{ "best_metric": 0.5440191626548767, "best_model_checkpoint": "upernet-convnext-small-finetuned/checkpoint-1725", "epoch": 3.0, "eval_steps": 500, "global_step": 1725, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0.0005996521739130435, "loss": 2.1856, "step": 1 }, { "epoch": 0.0, "learning_rate": 0.0005993043478260869, "loss": 1.8053, "step": 2 }, { "epoch": 0.01, "learning_rate": 0.0005989565217391304, "loss": 1.5752, "step": 3 }, { "epoch": 0.01, "learning_rate": 0.0005986086956521739, "loss": 1.1745, "step": 4 }, { "epoch": 0.01, "learning_rate": 0.0005982608695652173, "loss": 1.0086, "step": 5 }, { "epoch": 0.01, "learning_rate": 0.0005979130434782608, "loss": 1.0224, "step": 6 }, { "epoch": 0.01, "learning_rate": 0.0005975652173913043, "loss": 0.9375, "step": 7 }, { "epoch": 0.01, "learning_rate": 0.0005972173913043478, "loss": 0.6089, "step": 8 }, { "epoch": 0.02, "learning_rate": 0.0005968695652173912, "loss": 0.6579, "step": 9 }, { "epoch": 0.02, "learning_rate": 0.0005965217391304347, "loss": 0.5931, "step": 10 }, { "epoch": 0.02, "learning_rate": 0.0005961739130434783, "loss": 0.6804, "step": 11 }, { "epoch": 0.02, "learning_rate": 0.0005958260869565217, "loss": 0.5951, "step": 12 }, { "epoch": 0.02, "learning_rate": 0.0005954782608695652, "loss": 0.473, "step": 13 }, { "epoch": 0.02, "learning_rate": 0.0005951304347826087, "loss": 0.7625, "step": 14 }, { "epoch": 0.03, "learning_rate": 0.0005947826086956521, "loss": 0.4651, "step": 15 }, { "epoch": 0.03, "learning_rate": 0.0005944347826086955, "loss": 0.6459, "step": 16 }, { "epoch": 0.03, "learning_rate": 0.0005940869565217391, "loss": 0.4976, "step": 17 }, { "epoch": 0.03, "learning_rate": 0.0005937391304347826, "loss": 0.6657, "step": 18 }, { "epoch": 0.03, "learning_rate": 0.000593391304347826, "loss": 0.6993, "step": 19 }, { "epoch": 0.03, "learning_rate": 0.0005930434782608695, "loss": 0.6763, "step": 20 }, { "epoch": 0.04, "learning_rate": 0.000592695652173913, "loss": 0.6459, "step": 21 }, { "epoch": 0.04, "learning_rate": 0.0005923478260869564, "loss": 0.3941, "step": 22 }, { "epoch": 0.04, "learning_rate": 0.000592, "loss": 0.5285, "step": 23 }, { "epoch": 0.04, "learning_rate": 0.0005916521739130435, "loss": 0.4065, "step": 24 }, { "epoch": 0.04, "learning_rate": 0.0005913043478260869, "loss": 0.3398, "step": 25 }, { "epoch": 0.05, "learning_rate": 0.0005909565217391303, "loss": 0.503, "step": 26 }, { "epoch": 0.05, "learning_rate": 0.0005906086956521738, "loss": 0.5469, "step": 27 }, { "epoch": 0.05, "learning_rate": 0.0005902608695652174, "loss": 0.8316, "step": 28 }, { "epoch": 0.05, "learning_rate": 0.0005899130434782608, "loss": 0.4833, "step": 29 }, { "epoch": 0.05, "learning_rate": 0.0005895652173913043, "loss": 0.5218, "step": 30 }, { "epoch": 0.05, "learning_rate": 0.0005892173913043478, "loss": 0.425, "step": 31 }, { "epoch": 0.06, "learning_rate": 0.0005888695652173912, "loss": 0.4683, "step": 32 }, { "epoch": 0.06, "learning_rate": 0.0005885217391304348, "loss": 0.4848, "step": 33 }, { "epoch": 0.06, "learning_rate": 0.0005881739130434783, "loss": 0.4143, "step": 34 }, { "epoch": 0.06, "learning_rate": 0.0005878260869565217, "loss": 0.4866, "step": 35 }, { "epoch": 0.06, "learning_rate": 0.0005874782608695652, "loss": 0.6084, "step": 36 }, { "epoch": 0.06, "learning_rate": 0.0005871304347826086, "loss": 0.5069, "step": 37 }, { "epoch": 0.07, "learning_rate": 0.0005867826086956521, "loss": 0.4934, "step": 38 }, { "epoch": 0.07, "learning_rate": 0.0005864347826086956, "loss": 0.455, "step": 39 }, { "epoch": 0.07, "learning_rate": 0.0005860869565217391, "loss": 0.3881, "step": 40 }, { "epoch": 0.07, "learning_rate": 0.0005857391304347826, "loss": 0.4883, "step": 41 }, { "epoch": 0.07, "learning_rate": 0.000585391304347826, "loss": 0.6074, "step": 42 }, { "epoch": 0.07, "learning_rate": 0.0005850434782608695, "loss": 0.427, "step": 43 }, { "epoch": 0.08, "learning_rate": 0.0005846956521739131, "loss": 0.4375, "step": 44 }, { "epoch": 0.08, "learning_rate": 0.0005843478260869565, "loss": 0.4417, "step": 45 }, { "epoch": 0.08, "learning_rate": 0.000584, "loss": 0.5273, "step": 46 }, { "epoch": 0.08, "learning_rate": 0.0005836521739130434, "loss": 0.3472, "step": 47 }, { "epoch": 0.08, "learning_rate": 0.0005833043478260869, "loss": 0.3796, "step": 48 }, { "epoch": 0.09, "learning_rate": 0.0005829565217391303, "loss": 0.3848, "step": 49 }, { "epoch": 0.09, "learning_rate": 0.0005826086956521739, "loss": 0.3735, "step": 50 }, { "epoch": 0.09, "learning_rate": 0.0005822608695652174, "loss": 0.4995, "step": 51 }, { "epoch": 0.09, "learning_rate": 0.0005819130434782608, "loss": 0.5001, "step": 52 }, { "epoch": 0.09, "learning_rate": 0.0005815652173913043, "loss": 0.365, "step": 53 }, { "epoch": 0.09, "learning_rate": 0.0005812173913043478, "loss": 0.5385, "step": 54 }, { "epoch": 0.1, "learning_rate": 0.0005808695652173913, "loss": 0.54, "step": 55 }, { "epoch": 0.1, "learning_rate": 0.0005805217391304348, "loss": 0.4025, "step": 56 }, { "epoch": 0.1, "learning_rate": 0.0005801739130434782, "loss": 0.511, "step": 57 }, { "epoch": 0.1, "learning_rate": 0.0005798260869565217, "loss": 0.3292, "step": 58 }, { "epoch": 0.1, "learning_rate": 0.0005794782608695651, "loss": 0.5285, "step": 59 }, { "epoch": 0.1, "learning_rate": 0.0005791304347826086, "loss": 0.4411, "step": 60 }, { "epoch": 0.11, "learning_rate": 0.0005787826086956522, "loss": 0.4315, "step": 61 }, { "epoch": 0.11, "learning_rate": 0.0005784347826086956, "loss": 0.4092, "step": 62 }, { "epoch": 0.11, "learning_rate": 0.0005780869565217391, "loss": 0.4468, "step": 63 }, { "epoch": 0.11, "learning_rate": 0.0005777391304347826, "loss": 0.5128, "step": 64 }, { "epoch": 0.11, "learning_rate": 0.000577391304347826, "loss": 0.5136, "step": 65 }, { "epoch": 0.11, "learning_rate": 0.0005770434782608696, "loss": 0.4, "step": 66 }, { "epoch": 0.12, "learning_rate": 0.000576695652173913, "loss": 0.3075, "step": 67 }, { "epoch": 0.12, "learning_rate": 0.0005763478260869565, "loss": 0.5756, "step": 68 }, { "epoch": 0.12, "learning_rate": 0.0005759999999999999, "loss": 0.4765, "step": 69 }, { "epoch": 0.12, "learning_rate": 0.0005756521739130434, "loss": 0.3477, "step": 70 }, { "epoch": 0.12, "learning_rate": 0.000575304347826087, "loss": 0.4781, "step": 71 }, { "epoch": 0.13, "learning_rate": 0.0005749565217391304, "loss": 0.4364, "step": 72 }, { "epoch": 0.13, "learning_rate": 0.0005746086956521739, "loss": 0.5798, "step": 73 }, { "epoch": 0.13, "learning_rate": 0.0005742608695652174, "loss": 0.5236, "step": 74 }, { "epoch": 0.13, "learning_rate": 0.0005739130434782608, "loss": 0.348, "step": 75 }, { "epoch": 0.13, "learning_rate": 0.0005735652173913043, "loss": 0.4795, "step": 76 }, { "epoch": 0.13, "learning_rate": 0.0005732173913043478, "loss": 0.525, "step": 77 }, { "epoch": 0.14, "learning_rate": 0.0005728695652173913, "loss": 0.3808, "step": 78 }, { "epoch": 0.14, "learning_rate": 0.0005725217391304347, "loss": 0.6183, "step": 79 }, { "epoch": 0.14, "learning_rate": 0.0005721739130434782, "loss": 0.528, "step": 80 }, { "epoch": 0.14, "learning_rate": 0.0005718260869565217, "loss": 0.3989, "step": 81 }, { "epoch": 0.14, "learning_rate": 0.0005714782608695651, "loss": 0.4159, "step": 82 }, { "epoch": 0.14, "learning_rate": 0.0005711304347826087, "loss": 0.3876, "step": 83 }, { "epoch": 0.15, "learning_rate": 0.0005707826086956521, "loss": 0.5781, "step": 84 }, { "epoch": 0.15, "learning_rate": 0.0005704347826086956, "loss": 0.5086, "step": 85 }, { "epoch": 0.15, "learning_rate": 0.0005700869565217391, "loss": 0.5343, "step": 86 }, { "epoch": 0.15, "learning_rate": 0.0005697391304347825, "loss": 0.5413, "step": 87 }, { "epoch": 0.15, "learning_rate": 0.0005693913043478261, "loss": 0.4632, "step": 88 }, { "epoch": 0.15, "learning_rate": 0.0005690434782608695, "loss": 0.4597, "step": 89 }, { "epoch": 0.16, "learning_rate": 0.000568695652173913, "loss": 0.6424, "step": 90 }, { "epoch": 0.16, "learning_rate": 0.0005683478260869564, "loss": 0.4235, "step": 91 }, { "epoch": 0.16, "learning_rate": 0.0005679999999999999, "loss": 0.5384, "step": 92 }, { "epoch": 0.16, "learning_rate": 0.0005676521739130435, "loss": 0.3902, "step": 93 }, { "epoch": 0.16, "learning_rate": 0.0005673043478260869, "loss": 0.4713, "step": 94 }, { "epoch": 0.17, "learning_rate": 0.0005669565217391304, "loss": 0.3048, "step": 95 }, { "epoch": 0.17, "learning_rate": 0.0005666086956521739, "loss": 0.3824, "step": 96 }, { "epoch": 0.17, "learning_rate": 0.0005662608695652173, "loss": 0.3625, "step": 97 }, { "epoch": 0.17, "learning_rate": 0.0005659130434782608, "loss": 0.3409, "step": 98 }, { "epoch": 0.17, "learning_rate": 0.0005655652173913043, "loss": 0.4191, "step": 99 }, { "epoch": 0.17, "learning_rate": 0.0005652173913043478, "loss": 0.4968, "step": 100 }, { "epoch": 0.18, "learning_rate": 0.0005648695652173912, "loss": 0.4378, "step": 101 }, { "epoch": 0.18, "learning_rate": 0.0005645217391304347, "loss": 0.6287, "step": 102 }, { "epoch": 0.18, "learning_rate": 0.0005641739130434782, "loss": 0.5649, "step": 103 }, { "epoch": 0.18, "learning_rate": 0.0005638260869565216, "loss": 0.4456, "step": 104 }, { "epoch": 0.18, "learning_rate": 0.0005634782608695652, "loss": 0.378, "step": 105 }, { "epoch": 0.18, "learning_rate": 0.0005631304347826087, "loss": 0.5281, "step": 106 }, { "epoch": 0.19, "learning_rate": 0.0005627826086956521, "loss": 0.4917, "step": 107 }, { "epoch": 0.19, "learning_rate": 0.0005624347826086956, "loss": 0.4609, "step": 108 }, { "epoch": 0.19, "learning_rate": 0.000562086956521739, "loss": 0.5439, "step": 109 }, { "epoch": 0.19, "learning_rate": 0.0005617391304347826, "loss": 0.4667, "step": 110 }, { "epoch": 0.19, "learning_rate": 0.000561391304347826, "loss": 0.3511, "step": 111 }, { "epoch": 0.19, "learning_rate": 0.0005610434782608695, "loss": 0.5089, "step": 112 }, { "epoch": 0.2, "learning_rate": 0.000560695652173913, "loss": 0.4884, "step": 113 }, { "epoch": 0.2, "learning_rate": 0.0005603478260869564, "loss": 0.393, "step": 114 }, { "epoch": 0.2, "learning_rate": 0.00056, "loss": 0.435, "step": 115 }, { "epoch": 0.2, "learning_rate": 0.0005596521739130435, "loss": 0.3567, "step": 116 }, { "epoch": 0.2, "learning_rate": 0.0005593043478260869, "loss": 0.5612, "step": 117 }, { "epoch": 0.21, "learning_rate": 0.0005589565217391304, "loss": 0.3585, "step": 118 }, { "epoch": 0.21, "learning_rate": 0.0005586086956521738, "loss": 0.4096, "step": 119 }, { "epoch": 0.21, "learning_rate": 0.0005582608695652173, "loss": 0.3628, "step": 120 }, { "epoch": 0.21, "learning_rate": 0.0005579130434782608, "loss": 0.3428, "step": 121 }, { "epoch": 0.21, "learning_rate": 0.0005575652173913043, "loss": 0.5759, "step": 122 }, { "epoch": 0.21, "learning_rate": 0.0005572173913043478, "loss": 0.552, "step": 123 }, { "epoch": 0.22, "learning_rate": 0.0005568695652173912, "loss": 0.5093, "step": 124 }, { "epoch": 0.22, "learning_rate": 0.0005565217391304347, "loss": 0.3613, "step": 125 }, { "epoch": 0.22, "learning_rate": 0.0005561739130434783, "loss": 0.3834, "step": 126 }, { "epoch": 0.22, "learning_rate": 0.0005558260869565217, "loss": 0.3717, "step": 127 }, { "epoch": 0.22, "learning_rate": 0.0005554782608695652, "loss": 0.4666, "step": 128 }, { "epoch": 0.22, "learning_rate": 0.0005551304347826087, "loss": 0.4819, "step": 129 }, { "epoch": 0.23, "learning_rate": 0.0005547826086956521, "loss": 0.5606, "step": 130 }, { "epoch": 0.23, "learning_rate": 0.0005544347826086955, "loss": 0.4164, "step": 131 }, { "epoch": 0.23, "learning_rate": 0.0005540869565217391, "loss": 0.4083, "step": 132 }, { "epoch": 0.23, "learning_rate": 0.0005537391304347826, "loss": 0.454, "step": 133 }, { "epoch": 0.23, "learning_rate": 0.000553391304347826, "loss": 0.4533, "step": 134 }, { "epoch": 0.23, "learning_rate": 0.0005530434782608695, "loss": 0.5167, "step": 135 }, { "epoch": 0.24, "learning_rate": 0.000552695652173913, "loss": 0.3254, "step": 136 }, { "epoch": 0.24, "learning_rate": 0.0005523478260869565, "loss": 0.4695, "step": 137 }, { "epoch": 0.24, "learning_rate": 0.000552, "loss": 0.2475, "step": 138 }, { "epoch": 0.24, "learning_rate": 0.0005516521739130435, "loss": 0.4745, "step": 139 }, { "epoch": 0.24, "learning_rate": 0.0005513043478260869, "loss": 0.2616, "step": 140 }, { "epoch": 0.25, "learning_rate": 0.0005509565217391303, "loss": 0.4784, "step": 141 }, { "epoch": 0.25, "learning_rate": 0.0005506086956521738, "loss": 0.4737, "step": 142 }, { "epoch": 0.25, "learning_rate": 0.0005502608695652174, "loss": 0.3403, "step": 143 }, { "epoch": 0.25, "learning_rate": 0.0005499130434782608, "loss": 0.4563, "step": 144 }, { "epoch": 0.25, "learning_rate": 0.0005495652173913043, "loss": 0.4445, "step": 145 }, { "epoch": 0.25, "learning_rate": 0.0005492173913043478, "loss": 0.3949, "step": 146 }, { "epoch": 0.26, "learning_rate": 0.0005488695652173912, "loss": 0.3556, "step": 147 }, { "epoch": 0.26, "learning_rate": 0.0005485217391304348, "loss": 0.4004, "step": 148 }, { "epoch": 0.26, "learning_rate": 0.0005481739130434783, "loss": 0.5179, "step": 149 }, { "epoch": 0.26, "learning_rate": 0.0005478260869565217, "loss": 0.4297, "step": 150 }, { "epoch": 0.26, "learning_rate": 0.0005474782608695651, "loss": 0.3168, "step": 151 }, { "epoch": 0.26, "learning_rate": 0.0005471304347826086, "loss": 0.3384, "step": 152 }, { "epoch": 0.27, "learning_rate": 0.0005467826086956521, "loss": 0.4387, "step": 153 }, { "epoch": 0.27, "learning_rate": 0.0005464347826086956, "loss": 0.4496, "step": 154 }, { "epoch": 0.27, "learning_rate": 0.0005460869565217391, "loss": 0.4544, "step": 155 }, { "epoch": 0.27, "learning_rate": 0.0005457391304347826, "loss": 0.4517, "step": 156 }, { "epoch": 0.27, "learning_rate": 0.000545391304347826, "loss": 0.433, "step": 157 }, { "epoch": 0.27, "learning_rate": 0.0005450434782608695, "loss": 0.3733, "step": 158 }, { "epoch": 0.28, "learning_rate": 0.0005446956521739131, "loss": 0.2987, "step": 159 }, { "epoch": 0.28, "learning_rate": 0.0005443478260869565, "loss": 0.4131, "step": 160 }, { "epoch": 0.28, "learning_rate": 0.0005439999999999999, "loss": 0.3554, "step": 161 }, { "epoch": 0.28, "learning_rate": 0.0005436521739130434, "loss": 0.477, "step": 162 }, { "epoch": 0.28, "learning_rate": 0.0005433043478260869, "loss": 0.3449, "step": 163 }, { "epoch": 0.29, "learning_rate": 0.0005429565217391303, "loss": 0.731, "step": 164 }, { "epoch": 0.29, "learning_rate": 0.0005426086956521739, "loss": 0.295, "step": 165 }, { "epoch": 0.29, "learning_rate": 0.0005422608695652174, "loss": 0.4337, "step": 166 }, { "epoch": 0.29, "learning_rate": 0.0005419130434782608, "loss": 0.3261, "step": 167 }, { "epoch": 0.29, "learning_rate": 0.0005415652173913043, "loss": 0.4041, "step": 168 }, { "epoch": 0.29, "learning_rate": 0.0005412173913043478, "loss": 0.4937, "step": 169 }, { "epoch": 0.3, "learning_rate": 0.0005408695652173913, "loss": 0.3729, "step": 170 }, { "epoch": 0.3, "learning_rate": 0.0005405217391304347, "loss": 0.3224, "step": 171 }, { "epoch": 0.3, "learning_rate": 0.0005401739130434782, "loss": 0.4806, "step": 172 }, { "epoch": 0.3, "learning_rate": 0.0005398260869565217, "loss": 0.4868, "step": 173 }, { "epoch": 0.3, "learning_rate": 0.0005394782608695651, "loss": 0.3315, "step": 174 }, { "epoch": 0.3, "learning_rate": 0.0005391304347826086, "loss": 0.3692, "step": 175 }, { "epoch": 0.31, "learning_rate": 0.0005387826086956522, "loss": 0.3518, "step": 176 }, { "epoch": 0.31, "learning_rate": 0.0005384347826086956, "loss": 0.5419, "step": 177 }, { "epoch": 0.31, "learning_rate": 0.0005380869565217391, "loss": 0.5138, "step": 178 }, { "epoch": 0.31, "learning_rate": 0.0005377391304347826, "loss": 0.3229, "step": 179 }, { "epoch": 0.31, "learning_rate": 0.000537391304347826, "loss": 0.3837, "step": 180 }, { "epoch": 0.31, "learning_rate": 0.0005370434782608695, "loss": 0.3869, "step": 181 }, { "epoch": 0.32, "learning_rate": 0.000536695652173913, "loss": 0.488, "step": 182 }, { "epoch": 0.32, "learning_rate": 0.0005363478260869565, "loss": 0.5541, "step": 183 }, { "epoch": 0.32, "learning_rate": 0.0005359999999999999, "loss": 0.5053, "step": 184 }, { "epoch": 0.32, "learning_rate": 0.0005356521739130434, "loss": 0.2738, "step": 185 }, { "epoch": 0.32, "learning_rate": 0.000535304347826087, "loss": 0.5393, "step": 186 }, { "epoch": 0.33, "learning_rate": 0.0005349565217391304, "loss": 0.4429, "step": 187 }, { "epoch": 0.33, "learning_rate": 0.0005346086956521739, "loss": 0.3537, "step": 188 }, { "epoch": 0.33, "learning_rate": 0.0005342608695652174, "loss": 0.3719, "step": 189 }, { "epoch": 0.33, "learning_rate": 0.0005339130434782608, "loss": 0.3739, "step": 190 }, { "epoch": 0.33, "learning_rate": 0.0005335652173913043, "loss": 0.349, "step": 191 }, { "epoch": 0.33, "learning_rate": 0.0005332173913043478, "loss": 0.3435, "step": 192 }, { "epoch": 0.34, "learning_rate": 0.0005328695652173913, "loss": 0.3479, "step": 193 }, { "epoch": 0.34, "learning_rate": 0.0005325217391304347, "loss": 0.304, "step": 194 }, { "epoch": 0.34, "learning_rate": 0.0005321739130434782, "loss": 0.3976, "step": 195 }, { "epoch": 0.34, "learning_rate": 0.0005318260869565217, "loss": 0.4209, "step": 196 }, { "epoch": 0.34, "learning_rate": 0.0005314782608695651, "loss": 0.4334, "step": 197 }, { "epoch": 0.34, "learning_rate": 0.0005311304347826087, "loss": 0.389, "step": 198 }, { "epoch": 0.35, "learning_rate": 0.0005307826086956522, "loss": 0.3164, "step": 199 }, { "epoch": 0.35, "learning_rate": 0.0005304347826086956, "loss": 0.7116, "step": 200 }, { "epoch": 0.35, "learning_rate": 0.0005300869565217391, "loss": 0.3852, "step": 201 }, { "epoch": 0.35, "learning_rate": 0.0005297391304347825, "loss": 0.3146, "step": 202 }, { "epoch": 0.35, "learning_rate": 0.0005293913043478261, "loss": 0.4017, "step": 203 }, { "epoch": 0.35, "learning_rate": 0.0005290434782608695, "loss": 0.4697, "step": 204 }, { "epoch": 0.36, "learning_rate": 0.000528695652173913, "loss": 0.3811, "step": 205 }, { "epoch": 0.36, "learning_rate": 0.0005283478260869565, "loss": 0.3434, "step": 206 }, { "epoch": 0.36, "learning_rate": 0.0005279999999999999, "loss": 0.307, "step": 207 }, { "epoch": 0.36, "learning_rate": 0.0005276521739130435, "loss": 0.4292, "step": 208 }, { "epoch": 0.36, "learning_rate": 0.000527304347826087, "loss": 0.3428, "step": 209 }, { "epoch": 0.37, "learning_rate": 0.0005269565217391304, "loss": 0.4429, "step": 210 }, { "epoch": 0.37, "learning_rate": 0.0005266086956521739, "loss": 0.3297, "step": 211 }, { "epoch": 0.37, "learning_rate": 0.0005262608695652173, "loss": 0.4362, "step": 212 }, { "epoch": 0.37, "learning_rate": 0.0005259130434782608, "loss": 0.3154, "step": 213 }, { "epoch": 0.37, "learning_rate": 0.0005255652173913043, "loss": 0.2614, "step": 214 }, { "epoch": 0.37, "learning_rate": 0.0005252173913043478, "loss": 0.3937, "step": 215 }, { "epoch": 0.38, "learning_rate": 0.0005248695652173913, "loss": 0.5459, "step": 216 }, { "epoch": 0.38, "learning_rate": 0.0005245217391304347, "loss": 0.3998, "step": 217 }, { "epoch": 0.38, "learning_rate": 0.0005241739130434782, "loss": 0.3943, "step": 218 }, { "epoch": 0.38, "learning_rate": 0.0005238260869565218, "loss": 0.6246, "step": 219 }, { "epoch": 0.38, "learning_rate": 0.0005234782608695652, "loss": 0.4418, "step": 220 }, { "epoch": 0.38, "learning_rate": 0.0005231304347826087, "loss": 0.4283, "step": 221 }, { "epoch": 0.39, "learning_rate": 0.0005227826086956522, "loss": 0.4115, "step": 222 }, { "epoch": 0.39, "learning_rate": 0.0005224347826086956, "loss": 0.4461, "step": 223 }, { "epoch": 0.39, "learning_rate": 0.000522086956521739, "loss": 0.3147, "step": 224 }, { "epoch": 0.39, "learning_rate": 0.0005217391304347826, "loss": 0.431, "step": 225 }, { "epoch": 0.39, "learning_rate": 0.0005213913043478261, "loss": 0.3645, "step": 226 }, { "epoch": 0.39, "learning_rate": 0.0005210434782608695, "loss": 0.5096, "step": 227 }, { "epoch": 0.4, "learning_rate": 0.000520695652173913, "loss": 0.456, "step": 228 }, { "epoch": 0.4, "learning_rate": 0.0005203478260869565, "loss": 0.3486, "step": 229 }, { "epoch": 0.4, "learning_rate": 0.00052, "loss": 0.3791, "step": 230 }, { "epoch": 0.4, "learning_rate": 0.0005196521739130435, "loss": 0.403, "step": 231 }, { "epoch": 0.4, "learning_rate": 0.0005193043478260869, "loss": 0.3298, "step": 232 }, { "epoch": 0.41, "learning_rate": 0.0005189565217391304, "loss": 0.3204, "step": 233 }, { "epoch": 0.41, "learning_rate": 0.0005186086956521738, "loss": 0.387, "step": 234 }, { "epoch": 0.41, "learning_rate": 0.0005182608695652173, "loss": 0.4611, "step": 235 }, { "epoch": 0.41, "learning_rate": 0.0005179130434782609, "loss": 0.4694, "step": 236 }, { "epoch": 0.41, "learning_rate": 0.0005175652173913043, "loss": 0.4213, "step": 237 }, { "epoch": 0.41, "learning_rate": 0.0005172173913043478, "loss": 0.5201, "step": 238 }, { "epoch": 0.42, "learning_rate": 0.0005168695652173912, "loss": 0.5185, "step": 239 }, { "epoch": 0.42, "learning_rate": 0.0005165217391304347, "loss": 0.3898, "step": 240 }, { "epoch": 0.42, "learning_rate": 0.0005161739130434783, "loss": 0.3413, "step": 241 }, { "epoch": 0.42, "learning_rate": 0.0005158260869565217, "loss": 0.3748, "step": 242 }, { "epoch": 0.42, "learning_rate": 0.0005154782608695652, "loss": 0.4882, "step": 243 }, { "epoch": 0.42, "learning_rate": 0.0005151304347826086, "loss": 0.3164, "step": 244 }, { "epoch": 0.43, "learning_rate": 0.0005147826086956521, "loss": 0.5037, "step": 245 }, { "epoch": 0.43, "learning_rate": 0.0005144347826086955, "loss": 0.474, "step": 246 }, { "epoch": 0.43, "learning_rate": 0.0005140869565217391, "loss": 0.6393, "step": 247 }, { "epoch": 0.43, "learning_rate": 0.0005137391304347826, "loss": 0.5011, "step": 248 }, { "epoch": 0.43, "learning_rate": 0.000513391304347826, "loss": 0.4485, "step": 249 }, { "epoch": 0.43, "learning_rate": 0.0005130434782608695, "loss": 0.3244, "step": 250 }, { "epoch": 0.44, "learning_rate": 0.000512695652173913, "loss": 0.4114, "step": 251 }, { "epoch": 0.44, "learning_rate": 0.0005123478260869565, "loss": 0.3274, "step": 252 }, { "epoch": 0.44, "learning_rate": 0.000512, "loss": 0.3129, "step": 253 }, { "epoch": 0.44, "learning_rate": 0.0005116521739130434, "loss": 0.3908, "step": 254 }, { "epoch": 0.44, "learning_rate": 0.0005113043478260869, "loss": 0.3369, "step": 255 }, { "epoch": 0.45, "learning_rate": 0.0005109565217391303, "loss": 0.3682, "step": 256 }, { "epoch": 0.45, "learning_rate": 0.0005106086956521738, "loss": 0.4161, "step": 257 }, { "epoch": 0.45, "learning_rate": 0.0005102608695652174, "loss": 0.3206, "step": 258 }, { "epoch": 0.45, "learning_rate": 0.0005099130434782608, "loss": 0.4214, "step": 259 }, { "epoch": 0.45, "learning_rate": 0.0005095652173913043, "loss": 0.3284, "step": 260 }, { "epoch": 0.45, "learning_rate": 0.0005092173913043478, "loss": 0.3824, "step": 261 }, { "epoch": 0.46, "learning_rate": 0.0005088695652173912, "loss": 0.3291, "step": 262 }, { "epoch": 0.46, "learning_rate": 0.0005085217391304348, "loss": 0.256, "step": 263 }, { "epoch": 0.46, "learning_rate": 0.0005081739130434782, "loss": 0.3193, "step": 264 }, { "epoch": 0.46, "learning_rate": 0.0005078260869565217, "loss": 0.4613, "step": 265 }, { "epoch": 0.46, "learning_rate": 0.0005074782608695651, "loss": 0.3299, "step": 266 }, { "epoch": 0.46, "learning_rate": 0.0005071304347826086, "loss": 0.2747, "step": 267 }, { "epoch": 0.47, "learning_rate": 0.0005067826086956522, "loss": 0.3887, "step": 268 }, { "epoch": 0.47, "learning_rate": 0.0005064347826086956, "loss": 0.5639, "step": 269 }, { "epoch": 0.47, "learning_rate": 0.0005060869565217391, "loss": 0.4741, "step": 270 }, { "epoch": 0.47, "learning_rate": 0.0005057391304347826, "loss": 0.4658, "step": 271 }, { "epoch": 0.47, "learning_rate": 0.000505391304347826, "loss": 0.4417, "step": 272 }, { "epoch": 0.47, "learning_rate": 0.0005050434782608695, "loss": 0.3463, "step": 273 }, { "epoch": 0.48, "learning_rate": 0.000504695652173913, "loss": 0.4364, "step": 274 }, { "epoch": 0.48, "learning_rate": 0.0005043478260869565, "loss": 0.3228, "step": 275 }, { "epoch": 0.48, "learning_rate": 0.0005039999999999999, "loss": 0.4163, "step": 276 }, { "epoch": 0.48, "learning_rate": 0.0005036521739130434, "loss": 0.3995, "step": 277 }, { "epoch": 0.48, "learning_rate": 0.0005033043478260869, "loss": 0.4739, "step": 278 }, { "epoch": 0.49, "learning_rate": 0.0005029565217391303, "loss": 0.5107, "step": 279 }, { "epoch": 0.49, "learning_rate": 0.0005026086956521739, "loss": 0.3085, "step": 280 }, { "epoch": 0.49, "learning_rate": 0.0005022608695652174, "loss": 0.5414, "step": 281 }, { "epoch": 0.49, "learning_rate": 0.0005019130434782608, "loss": 0.355, "step": 282 }, { "epoch": 0.49, "learning_rate": 0.0005015652173913043, "loss": 0.5159, "step": 283 }, { "epoch": 0.49, "learning_rate": 0.0005012173913043478, "loss": 0.4088, "step": 284 }, { "epoch": 0.5, "learning_rate": 0.0005008695652173913, "loss": 0.5618, "step": 285 }, { "epoch": 0.5, "learning_rate": 0.0005005217391304347, "loss": 0.3853, "step": 286 }, { "epoch": 0.5, "learning_rate": 0.0005001739130434782, "loss": 0.4689, "step": 287 }, { "epoch": 0.5, "learning_rate": 0.0004998260869565217, "loss": 0.5098, "step": 288 }, { "epoch": 0.5, "learning_rate": 0.0004994782608695651, "loss": 0.3794, "step": 289 }, { "epoch": 0.5, "learning_rate": 0.0004991304347826087, "loss": 0.4242, "step": 290 }, { "epoch": 0.51, "learning_rate": 0.0004987826086956522, "loss": 0.5362, "step": 291 }, { "epoch": 0.51, "learning_rate": 0.0004984347826086956, "loss": 0.393, "step": 292 }, { "epoch": 0.51, "learning_rate": 0.0004980869565217391, "loss": 0.3632, "step": 293 }, { "epoch": 0.51, "learning_rate": 0.0004977391304347826, "loss": 0.4277, "step": 294 }, { "epoch": 0.51, "learning_rate": 0.000497391304347826, "loss": 0.4004, "step": 295 }, { "epoch": 0.51, "learning_rate": 0.0004970434782608695, "loss": 0.5054, "step": 296 }, { "epoch": 0.52, "learning_rate": 0.000496695652173913, "loss": 0.3996, "step": 297 }, { "epoch": 0.52, "learning_rate": 0.0004963478260869565, "loss": 0.4602, "step": 298 }, { "epoch": 0.52, "learning_rate": 0.0004959999999999999, "loss": 0.3782, "step": 299 }, { "epoch": 0.52, "learning_rate": 0.0004956521739130434, "loss": 0.3904, "step": 300 }, { "epoch": 0.52, "learning_rate": 0.000495304347826087, "loss": 0.4157, "step": 301 }, { "epoch": 0.53, "learning_rate": 0.0004949565217391304, "loss": 0.3796, "step": 302 }, { "epoch": 0.53, "learning_rate": 0.0004946086956521739, "loss": 0.4296, "step": 303 }, { "epoch": 0.53, "learning_rate": 0.0004942608695652174, "loss": 0.4334, "step": 304 }, { "epoch": 0.53, "learning_rate": 0.0004939130434782608, "loss": 0.5047, "step": 305 }, { "epoch": 0.53, "learning_rate": 0.0004935652173913042, "loss": 0.4333, "step": 306 }, { "epoch": 0.53, "learning_rate": 0.0004932173913043478, "loss": 0.3288, "step": 307 }, { "epoch": 0.54, "learning_rate": 0.0004928695652173913, "loss": 0.3563, "step": 308 }, { "epoch": 0.54, "learning_rate": 0.0004925217391304347, "loss": 0.3758, "step": 309 }, { "epoch": 0.54, "learning_rate": 0.0004921739130434782, "loss": 0.435, "step": 310 }, { "epoch": 0.54, "learning_rate": 0.0004918260869565217, "loss": 0.5123, "step": 311 }, { "epoch": 0.54, "learning_rate": 0.0004914782608695652, "loss": 0.4267, "step": 312 }, { "epoch": 0.54, "learning_rate": 0.0004911304347826087, "loss": 0.4387, "step": 313 }, { "epoch": 0.55, "learning_rate": 0.0004907826086956522, "loss": 0.3285, "step": 314 }, { "epoch": 0.55, "learning_rate": 0.0004904347826086956, "loss": 0.4338, "step": 315 }, { "epoch": 0.55, "learning_rate": 0.000490086956521739, "loss": 0.2459, "step": 316 }, { "epoch": 0.55, "learning_rate": 0.0004897391304347825, "loss": 0.3638, "step": 317 }, { "epoch": 0.55, "learning_rate": 0.0004893913043478261, "loss": 0.417, "step": 318 }, { "epoch": 0.55, "learning_rate": 0.0004890434782608695, "loss": 0.3581, "step": 319 }, { "epoch": 0.56, "learning_rate": 0.000488695652173913, "loss": 0.4019, "step": 320 }, { "epoch": 0.56, "learning_rate": 0.0004883478260869565, "loss": 0.3594, "step": 321 }, { "epoch": 0.56, "learning_rate": 0.000488, "loss": 0.5029, "step": 322 }, { "epoch": 0.56, "learning_rate": 0.00048765217391304346, "loss": 0.3217, "step": 323 }, { "epoch": 0.56, "learning_rate": 0.0004873043478260869, "loss": 0.2806, "step": 324 }, { "epoch": 0.57, "learning_rate": 0.00048695652173913045, "loss": 0.3502, "step": 325 }, { "epoch": 0.57, "learning_rate": 0.00048660869565217386, "loss": 0.3045, "step": 326 }, { "epoch": 0.57, "learning_rate": 0.0004862608695652173, "loss": 0.5014, "step": 327 }, { "epoch": 0.57, "learning_rate": 0.0004859130434782608, "loss": 0.4246, "step": 328 }, { "epoch": 0.57, "learning_rate": 0.0004855652173913043, "loss": 0.3105, "step": 329 }, { "epoch": 0.57, "learning_rate": 0.0004852173913043478, "loss": 0.3379, "step": 330 }, { "epoch": 0.58, "learning_rate": 0.00048486956521739125, "loss": 0.4452, "step": 331 }, { "epoch": 0.58, "learning_rate": 0.00048452173913043477, "loss": 0.3854, "step": 332 }, { "epoch": 0.58, "learning_rate": 0.00048417391304347824, "loss": 0.35, "step": 333 }, { "epoch": 0.58, "learning_rate": 0.0004838260869565217, "loss": 0.3776, "step": 334 }, { "epoch": 0.58, "learning_rate": 0.00048347826086956523, "loss": 0.5938, "step": 335 }, { "epoch": 0.58, "learning_rate": 0.00048313043478260864, "loss": 0.2886, "step": 336 }, { "epoch": 0.59, "learning_rate": 0.0004827826086956521, "loss": 0.3655, "step": 337 }, { "epoch": 0.59, "learning_rate": 0.0004824347826086956, "loss": 0.2544, "step": 338 }, { "epoch": 0.59, "learning_rate": 0.0004820869565217391, "loss": 0.3777, "step": 339 }, { "epoch": 0.59, "learning_rate": 0.00048173913043478257, "loss": 0.4824, "step": 340 }, { "epoch": 0.59, "learning_rate": 0.00048139130434782603, "loss": 0.3526, "step": 341 }, { "epoch": 0.59, "learning_rate": 0.00048104347826086956, "loss": 0.3574, "step": 342 }, { "epoch": 0.6, "learning_rate": 0.000480695652173913, "loss": 0.3135, "step": 343 }, { "epoch": 0.6, "learning_rate": 0.0004803478260869565, "loss": 0.3698, "step": 344 }, { "epoch": 0.6, "learning_rate": 0.00047999999999999996, "loss": 0.3259, "step": 345 }, { "epoch": 0.6, "learning_rate": 0.0004796521739130435, "loss": 0.3287, "step": 346 }, { "epoch": 0.6, "learning_rate": 0.0004793043478260869, "loss": 0.4842, "step": 347 }, { "epoch": 0.61, "learning_rate": 0.00047895652173913036, "loss": 0.3711, "step": 348 }, { "epoch": 0.61, "learning_rate": 0.0004786086956521739, "loss": 0.3369, "step": 349 }, { "epoch": 0.61, "learning_rate": 0.00047826086956521735, "loss": 0.4588, "step": 350 }, { "epoch": 0.61, "learning_rate": 0.0004779130434782608, "loss": 0.6008, "step": 351 }, { "epoch": 0.61, "learning_rate": 0.0004775652173913043, "loss": 0.5447, "step": 352 }, { "epoch": 0.61, "learning_rate": 0.0004772173913043478, "loss": 0.3756, "step": 353 }, { "epoch": 0.62, "learning_rate": 0.0004768695652173913, "loss": 0.437, "step": 354 }, { "epoch": 0.62, "learning_rate": 0.00047652173913043474, "loss": 0.4355, "step": 355 }, { "epoch": 0.62, "learning_rate": 0.00047617391304347826, "loss": 0.4498, "step": 356 }, { "epoch": 0.62, "learning_rate": 0.0004758260869565217, "loss": 0.4822, "step": 357 }, { "epoch": 0.62, "learning_rate": 0.00047547826086956514, "loss": 0.4175, "step": 358 }, { "epoch": 0.62, "learning_rate": 0.0004751304347826086, "loss": 0.3622, "step": 359 }, { "epoch": 0.63, "learning_rate": 0.00047478260869565213, "loss": 0.3803, "step": 360 }, { "epoch": 0.63, "learning_rate": 0.0004744347826086956, "loss": 0.4773, "step": 361 }, { "epoch": 0.63, "learning_rate": 0.00047408695652173907, "loss": 0.3765, "step": 362 }, { "epoch": 0.63, "learning_rate": 0.0004737391304347826, "loss": 0.396, "step": 363 }, { "epoch": 0.63, "learning_rate": 0.00047339130434782606, "loss": 0.3553, "step": 364 }, { "epoch": 0.63, "learning_rate": 0.0004730434782608695, "loss": 0.3783, "step": 365 }, { "epoch": 0.64, "learning_rate": 0.00047269565217391305, "loss": 0.5343, "step": 366 }, { "epoch": 0.64, "learning_rate": 0.00047234782608695646, "loss": 0.4224, "step": 367 }, { "epoch": 0.64, "learning_rate": 0.0004719999999999999, "loss": 0.431, "step": 368 }, { "epoch": 0.64, "learning_rate": 0.0004716521739130434, "loss": 0.3127, "step": 369 }, { "epoch": 0.64, "learning_rate": 0.0004713043478260869, "loss": 0.4126, "step": 370 }, { "epoch": 0.65, "learning_rate": 0.0004709565217391304, "loss": 0.5457, "step": 371 }, { "epoch": 0.65, "learning_rate": 0.00047060869565217385, "loss": 0.3459, "step": 372 }, { "epoch": 0.65, "learning_rate": 0.00047026086956521737, "loss": 0.3699, "step": 373 }, { "epoch": 0.65, "learning_rate": 0.00046991304347826084, "loss": 0.7105, "step": 374 }, { "epoch": 0.65, "learning_rate": 0.0004695652173913043, "loss": 0.4423, "step": 375 }, { "epoch": 0.65, "learning_rate": 0.00046921739130434783, "loss": 0.3803, "step": 376 }, { "epoch": 0.66, "learning_rate": 0.0004688695652173913, "loss": 0.341, "step": 377 }, { "epoch": 0.66, "learning_rate": 0.0004685217391304347, "loss": 0.3708, "step": 378 }, { "epoch": 0.66, "learning_rate": 0.0004681739130434782, "loss": 0.3647, "step": 379 }, { "epoch": 0.66, "learning_rate": 0.0004678260869565217, "loss": 0.4029, "step": 380 }, { "epoch": 0.66, "learning_rate": 0.00046747826086956517, "loss": 0.4387, "step": 381 }, { "epoch": 0.66, "learning_rate": 0.00046713043478260863, "loss": 0.3224, "step": 382 }, { "epoch": 0.67, "learning_rate": 0.00046678260869565215, "loss": 0.3468, "step": 383 }, { "epoch": 0.67, "learning_rate": 0.0004664347826086956, "loss": 0.3726, "step": 384 }, { "epoch": 0.67, "learning_rate": 0.0004660869565217391, "loss": 0.4739, "step": 385 }, { "epoch": 0.67, "learning_rate": 0.0004657391304347826, "loss": 0.382, "step": 386 }, { "epoch": 0.67, "learning_rate": 0.0004653913043478261, "loss": 0.3437, "step": 387 }, { "epoch": 0.67, "learning_rate": 0.0004650434782608695, "loss": 0.348, "step": 388 }, { "epoch": 0.68, "learning_rate": 0.00046469565217391296, "loss": 0.4077, "step": 389 }, { "epoch": 0.68, "learning_rate": 0.0004643478260869565, "loss": 0.4665, "step": 390 }, { "epoch": 0.68, "learning_rate": 0.00046399999999999995, "loss": 0.404, "step": 391 }, { "epoch": 0.68, "learning_rate": 0.0004636521739130434, "loss": 0.5046, "step": 392 }, { "epoch": 0.68, "learning_rate": 0.00046330434782608694, "loss": 0.4719, "step": 393 }, { "epoch": 0.69, "learning_rate": 0.0004629565217391304, "loss": 0.4336, "step": 394 }, { "epoch": 0.69, "learning_rate": 0.00046260869565217387, "loss": 0.6566, "step": 395 }, { "epoch": 0.69, "learning_rate": 0.0004622608695652174, "loss": 0.504, "step": 396 }, { "epoch": 0.69, "learning_rate": 0.00046191304347826086, "loss": 0.3583, "step": 397 }, { "epoch": 0.69, "learning_rate": 0.0004615652173913043, "loss": 0.4653, "step": 398 }, { "epoch": 0.69, "learning_rate": 0.00046121739130434774, "loss": 0.5052, "step": 399 }, { "epoch": 0.7, "learning_rate": 0.00046086956521739126, "loss": 0.351, "step": 400 }, { "epoch": 0.7, "learning_rate": 0.00046052173913043473, "loss": 0.364, "step": 401 }, { "epoch": 0.7, "learning_rate": 0.0004601739130434782, "loss": 0.4448, "step": 402 }, { "epoch": 0.7, "learning_rate": 0.0004598260869565217, "loss": 0.3708, "step": 403 }, { "epoch": 0.7, "learning_rate": 0.0004594782608695652, "loss": 0.3067, "step": 404 }, { "epoch": 0.7, "learning_rate": 0.00045913043478260866, "loss": 0.3807, "step": 405 }, { "epoch": 0.71, "learning_rate": 0.0004587826086956522, "loss": 0.232, "step": 406 }, { "epoch": 0.71, "learning_rate": 0.00045843478260869564, "loss": 0.423, "step": 407 }, { "epoch": 0.71, "learning_rate": 0.0004580869565217391, "loss": 0.3634, "step": 408 }, { "epoch": 0.71, "learning_rate": 0.0004577391304347825, "loss": 0.2996, "step": 409 }, { "epoch": 0.71, "learning_rate": 0.00045739130434782605, "loss": 0.4093, "step": 410 }, { "epoch": 0.71, "learning_rate": 0.0004570434782608695, "loss": 0.5011, "step": 411 }, { "epoch": 0.72, "learning_rate": 0.000456695652173913, "loss": 0.3568, "step": 412 }, { "epoch": 0.72, "learning_rate": 0.0004563478260869565, "loss": 0.4513, "step": 413 }, { "epoch": 0.72, "learning_rate": 0.00045599999999999997, "loss": 0.3784, "step": 414 }, { "epoch": 0.72, "learning_rate": 0.00045565217391304344, "loss": 0.3018, "step": 415 }, { "epoch": 0.72, "learning_rate": 0.00045530434782608696, "loss": 0.417, "step": 416 }, { "epoch": 0.73, "learning_rate": 0.00045495652173913043, "loss": 0.4267, "step": 417 }, { "epoch": 0.73, "learning_rate": 0.0004546086956521739, "loss": 0.3231, "step": 418 }, { "epoch": 0.73, "learning_rate": 0.0004542608695652173, "loss": 0.4543, "step": 419 }, { "epoch": 0.73, "learning_rate": 0.00045391304347826083, "loss": 0.4274, "step": 420 }, { "epoch": 0.73, "learning_rate": 0.0004535652173913043, "loss": 0.4032, "step": 421 }, { "epoch": 0.73, "learning_rate": 0.00045321739130434776, "loss": 0.3877, "step": 422 }, { "epoch": 0.74, "learning_rate": 0.0004528695652173913, "loss": 0.3042, "step": 423 }, { "epoch": 0.74, "learning_rate": 0.00045252173913043475, "loss": 0.2819, "step": 424 }, { "epoch": 0.74, "learning_rate": 0.0004521739130434782, "loss": 0.2814, "step": 425 }, { "epoch": 0.74, "learning_rate": 0.0004518260869565217, "loss": 0.361, "step": 426 }, { "epoch": 0.74, "learning_rate": 0.0004514782608695652, "loss": 0.4277, "step": 427 }, { "epoch": 0.74, "learning_rate": 0.0004511304347826087, "loss": 0.4062, "step": 428 }, { "epoch": 0.75, "learning_rate": 0.0004507826086956521, "loss": 0.471, "step": 429 }, { "epoch": 0.75, "learning_rate": 0.0004504347826086956, "loss": 0.4185, "step": 430 }, { "epoch": 0.75, "learning_rate": 0.0004500869565217391, "loss": 0.2808, "step": 431 }, { "epoch": 0.75, "learning_rate": 0.00044973913043478255, "loss": 0.3838, "step": 432 }, { "epoch": 0.75, "learning_rate": 0.00044939130434782607, "loss": 0.4828, "step": 433 }, { "epoch": 0.75, "learning_rate": 0.00044904347826086954, "loss": 0.3392, "step": 434 }, { "epoch": 0.76, "learning_rate": 0.000448695652173913, "loss": 0.4486, "step": 435 }, { "epoch": 0.76, "learning_rate": 0.00044834782608695647, "loss": 0.3491, "step": 436 }, { "epoch": 0.76, "learning_rate": 0.000448, "loss": 0.2978, "step": 437 }, { "epoch": 0.76, "learning_rate": 0.00044765217391304346, "loss": 0.3553, "step": 438 }, { "epoch": 0.76, "learning_rate": 0.0004473043478260869, "loss": 0.3361, "step": 439 }, { "epoch": 0.77, "learning_rate": 0.0004469565217391304, "loss": 0.2926, "step": 440 }, { "epoch": 0.77, "learning_rate": 0.00044660869565217386, "loss": 0.6323, "step": 441 }, { "epoch": 0.77, "learning_rate": 0.00044626086956521733, "loss": 0.3964, "step": 442 }, { "epoch": 0.77, "learning_rate": 0.0004459130434782608, "loss": 0.4509, "step": 443 }, { "epoch": 0.77, "learning_rate": 0.0004455652173913043, "loss": 0.3596, "step": 444 }, { "epoch": 0.77, "learning_rate": 0.0004452173913043478, "loss": 0.5505, "step": 445 }, { "epoch": 0.78, "learning_rate": 0.00044486956521739125, "loss": 0.4808, "step": 446 }, { "epoch": 0.78, "learning_rate": 0.0004445217391304348, "loss": 0.2984, "step": 447 }, { "epoch": 0.78, "learning_rate": 0.00044417391304347824, "loss": 0.4396, "step": 448 }, { "epoch": 0.78, "learning_rate": 0.0004438260869565217, "loss": 0.3515, "step": 449 }, { "epoch": 0.78, "learning_rate": 0.0004434782608695651, "loss": 0.4259, "step": 450 }, { "epoch": 0.78, "learning_rate": 0.00044313043478260865, "loss": 0.4089, "step": 451 }, { "epoch": 0.79, "learning_rate": 0.0004427826086956521, "loss": 0.3625, "step": 452 }, { "epoch": 0.79, "learning_rate": 0.0004424347826086956, "loss": 0.4137, "step": 453 }, { "epoch": 0.79, "learning_rate": 0.0004420869565217391, "loss": 0.3523, "step": 454 }, { "epoch": 0.79, "learning_rate": 0.00044173913043478257, "loss": 0.4537, "step": 455 }, { "epoch": 0.79, "learning_rate": 0.00044139130434782604, "loss": 0.3358, "step": 456 }, { "epoch": 0.79, "learning_rate": 0.00044104347826086956, "loss": 0.2377, "step": 457 }, { "epoch": 0.8, "learning_rate": 0.000440695652173913, "loss": 0.3512, "step": 458 }, { "epoch": 0.8, "learning_rate": 0.0004403478260869565, "loss": 0.3196, "step": 459 }, { "epoch": 0.8, "learning_rate": 0.0004399999999999999, "loss": 0.5847, "step": 460 }, { "epoch": 0.8, "learning_rate": 0.00043965217391304343, "loss": 0.5025, "step": 461 }, { "epoch": 0.8, "learning_rate": 0.0004393043478260869, "loss": 0.2878, "step": 462 }, { "epoch": 0.81, "learning_rate": 0.00043895652173913036, "loss": 0.391, "step": 463 }, { "epoch": 0.81, "learning_rate": 0.0004386086956521739, "loss": 0.337, "step": 464 }, { "epoch": 0.81, "learning_rate": 0.00043826086956521735, "loss": 0.3813, "step": 465 }, { "epoch": 0.81, "learning_rate": 0.0004379130434782608, "loss": 0.2528, "step": 466 }, { "epoch": 0.81, "learning_rate": 0.00043756521739130434, "loss": 0.3866, "step": 467 }, { "epoch": 0.81, "learning_rate": 0.0004372173913043478, "loss": 0.4909, "step": 468 }, { "epoch": 0.82, "learning_rate": 0.0004368695652173913, "loss": 0.3329, "step": 469 }, { "epoch": 0.82, "learning_rate": 0.0004365217391304347, "loss": 0.2729, "step": 470 }, { "epoch": 0.82, "learning_rate": 0.0004361739130434782, "loss": 0.316, "step": 471 }, { "epoch": 0.82, "learning_rate": 0.0004358260869565217, "loss": 0.446, "step": 472 }, { "epoch": 0.82, "learning_rate": 0.00043547826086956515, "loss": 0.4399, "step": 473 }, { "epoch": 0.82, "learning_rate": 0.00043513043478260867, "loss": 0.3535, "step": 474 }, { "epoch": 0.83, "learning_rate": 0.00043478260869565214, "loss": 0.4193, "step": 475 }, { "epoch": 0.83, "learning_rate": 0.0004344347826086956, "loss": 0.3247, "step": 476 }, { "epoch": 0.83, "learning_rate": 0.0004340869565217391, "loss": 0.3448, "step": 477 }, { "epoch": 0.83, "learning_rate": 0.0004337391304347826, "loss": 0.5503, "step": 478 }, { "epoch": 0.83, "learning_rate": 0.00043339130434782606, "loss": 0.5168, "step": 479 }, { "epoch": 0.83, "learning_rate": 0.0004330434782608696, "loss": 0.5161, "step": 480 }, { "epoch": 0.84, "learning_rate": 0.000432695652173913, "loss": 0.3386, "step": 481 }, { "epoch": 0.84, "learning_rate": 0.00043234782608695646, "loss": 0.3369, "step": 482 }, { "epoch": 0.84, "learning_rate": 0.00043199999999999993, "loss": 0.5102, "step": 483 }, { "epoch": 0.84, "learning_rate": 0.00043165217391304345, "loss": 0.4622, "step": 484 }, { "epoch": 0.84, "learning_rate": 0.0004313043478260869, "loss": 0.426, "step": 485 }, { "epoch": 0.85, "learning_rate": 0.0004309565217391304, "loss": 0.3391, "step": 486 }, { "epoch": 0.85, "learning_rate": 0.0004306086956521739, "loss": 0.52, "step": 487 }, { "epoch": 0.85, "learning_rate": 0.0004302608695652174, "loss": 0.3275, "step": 488 }, { "epoch": 0.85, "learning_rate": 0.00042991304347826084, "loss": 0.3569, "step": 489 }, { "epoch": 0.85, "learning_rate": 0.00042956521739130436, "loss": 0.4486, "step": 490 }, { "epoch": 0.85, "learning_rate": 0.0004292173913043478, "loss": 0.453, "step": 491 }, { "epoch": 0.86, "learning_rate": 0.00042886956521739125, "loss": 0.2868, "step": 492 }, { "epoch": 0.86, "learning_rate": 0.0004285217391304347, "loss": 0.3756, "step": 493 }, { "epoch": 0.86, "learning_rate": 0.00042817391304347823, "loss": 0.31, "step": 494 }, { "epoch": 0.86, "learning_rate": 0.0004278260869565217, "loss": 0.3135, "step": 495 }, { "epoch": 0.86, "learning_rate": 0.00042747826086956517, "loss": 0.4106, "step": 496 }, { "epoch": 0.86, "learning_rate": 0.0004271304347826087, "loss": 0.2658, "step": 497 }, { "epoch": 0.87, "learning_rate": 0.00042678260869565216, "loss": 0.3895, "step": 498 }, { "epoch": 0.87, "learning_rate": 0.0004264347826086956, "loss": 0.3984, "step": 499 }, { "epoch": 0.87, "learning_rate": 0.0004260869565217391, "loss": 0.369, "step": 500 }, { "epoch": 0.87, "learning_rate": 0.00042573913043478256, "loss": 0.2953, "step": 501 }, { "epoch": 0.87, "learning_rate": 0.00042539130434782603, "loss": 0.3431, "step": 502 }, { "epoch": 0.87, "learning_rate": 0.0004250434782608695, "loss": 0.428, "step": 503 }, { "epoch": 0.88, "learning_rate": 0.000424695652173913, "loss": 0.5424, "step": 504 }, { "epoch": 0.88, "learning_rate": 0.0004243478260869565, "loss": 0.4321, "step": 505 }, { "epoch": 0.88, "learning_rate": 0.00042399999999999995, "loss": 0.5802, "step": 506 }, { "epoch": 0.88, "learning_rate": 0.0004236521739130435, "loss": 0.5399, "step": 507 }, { "epoch": 0.88, "learning_rate": 0.00042330434782608694, "loss": 0.3867, "step": 508 }, { "epoch": 0.89, "learning_rate": 0.0004229565217391304, "loss": 0.4919, "step": 509 }, { "epoch": 0.89, "learning_rate": 0.0004226086956521739, "loss": 0.3548, "step": 510 }, { "epoch": 0.89, "learning_rate": 0.0004222608695652174, "loss": 0.4161, "step": 511 }, { "epoch": 0.89, "learning_rate": 0.0004219130434782608, "loss": 0.2782, "step": 512 }, { "epoch": 0.89, "learning_rate": 0.0004215652173913043, "loss": 0.5098, "step": 513 }, { "epoch": 0.89, "learning_rate": 0.0004212173913043478, "loss": 0.3401, "step": 514 }, { "epoch": 0.9, "learning_rate": 0.00042086956521739127, "loss": 0.4083, "step": 515 }, { "epoch": 0.9, "learning_rate": 0.00042052173913043473, "loss": 0.4416, "step": 516 }, { "epoch": 0.9, "learning_rate": 0.0004201739130434782, "loss": 0.3012, "step": 517 }, { "epoch": 0.9, "learning_rate": 0.0004198260869565217, "loss": 0.4786, "step": 518 }, { "epoch": 0.9, "learning_rate": 0.0004194782608695652, "loss": 0.4125, "step": 519 }, { "epoch": 0.9, "learning_rate": 0.00041913043478260866, "loss": 0.2584, "step": 520 }, { "epoch": 0.91, "learning_rate": 0.0004187826086956522, "loss": 0.3259, "step": 521 }, { "epoch": 0.91, "learning_rate": 0.0004184347826086956, "loss": 0.2705, "step": 522 }, { "epoch": 0.91, "learning_rate": 0.00041808695652173906, "loss": 0.3585, "step": 523 }, { "epoch": 0.91, "learning_rate": 0.00041773913043478253, "loss": 0.3614, "step": 524 }, { "epoch": 0.91, "learning_rate": 0.00041739130434782605, "loss": 0.3243, "step": 525 }, { "epoch": 0.91, "learning_rate": 0.0004170434782608695, "loss": 0.3566, "step": 526 }, { "epoch": 0.92, "learning_rate": 0.000416695652173913, "loss": 0.3798, "step": 527 }, { "epoch": 0.92, "learning_rate": 0.0004163478260869565, "loss": 0.4926, "step": 528 }, { "epoch": 0.92, "learning_rate": 0.000416, "loss": 0.3522, "step": 529 }, { "epoch": 0.92, "learning_rate": 0.00041565217391304344, "loss": 0.4664, "step": 530 }, { "epoch": 0.92, "learning_rate": 0.00041530434782608696, "loss": 0.3307, "step": 531 }, { "epoch": 0.93, "learning_rate": 0.0004149565217391304, "loss": 0.3522, "step": 532 }, { "epoch": 0.93, "learning_rate": 0.00041460869565217384, "loss": 0.5061, "step": 533 }, { "epoch": 0.93, "learning_rate": 0.0004142608695652173, "loss": 0.4236, "step": 534 }, { "epoch": 0.93, "learning_rate": 0.00041391304347826083, "loss": 0.4334, "step": 535 }, { "epoch": 0.93, "learning_rate": 0.0004135652173913043, "loss": 0.3511, "step": 536 }, { "epoch": 0.93, "learning_rate": 0.00041321739130434777, "loss": 0.3238, "step": 537 }, { "epoch": 0.94, "learning_rate": 0.0004128695652173913, "loss": 0.288, "step": 538 }, { "epoch": 0.94, "learning_rate": 0.00041252173913043476, "loss": 0.3685, "step": 539 }, { "epoch": 0.94, "learning_rate": 0.0004121739130434782, "loss": 0.3725, "step": 540 }, { "epoch": 0.94, "learning_rate": 0.00041182608695652175, "loss": 0.2928, "step": 541 }, { "epoch": 0.94, "learning_rate": 0.0004114782608695652, "loss": 0.4039, "step": 542 }, { "epoch": 0.94, "learning_rate": 0.00041113043478260863, "loss": 0.5417, "step": 543 }, { "epoch": 0.95, "learning_rate": 0.0004107826086956521, "loss": 0.3801, "step": 544 }, { "epoch": 0.95, "learning_rate": 0.0004104347826086956, "loss": 0.4705, "step": 545 }, { "epoch": 0.95, "learning_rate": 0.0004100869565217391, "loss": 0.6205, "step": 546 }, { "epoch": 0.95, "learning_rate": 0.00040973913043478255, "loss": 0.4249, "step": 547 }, { "epoch": 0.95, "learning_rate": 0.00040939130434782607, "loss": 0.348, "step": 548 }, { "epoch": 0.95, "learning_rate": 0.00040904347826086954, "loss": 0.3395, "step": 549 }, { "epoch": 0.96, "learning_rate": 0.000408695652173913, "loss": 0.5281, "step": 550 }, { "epoch": 0.96, "learning_rate": 0.00040834782608695653, "loss": 0.3491, "step": 551 }, { "epoch": 0.96, "learning_rate": 0.000408, "loss": 0.4315, "step": 552 }, { "epoch": 0.96, "learning_rate": 0.0004076521739130434, "loss": 0.5667, "step": 553 }, { "epoch": 0.96, "learning_rate": 0.0004073043478260869, "loss": 0.3469, "step": 554 }, { "epoch": 0.97, "learning_rate": 0.0004069565217391304, "loss": 0.4372, "step": 555 }, { "epoch": 0.97, "learning_rate": 0.00040660869565217387, "loss": 0.4783, "step": 556 }, { "epoch": 0.97, "learning_rate": 0.00040626086956521733, "loss": 0.4213, "step": 557 }, { "epoch": 0.97, "learning_rate": 0.00040591304347826086, "loss": 0.4771, "step": 558 }, { "epoch": 0.97, "learning_rate": 0.0004055652173913043, "loss": 0.3926, "step": 559 }, { "epoch": 0.97, "learning_rate": 0.0004052173913043478, "loss": 0.2505, "step": 560 }, { "epoch": 0.98, "learning_rate": 0.0004048695652173913, "loss": 0.358, "step": 561 }, { "epoch": 0.98, "learning_rate": 0.0004045217391304348, "loss": 0.4851, "step": 562 }, { "epoch": 0.98, "learning_rate": 0.0004041739130434782, "loss": 0.2478, "step": 563 }, { "epoch": 0.98, "learning_rate": 0.00040382608695652166, "loss": 0.404, "step": 564 }, { "epoch": 0.98, "learning_rate": 0.0004034782608695652, "loss": 0.4729, "step": 565 }, { "epoch": 0.98, "learning_rate": 0.00040313043478260865, "loss": 0.394, "step": 566 }, { "epoch": 0.99, "learning_rate": 0.0004027826086956521, "loss": 0.2706, "step": 567 }, { "epoch": 0.99, "learning_rate": 0.00040243478260869564, "loss": 0.25, "step": 568 }, { "epoch": 0.99, "learning_rate": 0.0004020869565217391, "loss": 0.3064, "step": 569 }, { "epoch": 0.99, "learning_rate": 0.0004017391304347826, "loss": 0.4389, "step": 570 }, { "epoch": 0.99, "learning_rate": 0.0004013913043478261, "loss": 0.3719, "step": 571 }, { "epoch": 0.99, "learning_rate": 0.00040104347826086956, "loss": 0.3651, "step": 572 }, { "epoch": 1.0, "learning_rate": 0.00040069565217391303, "loss": 0.3484, "step": 573 }, { "epoch": 1.0, "learning_rate": 0.00040034782608695644, "loss": 0.3339, "step": 574 }, { "epoch": 1.0, "learning_rate": 0.00039999999999999996, "loss": 0.8456, "step": 575 }, { "epoch": 1.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.8110184870018815, "eval_accuracy_Leaf": 0.6950893908357898, "eval_accuracy_Stem": 0.48876976417160484, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.8087739765339865, "eval_iou_Leaf": 0.6852369777662191, "eval_iou_Stem": 0.47911690075300445, "eval_iou_void": 0.0, "eval_loss": 0.307381808757782, "eval_mean_accuracy": 0.49871941050231905, "eval_mean_iou": 0.394625571010642, "eval_median_iou": 0.47911688685417175, "eval_overall_accuracy": 0.7053780840970094, "eval_runtime": 686.8098, "eval_samples_per_second": 2.092, "eval_steps_per_second": 0.21, "step": 575 }, { "epoch": 1.0, "learning_rate": 0.00039965217391304343, "loss": 0.2571, "step": 576 }, { "epoch": 1.0, "learning_rate": 0.0003993043478260869, "loss": 0.3624, "step": 577 }, { "epoch": 1.01, "learning_rate": 0.0003989565217391304, "loss": 0.3449, "step": 578 }, { "epoch": 1.01, "learning_rate": 0.0003986086956521739, "loss": 0.4753, "step": 579 }, { "epoch": 1.01, "learning_rate": 0.00039826086956521736, "loss": 0.3468, "step": 580 }, { "epoch": 1.01, "learning_rate": 0.0003979130434782609, "loss": 0.2702, "step": 581 }, { "epoch": 1.01, "learning_rate": 0.00039756521739130435, "loss": 0.3291, "step": 582 }, { "epoch": 1.01, "learning_rate": 0.0003972173913043478, "loss": 0.3126, "step": 583 }, { "epoch": 1.02, "learning_rate": 0.0003968695652173912, "loss": 0.3625, "step": 584 }, { "epoch": 1.02, "learning_rate": 0.00039652173913043475, "loss": 0.3841, "step": 585 }, { "epoch": 1.02, "learning_rate": 0.0003961739130434782, "loss": 0.3767, "step": 586 }, { "epoch": 1.02, "learning_rate": 0.0003958260869565217, "loss": 0.317, "step": 587 }, { "epoch": 1.02, "learning_rate": 0.0003954782608695652, "loss": 0.3128, "step": 588 }, { "epoch": 1.02, "learning_rate": 0.00039513043478260867, "loss": 0.5235, "step": 589 }, { "epoch": 1.03, "learning_rate": 0.00039478260869565214, "loss": 0.2397, "step": 590 }, { "epoch": 1.03, "learning_rate": 0.0003944347826086956, "loss": 0.2881, "step": 591 }, { "epoch": 1.03, "learning_rate": 0.00039408695652173913, "loss": 0.3104, "step": 592 }, { "epoch": 1.03, "learning_rate": 0.0003937391304347826, "loss": 0.4002, "step": 593 }, { "epoch": 1.03, "learning_rate": 0.000393391304347826, "loss": 0.4242, "step": 594 }, { "epoch": 1.03, "learning_rate": 0.00039304347826086953, "loss": 0.5364, "step": 595 }, { "epoch": 1.04, "learning_rate": 0.000392695652173913, "loss": 0.2638, "step": 596 }, { "epoch": 1.04, "learning_rate": 0.00039234782608695647, "loss": 0.359, "step": 597 }, { "epoch": 1.04, "learning_rate": 0.00039199999999999993, "loss": 0.6339, "step": 598 }, { "epoch": 1.04, "learning_rate": 0.00039165217391304345, "loss": 0.3221, "step": 599 }, { "epoch": 1.04, "learning_rate": 0.0003913043478260869, "loss": 0.4058, "step": 600 }, { "epoch": 1.05, "learning_rate": 0.0003909565217391304, "loss": 0.2303, "step": 601 }, { "epoch": 1.05, "learning_rate": 0.0003906086956521739, "loss": 0.4673, "step": 602 }, { "epoch": 1.05, "learning_rate": 0.0003902608695652174, "loss": 0.3878, "step": 603 }, { "epoch": 1.05, "learning_rate": 0.00038991304347826085, "loss": 0.5696, "step": 604 }, { "epoch": 1.05, "learning_rate": 0.00038956521739130426, "loss": 0.4453, "step": 605 }, { "epoch": 1.05, "learning_rate": 0.0003892173913043478, "loss": 0.3034, "step": 606 }, { "epoch": 1.06, "learning_rate": 0.00038886956521739125, "loss": 0.4836, "step": 607 }, { "epoch": 1.06, "learning_rate": 0.0003885217391304347, "loss": 0.3305, "step": 608 }, { "epoch": 1.06, "learning_rate": 0.00038817391304347824, "loss": 0.5087, "step": 609 }, { "epoch": 1.06, "learning_rate": 0.0003878260869565217, "loss": 0.4124, "step": 610 }, { "epoch": 1.06, "learning_rate": 0.00038747826086956517, "loss": 0.3658, "step": 611 }, { "epoch": 1.06, "learning_rate": 0.0003871304347826087, "loss": 0.3526, "step": 612 }, { "epoch": 1.07, "learning_rate": 0.00038678260869565216, "loss": 0.3807, "step": 613 }, { "epoch": 1.07, "learning_rate": 0.00038643478260869563, "loss": 0.2891, "step": 614 }, { "epoch": 1.07, "learning_rate": 0.00038608695652173904, "loss": 0.3146, "step": 615 }, { "epoch": 1.07, "learning_rate": 0.00038573913043478256, "loss": 0.3742, "step": 616 }, { "epoch": 1.07, "learning_rate": 0.00038539130434782603, "loss": 0.2868, "step": 617 }, { "epoch": 1.07, "learning_rate": 0.0003850434782608695, "loss": 0.4354, "step": 618 }, { "epoch": 1.08, "learning_rate": 0.000384695652173913, "loss": 0.444, "step": 619 }, { "epoch": 1.08, "learning_rate": 0.0003843478260869565, "loss": 0.3942, "step": 620 }, { "epoch": 1.08, "learning_rate": 0.00038399999999999996, "loss": 0.3801, "step": 621 }, { "epoch": 1.08, "learning_rate": 0.0003836521739130435, "loss": 0.4235, "step": 622 }, { "epoch": 1.08, "learning_rate": 0.00038330434782608694, "loss": 0.3242, "step": 623 }, { "epoch": 1.09, "learning_rate": 0.0003829565217391304, "loss": 0.4009, "step": 624 }, { "epoch": 1.09, "learning_rate": 0.0003826086956521738, "loss": 0.3567, "step": 625 }, { "epoch": 1.09, "learning_rate": 0.00038226086956521735, "loss": 0.3254, "step": 626 }, { "epoch": 1.09, "learning_rate": 0.0003819130434782608, "loss": 0.3807, "step": 627 }, { "epoch": 1.09, "learning_rate": 0.0003815652173913043, "loss": 0.5237, "step": 628 }, { "epoch": 1.09, "learning_rate": 0.0003812173913043478, "loss": 0.2505, "step": 629 }, { "epoch": 1.1, "learning_rate": 0.00038086956521739127, "loss": 0.3653, "step": 630 }, { "epoch": 1.1, "learning_rate": 0.00038052173913043474, "loss": 0.4238, "step": 631 }, { "epoch": 1.1, "learning_rate": 0.00038017391304347826, "loss": 0.3175, "step": 632 }, { "epoch": 1.1, "learning_rate": 0.00037982608695652173, "loss": 0.5053, "step": 633 }, { "epoch": 1.1, "learning_rate": 0.0003794782608695652, "loss": 0.4471, "step": 634 }, { "epoch": 1.1, "learning_rate": 0.0003791304347826087, "loss": 0.2352, "step": 635 }, { "epoch": 1.11, "learning_rate": 0.00037878260869565213, "loss": 0.3538, "step": 636 }, { "epoch": 1.11, "learning_rate": 0.0003784347826086956, "loss": 0.3092, "step": 637 }, { "epoch": 1.11, "learning_rate": 0.00037808695652173906, "loss": 0.3365, "step": 638 }, { "epoch": 1.11, "learning_rate": 0.0003777391304347826, "loss": 0.3143, "step": 639 }, { "epoch": 1.11, "learning_rate": 0.00037739130434782605, "loss": 0.3519, "step": 640 }, { "epoch": 1.11, "learning_rate": 0.0003770434782608695, "loss": 0.4772, "step": 641 }, { "epoch": 1.12, "learning_rate": 0.00037669565217391304, "loss": 0.4057, "step": 642 }, { "epoch": 1.12, "learning_rate": 0.0003763478260869565, "loss": 0.3467, "step": 643 }, { "epoch": 1.12, "learning_rate": 0.000376, "loss": 0.3516, "step": 644 }, { "epoch": 1.12, "learning_rate": 0.0003756521739130435, "loss": 0.4267, "step": 645 }, { "epoch": 1.12, "learning_rate": 0.0003753043478260869, "loss": 0.3348, "step": 646 }, { "epoch": 1.13, "learning_rate": 0.0003749565217391304, "loss": 0.4179, "step": 647 }, { "epoch": 1.13, "learning_rate": 0.00037460869565217385, "loss": 0.3291, "step": 648 }, { "epoch": 1.13, "learning_rate": 0.00037426086956521737, "loss": 0.3084, "step": 649 }, { "epoch": 1.13, "learning_rate": 0.00037391304347826084, "loss": 0.5092, "step": 650 }, { "epoch": 1.13, "learning_rate": 0.0003735652173913043, "loss": 0.3122, "step": 651 }, { "epoch": 1.13, "learning_rate": 0.0003732173913043478, "loss": 0.2745, "step": 652 }, { "epoch": 1.14, "learning_rate": 0.0003728695652173913, "loss": 0.3567, "step": 653 }, { "epoch": 1.14, "learning_rate": 0.00037252173913043476, "loss": 0.415, "step": 654 }, { "epoch": 1.14, "learning_rate": 0.0003721739130434783, "loss": 0.3022, "step": 655 }, { "epoch": 1.14, "learning_rate": 0.0003718260869565217, "loss": 0.3937, "step": 656 }, { "epoch": 1.14, "learning_rate": 0.00037147826086956516, "loss": 0.2756, "step": 657 }, { "epoch": 1.14, "learning_rate": 0.00037113043478260863, "loss": 0.3414, "step": 658 }, { "epoch": 1.15, "learning_rate": 0.00037078260869565215, "loss": 0.2865, "step": 659 }, { "epoch": 1.15, "learning_rate": 0.0003704347826086956, "loss": 0.397, "step": 660 }, { "epoch": 1.15, "learning_rate": 0.0003700869565217391, "loss": 0.3779, "step": 661 }, { "epoch": 1.15, "learning_rate": 0.0003697391304347826, "loss": 0.2735, "step": 662 }, { "epoch": 1.15, "learning_rate": 0.0003693913043478261, "loss": 0.5073, "step": 663 }, { "epoch": 1.15, "learning_rate": 0.00036904347826086954, "loss": 0.3262, "step": 664 }, { "epoch": 1.16, "learning_rate": 0.000368695652173913, "loss": 0.4989, "step": 665 }, { "epoch": 1.16, "learning_rate": 0.00036834782608695653, "loss": 0.3093, "step": 666 }, { "epoch": 1.16, "learning_rate": 0.00036799999999999995, "loss": 0.4083, "step": 667 }, { "epoch": 1.16, "learning_rate": 0.0003676521739130434, "loss": 0.3629, "step": 668 }, { "epoch": 1.16, "learning_rate": 0.00036730434782608694, "loss": 0.2459, "step": 669 }, { "epoch": 1.17, "learning_rate": 0.0003669565217391304, "loss": 0.314, "step": 670 }, { "epoch": 1.17, "learning_rate": 0.00036660869565217387, "loss": 0.4387, "step": 671 }, { "epoch": 1.17, "learning_rate": 0.00036626086956521734, "loss": 0.3654, "step": 672 }, { "epoch": 1.17, "learning_rate": 0.00036591304347826086, "loss": 0.3083, "step": 673 }, { "epoch": 1.17, "learning_rate": 0.0003655652173913043, "loss": 0.383, "step": 674 }, { "epoch": 1.17, "learning_rate": 0.0003652173913043478, "loss": 0.3955, "step": 675 }, { "epoch": 1.18, "learning_rate": 0.0003648695652173913, "loss": 0.3338, "step": 676 }, { "epoch": 1.18, "learning_rate": 0.00036452173913043473, "loss": 0.2822, "step": 677 }, { "epoch": 1.18, "learning_rate": 0.0003641739130434782, "loss": 0.278, "step": 678 }, { "epoch": 1.18, "learning_rate": 0.00036382608695652166, "loss": 0.4095, "step": 679 }, { "epoch": 1.18, "learning_rate": 0.0003634782608695652, "loss": 0.3356, "step": 680 }, { "epoch": 1.18, "learning_rate": 0.00036313043478260865, "loss": 0.4347, "step": 681 }, { "epoch": 1.19, "learning_rate": 0.0003627826086956521, "loss": 0.3884, "step": 682 }, { "epoch": 1.19, "learning_rate": 0.00036243478260869564, "loss": 0.4308, "step": 683 }, { "epoch": 1.19, "learning_rate": 0.0003620869565217391, "loss": 0.304, "step": 684 }, { "epoch": 1.19, "learning_rate": 0.0003617391304347826, "loss": 0.56, "step": 685 }, { "epoch": 1.19, "learning_rate": 0.0003613913043478261, "loss": 0.3017, "step": 686 }, { "epoch": 1.19, "learning_rate": 0.0003610434782608695, "loss": 0.2648, "step": 687 }, { "epoch": 1.2, "learning_rate": 0.000360695652173913, "loss": 0.2441, "step": 688 }, { "epoch": 1.2, "learning_rate": 0.00036034782608695645, "loss": 0.3861, "step": 689 }, { "epoch": 1.2, "learning_rate": 0.00035999999999999997, "loss": 0.3647, "step": 690 }, { "epoch": 1.2, "learning_rate": 0.00035965217391304344, "loss": 0.2324, "step": 691 }, { "epoch": 1.2, "learning_rate": 0.0003593043478260869, "loss": 0.3163, "step": 692 }, { "epoch": 1.21, "learning_rate": 0.0003589565217391304, "loss": 0.402, "step": 693 }, { "epoch": 1.21, "learning_rate": 0.0003586086956521739, "loss": 0.3335, "step": 694 }, { "epoch": 1.21, "learning_rate": 0.00035826086956521736, "loss": 0.594, "step": 695 }, { "epoch": 1.21, "learning_rate": 0.0003579130434782609, "loss": 0.4758, "step": 696 }, { "epoch": 1.21, "learning_rate": 0.00035756521739130435, "loss": 0.406, "step": 697 }, { "epoch": 1.21, "learning_rate": 0.00035721739130434776, "loss": 0.3677, "step": 698 }, { "epoch": 1.22, "learning_rate": 0.00035686956521739123, "loss": 0.326, "step": 699 }, { "epoch": 1.22, "learning_rate": 0.00035652173913043475, "loss": 0.4241, "step": 700 }, { "epoch": 1.22, "learning_rate": 0.0003561739130434782, "loss": 0.4543, "step": 701 }, { "epoch": 1.22, "learning_rate": 0.0003558260869565217, "loss": 0.4049, "step": 702 }, { "epoch": 1.22, "learning_rate": 0.0003554782608695652, "loss": 0.4055, "step": 703 }, { "epoch": 1.22, "learning_rate": 0.0003551304347826087, "loss": 0.4056, "step": 704 }, { "epoch": 1.23, "learning_rate": 0.00035478260869565214, "loss": 0.2938, "step": 705 }, { "epoch": 1.23, "learning_rate": 0.00035443478260869566, "loss": 0.2996, "step": 706 }, { "epoch": 1.23, "learning_rate": 0.00035408695652173913, "loss": 0.3949, "step": 707 }, { "epoch": 1.23, "learning_rate": 0.00035373913043478255, "loss": 0.3953, "step": 708 }, { "epoch": 1.23, "learning_rate": 0.000353391304347826, "loss": 0.3817, "step": 709 }, { "epoch": 1.23, "learning_rate": 0.00035304347826086953, "loss": 0.3552, "step": 710 }, { "epoch": 1.24, "learning_rate": 0.000352695652173913, "loss": 0.223, "step": 711 }, { "epoch": 1.24, "learning_rate": 0.00035234782608695647, "loss": 0.2821, "step": 712 }, { "epoch": 1.24, "learning_rate": 0.000352, "loss": 0.4723, "step": 713 }, { "epoch": 1.24, "learning_rate": 0.00035165217391304346, "loss": 0.3247, "step": 714 }, { "epoch": 1.24, "learning_rate": 0.0003513043478260869, "loss": 0.274, "step": 715 }, { "epoch": 1.25, "learning_rate": 0.00035095652173913045, "loss": 0.2609, "step": 716 }, { "epoch": 1.25, "learning_rate": 0.0003506086956521739, "loss": 0.2119, "step": 717 }, { "epoch": 1.25, "learning_rate": 0.00035026086956521733, "loss": 0.3938, "step": 718 }, { "epoch": 1.25, "learning_rate": 0.0003499130434782608, "loss": 0.3972, "step": 719 }, { "epoch": 1.25, "learning_rate": 0.0003495652173913043, "loss": 0.3058, "step": 720 }, { "epoch": 1.25, "learning_rate": 0.0003492173913043478, "loss": 0.25, "step": 721 }, { "epoch": 1.26, "learning_rate": 0.00034886956521739125, "loss": 0.3588, "step": 722 }, { "epoch": 1.26, "learning_rate": 0.0003485217391304348, "loss": 0.4598, "step": 723 }, { "epoch": 1.26, "learning_rate": 0.00034817391304347824, "loss": 0.3632, "step": 724 }, { "epoch": 1.26, "learning_rate": 0.0003478260869565217, "loss": 0.3338, "step": 725 }, { "epoch": 1.26, "learning_rate": 0.00034747826086956523, "loss": 0.3525, "step": 726 }, { "epoch": 1.26, "learning_rate": 0.0003471304347826087, "loss": 0.3905, "step": 727 }, { "epoch": 1.27, "learning_rate": 0.00034678260869565217, "loss": 0.4038, "step": 728 }, { "epoch": 1.27, "learning_rate": 0.0003464347826086956, "loss": 0.3371, "step": 729 }, { "epoch": 1.27, "learning_rate": 0.0003460869565217391, "loss": 0.2324, "step": 730 }, { "epoch": 1.27, "learning_rate": 0.00034573913043478257, "loss": 0.2718, "step": 731 }, { "epoch": 1.27, "learning_rate": 0.00034539130434782604, "loss": 0.3045, "step": 732 }, { "epoch": 1.27, "learning_rate": 0.00034504347826086956, "loss": 0.5286, "step": 733 }, { "epoch": 1.28, "learning_rate": 0.000344695652173913, "loss": 0.4424, "step": 734 }, { "epoch": 1.28, "learning_rate": 0.0003443478260869565, "loss": 0.2662, "step": 735 }, { "epoch": 1.28, "learning_rate": 0.000344, "loss": 0.4038, "step": 736 }, { "epoch": 1.28, "learning_rate": 0.0003436521739130435, "loss": 0.3982, "step": 737 }, { "epoch": 1.28, "learning_rate": 0.00034330434782608695, "loss": 0.3814, "step": 738 }, { "epoch": 1.29, "learning_rate": 0.00034295652173913036, "loss": 0.3256, "step": 739 }, { "epoch": 1.29, "learning_rate": 0.0003426086956521739, "loss": 0.388, "step": 740 }, { "epoch": 1.29, "learning_rate": 0.00034226086956521735, "loss": 0.4011, "step": 741 }, { "epoch": 1.29, "learning_rate": 0.0003419130434782608, "loss": 0.4138, "step": 742 }, { "epoch": 1.29, "learning_rate": 0.00034156521739130434, "loss": 0.5035, "step": 743 }, { "epoch": 1.29, "learning_rate": 0.0003412173913043478, "loss": 0.6184, "step": 744 }, { "epoch": 1.3, "learning_rate": 0.0003408695652173913, "loss": 0.2971, "step": 745 }, { "epoch": 1.3, "learning_rate": 0.00034052173913043474, "loss": 0.3584, "step": 746 }, { "epoch": 1.3, "learning_rate": 0.00034017391304347826, "loss": 0.3305, "step": 747 }, { "epoch": 1.3, "learning_rate": 0.00033982608695652173, "loss": 0.3788, "step": 748 }, { "epoch": 1.3, "learning_rate": 0.00033947826086956514, "loss": 0.379, "step": 749 }, { "epoch": 1.3, "learning_rate": 0.00033913043478260867, "loss": 0.3836, "step": 750 }, { "epoch": 1.31, "learning_rate": 0.00033878260869565213, "loss": 0.4517, "step": 751 }, { "epoch": 1.31, "learning_rate": 0.0003384347826086956, "loss": 0.4163, "step": 752 }, { "epoch": 1.31, "learning_rate": 0.00033808695652173907, "loss": 0.3844, "step": 753 }, { "epoch": 1.31, "learning_rate": 0.0003377391304347826, "loss": 0.4502, "step": 754 }, { "epoch": 1.31, "learning_rate": 0.00033739130434782606, "loss": 0.4209, "step": 755 }, { "epoch": 1.31, "learning_rate": 0.0003370434782608695, "loss": 0.484, "step": 756 }, { "epoch": 1.32, "learning_rate": 0.00033669565217391305, "loss": 0.3193, "step": 757 }, { "epoch": 1.32, "learning_rate": 0.0003363478260869565, "loss": 0.3006, "step": 758 }, { "epoch": 1.32, "learning_rate": 0.000336, "loss": 0.2794, "step": 759 }, { "epoch": 1.32, "learning_rate": 0.0003356521739130434, "loss": 0.4736, "step": 760 }, { "epoch": 1.32, "learning_rate": 0.0003353043478260869, "loss": 0.366, "step": 761 }, { "epoch": 1.33, "learning_rate": 0.0003349565217391304, "loss": 0.3338, "step": 762 }, { "epoch": 1.33, "learning_rate": 0.00033460869565217385, "loss": 0.2505, "step": 763 }, { "epoch": 1.33, "learning_rate": 0.00033426086956521737, "loss": 0.4106, "step": 764 }, { "epoch": 1.33, "learning_rate": 0.00033391304347826084, "loss": 0.3429, "step": 765 }, { "epoch": 1.33, "learning_rate": 0.0003335652173913043, "loss": 0.3044, "step": 766 }, { "epoch": 1.33, "learning_rate": 0.00033321739130434783, "loss": 0.3303, "step": 767 }, { "epoch": 1.34, "learning_rate": 0.0003328695652173913, "loss": 0.4461, "step": 768 }, { "epoch": 1.34, "learning_rate": 0.00033252173913043476, "loss": 0.3489, "step": 769 }, { "epoch": 1.34, "learning_rate": 0.0003321739130434782, "loss": 0.2388, "step": 770 }, { "epoch": 1.34, "learning_rate": 0.0003318260869565217, "loss": 0.4526, "step": 771 }, { "epoch": 1.34, "learning_rate": 0.00033147826086956517, "loss": 0.4578, "step": 772 }, { "epoch": 1.34, "learning_rate": 0.00033113043478260863, "loss": 0.2629, "step": 773 }, { "epoch": 1.35, "learning_rate": 0.00033078260869565216, "loss": 0.2679, "step": 774 }, { "epoch": 1.35, "learning_rate": 0.0003304347826086956, "loss": 0.2243, "step": 775 }, { "epoch": 1.35, "learning_rate": 0.0003300869565217391, "loss": 0.2146, "step": 776 }, { "epoch": 1.35, "learning_rate": 0.0003297391304347826, "loss": 0.2478, "step": 777 }, { "epoch": 1.35, "learning_rate": 0.0003293913043478261, "loss": 0.3407, "step": 778 }, { "epoch": 1.35, "learning_rate": 0.00032904347826086955, "loss": 0.3569, "step": 779 }, { "epoch": 1.36, "learning_rate": 0.00032869565217391296, "loss": 0.2524, "step": 780 }, { "epoch": 1.36, "learning_rate": 0.0003283478260869565, "loss": 0.332, "step": 781 }, { "epoch": 1.36, "learning_rate": 0.00032799999999999995, "loss": 0.348, "step": 782 }, { "epoch": 1.36, "learning_rate": 0.0003276521739130434, "loss": 0.2962, "step": 783 }, { "epoch": 1.36, "learning_rate": 0.00032730434782608694, "loss": 0.2492, "step": 784 }, { "epoch": 1.37, "learning_rate": 0.0003269565217391304, "loss": 0.318, "step": 785 }, { "epoch": 1.37, "learning_rate": 0.0003266086956521739, "loss": 0.2985, "step": 786 }, { "epoch": 1.37, "learning_rate": 0.0003262608695652174, "loss": 0.3436, "step": 787 }, { "epoch": 1.37, "learning_rate": 0.00032591304347826086, "loss": 0.3187, "step": 788 }, { "epoch": 1.37, "learning_rate": 0.00032556521739130433, "loss": 0.3607, "step": 789 }, { "epoch": 1.37, "learning_rate": 0.00032521739130434785, "loss": 0.3353, "step": 790 }, { "epoch": 1.38, "learning_rate": 0.00032486956521739127, "loss": 0.4273, "step": 791 }, { "epoch": 1.38, "learning_rate": 0.00032452173913043473, "loss": 0.3242, "step": 792 }, { "epoch": 1.38, "learning_rate": 0.0003241739130434782, "loss": 0.3407, "step": 793 }, { "epoch": 1.38, "learning_rate": 0.0003238260869565217, "loss": 0.3037, "step": 794 }, { "epoch": 1.38, "learning_rate": 0.0003234782608695652, "loss": 0.21, "step": 795 }, { "epoch": 1.38, "learning_rate": 0.00032313043478260866, "loss": 0.3219, "step": 796 }, { "epoch": 1.39, "learning_rate": 0.0003227826086956522, "loss": 0.3225, "step": 797 }, { "epoch": 1.39, "learning_rate": 0.00032243478260869565, "loss": 0.327, "step": 798 }, { "epoch": 1.39, "learning_rate": 0.0003220869565217391, "loss": 0.6016, "step": 799 }, { "epoch": 1.39, "learning_rate": 0.00032173913043478263, "loss": 0.23, "step": 800 }, { "epoch": 1.39, "learning_rate": 0.00032139130434782605, "loss": 0.2519, "step": 801 }, { "epoch": 1.39, "learning_rate": 0.0003210434782608695, "loss": 0.3434, "step": 802 }, { "epoch": 1.4, "learning_rate": 0.000320695652173913, "loss": 0.3145, "step": 803 }, { "epoch": 1.4, "learning_rate": 0.0003203478260869565, "loss": 0.3804, "step": 804 }, { "epoch": 1.4, "learning_rate": 0.00031999999999999997, "loss": 0.4655, "step": 805 }, { "epoch": 1.4, "learning_rate": 0.00031965217391304344, "loss": 0.4793, "step": 806 }, { "epoch": 1.4, "learning_rate": 0.00031930434782608696, "loss": 0.273, "step": 807 }, { "epoch": 1.41, "learning_rate": 0.00031895652173913043, "loss": 0.488, "step": 808 }, { "epoch": 1.41, "learning_rate": 0.0003186086956521739, "loss": 0.2987, "step": 809 }, { "epoch": 1.41, "learning_rate": 0.0003182608695652174, "loss": 0.2456, "step": 810 }, { "epoch": 1.41, "learning_rate": 0.00031791304347826083, "loss": 0.3452, "step": 811 }, { "epoch": 1.41, "learning_rate": 0.0003175652173913043, "loss": 0.3435, "step": 812 }, { "epoch": 1.41, "learning_rate": 0.00031721739130434777, "loss": 0.4191, "step": 813 }, { "epoch": 1.42, "learning_rate": 0.0003168695652173913, "loss": 0.3411, "step": 814 }, { "epoch": 1.42, "learning_rate": 0.00031652173913043475, "loss": 0.2708, "step": 815 }, { "epoch": 1.42, "learning_rate": 0.0003161739130434782, "loss": 0.3642, "step": 816 }, { "epoch": 1.42, "learning_rate": 0.00031582608695652174, "loss": 0.3442, "step": 817 }, { "epoch": 1.42, "learning_rate": 0.0003154782608695652, "loss": 0.4794, "step": 818 }, { "epoch": 1.42, "learning_rate": 0.0003151304347826087, "loss": 0.3589, "step": 819 }, { "epoch": 1.43, "learning_rate": 0.00031478260869565215, "loss": 0.3477, "step": 820 }, { "epoch": 1.43, "learning_rate": 0.00031443478260869567, "loss": 0.24, "step": 821 }, { "epoch": 1.43, "learning_rate": 0.0003140869565217391, "loss": 0.3263, "step": 822 }, { "epoch": 1.43, "learning_rate": 0.00031373913043478255, "loss": 0.3876, "step": 823 }, { "epoch": 1.43, "learning_rate": 0.00031339130434782607, "loss": 0.2837, "step": 824 }, { "epoch": 1.43, "learning_rate": 0.00031304347826086954, "loss": 0.2717, "step": 825 }, { "epoch": 1.44, "learning_rate": 0.000312695652173913, "loss": 0.3431, "step": 826 }, { "epoch": 1.44, "learning_rate": 0.00031234782608695647, "loss": 0.2522, "step": 827 }, { "epoch": 1.44, "learning_rate": 0.000312, "loss": 0.3218, "step": 828 }, { "epoch": 1.44, "learning_rate": 0.00031165217391304346, "loss": 0.2647, "step": 829 }, { "epoch": 1.44, "learning_rate": 0.00031130434782608693, "loss": 0.2915, "step": 830 }, { "epoch": 1.45, "learning_rate": 0.00031095652173913045, "loss": 0.5041, "step": 831 }, { "epoch": 1.45, "learning_rate": 0.00031060869565217386, "loss": 0.3255, "step": 832 }, { "epoch": 1.45, "learning_rate": 0.00031026086956521733, "loss": 0.58, "step": 833 }, { "epoch": 1.45, "learning_rate": 0.0003099130434782608, "loss": 0.2958, "step": 834 }, { "epoch": 1.45, "learning_rate": 0.0003095652173913043, "loss": 0.3019, "step": 835 }, { "epoch": 1.45, "learning_rate": 0.0003092173913043478, "loss": 0.3356, "step": 836 }, { "epoch": 1.46, "learning_rate": 0.00030886956521739126, "loss": 0.5016, "step": 837 }, { "epoch": 1.46, "learning_rate": 0.0003085217391304348, "loss": 0.3581, "step": 838 }, { "epoch": 1.46, "learning_rate": 0.00030817391304347824, "loss": 0.3394, "step": 839 }, { "epoch": 1.46, "learning_rate": 0.0003078260869565217, "loss": 0.3134, "step": 840 }, { "epoch": 1.46, "learning_rate": 0.00030747826086956523, "loss": 0.2886, "step": 841 }, { "epoch": 1.46, "learning_rate": 0.00030713043478260865, "loss": 0.3338, "step": 842 }, { "epoch": 1.47, "learning_rate": 0.0003067826086956521, "loss": 0.3502, "step": 843 }, { "epoch": 1.47, "learning_rate": 0.0003064347826086956, "loss": 0.2426, "step": 844 }, { "epoch": 1.47, "learning_rate": 0.0003060869565217391, "loss": 0.448, "step": 845 }, { "epoch": 1.47, "learning_rate": 0.00030573913043478257, "loss": 0.363, "step": 846 }, { "epoch": 1.47, "learning_rate": 0.00030539130434782604, "loss": 0.4046, "step": 847 }, { "epoch": 1.47, "learning_rate": 0.00030504347826086956, "loss": 0.3373, "step": 848 }, { "epoch": 1.48, "learning_rate": 0.00030469565217391303, "loss": 0.4224, "step": 849 }, { "epoch": 1.48, "learning_rate": 0.0003043478260869565, "loss": 0.4995, "step": 850 }, { "epoch": 1.48, "learning_rate": 0.000304, "loss": 0.4614, "step": 851 }, { "epoch": 1.48, "learning_rate": 0.0003036521739130435, "loss": 0.492, "step": 852 }, { "epoch": 1.48, "learning_rate": 0.0003033043478260869, "loss": 0.282, "step": 853 }, { "epoch": 1.49, "learning_rate": 0.00030295652173913036, "loss": 0.3267, "step": 854 }, { "epoch": 1.49, "learning_rate": 0.0003026086956521739, "loss": 0.3201, "step": 855 }, { "epoch": 1.49, "learning_rate": 0.00030226086956521735, "loss": 0.2887, "step": 856 }, { "epoch": 1.49, "learning_rate": 0.0003019130434782608, "loss": 0.3129, "step": 857 }, { "epoch": 1.49, "learning_rate": 0.00030156521739130434, "loss": 0.4633, "step": 858 }, { "epoch": 1.49, "learning_rate": 0.0003012173913043478, "loss": 0.2506, "step": 859 }, { "epoch": 1.5, "learning_rate": 0.0003008695652173913, "loss": 0.3127, "step": 860 }, { "epoch": 1.5, "learning_rate": 0.0003005217391304348, "loss": 0.4365, "step": 861 }, { "epoch": 1.5, "learning_rate": 0.00030017391304347827, "loss": 0.398, "step": 862 }, { "epoch": 1.5, "learning_rate": 0.00029982608695652173, "loss": 0.3683, "step": 863 }, { "epoch": 1.5, "learning_rate": 0.0002994782608695652, "loss": 0.6646, "step": 864 }, { "epoch": 1.5, "learning_rate": 0.00029913043478260867, "loss": 0.2647, "step": 865 }, { "epoch": 1.51, "learning_rate": 0.00029878260869565214, "loss": 0.2862, "step": 866 }, { "epoch": 1.51, "learning_rate": 0.0002984347826086956, "loss": 0.4037, "step": 867 }, { "epoch": 1.51, "learning_rate": 0.0002980869565217391, "loss": 0.4184, "step": 868 }, { "epoch": 1.51, "learning_rate": 0.0002977391304347826, "loss": 0.3171, "step": 869 }, { "epoch": 1.51, "learning_rate": 0.00029739130434782606, "loss": 0.3368, "step": 870 }, { "epoch": 1.51, "learning_rate": 0.00029704347826086953, "loss": 0.2614, "step": 871 }, { "epoch": 1.52, "learning_rate": 0.000296695652173913, "loss": 0.4808, "step": 872 }, { "epoch": 1.52, "learning_rate": 0.0002963478260869565, "loss": 0.2622, "step": 873 }, { "epoch": 1.52, "learning_rate": 0.000296, "loss": 0.3648, "step": 874 }, { "epoch": 1.52, "learning_rate": 0.00029565217391304345, "loss": 0.2683, "step": 875 }, { "epoch": 1.52, "learning_rate": 0.0002953043478260869, "loss": 0.3897, "step": 876 }, { "epoch": 1.53, "learning_rate": 0.0002949565217391304, "loss": 0.3407, "step": 877 }, { "epoch": 1.53, "learning_rate": 0.0002946086956521739, "loss": 0.2931, "step": 878 }, { "epoch": 1.53, "learning_rate": 0.0002942608695652174, "loss": 0.4788, "step": 879 }, { "epoch": 1.53, "learning_rate": 0.00029391304347826084, "loss": 0.2492, "step": 880 }, { "epoch": 1.53, "learning_rate": 0.0002935652173913043, "loss": 0.4518, "step": 881 }, { "epoch": 1.53, "learning_rate": 0.0002932173913043478, "loss": 0.3942, "step": 882 }, { "epoch": 1.54, "learning_rate": 0.0002928695652173913, "loss": 0.2902, "step": 883 }, { "epoch": 1.54, "learning_rate": 0.00029252173913043477, "loss": 0.325, "step": 884 }, { "epoch": 1.54, "learning_rate": 0.00029217391304347824, "loss": 0.3547, "step": 885 }, { "epoch": 1.54, "learning_rate": 0.0002918260869565217, "loss": 0.3568, "step": 886 }, { "epoch": 1.54, "learning_rate": 0.00029147826086956517, "loss": 0.3321, "step": 887 }, { "epoch": 1.54, "learning_rate": 0.0002911304347826087, "loss": 0.4452, "step": 888 }, { "epoch": 1.55, "learning_rate": 0.00029078260869565216, "loss": 0.4346, "step": 889 }, { "epoch": 1.55, "learning_rate": 0.0002904347826086956, "loss": 0.3435, "step": 890 }, { "epoch": 1.55, "learning_rate": 0.0002900869565217391, "loss": 0.3092, "step": 891 }, { "epoch": 1.55, "learning_rate": 0.00028973913043478256, "loss": 0.3276, "step": 892 }, { "epoch": 1.55, "learning_rate": 0.0002893913043478261, "loss": 0.3033, "step": 893 }, { "epoch": 1.55, "learning_rate": 0.00028904347826086955, "loss": 0.3189, "step": 894 }, { "epoch": 1.56, "learning_rate": 0.000288695652173913, "loss": 0.2963, "step": 895 }, { "epoch": 1.56, "learning_rate": 0.0002883478260869565, "loss": 0.215, "step": 896 }, { "epoch": 1.56, "learning_rate": 0.00028799999999999995, "loss": 0.3597, "step": 897 }, { "epoch": 1.56, "learning_rate": 0.0002876521739130435, "loss": 0.439, "step": 898 }, { "epoch": 1.56, "learning_rate": 0.00028730434782608694, "loss": 0.5415, "step": 899 }, { "epoch": 1.57, "learning_rate": 0.0002869565217391304, "loss": 0.2693, "step": 900 }, { "epoch": 1.57, "learning_rate": 0.0002866086956521739, "loss": 0.3523, "step": 901 }, { "epoch": 1.57, "learning_rate": 0.00028626086956521734, "loss": 0.2509, "step": 902 }, { "epoch": 1.57, "learning_rate": 0.00028591304347826087, "loss": 0.3695, "step": 903 }, { "epoch": 1.57, "learning_rate": 0.00028556521739130433, "loss": 0.4849, "step": 904 }, { "epoch": 1.57, "learning_rate": 0.0002852173913043478, "loss": 0.3107, "step": 905 }, { "epoch": 1.58, "learning_rate": 0.00028486956521739127, "loss": 0.3214, "step": 906 }, { "epoch": 1.58, "learning_rate": 0.00028452173913043474, "loss": 0.3098, "step": 907 }, { "epoch": 1.58, "learning_rate": 0.0002841739130434782, "loss": 0.3416, "step": 908 }, { "epoch": 1.58, "learning_rate": 0.0002838260869565217, "loss": 0.3165, "step": 909 }, { "epoch": 1.58, "learning_rate": 0.0002834782608695652, "loss": 0.3337, "step": 910 }, { "epoch": 1.58, "learning_rate": 0.00028313043478260866, "loss": 0.3711, "step": 911 }, { "epoch": 1.59, "learning_rate": 0.00028278260869565213, "loss": 0.496, "step": 912 }, { "epoch": 1.59, "learning_rate": 0.0002824347826086956, "loss": 0.413, "step": 913 }, { "epoch": 1.59, "learning_rate": 0.0002820869565217391, "loss": 0.3664, "step": 914 }, { "epoch": 1.59, "learning_rate": 0.0002817391304347826, "loss": 0.3327, "step": 915 }, { "epoch": 1.59, "learning_rate": 0.00028139130434782605, "loss": 0.2906, "step": 916 }, { "epoch": 1.59, "learning_rate": 0.0002810434782608695, "loss": 0.3398, "step": 917 }, { "epoch": 1.6, "learning_rate": 0.000280695652173913, "loss": 0.3101, "step": 918 }, { "epoch": 1.6, "learning_rate": 0.0002803478260869565, "loss": 0.3177, "step": 919 }, { "epoch": 1.6, "learning_rate": 0.00028, "loss": 0.2191, "step": 920 }, { "epoch": 1.6, "learning_rate": 0.00027965217391304344, "loss": 0.4353, "step": 921 }, { "epoch": 1.6, "learning_rate": 0.0002793043478260869, "loss": 0.4265, "step": 922 }, { "epoch": 1.61, "learning_rate": 0.0002789565217391304, "loss": 0.4226, "step": 923 }, { "epoch": 1.61, "learning_rate": 0.0002786086956521739, "loss": 0.3954, "step": 924 }, { "epoch": 1.61, "learning_rate": 0.00027826086956521737, "loss": 0.3186, "step": 925 }, { "epoch": 1.61, "learning_rate": 0.00027791304347826083, "loss": 0.3276, "step": 926 }, { "epoch": 1.61, "learning_rate": 0.00027756521739130436, "loss": 0.3975, "step": 927 }, { "epoch": 1.61, "learning_rate": 0.00027721739130434777, "loss": 0.3396, "step": 928 }, { "epoch": 1.62, "learning_rate": 0.0002768695652173913, "loss": 0.3777, "step": 929 }, { "epoch": 1.62, "learning_rate": 0.00027652173913043476, "loss": 0.3257, "step": 930 }, { "epoch": 1.62, "learning_rate": 0.0002761739130434782, "loss": 0.344, "step": 931 }, { "epoch": 1.62, "learning_rate": 0.00027582608695652175, "loss": 0.478, "step": 932 }, { "epoch": 1.62, "learning_rate": 0.00027547826086956516, "loss": 0.3339, "step": 933 }, { "epoch": 1.62, "learning_rate": 0.0002751304347826087, "loss": 0.2559, "step": 934 }, { "epoch": 1.63, "learning_rate": 0.00027478260869565215, "loss": 0.3054, "step": 935 }, { "epoch": 1.63, "learning_rate": 0.0002744347826086956, "loss": 0.2192, "step": 936 }, { "epoch": 1.63, "learning_rate": 0.00027408695652173914, "loss": 0.3356, "step": 937 }, { "epoch": 1.63, "learning_rate": 0.00027373913043478255, "loss": 0.3313, "step": 938 }, { "epoch": 1.63, "learning_rate": 0.0002733913043478261, "loss": 0.3927, "step": 939 }, { "epoch": 1.63, "learning_rate": 0.00027304347826086954, "loss": 0.2222, "step": 940 }, { "epoch": 1.64, "learning_rate": 0.000272695652173913, "loss": 0.3326, "step": 941 }, { "epoch": 1.64, "learning_rate": 0.00027234782608695653, "loss": 0.3389, "step": 942 }, { "epoch": 1.64, "learning_rate": 0.00027199999999999994, "loss": 0.2922, "step": 943 }, { "epoch": 1.64, "learning_rate": 0.00027165217391304347, "loss": 0.3532, "step": 944 }, { "epoch": 1.64, "learning_rate": 0.00027130434782608693, "loss": 0.4425, "step": 945 }, { "epoch": 1.65, "learning_rate": 0.0002709565217391304, "loss": 0.3157, "step": 946 }, { "epoch": 1.65, "learning_rate": 0.0002706086956521739, "loss": 0.3705, "step": 947 }, { "epoch": 1.65, "learning_rate": 0.00027026086956521734, "loss": 0.561, "step": 948 }, { "epoch": 1.65, "learning_rate": 0.00026991304347826086, "loss": 0.4315, "step": 949 }, { "epoch": 1.65, "learning_rate": 0.0002695652173913043, "loss": 0.3941, "step": 950 }, { "epoch": 1.65, "learning_rate": 0.0002692173913043478, "loss": 0.3158, "step": 951 }, { "epoch": 1.66, "learning_rate": 0.0002688695652173913, "loss": 0.5362, "step": 952 }, { "epoch": 1.66, "learning_rate": 0.0002685217391304347, "loss": 0.4373, "step": 953 }, { "epoch": 1.66, "learning_rate": 0.00026817391304347825, "loss": 0.391, "step": 954 }, { "epoch": 1.66, "learning_rate": 0.0002678260869565217, "loss": 0.3698, "step": 955 }, { "epoch": 1.66, "learning_rate": 0.0002674782608695652, "loss": 0.3701, "step": 956 }, { "epoch": 1.66, "learning_rate": 0.0002671304347826087, "loss": 0.2698, "step": 957 }, { "epoch": 1.67, "learning_rate": 0.00026678260869565217, "loss": 0.3169, "step": 958 }, { "epoch": 1.67, "learning_rate": 0.00026643478260869564, "loss": 0.2967, "step": 959 }, { "epoch": 1.67, "learning_rate": 0.0002660869565217391, "loss": 0.3969, "step": 960 }, { "epoch": 1.67, "learning_rate": 0.0002657391304347826, "loss": 0.2973, "step": 961 }, { "epoch": 1.67, "learning_rate": 0.0002653913043478261, "loss": 0.357, "step": 962 }, { "epoch": 1.67, "learning_rate": 0.00026504347826086956, "loss": 0.273, "step": 963 }, { "epoch": 1.68, "learning_rate": 0.00026469565217391303, "loss": 0.3449, "step": 964 }, { "epoch": 1.68, "learning_rate": 0.0002643478260869565, "loss": 0.4353, "step": 965 }, { "epoch": 1.68, "learning_rate": 0.00026399999999999997, "loss": 0.401, "step": 966 }, { "epoch": 1.68, "learning_rate": 0.0002636521739130435, "loss": 0.464, "step": 967 }, { "epoch": 1.68, "learning_rate": 0.00026330434782608696, "loss": 0.3986, "step": 968 }, { "epoch": 1.69, "learning_rate": 0.0002629565217391304, "loss": 0.3823, "step": 969 }, { "epoch": 1.69, "learning_rate": 0.0002626086956521739, "loss": 0.332, "step": 970 }, { "epoch": 1.69, "learning_rate": 0.00026226086956521736, "loss": 0.3691, "step": 971 }, { "epoch": 1.69, "learning_rate": 0.0002619130434782609, "loss": 0.4882, "step": 972 }, { "epoch": 1.69, "learning_rate": 0.00026156521739130435, "loss": 0.4955, "step": 973 }, { "epoch": 1.69, "learning_rate": 0.0002612173913043478, "loss": 0.553, "step": 974 }, { "epoch": 1.7, "learning_rate": 0.0002608695652173913, "loss": 0.2793, "step": 975 }, { "epoch": 1.7, "learning_rate": 0.00026052173913043475, "loss": 0.2847, "step": 976 }, { "epoch": 1.7, "learning_rate": 0.00026017391304347827, "loss": 0.3155, "step": 977 }, { "epoch": 1.7, "learning_rate": 0.00025982608695652174, "loss": 0.3096, "step": 978 }, { "epoch": 1.7, "learning_rate": 0.0002594782608695652, "loss": 0.3594, "step": 979 }, { "epoch": 1.7, "learning_rate": 0.0002591304347826087, "loss": 0.3836, "step": 980 }, { "epoch": 1.71, "learning_rate": 0.00025878260869565214, "loss": 0.4389, "step": 981 }, { "epoch": 1.71, "learning_rate": 0.0002584347826086956, "loss": 0.2998, "step": 982 }, { "epoch": 1.71, "learning_rate": 0.00025808695652173913, "loss": 0.3825, "step": 983 }, { "epoch": 1.71, "learning_rate": 0.0002577391304347826, "loss": 0.3447, "step": 984 }, { "epoch": 1.71, "learning_rate": 0.00025739130434782606, "loss": 0.3036, "step": 985 }, { "epoch": 1.71, "learning_rate": 0.00025704347826086953, "loss": 0.437, "step": 986 }, { "epoch": 1.72, "learning_rate": 0.000256695652173913, "loss": 0.3951, "step": 987 }, { "epoch": 1.72, "learning_rate": 0.0002563478260869565, "loss": 0.3229, "step": 988 }, { "epoch": 1.72, "learning_rate": 0.000256, "loss": 0.3605, "step": 989 }, { "epoch": 1.72, "learning_rate": 0.00025565217391304346, "loss": 0.3592, "step": 990 }, { "epoch": 1.72, "learning_rate": 0.0002553043478260869, "loss": 0.3971, "step": 991 }, { "epoch": 1.73, "learning_rate": 0.0002549565217391304, "loss": 0.3523, "step": 992 }, { "epoch": 1.73, "learning_rate": 0.0002546086956521739, "loss": 0.4208, "step": 993 }, { "epoch": 1.73, "learning_rate": 0.0002542608695652174, "loss": 0.337, "step": 994 }, { "epoch": 1.73, "learning_rate": 0.00025391304347826085, "loss": 0.3901, "step": 995 }, { "epoch": 1.73, "learning_rate": 0.0002535652173913043, "loss": 0.2341, "step": 996 }, { "epoch": 1.73, "learning_rate": 0.0002532173913043478, "loss": 0.2716, "step": 997 }, { "epoch": 1.74, "learning_rate": 0.0002528695652173913, "loss": 0.4546, "step": 998 }, { "epoch": 1.74, "learning_rate": 0.00025252173913043477, "loss": 0.2678, "step": 999 }, { "epoch": 1.74, "learning_rate": 0.00025217391304347824, "loss": 0.4069, "step": 1000 }, { "epoch": 1.74, "learning_rate": 0.0002518260869565217, "loss": 0.3137, "step": 1001 }, { "epoch": 1.74, "learning_rate": 0.0002514782608695652, "loss": 0.2574, "step": 1002 }, { "epoch": 1.74, "learning_rate": 0.0002511304347826087, "loss": 0.4708, "step": 1003 }, { "epoch": 1.75, "learning_rate": 0.00025078260869565216, "loss": 0.3406, "step": 1004 }, { "epoch": 1.75, "learning_rate": 0.00025043478260869563, "loss": 0.2784, "step": 1005 }, { "epoch": 1.75, "learning_rate": 0.0002500869565217391, "loss": 0.2939, "step": 1006 }, { "epoch": 1.75, "learning_rate": 0.00024973913043478257, "loss": 0.4822, "step": 1007 }, { "epoch": 1.75, "learning_rate": 0.0002493913043478261, "loss": 0.3539, "step": 1008 }, { "epoch": 1.75, "learning_rate": 0.00024904347826086955, "loss": 0.3372, "step": 1009 }, { "epoch": 1.76, "learning_rate": 0.000248695652173913, "loss": 0.2061, "step": 1010 }, { "epoch": 1.76, "learning_rate": 0.0002483478260869565, "loss": 0.2388, "step": 1011 }, { "epoch": 1.76, "learning_rate": 0.00024799999999999996, "loss": 0.5142, "step": 1012 }, { "epoch": 1.76, "learning_rate": 0.0002476521739130435, "loss": 0.2969, "step": 1013 }, { "epoch": 1.76, "learning_rate": 0.00024730434782608695, "loss": 0.2295, "step": 1014 }, { "epoch": 1.77, "learning_rate": 0.0002469565217391304, "loss": 0.3207, "step": 1015 }, { "epoch": 1.77, "learning_rate": 0.0002466086956521739, "loss": 0.3808, "step": 1016 }, { "epoch": 1.77, "learning_rate": 0.00024626086956521735, "loss": 0.2847, "step": 1017 }, { "epoch": 1.77, "learning_rate": 0.00024591304347826087, "loss": 0.2881, "step": 1018 }, { "epoch": 1.77, "learning_rate": 0.00024556521739130434, "loss": 0.3686, "step": 1019 }, { "epoch": 1.77, "learning_rate": 0.0002452173913043478, "loss": 0.2683, "step": 1020 }, { "epoch": 1.78, "learning_rate": 0.00024486956521739127, "loss": 0.5033, "step": 1021 }, { "epoch": 1.78, "learning_rate": 0.00024452173913043474, "loss": 0.3996, "step": 1022 }, { "epoch": 1.78, "learning_rate": 0.00024417391304347826, "loss": 0.3607, "step": 1023 }, { "epoch": 1.78, "learning_rate": 0.00024382608695652173, "loss": 0.4346, "step": 1024 }, { "epoch": 1.78, "learning_rate": 0.00024347826086956522, "loss": 0.3693, "step": 1025 }, { "epoch": 1.78, "learning_rate": 0.00024313043478260866, "loss": 0.2734, "step": 1026 }, { "epoch": 1.79, "learning_rate": 0.00024278260869565216, "loss": 0.3059, "step": 1027 }, { "epoch": 1.79, "learning_rate": 0.00024243478260869563, "loss": 0.2867, "step": 1028 }, { "epoch": 1.79, "learning_rate": 0.00024208695652173912, "loss": 0.2554, "step": 1029 }, { "epoch": 1.79, "learning_rate": 0.00024173913043478261, "loss": 0.343, "step": 1030 }, { "epoch": 1.79, "learning_rate": 0.00024139130434782605, "loss": 0.2843, "step": 1031 }, { "epoch": 1.79, "learning_rate": 0.00024104347826086955, "loss": 0.3353, "step": 1032 }, { "epoch": 1.8, "learning_rate": 0.00024069565217391302, "loss": 0.3459, "step": 1033 }, { "epoch": 1.8, "learning_rate": 0.0002403478260869565, "loss": 0.3811, "step": 1034 }, { "epoch": 1.8, "learning_rate": 0.00023999999999999998, "loss": 0.2954, "step": 1035 }, { "epoch": 1.8, "learning_rate": 0.00023965217391304345, "loss": 0.2563, "step": 1036 }, { "epoch": 1.8, "learning_rate": 0.00023930434782608694, "loss": 0.3566, "step": 1037 }, { "epoch": 1.81, "learning_rate": 0.0002389565217391304, "loss": 0.3634, "step": 1038 }, { "epoch": 1.81, "learning_rate": 0.0002386086956521739, "loss": 0.2767, "step": 1039 }, { "epoch": 1.81, "learning_rate": 0.00023826086956521737, "loss": 0.4205, "step": 1040 }, { "epoch": 1.81, "learning_rate": 0.00023791304347826084, "loss": 0.3618, "step": 1041 }, { "epoch": 1.81, "learning_rate": 0.0002375652173913043, "loss": 0.3112, "step": 1042 }, { "epoch": 1.81, "learning_rate": 0.0002372173913043478, "loss": 0.3396, "step": 1043 }, { "epoch": 1.82, "learning_rate": 0.0002368695652173913, "loss": 0.2922, "step": 1044 }, { "epoch": 1.82, "learning_rate": 0.00023652173913043476, "loss": 0.2501, "step": 1045 }, { "epoch": 1.82, "learning_rate": 0.00023617391304347823, "loss": 0.3225, "step": 1046 }, { "epoch": 1.82, "learning_rate": 0.0002358260869565217, "loss": 0.4191, "step": 1047 }, { "epoch": 1.82, "learning_rate": 0.0002354782608695652, "loss": 0.3188, "step": 1048 }, { "epoch": 1.82, "learning_rate": 0.00023513043478260869, "loss": 0.2666, "step": 1049 }, { "epoch": 1.83, "learning_rate": 0.00023478260869565215, "loss": 0.3352, "step": 1050 }, { "epoch": 1.83, "learning_rate": 0.00023443478260869565, "loss": 0.2264, "step": 1051 }, { "epoch": 1.83, "learning_rate": 0.0002340869565217391, "loss": 0.2726, "step": 1052 }, { "epoch": 1.83, "learning_rate": 0.00023373913043478258, "loss": 0.294, "step": 1053 }, { "epoch": 1.83, "learning_rate": 0.00023339130434782608, "loss": 0.3505, "step": 1054 }, { "epoch": 1.83, "learning_rate": 0.00023304347826086954, "loss": 0.2769, "step": 1055 }, { "epoch": 1.84, "learning_rate": 0.00023269565217391304, "loss": 0.3305, "step": 1056 }, { "epoch": 1.84, "learning_rate": 0.00023234782608695648, "loss": 0.2018, "step": 1057 }, { "epoch": 1.84, "learning_rate": 0.00023199999999999997, "loss": 0.41, "step": 1058 }, { "epoch": 1.84, "learning_rate": 0.00023165217391304347, "loss": 0.3117, "step": 1059 }, { "epoch": 1.84, "learning_rate": 0.00023130434782608694, "loss": 0.2729, "step": 1060 }, { "epoch": 1.85, "learning_rate": 0.00023095652173913043, "loss": 0.2457, "step": 1061 }, { "epoch": 1.85, "learning_rate": 0.00023060869565217387, "loss": 0.4961, "step": 1062 }, { "epoch": 1.85, "learning_rate": 0.00023026086956521737, "loss": 0.4638, "step": 1063 }, { "epoch": 1.85, "learning_rate": 0.00022991304347826086, "loss": 0.2526, "step": 1064 }, { "epoch": 1.85, "learning_rate": 0.00022956521739130433, "loss": 0.4157, "step": 1065 }, { "epoch": 1.85, "learning_rate": 0.00022921739130434782, "loss": 0.3303, "step": 1066 }, { "epoch": 1.86, "learning_rate": 0.00022886956521739126, "loss": 0.3238, "step": 1067 }, { "epoch": 1.86, "learning_rate": 0.00022852173913043476, "loss": 0.3999, "step": 1068 }, { "epoch": 1.86, "learning_rate": 0.00022817391304347825, "loss": 0.2353, "step": 1069 }, { "epoch": 1.86, "learning_rate": 0.00022782608695652172, "loss": 0.3044, "step": 1070 }, { "epoch": 1.86, "learning_rate": 0.00022747826086956521, "loss": 0.314, "step": 1071 }, { "epoch": 1.86, "learning_rate": 0.00022713043478260865, "loss": 0.3304, "step": 1072 }, { "epoch": 1.87, "learning_rate": 0.00022678260869565215, "loss": 0.4017, "step": 1073 }, { "epoch": 1.87, "learning_rate": 0.00022643478260869564, "loss": 0.311, "step": 1074 }, { "epoch": 1.87, "learning_rate": 0.0002260869565217391, "loss": 0.5809, "step": 1075 }, { "epoch": 1.87, "learning_rate": 0.0002257391304347826, "loss": 0.3818, "step": 1076 }, { "epoch": 1.87, "learning_rate": 0.00022539130434782605, "loss": 0.3422, "step": 1077 }, { "epoch": 1.87, "learning_rate": 0.00022504347826086954, "loss": 0.2452, "step": 1078 }, { "epoch": 1.88, "learning_rate": 0.00022469565217391303, "loss": 0.3727, "step": 1079 }, { "epoch": 1.88, "learning_rate": 0.0002243478260869565, "loss": 0.3351, "step": 1080 }, { "epoch": 1.88, "learning_rate": 0.000224, "loss": 0.3948, "step": 1081 }, { "epoch": 1.88, "learning_rate": 0.00022365217391304344, "loss": 0.3886, "step": 1082 }, { "epoch": 1.88, "learning_rate": 0.00022330434782608693, "loss": 0.4294, "step": 1083 }, { "epoch": 1.89, "learning_rate": 0.0002229565217391304, "loss": 0.2652, "step": 1084 }, { "epoch": 1.89, "learning_rate": 0.0002226086956521739, "loss": 0.2729, "step": 1085 }, { "epoch": 1.89, "learning_rate": 0.0002222608695652174, "loss": 0.3338, "step": 1086 }, { "epoch": 1.89, "learning_rate": 0.00022191304347826086, "loss": 0.3129, "step": 1087 }, { "epoch": 1.89, "learning_rate": 0.00022156521739130432, "loss": 0.325, "step": 1088 }, { "epoch": 1.89, "learning_rate": 0.0002212173913043478, "loss": 0.3535, "step": 1089 }, { "epoch": 1.9, "learning_rate": 0.00022086956521739128, "loss": 0.3733, "step": 1090 }, { "epoch": 1.9, "learning_rate": 0.00022052173913043478, "loss": 0.4352, "step": 1091 }, { "epoch": 1.9, "learning_rate": 0.00022017391304347825, "loss": 0.3028, "step": 1092 }, { "epoch": 1.9, "learning_rate": 0.00021982608695652171, "loss": 0.2932, "step": 1093 }, { "epoch": 1.9, "learning_rate": 0.00021947826086956518, "loss": 0.3209, "step": 1094 }, { "epoch": 1.9, "learning_rate": 0.00021913043478260868, "loss": 0.2974, "step": 1095 }, { "epoch": 1.91, "learning_rate": 0.00021878260869565217, "loss": 0.445, "step": 1096 }, { "epoch": 1.91, "learning_rate": 0.00021843478260869564, "loss": 0.3197, "step": 1097 }, { "epoch": 1.91, "learning_rate": 0.0002180869565217391, "loss": 0.3133, "step": 1098 }, { "epoch": 1.91, "learning_rate": 0.00021773913043478257, "loss": 0.293, "step": 1099 }, { "epoch": 1.91, "learning_rate": 0.00021739130434782607, "loss": 0.3651, "step": 1100 }, { "epoch": 1.91, "learning_rate": 0.00021704347826086956, "loss": 0.2625, "step": 1101 }, { "epoch": 1.92, "learning_rate": 0.00021669565217391303, "loss": 0.302, "step": 1102 }, { "epoch": 1.92, "learning_rate": 0.0002163478260869565, "loss": 0.4953, "step": 1103 }, { "epoch": 1.92, "learning_rate": 0.00021599999999999996, "loss": 0.3553, "step": 1104 }, { "epoch": 1.92, "learning_rate": 0.00021565217391304346, "loss": 0.4508, "step": 1105 }, { "epoch": 1.92, "learning_rate": 0.00021530434782608695, "loss": 0.3719, "step": 1106 }, { "epoch": 1.93, "learning_rate": 0.00021495652173913042, "loss": 0.3569, "step": 1107 }, { "epoch": 1.93, "learning_rate": 0.0002146086956521739, "loss": 0.3924, "step": 1108 }, { "epoch": 1.93, "learning_rate": 0.00021426086956521736, "loss": 0.3093, "step": 1109 }, { "epoch": 1.93, "learning_rate": 0.00021391304347826085, "loss": 0.4371, "step": 1110 }, { "epoch": 1.93, "learning_rate": 0.00021356521739130435, "loss": 0.3084, "step": 1111 }, { "epoch": 1.93, "learning_rate": 0.0002132173913043478, "loss": 0.4002, "step": 1112 }, { "epoch": 1.94, "learning_rate": 0.00021286956521739128, "loss": 0.2347, "step": 1113 }, { "epoch": 1.94, "learning_rate": 0.00021252173913043475, "loss": 0.3529, "step": 1114 }, { "epoch": 1.94, "learning_rate": 0.00021217391304347824, "loss": 0.3925, "step": 1115 }, { "epoch": 1.94, "learning_rate": 0.00021182608695652174, "loss": 0.3577, "step": 1116 }, { "epoch": 1.94, "learning_rate": 0.0002114782608695652, "loss": 0.3798, "step": 1117 }, { "epoch": 1.94, "learning_rate": 0.0002111304347826087, "loss": 0.2596, "step": 1118 }, { "epoch": 1.95, "learning_rate": 0.00021078260869565214, "loss": 0.2057, "step": 1119 }, { "epoch": 1.95, "learning_rate": 0.00021043478260869563, "loss": 0.2684, "step": 1120 }, { "epoch": 1.95, "learning_rate": 0.0002100869565217391, "loss": 0.3752, "step": 1121 }, { "epoch": 1.95, "learning_rate": 0.0002097391304347826, "loss": 0.4329, "step": 1122 }, { "epoch": 1.95, "learning_rate": 0.0002093913043478261, "loss": 0.4141, "step": 1123 }, { "epoch": 1.95, "learning_rate": 0.00020904347826086953, "loss": 0.2743, "step": 1124 }, { "epoch": 1.96, "learning_rate": 0.00020869565217391303, "loss": 0.2414, "step": 1125 }, { "epoch": 1.96, "learning_rate": 0.0002083478260869565, "loss": 0.3033, "step": 1126 }, { "epoch": 1.96, "learning_rate": 0.000208, "loss": 0.3001, "step": 1127 }, { "epoch": 1.96, "learning_rate": 0.00020765217391304348, "loss": 0.266, "step": 1128 }, { "epoch": 1.96, "learning_rate": 0.00020730434782608692, "loss": 0.3394, "step": 1129 }, { "epoch": 1.97, "learning_rate": 0.00020695652173913042, "loss": 0.2839, "step": 1130 }, { "epoch": 1.97, "learning_rate": 0.00020660869565217388, "loss": 0.4041, "step": 1131 }, { "epoch": 1.97, "learning_rate": 0.00020626086956521738, "loss": 0.3264, "step": 1132 }, { "epoch": 1.97, "learning_rate": 0.00020591304347826087, "loss": 0.2031, "step": 1133 }, { "epoch": 1.97, "learning_rate": 0.00020556521739130431, "loss": 0.2864, "step": 1134 }, { "epoch": 1.97, "learning_rate": 0.0002052173913043478, "loss": 0.2983, "step": 1135 }, { "epoch": 1.98, "learning_rate": 0.00020486956521739128, "loss": 0.2338, "step": 1136 }, { "epoch": 1.98, "learning_rate": 0.00020452173913043477, "loss": 0.4693, "step": 1137 }, { "epoch": 1.98, "learning_rate": 0.00020417391304347826, "loss": 0.3161, "step": 1138 }, { "epoch": 1.98, "learning_rate": 0.0002038260869565217, "loss": 0.3074, "step": 1139 }, { "epoch": 1.98, "learning_rate": 0.0002034782608695652, "loss": 0.3716, "step": 1140 }, { "epoch": 1.98, "learning_rate": 0.00020313043478260867, "loss": 0.243, "step": 1141 }, { "epoch": 1.99, "learning_rate": 0.00020278260869565216, "loss": 0.5021, "step": 1142 }, { "epoch": 1.99, "learning_rate": 0.00020243478260869566, "loss": 0.3433, "step": 1143 }, { "epoch": 1.99, "learning_rate": 0.0002020869565217391, "loss": 0.2027, "step": 1144 }, { "epoch": 1.99, "learning_rate": 0.0002017391304347826, "loss": 0.2163, "step": 1145 }, { "epoch": 1.99, "learning_rate": 0.00020139130434782606, "loss": 0.2879, "step": 1146 }, { "epoch": 1.99, "learning_rate": 0.00020104347826086955, "loss": 0.3489, "step": 1147 }, { "epoch": 2.0, "learning_rate": 0.00020069565217391305, "loss": 0.2787, "step": 1148 }, { "epoch": 2.0, "learning_rate": 0.00020034782608695652, "loss": 0.4153, "step": 1149 }, { "epoch": 2.0, "learning_rate": 0.00019999999999999998, "loss": 0.3006, "step": 1150 }, { "epoch": 2.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.8532984879206978, "eval_accuracy_Leaf": 0.7185656810575579, "eval_accuracy_Stem": 0.4139498396674361, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.8493599925969277, "eval_iou_Leaf": 0.7138631136302652, "eval_iou_Stem": 0.4091853222600305, "eval_iou_void": 0.0, "eval_loss": 0.28675782680511475, "eval_mean_accuracy": 0.496453502161423, "eval_mean_iou": 0.39448168569744474, "eval_median_iou": 0.4091853201389313, "eval_overall_accuracy": 0.722749354872955, "eval_runtime": 714.3323, "eval_samples_per_second": 2.012, "eval_steps_per_second": 0.202, "step": 1150 }, { "epoch": 2.0, "learning_rate": 0.00019965217391304345, "loss": 0.3233, "step": 1151 }, { "epoch": 2.0, "learning_rate": 0.00019930434782608694, "loss": 0.2777, "step": 1152 }, { "epoch": 2.01, "learning_rate": 0.00019895652173913044, "loss": 0.3289, "step": 1153 }, { "epoch": 2.01, "learning_rate": 0.0001986086956521739, "loss": 0.4115, "step": 1154 }, { "epoch": 2.01, "learning_rate": 0.00019826086956521737, "loss": 0.3861, "step": 1155 }, { "epoch": 2.01, "learning_rate": 0.00019791304347826084, "loss": 0.2473, "step": 1156 }, { "epoch": 2.01, "learning_rate": 0.00019756521739130434, "loss": 0.2871, "step": 1157 }, { "epoch": 2.01, "learning_rate": 0.0001972173913043478, "loss": 0.3825, "step": 1158 }, { "epoch": 2.02, "learning_rate": 0.0001968695652173913, "loss": 0.2056, "step": 1159 }, { "epoch": 2.02, "learning_rate": 0.00019652173913043477, "loss": 0.3283, "step": 1160 }, { "epoch": 2.02, "learning_rate": 0.00019617391304347823, "loss": 0.3827, "step": 1161 }, { "epoch": 2.02, "learning_rate": 0.00019582608695652173, "loss": 0.3609, "step": 1162 }, { "epoch": 2.02, "learning_rate": 0.0001954782608695652, "loss": 0.2806, "step": 1163 }, { "epoch": 2.02, "learning_rate": 0.0001951304347826087, "loss": 0.2534, "step": 1164 }, { "epoch": 2.03, "learning_rate": 0.00019478260869565213, "loss": 0.232, "step": 1165 }, { "epoch": 2.03, "learning_rate": 0.00019443478260869562, "loss": 0.2508, "step": 1166 }, { "epoch": 2.03, "learning_rate": 0.00019408695652173912, "loss": 0.2752, "step": 1167 }, { "epoch": 2.03, "learning_rate": 0.00019373913043478259, "loss": 0.2366, "step": 1168 }, { "epoch": 2.03, "learning_rate": 0.00019339130434782608, "loss": 0.278, "step": 1169 }, { "epoch": 2.03, "learning_rate": 0.00019304347826086952, "loss": 0.3245, "step": 1170 }, { "epoch": 2.04, "learning_rate": 0.00019269565217391302, "loss": 0.401, "step": 1171 }, { "epoch": 2.04, "learning_rate": 0.0001923478260869565, "loss": 0.288, "step": 1172 }, { "epoch": 2.04, "learning_rate": 0.00019199999999999998, "loss": 0.2014, "step": 1173 }, { "epoch": 2.04, "learning_rate": 0.00019165217391304347, "loss": 0.4144, "step": 1174 }, { "epoch": 2.04, "learning_rate": 0.0001913043478260869, "loss": 0.3711, "step": 1175 }, { "epoch": 2.05, "learning_rate": 0.0001909565217391304, "loss": 0.2766, "step": 1176 }, { "epoch": 2.05, "learning_rate": 0.0001906086956521739, "loss": 0.1841, "step": 1177 }, { "epoch": 2.05, "learning_rate": 0.00019026086956521737, "loss": 0.3294, "step": 1178 }, { "epoch": 2.05, "learning_rate": 0.00018991304347826086, "loss": 0.3815, "step": 1179 }, { "epoch": 2.05, "learning_rate": 0.00018956521739130436, "loss": 0.3552, "step": 1180 }, { "epoch": 2.05, "learning_rate": 0.0001892173913043478, "loss": 0.3285, "step": 1181 }, { "epoch": 2.06, "learning_rate": 0.0001888695652173913, "loss": 0.286, "step": 1182 }, { "epoch": 2.06, "learning_rate": 0.00018852173913043476, "loss": 0.4645, "step": 1183 }, { "epoch": 2.06, "learning_rate": 0.00018817391304347826, "loss": 0.2712, "step": 1184 }, { "epoch": 2.06, "learning_rate": 0.00018782608695652175, "loss": 0.3943, "step": 1185 }, { "epoch": 2.06, "learning_rate": 0.0001874782608695652, "loss": 0.3098, "step": 1186 }, { "epoch": 2.06, "learning_rate": 0.00018713043478260868, "loss": 0.2866, "step": 1187 }, { "epoch": 2.07, "learning_rate": 0.00018678260869565215, "loss": 0.2905, "step": 1188 }, { "epoch": 2.07, "learning_rate": 0.00018643478260869565, "loss": 0.3069, "step": 1189 }, { "epoch": 2.07, "learning_rate": 0.00018608695652173914, "loss": 0.3854, "step": 1190 }, { "epoch": 2.07, "learning_rate": 0.00018573913043478258, "loss": 0.318, "step": 1191 }, { "epoch": 2.07, "learning_rate": 0.00018539130434782608, "loss": 0.348, "step": 1192 }, { "epoch": 2.07, "learning_rate": 0.00018504347826086954, "loss": 0.3091, "step": 1193 }, { "epoch": 2.08, "learning_rate": 0.00018469565217391304, "loss": 0.3456, "step": 1194 }, { "epoch": 2.08, "learning_rate": 0.0001843478260869565, "loss": 0.3581, "step": 1195 }, { "epoch": 2.08, "learning_rate": 0.00018399999999999997, "loss": 0.2583, "step": 1196 }, { "epoch": 2.08, "learning_rate": 0.00018365217391304347, "loss": 0.3574, "step": 1197 }, { "epoch": 2.08, "learning_rate": 0.00018330434782608694, "loss": 0.3428, "step": 1198 }, { "epoch": 2.09, "learning_rate": 0.00018295652173913043, "loss": 0.3415, "step": 1199 }, { "epoch": 2.09, "learning_rate": 0.0001826086956521739, "loss": 0.3328, "step": 1200 }, { "epoch": 2.09, "learning_rate": 0.00018226086956521736, "loss": 0.3395, "step": 1201 }, { "epoch": 2.09, "learning_rate": 0.00018191304347826083, "loss": 0.3836, "step": 1202 }, { "epoch": 2.09, "learning_rate": 0.00018156521739130433, "loss": 0.434, "step": 1203 }, { "epoch": 2.09, "learning_rate": 0.00018121739130434782, "loss": 0.2106, "step": 1204 }, { "epoch": 2.1, "learning_rate": 0.0001808695652173913, "loss": 0.3312, "step": 1205 }, { "epoch": 2.1, "learning_rate": 0.00018052173913043476, "loss": 0.2948, "step": 1206 }, { "epoch": 2.1, "learning_rate": 0.00018017391304347822, "loss": 0.3724, "step": 1207 }, { "epoch": 2.1, "learning_rate": 0.00017982608695652172, "loss": 0.2751, "step": 1208 }, { "epoch": 2.1, "learning_rate": 0.0001794782608695652, "loss": 0.2986, "step": 1209 }, { "epoch": 2.1, "learning_rate": 0.00017913043478260868, "loss": 0.2708, "step": 1210 }, { "epoch": 2.11, "learning_rate": 0.00017878260869565217, "loss": 0.2941, "step": 1211 }, { "epoch": 2.11, "learning_rate": 0.00017843478260869561, "loss": 0.1984, "step": 1212 }, { "epoch": 2.11, "learning_rate": 0.0001780869565217391, "loss": 0.323, "step": 1213 }, { "epoch": 2.11, "learning_rate": 0.0001777391304347826, "loss": 0.2194, "step": 1214 }, { "epoch": 2.11, "learning_rate": 0.00017739130434782607, "loss": 0.3589, "step": 1215 }, { "epoch": 2.11, "learning_rate": 0.00017704347826086957, "loss": 0.2829, "step": 1216 }, { "epoch": 2.12, "learning_rate": 0.000176695652173913, "loss": 0.3372, "step": 1217 }, { "epoch": 2.12, "learning_rate": 0.0001763478260869565, "loss": 0.4135, "step": 1218 }, { "epoch": 2.12, "learning_rate": 0.000176, "loss": 0.3162, "step": 1219 }, { "epoch": 2.12, "learning_rate": 0.00017565217391304346, "loss": 0.3747, "step": 1220 }, { "epoch": 2.12, "learning_rate": 0.00017530434782608696, "loss": 0.2738, "step": 1221 }, { "epoch": 2.13, "learning_rate": 0.0001749565217391304, "loss": 0.3831, "step": 1222 }, { "epoch": 2.13, "learning_rate": 0.0001746086956521739, "loss": 0.4364, "step": 1223 }, { "epoch": 2.13, "learning_rate": 0.0001742608695652174, "loss": 0.2703, "step": 1224 }, { "epoch": 2.13, "learning_rate": 0.00017391304347826085, "loss": 0.3936, "step": 1225 }, { "epoch": 2.13, "learning_rate": 0.00017356521739130435, "loss": 0.3247, "step": 1226 }, { "epoch": 2.13, "learning_rate": 0.0001732173913043478, "loss": 0.2325, "step": 1227 }, { "epoch": 2.14, "learning_rate": 0.00017286956521739128, "loss": 0.3442, "step": 1228 }, { "epoch": 2.14, "learning_rate": 0.00017252173913043478, "loss": 0.3542, "step": 1229 }, { "epoch": 2.14, "learning_rate": 0.00017217391304347825, "loss": 0.3225, "step": 1230 }, { "epoch": 2.14, "learning_rate": 0.00017182608695652174, "loss": 0.4175, "step": 1231 }, { "epoch": 2.14, "learning_rate": 0.00017147826086956518, "loss": 0.2656, "step": 1232 }, { "epoch": 2.14, "learning_rate": 0.00017113043478260868, "loss": 0.2572, "step": 1233 }, { "epoch": 2.15, "learning_rate": 0.00017078260869565217, "loss": 0.3797, "step": 1234 }, { "epoch": 2.15, "learning_rate": 0.00017043478260869564, "loss": 0.3, "step": 1235 }, { "epoch": 2.15, "learning_rate": 0.00017008695652173913, "loss": 0.339, "step": 1236 }, { "epoch": 2.15, "learning_rate": 0.00016973913043478257, "loss": 0.2814, "step": 1237 }, { "epoch": 2.15, "learning_rate": 0.00016939130434782607, "loss": 0.3351, "step": 1238 }, { "epoch": 2.15, "learning_rate": 0.00016904347826086953, "loss": 0.2523, "step": 1239 }, { "epoch": 2.16, "learning_rate": 0.00016869565217391303, "loss": 0.3055, "step": 1240 }, { "epoch": 2.16, "learning_rate": 0.00016834782608695652, "loss": 0.2741, "step": 1241 }, { "epoch": 2.16, "learning_rate": 0.000168, "loss": 0.3143, "step": 1242 }, { "epoch": 2.16, "learning_rate": 0.00016765217391304346, "loss": 0.2097, "step": 1243 }, { "epoch": 2.16, "learning_rate": 0.00016730434782608693, "loss": 0.442, "step": 1244 }, { "epoch": 2.17, "learning_rate": 0.00016695652173913042, "loss": 0.3511, "step": 1245 }, { "epoch": 2.17, "learning_rate": 0.00016660869565217391, "loss": 0.2895, "step": 1246 }, { "epoch": 2.17, "learning_rate": 0.00016626086956521738, "loss": 0.4038, "step": 1247 }, { "epoch": 2.17, "learning_rate": 0.00016591304347826085, "loss": 0.3496, "step": 1248 }, { "epoch": 2.17, "learning_rate": 0.00016556521739130432, "loss": 0.2067, "step": 1249 }, { "epoch": 2.17, "learning_rate": 0.0001652173913043478, "loss": 0.3521, "step": 1250 }, { "epoch": 2.18, "learning_rate": 0.0001648695652173913, "loss": 0.2784, "step": 1251 }, { "epoch": 2.18, "learning_rate": 0.00016452173913043477, "loss": 0.3462, "step": 1252 }, { "epoch": 2.18, "learning_rate": 0.00016417391304347824, "loss": 0.3638, "step": 1253 }, { "epoch": 2.18, "learning_rate": 0.0001638260869565217, "loss": 0.3441, "step": 1254 }, { "epoch": 2.18, "learning_rate": 0.0001634782608695652, "loss": 0.2474, "step": 1255 }, { "epoch": 2.18, "learning_rate": 0.0001631304347826087, "loss": 0.2831, "step": 1256 }, { "epoch": 2.19, "learning_rate": 0.00016278260869565217, "loss": 0.5149, "step": 1257 }, { "epoch": 2.19, "learning_rate": 0.00016243478260869563, "loss": 0.329, "step": 1258 }, { "epoch": 2.19, "learning_rate": 0.0001620869565217391, "loss": 0.3152, "step": 1259 }, { "epoch": 2.19, "learning_rate": 0.0001617391304347826, "loss": 0.423, "step": 1260 }, { "epoch": 2.19, "learning_rate": 0.0001613913043478261, "loss": 0.2712, "step": 1261 }, { "epoch": 2.19, "learning_rate": 0.00016104347826086956, "loss": 0.2788, "step": 1262 }, { "epoch": 2.2, "learning_rate": 0.00016069565217391302, "loss": 0.2314, "step": 1263 }, { "epoch": 2.2, "learning_rate": 0.0001603478260869565, "loss": 0.2858, "step": 1264 }, { "epoch": 2.2, "learning_rate": 0.00015999999999999999, "loss": 0.3413, "step": 1265 }, { "epoch": 2.2, "learning_rate": 0.00015965217391304348, "loss": 0.2168, "step": 1266 }, { "epoch": 2.2, "learning_rate": 0.00015930434782608695, "loss": 0.2266, "step": 1267 }, { "epoch": 2.21, "learning_rate": 0.00015895652173913042, "loss": 0.3422, "step": 1268 }, { "epoch": 2.21, "learning_rate": 0.00015860869565217388, "loss": 0.2725, "step": 1269 }, { "epoch": 2.21, "learning_rate": 0.00015826086956521738, "loss": 0.2999, "step": 1270 }, { "epoch": 2.21, "learning_rate": 0.00015791304347826087, "loss": 0.2494, "step": 1271 }, { "epoch": 2.21, "learning_rate": 0.00015756521739130434, "loss": 0.205, "step": 1272 }, { "epoch": 2.21, "learning_rate": 0.00015721739130434783, "loss": 0.4418, "step": 1273 }, { "epoch": 2.22, "learning_rate": 0.00015686956521739127, "loss": 0.4359, "step": 1274 }, { "epoch": 2.22, "learning_rate": 0.00015652173913043477, "loss": 0.2698, "step": 1275 }, { "epoch": 2.22, "learning_rate": 0.00015617391304347824, "loss": 0.3063, "step": 1276 }, { "epoch": 2.22, "learning_rate": 0.00015582608695652173, "loss": 0.4975, "step": 1277 }, { "epoch": 2.22, "learning_rate": 0.00015547826086956523, "loss": 0.5001, "step": 1278 }, { "epoch": 2.22, "learning_rate": 0.00015513043478260867, "loss": 0.3068, "step": 1279 }, { "epoch": 2.23, "learning_rate": 0.00015478260869565216, "loss": 0.3207, "step": 1280 }, { "epoch": 2.23, "learning_rate": 0.00015443478260869563, "loss": 0.2101, "step": 1281 }, { "epoch": 2.23, "learning_rate": 0.00015408695652173912, "loss": 0.2542, "step": 1282 }, { "epoch": 2.23, "learning_rate": 0.00015373913043478262, "loss": 0.3934, "step": 1283 }, { "epoch": 2.23, "learning_rate": 0.00015339130434782606, "loss": 0.32, "step": 1284 }, { "epoch": 2.23, "learning_rate": 0.00015304347826086955, "loss": 0.2027, "step": 1285 }, { "epoch": 2.24, "learning_rate": 0.00015269565217391302, "loss": 0.3679, "step": 1286 }, { "epoch": 2.24, "learning_rate": 0.00015234782608695651, "loss": 0.3099, "step": 1287 }, { "epoch": 2.24, "learning_rate": 0.000152, "loss": 0.1998, "step": 1288 }, { "epoch": 2.24, "learning_rate": 0.00015165217391304345, "loss": 0.4469, "step": 1289 }, { "epoch": 2.24, "learning_rate": 0.00015130434782608694, "loss": 0.3395, "step": 1290 }, { "epoch": 2.25, "learning_rate": 0.0001509565217391304, "loss": 0.3683, "step": 1291 }, { "epoch": 2.25, "learning_rate": 0.0001506086956521739, "loss": 0.3085, "step": 1292 }, { "epoch": 2.25, "learning_rate": 0.0001502608695652174, "loss": 0.2675, "step": 1293 }, { "epoch": 2.25, "learning_rate": 0.00014991304347826087, "loss": 0.2731, "step": 1294 }, { "epoch": 2.25, "learning_rate": 0.00014956521739130433, "loss": 0.254, "step": 1295 }, { "epoch": 2.25, "learning_rate": 0.0001492173913043478, "loss": 0.3972, "step": 1296 }, { "epoch": 2.26, "learning_rate": 0.0001488695652173913, "loss": 0.2254, "step": 1297 }, { "epoch": 2.26, "learning_rate": 0.00014852173913043476, "loss": 0.297, "step": 1298 }, { "epoch": 2.26, "learning_rate": 0.00014817391304347826, "loss": 0.3231, "step": 1299 }, { "epoch": 2.26, "learning_rate": 0.00014782608695652173, "loss": 0.3312, "step": 1300 }, { "epoch": 2.26, "learning_rate": 0.0001474782608695652, "loss": 0.2958, "step": 1301 }, { "epoch": 2.26, "learning_rate": 0.0001471304347826087, "loss": 0.3097, "step": 1302 }, { "epoch": 2.27, "learning_rate": 0.00014678260869565216, "loss": 0.326, "step": 1303 }, { "epoch": 2.27, "learning_rate": 0.00014643478260869565, "loss": 0.3007, "step": 1304 }, { "epoch": 2.27, "learning_rate": 0.00014608695652173912, "loss": 0.421, "step": 1305 }, { "epoch": 2.27, "learning_rate": 0.00014573913043478259, "loss": 0.3162, "step": 1306 }, { "epoch": 2.27, "learning_rate": 0.00014539130434782608, "loss": 0.4608, "step": 1307 }, { "epoch": 2.27, "learning_rate": 0.00014504347826086955, "loss": 0.311, "step": 1308 }, { "epoch": 2.28, "learning_rate": 0.00014469565217391304, "loss": 0.3189, "step": 1309 }, { "epoch": 2.28, "learning_rate": 0.0001443478260869565, "loss": 0.3136, "step": 1310 }, { "epoch": 2.28, "learning_rate": 0.00014399999999999998, "loss": 0.2649, "step": 1311 }, { "epoch": 2.28, "learning_rate": 0.00014365217391304347, "loss": 0.3466, "step": 1312 }, { "epoch": 2.28, "learning_rate": 0.00014330434782608694, "loss": 0.4083, "step": 1313 }, { "epoch": 2.29, "learning_rate": 0.00014295652173913043, "loss": 0.319, "step": 1314 }, { "epoch": 2.29, "learning_rate": 0.0001426086956521739, "loss": 0.3381, "step": 1315 }, { "epoch": 2.29, "learning_rate": 0.00014226086956521737, "loss": 0.3564, "step": 1316 }, { "epoch": 2.29, "learning_rate": 0.00014191304347826086, "loss": 0.3181, "step": 1317 }, { "epoch": 2.29, "learning_rate": 0.00014156521739130433, "loss": 0.4336, "step": 1318 }, { "epoch": 2.29, "learning_rate": 0.0001412173913043478, "loss": 0.2732, "step": 1319 }, { "epoch": 2.3, "learning_rate": 0.0001408695652173913, "loss": 0.2792, "step": 1320 }, { "epoch": 2.3, "learning_rate": 0.00014052173913043476, "loss": 0.3937, "step": 1321 }, { "epoch": 2.3, "learning_rate": 0.00014017391304347825, "loss": 0.2907, "step": 1322 }, { "epoch": 2.3, "learning_rate": 0.00013982608695652172, "loss": 0.3378, "step": 1323 }, { "epoch": 2.3, "learning_rate": 0.0001394782608695652, "loss": 0.3276, "step": 1324 }, { "epoch": 2.3, "learning_rate": 0.00013913043478260868, "loss": 0.3007, "step": 1325 }, { "epoch": 2.31, "learning_rate": 0.00013878260869565218, "loss": 0.2826, "step": 1326 }, { "epoch": 2.31, "learning_rate": 0.00013843478260869565, "loss": 0.2056, "step": 1327 }, { "epoch": 2.31, "learning_rate": 0.0001380869565217391, "loss": 0.246, "step": 1328 }, { "epoch": 2.31, "learning_rate": 0.00013773913043478258, "loss": 0.2498, "step": 1329 }, { "epoch": 2.31, "learning_rate": 0.00013739130434782607, "loss": 0.4314, "step": 1330 }, { "epoch": 2.31, "learning_rate": 0.00013704347826086957, "loss": 0.3118, "step": 1331 }, { "epoch": 2.32, "learning_rate": 0.00013669565217391304, "loss": 0.2698, "step": 1332 }, { "epoch": 2.32, "learning_rate": 0.0001363478260869565, "loss": 0.2789, "step": 1333 }, { "epoch": 2.32, "learning_rate": 0.00013599999999999997, "loss": 0.3942, "step": 1334 }, { "epoch": 2.32, "learning_rate": 0.00013565217391304347, "loss": 0.2404, "step": 1335 }, { "epoch": 2.32, "learning_rate": 0.00013530434782608696, "loss": 0.3542, "step": 1336 }, { "epoch": 2.33, "learning_rate": 0.00013495652173913043, "loss": 0.5125, "step": 1337 }, { "epoch": 2.33, "learning_rate": 0.0001346086956521739, "loss": 0.2843, "step": 1338 }, { "epoch": 2.33, "learning_rate": 0.00013426086956521736, "loss": 0.217, "step": 1339 }, { "epoch": 2.33, "learning_rate": 0.00013391304347826086, "loss": 0.1839, "step": 1340 }, { "epoch": 2.33, "learning_rate": 0.00013356521739130435, "loss": 0.36, "step": 1341 }, { "epoch": 2.33, "learning_rate": 0.00013321739130434782, "loss": 0.3945, "step": 1342 }, { "epoch": 2.34, "learning_rate": 0.0001328695652173913, "loss": 0.3845, "step": 1343 }, { "epoch": 2.34, "learning_rate": 0.00013252173913043478, "loss": 0.3559, "step": 1344 }, { "epoch": 2.34, "learning_rate": 0.00013217391304347825, "loss": 0.244, "step": 1345 }, { "epoch": 2.34, "learning_rate": 0.00013182608695652174, "loss": 0.2651, "step": 1346 }, { "epoch": 2.34, "learning_rate": 0.0001314782608695652, "loss": 0.351, "step": 1347 }, { "epoch": 2.34, "learning_rate": 0.00013113043478260868, "loss": 0.2782, "step": 1348 }, { "epoch": 2.35, "learning_rate": 0.00013078260869565217, "loss": 0.2919, "step": 1349 }, { "epoch": 2.35, "learning_rate": 0.00013043478260869564, "loss": 0.269, "step": 1350 }, { "epoch": 2.35, "learning_rate": 0.00013008695652173914, "loss": 0.2785, "step": 1351 }, { "epoch": 2.35, "learning_rate": 0.0001297391304347826, "loss": 0.2727, "step": 1352 }, { "epoch": 2.35, "learning_rate": 0.00012939130434782607, "loss": 0.2564, "step": 1353 }, { "epoch": 2.35, "learning_rate": 0.00012904347826086956, "loss": 0.191, "step": 1354 }, { "epoch": 2.36, "learning_rate": 0.00012869565217391303, "loss": 0.2518, "step": 1355 }, { "epoch": 2.36, "learning_rate": 0.0001283478260869565, "loss": 0.3767, "step": 1356 }, { "epoch": 2.36, "learning_rate": 0.000128, "loss": 0.277, "step": 1357 }, { "epoch": 2.36, "learning_rate": 0.00012765217391304346, "loss": 0.3001, "step": 1358 }, { "epoch": 2.36, "learning_rate": 0.00012730434782608696, "loss": 0.3967, "step": 1359 }, { "epoch": 2.37, "learning_rate": 0.00012695652173913042, "loss": 0.2453, "step": 1360 }, { "epoch": 2.37, "learning_rate": 0.0001266086956521739, "loss": 0.2934, "step": 1361 }, { "epoch": 2.37, "learning_rate": 0.00012626086956521739, "loss": 0.2333, "step": 1362 }, { "epoch": 2.37, "learning_rate": 0.00012591304347826085, "loss": 0.2827, "step": 1363 }, { "epoch": 2.37, "learning_rate": 0.00012556521739130435, "loss": 0.3562, "step": 1364 }, { "epoch": 2.37, "learning_rate": 0.00012521739130434782, "loss": 0.2655, "step": 1365 }, { "epoch": 2.38, "learning_rate": 0.00012486956521739128, "loss": 0.2782, "step": 1366 }, { "epoch": 2.38, "learning_rate": 0.00012452173913043478, "loss": 0.3244, "step": 1367 }, { "epoch": 2.38, "learning_rate": 0.00012417391304347824, "loss": 0.2779, "step": 1368 }, { "epoch": 2.38, "learning_rate": 0.00012382608695652174, "loss": 0.2089, "step": 1369 }, { "epoch": 2.38, "learning_rate": 0.0001234782608695652, "loss": 0.2195, "step": 1370 }, { "epoch": 2.38, "learning_rate": 0.00012313043478260867, "loss": 0.3294, "step": 1371 }, { "epoch": 2.39, "learning_rate": 0.00012278260869565217, "loss": 0.3117, "step": 1372 }, { "epoch": 2.39, "learning_rate": 0.00012243478260869564, "loss": 0.3235, "step": 1373 }, { "epoch": 2.39, "learning_rate": 0.00012208695652173913, "loss": 0.296, "step": 1374 }, { "epoch": 2.39, "learning_rate": 0.00012173913043478261, "loss": 0.3479, "step": 1375 }, { "epoch": 2.39, "learning_rate": 0.00012139130434782608, "loss": 0.4613, "step": 1376 }, { "epoch": 2.39, "learning_rate": 0.00012104347826086956, "loss": 0.2355, "step": 1377 }, { "epoch": 2.4, "learning_rate": 0.00012069565217391303, "loss": 0.2734, "step": 1378 }, { "epoch": 2.4, "learning_rate": 0.00012034782608695651, "loss": 0.3967, "step": 1379 }, { "epoch": 2.4, "learning_rate": 0.00011999999999999999, "loss": 0.2016, "step": 1380 }, { "epoch": 2.4, "learning_rate": 0.00011965217391304347, "loss": 0.321, "step": 1381 }, { "epoch": 2.4, "learning_rate": 0.00011930434782608695, "loss": 0.3061, "step": 1382 }, { "epoch": 2.41, "learning_rate": 0.00011895652173913042, "loss": 0.2859, "step": 1383 }, { "epoch": 2.41, "learning_rate": 0.0001186086956521739, "loss": 0.2818, "step": 1384 }, { "epoch": 2.41, "learning_rate": 0.00011826086956521738, "loss": 0.3075, "step": 1385 }, { "epoch": 2.41, "learning_rate": 0.00011791304347826085, "loss": 0.2353, "step": 1386 }, { "epoch": 2.41, "learning_rate": 0.00011756521739130434, "loss": 0.2659, "step": 1387 }, { "epoch": 2.41, "learning_rate": 0.00011721739130434782, "loss": 0.2778, "step": 1388 }, { "epoch": 2.42, "learning_rate": 0.00011686956521739129, "loss": 0.2713, "step": 1389 }, { "epoch": 2.42, "learning_rate": 0.00011652173913043477, "loss": 0.3727, "step": 1390 }, { "epoch": 2.42, "learning_rate": 0.00011617391304347824, "loss": 0.335, "step": 1391 }, { "epoch": 2.42, "learning_rate": 0.00011582608695652173, "loss": 0.2139, "step": 1392 }, { "epoch": 2.42, "learning_rate": 0.00011547826086956522, "loss": 0.3357, "step": 1393 }, { "epoch": 2.42, "learning_rate": 0.00011513043478260868, "loss": 0.4045, "step": 1394 }, { "epoch": 2.43, "learning_rate": 0.00011478260869565216, "loss": 0.3843, "step": 1395 }, { "epoch": 2.43, "learning_rate": 0.00011443478260869563, "loss": 0.2346, "step": 1396 }, { "epoch": 2.43, "learning_rate": 0.00011408695652173913, "loss": 0.3184, "step": 1397 }, { "epoch": 2.43, "learning_rate": 0.00011373913043478261, "loss": 0.3259, "step": 1398 }, { "epoch": 2.43, "learning_rate": 0.00011339130434782607, "loss": 0.3341, "step": 1399 }, { "epoch": 2.43, "learning_rate": 0.00011304347826086956, "loss": 0.1949, "step": 1400 }, { "epoch": 2.44, "learning_rate": 0.00011269565217391302, "loss": 0.4095, "step": 1401 }, { "epoch": 2.44, "learning_rate": 0.00011234782608695652, "loss": 0.3426, "step": 1402 }, { "epoch": 2.44, "learning_rate": 0.000112, "loss": 0.2845, "step": 1403 }, { "epoch": 2.44, "learning_rate": 0.00011165217391304347, "loss": 0.303, "step": 1404 }, { "epoch": 2.44, "learning_rate": 0.00011130434782608695, "loss": 0.4721, "step": 1405 }, { "epoch": 2.45, "learning_rate": 0.00011095652173913043, "loss": 0.2737, "step": 1406 }, { "epoch": 2.45, "learning_rate": 0.0001106086956521739, "loss": 0.2195, "step": 1407 }, { "epoch": 2.45, "learning_rate": 0.00011026086956521739, "loss": 0.2645, "step": 1408 }, { "epoch": 2.45, "learning_rate": 0.00010991304347826086, "loss": 0.334, "step": 1409 }, { "epoch": 2.45, "learning_rate": 0.00010956521739130434, "loss": 0.3187, "step": 1410 }, { "epoch": 2.45, "learning_rate": 0.00010921739130434782, "loss": 0.2074, "step": 1411 }, { "epoch": 2.46, "learning_rate": 0.00010886956521739129, "loss": 0.2957, "step": 1412 }, { "epoch": 2.46, "learning_rate": 0.00010852173913043478, "loss": 0.3269, "step": 1413 }, { "epoch": 2.46, "learning_rate": 0.00010817391304347825, "loss": 0.4065, "step": 1414 }, { "epoch": 2.46, "learning_rate": 0.00010782608695652173, "loss": 0.3388, "step": 1415 }, { "epoch": 2.46, "learning_rate": 0.00010747826086956521, "loss": 0.3202, "step": 1416 }, { "epoch": 2.46, "learning_rate": 0.00010713043478260868, "loss": 0.3209, "step": 1417 }, { "epoch": 2.47, "learning_rate": 0.00010678260869565217, "loss": 0.3414, "step": 1418 }, { "epoch": 2.47, "learning_rate": 0.00010643478260869564, "loss": 0.3081, "step": 1419 }, { "epoch": 2.47, "learning_rate": 0.00010608695652173912, "loss": 0.2844, "step": 1420 }, { "epoch": 2.47, "learning_rate": 0.0001057391304347826, "loss": 0.1949, "step": 1421 }, { "epoch": 2.47, "learning_rate": 0.00010539130434782607, "loss": 0.2592, "step": 1422 }, { "epoch": 2.47, "learning_rate": 0.00010504347826086955, "loss": 0.3498, "step": 1423 }, { "epoch": 2.48, "learning_rate": 0.00010469565217391305, "loss": 0.2302, "step": 1424 }, { "epoch": 2.48, "learning_rate": 0.00010434782608695651, "loss": 0.2471, "step": 1425 }, { "epoch": 2.48, "learning_rate": 0.000104, "loss": 0.2638, "step": 1426 }, { "epoch": 2.48, "learning_rate": 0.00010365217391304346, "loss": 0.3004, "step": 1427 }, { "epoch": 2.48, "learning_rate": 0.00010330434782608694, "loss": 0.2579, "step": 1428 }, { "epoch": 2.49, "learning_rate": 0.00010295652173913044, "loss": 0.2953, "step": 1429 }, { "epoch": 2.49, "learning_rate": 0.0001026086956521739, "loss": 0.4076, "step": 1430 }, { "epoch": 2.49, "learning_rate": 0.00010226086956521739, "loss": 0.2334, "step": 1431 }, { "epoch": 2.49, "learning_rate": 0.00010191304347826085, "loss": 0.1946, "step": 1432 }, { "epoch": 2.49, "learning_rate": 0.00010156521739130433, "loss": 0.2564, "step": 1433 }, { "epoch": 2.49, "learning_rate": 0.00010121739130434783, "loss": 0.2803, "step": 1434 }, { "epoch": 2.5, "learning_rate": 0.0001008695652173913, "loss": 0.395, "step": 1435 }, { "epoch": 2.5, "learning_rate": 0.00010052173913043478, "loss": 0.2418, "step": 1436 }, { "epoch": 2.5, "learning_rate": 0.00010017391304347826, "loss": 0.2454, "step": 1437 }, { "epoch": 2.5, "learning_rate": 9.982608695652172e-05, "loss": 0.2656, "step": 1438 }, { "epoch": 2.5, "learning_rate": 9.947826086956522e-05, "loss": 0.2689, "step": 1439 }, { "epoch": 2.5, "learning_rate": 9.913043478260869e-05, "loss": 0.2325, "step": 1440 }, { "epoch": 2.51, "learning_rate": 9.878260869565217e-05, "loss": 0.2857, "step": 1441 }, { "epoch": 2.51, "learning_rate": 9.843478260869565e-05, "loss": 0.3396, "step": 1442 }, { "epoch": 2.51, "learning_rate": 9.808695652173912e-05, "loss": 0.3224, "step": 1443 }, { "epoch": 2.51, "learning_rate": 9.77391304347826e-05, "loss": 0.3051, "step": 1444 }, { "epoch": 2.51, "learning_rate": 9.739130434782606e-05, "loss": 0.3339, "step": 1445 }, { "epoch": 2.51, "learning_rate": 9.704347826086956e-05, "loss": 0.3053, "step": 1446 }, { "epoch": 2.52, "learning_rate": 9.669565217391304e-05, "loss": 0.2991, "step": 1447 }, { "epoch": 2.52, "learning_rate": 9.634782608695651e-05, "loss": 0.3333, "step": 1448 }, { "epoch": 2.52, "learning_rate": 9.599999999999999e-05, "loss": 0.2947, "step": 1449 }, { "epoch": 2.52, "learning_rate": 9.565217391304346e-05, "loss": 0.2476, "step": 1450 }, { "epoch": 2.52, "learning_rate": 9.530434782608695e-05, "loss": 0.3409, "step": 1451 }, { "epoch": 2.53, "learning_rate": 9.495652173913043e-05, "loss": 0.3909, "step": 1452 }, { "epoch": 2.53, "learning_rate": 9.46086956521739e-05, "loss": 0.2365, "step": 1453 }, { "epoch": 2.53, "learning_rate": 9.426086956521738e-05, "loss": 0.3003, "step": 1454 }, { "epoch": 2.53, "learning_rate": 9.391304347826087e-05, "loss": 0.2659, "step": 1455 }, { "epoch": 2.53, "learning_rate": 9.356521739130434e-05, "loss": 0.2712, "step": 1456 }, { "epoch": 2.53, "learning_rate": 9.321739130434782e-05, "loss": 0.3558, "step": 1457 }, { "epoch": 2.54, "learning_rate": 9.286956521739129e-05, "loss": 0.2234, "step": 1458 }, { "epoch": 2.54, "learning_rate": 9.252173913043477e-05, "loss": 0.3003, "step": 1459 }, { "epoch": 2.54, "learning_rate": 9.217391304347825e-05, "loss": 0.2507, "step": 1460 }, { "epoch": 2.54, "learning_rate": 9.182608695652173e-05, "loss": 0.3512, "step": 1461 }, { "epoch": 2.54, "learning_rate": 9.147826086956521e-05, "loss": 0.2121, "step": 1462 }, { "epoch": 2.54, "learning_rate": 9.113043478260868e-05, "loss": 0.326, "step": 1463 }, { "epoch": 2.55, "learning_rate": 9.078260869565216e-05, "loss": 0.3127, "step": 1464 }, { "epoch": 2.55, "learning_rate": 9.043478260869564e-05, "loss": 0.5193, "step": 1465 }, { "epoch": 2.55, "learning_rate": 9.008695652173911e-05, "loss": 0.2808, "step": 1466 }, { "epoch": 2.55, "learning_rate": 8.97391304347826e-05, "loss": 0.2331, "step": 1467 }, { "epoch": 2.55, "learning_rate": 8.939130434782609e-05, "loss": 0.3406, "step": 1468 }, { "epoch": 2.55, "learning_rate": 8.904347826086955e-05, "loss": 0.2811, "step": 1469 }, { "epoch": 2.56, "learning_rate": 8.869565217391304e-05, "loss": 0.4295, "step": 1470 }, { "epoch": 2.56, "learning_rate": 8.83478260869565e-05, "loss": 0.2397, "step": 1471 }, { "epoch": 2.56, "learning_rate": 8.8e-05, "loss": 0.2997, "step": 1472 }, { "epoch": 2.56, "learning_rate": 8.765217391304348e-05, "loss": 0.2767, "step": 1473 }, { "epoch": 2.56, "learning_rate": 8.730434782608695e-05, "loss": 0.2289, "step": 1474 }, { "epoch": 2.57, "learning_rate": 8.695652173913043e-05, "loss": 0.3861, "step": 1475 }, { "epoch": 2.57, "learning_rate": 8.66086956521739e-05, "loss": 0.3182, "step": 1476 }, { "epoch": 2.57, "learning_rate": 8.626086956521739e-05, "loss": 0.4049, "step": 1477 }, { "epoch": 2.57, "learning_rate": 8.591304347826087e-05, "loss": 0.2906, "step": 1478 }, { "epoch": 2.57, "learning_rate": 8.556521739130434e-05, "loss": 0.4008, "step": 1479 }, { "epoch": 2.57, "learning_rate": 8.521739130434782e-05, "loss": 0.3128, "step": 1480 }, { "epoch": 2.58, "learning_rate": 8.486956521739129e-05, "loss": 0.2515, "step": 1481 }, { "epoch": 2.58, "learning_rate": 8.452173913043477e-05, "loss": 0.295, "step": 1482 }, { "epoch": 2.58, "learning_rate": 8.417391304347826e-05, "loss": 0.3128, "step": 1483 }, { "epoch": 2.58, "learning_rate": 8.382608695652173e-05, "loss": 0.3944, "step": 1484 }, { "epoch": 2.58, "learning_rate": 8.347826086956521e-05, "loss": 0.3245, "step": 1485 }, { "epoch": 2.58, "learning_rate": 8.313043478260869e-05, "loss": 0.3515, "step": 1486 }, { "epoch": 2.59, "learning_rate": 8.278260869565216e-05, "loss": 0.3659, "step": 1487 }, { "epoch": 2.59, "learning_rate": 8.243478260869565e-05, "loss": 0.3487, "step": 1488 }, { "epoch": 2.59, "learning_rate": 8.208695652173912e-05, "loss": 0.318, "step": 1489 }, { "epoch": 2.59, "learning_rate": 8.17391304347826e-05, "loss": 0.2617, "step": 1490 }, { "epoch": 2.59, "learning_rate": 8.139130434782608e-05, "loss": 0.3491, "step": 1491 }, { "epoch": 2.59, "learning_rate": 8.104347826086955e-05, "loss": 0.2493, "step": 1492 }, { "epoch": 2.6, "learning_rate": 8.069565217391304e-05, "loss": 0.2479, "step": 1493 }, { "epoch": 2.6, "learning_rate": 8.034782608695651e-05, "loss": 0.1922, "step": 1494 }, { "epoch": 2.6, "learning_rate": 7.999999999999999e-05, "loss": 0.2336, "step": 1495 }, { "epoch": 2.6, "learning_rate": 7.965217391304347e-05, "loss": 0.3178, "step": 1496 }, { "epoch": 2.6, "learning_rate": 7.930434782608694e-05, "loss": 0.235, "step": 1497 }, { "epoch": 2.61, "learning_rate": 7.895652173913044e-05, "loss": 0.3054, "step": 1498 }, { "epoch": 2.61, "learning_rate": 7.860869565217392e-05, "loss": 0.3015, "step": 1499 }, { "epoch": 2.61, "learning_rate": 7.826086956521738e-05, "loss": 0.1912, "step": 1500 }, { "epoch": 2.61, "learning_rate": 7.791304347826087e-05, "loss": 0.4272, "step": 1501 }, { "epoch": 2.61, "learning_rate": 7.756521739130433e-05, "loss": 0.2557, "step": 1502 }, { "epoch": 2.61, "learning_rate": 7.721739130434781e-05, "loss": 0.3926, "step": 1503 }, { "epoch": 2.62, "learning_rate": 7.686956521739131e-05, "loss": 0.2378, "step": 1504 }, { "epoch": 2.62, "learning_rate": 7.652173913043478e-05, "loss": 0.2395, "step": 1505 }, { "epoch": 2.62, "learning_rate": 7.617391304347826e-05, "loss": 0.3099, "step": 1506 }, { "epoch": 2.62, "learning_rate": 7.582608695652172e-05, "loss": 0.1828, "step": 1507 }, { "epoch": 2.62, "learning_rate": 7.54782608695652e-05, "loss": 0.2712, "step": 1508 }, { "epoch": 2.62, "learning_rate": 7.51304347826087e-05, "loss": 0.3329, "step": 1509 }, { "epoch": 2.63, "learning_rate": 7.478260869565217e-05, "loss": 0.3156, "step": 1510 }, { "epoch": 2.63, "learning_rate": 7.443478260869565e-05, "loss": 0.317, "step": 1511 }, { "epoch": 2.63, "learning_rate": 7.408695652173913e-05, "loss": 0.2439, "step": 1512 }, { "epoch": 2.63, "learning_rate": 7.37391304347826e-05, "loss": 0.4082, "step": 1513 }, { "epoch": 2.63, "learning_rate": 7.339130434782608e-05, "loss": 0.2912, "step": 1514 }, { "epoch": 2.63, "learning_rate": 7.304347826086956e-05, "loss": 0.2324, "step": 1515 }, { "epoch": 2.64, "learning_rate": 7.269565217391304e-05, "loss": 0.3606, "step": 1516 }, { "epoch": 2.64, "learning_rate": 7.234782608695652e-05, "loss": 0.2016, "step": 1517 }, { "epoch": 2.64, "learning_rate": 7.199999999999999e-05, "loss": 0.2269, "step": 1518 }, { "epoch": 2.64, "learning_rate": 7.165217391304347e-05, "loss": 0.3401, "step": 1519 }, { "epoch": 2.64, "learning_rate": 7.130434782608695e-05, "loss": 0.2014, "step": 1520 }, { "epoch": 2.65, "learning_rate": 7.095652173913043e-05, "loss": 0.1915, "step": 1521 }, { "epoch": 2.65, "learning_rate": 7.06086956521739e-05, "loss": 0.2989, "step": 1522 }, { "epoch": 2.65, "learning_rate": 7.026086956521738e-05, "loss": 0.3301, "step": 1523 }, { "epoch": 2.65, "learning_rate": 6.991304347826086e-05, "loss": 0.3001, "step": 1524 }, { "epoch": 2.65, "learning_rate": 6.956521739130434e-05, "loss": 0.3358, "step": 1525 }, { "epoch": 2.65, "learning_rate": 6.921739130434782e-05, "loss": 0.3449, "step": 1526 }, { "epoch": 2.66, "learning_rate": 6.886956521739129e-05, "loss": 0.2583, "step": 1527 }, { "epoch": 2.66, "learning_rate": 6.852173913043478e-05, "loss": 0.2955, "step": 1528 }, { "epoch": 2.66, "learning_rate": 6.817391304347825e-05, "loss": 0.2425, "step": 1529 }, { "epoch": 2.66, "learning_rate": 6.782608695652173e-05, "loss": 0.24, "step": 1530 }, { "epoch": 2.66, "learning_rate": 6.747826086956521e-05, "loss": 0.2334, "step": 1531 }, { "epoch": 2.66, "learning_rate": 6.713043478260868e-05, "loss": 0.3057, "step": 1532 }, { "epoch": 2.67, "learning_rate": 6.678260869565218e-05, "loss": 0.3853, "step": 1533 }, { "epoch": 2.67, "learning_rate": 6.643478260869564e-05, "loss": 0.2729, "step": 1534 }, { "epoch": 2.67, "learning_rate": 6.608695652173912e-05, "loss": 0.2813, "step": 1535 }, { "epoch": 2.67, "learning_rate": 6.57391304347826e-05, "loss": 0.4779, "step": 1536 }, { "epoch": 2.67, "learning_rate": 6.539130434782609e-05, "loss": 0.3637, "step": 1537 }, { "epoch": 2.67, "learning_rate": 6.504347826086957e-05, "loss": 0.2781, "step": 1538 }, { "epoch": 2.68, "learning_rate": 6.469565217391304e-05, "loss": 0.2749, "step": 1539 }, { "epoch": 2.68, "learning_rate": 6.434782608695652e-05, "loss": 0.1929, "step": 1540 }, { "epoch": 2.68, "learning_rate": 6.4e-05, "loss": 0.3206, "step": 1541 }, { "epoch": 2.68, "learning_rate": 6.365217391304348e-05, "loss": 0.2598, "step": 1542 }, { "epoch": 2.68, "learning_rate": 6.330434782608695e-05, "loss": 0.3478, "step": 1543 }, { "epoch": 2.69, "learning_rate": 6.295652173913043e-05, "loss": 0.2381, "step": 1544 }, { "epoch": 2.69, "learning_rate": 6.260869565217391e-05, "loss": 0.2978, "step": 1545 }, { "epoch": 2.69, "learning_rate": 6.226086956521739e-05, "loss": 0.2803, "step": 1546 }, { "epoch": 2.69, "learning_rate": 6.191304347826087e-05, "loss": 0.2075, "step": 1547 }, { "epoch": 2.69, "learning_rate": 6.156521739130434e-05, "loss": 0.2317, "step": 1548 }, { "epoch": 2.69, "learning_rate": 6.121739130434782e-05, "loss": 0.399, "step": 1549 }, { "epoch": 2.7, "learning_rate": 6.0869565217391306e-05, "loss": 0.3344, "step": 1550 }, { "epoch": 2.7, "learning_rate": 6.052173913043478e-05, "loss": 0.2776, "step": 1551 }, { "epoch": 2.7, "learning_rate": 6.0173913043478254e-05, "loss": 0.2426, "step": 1552 }, { "epoch": 2.7, "learning_rate": 5.9826086956521735e-05, "loss": 0.223, "step": 1553 }, { "epoch": 2.7, "learning_rate": 5.947826086956521e-05, "loss": 0.2545, "step": 1554 }, { "epoch": 2.7, "learning_rate": 5.913043478260869e-05, "loss": 0.2278, "step": 1555 }, { "epoch": 2.71, "learning_rate": 5.878260869565217e-05, "loss": 0.2844, "step": 1556 }, { "epoch": 2.71, "learning_rate": 5.8434782608695646e-05, "loss": 0.3422, "step": 1557 }, { "epoch": 2.71, "learning_rate": 5.808695652173912e-05, "loss": 0.2209, "step": 1558 }, { "epoch": 2.71, "learning_rate": 5.773913043478261e-05, "loss": 0.2698, "step": 1559 }, { "epoch": 2.71, "learning_rate": 5.739130434782608e-05, "loss": 0.3442, "step": 1560 }, { "epoch": 2.71, "learning_rate": 5.704347826086956e-05, "loss": 0.2816, "step": 1561 }, { "epoch": 2.72, "learning_rate": 5.669565217391304e-05, "loss": 0.2404, "step": 1562 }, { "epoch": 2.72, "learning_rate": 5.634782608695651e-05, "loss": 0.542, "step": 1563 }, { "epoch": 2.72, "learning_rate": 5.6e-05, "loss": 0.2879, "step": 1564 }, { "epoch": 2.72, "learning_rate": 5.565217391304347e-05, "loss": 0.3866, "step": 1565 }, { "epoch": 2.72, "learning_rate": 5.530434782608695e-05, "loss": 0.2963, "step": 1566 }, { "epoch": 2.73, "learning_rate": 5.495652173913043e-05, "loss": 0.3033, "step": 1567 }, { "epoch": 2.73, "learning_rate": 5.460869565217391e-05, "loss": 0.2092, "step": 1568 }, { "epoch": 2.73, "learning_rate": 5.426086956521739e-05, "loss": 0.2667, "step": 1569 }, { "epoch": 2.73, "learning_rate": 5.3913043478260865e-05, "loss": 0.2471, "step": 1570 }, { "epoch": 2.73, "learning_rate": 5.356521739130434e-05, "loss": 0.2676, "step": 1571 }, { "epoch": 2.73, "learning_rate": 5.321739130434782e-05, "loss": 0.3182, "step": 1572 }, { "epoch": 2.74, "learning_rate": 5.28695652173913e-05, "loss": 0.3386, "step": 1573 }, { "epoch": 2.74, "learning_rate": 5.2521739130434775e-05, "loss": 0.3509, "step": 1574 }, { "epoch": 2.74, "learning_rate": 5.2173913043478256e-05, "loss": 0.2789, "step": 1575 }, { "epoch": 2.74, "learning_rate": 5.182608695652173e-05, "loss": 0.3167, "step": 1576 }, { "epoch": 2.74, "learning_rate": 5.147826086956522e-05, "loss": 0.1909, "step": 1577 }, { "epoch": 2.74, "learning_rate": 5.113043478260869e-05, "loss": 0.2356, "step": 1578 }, { "epoch": 2.75, "learning_rate": 5.078260869565217e-05, "loss": 0.2591, "step": 1579 }, { "epoch": 2.75, "learning_rate": 5.043478260869565e-05, "loss": 0.2463, "step": 1580 }, { "epoch": 2.75, "learning_rate": 5.008695652173913e-05, "loss": 0.2396, "step": 1581 }, { "epoch": 2.75, "learning_rate": 4.973913043478261e-05, "loss": 0.239, "step": 1582 }, { "epoch": 2.75, "learning_rate": 4.9391304347826084e-05, "loss": 0.2384, "step": 1583 }, { "epoch": 2.75, "learning_rate": 4.904347826086956e-05, "loss": 0.2505, "step": 1584 }, { "epoch": 2.76, "learning_rate": 4.869565217391303e-05, "loss": 0.2682, "step": 1585 }, { "epoch": 2.76, "learning_rate": 4.834782608695652e-05, "loss": 0.2181, "step": 1586 }, { "epoch": 2.76, "learning_rate": 4.7999999999999994e-05, "loss": 0.2555, "step": 1587 }, { "epoch": 2.76, "learning_rate": 4.7652173913043475e-05, "loss": 0.2495, "step": 1588 }, { "epoch": 2.76, "learning_rate": 4.730434782608695e-05, "loss": 0.3433, "step": 1589 }, { "epoch": 2.77, "learning_rate": 4.695652173913044e-05, "loss": 0.3357, "step": 1590 }, { "epoch": 2.77, "learning_rate": 4.660869565217391e-05, "loss": 0.3176, "step": 1591 }, { "epoch": 2.77, "learning_rate": 4.6260869565217386e-05, "loss": 0.401, "step": 1592 }, { "epoch": 2.77, "learning_rate": 4.591304347826087e-05, "loss": 0.2868, "step": 1593 }, { "epoch": 2.77, "learning_rate": 4.556521739130434e-05, "loss": 0.3916, "step": 1594 }, { "epoch": 2.77, "learning_rate": 4.521739130434782e-05, "loss": 0.2295, "step": 1595 }, { "epoch": 2.78, "learning_rate": 4.48695652173913e-05, "loss": 0.3138, "step": 1596 }, { "epoch": 2.78, "learning_rate": 4.452173913043478e-05, "loss": 0.2069, "step": 1597 }, { "epoch": 2.78, "learning_rate": 4.417391304347825e-05, "loss": 0.2486, "step": 1598 }, { "epoch": 2.78, "learning_rate": 4.382608695652174e-05, "loss": 0.343, "step": 1599 }, { "epoch": 2.78, "learning_rate": 4.3478260869565214e-05, "loss": 0.2816, "step": 1600 }, { "epoch": 2.78, "learning_rate": 4.3130434782608695e-05, "loss": 0.2336, "step": 1601 }, { "epoch": 2.79, "learning_rate": 4.278260869565217e-05, "loss": 0.2397, "step": 1602 }, { "epoch": 2.79, "learning_rate": 4.243478260869564e-05, "loss": 0.3141, "step": 1603 }, { "epoch": 2.79, "learning_rate": 4.208695652173913e-05, "loss": 0.4837, "step": 1604 }, { "epoch": 2.79, "learning_rate": 4.1739130434782605e-05, "loss": 0.2256, "step": 1605 }, { "epoch": 2.79, "learning_rate": 4.139130434782608e-05, "loss": 0.3139, "step": 1606 }, { "epoch": 2.79, "learning_rate": 4.104347826086956e-05, "loss": 0.301, "step": 1607 }, { "epoch": 2.8, "learning_rate": 4.069565217391304e-05, "loss": 0.3984, "step": 1608 }, { "epoch": 2.8, "learning_rate": 4.034782608695652e-05, "loss": 0.3004, "step": 1609 }, { "epoch": 2.8, "learning_rate": 3.9999999999999996e-05, "loss": 0.3901, "step": 1610 }, { "epoch": 2.8, "learning_rate": 3.965217391304347e-05, "loss": 0.2378, "step": 1611 }, { "epoch": 2.8, "learning_rate": 3.930434782608696e-05, "loss": 0.2739, "step": 1612 }, { "epoch": 2.81, "learning_rate": 3.895652173913043e-05, "loss": 0.3092, "step": 1613 }, { "epoch": 2.81, "learning_rate": 3.860869565217391e-05, "loss": 0.2732, "step": 1614 }, { "epoch": 2.81, "learning_rate": 3.826086956521739e-05, "loss": 0.3298, "step": 1615 }, { "epoch": 2.81, "learning_rate": 3.791304347826086e-05, "loss": 0.2385, "step": 1616 }, { "epoch": 2.81, "learning_rate": 3.756521739130435e-05, "loss": 0.2441, "step": 1617 }, { "epoch": 2.81, "learning_rate": 3.7217391304347824e-05, "loss": 0.2894, "step": 1618 }, { "epoch": 2.82, "learning_rate": 3.68695652173913e-05, "loss": 0.2321, "step": 1619 }, { "epoch": 2.82, "learning_rate": 3.652173913043478e-05, "loss": 0.3086, "step": 1620 }, { "epoch": 2.82, "learning_rate": 3.617391304347826e-05, "loss": 0.3151, "step": 1621 }, { "epoch": 2.82, "learning_rate": 3.5826086956521735e-05, "loss": 0.3295, "step": 1622 }, { "epoch": 2.82, "learning_rate": 3.5478260869565216e-05, "loss": 0.3315, "step": 1623 }, { "epoch": 2.82, "learning_rate": 3.513043478260869e-05, "loss": 0.2985, "step": 1624 }, { "epoch": 2.83, "learning_rate": 3.478260869565217e-05, "loss": 0.2171, "step": 1625 }, { "epoch": 2.83, "learning_rate": 3.4434782608695645e-05, "loss": 0.2541, "step": 1626 }, { "epoch": 2.83, "learning_rate": 3.4086956521739126e-05, "loss": 0.2957, "step": 1627 }, { "epoch": 2.83, "learning_rate": 3.373913043478261e-05, "loss": 0.2801, "step": 1628 }, { "epoch": 2.83, "learning_rate": 3.339130434782609e-05, "loss": 0.2779, "step": 1629 }, { "epoch": 2.83, "learning_rate": 3.304347826086956e-05, "loss": 0.2996, "step": 1630 }, { "epoch": 2.84, "learning_rate": 3.269565217391304e-05, "loss": 0.4337, "step": 1631 }, { "epoch": 2.84, "learning_rate": 3.234782608695652e-05, "loss": 0.2759, "step": 1632 }, { "epoch": 2.84, "learning_rate": 3.2e-05, "loss": 0.3173, "step": 1633 }, { "epoch": 2.84, "learning_rate": 3.165217391304347e-05, "loss": 0.3446, "step": 1634 }, { "epoch": 2.84, "learning_rate": 3.1304347826086954e-05, "loss": 0.1774, "step": 1635 }, { "epoch": 2.85, "learning_rate": 3.0956521739130435e-05, "loss": 0.3119, "step": 1636 }, { "epoch": 2.85, "learning_rate": 3.060869565217391e-05, "loss": 0.2941, "step": 1637 }, { "epoch": 2.85, "learning_rate": 3.026086956521739e-05, "loss": 0.317, "step": 1638 }, { "epoch": 2.85, "learning_rate": 2.9913043478260868e-05, "loss": 0.2171, "step": 1639 }, { "epoch": 2.85, "learning_rate": 2.9565217391304345e-05, "loss": 0.2673, "step": 1640 }, { "epoch": 2.85, "learning_rate": 2.9217391304347823e-05, "loss": 0.2218, "step": 1641 }, { "epoch": 2.86, "learning_rate": 2.8869565217391304e-05, "loss": 0.3178, "step": 1642 }, { "epoch": 2.86, "learning_rate": 2.852173913043478e-05, "loss": 0.2724, "step": 1643 }, { "epoch": 2.86, "learning_rate": 2.8173913043478256e-05, "loss": 0.3063, "step": 1644 }, { "epoch": 2.86, "learning_rate": 2.7826086956521737e-05, "loss": 0.2644, "step": 1645 }, { "epoch": 2.86, "learning_rate": 2.7478260869565214e-05, "loss": 0.2347, "step": 1646 }, { "epoch": 2.86, "learning_rate": 2.7130434782608695e-05, "loss": 0.2468, "step": 1647 }, { "epoch": 2.87, "learning_rate": 2.678260869565217e-05, "loss": 0.2708, "step": 1648 }, { "epoch": 2.87, "learning_rate": 2.643478260869565e-05, "loss": 0.3126, "step": 1649 }, { "epoch": 2.87, "learning_rate": 2.6086956521739128e-05, "loss": 0.2951, "step": 1650 }, { "epoch": 2.87, "learning_rate": 2.573913043478261e-05, "loss": 0.3103, "step": 1651 }, { "epoch": 2.87, "learning_rate": 2.5391304347826083e-05, "loss": 0.3051, "step": 1652 }, { "epoch": 2.87, "learning_rate": 2.5043478260869564e-05, "loss": 0.274, "step": 1653 }, { "epoch": 2.88, "learning_rate": 2.4695652173913042e-05, "loss": 0.2807, "step": 1654 }, { "epoch": 2.88, "learning_rate": 2.4347826086956516e-05, "loss": 0.2707, "step": 1655 }, { "epoch": 2.88, "learning_rate": 2.3999999999999997e-05, "loss": 0.3818, "step": 1656 }, { "epoch": 2.88, "learning_rate": 2.3652173913043475e-05, "loss": 0.2726, "step": 1657 }, { "epoch": 2.88, "learning_rate": 2.3304347826086956e-05, "loss": 0.2738, "step": 1658 }, { "epoch": 2.89, "learning_rate": 2.2956521739130433e-05, "loss": 0.2261, "step": 1659 }, { "epoch": 2.89, "learning_rate": 2.260869565217391e-05, "loss": 0.2226, "step": 1660 }, { "epoch": 2.89, "learning_rate": 2.226086956521739e-05, "loss": 0.2625, "step": 1661 }, { "epoch": 2.89, "learning_rate": 2.191304347826087e-05, "loss": 0.2189, "step": 1662 }, { "epoch": 2.89, "learning_rate": 2.1565217391304347e-05, "loss": 0.2388, "step": 1663 }, { "epoch": 2.89, "learning_rate": 2.121739130434782e-05, "loss": 0.3177, "step": 1664 }, { "epoch": 2.9, "learning_rate": 2.0869565217391303e-05, "loss": 0.3007, "step": 1665 }, { "epoch": 2.9, "learning_rate": 2.052173913043478e-05, "loss": 0.3468, "step": 1666 }, { "epoch": 2.9, "learning_rate": 2.017391304347826e-05, "loss": 0.2562, "step": 1667 }, { "epoch": 2.9, "learning_rate": 1.9826086956521735e-05, "loss": 0.1991, "step": 1668 }, { "epoch": 2.9, "learning_rate": 1.9478260869565216e-05, "loss": 0.2853, "step": 1669 }, { "epoch": 2.9, "learning_rate": 1.9130434782608694e-05, "loss": 0.4277, "step": 1670 }, { "epoch": 2.91, "learning_rate": 1.8782608695652175e-05, "loss": 0.2941, "step": 1671 }, { "epoch": 2.91, "learning_rate": 1.843478260869565e-05, "loss": 0.2767, "step": 1672 }, { "epoch": 2.91, "learning_rate": 1.808695652173913e-05, "loss": 0.2361, "step": 1673 }, { "epoch": 2.91, "learning_rate": 1.7739130434782608e-05, "loss": 0.2147, "step": 1674 }, { "epoch": 2.91, "learning_rate": 1.7391304347826085e-05, "loss": 0.3086, "step": 1675 }, { "epoch": 2.91, "learning_rate": 1.7043478260869563e-05, "loss": 0.4273, "step": 1676 }, { "epoch": 2.92, "learning_rate": 1.6695652173913044e-05, "loss": 0.3143, "step": 1677 }, { "epoch": 2.92, "learning_rate": 1.634782608695652e-05, "loss": 0.4031, "step": 1678 }, { "epoch": 2.92, "learning_rate": 1.6e-05, "loss": 0.243, "step": 1679 }, { "epoch": 2.92, "learning_rate": 1.5652173913043477e-05, "loss": 0.4075, "step": 1680 }, { "epoch": 2.92, "learning_rate": 1.5304347826086955e-05, "loss": 0.3498, "step": 1681 }, { "epoch": 2.93, "learning_rate": 1.4956521739130434e-05, "loss": 0.246, "step": 1682 }, { "epoch": 2.93, "learning_rate": 1.4608695652173911e-05, "loss": 0.293, "step": 1683 }, { "epoch": 2.93, "learning_rate": 1.426086956521739e-05, "loss": 0.2561, "step": 1684 }, { "epoch": 2.93, "learning_rate": 1.3913043478260868e-05, "loss": 0.5954, "step": 1685 }, { "epoch": 2.93, "learning_rate": 1.3565217391304348e-05, "loss": 0.2142, "step": 1686 }, { "epoch": 2.93, "learning_rate": 1.3217391304347825e-05, "loss": 0.2173, "step": 1687 }, { "epoch": 2.94, "learning_rate": 1.2869565217391305e-05, "loss": 0.2143, "step": 1688 }, { "epoch": 2.94, "learning_rate": 1.2521739130434782e-05, "loss": 0.2953, "step": 1689 }, { "epoch": 2.94, "learning_rate": 1.2173913043478258e-05, "loss": 0.3351, "step": 1690 }, { "epoch": 2.94, "learning_rate": 1.1826086956521737e-05, "loss": 0.2922, "step": 1691 }, { "epoch": 2.94, "learning_rate": 1.1478260869565217e-05, "loss": 0.2405, "step": 1692 }, { "epoch": 2.94, "learning_rate": 1.1130434782608694e-05, "loss": 0.3797, "step": 1693 }, { "epoch": 2.95, "learning_rate": 1.0782608695652174e-05, "loss": 0.3694, "step": 1694 }, { "epoch": 2.95, "learning_rate": 1.0434782608695651e-05, "loss": 0.2237, "step": 1695 }, { "epoch": 2.95, "learning_rate": 1.008695652173913e-05, "loss": 0.216, "step": 1696 }, { "epoch": 2.95, "learning_rate": 9.739130434782608e-06, "loss": 0.2825, "step": 1697 }, { "epoch": 2.95, "learning_rate": 9.391304347826087e-06, "loss": 0.2603, "step": 1698 }, { "epoch": 2.95, "learning_rate": 9.043478260869565e-06, "loss": 0.2748, "step": 1699 }, { "epoch": 2.96, "learning_rate": 8.695652173913043e-06, "loss": 0.2295, "step": 1700 }, { "epoch": 2.96, "learning_rate": 8.347826086956522e-06, "loss": 0.3298, "step": 1701 }, { "epoch": 2.96, "learning_rate": 8e-06, "loss": 0.3812, "step": 1702 }, { "epoch": 2.96, "learning_rate": 7.652173913043477e-06, "loss": 0.2487, "step": 1703 }, { "epoch": 2.96, "learning_rate": 7.304347826086956e-06, "loss": 0.3211, "step": 1704 }, { "epoch": 2.97, "learning_rate": 6.956521739130434e-06, "loss": 0.2762, "step": 1705 }, { "epoch": 2.97, "learning_rate": 6.608695652173913e-06, "loss": 0.2873, "step": 1706 }, { "epoch": 2.97, "learning_rate": 6.260869565217391e-06, "loss": 0.2798, "step": 1707 }, { "epoch": 2.97, "learning_rate": 5.913043478260869e-06, "loss": 0.1873, "step": 1708 }, { "epoch": 2.97, "learning_rate": 5.565217391304347e-06, "loss": 0.2583, "step": 1709 }, { "epoch": 2.97, "learning_rate": 5.217391304347826e-06, "loss": 0.2204, "step": 1710 }, { "epoch": 2.98, "learning_rate": 4.869565217391304e-06, "loss": 0.2843, "step": 1711 }, { "epoch": 2.98, "learning_rate": 4.5217391304347826e-06, "loss": 0.2502, "step": 1712 }, { "epoch": 2.98, "learning_rate": 4.173913043478261e-06, "loss": 0.2761, "step": 1713 }, { "epoch": 2.98, "learning_rate": 3.826086956521739e-06, "loss": 0.279, "step": 1714 }, { "epoch": 2.98, "learning_rate": 3.478260869565217e-06, "loss": 0.2164, "step": 1715 }, { "epoch": 2.98, "learning_rate": 3.1304347826086955e-06, "loss": 0.2541, "step": 1716 }, { "epoch": 2.99, "learning_rate": 2.7826086956521736e-06, "loss": 0.3425, "step": 1717 }, { "epoch": 2.99, "learning_rate": 2.434782608695652e-06, "loss": 0.2292, "step": 1718 }, { "epoch": 2.99, "learning_rate": 2.0869565217391305e-06, "loss": 0.4282, "step": 1719 }, { "epoch": 2.99, "learning_rate": 1.7391304347826085e-06, "loss": 0.2143, "step": 1720 }, { "epoch": 2.99, "learning_rate": 1.3913043478260868e-06, "loss": 0.253, "step": 1721 }, { "epoch": 2.99, "learning_rate": 1.0434782608695653e-06, "loss": 0.3221, "step": 1722 }, { "epoch": 3.0, "learning_rate": 6.956521739130434e-07, "loss": 0.3156, "step": 1723 }, { "epoch": 3.0, "learning_rate": 3.478260869565217e-07, "loss": 0.2779, "step": 1724 }, { "epoch": 3.0, "learning_rate": 0.0, "loss": 0.3315, "step": 1725 }, { "epoch": 3.0, "eval_accuracy_Flower": 0.0, "eval_accuracy_Fruit": 0.8642061740279863, "eval_accuracy_Leaf": 0.7167276511452706, "eval_accuracy_Stem": 0.5563431060988867, "eval_accuracy_void": NaN, "eval_iou_Flower": 0.0, "eval_iou_Fruit": 0.8604970362487987, "eval_iou_Leaf": 0.7107696156044375, "eval_iou_Stem": 0.544019147132753, "eval_iou_void": 0.0, "eval_loss": 0.2873678207397461, "eval_mean_accuracy": 0.5343192328180358, "eval_mean_iou": 0.4230571597971979, "eval_median_iou": 0.5440191626548767, "eval_overall_accuracy": 0.7437177825812409, "eval_runtime": 719.9951, "eval_samples_per_second": 1.996, "eval_steps_per_second": 0.2, "step": 1725 }, { "epoch": 3.0, "step": 1725, "total_flos": 6.649577559970284e+18, "train_loss": 0.3622851570447286, "train_runtime": 4827.2765, "train_samples_per_second": 3.57, "train_steps_per_second": 0.357 } ], "logging_steps": 1, "max_steps": 1725, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 500, "total_flos": 6.649577559970284e+18, "train_batch_size": 10, "trial_name": null, "trial_params": null }