0.9585, + "step": 26669 + }, + { + "epoch": 0.85, + "learning_rate": 1.1133289015511905e-06, + "loss": 0.8867, + "step": 26670 + }, + { + "epoch": 0.85, + "learning_rate": 1.1128539784223568e-06, + "loss": 0.8481, + "step": 26671 + }, + { + "epoch": 0.85, + "learning_rate": 1.1123791506427628e-06, + "loss": 0.9443, + "step": 26672 + }, + { + "epoch": 0.85, + "learning_rate": 1.1119044182175054e-06, + "loss": 1.0703, + "step": 26673 + }, + { + "epoch": 0.85, + "learning_rate": 1.111429781151675e-06, + "loss": 1.0044, + "step": 26674 + }, + { + "epoch": 0.85, + "learning_rate": 1.1109552394503676e-06, + "loss": 0.9082, + "step": 26675 + }, + { + "epoch": 0.85, + "learning_rate": 1.1104807931186724e-06, + "loss": 0.9116, + "step": 26676 + }, + { + "epoch": 0.85, + "learning_rate": 1.1100064421616808e-06, + "loss": 0.8691, + "step": 26677 + }, + { + "epoch": 0.85, + "learning_rate": 1.1095321865844777e-06, + "loss": 0.8862, + "step": 26678 + }, + { + "epoch": 0.85, + "learning_rate": 1.109058026392158e-06, + "loss": 0.9321, + "step": 26679 + }, + { + "epoch": 0.85, + "learning_rate": 1.108583961589803e-06, + "loss": 0.772, + "step": 26680 + }, + { + "epoch": 0.85, + "learning_rate": 1.1081099921825033e-06, + "loss": 0.8647, + "step": 26681 + }, + { + "epoch": 0.85, + "learning_rate": 1.1076361181753425e-06, + "loss": 0.8994, + "step": 26682 + }, + { + "epoch": 0.85, + "learning_rate": 1.1071623395734032e-06, + "loss": 0.9312, + "step": 26683 + }, + { + "epoch": 0.85, + "learning_rate": 1.1066886563817714e-06, + "loss": 0.9277, + "step": 26684 + }, + { + "epoch": 0.85, + "learning_rate": 1.1062150686055272e-06, + "loss": 0.8491, + "step": 26685 + }, + { + "epoch": 0.85, + "learning_rate": 1.1057415762497526e-06, + "loss": 0.8843, + "step": 26686 + }, + { + "epoch": 0.85, + "learning_rate": 1.1052681793195253e-06, + "loss": 0.9404, + "step": 26687 + }, + { + "epoch": 0.85, + "learning_rate": 1.104794877819928e-06, + "loss": 1.0283, + "step": 26688 + }, + { + "epoch": 0.85, + "learning_rate": 1.1043216717560357e-06, + "loss": 0.9556, + "step": 26689 + }, + { + "epoch": 0.85, + "learning_rate": 1.1038485611329274e-06, + "loss": 0.9761, + "step": 26690 + }, + { + "epoch": 0.85, + "learning_rate": 1.1033755459556793e-06, + "loss": 0.8896, + "step": 26691 + }, + { + "epoch": 0.85, + "learning_rate": 1.1029026262293651e-06, + "loss": 0.8159, + "step": 26692 + }, + { + "epoch": 0.85, + "learning_rate": 1.1024298019590563e-06, + "loss": 0.9272, + "step": 26693 + }, + { + "epoch": 0.85, + "learning_rate": 1.1019570731498308e-06, + "loss": 0.9917, + "step": 26694 + }, + { + "epoch": 0.85, + "learning_rate": 1.1014844398067558e-06, + "loss": 0.9902, + "step": 26695 + }, + { + "epoch": 0.85, + "learning_rate": 1.1010119019349075e-06, + "loss": 0.8994, + "step": 26696 + }, + { + "epoch": 0.85, + "learning_rate": 1.1005394595393493e-06, + "loss": 0.9897, + "step": 26697 + }, + { + "epoch": 0.85, + "learning_rate": 1.100067112625156e-06, + "loss": 0.9116, + "step": 26698 + }, + { + "epoch": 0.85, + "learning_rate": 1.0995948611973916e-06, + "loss": 0.9541, + "step": 26699 + }, + { + "epoch": 0.85, + "learning_rate": 1.0991227052611253e-06, + "loss": 0.9639, + "step": 26700 + }, + { + "epoch": 0.85, + "learning_rate": 1.0986506448214185e-06, + "loss": 0.8311, + "step": 26701 + }, + { + "epoch": 0.85, + "learning_rate": 1.0981786798833406e-06, + "loss": 0.9536, + "step": 26702 + }, + { + "epoch": 0.85, + "learning_rate": 1.0977068104519516e-06, + "loss": 0.874, + "step": 26703 + }, + { + "epoch": 0.85, + "learning_rate": 1.0972350365323182e-06, + "loss": 0.9312, + "step": 26704 + }, + { + "epoch": 0.85, + "learning_rate": 1.0967633581294979e-06, + "loss": 0.9087, + "step": 26705 + }, + { + "epoch": 0.85, + "learning_rate": 1.0962917752485569e-06, + "loss": 0.9097, + "step": 26706 + }, + { + "epoch": 0.85, + "learning_rate": 1.0958202878945468e-06, + "loss": 0.8379, + "step": 26707 + }, + { + "epoch": 0.85, + "learning_rate": 1.0953488960725323e-06, + "loss": 0.9849, + "step": 26708 + }, + { + "epoch": 0.85, + "learning_rate": 1.094877599787567e-06, + "loss": 0.7595, + "step": 26709 + }, + { + "epoch": 0.85, + "learning_rate": 1.0944063990447118e-06, + "loss": 0.9614, + "step": 26710 + }, + { + "epoch": 0.85, + "learning_rate": 1.0939352938490178e-06, + "loss": 0.9697, + "step": 26711 + }, + { + "epoch": 0.85, + "learning_rate": 1.0934642842055432e-06, + "loss": 0.9453, + "step": 26712 + }, + { + "epoch": 0.85, + "learning_rate": 1.0929933701193375e-06, + "loss": 0.8062, + "step": 26713 + }, + { + "epoch": 0.85, + "learning_rate": 1.0925225515954585e-06, + "loss": 0.9507, + "step": 26714 + }, + { + "epoch": 0.85, + "learning_rate": 1.0920518286389503e-06, + "loss": 0.8799, + "step": 26715 + }, + { + "epoch": 0.85, + "learning_rate": 1.0915812012548698e-06, + "loss": 0.8818, + "step": 26716 + }, + { + "epoch": 0.85, + "learning_rate": 1.091110669448262e-06, + "loss": 0.8225, + "step": 26717 + }, + { + "epoch": 0.85, + "learning_rate": 1.0906402332241784e-06, + "loss": 0.8843, + "step": 26718 + }, + { + "epoch": 0.85, + "learning_rate": 1.0901698925876626e-06, + "loss": 0.8677, + "step": 26719 + }, + { + "epoch": 0.85, + "learning_rate": 1.0896996475437649e-06, + "loss": 0.915, + "step": 26720 + }, + { + "epoch": 0.85, + "learning_rate": 1.0892294980975282e-06, + "loss": 0.9497, + "step": 26721 + }, + { + "epoch": 0.85, + "learning_rate": 1.0887594442539973e-06, + "loss": 0.9058, + "step": 26722 + }, + { + "epoch": 0.85, + "learning_rate": 1.0882894860182135e-06, + "loss": 0.9712, + "step": 26723 + }, + { + "epoch": 0.85, + "learning_rate": 1.087819623395222e-06, + "loss": 0.9785, + "step": 26724 + }, + { + "epoch": 0.85, + "learning_rate": 1.087349856390063e-06, + "loss": 0.9644, + "step": 26725 + }, + { + "epoch": 0.85, + "learning_rate": 1.0868801850077738e-06, + "loss": 0.9033, + "step": 26726 + }, + { + "epoch": 0.85, + "learning_rate": 1.0864106092533977e-06, + "loss": 0.9863, + "step": 26727 + }, + { + "epoch": 0.85, + "learning_rate": 1.0859411291319678e-06, + "loss": 0.9536, + "step": 26728 + }, + { + "epoch": 0.85, + "learning_rate": 1.0854717446485274e-06, + "loss": 0.8618, + "step": 26729 + }, + { + "epoch": 0.85, + "learning_rate": 1.0850024558081085e-06, + "loss": 0.9126, + "step": 26730 + }, + { + "epoch": 0.85, + "learning_rate": 1.0845332626157456e-06, + "loss": 0.8872, + "step": 26731 + }, + { + "epoch": 0.85, + "learning_rate": 1.0840641650764727e-06, + "loss": 0.9014, + "step": 26732 + }, + { + "epoch": 0.85, + "learning_rate": 1.0835951631953256e-06, + "loss": 0.9102, + "step": 26733 + }, + { + "epoch": 0.85, + "learning_rate": 1.0831262569773327e-06, + "loss": 0.9556, + "step": 26734 + }, + { + "epoch": 0.86, + "learning_rate": 1.0826574464275274e-06, + "loss": 0.9194, + "step": 26735 + }, + { + "epoch": 0.86, + "learning_rate": 1.0821887315509372e-06, + "loss": 0.8391, + "step": 26736 + }, + { + "epoch": 0.86, + "learning_rate": 1.0817201123525955e-06, + "loss": 0.8857, + "step": 26737 + }, + { + "epoch": 0.86, + "learning_rate": 1.081251588837523e-06, + "loss": 0.9028, + "step": 26738 + }, + { + "epoch": 0.86, + "learning_rate": 1.0807831610107521e-06, + "loss": 0.8667, + "step": 26739 + }, + { + "epoch": 0.86, + "learning_rate": 1.0803148288773047e-06, + "loss": 0.9077, + "step": 26740 + }, + { + "epoch": 0.86, + "learning_rate": 1.0798465924422097e-06, + "loss": 0.96, + "step": 26741 + }, + { + "epoch": 0.86, + "learning_rate": 1.0793784517104856e-06, + "loss": 0.9189, + "step": 26742 + }, + { + "epoch": 0.86, + "learning_rate": 1.0789104066871604e-06, + "loss": 0.874, + "step": 26743 + }, + { + "epoch": 0.86, + "learning_rate": 1.0784424573772533e-06, + "loss": 0.7671, + "step": 26744 + }, + { + "epoch": 0.86, + "learning_rate": 1.0779746037857842e-06, + "loss": 0.8564, + "step": 26745 + }, + { + "epoch": 0.86, + "learning_rate": 1.0775068459177719e-06, + "loss": 0.9429, + "step": 26746 + }, + { + "epoch": 0.86, + "learning_rate": 1.077039183778238e-06, + "loss": 0.9604, + "step": 26747 + }, + { + "epoch": 0.86, + "learning_rate": 1.0765716173721963e-06, + "loss": 0.8569, + "step": 26748 + }, + { + "epoch": 0.86, + "learning_rate": 1.0761041467046661e-06, + "loss": 0.9053, + "step": 26749 + }, + { + "epoch": 0.86, + "learning_rate": 1.0756367717806616e-06, + "loss": 0.9146, + "step": 26750 + }, + { + "epoch": 0.86, + "learning_rate": 1.075169492605198e-06, + "loss": 0.9028, + "step": 26751 + }, + { + "epoch": 0.86, + "learning_rate": 1.0747023091832887e-06, + "loss": 0.9185, + "step": 26752 + }, + { + "epoch": 0.86, + "learning_rate": 1.0742352215199459e-06, + "loss": 0.981, + "step": 26753 + }, + { + "epoch": 0.86, + "learning_rate": 1.073768229620179e-06, + "loss": 0.8726, + "step": 26754 + }, + { + "epoch": 0.86, + "learning_rate": 1.0733013334890019e-06, + "loss": 0.9995, + "step": 26755 + }, + { + "epoch": 0.86, + "learning_rate": 1.0728345331314194e-06, + "loss": 0.4658, + "step": 26756 + }, + { + "epoch": 0.86, + "learning_rate": 1.072367828552444e-06, + "loss": 0.9253, + "step": 26757 + }, + { + "epoch": 0.86, + "learning_rate": 1.0719012197570799e-06, + "loss": 1.0171, + "step": 26758 + }, + { + "epoch": 0.86, + "learning_rate": 1.0714347067503383e-06, + "loss": 1.0137, + "step": 26759 + }, + { + "epoch": 0.86, + "learning_rate": 1.0709682895372165e-06, + "loss": 1.021, + "step": 26760 + }, + { + "epoch": 0.86, + "learning_rate": 1.0705019681227246e-06, + "loss": 0.8291, + "step": 26761 + }, + { + "epoch": 0.86, + "learning_rate": 1.0700357425118624e-06, + "loss": 0.8687, + "step": 26762 + }, + { + "epoch": 0.86, + "learning_rate": 1.0695696127096355e-06, + "loss": 0.9956, + "step": 26763 + }, + { + "epoch": 0.86, + "learning_rate": 1.0691035787210424e-06, + "loss": 0.9531, + "step": 26764 + }, + { + "epoch": 0.86, + "learning_rate": 1.0686376405510824e-06, + "loss": 1.0791, + "step": 26765 + }, + { + "epoch": 0.86, + "learning_rate": 1.0681717982047568e-06, + "loss": 0.9243, + "step": 26766 + }, + { + "epoch": 0.86, + "learning_rate": 1.067706051687063e-06, + "loss": 0.9438, + "step": 26767 + }, + { + "epoch": 0.86, + "learning_rate": 1.0672404010029958e-06, + "loss": 0.4844, + "step": 26768 + }, + { + "epoch": 0.86, + "learning_rate": 1.0667748461575544e-06, + "loss": 0.874, + "step": 26769 + }, + { + "epoch": 0.86, + "learning_rate": 1.0663093871557329e-06, + "loss": 0.9473, + "step": 26770 + }, + { + "epoch": 0.86, + "learning_rate": 1.0658440240025215e-06, + "loss": 0.8765, + "step": 26771 + }, + { + "epoch": 0.86, + "learning_rate": 1.0653787567029184e-06, + "loss": 0.8662, + "step": 26772 + }, + { + "epoch": 0.86, + "learning_rate": 1.0649135852619109e-06, + "loss": 0.9824, + "step": 26773 + }, + { + "epoch": 0.86, + "learning_rate": 1.0644485096844926e-06, + "loss": 0.855, + "step": 26774 + }, + { + "epoch": 0.86, + "learning_rate": 1.063983529975654e-06, + "loss": 0.9331, + "step": 26775 + }, + { + "epoch": 0.86, + "learning_rate": 1.0635186461403813e-06, + "loss": 0.9756, + "step": 26776 + }, + { + "epoch": 0.86, + "learning_rate": 1.0630538581836613e-06, + "loss": 0.9287, + "step": 26777 + }, + { + "epoch": 0.86, + "learning_rate": 1.0625891661104847e-06, + "loss": 0.8823, + "step": 26778 + }, + { + "epoch": 0.86, + "learning_rate": 1.062124569925833e-06, + "loss": 0.9819, + "step": 26779 + }, + { + "epoch": 0.86, + "learning_rate": 1.0616600696346945e-06, + "loss": 0.886, + "step": 26780 + }, + { + "epoch": 0.86, + "learning_rate": 1.0611956652420485e-06, + "loss": 0.9854, + "step": 26781 + }, + { + "epoch": 0.86, + "learning_rate": 1.0607313567528843e-06, + "loss": 0.8828, + "step": 26782 + }, + { + "epoch": 0.86, + "learning_rate": 1.0602671441721756e-06, + "loss": 0.876, + "step": 26783 + }, + { + "epoch": 0.86, + "learning_rate": 1.0598030275049088e-06, + "loss": 0.9912, + "step": 26784 + }, + { + "epoch": 0.86, + "learning_rate": 1.0593390067560582e-06, + "loss": 0.8569, + "step": 26785 + }, + { + "epoch": 0.86, + "learning_rate": 1.058875081930607e-06, + "loss": 0.9346, + "step": 26786 + }, + { + "epoch": 0.86, + "learning_rate": 1.058411253033529e-06, + "loss": 0.4243, + "step": 26787 + }, + { + "epoch": 0.86, + "learning_rate": 1.0579475200698043e-06, + "loss": 0.9194, + "step": 26788 + }, + { + "epoch": 0.86, + "learning_rate": 1.0574838830444056e-06, + "loss": 0.9922, + "step": 26789 + }, + { + "epoch": 0.86, + "learning_rate": 1.0570203419623082e-06, + "loss": 0.8101, + "step": 26790 + }, + { + "epoch": 0.86, + "learning_rate": 1.0565568968284822e-06, + "loss": 1.0142, + "step": 26791 + }, + { + "epoch": 0.86, + "learning_rate": 1.056093547647905e-06, + "loss": 0.8359, + "step": 26792 + }, + { + "epoch": 0.86, + "learning_rate": 1.0556302944255437e-06, + "loss": 0.9961, + "step": 26793 + }, + { + "epoch": 0.86, + "learning_rate": 1.0551671371663719e-06, + "loss": 0.98, + "step": 26794 + }, + { + "epoch": 0.86, + "learning_rate": 1.0547040758753557e-06, + "loss": 0.9272, + "step": 26795 + }, + { + "epoch": 0.86, + "learning_rate": 1.0542411105574657e-06, + "loss": 0.9927, + "step": 26796 + }, + { + "epoch": 0.86, + "learning_rate": 1.053778241217669e-06, + "loss": 0.8672, + "step": 26797 + }, + { + "epoch": 0.86, + "learning_rate": 1.0533154678609292e-06, + "loss": 1.0718, + "step": 26798 + }, + { + "epoch": 0.86, + "learning_rate": 1.0528527904922114e-06, + "loss": 0.876, + "step": 26799 + }, + { + "epoch": 0.86, + "learning_rate": 1.0523902091164827e-06, + "loss": 0.958, + "step": 26800 + }, + { + "epoch": 0.86, + "learning_rate": 1.0519277237387037e-06, + "loss": 0.9702, + "step": 26801 + }, + { + "epoch": 0.86, + "learning_rate": 1.0514653343638371e-06, + "loss": 0.9644, + "step": 26802 + }, + { + "epoch": 0.86, + "learning_rate": 1.0510030409968453e-06, + "loss": 0.8774, + "step": 26803 + }, + { + "epoch": 0.86, + "learning_rate": 1.0505408436426835e-06, + "loss": 0.9297, + "step": 26804 + }, + { + "epoch": 0.86, + "learning_rate": 1.0500787423063164e-06, + "loss": 0.9551, + "step": 26805 + }, + { + "epoch": 0.86, + "learning_rate": 1.0496167369926979e-06, + "loss": 0.8472, + "step": 26806 + }, + { + "epoch": 0.86, + "learning_rate": 1.0491548277067852e-06, + "loss": 1.0083, + "step": 26807 + }, + { + "epoch": 0.86, + "learning_rate": 1.0486930144535367e-06, + "loss": 0.8423, + "step": 26808 + }, + { + "epoch": 0.86, + "learning_rate": 1.048231297237905e-06, + "loss": 0.8862, + "step": 26809 + }, + { + "epoch": 0.86, + "learning_rate": 1.0477696760648425e-06, + "loss": 0.9233, + "step": 26810 + }, + { + "epoch": 0.86, + "learning_rate": 1.0473081509393045e-06, + "loss": 0.9136, + "step": 26811 + }, + { + "epoch": 0.86, + "learning_rate": 1.0468467218662437e-06, + "loss": 0.749, + "step": 26812 + }, + { + "epoch": 0.86, + "learning_rate": 1.0463853888506049e-06, + "loss": 0.8594, + "step": 26813 + }, + { + "epoch": 0.86, + "learning_rate": 1.045924151897344e-06, + "loss": 0.9028, + "step": 26814 + }, + { + "epoch": 0.86, + "learning_rate": 1.0454630110114072e-06, + "loss": 0.8369, + "step": 26815 + }, + { + "epoch": 0.86, + "learning_rate": 1.0450019661977406e-06, + "loss": 0.9077, + "step": 26816 + }, + { + "epoch": 0.86, + "learning_rate": 1.0445410174612934e-06, + "loss": 1.0063, + "step": 26817 + }, + { + "epoch": 0.86, + "learning_rate": 1.0440801648070076e-06, + "loss": 0.8862, + "step": 26818 + }, + { + "epoch": 0.86, + "learning_rate": 1.043619408239831e-06, + "loss": 0.8872, + "step": 26819 + }, + { + "epoch": 0.86, + "learning_rate": 1.0431587477647064e-06, + "loss": 0.9932, + "step": 26820 + }, + { + "epoch": 0.86, + "learning_rate": 1.0426981833865757e-06, + "loss": 0.8271, + "step": 26821 + }, + { + "epoch": 0.86, + "learning_rate": 1.042237715110378e-06, + "loss": 0.9888, + "step": 26822 + }, + { + "epoch": 0.86, + "learning_rate": 1.0417773429410582e-06, + "loss": 0.8667, + "step": 26823 + }, + { + "epoch": 0.86, + "learning_rate": 1.0413170668835504e-06, + "loss": 0.9014, + "step": 26824 + }, + { + "epoch": 0.86, + "learning_rate": 1.0408568869427972e-06, + "loss": 0.9258, + "step": 26825 + }, + { + "epoch": 0.86, + "learning_rate": 1.0403968031237322e-06, + "loss": 0.981, + "step": 26826 + }, + { + "epoch": 0.86, + "learning_rate": 1.0399368154312971e-06, + "loss": 0.8779, + "step": 26827 + }, + { + "epoch": 0.86, + "learning_rate": 1.0394769238704206e-06, + "loss": 0.9658, + "step": 26828 + }, + { + "epoch": 0.86, + "learning_rate": 1.0390171284460416e-06, + "loss": 0.8647, + "step": 26829 + }, + { + "epoch": 0.86, + "learning_rate": 1.0385574291630885e-06, + "loss": 0.9351, + "step": 26830 + }, + { + "epoch": 0.86, + "learning_rate": 1.0380978260264984e-06, + "loss": 0.9648, + "step": 26831 + }, + { + "epoch": 0.86, + "learning_rate": 1.0376383190411986e-06, + "loss": 0.9023, + "step": 26832 + }, + { + "epoch": 0.86, + "learning_rate": 1.037178908212122e-06, + "loss": 1.0122, + "step": 26833 + }, + { + "epoch": 0.86, + "learning_rate": 1.0367195935441943e-06, + "loss": 0.9219, + "step": 26834 + }, + { + "epoch": 0.86, + "learning_rate": 1.0362603750423494e-06, + "loss": 0.9487, + "step": 26835 + }, + { + "epoch": 0.86, + "learning_rate": 1.0358012527115058e-06, + "loss": 0.8657, + "step": 26836 + }, + { + "epoch": 0.86, + "learning_rate": 1.035342226556596e-06, + "loss": 0.8989, + "step": 26837 + }, + { + "epoch": 0.86, + "learning_rate": 1.0348832965825406e-06, + "loss": 0.8809, + "step": 26838 + }, + { + "epoch": 0.86, + "learning_rate": 1.0344244627942668e-06, + "loss": 0.9688, + "step": 26839 + }, + { + "epoch": 0.86, + "learning_rate": 1.033965725196694e-06, + "loss": 0.8494, + "step": 26840 + }, + { + "epoch": 0.86, + "learning_rate": 1.033507083794748e-06, + "loss": 0.8196, + "step": 26841 + }, + { + "epoch": 0.86, + "learning_rate": 1.0330485385933476e-06, + "loss": 0.9214, + "step": 26842 + }, + { + "epoch": 0.86, + "learning_rate": 1.0325900895974128e-06, + "loss": 0.9609, + "step": 26843 + }, + { + "epoch": 0.86, + "learning_rate": 1.0321317368118588e-06, + "loss": 0.8384, + "step": 26844 + }, + { + "epoch": 0.86, + "learning_rate": 1.0316734802416085e-06, + "loss": 0.9424, + "step": 26845 + }, + { + "epoch": 0.86, + "learning_rate": 1.031215319891573e-06, + "loss": 0.9189, + "step": 26846 + }, + { + "epoch": 0.86, + "learning_rate": 1.0307572557666745e-06, + "loss": 0.835, + "step": 26847 + }, + { + "epoch": 0.86, + "learning_rate": 1.0302992878718232e-06, + "loss": 1.0249, + "step": 26848 + }, + { + "epoch": 0.86, + "learning_rate": 1.0298414162119318e-06, + "loss": 0.9614, + "step": 26849 + }, + { + "epoch": 0.86, + "learning_rate": 1.0293836407919154e-06, + "loss": 0.9756, + "step": 26850 + }, + { + "epoch": 0.86, + "learning_rate": 1.0289259616166858e-06, + "loss": 0.8999, + "step": 26851 + }, + { + "epoch": 0.86, + "learning_rate": 1.0284683786911486e-06, + "loss": 1.0278, + "step": 26852 + }, + { + "epoch": 0.86, + "learning_rate": 1.0280108920202192e-06, + "loss": 0.8125, + "step": 26853 + }, + { + "epoch": 0.86, + "learning_rate": 1.0275535016088033e-06, + "loss": 0.781, + "step": 26854 + }, + { + "epoch": 0.86, + "learning_rate": 1.0270962074618064e-06, + "loss": 0.918, + "step": 26855 + }, + { + "epoch": 0.86, + "learning_rate": 1.0266390095841372e-06, + "loss": 0.8286, + "step": 26856 + }, + { + "epoch": 0.86, + "learning_rate": 1.0261819079807012e-06, + "loss": 0.917, + "step": 26857 + }, + { + "epoch": 0.86, + "learning_rate": 1.0257249026564009e-06, + "loss": 0.9707, + "step": 26858 + }, + { + "epoch": 0.86, + "learning_rate": 1.0252679936161392e-06, + "loss": 0.9175, + "step": 26859 + }, + { + "epoch": 0.86, + "learning_rate": 1.0248111808648209e-06, + "loss": 0.9546, + "step": 26860 + }, + { + "epoch": 0.86, + "learning_rate": 1.0243544644073423e-06, + "loss": 0.9688, + "step": 26861 + }, + { + "epoch": 0.86, + "learning_rate": 1.0238978442486091e-06, + "loss": 0.9917, + "step": 26862 + }, + { + "epoch": 0.86, + "learning_rate": 1.0234413203935167e-06, + "loss": 0.9365, + "step": 26863 + }, + { + "epoch": 0.86, + "learning_rate": 1.0229848928469654e-06, + "loss": 0.8198, + "step": 26864 + }, + { + "epoch": 0.86, + "learning_rate": 1.0225285616138515e-06, + "loss": 1.0498, + "step": 26865 + }, + { + "epoch": 0.86, + "learning_rate": 1.0220723266990695e-06, + "loss": 0.959, + "step": 26866 + }, + { + "epoch": 0.86, + "learning_rate": 1.0216161881075136e-06, + "loss": 0.8613, + "step": 26867 + }, + { + "epoch": 0.86, + "learning_rate": 1.0211601458440813e-06, + "loss": 0.8213, + "step": 26868 + }, + { + "epoch": 0.86, + "learning_rate": 1.0207041999136613e-06, + "loss": 0.9546, + "step": 26869 + }, + { + "epoch": 0.86, + "learning_rate": 1.0202483503211492e-06, + "loss": 0.9429, + "step": 26870 + }, + { + "epoch": 0.86, + "learning_rate": 1.0197925970714317e-06, + "loss": 0.791, + "step": 26871 + }, + { + "epoch": 0.86, + "learning_rate": 1.0193369401694054e-06, + "loss": 0.9819, + "step": 26872 + }, + { + "epoch": 0.86, + "learning_rate": 1.0188813796199504e-06, + "loss": 0.8716, + "step": 26873 + }, + { + "epoch": 0.86, + "learning_rate": 1.0184259154279608e-06, + "loss": 0.9092, + "step": 26874 + }, + { + "epoch": 0.86, + "learning_rate": 1.017970547598318e-06, + "loss": 0.8491, + "step": 26875 + }, + { + "epoch": 0.86, + "learning_rate": 1.0175152761359119e-06, + "loss": 0.8257, + "step": 26876 + }, + { + "epoch": 0.86, + "learning_rate": 1.017060101045625e-06, + "loss": 0.8838, + "step": 26877 + }, + { + "epoch": 0.86, + "learning_rate": 1.016605022332343e-06, + "loss": 0.9604, + "step": 26878 + }, + { + "epoch": 0.86, + "learning_rate": 1.016150040000944e-06, + "loss": 0.9155, + "step": 26879 + }, + { + "epoch": 0.86, + "learning_rate": 1.0156951540563176e-06, + "loss": 0.9917, + "step": 26880 + }, + { + "epoch": 0.86, + "learning_rate": 1.0152403645033338e-06, + "loss": 1.0503, + "step": 26881 + }, + { + "epoch": 0.86, + "learning_rate": 1.0147856713468795e-06, + "loss": 0.8408, + "step": 26882 + }, + { + "epoch": 0.86, + "learning_rate": 1.0143310745918279e-06, + "loss": 0.8394, + "step": 26883 + }, + { + "epoch": 0.86, + "learning_rate": 1.0138765742430613e-06, + "loss": 0.9702, + "step": 26884 + }, + { + "epoch": 0.86, + "learning_rate": 1.0134221703054524e-06, + "loss": 0.9482, + "step": 26885 + }, + { + "epoch": 0.86, + "learning_rate": 1.012967862783879e-06, + "loss": 0.9502, + "step": 26886 + }, + { + "epoch": 0.86, + "learning_rate": 1.0125136516832146e-06, + "loss": 0.9424, + "step": 26887 + }, + { + "epoch": 0.86, + "learning_rate": 1.012059537008332e-06, + "loss": 1.0161, + "step": 26888 + }, + { + "epoch": 0.86, + "learning_rate": 1.0116055187641005e-06, + "loss": 0.9775, + "step": 26889 + }, + { + "epoch": 0.86, + "learning_rate": 1.0111515969553965e-06, + "loss": 0.8076, + "step": 26890 + }, + { + "epoch": 0.86, + "learning_rate": 1.0106977715870847e-06, + "loss": 0.8579, + "step": 26891 + }, + { + "epoch": 0.86, + "learning_rate": 1.0102440426640393e-06, + "loss": 0.9136, + "step": 26892 + }, + { + "epoch": 0.86, + "learning_rate": 1.009790410191126e-06, + "loss": 0.9673, + "step": 26893 + }, + { + "epoch": 0.86, + "learning_rate": 1.0093368741732101e-06, + "loss": 0.469, + "step": 26894 + }, + { + "epoch": 0.86, + "learning_rate": 1.008883434615161e-06, + "loss": 0.8633, + "step": 26895 + }, + { + "epoch": 0.86, + "learning_rate": 1.0084300915218415e-06, + "loss": 0.9189, + "step": 26896 + }, + { + "epoch": 0.86, + "learning_rate": 1.0079768448981141e-06, + "loss": 0.9316, + "step": 26897 + }, + { + "epoch": 0.86, + "learning_rate": 1.0075236947488454e-06, + "loss": 0.9175, + "step": 26898 + }, + { + "epoch": 0.86, + "learning_rate": 1.0070706410788943e-06, + "loss": 0.8462, + "step": 26899 + }, + { + "epoch": 0.86, + "learning_rate": 1.0066176838931218e-06, + "loss": 0.8657, + "step": 26900 + }, + { + "epoch": 0.86, + "learning_rate": 1.0061648231963894e-06, + "loss": 0.8862, + "step": 26901 + }, + { + "epoch": 0.86, + "learning_rate": 1.0057120589935521e-06, + "loss": 0.5156, + "step": 26902 + }, + { + "epoch": 0.86, + "learning_rate": 1.0052593912894736e-06, + "loss": 0.9878, + "step": 26903 + }, + { + "epoch": 0.86, + "learning_rate": 1.0048068200890037e-06, + "loss": 0.9536, + "step": 26904 + }, + { + "epoch": 0.86, + "learning_rate": 1.0043543453970039e-06, + "loss": 0.873, + "step": 26905 + }, + { + "epoch": 0.86, + "learning_rate": 1.0039019672183237e-06, + "loss": 0.9434, + "step": 26906 + }, + { + "epoch": 0.86, + "learning_rate": 1.0034496855578202e-06, + "loss": 0.9111, + "step": 26907 + }, + { + "epoch": 0.86, + "learning_rate": 1.0029975004203429e-06, + "loss": 0.9517, + "step": 26908 + }, + { + "epoch": 0.86, + "learning_rate": 1.002545411810748e-06, + "loss": 1.0107, + "step": 26909 + }, + { + "epoch": 0.86, + "learning_rate": 1.0020934197338817e-06, + "loss": 0.9463, + "step": 26910 + }, + { + "epoch": 0.86, + "learning_rate": 1.0016415241945954e-06, + "loss": 0.8154, + "step": 26911 + }, + { + "epoch": 0.86, + "learning_rate": 1.0011897251977343e-06, + "loss": 0.9434, + "step": 26912 + }, + { + "epoch": 0.86, + "learning_rate": 1.00073802274815e-06, + "loss": 0.9199, + "step": 26913 + }, + { + "epoch": 0.86, + "learning_rate": 1.0002864168506854e-06, + "loss": 0.9331, + "step": 26914 + }, + { + "epoch": 0.86, + "learning_rate": 9.998349075101887e-07, + "loss": 0.9648, + "step": 26915 + }, + { + "epoch": 0.86, + "learning_rate": 9.993834947315006e-07, + "loss": 0.9199, + "step": 26916 + }, + { + "epoch": 0.86, + "learning_rate": 9.989321785194683e-07, + "loss": 0.8989, + "step": 26917 + }, + { + "epoch": 0.86, + "learning_rate": 9.984809588789312e-07, + "loss": 0.9487, + "step": 26918 + }, + { + "epoch": 0.86, + "learning_rate": 9.980298358147323e-07, + "loss": 0.8232, + "step": 26919 + }, + { + "epoch": 0.86, + "learning_rate": 9.975788093317073e-07, + "loss": 0.8921, + "step": 26920 + }, + { + "epoch": 0.86, + "learning_rate": 9.971278794347007e-07, + "loss": 0.9067, + "step": 26921 + }, + { + "epoch": 0.86, + "learning_rate": 9.966770461285458e-07, + "loss": 0.8398, + "step": 26922 + }, + { + "epoch": 0.86, + "learning_rate": 9.962263094180846e-07, + "loss": 0.9561, + "step": 26923 + }, + { + "epoch": 0.86, + "learning_rate": 9.957756693081478e-07, + "loss": 0.9238, + "step": 26924 + }, + { + "epoch": 0.86, + "learning_rate": 9.953251258035757e-07, + "loss": 0.9531, + "step": 26925 + }, + { + "epoch": 0.86, + "learning_rate": 9.948746789091967e-07, + "loss": 0.8586, + "step": 26926 + }, + { + "epoch": 0.86, + "learning_rate": 9.944243286298473e-07, + "loss": 0.4868, + "step": 26927 + }, + { + "epoch": 0.86, + "learning_rate": 9.939740749703553e-07, + "loss": 0.9043, + "step": 26928 + }, + { + "epoch": 0.86, + "learning_rate": 9.935239179355571e-07, + "loss": 1.0342, + "step": 26929 + }, + { + "epoch": 0.86, + "learning_rate": 9.930738575302767e-07, + "loss": 0.9287, + "step": 26930 + }, + { + "epoch": 0.86, + "learning_rate": 9.926238937593479e-07, + "loss": 1.0161, + "step": 26931 + }, + { + "epoch": 0.86, + "learning_rate": 9.92174026627596e-07, + "loss": 0.8022, + "step": 26932 + }, + { + "epoch": 0.86, + "learning_rate": 9.917242561398465e-07, + "loss": 0.9146, + "step": 26933 + }, + { + "epoch": 0.86, + "learning_rate": 9.91274582300925e-07, + "loss": 0.979, + "step": 26934 + }, + { + "epoch": 0.86, + "learning_rate": 9.908250051156586e-07, + "loss": 1.0078, + "step": 26935 + }, + { + "epoch": 0.86, + "learning_rate": 9.903755245888669e-07, + "loss": 0.9932, + "step": 26936 + }, + { + "epoch": 0.86, + "learning_rate": 9.899261407253757e-07, + "loss": 0.8081, + "step": 26937 + }, + { + "epoch": 0.86, + "learning_rate": 9.894768535300059e-07, + "loss": 0.811, + "step": 26938 + }, + { + "epoch": 0.86, + "learning_rate": 9.890276630075757e-07, + "loss": 0.8403, + "step": 26939 + }, + { + "epoch": 0.86, + "learning_rate": 9.88578569162907e-07, + "loss": 0.9717, + "step": 26940 + }, + { + "epoch": 0.86, + "learning_rate": 9.881295720008166e-07, + "loss": 0.8262, + "step": 26941 + }, + { + "epoch": 0.86, + "learning_rate": 9.876806715261222e-07, + "loss": 0.9717, + "step": 26942 + }, + { + "epoch": 0.86, + "learning_rate": 9.872318677436387e-07, + "loss": 0.9321, + "step": 26943 + }, + { + "epoch": 0.86, + "learning_rate": 9.867831606581835e-07, + "loss": 0.8296, + "step": 26944 + }, + { + "epoch": 0.86, + "learning_rate": 9.86334550274568e-07, + "loss": 1.0557, + "step": 26945 + }, + { + "epoch": 0.86, + "learning_rate": 9.858860365976098e-07, + "loss": 0.8921, + "step": 26946 + }, + { + "epoch": 0.86, + "learning_rate": 9.854376196321148e-07, + "loss": 0.9878, + "step": 26947 + }, + { + "epoch": 0.86, + "learning_rate": 9.849892993829014e-07, + "loss": 0.8823, + "step": 26948 + }, + { + "epoch": 0.86, + "learning_rate": 9.845410758547724e-07, + "loss": 0.9048, + "step": 26949 + }, + { + "epoch": 0.86, + "learning_rate": 9.840929490525408e-07, + "loss": 0.9019, + "step": 26950 + }, + { + "epoch": 0.86, + "learning_rate": 9.836449189810127e-07, + "loss": 1.04, + "step": 26951 + }, + { + "epoch": 0.86, + "learning_rate": 9.831969856449975e-07, + "loss": 0.7834, + "step": 26952 + }, + { + "epoch": 0.86, + "learning_rate": 9.82749149049297e-07, + "loss": 1.0171, + "step": 26953 + }, + { + "epoch": 0.86, + "learning_rate": 9.823014091987203e-07, + "loss": 0.9302, + "step": 26954 + }, + { + "epoch": 0.86, + "learning_rate": 9.818537660980666e-07, + "loss": 0.9482, + "step": 26955 + }, + { + "epoch": 0.86, + "learning_rate": 9.814062197521456e-07, + "loss": 0.9561, + "step": 26956 + }, + { + "epoch": 0.86, + "learning_rate": 9.80958770165752e-07, + "loss": 1.0522, + "step": 26957 + }, + { + "epoch": 0.86, + "learning_rate": 9.805114173436891e-07, + "loss": 0.8252, + "step": 26958 + }, + { + "epoch": 0.86, + "learning_rate": 9.800641612907558e-07, + "loss": 0.9443, + "step": 26959 + }, + { + "epoch": 0.86, + "learning_rate": 9.796170020117523e-07, + "loss": 0.9399, + "step": 26960 + }, + { + "epoch": 0.86, + "learning_rate": 9.791699395114728e-07, + "loss": 0.8418, + "step": 26961 + }, + { + "epoch": 0.86, + "learning_rate": 9.78722973794719e-07, + "loss": 0.897, + "step": 26962 + }, + { + "epoch": 0.86, + "learning_rate": 9.782761048662836e-07, + "loss": 0.9224, + "step": 26963 + }, + { + "epoch": 0.86, + "learning_rate": 9.778293327309607e-07, + "loss": 0.8525, + "step": 26964 + }, + { + "epoch": 0.86, + "learning_rate": 9.773826573935409e-07, + "loss": 0.9746, + "step": 26965 + }, + { + "epoch": 0.86, + "learning_rate": 9.769360788588223e-07, + "loss": 0.8843, + "step": 26966 + }, + { + "epoch": 0.86, + "learning_rate": 9.764895971315914e-07, + "loss": 0.9106, + "step": 26967 + }, + { + "epoch": 0.86, + "learning_rate": 9.76043212216642e-07, + "loss": 0.8337, + "step": 26968 + }, + { + "epoch": 0.86, + "learning_rate": 9.755969241187603e-07, + "loss": 1.0332, + "step": 26969 + }, + { + "epoch": 0.86, + "learning_rate": 9.751507328427368e-07, + "loss": 1.0029, + "step": 26970 + }, + { + "epoch": 0.86, + "learning_rate": 9.74704638393359e-07, + "loss": 0.8672, + "step": 26971 + }, + { + "epoch": 0.86, + "learning_rate": 9.742586407754106e-07, + "loss": 0.9751, + "step": 26972 + }, + { + "epoch": 0.86, + "learning_rate": 9.738127399936758e-07, + "loss": 0.9521, + "step": 26973 + }, + { + "epoch": 0.86, + "learning_rate": 9.73366936052943e-07, + "loss": 0.9263, + "step": 26974 + }, + { + "epoch": 0.86, + "learning_rate": 9.729212289579915e-07, + "loss": 0.9639, + "step": 26975 + }, + { + "epoch": 0.86, + "learning_rate": 9.724756187136052e-07, + "loss": 0.9326, + "step": 26976 + }, + { + "epoch": 0.86, + "learning_rate": 9.72030105324565e-07, + "loss": 0.8262, + "step": 26977 + }, + { + "epoch": 0.86, + "learning_rate": 9.715846887956503e-07, + "loss": 0.9312, + "step": 26978 + }, + { + "epoch": 0.86, + "learning_rate": 9.711393691316383e-07, + "loss": 0.8916, + "step": 26979 + }, + { + "epoch": 0.86, + "learning_rate": 9.706941463373087e-07, + "loss": 0.9297, + "step": 26980 + }, + { + "epoch": 0.86, + "learning_rate": 9.702490204174386e-07, + "loss": 1.0034, + "step": 26981 + }, + { + "epoch": 0.86, + "learning_rate": 9.69803991376801e-07, + "loss": 0.9585, + "step": 26982 + }, + { + "epoch": 0.86, + "learning_rate": 9.693590592201752e-07, + "loss": 0.8818, + "step": 26983 + }, + { + "epoch": 0.86, + "learning_rate": 9.689142239523287e-07, + "loss": 1.001, + "step": 26984 + }, + { + "epoch": 0.86, + "learning_rate": 9.684694855780397e-07, + "loss": 0.8828, + "step": 26985 + }, + { + "epoch": 0.86, + "learning_rate": 9.680248441020779e-07, + "loss": 0.8462, + "step": 26986 + }, + { + "epoch": 0.86, + "learning_rate": 9.675802995292138e-07, + "loss": 0.8589, + "step": 26987 + }, + { + "epoch": 0.86, + "learning_rate": 9.671358518642137e-07, + "loss": 0.9521, + "step": 26988 + }, + { + "epoch": 0.86, + "learning_rate": 9.666915011118505e-07, + "loss": 0.9648, + "step": 26989 + }, + { + "epoch": 0.86, + "learning_rate": 9.66247247276889e-07, + "loss": 1.0005, + "step": 26990 + }, + { + "epoch": 0.86, + "learning_rate": 9.65803090364098e-07, + "loss": 0.8921, + "step": 26991 + }, + { + "epoch": 0.86, + "learning_rate": 9.65359030378239e-07, + "loss": 0.9487, + "step": 26992 + }, + { + "epoch": 0.86, + "learning_rate": 9.649150673240837e-07, + "loss": 0.9014, + "step": 26993 + }, + { + "epoch": 0.86, + "learning_rate": 9.64471201206385e-07, + "loss": 0.8696, + "step": 26994 + }, + { + "epoch": 0.86, + "learning_rate": 9.640274320299125e-07, + "loss": 0.8911, + "step": 26995 + }, + { + "epoch": 0.86, + "learning_rate": 9.635837597994236e-07, + "loss": 0.9229, + "step": 26996 + }, + { + "epoch": 0.86, + "learning_rate": 9.63140184519682e-07, + "loss": 0.8936, + "step": 26997 + }, + { + "epoch": 0.86, + "learning_rate": 9.626967061954417e-07, + "loss": 0.9478, + "step": 26998 + }, + { + "epoch": 0.86, + "learning_rate": 9.62253324831467e-07, + "loss": 0.875, + "step": 26999 + }, + { + "epoch": 0.86, + "learning_rate": 9.618100404325082e-07, + "loss": 0.8394, + "step": 27000 + }, + { + "epoch": 0.86, + "learning_rate": 9.613668530033294e-07, + "loss": 0.8896, + "step": 27001 + }, + { + "epoch": 0.86, + "learning_rate": 9.609237625486779e-07, + "loss": 0.8994, + "step": 27002 + }, + { + "epoch": 0.86, + "learning_rate": 9.604807690733109e-07, + "loss": 0.8511, + "step": 27003 + }, + { + "epoch": 0.86, + "learning_rate": 9.600378725819793e-07, + "loss": 0.8589, + "step": 27004 + }, + { + "epoch": 0.86, + "learning_rate": 9.59595073079439e-07, + "loss": 0.8525, + "step": 27005 + }, + { + "epoch": 0.86, + "learning_rate": 9.591523705704353e-07, + "loss": 0.8604, + "step": 27006 + }, + { + "epoch": 0.86, + "learning_rate": 9.587097650597232e-07, + "loss": 0.8118, + "step": 27007 + }, + { + "epoch": 0.86, + "learning_rate": 9.58267256552049e-07, + "loss": 0.7664, + "step": 27008 + }, + { + "epoch": 0.86, + "learning_rate": 9.578248450521598e-07, + "loss": 1.0112, + "step": 27009 + }, + { + "epoch": 0.86, + "learning_rate": 9.57382530564801e-07, + "loss": 0.8906, + "step": 27010 + }, + { + "epoch": 0.86, + "learning_rate": 9.56940313094722e-07, + "loss": 0.8799, + "step": 27011 + }, + { + "epoch": 0.86, + "learning_rate": 9.56498192646662e-07, + "loss": 0.9453, + "step": 27012 + }, + { + "epoch": 0.86, + "learning_rate": 9.56056169225371e-07, + "loss": 0.9326, + "step": 27013 + }, + { + "epoch": 0.86, + "learning_rate": 9.556142428355853e-07, + "loss": 0.873, + "step": 27014 + }, + { + "epoch": 0.86, + "learning_rate": 9.551724134820506e-07, + "loss": 0.7852, + "step": 27015 + }, + { + "epoch": 0.86, + "learning_rate": 9.547306811695056e-07, + "loss": 0.7456, + "step": 27016 + }, + { + "epoch": 0.86, + "learning_rate": 9.542890459026899e-07, + "loss": 0.8574, + "step": 27017 + }, + { + "epoch": 0.86, + "learning_rate": 9.538475076863395e-07, + "loss": 0.8281, + "step": 27018 + }, + { + "epoch": 0.86, + "learning_rate": 9.534060665251954e-07, + "loss": 0.9917, + "step": 27019 + }, + { + "epoch": 0.86, + "learning_rate": 9.529647224239924e-07, + "loss": 0.8706, + "step": 27020 + }, + { + "epoch": 0.86, + "learning_rate": 9.525234753874623e-07, + "loss": 0.9707, + "step": 27021 + }, + { + "epoch": 0.86, + "learning_rate": 9.520823254203449e-07, + "loss": 0.9277, + "step": 27022 + }, + { + "epoch": 0.86, + "learning_rate": 9.516412725273694e-07, + "loss": 0.999, + "step": 27023 + }, + { + "epoch": 0.86, + "learning_rate": 9.512003167132678e-07, + "loss": 0.8384, + "step": 27024 + }, + { + "epoch": 0.86, + "learning_rate": 9.50759457982774e-07, + "loss": 0.9014, + "step": 27025 + }, + { + "epoch": 0.86, + "learning_rate": 9.503186963406152e-07, + "loss": 0.7532, + "step": 27026 + }, + { + "epoch": 0.86, + "learning_rate": 9.498780317915201e-07, + "loss": 0.7964, + "step": 27027 + }, + { + "epoch": 0.86, + "learning_rate": 9.494374643402182e-07, + "loss": 0.7898, + "step": 27028 + }, + { + "epoch": 0.86, + "learning_rate": 9.489969939914356e-07, + "loss": 0.9102, + "step": 27029 + }, + { + "epoch": 0.86, + "learning_rate": 9.485566207498986e-07, + "loss": 0.8572, + "step": 27030 + }, + { + "epoch": 0.86, + "learning_rate": 9.481163446203312e-07, + "loss": 0.9443, + "step": 27031 + }, + { + "epoch": 0.86, + "learning_rate": 9.476761656074585e-07, + "loss": 0.9072, + "step": 27032 + }, + { + "epoch": 0.86, + "learning_rate": 9.472360837159989e-07, + "loss": 0.9873, + "step": 27033 + }, + { + "epoch": 0.86, + "learning_rate": 9.467960989506797e-07, + "loss": 0.96, + "step": 27034 + }, + { + "epoch": 0.86, + "learning_rate": 9.463562113162172e-07, + "loss": 0.9741, + "step": 27035 + }, + { + "epoch": 0.86, + "learning_rate": 9.459164208173333e-07, + "loss": 0.9385, + "step": 27036 + }, + { + "epoch": 0.86, + "learning_rate": 9.45476727458744e-07, + "loss": 0.9204, + "step": 27037 + }, + { + "epoch": 0.86, + "learning_rate": 9.450371312451711e-07, + "loss": 0.9141, + "step": 27038 + }, + { + "epoch": 0.86, + "learning_rate": 9.445976321813277e-07, + "loss": 0.8145, + "step": 27039 + }, + { + "epoch": 0.86, + "learning_rate": 9.441582302719288e-07, + "loss": 0.8555, + "step": 27040 + }, + { + "epoch": 0.86, + "learning_rate": 9.437189255216883e-07, + "loss": 0.8809, + "step": 27041 + }, + { + "epoch": 0.86, + "learning_rate": 9.432797179353237e-07, + "loss": 0.8887, + "step": 27042 + }, + { + "epoch": 0.86, + "learning_rate": 9.428406075175412e-07, + "loss": 0.9868, + "step": 27043 + }, + { + "epoch": 0.86, + "learning_rate": 9.424015942730569e-07, + "loss": 0.8774, + "step": 27044 + }, + { + "epoch": 0.86, + "learning_rate": 9.419626782065771e-07, + "loss": 0.9121, + "step": 27045 + }, + { + "epoch": 0.86, + "learning_rate": 9.415238593228171e-07, + "loss": 0.96, + "step": 27046 + }, + { + "epoch": 0.86, + "learning_rate": 9.410851376264774e-07, + "loss": 0.895, + "step": 27047 + }, + { + "epoch": 0.87, + "learning_rate": 9.406465131222687e-07, + "loss": 0.8755, + "step": 27048 + }, + { + "epoch": 0.87, + "learning_rate": 9.402079858148961e-07, + "loss": 0.9614, + "step": 27049 + }, + { + "epoch": 0.87, + "learning_rate": 9.397695557090658e-07, + "loss": 0.8877, + "step": 27050 + }, + { + "epoch": 0.87, + "learning_rate": 9.393312228094798e-07, + "loss": 1.0552, + "step": 27051 + }, + { + "epoch": 0.87, + "learning_rate": 9.38892987120843e-07, + "loss": 0.9014, + "step": 27052 + }, + { + "epoch": 0.87, + "learning_rate": 9.38454848647854e-07, + "loss": 1.0298, + "step": 27053 + }, + { + "epoch": 0.87, + "learning_rate": 9.3801680739522e-07, + "loss": 0.877, + "step": 27054 + }, + { + "epoch": 0.87, + "learning_rate": 9.375788633676331e-07, + "loss": 0.9878, + "step": 27055 + }, + { + "epoch": 0.87, + "learning_rate": 9.371410165697958e-07, + "loss": 0.9312, + "step": 27056 + }, + { + "epoch": 0.87, + "learning_rate": 9.367032670064035e-07, + "loss": 0.4778, + "step": 27057 + }, + { + "epoch": 0.87, + "learning_rate": 9.36265614682157e-07, + "loss": 0.9023, + "step": 27058 + }, + { + "epoch": 0.87, + "learning_rate": 9.358280596017489e-07, + "loss": 0.8887, + "step": 27059 + }, + { + "epoch": 0.87, + "learning_rate": 9.353906017698721e-07, + "loss": 0.8638, + "step": 27060 + }, + { + "epoch": 0.87, + "learning_rate": 9.349532411912243e-07, + "loss": 0.9663, + "step": 27061 + }, + { + "epoch": 0.87, + "learning_rate": 9.345159778704949e-07, + "loss": 0.9355, + "step": 27062 + }, + { + "epoch": 0.87, + "learning_rate": 9.340788118123745e-07, + "loss": 0.875, + "step": 27063 + }, + { + "epoch": 0.87, + "learning_rate": 9.336417430215561e-07, + "loss": 0.9902, + "step": 27064 + }, + { + "epoch": 0.87, + "learning_rate": 9.33204771502727e-07, + "loss": 1.0049, + "step": 27065 + }, + { + "epoch": 0.87, + "learning_rate": 9.327678972605758e-07, + "loss": 0.8599, + "step": 27066 + }, + { + "epoch": 0.87, + "learning_rate": 9.323311202997909e-07, + "loss": 0.7798, + "step": 27067 + }, + { + "epoch": 0.87, + "learning_rate": 9.318944406250552e-07, + "loss": 0.8408, + "step": 27068 + }, + { + "epoch": 0.87, + "learning_rate": 9.314578582410593e-07, + "loss": 0.875, + "step": 27069 + }, + { + "epoch": 0.87, + "learning_rate": 9.310213731524831e-07, + "loss": 1.0176, + "step": 27070 + }, + { + "epoch": 0.87, + "learning_rate": 9.305849853640103e-07, + "loss": 0.8872, + "step": 27071 + }, + { + "epoch": 0.87, + "learning_rate": 9.301486948803218e-07, + "loss": 0.9004, + "step": 27072 + }, + { + "epoch": 0.87, + "learning_rate": 9.297125017061015e-07, + "loss": 0.9526, + "step": 27073 + }, + { + "epoch": 0.87, + "learning_rate": 9.292764058460257e-07, + "loss": 0.7305, + "step": 27074 + }, + { + "epoch": 0.87, + "learning_rate": 9.288404073047763e-07, + "loss": 0.9775, + "step": 27075 + }, + { + "epoch": 0.87, + "learning_rate": 9.284045060870306e-07, + "loss": 0.9155, + "step": 27076 + }, + { + "epoch": 0.87, + "learning_rate": 9.279687021974648e-07, + "loss": 0.896, + "step": 27077 + }, + { + "epoch": 0.87, + "learning_rate": 9.275329956407519e-07, + "loss": 1.0249, + "step": 27078 + }, + { + "epoch": 0.87, + "learning_rate": 9.270973864215715e-07, + "loss": 0.9907, + "step": 27079 + }, + { + "epoch": 0.87, + "learning_rate": 9.266618745445921e-07, + "loss": 1.0229, + "step": 27080 + }, + { + "epoch": 0.87, + "learning_rate": 9.26226460014491e-07, + "loss": 0.8926, + "step": 27081 + }, + { + "epoch": 0.87, + "learning_rate": 9.257911428359368e-07, + "loss": 0.9937, + "step": 27082 + }, + { + "epoch": 0.87, + "learning_rate": 9.253559230136011e-07, + "loss": 0.926, + "step": 27083 + }, + { + "epoch": 0.87, + "learning_rate": 9.249208005521538e-07, + "loss": 0.9375, + "step": 27084 + }, + { + "epoch": 0.87, + "learning_rate": 9.24485775456263e-07, + "loss": 1.0117, + "step": 27085 + }, + { + "epoch": 0.87, + "learning_rate": 9.240508477305932e-07, + "loss": 0.9834, + "step": 27086 + }, + { + "epoch": 0.87, + "learning_rate": 9.236160173798148e-07, + "loss": 0.8867, + "step": 27087 + }, + { + "epoch": 0.87, + "learning_rate": 9.231812844085897e-07, + "loss": 0.8486, + "step": 27088 + }, + { + "epoch": 0.87, + "learning_rate": 9.227466488215853e-07, + "loss": 0.9028, + "step": 27089 + }, + { + "epoch": 0.87, + "learning_rate": 9.223121106234612e-07, + "loss": 0.79, + "step": 27090 + }, + { + "epoch": 0.87, + "learning_rate": 9.218776698188847e-07, + "loss": 0.8394, + "step": 27091 + }, + { + "epoch": 0.87, + "learning_rate": 9.214433264125111e-07, + "loss": 0.8511, + "step": 27092 + }, + { + "epoch": 0.87, + "learning_rate": 9.210090804090044e-07, + "loss": 0.8457, + "step": 27093 + }, + { + "epoch": 0.87, + "learning_rate": 9.205749318130197e-07, + "loss": 0.8975, + "step": 27094 + }, + { + "epoch": 0.87, + "learning_rate": 9.201408806292201e-07, + "loss": 0.7603, + "step": 27095 + }, + { + "epoch": 0.87, + "learning_rate": 9.197069268622572e-07, + "loss": 0.939, + "step": 27096 + }, + { + "epoch": 0.87, + "learning_rate": 9.192730705167918e-07, + "loss": 0.9702, + "step": 27097 + }, + { + "epoch": 0.87, + "learning_rate": 9.188393115974758e-07, + "loss": 0.9324, + "step": 27098 + }, + { + "epoch": 0.87, + "learning_rate": 9.184056501089633e-07, + "loss": 0.9033, + "step": 27099 + }, + { + "epoch": 0.87, + "learning_rate": 9.179720860559061e-07, + "loss": 0.9033, + "step": 27100 + }, + { + "epoch": 0.87, + "learning_rate": 9.175386194429581e-07, + "loss": 0.9502, + "step": 27101 + }, + { + "epoch": 0.87, + "learning_rate": 9.171052502747679e-07, + "loss": 0.957, + "step": 27102 + }, + { + "epoch": 0.87, + "learning_rate": 9.166719785559863e-07, + "loss": 0.9116, + "step": 27103 + }, + { + "epoch": 0.87, + "learning_rate": 9.162388042912629e-07, + "loss": 0.8213, + "step": 27104 + }, + { + "epoch": 0.87, + "learning_rate": 9.158057274852405e-07, + "loss": 0.9009, + "step": 27105 + }, + { + "epoch": 0.87, + "learning_rate": 9.153727481425723e-07, + "loss": 0.8979, + "step": 27106 + }, + { + "epoch": 0.87, + "learning_rate": 9.149398662678987e-07, + "loss": 0.8618, + "step": 27107 + }, + { + "epoch": 0.87, + "learning_rate": 9.145070818658641e-07, + "loss": 0.9614, + "step": 27108 + }, + { + "epoch": 0.87, + "learning_rate": 9.140743949411146e-07, + "loss": 0.9971, + "step": 27109 + }, + { + "epoch": 0.87, + "learning_rate": 9.136418054982921e-07, + "loss": 0.959, + "step": 27110 + }, + { + "epoch": 0.87, + "learning_rate": 9.132093135420339e-07, + "loss": 0.8213, + "step": 27111 + }, + { + "epoch": 0.87, + "learning_rate": 9.127769190769853e-07, + "loss": 1.001, + "step": 27112 + }, + { + "epoch": 0.87, + "learning_rate": 9.123446221077814e-07, + "loss": 0.45, + "step": 27113 + }, + { + "epoch": 0.87, + "learning_rate": 9.11912422639063e-07, + "loss": 0.9321, + "step": 27114 + }, + { + "epoch": 0.87, + "learning_rate": 9.114803206754664e-07, + "loss": 0.9338, + "step": 27115 + }, + { + "epoch": 0.87, + "learning_rate": 9.110483162216277e-07, + "loss": 0.916, + "step": 27116 + }, + { + "epoch": 0.87, + "learning_rate": 9.106164092821801e-07, + "loss": 0.8965, + "step": 27117 + }, + { + "epoch": 0.87, + "learning_rate": 9.101845998617598e-07, + "loss": 0.9023, + "step": 27118 + }, + { + "epoch": 0.87, + "learning_rate": 9.097528879649964e-07, + "loss": 0.9365, + "step": 27119 + }, + { + "epoch": 0.87, + "learning_rate": 9.093212735965262e-07, + "loss": 0.9141, + "step": 27120 + }, + { + "epoch": 0.87, + "learning_rate": 9.088897567609767e-07, + "loss": 0.9326, + "step": 27121 + }, + { + "epoch": 0.87, + "learning_rate": 9.084583374629807e-07, + "loss": 0.8467, + "step": 27122 + }, + { + "epoch": 0.87, + "learning_rate": 9.080270157071625e-07, + "loss": 0.9331, + "step": 27123 + }, + { + "epoch": 0.87, + "learning_rate": 9.075957914981525e-07, + "loss": 0.8711, + "step": 27124 + }, + { + "epoch": 0.87, + "learning_rate": 9.071646648405763e-07, + "loss": 0.9456, + "step": 27125 + }, + { + "epoch": 0.87, + "learning_rate": 9.06733635739061e-07, + "loss": 0.9751, + "step": 27126 + }, + { + "epoch": 0.87, + "learning_rate": 9.063027041982275e-07, + "loss": 0.8726, + "step": 27127 + }, + { + "epoch": 0.87, + "learning_rate": 9.058718702227032e-07, + "loss": 0.4614, + "step": 27128 + }, + { + "epoch": 0.87, + "learning_rate": 9.054411338171099e-07, + "loss": 0.9185, + "step": 27129 + }, + { + "epoch": 0.87, + "learning_rate": 9.050104949860672e-07, + "loss": 0.748, + "step": 27130 + }, + { + "epoch": 0.87, + "learning_rate": 9.045799537341948e-07, + "loss": 0.9189, + "step": 27131 + }, + { + "epoch": 0.87, + "learning_rate": 9.041495100661146e-07, + "loss": 0.9253, + "step": 27132 + }, + { + "epoch": 0.87, + "learning_rate": 9.037191639864407e-07, + "loss": 0.8989, + "step": 27133 + }, + { + "epoch": 0.87, + "learning_rate": 9.032889154997959e-07, + "loss": 0.9683, + "step": 27134 + }, + { + "epoch": 0.87, + "learning_rate": 9.028587646107911e-07, + "loss": 0.9248, + "step": 27135 + }, + { + "epoch": 0.87, + "learning_rate": 9.02428711324046e-07, + "loss": 0.8203, + "step": 27136 + }, + { + "epoch": 0.87, + "learning_rate": 9.019987556441711e-07, + "loss": 0.8281, + "step": 27137 + }, + { + "epoch": 0.87, + "learning_rate": 9.015688975757808e-07, + "loss": 0.9404, + "step": 27138 + }, + { + "epoch": 0.87, + "learning_rate": 9.011391371234845e-07, + "loss": 0.9258, + "step": 27139 + }, + { + "epoch": 0.87, + "learning_rate": 9.007094742918976e-07, + "loss": 0.875, + "step": 27140 + }, + { + "epoch": 0.87, + "learning_rate": 9.002799090856251e-07, + "loss": 0.9331, + "step": 27141 + }, + { + "epoch": 0.87, + "learning_rate": 8.998504415092801e-07, + "loss": 0.9175, + "step": 27142 + }, + { + "epoch": 0.87, + "learning_rate": 8.994210715674678e-07, + "loss": 0.9248, + "step": 27143 + }, + { + "epoch": 0.87, + "learning_rate": 8.989917992647968e-07, + "loss": 0.8911, + "step": 27144 + }, + { + "epoch": 0.87, + "learning_rate": 8.985626246058676e-07, + "loss": 0.9033, + "step": 27145 + }, + { + "epoch": 0.87, + "learning_rate": 8.981335475952924e-07, + "loss": 0.854, + "step": 27146 + }, + { + "epoch": 0.87, + "learning_rate": 8.977045682376672e-07, + "loss": 0.8994, + "step": 27147 + }, + { + "epoch": 0.87, + "learning_rate": 8.972756865376009e-07, + "loss": 0.8008, + "step": 27148 + }, + { + "epoch": 0.87, + "learning_rate": 8.968469024996918e-07, + "loss": 0.9746, + "step": 27149 + }, + { + "epoch": 0.87, + "learning_rate": 8.964182161285384e-07, + "loss": 0.8613, + "step": 27150 + }, + { + "epoch": 0.87, + "learning_rate": 8.959896274287439e-07, + "loss": 0.9087, + "step": 27151 + }, + { + "epoch": 0.87, + "learning_rate": 8.955611364049055e-07, + "loss": 0.9692, + "step": 27152 + }, + { + "epoch": 0.87, + "learning_rate": 8.951327430616175e-07, + "loss": 0.959, + "step": 27153 + }, + { + "epoch": 0.87, + "learning_rate": 8.947044474034805e-07, + "loss": 0.8408, + "step": 27154 + }, + { + "epoch": 0.87, + "learning_rate": 8.942762494350877e-07, + "loss": 0.9595, + "step": 27155 + }, + { + "epoch": 0.87, + "learning_rate": 8.938481491610307e-07, + "loss": 1.02, + "step": 27156 + }, + { + "epoch": 0.87, + "learning_rate": 8.934201465859071e-07, + "loss": 0.9106, + "step": 27157 + }, + { + "epoch": 0.87, + "learning_rate": 8.929922417143045e-07, + "loss": 0.9033, + "step": 27158 + }, + { + "epoch": 0.87, + "learning_rate": 8.9256443455082e-07, + "loss": 0.8862, + "step": 27159 + }, + { + "epoch": 0.87, + "learning_rate": 8.921367251000357e-07, + "loss": 0.9731, + "step": 27160 + }, + { + "epoch": 0.87, + "learning_rate": 8.917091133665457e-07, + "loss": 0.9189, + "step": 27161 + }, + { + "epoch": 0.87, + "learning_rate": 8.912815993549351e-07, + "loss": 0.9458, + "step": 27162 + }, + { + "epoch": 0.87, + "learning_rate": 8.908541830697937e-07, + "loss": 0.8926, + "step": 27163 + }, + { + "epoch": 0.87, + "learning_rate": 8.904268645157032e-07, + "loss": 0.9487, + "step": 27164 + }, + { + "epoch": 0.87, + "learning_rate": 8.899996436972525e-07, + "loss": 0.9868, + "step": 27165 + }, + { + "epoch": 0.87, + "learning_rate": 8.89572520619022e-07, + "loss": 0.9561, + "step": 27166 + }, + { + "epoch": 0.87, + "learning_rate": 8.891454952855983e-07, + "loss": 0.9106, + "step": 27167 + }, + { + "epoch": 0.87, + "learning_rate": 8.887185677015564e-07, + "loss": 0.855, + "step": 27168 + }, + { + "epoch": 0.87, + "learning_rate": 8.882917378714828e-07, + "loss": 0.9961, + "step": 27169 + }, + { + "epoch": 0.87, + "learning_rate": 8.878650057999516e-07, + "loss": 0.9395, + "step": 27170 + }, + { + "epoch": 0.87, + "learning_rate": 8.874383714915469e-07, + "loss": 0.8091, + "step": 27171 + }, + { + "epoch": 0.87, + "learning_rate": 8.870118349508417e-07, + "loss": 0.8994, + "step": 27172 + }, + { + "epoch": 0.87, + "learning_rate": 8.865853961824144e-07, + "loss": 0.874, + "step": 27173 + }, + { + "epoch": 0.87, + "learning_rate": 8.861590551908405e-07, + "loss": 0.8965, + "step": 27174 + }, + { + "epoch": 0.87, + "learning_rate": 8.857328119806929e-07, + "loss": 0.8901, + "step": 27175 + }, + { + "epoch": 0.87, + "learning_rate": 8.853066665565424e-07, + "loss": 0.936, + "step": 27176 + }, + { + "epoch": 0.87, + "learning_rate": 8.848806189229664e-07, + "loss": 0.8882, + "step": 27177 + }, + { + "epoch": 0.87, + "learning_rate": 8.844546690845313e-07, + "loss": 0.9419, + "step": 27178 + }, + { + "epoch": 0.87, + "learning_rate": 8.840288170458111e-07, + "loss": 0.9141, + "step": 27179 + }, + { + "epoch": 0.87, + "learning_rate": 8.836030628113712e-07, + "loss": 0.9102, + "step": 27180 + }, + { + "epoch": 0.87, + "learning_rate": 8.831774063857812e-07, + "loss": 0.9697, + "step": 27181 + }, + { + "epoch": 0.87, + "learning_rate": 8.827518477736086e-07, + "loss": 0.8369, + "step": 27182 + }, + { + "epoch": 0.87, + "learning_rate": 8.823263869794185e-07, + "loss": 0.9395, + "step": 27183 + }, + { + "epoch": 0.87, + "learning_rate": 8.81901024007773e-07, + "loss": 0.9009, + "step": 27184 + }, + { + "epoch": 0.87, + "learning_rate": 8.814757588632406e-07, + "loss": 0.9399, + "step": 27185 + }, + { + "epoch": 0.87, + "learning_rate": 8.810505915503787e-07, + "loss": 1.0073, + "step": 27186 + }, + { + "epoch": 0.87, + "learning_rate": 8.806255220737536e-07, + "loss": 0.9478, + "step": 27187 + }, + { + "epoch": 0.87, + "learning_rate": 8.802005504379241e-07, + "loss": 0.906, + "step": 27188 + }, + { + "epoch": 0.87, + "learning_rate": 8.797756766474474e-07, + "loss": 0.821, + "step": 27189 + }, + { + "epoch": 0.87, + "learning_rate": 8.793509007068868e-07, + "loss": 0.8799, + "step": 27190 + }, + { + "epoch": 0.87, + "learning_rate": 8.789262226207951e-07, + "loss": 1.0225, + "step": 27191 + }, + { + "epoch": 0.87, + "learning_rate": 8.785016423937298e-07, + "loss": 0.8237, + "step": 27192 + }, + { + "epoch": 0.87, + "learning_rate": 8.780771600302474e-07, + "loss": 0.894, + "step": 27193 + }, + { + "epoch": 0.87, + "learning_rate": 8.776527755349029e-07, + "loss": 1.0132, + "step": 27194 + }, + { + "epoch": 0.87, + "learning_rate": 8.772284889122451e-07, + "loss": 0.9473, + "step": 27195 + }, + { + "epoch": 0.87, + "learning_rate": 8.768043001668303e-07, + "loss": 1.0601, + "step": 27196 + }, + { + "epoch": 0.87, + "learning_rate": 8.763802093032092e-07, + "loss": 0.9775, + "step": 27197 + }, + { + "epoch": 0.87, + "learning_rate": 8.759562163259317e-07, + "loss": 1.085, + "step": 27198 + }, + { + "epoch": 0.87, + "learning_rate": 8.755323212395428e-07, + "loss": 0.9722, + "step": 27199 + }, + { + "epoch": 0.87, + "learning_rate": 8.751085240485957e-07, + "loss": 0.8926, + "step": 27200 + }, + { + "epoch": 0.87, + "learning_rate": 8.746848247576334e-07, + "loss": 0.8643, + "step": 27201 + }, + { + "epoch": 0.87, + "learning_rate": 8.742612233712056e-07, + "loss": 0.9761, + "step": 27202 + }, + { + "epoch": 0.87, + "learning_rate": 8.73837719893853e-07, + "loss": 1.0234, + "step": 27203 + }, + { + "epoch": 0.87, + "learning_rate": 8.734143143301243e-07, + "loss": 0.8921, + "step": 27204 + }, + { + "epoch": 0.87, + "learning_rate": 8.729910066845582e-07, + "loss": 0.9429, + "step": 27205 + }, + { + "epoch": 0.87, + "learning_rate": 8.725677969616963e-07, + "loss": 0.9316, + "step": 27206 + }, + { + "epoch": 0.87, + "learning_rate": 8.721446851660797e-07, + "loss": 0.938, + "step": 27207 + }, + { + "epoch": 0.87, + "learning_rate": 8.717216713022503e-07, + "loss": 0.7964, + "step": 27208 + }, + { + "epoch": 0.87, + "learning_rate": 8.71298755374742e-07, + "loss": 0.8784, + "step": 27209 + }, + { + "epoch": 0.87, + "learning_rate": 8.708759373880982e-07, + "loss": 0.9443, + "step": 27210 + }, + { + "epoch": 0.87, + "learning_rate": 8.704532173468483e-07, + "loss": 0.8186, + "step": 27211 + }, + { + "epoch": 0.87, + "learning_rate": 8.700305952555366e-07, + "loss": 0.8667, + "step": 27212 + }, + { + "epoch": 0.87, + "learning_rate": 8.696080711186872e-07, + "loss": 0.9116, + "step": 27213 + }, + { + "epoch": 0.87, + "learning_rate": 8.691856449408409e-07, + "loss": 0.9106, + "step": 27214 + }, + { + "epoch": 0.87, + "learning_rate": 8.687633167265242e-07, + "loss": 0.9624, + "step": 27215 + }, + { + "epoch": 0.87, + "learning_rate": 8.683410864802743e-07, + "loss": 0.8916, + "step": 27216 + }, + { + "epoch": 0.87, + "learning_rate": 8.679189542066157e-07, + "loss": 1.0479, + "step": 27217 + }, + { + "epoch": 0.87, + "learning_rate": 8.674969199100813e-07, + "loss": 1.0229, + "step": 27218 + }, + { + "epoch": 0.87, + "learning_rate": 8.670749835951964e-07, + "loss": 0.897, + "step": 27219 + }, + { + "epoch": 0.87, + "learning_rate": 8.666531452664928e-07, + "loss": 0.9966, + "step": 27220 + }, + { + "epoch": 0.87, + "learning_rate": 8.662314049284881e-07, + "loss": 1.0688, + "step": 27221 + }, + { + "epoch": 0.87, + "learning_rate": 8.658097625857143e-07, + "loss": 1.0396, + "step": 27222 + }, + { + "epoch": 0.87, + "learning_rate": 8.653882182426898e-07, + "loss": 0.8789, + "step": 27223 + }, + { + "epoch": 0.87, + "learning_rate": 8.649667719039434e-07, + "loss": 0.9453, + "step": 27224 + }, + { + "epoch": 0.87, + "learning_rate": 8.645454235739903e-07, + "loss": 0.9238, + "step": 27225 + }, + { + "epoch": 0.87, + "learning_rate": 8.641241732573569e-07, + "loss": 0.8882, + "step": 27226 + }, + { + "epoch": 0.87, + "learning_rate": 8.637030209585595e-07, + "loss": 0.9966, + "step": 27227 + }, + { + "epoch": 0.87, + "learning_rate": 8.632819666821169e-07, + "loss": 0.8501, + "step": 27228 + }, + { + "epoch": 0.87, + "learning_rate": 8.628610104325453e-07, + "loss": 0.9736, + "step": 27229 + }, + { + "epoch": 0.87, + "learning_rate": 8.624401522143644e-07, + "loss": 0.894, + "step": 27230 + }, + { + "epoch": 0.87, + "learning_rate": 8.620193920320852e-07, + "loss": 0.8823, + "step": 27231 + }, + { + "epoch": 0.87, + "learning_rate": 8.615987298902273e-07, + "loss": 0.9248, + "step": 27232 + }, + { + "epoch": 0.87, + "learning_rate": 8.611781657933005e-07, + "loss": 0.9673, + "step": 27233 + }, + { + "epoch": 0.87, + "learning_rate": 8.607576997458167e-07, + "loss": 0.7983, + "step": 27234 + }, + { + "epoch": 0.87, + "learning_rate": 8.60337331752289e-07, + "loss": 0.9604, + "step": 27235 + }, + { + "epoch": 0.87, + "learning_rate": 8.599170618172259e-07, + "loss": 0.8926, + "step": 27236 + }, + { + "epoch": 0.87, + "learning_rate": 8.594968899451383e-07, + "loss": 0.9194, + "step": 27237 + }, + { + "epoch": 0.87, + "learning_rate": 8.590768161405305e-07, + "loss": 0.9448, + "step": 27238 + }, + { + "epoch": 0.87, + "learning_rate": 8.586568404079132e-07, + "loss": 1.0522, + "step": 27239 + }, + { + "epoch": 0.87, + "learning_rate": 8.582369627517894e-07, + "loss": 1.0664, + "step": 27240 + }, + { + "epoch": 0.87, + "learning_rate": 8.578171831766679e-07, + "loss": 1.0376, + "step": 27241 + }, + { + "epoch": 0.87, + "learning_rate": 8.573975016870484e-07, + "loss": 0.8906, + "step": 27242 + }, + { + "epoch": 0.87, + "learning_rate": 8.56977918287436e-07, + "loss": 0.8911, + "step": 27243 + }, + { + "epoch": 0.87, + "learning_rate": 8.565584329823306e-07, + "loss": 0.9043, + "step": 27244 + }, + { + "epoch": 0.87, + "learning_rate": 8.561390457762341e-07, + "loss": 0.8638, + "step": 27245 + }, + { + "epoch": 0.87, + "learning_rate": 8.557197566736453e-07, + "loss": 0.8926, + "step": 27246 + }, + { + "epoch": 0.87, + "learning_rate": 8.553005656790636e-07, + "loss": 0.998, + "step": 27247 + }, + { + "epoch": 0.87, + "learning_rate": 8.548814727969845e-07, + "loss": 0.8901, + "step": 27248 + }, + { + "epoch": 0.87, + "learning_rate": 8.544624780319089e-07, + "loss": 0.9771, + "step": 27249 + }, + { + "epoch": 0.87, + "learning_rate": 8.540435813883275e-07, + "loss": 0.7429, + "step": 27250 + }, + { + "epoch": 0.87, + "learning_rate": 8.536247828707367e-07, + "loss": 0.9443, + "step": 27251 + }, + { + "epoch": 0.87, + "learning_rate": 8.532060824836286e-07, + "loss": 0.8672, + "step": 27252 + }, + { + "epoch": 0.87, + "learning_rate": 8.527874802314961e-07, + "loss": 0.9385, + "step": 27253 + }, + { + "epoch": 0.87, + "learning_rate": 8.523689761188302e-07, + "loss": 0.9648, + "step": 27254 + }, + { + "epoch": 0.87, + "learning_rate": 8.519505701501219e-07, + "loss": 1.0254, + "step": 27255 + }, + { + "epoch": 0.87, + "learning_rate": 8.515322623298572e-07, + "loss": 0.7439, + "step": 27256 + }, + { + "epoch": 0.87, + "learning_rate": 8.511140526625306e-07, + "loss": 0.9268, + "step": 27257 + }, + { + "epoch": 0.87, + "learning_rate": 8.506959411526216e-07, + "loss": 0.8462, + "step": 27258 + }, + { + "epoch": 0.87, + "learning_rate": 8.502779278046202e-07, + "loss": 0.9224, + "step": 27259 + }, + { + "epoch": 0.87, + "learning_rate": 8.498600126230083e-07, + "loss": 0.9634, + "step": 27260 + }, + { + "epoch": 0.87, + "learning_rate": 8.494421956122734e-07, + "loss": 0.9526, + "step": 27261 + }, + { + "epoch": 0.87, + "learning_rate": 8.490244767768951e-07, + "loss": 0.9976, + "step": 27262 + }, + { + "epoch": 0.87, + "learning_rate": 8.486068561213578e-07, + "loss": 0.9038, + "step": 27263 + }, + { + "epoch": 0.87, + "learning_rate": 8.48189333650139e-07, + "loss": 0.9707, + "step": 27264 + }, + { + "epoch": 0.87, + "learning_rate": 8.477719093677239e-07, + "loss": 0.8403, + "step": 27265 + }, + { + "epoch": 0.87, + "learning_rate": 8.473545832785834e-07, + "loss": 0.8088, + "step": 27266 + }, + { + "epoch": 0.87, + "learning_rate": 8.469373553871996e-07, + "loss": 0.8838, + "step": 27267 + }, + { + "epoch": 0.87, + "learning_rate": 8.465202256980465e-07, + "loss": 1.0503, + "step": 27268 + }, + { + "epoch": 0.87, + "learning_rate": 8.461031942156028e-07, + "loss": 0.8696, + "step": 27269 + }, + { + "epoch": 0.87, + "learning_rate": 8.456862609443395e-07, + "loss": 1.0273, + "step": 27270 + }, + { + "epoch": 0.87, + "learning_rate": 8.452694258887318e-07, + "loss": 0.9741, + "step": 27271 + }, + { + "epoch": 0.87, + "learning_rate": 8.448526890532515e-07, + "loss": 0.8691, + "step": 27272 + }, + { + "epoch": 0.87, + "learning_rate": 8.444360504423699e-07, + "loss": 0.9248, + "step": 27273 + }, + { + "epoch": 0.87, + "learning_rate": 8.440195100605542e-07, + "loss": 1.0132, + "step": 27274 + }, + { + "epoch": 0.87, + "learning_rate": 8.436030679122786e-07, + "loss": 0.9878, + "step": 27275 + }, + { + "epoch": 0.87, + "learning_rate": 8.431867240020075e-07, + "loss": 0.9268, + "step": 27276 + }, + { + "epoch": 0.87, + "learning_rate": 8.427704783342072e-07, + "loss": 0.9492, + "step": 27277 + }, + { + "epoch": 0.87, + "learning_rate": 8.423543309133464e-07, + "loss": 0.9385, + "step": 27278 + }, + { + "epoch": 0.87, + "learning_rate": 8.419382817438859e-07, + "loss": 0.7383, + "step": 27279 + }, + { + "epoch": 0.87, + "learning_rate": 8.415223308302956e-07, + "loss": 0.8628, + "step": 27280 + }, + { + "epoch": 0.87, + "learning_rate": 8.411064781770328e-07, + "loss": 0.8628, + "step": 27281 + }, + { + "epoch": 0.87, + "learning_rate": 8.40690723788562e-07, + "loss": 0.7212, + "step": 27282 + }, + { + "epoch": 0.87, + "learning_rate": 8.402750676693406e-07, + "loss": 0.937, + "step": 27283 + }, + { + "epoch": 0.87, + "learning_rate": 8.398595098238327e-07, + "loss": 0.9536, + "step": 27284 + }, + { + "epoch": 0.87, + "learning_rate": 8.394440502564915e-07, + "loss": 0.9717, + "step": 27285 + }, + { + "epoch": 0.87, + "learning_rate": 8.390286889717791e-07, + "loss": 0.7783, + "step": 27286 + }, + { + "epoch": 0.87, + "learning_rate": 8.386134259741486e-07, + "loss": 0.4536, + "step": 27287 + }, + { + "epoch": 0.87, + "learning_rate": 8.381982612680605e-07, + "loss": 1.0107, + "step": 27288 + }, + { + "epoch": 0.87, + "learning_rate": 8.377831948579618e-07, + "loss": 0.9458, + "step": 27289 + }, + { + "epoch": 0.87, + "learning_rate": 8.373682267483108e-07, + "loss": 0.9697, + "step": 27290 + }, + { + "epoch": 0.87, + "learning_rate": 8.369533569435551e-07, + "loss": 0.9214, + "step": 27291 + }, + { + "epoch": 0.87, + "learning_rate": 8.365385854481523e-07, + "loss": 0.9785, + "step": 27292 + }, + { + "epoch": 0.87, + "learning_rate": 8.361239122665454e-07, + "loss": 0.9624, + "step": 27293 + }, + { + "epoch": 0.87, + "learning_rate": 8.357093374031899e-07, + "loss": 1.0034, + "step": 27294 + }, + { + "epoch": 0.87, + "learning_rate": 8.3529486086253e-07, + "loss": 0.812, + "step": 27295 + }, + { + "epoch": 0.87, + "learning_rate": 8.348804826490142e-07, + "loss": 0.937, + "step": 27296 + }, + { + "epoch": 0.87, + "learning_rate": 8.344662027670847e-07, + "loss": 0.9233, + "step": 27297 + }, + { + "epoch": 0.87, + "learning_rate": 8.340520212211911e-07, + "loss": 0.9697, + "step": 27298 + }, + { + "epoch": 0.87, + "learning_rate": 8.336379380157722e-07, + "loss": 0.8491, + "step": 27299 + }, + { + "epoch": 0.87, + "learning_rate": 8.332239531552766e-07, + "loss": 0.8408, + "step": 27300 + }, + { + "epoch": 0.87, + "learning_rate": 8.328100666441396e-07, + "loss": 0.9214, + "step": 27301 + }, + { + "epoch": 0.87, + "learning_rate": 8.323962784868078e-07, + "loss": 0.8535, + "step": 27302 + }, + { + "epoch": 0.87, + "learning_rate": 8.319825886877176e-07, + "loss": 0.8525, + "step": 27303 + }, + { + "epoch": 0.87, + "learning_rate": 8.315689972513064e-07, + "loss": 0.8218, + "step": 27304 + }, + { + "epoch": 0.87, + "learning_rate": 8.311555041820119e-07, + "loss": 0.8579, + "step": 27305 + }, + { + "epoch": 0.87, + "learning_rate": 8.307421094842727e-07, + "loss": 0.8647, + "step": 27306 + }, + { + "epoch": 0.87, + "learning_rate": 8.303288131625209e-07, + "loss": 0.897, + "step": 27307 + }, + { + "epoch": 0.87, + "learning_rate": 8.299156152211951e-07, + "loss": 0.9111, + "step": 27308 + }, + { + "epoch": 0.87, + "learning_rate": 8.29502515664723e-07, + "loss": 0.8618, + "step": 27309 + }, + { + "epoch": 0.87, + "learning_rate": 8.29089514497543e-07, + "loss": 0.8921, + "step": 27310 + }, + { + "epoch": 0.87, + "learning_rate": 8.286766117240796e-07, + "loss": 0.9487, + "step": 27311 + }, + { + "epoch": 0.87, + "learning_rate": 8.282638073487681e-07, + "loss": 0.8667, + "step": 27312 + }, + { + "epoch": 0.87, + "learning_rate": 8.278511013760327e-07, + "loss": 0.9854, + "step": 27313 + }, + { + "epoch": 0.87, + "learning_rate": 8.274384938103053e-07, + "loss": 0.9951, + "step": 27314 + }, + { + "epoch": 0.87, + "learning_rate": 8.270259846560114e-07, + "loss": 0.9092, + "step": 27315 + }, + { + "epoch": 0.87, + "learning_rate": 8.266135739175752e-07, + "loss": 1.0859, + "step": 27316 + }, + { + "epoch": 0.87, + "learning_rate": 8.262012615994242e-07, + "loss": 0.9653, + "step": 27317 + }, + { + "epoch": 0.87, + "learning_rate": 8.257890477059804e-07, + "loss": 0.896, + "step": 27318 + }, + { + "epoch": 0.87, + "learning_rate": 8.253769322416649e-07, + "loss": 0.8457, + "step": 27319 + }, + { + "epoch": 0.87, + "learning_rate": 8.24964915210903e-07, + "loss": 0.9111, + "step": 27320 + }, + { + "epoch": 0.87, + "learning_rate": 8.245529966181121e-07, + "loss": 0.8755, + "step": 27321 + }, + { + "epoch": 0.87, + "learning_rate": 8.241411764677121e-07, + "loss": 0.812, + "step": 27322 + }, + { + "epoch": 0.87, + "learning_rate": 8.237294547641239e-07, + "loss": 0.9062, + "step": 27323 + }, + { + "epoch": 0.87, + "learning_rate": 8.233178315117596e-07, + "loss": 0.9727, + "step": 27324 + }, + { + "epoch": 0.87, + "learning_rate": 8.229063067150422e-07, + "loss": 0.8398, + "step": 27325 + }, + { + "epoch": 0.87, + "learning_rate": 8.224948803783828e-07, + "loss": 0.8486, + "step": 27326 + }, + { + "epoch": 0.87, + "learning_rate": 8.220835525061954e-07, + "loss": 0.9775, + "step": 27327 + }, + { + "epoch": 0.87, + "learning_rate": 8.216723231028933e-07, + "loss": 0.979, + "step": 27328 + }, + { + "epoch": 0.87, + "learning_rate": 8.212611921728897e-07, + "loss": 0.9585, + "step": 27329 + }, + { + "epoch": 0.87, + "learning_rate": 8.208501597205942e-07, + "loss": 0.7747, + "step": 27330 + }, + { + "epoch": 0.87, + "learning_rate": 8.20439225750419e-07, + "loss": 0.8062, + "step": 27331 + }, + { + "epoch": 0.87, + "learning_rate": 8.200283902667694e-07, + "loss": 0.9302, + "step": 27332 + }, + { + "epoch": 0.87, + "learning_rate": 8.196176532740585e-07, + "loss": 0.8979, + "step": 27333 + }, + { + "epoch": 0.87, + "learning_rate": 8.192070147766872e-07, + "loss": 0.9937, + "step": 27334 + }, + { + "epoch": 0.87, + "learning_rate": 8.187964747790656e-07, + "loss": 1.0029, + "step": 27335 + }, + { + "epoch": 0.87, + "learning_rate": 8.183860332855942e-07, + "loss": 0.9741, + "step": 27336 + }, + { + "epoch": 0.87, + "learning_rate": 8.179756903006808e-07, + "loss": 0.9229, + "step": 27337 + }, + { + "epoch": 0.87, + "learning_rate": 8.175654458287241e-07, + "loss": 0.9219, + "step": 27338 + }, + { + "epoch": 0.87, + "learning_rate": 8.171552998741295e-07, + "loss": 0.8701, + "step": 27339 + }, + { + "epoch": 0.87, + "learning_rate": 8.167452524412956e-07, + "loss": 0.9888, + "step": 27340 + }, + { + "epoch": 0.87, + "learning_rate": 8.163353035346222e-07, + "loss": 1.0029, + "step": 27341 + }, + { + "epoch": 0.87, + "learning_rate": 8.159254531585048e-07, + "loss": 0.8999, + "step": 27342 + }, + { + "epoch": 0.87, + "learning_rate": 8.155157013173442e-07, + "loss": 0.8647, + "step": 27343 + }, + { + "epoch": 0.87, + "learning_rate": 8.151060480155337e-07, + "loss": 0.8921, + "step": 27344 + }, + { + "epoch": 0.87, + "learning_rate": 8.146964932574719e-07, + "loss": 0.8101, + "step": 27345 + }, + { + "epoch": 0.87, + "learning_rate": 8.142870370475497e-07, + "loss": 0.8774, + "step": 27346 + }, + { + "epoch": 0.87, + "learning_rate": 8.138776793901626e-07, + "loss": 0.8555, + "step": 27347 + }, + { + "epoch": 0.87, + "learning_rate": 8.134684202897003e-07, + "loss": 0.9585, + "step": 27348 + }, + { + "epoch": 0.87, + "learning_rate": 8.13059259750556e-07, + "loss": 0.9397, + "step": 27349 + }, + { + "epoch": 0.87, + "learning_rate": 8.126501977771161e-07, + "loss": 0.9307, + "step": 27350 + }, + { + "epoch": 0.87, + "learning_rate": 8.122412343737729e-07, + "loss": 0.9375, + "step": 27351 + }, + { + "epoch": 0.87, + "learning_rate": 8.118323695449115e-07, + "loss": 0.9087, + "step": 27352 + }, + { + "epoch": 0.87, + "learning_rate": 8.114236032949219e-07, + "loss": 0.9761, + "step": 27353 + }, + { + "epoch": 0.87, + "learning_rate": 8.110149356281848e-07, + "loss": 0.8257, + "step": 27354 + }, + { + "epoch": 0.87, + "learning_rate": 8.106063665490904e-07, + "loss": 0.9443, + "step": 27355 + }, + { + "epoch": 0.87, + "learning_rate": 8.101978960620194e-07, + "loss": 0.895, + "step": 27356 + }, + { + "epoch": 0.87, + "learning_rate": 8.097895241713527e-07, + "loss": 0.8525, + "step": 27357 + }, + { + "epoch": 0.87, + "learning_rate": 8.093812508814725e-07, + "loss": 0.9407, + "step": 27358 + }, + { + "epoch": 0.87, + "learning_rate": 8.089730761967618e-07, + "loss": 0.8398, + "step": 27359 + }, + { + "epoch": 0.87, + "learning_rate": 8.085650001215973e-07, + "loss": 0.9121, + "step": 27360 + }, + { + "epoch": 0.88, + "learning_rate": 8.081570226603563e-07, + "loss": 0.8193, + "step": 27361 + }, + { + "epoch": 0.88, + "learning_rate": 8.077491438174201e-07, + "loss": 0.8608, + "step": 27362 + }, + { + "epoch": 0.88, + "learning_rate": 8.073413635971606e-07, + "loss": 0.9414, + "step": 27363 + }, + { + "epoch": 0.88, + "learning_rate": 8.06933682003953e-07, + "loss": 1.062, + "step": 27364 + }, + { + "epoch": 0.88, + "learning_rate": 8.065260990421753e-07, + "loss": 0.9771, + "step": 27365 + }, + { + "epoch": 0.88, + "learning_rate": 8.061186147161959e-07, + "loss": 1.0122, + "step": 27366 + }, + { + "epoch": 0.88, + "learning_rate": 8.057112290303881e-07, + "loss": 0.9395, + "step": 27367 + }, + { + "epoch": 0.88, + "learning_rate": 8.053039419891251e-07, + "loss": 0.9141, + "step": 27368 + }, + { + "epoch": 0.88, + "learning_rate": 8.048967535967722e-07, + "loss": 0.9775, + "step": 27369 + }, + { + "epoch": 0.88, + "learning_rate": 8.044896638577015e-07, + "loss": 0.9263, + "step": 27370 + }, + { + "epoch": 0.88, + "learning_rate": 8.040826727762807e-07, + "loss": 0.8279, + "step": 27371 + }, + { + "epoch": 0.88, + "learning_rate": 8.036757803568751e-07, + "loss": 0.8965, + "step": 27372 + }, + { + "epoch": 0.88, + "learning_rate": 8.03268986603849e-07, + "loss": 0.9414, + "step": 27373 + }, + { + "epoch": 0.88, + "learning_rate": 8.028622915215689e-07, + "loss": 0.8774, + "step": 27374 + }, + { + "epoch": 0.88, + "learning_rate": 8.024556951143969e-07, + "loss": 0.8496, + "step": 27375 + }, + { + "epoch": 0.88, + "learning_rate": 8.020491973866973e-07, + "loss": 0.8374, + "step": 27376 + }, + { + "epoch": 0.88, + "learning_rate": 8.016427983428277e-07, + "loss": 0.897, + "step": 27377 + }, + { + "epoch": 0.88, + "learning_rate": 8.012364979871545e-07, + "loss": 0.916, + "step": 27378 + }, + { + "epoch": 0.88, + "learning_rate": 8.008302963240299e-07, + "loss": 0.9854, + "step": 27379 + }, + { + "epoch": 0.88, + "learning_rate": 8.004241933578172e-07, + "loss": 1.0049, + "step": 27380 + }, + { + "epoch": 0.88, + "learning_rate": 8.000181890928683e-07, + "loss": 0.9849, + "step": 27381 + }, + { + "epoch": 0.88, + "learning_rate": 7.996122835335451e-07, + "loss": 0.981, + "step": 27382 + }, + { + "epoch": 0.88, + "learning_rate": 7.992064766841967e-07, + "loss": 0.895, + "step": 27383 + }, + { + "epoch": 0.88, + "learning_rate": 7.988007685491827e-07, + "loss": 0.8823, + "step": 27384 + }, + { + "epoch": 0.88, + "learning_rate": 7.983951591328509e-07, + "loss": 0.939, + "step": 27385 + }, + { + "epoch": 0.88, + "learning_rate": 7.979896484395588e-07, + "loss": 0.8608, + "step": 27386 + }, + { + "epoch": 0.88, + "learning_rate": 7.975842364736497e-07, + "loss": 0.9165, + "step": 27387 + }, + { + "epoch": 0.88, + "learning_rate": 7.971789232394788e-07, + "loss": 0.8896, + "step": 27388 + }, + { + "epoch": 0.88, + "learning_rate": 7.967737087413918e-07, + "loss": 0.8604, + "step": 27389 + }, + { + "epoch": 0.88, + "learning_rate": 7.963685929837384e-07, + "loss": 0.9468, + "step": 27390 + }, + { + "epoch": 0.88, + "learning_rate": 7.959635759708617e-07, + "loss": 0.9951, + "step": 27391 + }, + { + "epoch": 0.88, + "learning_rate": 7.955586577071117e-07, + "loss": 0.8557, + "step": 27392 + }, + { + "epoch": 0.88, + "learning_rate": 7.951538381968304e-07, + "loss": 0.9321, + "step": 27393 + }, + { + "epoch": 0.88, + "learning_rate": 7.947491174443611e-07, + "loss": 0.9092, + "step": 27394 + }, + { + "epoch": 0.88, + "learning_rate": 7.943444954540435e-07, + "loss": 0.8779, + "step": 27395 + }, + { + "epoch": 0.88, + "learning_rate": 7.939399722302232e-07, + "loss": 0.9365, + "step": 27396 + }, + { + "epoch": 0.88, + "learning_rate": 7.935355477772366e-07, + "loss": 0.9429, + "step": 27397 + }, + { + "epoch": 0.88, + "learning_rate": 7.931312220994258e-07, + "loss": 0.9214, + "step": 27398 + }, + { + "epoch": 0.88, + "learning_rate": 7.927269952011285e-07, + "loss": 0.9541, + "step": 27399 + }, + { + "epoch": 0.88, + "learning_rate": 7.923228670866778e-07, + "loss": 0.8994, + "step": 27400 + }, + { + "epoch": 0.88, + "learning_rate": 7.919188377604137e-07, + "loss": 0.7566, + "step": 27401 + }, + { + "epoch": 0.88, + "learning_rate": 7.915149072266692e-07, + "loss": 0.8906, + "step": 27402 + }, + { + "epoch": 0.88, + "learning_rate": 7.911110754897777e-07, + "loss": 0.9458, + "step": 27403 + }, + { + "epoch": 0.88, + "learning_rate": 7.907073425540734e-07, + "loss": 1.0098, + "step": 27404 + }, + { + "epoch": 0.88, + "learning_rate": 7.903037084238873e-07, + "loss": 0.9961, + "step": 27405 + }, + { + "epoch": 0.88, + "learning_rate": 7.899001731035472e-07, + "loss": 0.8091, + "step": 27406 + }, + { + "epoch": 0.88, + "learning_rate": 7.894967365973882e-07, + "loss": 1.0654, + "step": 27407 + }, + { + "epoch": 0.88, + "learning_rate": 7.890933989097338e-07, + "loss": 0.9194, + "step": 27408 + }, + { + "epoch": 0.88, + "learning_rate": 7.886901600449126e-07, + "loss": 0.8501, + "step": 27409 + }, + { + "epoch": 0.88, + "learning_rate": 7.882870200072524e-07, + "loss": 0.8916, + "step": 27410 + }, + { + "epoch": 0.88, + "learning_rate": 7.878839788010784e-07, + "loss": 0.9443, + "step": 27411 + }, + { + "epoch": 0.88, + "learning_rate": 7.874810364307117e-07, + "loss": 0.9814, + "step": 27412 + }, + { + "epoch": 0.88, + "learning_rate": 7.87078192900479e-07, + "loss": 0.7966, + "step": 27413 + }, + { + "epoch": 0.88, + "learning_rate": 7.866754482146987e-07, + "loss": 0.9067, + "step": 27414 + }, + { + "epoch": 0.88, + "learning_rate": 7.862728023776967e-07, + "loss": 0.9272, + "step": 27415 + }, + { + "epoch": 0.88, + "learning_rate": 7.858702553937902e-07, + "loss": 0.8857, + "step": 27416 + }, + { + "epoch": 0.88, + "learning_rate": 7.854678072672972e-07, + "loss": 0.9043, + "step": 27417 + }, + { + "epoch": 0.88, + "learning_rate": 7.850654580025363e-07, + "loss": 0.8408, + "step": 27418 + }, + { + "epoch": 0.88, + "learning_rate": 7.846632076038251e-07, + "loss": 0.9229, + "step": 27419 + }, + { + "epoch": 0.88, + "learning_rate": 7.842610560754771e-07, + "loss": 1.0654, + "step": 27420 + }, + { + "epoch": 0.88, + "learning_rate": 7.838590034218107e-07, + "loss": 0.9424, + "step": 27421 + }, + { + "epoch": 0.88, + "learning_rate": 7.83457049647135e-07, + "loss": 0.8735, + "step": 27422 + }, + { + "epoch": 0.88, + "learning_rate": 7.830551947557663e-07, + "loss": 0.9106, + "step": 27423 + }, + { + "epoch": 0.88, + "learning_rate": 7.826534387520146e-07, + "loss": 0.9263, + "step": 27424 + }, + { + "epoch": 0.88, + "learning_rate": 7.822517816401909e-07, + "loss": 0.8267, + "step": 27425 + }, + { + "epoch": 0.88, + "learning_rate": 7.818502234246017e-07, + "loss": 1.0195, + "step": 27426 + }, + { + "epoch": 0.88, + "learning_rate": 7.814487641095592e-07, + "loss": 0.8438, + "step": 27427 + }, + { + "epoch": 0.88, + "learning_rate": 7.810474036993676e-07, + "loss": 0.9458, + "step": 27428 + }, + { + "epoch": 0.88, + "learning_rate": 7.806461421983358e-07, + "loss": 0.9502, + "step": 27429 + }, + { + "epoch": 0.88, + "learning_rate": 7.802449796107659e-07, + "loss": 0.9595, + "step": 27430 + }, + { + "epoch": 0.88, + "learning_rate": 7.798439159409666e-07, + "loss": 0.9434, + "step": 27431 + }, + { + "epoch": 0.88, + "learning_rate": 7.794429511932344e-07, + "loss": 0.896, + "step": 27432 + }, + { + "epoch": 0.88, + "learning_rate": 7.790420853718761e-07, + "loss": 0.8569, + "step": 27433 + }, + { + "epoch": 0.88, + "learning_rate": 7.786413184811892e-07, + "loss": 0.9492, + "step": 27434 + }, + { + "epoch": 0.88, + "learning_rate": 7.782406505254769e-07, + "loss": 0.8208, + "step": 27435 + }, + { + "epoch": 0.88, + "learning_rate": 7.778400815090347e-07, + "loss": 0.8281, + "step": 27436 + }, + { + "epoch": 0.88, + "learning_rate": 7.774396114361637e-07, + "loss": 0.9409, + "step": 27437 + }, + { + "epoch": 0.88, + "learning_rate": 7.770392403111582e-07, + "loss": 0.8911, + "step": 27438 + }, + { + "epoch": 0.88, + "learning_rate": 7.766389681383147e-07, + "loss": 1.1216, + "step": 27439 + }, + { + "epoch": 0.88, + "learning_rate": 7.762387949219253e-07, + "loss": 0.9521, + "step": 27440 + }, + { + "epoch": 0.88, + "learning_rate": 7.758387206662865e-07, + "loss": 0.8789, + "step": 27441 + }, + { + "epoch": 0.88, + "learning_rate": 7.754387453756884e-07, + "loss": 0.9688, + "step": 27442 + }, + { + "epoch": 0.88, + "learning_rate": 7.750388690544253e-07, + "loss": 1.0122, + "step": 27443 + }, + { + "epoch": 0.88, + "learning_rate": 7.746390917067847e-07, + "loss": 0.8818, + "step": 27444 + }, + { + "epoch": 0.88, + "learning_rate": 7.742394133370557e-07, + "loss": 0.8882, + "step": 27445 + }, + { + "epoch": 0.88, + "learning_rate": 7.738398339495279e-07, + "loss": 0.9331, + "step": 27446 + }, + { + "epoch": 0.88, + "learning_rate": 7.73440353548488e-07, + "loss": 0.9326, + "step": 27447 + }, + { + "epoch": 0.88, + "learning_rate": 7.730409721382203e-07, + "loss": 0.7207, + "step": 27448 + }, + { + "epoch": 0.88, + "learning_rate": 7.726416897230127e-07, + "loss": 0.9873, + "step": 27449 + }, + { + "epoch": 0.88, + "learning_rate": 7.72242506307147e-07, + "loss": 0.936, + "step": 27450 + }, + { + "epoch": 0.88, + "learning_rate": 7.718434218949044e-07, + "loss": 0.9683, + "step": 27451 + }, + { + "epoch": 0.88, + "learning_rate": 7.714444364905715e-07, + "loss": 0.9512, + "step": 27452 + }, + { + "epoch": 0.88, + "learning_rate": 7.710455500984237e-07, + "loss": 0.9014, + "step": 27453 + }, + { + "epoch": 0.88, + "learning_rate": 7.706467627227465e-07, + "loss": 0.8979, + "step": 27454 + }, + { + "epoch": 0.88, + "learning_rate": 7.70248074367812e-07, + "loss": 0.9839, + "step": 27455 + }, + { + "epoch": 0.88, + "learning_rate": 7.698494850379024e-07, + "loss": 0.8657, + "step": 27456 + }, + { + "epoch": 0.88, + "learning_rate": 7.694509947372897e-07, + "loss": 0.9946, + "step": 27457 + }, + { + "epoch": 0.88, + "learning_rate": 7.69052603470255e-07, + "loss": 0.9238, + "step": 27458 + }, + { + "epoch": 0.88, + "learning_rate": 7.686543112410672e-07, + "loss": 0.8794, + "step": 27459 + }, + { + "epoch": 0.88, + "learning_rate": 7.682561180540038e-07, + "loss": 0.9326, + "step": 27460 + }, + { + "epoch": 0.88, + "learning_rate": 7.678580239133348e-07, + "loss": 0.8882, + "step": 27461 + }, + { + "epoch": 0.88, + "learning_rate": 7.674600288233314e-07, + "loss": 0.9248, + "step": 27462 + }, + { + "epoch": 0.88, + "learning_rate": 7.670621327882621e-07, + "loss": 1.0664, + "step": 27463 + }, + { + "epoch": 0.88, + "learning_rate": 7.666643358123993e-07, + "loss": 1.0068, + "step": 27464 + }, + { + "epoch": 0.88, + "learning_rate": 7.662666379000084e-07, + "loss": 0.7695, + "step": 27465 + }, + { + "epoch": 0.88, + "learning_rate": 7.65869039055358e-07, + "loss": 0.8159, + "step": 27466 + }, + { + "epoch": 0.88, + "learning_rate": 7.654715392827117e-07, + "loss": 0.5044, + "step": 27467 + }, + { + "epoch": 0.88, + "learning_rate": 7.650741385863369e-07, + "loss": 0.8779, + "step": 27468 + }, + { + "epoch": 0.88, + "learning_rate": 7.64676836970496e-07, + "loss": 0.9551, + "step": 27469 + }, + { + "epoch": 0.88, + "learning_rate": 7.642796344394509e-07, + "loss": 0.8345, + "step": 27470 + }, + { + "epoch": 0.88, + "learning_rate": 7.638825309974629e-07, + "loss": 0.9448, + "step": 27471 + }, + { + "epoch": 0.88, + "learning_rate": 7.63485526648795e-07, + "loss": 0.8618, + "step": 27472 + }, + { + "epoch": 0.88, + "learning_rate": 7.630886213977029e-07, + "loss": 1.0146, + "step": 27473 + }, + { + "epoch": 0.88, + "learning_rate": 7.626918152484486e-07, + "loss": 0.9614, + "step": 27474 + }, + { + "epoch": 0.88, + "learning_rate": 7.622951082052865e-07, + "loss": 0.8823, + "step": 27475 + }, + { + "epoch": 0.88, + "learning_rate": 7.618985002724777e-07, + "loss": 0.9678, + "step": 27476 + }, + { + "epoch": 0.88, + "learning_rate": 7.615019914542699e-07, + "loss": 0.9185, + "step": 27477 + }, + { + "epoch": 0.88, + "learning_rate": 7.611055817549217e-07, + "loss": 0.8872, + "step": 27478 + }, + { + "epoch": 0.88, + "learning_rate": 7.607092711786857e-07, + "loss": 0.8784, + "step": 27479 + }, + { + "epoch": 0.88, + "learning_rate": 7.603130597298147e-07, + "loss": 0.8706, + "step": 27480 + }, + { + "epoch": 0.88, + "learning_rate": 7.599169474125567e-07, + "loss": 0.8657, + "step": 27481 + }, + { + "epoch": 0.88, + "learning_rate": 7.595209342311649e-07, + "loss": 0.7979, + "step": 27482 + }, + { + "epoch": 0.88, + "learning_rate": 7.59125020189887e-07, + "loss": 0.8794, + "step": 27483 + }, + { + "epoch": 0.88, + "learning_rate": 7.587292052929695e-07, + "loss": 0.9951, + "step": 27484 + }, + { + "epoch": 0.88, + "learning_rate": 7.583334895446593e-07, + "loss": 0.9458, + "step": 27485 + }, + { + "epoch": 0.88, + "learning_rate": 7.579378729492037e-07, + "loss": 0.9302, + "step": 27486 + }, + { + "epoch": 0.88, + "learning_rate": 7.575423555108452e-07, + "loss": 0.8599, + "step": 27487 + }, + { + "epoch": 0.88, + "learning_rate": 7.571469372338292e-07, + "loss": 0.9863, + "step": 27488 + }, + { + "epoch": 0.88, + "learning_rate": 7.567516181223966e-07, + "loss": 0.9126, + "step": 27489 + }, + { + "epoch": 0.88, + "learning_rate": 7.563563981807887e-07, + "loss": 0.9463, + "step": 27490 + }, + { + "epoch": 0.88, + "learning_rate": 7.559612774132474e-07, + "loss": 0.9385, + "step": 27491 + }, + { + "epoch": 0.88, + "learning_rate": 7.555662558240107e-07, + "loss": 0.8765, + "step": 27492 + }, + { + "epoch": 0.88, + "learning_rate": 7.551713334173172e-07, + "loss": 0.8008, + "step": 27493 + }, + { + "epoch": 0.88, + "learning_rate": 7.547765101974014e-07, + "loss": 0.8188, + "step": 27494 + }, + { + "epoch": 0.88, + "learning_rate": 7.543817861685033e-07, + "loss": 0.9609, + "step": 27495 + }, + { + "epoch": 0.88, + "learning_rate": 7.539871613348548e-07, + "loss": 0.9565, + "step": 27496 + }, + { + "epoch": 0.88, + "learning_rate": 7.535926357006928e-07, + "loss": 0.9658, + "step": 27497 + }, + { + "epoch": 0.88, + "learning_rate": 7.531982092702461e-07, + "loss": 1.0181, + "step": 27498 + }, + { + "epoch": 0.88, + "learning_rate": 7.528038820477523e-07, + "loss": 0.9277, + "step": 27499 + }, + { + "epoch": 0.88, + "learning_rate": 7.52409654037436e-07, + "loss": 0.9497, + "step": 27500 + }, + { + "epoch": 0.88, + "learning_rate": 7.520155252435302e-07, + "loss": 0.7676, + "step": 27501 + }, + { + "epoch": 0.88, + "learning_rate": 7.516214956702605e-07, + "loss": 0.9175, + "step": 27502 + }, + { + "epoch": 0.88, + "learning_rate": 7.512275653218592e-07, + "loss": 0.7744, + "step": 27503 + }, + { + "epoch": 0.88, + "learning_rate": 7.508337342025485e-07, + "loss": 0.9536, + "step": 27504 + }, + { + "epoch": 0.88, + "learning_rate": 7.50440002316557e-07, + "loss": 0.8364, + "step": 27505 + }, + { + "epoch": 0.88, + "learning_rate": 7.500463696681048e-07, + "loss": 0.7769, + "step": 27506 + }, + { + "epoch": 0.88, + "learning_rate": 7.496528362614219e-07, + "loss": 0.9141, + "step": 27507 + }, + { + "epoch": 0.88, + "learning_rate": 7.492594021007227e-07, + "loss": 0.895, + "step": 27508 + }, + { + "epoch": 0.88, + "learning_rate": 7.488660671902337e-07, + "loss": 0.9893, + "step": 27509 + }, + { + "epoch": 0.88, + "learning_rate": 7.484728315341716e-07, + "loss": 0.8013, + "step": 27510 + }, + { + "epoch": 0.88, + "learning_rate": 7.480796951367586e-07, + "loss": 0.959, + "step": 27511 + }, + { + "epoch": 0.88, + "learning_rate": 7.476866580022091e-07, + "loss": 0.916, + "step": 27512 + }, + { + "epoch": 0.88, + "learning_rate": 7.472937201347429e-07, + "loss": 1.0869, + "step": 27513 + }, + { + "epoch": 0.88, + "learning_rate": 7.469008815385748e-07, + "loss": 0.9233, + "step": 27514 + }, + { + "epoch": 0.88, + "learning_rate": 7.465081422179199e-07, + "loss": 0.9487, + "step": 27515 + }, + { + "epoch": 0.88, + "learning_rate": 7.461155021769906e-07, + "loss": 0.9121, + "step": 27516 + }, + { + "epoch": 0.88, + "learning_rate": 7.457229614200001e-07, + "loss": 0.9097, + "step": 27517 + }, + { + "epoch": 0.88, + "learning_rate": 7.453305199511596e-07, + "loss": 0.9209, + "step": 27518 + }, + { + "epoch": 0.88, + "learning_rate": 7.449381777746811e-07, + "loss": 0.855, + "step": 27519 + }, + { + "epoch": 0.88, + "learning_rate": 7.445459348947726e-07, + "loss": 0.8574, + "step": 27520 + }, + { + "epoch": 0.88, + "learning_rate": 7.441537913156427e-07, + "loss": 1.042, + "step": 27521 + }, + { + "epoch": 0.88, + "learning_rate": 7.437617470414992e-07, + "loss": 0.874, + "step": 27522 + }, + { + "epoch": 0.88, + "learning_rate": 7.43369802076549e-07, + "loss": 0.8853, + "step": 27523 + }, + { + "epoch": 0.88, + "learning_rate": 7.429779564249939e-07, + "loss": 0.9819, + "step": 27524 + }, + { + "epoch": 0.88, + "learning_rate": 7.425862100910408e-07, + "loss": 0.876, + "step": 27525 + }, + { + "epoch": 0.88, + "learning_rate": 7.421945630788918e-07, + "loss": 0.8257, + "step": 27526 + }, + { + "epoch": 0.88, + "learning_rate": 7.418030153927502e-07, + "loss": 0.9097, + "step": 27527 + }, + { + "epoch": 0.88, + "learning_rate": 7.41411567036815e-07, + "loss": 0.9521, + "step": 27528 + }, + { + "epoch": 0.88, + "learning_rate": 7.410202180152881e-07, + "loss": 0.9067, + "step": 27529 + }, + { + "epoch": 0.88, + "learning_rate": 7.406289683323642e-07, + "loss": 0.957, + "step": 27530 + }, + { + "epoch": 0.88, + "learning_rate": 7.402378179922465e-07, + "loss": 0.9531, + "step": 27531 + }, + { + "epoch": 0.88, + "learning_rate": 7.398467669991261e-07, + "loss": 0.9741, + "step": 27532 + }, + { + "epoch": 0.88, + "learning_rate": 7.394558153572029e-07, + "loss": 1.0073, + "step": 27533 + }, + { + "epoch": 0.88, + "learning_rate": 7.390649630706703e-07, + "loss": 0.9639, + "step": 27534 + }, + { + "epoch": 0.88, + "learning_rate": 7.386742101437194e-07, + "loss": 0.8823, + "step": 27535 + }, + { + "epoch": 0.88, + "learning_rate": 7.382835565805457e-07, + "loss": 0.8599, + "step": 27536 + }, + { + "epoch": 0.88, + "learning_rate": 7.378930023853392e-07, + "loss": 1.0, + "step": 27537 + }, + { + "epoch": 0.88, + "learning_rate": 7.37502547562291e-07, + "loss": 0.8867, + "step": 27538 + }, + { + "epoch": 0.88, + "learning_rate": 7.371121921155866e-07, + "loss": 0.8735, + "step": 27539 + }, + { + "epoch": 0.88, + "learning_rate": 7.367219360494183e-07, + "loss": 1.0273, + "step": 27540 + }, + { + "epoch": 0.88, + "learning_rate": 7.363317793679703e-07, + "loss": 0.9155, + "step": 27541 + }, + { + "epoch": 0.88, + "learning_rate": 7.359417220754306e-07, + "loss": 0.9023, + "step": 27542 + }, + { + "epoch": 0.88, + "learning_rate": 7.355517641759824e-07, + "loss": 0.499, + "step": 27543 + }, + { + "epoch": 0.88, + "learning_rate": 7.351619056738135e-07, + "loss": 0.9727, + "step": 27544 + }, + { + "epoch": 0.88, + "learning_rate": 7.347721465730994e-07, + "loss": 0.7766, + "step": 27545 + }, + { + "epoch": 0.88, + "learning_rate": 7.343824868780292e-07, + "loss": 0.9067, + "step": 27546 + }, + { + "epoch": 0.88, + "learning_rate": 7.339929265927781e-07, + "loss": 0.8735, + "step": 27547 + }, + { + "epoch": 0.88, + "learning_rate": 7.336034657215285e-07, + "loss": 0.8306, + "step": 27548 + }, + { + "epoch": 0.88, + "learning_rate": 7.332141042684571e-07, + "loss": 0.8545, + "step": 27549 + }, + { + "epoch": 0.88, + "learning_rate": 7.328248422377437e-07, + "loss": 0.9663, + "step": 27550 + }, + { + "epoch": 0.88, + "learning_rate": 7.32435679633563e-07, + "loss": 0.9355, + "step": 27551 + }, + { + "epoch": 0.88, + "learning_rate": 7.320466164600926e-07, + "loss": 0.8457, + "step": 27552 + }, + { + "epoch": 0.88, + "learning_rate": 7.316576527215013e-07, + "loss": 0.9507, + "step": 27553 + }, + { + "epoch": 0.88, + "learning_rate": 7.312687884219683e-07, + "loss": 1.0137, + "step": 27554 + }, + { + "epoch": 0.88, + "learning_rate": 7.30880023565661e-07, + "loss": 0.8906, + "step": 27555 + }, + { + "epoch": 0.88, + "learning_rate": 7.304913581567541e-07, + "loss": 0.875, + "step": 27556 + }, + { + "epoch": 0.88, + "learning_rate": 7.301027921994142e-07, + "loss": 0.9429, + "step": 27557 + }, + { + "epoch": 0.88, + "learning_rate": 7.297143256978134e-07, + "loss": 0.8867, + "step": 27558 + }, + { + "epoch": 0.88, + "learning_rate": 7.293259586561186e-07, + "loss": 0.4226, + "step": 27559 + }, + { + "epoch": 0.88, + "learning_rate": 7.289376910784962e-07, + "loss": 0.8223, + "step": 27560 + }, + { + "epoch": 0.88, + "learning_rate": 7.285495229691097e-07, + "loss": 0.8926, + "step": 27561 + }, + { + "epoch": 0.88, + "learning_rate": 7.281614543321269e-07, + "loss": 0.9424, + "step": 27562 + }, + { + "epoch": 0.88, + "learning_rate": 7.277734851717088e-07, + "loss": 0.79, + "step": 27563 + }, + { + "epoch": 0.88, + "learning_rate": 7.273856154920211e-07, + "loss": 0.9419, + "step": 27564 + }, + { + "epoch": 0.88, + "learning_rate": 7.269978452972215e-07, + "loss": 0.9604, + "step": 27565 + }, + { + "epoch": 0.88, + "learning_rate": 7.266101745914733e-07, + "loss": 0.7827, + "step": 27566 + }, + { + "epoch": 0.88, + "learning_rate": 7.262226033789355e-07, + "loss": 0.8235, + "step": 27567 + }, + { + "epoch": 0.88, + "learning_rate": 7.258351316637646e-07, + "loss": 0.8843, + "step": 27568 + }, + { + "epoch": 0.88, + "learning_rate": 7.254477594501174e-07, + "loss": 0.9043, + "step": 27569 + }, + { + "epoch": 0.88, + "learning_rate": 7.250604867421529e-07, + "loss": 0.8823, + "step": 27570 + }, + { + "epoch": 0.88, + "learning_rate": 7.24673313544022e-07, + "loss": 0.72, + "step": 27571 + }, + { + "epoch": 0.88, + "learning_rate": 7.242862398598838e-07, + "loss": 0.9268, + "step": 27572 + }, + { + "epoch": 0.88, + "learning_rate": 7.238992656938882e-07, + "loss": 0.9326, + "step": 27573 + }, + { + "epoch": 0.88, + "learning_rate": 7.235123910501863e-07, + "loss": 0.9468, + "step": 27574 + }, + { + "epoch": 0.88, + "learning_rate": 7.231256159329291e-07, + "loss": 0.9478, + "step": 27575 + }, + { + "epoch": 0.88, + "learning_rate": 7.227389403462681e-07, + "loss": 1.083, + "step": 27576 + }, + { + "epoch": 0.88, + "learning_rate": 7.223523642943509e-07, + "loss": 1.0674, + "step": 27577 + }, + { + "epoch": 0.88, + "learning_rate": 7.219658877813229e-07, + "loss": 0.9678, + "step": 27578 + }, + { + "epoch": 0.88, + "learning_rate": 7.215795108113343e-07, + "loss": 0.8271, + "step": 27579 + }, + { + "epoch": 0.88, + "learning_rate": 7.211932333885274e-07, + "loss": 0.9136, + "step": 27580 + }, + { + "epoch": 0.88, + "learning_rate": 7.208070555170487e-07, + "loss": 1.0063, + "step": 27581 + }, + { + "epoch": 0.88, + "learning_rate": 7.204209772010418e-07, + "loss": 0.8711, + "step": 27582 + }, + { + "epoch": 0.88, + "learning_rate": 7.200349984446464e-07, + "loss": 0.9526, + "step": 27583 + }, + { + "epoch": 0.88, + "learning_rate": 7.196491192520039e-07, + "loss": 0.5122, + "step": 27584 + }, + { + "epoch": 0.88, + "learning_rate": 7.192633396272563e-07, + "loss": 0.9336, + "step": 27585 + }, + { + "epoch": 0.88, + "learning_rate": 7.188776595745406e-07, + "loss": 0.9414, + "step": 27586 + }, + { + "epoch": 0.88, + "learning_rate": 7.184920790979976e-07, + "loss": 1.0552, + "step": 27587 + }, + { + "epoch": 0.88, + "learning_rate": 7.18106598201761e-07, + "loss": 0.8877, + "step": 27588 + }, + { + "epoch": 0.88, + "learning_rate": 7.177212168899683e-07, + "loss": 0.895, + "step": 27589 + }, + { + "epoch": 0.88, + "learning_rate": 7.173359351667553e-07, + "loss": 0.9038, + "step": 27590 + }, + { + "epoch": 0.88, + "learning_rate": 7.16950753036253e-07, + "loss": 0.8931, + "step": 27591 + }, + { + "epoch": 0.88, + "learning_rate": 7.165656705025948e-07, + "loss": 0.8809, + "step": 27592 + }, + { + "epoch": 0.88, + "learning_rate": 7.161806875699128e-07, + "loss": 0.9463, + "step": 27593 + }, + { + "epoch": 0.88, + "learning_rate": 7.157958042423362e-07, + "loss": 1.0098, + "step": 27594 + }, + { + "epoch": 0.88, + "learning_rate": 7.154110205239973e-07, + "loss": 0.9268, + "step": 27595 + }, + { + "epoch": 0.88, + "learning_rate": 7.150263364190202e-07, + "loss": 0.8335, + "step": 27596 + }, + { + "epoch": 0.88, + "learning_rate": 7.146417519315373e-07, + "loss": 0.8501, + "step": 27597 + }, + { + "epoch": 0.88, + "learning_rate": 7.1425726706567e-07, + "loss": 0.9604, + "step": 27598 + }, + { + "epoch": 0.88, + "learning_rate": 7.138728818255458e-07, + "loss": 0.8647, + "step": 27599 + }, + { + "epoch": 0.88, + "learning_rate": 7.134885962152871e-07, + "loss": 0.8955, + "step": 27600 + }, + { + "epoch": 0.88, + "learning_rate": 7.131044102390195e-07, + "loss": 0.8252, + "step": 27601 + }, + { + "epoch": 0.88, + "learning_rate": 7.127203239008618e-07, + "loss": 0.9614, + "step": 27602 + }, + { + "epoch": 0.88, + "learning_rate": 7.123363372049374e-07, + "loss": 0.8574, + "step": 27603 + }, + { + "epoch": 0.88, + "learning_rate": 7.119524501553643e-07, + "loss": 0.98, + "step": 27604 + }, + { + "epoch": 0.88, + "learning_rate": 7.115686627562646e-07, + "loss": 0.9048, + "step": 27605 + }, + { + "epoch": 0.88, + "learning_rate": 7.111849750117494e-07, + "loss": 0.833, + "step": 27606 + }, + { + "epoch": 0.88, + "learning_rate": 7.108013869259422e-07, + "loss": 0.9199, + "step": 27607 + }, + { + "epoch": 0.88, + "learning_rate": 7.10417898502953e-07, + "loss": 0.4641, + "step": 27608 + }, + { + "epoch": 0.88, + "learning_rate": 7.100345097468997e-07, + "loss": 0.7671, + "step": 27609 + }, + { + "epoch": 0.88, + "learning_rate": 7.096512206618933e-07, + "loss": 0.9355, + "step": 27610 + }, + { + "epoch": 0.88, + "learning_rate": 7.092680312520483e-07, + "loss": 0.8491, + "step": 27611 + }, + { + "epoch": 0.88, + "learning_rate": 7.088849415214761e-07, + "loss": 1.1265, + "step": 27612 + }, + { + "epoch": 0.88, + "learning_rate": 7.085019514742841e-07, + "loss": 0.7195, + "step": 27613 + }, + { + "epoch": 0.88, + "learning_rate": 7.081190611145828e-07, + "loss": 0.915, + "step": 27614 + }, + { + "epoch": 0.88, + "learning_rate": 7.077362704464808e-07, + "loss": 0.9585, + "step": 27615 + }, + { + "epoch": 0.88, + "learning_rate": 7.073535794740849e-07, + "loss": 0.9131, + "step": 27616 + }, + { + "epoch": 0.88, + "learning_rate": 7.069709882014985e-07, + "loss": 0.958, + "step": 27617 + }, + { + "epoch": 0.88, + "learning_rate": 7.065884966328296e-07, + "loss": 0.9141, + "step": 27618 + }, + { + "epoch": 0.88, + "learning_rate": 7.062061047721802e-07, + "loss": 0.8604, + "step": 27619 + }, + { + "epoch": 0.88, + "learning_rate": 7.058238126236538e-07, + "loss": 0.8479, + "step": 27620 + }, + { + "epoch": 0.88, + "learning_rate": 7.054416201913527e-07, + "loss": 0.9688, + "step": 27621 + }, + { + "epoch": 0.88, + "learning_rate": 7.050595274793759e-07, + "loss": 0.9397, + "step": 27622 + }, + { + "epoch": 0.88, + "learning_rate": 7.04677534491821e-07, + "loss": 0.8506, + "step": 27623 + }, + { + "epoch": 0.88, + "learning_rate": 7.042956412327917e-07, + "loss": 1.0034, + "step": 27624 + }, + { + "epoch": 0.88, + "learning_rate": 7.03913847706379e-07, + "loss": 0.7471, + "step": 27625 + }, + { + "epoch": 0.88, + "learning_rate": 7.035321539166851e-07, + "loss": 0.9565, + "step": 27626 + }, + { + "epoch": 0.88, + "learning_rate": 7.031505598678013e-07, + "loss": 0.9292, + "step": 27627 + }, + { + "epoch": 0.88, + "learning_rate": 7.027690655638242e-07, + "loss": 0.9839, + "step": 27628 + }, + { + "epoch": 0.88, + "learning_rate": 7.023876710088428e-07, + "loss": 0.8237, + "step": 27629 + }, + { + "epoch": 0.88, + "learning_rate": 7.020063762069529e-07, + "loss": 0.8999, + "step": 27630 + }, + { + "epoch": 0.88, + "learning_rate": 7.016251811622432e-07, + "loss": 0.8955, + "step": 27631 + }, + { + "epoch": 0.88, + "learning_rate": 7.01244085878805e-07, + "loss": 1.0122, + "step": 27632 + }, + { + "epoch": 0.88, + "learning_rate": 7.008630903607261e-07, + "loss": 0.7212, + "step": 27633 + }, + { + "epoch": 0.88, + "learning_rate": 7.004821946120954e-07, + "loss": 0.9937, + "step": 27634 + }, + { + "epoch": 0.88, + "learning_rate": 7.001013986369987e-07, + "loss": 0.9346, + "step": 27635 + }, + { + "epoch": 0.88, + "learning_rate": 6.997207024395213e-07, + "loss": 1.0264, + "step": 27636 + }, + { + "epoch": 0.88, + "learning_rate": 6.99340106023746e-07, + "loss": 0.8945, + "step": 27637 + }, + { + "epoch": 0.88, + "learning_rate": 6.98959609393759e-07, + "loss": 0.897, + "step": 27638 + }, + { + "epoch": 0.88, + "learning_rate": 6.985792125536406e-07, + "loss": 0.915, + "step": 27639 + }, + { + "epoch": 0.88, + "learning_rate": 6.981989155074742e-07, + "loss": 0.8301, + "step": 27640 + }, + { + "epoch": 0.88, + "learning_rate": 6.978187182593377e-07, + "loss": 0.9199, + "step": 27641 + }, + { + "epoch": 0.88, + "learning_rate": 6.974386208133144e-07, + "loss": 0.918, + "step": 27642 + }, + { + "epoch": 0.88, + "learning_rate": 6.970586231734755e-07, + "loss": 1.0605, + "step": 27643 + }, + { + "epoch": 0.88, + "learning_rate": 6.966787253439034e-07, + "loss": 0.9902, + "step": 27644 + }, + { + "epoch": 0.88, + "learning_rate": 6.962989273286702e-07, + "loss": 0.8564, + "step": 27645 + }, + { + "epoch": 0.88, + "learning_rate": 6.95919229131854e-07, + "loss": 0.8975, + "step": 27646 + }, + { + "epoch": 0.88, + "learning_rate": 6.955396307575268e-07, + "loss": 0.7808, + "step": 27647 + }, + { + "epoch": 0.88, + "learning_rate": 6.951601322097624e-07, + "loss": 0.8667, + "step": 27648 + }, + { + "epoch": 0.88, + "learning_rate": 6.947807334926304e-07, + "loss": 0.7666, + "step": 27649 + }, + { + "epoch": 0.88, + "learning_rate": 6.944014346102057e-07, + "loss": 0.9058, + "step": 27650 + }, + { + "epoch": 0.88, + "learning_rate": 6.940222355665516e-07, + "loss": 0.8745, + "step": 27651 + }, + { + "epoch": 0.88, + "learning_rate": 6.936431363657414e-07, + "loss": 0.8252, + "step": 27652 + }, + { + "epoch": 0.88, + "learning_rate": 6.932641370118387e-07, + "loss": 0.8179, + "step": 27653 + }, + { + "epoch": 0.88, + "learning_rate": 6.928852375089135e-07, + "loss": 0.894, + "step": 27654 + }, + { + "epoch": 0.88, + "learning_rate": 6.92506437861028e-07, + "loss": 0.895, + "step": 27655 + }, + { + "epoch": 0.88, + "learning_rate": 6.921277380722469e-07, + "loss": 0.9224, + "step": 27656 + }, + { + "epoch": 0.88, + "learning_rate": 6.917491381466346e-07, + "loss": 0.4966, + "step": 27657 + }, + { + "epoch": 0.88, + "learning_rate": 6.913706380882535e-07, + "loss": 0.8008, + "step": 27658 + }, + { + "epoch": 0.88, + "learning_rate": 6.909922379011602e-07, + "loss": 0.5066, + "step": 27659 + }, + { + "epoch": 0.88, + "learning_rate": 6.906139375894194e-07, + "loss": 0.9019, + "step": 27660 + }, + { + "epoch": 0.88, + "learning_rate": 6.902357371570889e-07, + "loss": 0.4714, + "step": 27661 + }, + { + "epoch": 0.88, + "learning_rate": 6.898576366082222e-07, + "loss": 0.9771, + "step": 27662 + }, + { + "epoch": 0.88, + "learning_rate": 6.894796359468825e-07, + "loss": 0.8901, + "step": 27663 + }, + { + "epoch": 0.88, + "learning_rate": 6.89101735177119e-07, + "loss": 0.8442, + "step": 27664 + }, + { + "epoch": 0.88, + "learning_rate": 6.887239343029905e-07, + "loss": 0.9351, + "step": 27665 + }, + { + "epoch": 0.88, + "learning_rate": 6.883462333285495e-07, + "loss": 0.8955, + "step": 27666 + }, + { + "epoch": 0.88, + "learning_rate": 6.879686322578483e-07, + "loss": 0.7891, + "step": 27667 + }, + { + "epoch": 0.88, + "learning_rate": 6.875911310949357e-07, + "loss": 0.8564, + "step": 27668 + }, + { + "epoch": 0.88, + "learning_rate": 6.872137298438653e-07, + "loss": 0.9189, + "step": 27669 + }, + { + "epoch": 0.88, + "learning_rate": 6.868364285086826e-07, + "loss": 0.9146, + "step": 27670 + }, + { + "epoch": 0.88, + "learning_rate": 6.8645922709344e-07, + "loss": 0.5232, + "step": 27671 + }, + { + "epoch": 0.88, + "learning_rate": 6.860821256021799e-07, + "loss": 0.8042, + "step": 27672 + }, + { + "epoch": 0.88, + "learning_rate": 6.857051240389545e-07, + "loss": 0.9316, + "step": 27673 + }, + { + "epoch": 0.89, + "learning_rate": 6.853282224078006e-07, + "loss": 0.8662, + "step": 27674 + }, + { + "epoch": 0.89, + "learning_rate": 6.849514207127683e-07, + "loss": 0.875, + "step": 27675 + }, + { + "epoch": 0.89, + "learning_rate": 6.845747189578955e-07, + "loss": 0.842, + "step": 27676 + }, + { + "epoch": 0.89, + "learning_rate": 6.841981171472289e-07, + "loss": 0.9014, + "step": 27677 + }, + { + "epoch": 0.89, + "learning_rate": 6.838216152848032e-07, + "loss": 0.9365, + "step": 27678 + }, + { + "epoch": 0.89, + "learning_rate": 6.834452133746638e-07, + "loss": 1.0024, + "step": 27679 + }, + { + "epoch": 0.89, + "learning_rate": 6.830689114208455e-07, + "loss": 0.8413, + "step": 27680 + }, + { + "epoch": 0.89, + "learning_rate": 6.826927094273861e-07, + "loss": 0.7852, + "step": 27681 + }, + { + "epoch": 0.89, + "learning_rate": 6.823166073983212e-07, + "loss": 0.9639, + "step": 27682 + }, + { + "epoch": 0.89, + "learning_rate": 6.819406053376876e-07, + "loss": 0.9341, + "step": 27683 + }, + { + "epoch": 0.89, + "learning_rate": 6.815647032495176e-07, + "loss": 0.5237, + "step": 27684 + }, + { + "epoch": 0.89, + "learning_rate": 6.81188901137847e-07, + "loss": 0.8584, + "step": 27685 + }, + { + "epoch": 0.89, + "learning_rate": 6.808131990067036e-07, + "loss": 0.8823, + "step": 27686 + }, + { + "epoch": 0.89, + "learning_rate": 6.804375968601229e-07, + "loss": 0.8965, + "step": 27687 + }, + { + "epoch": 0.89, + "learning_rate": 6.80062094702132e-07, + "loss": 0.9302, + "step": 27688 + }, + { + "epoch": 0.89, + "learning_rate": 6.796866925367595e-07, + "loss": 0.915, + "step": 27689 + }, + { + "epoch": 0.89, + "learning_rate": 6.793113903680327e-07, + "loss": 0.8633, + "step": 27690 + }, + { + "epoch": 0.89, + "learning_rate": 6.789361881999801e-07, + "loss": 0.8687, + "step": 27691 + }, + { + "epoch": 0.89, + "learning_rate": 6.785610860366254e-07, + "loss": 0.9517, + "step": 27692 + }, + { + "epoch": 0.89, + "learning_rate": 6.781860838819953e-07, + "loss": 0.9897, + "step": 27693 + }, + { + "epoch": 0.89, + "learning_rate": 6.778111817401111e-07, + "loss": 0.792, + "step": 27694 + }, + { + "epoch": 0.89, + "learning_rate": 6.774363796149952e-07, + "loss": 0.8911, + "step": 27695 + }, + { + "epoch": 0.89, + "learning_rate": 6.770616775106687e-07, + "loss": 0.9805, + "step": 27696 + }, + { + "epoch": 0.89, + "learning_rate": 6.766870754311539e-07, + "loss": 0.998, + "step": 27697 + }, + { + "epoch": 0.89, + "learning_rate": 6.763125733804655e-07, + "loss": 0.9116, + "step": 27698 + }, + { + "epoch": 0.89, + "learning_rate": 6.759381713626267e-07, + "loss": 1.0269, + "step": 27699 + }, + { + "epoch": 0.89, + "learning_rate": 6.755638693816525e-07, + "loss": 0.9863, + "step": 27700 + }, + { + "epoch": 0.89, + "learning_rate": 6.751896674415548e-07, + "loss": 0.875, + "step": 27701 + }, + { + "epoch": 0.89, + "learning_rate": 6.748155655463551e-07, + "loss": 0.8799, + "step": 27702 + }, + { + "epoch": 0.89, + "learning_rate": 6.744415637000635e-07, + "loss": 1.0298, + "step": 27703 + }, + { + "epoch": 0.89, + "learning_rate": 6.74067661906691e-07, + "loss": 1.0127, + "step": 27704 + }, + { + "epoch": 0.89, + "learning_rate": 6.736938601702525e-07, + "loss": 0.7917, + "step": 27705 + }, + { + "epoch": 0.89, + "learning_rate": 6.733201584947569e-07, + "loss": 0.9912, + "step": 27706 + }, + { + "epoch": 0.89, + "learning_rate": 6.729465568842119e-07, + "loss": 0.8428, + "step": 27707 + }, + { + "epoch": 0.89, + "learning_rate": 6.725730553426301e-07, + "loss": 0.8655, + "step": 27708 + }, + { + "epoch": 0.89, + "learning_rate": 6.72199653874015e-07, + "loss": 0.9214, + "step": 27709 + }, + { + "epoch": 0.89, + "learning_rate": 6.718263524823765e-07, + "loss": 0.9639, + "step": 27710 + }, + { + "epoch": 0.89, + "learning_rate": 6.714531511717137e-07, + "loss": 0.9443, + "step": 27711 + }, + { + "epoch": 0.89, + "learning_rate": 6.710800499460368e-07, + "loss": 0.811, + "step": 27712 + }, + { + "epoch": 0.89, + "learning_rate": 6.707070488093448e-07, + "loss": 0.9307, + "step": 27713 + }, + { + "epoch": 0.89, + "learning_rate": 6.703341477656422e-07, + "loss": 0.7063, + "step": 27714 + }, + { + "epoch": 0.89, + "learning_rate": 6.69961346818927e-07, + "loss": 0.8589, + "step": 27715 + }, + { + "epoch": 0.89, + "learning_rate": 6.695886459732015e-07, + "loss": 1.0254, + "step": 27716 + }, + { + "epoch": 0.89, + "learning_rate": 6.692160452324625e-07, + "loss": 0.979, + "step": 27717 + }, + { + "epoch": 0.89, + "learning_rate": 6.688435446007124e-07, + "loss": 0.9065, + "step": 27718 + }, + { + "epoch": 0.89, + "learning_rate": 6.684711440819402e-07, + "loss": 0.9956, + "step": 27719 + }, + { + "epoch": 0.89, + "learning_rate": 6.680988436801461e-07, + "loss": 0.9526, + "step": 27720 + }, + { + "epoch": 0.89, + "learning_rate": 6.677266433993223e-07, + "loss": 0.9253, + "step": 27721 + }, + { + "epoch": 0.89, + "learning_rate": 6.673545432434647e-07, + "loss": 0.894, + "step": 27722 + }, + { + "epoch": 0.89, + "learning_rate": 6.669825432165622e-07, + "loss": 0.9028, + "step": 27723 + }, + { + "epoch": 0.89, + "learning_rate": 6.666106433226094e-07, + "loss": 0.8623, + "step": 27724 + }, + { + "epoch": 0.89, + "learning_rate": 6.662388435655942e-07, + "loss": 0.9453, + "step": 27725 + }, + { + "epoch": 0.89, + "learning_rate": 6.658671439495068e-07, + "loss": 0.9282, + "step": 27726 + }, + { + "epoch": 0.89, + "learning_rate": 6.654955444783329e-07, + "loss": 0.9211, + "step": 27727 + }, + { + "epoch": 0.89, + "learning_rate": 6.651240451560615e-07, + "loss": 0.7659, + "step": 27728 + }, + { + "epoch": 0.89, + "learning_rate": 6.647526459866771e-07, + "loss": 0.8945, + "step": 27729 + }, + { + "epoch": 0.89, + "learning_rate": 6.643813469741667e-07, + "loss": 0.9199, + "step": 27730 + }, + { + "epoch": 0.89, + "learning_rate": 6.640101481225103e-07, + "loss": 0.8931, + "step": 27731 + }, + { + "epoch": 0.89, + "learning_rate": 6.636390494356937e-07, + "loss": 0.8569, + "step": 27732 + }, + { + "epoch": 0.89, + "learning_rate": 6.63268050917698e-07, + "loss": 0.9458, + "step": 27733 + }, + { + "epoch": 0.89, + "learning_rate": 6.628971525725026e-07, + "loss": 0.9448, + "step": 27734 + }, + { + "epoch": 0.89, + "learning_rate": 6.62526354404085e-07, + "loss": 0.9072, + "step": 27735 + }, + { + "epoch": 0.89, + "learning_rate": 6.621556564164267e-07, + "loss": 0.8994, + "step": 27736 + }, + { + "epoch": 0.89, + "learning_rate": 6.617850586135033e-07, + "loss": 0.8896, + "step": 27737 + }, + { + "epoch": 0.89, + "learning_rate": 6.614145609992917e-07, + "loss": 0.8708, + "step": 27738 + }, + { + "epoch": 0.89, + "learning_rate": 6.610441635777665e-07, + "loss": 0.8936, + "step": 27739 + }, + { + "epoch": 0.89, + "learning_rate": 6.606738663529e-07, + "loss": 0.4763, + "step": 27740 + }, + { + "epoch": 0.89, + "learning_rate": 6.60303669328668e-07, + "loss": 0.9956, + "step": 27741 + }, + { + "epoch": 0.89, + "learning_rate": 6.599335725090416e-07, + "loss": 1.0317, + "step": 27742 + }, + { + "epoch": 0.89, + "learning_rate": 6.59563575897989e-07, + "loss": 0.8384, + "step": 27743 + }, + { + "epoch": 0.89, + "learning_rate": 6.591936794994835e-07, + "loss": 0.8774, + "step": 27744 + }, + { + "epoch": 0.89, + "learning_rate": 6.588238833174921e-07, + "loss": 0.8789, + "step": 27745 + }, + { + "epoch": 0.89, + "learning_rate": 6.584541873559802e-07, + "loss": 0.8618, + "step": 27746 + }, + { + "epoch": 0.89, + "learning_rate": 6.580845916189183e-07, + "loss": 0.8818, + "step": 27747 + }, + { + "epoch": 0.89, + "learning_rate": 6.577150961102685e-07, + "loss": 0.9985, + "step": 27748 + }, + { + "epoch": 0.89, + "learning_rate": 6.573457008339956e-07, + "loss": 0.8472, + "step": 27749 + }, + { + "epoch": 0.89, + "learning_rate": 6.569764057940641e-07, + "loss": 0.9395, + "step": 27750 + }, + { + "epoch": 0.89, + "learning_rate": 6.566072109944366e-07, + "loss": 0.8813, + "step": 27751 + }, + { + "epoch": 0.89, + "learning_rate": 6.562381164390696e-07, + "loss": 0.9482, + "step": 27752 + }, + { + "epoch": 0.89, + "learning_rate": 6.558691221319302e-07, + "loss": 0.9038, + "step": 27753 + }, + { + "epoch": 0.89, + "learning_rate": 6.555002280769706e-07, + "loss": 0.9702, + "step": 27754 + }, + { + "epoch": 0.89, + "learning_rate": 6.551314342781534e-07, + "loss": 0.8838, + "step": 27755 + }, + { + "epoch": 0.89, + "learning_rate": 6.54762740739433e-07, + "loss": 1.0488, + "step": 27756 + }, + { + "epoch": 0.89, + "learning_rate": 6.543941474647663e-07, + "loss": 0.9727, + "step": 27757 + }, + { + "epoch": 0.89, + "learning_rate": 6.540256544581058e-07, + "loss": 0.9219, + "step": 27758 + }, + { + "epoch": 0.89, + "learning_rate": 6.536572617234082e-07, + "loss": 0.9053, + "step": 27759 + }, + { + "epoch": 0.89, + "learning_rate": 6.532889692646216e-07, + "loss": 0.9951, + "step": 27760 + }, + { + "epoch": 0.89, + "learning_rate": 6.529207770857015e-07, + "loss": 0.8413, + "step": 27761 + }, + { + "epoch": 0.89, + "learning_rate": 6.525526851905961e-07, + "loss": 0.959, + "step": 27762 + }, + { + "epoch": 0.89, + "learning_rate": 6.521846935832576e-07, + "loss": 0.8608, + "step": 27763 + }, + { + "epoch": 0.89, + "learning_rate": 6.518168022676285e-07, + "loss": 0.9561, + "step": 27764 + }, + { + "epoch": 0.89, + "learning_rate": 6.514490112476612e-07, + "loss": 1.0098, + "step": 27765 + }, + { + "epoch": 0.89, + "learning_rate": 6.51081320527297e-07, + "loss": 0.7886, + "step": 27766 + }, + { + "epoch": 0.89, + "learning_rate": 6.50713730110486e-07, + "loss": 1.0273, + "step": 27767 + }, + { + "epoch": 0.89, + "learning_rate": 6.503462400011673e-07, + "loss": 0.7568, + "step": 27768 + }, + { + "epoch": 0.89, + "learning_rate": 6.499788502032878e-07, + "loss": 1.0063, + "step": 27769 + }, + { + "epoch": 0.89, + "learning_rate": 6.496115607207843e-07, + "loss": 0.9458, + "step": 27770 + }, + { + "epoch": 0.89, + "learning_rate": 6.492443715576046e-07, + "loss": 0.9067, + "step": 27771 + }, + { + "epoch": 0.89, + "learning_rate": 6.488772827176803e-07, + "loss": 0.8823, + "step": 27772 + }, + { + "epoch": 0.89, + "learning_rate": 6.485102942049549e-07, + "loss": 0.8589, + "step": 27773 + }, + { + "epoch": 0.89, + "learning_rate": 6.481434060233627e-07, + "loss": 0.7695, + "step": 27774 + }, + { + "epoch": 0.89, + "learning_rate": 6.477766181768441e-07, + "loss": 0.8906, + "step": 27775 + }, + { + "epoch": 0.89, + "learning_rate": 6.474099306693293e-07, + "loss": 0.9497, + "step": 27776 + }, + { + "epoch": 0.89, + "learning_rate": 6.470433435047574e-07, + "loss": 1.0254, + "step": 27777 + }, + { + "epoch": 0.89, + "learning_rate": 6.466768566870585e-07, + "loss": 0.8584, + "step": 27778 + }, + { + "epoch": 0.89, + "learning_rate": 6.463104702201651e-07, + "loss": 0.9409, + "step": 27779 + }, + { + "epoch": 0.89, + "learning_rate": 6.459441841080072e-07, + "loss": 1.0405, + "step": 27780 + }, + { + "epoch": 0.89, + "learning_rate": 6.455779983545162e-07, + "loss": 1.0361, + "step": 27781 + }, + { + "epoch": 0.89, + "learning_rate": 6.452119129636192e-07, + "loss": 0.8535, + "step": 27782 + }, + { + "epoch": 0.89, + "learning_rate": 6.448459279392472e-07, + "loss": 0.9141, + "step": 27783 + }, + { + "epoch": 0.89, + "learning_rate": 6.444800432853237e-07, + "loss": 0.9136, + "step": 27784 + }, + { + "epoch": 0.89, + "learning_rate": 6.441142590057747e-07, + "loss": 0.9946, + "step": 27785 + }, + { + "epoch": 0.89, + "learning_rate": 6.437485751045258e-07, + "loss": 0.8618, + "step": 27786 + }, + { + "epoch": 0.89, + "learning_rate": 6.433829915855006e-07, + "loss": 0.9453, + "step": 27787 + }, + { + "epoch": 0.89, + "learning_rate": 6.43017508452618e-07, + "loss": 0.936, + "step": 27788 + }, + { + "epoch": 0.89, + "learning_rate": 6.42652125709804e-07, + "loss": 1.002, + "step": 27789 + }, + { + "epoch": 0.89, + "learning_rate": 6.422868433609753e-07, + "loss": 0.8623, + "step": 27790 + }, + { + "epoch": 0.89, + "learning_rate": 6.419216614100521e-07, + "loss": 0.8042, + "step": 27791 + }, + { + "epoch": 0.89, + "learning_rate": 6.415565798609524e-07, + "loss": 1.0088, + "step": 27792 + }, + { + "epoch": 0.89, + "learning_rate": 6.411915987175943e-07, + "loss": 0.9136, + "step": 27793 + }, + { + "epoch": 0.89, + "learning_rate": 6.408267179838923e-07, + "loss": 0.7983, + "step": 27794 + }, + { + "epoch": 0.89, + "learning_rate": 6.404619376637599e-07, + "loss": 0.9805, + "step": 27795 + }, + { + "epoch": 0.89, + "learning_rate": 6.400972577611142e-07, + "loss": 0.979, + "step": 27796 + }, + { + "epoch": 0.89, + "learning_rate": 6.397326782798641e-07, + "loss": 0.9731, + "step": 27797 + }, + { + "epoch": 0.89, + "learning_rate": 6.393681992239242e-07, + "loss": 0.936, + "step": 27798 + }, + { + "epoch": 0.89, + "learning_rate": 6.390038205972026e-07, + "loss": 0.5037, + "step": 27799 + }, + { + "epoch": 0.89, + "learning_rate": 6.386395424036107e-07, + "loss": 0.4539, + "step": 27800 + }, + { + "epoch": 0.89, + "learning_rate": 6.382753646470563e-07, + "loss": 0.9956, + "step": 27801 + }, + { + "epoch": 0.89, + "learning_rate": 6.379112873314464e-07, + "loss": 0.873, + "step": 27802 + }, + { + "epoch": 0.89, + "learning_rate": 6.375473104606844e-07, + "loss": 0.833, + "step": 27803 + }, + { + "epoch": 0.89, + "learning_rate": 6.371834340386807e-07, + "loss": 0.8008, + "step": 27804 + }, + { + "epoch": 0.89, + "learning_rate": 6.368196580693342e-07, + "loss": 0.9165, + "step": 27805 + }, + { + "epoch": 0.89, + "learning_rate": 6.36455982556552e-07, + "loss": 0.9326, + "step": 27806 + }, + { + "epoch": 0.89, + "learning_rate": 6.360924075042319e-07, + "loss": 0.9058, + "step": 27807 + }, + { + "epoch": 0.89, + "learning_rate": 6.357289329162808e-07, + "loss": 0.9365, + "step": 27808 + }, + { + "epoch": 0.89, + "learning_rate": 6.353655587965901e-07, + "loss": 1.0015, + "step": 27809 + }, + { + "epoch": 0.89, + "learning_rate": 6.350022851490656e-07, + "loss": 0.8643, + "step": 27810 + }, + { + "epoch": 0.89, + "learning_rate": 6.346391119775996e-07, + "loss": 0.9302, + "step": 27811 + }, + { + "epoch": 0.89, + "learning_rate": 6.342760392860925e-07, + "loss": 0.8384, + "step": 27812 + }, + { + "epoch": 0.89, + "learning_rate": 6.339130670784366e-07, + "loss": 0.9263, + "step": 27813 + }, + { + "epoch": 0.89, + "learning_rate": 6.335501953585288e-07, + "loss": 0.9399, + "step": 27814 + }, + { + "epoch": 0.89, + "learning_rate": 6.331874241302605e-07, + "loss": 1.1133, + "step": 27815 + }, + { + "epoch": 0.89, + "learning_rate": 6.328247533975262e-07, + "loss": 0.9717, + "step": 27816 + }, + { + "epoch": 0.89, + "learning_rate": 6.32462183164213e-07, + "loss": 0.9229, + "step": 27817 + }, + { + "epoch": 0.89, + "learning_rate": 6.320997134342144e-07, + "loss": 0.9224, + "step": 27818 + }, + { + "epoch": 0.89, + "learning_rate": 6.317373442114172e-07, + "loss": 0.9961, + "step": 27819 + }, + { + "epoch": 0.89, + "learning_rate": 6.313750754997106e-07, + "loss": 0.9146, + "step": 27820 + }, + { + "epoch": 0.89, + "learning_rate": 6.310129073029791e-07, + "loss": 1.0044, + "step": 27821 + }, + { + "epoch": 0.89, + "learning_rate": 6.306508396251121e-07, + "loss": 0.9468, + "step": 27822 + }, + { + "epoch": 0.89, + "learning_rate": 6.30288872469993e-07, + "loss": 0.8877, + "step": 27823 + }, + { + "epoch": 0.89, + "learning_rate": 6.29927005841503e-07, + "loss": 0.7944, + "step": 27824 + }, + { + "epoch": 0.89, + "learning_rate": 6.295652397435248e-07, + "loss": 0.9595, + "step": 27825 + }, + { + "epoch": 0.89, + "learning_rate": 6.292035741799429e-07, + "loss": 0.9629, + "step": 27826 + }, + { + "epoch": 0.89, + "learning_rate": 6.288420091546332e-07, + "loss": 0.8003, + "step": 27827 + }, + { + "epoch": 0.89, + "learning_rate": 6.284805446714793e-07, + "loss": 0.98, + "step": 27828 + }, + { + "epoch": 0.89, + "learning_rate": 6.281191807343567e-07, + "loss": 0.8613, + "step": 27829 + }, + { + "epoch": 0.89, + "learning_rate": 6.277579173471427e-07, + "loss": 0.9951, + "step": 27830 + }, + { + "epoch": 0.89, + "learning_rate": 6.273967545137138e-07, + "loss": 0.9229, + "step": 27831 + }, + { + "epoch": 0.89, + "learning_rate": 6.27035692237945e-07, + "loss": 0.8149, + "step": 27832 + }, + { + "epoch": 0.89, + "learning_rate": 6.266747305237098e-07, + "loss": 0.9287, + "step": 27833 + }, + { + "epoch": 0.89, + "learning_rate": 6.263138693748794e-07, + "loss": 0.8281, + "step": 27834 + }, + { + "epoch": 0.89, + "learning_rate": 6.259531087953285e-07, + "loss": 0.5029, + "step": 27835 + }, + { + "epoch": 0.89, + "learning_rate": 6.255924487889253e-07, + "loss": 1.0132, + "step": 27836 + }, + { + "epoch": 0.89, + "learning_rate": 6.2523188935954e-07, + "loss": 0.8833, + "step": 27837 + }, + { + "epoch": 0.89, + "learning_rate": 6.248714305110403e-07, + "loss": 0.8037, + "step": 27838 + }, + { + "epoch": 0.89, + "learning_rate": 6.24511072247298e-07, + "loss": 0.9316, + "step": 27839 + }, + { + "epoch": 0.89, + "learning_rate": 6.24150814572172e-07, + "loss": 0.8848, + "step": 27840 + }, + { + "epoch": 0.89, + "learning_rate": 6.237906574895325e-07, + "loss": 0.9312, + "step": 27841 + }, + { + "epoch": 0.89, + "learning_rate": 6.234306010032398e-07, + "loss": 0.9355, + "step": 27842 + }, + { + "epoch": 0.89, + "learning_rate": 6.23070645117162e-07, + "loss": 0.7778, + "step": 27843 + }, + { + "epoch": 0.89, + "learning_rate": 6.22710789835157e-07, + "loss": 0.9565, + "step": 27844 + }, + { + "epoch": 0.89, + "learning_rate": 6.223510351610873e-07, + "loss": 0.8403, + "step": 27845 + }, + { + "epoch": 0.89, + "learning_rate": 6.219913810988132e-07, + "loss": 0.8848, + "step": 27846 + }, + { + "epoch": 0.89, + "learning_rate": 6.216318276521927e-07, + "loss": 0.9033, + "step": 27847 + }, + { + "epoch": 0.89, + "learning_rate": 6.212723748250804e-07, + "loss": 0.916, + "step": 27848 + }, + { + "epoch": 0.89, + "learning_rate": 6.209130226213378e-07, + "loss": 0.8823, + "step": 27849 + }, + { + "epoch": 0.89, + "learning_rate": 6.205537710448173e-07, + "loss": 0.9316, + "step": 27850 + }, + { + "epoch": 0.89, + "learning_rate": 6.201946200993758e-07, + "loss": 0.9185, + "step": 27851 + }, + { + "epoch": 0.89, + "learning_rate": 6.198355697888625e-07, + "loss": 0.9136, + "step": 27852 + }, + { + "epoch": 0.89, + "learning_rate": 6.194766201171342e-07, + "loss": 0.7617, + "step": 27853 + }, + { + "epoch": 0.89, + "learning_rate": 6.191177710880403e-07, + "loss": 0.7993, + "step": 27854 + }, + { + "epoch": 0.89, + "learning_rate": 6.187590227054296e-07, + "loss": 0.8354, + "step": 27855 + }, + { + "epoch": 0.89, + "learning_rate": 6.184003749731504e-07, + "loss": 0.8984, + "step": 27856 + }, + { + "epoch": 0.89, + "learning_rate": 6.18041827895054e-07, + "loss": 0.8276, + "step": 27857 + }, + { + "epoch": 0.89, + "learning_rate": 6.176833814749839e-07, + "loss": 0.8989, + "step": 27858 + }, + { + "epoch": 0.89, + "learning_rate": 6.173250357167892e-07, + "loss": 0.8721, + "step": 27859 + }, + { + "epoch": 0.89, + "learning_rate": 6.169667906243104e-07, + "loss": 0.8838, + "step": 27860 + }, + { + "epoch": 0.89, + "learning_rate": 6.166086462013965e-07, + "loss": 0.8662, + "step": 27861 + }, + { + "epoch": 0.89, + "learning_rate": 6.162506024518844e-07, + "loss": 0.8311, + "step": 27862 + }, + { + "epoch": 0.89, + "learning_rate": 6.158926593796188e-07, + "loss": 0.8403, + "step": 27863 + }, + { + "epoch": 0.89, + "learning_rate": 6.155348169884379e-07, + "loss": 0.9663, + "step": 27864 + }, + { + "epoch": 0.89, + "learning_rate": 6.151770752821829e-07, + "loss": 0.833, + "step": 27865 + }, + { + "epoch": 0.89, + "learning_rate": 6.148194342646907e-07, + "loss": 0.4883, + "step": 27866 + }, + { + "epoch": 0.89, + "learning_rate": 6.144618939398006e-07, + "loss": 0.8457, + "step": 27867 + }, + { + "epoch": 0.89, + "learning_rate": 6.141044543113462e-07, + "loss": 0.9492, + "step": 27868 + }, + { + "epoch": 0.89, + "learning_rate": 6.137471153831642e-07, + "loss": 0.812, + "step": 27869 + }, + { + "epoch": 0.89, + "learning_rate": 6.133898771590851e-07, + "loss": 0.8086, + "step": 27870 + }, + { + "epoch": 0.89, + "learning_rate": 6.130327396429458e-07, + "loss": 0.8689, + "step": 27871 + }, + { + "epoch": 0.89, + "learning_rate": 6.126757028385766e-07, + "loss": 0.9365, + "step": 27872 + }, + { + "epoch": 0.89, + "learning_rate": 6.123187667498054e-07, + "loss": 0.9072, + "step": 27873 + }, + { + "epoch": 0.89, + "learning_rate": 6.119619313804659e-07, + "loss": 0.8096, + "step": 27874 + }, + { + "epoch": 0.89, + "learning_rate": 6.116051967343839e-07, + "loss": 0.9658, + "step": 27875 + }, + { + "epoch": 0.89, + "learning_rate": 6.112485628153886e-07, + "loss": 0.8359, + "step": 27876 + }, + { + "epoch": 0.89, + "learning_rate": 6.108920296273047e-07, + "loss": 0.9668, + "step": 27877 + }, + { + "epoch": 0.89, + "learning_rate": 6.10535597173958e-07, + "loss": 0.9546, + "step": 27878 + }, + { + "epoch": 0.89, + "learning_rate": 6.101792654591721e-07, + "loss": 0.9189, + "step": 27879 + }, + { + "epoch": 0.89, + "learning_rate": 6.098230344867717e-07, + "loss": 0.9111, + "step": 27880 + }, + { + "epoch": 0.89, + "learning_rate": 6.094669042605761e-07, + "loss": 0.936, + "step": 27881 + }, + { + "epoch": 0.89, + "learning_rate": 6.091108747844088e-07, + "loss": 0.9785, + "step": 27882 + }, + { + "epoch": 0.89, + "learning_rate": 6.087549460620867e-07, + "loss": 0.918, + "step": 27883 + }, + { + "epoch": 0.89, + "learning_rate": 6.083991180974336e-07, + "loss": 0.8948, + "step": 27884 + }, + { + "epoch": 0.89, + "learning_rate": 6.080433908942618e-07, + "loss": 0.8896, + "step": 27885 + }, + { + "epoch": 0.89, + "learning_rate": 6.076877644563906e-07, + "loss": 0.9272, + "step": 27886 + }, + { + "epoch": 0.89, + "learning_rate": 6.073322387876334e-07, + "loss": 0.4592, + "step": 27887 + }, + { + "epoch": 0.89, + "learning_rate": 6.069768138918076e-07, + "loss": 0.8931, + "step": 27888 + }, + { + "epoch": 0.89, + "learning_rate": 6.066214897727229e-07, + "loss": 0.9575, + "step": 27889 + }, + { + "epoch": 0.89, + "learning_rate": 6.062662664341956e-07, + "loss": 0.8823, + "step": 27890 + }, + { + "epoch": 0.89, + "learning_rate": 6.059111438800325e-07, + "loss": 0.8423, + "step": 27891 + }, + { + "epoch": 0.89, + "learning_rate": 6.055561221140493e-07, + "loss": 0.8364, + "step": 27892 + }, + { + "epoch": 0.89, + "learning_rate": 6.052012011400488e-07, + "loss": 1.042, + "step": 27893 + }, + { + "epoch": 0.89, + "learning_rate": 6.048463809618444e-07, + "loss": 0.9131, + "step": 27894 + }, + { + "epoch": 0.89, + "learning_rate": 6.044916615832375e-07, + "loss": 0.9912, + "step": 27895 + }, + { + "epoch": 0.89, + "learning_rate": 6.041370430080384e-07, + "loss": 0.7441, + "step": 27896 + }, + { + "epoch": 0.89, + "learning_rate": 6.037825252400487e-07, + "loss": 0.7539, + "step": 27897 + }, + { + "epoch": 0.89, + "learning_rate": 6.03428108283074e-07, + "loss": 0.9888, + "step": 27898 + }, + { + "epoch": 0.89, + "learning_rate": 6.030737921409169e-07, + "loss": 1.0278, + "step": 27899 + }, + { + "epoch": 0.89, + "learning_rate": 6.027195768173777e-07, + "loss": 0.8418, + "step": 27900 + }, + { + "epoch": 0.89, + "learning_rate": 6.023654623162555e-07, + "loss": 0.9102, + "step": 27901 + }, + { + "epoch": 0.89, + "learning_rate": 6.020114486413531e-07, + "loss": 0.8892, + "step": 27902 + }, + { + "epoch": 0.89, + "learning_rate": 6.016575357964649e-07, + "loss": 0.9487, + "step": 27903 + }, + { + "epoch": 0.89, + "learning_rate": 6.013037237853903e-07, + "loss": 0.9106, + "step": 27904 + }, + { + "epoch": 0.89, + "learning_rate": 6.00950012611925e-07, + "loss": 0.8535, + "step": 27905 + }, + { + "epoch": 0.89, + "learning_rate": 6.005964022798638e-07, + "loss": 0.874, + "step": 27906 + }, + { + "epoch": 0.89, + "learning_rate": 6.002428927930015e-07, + "loss": 0.8652, + "step": 27907 + }, + { + "epoch": 0.89, + "learning_rate": 5.998894841551295e-07, + "loss": 0.8408, + "step": 27908 + }, + { + "epoch": 0.89, + "learning_rate": 5.995361763700381e-07, + "loss": 1.001, + "step": 27909 + }, + { + "epoch": 0.89, + "learning_rate": 5.991829694415208e-07, + "loss": 0.8828, + "step": 27910 + }, + { + "epoch": 0.89, + "learning_rate": 5.988298633733658e-07, + "loss": 0.9302, + "step": 27911 + }, + { + "epoch": 0.89, + "learning_rate": 5.984768581693611e-07, + "loss": 0.873, + "step": 27912 + }, + { + "epoch": 0.89, + "learning_rate": 5.981239538332961e-07, + "loss": 0.9546, + "step": 27913 + }, + { + "epoch": 0.89, + "learning_rate": 5.977711503689543e-07, + "loss": 0.8042, + "step": 27914 + }, + { + "epoch": 0.89, + "learning_rate": 5.974184477801215e-07, + "loss": 0.8057, + "step": 27915 + }, + { + "epoch": 0.89, + "learning_rate": 5.970658460705847e-07, + "loss": 0.8071, + "step": 27916 + }, + { + "epoch": 0.89, + "learning_rate": 5.967133452441243e-07, + "loss": 0.8145, + "step": 27917 + }, + { + "epoch": 0.89, + "learning_rate": 5.963609453045205e-07, + "loss": 0.8408, + "step": 27918 + }, + { + "epoch": 0.89, + "learning_rate": 5.960086462555581e-07, + "loss": 0.7988, + "step": 27919 + }, + { + "epoch": 0.89, + "learning_rate": 5.95656448101014e-07, + "loss": 0.9062, + "step": 27920 + }, + { + "epoch": 0.89, + "learning_rate": 5.953043508446687e-07, + "loss": 0.8125, + "step": 27921 + }, + { + "epoch": 0.89, + "learning_rate": 5.949523544903002e-07, + "loss": 0.957, + "step": 27922 + }, + { + "epoch": 0.89, + "learning_rate": 5.946004590416832e-07, + "loss": 0.8706, + "step": 27923 + }, + { + "epoch": 0.89, + "learning_rate": 5.942486645025925e-07, + "loss": 0.917, + "step": 27924 + }, + { + "epoch": 0.89, + "learning_rate": 5.938969708768061e-07, + "loss": 0.9697, + "step": 27925 + }, + { + "epoch": 0.89, + "learning_rate": 5.935453781680922e-07, + "loss": 0.9434, + "step": 27926 + }, + { + "epoch": 0.89, + "learning_rate": 5.931938863802289e-07, + "loss": 0.8823, + "step": 27927 + }, + { + "epoch": 0.89, + "learning_rate": 5.928424955169821e-07, + "loss": 0.8325, + "step": 27928 + }, + { + "epoch": 0.89, + "learning_rate": 5.924912055821264e-07, + "loss": 0.8779, + "step": 27929 + }, + { + "epoch": 0.89, + "learning_rate": 5.921400165794255e-07, + "loss": 0.9624, + "step": 27930 + }, + { + "epoch": 0.89, + "learning_rate": 5.917889285126521e-07, + "loss": 0.8408, + "step": 27931 + }, + { + "epoch": 0.89, + "learning_rate": 5.914379413855698e-07, + "loss": 0.8428, + "step": 27932 + }, + { + "epoch": 0.89, + "learning_rate": 5.910870552019455e-07, + "loss": 0.8284, + "step": 27933 + }, + { + "epoch": 0.89, + "learning_rate": 5.907362699655439e-07, + "loss": 0.8911, + "step": 27934 + }, + { + "epoch": 0.89, + "learning_rate": 5.903855856801288e-07, + "loss": 0.9775, + "step": 27935 + }, + { + "epoch": 0.89, + "learning_rate": 5.900350023494616e-07, + "loss": 0.8545, + "step": 27936 + }, + { + "epoch": 0.89, + "learning_rate": 5.896845199773071e-07, + "loss": 1.0776, + "step": 27937 + }, + { + "epoch": 0.89, + "learning_rate": 5.8933413856742e-07, + "loss": 0.8828, + "step": 27938 + }, + { + "epoch": 0.89, + "learning_rate": 5.889838581235641e-07, + "loss": 0.9116, + "step": 27939 + }, + { + "epoch": 0.89, + "learning_rate": 5.886336786494939e-07, + "loss": 0.8833, + "step": 27940 + }, + { + "epoch": 0.89, + "learning_rate": 5.8828360014897e-07, + "loss": 0.9229, + "step": 27941 + }, + { + "epoch": 0.89, + "learning_rate": 5.879336226257459e-07, + "loss": 1.0371, + "step": 27942 + }, + { + "epoch": 0.89, + "learning_rate": 5.875837460835776e-07, + "loss": 0.9521, + "step": 27943 + }, + { + "epoch": 0.89, + "learning_rate": 5.872339705262198e-07, + "loss": 0.9346, + "step": 27944 + }, + { + "epoch": 0.89, + "learning_rate": 5.868842959574228e-07, + "loss": 0.4727, + "step": 27945 + }, + { + "epoch": 0.89, + "learning_rate": 5.865347223809381e-07, + "loss": 1.0435, + "step": 27946 + }, + { + "epoch": 0.89, + "learning_rate": 5.861852498005194e-07, + "loss": 0.9351, + "step": 27947 + }, + { + "epoch": 0.89, + "learning_rate": 5.858358782199125e-07, + "loss": 1.0098, + "step": 27948 + }, + { + "epoch": 0.89, + "learning_rate": 5.85486607642869e-07, + "loss": 0.9238, + "step": 27949 + }, + { + "epoch": 0.89, + "learning_rate": 5.851374380731345e-07, + "loss": 0.8218, + "step": 27950 + }, + { + "epoch": 0.89, + "learning_rate": 5.847883695144551e-07, + "loss": 0.895, + "step": 27951 + }, + { + "epoch": 0.89, + "learning_rate": 5.844394019705768e-07, + "loss": 1.0488, + "step": 27952 + }, + { + "epoch": 0.89, + "learning_rate": 5.840905354452431e-07, + "loss": 0.9194, + "step": 27953 + }, + { + "epoch": 0.89, + "learning_rate": 5.837417699421954e-07, + "loss": 0.96, + "step": 27954 + }, + { + "epoch": 0.89, + "learning_rate": 5.833931054651787e-07, + "loss": 0.9712, + "step": 27955 + }, + { + "epoch": 0.89, + "learning_rate": 5.830445420179331e-07, + "loss": 0.9404, + "step": 27956 + }, + { + "epoch": 0.89, + "learning_rate": 5.826960796041948e-07, + "loss": 0.9116, + "step": 27957 + }, + { + "epoch": 0.89, + "learning_rate": 5.823477182277071e-07, + "loss": 0.9414, + "step": 27958 + }, + { + "epoch": 0.89, + "learning_rate": 5.81999457892205e-07, + "loss": 0.9468, + "step": 27959 + }, + { + "epoch": 0.89, + "learning_rate": 5.816512986014234e-07, + "loss": 0.9043, + "step": 27960 + }, + { + "epoch": 0.89, + "learning_rate": 5.813032403591023e-07, + "loss": 1.0068, + "step": 27961 + }, + { + "epoch": 0.89, + "learning_rate": 5.809552831689724e-07, + "loss": 1.0098, + "step": 27962 + }, + { + "epoch": 0.89, + "learning_rate": 5.806074270347673e-07, + "loss": 0.957, + "step": 27963 + }, + { + "epoch": 0.89, + "learning_rate": 5.802596719602205e-07, + "loss": 0.9326, + "step": 27964 + }, + { + "epoch": 0.89, + "learning_rate": 5.799120179490603e-07, + "loss": 0.9058, + "step": 27965 + }, + { + "epoch": 0.89, + "learning_rate": 5.795644650050202e-07, + "loss": 1.0073, + "step": 27966 + }, + { + "epoch": 0.89, + "learning_rate": 5.792170131318287e-07, + "loss": 0.9043, + "step": 27967 + }, + { + "epoch": 0.89, + "learning_rate": 5.788696623332102e-07, + "loss": 0.7881, + "step": 27968 + }, + { + "epoch": 0.89, + "learning_rate": 5.78522412612893e-07, + "loss": 0.7969, + "step": 27969 + }, + { + "epoch": 0.89, + "learning_rate": 5.781752639746052e-07, + "loss": 0.9873, + "step": 27970 + }, + { + "epoch": 0.89, + "learning_rate": 5.778282164220672e-07, + "loss": 0.8613, + "step": 27971 + }, + { + "epoch": 0.89, + "learning_rate": 5.77481269959006e-07, + "loss": 0.8799, + "step": 27972 + }, + { + "epoch": 0.89, + "learning_rate": 5.771344245891408e-07, + "loss": 0.8599, + "step": 27973 + }, + { + "epoch": 0.89, + "learning_rate": 5.767876803161954e-07, + "loss": 1.0127, + "step": 27974 + }, + { + "epoch": 0.89, + "learning_rate": 5.7644103714389e-07, + "loss": 0.8472, + "step": 27975 + }, + { + "epoch": 0.89, + "learning_rate": 5.760944950759429e-07, + "loss": 0.9834, + "step": 27976 + }, + { + "epoch": 0.89, + "learning_rate": 5.757480541160699e-07, + "loss": 0.918, + "step": 27977 + }, + { + "epoch": 0.89, + "learning_rate": 5.754017142679913e-07, + "loss": 0.8398, + "step": 27978 + }, + { + "epoch": 0.89, + "learning_rate": 5.750554755354209e-07, + "loss": 0.8379, + "step": 27979 + }, + { + "epoch": 0.89, + "learning_rate": 5.747093379220758e-07, + "loss": 0.9434, + "step": 27980 + }, + { + "epoch": 0.89, + "learning_rate": 5.743633014316663e-07, + "loss": 0.8896, + "step": 27981 + }, + { + "epoch": 0.89, + "learning_rate": 5.740173660679094e-07, + "loss": 0.978, + "step": 27982 + }, + { + "epoch": 0.89, + "learning_rate": 5.736715318345121e-07, + "loss": 0.9644, + "step": 27983 + }, + { + "epoch": 0.89, + "learning_rate": 5.73325798735187e-07, + "loss": 0.9438, + "step": 27984 + }, + { + "epoch": 0.89, + "learning_rate": 5.729801667736423e-07, + "loss": 0.9512, + "step": 27985 + }, + { + "epoch": 0.9, + "learning_rate": 5.726346359535884e-07, + "loss": 0.8975, + "step": 27986 + }, + { + "epoch": 0.9, + "learning_rate": 5.7228920627873e-07, + "loss": 0.8618, + "step": 27987 + }, + { + "epoch": 0.9, + "learning_rate": 5.719438777527752e-07, + "loss": 0.9546, + "step": 27988 + }, + { + "epoch": 0.9, + "learning_rate": 5.715986503794279e-07, + "loss": 0.8984, + "step": 27989 + }, + { + "epoch": 0.9, + "learning_rate": 5.71253524162394e-07, + "loss": 0.9492, + "step": 27990 + }, + { + "epoch": 0.9, + "learning_rate": 5.709084991053715e-07, + "loss": 0.9219, + "step": 27991 + }, + { + "epoch": 0.9, + "learning_rate": 5.705635752120675e-07, + "loss": 0.938, + "step": 27992 + }, + { + "epoch": 0.9, + "learning_rate": 5.702187524861791e-07, + "loss": 0.9468, + "step": 27993 + }, + { + "epoch": 0.9, + "learning_rate": 5.698740309314077e-07, + "loss": 0.9321, + "step": 27994 + }, + { + "epoch": 0.9, + "learning_rate": 5.695294105514515e-07, + "loss": 0.9844, + "step": 27995 + }, + { + "epoch": 0.9, + "learning_rate": 5.691848913500064e-07, + "loss": 1.0479, + "step": 27996 + }, + { + "epoch": 0.9, + "learning_rate": 5.688404733307706e-07, + "loss": 0.8999, + "step": 27997 + }, + { + "epoch": 0.9, + "learning_rate": 5.684961564974389e-07, + "loss": 0.8281, + "step": 27998 + }, + { + "epoch": 0.9, + "learning_rate": 5.681519408537039e-07, + "loss": 0.8716, + "step": 27999 + }, + { + "epoch": 0.9, + "learning_rate": 5.678078264032616e-07, + "loss": 0.9985, + "step": 28000 + }, + { + "epoch": 0.9, + "learning_rate": 5.674638131498022e-07, + "loss": 0.9153, + "step": 28001 + }, + { + "epoch": 0.9, + "learning_rate": 5.671199010970152e-07, + "loss": 0.9087, + "step": 28002 + }, + { + "epoch": 0.9, + "learning_rate": 5.667760902485931e-07, + "loss": 0.939, + "step": 28003 + }, + { + "epoch": 0.9, + "learning_rate": 5.664323806082228e-07, + "loss": 0.8325, + "step": 28004 + }, + { + "epoch": 0.9, + "learning_rate": 5.660887721795927e-07, + "loss": 0.8269, + "step": 28005 + }, + { + "epoch": 0.9, + "learning_rate": 5.657452649663898e-07, + "loss": 0.8374, + "step": 28006 + }, + { + "epoch": 0.9, + "learning_rate": 5.654018589722987e-07, + "loss": 0.8203, + "step": 28007 + }, + { + "epoch": 0.9, + "learning_rate": 5.650585542010023e-07, + "loss": 0.9839, + "step": 28008 + }, + { + "epoch": 0.9, + "learning_rate": 5.647153506561876e-07, + "loss": 1.042, + "step": 28009 + }, + { + "epoch": 0.9, + "learning_rate": 5.643722483415325e-07, + "loss": 1.0801, + "step": 28010 + }, + { + "epoch": 0.9, + "learning_rate": 5.640292472607223e-07, + "loss": 0.7446, + "step": 28011 + }, + { + "epoch": 0.9, + "learning_rate": 5.636863474174359e-07, + "loss": 0.8718, + "step": 28012 + }, + { + "epoch": 0.9, + "learning_rate": 5.633435488153505e-07, + "loss": 1.0356, + "step": 28013 + }, + { + "epoch": 0.9, + "learning_rate": 5.630008514581431e-07, + "loss": 0.937, + "step": 28014 + }, + { + "epoch": 0.9, + "learning_rate": 5.626582553494942e-07, + "loss": 0.9146, + "step": 28015 + }, + { + "epoch": 0.9, + "learning_rate": 5.623157604930751e-07, + "loss": 1.0127, + "step": 28016 + }, + { + "epoch": 0.9, + "learning_rate": 5.619733668925653e-07, + "loss": 0.9102, + "step": 28017 + }, + { + "epoch": 0.9, + "learning_rate": 5.61631074551634e-07, + "loss": 1.0332, + "step": 28018 + }, + { + "epoch": 0.9, + "learning_rate": 5.612888834739571e-07, + "loss": 0.9995, + "step": 28019 + }, + { + "epoch": 0.9, + "learning_rate": 5.60946793663204e-07, + "loss": 0.9653, + "step": 28020 + }, + { + "epoch": 0.9, + "learning_rate": 5.60604805123045e-07, + "loss": 0.9199, + "step": 28021 + }, + { + "epoch": 0.9, + "learning_rate": 5.602629178571473e-07, + "loss": 0.8301, + "step": 28022 + }, + { + "epoch": 0.9, + "learning_rate": 5.599211318691833e-07, + "loss": 0.9111, + "step": 28023 + }, + { + "epoch": 0.9, + "learning_rate": 5.595794471628169e-07, + "loss": 0.8252, + "step": 28024 + }, + { + "epoch": 0.9, + "learning_rate": 5.592378637417162e-07, + "loss": 0.8901, + "step": 28025 + }, + { + "epoch": 0.9, + "learning_rate": 5.588963816095428e-07, + "loss": 0.9316, + "step": 28026 + }, + { + "epoch": 0.9, + "learning_rate": 5.585550007699648e-07, + "loss": 0.9238, + "step": 28027 + }, + { + "epoch": 0.9, + "learning_rate": 5.582137212266403e-07, + "loss": 0.9482, + "step": 28028 + }, + { + "epoch": 0.9, + "learning_rate": 5.578725429832344e-07, + "loss": 0.8525, + "step": 28029 + }, + { + "epoch": 0.9, + "learning_rate": 5.575314660434039e-07, + "loss": 0.9644, + "step": 28030 + }, + { + "epoch": 0.9, + "learning_rate": 5.571904904108127e-07, + "loss": 0.8418, + "step": 28031 + }, + { + "epoch": 0.9, + "learning_rate": 5.568496160891146e-07, + "loss": 0.8511, + "step": 28032 + }, + { + "epoch": 0.9, + "learning_rate": 5.56508843081971e-07, + "loss": 1.0005, + "step": 28033 + }, + { + "epoch": 0.9, + "learning_rate": 5.561681713930344e-07, + "loss": 0.9531, + "step": 28034 + }, + { + "epoch": 0.9, + "learning_rate": 5.558276010259634e-07, + "loss": 0.9253, + "step": 28035 + }, + { + "epoch": 0.9, + "learning_rate": 5.55487131984408e-07, + "loss": 1.0635, + "step": 28036 + }, + { + "epoch": 0.9, + "learning_rate": 5.551467642720232e-07, + "loss": 0.7954, + "step": 28037 + }, + { + "epoch": 0.9, + "learning_rate": 5.548064978924605e-07, + "loss": 0.9575, + "step": 28038 + }, + { + "epoch": 0.9, + "learning_rate": 5.544663328493716e-07, + "loss": 0.981, + "step": 28039 + }, + { + "epoch": 0.9, + "learning_rate": 5.541262691464055e-07, + "loss": 0.8481, + "step": 28040 + }, + { + "epoch": 0.9, + "learning_rate": 5.537863067872085e-07, + "loss": 0.8926, + "step": 28041 + }, + { + "epoch": 0.9, + "learning_rate": 5.534464457754319e-07, + "loss": 0.9893, + "step": 28042 + }, + { + "epoch": 0.9, + "learning_rate": 5.531066861147194e-07, + "loss": 0.9297, + "step": 28043 + }, + { + "epoch": 0.9, + "learning_rate": 5.527670278087149e-07, + "loss": 0.8345, + "step": 28044 + }, + { + "epoch": 0.9, + "learning_rate": 5.524274708610666e-07, + "loss": 1.0542, + "step": 28045 + }, + { + "epoch": 0.9, + "learning_rate": 5.52088015275416e-07, + "loss": 0.8794, + "step": 28046 + }, + { + "epoch": 0.9, + "learning_rate": 5.517486610554024e-07, + "loss": 1.0503, + "step": 28047 + }, + { + "epoch": 0.9, + "learning_rate": 5.514094082046706e-07, + "loss": 0.8711, + "step": 28048 + }, + { + "epoch": 0.9, + "learning_rate": 5.510702567268578e-07, + "loss": 0.8071, + "step": 28049 + }, + { + "epoch": 0.9, + "learning_rate": 5.507312066256054e-07, + "loss": 0.8726, + "step": 28050 + }, + { + "epoch": 0.9, + "learning_rate": 5.503922579045474e-07, + "loss": 0.957, + "step": 28051 + }, + { + "epoch": 0.9, + "learning_rate": 5.500534105673228e-07, + "loss": 0.9082, + "step": 28052 + }, + { + "epoch": 0.9, + "learning_rate": 5.497146646175644e-07, + "loss": 0.9062, + "step": 28053 + }, + { + "epoch": 0.9, + "learning_rate": 5.493760200589116e-07, + "loss": 0.877, + "step": 28054 + }, + { + "epoch": 0.9, + "learning_rate": 5.490374768949924e-07, + "loss": 0.9712, + "step": 28055 + }, + { + "epoch": 0.9, + "learning_rate": 5.48699035129443e-07, + "loss": 0.9707, + "step": 28056 + }, + { + "epoch": 0.9, + "learning_rate": 5.483606947658904e-07, + "loss": 0.9136, + "step": 28057 + }, + { + "epoch": 0.9, + "learning_rate": 5.480224558079706e-07, + "loss": 0.9731, + "step": 28058 + }, + { + "epoch": 0.9, + "learning_rate": 5.476843182593062e-07, + "loss": 0.9468, + "step": 28059 + }, + { + "epoch": 0.9, + "learning_rate": 5.473462821235287e-07, + "loss": 0.8789, + "step": 28060 + }, + { + "epoch": 0.9, + "learning_rate": 5.470083474042631e-07, + "loss": 0.825, + "step": 28061 + }, + { + "epoch": 0.9, + "learning_rate": 5.466705141051365e-07, + "loss": 0.981, + "step": 28062 + }, + { + "epoch": 0.9, + "learning_rate": 5.463327822297726e-07, + "loss": 0.8682, + "step": 28063 + }, + { + "epoch": 0.9, + "learning_rate": 5.459951517817951e-07, + "loss": 0.9185, + "step": 28064 + }, + { + "epoch": 0.9, + "learning_rate": 5.45657622764828e-07, + "loss": 1.0161, + "step": 28065 + }, + { + "epoch": 0.9, + "learning_rate": 5.453201951824893e-07, + "loss": 0.8545, + "step": 28066 + }, + { + "epoch": 0.9, + "learning_rate": 5.449828690384007e-07, + "loss": 0.9756, + "step": 28067 + }, + { + "epoch": 0.9, + "learning_rate": 5.446456443361836e-07, + "loss": 0.9194, + "step": 28068 + }, + { + "epoch": 0.9, + "learning_rate": 5.443085210794519e-07, + "loss": 1.0088, + "step": 28069 + }, + { + "epoch": 0.9, + "learning_rate": 5.43971499271826e-07, + "loss": 0.9229, + "step": 28070 + }, + { + "epoch": 0.9, + "learning_rate": 5.436345789169196e-07, + "loss": 0.9634, + "step": 28071 + }, + { + "epoch": 0.9, + "learning_rate": 5.432977600183486e-07, + "loss": 0.8613, + "step": 28072 + }, + { + "epoch": 0.9, + "learning_rate": 5.429610425797283e-07, + "loss": 0.833, + "step": 28073 + }, + { + "epoch": 0.9, + "learning_rate": 5.426244266046676e-07, + "loss": 0.9663, + "step": 28074 + }, + { + "epoch": 0.9, + "learning_rate": 5.422879120967794e-07, + "loss": 0.8872, + "step": 28075 + }, + { + "epoch": 0.9, + "learning_rate": 5.419514990596763e-07, + "loss": 0.9038, + "step": 28076 + }, + { + "epoch": 0.9, + "learning_rate": 5.416151874969644e-07, + "loss": 0.9556, + "step": 28077 + }, + { + "epoch": 0.9, + "learning_rate": 5.412789774122552e-07, + "loss": 0.959, + "step": 28078 + }, + { + "epoch": 0.9, + "learning_rate": 5.409428688091534e-07, + "loss": 0.9199, + "step": 28079 + }, + { + "epoch": 0.9, + "learning_rate": 5.406068616912663e-07, + "loss": 0.8862, + "step": 28080 + }, + { + "epoch": 0.9, + "learning_rate": 5.402709560621966e-07, + "loss": 0.8848, + "step": 28081 + }, + { + "epoch": 0.9, + "learning_rate": 5.399351519255524e-07, + "loss": 0.9043, + "step": 28082 + }, + { + "epoch": 0.9, + "learning_rate": 5.395994492849321e-07, + "loss": 0.894, + "step": 28083 + }, + { + "epoch": 0.9, + "learning_rate": 5.392638481439406e-07, + "loss": 0.9556, + "step": 28084 + }, + { + "epoch": 0.9, + "learning_rate": 5.389283485061781e-07, + "loss": 0.8755, + "step": 28085 + }, + { + "epoch": 0.9, + "learning_rate": 5.385929503752418e-07, + "loss": 0.855, + "step": 28086 + }, + { + "epoch": 0.9, + "learning_rate": 5.382576537547324e-07, + "loss": 1.0, + "step": 28087 + }, + { + "epoch": 0.9, + "learning_rate": 5.379224586482479e-07, + "loss": 0.9648, + "step": 28088 + }, + { + "epoch": 0.9, + "learning_rate": 5.375873650593822e-07, + "loss": 0.833, + "step": 28089 + }, + { + "epoch": 0.9, + "learning_rate": 5.372523729917312e-07, + "loss": 0.9404, + "step": 28090 + }, + { + "epoch": 0.9, + "learning_rate": 5.369174824488899e-07, + "loss": 0.7886, + "step": 28091 + }, + { + "epoch": 0.9, + "learning_rate": 5.365826934344498e-07, + "loss": 0.9922, + "step": 28092 + }, + { + "epoch": 0.9, + "learning_rate": 5.362480059520048e-07, + "loss": 0.9639, + "step": 28093 + }, + { + "epoch": 0.9, + "learning_rate": 5.359134200051441e-07, + "loss": 0.9697, + "step": 28094 + }, + { + "epoch": 0.9, + "learning_rate": 5.355789355974605e-07, + "loss": 0.8408, + "step": 28095 + }, + { + "epoch": 0.9, + "learning_rate": 5.352445527325368e-07, + "loss": 0.8555, + "step": 28096 + }, + { + "epoch": 0.9, + "learning_rate": 5.349102714139664e-07, + "loss": 0.9663, + "step": 28097 + }, + { + "epoch": 0.9, + "learning_rate": 5.345760916453313e-07, + "loss": 0.9814, + "step": 28098 + }, + { + "epoch": 0.9, + "learning_rate": 5.342420134302207e-07, + "loss": 0.9214, + "step": 28099 + }, + { + "epoch": 0.9, + "learning_rate": 5.33908036772215e-07, + "loss": 0.709, + "step": 28100 + }, + { + "epoch": 0.9, + "learning_rate": 5.335741616749013e-07, + "loss": 0.9995, + "step": 28101 + }, + { + "epoch": 0.9, + "learning_rate": 5.332403881418591e-07, + "loss": 0.9253, + "step": 28102 + }, + { + "epoch": 0.9, + "learning_rate": 5.329067161766721e-07, + "loss": 0.9424, + "step": 28103 + }, + { + "epoch": 0.9, + "learning_rate": 5.325731457829165e-07, + "loss": 0.4731, + "step": 28104 + }, + { + "epoch": 0.9, + "learning_rate": 5.322396769641736e-07, + "loss": 0.9141, + "step": 28105 + }, + { + "epoch": 0.9, + "learning_rate": 5.319063097240196e-07, + "loss": 0.9336, + "step": 28106 + }, + { + "epoch": 0.9, + "learning_rate": 5.315730440660338e-07, + "loss": 0.8782, + "step": 28107 + }, + { + "epoch": 0.9, + "learning_rate": 5.312398799937879e-07, + "loss": 0.8232, + "step": 28108 + }, + { + "epoch": 0.9, + "learning_rate": 5.309068175108612e-07, + "loss": 0.7407, + "step": 28109 + }, + { + "epoch": 0.9, + "learning_rate": 5.305738566208229e-07, + "loss": 0.8171, + "step": 28110 + }, + { + "epoch": 0.9, + "learning_rate": 5.302409973272482e-07, + "loss": 0.5115, + "step": 28111 + }, + { + "epoch": 0.9, + "learning_rate": 5.299082396337052e-07, + "loss": 0.8242, + "step": 28112 + }, + { + "epoch": 0.9, + "learning_rate": 5.295755835437666e-07, + "loss": 0.4915, + "step": 28113 + }, + { + "epoch": 0.9, + "learning_rate": 5.292430290610006e-07, + "loss": 0.8936, + "step": 28114 + }, + { + "epoch": 0.9, + "learning_rate": 5.289105761889757e-07, + "loss": 0.8843, + "step": 28115 + }, + { + "epoch": 0.9, + "learning_rate": 5.285782249312565e-07, + "loss": 0.9194, + "step": 28116 + }, + { + "epoch": 0.9, + "learning_rate": 5.282459752914127e-07, + "loss": 1.0176, + "step": 28117 + }, + { + "epoch": 0.9, + "learning_rate": 5.279138272730055e-07, + "loss": 0.9116, + "step": 28118 + }, + { + "epoch": 0.9, + "learning_rate": 5.275817808796013e-07, + "loss": 0.9033, + "step": 28119 + }, + { + "epoch": 0.9, + "learning_rate": 5.272498361147582e-07, + "loss": 0.897, + "step": 28120 + }, + { + "epoch": 0.9, + "learning_rate": 5.269179929820434e-07, + "loss": 0.9644, + "step": 28121 + }, + { + "epoch": 0.9, + "learning_rate": 5.265862514850118e-07, + "loss": 0.9805, + "step": 28122 + }, + { + "epoch": 0.9, + "learning_rate": 5.26254611627226e-07, + "loss": 0.896, + "step": 28123 + }, + { + "epoch": 0.9, + "learning_rate": 5.259230734122444e-07, + "loss": 0.8384, + "step": 28124 + }, + { + "epoch": 0.9, + "learning_rate": 5.255916368436209e-07, + "loss": 0.8555, + "step": 28125 + }, + { + "epoch": 0.9, + "learning_rate": 5.252603019249147e-07, + "loss": 0.9702, + "step": 28126 + }, + { + "epoch": 0.9, + "learning_rate": 5.249290686596798e-07, + "loss": 0.9619, + "step": 28127 + }, + { + "epoch": 0.9, + "learning_rate": 5.245979370514686e-07, + "loss": 0.887, + "step": 28128 + }, + { + "epoch": 0.9, + "learning_rate": 5.242669071038342e-07, + "loss": 0.9209, + "step": 28129 + }, + { + "epoch": 0.9, + "learning_rate": 5.239359788203291e-07, + "loss": 0.9858, + "step": 28130 + }, + { + "epoch": 0.9, + "learning_rate": 5.236051522045027e-07, + "loss": 0.8604, + "step": 28131 + }, + { + "epoch": 0.9, + "learning_rate": 5.232744272599067e-07, + "loss": 0.9272, + "step": 28132 + }, + { + "epoch": 0.9, + "learning_rate": 5.229438039900869e-07, + "loss": 0.8047, + "step": 28133 + }, + { + "epoch": 0.9, + "learning_rate": 5.226132823985919e-07, + "loss": 0.981, + "step": 28134 + }, + { + "epoch": 0.9, + "learning_rate": 5.222828624889664e-07, + "loss": 0.8965, + "step": 28135 + }, + { + "epoch": 0.9, + "learning_rate": 5.219525442647566e-07, + "loss": 1.0151, + "step": 28136 + }, + { + "epoch": 0.9, + "learning_rate": 5.216223277295063e-07, + "loss": 0.9316, + "step": 28137 + }, + { + "epoch": 0.9, + "learning_rate": 5.212922128867582e-07, + "loss": 0.9331, + "step": 28138 + }, + { + "epoch": 0.9, + "learning_rate": 5.209621997400538e-07, + "loss": 1.0078, + "step": 28139 + }, + { + "epoch": 0.9, + "learning_rate": 5.206322882929349e-07, + "loss": 0.916, + "step": 28140 + }, + { + "epoch": 0.9, + "learning_rate": 5.203024785489408e-07, + "loss": 0.8862, + "step": 28141 + }, + { + "epoch": 0.9, + "learning_rate": 5.199727705116086e-07, + "loss": 1.04, + "step": 28142 + }, + { + "epoch": 0.9, + "learning_rate": 5.196431641844768e-07, + "loss": 0.9473, + "step": 28143 + }, + { + "epoch": 0.9, + "learning_rate": 5.193136595710813e-07, + "loss": 0.9077, + "step": 28144 + }, + { + "epoch": 0.9, + "learning_rate": 5.18984256674957e-07, + "loss": 1.0005, + "step": 28145 + }, + { + "epoch": 0.9, + "learning_rate": 5.1865495549964e-07, + "loss": 0.7324, + "step": 28146 + }, + { + "epoch": 0.9, + "learning_rate": 5.183257560486598e-07, + "loss": 0.7866, + "step": 28147 + }, + { + "epoch": 0.9, + "learning_rate": 5.179966583255535e-07, + "loss": 0.9443, + "step": 28148 + }, + { + "epoch": 0.9, + "learning_rate": 5.176676623338461e-07, + "loss": 0.8008, + "step": 28149 + }, + { + "epoch": 0.9, + "learning_rate": 5.173387680770725e-07, + "loss": 0.8379, + "step": 28150 + }, + { + "epoch": 0.9, + "learning_rate": 5.170099755587566e-07, + "loss": 1.0244, + "step": 28151 + }, + { + "epoch": 0.9, + "learning_rate": 5.166812847824299e-07, + "loss": 1.0547, + "step": 28152 + }, + { + "epoch": 0.9, + "learning_rate": 5.163526957516162e-07, + "loss": 0.9658, + "step": 28153 + }, + { + "epoch": 0.9, + "learning_rate": 5.16024208469843e-07, + "loss": 0.9565, + "step": 28154 + }, + { + "epoch": 0.9, + "learning_rate": 5.156958229406328e-07, + "loss": 0.9692, + "step": 28155 + }, + { + "epoch": 0.9, + "learning_rate": 5.153675391675128e-07, + "loss": 0.936, + "step": 28156 + }, + { + "epoch": 0.9, + "learning_rate": 5.150393571539991e-07, + "loss": 0.9033, + "step": 28157 + }, + { + "epoch": 0.9, + "learning_rate": 5.147112769036167e-07, + "loss": 0.8999, + "step": 28158 + }, + { + "epoch": 0.9, + "learning_rate": 5.143832984198827e-07, + "loss": 0.9229, + "step": 28159 + }, + { + "epoch": 0.9, + "learning_rate": 5.140554217063199e-07, + "loss": 0.8496, + "step": 28160 + }, + { + "epoch": 0.9, + "learning_rate": 5.137276467664421e-07, + "loss": 0.9185, + "step": 28161 + }, + { + "epoch": 0.9, + "learning_rate": 5.133999736037698e-07, + "loss": 0.8447, + "step": 28162 + }, + { + "epoch": 0.9, + "learning_rate": 5.130724022218159e-07, + "loss": 0.9072, + "step": 28163 + }, + { + "epoch": 0.9, + "learning_rate": 5.127449326240952e-07, + "loss": 0.9399, + "step": 28164 + }, + { + "epoch": 0.9, + "learning_rate": 5.124175648141206e-07, + "loss": 0.9883, + "step": 28165 + }, + { + "epoch": 0.9, + "learning_rate": 5.120902987954068e-07, + "loss": 0.7725, + "step": 28166 + }, + { + "epoch": 0.9, + "learning_rate": 5.117631345714635e-07, + "loss": 0.9624, + "step": 28167 + }, + { + "epoch": 0.9, + "learning_rate": 5.114360721457989e-07, + "loss": 0.7842, + "step": 28168 + }, + { + "epoch": 0.9, + "learning_rate": 5.111091115219246e-07, + "loss": 0.9751, + "step": 28169 + }, + { + "epoch": 0.9, + "learning_rate": 5.107822527033468e-07, + "loss": 0.9546, + "step": 28170 + }, + { + "epoch": 0.9, + "learning_rate": 5.104554956935748e-07, + "loss": 0.8362, + "step": 28171 + }, + { + "epoch": 0.9, + "learning_rate": 5.101288404961136e-07, + "loss": 0.8984, + "step": 28172 + }, + { + "epoch": 0.9, + "learning_rate": 5.098022871144659e-07, + "loss": 1.002, + "step": 28173 + }, + { + "epoch": 0.9, + "learning_rate": 5.094758355521356e-07, + "loss": 0.8965, + "step": 28174 + }, + { + "epoch": 0.9, + "learning_rate": 5.091494858126266e-07, + "loss": 0.9634, + "step": 28175 + }, + { + "epoch": 0.9, + "learning_rate": 5.088232378994395e-07, + "loss": 0.8818, + "step": 28176 + }, + { + "epoch": 0.9, + "learning_rate": 5.084970918160759e-07, + "loss": 0.9624, + "step": 28177 + }, + { + "epoch": 0.9, + "learning_rate": 5.081710475660329e-07, + "loss": 0.9263, + "step": 28178 + }, + { + "epoch": 0.9, + "learning_rate": 5.078451051528099e-07, + "loss": 0.8232, + "step": 28179 + }, + { + "epoch": 0.9, + "learning_rate": 5.075192645799032e-07, + "loss": 0.897, + "step": 28180 + }, + { + "epoch": 0.9, + "learning_rate": 5.071935258508087e-07, + "loss": 0.9741, + "step": 28181 + }, + { + "epoch": 0.9, + "learning_rate": 5.068678889690215e-07, + "loss": 0.9375, + "step": 28182 + }, + { + "epoch": 0.9, + "learning_rate": 5.065423539380365e-07, + "loss": 0.9248, + "step": 28183 + }, + { + "epoch": 0.9, + "learning_rate": 5.062169207613432e-07, + "loss": 0.8328, + "step": 28184 + }, + { + "epoch": 0.9, + "learning_rate": 5.058915894424377e-07, + "loss": 0.9536, + "step": 28185 + }, + { + "epoch": 0.9, + "learning_rate": 5.055663599848071e-07, + "loss": 0.9634, + "step": 28186 + }, + { + "epoch": 0.9, + "learning_rate": 5.052412323919409e-07, + "loss": 0.9663, + "step": 28187 + }, + { + "epoch": 0.9, + "learning_rate": 5.049162066673274e-07, + "loss": 0.8926, + "step": 28188 + }, + { + "epoch": 0.9, + "learning_rate": 5.045912828144561e-07, + "loss": 0.9429, + "step": 28189 + }, + { + "epoch": 0.9, + "learning_rate": 5.042664608368086e-07, + "loss": 0.9058, + "step": 28190 + }, + { + "epoch": 0.9, + "learning_rate": 5.039417407378755e-07, + "loss": 0.8379, + "step": 28191 + }, + { + "epoch": 0.9, + "learning_rate": 5.036171225211362e-07, + "loss": 0.8369, + "step": 28192 + }, + { + "epoch": 0.9, + "learning_rate": 5.032926061900767e-07, + "loss": 0.9302, + "step": 28193 + }, + { + "epoch": 0.9, + "learning_rate": 5.029681917481754e-07, + "loss": 0.9766, + "step": 28194 + }, + { + "epoch": 0.9, + "learning_rate": 5.026438791989152e-07, + "loss": 0.8711, + "step": 28195 + }, + { + "epoch": 0.9, + "learning_rate": 5.023196685457743e-07, + "loss": 0.9126, + "step": 28196 + }, + { + "epoch": 0.9, + "learning_rate": 5.019955597922332e-07, + "loss": 0.4717, + "step": 28197 + }, + { + "epoch": 0.9, + "learning_rate": 5.016715529417671e-07, + "loss": 0.8013, + "step": 28198 + }, + { + "epoch": 0.9, + "learning_rate": 5.01347647997853e-07, + "loss": 0.7407, + "step": 28199 + }, + { + "epoch": 0.9, + "learning_rate": 5.01023844963966e-07, + "loss": 0.8843, + "step": 28200 + }, + { + "epoch": 0.9, + "learning_rate": 5.007001438435822e-07, + "loss": 0.4561, + "step": 28201 + }, + { + "epoch": 0.9, + "learning_rate": 5.003765446401699e-07, + "loss": 1.0229, + "step": 28202 + }, + { + "epoch": 0.9, + "learning_rate": 5.000530473572041e-07, + "loss": 0.8794, + "step": 28203 + }, + { + "epoch": 0.9, + "learning_rate": 4.997296519981554e-07, + "loss": 0.9575, + "step": 28204 + }, + { + "epoch": 0.9, + "learning_rate": 4.994063585664932e-07, + "loss": 0.9448, + "step": 28205 + }, + { + "epoch": 0.9, + "learning_rate": 4.990831670656848e-07, + "loss": 0.9043, + "step": 28206 + }, + { + "epoch": 0.9, + "learning_rate": 4.987600774992008e-07, + "loss": 0.957, + "step": 28207 + }, + { + "epoch": 0.9, + "learning_rate": 4.98437089870506e-07, + "loss": 0.9258, + "step": 28208 + }, + { + "epoch": 0.9, + "learning_rate": 4.981142041830645e-07, + "loss": 0.8599, + "step": 28209 + }, + { + "epoch": 0.9, + "learning_rate": 4.977914204403411e-07, + "loss": 0.9854, + "step": 28210 + }, + { + "epoch": 0.9, + "learning_rate": 4.974687386457999e-07, + "loss": 0.9756, + "step": 28211 + }, + { + "epoch": 0.9, + "learning_rate": 4.971461588029025e-07, + "loss": 0.937, + "step": 28212 + }, + { + "epoch": 0.9, + "learning_rate": 4.968236809151083e-07, + "loss": 0.9614, + "step": 28213 + }, + { + "epoch": 0.9, + "learning_rate": 4.965013049858791e-07, + "loss": 0.9102, + "step": 28214 + }, + { + "epoch": 0.9, + "learning_rate": 4.961790310186721e-07, + "loss": 0.7861, + "step": 28215 + }, + { + "epoch": 0.9, + "learning_rate": 4.958568590169477e-07, + "loss": 0.9023, + "step": 28216 + }, + { + "epoch": 0.9, + "learning_rate": 4.9553478898416e-07, + "loss": 0.9717, + "step": 28217 + }, + { + "epoch": 0.9, + "learning_rate": 4.95212820923765e-07, + "loss": 0.897, + "step": 28218 + }, + { + "epoch": 0.9, + "learning_rate": 4.948909548392156e-07, + "loss": 0.9233, + "step": 28219 + }, + { + "epoch": 0.9, + "learning_rate": 4.945691907339679e-07, + "loss": 1.0054, + "step": 28220 + }, + { + "epoch": 0.9, + "learning_rate": 4.942475286114712e-07, + "loss": 0.8203, + "step": 28221 + }, + { + "epoch": 0.9, + "learning_rate": 4.939259684751796e-07, + "loss": 0.9033, + "step": 28222 + }, + { + "epoch": 0.9, + "learning_rate": 4.936045103285403e-07, + "loss": 0.9395, + "step": 28223 + }, + { + "epoch": 0.9, + "learning_rate": 4.932831541750061e-07, + "loss": 0.874, + "step": 28224 + }, + { + "epoch": 0.9, + "learning_rate": 4.929619000180186e-07, + "loss": 0.7588, + "step": 28225 + }, + { + "epoch": 0.9, + "learning_rate": 4.926407478610306e-07, + "loss": 0.9121, + "step": 28226 + }, + { + "epoch": 0.9, + "learning_rate": 4.923196977074829e-07, + "loss": 0.981, + "step": 28227 + }, + { + "epoch": 0.9, + "learning_rate": 4.919987495608247e-07, + "loss": 0.9492, + "step": 28228 + }, + { + "epoch": 0.9, + "learning_rate": 4.916779034244945e-07, + "loss": 0.9331, + "step": 28229 + }, + { + "epoch": 0.9, + "learning_rate": 4.913571593019395e-07, + "loss": 1.0298, + "step": 28230 + }, + { + "epoch": 0.9, + "learning_rate": 4.910365171965981e-07, + "loss": 0.9888, + "step": 28231 + }, + { + "epoch": 0.9, + "learning_rate": 4.907159771119108e-07, + "loss": 0.9263, + "step": 28232 + }, + { + "epoch": 0.9, + "learning_rate": 4.903955390513149e-07, + "loss": 0.9785, + "step": 28233 + }, + { + "epoch": 0.9, + "learning_rate": 4.900752030182521e-07, + "loss": 0.9209, + "step": 28234 + }, + { + "epoch": 0.9, + "learning_rate": 4.897549690161563e-07, + "loss": 1.0439, + "step": 28235 + }, + { + "epoch": 0.9, + "learning_rate": 4.894348370484648e-07, + "loss": 0.8818, + "step": 28236 + }, + { + "epoch": 0.9, + "learning_rate": 4.891148071186103e-07, + "loss": 0.4229, + "step": 28237 + }, + { + "epoch": 0.9, + "learning_rate": 4.887948792300302e-07, + "loss": 0.9092, + "step": 28238 + }, + { + "epoch": 0.9, + "learning_rate": 4.884750533861537e-07, + "loss": 0.9141, + "step": 28239 + }, + { + "epoch": 0.9, + "learning_rate": 4.881553295904129e-07, + "loss": 0.8667, + "step": 28240 + }, + { + "epoch": 0.9, + "learning_rate": 4.87835707846237e-07, + "loss": 0.8569, + "step": 28241 + }, + { + "epoch": 0.9, + "learning_rate": 4.875161881570589e-07, + "loss": 0.9204, + "step": 28242 + }, + { + "epoch": 0.9, + "learning_rate": 4.871967705263014e-07, + "loss": 0.8428, + "step": 28243 + }, + { + "epoch": 0.9, + "learning_rate": 4.868774549573963e-07, + "loss": 0.9619, + "step": 28244 + }, + { + "epoch": 0.9, + "learning_rate": 4.865582414537662e-07, + "loss": 1.0122, + "step": 28245 + }, + { + "epoch": 0.9, + "learning_rate": 4.862391300188396e-07, + "loss": 0.9585, + "step": 28246 + }, + { + "epoch": 0.9, + "learning_rate": 4.859201206560349e-07, + "loss": 0.8965, + "step": 28247 + }, + { + "epoch": 0.9, + "learning_rate": 4.856012133687804e-07, + "loss": 0.8892, + "step": 28248 + }, + { + "epoch": 0.9, + "learning_rate": 4.852824081604924e-07, + "loss": 0.8921, + "step": 28249 + }, + { + "epoch": 0.9, + "learning_rate": 4.849637050345946e-07, + "loss": 1.0, + "step": 28250 + }, + { + "epoch": 0.9, + "learning_rate": 4.846451039945066e-07, + "loss": 0.9092, + "step": 28251 + }, + { + "epoch": 0.9, + "learning_rate": 4.843266050436435e-07, + "loss": 0.8574, + "step": 28252 + }, + { + "epoch": 0.9, + "learning_rate": 4.840082081854259e-07, + "loss": 0.96, + "step": 28253 + }, + { + "epoch": 0.9, + "learning_rate": 4.836899134232687e-07, + "loss": 0.8735, + "step": 28254 + }, + { + "epoch": 0.9, + "learning_rate": 4.833717207605859e-07, + "loss": 0.9302, + "step": 28255 + }, + { + "epoch": 0.9, + "learning_rate": 4.830536302007927e-07, + "loss": 0.856, + "step": 28256 + }, + { + "epoch": 0.9, + "learning_rate": 4.827356417473006e-07, + "loss": 0.8967, + "step": 28257 + }, + { + "epoch": 0.9, + "learning_rate": 4.824177554035215e-07, + "loss": 0.9556, + "step": 28258 + }, + { + "epoch": 0.9, + "learning_rate": 4.82099971172868e-07, + "loss": 0.447, + "step": 28259 + }, + { + "epoch": 0.9, + "learning_rate": 4.817822890587453e-07, + "loss": 0.8555, + "step": 28260 + }, + { + "epoch": 0.9, + "learning_rate": 4.814647090645663e-07, + "loss": 0.8599, + "step": 28261 + }, + { + "epoch": 0.9, + "learning_rate": 4.81147231193737e-07, + "loss": 0.9053, + "step": 28262 + }, + { + "epoch": 0.9, + "learning_rate": 4.808298554496627e-07, + "loss": 0.9072, + "step": 28263 + }, + { + "epoch": 0.9, + "learning_rate": 4.80512581835747e-07, + "loss": 0.8687, + "step": 28264 + }, + { + "epoch": 0.9, + "learning_rate": 4.801954103553963e-07, + "loss": 0.9004, + "step": 28265 + }, + { + "epoch": 0.9, + "learning_rate": 4.798783410120122e-07, + "loss": 0.8877, + "step": 28266 + }, + { + "epoch": 0.9, + "learning_rate": 4.795613738089988e-07, + "loss": 0.9355, + "step": 28267 + }, + { + "epoch": 0.9, + "learning_rate": 4.792445087497544e-07, + "loss": 0.9023, + "step": 28268 + }, + { + "epoch": 0.9, + "learning_rate": 4.789277458376807e-07, + "loss": 0.9243, + "step": 28269 + }, + { + "epoch": 0.9, + "learning_rate": 4.786110850761728e-07, + "loss": 0.8765, + "step": 28270 + }, + { + "epoch": 0.9, + "learning_rate": 4.782945264686323e-07, + "loss": 0.8203, + "step": 28271 + }, + { + "epoch": 0.9, + "learning_rate": 4.779780700184511e-07, + "loss": 0.4844, + "step": 28272 + }, + { + "epoch": 0.9, + "learning_rate": 4.776617157290287e-07, + "loss": 0.895, + "step": 28273 + }, + { + "epoch": 0.9, + "learning_rate": 4.773454636037556e-07, + "loss": 0.7988, + "step": 28274 + }, + { + "epoch": 0.9, + "learning_rate": 4.770293136460269e-07, + "loss": 1.0474, + "step": 28275 + }, + { + "epoch": 0.9, + "learning_rate": 4.767132658592344e-07, + "loss": 0.9722, + "step": 28276 + }, + { + "epoch": 0.9, + "learning_rate": 4.763973202467709e-07, + "loss": 0.9458, + "step": 28277 + }, + { + "epoch": 0.9, + "learning_rate": 4.7608147681202033e-07, + "loss": 0.9028, + "step": 28278 + }, + { + "epoch": 0.9, + "learning_rate": 4.757657355583778e-07, + "loss": 0.9072, + "step": 28279 + }, + { + "epoch": 0.9, + "learning_rate": 4.7545009648922613e-07, + "loss": 0.9443, + "step": 28280 + }, + { + "epoch": 0.9, + "learning_rate": 4.7513455960795484e-07, + "loss": 1.0151, + "step": 28281 + }, + { + "epoch": 0.9, + "learning_rate": 4.7481912491794677e-07, + "loss": 0.9517, + "step": 28282 + }, + { + "epoch": 0.9, + "learning_rate": 4.7450379242258925e-07, + "loss": 0.792, + "step": 28283 + }, + { + "epoch": 0.9, + "learning_rate": 4.741885621252629e-07, + "loss": 0.9302, + "step": 28284 + }, + { + "epoch": 0.9, + "learning_rate": 4.738734340293516e-07, + "loss": 0.9478, + "step": 28285 + }, + { + "epoch": 0.9, + "learning_rate": 4.735584081382327e-07, + "loss": 0.9697, + "step": 28286 + }, + { + "epoch": 0.9, + "learning_rate": 4.7324348445529133e-07, + "loss": 0.998, + "step": 28287 + }, + { + "epoch": 0.9, + "learning_rate": 4.729286629839025e-07, + "loss": 0.9536, + "step": 28288 + }, + { + "epoch": 0.9, + "learning_rate": 4.726139437274457e-07, + "loss": 0.9946, + "step": 28289 + }, + { + "epoch": 0.9, + "learning_rate": 4.7229932668929726e-07, + "loss": 0.9111, + "step": 28290 + }, + { + "epoch": 0.9, + "learning_rate": 4.71984811872831e-07, + "loss": 0.9575, + "step": 28291 + }, + { + "epoch": 0.9, + "learning_rate": 4.716703992814231e-07, + "loss": 0.8999, + "step": 28292 + }, + { + "epoch": 0.9, + "learning_rate": 4.713560889184476e-07, + "loss": 0.8892, + "step": 28293 + }, + { + "epoch": 0.9, + "learning_rate": 4.71041880787273e-07, + "loss": 0.9546, + "step": 28294 + }, + { + "epoch": 0.9, + "learning_rate": 4.707277748912753e-07, + "loss": 0.856, + "step": 28295 + }, + { + "epoch": 0.9, + "learning_rate": 4.704137712338208e-07, + "loss": 0.9746, + "step": 28296 + }, + { + "epoch": 0.9, + "learning_rate": 4.7009986981828016e-07, + "loss": 0.9614, + "step": 28297 + }, + { + "epoch": 0.9, + "learning_rate": 4.697860706480206e-07, + "loss": 0.9341, + "step": 28298 + }, + { + "epoch": 0.91, + "learning_rate": 4.6947237372640954e-07, + "loss": 1.0122, + "step": 28299 + }, + { + "epoch": 0.91, + "learning_rate": 4.6915877905681084e-07, + "loss": 0.957, + "step": 28300 + }, + { + "epoch": 0.91, + "learning_rate": 4.6884528664259076e-07, + "loss": 0.9893, + "step": 28301 + }, + { + "epoch": 0.91, + "learning_rate": 4.6853189648711327e-07, + "loss": 0.9668, + "step": 28302 + }, + { + "epoch": 0.91, + "learning_rate": 4.6821860859373793e-07, + "loss": 0.8086, + "step": 28303 + }, + { + "epoch": 0.91, + "learning_rate": 4.6790542296582866e-07, + "loss": 0.9189, + "step": 28304 + }, + { + "epoch": 0.91, + "learning_rate": 4.6759233960674387e-07, + "loss": 0.8049, + "step": 28305 + }, + { + "epoch": 0.91, + "learning_rate": 4.672793585198443e-07, + "loss": 0.8564, + "step": 28306 + }, + { + "epoch": 0.91, + "learning_rate": 4.669664797084872e-07, + "loss": 0.9775, + "step": 28307 + }, + { + "epoch": 0.91, + "learning_rate": 4.6665370317602874e-07, + "loss": 0.8755, + "step": 28308 + }, + { + "epoch": 0.91, + "learning_rate": 4.663410289258241e-07, + "loss": 0.8477, + "step": 28309 + }, + { + "epoch": 0.91, + "learning_rate": 4.6602845696122946e-07, + "loss": 0.897, + "step": 28310 + }, + { + "epoch": 0.91, + "learning_rate": 4.6571598728559655e-07, + "loss": 0.9155, + "step": 28311 + }, + { + "epoch": 0.91, + "learning_rate": 4.6540361990228154e-07, + "loss": 0.9883, + "step": 28312 + }, + { + "epoch": 0.91, + "learning_rate": 4.6509135481463074e-07, + "loss": 0.8594, + "step": 28313 + }, + { + "epoch": 0.91, + "learning_rate": 4.647791920260003e-07, + "loss": 0.8801, + "step": 28314 + }, + { + "epoch": 0.91, + "learning_rate": 4.644671315397331e-07, + "loss": 0.9946, + "step": 28315 + }, + { + "epoch": 0.91, + "learning_rate": 4.6415517335918205e-07, + "loss": 0.9863, + "step": 28316 + }, + { + "epoch": 0.91, + "learning_rate": 4.6384331748769104e-07, + "loss": 0.9341, + "step": 28317 + }, + { + "epoch": 0.91, + "learning_rate": 4.6353156392860754e-07, + "loss": 1.0649, + "step": 28318 + }, + { + "epoch": 0.91, + "learning_rate": 4.632199126852754e-07, + "loss": 0.8745, + "step": 28319 + }, + { + "epoch": 0.91, + "learning_rate": 4.629083637610399e-07, + "loss": 0.9209, + "step": 28320 + }, + { + "epoch": 0.91, + "learning_rate": 4.6259691715924035e-07, + "loss": 0.5029, + "step": 28321 + }, + { + "epoch": 0.91, + "learning_rate": 4.622855728832243e-07, + "loss": 0.7939, + "step": 28322 + }, + { + "epoch": 0.91, + "learning_rate": 4.619743309363245e-07, + "loss": 1.0283, + "step": 28323 + }, + { + "epoch": 0.91, + "learning_rate": 4.6166319132188607e-07, + "loss": 0.8335, + "step": 28324 + }, + { + "epoch": 0.91, + "learning_rate": 4.6135215404324417e-07, + "loss": 0.9492, + "step": 28325 + }, + { + "epoch": 0.91, + "learning_rate": 4.610412191037372e-07, + "loss": 0.9771, + "step": 28326 + }, + { + "epoch": 0.91, + "learning_rate": 4.6073038650670034e-07, + "loss": 0.7529, + "step": 28327 + }, + { + "epoch": 0.91, + "learning_rate": 4.604196562554686e-07, + "loss": 0.8887, + "step": 28328 + }, + { + "epoch": 0.91, + "learning_rate": 4.601090283533771e-07, + "loss": 0.7856, + "step": 28329 + }, + { + "epoch": 0.91, + "learning_rate": 4.597985028037577e-07, + "loss": 0.7373, + "step": 28330 + }, + { + "epoch": 0.91, + "learning_rate": 4.5948807960994103e-07, + "loss": 0.8726, + "step": 28331 + }, + { + "epoch": 0.91, + "learning_rate": 4.5917775877525885e-07, + "loss": 0.8325, + "step": 28332 + }, + { + "epoch": 0.91, + "learning_rate": 4.588675403030385e-07, + "loss": 0.9014, + "step": 28333 + }, + { + "epoch": 0.91, + "learning_rate": 4.585574241966118e-07, + "loss": 1.0259, + "step": 28334 + }, + { + "epoch": 0.91, + "learning_rate": 4.5824741045930265e-07, + "loss": 0.8916, + "step": 28335 + }, + { + "epoch": 0.91, + "learning_rate": 4.579374990944385e-07, + "loss": 0.8237, + "step": 28336 + }, + { + "epoch": 0.91, + "learning_rate": 4.576276901053456e-07, + "loss": 0.9126, + "step": 28337 + }, + { + "epoch": 0.91, + "learning_rate": 4.5731798349534673e-07, + "loss": 0.9048, + "step": 28338 + }, + { + "epoch": 0.91, + "learning_rate": 4.5700837926776266e-07, + "loss": 1.0137, + "step": 28339 + }, + { + "epoch": 0.91, + "learning_rate": 4.5669887742591846e-07, + "loss": 0.9443, + "step": 28340 + }, + { + "epoch": 0.91, + "learning_rate": 4.563894779731337e-07, + "loss": 0.9062, + "step": 28341 + }, + { + "epoch": 0.91, + "learning_rate": 4.5608018091272576e-07, + "loss": 0.957, + "step": 28342 + }, + { + "epoch": 0.91, + "learning_rate": 4.557709862480164e-07, + "loss": 0.9521, + "step": 28343 + }, + { + "epoch": 0.91, + "learning_rate": 4.5546189398232075e-07, + "loss": 0.9829, + "step": 28344 + }, + { + "epoch": 0.91, + "learning_rate": 4.551529041189562e-07, + "loss": 0.855, + "step": 28345 + }, + { + "epoch": 0.91, + "learning_rate": 4.5484401666123555e-07, + "loss": 0.9478, + "step": 28346 + }, + { + "epoch": 0.91, + "learning_rate": 4.5453523161247624e-07, + "loss": 0.8867, + "step": 28347 + }, + { + "epoch": 0.91, + "learning_rate": 4.542265489759878e-07, + "loss": 1.0229, + "step": 28348 + }, + { + "epoch": 0.91, + "learning_rate": 4.539179687550854e-07, + "loss": 0.8833, + "step": 28349 + }, + { + "epoch": 0.91, + "learning_rate": 4.536094909530764e-07, + "loss": 0.9917, + "step": 28350 + }, + { + "epoch": 0.91, + "learning_rate": 4.5330111557327363e-07, + "loss": 0.8467, + "step": 28351 + }, + { + "epoch": 0.91, + "learning_rate": 4.529928426189845e-07, + "loss": 0.8906, + "step": 28352 + }, + { + "epoch": 0.91, + "learning_rate": 4.526846720935163e-07, + "loss": 0.9399, + "step": 28353 + }, + { + "epoch": 0.91, + "learning_rate": 4.5237660400017316e-07, + "loss": 0.9951, + "step": 28354 + }, + { + "epoch": 0.91, + "learning_rate": 4.5206863834226344e-07, + "loss": 0.8208, + "step": 28355 + }, + { + "epoch": 0.91, + "learning_rate": 4.5176077512308905e-07, + "loss": 1.0552, + "step": 28356 + }, + { + "epoch": 0.91, + "learning_rate": 4.514530143459561e-07, + "loss": 0.957, + "step": 28357 + }, + { + "epoch": 0.91, + "learning_rate": 4.511453560141621e-07, + "loss": 0.9673, + "step": 28358 + }, + { + "epoch": 0.91, + "learning_rate": 4.5083780013101206e-07, + "loss": 0.8901, + "step": 28359 + }, + { + "epoch": 0.91, + "learning_rate": 4.5053034669980346e-07, + "loss": 0.8701, + "step": 28360 + }, + { + "epoch": 0.91, + "learning_rate": 4.5022299572383574e-07, + "loss": 0.9336, + "step": 28361 + }, + { + "epoch": 0.91, + "learning_rate": 4.499157472064053e-07, + "loss": 0.8623, + "step": 28362 + }, + { + "epoch": 0.91, + "learning_rate": 4.4960860115081053e-07, + "loss": 0.939, + "step": 28363 + }, + { + "epoch": 0.91, + "learning_rate": 4.4930155756034434e-07, + "loss": 0.9531, + "step": 28364 + }, + { + "epoch": 0.91, + "learning_rate": 4.48994616438303e-07, + "loss": 0.9009, + "step": 28365 + }, + { + "epoch": 0.91, + "learning_rate": 4.4868777778797724e-07, + "loss": 0.814, + "step": 28366 + }, + { + "epoch": 0.91, + "learning_rate": 4.483810416126644e-07, + "loss": 1.0273, + "step": 28367 + }, + { + "epoch": 0.91, + "learning_rate": 4.480744079156485e-07, + "loss": 0.9604, + "step": 28368 + }, + { + "epoch": 0.91, + "learning_rate": 4.477678767002247e-07, + "loss": 0.8301, + "step": 28369 + }, + { + "epoch": 0.91, + "learning_rate": 4.4746144796967815e-07, + "loss": 0.917, + "step": 28370 + }, + { + "epoch": 0.91, + "learning_rate": 4.4715512172729846e-07, + "loss": 0.8198, + "step": 28371 + }, + { + "epoch": 0.91, + "learning_rate": 4.468488979763719e-07, + "loss": 0.9092, + "step": 28372 + }, + { + "epoch": 0.91, + "learning_rate": 4.4654277672018355e-07, + "loss": 0.9468, + "step": 28373 + }, + { + "epoch": 0.91, + "learning_rate": 4.462367579620186e-07, + "loss": 0.4937, + "step": 28374 + }, + { + "epoch": 0.91, + "learning_rate": 4.4593084170516e-07, + "loss": 0.9673, + "step": 28375 + }, + { + "epoch": 0.91, + "learning_rate": 4.456250279528884e-07, + "loss": 0.8931, + "step": 28376 + }, + { + "epoch": 0.91, + "learning_rate": 4.453193167084868e-07, + "loss": 0.9146, + "step": 28377 + }, + { + "epoch": 0.91, + "learning_rate": 4.450137079752326e-07, + "loss": 0.8853, + "step": 28378 + }, + { + "epoch": 0.91, + "learning_rate": 4.4470820175640863e-07, + "loss": 0.9023, + "step": 28379 + }, + { + "epoch": 0.91, + "learning_rate": 4.444027980552901e-07, + "loss": 0.9229, + "step": 28380 + }, + { + "epoch": 0.91, + "learning_rate": 4.4409749687515215e-07, + "loss": 0.896, + "step": 28381 + }, + { + "epoch": 0.91, + "learning_rate": 4.437922982192733e-07, + "loss": 0.8726, + "step": 28382 + }, + { + "epoch": 0.91, + "learning_rate": 4.4348720209092753e-07, + "loss": 0.9873, + "step": 28383 + }, + { + "epoch": 0.91, + "learning_rate": 4.431822084933868e-07, + "loss": 0.9072, + "step": 28384 + }, + { + "epoch": 0.91, + "learning_rate": 4.428773174299239e-07, + "loss": 0.8452, + "step": 28385 + }, + { + "epoch": 0.91, + "learning_rate": 4.4257252890381077e-07, + "loss": 0.9194, + "step": 28386 + }, + { + "epoch": 0.91, + "learning_rate": 4.422678429183147e-07, + "loss": 0.8545, + "step": 28387 + }, + { + "epoch": 0.91, + "learning_rate": 4.419632594767087e-07, + "loss": 0.7976, + "step": 28388 + }, + { + "epoch": 0.91, + "learning_rate": 4.416587785822568e-07, + "loss": 0.9541, + "step": 28389 + }, + { + "epoch": 0.91, + "learning_rate": 4.4135440023823087e-07, + "loss": 0.978, + "step": 28390 + }, + { + "epoch": 0.91, + "learning_rate": 4.410501244478904e-07, + "loss": 0.8462, + "step": 28391 + }, + { + "epoch": 0.91, + "learning_rate": 4.4074595121450406e-07, + "loss": 0.9507, + "step": 28392 + }, + { + "epoch": 0.91, + "learning_rate": 4.404418805413324e-07, + "loss": 0.9932, + "step": 28393 + }, + { + "epoch": 0.91, + "learning_rate": 4.401379124316407e-07, + "loss": 0.8413, + "step": 28394 + }, + { + "epoch": 0.91, + "learning_rate": 4.398340468886875e-07, + "loss": 0.9644, + "step": 28395 + }, + { + "epoch": 0.91, + "learning_rate": 4.3953028391573673e-07, + "loss": 0.4482, + "step": 28396 + }, + { + "epoch": 0.91, + "learning_rate": 4.3922662351604363e-07, + "loss": 0.9688, + "step": 28397 + }, + { + "epoch": 0.91, + "learning_rate": 4.389230656928678e-07, + "loss": 0.8945, + "step": 28398 + }, + { + "epoch": 0.91, + "learning_rate": 4.3861961044946555e-07, + "loss": 0.9158, + "step": 28399 + }, + { + "epoch": 0.91, + "learning_rate": 4.3831625778909314e-07, + "loss": 0.8735, + "step": 28400 + }, + { + "epoch": 0.91, + "learning_rate": 4.3801300771500353e-07, + "loss": 0.9854, + "step": 28401 + }, + { + "epoch": 0.91, + "learning_rate": 4.3770986023045304e-07, + "loss": 0.8213, + "step": 28402 + }, + { + "epoch": 0.91, + "learning_rate": 4.3740681533869236e-07, + "loss": 0.7417, + "step": 28403 + }, + { + "epoch": 0.91, + "learning_rate": 4.371038730429733e-07, + "loss": 0.9097, + "step": 28404 + }, + { + "epoch": 0.91, + "learning_rate": 4.368010333465467e-07, + "loss": 0.9409, + "step": 28405 + }, + { + "epoch": 0.91, + "learning_rate": 4.364982962526609e-07, + "loss": 0.9404, + "step": 28406 + }, + { + "epoch": 0.91, + "learning_rate": 4.361956617645624e-07, + "loss": 0.9365, + "step": 28407 + }, + { + "epoch": 0.91, + "learning_rate": 4.358931298855018e-07, + "loss": 0.9487, + "step": 28408 + }, + { + "epoch": 0.91, + "learning_rate": 4.35590700618721e-07, + "loss": 0.9404, + "step": 28409 + }, + { + "epoch": 0.91, + "learning_rate": 4.352883739674674e-07, + "loss": 0.9053, + "step": 28410 + }, + { + "epoch": 0.91, + "learning_rate": 4.3498614993498277e-07, + "loss": 0.9629, + "step": 28411 + }, + { + "epoch": 0.91, + "learning_rate": 4.3468402852451464e-07, + "loss": 0.9248, + "step": 28412 + }, + { + "epoch": 0.91, + "learning_rate": 4.34382009739297e-07, + "loss": 0.9355, + "step": 28413 + }, + { + "epoch": 0.91, + "learning_rate": 4.3408009358257396e-07, + "loss": 1.0371, + "step": 28414 + }, + { + "epoch": 0.91, + "learning_rate": 4.3377828005758513e-07, + "loss": 0.9585, + "step": 28415 + }, + { + "epoch": 0.91, + "learning_rate": 4.33476569167568e-07, + "loss": 1.082, + "step": 28416 + }, + { + "epoch": 0.91, + "learning_rate": 4.331749609157587e-07, + "loss": 0.823, + "step": 28417 + }, + { + "epoch": 0.91, + "learning_rate": 4.3287345530539484e-07, + "loss": 0.9214, + "step": 28418 + }, + { + "epoch": 0.91, + "learning_rate": 4.3257205233971145e-07, + "loss": 0.9194, + "step": 28419 + }, + { + "epoch": 0.91, + "learning_rate": 4.3227075202194045e-07, + "loss": 0.9526, + "step": 28420 + }, + { + "epoch": 0.91, + "learning_rate": 4.3196955435531374e-07, + "loss": 0.9482, + "step": 28421 + }, + { + "epoch": 0.91, + "learning_rate": 4.3166845934306647e-07, + "loss": 0.8711, + "step": 28422 + }, + { + "epoch": 0.91, + "learning_rate": 4.313674669884249e-07, + "loss": 0.9185, + "step": 28423 + }, + { + "epoch": 0.91, + "learning_rate": 4.3106657729462207e-07, + "loss": 0.8955, + "step": 28424 + }, + { + "epoch": 0.91, + "learning_rate": 4.307657902648843e-07, + "loss": 0.5127, + "step": 28425 + }, + { + "epoch": 0.91, + "learning_rate": 4.3046510590243784e-07, + "loss": 0.9131, + "step": 28426 + }, + { + "epoch": 0.91, + "learning_rate": 4.301645242105101e-07, + "loss": 0.8105, + "step": 28427 + }, + { + "epoch": 0.91, + "learning_rate": 4.2986404519232637e-07, + "loss": 0.9395, + "step": 28428 + }, + { + "epoch": 0.91, + "learning_rate": 4.295636688511107e-07, + "loss": 0.9658, + "step": 28429 + }, + { + "epoch": 0.91, + "learning_rate": 4.2926339519008155e-07, + "loss": 0.8203, + "step": 28430 + }, + { + "epoch": 0.91, + "learning_rate": 4.289632242124664e-07, + "loss": 0.4675, + "step": 28431 + }, + { + "epoch": 0.91, + "learning_rate": 4.2866315592148157e-07, + "loss": 0.8965, + "step": 28432 + }, + { + "epoch": 0.91, + "learning_rate": 4.2836319032035e-07, + "loss": 0.8154, + "step": 28433 + }, + { + "epoch": 0.91, + "learning_rate": 4.2806332741228586e-07, + "loss": 0.8247, + "step": 28434 + }, + { + "epoch": 0.91, + "learning_rate": 4.2776356720051095e-07, + "loss": 0.9995, + "step": 28435 + }, + { + "epoch": 0.91, + "learning_rate": 4.274639096882371e-07, + "loss": 0.917, + "step": 28436 + }, + { + "epoch": 0.91, + "learning_rate": 4.271643548786819e-07, + "loss": 0.9927, + "step": 28437 + }, + { + "epoch": 0.91, + "learning_rate": 4.268649027750582e-07, + "loss": 0.8428, + "step": 28438 + }, + { + "epoch": 0.91, + "learning_rate": 4.2656555338057905e-07, + "loss": 0.9707, + "step": 28439 + }, + { + "epoch": 0.91, + "learning_rate": 4.262663066984551e-07, + "loss": 0.8069, + "step": 28440 + }, + { + "epoch": 0.91, + "learning_rate": 4.2596716273189954e-07, + "loss": 0.8711, + "step": 28441 + }, + { + "epoch": 0.91, + "learning_rate": 4.256681214841196e-07, + "loss": 0.9385, + "step": 28442 + }, + { + "epoch": 0.91, + "learning_rate": 4.253691829583262e-07, + "loss": 0.8848, + "step": 28443 + }, + { + "epoch": 0.91, + "learning_rate": 4.250703471577222e-07, + "loss": 0.9277, + "step": 28444 + }, + { + "epoch": 0.91, + "learning_rate": 4.247716140855185e-07, + "loss": 1.0669, + "step": 28445 + }, + { + "epoch": 0.91, + "learning_rate": 4.244729837449157e-07, + "loss": 1.02, + "step": 28446 + }, + { + "epoch": 0.91, + "learning_rate": 4.2417445613912144e-07, + "loss": 0.9897, + "step": 28447 + }, + { + "epoch": 0.91, + "learning_rate": 4.238760312713364e-07, + "loss": 0.9409, + "step": 28448 + }, + { + "epoch": 0.91, + "learning_rate": 4.235777091447646e-07, + "loss": 0.8809, + "step": 28449 + }, + { + "epoch": 0.91, + "learning_rate": 4.232794897626058e-07, + "loss": 0.9585, + "step": 28450 + }, + { + "epoch": 0.91, + "learning_rate": 4.229813731280585e-07, + "loss": 0.824, + "step": 28451 + }, + { + "epoch": 0.91, + "learning_rate": 4.2268335924432023e-07, + "loss": 0.8154, + "step": 28452 + }, + { + "epoch": 0.91, + "learning_rate": 4.223854481145917e-07, + "loss": 0.96, + "step": 28453 + }, + { + "epoch": 0.91, + "learning_rate": 4.220876397420648e-07, + "loss": 0.9272, + "step": 28454 + }, + { + "epoch": 0.91, + "learning_rate": 4.217899341299403e-07, + "loss": 0.8237, + "step": 28455 + }, + { + "epoch": 0.91, + "learning_rate": 4.214923312814068e-07, + "loss": 0.9199, + "step": 28456 + }, + { + "epoch": 0.91, + "learning_rate": 4.211948311996605e-07, + "loss": 0.832, + "step": 28457 + }, + { + "epoch": 0.91, + "learning_rate": 4.208974338878935e-07, + "loss": 0.8188, + "step": 28458 + }, + { + "epoch": 0.91, + "learning_rate": 4.2060013934929423e-07, + "loss": 0.8699, + "step": 28459 + }, + { + "epoch": 0.91, + "learning_rate": 4.2030294758705346e-07, + "loss": 0.9106, + "step": 28460 + }, + { + "epoch": 0.91, + "learning_rate": 4.2000585860435983e-07, + "loss": 0.9609, + "step": 28461 + }, + { + "epoch": 0.91, + "learning_rate": 4.197088724044007e-07, + "loss": 0.897, + "step": 28462 + }, + { + "epoch": 0.91, + "learning_rate": 4.194119889903625e-07, + "loss": 0.9077, + "step": 28463 + }, + { + "epoch": 0.91, + "learning_rate": 4.191152083654315e-07, + "loss": 0.8789, + "step": 28464 + }, + { + "epoch": 0.91, + "learning_rate": 4.1881853053278966e-07, + "loss": 0.8657, + "step": 28465 + }, + { + "epoch": 0.91, + "learning_rate": 4.18521955495621e-07, + "loss": 0.8984, + "step": 28466 + }, + { + "epoch": 0.91, + "learning_rate": 4.1822548325710757e-07, + "loss": 0.9355, + "step": 28467 + }, + { + "epoch": 0.91, + "learning_rate": 4.1792911382043113e-07, + "loss": 0.9495, + "step": 28468 + }, + { + "epoch": 0.91, + "learning_rate": 4.176328471887681e-07, + "loss": 0.9414, + "step": 28469 + }, + { + "epoch": 0.91, + "learning_rate": 4.173366833653014e-07, + "loss": 0.822, + "step": 28470 + }, + { + "epoch": 0.91, + "learning_rate": 4.170406223532053e-07, + "loss": 0.8127, + "step": 28471 + }, + { + "epoch": 0.91, + "learning_rate": 4.167446641556583e-07, + "loss": 0.791, + "step": 28472 + }, + { + "epoch": 0.91, + "learning_rate": 4.164488087758345e-07, + "loss": 0.8945, + "step": 28473 + }, + { + "epoch": 0.91, + "learning_rate": 4.1615305621690915e-07, + "loss": 0.9517, + "step": 28474 + }, + { + "epoch": 0.91, + "learning_rate": 4.158574064820531e-07, + "loss": 0.9185, + "step": 28475 + }, + { + "epoch": 0.91, + "learning_rate": 4.155618595744415e-07, + "loss": 0.958, + "step": 28476 + }, + { + "epoch": 0.91, + "learning_rate": 4.152664154972419e-07, + "loss": 0.9541, + "step": 28477 + }, + { + "epoch": 0.91, + "learning_rate": 4.149710742536273e-07, + "loss": 0.8848, + "step": 28478 + }, + { + "epoch": 0.91, + "learning_rate": 4.1467583584676395e-07, + "loss": 0.9751, + "step": 28479 + }, + { + "epoch": 0.91, + "learning_rate": 4.143807002798217e-07, + "loss": 0.8726, + "step": 28480 + }, + { + "epoch": 0.91, + "learning_rate": 4.1408566755596456e-07, + "loss": 0.8887, + "step": 28481 + }, + { + "epoch": 0.91, + "learning_rate": 4.1379073767835896e-07, + "loss": 0.9409, + "step": 28482 + }, + { + "epoch": 0.91, + "learning_rate": 4.13495910650169e-07, + "loss": 0.8784, + "step": 28483 + }, + { + "epoch": 0.91, + "learning_rate": 4.1320118647455887e-07, + "loss": 0.9263, + "step": 28484 + }, + { + "epoch": 0.91, + "learning_rate": 4.129065651546893e-07, + "loss": 0.9414, + "step": 28485 + }, + { + "epoch": 0.91, + "learning_rate": 4.1261204669372226e-07, + "loss": 0.9307, + "step": 28486 + }, + { + "epoch": 0.91, + "learning_rate": 4.1231763109481624e-07, + "loss": 0.9136, + "step": 28487 + }, + { + "epoch": 0.91, + "learning_rate": 4.120233183611333e-07, + "loss": 0.9165, + "step": 28488 + }, + { + "epoch": 0.91, + "learning_rate": 4.117291084958264e-07, + "loss": 0.4629, + "step": 28489 + }, + { + "epoch": 0.91, + "learning_rate": 4.114350015020552e-07, + "loss": 0.8584, + "step": 28490 + }, + { + "epoch": 0.91, + "learning_rate": 4.1114099738297385e-07, + "loss": 1.0845, + "step": 28491 + }, + { + "epoch": 0.91, + "learning_rate": 4.108470961417377e-07, + "loss": 0.8267, + "step": 28492 + }, + { + "epoch": 0.91, + "learning_rate": 4.1055329778149857e-07, + "loss": 0.9521, + "step": 28493 + }, + { + "epoch": 0.91, + "learning_rate": 4.102596023054106e-07, + "loss": 0.978, + "step": 28494 + }, + { + "epoch": 0.91, + "learning_rate": 4.099660097166236e-07, + "loss": 0.8145, + "step": 28495 + }, + { + "epoch": 0.91, + "learning_rate": 4.0967252001828826e-07, + "loss": 0.9253, + "step": 28496 + }, + { + "epoch": 0.91, + "learning_rate": 4.0937913321355104e-07, + "loss": 0.7349, + "step": 28497 + }, + { + "epoch": 0.91, + "learning_rate": 4.0908584930556274e-07, + "loss": 0.4707, + "step": 28498 + }, + { + "epoch": 0.91, + "learning_rate": 4.087926682974674e-07, + "loss": 0.938, + "step": 28499 + }, + { + "epoch": 0.91, + "learning_rate": 4.0849959019241267e-07, + "loss": 1.0146, + "step": 28500 + }, + { + "epoch": 0.91, + "learning_rate": 4.082066149935404e-07, + "loss": 0.8857, + "step": 28501 + }, + { + "epoch": 0.91, + "learning_rate": 4.0791374270399695e-07, + "loss": 0.9102, + "step": 28502 + }, + { + "epoch": 0.91, + "learning_rate": 4.076209733269232e-07, + "loss": 0.9873, + "step": 28503 + }, + { + "epoch": 0.91, + "learning_rate": 4.0732830686545986e-07, + "loss": 0.9014, + "step": 28504 + }, + { + "epoch": 0.91, + "learning_rate": 4.0703574332274675e-07, + "loss": 0.8799, + "step": 28505 + }, + { + "epoch": 0.91, + "learning_rate": 4.0674328270192355e-07, + "loss": 0.9282, + "step": 28506 + }, + { + "epoch": 0.91, + "learning_rate": 4.064509250061277e-07, + "loss": 0.9771, + "step": 28507 + }, + { + "epoch": 0.91, + "learning_rate": 4.0615867023849454e-07, + "loss": 0.8799, + "step": 28508 + }, + { + "epoch": 0.91, + "learning_rate": 4.058665184021626e-07, + "loss": 0.9277, + "step": 28509 + }, + { + "epoch": 0.91, + "learning_rate": 4.055744695002628e-07, + "loss": 0.8672, + "step": 28510 + }, + { + "epoch": 0.91, + "learning_rate": 4.0528252353593255e-07, + "loss": 0.9175, + "step": 28511 + }, + { + "epoch": 0.91, + "learning_rate": 4.049906805123005e-07, + "loss": 0.8623, + "step": 28512 + }, + { + "epoch": 0.91, + "learning_rate": 4.046989404325008e-07, + "loss": 0.9482, + "step": 28513 + }, + { + "epoch": 0.91, + "learning_rate": 4.0440730329965984e-07, + "loss": 1.0269, + "step": 28514 + }, + { + "epoch": 0.91, + "learning_rate": 4.0411576911690955e-07, + "loss": 0.9731, + "step": 28515 + }, + { + "epoch": 0.91, + "learning_rate": 4.038243378873763e-07, + "loss": 0.8596, + "step": 28516 + }, + { + "epoch": 0.91, + "learning_rate": 4.0353300961418764e-07, + "loss": 0.9331, + "step": 28517 + }, + { + "epoch": 0.91, + "learning_rate": 4.032417843004699e-07, + "loss": 0.4897, + "step": 28518 + }, + { + "epoch": 0.91, + "learning_rate": 4.0295066194934617e-07, + "loss": 0.856, + "step": 28519 + }, + { + "epoch": 0.91, + "learning_rate": 4.026596425639395e-07, + "loss": 0.8198, + "step": 28520 + }, + { + "epoch": 0.91, + "learning_rate": 4.023687261473741e-07, + "loss": 0.8555, + "step": 28521 + }, + { + "epoch": 0.91, + "learning_rate": 4.0207791270276964e-07, + "loss": 0.9121, + "step": 28522 + }, + { + "epoch": 0.91, + "learning_rate": 4.0178720223324695e-07, + "loss": 0.8789, + "step": 28523 + }, + { + "epoch": 0.91, + "learning_rate": 4.014965947419236e-07, + "loss": 0.897, + "step": 28524 + }, + { + "epoch": 0.91, + "learning_rate": 4.0120609023192036e-07, + "loss": 0.9209, + "step": 28525 + }, + { + "epoch": 0.91, + "learning_rate": 4.009156887063526e-07, + "loss": 0.7959, + "step": 28526 + }, + { + "epoch": 0.91, + "learning_rate": 4.0062539016833437e-07, + "loss": 0.792, + "step": 28527 + }, + { + "epoch": 0.91, + "learning_rate": 4.003351946209821e-07, + "loss": 0.8018, + "step": 28528 + }, + { + "epoch": 0.91, + "learning_rate": 4.00045102067409e-07, + "loss": 0.9844, + "step": 28529 + }, + { + "epoch": 0.91, + "learning_rate": 3.9975511251072686e-07, + "loss": 0.9302, + "step": 28530 + }, + { + "epoch": 0.91, + "learning_rate": 3.994652259540477e-07, + "loss": 0.8975, + "step": 28531 + }, + { + "epoch": 0.91, + "learning_rate": 3.9917544240048013e-07, + "loss": 0.9399, + "step": 28532 + }, + { + "epoch": 0.91, + "learning_rate": 3.988857618531372e-07, + "loss": 0.9951, + "step": 28533 + }, + { + "epoch": 0.91, + "learning_rate": 3.9859618431512205e-07, + "loss": 0.9102, + "step": 28534 + }, + { + "epoch": 0.91, + "learning_rate": 3.9830670978954434e-07, + "loss": 0.9863, + "step": 28535 + }, + { + "epoch": 0.91, + "learning_rate": 3.9801733827950715e-07, + "loss": 0.9468, + "step": 28536 + }, + { + "epoch": 0.91, + "learning_rate": 3.977280697881192e-07, + "loss": 1.0376, + "step": 28537 + }, + { + "epoch": 0.91, + "learning_rate": 3.974389043184801e-07, + "loss": 0.874, + "step": 28538 + }, + { + "epoch": 0.91, + "learning_rate": 3.971498418736952e-07, + "loss": 1.0117, + "step": 28539 + }, + { + "epoch": 0.91, + "learning_rate": 3.9686088245686315e-07, + "loss": 0.9985, + "step": 28540 + }, + { + "epoch": 0.91, + "learning_rate": 3.9657202607108814e-07, + "loss": 0.8765, + "step": 28541 + }, + { + "epoch": 0.91, + "learning_rate": 3.9628327271946434e-07, + "loss": 1.0625, + "step": 28542 + }, + { + "epoch": 0.91, + "learning_rate": 3.959946224050937e-07, + "loss": 1.0215, + "step": 28543 + }, + { + "epoch": 0.91, + "learning_rate": 3.957060751310704e-07, + "loss": 0.957, + "step": 28544 + }, + { + "epoch": 0.91, + "learning_rate": 3.95417630900492e-07, + "loss": 0.9673, + "step": 28545 + }, + { + "epoch": 0.91, + "learning_rate": 3.951292897164527e-07, + "loss": 0.9214, + "step": 28546 + }, + { + "epoch": 0.91, + "learning_rate": 3.948410515820456e-07, + "loss": 0.9507, + "step": 28547 + }, + { + "epoch": 0.91, + "learning_rate": 3.9455291650036363e-07, + "loss": 0.8477, + "step": 28548 + }, + { + "epoch": 0.91, + "learning_rate": 3.9426488447449893e-07, + "loss": 0.7856, + "step": 28549 + }, + { + "epoch": 0.91, + "learning_rate": 3.93976955507539e-07, + "loss": 0.8496, + "step": 28550 + }, + { + "epoch": 0.91, + "learning_rate": 3.936891296025769e-07, + "loss": 0.9585, + "step": 28551 + }, + { + "epoch": 0.91, + "learning_rate": 3.93401406762699e-07, + "loss": 0.4587, + "step": 28552 + }, + { + "epoch": 0.91, + "learning_rate": 3.9311378699098956e-07, + "loss": 0.8921, + "step": 28553 + }, + { + "epoch": 0.91, + "learning_rate": 3.928262702905383e-07, + "loss": 0.9316, + "step": 28554 + }, + { + "epoch": 0.91, + "learning_rate": 3.925388566644273e-07, + "loss": 0.9341, + "step": 28555 + }, + { + "epoch": 0.91, + "learning_rate": 3.922515461157428e-07, + "loss": 0.9067, + "step": 28556 + }, + { + "epoch": 0.91, + "learning_rate": 3.919643386475658e-07, + "loss": 0.8784, + "step": 28557 + }, + { + "epoch": 0.91, + "learning_rate": 3.916772342629782e-07, + "loss": 0.9199, + "step": 28558 + }, + { + "epoch": 0.91, + "learning_rate": 3.9139023296505874e-07, + "loss": 0.9463, + "step": 28559 + }, + { + "epoch": 0.91, + "learning_rate": 3.9110333475688823e-07, + "loss": 0.9111, + "step": 28560 + }, + { + "epoch": 0.91, + "learning_rate": 3.9081653964154307e-07, + "loss": 0.8135, + "step": 28561 + }, + { + "epoch": 0.91, + "learning_rate": 3.905298476221042e-07, + "loss": 0.8755, + "step": 28562 + }, + { + "epoch": 0.91, + "learning_rate": 3.902432587016436e-07, + "loss": 0.9016, + "step": 28563 + }, + { + "epoch": 0.91, + "learning_rate": 3.8995677288323765e-07, + "loss": 0.9077, + "step": 28564 + }, + { + "epoch": 0.91, + "learning_rate": 3.8967039016995833e-07, + "loss": 0.9385, + "step": 28565 + }, + { + "epoch": 0.91, + "learning_rate": 3.893841105648799e-07, + "loss": 0.9053, + "step": 28566 + }, + { + "epoch": 0.91, + "learning_rate": 3.890979340710732e-07, + "loss": 0.8359, + "step": 28567 + }, + { + "epoch": 0.91, + "learning_rate": 3.8881186069161026e-07, + "loss": 0.8813, + "step": 28568 + }, + { + "epoch": 0.91, + "learning_rate": 3.885258904295575e-07, + "loss": 0.8159, + "step": 28569 + }, + { + "epoch": 0.91, + "learning_rate": 3.882400232879857e-07, + "loss": 0.876, + "step": 28570 + }, + { + "epoch": 0.91, + "learning_rate": 3.8795425926996033e-07, + "loss": 0.783, + "step": 28571 + }, + { + "epoch": 0.91, + "learning_rate": 3.8766859837854773e-07, + "loss": 0.8423, + "step": 28572 + }, + { + "epoch": 0.91, + "learning_rate": 3.8738304061681107e-07, + "loss": 0.8809, + "step": 28573 + }, + { + "epoch": 0.91, + "learning_rate": 3.8709758598781786e-07, + "loss": 0.8037, + "step": 28574 + }, + { + "epoch": 0.91, + "learning_rate": 3.868122344946268e-07, + "loss": 0.9473, + "step": 28575 + }, + { + "epoch": 0.91, + "learning_rate": 3.8652698614030204e-07, + "loss": 0.7358, + "step": 28576 + }, + { + "epoch": 0.91, + "learning_rate": 3.862418409279023e-07, + "loss": 0.9595, + "step": 28577 + }, + { + "epoch": 0.91, + "learning_rate": 3.8595679886049063e-07, + "loss": 1.0488, + "step": 28578 + }, + { + "epoch": 0.91, + "learning_rate": 3.856718599411191e-07, + "loss": 1.0073, + "step": 28579 + }, + { + "epoch": 0.91, + "learning_rate": 3.8538702417284855e-07, + "loss": 0.9243, + "step": 28580 + }, + { + "epoch": 0.91, + "learning_rate": 3.8510229155873435e-07, + "loss": 0.8579, + "step": 28581 + }, + { + "epoch": 0.91, + "learning_rate": 3.848176621018318e-07, + "loss": 0.8853, + "step": 28582 + }, + { + "epoch": 0.91, + "learning_rate": 3.845331358051929e-07, + "loss": 0.939, + "step": 28583 + }, + { + "epoch": 0.91, + "learning_rate": 3.8424871267187305e-07, + "loss": 0.9526, + "step": 28584 + }, + { + "epoch": 0.91, + "learning_rate": 3.8396439270492194e-07, + "loss": 0.9492, + "step": 28585 + }, + { + "epoch": 0.91, + "learning_rate": 3.836801759073905e-07, + "loss": 0.8799, + "step": 28586 + }, + { + "epoch": 0.91, + "learning_rate": 3.833960622823263e-07, + "loss": 0.9258, + "step": 28587 + }, + { + "epoch": 0.91, + "learning_rate": 3.831120518327813e-07, + "loss": 0.825, + "step": 28588 + }, + { + "epoch": 0.91, + "learning_rate": 3.828281445617987e-07, + "loss": 0.939, + "step": 28589 + }, + { + "epoch": 0.91, + "learning_rate": 3.825443404724283e-07, + "loss": 0.9121, + "step": 28590 + }, + { + "epoch": 0.91, + "learning_rate": 3.822606395677131e-07, + "loss": 0.9922, + "step": 28591 + }, + { + "epoch": 0.91, + "learning_rate": 3.819770418506952e-07, + "loss": 0.9287, + "step": 28592 + }, + { + "epoch": 0.91, + "learning_rate": 3.816935473244199e-07, + "loss": 0.9126, + "step": 28593 + }, + { + "epoch": 0.91, + "learning_rate": 3.814101559919281e-07, + "loss": 0.9258, + "step": 28594 + }, + { + "epoch": 0.91, + "learning_rate": 3.811268678562585e-07, + "loss": 0.9133, + "step": 28595 + }, + { + "epoch": 0.91, + "learning_rate": 3.8084368292045316e-07, + "loss": 1.0029, + "step": 28596 + }, + { + "epoch": 0.91, + "learning_rate": 3.805606011875496e-07, + "loss": 0.8423, + "step": 28597 + }, + { + "epoch": 0.91, + "learning_rate": 3.802776226605831e-07, + "loss": 0.9438, + "step": 28598 + }, + { + "epoch": 0.91, + "learning_rate": 3.7999474734259245e-07, + "loss": 0.4976, + "step": 28599 + }, + { + "epoch": 0.91, + "learning_rate": 3.7971197523661073e-07, + "loss": 0.7905, + "step": 28600 + }, + { + "epoch": 0.91, + "learning_rate": 3.794293063456733e-07, + "loss": 0.9873, + "step": 28601 + }, + { + "epoch": 0.91, + "learning_rate": 3.7914674067281e-07, + "loss": 0.8604, + "step": 28602 + }, + { + "epoch": 0.91, + "learning_rate": 3.788642782210561e-07, + "loss": 1.0137, + "step": 28603 + }, + { + "epoch": 0.91, + "learning_rate": 3.785819189934381e-07, + "loss": 0.8809, + "step": 28604 + }, + { + "epoch": 0.91, + "learning_rate": 3.7829966299299026e-07, + "loss": 0.9409, + "step": 28605 + }, + { + "epoch": 0.91, + "learning_rate": 3.780175102227357e-07, + "loss": 0.9248, + "step": 28606 + }, + { + "epoch": 0.91, + "learning_rate": 3.7773546068570646e-07, + "loss": 0.9146, + "step": 28607 + }, + { + "epoch": 0.91, + "learning_rate": 3.774535143849245e-07, + "loss": 0.8027, + "step": 28608 + }, + { + "epoch": 0.91, + "learning_rate": 3.771716713234197e-07, + "loss": 0.8623, + "step": 28609 + }, + { + "epoch": 0.91, + "learning_rate": 3.768899315042096e-07, + "loss": 0.9019, + "step": 28610 + }, + { + "epoch": 0.91, + "learning_rate": 3.766082949303218e-07, + "loss": 0.4995, + "step": 28611 + }, + { + "epoch": 0.92, + "learning_rate": 3.76326761604775e-07, + "loss": 0.938, + "step": 28612 + }, + { + "epoch": 0.92, + "learning_rate": 3.7604533153059233e-07, + "loss": 0.8042, + "step": 28613 + }, + { + "epoch": 0.92, + "learning_rate": 3.7576400471079023e-07, + "loss": 0.8384, + "step": 28614 + }, + { + "epoch": 0.92, + "learning_rate": 3.7548278114839075e-07, + "loss": 0.9233, + "step": 28615 + }, + { + "epoch": 0.92, + "learning_rate": 3.752016608464093e-07, + "loss": 0.9971, + "step": 28616 + }, + { + "epoch": 0.92, + "learning_rate": 3.749206438078612e-07, + "loss": 0.9326, + "step": 28617 + }, + { + "epoch": 0.92, + "learning_rate": 3.7463973003576184e-07, + "loss": 0.8284, + "step": 28618 + }, + { + "epoch": 0.92, + "learning_rate": 3.743589195331254e-07, + "loss": 1.1304, + "step": 28619 + }, + { + "epoch": 0.92, + "learning_rate": 3.7407821230296404e-07, + "loss": 0.9575, + "step": 28620 + }, + { + "epoch": 0.92, + "learning_rate": 3.737976083482908e-07, + "loss": 0.9839, + "step": 28621 + }, + { + "epoch": 0.92, + "learning_rate": 3.7351710767211446e-07, + "loss": 0.8745, + "step": 28622 + }, + { + "epoch": 0.92, + "learning_rate": 3.73236710277447e-07, + "loss": 0.9565, + "step": 28623 + }, + { + "epoch": 0.92, + "learning_rate": 3.7295641616729493e-07, + "loss": 0.9199, + "step": 28624 + }, + { + "epoch": 0.92, + "learning_rate": 3.7267622534466585e-07, + "loss": 0.9546, + "step": 28625 + }, + { + "epoch": 0.92, + "learning_rate": 3.723961378125651e-07, + "loss": 1.0889, + "step": 28626 + }, + { + "epoch": 0.92, + "learning_rate": 3.7211615357399923e-07, + "loss": 0.9058, + "step": 28627 + }, + { + "epoch": 0.92, + "learning_rate": 3.7183627263197017e-07, + "loss": 0.9497, + "step": 28628 + }, + { + "epoch": 0.92, + "learning_rate": 3.715564949894834e-07, + "loss": 0.9565, + "step": 28629 + }, + { + "epoch": 0.92, + "learning_rate": 3.7127682064953984e-07, + "loss": 0.9424, + "step": 28630 + }, + { + "epoch": 0.92, + "learning_rate": 3.7099724961513814e-07, + "loss": 0.5127, + "step": 28631 + }, + { + "epoch": 0.92, + "learning_rate": 3.707177818892793e-07, + "loss": 0.9351, + "step": 28632 + }, + { + "epoch": 0.92, + "learning_rate": 3.7043841747496203e-07, + "loss": 0.8838, + "step": 28633 + }, + { + "epoch": 0.92, + "learning_rate": 3.7015915637518163e-07, + "loss": 0.8286, + "step": 28634 + }, + { + "epoch": 0.92, + "learning_rate": 3.698799985929369e-07, + "loss": 1.0151, + "step": 28635 + }, + { + "epoch": 0.92, + "learning_rate": 3.6960094413122204e-07, + "loss": 0.939, + "step": 28636 + }, + { + "epoch": 0.92, + "learning_rate": 3.693219929930292e-07, + "loss": 0.915, + "step": 28637 + }, + { + "epoch": 0.92, + "learning_rate": 3.690431451813536e-07, + "loss": 0.9507, + "step": 28638 + }, + { + "epoch": 0.92, + "learning_rate": 3.687644006991864e-07, + "loss": 0.9653, + "step": 28639 + }, + { + "epoch": 0.92, + "learning_rate": 3.684857595495162e-07, + "loss": 0.9575, + "step": 28640 + }, + { + "epoch": 0.92, + "learning_rate": 3.6820722173533605e-07, + "loss": 0.979, + "step": 28641 + }, + { + "epoch": 0.92, + "learning_rate": 3.679287872596327e-07, + "loss": 0.8774, + "step": 28642 + }, + { + "epoch": 0.92, + "learning_rate": 3.676504561253913e-07, + "loss": 0.7969, + "step": 28643 + }, + { + "epoch": 0.92, + "learning_rate": 3.6737222833560183e-07, + "loss": 0.9087, + "step": 28644 + }, + { + "epoch": 0.92, + "learning_rate": 3.6709410389324515e-07, + "loss": 0.998, + "step": 28645 + }, + { + "epoch": 0.92, + "learning_rate": 3.668160828013112e-07, + "loss": 0.8394, + "step": 28646 + }, + { + "epoch": 0.92, + "learning_rate": 3.665381650627764e-07, + "loss": 1.0425, + "step": 28647 + }, + { + "epoch": 0.92, + "learning_rate": 3.662603506806262e-07, + "loss": 0.9971, + "step": 28648 + }, + { + "epoch": 0.92, + "learning_rate": 3.659826396578392e-07, + "loss": 0.9219, + "step": 28649 + }, + { + "epoch": 0.92, + "learning_rate": 3.657050319973976e-07, + "loss": 0.5142, + "step": 28650 + }, + { + "epoch": 0.92, + "learning_rate": 3.6542752770227675e-07, + "loss": 0.9553, + "step": 28651 + }, + { + "epoch": 0.92, + "learning_rate": 3.651501267754565e-07, + "loss": 0.9312, + "step": 28652 + }, + { + "epoch": 0.92, + "learning_rate": 3.648728292199111e-07, + "loss": 0.9097, + "step": 28653 + }, + { + "epoch": 0.92, + "learning_rate": 3.6459563503861927e-07, + "loss": 0.9082, + "step": 28654 + }, + { + "epoch": 0.92, + "learning_rate": 3.643185442345498e-07, + "loss": 0.8882, + "step": 28655 + }, + { + "epoch": 0.92, + "learning_rate": 3.640415568106792e-07, + "loss": 0.9292, + "step": 28656 + }, + { + "epoch": 0.92, + "learning_rate": 3.637646727699773e-07, + "loss": 0.8374, + "step": 28657 + }, + { + "epoch": 0.92, + "learning_rate": 3.6348789211541614e-07, + "loss": 0.9878, + "step": 28658 + }, + { + "epoch": 0.92, + "learning_rate": 3.6321121484996447e-07, + "loss": 0.9326, + "step": 28659 + }, + { + "epoch": 0.92, + "learning_rate": 3.62934640976591e-07, + "loss": 0.9414, + "step": 28660 + }, + { + "epoch": 0.92, + "learning_rate": 3.6265817049826456e-07, + "loss": 0.8501, + "step": 28661 + }, + { + "epoch": 0.92, + "learning_rate": 3.6238180341794826e-07, + "loss": 0.9087, + "step": 28662 + }, + { + "epoch": 0.92, + "learning_rate": 3.621055397386086e-07, + "loss": 0.9556, + "step": 28663 + }, + { + "epoch": 0.92, + "learning_rate": 3.6182937946321106e-07, + "loss": 0.8809, + "step": 28664 + }, + { + "epoch": 0.92, + "learning_rate": 3.615533225947154e-07, + "loss": 0.9199, + "step": 28665 + }, + { + "epoch": 0.92, + "learning_rate": 3.612773691360871e-07, + "loss": 1.0391, + "step": 28666 + }, + { + "epoch": 0.92, + "learning_rate": 3.610015190902838e-07, + "loss": 0.9565, + "step": 28667 + }, + { + "epoch": 0.92, + "learning_rate": 3.6072577246026753e-07, + "loss": 0.9907, + "step": 28668 + }, + { + "epoch": 0.92, + "learning_rate": 3.6045012924899483e-07, + "loss": 0.9399, + "step": 28669 + }, + { + "epoch": 0.92, + "learning_rate": 3.601745894594244e-07, + "loss": 0.9277, + "step": 28670 + }, + { + "epoch": 0.92, + "learning_rate": 3.598991530945106e-07, + "loss": 0.9102, + "step": 28671 + }, + { + "epoch": 0.92, + "learning_rate": 3.5962382015721e-07, + "loss": 0.9111, + "step": 28672 + }, + { + "epoch": 0.92, + "learning_rate": 3.593485906504757e-07, + "loss": 0.8359, + "step": 28673 + }, + { + "epoch": 0.92, + "learning_rate": 3.590734645772631e-07, + "loss": 0.9287, + "step": 28674 + }, + { + "epoch": 0.92, + "learning_rate": 3.5879844194052216e-07, + "loss": 0.9507, + "step": 28675 + }, + { + "epoch": 0.92, + "learning_rate": 3.5852352274320156e-07, + "loss": 0.8677, + "step": 28676 + }, + { + "epoch": 0.92, + "learning_rate": 3.5824870698825455e-07, + "loss": 0.9839, + "step": 28677 + }, + { + "epoch": 0.92, + "learning_rate": 3.579739946786276e-07, + "loss": 0.9746, + "step": 28678 + }, + { + "epoch": 0.92, + "learning_rate": 3.576993858172673e-07, + "loss": 0.8994, + "step": 28679 + }, + { + "epoch": 0.92, + "learning_rate": 3.574248804071223e-07, + "loss": 0.8872, + "step": 28680 + }, + { + "epoch": 0.92, + "learning_rate": 3.5715047845113703e-07, + "loss": 0.9302, + "step": 28681 + }, + { + "epoch": 0.92, + "learning_rate": 3.5687617995225356e-07, + "loss": 0.9185, + "step": 28682 + }, + { + "epoch": 0.92, + "learning_rate": 3.5660198491341725e-07, + "loss": 0.918, + "step": 28683 + }, + { + "epoch": 0.92, + "learning_rate": 3.5632789333756804e-07, + "loss": 0.9487, + "step": 28684 + }, + { + "epoch": 0.92, + "learning_rate": 3.5605390522764904e-07, + "loss": 0.916, + "step": 28685 + }, + { + "epoch": 0.92, + "learning_rate": 3.5578002058659577e-07, + "loss": 0.855, + "step": 28686 + }, + { + "epoch": 0.92, + "learning_rate": 3.555062394173503e-07, + "loss": 1.0054, + "step": 28687 + }, + { + "epoch": 0.92, + "learning_rate": 3.5523256172284914e-07, + "loss": 1.0112, + "step": 28688 + }, + { + "epoch": 0.92, + "learning_rate": 3.5495898750602773e-07, + "loss": 0.9243, + "step": 28689 + }, + { + "epoch": 0.92, + "learning_rate": 3.5468551676982155e-07, + "loss": 1.0039, + "step": 28690 + }, + { + "epoch": 0.92, + "learning_rate": 3.5441214951716594e-07, + "loss": 0.8853, + "step": 28691 + }, + { + "epoch": 0.92, + "learning_rate": 3.5413888575099307e-07, + "loss": 0.8979, + "step": 28692 + }, + { + "epoch": 0.92, + "learning_rate": 3.5386572547423393e-07, + "loss": 0.9224, + "step": 28693 + }, + { + "epoch": 0.92, + "learning_rate": 3.5359266868981836e-07, + "loss": 0.8833, + "step": 28694 + }, + { + "epoch": 0.92, + "learning_rate": 3.5331971540067953e-07, + "loss": 0.876, + "step": 28695 + }, + { + "epoch": 0.92, + "learning_rate": 3.5304686560974187e-07, + "loss": 0.915, + "step": 28696 + }, + { + "epoch": 0.92, + "learning_rate": 3.527741193199352e-07, + "loss": 0.9062, + "step": 28697 + }, + { + "epoch": 0.92, + "learning_rate": 3.52501476534185e-07, + "loss": 0.9014, + "step": 28698 + }, + { + "epoch": 0.92, + "learning_rate": 3.5222893725541885e-07, + "loss": 0.8813, + "step": 28699 + }, + { + "epoch": 0.92, + "learning_rate": 3.5195650148655557e-07, + "loss": 0.9497, + "step": 28700 + }, + { + "epoch": 0.92, + "learning_rate": 3.5168416923052284e-07, + "loss": 0.9355, + "step": 28701 + }, + { + "epoch": 0.92, + "learning_rate": 3.514119404902394e-07, + "loss": 0.9819, + "step": 28702 + }, + { + "epoch": 0.92, + "learning_rate": 3.5113981526862853e-07, + "loss": 0.5039, + "step": 28703 + }, + { + "epoch": 0.92, + "learning_rate": 3.5086779356860777e-07, + "loss": 0.8926, + "step": 28704 + }, + { + "epoch": 0.92, + "learning_rate": 3.5059587539309713e-07, + "loss": 0.8848, + "step": 28705 + }, + { + "epoch": 0.92, + "learning_rate": 3.50324060745012e-07, + "loss": 0.936, + "step": 28706 + }, + { + "epoch": 0.92, + "learning_rate": 3.500523496272734e-07, + "loss": 0.874, + "step": 28707 + }, + { + "epoch": 0.92, + "learning_rate": 3.4978074204279014e-07, + "loss": 0.7573, + "step": 28708 + }, + { + "epoch": 0.92, + "learning_rate": 3.495092379944798e-07, + "loss": 0.9067, + "step": 28709 + }, + { + "epoch": 0.92, + "learning_rate": 3.4923783748525343e-07, + "loss": 0.9409, + "step": 28710 + }, + { + "epoch": 0.92, + "learning_rate": 3.489665405180265e-07, + "loss": 0.8853, + "step": 28711 + }, + { + "epoch": 0.92, + "learning_rate": 3.486953470957055e-07, + "loss": 0.999, + "step": 28712 + }, + { + "epoch": 0.92, + "learning_rate": 3.4842425722120263e-07, + "loss": 0.8882, + "step": 28713 + }, + { + "epoch": 0.92, + "learning_rate": 3.481532708974256e-07, + "loss": 0.8677, + "step": 28714 + }, + { + "epoch": 0.92, + "learning_rate": 3.4788238812728305e-07, + "loss": 0.8062, + "step": 28715 + }, + { + "epoch": 0.92, + "learning_rate": 3.476116089136783e-07, + "loss": 0.9868, + "step": 28716 + }, + { + "epoch": 0.92, + "learning_rate": 3.47340933259519e-07, + "loss": 0.9644, + "step": 28717 + }, + { + "epoch": 0.92, + "learning_rate": 3.470703611677073e-07, + "loss": 0.9482, + "step": 28718 + }, + { + "epoch": 0.92, + "learning_rate": 3.4679989264114866e-07, + "loss": 0.834, + "step": 28719 + }, + { + "epoch": 0.92, + "learning_rate": 3.4652952768274296e-07, + "loss": 0.8042, + "step": 28720 + }, + { + "epoch": 0.92, + "learning_rate": 3.462592662953912e-07, + "loss": 1.0254, + "step": 28721 + }, + { + "epoch": 0.92, + "learning_rate": 3.459891084819944e-07, + "loss": 0.7856, + "step": 28722 + }, + { + "epoch": 0.92, + "learning_rate": 3.4571905424545025e-07, + "loss": 1.0415, + "step": 28723 + }, + { + "epoch": 0.92, + "learning_rate": 3.4544910358865427e-07, + "loss": 0.9058, + "step": 28724 + }, + { + "epoch": 0.92, + "learning_rate": 3.4517925651450515e-07, + "loss": 0.8643, + "step": 28725 + }, + { + "epoch": 0.92, + "learning_rate": 3.4490951302589725e-07, + "loss": 0.8955, + "step": 28726 + }, + { + "epoch": 0.92, + "learning_rate": 3.446398731257228e-07, + "loss": 0.938, + "step": 28727 + }, + { + "epoch": 0.92, + "learning_rate": 3.4437033681687826e-07, + "loss": 0.8955, + "step": 28728 + }, + { + "epoch": 0.92, + "learning_rate": 3.441009041022536e-07, + "loss": 1.0308, + "step": 28729 + }, + { + "epoch": 0.92, + "learning_rate": 3.438315749847398e-07, + "loss": 0.981, + "step": 28730 + }, + { + "epoch": 0.92, + "learning_rate": 3.435623494672258e-07, + "loss": 0.9082, + "step": 28731 + }, + { + "epoch": 0.92, + "learning_rate": 3.4329322755260017e-07, + "loss": 0.9028, + "step": 28732 + }, + { + "epoch": 0.92, + "learning_rate": 3.430242092437508e-07, + "loss": 0.4873, + "step": 28733 + }, + { + "epoch": 0.92, + "learning_rate": 3.4275529454356413e-07, + "loss": 0.897, + "step": 28734 + }, + { + "epoch": 0.92, + "learning_rate": 3.4248648345492466e-07, + "loss": 0.9561, + "step": 28735 + }, + { + "epoch": 0.92, + "learning_rate": 3.422177759807177e-07, + "loss": 0.9746, + "step": 28736 + }, + { + "epoch": 0.92, + "learning_rate": 3.419491721238255e-07, + "loss": 0.8892, + "step": 28737 + }, + { + "epoch": 0.92, + "learning_rate": 3.416806718871291e-07, + "loss": 0.8945, + "step": 28738 + }, + { + "epoch": 0.92, + "learning_rate": 3.414122752735094e-07, + "loss": 0.8794, + "step": 28739 + }, + { + "epoch": 0.92, + "learning_rate": 3.4114398228584645e-07, + "loss": 0.9805, + "step": 28740 + }, + { + "epoch": 0.92, + "learning_rate": 3.408757929270179e-07, + "loss": 0.9722, + "step": 28741 + }, + { + "epoch": 0.92, + "learning_rate": 3.4060770719990366e-07, + "loss": 0.959, + "step": 28742 + }, + { + "epoch": 0.92, + "learning_rate": 3.40339725107377e-07, + "loss": 0.979, + "step": 28743 + }, + { + "epoch": 0.92, + "learning_rate": 3.400718466523156e-07, + "loss": 1.0449, + "step": 28744 + }, + { + "epoch": 0.92, + "learning_rate": 3.3980407183759166e-07, + "loss": 0.9058, + "step": 28745 + }, + { + "epoch": 0.92, + "learning_rate": 3.395364006660784e-07, + "loss": 0.7432, + "step": 28746 + }, + { + "epoch": 0.92, + "learning_rate": 3.3926883314064685e-07, + "loss": 0.8491, + "step": 28747 + }, + { + "epoch": 0.92, + "learning_rate": 3.3900136926417025e-07, + "loss": 0.855, + "step": 28748 + }, + { + "epoch": 0.92, + "learning_rate": 3.3873400903951636e-07, + "loss": 0.9238, + "step": 28749 + }, + { + "epoch": 0.92, + "learning_rate": 3.3846675246955394e-07, + "loss": 0.8647, + "step": 28750 + }, + { + "epoch": 0.92, + "learning_rate": 3.3819959955714966e-07, + "loss": 0.8389, + "step": 28751 + }, + { + "epoch": 0.92, + "learning_rate": 3.3793255030517225e-07, + "loss": 0.9556, + "step": 28752 + }, + { + "epoch": 0.92, + "learning_rate": 3.376656047164839e-07, + "loss": 1.0215, + "step": 28753 + }, + { + "epoch": 0.92, + "learning_rate": 3.3739876279395013e-07, + "loss": 0.8335, + "step": 28754 + }, + { + "epoch": 0.92, + "learning_rate": 3.3713202454043305e-07, + "loss": 0.9285, + "step": 28755 + }, + { + "epoch": 0.92, + "learning_rate": 3.368653899587959e-07, + "loss": 0.8193, + "step": 28756 + }, + { + "epoch": 0.92, + "learning_rate": 3.365988590518976e-07, + "loss": 0.8789, + "step": 28757 + }, + { + "epoch": 0.92, + "learning_rate": 3.363324318225991e-07, + "loss": 0.8247, + "step": 28758 + }, + { + "epoch": 0.92, + "learning_rate": 3.360661082737593e-07, + "loss": 1.0054, + "step": 28759 + }, + { + "epoch": 0.92, + "learning_rate": 3.357998884082336e-07, + "loss": 1.0068, + "step": 28760 + }, + { + "epoch": 0.92, + "learning_rate": 3.3553377222887874e-07, + "loss": 0.9062, + "step": 28761 + }, + { + "epoch": 0.92, + "learning_rate": 3.352677597385512e-07, + "loss": 0.9224, + "step": 28762 + }, + { + "epoch": 0.92, + "learning_rate": 3.350018509401054e-07, + "loss": 0.8672, + "step": 28763 + }, + { + "epoch": 0.92, + "learning_rate": 3.347360458363902e-07, + "loss": 0.8291, + "step": 28764 + }, + { + "epoch": 0.92, + "learning_rate": 3.3447034443026217e-07, + "loss": 0.8843, + "step": 28765 + }, + { + "epoch": 0.92, + "learning_rate": 3.3420474672456906e-07, + "loss": 0.834, + "step": 28766 + }, + { + "epoch": 0.92, + "learning_rate": 3.33939252722163e-07, + "loss": 0.9199, + "step": 28767 + }, + { + "epoch": 0.92, + "learning_rate": 3.336738624258895e-07, + "loss": 0.9897, + "step": 28768 + }, + { + "epoch": 0.92, + "learning_rate": 3.3340857583859855e-07, + "loss": 0.9873, + "step": 28769 + }, + { + "epoch": 0.92, + "learning_rate": 3.3314339296313337e-07, + "loss": 0.9448, + "step": 28770 + }, + { + "epoch": 0.92, + "learning_rate": 3.328783138023428e-07, + "loss": 0.8286, + "step": 28771 + }, + { + "epoch": 0.92, + "learning_rate": 3.3261333835906685e-07, + "loss": 0.9893, + "step": 28772 + }, + { + "epoch": 0.92, + "learning_rate": 3.3234846663615205e-07, + "loss": 0.8364, + "step": 28773 + }, + { + "epoch": 0.92, + "learning_rate": 3.320836986364373e-07, + "loss": 1.0425, + "step": 28774 + }, + { + "epoch": 0.92, + "learning_rate": 3.3181903436276696e-07, + "loss": 0.9243, + "step": 28775 + }, + { + "epoch": 0.92, + "learning_rate": 3.3155447381797543e-07, + "loss": 0.8682, + "step": 28776 + }, + { + "epoch": 0.92, + "learning_rate": 3.312900170049049e-07, + "loss": 0.7979, + "step": 28777 + }, + { + "epoch": 0.92, + "learning_rate": 3.310256639263909e-07, + "loss": 0.9409, + "step": 28778 + }, + { + "epoch": 0.92, + "learning_rate": 3.307614145852722e-07, + "loss": 0.8696, + "step": 28779 + }, + { + "epoch": 0.92, + "learning_rate": 3.3049726898437997e-07, + "loss": 0.9634, + "step": 28780 + }, + { + "epoch": 0.92, + "learning_rate": 3.3023322712655294e-07, + "loss": 0.9395, + "step": 28781 + }, + { + "epoch": 0.92, + "learning_rate": 3.2996928901462e-07, + "loss": 0.9346, + "step": 28782 + }, + { + "epoch": 0.92, + "learning_rate": 3.297054546514145e-07, + "loss": 0.9785, + "step": 28783 + }, + { + "epoch": 0.92, + "learning_rate": 3.2944172403976626e-07, + "loss": 1.0127, + "step": 28784 + }, + { + "epoch": 0.92, + "learning_rate": 3.291780971825065e-07, + "loss": 0.9121, + "step": 28785 + }, + { + "epoch": 0.92, + "learning_rate": 3.2891457408246175e-07, + "loss": 0.9268, + "step": 28786 + }, + { + "epoch": 0.92, + "learning_rate": 3.286511547424609e-07, + "loss": 0.8696, + "step": 28787 + }, + { + "epoch": 0.92, + "learning_rate": 3.283878391653283e-07, + "loss": 0.9019, + "step": 28788 + }, + { + "epoch": 0.92, + "learning_rate": 3.2812462735389074e-07, + "loss": 0.9321, + "step": 28789 + }, + { + "epoch": 0.92, + "learning_rate": 3.2786151931097245e-07, + "loss": 1.0635, + "step": 28790 + }, + { + "epoch": 0.92, + "learning_rate": 3.2759851503939567e-07, + "loss": 0.9595, + "step": 28791 + }, + { + "epoch": 0.92, + "learning_rate": 3.273356145419804e-07, + "loss": 0.9429, + "step": 28792 + }, + { + "epoch": 0.92, + "learning_rate": 3.270728178215499e-07, + "loss": 0.9883, + "step": 28793 + }, + { + "epoch": 0.92, + "learning_rate": 3.268101248809219e-07, + "loss": 1.0093, + "step": 28794 + }, + { + "epoch": 0.92, + "learning_rate": 3.2654753572291644e-07, + "loss": 0.875, + "step": 28795 + }, + { + "epoch": 0.92, + "learning_rate": 3.2628505035034784e-07, + "loss": 0.9858, + "step": 28796 + }, + { + "epoch": 0.92, + "learning_rate": 3.2602266876603727e-07, + "loss": 0.8633, + "step": 28797 + }, + { + "epoch": 0.92, + "learning_rate": 3.2576039097279464e-07, + "loss": 0.9756, + "step": 28798 + }, + { + "epoch": 0.92, + "learning_rate": 3.254982169734378e-07, + "loss": 0.8721, + "step": 28799 + }, + { + "epoch": 0.92, + "learning_rate": 3.2523614677077654e-07, + "loss": 0.9766, + "step": 28800 + }, + { + "epoch": 0.92, + "learning_rate": 3.249741803676243e-07, + "loss": 0.9683, + "step": 28801 + }, + { + "epoch": 0.92, + "learning_rate": 3.2471231776679213e-07, + "loss": 0.8994, + "step": 28802 + }, + { + "epoch": 0.92, + "learning_rate": 3.244505589710867e-07, + "loss": 0.8262, + "step": 28803 + }, + { + "epoch": 0.92, + "learning_rate": 3.2418890398332016e-07, + "loss": 0.9399, + "step": 28804 + }, + { + "epoch": 0.92, + "learning_rate": 3.2392735280629807e-07, + "loss": 0.8848, + "step": 28805 + }, + { + "epoch": 0.92, + "learning_rate": 3.236659054428248e-07, + "loss": 0.9717, + "step": 28806 + }, + { + "epoch": 0.92, + "learning_rate": 3.2340456189570825e-07, + "loss": 0.9805, + "step": 28807 + }, + { + "epoch": 0.92, + "learning_rate": 3.231433221677516e-07, + "loss": 0.9907, + "step": 28808 + }, + { + "epoch": 0.92, + "learning_rate": 3.2288218626175704e-07, + "loss": 0.936, + "step": 28809 + }, + { + "epoch": 0.92, + "learning_rate": 3.226211541805269e-07, + "loss": 0.9316, + "step": 28810 + }, + { + "epoch": 0.92, + "learning_rate": 3.223602259268599e-07, + "loss": 0.9551, + "step": 28811 + }, + { + "epoch": 0.92, + "learning_rate": 3.2209940150355834e-07, + "loss": 0.8623, + "step": 28812 + }, + { + "epoch": 0.92, + "learning_rate": 3.2183868091341884e-07, + "loss": 0.9863, + "step": 28813 + }, + { + "epoch": 0.92, + "learning_rate": 3.215780641592392e-07, + "loss": 0.9131, + "step": 28814 + }, + { + "epoch": 0.92, + "learning_rate": 3.213175512438149e-07, + "loss": 0.9878, + "step": 28815 + }, + { + "epoch": 0.92, + "learning_rate": 3.2105714216994266e-07, + "loss": 1.0244, + "step": 28816 + }, + { + "epoch": 0.92, + "learning_rate": 3.2079683694041354e-07, + "loss": 0.9419, + "step": 28817 + }, + { + "epoch": 0.92, + "learning_rate": 3.2053663555802304e-07, + "loss": 0.9165, + "step": 28818 + }, + { + "epoch": 0.92, + "learning_rate": 3.202765380255601e-07, + "loss": 0.6743, + "step": 28819 + }, + { + "epoch": 0.92, + "learning_rate": 3.200165443458203e-07, + "loss": 0.9536, + "step": 28820 + }, + { + "epoch": 0.92, + "learning_rate": 3.197566545215869e-07, + "loss": 0.9702, + "step": 28821 + }, + { + "epoch": 0.92, + "learning_rate": 3.194968685556521e-07, + "loss": 0.8774, + "step": 28822 + }, + { + "epoch": 0.92, + "learning_rate": 3.1923718645080036e-07, + "loss": 0.9365, + "step": 28823 + }, + { + "epoch": 0.92, + "learning_rate": 3.1897760820982173e-07, + "loss": 0.8496, + "step": 28824 + }, + { + "epoch": 0.92, + "learning_rate": 3.187181338354972e-07, + "loss": 0.9277, + "step": 28825 + }, + { + "epoch": 0.92, + "learning_rate": 3.1845876333061355e-07, + "loss": 0.8989, + "step": 28826 + }, + { + "epoch": 0.92, + "learning_rate": 3.181994966979518e-07, + "loss": 0.9146, + "step": 28827 + }, + { + "epoch": 0.92, + "learning_rate": 3.1794033394029534e-07, + "loss": 0.8755, + "step": 28828 + }, + { + "epoch": 0.92, + "learning_rate": 3.176812750604219e-07, + "loss": 0.8352, + "step": 28829 + }, + { + "epoch": 0.92, + "learning_rate": 3.1742232006111374e-07, + "loss": 0.9443, + "step": 28830 + }, + { + "epoch": 0.92, + "learning_rate": 3.1716346894514636e-07, + "loss": 0.8687, + "step": 28831 + }, + { + "epoch": 0.92, + "learning_rate": 3.1690472171529985e-07, + "loss": 0.9189, + "step": 28832 + }, + { + "epoch": 0.92, + "learning_rate": 3.166460783743486e-07, + "loss": 0.8696, + "step": 28833 + }, + { + "epoch": 0.92, + "learning_rate": 3.1638753892506814e-07, + "loss": 0.855, + "step": 28834 + }, + { + "epoch": 0.92, + "learning_rate": 3.1612910337023184e-07, + "loss": 0.9922, + "step": 28835 + }, + { + "epoch": 0.92, + "learning_rate": 3.1587077171261414e-07, + "loss": 0.8599, + "step": 28836 + }, + { + "epoch": 0.92, + "learning_rate": 3.1561254395498287e-07, + "loss": 0.9058, + "step": 28837 + }, + { + "epoch": 0.92, + "learning_rate": 3.1535442010011243e-07, + "loss": 0.9146, + "step": 28838 + }, + { + "epoch": 0.92, + "learning_rate": 3.1509640015076946e-07, + "loss": 0.9409, + "step": 28839 + }, + { + "epoch": 0.92, + "learning_rate": 3.1483848410972405e-07, + "loss": 0.9082, + "step": 28840 + }, + { + "epoch": 0.92, + "learning_rate": 3.1458067197974173e-07, + "loss": 0.957, + "step": 28841 + }, + { + "epoch": 0.92, + "learning_rate": 3.143229637635903e-07, + "loss": 0.8782, + "step": 28842 + }, + { + "epoch": 0.92, + "learning_rate": 3.1406535946403525e-07, + "loss": 0.7852, + "step": 28843 + }, + { + "epoch": 0.92, + "learning_rate": 3.1380785908383783e-07, + "loss": 0.9971, + "step": 28844 + }, + { + "epoch": 0.92, + "learning_rate": 3.135504626257613e-07, + "loss": 0.9941, + "step": 28845 + }, + { + "epoch": 0.92, + "learning_rate": 3.13293170092569e-07, + "loss": 0.9424, + "step": 28846 + }, + { + "epoch": 0.92, + "learning_rate": 3.1303598148701984e-07, + "loss": 0.509, + "step": 28847 + }, + { + "epoch": 0.92, + "learning_rate": 3.127788968118728e-07, + "loss": 0.8618, + "step": 28848 + }, + { + "epoch": 0.92, + "learning_rate": 3.125219160698878e-07, + "loss": 1.0503, + "step": 28849 + }, + { + "epoch": 0.92, + "learning_rate": 3.1226503926382157e-07, + "loss": 0.8428, + "step": 28850 + }, + { + "epoch": 0.92, + "learning_rate": 3.1200826639642747e-07, + "loss": 0.4746, + "step": 28851 + }, + { + "epoch": 0.92, + "learning_rate": 3.117515974704643e-07, + "loss": 0.9404, + "step": 28852 + }, + { + "epoch": 0.92, + "learning_rate": 3.114950324886845e-07, + "loss": 0.9575, + "step": 28853 + }, + { + "epoch": 0.92, + "learning_rate": 3.1123857145383907e-07, + "loss": 1.0386, + "step": 28854 + }, + { + "epoch": 0.92, + "learning_rate": 3.109822143686814e-07, + "loss": 0.7407, + "step": 28855 + }, + { + "epoch": 0.92, + "learning_rate": 3.1072596123596034e-07, + "loss": 0.8931, + "step": 28856 + }, + { + "epoch": 0.92, + "learning_rate": 3.1046981205842707e-07, + "loss": 0.8887, + "step": 28857 + }, + { + "epoch": 0.92, + "learning_rate": 3.102137668388294e-07, + "loss": 0.8521, + "step": 28858 + }, + { + "epoch": 0.92, + "learning_rate": 3.0995782557991407e-07, + "loss": 0.79, + "step": 28859 + }, + { + "epoch": 0.92, + "learning_rate": 3.0970198828442544e-07, + "loss": 0.9658, + "step": 28860 + }, + { + "epoch": 0.92, + "learning_rate": 3.0944625495511025e-07, + "loss": 0.9365, + "step": 28861 + }, + { + "epoch": 0.92, + "learning_rate": 3.091906255947119e-07, + "loss": 0.8755, + "step": 28862 + }, + { + "epoch": 0.92, + "learning_rate": 3.089351002059726e-07, + "loss": 0.9521, + "step": 28863 + }, + { + "epoch": 0.92, + "learning_rate": 3.0867967879163353e-07, + "loss": 0.8945, + "step": 28864 + }, + { + "epoch": 0.92, + "learning_rate": 3.0842436135443797e-07, + "loss": 0.9795, + "step": 28865 + }, + { + "epoch": 0.92, + "learning_rate": 3.081691478971216e-07, + "loss": 0.9907, + "step": 28866 + }, + { + "epoch": 0.92, + "learning_rate": 3.079140384224244e-07, + "loss": 0.9917, + "step": 28867 + }, + { + "epoch": 0.92, + "learning_rate": 3.076590329330809e-07, + "loss": 0.916, + "step": 28868 + }, + { + "epoch": 0.92, + "learning_rate": 3.07404131431831e-07, + "loss": 0.9155, + "step": 28869 + }, + { + "epoch": 0.92, + "learning_rate": 3.07149333921406e-07, + "loss": 0.9541, + "step": 28870 + }, + { + "epoch": 0.92, + "learning_rate": 3.068946404045425e-07, + "loss": 0.978, + "step": 28871 + }, + { + "epoch": 0.92, + "learning_rate": 3.0664005088397065e-07, + "loss": 0.8076, + "step": 28872 + }, + { + "epoch": 0.92, + "learning_rate": 3.063855653624248e-07, + "loss": 0.9316, + "step": 28873 + }, + { + "epoch": 0.92, + "learning_rate": 3.061311838426317e-07, + "loss": 0.981, + "step": 28874 + }, + { + "epoch": 0.92, + "learning_rate": 3.058769063273237e-07, + "loss": 0.9321, + "step": 28875 + }, + { + "epoch": 0.92, + "learning_rate": 3.056227328192263e-07, + "loss": 0.9673, + "step": 28876 + }, + { + "epoch": 0.92, + "learning_rate": 3.0536866332106843e-07, + "loss": 0.7979, + "step": 28877 + }, + { + "epoch": 0.92, + "learning_rate": 3.0511469783557455e-07, + "loss": 0.9038, + "step": 28878 + }, + { + "epoch": 0.92, + "learning_rate": 3.048608363654715e-07, + "loss": 0.9194, + "step": 28879 + }, + { + "epoch": 0.92, + "learning_rate": 3.046070789134814e-07, + "loss": 0.9453, + "step": 28880 + }, + { + "epoch": 0.92, + "learning_rate": 3.043534254823266e-07, + "loss": 0.9233, + "step": 28881 + }, + { + "epoch": 0.92, + "learning_rate": 3.040998760747282e-07, + "loss": 0.8599, + "step": 28882 + }, + { + "epoch": 0.92, + "learning_rate": 3.0384643069340857e-07, + "loss": 0.9185, + "step": 28883 + }, + { + "epoch": 0.92, + "learning_rate": 3.0359308934108435e-07, + "loss": 0.9653, + "step": 28884 + }, + { + "epoch": 0.92, + "learning_rate": 3.0333985202047555e-07, + "loss": 0.9229, + "step": 28885 + }, + { + "epoch": 0.92, + "learning_rate": 3.0308671873429895e-07, + "loss": 0.9404, + "step": 28886 + }, + { + "epoch": 0.92, + "learning_rate": 3.0283368948526794e-07, + "loss": 0.9849, + "step": 28887 + }, + { + "epoch": 0.92, + "learning_rate": 3.0258076427610137e-07, + "loss": 0.8164, + "step": 28888 + }, + { + "epoch": 0.92, + "learning_rate": 3.0232794310950943e-07, + "loss": 0.9341, + "step": 28889 + }, + { + "epoch": 0.92, + "learning_rate": 3.020752259882054e-07, + "loss": 0.8774, + "step": 28890 + }, + { + "epoch": 0.92, + "learning_rate": 3.0182261291490266e-07, + "loss": 0.9146, + "step": 28891 + }, + { + "epoch": 0.92, + "learning_rate": 3.015701038923091e-07, + "loss": 0.8765, + "step": 28892 + }, + { + "epoch": 0.92, + "learning_rate": 3.013176989231337e-07, + "loss": 0.9067, + "step": 28893 + }, + { + "epoch": 0.92, + "learning_rate": 3.010653980100864e-07, + "loss": 0.9072, + "step": 28894 + }, + { + "epoch": 0.92, + "learning_rate": 3.008132011558729e-07, + "loss": 0.9175, + "step": 28895 + }, + { + "epoch": 0.92, + "learning_rate": 3.0056110836319876e-07, + "loss": 0.8892, + "step": 28896 + }, + { + "epoch": 0.92, + "learning_rate": 3.003091196347696e-07, + "loss": 0.7529, + "step": 28897 + }, + { + "epoch": 0.92, + "learning_rate": 3.0005723497328885e-07, + "loss": 0.8613, + "step": 28898 + }, + { + "epoch": 0.92, + "learning_rate": 2.998054543814577e-07, + "loss": 0.8313, + "step": 28899 + }, + { + "epoch": 0.92, + "learning_rate": 2.995537778619795e-07, + "loss": 0.9136, + "step": 28900 + }, + { + "epoch": 0.92, + "learning_rate": 2.993022054175521e-07, + "loss": 0.9023, + "step": 28901 + }, + { + "epoch": 0.92, + "learning_rate": 2.990507370508766e-07, + "loss": 0.7705, + "step": 28902 + }, + { + "epoch": 0.92, + "learning_rate": 2.98799372764651e-07, + "loss": 1.0005, + "step": 28903 + }, + { + "epoch": 0.92, + "learning_rate": 2.985481125615708e-07, + "loss": 0.8657, + "step": 28904 + }, + { + "epoch": 0.92, + "learning_rate": 2.9829695644433055e-07, + "loss": 0.8511, + "step": 28905 + }, + { + "epoch": 0.92, + "learning_rate": 2.980459044156292e-07, + "loss": 0.9111, + "step": 28906 + }, + { + "epoch": 0.92, + "learning_rate": 2.977949564781557e-07, + "loss": 0.9092, + "step": 28907 + }, + { + "epoch": 0.92, + "learning_rate": 2.975441126346057e-07, + "loss": 0.9282, + "step": 28908 + }, + { + "epoch": 0.92, + "learning_rate": 2.972933728876681e-07, + "loss": 0.9531, + "step": 28909 + }, + { + "epoch": 0.92, + "learning_rate": 2.970427372400353e-07, + "loss": 0.9346, + "step": 28910 + }, + { + "epoch": 0.92, + "learning_rate": 2.96792205694395e-07, + "loss": 0.9888, + "step": 28911 + }, + { + "epoch": 0.92, + "learning_rate": 2.965417782534363e-07, + "loss": 0.8696, + "step": 28912 + }, + { + "epoch": 0.92, + "learning_rate": 2.9629145491984257e-07, + "loss": 0.4744, + "step": 28913 + }, + { + "epoch": 0.92, + "learning_rate": 2.9604123569630383e-07, + "loss": 0.8955, + "step": 28914 + }, + { + "epoch": 0.92, + "learning_rate": 2.9579112058550244e-07, + "loss": 0.8301, + "step": 28915 + }, + { + "epoch": 0.92, + "learning_rate": 2.955411095901217e-07, + "loss": 0.8691, + "step": 28916 + }, + { + "epoch": 0.92, + "learning_rate": 2.9529120271284515e-07, + "loss": 0.875, + "step": 28917 + }, + { + "epoch": 0.92, + "learning_rate": 2.9504139995635395e-07, + "loss": 0.8555, + "step": 28918 + }, + { + "epoch": 0.92, + "learning_rate": 2.947917013233259e-07, + "loss": 0.9248, + "step": 28919 + }, + { + "epoch": 0.92, + "learning_rate": 2.9454210681644336e-07, + "loss": 0.8232, + "step": 28920 + }, + { + "epoch": 0.92, + "learning_rate": 2.9429261643838194e-07, + "loss": 0.9351, + "step": 28921 + }, + { + "epoch": 0.92, + "learning_rate": 2.9404323019181836e-07, + "loss": 0.8662, + "step": 28922 + }, + { + "epoch": 0.92, + "learning_rate": 2.937939480794294e-07, + "loss": 0.4485, + "step": 28923 + }, + { + "epoch": 0.93, + "learning_rate": 2.935447701038896e-07, + "loss": 0.9912, + "step": 28924 + }, + { + "epoch": 0.93, + "learning_rate": 2.932956962678712e-07, + "loss": 0.7886, + "step": 28925 + }, + { + "epoch": 0.93, + "learning_rate": 2.9304672657404776e-07, + "loss": 0.8833, + "step": 28926 + }, + { + "epoch": 0.93, + "learning_rate": 2.927978610250892e-07, + "loss": 0.9536, + "step": 28927 + }, + { + "epoch": 0.93, + "learning_rate": 2.9254909962366684e-07, + "loss": 0.8379, + "step": 28928 + }, + { + "epoch": 0.93, + "learning_rate": 2.923004423724474e-07, + "loss": 0.9961, + "step": 28929 + }, + { + "epoch": 0.93, + "learning_rate": 2.92051889274102e-07, + "loss": 0.9165, + "step": 28930 + }, + { + "epoch": 0.93, + "learning_rate": 2.918034403312953e-07, + "loss": 0.8687, + "step": 28931 + }, + { + "epoch": 0.93, + "learning_rate": 2.915550955466917e-07, + "loss": 0.9468, + "step": 28932 + }, + { + "epoch": 0.93, + "learning_rate": 2.913068549229592e-07, + "loss": 0.8628, + "step": 28933 + }, + { + "epoch": 0.93, + "learning_rate": 2.910587184627578e-07, + "loss": 0.8994, + "step": 28934 + }, + { + "epoch": 0.93, + "learning_rate": 2.90810686168751e-07, + "loss": 1.0093, + "step": 28935 + }, + { + "epoch": 0.93, + "learning_rate": 2.90562758043601e-07, + "loss": 0.8872, + "step": 28936 + }, + { + "epoch": 0.93, + "learning_rate": 2.903149340899658e-07, + "loss": 0.855, + "step": 28937 + }, + { + "epoch": 0.93, + "learning_rate": 2.9006721431050545e-07, + "loss": 0.8813, + "step": 28938 + }, + { + "epoch": 0.93, + "learning_rate": 2.898195987078778e-07, + "loss": 1.0181, + "step": 28939 + }, + { + "epoch": 0.93, + "learning_rate": 2.895720872847385e-07, + "loss": 0.845, + "step": 28940 + }, + { + "epoch": 0.93, + "learning_rate": 2.893246800437455e-07, + "loss": 0.8843, + "step": 28941 + }, + { + "epoch": 0.93, + "learning_rate": 2.890773769875488e-07, + "loss": 0.9756, + "step": 28942 + }, + { + "epoch": 0.93, + "learning_rate": 2.888301781188063e-07, + "loss": 0.8604, + "step": 28943 + }, + { + "epoch": 0.93, + "learning_rate": 2.8858308344016704e-07, + "loss": 1.0039, + "step": 28944 + }, + { + "epoch": 0.93, + "learning_rate": 2.883360929542833e-07, + "loss": 0.958, + "step": 28945 + }, + { + "epoch": 0.93, + "learning_rate": 2.8808920666380523e-07, + "loss": 1.0249, + "step": 28946 + }, + { + "epoch": 0.93, + "learning_rate": 2.8784242457138176e-07, + "loss": 0.9468, + "step": 28947 + }, + { + "epoch": 0.93, + "learning_rate": 2.8759574667965975e-07, + "loss": 0.9473, + "step": 28948 + }, + { + "epoch": 0.93, + "learning_rate": 2.8734917299128695e-07, + "loss": 0.9839, + "step": 28949 + }, + { + "epoch": 0.93, + "learning_rate": 2.871027035089069e-07, + "loss": 0.874, + "step": 28950 + }, + { + "epoch": 0.93, + "learning_rate": 2.8685633823516633e-07, + "loss": 0.96, + "step": 28951 + }, + { + "epoch": 0.93, + "learning_rate": 2.8661007717270537e-07, + "loss": 0.8428, + "step": 28952 + }, + { + "epoch": 0.93, + "learning_rate": 2.863639203241697e-07, + "loss": 0.7405, + "step": 28953 + }, + { + "epoch": 0.93, + "learning_rate": 2.861178676921983e-07, + "loss": 1.02, + "step": 28954 + }, + { + "epoch": 0.93, + "learning_rate": 2.858719192794313e-07, + "loss": 0.9922, + "step": 28955 + }, + { + "epoch": 0.93, + "learning_rate": 2.856260750885087e-07, + "loss": 0.937, + "step": 28956 + }, + { + "epoch": 0.93, + "learning_rate": 2.853803351220652e-07, + "loss": 0.9468, + "step": 28957 + }, + { + "epoch": 0.93, + "learning_rate": 2.8513469938273974e-07, + "loss": 0.9292, + "step": 28958 + }, + { + "epoch": 0.93, + "learning_rate": 2.848891678731669e-07, + "loss": 0.9268, + "step": 28959 + }, + { + "epoch": 0.93, + "learning_rate": 2.846437405959812e-07, + "loss": 0.8672, + "step": 28960 + }, + { + "epoch": 0.93, + "learning_rate": 2.8439841755381614e-07, + "loss": 0.9121, + "step": 28961 + }, + { + "epoch": 0.93, + "learning_rate": 2.8415319874930294e-07, + "loss": 0.8794, + "step": 28962 + }, + { + "epoch": 0.93, + "learning_rate": 2.839080841850739e-07, + "loss": 0.9458, + "step": 28963 + }, + { + "epoch": 0.93, + "learning_rate": 2.836630738637569e-07, + "loss": 0.9785, + "step": 28964 + }, + { + "epoch": 0.93, + "learning_rate": 2.8341816778798215e-07, + "loss": 0.9663, + "step": 28965 + }, + { + "epoch": 0.93, + "learning_rate": 2.8317336596037526e-07, + "loss": 0.5833, + "step": 28966 + }, + { + "epoch": 0.93, + "learning_rate": 2.8292866838356637e-07, + "loss": 0.9082, + "step": 28967 + }, + { + "epoch": 0.93, + "learning_rate": 2.826840750601767e-07, + "loss": 0.979, + "step": 28968 + }, + { + "epoch": 0.93, + "learning_rate": 2.8243958599283306e-07, + "loss": 0.8804, + "step": 28969 + }, + { + "epoch": 0.93, + "learning_rate": 2.8219520118415776e-07, + "loss": 0.8652, + "step": 28970 + }, + { + "epoch": 0.93, + "learning_rate": 2.8195092063677323e-07, + "loss": 0.8491, + "step": 28971 + }, + { + "epoch": 0.93, + "learning_rate": 2.817067443532984e-07, + "loss": 1.0327, + "step": 28972 + }, + { + "epoch": 0.93, + "learning_rate": 2.814626723363567e-07, + "loss": 0.8306, + "step": 28973 + }, + { + "epoch": 0.93, + "learning_rate": 2.8121870458856284e-07, + "loss": 0.9585, + "step": 28974 + }, + { + "epoch": 0.93, + "learning_rate": 2.809748411125368e-07, + "loss": 0.9639, + "step": 28975 + }, + { + "epoch": 0.93, + "learning_rate": 2.8073108191089436e-07, + "loss": 0.915, + "step": 28976 + }, + { + "epoch": 0.93, + "learning_rate": 2.804874269862501e-07, + "loss": 0.9331, + "step": 28977 + }, + { + "epoch": 0.93, + "learning_rate": 2.8024387634121966e-07, + "loss": 1.0039, + "step": 28978 + }, + { + "epoch": 0.93, + "learning_rate": 2.8000042997841426e-07, + "loss": 0.9893, + "step": 28979 + }, + { + "epoch": 0.93, + "learning_rate": 2.797570879004474e-07, + "loss": 0.9033, + "step": 28980 + }, + { + "epoch": 0.93, + "learning_rate": 2.7951385010992813e-07, + "loss": 0.8257, + "step": 28981 + }, + { + "epoch": 0.93, + "learning_rate": 2.7927071660946767e-07, + "loss": 0.8093, + "step": 28982 + }, + { + "epoch": 0.93, + "learning_rate": 2.79027687401674e-07, + "loss": 0.9282, + "step": 28983 + }, + { + "epoch": 0.93, + "learning_rate": 2.787847624891549e-07, + "loss": 0.9341, + "step": 28984 + }, + { + "epoch": 0.93, + "learning_rate": 2.78541941874515e-07, + "loss": 0.9766, + "step": 28985 + }, + { + "epoch": 0.93, + "learning_rate": 2.782992255603645e-07, + "loss": 0.8809, + "step": 28986 + }, + { + "epoch": 0.93, + "learning_rate": 2.780566135493001e-07, + "loss": 1.0317, + "step": 28987 + }, + { + "epoch": 0.93, + "learning_rate": 2.77814105843931e-07, + "loss": 0.9092, + "step": 28988 + }, + { + "epoch": 0.93, + "learning_rate": 2.775717024468538e-07, + "loss": 0.8301, + "step": 28989 + }, + { + "epoch": 0.93, + "learning_rate": 2.7732940336067437e-07, + "loss": 0.9287, + "step": 28990 + }, + { + "epoch": 0.93, + "learning_rate": 2.7708720858798834e-07, + "loss": 1.0088, + "step": 28991 + }, + { + "epoch": 0.93, + "learning_rate": 2.768451181313969e-07, + "loss": 0.8413, + "step": 28992 + }, + { + "epoch": 0.93, + "learning_rate": 2.7660313199349585e-07, + "loss": 0.9902, + "step": 28993 + }, + { + "epoch": 0.93, + "learning_rate": 2.763612501768831e-07, + "loss": 0.9907, + "step": 28994 + }, + { + "epoch": 0.93, + "learning_rate": 2.7611947268415095e-07, + "loss": 1.0044, + "step": 28995 + }, + { + "epoch": 0.93, + "learning_rate": 2.758777995178963e-07, + "loss": 0.9541, + "step": 28996 + }, + { + "epoch": 0.93, + "learning_rate": 2.756362306807092e-07, + "loss": 0.9487, + "step": 28997 + }, + { + "epoch": 0.93, + "learning_rate": 2.7539476617518326e-07, + "loss": 0.9385, + "step": 28998 + }, + { + "epoch": 0.93, + "learning_rate": 2.751534060039085e-07, + "loss": 1.0024, + "step": 28999 + }, + { + "epoch": 0.93, + "learning_rate": 2.7491215016947624e-07, + "loss": 0.8936, + "step": 29000 + }, + { + "epoch": 0.93, + "learning_rate": 2.746709986744733e-07, + "loss": 1.0083, + "step": 29001 + }, + { + "epoch": 0.93, + "learning_rate": 2.7442995152148656e-07, + "loss": 0.9414, + "step": 29002 + }, + { + "epoch": 0.93, + "learning_rate": 2.7418900871310163e-07, + "loss": 0.9067, + "step": 29003 + }, + { + "epoch": 0.93, + "learning_rate": 2.7394817025190645e-07, + "loss": 0.8713, + "step": 29004 + }, + { + "epoch": 0.93, + "learning_rate": 2.737074361404812e-07, + "loss": 0.9551, + "step": 29005 + }, + { + "epoch": 0.93, + "learning_rate": 2.734668063814128e-07, + "loss": 0.9634, + "step": 29006 + }, + { + "epoch": 0.93, + "learning_rate": 2.73226280977279e-07, + "loss": 0.9263, + "step": 29007 + }, + { + "epoch": 0.93, + "learning_rate": 2.729858599306634e-07, + "loss": 0.936, + "step": 29008 + }, + { + "epoch": 0.93, + "learning_rate": 2.7274554324414394e-07, + "loss": 0.8662, + "step": 29009 + }, + { + "epoch": 0.93, + "learning_rate": 2.7250533092029965e-07, + "loss": 1.0249, + "step": 29010 + }, + { + "epoch": 0.93, + "learning_rate": 2.722652229617051e-07, + "loss": 0.834, + "step": 29011 + }, + { + "epoch": 0.93, + "learning_rate": 2.720252193709405e-07, + "loss": 0.9375, + "step": 29012 + }, + { + "epoch": 0.93, + "learning_rate": 2.7178532015057826e-07, + "loss": 0.9268, + "step": 29013 + }, + { + "epoch": 0.93, + "learning_rate": 2.71545525303194e-07, + "loss": 0.8896, + "step": 29014 + }, + { + "epoch": 0.93, + "learning_rate": 2.7130583483135796e-07, + "loss": 1.0283, + "step": 29015 + }, + { + "epoch": 0.93, + "learning_rate": 2.7106624873764473e-07, + "loss": 0.9268, + "step": 29016 + }, + { + "epoch": 0.93, + "learning_rate": 2.708267670246212e-07, + "loss": 0.897, + "step": 29017 + }, + { + "epoch": 0.93, + "learning_rate": 2.7058738969485966e-07, + "loss": 0.8794, + "step": 29018 + }, + { + "epoch": 0.93, + "learning_rate": 2.703481167509281e-07, + "loss": 0.8945, + "step": 29019 + }, + { + "epoch": 0.93, + "learning_rate": 2.7010894819539115e-07, + "loss": 0.9155, + "step": 29020 + }, + { + "epoch": 0.93, + "learning_rate": 2.6986988403081895e-07, + "loss": 0.9158, + "step": 29021 + }, + { + "epoch": 0.93, + "learning_rate": 2.6963092425977164e-07, + "loss": 0.9014, + "step": 29022 + }, + { + "epoch": 0.93, + "learning_rate": 2.693920688848173e-07, + "loss": 0.9233, + "step": 29023 + }, + { + "epoch": 0.93, + "learning_rate": 2.6915331790851707e-07, + "loss": 0.9702, + "step": 29024 + }, + { + "epoch": 0.93, + "learning_rate": 2.689146713334312e-07, + "loss": 0.9399, + "step": 29025 + }, + { + "epoch": 0.93, + "learning_rate": 2.686761291621209e-07, + "loss": 0.8921, + "step": 29026 + }, + { + "epoch": 0.93, + "learning_rate": 2.6843769139714646e-07, + "loss": 1.0117, + "step": 29027 + }, + { + "epoch": 0.93, + "learning_rate": 2.6819935804106467e-07, + "loss": 0.958, + "step": 29028 + }, + { + "epoch": 0.93, + "learning_rate": 2.679611290964335e-07, + "loss": 0.8442, + "step": 29029 + }, + { + "epoch": 0.93, + "learning_rate": 2.677230045658075e-07, + "loss": 0.9907, + "step": 29030 + }, + { + "epoch": 0.93, + "learning_rate": 2.674849844517446e-07, + "loss": 0.9067, + "step": 29031 + }, + { + "epoch": 0.93, + "learning_rate": 2.6724706875679405e-07, + "loss": 1.0352, + "step": 29032 + }, + { + "epoch": 0.93, + "learning_rate": 2.6700925748351257e-07, + "loss": 0.8242, + "step": 29033 + }, + { + "epoch": 0.93, + "learning_rate": 2.667715506344482e-07, + "loss": 0.9414, + "step": 29034 + }, + { + "epoch": 0.93, + "learning_rate": 2.665339482121543e-07, + "loss": 0.9766, + "step": 29035 + }, + { + "epoch": 0.93, + "learning_rate": 2.662964502191778e-07, + "loss": 0.9497, + "step": 29036 + }, + { + "epoch": 0.93, + "learning_rate": 2.6605905665806784e-07, + "loss": 0.9678, + "step": 29037 + }, + { + "epoch": 0.93, + "learning_rate": 2.6582176753137123e-07, + "loss": 0.9912, + "step": 29038 + }, + { + "epoch": 0.93, + "learning_rate": 2.655845828416348e-07, + "loss": 0.7993, + "step": 29039 + }, + { + "epoch": 0.93, + "learning_rate": 2.6534750259140097e-07, + "loss": 0.9033, + "step": 29040 + }, + { + "epoch": 0.93, + "learning_rate": 2.6511052678321545e-07, + "loss": 0.9233, + "step": 29041 + }, + { + "epoch": 0.93, + "learning_rate": 2.6487365541961963e-07, + "loss": 0.9014, + "step": 29042 + }, + { + "epoch": 0.93, + "learning_rate": 2.6463688850315473e-07, + "loss": 0.9136, + "step": 29043 + }, + { + "epoch": 0.93, + "learning_rate": 2.6440022603636094e-07, + "loss": 0.9614, + "step": 29044 + }, + { + "epoch": 0.93, + "learning_rate": 2.641636680217796e-07, + "loss": 0.9419, + "step": 29045 + }, + { + "epoch": 0.93, + "learning_rate": 2.639272144619465e-07, + "loss": 0.8794, + "step": 29046 + }, + { + "epoch": 0.93, + "learning_rate": 2.636908653593995e-07, + "loss": 0.8706, + "step": 29047 + }, + { + "epoch": 0.93, + "learning_rate": 2.6345462071667216e-07, + "loss": 0.8691, + "step": 29048 + }, + { + "epoch": 0.93, + "learning_rate": 2.632184805363025e-07, + "loss": 0.7983, + "step": 29049 + }, + { + "epoch": 0.93, + "learning_rate": 2.629824448208207e-07, + "loss": 0.894, + "step": 29050 + }, + { + "epoch": 0.93, + "learning_rate": 2.6274651357276246e-07, + "loss": 0.9819, + "step": 29051 + }, + { + "epoch": 0.93, + "learning_rate": 2.6251068679465697e-07, + "loss": 0.9351, + "step": 29052 + }, + { + "epoch": 0.93, + "learning_rate": 2.6227496448903435e-07, + "loss": 1.0581, + "step": 29053 + }, + { + "epoch": 0.93, + "learning_rate": 2.620393466584259e-07, + "loss": 0.9971, + "step": 29054 + }, + { + "epoch": 0.93, + "learning_rate": 2.618038333053563e-07, + "loss": 0.9507, + "step": 29055 + }, + { + "epoch": 0.93, + "learning_rate": 2.6156842443235355e-07, + "loss": 0.9312, + "step": 29056 + }, + { + "epoch": 0.93, + "learning_rate": 2.6133312004194556e-07, + "loss": 0.8535, + "step": 29057 + }, + { + "epoch": 0.93, + "learning_rate": 2.6109792013665257e-07, + "loss": 0.8555, + "step": 29058 + }, + { + "epoch": 0.93, + "learning_rate": 2.608628247190015e-07, + "loss": 0.8398, + "step": 29059 + }, + { + "epoch": 0.93, + "learning_rate": 2.606278337915147e-07, + "loss": 0.96, + "step": 29060 + }, + { + "epoch": 0.93, + "learning_rate": 2.6039294735671015e-07, + "loss": 0.8516, + "step": 29061 + }, + { + "epoch": 0.93, + "learning_rate": 2.6015816541711146e-07, + "loss": 0.9019, + "step": 29062 + }, + { + "epoch": 0.93, + "learning_rate": 2.599234879752366e-07, + "loss": 0.9829, + "step": 29063 + }, + { + "epoch": 0.93, + "learning_rate": 2.596889150336024e-07, + "loss": 0.9673, + "step": 29064 + }, + { + "epoch": 0.93, + "learning_rate": 2.5945444659472464e-07, + "loss": 1.0332, + "step": 29065 + }, + { + "epoch": 0.93, + "learning_rate": 2.5922008266112133e-07, + "loss": 0.4946, + "step": 29066 + }, + { + "epoch": 0.93, + "learning_rate": 2.5898582323530597e-07, + "loss": 0.9067, + "step": 29067 + }, + { + "epoch": 0.93, + "learning_rate": 2.5875166831979213e-07, + "loss": 0.9448, + "step": 29068 + }, + { + "epoch": 0.93, + "learning_rate": 2.585176179170923e-07, + "loss": 1.0181, + "step": 29069 + }, + { + "epoch": 0.93, + "learning_rate": 2.582836720297166e-07, + "loss": 0.9429, + "step": 29070 + }, + { + "epoch": 0.93, + "learning_rate": 2.580498306601753e-07, + "loss": 0.489, + "step": 29071 + }, + { + "epoch": 0.93, + "learning_rate": 2.578160938109775e-07, + "loss": 0.9155, + "step": 29072 + }, + { + "epoch": 0.93, + "learning_rate": 2.5758246148463005e-07, + "loss": 0.9624, + "step": 29073 + }, + { + "epoch": 0.93, + "learning_rate": 2.573489336836421e-07, + "loss": 0.8945, + "step": 29074 + }, + { + "epoch": 0.93, + "learning_rate": 2.571155104105161e-07, + "loss": 0.9502, + "step": 29075 + }, + { + "epoch": 0.93, + "learning_rate": 2.5688219166775886e-07, + "loss": 0.8687, + "step": 29076 + }, + { + "epoch": 0.93, + "learning_rate": 2.5664897745787177e-07, + "loss": 0.9932, + "step": 29077 + }, + { + "epoch": 0.93, + "learning_rate": 2.5641586778335836e-07, + "loss": 0.957, + "step": 29078 + }, + { + "epoch": 0.93, + "learning_rate": 2.561828626467189e-07, + "loss": 0.9731, + "step": 29079 + }, + { + "epoch": 0.93, + "learning_rate": 2.5594996205045355e-07, + "loss": 0.876, + "step": 29080 + }, + { + "epoch": 0.93, + "learning_rate": 2.557171659970592e-07, + "loss": 0.9336, + "step": 29081 + }, + { + "epoch": 0.93, + "learning_rate": 2.5548447448903726e-07, + "loss": 0.9258, + "step": 29082 + }, + { + "epoch": 0.93, + "learning_rate": 2.552518875288812e-07, + "loss": 0.9429, + "step": 29083 + }, + { + "epoch": 0.93, + "learning_rate": 2.5501940511908905e-07, + "loss": 0.978, + "step": 29084 + }, + { + "epoch": 0.93, + "learning_rate": 2.547870272621511e-07, + "loss": 0.9072, + "step": 29085 + }, + { + "epoch": 0.93, + "learning_rate": 2.5455475396056527e-07, + "loss": 0.9897, + "step": 29086 + }, + { + "epoch": 0.93, + "learning_rate": 2.5432258521681854e-07, + "loss": 0.9541, + "step": 29087 + }, + { + "epoch": 0.93, + "learning_rate": 2.540905210334066e-07, + "loss": 0.8579, + "step": 29088 + }, + { + "epoch": 0.93, + "learning_rate": 2.5385856141281526e-07, + "loss": 0.9033, + "step": 29089 + }, + { + "epoch": 0.93, + "learning_rate": 2.5362670635753704e-07, + "loss": 0.876, + "step": 29090 + }, + { + "epoch": 0.93, + "learning_rate": 2.5339495587005657e-07, + "loss": 0.9502, + "step": 29091 + }, + { + "epoch": 0.93, + "learning_rate": 2.531633099528619e-07, + "loss": 0.8545, + "step": 29092 + }, + { + "epoch": 0.93, + "learning_rate": 2.529317686084365e-07, + "loss": 0.4287, + "step": 29093 + }, + { + "epoch": 0.93, + "learning_rate": 2.5270033183926737e-07, + "loss": 0.792, + "step": 29094 + }, + { + "epoch": 0.93, + "learning_rate": 2.5246899964783357e-07, + "loss": 0.8857, + "step": 29095 + }, + { + "epoch": 0.93, + "learning_rate": 2.5223777203662095e-07, + "loss": 1.0918, + "step": 29096 + }, + { + "epoch": 0.93, + "learning_rate": 2.5200664900810744e-07, + "loss": 0.9385, + "step": 29097 + }, + { + "epoch": 0.93, + "learning_rate": 2.517756305647756e-07, + "loss": 0.98, + "step": 29098 + }, + { + "epoch": 0.93, + "learning_rate": 2.515447167091023e-07, + "loss": 0.8862, + "step": 29099 + }, + { + "epoch": 0.93, + "learning_rate": 2.5131390744356556e-07, + "loss": 0.8032, + "step": 29100 + }, + { + "epoch": 0.93, + "learning_rate": 2.510832027706389e-07, + "loss": 0.9609, + "step": 29101 + }, + { + "epoch": 0.93, + "learning_rate": 2.5085260269280266e-07, + "loss": 0.9404, + "step": 29102 + }, + { + "epoch": 0.93, + "learning_rate": 2.50622107212527e-07, + "loss": 1.0688, + "step": 29103 + }, + { + "epoch": 0.93, + "learning_rate": 2.5039171633228556e-07, + "loss": 0.9507, + "step": 29104 + }, + { + "epoch": 0.93, + "learning_rate": 2.5016143005455073e-07, + "loss": 0.8862, + "step": 29105 + }, + { + "epoch": 0.93, + "learning_rate": 2.4993124838179284e-07, + "loss": 0.8621, + "step": 29106 + }, + { + "epoch": 0.93, + "learning_rate": 2.497011713164832e-07, + "loss": 0.979, + "step": 29107 + }, + { + "epoch": 0.93, + "learning_rate": 2.4947119886108875e-07, + "loss": 0.8315, + "step": 29108 + }, + { + "epoch": 0.93, + "learning_rate": 2.4924133101807636e-07, + "loss": 0.8696, + "step": 29109 + }, + { + "epoch": 0.93, + "learning_rate": 2.4901156778991296e-07, + "loss": 0.895, + "step": 29110 + }, + { + "epoch": 0.93, + "learning_rate": 2.487819091790633e-07, + "loss": 0.8774, + "step": 29111 + }, + { + "epoch": 0.93, + "learning_rate": 2.4855235518799203e-07, + "loss": 0.8193, + "step": 29112 + }, + { + "epoch": 0.93, + "learning_rate": 2.4832290581916164e-07, + "loss": 0.8896, + "step": 29113 + }, + { + "epoch": 0.93, + "learning_rate": 2.4809356107503457e-07, + "loss": 0.9321, + "step": 29114 + }, + { + "epoch": 0.93, + "learning_rate": 2.4786432095807e-07, + "loss": 0.6899, + "step": 29115 + }, + { + "epoch": 0.93, + "learning_rate": 2.476351854707271e-07, + "loss": 0.9639, + "step": 29116 + }, + { + "epoch": 0.93, + "learning_rate": 2.4740615461546604e-07, + "loss": 0.8877, + "step": 29117 + }, + { + "epoch": 0.93, + "learning_rate": 2.471772283947438e-07, + "loss": 0.8638, + "step": 29118 + }, + { + "epoch": 0.93, + "learning_rate": 2.4694840681101505e-07, + "loss": 0.8398, + "step": 29119 + }, + { + "epoch": 0.93, + "learning_rate": 2.467196898667357e-07, + "loss": 0.8743, + "step": 29120 + }, + { + "epoch": 0.93, + "learning_rate": 2.464910775643614e-07, + "loss": 0.9321, + "step": 29121 + }, + { + "epoch": 0.93, + "learning_rate": 2.4626256990634255e-07, + "loss": 0.512, + "step": 29122 + }, + { + "epoch": 0.93, + "learning_rate": 2.460341668951316e-07, + "loss": 0.835, + "step": 29123 + }, + { + "epoch": 0.93, + "learning_rate": 2.458058685331777e-07, + "loss": 0.9028, + "step": 29124 + }, + { + "epoch": 0.93, + "learning_rate": 2.4557767482293215e-07, + "loss": 1.0103, + "step": 29125 + }, + { + "epoch": 0.93, + "learning_rate": 2.453495857668409e-07, + "loss": 0.8911, + "step": 29126 + }, + { + "epoch": 0.93, + "learning_rate": 2.4512160136735517e-07, + "loss": 0.9077, + "step": 29127 + }, + { + "epoch": 0.93, + "learning_rate": 2.4489372162691647e-07, + "loss": 0.8765, + "step": 29128 + }, + { + "epoch": 0.93, + "learning_rate": 2.4466594654797393e-07, + "loss": 0.8745, + "step": 29129 + }, + { + "epoch": 0.93, + "learning_rate": 2.444382761329678e-07, + "loss": 0.7998, + "step": 29130 + }, + { + "epoch": 0.93, + "learning_rate": 2.4421071038434276e-07, + "loss": 0.7847, + "step": 29131 + }, + { + "epoch": 0.93, + "learning_rate": 2.4398324930453796e-07, + "loss": 0.9092, + "step": 29132 + }, + { + "epoch": 0.93, + "learning_rate": 2.437558928959971e-07, + "loss": 0.9424, + "step": 29133 + }, + { + "epoch": 0.93, + "learning_rate": 2.435286411611559e-07, + "loss": 0.7896, + "step": 29134 + }, + { + "epoch": 0.93, + "learning_rate": 2.43301494102457e-07, + "loss": 0.9492, + "step": 29135 + }, + { + "epoch": 0.93, + "learning_rate": 2.4307445172233267e-07, + "loss": 1.0259, + "step": 29136 + }, + { + "epoch": 0.93, + "learning_rate": 2.428475140232234e-07, + "loss": 0.979, + "step": 29137 + }, + { + "epoch": 0.93, + "learning_rate": 2.426206810075593e-07, + "loss": 0.9067, + "step": 29138 + }, + { + "epoch": 0.93, + "learning_rate": 2.423939526777774e-07, + "loss": 0.9575, + "step": 29139 + }, + { + "epoch": 0.93, + "learning_rate": 2.4216732903630913e-07, + "loss": 1.0586, + "step": 29140 + }, + { + "epoch": 0.93, + "learning_rate": 2.419408100855858e-07, + "loss": 0.8931, + "step": 29141 + }, + { + "epoch": 0.93, + "learning_rate": 2.417143958280388e-07, + "loss": 0.9072, + "step": 29142 + }, + { + "epoch": 0.93, + "learning_rate": 2.414880862660951e-07, + "loss": 1.0073, + "step": 29143 + }, + { + "epoch": 0.93, + "learning_rate": 2.41261881402185e-07, + "loss": 0.47, + "step": 29144 + }, + { + "epoch": 0.93, + "learning_rate": 2.4103578123873427e-07, + "loss": 0.937, + "step": 29145 + }, + { + "epoch": 0.93, + "learning_rate": 2.408097857781688e-07, + "loss": 0.8135, + "step": 29146 + }, + { + "epoch": 0.93, + "learning_rate": 2.4058389502291335e-07, + "loss": 0.9189, + "step": 29147 + }, + { + "epoch": 0.93, + "learning_rate": 2.4035810897539146e-07, + "loss": 0.8701, + "step": 29148 + }, + { + "epoch": 0.93, + "learning_rate": 2.401324276380257e-07, + "loss": 0.8853, + "step": 29149 + }, + { + "epoch": 0.93, + "learning_rate": 2.399068510132374e-07, + "loss": 0.8379, + "step": 29150 + }, + { + "epoch": 0.93, + "learning_rate": 2.396813791034458e-07, + "loss": 0.9712, + "step": 29151 + }, + { + "epoch": 0.93, + "learning_rate": 2.394560119110723e-07, + "loss": 0.8574, + "step": 29152 + }, + { + "epoch": 0.93, + "learning_rate": 2.3923074943853264e-07, + "loss": 1.0405, + "step": 29153 + }, + { + "epoch": 0.93, + "learning_rate": 2.39005591688245e-07, + "loss": 1.0161, + "step": 29154 + }, + { + "epoch": 0.93, + "learning_rate": 2.38780538662623e-07, + "loss": 0.9531, + "step": 29155 + }, + { + "epoch": 0.93, + "learning_rate": 2.385555903640835e-07, + "loss": 0.8887, + "step": 29156 + }, + { + "epoch": 0.93, + "learning_rate": 2.383307467950391e-07, + "loss": 0.8027, + "step": 29157 + }, + { + "epoch": 0.93, + "learning_rate": 2.3810600795790228e-07, + "loss": 0.8643, + "step": 29158 + }, + { + "epoch": 0.93, + "learning_rate": 2.3788137385508226e-07, + "loss": 0.8604, + "step": 29159 + }, + { + "epoch": 0.93, + "learning_rate": 2.3765684448899372e-07, + "loss": 0.957, + "step": 29160 + }, + { + "epoch": 0.93, + "learning_rate": 2.3743241986204146e-07, + "loss": 0.7695, + "step": 29161 + }, + { + "epoch": 0.93, + "learning_rate": 2.3720809997663463e-07, + "loss": 0.9492, + "step": 29162 + }, + { + "epoch": 0.93, + "learning_rate": 2.369838848351802e-07, + "loss": 0.9692, + "step": 29163 + }, + { + "epoch": 0.93, + "learning_rate": 2.3675977444008291e-07, + "loss": 0.9463, + "step": 29164 + }, + { + "epoch": 0.93, + "learning_rate": 2.3653576879374862e-07, + "loss": 0.8818, + "step": 29165 + }, + { + "epoch": 0.93, + "learning_rate": 2.3631186789857985e-07, + "loss": 0.856, + "step": 29166 + }, + { + "epoch": 0.93, + "learning_rate": 2.36088071756978e-07, + "loss": 0.9077, + "step": 29167 + }, + { + "epoch": 0.93, + "learning_rate": 2.3586438037134674e-07, + "loss": 1.0986, + "step": 29168 + }, + { + "epoch": 0.93, + "learning_rate": 2.356407937440819e-07, + "loss": 0.874, + "step": 29169 + }, + { + "epoch": 0.93, + "learning_rate": 2.35417311877586e-07, + "loss": 0.8169, + "step": 29170 + }, + { + "epoch": 0.93, + "learning_rate": 2.351939347742549e-07, + "loss": 0.8638, + "step": 29171 + }, + { + "epoch": 0.93, + "learning_rate": 2.349706624364867e-07, + "loss": 0.8447, + "step": 29172 + }, + { + "epoch": 0.93, + "learning_rate": 2.3474749486667392e-07, + "loss": 0.8892, + "step": 29173 + }, + { + "epoch": 0.93, + "learning_rate": 2.3452443206721465e-07, + "loss": 0.9448, + "step": 29174 + }, + { + "epoch": 0.93, + "learning_rate": 2.343014740405003e-07, + "loss": 0.9399, + "step": 29175 + }, + { + "epoch": 0.93, + "learning_rate": 2.340786207889234e-07, + "loss": 0.9282, + "step": 29176 + }, + { + "epoch": 0.93, + "learning_rate": 2.3385587231487315e-07, + "loss": 0.9536, + "step": 29177 + }, + { + "epoch": 0.93, + "learning_rate": 2.33633228620741e-07, + "loss": 0.9937, + "step": 29178 + }, + { + "epoch": 0.93, + "learning_rate": 2.33410689708915e-07, + "loss": 0.8667, + "step": 29179 + }, + { + "epoch": 0.93, + "learning_rate": 2.3318825558178438e-07, + "loss": 0.8257, + "step": 29180 + }, + { + "epoch": 0.93, + "learning_rate": 2.3296592624173496e-07, + "loss": 0.8594, + "step": 29181 + }, + { + "epoch": 0.93, + "learning_rate": 2.3274370169115047e-07, + "loss": 0.8818, + "step": 29182 + }, + { + "epoch": 0.93, + "learning_rate": 2.3252158193241557e-07, + "loss": 0.8906, + "step": 29183 + }, + { + "epoch": 0.93, + "learning_rate": 2.322995669679151e-07, + "loss": 0.9985, + "step": 29184 + }, + { + "epoch": 0.93, + "learning_rate": 2.3207765680002936e-07, + "loss": 0.9536, + "step": 29185 + }, + { + "epoch": 0.93, + "learning_rate": 2.3185585143113977e-07, + "loss": 0.4678, + "step": 29186 + }, + { + "epoch": 0.93, + "learning_rate": 2.3163415086362662e-07, + "loss": 0.8423, + "step": 29187 + }, + { + "epoch": 0.93, + "learning_rate": 2.3141255509986694e-07, + "loss": 1.0239, + "step": 29188 + }, + { + "epoch": 0.93, + "learning_rate": 2.311910641422399e-07, + "loss": 0.9536, + "step": 29189 + }, + { + "epoch": 0.93, + "learning_rate": 2.309696779931214e-07, + "loss": 0.9473, + "step": 29190 + }, + { + "epoch": 0.93, + "learning_rate": 2.3074839665488514e-07, + "loss": 0.9883, + "step": 29191 + }, + { + "epoch": 0.93, + "learning_rate": 2.3052722012990803e-07, + "loss": 0.9424, + "step": 29192 + }, + { + "epoch": 0.93, + "learning_rate": 2.3030614842056153e-07, + "loss": 0.9629, + "step": 29193 + }, + { + "epoch": 0.93, + "learning_rate": 2.30085181529216e-07, + "loss": 0.8469, + "step": 29194 + }, + { + "epoch": 0.93, + "learning_rate": 2.2986431945824505e-07, + "loss": 1.0439, + "step": 29195 + }, + { + "epoch": 0.93, + "learning_rate": 2.2964356221001572e-07, + "loss": 0.8823, + "step": 29196 + }, + { + "epoch": 0.93, + "learning_rate": 2.294229097868994e-07, + "loss": 0.8652, + "step": 29197 + }, + { + "epoch": 0.93, + "learning_rate": 2.292023621912598e-07, + "loss": 0.9805, + "step": 29198 + }, + { + "epoch": 0.93, + "learning_rate": 2.289819194254661e-07, + "loss": 0.8149, + "step": 29199 + }, + { + "epoch": 0.93, + "learning_rate": 2.28761581491882e-07, + "loss": 0.9209, + "step": 29200 + }, + { + "epoch": 0.93, + "learning_rate": 2.2854134839287222e-07, + "loss": 0.9355, + "step": 29201 + }, + { + "epoch": 0.93, + "learning_rate": 2.2832122013079827e-07, + "loss": 0.9302, + "step": 29202 + }, + { + "epoch": 0.93, + "learning_rate": 2.281011967080249e-07, + "loss": 0.9863, + "step": 29203 + }, + { + "epoch": 0.93, + "learning_rate": 2.2788127812690796e-07, + "loss": 0.4993, + "step": 29204 + }, + { + "epoch": 0.93, + "learning_rate": 2.276614643898134e-07, + "loss": 0.9863, + "step": 29205 + }, + { + "epoch": 0.93, + "learning_rate": 2.2744175549909263e-07, + "loss": 0.8135, + "step": 29206 + }, + { + "epoch": 0.93, + "learning_rate": 2.2722215145710714e-07, + "loss": 0.9604, + "step": 29207 + }, + { + "epoch": 0.93, + "learning_rate": 2.2700265226621165e-07, + "loss": 0.916, + "step": 29208 + }, + { + "epoch": 0.93, + "learning_rate": 2.267832579287621e-07, + "loss": 0.7871, + "step": 29209 + }, + { + "epoch": 0.93, + "learning_rate": 2.2656396844711103e-07, + "loss": 0.9023, + "step": 29210 + }, + { + "epoch": 0.93, + "learning_rate": 2.2634478382361213e-07, + "loss": 1.0181, + "step": 29211 + }, + { + "epoch": 0.93, + "learning_rate": 2.2612570406061684e-07, + "loss": 0.8994, + "step": 29212 + }, + { + "epoch": 0.93, + "learning_rate": 2.2590672916047662e-07, + "loss": 0.8916, + "step": 29213 + }, + { + "epoch": 0.93, + "learning_rate": 2.2568785912553848e-07, + "loss": 0.9116, + "step": 29214 + }, + { + "epoch": 0.93, + "learning_rate": 2.2546909395815164e-07, + "loss": 0.9043, + "step": 29215 + }, + { + "epoch": 0.93, + "learning_rate": 2.2525043366066422e-07, + "loss": 0.9426, + "step": 29216 + }, + { + "epoch": 0.93, + "learning_rate": 2.2503187823542106e-07, + "loss": 0.832, + "step": 29217 + }, + { + "epoch": 0.93, + "learning_rate": 2.24813427684768e-07, + "loss": 0.9473, + "step": 29218 + }, + { + "epoch": 0.93, + "learning_rate": 2.2459508201104763e-07, + "loss": 0.9858, + "step": 29219 + }, + { + "epoch": 0.93, + "learning_rate": 2.2437684121660364e-07, + "loss": 0.9541, + "step": 29220 + }, + { + "epoch": 0.93, + "learning_rate": 2.241587053037775e-07, + "loss": 0.9878, + "step": 29221 + }, + { + "epoch": 0.93, + "learning_rate": 2.2394067427490728e-07, + "loss": 0.9756, + "step": 29222 + }, + { + "epoch": 0.93, + "learning_rate": 2.2372274813233562e-07, + "loss": 0.939, + "step": 29223 + }, + { + "epoch": 0.93, + "learning_rate": 2.2350492687839843e-07, + "loss": 0.8638, + "step": 29224 + }, + { + "epoch": 0.93, + "learning_rate": 2.232872105154338e-07, + "loss": 0.8691, + "step": 29225 + }, + { + "epoch": 0.93, + "learning_rate": 2.2306959904577652e-07, + "loss": 0.9062, + "step": 29226 + }, + { + "epoch": 0.93, + "learning_rate": 2.2285209247176144e-07, + "loss": 0.8989, + "step": 29227 + }, + { + "epoch": 0.93, + "learning_rate": 2.2263469079572442e-07, + "loss": 0.874, + "step": 29228 + }, + { + "epoch": 0.93, + "learning_rate": 2.2241739401999474e-07, + "loss": 0.9282, + "step": 29229 + }, + { + "epoch": 0.93, + "learning_rate": 2.2220020214690496e-07, + "loss": 0.7432, + "step": 29230 + }, + { + "epoch": 0.93, + "learning_rate": 2.2198311517878544e-07, + "loss": 0.7673, + "step": 29231 + }, + { + "epoch": 0.93, + "learning_rate": 2.2176613311796658e-07, + "loss": 0.8735, + "step": 29232 + }, + { + "epoch": 0.93, + "learning_rate": 2.2154925596677312e-07, + "loss": 0.9536, + "step": 29233 + }, + { + "epoch": 0.93, + "learning_rate": 2.2133248372753546e-07, + "loss": 0.7949, + "step": 29234 + }, + { + "epoch": 0.93, + "learning_rate": 2.211158164025784e-07, + "loss": 0.8379, + "step": 29235 + }, + { + "epoch": 0.93, + "learning_rate": 2.2089925399422453e-07, + "loss": 0.9155, + "step": 29236 + }, + { + "epoch": 0.94, + "learning_rate": 2.2068279650479862e-07, + "loss": 0.9824, + "step": 29237 + }, + { + "epoch": 0.94, + "learning_rate": 2.204664439366233e-07, + "loss": 0.9688, + "step": 29238 + }, + { + "epoch": 0.94, + "learning_rate": 2.2025019629201895e-07, + "loss": 0.9312, + "step": 29239 + }, + { + "epoch": 0.94, + "learning_rate": 2.2003405357330809e-07, + "loss": 0.8701, + "step": 29240 + }, + { + "epoch": 0.94, + "learning_rate": 2.198180157828056e-07, + "loss": 0.916, + "step": 29241 + }, + { + "epoch": 0.94, + "learning_rate": 2.1960208292283402e-07, + "loss": 0.8716, + "step": 29242 + }, + { + "epoch": 0.94, + "learning_rate": 2.19386254995706e-07, + "loss": 0.9941, + "step": 29243 + }, + { + "epoch": 0.94, + "learning_rate": 2.1917053200374073e-07, + "loss": 0.8081, + "step": 29244 + }, + { + "epoch": 0.94, + "learning_rate": 2.1895491394924861e-07, + "loss": 0.895, + "step": 29245 + }, + { + "epoch": 0.94, + "learning_rate": 2.187394008345456e-07, + "loss": 0.9419, + "step": 29246 + }, + { + "epoch": 0.94, + "learning_rate": 2.1852399266194312e-07, + "loss": 0.9619, + "step": 29247 + }, + { + "epoch": 0.94, + "learning_rate": 2.1830868943375383e-07, + "loss": 0.9961, + "step": 29248 + }, + { + "epoch": 0.94, + "learning_rate": 2.1809349115228583e-07, + "loss": 0.7627, + "step": 29249 + }, + { + "epoch": 0.94, + "learning_rate": 2.1787839781984954e-07, + "loss": 1.0137, + "step": 29250 + }, + { + "epoch": 0.94, + "learning_rate": 2.1766340943875087e-07, + "loss": 1.0747, + "step": 29251 + }, + { + "epoch": 0.94, + "learning_rate": 2.1744852601129686e-07, + "loss": 0.9072, + "step": 29252 + }, + { + "epoch": 0.94, + "learning_rate": 2.1723374753979343e-07, + "loss": 0.9429, + "step": 29253 + }, + { + "epoch": 0.94, + "learning_rate": 2.1701907402654543e-07, + "loss": 0.9243, + "step": 29254 + }, + { + "epoch": 0.94, + "learning_rate": 2.1680450547385434e-07, + "loss": 0.8174, + "step": 29255 + }, + { + "epoch": 0.94, + "learning_rate": 2.16590041884025e-07, + "loss": 0.9956, + "step": 29256 + }, + { + "epoch": 0.94, + "learning_rate": 2.1637568325935555e-07, + "loss": 0.9087, + "step": 29257 + }, + { + "epoch": 0.94, + "learning_rate": 2.161614296021486e-07, + "loss": 0.8633, + "step": 29258 + }, + { + "epoch": 0.94, + "learning_rate": 2.1594728091470008e-07, + "loss": 0.9033, + "step": 29259 + }, + { + "epoch": 0.94, + "learning_rate": 2.1573323719930926e-07, + "loss": 0.8574, + "step": 29260 + }, + { + "epoch": 0.94, + "learning_rate": 2.15519298458271e-07, + "loss": 0.9312, + "step": 29261 + }, + { + "epoch": 0.94, + "learning_rate": 2.1530546469388346e-07, + "loss": 1.0234, + "step": 29262 + }, + { + "epoch": 0.94, + "learning_rate": 2.15091735908437e-07, + "loss": 0.9658, + "step": 29263 + }, + { + "epoch": 0.94, + "learning_rate": 2.148781121042287e-07, + "loss": 0.8442, + "step": 29264 + }, + { + "epoch": 0.94, + "learning_rate": 2.1466459328354783e-07, + "loss": 0.9868, + "step": 29265 + }, + { + "epoch": 0.94, + "learning_rate": 2.14451179448687e-07, + "loss": 0.8472, + "step": 29266 + }, + { + "epoch": 0.94, + "learning_rate": 2.1423787060193324e-07, + "loss": 1.0425, + "step": 29267 + }, + { + "epoch": 0.94, + "learning_rate": 2.1402466674557809e-07, + "loss": 0.9697, + "step": 29268 + }, + { + "epoch": 0.94, + "learning_rate": 2.1381156788190638e-07, + "loss": 0.8267, + "step": 29269 + }, + { + "epoch": 0.94, + "learning_rate": 2.1359857401320738e-07, + "loss": 0.9854, + "step": 29270 + }, + { + "epoch": 0.94, + "learning_rate": 2.1338568514176372e-07, + "loss": 0.9277, + "step": 29271 + }, + { + "epoch": 0.94, + "learning_rate": 2.1317290126986023e-07, + "loss": 0.9426, + "step": 29272 + }, + { + "epoch": 0.94, + "learning_rate": 2.1296022239977955e-07, + "loss": 0.8228, + "step": 29273 + }, + { + "epoch": 0.94, + "learning_rate": 2.1274764853380537e-07, + "loss": 0.8833, + "step": 29274 + }, + { + "epoch": 0.94, + "learning_rate": 2.125351796742159e-07, + "loss": 0.9673, + "step": 29275 + }, + { + "epoch": 0.94, + "learning_rate": 2.1232281582329152e-07, + "loss": 0.9102, + "step": 29276 + }, + { + "epoch": 0.94, + "learning_rate": 2.121105569833115e-07, + "loss": 0.8931, + "step": 29277 + }, + { + "epoch": 0.94, + "learning_rate": 2.1189840315655185e-07, + "loss": 0.8188, + "step": 29278 + }, + { + "epoch": 0.94, + "learning_rate": 2.1168635434529073e-07, + "loss": 0.9146, + "step": 29279 + }, + { + "epoch": 0.94, + "learning_rate": 2.1147441055180074e-07, + "loss": 0.7935, + "step": 29280 + }, + { + "epoch": 0.94, + "learning_rate": 2.1126257177835785e-07, + "loss": 0.9058, + "step": 29281 + }, + { + "epoch": 0.94, + "learning_rate": 2.1105083802723248e-07, + "loss": 0.8677, + "step": 29282 + }, + { + "epoch": 0.94, + "learning_rate": 2.1083920930069835e-07, + "loss": 0.9453, + "step": 29283 + }, + { + "epoch": 0.94, + "learning_rate": 2.1062768560102585e-07, + "loss": 0.8198, + "step": 29284 + }, + { + "epoch": 0.94, + "learning_rate": 2.1041626693048322e-07, + "loss": 0.8296, + "step": 29285 + }, + { + "epoch": 0.94, + "learning_rate": 2.1020495329133972e-07, + "loss": 0.9497, + "step": 29286 + }, + { + "epoch": 0.94, + "learning_rate": 2.0999374468586242e-07, + "loss": 0.9326, + "step": 29287 + }, + { + "epoch": 0.94, + "learning_rate": 2.097826411163173e-07, + "loss": 0.8599, + "step": 29288 + }, + { + "epoch": 0.94, + "learning_rate": 2.0957164258497031e-07, + "loss": 0.9424, + "step": 29289 + }, + { + "epoch": 0.94, + "learning_rate": 2.093607490940819e-07, + "loss": 0.9363, + "step": 29290 + }, + { + "epoch": 0.94, + "learning_rate": 2.0914996064591797e-07, + "loss": 0.8628, + "step": 29291 + }, + { + "epoch": 0.94, + "learning_rate": 2.0893927724273787e-07, + "loss": 0.8706, + "step": 29292 + }, + { + "epoch": 0.94, + "learning_rate": 2.0872869888680536e-07, + "loss": 0.8599, + "step": 29293 + }, + { + "epoch": 0.94, + "learning_rate": 2.0851822558037526e-07, + "loss": 0.8643, + "step": 29294 + }, + { + "epoch": 0.94, + "learning_rate": 2.0830785732570913e-07, + "loss": 0.9229, + "step": 29295 + }, + { + "epoch": 0.94, + "learning_rate": 2.0809759412506293e-07, + "loss": 0.9512, + "step": 29296 + }, + { + "epoch": 0.94, + "learning_rate": 2.078874359806926e-07, + "loss": 0.8823, + "step": 29297 + }, + { + "epoch": 0.94, + "learning_rate": 2.0767738289485084e-07, + "loss": 0.9165, + "step": 29298 + }, + { + "epoch": 0.94, + "learning_rate": 2.074674348697947e-07, + "loss": 0.8823, + "step": 29299 + }, + { + "epoch": 0.94, + "learning_rate": 2.0725759190777462e-07, + "loss": 1.0073, + "step": 29300 + }, + { + "epoch": 0.94, + "learning_rate": 2.070478540110432e-07, + "loss": 0.8286, + "step": 29301 + }, + { + "epoch": 0.94, + "learning_rate": 2.068382211818487e-07, + "loss": 0.9031, + "step": 29302 + }, + { + "epoch": 0.94, + "learning_rate": 2.0662869342244375e-07, + "loss": 0.8677, + "step": 29303 + }, + { + "epoch": 0.94, + "learning_rate": 2.0641927073507207e-07, + "loss": 0.8838, + "step": 29304 + }, + { + "epoch": 0.94, + "learning_rate": 2.062099531219841e-07, + "loss": 0.8828, + "step": 29305 + }, + { + "epoch": 0.94, + "learning_rate": 2.060007405854225e-07, + "loss": 1.042, + "step": 29306 + }, + { + "epoch": 0.94, + "learning_rate": 2.0579163312763552e-07, + "loss": 0.8711, + "step": 29307 + }, + { + "epoch": 0.94, + "learning_rate": 2.0558263075086236e-07, + "loss": 0.8721, + "step": 29308 + }, + { + "epoch": 0.94, + "learning_rate": 2.0537373345734912e-07, + "loss": 0.895, + "step": 29309 + }, + { + "epoch": 0.94, + "learning_rate": 2.0516494124933616e-07, + "loss": 0.8652, + "step": 29310 + }, + { + "epoch": 0.94, + "learning_rate": 2.0495625412906172e-07, + "loss": 0.9463, + "step": 29311 + }, + { + "epoch": 0.94, + "learning_rate": 2.0474767209876621e-07, + "loss": 0.8911, + "step": 29312 + }, + { + "epoch": 0.94, + "learning_rate": 2.0453919516068786e-07, + "loss": 0.8501, + "step": 29313 + }, + { + "epoch": 0.94, + "learning_rate": 2.0433082331706156e-07, + "loss": 0.9233, + "step": 29314 + }, + { + "epoch": 0.94, + "learning_rate": 2.0412255657012547e-07, + "loss": 1.0229, + "step": 29315 + }, + { + "epoch": 0.94, + "learning_rate": 2.039143949221134e-07, + "loss": 1.0122, + "step": 29316 + }, + { + "epoch": 0.94, + "learning_rate": 2.0370633837525579e-07, + "loss": 0.8872, + "step": 29317 + }, + { + "epoch": 0.94, + "learning_rate": 2.034983869317897e-07, + "loss": 0.895, + "step": 29318 + }, + { + "epoch": 0.94, + "learning_rate": 2.0329054059394227e-07, + "loss": 0.9717, + "step": 29319 + }, + { + "epoch": 0.94, + "learning_rate": 2.0308279936394616e-07, + "loss": 0.9912, + "step": 29320 + }, + { + "epoch": 0.94, + "learning_rate": 2.0287516324402734e-07, + "loss": 0.875, + "step": 29321 + }, + { + "epoch": 0.94, + "learning_rate": 2.0266763223641517e-07, + "loss": 0.936, + "step": 29322 + }, + { + "epoch": 0.94, + "learning_rate": 2.0246020634333674e-07, + "loss": 0.9819, + "step": 29323 + }, + { + "epoch": 0.94, + "learning_rate": 2.0225288556701694e-07, + "loss": 0.8335, + "step": 29324 + }, + { + "epoch": 0.94, + "learning_rate": 2.0204566990967956e-07, + "loss": 1.0039, + "step": 29325 + }, + { + "epoch": 0.94, + "learning_rate": 2.0183855937354947e-07, + "loss": 0.8691, + "step": 29326 + }, + { + "epoch": 0.94, + "learning_rate": 2.01631553960846e-07, + "loss": 0.9111, + "step": 29327 + }, + { + "epoch": 0.94, + "learning_rate": 2.0142465367379293e-07, + "loss": 0.8135, + "step": 29328 + }, + { + "epoch": 0.94, + "learning_rate": 2.012178585146085e-07, + "loss": 0.9072, + "step": 29329 + }, + { + "epoch": 0.94, + "learning_rate": 2.0101116848551206e-07, + "loss": 0.9141, + "step": 29330 + }, + { + "epoch": 0.94, + "learning_rate": 2.0080458358871958e-07, + "loss": 0.9336, + "step": 29331 + }, + { + "epoch": 0.94, + "learning_rate": 2.005981038264493e-07, + "loss": 0.8652, + "step": 29332 + }, + { + "epoch": 0.94, + "learning_rate": 2.003917292009161e-07, + "loss": 1.0083, + "step": 29333 + }, + { + "epoch": 0.94, + "learning_rate": 2.001854597143349e-07, + "loss": 0.8967, + "step": 29334 + }, + { + "epoch": 0.94, + "learning_rate": 1.9997929536891613e-07, + "loss": 0.8462, + "step": 29335 + }, + { + "epoch": 0.94, + "learning_rate": 1.997732361668747e-07, + "loss": 0.8628, + "step": 29336 + }, + { + "epoch": 0.94, + "learning_rate": 1.9956728211041886e-07, + "loss": 0.9067, + "step": 29337 + }, + { + "epoch": 0.94, + "learning_rate": 1.9936143320176017e-07, + "loss": 0.8569, + "step": 29338 + }, + { + "epoch": 0.94, + "learning_rate": 1.991556894431057e-07, + "loss": 0.9741, + "step": 29339 + }, + { + "epoch": 0.94, + "learning_rate": 1.9895005083666485e-07, + "loss": 0.916, + "step": 29340 + }, + { + "epoch": 0.94, + "learning_rate": 1.987445173846425e-07, + "loss": 0.9512, + "step": 29341 + }, + { + "epoch": 0.94, + "learning_rate": 1.9853908908924358e-07, + "loss": 0.8267, + "step": 29342 + }, + { + "epoch": 0.94, + "learning_rate": 1.9833376595267294e-07, + "loss": 0.9258, + "step": 29343 + }, + { + "epoch": 0.94, + "learning_rate": 1.9812854797713222e-07, + "loss": 0.8857, + "step": 29344 + }, + { + "epoch": 0.94, + "learning_rate": 1.9792343516482404e-07, + "loss": 0.9053, + "step": 29345 + }, + { + "epoch": 0.94, + "learning_rate": 1.9771842751795002e-07, + "loss": 0.8672, + "step": 29346 + }, + { + "epoch": 0.94, + "learning_rate": 1.9751352503870835e-07, + "loss": 0.9102, + "step": 29347 + }, + { + "epoch": 0.94, + "learning_rate": 1.9730872772929844e-07, + "loss": 0.8877, + "step": 29348 + }, + { + "epoch": 0.94, + "learning_rate": 1.9710403559191516e-07, + "loss": 0.9536, + "step": 29349 + }, + { + "epoch": 0.94, + "learning_rate": 1.9689944862875677e-07, + "loss": 0.8857, + "step": 29350 + }, + { + "epoch": 0.94, + "learning_rate": 1.9669496684201815e-07, + "loss": 0.9951, + "step": 29351 + }, + { + "epoch": 0.94, + "learning_rate": 1.9649059023389206e-07, + "loss": 0.917, + "step": 29352 + }, + { + "epoch": 0.94, + "learning_rate": 1.9628631880657223e-07, + "loss": 0.8384, + "step": 29353 + }, + { + "epoch": 0.94, + "learning_rate": 1.9608215256225027e-07, + "loss": 0.8789, + "step": 29354 + }, + { + "epoch": 0.94, + "learning_rate": 1.9587809150311665e-07, + "loss": 0.8975, + "step": 29355 + }, + { + "epoch": 0.94, + "learning_rate": 1.9567413563136073e-07, + "loss": 0.9995, + "step": 29356 + }, + { + "epoch": 0.94, + "learning_rate": 1.9547028494916965e-07, + "loss": 0.9722, + "step": 29357 + }, + { + "epoch": 0.94, + "learning_rate": 1.9526653945873165e-07, + "loss": 0.7954, + "step": 29358 + }, + { + "epoch": 0.94, + "learning_rate": 1.9506289916223276e-07, + "loss": 0.4688, + "step": 29359 + }, + { + "epoch": 0.94, + "learning_rate": 1.9485936406185567e-07, + "loss": 0.9736, + "step": 29360 + }, + { + "epoch": 0.94, + "learning_rate": 1.9465593415978757e-07, + "loss": 0.9243, + "step": 29361 + }, + { + "epoch": 0.94, + "learning_rate": 1.9445260945820775e-07, + "loss": 0.915, + "step": 29362 + }, + { + "epoch": 0.94, + "learning_rate": 1.942493899593012e-07, + "loss": 0.9634, + "step": 29363 + }, + { + "epoch": 0.94, + "learning_rate": 1.9404627566524502e-07, + "loss": 0.9028, + "step": 29364 + }, + { + "epoch": 0.94, + "learning_rate": 1.9384326657821972e-07, + "loss": 0.8423, + "step": 29365 + }, + { + "epoch": 0.94, + "learning_rate": 1.9364036270040243e-07, + "loss": 0.9551, + "step": 29366 + }, + { + "epoch": 0.94, + "learning_rate": 1.934375640339725e-07, + "loss": 0.918, + "step": 29367 + }, + { + "epoch": 0.94, + "learning_rate": 1.932348705811027e-07, + "loss": 0.9844, + "step": 29368 + }, + { + "epoch": 0.94, + "learning_rate": 1.930322823439701e-07, + "loss": 0.96, + "step": 29369 + }, + { + "epoch": 0.94, + "learning_rate": 1.9282979932474632e-07, + "loss": 0.8984, + "step": 29370 + }, + { + "epoch": 0.94, + "learning_rate": 1.9262742152560744e-07, + "loss": 0.8223, + "step": 29371 + }, + { + "epoch": 0.94, + "learning_rate": 1.9242514894871945e-07, + "loss": 0.9072, + "step": 29372 + }, + { + "epoch": 0.94, + "learning_rate": 1.9222298159625508e-07, + "loss": 0.9453, + "step": 29373 + }, + { + "epoch": 0.94, + "learning_rate": 1.920209194703837e-07, + "loss": 0.9116, + "step": 29374 + }, + { + "epoch": 0.94, + "learning_rate": 1.918189625732736e-07, + "loss": 1.0742, + "step": 29375 + }, + { + "epoch": 0.94, + "learning_rate": 1.9161711090709078e-07, + "loss": 0.9111, + "step": 29376 + }, + { + "epoch": 0.94, + "learning_rate": 1.9141536447400021e-07, + "loss": 0.8977, + "step": 29377 + }, + { + "epoch": 0.94, + "learning_rate": 1.9121372327616793e-07, + "loss": 0.9351, + "step": 29378 + }, + { + "epoch": 0.94, + "learning_rate": 1.9101218731575777e-07, + "loss": 0.9243, + "step": 29379 + }, + { + "epoch": 0.94, + "learning_rate": 1.9081075659492798e-07, + "loss": 0.9453, + "step": 29380 + }, + { + "epoch": 0.94, + "learning_rate": 1.9060943111584463e-07, + "loss": 0.8989, + "step": 29381 + }, + { + "epoch": 0.94, + "learning_rate": 1.9040821088066374e-07, + "loss": 0.813, + "step": 29382 + }, + { + "epoch": 0.94, + "learning_rate": 1.9020709589154805e-07, + "loss": 0.9238, + "step": 29383 + }, + { + "epoch": 0.94, + "learning_rate": 1.900060861506514e-07, + "loss": 0.9141, + "step": 29384 + }, + { + "epoch": 0.94, + "learning_rate": 1.8980518166013427e-07, + "loss": 0.9512, + "step": 29385 + }, + { + "epoch": 0.94, + "learning_rate": 1.896043824221494e-07, + "loss": 0.9072, + "step": 29386 + }, + { + "epoch": 0.94, + "learning_rate": 1.8940368843885171e-07, + "loss": 1.0586, + "step": 29387 + }, + { + "epoch": 0.94, + "learning_rate": 1.8920309971239393e-07, + "loss": 1.0107, + "step": 29388 + }, + { + "epoch": 0.94, + "learning_rate": 1.8900261624492987e-07, + "loss": 1.0181, + "step": 29389 + }, + { + "epoch": 0.94, + "learning_rate": 1.8880223803860897e-07, + "loss": 0.8838, + "step": 29390 + }, + { + "epoch": 0.94, + "learning_rate": 1.886019650955817e-07, + "loss": 0.9697, + "step": 29391 + }, + { + "epoch": 0.94, + "learning_rate": 1.8840179741799636e-07, + "loss": 0.8892, + "step": 29392 + }, + { + "epoch": 0.94, + "learning_rate": 1.882017350080012e-07, + "loss": 0.8926, + "step": 29393 + }, + { + "epoch": 0.94, + "learning_rate": 1.8800177786774232e-07, + "loss": 0.8652, + "step": 29394 + }, + { + "epoch": 0.94, + "learning_rate": 1.8780192599936465e-07, + "loss": 0.8281, + "step": 29395 + }, + { + "epoch": 0.94, + "learning_rate": 1.87602179405012e-07, + "loss": 1.0063, + "step": 29396 + }, + { + "epoch": 0.94, + "learning_rate": 1.874025380868294e-07, + "loss": 0.6743, + "step": 29397 + }, + { + "epoch": 0.94, + "learning_rate": 1.872030020469573e-07, + "loss": 0.8618, + "step": 29398 + }, + { + "epoch": 0.94, + "learning_rate": 1.8700357128753622e-07, + "loss": 0.9785, + "step": 29399 + }, + { + "epoch": 0.94, + "learning_rate": 1.8680424581070667e-07, + "loss": 1.0229, + "step": 29400 + }, + { + "epoch": 0.94, + "learning_rate": 1.8660502561860694e-07, + "loss": 0.9111, + "step": 29401 + }, + { + "epoch": 0.94, + "learning_rate": 1.8640591071337422e-07, + "loss": 0.9077, + "step": 29402 + }, + { + "epoch": 0.94, + "learning_rate": 1.8620690109714569e-07, + "loss": 0.9731, + "step": 29403 + }, + { + "epoch": 0.94, + "learning_rate": 1.8600799677205517e-07, + "loss": 0.9653, + "step": 29404 + }, + { + "epoch": 0.94, + "learning_rate": 1.8580919774023654e-07, + "loss": 0.9546, + "step": 29405 + }, + { + "epoch": 0.94, + "learning_rate": 1.8561050400382362e-07, + "loss": 0.8459, + "step": 29406 + }, + { + "epoch": 0.94, + "learning_rate": 1.8541191556494808e-07, + "loss": 0.9062, + "step": 29407 + }, + { + "epoch": 0.94, + "learning_rate": 1.8521343242574042e-07, + "loss": 1.0278, + "step": 29408 + }, + { + "epoch": 0.94, + "learning_rate": 1.8501505458833114e-07, + "loss": 0.897, + "step": 29409 + }, + { + "epoch": 0.94, + "learning_rate": 1.8481678205484743e-07, + "loss": 0.9141, + "step": 29410 + }, + { + "epoch": 0.94, + "learning_rate": 1.846186148274154e-07, + "loss": 0.8647, + "step": 29411 + }, + { + "epoch": 0.94, + "learning_rate": 1.8442055290816329e-07, + "loss": 0.8301, + "step": 29412 + }, + { + "epoch": 0.94, + "learning_rate": 1.84222596299215e-07, + "loss": 0.894, + "step": 29413 + }, + { + "epoch": 0.94, + "learning_rate": 1.840247450026955e-07, + "loss": 0.8389, + "step": 29414 + }, + { + "epoch": 0.94, + "learning_rate": 1.8382699902072533e-07, + "loss": 1.0078, + "step": 29415 + }, + { + "epoch": 0.94, + "learning_rate": 1.8362935835542938e-07, + "loss": 0.8125, + "step": 29416 + }, + { + "epoch": 0.94, + "learning_rate": 1.834318230089249e-07, + "loss": 0.9268, + "step": 29417 + }, + { + "epoch": 0.94, + "learning_rate": 1.8323439298333246e-07, + "loss": 0.9805, + "step": 29418 + }, + { + "epoch": 0.94, + "learning_rate": 1.8303706828077028e-07, + "loss": 0.9243, + "step": 29419 + }, + { + "epoch": 0.94, + "learning_rate": 1.8283984890335671e-07, + "loss": 0.8574, + "step": 29420 + }, + { + "epoch": 0.94, + "learning_rate": 1.826427348532045e-07, + "loss": 0.9854, + "step": 29421 + }, + { + "epoch": 0.94, + "learning_rate": 1.8244572613243194e-07, + "loss": 0.856, + "step": 29422 + }, + { + "epoch": 0.94, + "learning_rate": 1.822488227431507e-07, + "loss": 0.9526, + "step": 29423 + }, + { + "epoch": 0.94, + "learning_rate": 1.8205202468747463e-07, + "loss": 0.9839, + "step": 29424 + }, + { + "epoch": 0.94, + "learning_rate": 1.8185533196751316e-07, + "loss": 0.8999, + "step": 29425 + }, + { + "epoch": 0.94, + "learning_rate": 1.8165874458537901e-07, + "loss": 0.7705, + "step": 29426 + }, + { + "epoch": 0.94, + "learning_rate": 1.8146226254317834e-07, + "loss": 0.8784, + "step": 29427 + }, + { + "epoch": 0.94, + "learning_rate": 1.8126588584302273e-07, + "loss": 0.9004, + "step": 29428 + }, + { + "epoch": 0.94, + "learning_rate": 1.810696144870161e-07, + "loss": 0.8896, + "step": 29429 + }, + { + "epoch": 0.94, + "learning_rate": 1.8087344847726563e-07, + "loss": 0.9531, + "step": 29430 + }, + { + "epoch": 0.94, + "learning_rate": 1.806773878158763e-07, + "loss": 0.9819, + "step": 29431 + }, + { + "epoch": 0.94, + "learning_rate": 1.8048143250495199e-07, + "loss": 0.9663, + "step": 29432 + }, + { + "epoch": 0.94, + "learning_rate": 1.8028558254659213e-07, + "loss": 0.9434, + "step": 29433 + }, + { + "epoch": 0.94, + "learning_rate": 1.8008983794290168e-07, + "loss": 0.9634, + "step": 29434 + }, + { + "epoch": 0.94, + "learning_rate": 1.798941986959779e-07, + "loss": 1.061, + "step": 29435 + }, + { + "epoch": 0.94, + "learning_rate": 1.7969866480792242e-07, + "loss": 0.9404, + "step": 29436 + }, + { + "epoch": 0.94, + "learning_rate": 1.7950323628083134e-07, + "loss": 0.7158, + "step": 29437 + }, + { + "epoch": 0.94, + "learning_rate": 1.7930791311680074e-07, + "loss": 0.9038, + "step": 29438 + }, + { + "epoch": 0.94, + "learning_rate": 1.7911269531792786e-07, + "loss": 0.4766, + "step": 29439 + }, + { + "epoch": 0.94, + "learning_rate": 1.7891758288630656e-07, + "loss": 0.8652, + "step": 29440 + }, + { + "epoch": 0.94, + "learning_rate": 1.7872257582402962e-07, + "loss": 0.9204, + "step": 29441 + }, + { + "epoch": 0.94, + "learning_rate": 1.7852767413319093e-07, + "loss": 0.876, + "step": 29442 + }, + { + "epoch": 0.94, + "learning_rate": 1.7833287781587994e-07, + "loss": 0.9033, + "step": 29443 + }, + { + "epoch": 0.94, + "learning_rate": 1.7813818687418606e-07, + "loss": 0.8862, + "step": 29444 + }, + { + "epoch": 0.94, + "learning_rate": 1.7794360131020094e-07, + "loss": 0.8682, + "step": 29445 + }, + { + "epoch": 0.94, + "learning_rate": 1.7774912112600851e-07, + "loss": 0.9458, + "step": 29446 + }, + { + "epoch": 0.94, + "learning_rate": 1.7755474632369818e-07, + "loss": 0.8975, + "step": 29447 + }, + { + "epoch": 0.94, + "learning_rate": 1.7736047690535495e-07, + "loss": 0.8843, + "step": 29448 + }, + { + "epoch": 0.94, + "learning_rate": 1.7716631287306275e-07, + "loss": 1.0049, + "step": 29449 + }, + { + "epoch": 0.94, + "learning_rate": 1.769722542289043e-07, + "loss": 0.8931, + "step": 29450 + }, + { + "epoch": 0.94, + "learning_rate": 1.7677830097496129e-07, + "loss": 1.0181, + "step": 29451 + }, + { + "epoch": 0.94, + "learning_rate": 1.765844531133154e-07, + "loss": 0.9702, + "step": 29452 + }, + { + "epoch": 0.94, + "learning_rate": 1.7639071064604718e-07, + "loss": 1.0073, + "step": 29453 + }, + { + "epoch": 0.94, + "learning_rate": 1.7619707357523498e-07, + "loss": 1.0171, + "step": 29454 + }, + { + "epoch": 0.94, + "learning_rate": 1.7600354190295488e-07, + "loss": 0.9409, + "step": 29455 + }, + { + "epoch": 0.94, + "learning_rate": 1.7581011563128302e-07, + "loss": 0.9473, + "step": 29456 + }, + { + "epoch": 0.94, + "learning_rate": 1.7561679476229664e-07, + "loss": 0.8745, + "step": 29457 + }, + { + "epoch": 0.94, + "learning_rate": 1.7542357929806852e-07, + "loss": 0.8867, + "step": 29458 + }, + { + "epoch": 0.94, + "learning_rate": 1.7523046924067366e-07, + "loss": 0.8518, + "step": 29459 + }, + { + "epoch": 0.94, + "learning_rate": 1.750374645921804e-07, + "loss": 0.9189, + "step": 29460 + }, + { + "epoch": 0.94, + "learning_rate": 1.7484456535466266e-07, + "loss": 0.918, + "step": 29461 + }, + { + "epoch": 0.94, + "learning_rate": 1.7465177153018874e-07, + "loss": 0.8845, + "step": 29462 + }, + { + "epoch": 0.94, + "learning_rate": 1.7445908312082705e-07, + "loss": 0.8511, + "step": 29463 + }, + { + "epoch": 0.94, + "learning_rate": 1.7426650012864476e-07, + "loss": 0.9966, + "step": 29464 + }, + { + "epoch": 0.94, + "learning_rate": 1.7407402255570804e-07, + "loss": 0.9478, + "step": 29465 + }, + { + "epoch": 0.94, + "learning_rate": 1.73881650404083e-07, + "loss": 1.0469, + "step": 29466 + }, + { + "epoch": 0.94, + "learning_rate": 1.7368938367583242e-07, + "loss": 0.9189, + "step": 29467 + }, + { + "epoch": 0.94, + "learning_rate": 1.7349722237301914e-07, + "loss": 0.999, + "step": 29468 + }, + { + "epoch": 0.94, + "learning_rate": 1.73305166497707e-07, + "loss": 0.917, + "step": 29469 + }, + { + "epoch": 0.94, + "learning_rate": 1.731132160519522e-07, + "loss": 0.8638, + "step": 29470 + }, + { + "epoch": 0.94, + "learning_rate": 1.7292137103781858e-07, + "loss": 0.9355, + "step": 29471 + }, + { + "epoch": 0.94, + "learning_rate": 1.727296314573612e-07, + "loss": 0.9272, + "step": 29472 + }, + { + "epoch": 0.94, + "learning_rate": 1.7253799731263954e-07, + "loss": 0.958, + "step": 29473 + }, + { + "epoch": 0.94, + "learning_rate": 1.7234646860570858e-07, + "loss": 0.8677, + "step": 29474 + }, + { + "epoch": 0.94, + "learning_rate": 1.7215504533862336e-07, + "loss": 0.8623, + "step": 29475 + }, + { + "epoch": 0.94, + "learning_rate": 1.719637275134367e-07, + "loss": 0.8896, + "step": 29476 + }, + { + "epoch": 0.94, + "learning_rate": 1.7177251513220472e-07, + "loss": 0.8779, + "step": 29477 + }, + { + "epoch": 0.94, + "learning_rate": 1.7158140819697355e-07, + "loss": 0.8838, + "step": 29478 + }, + { + "epoch": 0.94, + "learning_rate": 1.713904067097971e-07, + "loss": 0.8667, + "step": 29479 + }, + { + "epoch": 0.94, + "learning_rate": 1.7119951067272378e-07, + "loss": 0.7224, + "step": 29480 + }, + { + "epoch": 0.94, + "learning_rate": 1.7100872008780188e-07, + "loss": 1.0103, + "step": 29481 + }, + { + "epoch": 0.94, + "learning_rate": 1.708180349570787e-07, + "loss": 0.957, + "step": 29482 + }, + { + "epoch": 0.94, + "learning_rate": 1.7062745528259927e-07, + "loss": 0.8809, + "step": 29483 + }, + { + "epoch": 0.94, + "learning_rate": 1.704369810664086e-07, + "loss": 0.981, + "step": 29484 + }, + { + "epoch": 0.94, + "learning_rate": 1.7024661231055063e-07, + "loss": 0.9067, + "step": 29485 + }, + { + "epoch": 0.94, + "learning_rate": 1.700563490170659e-07, + "loss": 0.9121, + "step": 29486 + }, + { + "epoch": 0.94, + "learning_rate": 1.6986619118799952e-07, + "loss": 0.8647, + "step": 29487 + }, + { + "epoch": 0.94, + "learning_rate": 1.696761388253887e-07, + "loss": 0.4443, + "step": 29488 + }, + { + "epoch": 0.94, + "learning_rate": 1.694861919312718e-07, + "loss": 0.9707, + "step": 29489 + }, + { + "epoch": 0.94, + "learning_rate": 1.6929635050768945e-07, + "loss": 0.9717, + "step": 29490 + }, + { + "epoch": 0.94, + "learning_rate": 1.6910661455667665e-07, + "loss": 0.895, + "step": 29491 + }, + { + "epoch": 0.94, + "learning_rate": 1.689169840802707e-07, + "loss": 0.8506, + "step": 29492 + }, + { + "epoch": 0.94, + "learning_rate": 1.6872745908050436e-07, + "loss": 0.8848, + "step": 29493 + }, + { + "epoch": 0.94, + "learning_rate": 1.6853803955941161e-07, + "loss": 0.8301, + "step": 29494 + }, + { + "epoch": 0.94, + "learning_rate": 1.6834872551902415e-07, + "loss": 0.8735, + "step": 29495 + }, + { + "epoch": 0.94, + "learning_rate": 1.6815951696137478e-07, + "loss": 0.9517, + "step": 29496 + }, + { + "epoch": 0.94, + "learning_rate": 1.6797041388849188e-07, + "loss": 0.9863, + "step": 29497 + }, + { + "epoch": 0.94, + "learning_rate": 1.6778141630240496e-07, + "loss": 0.9658, + "step": 29498 + }, + { + "epoch": 0.94, + "learning_rate": 1.6759252420514238e-07, + "loss": 0.8457, + "step": 29499 + }, + { + "epoch": 0.94, + "learning_rate": 1.674037375987292e-07, + "loss": 0.9355, + "step": 29500 + }, + { + "epoch": 0.94, + "learning_rate": 1.6721505648519153e-07, + "loss": 0.9404, + "step": 29501 + }, + { + "epoch": 0.94, + "learning_rate": 1.6702648086655449e-07, + "loss": 0.8066, + "step": 29502 + }, + { + "epoch": 0.94, + "learning_rate": 1.6683801074484084e-07, + "loss": 0.7783, + "step": 29503 + }, + { + "epoch": 0.94, + "learning_rate": 1.6664964612207236e-07, + "loss": 0.9194, + "step": 29504 + }, + { + "epoch": 0.94, + "learning_rate": 1.664613870002696e-07, + "loss": 1.0322, + "step": 29505 + }, + { + "epoch": 0.94, + "learning_rate": 1.6627323338145318e-07, + "loss": 0.9634, + "step": 29506 + }, + { + "epoch": 0.94, + "learning_rate": 1.6608518526764262e-07, + "loss": 0.8838, + "step": 29507 + }, + { + "epoch": 0.94, + "learning_rate": 1.6589724266085404e-07, + "loss": 0.9712, + "step": 29508 + }, + { + "epoch": 0.94, + "learning_rate": 1.6570940556310256e-07, + "loss": 0.8271, + "step": 29509 + }, + { + "epoch": 0.94, + "learning_rate": 1.655216739764065e-07, + "loss": 0.918, + "step": 29510 + }, + { + "epoch": 0.94, + "learning_rate": 1.6533404790277762e-07, + "loss": 0.9375, + "step": 29511 + }, + { + "epoch": 0.94, + "learning_rate": 1.6514652734423098e-07, + "loss": 0.8301, + "step": 29512 + }, + { + "epoch": 0.94, + "learning_rate": 1.6495911230277717e-07, + "loss": 1.0166, + "step": 29513 + }, + { + "epoch": 0.94, + "learning_rate": 1.6477180278042793e-07, + "loss": 0.9341, + "step": 29514 + }, + { + "epoch": 0.94, + "learning_rate": 1.6458459877919052e-07, + "loss": 0.8789, + "step": 29515 + }, + { + "epoch": 0.94, + "learning_rate": 1.643975003010767e-07, + "loss": 0.8608, + "step": 29516 + }, + { + "epoch": 0.94, + "learning_rate": 1.642105073480904e-07, + "loss": 0.8408, + "step": 29517 + }, + { + "epoch": 0.94, + "learning_rate": 1.6402361992224114e-07, + "loss": 0.8916, + "step": 29518 + }, + { + "epoch": 0.94, + "learning_rate": 1.6383683802553175e-07, + "loss": 0.9614, + "step": 29519 + }, + { + "epoch": 0.94, + "learning_rate": 1.6365016165996727e-07, + "loss": 0.856, + "step": 29520 + }, + { + "epoch": 0.94, + "learning_rate": 1.634635908275506e-07, + "loss": 1.04, + "step": 29521 + }, + { + "epoch": 0.94, + "learning_rate": 1.632771255302834e-07, + "loss": 1.0264, + "step": 29522 + }, + { + "epoch": 0.94, + "learning_rate": 1.6309076577016413e-07, + "loss": 0.9287, + "step": 29523 + }, + { + "epoch": 0.94, + "learning_rate": 1.629045115491956e-07, + "loss": 0.762, + "step": 29524 + }, + { + "epoch": 0.94, + "learning_rate": 1.6271836286937292e-07, + "loss": 0.9224, + "step": 29525 + }, + { + "epoch": 0.94, + "learning_rate": 1.6253231973269666e-07, + "loss": 0.9072, + "step": 29526 + }, + { + "epoch": 0.94, + "learning_rate": 1.6234638214116082e-07, + "loss": 0.897, + "step": 29527 + }, + { + "epoch": 0.94, + "learning_rate": 1.6216055009675935e-07, + "loss": 0.9688, + "step": 29528 + }, + { + "epoch": 0.94, + "learning_rate": 1.6197482360148732e-07, + "loss": 0.8931, + "step": 29529 + }, + { + "epoch": 0.94, + "learning_rate": 1.6178920265733866e-07, + "loss": 0.979, + "step": 29530 + }, + { + "epoch": 0.94, + "learning_rate": 1.616036872663018e-07, + "loss": 0.8169, + "step": 29531 + }, + { + "epoch": 0.94, + "learning_rate": 1.6141827743036965e-07, + "loss": 0.9341, + "step": 29532 + }, + { + "epoch": 0.94, + "learning_rate": 1.6123297315153052e-07, + "loss": 0.4966, + "step": 29533 + }, + { + "epoch": 0.94, + "learning_rate": 1.6104777443177176e-07, + "loss": 0.9229, + "step": 29534 + }, + { + "epoch": 0.94, + "learning_rate": 1.608626812730829e-07, + "loss": 0.916, + "step": 29535 + }, + { + "epoch": 0.94, + "learning_rate": 1.6067769367744568e-07, + "loss": 0.8877, + "step": 29536 + }, + { + "epoch": 0.94, + "learning_rate": 1.604928116468496e-07, + "loss": 0.9458, + "step": 29537 + }, + { + "epoch": 0.94, + "learning_rate": 1.603080351832742e-07, + "loss": 0.9644, + "step": 29538 + }, + { + "epoch": 0.94, + "learning_rate": 1.6012336428870457e-07, + "loss": 0.9189, + "step": 29539 + }, + { + "epoch": 0.94, + "learning_rate": 1.5993879896512021e-07, + "loss": 0.8926, + "step": 29540 + }, + { + "epoch": 0.94, + "learning_rate": 1.597543392145029e-07, + "loss": 0.9639, + "step": 29541 + }, + { + "epoch": 0.94, + "learning_rate": 1.5956998503882991e-07, + "loss": 0.9878, + "step": 29542 + }, + { + "epoch": 0.94, + "learning_rate": 1.5938573644008083e-07, + "loss": 0.8145, + "step": 29543 + }, + { + "epoch": 0.94, + "learning_rate": 1.5920159342023068e-07, + "loss": 0.9331, + "step": 29544 + }, + { + "epoch": 0.94, + "learning_rate": 1.590175559812579e-07, + "loss": 0.7783, + "step": 29545 + }, + { + "epoch": 0.94, + "learning_rate": 1.5883362412513426e-07, + "loss": 0.9189, + "step": 29546 + }, + { + "epoch": 0.94, + "learning_rate": 1.5864979785383482e-07, + "loss": 0.8838, + "step": 29547 + }, + { + "epoch": 0.94, + "learning_rate": 1.5846607716933028e-07, + "loss": 0.9746, + "step": 29548 + }, + { + "epoch": 0.94, + "learning_rate": 1.5828246207359345e-07, + "loss": 0.8491, + "step": 29549 + }, + { + "epoch": 0.95, + "learning_rate": 1.5809895256859276e-07, + "loss": 0.8223, + "step": 29550 + }, + { + "epoch": 0.95, + "learning_rate": 1.5791554865629777e-07, + "loss": 0.9053, + "step": 29551 + }, + { + "epoch": 0.95, + "learning_rate": 1.57732250338678e-07, + "loss": 0.8169, + "step": 29552 + }, + { + "epoch": 0.95, + "learning_rate": 1.5754905761769634e-07, + "loss": 0.9883, + "step": 29553 + }, + { + "epoch": 0.95, + "learning_rate": 1.573659704953212e-07, + "loss": 0.8452, + "step": 29554 + }, + { + "epoch": 0.95, + "learning_rate": 1.5718298897351548e-07, + "loss": 0.9277, + "step": 29555 + }, + { + "epoch": 0.95, + "learning_rate": 1.57000113054242e-07, + "loss": 0.9219, + "step": 29556 + }, + { + "epoch": 0.95, + "learning_rate": 1.568173427394648e-07, + "loss": 0.8928, + "step": 29557 + }, + { + "epoch": 0.95, + "learning_rate": 1.566346780311434e-07, + "loss": 0.8828, + "step": 29558 + }, + { + "epoch": 0.95, + "learning_rate": 1.5645211893123846e-07, + "loss": 0.9619, + "step": 29559 + }, + { + "epoch": 0.95, + "learning_rate": 1.5626966544170729e-07, + "loss": 0.9399, + "step": 29560 + }, + { + "epoch": 0.95, + "learning_rate": 1.5608731756450834e-07, + "loss": 0.9644, + "step": 29561 + }, + { + "epoch": 0.95, + "learning_rate": 1.5590507530159672e-07, + "loss": 0.9453, + "step": 29562 + }, + { + "epoch": 0.95, + "learning_rate": 1.5572293865492972e-07, + "loss": 0.895, + "step": 29563 + }, + { + "epoch": 0.95, + "learning_rate": 1.5554090762646023e-07, + "loss": 0.728, + "step": 29564 + }, + { + "epoch": 0.95, + "learning_rate": 1.5535898221814227e-07, + "loss": 0.8979, + "step": 29565 + }, + { + "epoch": 0.95, + "learning_rate": 1.551771624319265e-07, + "loss": 0.9214, + "step": 29566 + }, + { + "epoch": 0.95, + "learning_rate": 1.5499544826976353e-07, + "loss": 0.8569, + "step": 29567 + }, + { + "epoch": 0.95, + "learning_rate": 1.5481383973360298e-07, + "loss": 0.9106, + "step": 29568 + }, + { + "epoch": 0.95, + "learning_rate": 1.5463233682539547e-07, + "loss": 0.8589, + "step": 29569 + }, + { + "epoch": 0.95, + "learning_rate": 1.54450939547085e-07, + "loss": 0.8779, + "step": 29570 + }, + { + "epoch": 0.95, + "learning_rate": 1.5426964790062004e-07, + "loss": 0.9414, + "step": 29571 + }, + { + "epoch": 0.95, + "learning_rate": 1.5408846188794567e-07, + "loss": 0.9766, + "step": 29572 + }, + { + "epoch": 0.95, + "learning_rate": 1.5390738151100482e-07, + "loss": 0.8398, + "step": 29573 + }, + { + "epoch": 0.95, + "learning_rate": 1.5372640677174032e-07, + "loss": 0.8843, + "step": 29574 + }, + { + "epoch": 0.95, + "learning_rate": 1.5354553767209513e-07, + "loss": 0.8799, + "step": 29575 + }, + { + "epoch": 0.95, + "learning_rate": 1.5336477421400875e-07, + "loss": 0.8701, + "step": 29576 + }, + { + "epoch": 0.95, + "learning_rate": 1.5318411639941855e-07, + "loss": 0.8428, + "step": 29577 + }, + { + "epoch": 0.95, + "learning_rate": 1.530035642302674e-07, + "loss": 0.4966, + "step": 29578 + }, + { + "epoch": 0.95, + "learning_rate": 1.5282311770848823e-07, + "loss": 0.8687, + "step": 29579 + }, + { + "epoch": 0.95, + "learning_rate": 1.5264277683601946e-07, + "loss": 0.5007, + "step": 29580 + }, + { + "epoch": 0.95, + "learning_rate": 1.5246254161479402e-07, + "loss": 0.9526, + "step": 29581 + }, + { + "epoch": 0.95, + "learning_rate": 1.522824120467492e-07, + "loss": 0.9741, + "step": 29582 + }, + { + "epoch": 0.95, + "learning_rate": 1.521023881338135e-07, + "loss": 0.9106, + "step": 29583 + }, + { + "epoch": 0.95, + "learning_rate": 1.519224698779198e-07, + "loss": 0.957, + "step": 29584 + }, + { + "epoch": 0.95, + "learning_rate": 1.517426572809988e-07, + "loss": 0.9946, + "step": 29585 + }, + { + "epoch": 0.95, + "learning_rate": 1.5156295034498002e-07, + "loss": 0.8794, + "step": 29586 + }, + { + "epoch": 0.95, + "learning_rate": 1.5138334907179085e-07, + "loss": 0.8979, + "step": 29587 + }, + { + "epoch": 0.95, + "learning_rate": 1.5120385346335863e-07, + "loss": 0.8662, + "step": 29588 + }, + { + "epoch": 0.95, + "learning_rate": 1.5102446352160848e-07, + "loss": 0.9805, + "step": 29589 + }, + { + "epoch": 0.95, + "learning_rate": 1.5084517924846665e-07, + "loss": 0.8198, + "step": 29590 + }, + { + "epoch": 0.95, + "learning_rate": 1.506660006458549e-07, + "loss": 0.8086, + "step": 29591 + }, + { + "epoch": 0.95, + "learning_rate": 1.5048692771569616e-07, + "loss": 0.9355, + "step": 29592 + }, + { + "epoch": 0.95, + "learning_rate": 1.5030796045991114e-07, + "loss": 0.9863, + "step": 29593 + }, + { + "epoch": 0.95, + "learning_rate": 1.501290988804216e-07, + "loss": 0.8555, + "step": 29594 + }, + { + "epoch": 0.95, + "learning_rate": 1.499503429791449e-07, + "loss": 0.9219, + "step": 29595 + }, + { + "epoch": 0.95, + "learning_rate": 1.4977169275800062e-07, + "loss": 0.9453, + "step": 29596 + }, + { + "epoch": 0.95, + "learning_rate": 1.4959314821890282e-07, + "loss": 0.918, + "step": 29597 + }, + { + "epoch": 0.95, + "learning_rate": 1.49414709363771e-07, + "loss": 0.9658, + "step": 29598 + }, + { + "epoch": 0.95, + "learning_rate": 1.492363761945148e-07, + "loss": 0.9277, + "step": 29599 + }, + { + "epoch": 0.95, + "learning_rate": 1.4905814871305157e-07, + "loss": 1.0474, + "step": 29600 + }, + { + "epoch": 0.95, + "learning_rate": 1.4888002692129088e-07, + "loss": 0.8608, + "step": 29601 + }, + { + "epoch": 0.95, + "learning_rate": 1.4870201082114454e-07, + "loss": 0.9365, + "step": 29602 + }, + { + "epoch": 0.95, + "learning_rate": 1.4852410041452327e-07, + "loss": 0.8496, + "step": 29603 + }, + { + "epoch": 0.95, + "learning_rate": 1.4834629570333548e-07, + "loss": 0.4412, + "step": 29604 + }, + { + "epoch": 0.95, + "learning_rate": 1.481685966894897e-07, + "loss": 0.9146, + "step": 29605 + }, + { + "epoch": 0.95, + "learning_rate": 1.4799100337488993e-07, + "loss": 0.8779, + "step": 29606 + }, + { + "epoch": 0.95, + "learning_rate": 1.4781351576144354e-07, + "loss": 0.9702, + "step": 29607 + }, + { + "epoch": 0.95, + "learning_rate": 1.4763613385105457e-07, + "loss": 0.8418, + "step": 29608 + }, + { + "epoch": 0.95, + "learning_rate": 1.4745885764562485e-07, + "loss": 0.7939, + "step": 29609 + }, + { + "epoch": 0.95, + "learning_rate": 1.4728168714705838e-07, + "loss": 1.0645, + "step": 29610 + }, + { + "epoch": 0.95, + "learning_rate": 1.4710462235725475e-07, + "loss": 0.8667, + "step": 29611 + }, + { + "epoch": 0.95, + "learning_rate": 1.4692766327811247e-07, + "loss": 0.9834, + "step": 29612 + }, + { + "epoch": 0.95, + "learning_rate": 1.4675080991153335e-07, + "loss": 0.9282, + "step": 29613 + }, + { + "epoch": 0.95, + "learning_rate": 1.465740622594125e-07, + "loss": 1.0713, + "step": 29614 + }, + { + "epoch": 0.95, + "learning_rate": 1.4639742032364734e-07, + "loss": 0.9736, + "step": 29615 + }, + { + "epoch": 0.95, + "learning_rate": 1.4622088410613077e-07, + "loss": 0.9155, + "step": 29616 + }, + { + "epoch": 0.95, + "learning_rate": 1.4604445360876018e-07, + "loss": 0.9146, + "step": 29617 + }, + { + "epoch": 0.95, + "learning_rate": 1.4586812883342516e-07, + "loss": 0.8965, + "step": 29618 + }, + { + "epoch": 0.95, + "learning_rate": 1.4569190978201975e-07, + "loss": 0.9165, + "step": 29619 + }, + { + "epoch": 0.95, + "learning_rate": 1.4551579645643465e-07, + "loss": 0.9683, + "step": 29620 + }, + { + "epoch": 0.95, + "learning_rate": 1.4533978885855837e-07, + "loss": 0.9121, + "step": 29621 + }, + { + "epoch": 0.95, + "learning_rate": 1.451638869902794e-07, + "loss": 0.8276, + "step": 29622 + }, + { + "epoch": 0.95, + "learning_rate": 1.449880908534851e-07, + "loss": 0.8843, + "step": 29623 + }, + { + "epoch": 0.95, + "learning_rate": 1.4481240045006062e-07, + "loss": 0.8838, + "step": 29624 + }, + { + "epoch": 0.95, + "learning_rate": 1.4463681578189338e-07, + "loss": 0.7959, + "step": 29625 + }, + { + "epoch": 0.95, + "learning_rate": 1.4446133685086518e-07, + "loss": 0.9189, + "step": 29626 + }, + { + "epoch": 0.95, + "learning_rate": 1.4428596365885894e-07, + "loss": 0.9795, + "step": 29627 + }, + { + "epoch": 0.95, + "learning_rate": 1.4411069620775764e-07, + "loss": 0.7986, + "step": 29628 + }, + { + "epoch": 0.95, + "learning_rate": 1.4393553449943975e-07, + "loss": 0.8901, + "step": 29629 + }, + { + "epoch": 0.95, + "learning_rate": 1.437604785357849e-07, + "loss": 0.5171, + "step": 29630 + }, + { + "epoch": 0.95, + "learning_rate": 1.4358552831867268e-07, + "loss": 0.8027, + "step": 29631 + }, + { + "epoch": 0.95, + "learning_rate": 1.4341068384997824e-07, + "loss": 0.8848, + "step": 29632 + }, + { + "epoch": 0.95, + "learning_rate": 1.4323594513158012e-07, + "loss": 0.9307, + "step": 29633 + }, + { + "epoch": 0.95, + "learning_rate": 1.4306131216534902e-07, + "loss": 0.7549, + "step": 29634 + }, + { + "epoch": 0.95, + "learning_rate": 1.4288678495316343e-07, + "loss": 1.02, + "step": 29635 + }, + { + "epoch": 0.95, + "learning_rate": 1.4271236349689187e-07, + "loss": 0.8574, + "step": 29636 + }, + { + "epoch": 0.95, + "learning_rate": 1.425380477984073e-07, + "loss": 0.9277, + "step": 29637 + }, + { + "epoch": 0.95, + "learning_rate": 1.4236383785958042e-07, + "loss": 0.9697, + "step": 29638 + }, + { + "epoch": 0.95, + "learning_rate": 1.4218973368227862e-07, + "loss": 1.0127, + "step": 29639 + }, + { + "epoch": 0.95, + "learning_rate": 1.4201573526837155e-07, + "loss": 0.8667, + "step": 29640 + }, + { + "epoch": 0.95, + "learning_rate": 1.4184184261972545e-07, + "loss": 0.9102, + "step": 29641 + }, + { + "epoch": 0.95, + "learning_rate": 1.4166805573820553e-07, + "loss": 0.9229, + "step": 29642 + }, + { + "epoch": 0.95, + "learning_rate": 1.4149437462567694e-07, + "loss": 0.9888, + "step": 29643 + }, + { + "epoch": 0.95, + "learning_rate": 1.4132079928400267e-07, + "loss": 0.957, + "step": 29644 + }, + { + "epoch": 0.95, + "learning_rate": 1.4114732971504564e-07, + "loss": 0.9221, + "step": 29645 + }, + { + "epoch": 0.95, + "learning_rate": 1.4097396592066548e-07, + "loss": 0.8545, + "step": 29646 + }, + { + "epoch": 0.95, + "learning_rate": 1.4080070790272405e-07, + "loss": 1.0322, + "step": 29647 + }, + { + "epoch": 0.95, + "learning_rate": 1.4062755566307873e-07, + "loss": 0.877, + "step": 29648 + }, + { + "epoch": 0.95, + "learning_rate": 1.4045450920358917e-07, + "loss": 0.9263, + "step": 29649 + }, + { + "epoch": 0.95, + "learning_rate": 1.4028156852611051e-07, + "loss": 0.9551, + "step": 29650 + }, + { + "epoch": 0.95, + "learning_rate": 1.40108733632498e-07, + "loss": 0.9663, + "step": 29651 + }, + { + "epoch": 0.95, + "learning_rate": 1.3993600452460565e-07, + "loss": 0.9131, + "step": 29652 + }, + { + "epoch": 0.95, + "learning_rate": 1.3976338120428866e-07, + "loss": 0.9111, + "step": 29653 + }, + { + "epoch": 0.95, + "learning_rate": 1.395908636733978e-07, + "loss": 0.8838, + "step": 29654 + }, + { + "epoch": 0.95, + "learning_rate": 1.3941845193378268e-07, + "loss": 0.8472, + "step": 29655 + }, + { + "epoch": 0.95, + "learning_rate": 1.3924614598729514e-07, + "loss": 0.9136, + "step": 29656 + }, + { + "epoch": 0.95, + "learning_rate": 1.3907394583578372e-07, + "loss": 0.7134, + "step": 29657 + }, + { + "epoch": 0.95, + "learning_rate": 1.389018514810947e-07, + "loss": 0.8555, + "step": 29658 + }, + { + "epoch": 0.95, + "learning_rate": 1.3872986292507552e-07, + "loss": 0.8945, + "step": 29659 + }, + { + "epoch": 0.95, + "learning_rate": 1.3855798016957022e-07, + "loss": 0.9922, + "step": 29660 + }, + { + "epoch": 0.95, + "learning_rate": 1.3838620321642404e-07, + "loss": 0.9148, + "step": 29661 + }, + { + "epoch": 0.95, + "learning_rate": 1.3821453206747993e-07, + "loss": 0.9341, + "step": 29662 + }, + { + "epoch": 0.95, + "learning_rate": 1.380429667245786e-07, + "loss": 0.9019, + "step": 29663 + }, + { + "epoch": 0.95, + "learning_rate": 1.378715071895631e-07, + "loss": 0.9448, + "step": 29664 + }, + { + "epoch": 0.95, + "learning_rate": 1.377001534642708e-07, + "loss": 0.9087, + "step": 29665 + }, + { + "epoch": 0.95, + "learning_rate": 1.3752890555054022e-07, + "loss": 0.6926, + "step": 29666 + }, + { + "epoch": 0.95, + "learning_rate": 1.373577634502088e-07, + "loss": 0.8691, + "step": 29667 + }, + { + "epoch": 0.95, + "learning_rate": 1.3718672716511506e-07, + "loss": 0.8638, + "step": 29668 + }, + { + "epoch": 0.95, + "learning_rate": 1.3701579669708976e-07, + "loss": 1.0171, + "step": 29669 + }, + { + "epoch": 0.95, + "learning_rate": 1.3684497204797032e-07, + "loss": 0.8521, + "step": 29670 + }, + { + "epoch": 0.95, + "learning_rate": 1.366742532195886e-07, + "loss": 0.9673, + "step": 29671 + }, + { + "epoch": 0.95, + "learning_rate": 1.3650364021377537e-07, + "loss": 0.8564, + "step": 29672 + }, + { + "epoch": 0.95, + "learning_rate": 1.363331330323625e-07, + "loss": 1.0356, + "step": 29673 + }, + { + "epoch": 0.95, + "learning_rate": 1.3616273167717852e-07, + "loss": 0.9336, + "step": 29674 + }, + { + "epoch": 0.95, + "learning_rate": 1.3599243615005087e-07, + "loss": 0.7617, + "step": 29675 + }, + { + "epoch": 0.95, + "learning_rate": 1.3582224645280807e-07, + "loss": 0.8638, + "step": 29676 + }, + { + "epoch": 0.95, + "learning_rate": 1.356521625872742e-07, + "loss": 0.9082, + "step": 29677 + }, + { + "epoch": 0.95, + "learning_rate": 1.3548218455527674e-07, + "loss": 0.9419, + "step": 29678 + }, + { + "epoch": 0.95, + "learning_rate": 1.3531231235863752e-07, + "loss": 0.8896, + "step": 29679 + }, + { + "epoch": 0.95, + "learning_rate": 1.3514254599917952e-07, + "loss": 0.9751, + "step": 29680 + }, + { + "epoch": 0.95, + "learning_rate": 1.3497288547872355e-07, + "loss": 0.9941, + "step": 29681 + }, + { + "epoch": 0.95, + "learning_rate": 1.3480333079909146e-07, + "loss": 0.8794, + "step": 29682 + }, + { + "epoch": 0.95, + "learning_rate": 1.346338819621007e-07, + "loss": 0.8604, + "step": 29683 + }, + { + "epoch": 0.95, + "learning_rate": 1.344645389695698e-07, + "loss": 0.9795, + "step": 29684 + }, + { + "epoch": 0.95, + "learning_rate": 1.342953018233162e-07, + "loss": 0.9219, + "step": 29685 + }, + { + "epoch": 0.95, + "learning_rate": 1.3412617052515508e-07, + "loss": 0.9106, + "step": 29686 + }, + { + "epoch": 0.95, + "learning_rate": 1.339571450769006e-07, + "loss": 0.9048, + "step": 29687 + }, + { + "epoch": 0.95, + "learning_rate": 1.3378822548036795e-07, + "loss": 0.8452, + "step": 29688 + }, + { + "epoch": 0.95, + "learning_rate": 1.336194117373668e-07, + "loss": 0.9731, + "step": 29689 + }, + { + "epoch": 0.95, + "learning_rate": 1.3345070384971127e-07, + "loss": 0.8145, + "step": 29690 + }, + { + "epoch": 0.95, + "learning_rate": 1.3328210181920875e-07, + "loss": 1.041, + "step": 29691 + }, + { + "epoch": 0.95, + "learning_rate": 1.3311360564767006e-07, + "loss": 0.7983, + "step": 29692 + }, + { + "epoch": 0.95, + "learning_rate": 1.3294521533690152e-07, + "loss": 0.8926, + "step": 29693 + }, + { + "epoch": 0.95, + "learning_rate": 1.327769308887117e-07, + "loss": 0.9531, + "step": 29694 + }, + { + "epoch": 0.95, + "learning_rate": 1.3260875230490356e-07, + "loss": 0.8723, + "step": 29695 + }, + { + "epoch": 0.95, + "learning_rate": 1.3244067958728456e-07, + "loss": 0.873, + "step": 29696 + }, + { + "epoch": 0.95, + "learning_rate": 1.3227271273765441e-07, + "loss": 0.8711, + "step": 29697 + }, + { + "epoch": 0.95, + "learning_rate": 1.3210485175781718e-07, + "loss": 0.9062, + "step": 29698 + }, + { + "epoch": 0.95, + "learning_rate": 1.3193709664957477e-07, + "loss": 0.937, + "step": 29699 + }, + { + "epoch": 0.95, + "learning_rate": 1.3176944741472464e-07, + "loss": 0.9043, + "step": 29700 + }, + { + "epoch": 0.95, + "learning_rate": 1.3160190405506756e-07, + "loss": 0.9185, + "step": 29701 + }, + { + "epoch": 0.95, + "learning_rate": 1.3143446657239989e-07, + "loss": 0.9683, + "step": 29702 + }, + { + "epoch": 0.95, + "learning_rate": 1.3126713496851906e-07, + "loss": 1.0112, + "step": 29703 + }, + { + "epoch": 0.95, + "learning_rate": 1.310999092452192e-07, + "loss": 0.999, + "step": 29704 + }, + { + "epoch": 0.95, + "learning_rate": 1.3093278940429555e-07, + "loss": 1.0078, + "step": 29705 + }, + { + "epoch": 0.95, + "learning_rate": 1.3076577544753888e-07, + "loss": 0.8901, + "step": 29706 + }, + { + "epoch": 0.95, + "learning_rate": 1.3059886737674444e-07, + "loss": 0.814, + "step": 29707 + }, + { + "epoch": 0.95, + "learning_rate": 1.3043206519369966e-07, + "loss": 0.8828, + "step": 29708 + }, + { + "epoch": 0.95, + "learning_rate": 1.302653689001965e-07, + "loss": 0.9585, + "step": 29709 + }, + { + "epoch": 0.95, + "learning_rate": 1.3009877849802233e-07, + "loss": 0.9219, + "step": 29710 + }, + { + "epoch": 0.95, + "learning_rate": 1.299322939889669e-07, + "loss": 0.98, + "step": 29711 + }, + { + "epoch": 0.95, + "learning_rate": 1.297659153748121e-07, + "loss": 1.0264, + "step": 29712 + }, + { + "epoch": 0.95, + "learning_rate": 1.2959964265734537e-07, + "loss": 0.958, + "step": 29713 + }, + { + "epoch": 0.95, + "learning_rate": 1.2943347583835085e-07, + "loss": 0.9199, + "step": 29714 + }, + { + "epoch": 0.95, + "learning_rate": 1.2926741491961047e-07, + "loss": 0.7798, + "step": 29715 + }, + { + "epoch": 0.95, + "learning_rate": 1.2910145990290722e-07, + "loss": 1.0625, + "step": 29716 + }, + { + "epoch": 0.95, + "learning_rate": 1.2893561079001971e-07, + "loss": 0.8584, + "step": 29717 + }, + { + "epoch": 0.95, + "learning_rate": 1.287698675827298e-07, + "loss": 0.9761, + "step": 29718 + }, + { + "epoch": 0.95, + "learning_rate": 1.286042302828139e-07, + "loss": 0.9331, + "step": 29719 + }, + { + "epoch": 0.95, + "learning_rate": 1.284386988920483e-07, + "loss": 0.937, + "step": 29720 + }, + { + "epoch": 0.95, + "learning_rate": 1.2827327341221053e-07, + "loss": 0.7979, + "step": 29721 + }, + { + "epoch": 0.95, + "learning_rate": 1.2810795384507467e-07, + "loss": 0.9375, + "step": 29722 + }, + { + "epoch": 0.95, + "learning_rate": 1.2794274019241603e-07, + "loss": 0.8804, + "step": 29723 + }, + { + "epoch": 0.95, + "learning_rate": 1.2777763245600538e-07, + "loss": 0.8569, + "step": 29724 + }, + { + "epoch": 0.95, + "learning_rate": 1.2761263063761464e-07, + "loss": 0.4529, + "step": 29725 + }, + { + "epoch": 0.95, + "learning_rate": 1.2744773473901463e-07, + "loss": 0.9507, + "step": 29726 + }, + { + "epoch": 0.95, + "learning_rate": 1.272829447619739e-07, + "loss": 0.8701, + "step": 29727 + }, + { + "epoch": 0.95, + "learning_rate": 1.2711826070825994e-07, + "loss": 0.9131, + "step": 29728 + }, + { + "epoch": 0.95, + "learning_rate": 1.2695368257964023e-07, + "loss": 0.8857, + "step": 29729 + }, + { + "epoch": 0.95, + "learning_rate": 1.2678921037788118e-07, + "loss": 0.9932, + "step": 29730 + }, + { + "epoch": 0.95, + "learning_rate": 1.2662484410474685e-07, + "loss": 0.9639, + "step": 29731 + }, + { + "epoch": 0.95, + "learning_rate": 1.264605837619992e-07, + "loss": 0.9424, + "step": 29732 + }, + { + "epoch": 0.95, + "learning_rate": 1.2629642935140462e-07, + "loss": 0.812, + "step": 29733 + }, + { + "epoch": 0.95, + "learning_rate": 1.2613238087471947e-07, + "loss": 0.9292, + "step": 29734 + }, + { + "epoch": 0.95, + "learning_rate": 1.2596843833370674e-07, + "loss": 0.877, + "step": 29735 + }, + { + "epoch": 0.95, + "learning_rate": 1.258046017301251e-07, + "loss": 0.8213, + "step": 29736 + }, + { + "epoch": 0.95, + "learning_rate": 1.2564087106573086e-07, + "loss": 0.9375, + "step": 29737 + }, + { + "epoch": 0.95, + "learning_rate": 1.2547724634228265e-07, + "loss": 0.9204, + "step": 29738 + }, + { + "epoch": 0.95, + "learning_rate": 1.2531372756153458e-07, + "loss": 0.9604, + "step": 29739 + }, + { + "epoch": 0.95, + "learning_rate": 1.2515031472524087e-07, + "loss": 1.0059, + "step": 29740 + }, + { + "epoch": 0.95, + "learning_rate": 1.249870078351567e-07, + "loss": 0.8892, + "step": 29741 + }, + { + "epoch": 0.95, + "learning_rate": 1.2482380689303186e-07, + "loss": 0.8076, + "step": 29742 + }, + { + "epoch": 0.95, + "learning_rate": 1.2466071190061823e-07, + "loss": 0.8184, + "step": 29743 + }, + { + "epoch": 0.95, + "learning_rate": 1.2449772285966666e-07, + "loss": 0.8828, + "step": 29744 + }, + { + "epoch": 0.95, + "learning_rate": 1.243348397719235e-07, + "loss": 0.9229, + "step": 29745 + }, + { + "epoch": 0.95, + "learning_rate": 1.241720626391385e-07, + "loss": 0.9619, + "step": 29746 + }, + { + "epoch": 0.95, + "learning_rate": 1.240093914630569e-07, + "loss": 0.9331, + "step": 29747 + }, + { + "epoch": 0.95, + "learning_rate": 1.23846826245424e-07, + "loss": 0.876, + "step": 29748 + }, + { + "epoch": 0.95, + "learning_rate": 1.2368436698798502e-07, + "loss": 0.8696, + "step": 29749 + }, + { + "epoch": 0.95, + "learning_rate": 1.2352201369248195e-07, + "loss": 0.917, + "step": 29750 + }, + { + "epoch": 0.95, + "learning_rate": 1.2335976636065672e-07, + "loss": 0.981, + "step": 29751 + }, + { + "epoch": 0.95, + "learning_rate": 1.2319762499425013e-07, + "loss": 0.8853, + "step": 29752 + }, + { + "epoch": 0.95, + "learning_rate": 1.2303558959500196e-07, + "loss": 0.7778, + "step": 29753 + }, + { + "epoch": 0.95, + "learning_rate": 1.2287366016465185e-07, + "loss": 0.9082, + "step": 29754 + }, + { + "epoch": 0.95, + "learning_rate": 1.2271183670493402e-07, + "loss": 0.9951, + "step": 29755 + }, + { + "epoch": 0.95, + "learning_rate": 1.2255011921758818e-07, + "loss": 0.9868, + "step": 29756 + }, + { + "epoch": 0.95, + "learning_rate": 1.2238850770434629e-07, + "loss": 0.8706, + "step": 29757 + }, + { + "epoch": 0.95, + "learning_rate": 1.2222700216694472e-07, + "loss": 1.0093, + "step": 29758 + }, + { + "epoch": 0.95, + "learning_rate": 1.220656026071154e-07, + "loss": 0.948, + "step": 29759 + }, + { + "epoch": 0.95, + "learning_rate": 1.2190430902658922e-07, + "loss": 0.8486, + "step": 29760 + }, + { + "epoch": 0.95, + "learning_rate": 1.2174312142709698e-07, + "loss": 0.8799, + "step": 29761 + }, + { + "epoch": 0.95, + "learning_rate": 1.2158203981036954e-07, + "loss": 0.8047, + "step": 29762 + }, + { + "epoch": 0.95, + "learning_rate": 1.2142106417813215e-07, + "loss": 0.9116, + "step": 29763 + }, + { + "epoch": 0.95, + "learning_rate": 1.2126019453211568e-07, + "loss": 0.8286, + "step": 29764 + }, + { + "epoch": 0.95, + "learning_rate": 1.210994308740421e-07, + "loss": 0.8667, + "step": 29765 + }, + { + "epoch": 0.95, + "learning_rate": 1.2093877320563995e-07, + "loss": 0.9951, + "step": 29766 + }, + { + "epoch": 0.95, + "learning_rate": 1.2077822152863017e-07, + "loss": 0.9629, + "step": 29767 + }, + { + "epoch": 0.95, + "learning_rate": 1.2061777584473689e-07, + "loss": 0.9385, + "step": 29768 + }, + { + "epoch": 0.95, + "learning_rate": 1.2045743615567984e-07, + "loss": 1.0435, + "step": 29769 + }, + { + "epoch": 0.95, + "learning_rate": 1.20297202463181e-07, + "loss": 0.8594, + "step": 29770 + }, + { + "epoch": 0.95, + "learning_rate": 1.2013707476895897e-07, + "loss": 0.8315, + "step": 29771 + }, + { + "epoch": 0.95, + "learning_rate": 1.199770530747324e-07, + "loss": 1.0952, + "step": 29772 + }, + { + "epoch": 0.95, + "learning_rate": 1.1981713738221545e-07, + "loss": 0.8267, + "step": 29773 + }, + { + "epoch": 0.95, + "learning_rate": 1.1965732769312787e-07, + "loss": 1.02, + "step": 29774 + }, + { + "epoch": 0.95, + "learning_rate": 1.1949762400918053e-07, + "loss": 1.1035, + "step": 29775 + }, + { + "epoch": 0.95, + "learning_rate": 1.1933802633208868e-07, + "loss": 0.9985, + "step": 29776 + }, + { + "epoch": 0.95, + "learning_rate": 1.1917853466356544e-07, + "loss": 0.8789, + "step": 29777 + }, + { + "epoch": 0.95, + "learning_rate": 1.1901914900532052e-07, + "loss": 0.8931, + "step": 29778 + }, + { + "epoch": 0.95, + "learning_rate": 1.1885986935906368e-07, + "loss": 0.9333, + "step": 29779 + }, + { + "epoch": 0.95, + "learning_rate": 1.1870069572650577e-07, + "loss": 0.8208, + "step": 29780 + }, + { + "epoch": 0.95, + "learning_rate": 1.185416281093521e-07, + "loss": 0.9468, + "step": 29781 + }, + { + "epoch": 0.95, + "learning_rate": 1.1838266650931129e-07, + "loss": 0.8486, + "step": 29782 + }, + { + "epoch": 0.95, + "learning_rate": 1.1822381092808865e-07, + "loss": 0.8984, + "step": 29783 + }, + { + "epoch": 0.95, + "learning_rate": 1.1806506136738616e-07, + "loss": 0.9541, + "step": 29784 + }, + { + "epoch": 0.95, + "learning_rate": 1.1790641782891022e-07, + "loss": 0.8242, + "step": 29785 + }, + { + "epoch": 0.95, + "learning_rate": 1.1774788031436168e-07, + "loss": 0.8989, + "step": 29786 + }, + { + "epoch": 0.95, + "learning_rate": 1.1758944882543921e-07, + "loss": 0.9741, + "step": 29787 + }, + { + "epoch": 0.95, + "learning_rate": 1.1743112336384698e-07, + "loss": 0.8716, + "step": 29788 + }, + { + "epoch": 0.95, + "learning_rate": 1.1727290393128033e-07, + "loss": 0.856, + "step": 29789 + }, + { + "epoch": 0.95, + "learning_rate": 1.1711479052943675e-07, + "loss": 0.9053, + "step": 29790 + }, + { + "epoch": 0.95, + "learning_rate": 1.1695678316001491e-07, + "loss": 0.9302, + "step": 29791 + }, + { + "epoch": 0.95, + "learning_rate": 1.1679888182470789e-07, + "loss": 0.8047, + "step": 29792 + }, + { + "epoch": 0.95, + "learning_rate": 1.1664108652521211e-07, + "loss": 0.8809, + "step": 29793 + }, + { + "epoch": 0.95, + "learning_rate": 1.1648339726321845e-07, + "loss": 0.833, + "step": 29794 + }, + { + "epoch": 0.95, + "learning_rate": 1.1632581404041887e-07, + "loss": 0.8657, + "step": 29795 + }, + { + "epoch": 0.95, + "learning_rate": 1.1616833685850537e-07, + "loss": 0.8823, + "step": 29796 + }, + { + "epoch": 0.95, + "learning_rate": 1.1601096571916659e-07, + "loss": 0.9092, + "step": 29797 + }, + { + "epoch": 0.95, + "learning_rate": 1.1585370062409007e-07, + "loss": 0.8379, + "step": 29798 + }, + { + "epoch": 0.95, + "learning_rate": 1.1569654157496447e-07, + "loss": 0.8828, + "step": 29799 + }, + { + "epoch": 0.95, + "learning_rate": 1.1553948857347619e-07, + "loss": 0.9277, + "step": 29800 + }, + { + "epoch": 0.95, + "learning_rate": 1.1538254162130946e-07, + "loss": 0.9121, + "step": 29801 + }, + { + "epoch": 0.95, + "learning_rate": 1.152257007201485e-07, + "loss": 0.9175, + "step": 29802 + }, + { + "epoch": 0.95, + "learning_rate": 1.1506896587167527e-07, + "loss": 0.8804, + "step": 29803 + }, + { + "epoch": 0.95, + "learning_rate": 1.1491233707757176e-07, + "loss": 1.0151, + "step": 29804 + }, + { + "epoch": 0.95, + "learning_rate": 1.1475581433951999e-07, + "loss": 0.7871, + "step": 29805 + }, + { + "epoch": 0.95, + "learning_rate": 1.1459939765919637e-07, + "loss": 0.8691, + "step": 29806 + }, + { + "epoch": 0.95, + "learning_rate": 1.1444308703828177e-07, + "loss": 0.9233, + "step": 29807 + }, + { + "epoch": 0.95, + "learning_rate": 1.1428688247845043e-07, + "loss": 0.9795, + "step": 29808 + }, + { + "epoch": 0.95, + "learning_rate": 1.1413078398138211e-07, + "loss": 0.938, + "step": 29809 + }, + { + "epoch": 0.95, + "learning_rate": 1.1397479154874768e-07, + "loss": 0.8823, + "step": 29810 + }, + { + "epoch": 0.95, + "learning_rate": 1.1381890518222361e-07, + "loss": 0.834, + "step": 29811 + }, + { + "epoch": 0.95, + "learning_rate": 1.1366312488347964e-07, + "loss": 0.7383, + "step": 29812 + }, + { + "epoch": 0.95, + "learning_rate": 1.1350745065419e-07, + "loss": 0.855, + "step": 29813 + }, + { + "epoch": 0.95, + "learning_rate": 1.1335188249602225e-07, + "loss": 0.8145, + "step": 29814 + }, + { + "epoch": 0.95, + "learning_rate": 1.1319642041064838e-07, + "loss": 0.8354, + "step": 29815 + }, + { + "epoch": 0.95, + "learning_rate": 1.1304106439973373e-07, + "loss": 0.8394, + "step": 29816 + }, + { + "epoch": 0.95, + "learning_rate": 1.1288581446494696e-07, + "loss": 0.9907, + "step": 29817 + }, + { + "epoch": 0.95, + "learning_rate": 1.1273067060795229e-07, + "loss": 0.9487, + "step": 29818 + }, + { + "epoch": 0.95, + "learning_rate": 1.1257563283041506e-07, + "loss": 0.9951, + "step": 29819 + }, + { + "epoch": 0.95, + "learning_rate": 1.1242070113399728e-07, + "loss": 0.9146, + "step": 29820 + }, + { + "epoch": 0.95, + "learning_rate": 1.1226587552036428e-07, + "loss": 0.9785, + "step": 29821 + }, + { + "epoch": 0.95, + "learning_rate": 1.1211115599117361e-07, + "loss": 0.8589, + "step": 29822 + }, + { + "epoch": 0.95, + "learning_rate": 1.119565425480873e-07, + "loss": 0.9702, + "step": 29823 + }, + { + "epoch": 0.95, + "learning_rate": 1.11802035192764e-07, + "loss": 0.7944, + "step": 29824 + }, + { + "epoch": 0.95, + "learning_rate": 1.1164763392686129e-07, + "loss": 0.894, + "step": 29825 + }, + { + "epoch": 0.95, + "learning_rate": 1.1149333875203561e-07, + "loss": 0.9277, + "step": 29826 + }, + { + "epoch": 0.95, + "learning_rate": 1.113391496699423e-07, + "loss": 0.9116, + "step": 29827 + }, + { + "epoch": 0.95, + "learning_rate": 1.1118506668223561e-07, + "loss": 0.9951, + "step": 29828 + }, + { + "epoch": 0.95, + "learning_rate": 1.1103108979056865e-07, + "loss": 0.939, + "step": 29829 + }, + { + "epoch": 0.95, + "learning_rate": 1.1087721899659343e-07, + "loss": 0.7847, + "step": 29830 + }, + { + "epoch": 0.95, + "learning_rate": 1.1072345430196085e-07, + "loss": 0.9387, + "step": 29831 + }, + { + "epoch": 0.95, + "learning_rate": 1.1056979570832183e-07, + "loss": 0.7847, + "step": 29832 + }, + { + "epoch": 0.95, + "learning_rate": 1.104162432173228e-07, + "loss": 0.8477, + "step": 29833 + }, + { + "epoch": 0.95, + "learning_rate": 1.1026279683061247e-07, + "loss": 0.9204, + "step": 29834 + }, + { + "epoch": 0.95, + "learning_rate": 1.1010945654983617e-07, + "loss": 0.9497, + "step": 29835 + }, + { + "epoch": 0.95, + "learning_rate": 1.0995622237664039e-07, + "loss": 0.8945, + "step": 29836 + }, + { + "epoch": 0.95, + "learning_rate": 1.0980309431266823e-07, + "loss": 0.894, + "step": 29837 + }, + { + "epoch": 0.95, + "learning_rate": 1.0965007235956393e-07, + "loss": 0.9814, + "step": 29838 + }, + { + "epoch": 0.95, + "learning_rate": 1.0949715651896731e-07, + "loss": 0.8103, + "step": 29839 + }, + { + "epoch": 0.95, + "learning_rate": 1.0934434679252037e-07, + "loss": 1.0044, + "step": 29840 + }, + { + "epoch": 0.95, + "learning_rate": 1.0919164318186182e-07, + "loss": 0.7979, + "step": 29841 + }, + { + "epoch": 0.95, + "learning_rate": 1.0903904568863144e-07, + "loss": 0.4817, + "step": 29842 + }, + { + "epoch": 0.95, + "learning_rate": 1.088865543144635e-07, + "loss": 0.8779, + "step": 29843 + }, + { + "epoch": 0.95, + "learning_rate": 1.0873416906099665e-07, + "loss": 1.0625, + "step": 29844 + }, + { + "epoch": 0.95, + "learning_rate": 1.0858188992986518e-07, + "loss": 0.9121, + "step": 29845 + }, + { + "epoch": 0.95, + "learning_rate": 1.0842971692270332e-07, + "loss": 0.9023, + "step": 29846 + }, + { + "epoch": 0.95, + "learning_rate": 1.082776500411431e-07, + "loss": 0.9136, + "step": 29847 + }, + { + "epoch": 0.95, + "learning_rate": 1.0812568928681544e-07, + "loss": 1.1167, + "step": 29848 + }, + { + "epoch": 0.95, + "learning_rate": 1.0797383466135125e-07, + "loss": 0.8994, + "step": 29849 + }, + { + "epoch": 0.95, + "learning_rate": 1.0782208616638034e-07, + "loss": 0.9009, + "step": 29850 + }, + { + "epoch": 0.95, + "learning_rate": 1.076704438035292e-07, + "loss": 0.752, + "step": 29851 + }, + { + "epoch": 0.95, + "learning_rate": 1.0751890757442763e-07, + "loss": 0.9072, + "step": 29852 + }, + { + "epoch": 0.95, + "learning_rate": 1.0736747748069876e-07, + "loss": 0.9253, + "step": 29853 + }, + { + "epoch": 0.95, + "learning_rate": 1.0721615352396907e-07, + "loss": 0.8833, + "step": 29854 + }, + { + "epoch": 0.95, + "learning_rate": 1.0706493570585952e-07, + "loss": 1.0176, + "step": 29855 + }, + { + "epoch": 0.95, + "learning_rate": 1.0691382402799655e-07, + "loss": 0.9473, + "step": 29856 + }, + { + "epoch": 0.95, + "learning_rate": 1.0676281849199665e-07, + "loss": 0.9155, + "step": 29857 + }, + { + "epoch": 0.95, + "learning_rate": 1.066119190994841e-07, + "loss": 0.9551, + "step": 29858 + }, + { + "epoch": 0.95, + "learning_rate": 1.0646112585207535e-07, + "loss": 0.9268, + "step": 29859 + }, + { + "epoch": 0.95, + "learning_rate": 1.0631043875139024e-07, + "loss": 0.9231, + "step": 29860 + }, + { + "epoch": 0.95, + "learning_rate": 1.0615985779904303e-07, + "loss": 1.0791, + "step": 29861 + }, + { + "epoch": 0.96, + "learning_rate": 1.0600938299665243e-07, + "loss": 1.0547, + "step": 29862 + }, + { + "epoch": 0.96, + "learning_rate": 1.0585901434582934e-07, + "loss": 0.9097, + "step": 29863 + }, + { + "epoch": 0.96, + "learning_rate": 1.0570875184818918e-07, + "loss": 0.9351, + "step": 29864 + }, + { + "epoch": 0.96, + "learning_rate": 1.0555859550534286e-07, + "loss": 0.9067, + "step": 29865 + }, + { + "epoch": 0.96, + "learning_rate": 1.0540854531890354e-07, + "loss": 0.8975, + "step": 29866 + }, + { + "epoch": 0.96, + "learning_rate": 1.0525860129047882e-07, + "loss": 0.8765, + "step": 29867 + }, + { + "epoch": 0.96, + "learning_rate": 1.0510876342167852e-07, + "loss": 0.8906, + "step": 29868 + }, + { + "epoch": 0.96, + "learning_rate": 1.0495903171411026e-07, + "loss": 0.8801, + "step": 29869 + }, + { + "epoch": 0.96, + "learning_rate": 1.0480940616938052e-07, + "loss": 0.9512, + "step": 29870 + }, + { + "epoch": 0.96, + "learning_rate": 1.0465988678909467e-07, + "loss": 0.8677, + "step": 29871 + }, + { + "epoch": 0.96, + "learning_rate": 1.0451047357485589e-07, + "loss": 0.9761, + "step": 29872 + }, + { + "epoch": 0.96, + "learning_rate": 1.0436116652826844e-07, + "loss": 0.8628, + "step": 29873 + }, + { + "epoch": 0.96, + "learning_rate": 1.0421196565093217e-07, + "loss": 0.8245, + "step": 29874 + }, + { + "epoch": 0.96, + "learning_rate": 1.0406287094445133e-07, + "loss": 0.9839, + "step": 29875 + }, + { + "epoch": 0.96, + "learning_rate": 1.0391388241042244e-07, + "loss": 1.0161, + "step": 29876 + }, + { + "epoch": 0.96, + "learning_rate": 1.0376500005044643e-07, + "loss": 1.022, + "step": 29877 + }, + { + "epoch": 0.96, + "learning_rate": 1.036162238661187e-07, + "loss": 0.8984, + "step": 29878 + }, + { + "epoch": 0.96, + "learning_rate": 1.0346755385903573e-07, + "loss": 0.9741, + "step": 29879 + }, + { + "epoch": 0.96, + "learning_rate": 1.0331899003079293e-07, + "loss": 0.9224, + "step": 29880 + }, + { + "epoch": 0.96, + "learning_rate": 1.0317053238298458e-07, + "loss": 0.9185, + "step": 29881 + }, + { + "epoch": 0.96, + "learning_rate": 1.0302218091720273e-07, + "loss": 0.8196, + "step": 29882 + }, + { + "epoch": 0.96, + "learning_rate": 1.0287393563503945e-07, + "loss": 0.9438, + "step": 29883 + }, + { + "epoch": 0.96, + "learning_rate": 1.0272579653808679e-07, + "loss": 0.833, + "step": 29884 + }, + { + "epoch": 0.96, + "learning_rate": 1.0257776362793126e-07, + "loss": 0.9238, + "step": 29885 + }, + { + "epoch": 0.96, + "learning_rate": 1.0242983690616159e-07, + "loss": 0.8901, + "step": 29886 + }, + { + "epoch": 0.96, + "learning_rate": 1.0228201637436763e-07, + "loss": 0.8628, + "step": 29887 + }, + { + "epoch": 0.96, + "learning_rate": 1.0213430203413143e-07, + "loss": 0.8765, + "step": 29888 + }, + { + "epoch": 0.96, + "learning_rate": 1.0198669388704064e-07, + "loss": 0.978, + "step": 29889 + }, + { + "epoch": 0.96, + "learning_rate": 1.0183919193467839e-07, + "loss": 0.9248, + "step": 29890 + }, + { + "epoch": 0.96, + "learning_rate": 1.0169179617862679e-07, + "loss": 0.8779, + "step": 29891 + }, + { + "epoch": 0.96, + "learning_rate": 1.0154450662046677e-07, + "loss": 0.8289, + "step": 29892 + }, + { + "epoch": 0.96, + "learning_rate": 1.0139732326178043e-07, + "loss": 0.8403, + "step": 29893 + }, + { + "epoch": 0.96, + "learning_rate": 1.0125024610414425e-07, + "loss": 1.0493, + "step": 29894 + }, + { + "epoch": 0.96, + "learning_rate": 1.0110327514913809e-07, + "loss": 0.9258, + "step": 29895 + }, + { + "epoch": 0.96, + "learning_rate": 1.0095641039833737e-07, + "loss": 0.9355, + "step": 29896 + }, + { + "epoch": 0.96, + "learning_rate": 1.008096518533197e-07, + "loss": 0.8555, + "step": 29897 + }, + { + "epoch": 0.96, + "learning_rate": 1.0066299951565717e-07, + "loss": 0.9722, + "step": 29898 + }, + { + "epoch": 0.96, + "learning_rate": 1.0051645338692739e-07, + "loss": 0.9468, + "step": 29899 + }, + { + "epoch": 0.96, + "learning_rate": 1.003700134686969e-07, + "loss": 0.9604, + "step": 29900 + }, + { + "epoch": 0.96, + "learning_rate": 1.002236797625411e-07, + "loss": 0.8726, + "step": 29901 + }, + { + "epoch": 0.96, + "learning_rate": 1.0007745227002875e-07, + "loss": 0.8643, + "step": 29902 + }, + { + "epoch": 0.96, + "learning_rate": 9.993133099272856e-08, + "loss": 0.9438, + "step": 29903 + }, + { + "epoch": 0.96, + "learning_rate": 9.978531593220819e-08, + "loss": 0.9053, + "step": 29904 + }, + { + "epoch": 0.96, + "learning_rate": 9.963940709003417e-08, + "loss": 0.9429, + "step": 29905 + }, + { + "epoch": 0.96, + "learning_rate": 9.9493604467773e-08, + "loss": 1.0479, + "step": 29906 + }, + { + "epoch": 0.96, + "learning_rate": 9.934790806698791e-08, + "loss": 0.8125, + "step": 29907 + }, + { + "epoch": 0.96, + "learning_rate": 9.920231788924095e-08, + "loss": 0.998, + "step": 29908 + }, + { + "epoch": 0.96, + "learning_rate": 9.905683393609755e-08, + "loss": 0.9624, + "step": 29909 + }, + { + "epoch": 0.96, + "learning_rate": 9.891145620911424e-08, + "loss": 0.793, + "step": 29910 + }, + { + "epoch": 0.96, + "learning_rate": 9.876618470985422e-08, + "loss": 0.8618, + "step": 29911 + }, + { + "epoch": 0.96, + "learning_rate": 9.862101943987512e-08, + "loss": 0.9546, + "step": 29912 + }, + { + "epoch": 0.96, + "learning_rate": 9.847596040073349e-08, + "loss": 0.9292, + "step": 29913 + }, + { + "epoch": 0.96, + "learning_rate": 9.833100759398694e-08, + "loss": 0.8809, + "step": 29914 + }, + { + "epoch": 0.96, + "learning_rate": 9.818616102118983e-08, + "loss": 0.9292, + "step": 29915 + }, + { + "epoch": 0.96, + "learning_rate": 9.804142068389755e-08, + "loss": 0.9502, + "step": 29916 + }, + { + "epoch": 0.96, + "learning_rate": 9.789678658365997e-08, + "loss": 0.9717, + "step": 29917 + }, + { + "epoch": 0.96, + "learning_rate": 9.775225872203142e-08, + "loss": 0.8779, + "step": 29918 + }, + { + "epoch": 0.96, + "learning_rate": 9.760783710056176e-08, + "loss": 1.002, + "step": 29919 + }, + { + "epoch": 0.96, + "learning_rate": 9.746352172080087e-08, + "loss": 1.0259, + "step": 29920 + }, + { + "epoch": 0.96, + "learning_rate": 9.731931258429638e-08, + "loss": 1.0088, + "step": 29921 + }, + { + "epoch": 0.96, + "learning_rate": 9.717520969259708e-08, + "loss": 0.9229, + "step": 29922 + }, + { + "epoch": 0.96, + "learning_rate": 9.703121304724727e-08, + "loss": 0.9497, + "step": 29923 + }, + { + "epoch": 0.96, + "learning_rate": 9.68873226497924e-08, + "loss": 0.9614, + "step": 29924 + }, + { + "epoch": 0.96, + "learning_rate": 9.674353850177565e-08, + "loss": 0.8799, + "step": 29925 + }, + { + "epoch": 0.96, + "learning_rate": 9.659986060474136e-08, + "loss": 0.9546, + "step": 29926 + }, + { + "epoch": 0.96, + "learning_rate": 9.645628896022941e-08, + "loss": 0.9736, + "step": 29927 + }, + { + "epoch": 0.96, + "learning_rate": 9.631282356978078e-08, + "loss": 0.8384, + "step": 29928 + }, + { + "epoch": 0.96, + "learning_rate": 9.616946443493425e-08, + "loss": 0.8877, + "step": 29929 + }, + { + "epoch": 0.96, + "learning_rate": 9.60262115572308e-08, + "loss": 1.0654, + "step": 29930 + }, + { + "epoch": 0.96, + "learning_rate": 9.588306493820254e-08, + "loss": 0.8975, + "step": 29931 + }, + { + "epoch": 0.96, + "learning_rate": 9.574002457938825e-08, + "loss": 1.0054, + "step": 29932 + }, + { + "epoch": 0.96, + "learning_rate": 9.559709048232113e-08, + "loss": 0.8594, + "step": 29933 + }, + { + "epoch": 0.96, + "learning_rate": 9.545426264853664e-08, + "loss": 0.9116, + "step": 29934 + }, + { + "epoch": 0.96, + "learning_rate": 9.531154107956575e-08, + "loss": 0.8755, + "step": 29935 + }, + { + "epoch": 0.96, + "learning_rate": 9.516892577694059e-08, + "loss": 0.8953, + "step": 29936 + }, + { + "epoch": 0.96, + "learning_rate": 9.502641674219104e-08, + "loss": 0.8677, + "step": 29937 + }, + { + "epoch": 0.96, + "learning_rate": 9.488401397684477e-08, + "loss": 0.8389, + "step": 29938 + }, + { + "epoch": 0.96, + "learning_rate": 9.474171748243055e-08, + "loss": 0.8628, + "step": 29939 + }, + { + "epoch": 0.96, + "learning_rate": 9.459952726047606e-08, + "loss": 0.9346, + "step": 29940 + }, + { + "epoch": 0.96, + "learning_rate": 9.445744331250451e-08, + "loss": 0.8394, + "step": 29941 + }, + { + "epoch": 0.96, + "learning_rate": 9.431546564004246e-08, + "loss": 1.0171, + "step": 29942 + }, + { + "epoch": 0.96, + "learning_rate": 9.417359424461203e-08, + "loss": 0.9951, + "step": 29943 + }, + { + "epoch": 0.96, + "learning_rate": 9.403182912773645e-08, + "loss": 0.8892, + "step": 29944 + }, + { + "epoch": 0.96, + "learning_rate": 9.389017029093562e-08, + "loss": 0.9531, + "step": 29945 + }, + { + "epoch": 0.96, + "learning_rate": 9.374861773573052e-08, + "loss": 0.8799, + "step": 29946 + }, + { + "epoch": 0.96, + "learning_rate": 9.360717146363773e-08, + "loss": 1.1143, + "step": 29947 + }, + { + "epoch": 0.96, + "learning_rate": 9.346583147617606e-08, + "loss": 0.9756, + "step": 29948 + }, + { + "epoch": 0.96, + "learning_rate": 9.332459777486314e-08, + "loss": 0.9263, + "step": 29949 + }, + { + "epoch": 0.96, + "learning_rate": 9.318347036121223e-08, + "loss": 0.9209, + "step": 29950 + }, + { + "epoch": 0.96, + "learning_rate": 9.304244923673877e-08, + "loss": 0.4761, + "step": 29951 + }, + { + "epoch": 0.96, + "learning_rate": 9.290153440295602e-08, + "loss": 1.0195, + "step": 29952 + }, + { + "epoch": 0.96, + "learning_rate": 9.276072586137385e-08, + "loss": 0.9688, + "step": 29953 + }, + { + "epoch": 0.96, + "learning_rate": 9.262002361350552e-08, + "loss": 0.9028, + "step": 29954 + }, + { + "epoch": 0.96, + "learning_rate": 9.247942766085871e-08, + "loss": 0.8789, + "step": 29955 + }, + { + "epoch": 0.96, + "learning_rate": 9.23389380049422e-08, + "loss": 0.8584, + "step": 29956 + }, + { + "epoch": 0.96, + "learning_rate": 9.219855464726368e-08, + "loss": 0.9258, + "step": 29957 + }, + { + "epoch": 0.96, + "learning_rate": 9.205827758932973e-08, + "loss": 0.9248, + "step": 29958 + }, + { + "epoch": 0.96, + "learning_rate": 9.191810683264468e-08, + "loss": 0.8496, + "step": 29959 + }, + { + "epoch": 0.96, + "learning_rate": 9.17780423787129e-08, + "loss": 0.9922, + "step": 29960 + }, + { + "epoch": 0.96, + "learning_rate": 9.163808422903653e-08, + "loss": 0.9045, + "step": 29961 + }, + { + "epoch": 0.96, + "learning_rate": 9.149823238511768e-08, + "loss": 0.8574, + "step": 29962 + }, + { + "epoch": 0.96, + "learning_rate": 9.135848684845627e-08, + "loss": 0.9116, + "step": 29963 + }, + { + "epoch": 0.96, + "learning_rate": 9.121884762055222e-08, + "loss": 0.9707, + "step": 29964 + }, + { + "epoch": 0.96, + "learning_rate": 9.107931470290322e-08, + "loss": 0.9673, + "step": 29965 + }, + { + "epoch": 0.96, + "learning_rate": 9.093988809700693e-08, + "loss": 0.8984, + "step": 29966 + }, + { + "epoch": 0.96, + "learning_rate": 9.080056780435887e-08, + "loss": 0.8892, + "step": 29967 + }, + { + "epoch": 0.96, + "learning_rate": 9.066135382645336e-08, + "loss": 0.937, + "step": 29968 + }, + { + "epoch": 0.96, + "learning_rate": 9.052224616478478e-08, + "loss": 0.8984, + "step": 29969 + }, + { + "epoch": 0.96, + "learning_rate": 9.038324482084526e-08, + "loss": 0.9783, + "step": 29970 + }, + { + "epoch": 0.96, + "learning_rate": 9.024434979612584e-08, + "loss": 0.8721, + "step": 29971 + }, + { + "epoch": 0.96, + "learning_rate": 9.010556109211643e-08, + "loss": 0.9258, + "step": 29972 + }, + { + "epoch": 0.96, + "learning_rate": 8.996687871030696e-08, + "loss": 0.9854, + "step": 29973 + }, + { + "epoch": 0.96, + "learning_rate": 8.982830265218511e-08, + "loss": 0.8857, + "step": 29974 + }, + { + "epoch": 0.96, + "learning_rate": 8.96898329192386e-08, + "loss": 0.978, + "step": 29975 + }, + { + "epoch": 0.96, + "learning_rate": 8.955146951295069e-08, + "loss": 0.832, + "step": 29976 + }, + { + "epoch": 0.96, + "learning_rate": 8.941321243480794e-08, + "loss": 0.9131, + "step": 29977 + }, + { + "epoch": 0.96, + "learning_rate": 8.927506168629252e-08, + "loss": 0.9248, + "step": 29978 + }, + { + "epoch": 0.96, + "learning_rate": 8.913701726888657e-08, + "loss": 0.9971, + "step": 29979 + }, + { + "epoch": 0.96, + "learning_rate": 8.899907918407225e-08, + "loss": 0.8442, + "step": 29980 + }, + { + "epoch": 0.96, + "learning_rate": 8.886124743332836e-08, + "loss": 0.9956, + "step": 29981 + }, + { + "epoch": 0.96, + "learning_rate": 8.872352201813484e-08, + "loss": 0.9253, + "step": 29982 + }, + { + "epoch": 0.96, + "learning_rate": 8.858590293996828e-08, + "loss": 0.8687, + "step": 29983 + }, + { + "epoch": 0.96, + "learning_rate": 8.844839020030526e-08, + "loss": 0.9307, + "step": 29984 + }, + { + "epoch": 0.96, + "learning_rate": 8.831098380062242e-08, + "loss": 1.043, + "step": 29985 + }, + { + "epoch": 0.96, + "learning_rate": 8.817368374239188e-08, + "loss": 0.8994, + "step": 29986 + }, + { + "epoch": 0.96, + "learning_rate": 8.803649002708914e-08, + "loss": 0.8599, + "step": 29987 + }, + { + "epoch": 0.96, + "learning_rate": 8.789940265618302e-08, + "loss": 0.8599, + "step": 29988 + }, + { + "epoch": 0.96, + "learning_rate": 8.77624216311479e-08, + "loss": 0.9946, + "step": 29989 + }, + { + "epoch": 0.96, + "learning_rate": 8.76255469534515e-08, + "loss": 0.9087, + "step": 29990 + }, + { + "epoch": 0.96, + "learning_rate": 8.748877862456152e-08, + "loss": 0.9277, + "step": 29991 + }, + { + "epoch": 0.96, + "learning_rate": 8.735211664594678e-08, + "loss": 0.9375, + "step": 29992 + }, + { + "epoch": 0.96, + "learning_rate": 8.72155610190728e-08, + "loss": 0.8633, + "step": 29993 + }, + { + "epoch": 0.96, + "learning_rate": 8.707911174540617e-08, + "loss": 0.9751, + "step": 29994 + }, + { + "epoch": 0.96, + "learning_rate": 8.694276882640795e-08, + "loss": 0.957, + "step": 29995 + }, + { + "epoch": 0.96, + "learning_rate": 8.680653226354363e-08, + "loss": 0.8577, + "step": 29996 + }, + { + "epoch": 0.96, + "learning_rate": 8.667040205827315e-08, + "loss": 0.9365, + "step": 29997 + }, + { + "epoch": 0.96, + "learning_rate": 8.653437821205868e-08, + "loss": 0.8398, + "step": 29998 + }, + { + "epoch": 0.96, + "learning_rate": 8.639846072635794e-08, + "loss": 0.9502, + "step": 29999 + }, + { + "epoch": 0.96, + "learning_rate": 8.626264960262975e-08, + "loss": 0.9316, + "step": 30000 + }, + { + "epoch": 0.96, + "learning_rate": 8.612694484233186e-08, + "loss": 0.917, + "step": 30001 + }, + { + "epoch": 0.96, + "learning_rate": 8.599134644691976e-08, + "loss": 0.8867, + "step": 30002 + }, + { + "epoch": 0.96, + "learning_rate": 8.585585441784672e-08, + "loss": 0.8921, + "step": 30003 + }, + { + "epoch": 0.96, + "learning_rate": 8.572046875656936e-08, + "loss": 0.8667, + "step": 30004 + }, + { + "epoch": 0.96, + "learning_rate": 8.558518946453875e-08, + "loss": 0.9814, + "step": 30005 + }, + { + "epoch": 0.96, + "learning_rate": 8.545001654320707e-08, + "loss": 0.9009, + "step": 30006 + }, + { + "epoch": 0.96, + "learning_rate": 8.531494999402201e-08, + "loss": 0.9077, + "step": 30007 + }, + { + "epoch": 0.96, + "learning_rate": 8.517998981843578e-08, + "loss": 0.9351, + "step": 30008 + }, + { + "epoch": 0.96, + "learning_rate": 8.504513601789388e-08, + "loss": 0.9233, + "step": 30009 + }, + { + "epoch": 0.96, + "learning_rate": 8.491038859384515e-08, + "loss": 0.918, + "step": 30010 + }, + { + "epoch": 0.96, + "learning_rate": 8.47757475477351e-08, + "loss": 0.9717, + "step": 30011 + }, + { + "epoch": 0.96, + "learning_rate": 8.464121288100701e-08, + "loss": 0.9229, + "step": 30012 + }, + { + "epoch": 0.96, + "learning_rate": 8.450678459510531e-08, + "loss": 0.9702, + "step": 30013 + }, + { + "epoch": 0.96, + "learning_rate": 8.437246269147104e-08, + "loss": 0.9282, + "step": 30014 + }, + { + "epoch": 0.96, + "learning_rate": 8.423824717154638e-08, + "loss": 0.9185, + "step": 30015 + }, + { + "epoch": 0.96, + "learning_rate": 8.410413803677131e-08, + "loss": 0.9536, + "step": 30016 + }, + { + "epoch": 0.96, + "learning_rate": 8.397013528858466e-08, + "loss": 0.9507, + "step": 30017 + }, + { + "epoch": 0.96, + "learning_rate": 8.383623892842418e-08, + "loss": 0.9868, + "step": 30018 + }, + { + "epoch": 0.96, + "learning_rate": 8.370244895772538e-08, + "loss": 0.9204, + "step": 30019 + }, + { + "epoch": 0.96, + "learning_rate": 8.356876537792602e-08, + "loss": 0.9214, + "step": 30020 + }, + { + "epoch": 0.96, + "learning_rate": 8.343518819045714e-08, + "loss": 0.9043, + "step": 30021 + }, + { + "epoch": 0.96, + "learning_rate": 8.33017173967543e-08, + "loss": 0.9722, + "step": 30022 + }, + { + "epoch": 0.96, + "learning_rate": 8.316835299824855e-08, + "loss": 0.9243, + "step": 30023 + }, + { + "epoch": 0.96, + "learning_rate": 8.303509499637097e-08, + "loss": 0.9668, + "step": 30024 + }, + { + "epoch": 0.96, + "learning_rate": 8.290194339255153e-08, + "loss": 0.8647, + "step": 30025 + }, + { + "epoch": 0.96, + "learning_rate": 8.276889818821798e-08, + "loss": 0.9243, + "step": 30026 + }, + { + "epoch": 0.96, + "learning_rate": 8.263595938479806e-08, + "loss": 1.0049, + "step": 30027 + }, + { + "epoch": 0.96, + "learning_rate": 8.250312698372065e-08, + "loss": 0.8921, + "step": 30028 + }, + { + "epoch": 0.96, + "learning_rate": 8.237040098640569e-08, + "loss": 0.8818, + "step": 30029 + }, + { + "epoch": 0.96, + "learning_rate": 8.223778139428207e-08, + "loss": 0.8853, + "step": 30030 + }, + { + "epoch": 0.96, + "learning_rate": 8.210526820876863e-08, + "loss": 0.4771, + "step": 30031 + }, + { + "epoch": 0.96, + "learning_rate": 8.19728614312909e-08, + "loss": 0.9995, + "step": 30032 + }, + { + "epoch": 0.96, + "learning_rate": 8.184056106326666e-08, + "loss": 0.8159, + "step": 30033 + }, + { + "epoch": 0.96, + "learning_rate": 8.170836710611696e-08, + "loss": 0.9292, + "step": 30034 + }, + { + "epoch": 0.96, + "learning_rate": 8.157627956125847e-08, + "loss": 0.9214, + "step": 30035 + }, + { + "epoch": 0.96, + "learning_rate": 8.144429843011115e-08, + "loss": 0.8223, + "step": 30036 + }, + { + "epoch": 0.96, + "learning_rate": 8.13124237140872e-08, + "loss": 0.4504, + "step": 30037 + }, + { + "epoch": 0.96, + "learning_rate": 8.11806554146044e-08, + "loss": 0.9087, + "step": 30038 + }, + { + "epoch": 0.96, + "learning_rate": 8.104899353307604e-08, + "loss": 1.0552, + "step": 30039 + }, + { + "epoch": 0.96, + "learning_rate": 8.091743807091324e-08, + "loss": 1.0, + "step": 30040 + }, + { + "epoch": 0.96, + "learning_rate": 8.078598902952928e-08, + "loss": 0.8467, + "step": 30041 + }, + { + "epoch": 0.96, + "learning_rate": 8.065464641033193e-08, + "loss": 0.9497, + "step": 30042 + }, + { + "epoch": 0.96, + "learning_rate": 8.052341021473342e-08, + "loss": 0.4958, + "step": 30043 + }, + { + "epoch": 0.96, + "learning_rate": 8.039228044414038e-08, + "loss": 0.9194, + "step": 30044 + }, + { + "epoch": 0.96, + "learning_rate": 8.026125709996058e-08, + "loss": 1.001, + "step": 30045 + }, + { + "epoch": 0.96, + "learning_rate": 8.013034018359734e-08, + "loss": 0.8225, + "step": 30046 + }, + { + "epoch": 0.96, + "learning_rate": 7.99995296964573e-08, + "loss": 0.8027, + "step": 30047 + }, + { + "epoch": 0.96, + "learning_rate": 7.98688256399427e-08, + "loss": 0.9858, + "step": 30048 + }, + { + "epoch": 0.96, + "learning_rate": 7.973822801545794e-08, + "loss": 0.9236, + "step": 30049 + }, + { + "epoch": 0.96, + "learning_rate": 7.960773682440192e-08, + "loss": 0.8486, + "step": 30050 + }, + { + "epoch": 0.96, + "learning_rate": 7.947735206817575e-08, + "loss": 0.8931, + "step": 30051 + }, + { + "epoch": 0.96, + "learning_rate": 7.934707374817829e-08, + "loss": 0.8838, + "step": 30052 + }, + { + "epoch": 0.96, + "learning_rate": 7.921690186580733e-08, + "loss": 0.9561, + "step": 30053 + }, + { + "epoch": 0.96, + "learning_rate": 7.90868364224584e-08, + "loss": 0.9961, + "step": 30054 + }, + { + "epoch": 0.96, + "learning_rate": 7.895687741952818e-08, + "loss": 0.9009, + "step": 30055 + }, + { + "epoch": 0.96, + "learning_rate": 7.882702485840998e-08, + "loss": 0.9775, + "step": 30056 + }, + { + "epoch": 0.96, + "learning_rate": 7.869727874049826e-08, + "loss": 0.8303, + "step": 30057 + }, + { + "epoch": 0.96, + "learning_rate": 7.856763906718412e-08, + "loss": 0.8403, + "step": 30058 + }, + { + "epoch": 0.96, + "learning_rate": 7.843810583985978e-08, + "loss": 0.9771, + "step": 30059 + }, + { + "epoch": 0.96, + "learning_rate": 7.830867905991301e-08, + "loss": 0.916, + "step": 30060 + }, + { + "epoch": 0.96, + "learning_rate": 7.817935872873273e-08, + "loss": 0.8423, + "step": 30061 + }, + { + "epoch": 0.96, + "learning_rate": 7.805014484770779e-08, + "loss": 0.8955, + "step": 30062 + }, + { + "epoch": 0.96, + "learning_rate": 7.792103741822376e-08, + "loss": 0.9932, + "step": 30063 + }, + { + "epoch": 0.96, + "learning_rate": 7.77920364416651e-08, + "loss": 0.9058, + "step": 30064 + }, + { + "epoch": 0.96, + "learning_rate": 7.766314191941737e-08, + "loss": 0.9907, + "step": 30065 + }, + { + "epoch": 0.96, + "learning_rate": 7.753435385286279e-08, + "loss": 0.8862, + "step": 30066 + }, + { + "epoch": 0.96, + "learning_rate": 7.740567224338246e-08, + "loss": 0.8853, + "step": 30067 + }, + { + "epoch": 0.96, + "learning_rate": 7.727709709235643e-08, + "loss": 0.8872, + "step": 30068 + }, + { + "epoch": 0.96, + "learning_rate": 7.714862840116689e-08, + "loss": 0.9565, + "step": 30069 + }, + { + "epoch": 0.96, + "learning_rate": 7.70202661711894e-08, + "loss": 0.8599, + "step": 30070 + }, + { + "epoch": 0.96, + "learning_rate": 7.68920104038029e-08, + "loss": 0.856, + "step": 30071 + }, + { + "epoch": 0.96, + "learning_rate": 7.676386110038292e-08, + "loss": 0.7515, + "step": 30072 + }, + { + "epoch": 0.96, + "learning_rate": 7.663581826230394e-08, + "loss": 0.8916, + "step": 30073 + }, + { + "epoch": 0.96, + "learning_rate": 7.650788189093927e-08, + "loss": 1.0679, + "step": 30074 + }, + { + "epoch": 0.96, + "learning_rate": 7.63800519876634e-08, + "loss": 0.874, + "step": 30075 + }, + { + "epoch": 0.96, + "learning_rate": 7.625232855384635e-08, + "loss": 1.0718, + "step": 30076 + }, + { + "epoch": 0.96, + "learning_rate": 7.61247115908581e-08, + "loss": 0.9434, + "step": 30077 + }, + { + "epoch": 0.96, + "learning_rate": 7.599720110006981e-08, + "loss": 0.8716, + "step": 30078 + }, + { + "epoch": 0.96, + "learning_rate": 7.586979708284703e-08, + "loss": 0.8691, + "step": 30079 + }, + { + "epoch": 0.96, + "learning_rate": 7.574249954055868e-08, + "loss": 0.8711, + "step": 30080 + }, + { + "epoch": 0.96, + "learning_rate": 7.561530847456921e-08, + "loss": 0.7798, + "step": 30081 + }, + { + "epoch": 0.96, + "learning_rate": 7.54882238862431e-08, + "loss": 0.916, + "step": 30082 + }, + { + "epoch": 0.96, + "learning_rate": 7.536124577694482e-08, + "loss": 0.7832, + "step": 30083 + }, + { + "epoch": 0.96, + "learning_rate": 7.523437414803769e-08, + "loss": 0.8647, + "step": 30084 + }, + { + "epoch": 0.96, + "learning_rate": 7.510760900087955e-08, + "loss": 0.9033, + "step": 30085 + }, + { + "epoch": 0.96, + "learning_rate": 7.498095033683372e-08, + "loss": 0.9077, + "step": 30086 + }, + { + "epoch": 0.96, + "learning_rate": 7.485439815725693e-08, + "loss": 0.8789, + "step": 30087 + }, + { + "epoch": 0.96, + "learning_rate": 7.472795246350916e-08, + "loss": 0.9556, + "step": 30088 + }, + { + "epoch": 0.96, + "learning_rate": 7.460161325694381e-08, + "loss": 0.8516, + "step": 30089 + }, + { + "epoch": 0.96, + "learning_rate": 7.447538053891979e-08, + "loss": 0.9268, + "step": 30090 + }, + { + "epoch": 0.96, + "learning_rate": 7.434925431078933e-08, + "loss": 1.0063, + "step": 30091 + }, + { + "epoch": 0.96, + "learning_rate": 7.42232345739069e-08, + "loss": 0.9023, + "step": 30092 + }, + { + "epoch": 0.96, + "learning_rate": 7.409732132962255e-08, + "loss": 0.9834, + "step": 30093 + }, + { + "epoch": 0.96, + "learning_rate": 7.397151457928964e-08, + "loss": 0.9429, + "step": 30094 + }, + { + "epoch": 0.96, + "learning_rate": 7.384581432425598e-08, + "loss": 0.8857, + "step": 30095 + }, + { + "epoch": 0.96, + "learning_rate": 7.372022056587158e-08, + "loss": 0.9561, + "step": 30096 + }, + { + "epoch": 0.96, + "learning_rate": 7.359473330548316e-08, + "loss": 0.9688, + "step": 30097 + }, + { + "epoch": 0.96, + "learning_rate": 7.346935254443743e-08, + "loss": 0.835, + "step": 30098 + }, + { + "epoch": 0.96, + "learning_rate": 7.334407828407885e-08, + "loss": 0.9473, + "step": 30099 + }, + { + "epoch": 0.96, + "learning_rate": 7.321891052575303e-08, + "loss": 0.9116, + "step": 30100 + }, + { + "epoch": 0.96, + "learning_rate": 7.309384927080109e-08, + "loss": 0.9048, + "step": 30101 + }, + { + "epoch": 0.96, + "learning_rate": 7.296889452056532e-08, + "loss": 0.8765, + "step": 30102 + }, + { + "epoch": 0.96, + "learning_rate": 7.284404627638797e-08, + "loss": 0.9927, + "step": 30103 + }, + { + "epoch": 0.96, + "learning_rate": 7.271930453960574e-08, + "loss": 0.8872, + "step": 30104 + }, + { + "epoch": 0.96, + "learning_rate": 7.259466931155868e-08, + "loss": 0.9453, + "step": 30105 + }, + { + "epoch": 0.96, + "learning_rate": 7.247014059358459e-08, + "loss": 0.9819, + "step": 30106 + }, + { + "epoch": 0.96, + "learning_rate": 7.234571838701687e-08, + "loss": 0.9658, + "step": 30107 + }, + { + "epoch": 0.96, + "learning_rate": 7.222140269319444e-08, + "loss": 0.9346, + "step": 30108 + }, + { + "epoch": 0.96, + "learning_rate": 7.209719351344735e-08, + "loss": 0.9404, + "step": 30109 + }, + { + "epoch": 0.96, + "learning_rate": 7.197309084911008e-08, + "loss": 0.8994, + "step": 30110 + }, + { + "epoch": 0.96, + "learning_rate": 7.18490947015138e-08, + "loss": 0.9561, + "step": 30111 + }, + { + "epoch": 0.96, + "learning_rate": 7.172520507198965e-08, + "loss": 0.8901, + "step": 30112 + }, + { + "epoch": 0.96, + "learning_rate": 7.160142196186437e-08, + "loss": 0.8516, + "step": 30113 + }, + { + "epoch": 0.96, + "learning_rate": 7.147774537246798e-08, + "loss": 0.9268, + "step": 30114 + }, + { + "epoch": 0.96, + "learning_rate": 7.135417530512723e-08, + "loss": 1.0269, + "step": 30115 + }, + { + "epoch": 0.96, + "learning_rate": 7.12307117611688e-08, + "loss": 0.8721, + "step": 30116 + }, + { + "epoch": 0.96, + "learning_rate": 7.1107354741915e-08, + "loss": 0.938, + "step": 30117 + }, + { + "epoch": 0.96, + "learning_rate": 7.09841042486914e-08, + "loss": 0.8862, + "step": 30118 + }, + { + "epoch": 0.96, + "learning_rate": 7.086096028281808e-08, + "loss": 0.9282, + "step": 30119 + }, + { + "epoch": 0.96, + "learning_rate": 7.073792284561953e-08, + "loss": 1.0171, + "step": 30120 + }, + { + "epoch": 0.96, + "learning_rate": 7.061499193841248e-08, + "loss": 0.9722, + "step": 30121 + }, + { + "epoch": 0.96, + "learning_rate": 7.049216756251698e-08, + "loss": 0.9106, + "step": 30122 + }, + { + "epoch": 0.96, + "learning_rate": 7.036944971925197e-08, + "loss": 0.9541, + "step": 30123 + }, + { + "epoch": 0.96, + "learning_rate": 7.024683840993307e-08, + "loss": 0.9023, + "step": 30124 + }, + { + "epoch": 0.96, + "learning_rate": 7.01243336358759e-08, + "loss": 0.854, + "step": 30125 + }, + { + "epoch": 0.96, + "learning_rate": 7.000193539839495e-08, + "loss": 0.7568, + "step": 30126 + }, + { + "epoch": 0.96, + "learning_rate": 6.987964369880252e-08, + "loss": 0.467, + "step": 30127 + }, + { + "epoch": 0.96, + "learning_rate": 6.975745853841198e-08, + "loss": 0.96, + "step": 30128 + }, + { + "epoch": 0.96, + "learning_rate": 6.963537991853342e-08, + "loss": 0.8965, + "step": 30129 + }, + { + "epoch": 0.96, + "learning_rate": 6.951340784047689e-08, + "loss": 0.9502, + "step": 30130 + }, + { + "epoch": 0.96, + "learning_rate": 6.939154230555134e-08, + "loss": 0.9707, + "step": 30131 + }, + { + "epoch": 0.96, + "learning_rate": 6.92697833150635e-08, + "loss": 1.0117, + "step": 30132 + }, + { + "epoch": 0.96, + "learning_rate": 6.914813087032124e-08, + "loss": 0.9922, + "step": 30133 + }, + { + "epoch": 0.96, + "learning_rate": 6.902658497262682e-08, + "loss": 0.9688, + "step": 30134 + }, + { + "epoch": 0.96, + "learning_rate": 6.890514562328699e-08, + "loss": 1.0, + "step": 30135 + }, + { + "epoch": 0.96, + "learning_rate": 6.878381282360291e-08, + "loss": 0.9043, + "step": 30136 + }, + { + "epoch": 0.96, + "learning_rate": 6.866258657487801e-08, + "loss": 0.9253, + "step": 30137 + }, + { + "epoch": 0.96, + "learning_rate": 6.854146687841234e-08, + "loss": 0.9409, + "step": 30138 + }, + { + "epoch": 0.96, + "learning_rate": 6.842045373550488e-08, + "loss": 0.9331, + "step": 30139 + }, + { + "epoch": 0.96, + "learning_rate": 6.829954714745346e-08, + "loss": 1.0093, + "step": 30140 + }, + { + "epoch": 0.96, + "learning_rate": 6.817874711555705e-08, + "loss": 0.8447, + "step": 30141 + }, + { + "epoch": 0.96, + "learning_rate": 6.805805364111018e-08, + "loss": 0.9551, + "step": 30142 + }, + { + "epoch": 0.96, + "learning_rate": 6.793746672540846e-08, + "loss": 1.0503, + "step": 30143 + }, + { + "epoch": 0.96, + "learning_rate": 6.781698636974532e-08, + "loss": 0.7939, + "step": 30144 + }, + { + "epoch": 0.96, + "learning_rate": 6.769661257541416e-08, + "loss": 0.9893, + "step": 30145 + }, + { + "epoch": 0.96, + "learning_rate": 6.757634534370505e-08, + "loss": 0.8916, + "step": 30146 + }, + { + "epoch": 0.96, + "learning_rate": 6.74561846759092e-08, + "loss": 0.9849, + "step": 30147 + }, + { + "epoch": 0.96, + "learning_rate": 6.733613057331556e-08, + "loss": 0.9658, + "step": 30148 + }, + { + "epoch": 0.96, + "learning_rate": 6.721618303721311e-08, + "loss": 0.9155, + "step": 30149 + }, + { + "epoch": 0.96, + "learning_rate": 6.709634206888749e-08, + "loss": 0.8833, + "step": 30150 + }, + { + "epoch": 0.96, + "learning_rate": 6.697660766962433e-08, + "loss": 0.7749, + "step": 30151 + }, + { + "epoch": 0.96, + "learning_rate": 6.685697984070927e-08, + "loss": 0.9326, + "step": 30152 + }, + { + "epoch": 0.96, + "learning_rate": 6.673745858342572e-08, + "loss": 0.9033, + "step": 30153 + }, + { + "epoch": 0.96, + "learning_rate": 6.661804389905491e-08, + "loss": 0.9277, + "step": 30154 + }, + { + "epoch": 0.96, + "learning_rate": 6.649873578888021e-08, + "loss": 0.8164, + "step": 30155 + }, + { + "epoch": 0.96, + "learning_rate": 6.637953425417842e-08, + "loss": 0.7825, + "step": 30156 + }, + { + "epoch": 0.96, + "learning_rate": 6.626043929623183e-08, + "loss": 0.918, + "step": 30157 + }, + { + "epoch": 0.96, + "learning_rate": 6.614145091631496e-08, + "loss": 0.9932, + "step": 30158 + }, + { + "epoch": 0.96, + "learning_rate": 6.602256911570792e-08, + "loss": 1.0112, + "step": 30159 + }, + { + "epoch": 0.96, + "learning_rate": 6.590379389568302e-08, + "loss": 0.9609, + "step": 30160 + }, + { + "epoch": 0.96, + "learning_rate": 6.57851252575159e-08, + "loss": 0.8916, + "step": 30161 + }, + { + "epoch": 0.96, + "learning_rate": 6.566656320247999e-08, + "loss": 0.9121, + "step": 30162 + }, + { + "epoch": 0.96, + "learning_rate": 6.554810773184761e-08, + "loss": 0.9346, + "step": 30163 + }, + { + "epoch": 0.96, + "learning_rate": 6.542975884688885e-08, + "loss": 0.9756, + "step": 30164 + }, + { + "epoch": 0.96, + "learning_rate": 6.53115165488738e-08, + "loss": 0.8511, + "step": 30165 + }, + { + "epoch": 0.96, + "learning_rate": 6.519338083907034e-08, + "loss": 0.8901, + "step": 30166 + }, + { + "epoch": 0.96, + "learning_rate": 6.507535171874746e-08, + "loss": 1.0332, + "step": 30167 + }, + { + "epoch": 0.96, + "learning_rate": 6.495742918916969e-08, + "loss": 1.083, + "step": 30168 + }, + { + "epoch": 0.96, + "learning_rate": 6.483961325160382e-08, + "loss": 1.002, + "step": 30169 + }, + { + "epoch": 0.96, + "learning_rate": 6.472190390731326e-08, + "loss": 0.8027, + "step": 30170 + }, + { + "epoch": 0.96, + "learning_rate": 6.460430115756034e-08, + "loss": 0.9385, + "step": 30171 + }, + { + "epoch": 0.96, + "learning_rate": 6.448680500360849e-08, + "loss": 0.9204, + "step": 30172 + }, + { + "epoch": 0.96, + "learning_rate": 6.43694154467156e-08, + "loss": 1.0605, + "step": 30173 + }, + { + "epoch": 0.96, + "learning_rate": 6.425213248814399e-08, + "loss": 0.8938, + "step": 30174 + }, + { + "epoch": 0.97, + "learning_rate": 6.413495612914933e-08, + "loss": 0.8379, + "step": 30175 + }, + { + "epoch": 0.97, + "learning_rate": 6.40178863709906e-08, + "loss": 0.8955, + "step": 30176 + }, + { + "epoch": 0.97, + "learning_rate": 6.390092321492348e-08, + "loss": 0.9263, + "step": 30177 + }, + { + "epoch": 0.97, + "learning_rate": 6.378406666220361e-08, + "loss": 0.7949, + "step": 30178 + }, + { + "epoch": 0.97, + "learning_rate": 6.366731671408332e-08, + "loss": 0.8799, + "step": 30179 + }, + { + "epoch": 0.97, + "learning_rate": 6.355067337181497e-08, + "loss": 0.7847, + "step": 30180 + }, + { + "epoch": 0.97, + "learning_rate": 6.343413663665088e-08, + "loss": 0.9211, + "step": 30181 + }, + { + "epoch": 0.97, + "learning_rate": 6.331770650984226e-08, + "loss": 0.9336, + "step": 30182 + }, + { + "epoch": 0.97, + "learning_rate": 6.320138299263589e-08, + "loss": 0.9785, + "step": 30183 + }, + { + "epoch": 0.97, + "learning_rate": 6.30851660862819e-08, + "loss": 0.8647, + "step": 30184 + }, + { + "epoch": 0.97, + "learning_rate": 6.296905579202705e-08, + "loss": 0.981, + "step": 30185 + }, + { + "epoch": 0.97, + "learning_rate": 6.285305211111592e-08, + "loss": 0.8921, + "step": 30186 + }, + { + "epoch": 0.97, + "learning_rate": 6.273715504479306e-08, + "loss": 1.1338, + "step": 30187 + }, + { + "epoch": 0.97, + "learning_rate": 6.262136459430412e-08, + "loss": 0.8813, + "step": 30188 + }, + { + "epoch": 0.97, + "learning_rate": 6.250568076088814e-08, + "loss": 0.8938, + "step": 30189 + }, + { + "epoch": 0.97, + "learning_rate": 6.239010354578856e-08, + "loss": 0.9814, + "step": 30190 + }, + { + "epoch": 0.97, + "learning_rate": 6.227463295024438e-08, + "loss": 0.9111, + "step": 30191 + }, + { + "epoch": 0.97, + "learning_rate": 6.215926897549574e-08, + "loss": 1.0117, + "step": 30192 + }, + { + "epoch": 0.97, + "learning_rate": 6.204401162277828e-08, + "loss": 0.9971, + "step": 30193 + }, + { + "epoch": 0.97, + "learning_rate": 6.192886089333106e-08, + "loss": 1.0469, + "step": 30194 + }, + { + "epoch": 0.97, + "learning_rate": 6.181381678838749e-08, + "loss": 0.8325, + "step": 30195 + }, + { + "epoch": 0.97, + "learning_rate": 6.169887930918328e-08, + "loss": 0.9121, + "step": 30196 + }, + { + "epoch": 0.97, + "learning_rate": 6.158404845694965e-08, + "loss": 0.873, + "step": 30197 + }, + { + "epoch": 0.97, + "learning_rate": 6.146932423292117e-08, + "loss": 0.8999, + "step": 30198 + }, + { + "epoch": 0.97, + "learning_rate": 6.135470663832688e-08, + "loss": 0.8872, + "step": 30199 + }, + { + "epoch": 0.97, + "learning_rate": 6.124019567439688e-08, + "loss": 0.8975, + "step": 30200 + }, + { + "epoch": 0.97, + "learning_rate": 6.112579134236018e-08, + "loss": 1.0127, + "step": 30201 + }, + { + "epoch": 0.97, + "learning_rate": 6.101149364344472e-08, + "loss": 0.9233, + "step": 30202 + }, + { + "epoch": 0.97, + "learning_rate": 6.089730257887505e-08, + "loss": 0.9785, + "step": 30203 + }, + { + "epoch": 0.97, + "learning_rate": 6.078321814987686e-08, + "loss": 0.9067, + "step": 30204 + }, + { + "epoch": 0.97, + "learning_rate": 6.066924035767475e-08, + "loss": 1.0767, + "step": 30205 + }, + { + "epoch": 0.97, + "learning_rate": 6.055536920349214e-08, + "loss": 0.9409, + "step": 30206 + }, + { + "epoch": 0.97, + "learning_rate": 6.044160468855032e-08, + "loss": 0.9258, + "step": 30207 + }, + { + "epoch": 0.97, + "learning_rate": 6.032794681406829e-08, + "loss": 0.9468, + "step": 30208 + }, + { + "epoch": 0.97, + "learning_rate": 6.02143955812673e-08, + "loss": 0.8687, + "step": 30209 + }, + { + "epoch": 0.97, + "learning_rate": 6.010095099136526e-08, + "loss": 0.814, + "step": 30210 + }, + { + "epoch": 0.97, + "learning_rate": 5.998761304557898e-08, + "loss": 0.939, + "step": 30211 + }, + { + "epoch": 0.97, + "learning_rate": 5.987438174512417e-08, + "loss": 0.8989, + "step": 30212 + }, + { + "epoch": 0.97, + "learning_rate": 5.97612570912165e-08, + "loss": 0.8291, + "step": 30213 + }, + { + "epoch": 0.97, + "learning_rate": 5.964823908506834e-08, + "loss": 0.9561, + "step": 30214 + }, + { + "epoch": 0.97, + "learning_rate": 5.9535327727894274e-08, + "loss": 0.917, + "step": 30215 + }, + { + "epoch": 0.97, + "learning_rate": 5.942252302090334e-08, + "loss": 0.8462, + "step": 30216 + }, + { + "epoch": 0.97, + "learning_rate": 5.9309824965307906e-08, + "loss": 0.9277, + "step": 30217 + }, + { + "epoch": 0.97, + "learning_rate": 5.919723356231588e-08, + "loss": 0.9424, + "step": 30218 + }, + { + "epoch": 0.97, + "learning_rate": 5.9084748813136305e-08, + "loss": 0.9629, + "step": 30219 + }, + { + "epoch": 0.97, + "learning_rate": 5.897237071897488e-08, + "loss": 0.9487, + "step": 30220 + }, + { + "epoch": 0.97, + "learning_rate": 5.886009928103731e-08, + "loss": 0.9995, + "step": 30221 + }, + { + "epoch": 0.97, + "learning_rate": 5.874793450052929e-08, + "loss": 0.9604, + "step": 30222 + }, + { + "epoch": 0.97, + "learning_rate": 5.86358763786532e-08, + "loss": 0.8506, + "step": 30223 + }, + { + "epoch": 0.97, + "learning_rate": 5.852392491661141e-08, + "loss": 0.8818, + "step": 30224 + }, + { + "epoch": 0.97, + "learning_rate": 5.8412080115605174e-08, + "loss": 0.9199, + "step": 30225 + }, + { + "epoch": 0.97, + "learning_rate": 5.830034197683465e-08, + "loss": 0.9019, + "step": 30226 + }, + { + "epoch": 0.97, + "learning_rate": 5.818871050149888e-08, + "loss": 0.8633, + "step": 30227 + }, + { + "epoch": 0.97, + "learning_rate": 5.807718569079468e-08, + "loss": 0.938, + "step": 30228 + }, + { + "epoch": 0.97, + "learning_rate": 5.7965767545918875e-08, + "loss": 0.8105, + "step": 30229 + }, + { + "epoch": 0.97, + "learning_rate": 5.785445606806828e-08, + "loss": 0.9175, + "step": 30230 + }, + { + "epoch": 0.97, + "learning_rate": 5.774325125843527e-08, + "loss": 0.8315, + "step": 30231 + }, + { + "epoch": 0.97, + "learning_rate": 5.763215311821335e-08, + "loss": 0.8975, + "step": 30232 + }, + { + "epoch": 0.97, + "learning_rate": 5.752116164859489e-08, + "loss": 0.8584, + "step": 30233 + }, + { + "epoch": 0.97, + "learning_rate": 5.7410276850770055e-08, + "loss": 0.8701, + "step": 30234 + }, + { + "epoch": 0.97, + "learning_rate": 5.72994987259301e-08, + "loss": 0.8867, + "step": 30235 + }, + { + "epoch": 0.97, + "learning_rate": 5.718882727526187e-08, + "loss": 0.8354, + "step": 30236 + }, + { + "epoch": 0.97, + "learning_rate": 5.7078262499953297e-08, + "loss": 0.7559, + "step": 30237 + }, + { + "epoch": 0.97, + "learning_rate": 5.696780440119121e-08, + "loss": 0.897, + "step": 30238 + }, + { + "epoch": 0.97, + "learning_rate": 5.685745298016021e-08, + "loss": 0.9985, + "step": 30239 + }, + { + "epoch": 0.97, + "learning_rate": 5.6747208238043805e-08, + "loss": 0.958, + "step": 30240 + }, + { + "epoch": 0.97, + "learning_rate": 5.6637070176025486e-08, + "loss": 0.959, + "step": 30241 + }, + { + "epoch": 0.97, + "learning_rate": 5.652703879528654e-08, + "loss": 0.9209, + "step": 30242 + }, + { + "epoch": 0.97, + "learning_rate": 5.641711409700712e-08, + "loss": 0.9082, + "step": 30243 + }, + { + "epoch": 0.97, + "learning_rate": 5.63072960823674e-08, + "loss": 0.9526, + "step": 30244 + }, + { + "epoch": 0.97, + "learning_rate": 5.619758475254533e-08, + "loss": 0.8513, + "step": 30245 + }, + { + "epoch": 0.97, + "learning_rate": 5.6087980108717743e-08, + "loss": 0.9438, + "step": 30246 + }, + { + "epoch": 0.97, + "learning_rate": 5.5978482152061474e-08, + "loss": 0.9146, + "step": 30247 + }, + { + "epoch": 0.97, + "learning_rate": 5.5869090883748925e-08, + "loss": 1.0591, + "step": 30248 + }, + { + "epoch": 0.97, + "learning_rate": 5.575980630495692e-08, + "loss": 1.0464, + "step": 30249 + }, + { + "epoch": 0.97, + "learning_rate": 5.565062841685676e-08, + "loss": 0.9126, + "step": 30250 + }, + { + "epoch": 0.97, + "learning_rate": 5.554155722061749e-08, + "loss": 0.8086, + "step": 30251 + }, + { + "epoch": 0.97, + "learning_rate": 5.5432592717413745e-08, + "loss": 1.0645, + "step": 30252 + }, + { + "epoch": 0.97, + "learning_rate": 5.532373490841014e-08, + "loss": 0.9634, + "step": 30253 + }, + { + "epoch": 0.97, + "learning_rate": 5.521498379477796e-08, + "loss": 0.8491, + "step": 30254 + }, + { + "epoch": 0.97, + "learning_rate": 5.5106339377681836e-08, + "loss": 0.9531, + "step": 30255 + }, + { + "epoch": 0.97, + "learning_rate": 5.499780165828861e-08, + "loss": 1.0742, + "step": 30256 + }, + { + "epoch": 0.97, + "learning_rate": 5.48893706377629e-08, + "loss": 0.8901, + "step": 30257 + }, + { + "epoch": 0.97, + "learning_rate": 5.4781046317267103e-08, + "loss": 0.9917, + "step": 30258 + }, + { + "epoch": 0.97, + "learning_rate": 5.467282869796364e-08, + "loss": 0.9114, + "step": 30259 + }, + { + "epoch": 0.97, + "learning_rate": 5.45647177810138e-08, + "loss": 0.8794, + "step": 30260 + }, + { + "epoch": 0.97, + "learning_rate": 5.445671356757776e-08, + "loss": 0.8618, + "step": 30261 + }, + { + "epoch": 0.97, + "learning_rate": 5.434881605881348e-08, + "loss": 0.8804, + "step": 30262 + }, + { + "epoch": 0.97, + "learning_rate": 5.4241025255880045e-08, + "loss": 0.9243, + "step": 30263 + }, + { + "epoch": 0.97, + "learning_rate": 5.4133341159932075e-08, + "loss": 0.8452, + "step": 30264 + }, + { + "epoch": 0.97, + "learning_rate": 5.4025763772126426e-08, + "loss": 0.8452, + "step": 30265 + }, + { + "epoch": 0.97, + "learning_rate": 5.3918293093615515e-08, + "loss": 0.8794, + "step": 30266 + }, + { + "epoch": 0.97, + "learning_rate": 5.381092912555508e-08, + "loss": 0.936, + "step": 30267 + }, + { + "epoch": 0.97, + "learning_rate": 5.37036718690942e-08, + "loss": 0.7954, + "step": 30268 + }, + { + "epoch": 0.97, + "learning_rate": 5.35965213253864e-08, + "loss": 0.9089, + "step": 30269 + }, + { + "epoch": 0.97, + "learning_rate": 5.3489477495578536e-08, + "loss": 0.8589, + "step": 30270 + }, + { + "epoch": 0.97, + "learning_rate": 5.33825403808208e-08, + "loss": 0.9463, + "step": 30271 + }, + { + "epoch": 0.97, + "learning_rate": 5.327570998226006e-08, + "loss": 0.8662, + "step": 30272 + }, + { + "epoch": 0.97, + "learning_rate": 5.316898630104206e-08, + "loss": 0.9214, + "step": 30273 + }, + { + "epoch": 0.97, + "learning_rate": 5.306236933831255e-08, + "loss": 0.9922, + "step": 30274 + }, + { + "epoch": 0.97, + "learning_rate": 5.295585909521506e-08, + "loss": 0.8911, + "step": 30275 + }, + { + "epoch": 0.97, + "learning_rate": 5.2849455572892006e-08, + "loss": 0.9414, + "step": 30276 + }, + { + "epoch": 0.97, + "learning_rate": 5.274315877248581e-08, + "loss": 0.9609, + "step": 30277 + }, + { + "epoch": 0.97, + "learning_rate": 5.263696869513668e-08, + "loss": 0.814, + "step": 30278 + }, + { + "epoch": 0.97, + "learning_rate": 5.2530885341982586e-08, + "loss": 0.9043, + "step": 30279 + }, + { + "epoch": 0.97, + "learning_rate": 5.2424908714163725e-08, + "loss": 0.8105, + "step": 30280 + }, + { + "epoch": 0.97, + "learning_rate": 5.231903881281586e-08, + "loss": 1.0269, + "step": 30281 + }, + { + "epoch": 0.97, + "learning_rate": 5.2213275639075854e-08, + "loss": 0.9341, + "step": 30282 + }, + { + "epoch": 0.97, + "learning_rate": 5.210761919407614e-08, + "loss": 0.8691, + "step": 30283 + }, + { + "epoch": 0.97, + "learning_rate": 5.2002069478953586e-08, + "loss": 0.8044, + "step": 30284 + }, + { + "epoch": 0.97, + "learning_rate": 5.18966264948384e-08, + "loss": 0.9233, + "step": 30285 + }, + { + "epoch": 0.97, + "learning_rate": 5.179129024286189e-08, + "loss": 0.9165, + "step": 30286 + }, + { + "epoch": 0.97, + "learning_rate": 5.168606072415538e-08, + "loss": 0.8857, + "step": 30287 + }, + { + "epoch": 0.97, + "learning_rate": 5.158093793984687e-08, + "loss": 0.4834, + "step": 30288 + }, + { + "epoch": 0.97, + "learning_rate": 5.1475921891064316e-08, + "loss": 0.9194, + "step": 30289 + }, + { + "epoch": 0.97, + "learning_rate": 5.137101257893462e-08, + "loss": 0.7964, + "step": 30290 + }, + { + "epoch": 0.97, + "learning_rate": 5.126621000458354e-08, + "loss": 0.896, + "step": 30291 + }, + { + "epoch": 0.97, + "learning_rate": 5.1161514169135727e-08, + "loss": 0.4619, + "step": 30292 + }, + { + "epoch": 0.97, + "learning_rate": 5.105692507371363e-08, + "loss": 0.7734, + "step": 30293 + }, + { + "epoch": 0.97, + "learning_rate": 5.095244271943966e-08, + "loss": 0.9849, + "step": 30294 + }, + { + "epoch": 0.97, + "learning_rate": 5.0848067107435175e-08, + "loss": 0.4756, + "step": 30295 + }, + { + "epoch": 0.97, + "learning_rate": 5.0743798238819255e-08, + "loss": 0.9707, + "step": 30296 + }, + { + "epoch": 0.97, + "learning_rate": 5.063963611471212e-08, + "loss": 0.8975, + "step": 30297 + }, + { + "epoch": 0.97, + "learning_rate": 5.053558073622844e-08, + "loss": 1.0225, + "step": 30298 + }, + { + "epoch": 0.97, + "learning_rate": 5.043163210448843e-08, + "loss": 0.8398, + "step": 30299 + }, + { + "epoch": 0.97, + "learning_rate": 5.032779022060452e-08, + "loss": 0.7881, + "step": 30300 + }, + { + "epoch": 0.97, + "learning_rate": 5.022405508569139e-08, + "loss": 0.8906, + "step": 30301 + }, + { + "epoch": 0.97, + "learning_rate": 5.012042670086259e-08, + "loss": 0.9902, + "step": 30302 + }, + { + "epoch": 0.97, + "learning_rate": 5.001690506723056e-08, + "loss": 0.9448, + "step": 30303 + }, + { + "epoch": 0.97, + "learning_rate": 4.99134901859033e-08, + "loss": 1.0317, + "step": 30304 + }, + { + "epoch": 0.97, + "learning_rate": 4.981018205799326e-08, + "loss": 0.9214, + "step": 30305 + }, + { + "epoch": 0.97, + "learning_rate": 4.970698068460733e-08, + "loss": 0.9624, + "step": 30306 + }, + { + "epoch": 0.97, + "learning_rate": 4.9603886066853513e-08, + "loss": 1.0503, + "step": 30307 + }, + { + "epoch": 0.97, + "learning_rate": 4.950089820583648e-08, + "loss": 1.0669, + "step": 30308 + }, + { + "epoch": 0.97, + "learning_rate": 4.939801710266312e-08, + "loss": 0.9204, + "step": 30309 + }, + { + "epoch": 0.97, + "learning_rate": 4.929524275843478e-08, + "loss": 0.978, + "step": 30310 + }, + { + "epoch": 0.97, + "learning_rate": 4.919257517425724e-08, + "loss": 0.9614, + "step": 30311 + }, + { + "epoch": 0.97, + "learning_rate": 4.909001435122962e-08, + "loss": 0.9609, + "step": 30312 + }, + { + "epoch": 0.97, + "learning_rate": 4.898756029045326e-08, + "loss": 0.9492, + "step": 30313 + }, + { + "epoch": 0.97, + "learning_rate": 4.8885212993026175e-08, + "loss": 1.0059, + "step": 30314 + }, + { + "epoch": 0.97, + "learning_rate": 4.878297246004859e-08, + "loss": 0.7983, + "step": 30315 + }, + { + "epoch": 0.97, + "learning_rate": 4.8680838692616305e-08, + "loss": 0.8789, + "step": 30316 + }, + { + "epoch": 0.97, + "learning_rate": 4.857881169182399e-08, + "loss": 0.9302, + "step": 30317 + }, + { + "epoch": 0.97, + "learning_rate": 4.847689145876855e-08, + "loss": 0.8237, + "step": 30318 + }, + { + "epoch": 0.97, + "learning_rate": 4.837507799454133e-08, + "loss": 0.9658, + "step": 30319 + }, + { + "epoch": 0.97, + "learning_rate": 4.8273371300235904e-08, + "loss": 0.8506, + "step": 30320 + }, + { + "epoch": 0.97, + "learning_rate": 4.8171771376943624e-08, + "loss": 0.8867, + "step": 30321 + }, + { + "epoch": 0.97, + "learning_rate": 4.8070278225753606e-08, + "loss": 0.9658, + "step": 30322 + }, + { + "epoch": 0.97, + "learning_rate": 4.7968891847754996e-08, + "loss": 1.0664, + "step": 30323 + }, + { + "epoch": 0.97, + "learning_rate": 4.7867612244036906e-08, + "loss": 0.9722, + "step": 30324 + }, + { + "epoch": 0.97, + "learning_rate": 4.776643941568404e-08, + "loss": 0.9087, + "step": 30325 + }, + { + "epoch": 0.97, + "learning_rate": 4.766537336378219e-08, + "loss": 0.9233, + "step": 30326 + }, + { + "epoch": 0.97, + "learning_rate": 4.756441408941715e-08, + "loss": 0.8486, + "step": 30327 + }, + { + "epoch": 0.97, + "learning_rate": 4.746356159367027e-08, + "loss": 0.9883, + "step": 30328 + }, + { + "epoch": 0.97, + "learning_rate": 4.7362815877625146e-08, + "loss": 0.4622, + "step": 30329 + }, + { + "epoch": 0.97, + "learning_rate": 4.726217694236201e-08, + "loss": 0.979, + "step": 30330 + }, + { + "epoch": 0.97, + "learning_rate": 4.716164478896001e-08, + "loss": 0.9482, + "step": 30331 + }, + { + "epoch": 0.97, + "learning_rate": 4.706121941849939e-08, + "loss": 0.9331, + "step": 30332 + }, + { + "epoch": 0.97, + "learning_rate": 4.6960900832054844e-08, + "loss": 0.9854, + "step": 30333 + }, + { + "epoch": 0.97, + "learning_rate": 4.686068903070551e-08, + "loss": 0.9478, + "step": 30334 + }, + { + "epoch": 0.97, + "learning_rate": 4.6760584015524966e-08, + "loss": 0.9297, + "step": 30335 + }, + { + "epoch": 0.97, + "learning_rate": 4.666058578758792e-08, + "loss": 0.9619, + "step": 30336 + }, + { + "epoch": 0.97, + "learning_rate": 4.656069434796795e-08, + "loss": 0.9175, + "step": 30337 + }, + { + "epoch": 0.97, + "learning_rate": 4.64609096977342e-08, + "loss": 0.8926, + "step": 30338 + }, + { + "epoch": 0.97, + "learning_rate": 4.636123183795915e-08, + "loss": 0.8994, + "step": 30339 + }, + { + "epoch": 0.97, + "learning_rate": 4.626166076971306e-08, + "loss": 0.9541, + "step": 30340 + }, + { + "epoch": 0.97, + "learning_rate": 4.616219649406173e-08, + "loss": 0.8232, + "step": 30341 + }, + { + "epoch": 0.97, + "learning_rate": 4.6062839012074314e-08, + "loss": 0.9941, + "step": 30342 + }, + { + "epoch": 0.97, + "learning_rate": 4.596358832481551e-08, + "loss": 0.9194, + "step": 30343 + }, + { + "epoch": 0.97, + "learning_rate": 4.5864444433351144e-08, + "loss": 0.981, + "step": 30344 + }, + { + "epoch": 0.97, + "learning_rate": 4.5765407338745907e-08, + "loss": 0.897, + "step": 30345 + }, + { + "epoch": 0.97, + "learning_rate": 4.5666477042060066e-08, + "loss": 0.9766, + "step": 30346 + }, + { + "epoch": 0.97, + "learning_rate": 4.5567653544354995e-08, + "loss": 0.8169, + "step": 30347 + }, + { + "epoch": 0.97, + "learning_rate": 4.546893684669318e-08, + "loss": 0.9873, + "step": 30348 + }, + { + "epoch": 0.97, + "learning_rate": 4.537032695013155e-08, + "loss": 0.9292, + "step": 30349 + }, + { + "epoch": 0.97, + "learning_rate": 4.527182385573037e-08, + "loss": 0.9424, + "step": 30350 + }, + { + "epoch": 0.97, + "learning_rate": 4.5173427564544347e-08, + "loss": 1.0015, + "step": 30351 + }, + { + "epoch": 0.97, + "learning_rate": 4.507513807763042e-08, + "loss": 0.9937, + "step": 30352 + }, + { + "epoch": 0.97, + "learning_rate": 4.497695539604219e-08, + "loss": 0.9531, + "step": 30353 + }, + { + "epoch": 0.97, + "learning_rate": 4.4878879520834364e-08, + "loss": 0.9834, + "step": 30354 + }, + { + "epoch": 0.97, + "learning_rate": 4.478091045305721e-08, + "loss": 0.8633, + "step": 30355 + }, + { + "epoch": 0.97, + "learning_rate": 4.4683048193764345e-08, + "loss": 0.9766, + "step": 30356 + }, + { + "epoch": 0.97, + "learning_rate": 4.45852927440038e-08, + "loss": 0.9102, + "step": 30357 + }, + { + "epoch": 0.97, + "learning_rate": 4.4487644104824754e-08, + "loss": 0.8701, + "step": 30358 + }, + { + "epoch": 0.97, + "learning_rate": 4.439010227727525e-08, + "loss": 0.8999, + "step": 30359 + }, + { + "epoch": 0.97, + "learning_rate": 4.4292667262401114e-08, + "loss": 0.854, + "step": 30360 + }, + { + "epoch": 0.97, + "learning_rate": 4.419533906124929e-08, + "loss": 0.9399, + "step": 30361 + }, + { + "epoch": 0.97, + "learning_rate": 4.4098117674862274e-08, + "loss": 0.7749, + "step": 30362 + }, + { + "epoch": 0.97, + "learning_rate": 4.400100310428368e-08, + "loss": 0.9502, + "step": 30363 + }, + { + "epoch": 0.97, + "learning_rate": 4.3903995350556e-08, + "loss": 0.9043, + "step": 30364 + }, + { + "epoch": 0.97, + "learning_rate": 4.380709441471842e-08, + "loss": 0.9399, + "step": 30365 + }, + { + "epoch": 0.97, + "learning_rate": 4.3710300297813425e-08, + "loss": 0.9458, + "step": 30366 + }, + { + "epoch": 0.97, + "learning_rate": 4.361361300087685e-08, + "loss": 0.9668, + "step": 30367 + }, + { + "epoch": 0.97, + "learning_rate": 4.3517032524946765e-08, + "loss": 0.8491, + "step": 30368 + }, + { + "epoch": 0.97, + "learning_rate": 4.3420558871060116e-08, + "loss": 0.8687, + "step": 30369 + }, + { + "epoch": 0.97, + "learning_rate": 4.332419204025162e-08, + "loss": 0.8818, + "step": 30370 + }, + { + "epoch": 0.97, + "learning_rate": 4.322793203355491e-08, + "loss": 0.855, + "step": 30371 + }, + { + "epoch": 0.97, + "learning_rate": 4.3131778852002484e-08, + "loss": 0.9282, + "step": 30372 + }, + { + "epoch": 0.97, + "learning_rate": 4.303573249662685e-08, + "loss": 0.9028, + "step": 30373 + }, + { + "epoch": 0.97, + "learning_rate": 4.2939792968458296e-08, + "loss": 0.8584, + "step": 30374 + }, + { + "epoch": 0.97, + "learning_rate": 4.2843960268524886e-08, + "loss": 0.9053, + "step": 30375 + }, + { + "epoch": 0.97, + "learning_rate": 4.2748234397856916e-08, + "loss": 0.8779, + "step": 30376 + }, + { + "epoch": 0.97, + "learning_rate": 4.265261535747911e-08, + "loss": 0.9282, + "step": 30377 + }, + { + "epoch": 0.97, + "learning_rate": 4.255710314841954e-08, + "loss": 0.9058, + "step": 30378 + }, + { + "epoch": 0.97, + "learning_rate": 4.246169777170073e-08, + "loss": 0.9551, + "step": 30379 + }, + { + "epoch": 0.97, + "learning_rate": 4.236639922834851e-08, + "loss": 0.8555, + "step": 30380 + }, + { + "epoch": 0.97, + "learning_rate": 4.2271207519383184e-08, + "loss": 1.0493, + "step": 30381 + }, + { + "epoch": 0.97, + "learning_rate": 4.2176122645827264e-08, + "loss": 1.0083, + "step": 30382 + }, + { + "epoch": 0.97, + "learning_rate": 4.208114460870216e-08, + "loss": 0.9053, + "step": 30383 + }, + { + "epoch": 0.97, + "learning_rate": 4.198627340902484e-08, + "loss": 0.9395, + "step": 30384 + }, + { + "epoch": 0.97, + "learning_rate": 4.189150904781336e-08, + "loss": 0.9038, + "step": 30385 + }, + { + "epoch": 0.97, + "learning_rate": 4.17968515260847e-08, + "loss": 0.9487, + "step": 30386 + }, + { + "epoch": 0.97, + "learning_rate": 4.170230084485582e-08, + "loss": 0.6982, + "step": 30387 + }, + { + "epoch": 0.97, + "learning_rate": 4.1607857005138143e-08, + "loss": 0.9331, + "step": 30388 + }, + { + "epoch": 0.97, + "learning_rate": 4.151352000794861e-08, + "loss": 0.8716, + "step": 30389 + }, + { + "epoch": 0.97, + "learning_rate": 4.141928985429644e-08, + "loss": 0.8604, + "step": 30390 + }, + { + "epoch": 0.97, + "learning_rate": 4.132516654519414e-08, + "loss": 0.979, + "step": 30391 + }, + { + "epoch": 0.97, + "learning_rate": 4.123115008165091e-08, + "loss": 0.8218, + "step": 30392 + }, + { + "epoch": 0.97, + "learning_rate": 4.113724046467593e-08, + "loss": 0.8623, + "step": 30393 + }, + { + "epoch": 0.97, + "learning_rate": 4.104343769527508e-08, + "loss": 1.0044, + "step": 30394 + }, + { + "epoch": 0.97, + "learning_rate": 4.094974177445754e-08, + "loss": 0.8584, + "step": 30395 + }, + { + "epoch": 0.97, + "learning_rate": 4.085615270322585e-08, + "loss": 0.7158, + "step": 30396 + }, + { + "epoch": 0.97, + "learning_rate": 4.0762670482586976e-08, + "loss": 0.8643, + "step": 30397 + }, + { + "epoch": 0.97, + "learning_rate": 4.066929511354123e-08, + "loss": 1.0576, + "step": 30398 + }, + { + "epoch": 0.97, + "learning_rate": 4.057602659709225e-08, + "loss": 0.9756, + "step": 30399 + }, + { + "epoch": 0.97, + "learning_rate": 4.048286493423925e-08, + "loss": 1.0166, + "step": 30400 + }, + { + "epoch": 0.97, + "learning_rate": 4.038981012598364e-08, + "loss": 0.8948, + "step": 30401 + }, + { + "epoch": 0.97, + "learning_rate": 4.0296862173321294e-08, + "loss": 0.8071, + "step": 30402 + }, + { + "epoch": 0.97, + "learning_rate": 4.020402107725141e-08, + "loss": 0.9805, + "step": 30403 + }, + { + "epoch": 0.97, + "learning_rate": 4.011128683876875e-08, + "loss": 0.9609, + "step": 30404 + }, + { + "epoch": 0.97, + "learning_rate": 4.001865945887029e-08, + "loss": 0.9619, + "step": 30405 + }, + { + "epoch": 0.97, + "learning_rate": 3.9926138938547466e-08, + "loss": 0.8706, + "step": 30406 + }, + { + "epoch": 0.97, + "learning_rate": 3.983372527879392e-08, + "loss": 0.936, + "step": 30407 + }, + { + "epoch": 0.97, + "learning_rate": 3.9741418480601094e-08, + "loss": 0.8306, + "step": 30408 + }, + { + "epoch": 0.97, + "learning_rate": 3.964921854495929e-08, + "loss": 0.9238, + "step": 30409 + }, + { + "epoch": 0.97, + "learning_rate": 3.955712547285773e-08, + "loss": 0.8779, + "step": 30410 + }, + { + "epoch": 0.97, + "learning_rate": 3.9465139265285615e-08, + "loss": 0.9048, + "step": 30411 + }, + { + "epoch": 0.97, + "learning_rate": 3.9373259923227714e-08, + "loss": 1.1021, + "step": 30412 + }, + { + "epoch": 0.97, + "learning_rate": 3.9281487447672125e-08, + "loss": 0.958, + "step": 30413 + }, + { + "epoch": 0.97, + "learning_rate": 3.9189821839600294e-08, + "loss": 0.9321, + "step": 30414 + }, + { + "epoch": 0.97, + "learning_rate": 3.90982630999992e-08, + "loss": 0.9116, + "step": 30415 + }, + { + "epoch": 0.97, + "learning_rate": 3.9006811229848064e-08, + "loss": 0.7402, + "step": 30416 + }, + { + "epoch": 0.97, + "learning_rate": 3.891546623013054e-08, + "loss": 0.8535, + "step": 30417 + }, + { + "epoch": 0.97, + "learning_rate": 3.8824228101825845e-08, + "loss": 0.9209, + "step": 30418 + }, + { + "epoch": 0.97, + "learning_rate": 3.873309684591209e-08, + "loss": 0.7583, + "step": 30419 + }, + { + "epoch": 0.97, + "learning_rate": 3.8642072463368486e-08, + "loss": 0.8809, + "step": 30420 + }, + { + "epoch": 0.97, + "learning_rate": 3.8551154955169814e-08, + "loss": 0.8584, + "step": 30421 + }, + { + "epoch": 0.97, + "learning_rate": 3.8460344322293064e-08, + "loss": 0.874, + "step": 30422 + }, + { + "epoch": 0.97, + "learning_rate": 3.836964056571191e-08, + "loss": 1.2007, + "step": 30423 + }, + { + "epoch": 0.97, + "learning_rate": 3.827904368640001e-08, + "loss": 0.8623, + "step": 30424 + }, + { + "epoch": 0.97, + "learning_rate": 3.81885536853277e-08, + "loss": 1.0142, + "step": 30425 + }, + { + "epoch": 0.97, + "learning_rate": 3.809817056346754e-08, + "loss": 0.915, + "step": 30426 + }, + { + "epoch": 0.97, + "learning_rate": 3.8007894321788754e-08, + "loss": 0.9111, + "step": 30427 + }, + { + "epoch": 0.97, + "learning_rate": 3.791772496126056e-08, + "loss": 0.9233, + "step": 30428 + }, + { + "epoch": 0.97, + "learning_rate": 3.782766248284775e-08, + "loss": 0.8872, + "step": 30429 + }, + { + "epoch": 0.97, + "learning_rate": 3.7737706887520655e-08, + "loss": 0.9355, + "step": 30430 + }, + { + "epoch": 0.97, + "learning_rate": 3.7647858176240724e-08, + "loss": 1.0234, + "step": 30431 + }, + { + "epoch": 0.97, + "learning_rate": 3.755811634997497e-08, + "loss": 0.8496, + "step": 30432 + }, + { + "epoch": 0.97, + "learning_rate": 3.746848140968373e-08, + "loss": 0.9072, + "step": 30433 + }, + { + "epoch": 0.97, + "learning_rate": 3.737895335632957e-08, + "loss": 0.9204, + "step": 30434 + }, + { + "epoch": 0.97, + "learning_rate": 3.7289532190873946e-08, + "loss": 0.979, + "step": 30435 + }, + { + "epoch": 0.97, + "learning_rate": 3.720021791427497e-08, + "loss": 0.9497, + "step": 30436 + }, + { + "epoch": 0.97, + "learning_rate": 3.711101052749078e-08, + "loss": 0.915, + "step": 30437 + }, + { + "epoch": 0.97, + "learning_rate": 3.702191003147948e-08, + "loss": 1.0225, + "step": 30438 + }, + { + "epoch": 0.97, + "learning_rate": 3.693291642719699e-08, + "loss": 0.9536, + "step": 30439 + }, + { + "epoch": 0.97, + "learning_rate": 3.6844029715598085e-08, + "loss": 0.9849, + "step": 30440 + }, + { + "epoch": 0.97, + "learning_rate": 3.675524989763535e-08, + "loss": 0.8203, + "step": 30441 + }, + { + "epoch": 0.97, + "learning_rate": 3.6666576974262455e-08, + "loss": 0.8623, + "step": 30442 + }, + { + "epoch": 0.97, + "learning_rate": 3.6578010946429764e-08, + "loss": 0.9185, + "step": 30443 + }, + { + "epoch": 0.97, + "learning_rate": 3.648955181508873e-08, + "loss": 0.7788, + "step": 30444 + }, + { + "epoch": 0.97, + "learning_rate": 3.6401199581187485e-08, + "loss": 0.8257, + "step": 30445 + }, + { + "epoch": 0.97, + "learning_rate": 3.631295424567416e-08, + "loss": 0.7363, + "step": 30446 + }, + { + "epoch": 0.97, + "learning_rate": 3.622481580949466e-08, + "loss": 0.9658, + "step": 30447 + }, + { + "epoch": 0.97, + "learning_rate": 3.6136784273597126e-08, + "loss": 0.9053, + "step": 30448 + }, + { + "epoch": 0.97, + "learning_rate": 3.604885963892302e-08, + "loss": 0.4607, + "step": 30449 + }, + { + "epoch": 0.97, + "learning_rate": 3.5961041906418245e-08, + "loss": 0.8594, + "step": 30450 + }, + { + "epoch": 0.97, + "learning_rate": 3.587333107702318e-08, + "loss": 0.8833, + "step": 30451 + }, + { + "epoch": 0.97, + "learning_rate": 3.578572715167927e-08, + "loss": 0.8989, + "step": 30452 + }, + { + "epoch": 0.97, + "learning_rate": 3.5698230131326893e-08, + "loss": 0.8564, + "step": 30453 + }, + { + "epoch": 0.97, + "learning_rate": 3.561084001690529e-08, + "loss": 0.8696, + "step": 30454 + }, + { + "epoch": 0.97, + "learning_rate": 3.552355680935038e-08, + "loss": 0.9375, + "step": 30455 + }, + { + "epoch": 0.97, + "learning_rate": 3.5436380509600296e-08, + "loss": 0.8408, + "step": 30456 + }, + { + "epoch": 0.97, + "learning_rate": 3.534931111858986e-08, + "loss": 0.8765, + "step": 30457 + }, + { + "epoch": 0.97, + "learning_rate": 3.526234863725164e-08, + "loss": 0.9253, + "step": 30458 + }, + { + "epoch": 0.97, + "learning_rate": 3.517549306652157e-08, + "loss": 0.9287, + "step": 30459 + }, + { + "epoch": 0.97, + "learning_rate": 3.50887444073289e-08, + "loss": 0.9609, + "step": 30460 + }, + { + "epoch": 0.97, + "learning_rate": 3.500210266060511e-08, + "loss": 0.8789, + "step": 30461 + }, + { + "epoch": 0.97, + "learning_rate": 3.491556782728056e-08, + "loss": 0.9551, + "step": 30462 + }, + { + "epoch": 0.97, + "learning_rate": 3.4829139908283404e-08, + "loss": 0.895, + "step": 30463 + }, + { + "epoch": 0.97, + "learning_rate": 3.474281890454068e-08, + "loss": 0.9072, + "step": 30464 + }, + { + "epoch": 0.97, + "learning_rate": 3.46566048169783e-08, + "loss": 0.8511, + "step": 30465 + }, + { + "epoch": 0.97, + "learning_rate": 3.457049764652109e-08, + "loss": 0.9438, + "step": 30466 + }, + { + "epoch": 0.97, + "learning_rate": 3.4484497394093874e-08, + "loss": 0.9492, + "step": 30467 + }, + { + "epoch": 0.97, + "learning_rate": 3.4398604060618127e-08, + "loss": 0.8926, + "step": 30468 + }, + { + "epoch": 0.97, + "learning_rate": 3.4312817647016436e-08, + "loss": 1.0049, + "step": 30469 + }, + { + "epoch": 0.97, + "learning_rate": 3.422713815420808e-08, + "loss": 0.9546, + "step": 30470 + }, + { + "epoch": 0.97, + "learning_rate": 3.4141565583114547e-08, + "loss": 0.8892, + "step": 30471 + }, + { + "epoch": 0.97, + "learning_rate": 3.405609993465064e-08, + "loss": 1.0044, + "step": 30472 + }, + { + "epoch": 0.97, + "learning_rate": 3.397074120973565e-08, + "loss": 1.0513, + "step": 30473 + }, + { + "epoch": 0.97, + "learning_rate": 3.388548940928438e-08, + "loss": 0.7886, + "step": 30474 + }, + { + "epoch": 0.97, + "learning_rate": 3.380034453421277e-08, + "loss": 0.9683, + "step": 30475 + }, + { + "epoch": 0.97, + "learning_rate": 3.371530658543232e-08, + "loss": 0.8545, + "step": 30476 + }, + { + "epoch": 0.97, + "learning_rate": 3.3630375563857843e-08, + "loss": 1.0405, + "step": 30477 + }, + { + "epoch": 0.97, + "learning_rate": 3.354555147039751e-08, + "loss": 0.9819, + "step": 30478 + }, + { + "epoch": 0.97, + "learning_rate": 3.3460834305963915e-08, + "loss": 0.9155, + "step": 30479 + }, + { + "epoch": 0.97, + "learning_rate": 3.3376224071465234e-08, + "loss": 0.8687, + "step": 30480 + }, + { + "epoch": 0.97, + "learning_rate": 3.329172076780962e-08, + "loss": 0.7539, + "step": 30481 + }, + { + "epoch": 0.97, + "learning_rate": 3.320732439590302e-08, + "loss": 0.9263, + "step": 30482 + }, + { + "epoch": 0.97, + "learning_rate": 3.3123034956650256e-08, + "loss": 0.9868, + "step": 30483 + }, + { + "epoch": 0.97, + "learning_rate": 3.303885245095728e-08, + "loss": 0.9336, + "step": 30484 + }, + { + "epoch": 0.97, + "learning_rate": 3.29547768797267e-08, + "loss": 0.8159, + "step": 30485 + }, + { + "epoch": 0.97, + "learning_rate": 3.287080824386002e-08, + "loss": 0.855, + "step": 30486 + }, + { + "epoch": 0.97, + "learning_rate": 3.2786946544258734e-08, + "loss": 0.8892, + "step": 30487 + }, + { + "epoch": 0.98, + "learning_rate": 3.270319178182213e-08, + "loss": 0.9595, + "step": 30488 + }, + { + "epoch": 0.98, + "learning_rate": 3.261954395744948e-08, + "loss": 0.8398, + "step": 30489 + }, + { + "epoch": 0.98, + "learning_rate": 3.253600307203675e-08, + "loss": 0.8594, + "step": 30490 + }, + { + "epoch": 0.98, + "learning_rate": 3.2452569126482094e-08, + "loss": 0.8877, + "step": 30491 + }, + { + "epoch": 0.98, + "learning_rate": 3.236924212167924e-08, + "loss": 0.9204, + "step": 30492 + }, + { + "epoch": 0.98, + "learning_rate": 3.228602205852305e-08, + "loss": 0.9634, + "step": 30493 + }, + { + "epoch": 0.98, + "learning_rate": 3.220290893790612e-08, + "loss": 0.8608, + "step": 30494 + }, + { + "epoch": 0.98, + "learning_rate": 3.2119902760719967e-08, + "loss": 0.939, + "step": 30495 + }, + { + "epoch": 0.98, + "learning_rate": 3.2037003527856105e-08, + "loss": 0.9507, + "step": 30496 + }, + { + "epoch": 0.98, + "learning_rate": 3.19542112402027e-08, + "loss": 0.8701, + "step": 30497 + }, + { + "epoch": 0.98, + "learning_rate": 3.187152589864906e-08, + "loss": 0.9243, + "step": 30498 + }, + { + "epoch": 0.98, + "learning_rate": 3.1788947504081126e-08, + "loss": 0.9746, + "step": 30499 + }, + { + "epoch": 0.98, + "learning_rate": 3.170647605738486e-08, + "loss": 0.9551, + "step": 30500 + }, + { + "epoch": 0.98, + "learning_rate": 3.162411155944733e-08, + "loss": 0.9927, + "step": 30501 + }, + { + "epoch": 0.98, + "learning_rate": 3.154185401115006e-08, + "loss": 0.9497, + "step": 30502 + }, + { + "epoch": 0.98, + "learning_rate": 3.1459703413375676e-08, + "loss": 0.9229, + "step": 30503 + }, + { + "epoch": 0.98, + "learning_rate": 3.1377659767006795e-08, + "loss": 0.9536, + "step": 30504 + }, + { + "epoch": 0.98, + "learning_rate": 3.1295723072921615e-08, + "loss": 0.9229, + "step": 30505 + }, + { + "epoch": 0.98, + "learning_rate": 3.1213893332001644e-08, + "loss": 0.8271, + "step": 30506 + }, + { + "epoch": 0.98, + "learning_rate": 3.1132170545122855e-08, + "loss": 0.8286, + "step": 30507 + }, + { + "epoch": 0.98, + "learning_rate": 3.105055471316454e-08, + "loss": 0.8457, + "step": 30508 + }, + { + "epoch": 0.98, + "learning_rate": 3.0969045836999334e-08, + "loss": 0.9224, + "step": 30509 + }, + { + "epoch": 0.98, + "learning_rate": 3.088764391750321e-08, + "loss": 0.9492, + "step": 30510 + }, + { + "epoch": 0.98, + "learning_rate": 3.080634895554879e-08, + "loss": 0.8711, + "step": 30511 + }, + { + "epoch": 0.98, + "learning_rate": 3.0725160952009834e-08, + "loss": 0.9189, + "step": 30512 + }, + { + "epoch": 0.98, + "learning_rate": 3.0644079907756754e-08, + "loss": 1.0039, + "step": 30513 + }, + { + "epoch": 0.98, + "learning_rate": 3.0563105823658846e-08, + "loss": 0.9038, + "step": 30514 + }, + { + "epoch": 0.98, + "learning_rate": 3.048223870058431e-08, + "loss": 1.0039, + "step": 30515 + }, + { + "epoch": 0.98, + "learning_rate": 3.0401478539402455e-08, + "loss": 0.8486, + "step": 30516 + }, + { + "epoch": 0.98, + "learning_rate": 3.0320825340978133e-08, + "loss": 0.8428, + "step": 30517 + }, + { + "epoch": 0.98, + "learning_rate": 3.0240279106178436e-08, + "loss": 0.8218, + "step": 30518 + }, + { + "epoch": 0.98, + "learning_rate": 3.0159839835866014e-08, + "loss": 0.8147, + "step": 30519 + }, + { + "epoch": 0.98, + "learning_rate": 3.007950753090461e-08, + "loss": 0.9258, + "step": 30520 + }, + { + "epoch": 0.98, + "learning_rate": 2.999928219215576e-08, + "loss": 0.9521, + "step": 30521 + }, + { + "epoch": 0.98, + "learning_rate": 2.991916382048099e-08, + "loss": 0.7495, + "step": 30522 + }, + { + "epoch": 0.98, + "learning_rate": 2.983915241673852e-08, + "loss": 0.8877, + "step": 30523 + }, + { + "epoch": 0.98, + "learning_rate": 2.9759247981787642e-08, + "loss": 1.0151, + "step": 30524 + }, + { + "epoch": 0.98, + "learning_rate": 2.9679450516485465e-08, + "loss": 0.9468, + "step": 30525 + }, + { + "epoch": 0.98, + "learning_rate": 2.9599760021689073e-08, + "loss": 0.7861, + "step": 30526 + }, + { + "epoch": 0.98, + "learning_rate": 2.9520176498252228e-08, + "loss": 1.0117, + "step": 30527 + }, + { + "epoch": 0.98, + "learning_rate": 2.9440699947028696e-08, + "loss": 0.9307, + "step": 30528 + }, + { + "epoch": 0.98, + "learning_rate": 2.9361330368871122e-08, + "loss": 0.8491, + "step": 30529 + }, + { + "epoch": 0.98, + "learning_rate": 2.928206776463216e-08, + "loss": 1.0317, + "step": 30530 + }, + { + "epoch": 0.98, + "learning_rate": 2.9202912135161132e-08, + "loss": 0.9902, + "step": 30531 + }, + { + "epoch": 0.98, + "learning_rate": 2.912386348130847e-08, + "loss": 1.0249, + "step": 30532 + }, + { + "epoch": 0.98, + "learning_rate": 2.9044921803921267e-08, + "loss": 0.8081, + "step": 30533 + }, + { + "epoch": 0.98, + "learning_rate": 2.896608710384663e-08, + "loss": 0.916, + "step": 30534 + }, + { + "epoch": 0.98, + "learning_rate": 2.8887359381930545e-08, + "loss": 0.8682, + "step": 30535 + }, + { + "epoch": 0.98, + "learning_rate": 2.8808738639017897e-08, + "loss": 0.7944, + "step": 30536 + }, + { + "epoch": 0.98, + "learning_rate": 2.873022487595134e-08, + "loss": 0.8135, + "step": 30537 + }, + { + "epoch": 0.98, + "learning_rate": 2.8651818093573535e-08, + "loss": 0.9062, + "step": 30538 + }, + { + "epoch": 0.98, + "learning_rate": 2.8573518292726034e-08, + "loss": 0.9351, + "step": 30539 + }, + { + "epoch": 0.98, + "learning_rate": 2.849532547424927e-08, + "loss": 0.9634, + "step": 30540 + }, + { + "epoch": 0.98, + "learning_rate": 2.8417239638982573e-08, + "loss": 0.8794, + "step": 30541 + }, + { + "epoch": 0.98, + "learning_rate": 2.8339260787761946e-08, + "loss": 0.9756, + "step": 30542 + }, + { + "epoch": 0.98, + "learning_rate": 2.826138892142449e-08, + "loss": 1.0312, + "step": 30543 + }, + { + "epoch": 0.98, + "learning_rate": 2.8183624040807323e-08, + "loss": 0.7988, + "step": 30544 + }, + { + "epoch": 0.98, + "learning_rate": 2.81059661467431e-08, + "loss": 0.8892, + "step": 30545 + }, + { + "epoch": 0.98, + "learning_rate": 2.802841524006561e-08, + "loss": 0.9702, + "step": 30546 + }, + { + "epoch": 0.98, + "learning_rate": 2.79509713216064e-08, + "loss": 0.9004, + "step": 30547 + }, + { + "epoch": 0.98, + "learning_rate": 2.7873634392197034e-08, + "loss": 0.7947, + "step": 30548 + }, + { + "epoch": 0.98, + "learning_rate": 2.7796404452666847e-08, + "loss": 0.9702, + "step": 30549 + }, + { + "epoch": 0.98, + "learning_rate": 2.7719281503845176e-08, + "loss": 0.8984, + "step": 30550 + }, + { + "epoch": 0.98, + "learning_rate": 2.7642265546558023e-08, + "loss": 0.9541, + "step": 30551 + }, + { + "epoch": 0.98, + "learning_rate": 2.7565356581632508e-08, + "loss": 0.8064, + "step": 30552 + }, + { + "epoch": 0.98, + "learning_rate": 2.7488554609894634e-08, + "loss": 0.9932, + "step": 30553 + }, + { + "epoch": 0.98, + "learning_rate": 2.7411859632165973e-08, + "loss": 0.8911, + "step": 30554 + }, + { + "epoch": 0.98, + "learning_rate": 2.7335271649272522e-08, + "loss": 0.9077, + "step": 30555 + }, + { + "epoch": 0.98, + "learning_rate": 2.7258790662033628e-08, + "loss": 0.896, + "step": 30556 + }, + { + "epoch": 0.98, + "learning_rate": 2.7182416671270863e-08, + "loss": 0.9951, + "step": 30557 + }, + { + "epoch": 0.98, + "learning_rate": 2.7106149677803562e-08, + "loss": 0.9814, + "step": 30558 + }, + { + "epoch": 0.98, + "learning_rate": 2.7029989682449963e-08, + "loss": 0.9888, + "step": 30559 + }, + { + "epoch": 0.98, + "learning_rate": 2.6953936686026083e-08, + "loss": 0.8386, + "step": 30560 + }, + { + "epoch": 0.98, + "learning_rate": 2.6877990689350152e-08, + "loss": 0.9277, + "step": 30561 + }, + { + "epoch": 0.98, + "learning_rate": 2.6802151693233746e-08, + "loss": 0.9058, + "step": 30562 + }, + { + "epoch": 0.98, + "learning_rate": 2.672641969849399e-08, + "loss": 0.9146, + "step": 30563 + }, + { + "epoch": 0.98, + "learning_rate": 2.665079470594245e-08, + "loss": 0.9194, + "step": 30564 + }, + { + "epoch": 0.98, + "learning_rate": 2.6575276716388488e-08, + "loss": 1.0044, + "step": 30565 + }, + { + "epoch": 0.98, + "learning_rate": 2.649986573064478e-08, + "loss": 0.8857, + "step": 30566 + }, + { + "epoch": 0.98, + "learning_rate": 2.6424561749518464e-08, + "loss": 0.9155, + "step": 30567 + }, + { + "epoch": 0.98, + "learning_rate": 2.6349364773818887e-08, + "loss": 0.9146, + "step": 30568 + }, + { + "epoch": 0.98, + "learning_rate": 2.6274274804352074e-08, + "loss": 0.8433, + "step": 30569 + }, + { + "epoch": 0.98, + "learning_rate": 2.6199291841924047e-08, + "loss": 0.8706, + "step": 30570 + }, + { + "epoch": 0.98, + "learning_rate": 2.6124415887339714e-08, + "loss": 0.8452, + "step": 30571 + }, + { + "epoch": 0.98, + "learning_rate": 2.6049646941401773e-08, + "loss": 0.9829, + "step": 30572 + }, + { + "epoch": 0.98, + "learning_rate": 2.5974985004911803e-08, + "loss": 0.937, + "step": 30573 + }, + { + "epoch": 0.98, + "learning_rate": 2.590043007867138e-08, + "loss": 0.8936, + "step": 30574 + }, + { + "epoch": 0.98, + "learning_rate": 2.5825982163482087e-08, + "loss": 0.9009, + "step": 30575 + }, + { + "epoch": 0.98, + "learning_rate": 2.575164126013996e-08, + "loss": 0.9087, + "step": 30576 + }, + { + "epoch": 0.98, + "learning_rate": 2.567740736944324e-08, + "loss": 0.8091, + "step": 30577 + }, + { + "epoch": 0.98, + "learning_rate": 2.5603280492190187e-08, + "loss": 0.8691, + "step": 30578 + }, + { + "epoch": 0.98, + "learning_rate": 2.55292606291746e-08, + "loss": 0.9556, + "step": 30579 + }, + { + "epoch": 0.98, + "learning_rate": 2.5455347781190297e-08, + "loss": 0.9116, + "step": 30580 + }, + { + "epoch": 0.98, + "learning_rate": 2.5381541949031084e-08, + "loss": 0.9658, + "step": 30581 + }, + { + "epoch": 0.98, + "learning_rate": 2.5307843133487443e-08, + "loss": 0.9878, + "step": 30582 + }, + { + "epoch": 0.98, + "learning_rate": 2.523425133535318e-08, + "loss": 0.7974, + "step": 30583 + }, + { + "epoch": 0.98, + "learning_rate": 2.5160766555414328e-08, + "loss": 0.9053, + "step": 30584 + }, + { + "epoch": 0.98, + "learning_rate": 2.5087388794461377e-08, + "loss": 0.9844, + "step": 30585 + }, + { + "epoch": 0.98, + "learning_rate": 2.5014118053281465e-08, + "loss": 0.9604, + "step": 30586 + }, + { + "epoch": 0.98, + "learning_rate": 2.4940954332659528e-08, + "loss": 0.9277, + "step": 30587 + }, + { + "epoch": 0.98, + "learning_rate": 2.48678976333816e-08, + "loss": 0.9556, + "step": 30588 + }, + { + "epoch": 0.98, + "learning_rate": 2.4794947956231496e-08, + "loss": 0.9727, + "step": 30589 + }, + { + "epoch": 0.98, + "learning_rate": 2.4722105301991927e-08, + "loss": 0.8789, + "step": 30590 + }, + { + "epoch": 0.98, + "learning_rate": 2.464936967144338e-08, + "loss": 0.9331, + "step": 30591 + }, + { + "epoch": 0.98, + "learning_rate": 2.4576741065367447e-08, + "loss": 0.9951, + "step": 30592 + }, + { + "epoch": 0.98, + "learning_rate": 2.45042194845424e-08, + "loss": 0.8184, + "step": 30593 + }, + { + "epoch": 0.98, + "learning_rate": 2.4431804929746506e-08, + "loss": 0.9302, + "step": 30594 + }, + { + "epoch": 0.98, + "learning_rate": 2.4359497401758026e-08, + "loss": 0.8447, + "step": 30595 + }, + { + "epoch": 0.98, + "learning_rate": 2.4287296901350787e-08, + "loss": 0.8608, + "step": 30596 + }, + { + "epoch": 0.98, + "learning_rate": 2.4215203429299727e-08, + "loss": 0.9126, + "step": 30597 + }, + { + "epoch": 0.98, + "learning_rate": 2.414321698637978e-08, + "loss": 0.9946, + "step": 30598 + }, + { + "epoch": 0.98, + "learning_rate": 2.4071337573362554e-08, + "loss": 0.9341, + "step": 30599 + }, + { + "epoch": 0.98, + "learning_rate": 2.3999565191018536e-08, + "loss": 0.979, + "step": 30600 + }, + { + "epoch": 0.98, + "learning_rate": 2.392789984011823e-08, + "loss": 1.0322, + "step": 30601 + }, + { + "epoch": 0.98, + "learning_rate": 2.3856341521431016e-08, + "loss": 0.9458, + "step": 30602 + }, + { + "epoch": 0.98, + "learning_rate": 2.3784890235724057e-08, + "loss": 0.8772, + "step": 30603 + }, + { + "epoch": 0.98, + "learning_rate": 2.371354598376452e-08, + "loss": 0.8435, + "step": 30604 + }, + { + "epoch": 0.98, + "learning_rate": 2.3642308766316236e-08, + "loss": 0.9751, + "step": 30605 + }, + { + "epoch": 0.98, + "learning_rate": 2.357117858414526e-08, + "loss": 1.0156, + "step": 30606 + }, + { + "epoch": 0.98, + "learning_rate": 2.3500155438014314e-08, + "loss": 0.8652, + "step": 30607 + }, + { + "epoch": 0.98, + "learning_rate": 2.3429239328685015e-08, + "loss": 0.9658, + "step": 30608 + }, + { + "epoch": 0.98, + "learning_rate": 2.3358430256918974e-08, + "loss": 0.8962, + "step": 30609 + }, + { + "epoch": 0.98, + "learning_rate": 2.328772822347558e-08, + "loss": 1.0796, + "step": 30610 + }, + { + "epoch": 0.98, + "learning_rate": 2.321713322911201e-08, + "loss": 0.8604, + "step": 30611 + }, + { + "epoch": 0.98, + "learning_rate": 2.3146645274587655e-08, + "loss": 0.8618, + "step": 30612 + }, + { + "epoch": 0.98, + "learning_rate": 2.3076264360658575e-08, + "loss": 0.8433, + "step": 30613 + }, + { + "epoch": 0.98, + "learning_rate": 2.300599048807861e-08, + "loss": 0.8267, + "step": 30614 + }, + { + "epoch": 0.98, + "learning_rate": 2.2935823657601606e-08, + "loss": 0.8657, + "step": 30615 + }, + { + "epoch": 0.98, + "learning_rate": 2.2865763869981404e-08, + "loss": 0.9375, + "step": 30616 + }, + { + "epoch": 0.98, + "learning_rate": 2.2795811125970734e-08, + "loss": 0.4602, + "step": 30617 + }, + { + "epoch": 0.98, + "learning_rate": 2.2725965426317886e-08, + "loss": 0.874, + "step": 30618 + }, + { + "epoch": 0.98, + "learning_rate": 2.2656226771773372e-08, + "loss": 0.9199, + "step": 30619 + }, + { + "epoch": 0.98, + "learning_rate": 2.2586595163084368e-08, + "loss": 0.8916, + "step": 30620 + }, + { + "epoch": 0.98, + "learning_rate": 2.251707060099917e-08, + "loss": 0.731, + "step": 30621 + }, + { + "epoch": 0.98, + "learning_rate": 2.2447653086263843e-08, + "loss": 0.8721, + "step": 30622 + }, + { + "epoch": 0.98, + "learning_rate": 2.237834261962224e-08, + "loss": 1.0098, + "step": 30623 + }, + { + "epoch": 0.98, + "learning_rate": 2.230913920181821e-08, + "loss": 0.9116, + "step": 30624 + }, + { + "epoch": 0.98, + "learning_rate": 2.2240042833594488e-08, + "loss": 0.8643, + "step": 30625 + }, + { + "epoch": 0.98, + "learning_rate": 2.21710535156916e-08, + "loss": 0.813, + "step": 30626 + }, + { + "epoch": 0.98, + "learning_rate": 2.2102171248851166e-08, + "loss": 0.8857, + "step": 30627 + }, + { + "epoch": 0.98, + "learning_rate": 2.2033396033811494e-08, + "loss": 0.834, + "step": 30628 + }, + { + "epoch": 0.98, + "learning_rate": 2.1964727871309766e-08, + "loss": 0.9087, + "step": 30629 + }, + { + "epoch": 0.98, + "learning_rate": 2.189616676208428e-08, + "loss": 0.8911, + "step": 30630 + }, + { + "epoch": 0.98, + "learning_rate": 2.18277127068689e-08, + "loss": 0.8838, + "step": 30631 + }, + { + "epoch": 0.98, + "learning_rate": 2.1759365706399694e-08, + "loss": 0.8823, + "step": 30632 + }, + { + "epoch": 0.98, + "learning_rate": 2.1691125761408305e-08, + "loss": 0.9668, + "step": 30633 + }, + { + "epoch": 0.98, + "learning_rate": 2.162299287262748e-08, + "loss": 0.9121, + "step": 30634 + }, + { + "epoch": 0.98, + "learning_rate": 2.1554967040789966e-08, + "loss": 1.0215, + "step": 30635 + }, + { + "epoch": 0.98, + "learning_rate": 2.1487048266622957e-08, + "loss": 0.8711, + "step": 30636 + }, + { + "epoch": 0.98, + "learning_rate": 2.141923655085587e-08, + "loss": 1.0166, + "step": 30637 + }, + { + "epoch": 0.98, + "learning_rate": 2.1351531894217014e-08, + "loss": 0.8853, + "step": 30638 + }, + { + "epoch": 0.98, + "learning_rate": 2.1283934297432472e-08, + "loss": 0.9893, + "step": 30639 + }, + { + "epoch": 0.98, + "learning_rate": 2.1216443761227224e-08, + "loss": 0.9468, + "step": 30640 + }, + { + "epoch": 0.98, + "learning_rate": 2.114906028632624e-08, + "loss": 0.4644, + "step": 30641 + }, + { + "epoch": 0.98, + "learning_rate": 2.1081783873451167e-08, + "loss": 0.8604, + "step": 30642 + }, + { + "epoch": 0.98, + "learning_rate": 2.101461452332476e-08, + "loss": 0.9346, + "step": 30643 + }, + { + "epoch": 0.98, + "learning_rate": 2.0947552236666448e-08, + "loss": 0.9868, + "step": 30644 + }, + { + "epoch": 0.98, + "learning_rate": 2.0880597014197868e-08, + "loss": 0.9331, + "step": 30645 + }, + { + "epoch": 0.98, + "learning_rate": 2.0813748856635117e-08, + "loss": 0.9951, + "step": 30646 + }, + { + "epoch": 0.98, + "learning_rate": 2.0747007764697625e-08, + "loss": 0.9946, + "step": 30647 + }, + { + "epoch": 0.98, + "learning_rate": 2.0680373739099258e-08, + "loss": 0.8623, + "step": 30648 + }, + { + "epoch": 0.98, + "learning_rate": 2.0613846780556113e-08, + "loss": 0.9326, + "step": 30649 + }, + { + "epoch": 0.98, + "learning_rate": 2.0547426889782064e-08, + "loss": 0.8872, + "step": 30650 + }, + { + "epoch": 0.98, + "learning_rate": 2.048111406748876e-08, + "loss": 0.918, + "step": 30651 + }, + { + "epoch": 0.98, + "learning_rate": 2.041490831438897e-08, + "loss": 0.9917, + "step": 30652 + }, + { + "epoch": 0.98, + "learning_rate": 2.0348809631192122e-08, + "loss": 0.8975, + "step": 30653 + }, + { + "epoch": 0.98, + "learning_rate": 2.028281801860765e-08, + "loss": 1.0405, + "step": 30654 + }, + { + "epoch": 0.98, + "learning_rate": 2.0216933477343882e-08, + "loss": 0.8735, + "step": 30655 + }, + { + "epoch": 0.98, + "learning_rate": 2.015115600810691e-08, + "loss": 0.9609, + "step": 30656 + }, + { + "epoch": 0.98, + "learning_rate": 2.008548561160284e-08, + "loss": 0.8906, + "step": 30657 + }, + { + "epoch": 0.98, + "learning_rate": 2.0019922288536663e-08, + "loss": 0.8828, + "step": 30658 + }, + { + "epoch": 0.98, + "learning_rate": 1.9954466039611153e-08, + "loss": 0.9751, + "step": 30659 + }, + { + "epoch": 0.98, + "learning_rate": 1.988911686552797e-08, + "loss": 0.957, + "step": 30660 + }, + { + "epoch": 0.98, + "learning_rate": 1.9823874766990993e-08, + "loss": 0.8618, + "step": 30661 + }, + { + "epoch": 0.98, + "learning_rate": 1.9758739744697442e-08, + "loss": 0.8589, + "step": 30662 + }, + { + "epoch": 0.98, + "learning_rate": 1.969371179934676e-08, + "loss": 0.894, + "step": 30663 + }, + { + "epoch": 0.98, + "learning_rate": 1.9628790931636166e-08, + "loss": 0.8213, + "step": 30664 + }, + { + "epoch": 0.98, + "learning_rate": 1.956397714226399e-08, + "loss": 0.9688, + "step": 30665 + }, + { + "epoch": 0.98, + "learning_rate": 1.949927043192412e-08, + "loss": 0.8765, + "step": 30666 + }, + { + "epoch": 0.98, + "learning_rate": 1.943467080131156e-08, + "loss": 0.9639, + "step": 30667 + }, + { + "epoch": 0.98, + "learning_rate": 1.9370178251119088e-08, + "loss": 1.0215, + "step": 30668 + }, + { + "epoch": 0.98, + "learning_rate": 1.9305792782039478e-08, + "loss": 0.8794, + "step": 30669 + }, + { + "epoch": 0.98, + "learning_rate": 1.9241514394761073e-08, + "loss": 1.0107, + "step": 30670 + }, + { + "epoch": 0.98, + "learning_rate": 1.9177343089975543e-08, + "loss": 0.9302, + "step": 30671 + }, + { + "epoch": 0.98, + "learning_rate": 1.9113278868371222e-08, + "loss": 0.8271, + "step": 30672 + }, + { + "epoch": 0.98, + "learning_rate": 1.904932173063423e-08, + "loss": 0.9053, + "step": 30673 + }, + { + "epoch": 0.98, + "learning_rate": 1.8985471677452904e-08, + "loss": 0.8633, + "step": 30674 + }, + { + "epoch": 0.98, + "learning_rate": 1.8921728709510033e-08, + "loss": 0.7617, + "step": 30675 + }, + { + "epoch": 0.98, + "learning_rate": 1.8858092827490625e-08, + "loss": 1.0171, + "step": 30676 + }, + { + "epoch": 0.98, + "learning_rate": 1.8794564032077466e-08, + "loss": 0.9072, + "step": 30677 + }, + { + "epoch": 0.98, + "learning_rate": 1.8731142323952234e-08, + "loss": 0.9038, + "step": 30678 + }, + { + "epoch": 0.98, + "learning_rate": 1.8667827703795494e-08, + "loss": 0.9609, + "step": 30679 + }, + { + "epoch": 0.98, + "learning_rate": 1.8604620172285593e-08, + "loss": 0.8906, + "step": 30680 + }, + { + "epoch": 0.98, + "learning_rate": 1.8541519730100876e-08, + "loss": 0.9751, + "step": 30681 + }, + { + "epoch": 0.98, + "learning_rate": 1.8478526377918583e-08, + "loss": 0.9434, + "step": 30682 + }, + { + "epoch": 0.98, + "learning_rate": 1.8415640116415946e-08, + "loss": 0.9053, + "step": 30683 + }, + { + "epoch": 0.98, + "learning_rate": 1.835286094626576e-08, + "loss": 0.8857, + "step": 30684 + }, + { + "epoch": 0.98, + "learning_rate": 1.8290188868141932e-08, + "loss": 0.8042, + "step": 30685 + }, + { + "epoch": 0.98, + "learning_rate": 1.822762388271726e-08, + "loss": 0.9668, + "step": 30686 + }, + { + "epoch": 0.98, + "learning_rate": 1.8165165990663425e-08, + "loss": 0.8804, + "step": 30687 + }, + { + "epoch": 0.98, + "learning_rate": 1.8102815192649892e-08, + "loss": 0.8403, + "step": 30688 + }, + { + "epoch": 0.98, + "learning_rate": 1.8040571489345017e-08, + "loss": 0.854, + "step": 30689 + }, + { + "epoch": 0.98, + "learning_rate": 1.797843488141826e-08, + "loss": 0.9243, + "step": 30690 + }, + { + "epoch": 0.98, + "learning_rate": 1.791640536953465e-08, + "loss": 1.0684, + "step": 30691 + }, + { + "epoch": 0.98, + "learning_rate": 1.785448295436032e-08, + "loss": 0.9419, + "step": 30692 + }, + { + "epoch": 0.98, + "learning_rate": 1.779266763656029e-08, + "loss": 0.9448, + "step": 30693 + }, + { + "epoch": 0.98, + "learning_rate": 1.7730959416796255e-08, + "loss": 0.8887, + "step": 30694 + }, + { + "epoch": 0.98, + "learning_rate": 1.766935829573213e-08, + "loss": 0.9551, + "step": 30695 + }, + { + "epoch": 0.98, + "learning_rate": 1.7607864274027385e-08, + "loss": 0.9849, + "step": 30696 + }, + { + "epoch": 0.98, + "learning_rate": 1.7546477352342605e-08, + "loss": 0.937, + "step": 30697 + }, + { + "epoch": 0.98, + "learning_rate": 1.7485197531336152e-08, + "loss": 0.8552, + "step": 30698 + }, + { + "epoch": 0.98, + "learning_rate": 1.7424024811665274e-08, + "loss": 0.9565, + "step": 30699 + }, + { + "epoch": 0.98, + "learning_rate": 1.7362959193986117e-08, + "loss": 0.4705, + "step": 30700 + }, + { + "epoch": 0.98, + "learning_rate": 1.730200067895482e-08, + "loss": 1.002, + "step": 30701 + }, + { + "epoch": 0.98, + "learning_rate": 1.72411492672242e-08, + "loss": 0.9131, + "step": 30702 + }, + { + "epoch": 0.98, + "learning_rate": 1.718040495944817e-08, + "loss": 0.8442, + "step": 30703 + }, + { + "epoch": 0.98, + "learning_rate": 1.7119767756277326e-08, + "loss": 0.9722, + "step": 30704 + }, + { + "epoch": 0.98, + "learning_rate": 1.705923765836337e-08, + "loss": 0.9541, + "step": 30705 + }, + { + "epoch": 0.98, + "learning_rate": 1.6998814666354668e-08, + "loss": 0.8516, + "step": 30706 + }, + { + "epoch": 0.98, + "learning_rate": 1.6938498780900703e-08, + "loss": 0.8809, + "step": 30707 + }, + { + "epoch": 0.98, + "learning_rate": 1.6878290002647624e-08, + "loss": 0.8105, + "step": 30708 + }, + { + "epoch": 0.98, + "learning_rate": 1.681818833224269e-08, + "loss": 0.9287, + "step": 30709 + }, + { + "epoch": 0.98, + "learning_rate": 1.6758193770328723e-08, + "loss": 0.8623, + "step": 30710 + }, + { + "epoch": 0.98, + "learning_rate": 1.669830631754965e-08, + "loss": 1.061, + "step": 30711 + }, + { + "epoch": 0.98, + "learning_rate": 1.6638525974550513e-08, + "loss": 0.8257, + "step": 30712 + }, + { + "epoch": 0.98, + "learning_rate": 1.6578852741969685e-08, + "loss": 1.02, + "step": 30713 + }, + { + "epoch": 0.98, + "learning_rate": 1.6519286620448883e-08, + "loss": 0.7998, + "step": 30714 + }, + { + "epoch": 0.98, + "learning_rate": 1.6459827610627588e-08, + "loss": 0.9106, + "step": 30715 + }, + { + "epoch": 0.98, + "learning_rate": 1.6400475713143072e-08, + "loss": 0.8965, + "step": 30716 + }, + { + "epoch": 0.98, + "learning_rate": 1.634123092863149e-08, + "loss": 0.9663, + "step": 30717 + }, + { + "epoch": 0.98, + "learning_rate": 1.6282093257729004e-08, + "loss": 0.9326, + "step": 30718 + }, + { + "epoch": 0.98, + "learning_rate": 1.622306270107177e-08, + "loss": 0.9502, + "step": 30719 + }, + { + "epoch": 0.98, + "learning_rate": 1.616413925928928e-08, + "loss": 0.9126, + "step": 30720 + }, + { + "epoch": 0.98, + "learning_rate": 1.6105322933017698e-08, + "loss": 0.8794, + "step": 30721 + }, + { + "epoch": 0.98, + "learning_rate": 1.6046613722886518e-08, + "loss": 0.9229, + "step": 30722 + }, + { + "epoch": 0.98, + "learning_rate": 1.5988011629524125e-08, + "loss": 0.957, + "step": 30723 + }, + { + "epoch": 0.98, + "learning_rate": 1.5929516653561128e-08, + "loss": 0.8013, + "step": 30724 + }, + { + "epoch": 0.98, + "learning_rate": 1.5871128795624802e-08, + "loss": 0.9219, + "step": 30725 + }, + { + "epoch": 0.98, + "learning_rate": 1.581284805634131e-08, + "loss": 0.8623, + "step": 30726 + }, + { + "epoch": 0.98, + "learning_rate": 1.5754674436336827e-08, + "loss": 0.4836, + "step": 30727 + }, + { + "epoch": 0.98, + "learning_rate": 1.5696607936233066e-08, + "loss": 0.998, + "step": 30728 + }, + { + "epoch": 0.98, + "learning_rate": 1.5638648556656198e-08, + "loss": 0.9082, + "step": 30729 + }, + { + "epoch": 0.98, + "learning_rate": 1.5580796298225733e-08, + "loss": 0.8184, + "step": 30730 + }, + { + "epoch": 0.98, + "learning_rate": 1.5523051161563384e-08, + "loss": 0.8857, + "step": 30731 + }, + { + "epoch": 0.98, + "learning_rate": 1.5465413147287557e-08, + "loss": 0.9653, + "step": 30732 + }, + { + "epoch": 0.98, + "learning_rate": 1.540788225601886e-08, + "loss": 0.9634, + "step": 30733 + }, + { + "epoch": 0.98, + "learning_rate": 1.5350458488372356e-08, + "loss": 0.8667, + "step": 30734 + }, + { + "epoch": 0.98, + "learning_rate": 1.529314184496422e-08, + "loss": 0.4658, + "step": 30735 + }, + { + "epoch": 0.98, + "learning_rate": 1.523593232641174e-08, + "loss": 0.8652, + "step": 30736 + }, + { + "epoch": 0.98, + "learning_rate": 1.5178829933326644e-08, + "loss": 1.0156, + "step": 30737 + }, + { + "epoch": 0.98, + "learning_rate": 1.5121834666321778e-08, + "loss": 0.8662, + "step": 30738 + }, + { + "epoch": 0.98, + "learning_rate": 1.5064946526008872e-08, + "loss": 0.8972, + "step": 30739 + }, + { + "epoch": 0.98, + "learning_rate": 1.5008165512997442e-08, + "loss": 0.8584, + "step": 30740 + }, + { + "epoch": 0.98, + "learning_rate": 1.4951491627899218e-08, + "loss": 0.9893, + "step": 30741 + }, + { + "epoch": 0.98, + "learning_rate": 1.4894924871319272e-08, + "loss": 0.9487, + "step": 30742 + }, + { + "epoch": 0.98, + "learning_rate": 1.4838465243867118e-08, + "loss": 0.9312, + "step": 30743 + }, + { + "epoch": 0.98, + "learning_rate": 1.4782112746145605e-08, + "loss": 0.8096, + "step": 30744 + }, + { + "epoch": 0.98, + "learning_rate": 1.4725867378762027e-08, + "loss": 0.8711, + "step": 30745 + }, + { + "epoch": 0.98, + "learning_rate": 1.4669729142319233e-08, + "loss": 0.9985, + "step": 30746 + }, + { + "epoch": 0.98, + "learning_rate": 1.4613698037417857e-08, + "loss": 0.9268, + "step": 30747 + }, + { + "epoch": 0.98, + "learning_rate": 1.4557774064660746e-08, + "loss": 1.1133, + "step": 30748 + }, + { + "epoch": 0.98, + "learning_rate": 1.4501957224647422e-08, + "loss": 0.8945, + "step": 30749 + }, + { + "epoch": 0.98, + "learning_rate": 1.4446247517976298e-08, + "loss": 0.9424, + "step": 30750 + }, + { + "epoch": 0.98, + "learning_rate": 1.439064494524578e-08, + "loss": 0.8352, + "step": 30751 + }, + { + "epoch": 0.98, + "learning_rate": 1.4335149507052059e-08, + "loss": 0.8735, + "step": 30752 + }, + { + "epoch": 0.98, + "learning_rate": 1.4279761203990216e-08, + "loss": 0.7361, + "step": 30753 + }, + { + "epoch": 0.98, + "learning_rate": 1.422448003665533e-08, + "loss": 1.0713, + "step": 30754 + }, + { + "epoch": 0.98, + "learning_rate": 1.416930600564026e-08, + "loss": 0.8643, + "step": 30755 + }, + { + "epoch": 0.98, + "learning_rate": 1.4114239111536754e-08, + "loss": 0.8613, + "step": 30756 + }, + { + "epoch": 0.98, + "learning_rate": 1.4059279354935452e-08, + "loss": 0.9253, + "step": 30757 + }, + { + "epoch": 0.98, + "learning_rate": 1.400442673642588e-08, + "loss": 0.7983, + "step": 30758 + }, + { + "epoch": 0.98, + "learning_rate": 1.3949681256597569e-08, + "loss": 0.9651, + "step": 30759 + }, + { + "epoch": 0.98, + "learning_rate": 1.3895042916036716e-08, + "loss": 0.8682, + "step": 30760 + }, + { + "epoch": 0.98, + "learning_rate": 1.3840511715329519e-08, + "loss": 0.8608, + "step": 30761 + }, + { + "epoch": 0.98, + "learning_rate": 1.3786087655062174e-08, + "loss": 0.9946, + "step": 30762 + }, + { + "epoch": 0.98, + "learning_rate": 1.373177073581755e-08, + "loss": 0.8862, + "step": 30763 + }, + { + "epoch": 0.98, + "learning_rate": 1.3677560958179625e-08, + "loss": 0.7695, + "step": 30764 + }, + { + "epoch": 0.98, + "learning_rate": 1.3623458322727933e-08, + "loss": 1.0, + "step": 30765 + }, + { + "epoch": 0.98, + "learning_rate": 1.3569462830044233e-08, + "loss": 0.855, + "step": 30766 + }, + { + "epoch": 0.98, + "learning_rate": 1.351557448070806e-08, + "loss": 0.9429, + "step": 30767 + }, + { + "epoch": 0.98, + "learning_rate": 1.3461793275297841e-08, + "loss": 0.9141, + "step": 30768 + }, + { + "epoch": 0.98, + "learning_rate": 1.3408119214389781e-08, + "loss": 0.9639, + "step": 30769 + }, + { + "epoch": 0.98, + "learning_rate": 1.3354552298560086e-08, + "loss": 0.8691, + "step": 30770 + }, + { + "epoch": 0.98, + "learning_rate": 1.3301092528382741e-08, + "loss": 0.8359, + "step": 30771 + }, + { + "epoch": 0.98, + "learning_rate": 1.3247739904432843e-08, + "loss": 0.7817, + "step": 30772 + }, + { + "epoch": 0.98, + "learning_rate": 1.3194494427281046e-08, + "loss": 0.9614, + "step": 30773 + }, + { + "epoch": 0.98, + "learning_rate": 1.3141356097500225e-08, + "loss": 0.7764, + "step": 30774 + }, + { + "epoch": 0.98, + "learning_rate": 1.3088324915658812e-08, + "loss": 0.7935, + "step": 30775 + }, + { + "epoch": 0.98, + "learning_rate": 1.3035400882327464e-08, + "loss": 0.9268, + "step": 30776 + }, + { + "epoch": 0.98, + "learning_rate": 1.2982583998072395e-08, + "loss": 0.915, + "step": 30777 + }, + { + "epoch": 0.98, + "learning_rate": 1.292987426346204e-08, + "loss": 1.0225, + "step": 30778 + }, + { + "epoch": 0.98, + "learning_rate": 1.2877271679060388e-08, + "loss": 0.8921, + "step": 30779 + }, + { + "epoch": 0.98, + "learning_rate": 1.2824776245433657e-08, + "loss": 0.9736, + "step": 30780 + }, + { + "epoch": 0.98, + "learning_rate": 1.2772387963142508e-08, + "loss": 0.8784, + "step": 30781 + }, + { + "epoch": 0.98, + "learning_rate": 1.2720106832750934e-08, + "loss": 0.8979, + "step": 30782 + }, + { + "epoch": 0.98, + "learning_rate": 1.2667932854818488e-08, + "loss": 0.916, + "step": 30783 + }, + { + "epoch": 0.98, + "learning_rate": 1.2615866029906942e-08, + "loss": 0.9121, + "step": 30784 + }, + { + "epoch": 0.98, + "learning_rate": 1.256390635857252e-08, + "loss": 0.9067, + "step": 30785 + }, + { + "epoch": 0.98, + "learning_rate": 1.2512053841373662e-08, + "loss": 0.98, + "step": 30786 + }, + { + "epoch": 0.98, + "learning_rate": 1.24603084788677e-08, + "loss": 0.9673, + "step": 30787 + }, + { + "epoch": 0.98, + "learning_rate": 1.2408670271608636e-08, + "loss": 0.9595, + "step": 30788 + }, + { + "epoch": 0.98, + "learning_rate": 1.2357139220150471e-08, + "loss": 0.469, + "step": 30789 + }, + { + "epoch": 0.98, + "learning_rate": 1.2305715325047207e-08, + "loss": 0.9829, + "step": 30790 + }, + { + "epoch": 0.98, + "learning_rate": 1.2254398586849514e-08, + "loss": 0.8931, + "step": 30791 + }, + { + "epoch": 0.98, + "learning_rate": 1.2203189006108063e-08, + "loss": 0.9336, + "step": 30792 + }, + { + "epoch": 0.98, + "learning_rate": 1.2152086583372414e-08, + "loss": 0.9639, + "step": 30793 + }, + { + "epoch": 0.98, + "learning_rate": 1.2101091319191017e-08, + "loss": 0.9351, + "step": 30794 + }, + { + "epoch": 0.98, + "learning_rate": 1.205020321411121e-08, + "loss": 0.8979, + "step": 30795 + }, + { + "epoch": 0.98, + "learning_rate": 1.1999422268678117e-08, + "loss": 0.8752, + "step": 30796 + }, + { + "epoch": 0.98, + "learning_rate": 1.1948748483436856e-08, + "loss": 0.8252, + "step": 30797 + }, + { + "epoch": 0.98, + "learning_rate": 1.1898181858931434e-08, + "loss": 0.999, + "step": 30798 + }, + { + "epoch": 0.98, + "learning_rate": 1.1847722395704753e-08, + "loss": 0.8804, + "step": 30799 + }, + { + "epoch": 0.99, + "learning_rate": 1.1797370094297488e-08, + "loss": 0.9004, + "step": 30800 + }, + { + "epoch": 0.99, + "learning_rate": 1.1747124955249212e-08, + "loss": 0.8076, + "step": 30801 + }, + { + "epoch": 0.99, + "learning_rate": 1.16969869791006e-08, + "loss": 0.9331, + "step": 30802 + }, + { + "epoch": 0.99, + "learning_rate": 1.1646956166389e-08, + "loss": 0.9126, + "step": 30803 + }, + { + "epoch": 0.99, + "learning_rate": 1.1597032517650653e-08, + "loss": 0.9146, + "step": 30804 + }, + { + "epoch": 0.99, + "learning_rate": 1.1547216033421793e-08, + "loss": 1.0083, + "step": 30805 + }, + { + "epoch": 0.99, + "learning_rate": 1.149750671423644e-08, + "loss": 0.9463, + "step": 30806 + }, + { + "epoch": 0.99, + "learning_rate": 1.1447904560627499e-08, + "loss": 0.9541, + "step": 30807 + }, + { + "epoch": 0.99, + "learning_rate": 1.1398409573128989e-08, + "loss": 1.0278, + "step": 30808 + }, + { + "epoch": 0.99, + "learning_rate": 1.1349021752269374e-08, + "loss": 0.4775, + "step": 30809 + }, + { + "epoch": 0.99, + "learning_rate": 1.1299741098580453e-08, + "loss": 0.8018, + "step": 30810 + }, + { + "epoch": 0.99, + "learning_rate": 1.125056761259069e-08, + "loss": 0.9121, + "step": 30811 + }, + { + "epoch": 0.99, + "learning_rate": 1.1201501294827444e-08, + "loss": 0.853, + "step": 30812 + }, + { + "epoch": 0.99, + "learning_rate": 1.1152542145816959e-08, + "loss": 0.8745, + "step": 30813 + }, + { + "epoch": 0.99, + "learning_rate": 1.110369016608437e-08, + "loss": 0.7166, + "step": 30814 + }, + { + "epoch": 0.99, + "learning_rate": 1.1054945356153701e-08, + "loss": 0.5217, + "step": 30815 + }, + { + "epoch": 0.99, + "learning_rate": 1.100630771654787e-08, + "loss": 0.9463, + "step": 30816 + }, + { + "epoch": 0.99, + "learning_rate": 1.0957777247790902e-08, + "loss": 0.8647, + "step": 30817 + }, + { + "epoch": 0.99, + "learning_rate": 1.0909353950400158e-08, + "loss": 1.0132, + "step": 30818 + }, + { + "epoch": 0.99, + "learning_rate": 1.0861037824896337e-08, + "loss": 0.8857, + "step": 30819 + }, + { + "epoch": 0.99, + "learning_rate": 1.081282887179902e-08, + "loss": 0.9658, + "step": 30820 + }, + { + "epoch": 0.99, + "learning_rate": 1.076472709162446e-08, + "loss": 0.8159, + "step": 30821 + }, + { + "epoch": 0.99, + "learning_rate": 1.0716732484888914e-08, + "loss": 0.9907, + "step": 30822 + }, + { + "epoch": 0.99, + "learning_rate": 1.0668845052107524e-08, + "loss": 0.9131, + "step": 30823 + }, + { + "epoch": 0.99, + "learning_rate": 1.0621064793793213e-08, + "loss": 0.6971, + "step": 30824 + }, + { + "epoch": 0.99, + "learning_rate": 1.0573391710458902e-08, + "loss": 0.8994, + "step": 30825 + }, + { + "epoch": 0.99, + "learning_rate": 1.0525825802616407e-08, + "loss": 0.8921, + "step": 30826 + }, + { + "epoch": 0.99, + "learning_rate": 1.047836707077643e-08, + "loss": 0.8896, + "step": 30827 + }, + { + "epoch": 0.99, + "learning_rate": 1.0431015515448562e-08, + "loss": 0.7944, + "step": 30828 + }, + { + "epoch": 0.99, + "learning_rate": 1.0383771137139065e-08, + "loss": 0.874, + "step": 30829 + }, + { + "epoch": 0.99, + "learning_rate": 1.0336633936355312e-08, + "loss": 0.9409, + "step": 30830 + }, + { + "epoch": 0.99, + "learning_rate": 1.0289603913603563e-08, + "loss": 0.4434, + "step": 30831 + }, + { + "epoch": 0.99, + "learning_rate": 1.0242681069388971e-08, + "loss": 0.9341, + "step": 30832 + }, + { + "epoch": 0.99, + "learning_rate": 1.0195865404213356e-08, + "loss": 0.9062, + "step": 30833 + }, + { + "epoch": 0.99, + "learning_rate": 1.014915691858076e-08, + "loss": 0.9497, + "step": 30834 + }, + { + "epoch": 0.99, + "learning_rate": 1.0102555612991894e-08, + "loss": 0.9785, + "step": 30835 + }, + { + "epoch": 0.99, + "learning_rate": 1.0056061487945246e-08, + "loss": 0.9741, + "step": 30836 + }, + { + "epoch": 0.99, + "learning_rate": 1.0009674543941527e-08, + "loss": 0.9214, + "step": 30837 + }, + { + "epoch": 0.99, + "learning_rate": 9.963394781477009e-09, + "loss": 0.9595, + "step": 30838 + }, + { + "epoch": 0.99, + "learning_rate": 9.91722220104907e-09, + "loss": 1.0273, + "step": 30839 + }, + { + "epoch": 0.99, + "learning_rate": 9.87115680315287e-09, + "loss": 0.9302, + "step": 30840 + }, + { + "epoch": 0.99, + "learning_rate": 9.825198588282458e-09, + "loss": 0.8687, + "step": 30841 + }, + { + "epoch": 0.99, + "learning_rate": 9.779347556931885e-09, + "loss": 0.8726, + "step": 30842 + }, + { + "epoch": 0.99, + "learning_rate": 9.733603709591866e-09, + "loss": 0.8701, + "step": 30843 + }, + { + "epoch": 0.99, + "learning_rate": 9.687967046753122e-09, + "loss": 1.0713, + "step": 30844 + }, + { + "epoch": 0.99, + "learning_rate": 9.64243756890637e-09, + "loss": 0.9746, + "step": 30845 + }, + { + "epoch": 0.99, + "learning_rate": 9.597015276539002e-09, + "loss": 0.981, + "step": 30846 + }, + { + "epoch": 0.99, + "learning_rate": 9.5517001701384e-09, + "loss": 0.9456, + "step": 30847 + }, + { + "epoch": 0.99, + "learning_rate": 9.506492250191957e-09, + "loss": 0.8945, + "step": 30848 + }, + { + "epoch": 0.99, + "learning_rate": 9.461391517183726e-09, + "loss": 0.9395, + "step": 30849 + }, + { + "epoch": 0.99, + "learning_rate": 9.416397971597769e-09, + "loss": 0.9365, + "step": 30850 + }, + { + "epoch": 0.99, + "learning_rate": 9.371511613917029e-09, + "loss": 0.9219, + "step": 30851 + }, + { + "epoch": 0.99, + "learning_rate": 9.326732444623344e-09, + "loss": 0.9722, + "step": 30852 + }, + { + "epoch": 0.99, + "learning_rate": 9.28206046419633e-09, + "loss": 0.8608, + "step": 30853 + }, + { + "epoch": 0.99, + "learning_rate": 9.237495673114494e-09, + "loss": 0.9595, + "step": 30854 + }, + { + "epoch": 0.99, + "learning_rate": 9.193038071858562e-09, + "loss": 0.7876, + "step": 30855 + }, + { + "epoch": 0.99, + "learning_rate": 9.14868766090371e-09, + "loss": 0.9497, + "step": 30856 + }, + { + "epoch": 0.99, + "learning_rate": 9.104444440725114e-09, + "loss": 1.0273, + "step": 30857 + }, + { + "epoch": 0.99, + "learning_rate": 9.060308411800167e-09, + "loss": 0.9634, + "step": 30858 + }, + { + "epoch": 0.99, + "learning_rate": 9.016279574599607e-09, + "loss": 0.9824, + "step": 30859 + }, + { + "epoch": 0.99, + "learning_rate": 8.972357929596387e-09, + "loss": 0.9644, + "step": 30860 + }, + { + "epoch": 0.99, + "learning_rate": 8.928543477263462e-09, + "loss": 0.9326, + "step": 30861 + }, + { + "epoch": 0.99, + "learning_rate": 8.884836218069347e-09, + "loss": 0.8818, + "step": 30862 + }, + { + "epoch": 0.99, + "learning_rate": 8.841236152482557e-09, + "loss": 0.8887, + "step": 30863 + }, + { + "epoch": 0.99, + "learning_rate": 8.797743280972715e-09, + "loss": 0.8701, + "step": 30864 + }, + { + "epoch": 0.99, + "learning_rate": 8.754357604003893e-09, + "loss": 0.9722, + "step": 30865 + }, + { + "epoch": 0.99, + "learning_rate": 8.711079122044608e-09, + "loss": 0.8892, + "step": 30866 + }, + { + "epoch": 0.99, + "learning_rate": 8.66790783555671e-09, + "loss": 0.5349, + "step": 30867 + }, + { + "epoch": 0.99, + "learning_rate": 8.624843745004274e-09, + "loss": 1.0903, + "step": 30868 + }, + { + "epoch": 0.99, + "learning_rate": 8.581886850850262e-09, + "loss": 0.7988, + "step": 30869 + }, + { + "epoch": 0.99, + "learning_rate": 8.539037153554308e-09, + "loss": 0.916, + "step": 30870 + }, + { + "epoch": 0.99, + "learning_rate": 8.496294653576043e-09, + "loss": 0.9033, + "step": 30871 + }, + { + "epoch": 0.99, + "learning_rate": 8.453659351375099e-09, + "loss": 0.9141, + "step": 30872 + }, + { + "epoch": 0.99, + "learning_rate": 8.411131247407778e-09, + "loss": 1.0312, + "step": 30873 + }, + { + "epoch": 0.99, + "learning_rate": 8.368710342130382e-09, + "loss": 0.9131, + "step": 30874 + }, + { + "epoch": 0.99, + "learning_rate": 8.326396635999212e-09, + "loss": 0.8452, + "step": 30875 + }, + { + "epoch": 0.99, + "learning_rate": 8.28419012946724e-09, + "loss": 0.9116, + "step": 30876 + }, + { + "epoch": 0.99, + "learning_rate": 8.242090822988546e-09, + "loss": 0.8418, + "step": 30877 + }, + { + "epoch": 0.99, + "learning_rate": 8.200098717012772e-09, + "loss": 0.8984, + "step": 30878 + }, + { + "epoch": 0.99, + "learning_rate": 8.158213811991778e-09, + "loss": 1.0396, + "step": 30879 + }, + { + "epoch": 0.99, + "learning_rate": 8.116436108375203e-09, + "loss": 0.8445, + "step": 30880 + }, + { + "epoch": 0.99, + "learning_rate": 8.074765606610468e-09, + "loss": 0.9185, + "step": 30881 + }, + { + "epoch": 0.99, + "learning_rate": 8.033202307146104e-09, + "loss": 0.9756, + "step": 30882 + }, + { + "epoch": 0.99, + "learning_rate": 7.99174621042509e-09, + "loss": 0.9526, + "step": 30883 + }, + { + "epoch": 0.99, + "learning_rate": 7.950397316894842e-09, + "loss": 1.0078, + "step": 30884 + }, + { + "epoch": 0.99, + "learning_rate": 7.909155626998345e-09, + "loss": 0.8262, + "step": 30885 + }, + { + "epoch": 0.99, + "learning_rate": 7.868021141177463e-09, + "loss": 0.9097, + "step": 30886 + }, + { + "epoch": 0.99, + "learning_rate": 7.826993859875177e-09, + "loss": 0.873, + "step": 30887 + }, + { + "epoch": 0.99, + "learning_rate": 7.786073783528913e-09, + "loss": 0.8652, + "step": 30888 + }, + { + "epoch": 0.99, + "learning_rate": 7.745260912579433e-09, + "loss": 0.4419, + "step": 30889 + }, + { + "epoch": 0.99, + "learning_rate": 7.704555247465274e-09, + "loss": 0.8848, + "step": 30890 + }, + { + "epoch": 0.99, + "learning_rate": 7.66395678862164e-09, + "loss": 0.8999, + "step": 30891 + }, + { + "epoch": 0.99, + "learning_rate": 7.623465536484853e-09, + "loss": 0.9048, + "step": 30892 + }, + { + "epoch": 0.99, + "learning_rate": 7.583081491490118e-09, + "loss": 0.8604, + "step": 30893 + }, + { + "epoch": 0.99, + "learning_rate": 7.542804654069313e-09, + "loss": 0.8613, + "step": 30894 + }, + { + "epoch": 0.99, + "learning_rate": 7.502635024654314e-09, + "loss": 0.9121, + "step": 30895 + }, + { + "epoch": 0.99, + "learning_rate": 7.462572603678108e-09, + "loss": 1.0151, + "step": 30896 + }, + { + "epoch": 0.99, + "learning_rate": 7.422617391569242e-09, + "loss": 0.9849, + "step": 30897 + }, + { + "epoch": 0.99, + "learning_rate": 7.382769388755151e-09, + "loss": 0.8589, + "step": 30898 + }, + { + "epoch": 0.99, + "learning_rate": 7.343028595665491e-09, + "loss": 0.9023, + "step": 30899 + }, + { + "epoch": 0.99, + "learning_rate": 7.303395012725478e-09, + "loss": 0.9258, + "step": 30900 + }, + { + "epoch": 0.99, + "learning_rate": 7.263868640361438e-09, + "loss": 0.7839, + "step": 30901 + }, + { + "epoch": 0.99, + "learning_rate": 7.224449478996365e-09, + "loss": 0.9023, + "step": 30902 + }, + { + "epoch": 0.99, + "learning_rate": 7.185137529053254e-09, + "loss": 0.7979, + "step": 30903 + }, + { + "epoch": 0.99, + "learning_rate": 7.145932790953991e-09, + "loss": 1.0073, + "step": 30904 + }, + { + "epoch": 0.99, + "learning_rate": 7.1068352651204596e-09, + "loss": 0.4924, + "step": 30905 + }, + { + "epoch": 0.99, + "learning_rate": 7.067844951968994e-09, + "loss": 0.9629, + "step": 30906 + }, + { + "epoch": 0.99, + "learning_rate": 7.028961851921478e-09, + "loss": 0.9707, + "step": 30907 + }, + { + "epoch": 0.99, + "learning_rate": 6.990185965393137e-09, + "loss": 0.9062, + "step": 30908 + }, + { + "epoch": 0.99, + "learning_rate": 6.951517292800303e-09, + "loss": 0.9595, + "step": 30909 + }, + { + "epoch": 0.99, + "learning_rate": 6.912955834558199e-09, + "loss": 0.9961, + "step": 30910 + }, + { + "epoch": 0.99, + "learning_rate": 6.874501591079829e-09, + "loss": 0.896, + "step": 30911 + }, + { + "epoch": 0.99, + "learning_rate": 6.836154562778196e-09, + "loss": 0.8394, + "step": 30912 + }, + { + "epoch": 0.99, + "learning_rate": 6.7979147500651934e-09, + "loss": 1.0391, + "step": 30913 + }, + { + "epoch": 0.99, + "learning_rate": 6.759782153350492e-09, + "loss": 0.8066, + "step": 30914 + }, + { + "epoch": 0.99, + "learning_rate": 6.721756773043764e-09, + "loss": 0.9351, + "step": 30915 + }, + { + "epoch": 0.99, + "learning_rate": 6.683838609551352e-09, + "loss": 1.0137, + "step": 30916 + }, + { + "epoch": 0.99, + "learning_rate": 6.6460276632818175e-09, + "loss": 0.8516, + "step": 30917 + }, + { + "epoch": 0.99, + "learning_rate": 6.608323934640393e-09, + "loss": 0.8735, + "step": 30918 + }, + { + "epoch": 0.99, + "learning_rate": 6.570727424032308e-09, + "loss": 0.9761, + "step": 30919 + }, + { + "epoch": 0.99, + "learning_rate": 6.5332381318594654e-09, + "loss": 0.895, + "step": 30920 + }, + { + "epoch": 0.99, + "learning_rate": 6.495856058524874e-09, + "loss": 0.8262, + "step": 30921 + }, + { + "epoch": 0.99, + "learning_rate": 6.458581204429326e-09, + "loss": 0.8799, + "step": 30922 + }, + { + "epoch": 0.99, + "learning_rate": 6.421413569972501e-09, + "loss": 0.9502, + "step": 30923 + }, + { + "epoch": 0.99, + "learning_rate": 6.38435315555408e-09, + "loss": 0.8774, + "step": 30924 + }, + { + "epoch": 0.99, + "learning_rate": 6.347399961571521e-09, + "loss": 0.8325, + "step": 30925 + }, + { + "epoch": 0.99, + "learning_rate": 6.310553988420065e-09, + "loss": 0.9229, + "step": 30926 + }, + { + "epoch": 0.99, + "learning_rate": 6.273815236496062e-09, + "loss": 0.9395, + "step": 30927 + }, + { + "epoch": 0.99, + "learning_rate": 6.2371837061936394e-09, + "loss": 0.9912, + "step": 30928 + }, + { + "epoch": 0.99, + "learning_rate": 6.200659397906927e-09, + "loss": 0.8599, + "step": 30929 + }, + { + "epoch": 0.99, + "learning_rate": 6.164242312024505e-09, + "loss": 0.8457, + "step": 30930 + }, + { + "epoch": 0.99, + "learning_rate": 6.127932448940499e-09, + "loss": 0.751, + "step": 30931 + }, + { + "epoch": 0.99, + "learning_rate": 6.091729809042379e-09, + "loss": 0.8921, + "step": 30932 + }, + { + "epoch": 0.99, + "learning_rate": 6.0556343927198336e-09, + "loss": 1.0142, + "step": 30933 + }, + { + "epoch": 0.99, + "learning_rate": 6.019646200359219e-09, + "loss": 0.9136, + "step": 30934 + }, + { + "epoch": 0.99, + "learning_rate": 5.983765232346894e-09, + "loss": 0.8159, + "step": 30935 + }, + { + "epoch": 0.99, + "learning_rate": 5.9479914890692155e-09, + "loss": 0.6907, + "step": 30936 + }, + { + "epoch": 0.99, + "learning_rate": 5.9123249709069904e-09, + "loss": 0.9033, + "step": 30937 + }, + { + "epoch": 0.99, + "learning_rate": 5.876765678245466e-09, + "loss": 0.8472, + "step": 30938 + }, + { + "epoch": 0.99, + "learning_rate": 5.841313611465449e-09, + "loss": 0.9067, + "step": 30939 + }, + { + "epoch": 0.99, + "learning_rate": 5.805968770946635e-09, + "loss": 0.8301, + "step": 30940 + }, + { + "epoch": 0.99, + "learning_rate": 5.770731157068721e-09, + "loss": 0.896, + "step": 30941 + }, + { + "epoch": 0.99, + "learning_rate": 5.735600770210292e-09, + "loss": 1.0576, + "step": 30942 + }, + { + "epoch": 0.99, + "learning_rate": 5.700577610747715e-09, + "loss": 0.9248, + "step": 30943 + }, + { + "epoch": 0.99, + "learning_rate": 5.665661679056245e-09, + "loss": 1.0151, + "step": 30944 + }, + { + "epoch": 0.99, + "learning_rate": 5.630852975511136e-09, + "loss": 0.936, + "step": 30945 + }, + { + "epoch": 0.99, + "learning_rate": 5.596151500486535e-09, + "loss": 0.9922, + "step": 30946 + }, + { + "epoch": 0.99, + "learning_rate": 5.561557254353256e-09, + "loss": 0.9453, + "step": 30947 + }, + { + "epoch": 0.99, + "learning_rate": 5.5270702374832234e-09, + "loss": 0.8936, + "step": 30948 + }, + { + "epoch": 0.99, + "learning_rate": 5.4926904502461405e-09, + "loss": 0.8931, + "step": 30949 + }, + { + "epoch": 0.99, + "learning_rate": 5.4584178930117135e-09, + "loss": 0.9634, + "step": 30950 + }, + { + "epoch": 0.99, + "learning_rate": 5.4242525661474256e-09, + "loss": 0.9385, + "step": 30951 + }, + { + "epoch": 0.99, + "learning_rate": 5.390194470018539e-09, + "loss": 0.7952, + "step": 30952 + }, + { + "epoch": 0.99, + "learning_rate": 5.356243604991429e-09, + "loss": 0.936, + "step": 30953 + }, + { + "epoch": 0.99, + "learning_rate": 5.322399971431358e-09, + "loss": 0.7893, + "step": 30954 + }, + { + "epoch": 0.99, + "learning_rate": 5.288663569699148e-09, + "loss": 0.9883, + "step": 30955 + }, + { + "epoch": 0.99, + "learning_rate": 5.2550344001589535e-09, + "loss": 0.9121, + "step": 30956 + }, + { + "epoch": 0.99, + "learning_rate": 5.221512463169376e-09, + "loss": 0.8223, + "step": 30957 + }, + { + "epoch": 0.99, + "learning_rate": 5.188097759092347e-09, + "loss": 0.9282, + "step": 30958 + }, + { + "epoch": 0.99, + "learning_rate": 5.1547902882853605e-09, + "loss": 0.9009, + "step": 30959 + }, + { + "epoch": 0.99, + "learning_rate": 5.121590051104797e-09, + "loss": 0.8823, + "step": 30960 + }, + { + "epoch": 0.99, + "learning_rate": 5.088497047908147e-09, + "loss": 0.8281, + "step": 30961 + }, + { + "epoch": 0.99, + "learning_rate": 5.0555112790506845e-09, + "loss": 0.9795, + "step": 30962 + }, + { + "epoch": 0.99, + "learning_rate": 5.022632744885458e-09, + "loss": 0.9424, + "step": 30963 + }, + { + "epoch": 0.99, + "learning_rate": 4.98986144576552e-09, + "loss": 0.4507, + "step": 30964 + }, + { + "epoch": 0.99, + "learning_rate": 4.957197382041701e-09, + "loss": 0.8052, + "step": 30965 + }, + { + "epoch": 0.99, + "learning_rate": 4.92464055406594e-09, + "loss": 0.9639, + "step": 30966 + }, + { + "epoch": 0.99, + "learning_rate": 4.892190962186849e-09, + "loss": 0.9351, + "step": 30967 + }, + { + "epoch": 0.99, + "learning_rate": 4.859848606751927e-09, + "loss": 0.9458, + "step": 30968 + }, + { + "epoch": 0.99, + "learning_rate": 4.827613488109784e-09, + "loss": 0.9204, + "step": 30969 + }, + { + "epoch": 0.99, + "learning_rate": 4.795485606603478e-09, + "loss": 0.936, + "step": 30970 + }, + { + "epoch": 0.99, + "learning_rate": 4.763464962581621e-09, + "loss": 0.9302, + "step": 30971 + }, + { + "epoch": 0.99, + "learning_rate": 4.73155155638394e-09, + "loss": 0.8662, + "step": 30972 + }, + { + "epoch": 0.99, + "learning_rate": 4.699745388355715e-09, + "loss": 0.7876, + "step": 30973 + }, + { + "epoch": 0.99, + "learning_rate": 4.668046458835562e-09, + "loss": 0.9937, + "step": 30974 + }, + { + "epoch": 0.99, + "learning_rate": 4.636454768166543e-09, + "loss": 0.9395, + "step": 30975 + }, + { + "epoch": 0.99, + "learning_rate": 4.604970316685054e-09, + "loss": 0.9941, + "step": 30976 + }, + { + "epoch": 0.99, + "learning_rate": 4.573593104729712e-09, + "loss": 0.8428, + "step": 30977 + }, + { + "epoch": 0.99, + "learning_rate": 4.542323132638027e-09, + "loss": 1.0352, + "step": 30978 + }, + { + "epoch": 0.99, + "learning_rate": 4.511160400744174e-09, + "loss": 0.8066, + "step": 30979 + }, + { + "epoch": 0.99, + "learning_rate": 4.480104909383442e-09, + "loss": 0.9258, + "step": 30980 + }, + { + "epoch": 0.99, + "learning_rate": 4.449156658888898e-09, + "loss": 0.8711, + "step": 30981 + }, + { + "epoch": 0.99, + "learning_rate": 4.418315649592497e-09, + "loss": 1.0186, + "step": 30982 + }, + { + "epoch": 0.99, + "learning_rate": 4.387581881823977e-09, + "loss": 0.8481, + "step": 30983 + }, + { + "epoch": 0.99, + "learning_rate": 4.356955355915293e-09, + "loss": 0.9263, + "step": 30984 + }, + { + "epoch": 0.99, + "learning_rate": 4.326436072192852e-09, + "loss": 0.7876, + "step": 30985 + }, + { + "epoch": 0.99, + "learning_rate": 4.2960240309852794e-09, + "loss": 0.9849, + "step": 30986 + }, + { + "epoch": 0.99, + "learning_rate": 4.265719232620091e-09, + "loss": 0.9478, + "step": 30987 + }, + { + "epoch": 0.99, + "learning_rate": 4.235521677419252e-09, + "loss": 0.9766, + "step": 30988 + }, + { + "epoch": 0.99, + "learning_rate": 4.205431365709167e-09, + "loss": 0.9966, + "step": 30989 + }, + { + "epoch": 0.99, + "learning_rate": 4.175448297811802e-09, + "loss": 0.9297, + "step": 30990 + }, + { + "epoch": 0.99, + "learning_rate": 4.14557247404912e-09, + "loss": 0.9844, + "step": 30991 + }, + { + "epoch": 0.99, + "learning_rate": 4.115803894741977e-09, + "loss": 0.9668, + "step": 30992 + }, + { + "epoch": 0.99, + "learning_rate": 4.086142560209005e-09, + "loss": 0.9937, + "step": 30993 + }, + { + "epoch": 0.99, + "learning_rate": 4.056588470768841e-09, + "loss": 0.9512, + "step": 30994 + }, + { + "epoch": 0.99, + "learning_rate": 4.027141626739006e-09, + "loss": 0.9927, + "step": 30995 + }, + { + "epoch": 0.99, + "learning_rate": 3.997802028433695e-09, + "loss": 0.9873, + "step": 30996 + }, + { + "epoch": 0.99, + "learning_rate": 3.96856967617043e-09, + "loss": 0.4543, + "step": 30997 + }, + { + "epoch": 0.99, + "learning_rate": 3.939444570260076e-09, + "loss": 0.8828, + "step": 30998 + }, + { + "epoch": 0.99, + "learning_rate": 3.9104267110168235e-09, + "loss": 0.8496, + "step": 30999 + }, + { + "epoch": 0.99, + "learning_rate": 3.881516098751536e-09, + "loss": 0.7837, + "step": 31000 + }, + { + "epoch": 0.99, + "learning_rate": 3.8527127337750765e-09, + "loss": 0.9756, + "step": 31001 + }, + { + "epoch": 0.99, + "learning_rate": 3.8240166163949765e-09, + "loss": 0.9277, + "step": 31002 + }, + { + "epoch": 0.99, + "learning_rate": 3.795427746920987e-09, + "loss": 0.4634, + "step": 31003 + }, + { + "epoch": 0.99, + "learning_rate": 3.766946125657311e-09, + "loss": 0.8916, + "step": 31004 + }, + { + "epoch": 0.99, + "learning_rate": 3.73857175291259e-09, + "loss": 1.064, + "step": 31005 + }, + { + "epoch": 0.99, + "learning_rate": 3.7103046289888036e-09, + "loss": 1.002, + "step": 31006 + }, + { + "epoch": 0.99, + "learning_rate": 3.6821447541901534e-09, + "loss": 0.9678, + "step": 31007 + }, + { + "epoch": 0.99, + "learning_rate": 3.6540921288186206e-09, + "loss": 0.9155, + "step": 31008 + }, + { + "epoch": 0.99, + "learning_rate": 3.626146753176185e-09, + "loss": 0.8745, + "step": 31009 + }, + { + "epoch": 0.99, + "learning_rate": 3.598308627560387e-09, + "loss": 0.9746, + "step": 31010 + }, + { + "epoch": 0.99, + "learning_rate": 3.5705777522720976e-09, + "loss": 0.9219, + "step": 31011 + }, + { + "epoch": 0.99, + "learning_rate": 3.542954127607745e-09, + "loss": 0.8335, + "step": 31012 + }, + { + "epoch": 0.99, + "learning_rate": 3.5154377538637595e-09, + "loss": 0.8857, + "step": 31013 + }, + { + "epoch": 0.99, + "learning_rate": 3.4880286313365707e-09, + "loss": 0.9414, + "step": 31014 + }, + { + "epoch": 0.99, + "learning_rate": 3.460726760318167e-09, + "loss": 0.8374, + "step": 31015 + }, + { + "epoch": 0.99, + "learning_rate": 3.4335321411027577e-09, + "loss": 0.9429, + "step": 31016 + }, + { + "epoch": 0.99, + "learning_rate": 3.4064447739823312e-09, + "loss": 0.8984, + "step": 31017 + }, + { + "epoch": 0.99, + "learning_rate": 3.3794646592466563e-09, + "loss": 0.9736, + "step": 31018 + }, + { + "epoch": 0.99, + "learning_rate": 3.3525917971855002e-09, + "loss": 0.9795, + "step": 31019 + }, + { + "epoch": 0.99, + "learning_rate": 3.325826188087522e-09, + "loss": 0.9712, + "step": 31020 + }, + { + "epoch": 0.99, + "learning_rate": 3.2991678322391584e-09, + "loss": 0.8877, + "step": 31021 + }, + { + "epoch": 0.99, + "learning_rate": 3.2726167299279575e-09, + "loss": 0.9058, + "step": 31022 + }, + { + "epoch": 0.99, + "learning_rate": 3.2461728814370263e-09, + "loss": 0.9312, + "step": 31023 + }, + { + "epoch": 0.99, + "learning_rate": 3.2198362870505818e-09, + "loss": 0.9312, + "step": 31024 + }, + { + "epoch": 0.99, + "learning_rate": 3.1936069470517306e-09, + "loss": 0.8687, + "step": 31025 + }, + { + "epoch": 0.99, + "learning_rate": 3.1674848617224697e-09, + "loss": 0.9146, + "step": 31026 + }, + { + "epoch": 0.99, + "learning_rate": 3.1414700313414647e-09, + "loss": 0.8916, + "step": 31027 + }, + { + "epoch": 0.99, + "learning_rate": 3.115562456187382e-09, + "loss": 0.9282, + "step": 31028 + }, + { + "epoch": 0.99, + "learning_rate": 3.0897621365411078e-09, + "loss": 0.9731, + "step": 31029 + }, + { + "epoch": 0.99, + "learning_rate": 3.0640690726768672e-09, + "loss": 0.9287, + "step": 31030 + }, + { + "epoch": 0.99, + "learning_rate": 3.038483264871106e-09, + "loss": 0.8569, + "step": 31031 + }, + { + "epoch": 0.99, + "learning_rate": 3.0130047133980488e-09, + "loss": 0.7749, + "step": 31032 + }, + { + "epoch": 0.99, + "learning_rate": 2.987633418531921e-09, + "loss": 0.7603, + "step": 31033 + }, + { + "epoch": 0.99, + "learning_rate": 2.962369380543617e-09, + "loss": 0.917, + "step": 31034 + }, + { + "epoch": 0.99, + "learning_rate": 2.9372125997051416e-09, + "loss": 0.897, + "step": 31035 + }, + { + "epoch": 0.99, + "learning_rate": 2.9121630762862784e-09, + "loss": 0.8784, + "step": 31036 + }, + { + "epoch": 0.99, + "learning_rate": 2.887220810555702e-09, + "loss": 0.8281, + "step": 31037 + }, + { + "epoch": 0.99, + "learning_rate": 2.862385802780976e-09, + "loss": 0.9741, + "step": 31038 + }, + { + "epoch": 0.99, + "learning_rate": 2.8376580532285534e-09, + "loss": 0.8628, + "step": 31039 + }, + { + "epoch": 0.99, + "learning_rate": 2.8130375621637783e-09, + "loss": 0.8596, + "step": 31040 + }, + { + "epoch": 0.99, + "learning_rate": 2.7885243298508836e-09, + "loss": 0.8447, + "step": 31041 + }, + { + "epoch": 0.99, + "learning_rate": 2.7641183565529915e-09, + "loss": 0.9097, + "step": 31042 + }, + { + "epoch": 0.99, + "learning_rate": 2.7398196425310054e-09, + "loss": 0.4675, + "step": 31043 + }, + { + "epoch": 0.99, + "learning_rate": 2.715628188046937e-09, + "loss": 0.8511, + "step": 31044 + }, + { + "epoch": 0.99, + "learning_rate": 2.691543993359469e-09, + "loss": 1.0464, + "step": 31045 + }, + { + "epoch": 0.99, + "learning_rate": 2.667567058727283e-09, + "loss": 1.0591, + "step": 31046 + }, + { + "epoch": 0.99, + "learning_rate": 2.6436973844079506e-09, + "loss": 0.9043, + "step": 31047 + }, + { + "epoch": 0.99, + "learning_rate": 2.6199349706568233e-09, + "loss": 0.9272, + "step": 31048 + }, + { + "epoch": 0.99, + "learning_rate": 2.5962798177292524e-09, + "loss": 1.001, + "step": 31049 + }, + { + "epoch": 0.99, + "learning_rate": 2.5727319258794794e-09, + "loss": 1.0034, + "step": 31050 + }, + { + "epoch": 0.99, + "learning_rate": 2.5492912953584137e-09, + "loss": 0.855, + "step": 31051 + }, + { + "epoch": 0.99, + "learning_rate": 2.5259579264202972e-09, + "loss": 0.8569, + "step": 31052 + }, + { + "epoch": 0.99, + "learning_rate": 2.5027318193138193e-09, + "loss": 0.9512, + "step": 31053 + }, + { + "epoch": 0.99, + "learning_rate": 2.4796129742876705e-09, + "loss": 0.9023, + "step": 31054 + }, + { + "epoch": 0.99, + "learning_rate": 2.4566013915905407e-09, + "loss": 0.9609, + "step": 31055 + }, + { + "epoch": 0.99, + "learning_rate": 2.433697071470009e-09, + "loss": 0.8569, + "step": 31056 + }, + { + "epoch": 0.99, + "learning_rate": 2.4109000141703252e-09, + "loss": 0.9897, + "step": 31057 + }, + { + "epoch": 0.99, + "learning_rate": 2.3882102199379587e-09, + "loss": 0.9116, + "step": 31058 + }, + { + "epoch": 0.99, + "learning_rate": 2.3656276890138275e-09, + "loss": 0.9854, + "step": 31059 + }, + { + "epoch": 0.99, + "learning_rate": 2.3431524216432912e-09, + "loss": 1.0044, + "step": 31060 + }, + { + "epoch": 0.99, + "learning_rate": 2.3207844180650473e-09, + "loss": 0.9692, + "step": 31061 + }, + { + "epoch": 0.99, + "learning_rate": 2.298523678520015e-09, + "loss": 0.8691, + "step": 31062 + }, + { + "epoch": 0.99, + "learning_rate": 2.2763702032468917e-09, + "loss": 0.7822, + "step": 31063 + }, + { + "epoch": 0.99, + "learning_rate": 2.2543239924832648e-09, + "loss": 0.8999, + "step": 31064 + }, + { + "epoch": 0.99, + "learning_rate": 2.232385046465613e-09, + "loss": 1.1274, + "step": 31065 + }, + { + "epoch": 0.99, + "learning_rate": 2.210553365429302e-09, + "loss": 0.9375, + "step": 31066 + }, + { + "epoch": 0.99, + "learning_rate": 2.1888289496097005e-09, + "loss": 0.9082, + "step": 31067 + }, + { + "epoch": 0.99, + "learning_rate": 2.1672117992388443e-09, + "loss": 0.8953, + "step": 31068 + }, + { + "epoch": 0.99, + "learning_rate": 2.1457019145476597e-09, + "loss": 0.8965, + "step": 31069 + }, + { + "epoch": 0.99, + "learning_rate": 2.124299295769294e-09, + "loss": 0.835, + "step": 31070 + }, + { + "epoch": 0.99, + "learning_rate": 2.103003943131343e-09, + "loss": 0.9482, + "step": 31071 + }, + { + "epoch": 0.99, + "learning_rate": 2.0818158568625125e-09, + "loss": 0.9663, + "step": 31072 + }, + { + "epoch": 0.99, + "learning_rate": 2.060735037191508e-09, + "loss": 0.896, + "step": 31073 + }, + { + "epoch": 0.99, + "learning_rate": 2.0397614843437053e-09, + "loss": 0.8535, + "step": 31074 + }, + { + "epoch": 0.99, + "learning_rate": 2.018895198543369e-09, + "loss": 0.8862, + "step": 31075 + }, + { + "epoch": 0.99, + "learning_rate": 1.998136180015875e-09, + "loss": 0.9307, + "step": 31076 + }, + { + "epoch": 0.99, + "learning_rate": 1.9774844289832674e-09, + "loss": 0.9229, + "step": 31077 + }, + { + "epoch": 0.99, + "learning_rate": 1.9569399456664806e-09, + "loss": 0.9219, + "step": 31078 + }, + { + "epoch": 0.99, + "learning_rate": 1.9365027302864494e-09, + "loss": 0.8926, + "step": 31079 + }, + { + "epoch": 0.99, + "learning_rate": 1.916172783061887e-09, + "loss": 0.9404, + "step": 31080 + }, + { + "epoch": 0.99, + "learning_rate": 1.895950104212618e-09, + "loss": 0.9409, + "step": 31081 + }, + { + "epoch": 0.99, + "learning_rate": 1.875834693954026e-09, + "loss": 0.9209, + "step": 31082 + }, + { + "epoch": 0.99, + "learning_rate": 1.855826552501494e-09, + "loss": 0.853, + "step": 31083 + }, + { + "epoch": 0.99, + "learning_rate": 1.8359256800715152e-09, + "loss": 0.9048, + "step": 31084 + }, + { + "epoch": 0.99, + "learning_rate": 1.8161320768761427e-09, + "loss": 0.998, + "step": 31085 + }, + { + "epoch": 0.99, + "learning_rate": 1.7964457431285387e-09, + "loss": 0.8008, + "step": 31086 + }, + { + "epoch": 0.99, + "learning_rate": 1.776866679039646e-09, + "loss": 0.4766, + "step": 31087 + }, + { + "epoch": 0.99, + "learning_rate": 1.7573948848192968e-09, + "loss": 0.9561, + "step": 31088 + }, + { + "epoch": 0.99, + "learning_rate": 1.738030360677323e-09, + "loss": 0.9033, + "step": 31089 + }, + { + "epoch": 0.99, + "learning_rate": 1.718773106819116e-09, + "loss": 0.896, + "step": 31090 + }, + { + "epoch": 0.99, + "learning_rate": 1.6996231234545079e-09, + "loss": 0.897, + "step": 31091 + }, + { + "epoch": 0.99, + "learning_rate": 1.6805804107866696e-09, + "loss": 1.0317, + "step": 31092 + }, + { + "epoch": 0.99, + "learning_rate": 1.6616449690209923e-09, + "loss": 0.8149, + "step": 31093 + }, + { + "epoch": 0.99, + "learning_rate": 1.6428167983595366e-09, + "loss": 0.9702, + "step": 31094 + }, + { + "epoch": 0.99, + "learning_rate": 1.6240958990054735e-09, + "loss": 0.7798, + "step": 31095 + }, + { + "epoch": 0.99, + "learning_rate": 1.6054822711597528e-09, + "loss": 0.8047, + "step": 31096 + }, + { + "epoch": 0.99, + "learning_rate": 1.586975915021105e-09, + "loss": 0.8442, + "step": 31097 + }, + { + "epoch": 0.99, + "learning_rate": 1.5685768307882598e-09, + "loss": 0.9639, + "step": 31098 + }, + { + "epoch": 0.99, + "learning_rate": 1.550285018658837e-09, + "loss": 0.9663, + "step": 31099 + }, + { + "epoch": 0.99, + "learning_rate": 1.5321004788293459e-09, + "loss": 0.4407, + "step": 31100 + }, + { + "epoch": 0.99, + "learning_rate": 1.5140232114940757e-09, + "loss": 0.894, + "step": 31101 + }, + { + "epoch": 0.99, + "learning_rate": 1.4960532168484254e-09, + "loss": 0.7942, + "step": 31102 + }, + { + "epoch": 0.99, + "learning_rate": 1.4781904950833536e-09, + "loss": 0.9712, + "step": 31103 + }, + { + "epoch": 0.99, + "learning_rate": 1.4604350463920392e-09, + "loss": 0.9253, + "step": 31104 + }, + { + "epoch": 0.99, + "learning_rate": 1.4427868709654402e-09, + "loss": 0.9116, + "step": 31105 + }, + { + "epoch": 0.99, + "learning_rate": 1.4252459689900744e-09, + "loss": 0.8862, + "step": 31106 + }, + { + "epoch": 0.99, + "learning_rate": 1.4078123406569e-09, + "loss": 0.9033, + "step": 31107 + }, + { + "epoch": 0.99, + "learning_rate": 1.3904859861524345e-09, + "loss": 0.9985, + "step": 31108 + }, + { + "epoch": 0.99, + "learning_rate": 1.373266905660975e-09, + "loss": 0.9287, + "step": 31109 + }, + { + "epoch": 0.99, + "learning_rate": 1.3561550993690387e-09, + "loss": 0.9302, + "step": 31110 + }, + { + "epoch": 0.99, + "learning_rate": 1.3391505674598126e-09, + "loss": 0.9028, + "step": 31111 + }, + { + "epoch": 0.99, + "learning_rate": 1.3222533101153734e-09, + "loss": 0.8516, + "step": 31112 + }, + { + "epoch": 1.0, + "learning_rate": 1.3054633275177976e-09, + "loss": 1.0732, + "step": 31113 + }, + { + "epoch": 1.0, + "learning_rate": 1.2887806198458307e-09, + "loss": 0.9146, + "step": 31114 + }, + { + "epoch": 1.0, + "learning_rate": 1.2722051872793295e-09, + "loss": 0.9131, + "step": 31115 + }, + { + "epoch": 1.0, + "learning_rate": 1.2557370299959293e-09, + "loss": 0.8794, + "step": 31116 + }, + { + "epoch": 1.0, + "learning_rate": 1.2393761481721556e-09, + "loss": 0.9961, + "step": 31117 + }, + { + "epoch": 1.0, + "learning_rate": 1.2231225419845338e-09, + "loss": 0.8682, + "step": 31118 + }, + { + "epoch": 1.0, + "learning_rate": 1.2069762116062588e-09, + "loss": 0.8657, + "step": 31119 + }, + { + "epoch": 1.0, + "learning_rate": 1.1909371572105255e-09, + "loss": 0.9326, + "step": 31120 + }, + { + "epoch": 1.0, + "learning_rate": 1.1750053789705284e-09, + "loss": 0.9692, + "step": 31121 + }, + { + "epoch": 1.0, + "learning_rate": 1.1591808770572422e-09, + "loss": 0.9106, + "step": 31122 + }, + { + "epoch": 1.0, + "learning_rate": 1.1434636516383102e-09, + "loss": 1.0161, + "step": 31123 + }, + { + "epoch": 1.0, + "learning_rate": 1.127853702884707e-09, + "loss": 0.8838, + "step": 31124 + }, + { + "epoch": 1.0, + "learning_rate": 1.1123510309629659e-09, + "loss": 0.9658, + "step": 31125 + }, + { + "epoch": 1.0, + "learning_rate": 1.0969556360396204e-09, + "loss": 0.9062, + "step": 31126 + }, + { + "epoch": 1.0, + "learning_rate": 1.0816675182789837e-09, + "loss": 1.0024, + "step": 31127 + }, + { + "epoch": 1.0, + "learning_rate": 1.066486677846479e-09, + "loss": 0.8926, + "step": 31128 + }, + { + "epoch": 1.0, + "learning_rate": 1.0514131149041983e-09, + "loss": 0.8843, + "step": 31129 + }, + { + "epoch": 1.0, + "learning_rate": 1.0364468296142348e-09, + "loss": 0.8208, + "step": 31130 + }, + { + "epoch": 1.0, + "learning_rate": 1.0215878221364605e-09, + "loss": 0.9399, + "step": 31131 + }, + { + "epoch": 1.0, + "learning_rate": 1.0068360926307474e-09, + "loss": 1.0405, + "step": 31132 + }, + { + "epoch": 1.0, + "learning_rate": 9.921916412547473e-10, + "loss": 0.9707, + "step": 31133 + }, + { + "epoch": 1.0, + "learning_rate": 9.776544681672218e-10, + "loss": 0.9443, + "step": 31134 + }, + { + "epoch": 1.0, + "learning_rate": 9.632245735224922e-10, + "loss": 0.814, + "step": 31135 + }, + { + "epoch": 1.0, + "learning_rate": 9.489019574759894e-10, + "loss": 0.7661, + "step": 31136 + }, + { + "epoch": 1.0, + "learning_rate": 9.346866201820348e-10, + "loss": 0.8267, + "step": 31137 + }, + { + "epoch": 1.0, + "learning_rate": 9.205785617916185e-10, + "loss": 1.0083, + "step": 31138 + }, + { + "epoch": 1.0, + "learning_rate": 9.065777824579514e-10, + "loss": 0.8896, + "step": 31139 + }, + { + "epoch": 1.0, + "learning_rate": 8.926842823298032e-10, + "loss": 1.0132, + "step": 31140 + }, + { + "epoch": 1.0, + "learning_rate": 8.78898061555944e-10, + "loss": 0.9946, + "step": 31141 + }, + { + "epoch": 1.0, + "learning_rate": 8.652191202862537e-10, + "loss": 0.9307, + "step": 31142 + }, + { + "epoch": 1.0, + "learning_rate": 8.516474586650614e-10, + "loss": 0.9805, + "step": 31143 + }, + { + "epoch": 1.0, + "learning_rate": 8.381830768400267e-10, + "loss": 0.9585, + "step": 31144 + }, + { + "epoch": 1.0, + "learning_rate": 8.248259749543686e-10, + "loss": 0.9507, + "step": 31145 + }, + { + "epoch": 1.0, + "learning_rate": 8.115761531524157e-10, + "loss": 0.8799, + "step": 31146 + }, + { + "epoch": 1.0, + "learning_rate": 7.984336115751667e-10, + "loss": 0.896, + "step": 31147 + }, + { + "epoch": 1.0, + "learning_rate": 7.853983503636198e-10, + "loss": 0.9849, + "step": 31148 + }, + { + "epoch": 1.0, + "learning_rate": 7.724703696587732e-10, + "loss": 0.8745, + "step": 31149 + }, + { + "epoch": 1.0, + "learning_rate": 7.596496695982947e-10, + "loss": 1.0259, + "step": 31150 + }, + { + "epoch": 1.0, + "learning_rate": 7.469362503209621e-10, + "loss": 0.8325, + "step": 31151 + }, + { + "epoch": 1.0, + "learning_rate": 7.343301119622226e-10, + "loss": 0.8667, + "step": 31152 + }, + { + "epoch": 1.0, + "learning_rate": 7.218312546575234e-10, + "loss": 0.8755, + "step": 31153 + }, + { + "epoch": 1.0, + "learning_rate": 7.094396785400915e-10, + "loss": 0.8406, + "step": 31154 + }, + { + "epoch": 1.0, + "learning_rate": 6.971553837442635e-10, + "loss": 0.8647, + "step": 31155 + }, + { + "epoch": 1.0, + "learning_rate": 6.849783704010459e-10, + "loss": 0.8701, + "step": 31156 + }, + { + "epoch": 1.0, + "learning_rate": 6.729086386414452e-10, + "loss": 0.7844, + "step": 31157 + }, + { + "epoch": 1.0, + "learning_rate": 6.609461885953572e-10, + "loss": 0.8574, + "step": 31158 + }, + { + "epoch": 1.0, + "learning_rate": 6.490910203904577e-10, + "loss": 0.8911, + "step": 31159 + }, + { + "epoch": 1.0, + "learning_rate": 6.373431341533121e-10, + "loss": 0.8789, + "step": 31160 + }, + { + "epoch": 1.0, + "learning_rate": 6.25702530011596e-10, + "loss": 0.8574, + "step": 31161 + }, + { + "epoch": 1.0, + "learning_rate": 6.141692080885442e-10, + "loss": 0.7935, + "step": 31162 + }, + { + "epoch": 1.0, + "learning_rate": 6.027431685085016e-10, + "loss": 0.8247, + "step": 31163 + }, + { + "epoch": 1.0, + "learning_rate": 5.914244113958134e-10, + "loss": 0.8447, + "step": 31164 + }, + { + "epoch": 1.0, + "learning_rate": 5.802129368692732e-10, + "loss": 0.9473, + "step": 31165 + }, + { + "epoch": 1.0, + "learning_rate": 5.691087450498955e-10, + "loss": 0.9014, + "step": 31166 + }, + { + "epoch": 1.0, + "learning_rate": 5.581118360575844e-10, + "loss": 0.929, + "step": 31167 + }, + { + "epoch": 1.0, + "learning_rate": 5.472222100100233e-10, + "loss": 0.8804, + "step": 31168 + }, + { + "epoch": 1.0, + "learning_rate": 5.364398670237859e-10, + "loss": 0.9453, + "step": 31169 + }, + { + "epoch": 1.0, + "learning_rate": 5.25764807214335e-10, + "loss": 0.8569, + "step": 31170 + }, + { + "epoch": 1.0, + "learning_rate": 5.151970306971343e-10, + "loss": 0.978, + "step": 31171 + }, + { + "epoch": 1.0, + "learning_rate": 5.047365375843161e-10, + "loss": 0.9727, + "step": 31172 + }, + { + "epoch": 1.0, + "learning_rate": 4.94383327989123e-10, + "loss": 1.0029, + "step": 31173 + }, + { + "epoch": 1.0, + "learning_rate": 4.841374020225775e-10, + "loss": 0.9385, + "step": 31174 + }, + { + "epoch": 1.0, + "learning_rate": 4.739987597945917e-10, + "loss": 0.9717, + "step": 31175 + }, + { + "epoch": 1.0, + "learning_rate": 4.63967401412857e-10, + "loss": 1.0464, + "step": 31176 + }, + { + "epoch": 1.0, + "learning_rate": 4.5404332698617546e-10, + "loss": 0.7363, + "step": 31177 + }, + { + "epoch": 1.0, + "learning_rate": 4.4422653662112847e-10, + "loss": 0.8911, + "step": 31178 + }, + { + "epoch": 1.0, + "learning_rate": 4.3451703042207694e-10, + "loss": 0.8862, + "step": 31179 + }, + { + "epoch": 1.0, + "learning_rate": 4.2491480849338187e-10, + "loss": 0.9814, + "step": 31180 + }, + { + "epoch": 1.0, + "learning_rate": 4.1541987093829396e-10, + "loss": 0.9331, + "step": 31181 + }, + { + "epoch": 1.0, + "learning_rate": 4.0603221786006396e-10, + "loss": 0.9624, + "step": 31182 + }, + { + "epoch": 1.0, + "learning_rate": 3.9675184935639157e-10, + "loss": 0.9844, + "step": 31183 + }, + { + "epoch": 1.0, + "learning_rate": 3.875787655294172e-10, + "loss": 0.9111, + "step": 31184 + }, + { + "epoch": 1.0, + "learning_rate": 3.7851296647684056e-10, + "loss": 0.8955, + "step": 31185 + }, + { + "epoch": 1.0, + "learning_rate": 3.69554452295251e-10, + "loss": 0.9346, + "step": 31186 + }, + { + "epoch": 1.0, + "learning_rate": 3.6070322308234817e-10, + "loss": 0.937, + "step": 31187 + }, + { + "epoch": 1.0, + "learning_rate": 3.519592789313908e-10, + "loss": 0.8594, + "step": 31188 + }, + { + "epoch": 1.0, + "learning_rate": 3.4332261993674787e-10, + "loss": 0.9697, + "step": 31189 + }, + { + "epoch": 1.0, + "learning_rate": 3.3479324619056785e-10, + "loss": 0.8936, + "step": 31190 + }, + { + "epoch": 1.0, + "learning_rate": 3.263711577861095e-10, + "loss": 0.9048, + "step": 31191 + }, + { + "epoch": 1.0, + "learning_rate": 3.1805635481219064e-10, + "loss": 0.9932, + "step": 31192 + }, + { + "epoch": 1.0, + "learning_rate": 3.0984883735762915e-10, + "loss": 0.8877, + "step": 31193 + }, + { + "epoch": 1.0, + "learning_rate": 3.017486055123531e-10, + "loss": 0.7678, + "step": 31194 + }, + { + "epoch": 1.0, + "learning_rate": 2.937556593618496e-10, + "loss": 0.874, + "step": 31195 + }, + { + "epoch": 1.0, + "learning_rate": 2.858699989927161e-10, + "loss": 0.9253, + "step": 31196 + }, + { + "epoch": 1.0, + "learning_rate": 2.7809162448821924e-10, + "loss": 0.918, + "step": 31197 + }, + { + "epoch": 1.0, + "learning_rate": 2.704205359327361e-10, + "loss": 0.9639, + "step": 31198 + }, + { + "epoch": 1.0, + "learning_rate": 2.6285673340953335e-10, + "loss": 1.0034, + "step": 31199 + }, + { + "epoch": 1.0, + "learning_rate": 2.5540021699743676e-10, + "loss": 0.8496, + "step": 31200 + }, + { + "epoch": 1.0, + "learning_rate": 2.4805098677860296e-10, + "loss": 0.9204, + "step": 31201 + }, + { + "epoch": 1.0, + "learning_rate": 2.408090428307475e-10, + "loss": 0.8677, + "step": 31202 + }, + { + "epoch": 1.0, + "learning_rate": 2.3367438523158593e-10, + "loss": 0.8555, + "step": 31203 + }, + { + "epoch": 1.0, + "learning_rate": 2.2664701405772372e-10, + "loss": 1.0225, + "step": 31204 + }, + { + "epoch": 1.0, + "learning_rate": 2.1972692938576623e-10, + "loss": 0.8892, + "step": 31205 + }, + { + "epoch": 1.0, + "learning_rate": 2.1291413128787796e-10, + "loss": 0.855, + "step": 31206 + }, + { + "epoch": 1.0, + "learning_rate": 2.0620861983844388e-10, + "loss": 0.8716, + "step": 31207 + }, + { + "epoch": 1.0, + "learning_rate": 1.9961039510962844e-10, + "loss": 0.9507, + "step": 31208 + }, + { + "epoch": 1.0, + "learning_rate": 1.931194571713757e-10, + "loss": 0.854, + "step": 31209 + }, + { + "epoch": 1.0, + "learning_rate": 1.8673580609362973e-10, + "loss": 0.9062, + "step": 31210 + }, + { + "epoch": 1.0, + "learning_rate": 1.8045944194522435e-10, + "loss": 0.8152, + "step": 31211 + }, + { + "epoch": 1.0, + "learning_rate": 1.7429036479277295e-10, + "loss": 0.4468, + "step": 31212 + }, + { + "epoch": 1.0, + "learning_rate": 1.6822857470288888e-10, + "loss": 0.9771, + "step": 31213 + }, + { + "epoch": 1.0, + "learning_rate": 1.6227407174107535e-10, + "loss": 0.8486, + "step": 31214 + }, + { + "epoch": 1.0, + "learning_rate": 1.5642685597061502e-10, + "loss": 0.9106, + "step": 31215 + }, + { + "epoch": 1.0, + "learning_rate": 1.5068692745479064e-10, + "loss": 0.9062, + "step": 31216 + }, + { + "epoch": 1.0, + "learning_rate": 1.4505428625466444e-10, + "loss": 0.9219, + "step": 31217 + }, + { + "epoch": 1.0, + "learning_rate": 1.395289324301885e-10, + "loss": 0.874, + "step": 31218 + }, + { + "epoch": 1.0, + "learning_rate": 1.3411086604242507e-10, + "loss": 0.9307, + "step": 31219 + }, + { + "epoch": 1.0, + "learning_rate": 1.288000871468853e-10, + "loss": 0.96, + "step": 31220 + }, + { + "epoch": 1.0, + "learning_rate": 1.2359659580352123e-10, + "loss": 0.9199, + "step": 31221 + }, + { + "epoch": 1.0, + "learning_rate": 1.1850039206562357e-10, + "loss": 0.8389, + "step": 31222 + }, + { + "epoch": 1.0, + "learning_rate": 1.1351147598981371e-10, + "loss": 0.9629, + "step": 31223 + }, + { + "epoch": 1.0, + "learning_rate": 1.0862984762716189e-10, + "loss": 0.8726, + "step": 31224 + }, + { + "epoch": 1.0, + "learning_rate": 1.0385550703317926e-10, + "loss": 0.9248, + "step": 31225 + }, + { + "epoch": 1.0, + "learning_rate": 9.918845425671564e-11, + "loss": 0.8706, + "step": 31226 + }, + { + "epoch": 1.0, + "learning_rate": 9.462868934884128e-11, + "loss": 0.8218, + "step": 31227 + }, + { + "epoch": 1.0, + "learning_rate": 9.017621235840601e-11, + "loss": 1.0063, + "step": 31228 + }, + { + "epoch": 1.0, + "learning_rate": 8.583102333203918e-11, + "loss": 0.9966, + "step": 31229 + }, + { + "epoch": 1.0, + "learning_rate": 8.159312231859062e-11, + "loss": 0.9194, + "step": 31230 + }, + { + "epoch": 1.0, + "learning_rate": 7.746250936135901e-11, + "loss": 0.771, + "step": 31231 + }, + { + "epoch": 1.0, + "learning_rate": 7.343918450586352e-11, + "loss": 1.0161, + "step": 31232 + }, + { + "epoch": 1.0, + "learning_rate": 6.952314779540281e-11, + "loss": 0.9229, + "step": 31233 + }, + { + "epoch": 1.0, + "learning_rate": 6.571439927105517e-11, + "loss": 0.9668, + "step": 31234 + }, + { + "epoch": 1.0, + "learning_rate": 6.201293897500904e-11, + "loss": 0.8677, + "step": 31235 + }, + { + "epoch": 1.0, + "learning_rate": 5.841876694501203e-11, + "loss": 0.918, + "step": 31236 + }, + { + "epoch": 1.0, + "learning_rate": 5.493188322214238e-11, + "loss": 0.9653, + "step": 31237 + }, + { + "epoch": 1.0, + "learning_rate": 5.155228784192723e-11, + "loss": 0.9019, + "step": 31238 + }, + { + "epoch": 1.0, + "learning_rate": 4.827998084100394e-11, + "loss": 0.9146, + "step": 31239 + }, + { + "epoch": 1.0, + "learning_rate": 4.511496225378942e-11, + "loss": 1.0073, + "step": 31240 + }, + { + "epoch": 1.0, + "learning_rate": 4.205723211581081e-11, + "loss": 0.9497, + "step": 31241 + }, + { + "epoch": 1.0, + "learning_rate": 3.910679045926458e-11, + "loss": 0.9658, + "step": 31242 + }, + { + "epoch": 1.0, + "learning_rate": 3.626363731634719e-11, + "loss": 0.9463, + "step": 31243 + }, + { + "epoch": 1.0, + "learning_rate": 3.3527772715924445e-11, + "loss": 1.0508, + "step": 31244 + }, + { + "epoch": 1.0, + "learning_rate": 3.089919668908259e-11, + "loss": 0.9692, + "step": 31245 + }, + { + "epoch": 1.0, + "learning_rate": 2.8377909262466975e-11, + "loss": 0.8083, + "step": 31246 + }, + { + "epoch": 1.0, + "learning_rate": 2.59639104649434e-11, + "loss": 0.9561, + "step": 31247 + }, + { + "epoch": 1.0, + "learning_rate": 2.3657200319826545e-11, + "loss": 0.8643, + "step": 31248 + }, + { + "epoch": 1.0, + "learning_rate": 2.1457778854871992e-11, + "loss": 0.938, + "step": 31249 + }, + { + "epoch": 1.0, + "learning_rate": 1.9365646092284196e-11, + "loss": 0.9062, + "step": 31250 + }, + { + "epoch": 1.0, + "learning_rate": 1.73808020531574e-11, + "loss": 0.9438, + "step": 31251 + }, + { + "epoch": 1.0, + "learning_rate": 1.5503246760806278e-11, + "loss": 0.9058, + "step": 31252 + }, + { + "epoch": 1.0, + "learning_rate": 1.3732980235214855e-11, + "loss": 0.7856, + "step": 31253 + }, + { + "epoch": 1.0, + "learning_rate": 1.2070002494146694e-11, + "loss": 1.0078, + "step": 31254 + }, + { + "epoch": 1.0, + "learning_rate": 1.0514313556475587e-11, + "loss": 0.876, + "step": 31255 + }, + { + "epoch": 1.0, + "learning_rate": 9.065913438854879e-12, + "loss": 0.8901, + "step": 31256 + }, + { + "epoch": 1.0, + "learning_rate": 7.72480215571747e-12, + "loss": 0.999, + "step": 31257 + }, + { + "epoch": 1.0, + "learning_rate": 6.490979722606483e-12, + "loss": 0.8828, + "step": 31258 + }, + { + "epoch": 1.0, + "learning_rate": 5.364446152844593e-12, + "loss": 0.8979, + "step": 31259 + }, + { + "epoch": 1.0, + "learning_rate": 4.345201457534032e-12, + "loss": 0.7759, + "step": 31260 + }, + { + "epoch": 1.0, + "learning_rate": 3.433245647777028e-12, + "loss": 0.7168, + "step": 31261 + }, + { + "epoch": 1.0, + "learning_rate": 2.628578734675813e-12, + "loss": 0.9805, + "step": 31262 + }, + { + "epoch": 1.0, + "learning_rate": 1.9312007248917243e-12, + "loss": 0.8965, + "step": 31263 + }, + { + "epoch": 1.0, + "learning_rate": 1.3411116284167691e-12, + "loss": 0.8711, + "step": 31264 + }, + { + "epoch": 1.0, + "learning_rate": 8.583114485816168e-13, + "loss": 0.9302, + "step": 31265 + }, + { + "epoch": 1.0, + "learning_rate": 4.828001931578286e-13, + "loss": 0.916, + "step": 31266 + }, + { + "epoch": 1.0, + "learning_rate": 2.1457786436585027e-13, + "loss": 0.96, + "step": 31267 + }, + { + "epoch": 1.0, + "learning_rate": 5.364446664657408e-14, + "loss": 0.8325, + "step": 31268 + }, + { + "epoch": 1.0, + "learning_rate": 0.0, + "loss": 0.8911, + "step": 31269 + }, + { + "epoch": 1.0, + "step": 31269, + "total_flos": 3798137260081152.0, + "train_loss": 0.9928542430577209, + "train_runtime": 293777.8936, + "train_samples_per_second": 13.624, + "train_steps_per_second": 0.106 + } + ], + "max_steps": 31269, + "num_train_epochs": 1, + "total_flos": 3798137260081152.0, + "trial_name": null, + "trial_params": null +}