| {"current_steps": 1, "total_steps": 190, "loss": 14.1364, "learning_rate": 5.000000000000001e-07, "epoch": 0.02572347266881029, "percentage": 0.53, "elapsed_time": "0:00:13", "remaining_time": "0:43:08", "throughput": "475.41", "total_tokens": 6512} | |
| {"current_steps": 2, "total_steps": 190, "loss": 13.7804, "learning_rate": 1.0000000000000002e-06, "epoch": 0.05144694533762058, "percentage": 1.05, "elapsed_time": "0:00:26", "remaining_time": "0:42:08", "throughput": "477.72", "total_tokens": 12848} | |
| {"current_steps": 3, "total_steps": 190, "loss": 13.4871, "learning_rate": 1.5e-06, "epoch": 0.07717041800643087, "percentage": 1.58, "elapsed_time": "0:00:40", "remaining_time": "0:41:38", "throughput": "480.28", "total_tokens": 19248} | |
| {"current_steps": 4, "total_steps": 190, "loss": 12.79, "learning_rate": 2.0000000000000003e-06, "epoch": 0.10289389067524116, "percentage": 2.11, "elapsed_time": "0:00:53", "remaining_time": "0:41:16", "throughput": "475.82", "total_tokens": 25344} | |
| {"current_steps": 5, "total_steps": 190, "loss": 9.2748, "learning_rate": 2.5e-06, "epoch": 0.12861736334405144, "percentage": 2.63, "elapsed_time": "0:01:06", "remaining_time": "0:40:57", "throughput": "481.26", "total_tokens": 31968} | |
| {"current_steps": 6, "total_steps": 190, "loss": 6.5585, "learning_rate": 3e-06, "epoch": 0.15434083601286175, "percentage": 3.16, "elapsed_time": "0:01:19", "remaining_time": "0:40:40", "throughput": "479.46", "total_tokens": 38160} | |
| {"current_steps": 7, "total_steps": 190, "loss": 5.3984, "learning_rate": 3.5e-06, "epoch": 0.18006430868167203, "percentage": 3.68, "elapsed_time": "0:01:32", "remaining_time": "0:40:24", "throughput": "475.92", "total_tokens": 44144} | |
| {"current_steps": 8, "total_steps": 190, "loss": 1.9363, "learning_rate": 4.000000000000001e-06, "epoch": 0.2057877813504823, "percentage": 4.21, "elapsed_time": "0:01:45", "remaining_time": "0:40:10", "throughput": "476.05", "total_tokens": 50432} | |
| {"current_steps": 9, "total_steps": 190, "loss": 0.6783, "learning_rate": 4.5e-06, "epoch": 0.2315112540192926, "percentage": 4.74, "elapsed_time": "0:01:59", "remaining_time": "0:39:55", "throughput": "477.93", "total_tokens": 56928} | |
| {"current_steps": 10, "total_steps": 190, "loss": 2.9945, "learning_rate": 5e-06, "epoch": 0.2572347266881029, "percentage": 5.26, "elapsed_time": "0:02:12", "remaining_time": "0:39:40", "throughput": "478.89", "total_tokens": 63344} | |
| {"current_steps": 11, "total_steps": 190, "loss": 0.2916, "learning_rate": 4.9996192378909785e-06, "epoch": 0.2829581993569132, "percentage": 5.79, "elapsed_time": "0:02:25", "remaining_time": "0:39:26", "throughput": "478.35", "total_tokens": 69568} | |
| {"current_steps": 12, "total_steps": 190, "loss": 2.2775, "learning_rate": 4.99847706754774e-06, "epoch": 0.3086816720257235, "percentage": 6.32, "elapsed_time": "0:02:38", "remaining_time": "0:39:12", "throughput": "478.05", "total_tokens": 75824} | |
| {"current_steps": 13, "total_steps": 190, "loss": 0.3757, "learning_rate": 4.9965738368864345e-06, "epoch": 0.33440514469453375, "percentage": 6.84, "elapsed_time": "0:02:51", "remaining_time": "0:38:58", "throughput": "478.29", "total_tokens": 82160} | |
| {"current_steps": 14, "total_steps": 190, "loss": 1.9543, "learning_rate": 4.993910125649561e-06, "epoch": 0.36012861736334406, "percentage": 7.37, "elapsed_time": "0:03:04", "remaining_time": "0:38:44", "throughput": "479.20", "total_tokens": 88624} | |
| {"current_steps": 15, "total_steps": 190, "loss": 0.7398, "learning_rate": 4.990486745229364e-06, "epoch": 0.3858520900321543, "percentage": 7.89, "elapsed_time": "0:03:18", "remaining_time": "0:38:31", "throughput": "478.49", "total_tokens": 94800} | |
| {"current_steps": 16, "total_steps": 190, "loss": 1.1868, "learning_rate": 4.986304738420684e-06, "epoch": 0.4115755627009646, "percentage": 8.42, "elapsed_time": "0:03:31", "remaining_time": "0:38:17", "throughput": "479.67", "total_tokens": 101360} | |
| {"current_steps": 17, "total_steps": 190, "loss": 0.5418, "learning_rate": 4.981365379103306e-06, "epoch": 0.43729903536977494, "percentage": 8.95, "elapsed_time": "0:03:44", "remaining_time": "0:38:04", "throughput": "478.83", "total_tokens": 107488} | |
| {"current_steps": 18, "total_steps": 190, "loss": 0.2263, "learning_rate": 4.975670171853926e-06, "epoch": 0.4630225080385852, "percentage": 9.47, "elapsed_time": "0:03:57", "remaining_time": "0:37:50", "throughput": "479.37", "total_tokens": 113920} | |
| {"current_steps": 19, "total_steps": 190, "loss": 0.1612, "learning_rate": 4.9692208514878445e-06, "epoch": 0.4887459807073955, "percentage": 10.0, "elapsed_time": "0:04:10", "remaining_time": "0:37:37", "throughput": "479.86", "total_tokens": 120352} | |
| {"current_steps": 20, "total_steps": 190, "loss": 0.3299, "learning_rate": 4.962019382530521e-06, "epoch": 0.5144694533762058, "percentage": 10.53, "elapsed_time": "0:04:23", "remaining_time": "0:37:23", "throughput": "480.64", "total_tokens": 126880} | |
| {"current_steps": 21, "total_steps": 190, "loss": 0.2013, "learning_rate": 4.9540679586191605e-06, "epoch": 0.5401929260450161, "percentage": 11.05, "elapsed_time": "0:04:37", "remaining_time": "0:37:10", "throughput": "481.12", "total_tokens": 133344} | |
| {"current_steps": 22, "total_steps": 190, "loss": 0.2446, "learning_rate": 4.9453690018345144e-06, "epoch": 0.5659163987138264, "percentage": 11.58, "elapsed_time": "0:04:50", "remaining_time": "0:36:56", "throughput": "481.36", "total_tokens": 139744} | |
| {"current_steps": 23, "total_steps": 190, "loss": 0.2235, "learning_rate": 4.935925161963089e-06, "epoch": 0.5916398713826366, "percentage": 12.11, "elapsed_time": "0:05:03", "remaining_time": "0:36:43", "throughput": "481.55", "total_tokens": 146144} | |
| {"current_steps": 24, "total_steps": 190, "loss": 0.116, "learning_rate": 4.925739315689991e-06, "epoch": 0.617363344051447, "percentage": 12.63, "elapsed_time": "0:05:16", "remaining_time": "0:36:30", "throughput": "480.78", "total_tokens": 152240} | |
| {"current_steps": 25, "total_steps": 190, "loss": 0.2179, "learning_rate": 4.914814565722671e-06, "epoch": 0.6430868167202572, "percentage": 13.16, "elapsed_time": "0:05:29", "remaining_time": "0:36:16", "throughput": "480.61", "total_tokens": 158512} | |
| {"current_steps": 26, "total_steps": 190, "loss": 0.1414, "learning_rate": 4.903154239845798e-06, "epoch": 0.6688102893890675, "percentage": 13.68, "elapsed_time": "0:05:42", "remaining_time": "0:36:03", "throughput": "480.45", "total_tokens": 164784} | |
| {"current_steps": 27, "total_steps": 190, "loss": 0.1181, "learning_rate": 4.890761889907589e-06, "epoch": 0.6945337620578779, "percentage": 14.21, "elapsed_time": "0:05:56", "remaining_time": "0:35:49", "throughput": "481.03", "total_tokens": 171312} | |
| {"current_steps": 28, "total_steps": 190, "loss": 0.2753, "learning_rate": 4.8776412907378845e-06, "epoch": 0.7202572347266881, "percentage": 14.74, "elapsed_time": "0:06:09", "remaining_time": "0:35:36", "throughput": "481.61", "total_tokens": 177856} | |
| {"current_steps": 29, "total_steps": 190, "loss": 0.3255, "learning_rate": 4.863796438998293e-06, "epoch": 0.7459807073954984, "percentage": 15.26, "elapsed_time": "0:06:22", "remaining_time": "0:35:23", "throughput": "482.08", "total_tokens": 184368} | |
| {"current_steps": 30, "total_steps": 190, "loss": 0.2352, "learning_rate": 4.849231551964771e-06, "epoch": 0.7717041800643086, "percentage": 15.79, "elapsed_time": "0:06:35", "remaining_time": "0:35:09", "throughput": "482.60", "total_tokens": 190928} | |
| {"current_steps": 31, "total_steps": 190, "loss": 0.063, "learning_rate": 4.833951066243004e-06, "epoch": 0.797427652733119, "percentage": 16.32, "elapsed_time": "0:06:48", "remaining_time": "0:34:56", "throughput": "482.48", "total_tokens": 197232} | |
| {"current_steps": 32, "total_steps": 190, "loss": 0.2042, "learning_rate": 4.817959636416969e-06, "epoch": 0.8231511254019293, "percentage": 16.84, "elapsed_time": "0:07:01", "remaining_time": "0:34:43", "throughput": "482.50", "total_tokens": 203584} | |
| {"current_steps": 33, "total_steps": 190, "loss": 0.1364, "learning_rate": 4.801262133631101e-06, "epoch": 0.8488745980707395, "percentage": 17.37, "elapsed_time": "0:07:15", "remaining_time": "0:34:30", "throughput": "482.93", "total_tokens": 210128} | |
| {"current_steps": 34, "total_steps": 190, "loss": 0.0934, "learning_rate": 4.783863644106502e-06, "epoch": 0.8745980707395499, "percentage": 17.89, "elapsed_time": "0:07:28", "remaining_time": "0:34:16", "throughput": "482.98", "total_tokens": 216512} | |
| {"current_steps": 35, "total_steps": 190, "loss": 0.1332, "learning_rate": 4.765769467591626e-06, "epoch": 0.9003215434083601, "percentage": 18.42, "elapsed_time": "0:07:41", "remaining_time": "0:34:03", "throughput": "483.11", "total_tokens": 222928} | |
| {"current_steps": 36, "total_steps": 190, "loss": 0.1595, "learning_rate": 4.746985115747918e-06, "epoch": 0.9260450160771704, "percentage": 18.95, "elapsed_time": "0:07:54", "remaining_time": "0:33:50", "throughput": "483.00", "total_tokens": 229232} | |
| {"current_steps": 37, "total_steps": 190, "loss": 0.1528, "learning_rate": 4.72751631047092e-06, "epoch": 0.9517684887459807, "percentage": 19.47, "elapsed_time": "0:08:07", "remaining_time": "0:33:37", "throughput": "483.18", "total_tokens": 235680} | |
| {"current_steps": 38, "total_steps": 190, "loss": 0.1342, "learning_rate": 4.707368982147318e-06, "epoch": 0.977491961414791, "percentage": 20.0, "elapsed_time": "0:08:20", "remaining_time": "0:33:23", "throughput": "483.48", "total_tokens": 242192} | |
| {"current_steps": 39, "total_steps": 190, "loss": 0.1586, "learning_rate": 4.68654926784849e-06, "epoch": 1.0032154340836013, "percentage": 20.53, "elapsed_time": "0:08:34", "remaining_time": "0:33:10", "throughput": "483.71", "total_tokens": 248672} | |
| {"current_steps": 40, "total_steps": 190, "loss": 0.1072, "learning_rate": 4.665063509461098e-06, "epoch": 1.0289389067524115, "percentage": 21.05, "elapsed_time": "0:08:47", "remaining_time": "0:32:57", "throughput": "483.77", "total_tokens": 255072} | |
| {"current_steps": 41, "total_steps": 190, "loss": 0.0357, "learning_rate": 4.642918251755281e-06, "epoch": 1.0546623794212218, "percentage": 21.58, "elapsed_time": "0:09:00", "remaining_time": "0:32:43", "throughput": "484.04", "total_tokens": 261584} | |
| {"current_steps": 42, "total_steps": 190, "loss": 0.06, "learning_rate": 4.620120240391065e-06, "epoch": 1.0803858520900322, "percentage": 22.11, "elapsed_time": "0:09:13", "remaining_time": "0:32:30", "throughput": "484.18", "total_tokens": 268032} | |
| {"current_steps": 43, "total_steps": 190, "loss": 0.0902, "learning_rate": 4.596676419863561e-06, "epoch": 1.1061093247588425, "percentage": 22.63, "elapsed_time": "0:09:26", "remaining_time": "0:32:17", "throughput": "484.46", "total_tokens": 274560} | |
| {"current_steps": 44, "total_steps": 190, "loss": 0.0202, "learning_rate": 4.572593931387604e-06, "epoch": 1.1318327974276527, "percentage": 23.16, "elapsed_time": "0:09:39", "remaining_time": "0:32:04", "throughput": "484.51", "total_tokens": 280960} | |
| {"current_steps": 45, "total_steps": 190, "loss": 0.038, "learning_rate": 4.54788011072248e-06, "epoch": 1.157556270096463, "percentage": 23.68, "elapsed_time": "0:09:53", "remaining_time": "0:31:51", "throughput": "484.10", "total_tokens": 287104} | |
| {"current_steps": 46, "total_steps": 190, "loss": 0.0379, "learning_rate": 4.522542485937369e-06, "epoch": 1.1832797427652733, "percentage": 24.21, "elapsed_time": "0:10:06", "remaining_time": "0:31:37", "throughput": "484.17", "total_tokens": 293520} | |
| {"current_steps": 47, "total_steps": 190, "loss": 0.0742, "learning_rate": 4.496588775118232e-06, "epoch": 1.2090032154340835, "percentage": 24.74, "elapsed_time": "0:10:19", "remaining_time": "0:31:24", "throughput": "484.24", "total_tokens": 299936} | |
| {"current_steps": 48, "total_steps": 190, "loss": 0.0658, "learning_rate": 4.470026884016805e-06, "epoch": 1.234726688102894, "percentage": 25.26, "elapsed_time": "0:10:32", "remaining_time": "0:31:11", "throughput": "483.64", "total_tokens": 305936} | |
| {"current_steps": 49, "total_steps": 190, "loss": 0.0336, "learning_rate": 4.442864903642428e-06, "epoch": 1.2604501607717042, "percentage": 25.79, "elapsed_time": "0:10:45", "remaining_time": "0:30:58", "throughput": "483.99", "total_tokens": 312528} | |
| {"current_steps": 50, "total_steps": 190, "loss": 0.1021, "learning_rate": 4.415111107797445e-06, "epoch": 1.2861736334405145, "percentage": 26.32, "elapsed_time": "0:10:58", "remaining_time": "0:30:44", "throughput": "483.77", "total_tokens": 318752} | |
| {"current_steps": 51, "total_steps": 190, "loss": 0.1312, "learning_rate": 4.386773950556931e-06, "epoch": 1.3118971061093248, "percentage": 26.84, "elapsed_time": "0:11:12", "remaining_time": "0:30:31", "throughput": "483.74", "total_tokens": 325088} | |
| {"current_steps": 52, "total_steps": 190, "loss": 0.0665, "learning_rate": 4.357862063693486e-06, "epoch": 1.337620578778135, "percentage": 27.37, "elapsed_time": "0:11:25", "remaining_time": "0:30:18", "throughput": "483.68", "total_tokens": 331424} | |
| {"current_steps": 53, "total_steps": 190, "loss": 0.0679, "learning_rate": 4.328384254047927e-06, "epoch": 1.3633440514469453, "percentage": 27.89, "elapsed_time": "0:11:38", "remaining_time": "0:30:05", "throughput": "483.66", "total_tokens": 337776} | |
| {"current_steps": 54, "total_steps": 190, "loss": 0.0579, "learning_rate": 4.2983495008466285e-06, "epoch": 1.3890675241157555, "percentage": 28.42, "elapsed_time": "0:11:51", "remaining_time": "0:29:52", "throughput": "483.46", "total_tokens": 344000} | |
| {"current_steps": 55, "total_steps": 190, "loss": 0.0542, "learning_rate": 4.267766952966369e-06, "epoch": 1.414790996784566, "percentage": 28.95, "elapsed_time": "0:12:04", "remaining_time": "0:29:38", "throughput": "483.69", "total_tokens": 350528} | |
| {"current_steps": 56, "total_steps": 190, "loss": 0.0476, "learning_rate": 4.236645926147493e-06, "epoch": 1.4405144694533762, "percentage": 29.47, "elapsed_time": "0:12:17", "remaining_time": "0:29:25", "throughput": "483.69", "total_tokens": 356896} | |
| {"current_steps": 57, "total_steps": 190, "loss": 0.0613, "learning_rate": 4.204995900156247e-06, "epoch": 1.4662379421221865, "percentage": 30.0, "elapsed_time": "0:12:31", "remaining_time": "0:29:12", "throughput": "483.84", "total_tokens": 363376} | |
| {"current_steps": 58, "total_steps": 190, "loss": 0.0995, "learning_rate": 4.172826515897146e-06, "epoch": 1.4919614147909968, "percentage": 30.53, "elapsed_time": "0:12:44", "remaining_time": "0:28:59", "throughput": "483.76", "total_tokens": 369680} | |
| {"current_steps": 59, "total_steps": 190, "loss": 0.0532, "learning_rate": 4.140147572476269e-06, "epoch": 1.517684887459807, "percentage": 31.05, "elapsed_time": "0:12:57", "remaining_time": "0:28:45", "throughput": "483.57", "total_tokens": 375904} | |
| {"current_steps": 60, "total_steps": 190, "loss": 0.0824, "learning_rate": 4.106969024216348e-06, "epoch": 1.5434083601286175, "percentage": 31.58, "elapsed_time": "0:13:10", "remaining_time": "0:28:32", "throughput": "483.60", "total_tokens": 382304} | |
| {"current_steps": 61, "total_steps": 190, "loss": 0.0499, "learning_rate": 4.073300977624594e-06, "epoch": 1.5691318327974275, "percentage": 32.11, "elapsed_time": "0:13:23", "remaining_time": "0:28:19", "throughput": "483.63", "total_tokens": 388688} | |
| {"current_steps": 62, "total_steps": 190, "loss": 0.0413, "learning_rate": 4.039153688314146e-06, "epoch": 1.594855305466238, "percentage": 32.63, "elapsed_time": "0:13:36", "remaining_time": "0:28:06", "throughput": "483.75", "total_tokens": 395152} | |
| {"current_steps": 63, "total_steps": 190, "loss": 0.0637, "learning_rate": 4.0045375578801216e-06, "epoch": 1.6205787781350482, "percentage": 33.16, "elapsed_time": "0:13:50", "remaining_time": "0:27:53", "throughput": "484.01", "total_tokens": 401728} | |
| {"current_steps": 64, "total_steps": 190, "loss": 0.0529, "learning_rate": 3.969463130731183e-06, "epoch": 1.6463022508038585, "percentage": 33.68, "elapsed_time": "0:14:03", "remaining_time": "0:27:40", "throughput": "483.77", "total_tokens": 407904} | |
| {"current_steps": 65, "total_steps": 190, "loss": 0.0474, "learning_rate": 3.933941090877615e-06, "epoch": 1.6720257234726688, "percentage": 34.21, "elapsed_time": "0:14:16", "remaining_time": "0:27:26", "throughput": "483.73", "total_tokens": 414240} | |
| {"current_steps": 66, "total_steps": 190, "loss": 0.0649, "learning_rate": 3.897982258676867e-06, "epoch": 1.697749196141479, "percentage": 34.74, "elapsed_time": "0:14:29", "remaining_time": "0:27:13", "throughput": "483.55", "total_tokens": 420448} | |
| {"current_steps": 67, "total_steps": 190, "loss": 0.0505, "learning_rate": 3.861597587537568e-06, "epoch": 1.7234726688102895, "percentage": 35.26, "elapsed_time": "0:14:42", "remaining_time": "0:27:00", "throughput": "483.51", "total_tokens": 426784} | |
| {"current_steps": 68, "total_steps": 190, "loss": 0.0621, "learning_rate": 3.824798160583012e-06, "epoch": 1.7491961414790995, "percentage": 35.79, "elapsed_time": "0:14:55", "remaining_time": "0:26:47", "throughput": "483.14", "total_tokens": 432816} | |
| {"current_steps": 69, "total_steps": 190, "loss": 0.0769, "learning_rate": 3.787595187275136e-06, "epoch": 1.77491961414791, "percentage": 36.32, "elapsed_time": "0:15:09", "remaining_time": "0:26:34", "throughput": "483.20", "total_tokens": 439232} | |
| {"current_steps": 70, "total_steps": 190, "loss": 0.0435, "learning_rate": 3.7500000000000005e-06, "epoch": 1.8006430868167203, "percentage": 36.84, "elapsed_time": "0:15:22", "remaining_time": "0:26:20", "throughput": "483.42", "total_tokens": 445792} | |
| {"current_steps": 71, "total_steps": 190, "loss": 0.0673, "learning_rate": 3.7120240506158433e-06, "epoch": 1.8263665594855305, "percentage": 37.37, "elapsed_time": "0:15:35", "remaining_time": "0:26:07", "throughput": "483.69", "total_tokens": 452400} | |
| {"current_steps": 72, "total_steps": 190, "loss": 0.1316, "learning_rate": 3.6736789069647273e-06, "epoch": 1.852090032154341, "percentage": 37.89, "elapsed_time": "0:15:48", "remaining_time": "0:25:54", "throughput": "483.44", "total_tokens": 458528} | |
| {"current_steps": 73, "total_steps": 190, "loss": 0.0531, "learning_rate": 3.634976249348867e-06, "epoch": 1.877813504823151, "percentage": 38.42, "elapsed_time": "0:16:01", "remaining_time": "0:25:41", "throughput": "483.53", "total_tokens": 464976} | |
| {"current_steps": 74, "total_steps": 190, "loss": 0.0287, "learning_rate": 3.595927866972694e-06, "epoch": 1.9035369774919615, "percentage": 38.95, "elapsed_time": "0:16:14", "remaining_time": "0:25:28", "throughput": "483.62", "total_tokens": 471440} | |
| {"current_steps": 75, "total_steps": 190, "loss": 0.0648, "learning_rate": 3.556545654351749e-06, "epoch": 1.9292604501607717, "percentage": 39.47, "elapsed_time": "0:16:27", "remaining_time": "0:25:14", "throughput": "483.59", "total_tokens": 477776} | |
| {"current_steps": 76, "total_steps": 190, "loss": 0.1211, "learning_rate": 3.516841607689501e-06, "epoch": 1.954983922829582, "percentage": 40.0, "elapsed_time": "0:16:41", "remaining_time": "0:25:01", "throughput": "483.54", "total_tokens": 484096} | |
| {"current_steps": 77, "total_steps": 190, "loss": 0.0879, "learning_rate": 3.476827821223184e-06, "epoch": 1.9807073954983923, "percentage": 40.53, "elapsed_time": "0:16:54", "remaining_time": "0:24:48", "throughput": "483.26", "total_tokens": 490176} | |
| {"current_steps": 78, "total_steps": 190, "loss": 0.0227, "learning_rate": 3.436516483539781e-06, "epoch": 2.0064308681672025, "percentage": 41.05, "elapsed_time": "0:17:07", "remaining_time": "0:24:35", "throughput": "483.39", "total_tokens": 496672} | |
| {"current_steps": 79, "total_steps": 190, "loss": 0.0228, "learning_rate": 3.39591987386325e-06, "epoch": 2.032154340836013, "percentage": 41.58, "elapsed_time": "0:17:20", "remaining_time": "0:24:22", "throughput": "483.45", "total_tokens": 503088} | |
| {"current_steps": 80, "total_steps": 190, "loss": 0.036, "learning_rate": 3.3550503583141726e-06, "epoch": 2.057877813504823, "percentage": 42.11, "elapsed_time": "0:17:33", "remaining_time": "0:24:08", "throughput": "483.47", "total_tokens": 509472} | |
| {"current_steps": 81, "total_steps": 190, "loss": 0.0138, "learning_rate": 3.313920386142892e-06, "epoch": 2.0836012861736335, "percentage": 42.63, "elapsed_time": "0:17:46", "remaining_time": "0:23:55", "throughput": "483.36", "total_tokens": 515728} | |
| {"current_steps": 82, "total_steps": 190, "loss": 0.0697, "learning_rate": 3.272542485937369e-06, "epoch": 2.1093247588424435, "percentage": 43.16, "elapsed_time": "0:18:00", "remaining_time": "0:23:42", "throughput": "483.18", "total_tokens": 521904} | |
| {"current_steps": 83, "total_steps": 190, "loss": 0.0508, "learning_rate": 3.230929261806842e-06, "epoch": 2.135048231511254, "percentage": 43.68, "elapsed_time": "0:18:13", "remaining_time": "0:23:29", "throughput": "482.89", "total_tokens": 527952} | |
| {"current_steps": 84, "total_steps": 190, "loss": 0.0088, "learning_rate": 3.189093389542498e-06, "epoch": 2.1607717041800645, "percentage": 44.21, "elapsed_time": "0:18:26", "remaining_time": "0:23:16", "throughput": "483.18", "total_tokens": 534624} | |
| {"current_steps": 85, "total_steps": 190, "loss": 0.0158, "learning_rate": 3.147047612756302e-06, "epoch": 2.1864951768488745, "percentage": 44.74, "elapsed_time": "0:18:39", "remaining_time": "0:23:03", "throughput": "483.34", "total_tokens": 541168} | |
| {"current_steps": 86, "total_steps": 190, "loss": 0.006, "learning_rate": 3.1048047389991693e-06, "epoch": 2.212218649517685, "percentage": 45.26, "elapsed_time": "0:18:52", "remaining_time": "0:22:49", "throughput": "483.30", "total_tokens": 547488} | |
| {"current_steps": 87, "total_steps": 190, "loss": 0.038, "learning_rate": 3.062377635859663e-06, "epoch": 2.237942122186495, "percentage": 45.79, "elapsed_time": "0:19:05", "remaining_time": "0:22:36", "throughput": "483.67", "total_tokens": 554272} | |
| {"current_steps": 88, "total_steps": 190, "loss": 0.0004, "learning_rate": 3.019779227044398e-06, "epoch": 2.2636655948553055, "percentage": 46.32, "elapsed_time": "0:19:19", "remaining_time": "0:22:23", "throughput": "483.58", "total_tokens": 560528} | |
| {"current_steps": 89, "total_steps": 190, "loss": 0.0111, "learning_rate": 2.9770224884413625e-06, "epoch": 2.289389067524116, "percentage": 46.84, "elapsed_time": "0:19:32", "remaining_time": "0:22:10", "throughput": "483.47", "total_tokens": 566784} | |
| {"current_steps": 90, "total_steps": 190, "loss": 0.0008, "learning_rate": 2.9341204441673267e-06, "epoch": 2.315112540192926, "percentage": 47.37, "elapsed_time": "0:19:45", "remaining_time": "0:21:57", "throughput": "483.64", "total_tokens": 573344} | |
| {"current_steps": 91, "total_steps": 190, "loss": 0.0182, "learning_rate": 2.8910861626005774e-06, "epoch": 2.3408360128617365, "percentage": 47.89, "elapsed_time": "0:19:58", "remaining_time": "0:21:43", "throughput": "483.70", "total_tokens": 579776} | |
| {"current_steps": 92, "total_steps": 190, "loss": 0.0491, "learning_rate": 2.847932752400164e-06, "epoch": 2.3665594855305465, "percentage": 48.42, "elapsed_time": "0:20:11", "remaining_time": "0:21:30", "throughput": "483.66", "total_tokens": 586096} | |
| {"current_steps": 93, "total_steps": 190, "loss": 0.004, "learning_rate": 2.804673358512869e-06, "epoch": 2.392282958199357, "percentage": 48.95, "elapsed_time": "0:20:24", "remaining_time": "0:21:17", "throughput": "483.71", "total_tokens": 592528} | |
| {"current_steps": 94, "total_steps": 190, "loss": 0.0176, "learning_rate": 2.761321158169134e-06, "epoch": 2.418006430868167, "percentage": 49.47, "elapsed_time": "0:20:38", "remaining_time": "0:21:04", "throughput": "483.76", "total_tokens": 598960} | |
| {"current_steps": 95, "total_steps": 190, "loss": 0.019, "learning_rate": 2.717889356869146e-06, "epoch": 2.4437299035369775, "percentage": 50.0, "elapsed_time": "0:20:51", "remaining_time": "0:20:51", "throughput": "483.69", "total_tokens": 605232} | |
| {"current_steps": 96, "total_steps": 190, "loss": 0.027, "learning_rate": 2.6743911843603134e-06, "epoch": 2.469453376205788, "percentage": 50.53, "elapsed_time": "0:21:04", "remaining_time": "0:20:38", "throughput": "483.49", "total_tokens": 611344} | |
| {"current_steps": 97, "total_steps": 190, "loss": 0.0354, "learning_rate": 2.6308398906073603e-06, "epoch": 2.495176848874598, "percentage": 51.05, "elapsed_time": "0:21:17", "remaining_time": "0:20:24", "throughput": "483.49", "total_tokens": 617712} | |
| {"current_steps": 98, "total_steps": 190, "loss": 0.0741, "learning_rate": 2.587248741756253e-06, "epoch": 2.5209003215434085, "percentage": 51.58, "elapsed_time": "0:21:30", "remaining_time": "0:20:11", "throughput": "483.59", "total_tokens": 624208} | |
| {"current_steps": 99, "total_steps": 190, "loss": 0.0582, "learning_rate": 2.543631016093209e-06, "epoch": 2.5466237942122185, "percentage": 52.11, "elapsed_time": "0:21:43", "remaining_time": "0:19:58", "throughput": "483.53", "total_tokens": 630496} | |
| {"current_steps": 100, "total_steps": 190, "loss": 0.0096, "learning_rate": 2.5e-06, "epoch": 2.572347266881029, "percentage": 52.63, "elapsed_time": "0:21:57", "remaining_time": "0:19:45", "throughput": "483.66", "total_tokens": 637040} | |
| {"current_steps": 101, "total_steps": 190, "loss": 0.0263, "learning_rate": 2.4563689839067913e-06, "epoch": 2.598070739549839, "percentage": 53.16, "elapsed_time": "0:22:10", "remaining_time": "0:19:32", "throughput": "483.71", "total_tokens": 643472} | |
| {"current_steps": 102, "total_steps": 190, "loss": 0.0121, "learning_rate": 2.4127512582437486e-06, "epoch": 2.6237942122186495, "percentage": 53.68, "elapsed_time": "0:22:23", "remaining_time": "0:19:19", "throughput": "483.65", "total_tokens": 649760} | |
| {"current_steps": 103, "total_steps": 190, "loss": 0.0204, "learning_rate": 2.3691601093926406e-06, "epoch": 2.64951768488746, "percentage": 54.21, "elapsed_time": "0:22:36", "remaining_time": "0:19:05", "throughput": "483.62", "total_tokens": 656096} | |
| {"current_steps": 104, "total_steps": 190, "loss": 0.0325, "learning_rate": 2.325608815639687e-06, "epoch": 2.67524115755627, "percentage": 54.74, "elapsed_time": "0:22:49", "remaining_time": "0:18:52", "throughput": "483.74", "total_tokens": 662624} | |
| {"current_steps": 105, "total_steps": 190, "loss": 0.0076, "learning_rate": 2.2821106431308546e-06, "epoch": 2.7009646302250805, "percentage": 55.26, "elapsed_time": "0:23:02", "remaining_time": "0:18:39", "throughput": "483.58", "total_tokens": 668768} | |
| {"current_steps": 106, "total_steps": 190, "loss": 0.0485, "learning_rate": 2.238678841830867e-06, "epoch": 2.7266881028938905, "percentage": 55.79, "elapsed_time": "0:23:16", "remaining_time": "0:18:26", "throughput": "483.48", "total_tokens": 674992} | |
| {"current_steps": 107, "total_steps": 190, "loss": 0.007, "learning_rate": 2.195326641487132e-06, "epoch": 2.752411575562701, "percentage": 56.32, "elapsed_time": "0:23:29", "remaining_time": "0:18:13", "throughput": "483.31", "total_tokens": 681120} | |
| {"current_steps": 108, "total_steps": 190, "loss": 0.0347, "learning_rate": 2.1520672475998374e-06, "epoch": 2.778135048231511, "percentage": 56.84, "elapsed_time": "0:23:42", "remaining_time": "0:18:00", "throughput": "483.23", "total_tokens": 687376} | |
| {"current_steps": 109, "total_steps": 190, "loss": 0.0142, "learning_rate": 2.1089138373994226e-06, "epoch": 2.8038585209003215, "percentage": 57.37, "elapsed_time": "0:23:55", "remaining_time": "0:17:46", "throughput": "483.41", "total_tokens": 693984} | |
| {"current_steps": 110, "total_steps": 190, "loss": 0.0414, "learning_rate": 2.0658795558326745e-06, "epoch": 2.829581993569132, "percentage": 57.89, "elapsed_time": "0:24:08", "remaining_time": "0:17:33", "throughput": "483.41", "total_tokens": 700352} | |
| {"current_steps": 111, "total_steps": 190, "loss": 0.0419, "learning_rate": 2.022977511558638e-06, "epoch": 2.855305466237942, "percentage": 58.42, "elapsed_time": "0:24:21", "remaining_time": "0:17:20", "throughput": "483.45", "total_tokens": 706768} | |
| {"current_steps": 112, "total_steps": 190, "loss": 0.043, "learning_rate": 1.9802207729556023e-06, "epoch": 2.8810289389067525, "percentage": 58.95, "elapsed_time": "0:24:35", "remaining_time": "0:17:07", "throughput": "483.52", "total_tokens": 713248} | |
| {"current_steps": 113, "total_steps": 190, "loss": 0.0192, "learning_rate": 1.937622364140338e-06, "epoch": 2.906752411575563, "percentage": 59.47, "elapsed_time": "0:24:48", "remaining_time": "0:16:54", "throughput": "483.49", "total_tokens": 719568} | |
| {"current_steps": 114, "total_steps": 190, "loss": 0.0427, "learning_rate": 1.895195261000831e-06, "epoch": 2.932475884244373, "percentage": 60.0, "elapsed_time": "0:25:01", "remaining_time": "0:16:40", "throughput": "483.53", "total_tokens": 725984} | |
| {"current_steps": 115, "total_steps": 190, "loss": 0.0116, "learning_rate": 1.852952387243698e-06, "epoch": 2.958199356913183, "percentage": 60.53, "elapsed_time": "0:25:14", "remaining_time": "0:16:27", "throughput": "483.44", "total_tokens": 732224} | |
| {"current_steps": 116, "total_steps": 190, "loss": 0.0135, "learning_rate": 1.8109066104575023e-06, "epoch": 2.9839228295819935, "percentage": 61.05, "elapsed_time": "0:25:27", "remaining_time": "0:16:14", "throughput": "483.38", "total_tokens": 738496} | |
| {"current_steps": 117, "total_steps": 190, "loss": 0.0128, "learning_rate": 1.7690707381931585e-06, "epoch": 3.009646302250804, "percentage": 61.58, "elapsed_time": "0:25:40", "remaining_time": "0:16:01", "throughput": "483.40", "total_tokens": 744880} | |
| {"current_steps": 118, "total_steps": 190, "loss": 0.0021, "learning_rate": 1.7274575140626318e-06, "epoch": 3.035369774919614, "percentage": 62.11, "elapsed_time": "0:25:54", "remaining_time": "0:15:48", "throughput": "483.50", "total_tokens": 751408} | |
| {"current_steps": 119, "total_steps": 190, "loss": 0.0057, "learning_rate": 1.686079613857109e-06, "epoch": 3.0610932475884245, "percentage": 62.63, "elapsed_time": "0:26:07", "remaining_time": "0:15:35", "throughput": "483.41", "total_tokens": 757632} | |
| {"current_steps": 120, "total_steps": 190, "loss": 0.0197, "learning_rate": 1.6449496416858285e-06, "epoch": 3.0868167202572345, "percentage": 63.16, "elapsed_time": "0:26:20", "remaining_time": "0:15:21", "throughput": "483.37", "total_tokens": 763936} | |
| {"current_steps": 121, "total_steps": 190, "loss": 0.0017, "learning_rate": 1.6040801261367494e-06, "epoch": 3.112540192926045, "percentage": 63.68, "elapsed_time": "0:26:33", "remaining_time": "0:15:08", "throughput": "483.22", "total_tokens": 770064} | |
| {"current_steps": 122, "total_steps": 190, "loss": 0.0068, "learning_rate": 1.56348351646022e-06, "epoch": 3.1382636655948555, "percentage": 64.21, "elapsed_time": "0:26:46", "remaining_time": "0:14:55", "throughput": "483.07", "total_tokens": 776176} | |
| {"current_steps": 123, "total_steps": 190, "loss": 0.0022, "learning_rate": 1.5231721787768162e-06, "epoch": 3.1639871382636655, "percentage": 64.74, "elapsed_time": "0:26:59", "remaining_time": "0:14:42", "throughput": "483.02", "total_tokens": 782464} | |
| {"current_steps": 124, "total_steps": 190, "loss": 0.0162, "learning_rate": 1.4831583923105e-06, "epoch": 3.189710610932476, "percentage": 65.26, "elapsed_time": "0:27:13", "remaining_time": "0:14:29", "throughput": "483.18", "total_tokens": 789072} | |
| {"current_steps": 125, "total_steps": 190, "loss": 0.0014, "learning_rate": 1.443454345648252e-06, "epoch": 3.215434083601286, "percentage": 65.79, "elapsed_time": "0:27:26", "remaining_time": "0:14:16", "throughput": "483.24", "total_tokens": 795536} | |
| {"current_steps": 126, "total_steps": 190, "loss": 0.0063, "learning_rate": 1.4040721330273063e-06, "epoch": 3.2411575562700965, "percentage": 66.32, "elapsed_time": "0:27:39", "remaining_time": "0:14:02", "throughput": "483.23", "total_tokens": 801888} | |
| {"current_steps": 127, "total_steps": 190, "loss": 0.0282, "learning_rate": 1.3650237506511333e-06, "epoch": 3.266881028938907, "percentage": 66.84, "elapsed_time": "0:27:52", "remaining_time": "0:13:49", "throughput": "483.34", "total_tokens": 808432} | |
| {"current_steps": 128, "total_steps": 190, "loss": 0.0003, "learning_rate": 1.3263210930352737e-06, "epoch": 3.292604501607717, "percentage": 67.37, "elapsed_time": "0:28:05", "remaining_time": "0:13:36", "throughput": "483.41", "total_tokens": 814896} | |
| {"current_steps": 129, "total_steps": 190, "loss": 0.0002, "learning_rate": 1.2879759493841577e-06, "epoch": 3.3183279742765275, "percentage": 67.89, "elapsed_time": "0:28:18", "remaining_time": "0:13:23", "throughput": "483.37", "total_tokens": 821200} | |
| {"current_steps": 130, "total_steps": 190, "loss": 0.0004, "learning_rate": 1.2500000000000007e-06, "epoch": 3.3440514469453375, "percentage": 68.42, "elapsed_time": "0:28:32", "remaining_time": "0:13:10", "throughput": "483.38", "total_tokens": 827584} | |
| {"current_steps": 131, "total_steps": 190, "loss": 0.0169, "learning_rate": 1.2124048127248644e-06, "epoch": 3.369774919614148, "percentage": 68.95, "elapsed_time": "0:28:45", "remaining_time": "0:12:57", "throughput": "483.44", "total_tokens": 834048} | |
| {"current_steps": 132, "total_steps": 190, "loss": 0.0127, "learning_rate": 1.1752018394169882e-06, "epoch": 3.395498392282958, "percentage": 69.47, "elapsed_time": "0:28:58", "remaining_time": "0:12:43", "throughput": "483.34", "total_tokens": 840240} | |
| {"current_steps": 133, "total_steps": 190, "loss": 0.0045, "learning_rate": 1.1384024124624324e-06, "epoch": 3.4212218649517685, "percentage": 70.0, "elapsed_time": "0:29:11", "remaining_time": "0:12:30", "throughput": "483.25", "total_tokens": 846448} | |
| {"current_steps": 134, "total_steps": 190, "loss": 0.0924, "learning_rate": 1.1020177413231334e-06, "epoch": 3.446945337620579, "percentage": 70.53, "elapsed_time": "0:29:24", "remaining_time": "0:12:17", "throughput": "483.31", "total_tokens": 852928} | |
| {"current_steps": 135, "total_steps": 190, "loss": 0.0067, "learning_rate": 1.0660589091223854e-06, "epoch": 3.472668810289389, "percentage": 71.05, "elapsed_time": "0:29:37", "remaining_time": "0:12:04", "throughput": "483.33", "total_tokens": 859312} | |
| {"current_steps": 136, "total_steps": 190, "loss": 0.003, "learning_rate": 1.0305368692688175e-06, "epoch": 3.4983922829581995, "percentage": 71.58, "elapsed_time": "0:29:51", "remaining_time": "0:11:51", "throughput": "483.19", "total_tokens": 865440} | |
| {"current_steps": 137, "total_steps": 190, "loss": 0.0164, "learning_rate": 9.95462442119879e-07, "epoch": 3.5241157556270095, "percentage": 72.11, "elapsed_time": "0:30:04", "remaining_time": "0:11:37", "throughput": "483.29", "total_tokens": 871968} | |
| {"current_steps": 138, "total_steps": 190, "loss": 0.0018, "learning_rate": 9.608463116858544e-07, "epoch": 3.54983922829582, "percentage": 72.63, "elapsed_time": "0:30:17", "remaining_time": "0:11:24", "throughput": "483.30", "total_tokens": 878352} | |
| {"current_steps": 139, "total_steps": 190, "loss": 0.0226, "learning_rate": 9.266990223754069e-07, "epoch": 3.57556270096463, "percentage": 73.16, "elapsed_time": "0:30:30", "remaining_time": "0:11:11", "throughput": "483.32", "total_tokens": 884736} | |
| {"current_steps": 140, "total_steps": 190, "loss": 0.0008, "learning_rate": 8.930309757836517e-07, "epoch": 3.6012861736334405, "percentage": 73.68, "elapsed_time": "0:30:43", "remaining_time": "0:10:58", "throughput": "483.26", "total_tokens": 891008} | |
| {"current_steps": 141, "total_steps": 190, "loss": 0.0004, "learning_rate": 8.598524275237321e-07, "epoch": 3.627009646302251, "percentage": 74.21, "elapsed_time": "0:30:56", "remaining_time": "0:10:45", "throughput": "483.13", "total_tokens": 897120} | |
| {"current_steps": 142, "total_steps": 190, "loss": 0.0008, "learning_rate": 8.271734841028553e-07, "epoch": 3.652733118971061, "percentage": 74.74, "elapsed_time": "0:31:10", "remaining_time": "0:10:32", "throughput": "483.16", "total_tokens": 903536} | |
| {"current_steps": 143, "total_steps": 190, "loss": 0.0256, "learning_rate": 7.950040998437541e-07, "epoch": 3.6784565916398715, "percentage": 75.26, "elapsed_time": "0:31:23", "remaining_time": "0:10:18", "throughput": "483.12", "total_tokens": 909824} | |
| {"current_steps": 144, "total_steps": 190, "loss": 0.0005, "learning_rate": 7.633540738525066e-07, "epoch": 3.7041800643086815, "percentage": 75.79, "elapsed_time": "0:31:36", "remaining_time": "0:10:05", "throughput": "483.08", "total_tokens": 916112} | |
| {"current_steps": 145, "total_steps": 190, "loss": 0.0045, "learning_rate": 7.322330470336314e-07, "epoch": 3.729903536977492, "percentage": 76.32, "elapsed_time": "0:31:49", "remaining_time": "0:09:52", "throughput": "483.15", "total_tokens": 922592} | |
| {"current_steps": 146, "total_steps": 190, "loss": 0.0005, "learning_rate": 7.016504991533727e-07, "epoch": 3.755627009646302, "percentage": 76.84, "elapsed_time": "0:32:02", "remaining_time": "0:09:39", "throughput": "482.98", "total_tokens": 928640} | |
| {"current_steps": 147, "total_steps": 190, "loss": 0.0069, "learning_rate": 6.716157459520739e-07, "epoch": 3.7813504823151125, "percentage": 77.37, "elapsed_time": "0:32:15", "remaining_time": "0:09:26", "throughput": "483.23", "total_tokens": 935488} | |
| {"current_steps": 148, "total_steps": 190, "loss": 0.015, "learning_rate": 6.421379363065142e-07, "epoch": 3.807073954983923, "percentage": 77.89, "elapsed_time": "0:32:29", "remaining_time": "0:09:13", "throughput": "483.29", "total_tokens": 941952} | |
| {"current_steps": 149, "total_steps": 190, "loss": 0.0012, "learning_rate": 6.1322604944307e-07, "epoch": 3.832797427652733, "percentage": 78.42, "elapsed_time": "0:32:42", "remaining_time": "0:08:59", "throughput": "483.32", "total_tokens": 948368} | |
| {"current_steps": 150, "total_steps": 190, "loss": 0.0095, "learning_rate": 5.848888922025553e-07, "epoch": 3.8585209003215435, "percentage": 78.95, "elapsed_time": "0:32:55", "remaining_time": "0:08:46", "throughput": "483.33", "total_tokens": 954752} | |
| {"current_steps": 151, "total_steps": 190, "loss": 0.0271, "learning_rate": 5.571350963575728e-07, "epoch": 3.884244372990354, "percentage": 79.47, "elapsed_time": "0:33:08", "remaining_time": "0:08:33", "throughput": "483.39", "total_tokens": 961248} | |
| {"current_steps": 152, "total_steps": 190, "loss": 0.0201, "learning_rate": 5.299731159831953e-07, "epoch": 3.909967845659164, "percentage": 80.0, "elapsed_time": "0:33:21", "remaining_time": "0:08:20", "throughput": "483.30", "total_tokens": 967424} | |
| {"current_steps": 153, "total_steps": 190, "loss": 0.012, "learning_rate": 5.034112248817685e-07, "epoch": 3.935691318327974, "percentage": 80.53, "elapsed_time": "0:33:34", "remaining_time": "0:08:07", "throughput": "483.25", "total_tokens": 973696} | |
| {"current_steps": 154, "total_steps": 190, "loss": 0.023, "learning_rate": 4.774575140626317e-07, "epoch": 3.9614147909967845, "percentage": 81.05, "elapsed_time": "0:33:48", "remaining_time": "0:07:54", "throughput": "483.29", "total_tokens": 980144} | |
| {"current_steps": 155, "total_steps": 190, "loss": 0.0156, "learning_rate": 4.5211988927752026e-07, "epoch": 3.987138263665595, "percentage": 81.58, "elapsed_time": "0:34:01", "remaining_time": "0:07:40", "throughput": "483.22", "total_tokens": 986352} | |
| {"current_steps": 156, "total_steps": 190, "loss": 0.0009, "learning_rate": 4.27406068612396e-07, "epoch": 4.012861736334405, "percentage": 82.11, "elapsed_time": "0:34:14", "remaining_time": "0:07:27", "throughput": "483.29", "total_tokens": 992880} | |
| {"current_steps": 157, "total_steps": 190, "loss": 0.0017, "learning_rate": 4.033235801364402e-07, "epoch": 4.038585209003215, "percentage": 82.63, "elapsed_time": "0:34:27", "remaining_time": "0:07:14", "throughput": "483.27", "total_tokens": 999200} | |
| {"current_steps": 158, "total_steps": 190, "loss": 0.0015, "learning_rate": 3.798797596089351e-07, "epoch": 4.064308681672026, "percentage": 83.16, "elapsed_time": "0:34:40", "remaining_time": "0:07:01", "throughput": "483.28", "total_tokens": 1005568} | |
| {"current_steps": 159, "total_steps": 190, "loss": 0.0035, "learning_rate": 3.5708174824471947e-07, "epoch": 4.090032154340836, "percentage": 83.68, "elapsed_time": "0:34:53", "remaining_time": "0:06:48", "throughput": "483.18", "total_tokens": 1011728} | |
| {"current_steps": 160, "total_steps": 190, "loss": 0.0016, "learning_rate": 3.3493649053890325e-07, "epoch": 4.115755627009646, "percentage": 84.21, "elapsed_time": "0:35:07", "remaining_time": "0:06:35", "throughput": "483.27", "total_tokens": 1018272} | |
| {"current_steps": 161, "total_steps": 190, "loss": 0.0028, "learning_rate": 3.134507321515107e-07, "epoch": 4.141479099678457, "percentage": 84.74, "elapsed_time": "0:35:20", "remaining_time": "0:06:21", "throughput": "483.30", "total_tokens": 1024688} | |
| {"current_steps": 162, "total_steps": 190, "loss": 0.0006, "learning_rate": 2.9263101785268253e-07, "epoch": 4.167202572347267, "percentage": 85.26, "elapsed_time": "0:35:33", "remaining_time": "0:06:08", "throughput": "483.34", "total_tokens": 1031152} | |
| {"current_steps": 163, "total_steps": 190, "loss": 0.0013, "learning_rate": 2.7248368952908055e-07, "epoch": 4.192926045016077, "percentage": 85.79, "elapsed_time": "0:35:46", "remaining_time": "0:05:55", "throughput": "483.39", "total_tokens": 1037632} | |
| {"current_steps": 164, "total_steps": 190, "loss": 0.0006, "learning_rate": 2.53014884252083e-07, "epoch": 4.218649517684887, "percentage": 86.32, "elapsed_time": "0:35:59", "remaining_time": "0:05:42", "throughput": "483.37", "total_tokens": 1043936} | |
| {"current_steps": 165, "total_steps": 190, "loss": 0.0017, "learning_rate": 2.3423053240837518e-07, "epoch": 4.244372990353698, "percentage": 86.84, "elapsed_time": "0:36:12", "remaining_time": "0:05:29", "throughput": "483.25", "total_tokens": 1050048} | |
| {"current_steps": 166, "total_steps": 190, "loss": 0.0004, "learning_rate": 2.1613635589349756e-07, "epoch": 4.270096463022508, "percentage": 87.37, "elapsed_time": "0:36:26", "remaining_time": "0:05:16", "throughput": "483.29", "total_tokens": 1056496} | |
| {"current_steps": 167, "total_steps": 190, "loss": 0.0049, "learning_rate": 1.9873786636889908e-07, "epoch": 4.295819935691318, "percentage": 87.89, "elapsed_time": "0:36:39", "remaining_time": "0:05:02", "throughput": "483.28", "total_tokens": 1062848} | |
| {"current_steps": 168, "total_steps": 190, "loss": 0.0071, "learning_rate": 1.8204036358303173e-07, "epoch": 4.321543408360129, "percentage": 88.42, "elapsed_time": "0:36:52", "remaining_time": "0:04:49", "throughput": "483.25", "total_tokens": 1069136} | |
| {"current_steps": 169, "total_steps": 190, "loss": 0.0011, "learning_rate": 1.6604893375699594e-07, "epoch": 4.347266881028939, "percentage": 88.95, "elapsed_time": "0:37:05", "remaining_time": "0:04:36", "throughput": "483.17", "total_tokens": 1075328} | |
| {"current_steps": 170, "total_steps": 190, "loss": 0.0004, "learning_rate": 1.507684480352292e-07, "epoch": 4.372990353697749, "percentage": 89.47, "elapsed_time": "0:37:18", "remaining_time": "0:04:23", "throughput": "483.17", "total_tokens": 1081696} | |
| {"current_steps": 171, "total_steps": 190, "loss": 0.0007, "learning_rate": 1.362035610017079e-07, "epoch": 4.39871382636656, "percentage": 90.0, "elapsed_time": "0:37:31", "remaining_time": "0:04:10", "throughput": "483.20", "total_tokens": 1088112} | |
| {"current_steps": 172, "total_steps": 190, "loss": 0.0017, "learning_rate": 1.223587092621162e-07, "epoch": 4.42443729903537, "percentage": 90.53, "elapsed_time": "0:37:45", "remaining_time": "0:03:57", "throughput": "483.21", "total_tokens": 1094512} | |
| {"current_steps": 173, "total_steps": 190, "loss": 0.0007, "learning_rate": 1.0923811009241142e-07, "epoch": 4.45016077170418, "percentage": 91.05, "elapsed_time": "0:37:58", "remaining_time": "0:03:43", "throughput": "483.29", "total_tokens": 1101040} | |
| {"current_steps": 174, "total_steps": 190, "loss": 0.0003, "learning_rate": 9.684576015420277e-08, "epoch": 4.47588424437299, "percentage": 91.58, "elapsed_time": "0:38:11", "remaining_time": "0:03:30", "throughput": "483.18", "total_tokens": 1107168} | |
| {"current_steps": 175, "total_steps": 190, "loss": 0.0046, "learning_rate": 8.518543427732951e-08, "epoch": 4.501607717041801, "percentage": 92.11, "elapsed_time": "0:38:24", "remaining_time": "0:03:17", "throughput": "483.13", "total_tokens": 1113408} | |
| {"current_steps": 176, "total_steps": 190, "loss": 0.0038, "learning_rate": 7.426068431000883e-08, "epoch": 4.527331189710611, "percentage": 92.63, "elapsed_time": "0:38:37", "remaining_time": "0:03:04", "throughput": "483.04", "total_tokens": 1119568} | |
| {"current_steps": 177, "total_steps": 190, "loss": 0.0036, "learning_rate": 6.407483803691216e-08, "epoch": 4.553054662379421, "percentage": 93.16, "elapsed_time": "0:38:50", "remaining_time": "0:02:51", "throughput": "483.09", "total_tokens": 1126048} | |
| {"current_steps": 178, "total_steps": 190, "loss": 0.0056, "learning_rate": 5.463099816548578e-08, "epoch": 4.578778135048232, "percentage": 93.68, "elapsed_time": "0:39:04", "remaining_time": "0:02:38", "throughput": "483.09", "total_tokens": 1132400} | |
| {"current_steps": 179, "total_steps": 190, "loss": 0.0057, "learning_rate": 4.593204138084006e-08, "epoch": 4.604501607717042, "percentage": 94.21, "elapsed_time": "0:39:17", "remaining_time": "0:02:24", "throughput": "483.12", "total_tokens": 1138832} | |
| {"current_steps": 180, "total_steps": 190, "loss": 0.002, "learning_rate": 3.798061746947995e-08, "epoch": 4.630225080385852, "percentage": 94.74, "elapsed_time": "0:39:30", "remaining_time": "0:02:11", "throughput": "483.19", "total_tokens": 1145344} | |
| {"current_steps": 181, "total_steps": 190, "loss": 0.0003, "learning_rate": 3.077914851215585e-08, "epoch": 4.655948553054662, "percentage": 95.26, "elapsed_time": "0:39:43", "remaining_time": "0:01:58", "throughput": "483.12", "total_tokens": 1151536} | |
| {"current_steps": 182, "total_steps": 190, "loss": 0.0002, "learning_rate": 2.4329828146074096e-08, "epoch": 4.681672025723473, "percentage": 95.79, "elapsed_time": "0:39:56", "remaining_time": "0:01:45", "throughput": "483.03", "total_tokens": 1157696} | |
| {"current_steps": 183, "total_steps": 190, "loss": 0.0043, "learning_rate": 1.8634620896695044e-08, "epoch": 4.707395498392283, "percentage": 96.32, "elapsed_time": "0:40:09", "remaining_time": "0:01:32", "throughput": "482.89", "total_tokens": 1163712} | |
| {"current_steps": 184, "total_steps": 190, "loss": 0.0002, "learning_rate": 1.3695261579316776e-08, "epoch": 4.733118971061093, "percentage": 96.84, "elapsed_time": "0:40:23", "remaining_time": "0:01:19", "throughput": "482.81", "total_tokens": 1169888} | |
| {"current_steps": 185, "total_steps": 190, "loss": 0.0013, "learning_rate": 9.513254770636138e-09, "epoch": 4.758842443729904, "percentage": 97.37, "elapsed_time": "0:40:36", "remaining_time": "0:01:05", "throughput": "482.82", "total_tokens": 1176272} | |
| {"current_steps": 186, "total_steps": 190, "loss": 0.0023, "learning_rate": 6.089874350439507e-09, "epoch": 4.784565916398714, "percentage": 97.89, "elapsed_time": "0:40:49", "remaining_time": "0:00:52", "throughput": "482.85", "total_tokens": 1182688} | |
| {"current_steps": 187, "total_steps": 190, "loss": 0.0002, "learning_rate": 3.4261631135654174e-09, "epoch": 4.810289389067524, "percentage": 98.42, "elapsed_time": "0:41:02", "remaining_time": "0:00:39", "throughput": "482.98", "total_tokens": 1189360} | |
| {"current_steps": 188, "total_steps": 190, "loss": 0.0015, "learning_rate": 1.5229324522605949e-09, "epoch": 4.836012861736334, "percentage": 98.95, "elapsed_time": "0:41:15", "remaining_time": "0:00:26", "throughput": "482.95", "total_tokens": 1195632} | |
| {"current_steps": 189, "total_steps": 190, "loss": 0.0002, "learning_rate": 3.8076210902182607e-10, "epoch": 4.861736334405145, "percentage": 99.47, "elapsed_time": "0:41:28", "remaining_time": "0:00:13", "throughput": "482.96", "total_tokens": 1202016} | |
| {"current_steps": 190, "total_steps": 190, "loss": 0.0028, "learning_rate": 0.0, "epoch": 4.887459807073955, "percentage": 100.0, "elapsed_time": "0:41:42", "remaining_time": "0:00:00", "throughput": "482.97", "total_tokens": 1208400} | |
| {"current_steps": 190, "total_steps": 190, "epoch": 4.887459807073955, "percentage": 100.0, "elapsed_time": "0:42:42", "remaining_time": "0:00:00", "throughput": "471.54", "total_tokens": 1208400} | |