trainer_log.jsonl

{"current_steps": 1, "total_steps": 312, "loss": 0.9262, "lr": 3.125e-07, "epoch": 0.0096, "percentage": 0.32, "elapsed_time": "0:02:00", "remaining_time": "10:23:34"}
{"current_steps": 2, "total_steps": 312, "loss": 0.9505, "lr": 6.25e-07, "epoch": 0.0192, "percentage": 0.64, "elapsed_time": "0:04:01", "remaining_time": "10:23:47"}
{"current_steps": 3, "total_steps": 312, "loss": 0.9138, "lr": 9.375000000000001e-07, "epoch": 0.0288, "percentage": 0.96, "elapsed_time": "0:05:56", "remaining_time": "10:12:11"}
{"current_steps": 4, "total_steps": 312, "loss": 0.9032, "lr": 1.25e-06, "epoch": 0.0384, "percentage": 1.28, "elapsed_time": "0:07:47", "remaining_time": "9:59:55"}
{"current_steps": 5, "total_steps": 312, "loss": 0.8911, "lr": 1.5625e-06, "epoch": 0.048, "percentage": 1.6, "elapsed_time": "0:09:40", "remaining_time": "9:54:09"}
{"current_steps": 6, "total_steps": 312, "loss": 0.919, "lr": 1.8750000000000003e-06, "epoch": 0.0576, "percentage": 1.92, "elapsed_time": "0:11:20", "remaining_time": "9:38:24"}
{"current_steps": 7, "total_steps": 312, "loss": 0.8684, "lr": 2.1875000000000002e-06, "epoch": 0.0672, "percentage": 2.24, "elapsed_time": "0:13:28", "remaining_time": "9:47:05"}
{"current_steps": 8, "total_steps": 312, "loss": 0.8867, "lr": 2.5e-06, "epoch": 0.0768, "percentage": 2.56, "elapsed_time": "0:15:19", "remaining_time": "9:42:35"}
{"current_steps": 9, "total_steps": 312, "loss": 0.848, "lr": 2.8125e-06, "epoch": 0.0864, "percentage": 2.88, "elapsed_time": "0:16:50", "remaining_time": "9:27:16"}
{"current_steps": 10, "total_steps": 312, "loss": 0.8326, "lr": 3.125e-06, "epoch": 0.096, "percentage": 3.21, "elapsed_time": "0:18:40", "remaining_time": "9:24:00"}
{"current_steps": 11, "total_steps": 312, "loss": 0.8688, "lr": 3.4375e-06, "epoch": 0.1056, "percentage": 3.53, "elapsed_time": "0:20:28", "remaining_time": "9:20:23"}
{"current_steps": 12, "total_steps": 312, "loss": 0.8271, "lr": 3.7500000000000005e-06, "epoch": 0.1152, "percentage": 3.85, "elapsed_time": "0:22:10", "remaining_time": "9:14:20"}
{"current_steps": 13, "total_steps": 312, "loss": 0.8368, "lr": 4.0625000000000005e-06, "epoch": 0.1248, "percentage": 4.17, "elapsed_time": "0:24:02", "remaining_time": "9:12:50"}
{"current_steps": 14, "total_steps": 312, "loss": 0.8211, "lr": 4.3750000000000005e-06, "epoch": 0.1344, "percentage": 4.49, "elapsed_time": "0:26:00", "remaining_time": "9:13:40"}
{"current_steps": 15, "total_steps": 312, "loss": 0.8001, "lr": 4.6875000000000004e-06, "epoch": 0.144, "percentage": 4.81, "elapsed_time": "0:27:53", "remaining_time": "9:12:07"}
{"current_steps": 16, "total_steps": 312, "loss": 0.7876, "lr": 5e-06, "epoch": 0.1536, "percentage": 5.13, "elapsed_time": "0:29:34", "remaining_time": "9:07:11"}
{"current_steps": 17, "total_steps": 312, "loss": 0.7792, "lr": 5.3125e-06, "epoch": 0.1632, "percentage": 5.45, "elapsed_time": "0:31:30", "remaining_time": "9:06:52"}
{"current_steps": 18, "total_steps": 312, "loss": 0.7999, "lr": 5.625e-06, "epoch": 0.1728, "percentage": 5.77, "elapsed_time": "0:33:10", "remaining_time": "9:01:59"}
{"current_steps": 19, "total_steps": 312, "loss": 0.7477, "lr": 5.9375e-06, "epoch": 0.1824, "percentage": 6.09, "elapsed_time": "0:35:18", "remaining_time": "9:04:35"}
{"current_steps": 20, "total_steps": 312, "loss": 0.727, "lr": 6.25e-06, "epoch": 0.192, "percentage": 6.41, "elapsed_time": "0:37:07", "remaining_time": "9:01:57"}
{"current_steps": 21, "total_steps": 312, "loss": 0.7473, "lr": 6.5625e-06, "epoch": 0.2016, "percentage": 6.73, "elapsed_time": "0:39:10", "remaining_time": "9:02:53"}
{"current_steps": 22, "total_steps": 312, "loss": 0.7575, "lr": 6.875e-06, "epoch": 0.2112, "percentage": 7.05, "elapsed_time": "0:41:04", "remaining_time": "9:01:20"}
{"current_steps": 23, "total_steps": 312, "loss": 0.7247, "lr": 7.1875e-06, "epoch": 0.2208, "percentage": 7.37, "elapsed_time": "0:42:57", "remaining_time": "8:59:45"}
{"current_steps": 24, "total_steps": 312, "loss": 0.7094, "lr": 7.500000000000001e-06, "epoch": 0.2304, "percentage": 7.69, "elapsed_time": "0:44:54", "remaining_time": "8:58:55"}
{"current_steps": 25, "total_steps": 312, "loss": 0.6891, "lr": 7.8125e-06, "epoch": 0.24, "percentage": 8.01, "elapsed_time": "0:46:53", "remaining_time": "8:58:24"}
{"current_steps": 26, "total_steps": 312, "loss": 0.6937, "lr": 8.125000000000001e-06, "epoch": 0.2496, "percentage": 8.33, "elapsed_time": "0:48:37", "remaining_time": "8:54:54"}
{"current_steps": 27, "total_steps": 312, "loss": 0.7369, "lr": 8.4375e-06, "epoch": 0.2592, "percentage": 8.65, "elapsed_time": "0:50:18", "remaining_time": "8:51:03"}
{"current_steps": 28, "total_steps": 312, "loss": 0.6974, "lr": 8.750000000000001e-06, "epoch": 0.2688, "percentage": 8.97, "elapsed_time": "0:51:58", "remaining_time": "8:47:14"}
{"current_steps": 29, "total_steps": 312, "loss": 0.7266, "lr": 9.0625e-06, "epoch": 0.2784, "percentage": 9.29, "elapsed_time": "0:53:53", "remaining_time": "8:45:56"}
{"current_steps": 30, "total_steps": 312, "loss": 0.6803, "lr": 9.375000000000001e-06, "epoch": 0.288, "percentage": 9.62, "elapsed_time": "0:55:41", "remaining_time": "8:43:33"}
{"current_steps": 31, "total_steps": 312, "loss": 0.6481, "lr": 9.6875e-06, "epoch": 0.2976, "percentage": 9.94, "elapsed_time": "0:57:23", "remaining_time": "8:40:15"}
{"current_steps": 32, "total_steps": 312, "loss": 0.6648, "lr": 1e-05, "epoch": 0.3072, "percentage": 10.26, "elapsed_time": "0:59:15", "remaining_time": "8:38:34"}
{"current_steps": 33, "total_steps": 312, "loss": 0.7003, "lr": 9.999685283773504e-06, "epoch": 0.3168, "percentage": 10.58, "elapsed_time": "1:01:08", "remaining_time": "8:36:53"}
{"current_steps": 34, "total_steps": 312, "loss": 0.659, "lr": 9.998741174712534e-06, "epoch": 0.3264, "percentage": 10.9, "elapsed_time": "1:03:08", "remaining_time": "8:36:14"}
{"current_steps": 35, "total_steps": 312, "loss": 0.6769, "lr": 9.997167791667668e-06, "epoch": 0.336, "percentage": 11.22, "elapsed_time": "1:05:02", "remaining_time": "8:34:46"}
{"current_steps": 36, "total_steps": 312, "loss": 0.6834, "lr": 9.994965332706574e-06, "epoch": 0.3456, "percentage": 11.54, "elapsed_time": "1:06:56", "remaining_time": "8:33:11"}
{"current_steps": 37, "total_steps": 312, "loss": 0.6857, "lr": 9.992134075089085e-06, "epoch": 0.3552, "percentage": 11.86, "elapsed_time": "1:08:42", "remaining_time": "8:30:42"}
{"current_steps": 38, "total_steps": 312, "loss": 0.6391, "lr": 9.98867437523228e-06, "epoch": 0.3648, "percentage": 12.18, "elapsed_time": "1:10:41", "remaining_time": "8:29:45"}
{"current_steps": 39, "total_steps": 312, "loss": 0.6642, "lr": 9.984586668665641e-06, "epoch": 0.3744, "percentage": 12.5, "elapsed_time": "1:12:36", "remaining_time": "8:28:16"}
{"current_steps": 40, "total_steps": 312, "loss": 0.6836, "lr": 9.979871469976197e-06, "epoch": 0.384, "percentage": 12.82, "elapsed_time": "1:14:24", "remaining_time": "8:26:00"}
{"current_steps": 41, "total_steps": 312, "loss": 0.6486, "lr": 9.974529372743762e-06, "epoch": 0.3936, "percentage": 13.14, "elapsed_time": "1:16:14", "remaining_time": "8:23:58"}
{"current_steps": 42, "total_steps": 312, "loss": 0.6614, "lr": 9.968561049466214e-06, "epoch": 0.4032, "percentage": 13.46, "elapsed_time": "1:17:53", "remaining_time": "8:20:44"}
{"current_steps": 43, "total_steps": 312, "loss": 0.6697, "lr": 9.961967251474823e-06, "epoch": 0.4128, "percentage": 13.78, "elapsed_time": "1:19:34", "remaining_time": "8:17:50"}
{"current_steps": 44, "total_steps": 312, "loss": 0.6512, "lr": 9.954748808839675e-06, "epoch": 0.4224, "percentage": 14.1, "elapsed_time": "1:21:24", "remaining_time": "8:15:51"}
{"current_steps": 45, "total_steps": 312, "loss": 0.6449, "lr": 9.946906630265184e-06, "epoch": 0.432, "percentage": 14.42, "elapsed_time": "1:23:16", "remaining_time": "8:14:07"}
{"current_steps": 46, "total_steps": 312, "loss": 0.6511, "lr": 9.938441702975689e-06, "epoch": 0.4416, "percentage": 14.74, "elapsed_time": "1:25:13", "remaining_time": "8:12:47"}
{"current_steps": 47, "total_steps": 312, "loss": 0.6531, "lr": 9.92935509259118e-06, "epoch": 0.4512, "percentage": 15.06, "elapsed_time": "1:27:12", "remaining_time": "8:11:44"}
{"current_steps": 48, "total_steps": 312, "loss": 0.6671, "lr": 9.91964794299315e-06, "epoch": 0.4608, "percentage": 15.38, "elapsed_time": "1:28:58", "remaining_time": "8:09:24"}
{"current_steps": 49, "total_steps": 312, "loss": 0.6459, "lr": 9.909321476180594e-06, "epoch": 0.4704, "percentage": 15.71, "elapsed_time": "1:30:54", "remaining_time": "8:07:56"}
{"current_steps": 50, "total_steps": 312, "loss": 0.6408, "lr": 9.898376992116179e-06, "epoch": 0.48, "percentage": 16.03, "elapsed_time": "1:32:48", "remaining_time": "8:06:20"}
{"current_steps": 51, "total_steps": 312, "loss": 0.6454, "lr": 9.886815868562596e-06, "epoch": 0.4896, "percentage": 16.35, "elapsed_time": "1:34:39", "remaining_time": "8:04:26"}
{"current_steps": 52, "total_steps": 312, "loss": 0.6619, "lr": 9.874639560909118e-06, "epoch": 0.4992, "percentage": 16.67, "elapsed_time": "1:36:41", "remaining_time": "8:03:28"}
{"current_steps": 53, "total_steps": 312, "loss": 0.6234, "lr": 9.861849601988384e-06, "epoch": 0.5088, "percentage": 16.99, "elapsed_time": "1:38:47", "remaining_time": "8:02:44"}
{"current_steps": 54, "total_steps": 312, "loss": 0.6384, "lr": 9.848447601883436e-06, "epoch": 0.5184, "percentage": 17.31, "elapsed_time": "1:40:28", "remaining_time": "8:00:02"}
{"current_steps": 55, "total_steps": 312, "loss": 0.6289, "lr": 9.834435247725032e-06, "epoch": 0.528, "percentage": 17.63, "elapsed_time": "1:42:13", "remaining_time": "7:57:39"}
{"current_steps": 56, "total_steps": 312, "loss": 0.6415, "lr": 9.819814303479268e-06, "epoch": 0.5376, "percentage": 17.95, "elapsed_time": "1:43:59", "remaining_time": "7:55:23"}
{"current_steps": 57, "total_steps": 312, "loss": 0.6554, "lr": 9.804586609725499e-06, "epoch": 0.5472, "percentage": 18.27, "elapsed_time": "1:45:46", "remaining_time": "7:53:10"}
{"current_steps": 58, "total_steps": 312, "loss": 0.6328, "lr": 9.788754083424654e-06, "epoch": 0.5568, "percentage": 18.59, "elapsed_time": "1:47:32", "remaining_time": "7:50:55"}
{"current_steps": 59, "total_steps": 312, "loss": 0.6115, "lr": 9.772318717677905e-06, "epoch": 0.5664, "percentage": 18.91, "elapsed_time": "1:49:44", "remaining_time": "7:50:35"}
{"current_steps": 60, "total_steps": 312, "loss": 0.6406, "lr": 9.755282581475769e-06, "epoch": 0.576, "percentage": 19.23, "elapsed_time": "1:51:34", "remaining_time": "7:48:37"}
{"current_steps": 61, "total_steps": 312, "loss": 0.6266, "lr": 9.737647819437645e-06, "epoch": 0.5856, "percentage": 19.55, "elapsed_time": "1:53:19", "remaining_time": "7:46:17"}
{"current_steps": 62, "total_steps": 312, "loss": 0.6564, "lr": 9.719416651541839e-06, "epoch": 0.5952, "percentage": 19.87, "elapsed_time": "1:55:11", "remaining_time": "7:44:30"}
{"current_steps": 63, "total_steps": 312, "loss": 0.6149, "lr": 9.700591372846096e-06, "epoch": 0.6048, "percentage": 20.19, "elapsed_time": "1:57:00", "remaining_time": "7:42:27"}
{"current_steps": 64, "total_steps": 312, "loss": 0.6551, "lr": 9.681174353198687e-06, "epoch": 0.6144, "percentage": 20.51, "elapsed_time": "1:58:46", "remaining_time": "7:40:16"}
{"current_steps": 65, "total_steps": 312, "loss": 0.6224, "lr": 9.661168036940071e-06, "epoch": 0.624, "percentage": 20.83, "elapsed_time": "2:00:25", "remaining_time": "7:37:36"}
{"current_steps": 66, "total_steps": 312, "loss": 0.6187, "lr": 9.640574942595195e-06, "epoch": 0.6336, "percentage": 21.15, "elapsed_time": "2:02:22", "remaining_time": "7:36:07"}
{"current_steps": 67, "total_steps": 312, "loss": 0.6304, "lr": 9.619397662556434e-06, "epoch": 0.6432, "percentage": 21.47, "elapsed_time": "2:04:18", "remaining_time": "7:34:33"}
{"current_steps": 68, "total_steps": 312, "loss": 0.6287, "lr": 9.597638862757255e-06, "epoch": 0.6528, "percentage": 21.79, "elapsed_time": "2:06:15", "remaining_time": "7:33:01"}
{"current_steps": 69, "total_steps": 312, "loss": 0.6264, "lr": 9.5753012823366e-06, "epoch": 0.6624, "percentage": 22.12, "elapsed_time": "2:07:58", "remaining_time": "7:30:41"}
{"current_steps": 70, "total_steps": 312, "loss": 0.6126, "lr": 9.552387733294081e-06, "epoch": 0.672, "percentage": 22.44, "elapsed_time": "2:09:46", "remaining_time": "7:28:38"}
{"current_steps": 71, "total_steps": 312, "loss": 0.6714, "lr": 9.528901100135971e-06, "epoch": 0.6816, "percentage": 22.76, "elapsed_time": "2:11:22", "remaining_time": "7:25:54"}
{"current_steps": 72, "total_steps": 312, "loss": 0.6224, "lr": 9.504844339512096e-06, "epoch": 0.6912, "percentage": 23.08, "elapsed_time": "2:13:21", "remaining_time": "7:24:30"}
{"current_steps": 73, "total_steps": 312, "loss": 0.6364, "lr": 9.480220479843627e-06, "epoch": 0.7008, "percentage": 23.4, "elapsed_time": "2:15:06", "remaining_time": "7:22:20"}
{"current_steps": 74, "total_steps": 312, "loss": 0.6396, "lr": 9.45503262094184e-06, "epoch": 0.7104, "percentage": 23.72, "elapsed_time": "2:16:53", "remaining_time": "7:20:15"}
{"current_steps": 75, "total_steps": 312, "loss": 0.6359, "lr": 9.4292839336179e-06, "epoch": 0.72, "percentage": 24.04, "elapsed_time": "2:18:39", "remaining_time": "7:18:10"}
{"current_steps": 76, "total_steps": 312, "loss": 0.6381, "lr": 9.40297765928369e-06, "epoch": 0.7296, "percentage": 24.36, "elapsed_time": "2:20:39", "remaining_time": "7:16:47"}
{"current_steps": 77, "total_steps": 312, "loss": 0.6079, "lr": 9.376117109543769e-06, "epoch": 0.7392, "percentage": 24.68, "elapsed_time": "2:22:31", "remaining_time": "7:14:59"}
{"current_steps": 78, "total_steps": 312, "loss": 0.6316, "lr": 9.348705665778479e-06, "epoch": 0.7488, "percentage": 25.0, "elapsed_time": "2:24:30", "remaining_time": "7:13:31"}
{"current_steps": 79, "total_steps": 312, "loss": 0.6268, "lr": 9.320746778718274e-06, "epoch": 0.7584, "percentage": 25.32, "elapsed_time": "2:26:20", "remaining_time": "7:11:35"}
{"current_steps": 80, "total_steps": 312, "loss": 0.6169, "lr": 9.292243968009332e-06, "epoch": 0.768, "percentage": 25.64, "elapsed_time": "2:28:16", "remaining_time": "7:09:59"}
{"current_steps": 81, "total_steps": 312, "loss": 0.643, "lr": 9.263200821770462e-06, "epoch": 0.7776, "percentage": 25.96, "elapsed_time": "2:30:20", "remaining_time": "7:08:43"}
{"current_steps": 82, "total_steps": 312, "loss": 0.611, "lr": 9.233620996141421e-06, "epoch": 0.7872, "percentage": 26.28, "elapsed_time": "2:32:19", "remaining_time": "7:07:15"}
{"current_steps": 83, "total_steps": 312, "loss": 0.6487, "lr": 9.203508214822652e-06, "epoch": 0.7968, "percentage": 26.6, "elapsed_time": "2:33:56", "remaining_time": "7:04:45"}
{"current_steps": 84, "total_steps": 312, "loss": 0.6115, "lr": 9.172866268606514e-06, "epoch": 0.8064, "percentage": 26.92, "elapsed_time": "2:36:00", "remaining_time": "7:03:25"}
{"current_steps": 85, "total_steps": 312, "loss": 0.6295, "lr": 9.141699014900084e-06, "epoch": 0.816, "percentage": 27.24, "elapsed_time": "2:37:58", "remaining_time": "7:01:52"}
{"current_steps": 86, "total_steps": 312, "loss": 0.5891, "lr": 9.110010377239552e-06, "epoch": 0.8256, "percentage": 27.56, "elapsed_time": "2:39:47", "remaining_time": "6:59:54"}
{"current_steps": 87, "total_steps": 312, "loss": 0.5975, "lr": 9.077804344796302e-06, "epoch": 0.8352, "percentage": 27.88, "elapsed_time": "2:41:42", "remaining_time": "6:58:12"}
{"current_steps": 88, "total_steps": 312, "loss": 0.6414, "lr": 9.045084971874738e-06, "epoch": 0.8448, "percentage": 28.21, "elapsed_time": "2:43:30", "remaining_time": "6:56:11"}
{"current_steps": 89, "total_steps": 312, "loss": 0.6191, "lr": 9.011856377401891e-06, "epoch": 0.8544, "percentage": 28.53, "elapsed_time": "2:45:29", "remaining_time": "6:54:39"}
{"current_steps": 90, "total_steps": 312, "loss": 0.6179, "lr": 8.978122744408905e-06, "epoch": 0.864, "percentage": 28.85, "elapsed_time": "2:47:17", "remaining_time": "6:52:38"}
{"current_steps": 91, "total_steps": 312, "loss": 0.5953, "lr": 8.943888319504456e-06, "epoch": 0.8736, "percentage": 29.17, "elapsed_time": "2:49:20", "remaining_time": "6:51:16"}
{"current_steps": 92, "total_steps": 312, "loss": 0.5965, "lr": 8.90915741234015e-06, "epoch": 0.8832, "percentage": 29.49, "elapsed_time": "2:51:02", "remaining_time": "6:49:01"}
{"current_steps": 93, "total_steps": 312, "loss": 0.6552, "lr": 8.873934395068006e-06, "epoch": 0.8928, "percentage": 29.81, "elapsed_time": "2:52:43", "remaining_time": "6:46:43"}
{"current_steps": 94, "total_steps": 312, "loss": 0.6208, "lr": 8.838223701790057e-06, "epoch": 0.9024, "percentage": 30.13, "elapsed_time": "2:54:40", "remaining_time": "6:45:05"}
{"current_steps": 95, "total_steps": 312, "loss": 0.6093, "lr": 8.802029828000157e-06, "epoch": 0.912, "percentage": 30.45, "elapsed_time": "2:56:12", "remaining_time": "6:42:28"}
{"current_steps": 96, "total_steps": 312, "loss": 0.6186, "lr": 8.765357330018056e-06, "epoch": 0.9216, "percentage": 30.77, "elapsed_time": "2:58:14", "remaining_time": "6:41:03"}
{"current_steps": 97, "total_steps": 312, "loss": 0.5871, "lr": 8.728210824415829e-06, "epoch": 0.9312, "percentage": 31.09, "elapsed_time": "3:00:09", "remaining_time": "6:39:19"}
{"current_steps": 98, "total_steps": 312, "loss": 0.6441, "lr": 8.690594987436705e-06, "epoch": 0.9408, "percentage": 31.41, "elapsed_time": "3:01:59", "remaining_time": "6:37:25"}
{"current_steps": 99, "total_steps": 312, "loss": 0.5956, "lr": 8.652514554406388e-06, "epoch": 0.9504, "percentage": 31.73, "elapsed_time": "3:03:54", "remaining_time": "6:35:40"}
{"current_steps": 100, "total_steps": 312, "loss": 0.6157, "lr": 8.613974319136959e-06, "epoch": 0.96, "percentage": 32.05, "elapsed_time": "3:05:39", "remaining_time": "6:33:35"}
{"current_steps": 101, "total_steps": 312, "loss": 0.6245, "lr": 8.574979133323378e-06, "epoch": 0.9696, "percentage": 32.37, "elapsed_time": "3:07:26", "remaining_time": "6:31:36"}
{"current_steps": 102, "total_steps": 312, "loss": 0.6153, "lr": 8.535533905932739e-06, "epoch": 0.9792, "percentage": 32.69, "elapsed_time": "3:09:19", "remaining_time": "6:29:47"}
{"current_steps": 103, "total_steps": 312, "loss": 0.5939, "lr": 8.495643602586287e-06, "epoch": 0.9888, "percentage": 33.01, "elapsed_time": "3:11:07", "remaining_time": "6:27:48"}
{"current_steps": 104, "total_steps": 312, "loss": 0.5773, "lr": 8.455313244934324e-06, "epoch": 0.9984, "percentage": 33.33, "elapsed_time": "3:12:50", "remaining_time": "6:25:41"}
{"current_steps": 105, "total_steps": 312, "loss": 1.1209, "lr": 8.414547910024035e-06, "epoch": 1.008, "percentage": 33.65, "elapsed_time": "3:15:34", "remaining_time": "6:25:34"}
{"current_steps": 106, "total_steps": 312, "loss": 0.5741, "lr": 8.373352729660373e-06, "epoch": 1.0176, "percentage": 33.97, "elapsed_time": "3:17:24", "remaining_time": "6:23:38"}
{"current_steps": 107, "total_steps": 312, "loss": 0.594, "lr": 8.331732889760021e-06, "epoch": 1.0272, "percentage": 34.29, "elapsed_time": "3:19:08", "remaining_time": "6:21:31"}
{"current_steps": 108, "total_steps": 312, "loss": 0.5962, "lr": 8.289693629698564e-06, "epoch": 1.0368, "percentage": 34.62, "elapsed_time": "3:21:08", "remaining_time": "6:19:56"}
{"current_steps": 109, "total_steps": 312, "loss": 0.5994, "lr": 8.247240241650918e-06, "epoch": 1.0464, "percentage": 34.94, "elapsed_time": "3:23:07", "remaining_time": "6:18:18"}
{"current_steps": 110, "total_steps": 312, "loss": 0.5558, "lr": 8.204378069925121e-06, "epoch": 1.056, "percentage": 35.26, "elapsed_time": "3:25:02", "remaining_time": "6:16:32"}
{"current_steps": 111, "total_steps": 312, "loss": 0.5637, "lr": 8.16111251028955e-06, "epoch": 1.0656, "percentage": 35.58, "elapsed_time": "3:26:58", "remaining_time": "6:14:46"}
{"current_steps": 112, "total_steps": 312, "loss": 0.5633, "lr": 8.117449009293668e-06, "epoch": 1.0752, "percentage": 35.9, "elapsed_time": "3:28:56", "remaining_time": "6:13:06"}
{"current_steps": 113, "total_steps": 312, "loss": 0.5632, "lr": 8.073393063582386e-06, "epoch": 1.0848, "percentage": 36.22, "elapsed_time": "3:30:56", "remaining_time": "6:11:29"}
{"current_steps": 114, "total_steps": 312, "loss": 0.5738, "lr": 8.0289502192041e-06, "epoch": 1.0944, "percentage": 36.54, "elapsed_time": "3:32:49", "remaining_time": "6:09:38"}
{"current_steps": 115, "total_steps": 312, "loss": 0.5862, "lr": 7.984126070912519e-06, "epoch": 1.104, "percentage": 36.86, "elapsed_time": "3:34:43", "remaining_time": "6:07:49"}
{"current_steps": 116, "total_steps": 312, "loss": 0.5706, "lr": 7.938926261462366e-06, "epoch": 1.1136, "percentage": 37.18, "elapsed_time": "3:36:34", "remaining_time": "6:05:56"}
{"current_steps": 117, "total_steps": 312, "loss": 0.5461, "lr": 7.89335648089903e-06, "epoch": 1.1232, "percentage": 37.5, "elapsed_time": "3:38:10", "remaining_time": "6:03:37"}
{"current_steps": 118, "total_steps": 312, "loss": 0.5556, "lr": 7.84742246584226e-06, "epoch": 1.1328, "percentage": 37.82, "elapsed_time": "3:39:50", "remaining_time": "6:01:26"}
{"current_steps": 119, "total_steps": 312, "loss": 0.5982, "lr": 7.801129998764014e-06, "epoch": 1.1424, "percentage": 38.14, "elapsed_time": "3:41:51", "remaining_time": "5:59:49"}
{"current_steps": 120, "total_steps": 312, "loss": 0.589, "lr": 7.754484907260513e-06, "epoch": 1.152, "percentage": 38.46, "elapsed_time": "3:43:43", "remaining_time": "5:57:57"}
{"current_steps": 121, "total_steps": 312, "loss": 0.5473, "lr": 7.70749306331863e-06, "epoch": 1.1616, "percentage": 38.78, "elapsed_time": "3:45:31", "remaining_time": "5:55:59"}
{"current_steps": 122, "total_steps": 312, "loss": 0.5741, "lr": 7.660160382576683e-06, "epoch": 1.1712, "percentage": 39.1, "elapsed_time": "3:47:30", "remaining_time": "5:54:19"}
{"current_steps": 123, "total_steps": 312, "loss": 0.5437, "lr": 7.612492823579744e-06, "epoch": 1.1808, "percentage": 39.42, "elapsed_time": "3:49:15", "remaining_time": "5:52:16"}
{"current_steps": 124, "total_steps": 312, "loss": 0.5739, "lr": 7.564496387029532e-06, "epoch": 1.1904, "percentage": 39.74, "elapsed_time": "3:51:01", "remaining_time": "5:50:15"}
{"current_steps": 125, "total_steps": 312, "loss": 0.5534, "lr": 7.516177115029002e-06, "epoch": 1.2, "percentage": 40.06, "elapsed_time": "3:53:06", "remaining_time": "5:48:43"}
{"current_steps": 126, "total_steps": 312, "loss": 0.5998, "lr": 7.467541090321735e-06, "epoch": 1.2096, "percentage": 40.38, "elapsed_time": "3:54:52", "remaining_time": "5:46:43"}
{"current_steps": 127, "total_steps": 312, "loss": 0.5852, "lr": 7.4185944355261996e-06, "epoch": 1.2192, "percentage": 40.71, "elapsed_time": "3:56:45", "remaining_time": "5:44:52"}
{"current_steps": 128, "total_steps": 312, "loss": 0.5524, "lr": 7.369343312364994e-06, "epoch": 1.2288000000000001, "percentage": 41.03, "elapsed_time": "3:58:37", "remaining_time": "5:43:01"}
{"current_steps": 129, "total_steps": 312, "loss": 0.573, "lr": 7.319793920889171e-06, "epoch": 1.2384, "percentage": 41.35, "elapsed_time": "4:00:24", "remaining_time": "5:41:03"}
{"current_steps": 130, "total_steps": 312, "loss": 0.5794, "lr": 7.269952498697734e-06, "epoch": 1.248, "percentage": 41.67, "elapsed_time": "4:02:13", "remaining_time": "5:39:07"}
{"current_steps": 131, "total_steps": 312, "loss": 0.5484, "lr": 7.219825320152411e-06, "epoch": 1.2576, "percentage": 41.99, "elapsed_time": "4:04:11", "remaining_time": "5:37:23"}
{"current_steps": 132, "total_steps": 312, "loss": 0.5735, "lr": 7.169418695587791e-06, "epoch": 1.2671999999999999, "percentage": 42.31, "elapsed_time": "4:06:04", "remaining_time": "5:35:33"}
{"current_steps": 133, "total_steps": 312, "loss": 0.5913, "lr": 7.118738970516944e-06, "epoch": 1.2768, "percentage": 42.63, "elapsed_time": "4:07:54", "remaining_time": "5:33:38"}
{"current_steps": 134, "total_steps": 312, "loss": 0.5399, "lr": 7.067792524832604e-06, "epoch": 1.2864, "percentage": 42.95, "elapsed_time": "4:09:49", "remaining_time": "5:31:51"}
{"current_steps": 135, "total_steps": 312, "loss": 0.5717, "lr": 7.016585772004026e-06, "epoch": 1.296, "percentage": 43.27, "elapsed_time": "4:11:34", "remaining_time": "5:29:50"}
{"current_steps": 136, "total_steps": 312, "loss": 0.5667, "lr": 6.965125158269619e-06, "epoch": 1.3056, "percentage": 43.59, "elapsed_time": "4:13:32", "remaining_time": "5:28:06"}
{"current_steps": 137, "total_steps": 312, "loss": 0.565, "lr": 6.913417161825449e-06, "epoch": 1.3152, "percentage": 43.91, "elapsed_time": "4:15:14", "remaining_time": "5:26:02"}
{"current_steps": 138, "total_steps": 312, "loss": 0.5624, "lr": 6.8614682920097265e-06, "epoch": 1.3248, "percentage": 44.23, "elapsed_time": "4:16:57", "remaining_time": "5:23:58"}
{"current_steps": 139, "total_steps": 312, "loss": 0.5565, "lr": 6.809285088483361e-06, "epoch": 1.3344, "percentage": 44.55, "elapsed_time": "4:18:48", "remaining_time": "5:22:07"}
{"current_steps": 140, "total_steps": 312, "loss": 0.5799, "lr": 6.7568741204067145e-06, "epoch": 1.3439999999999999, "percentage": 44.87, "elapsed_time": "4:20:38", "remaining_time": "5:20:13"}
{"current_steps": 141, "total_steps": 312, "loss": 0.5901, "lr": 6.704241985612625e-06, "epoch": 1.3536000000000001, "percentage": 45.19, "elapsed_time": "4:22:25", "remaining_time": "5:18:15"}
{"current_steps": 142, "total_steps": 312, "loss": 0.55, "lr": 6.651395309775837e-06, "epoch": 1.3632, "percentage": 45.51, "elapsed_time": "4:24:12", "remaining_time": "5:16:18"}
{"current_steps": 143, "total_steps": 312, "loss": 0.5999, "lr": 6.598340745578908e-06, "epoch": 1.3728, "percentage": 45.83, "elapsed_time": "4:26:11", "remaining_time": "5:14:34"}
{"current_steps": 144, "total_steps": 312, "loss": 0.5903, "lr": 6.545084971874738e-06, "epoch": 1.3824, "percentage": 46.15, "elapsed_time": "4:28:01", "remaining_time": "5:12:41"}
{"current_steps": 145, "total_steps": 312, "loss": 0.5751, "lr": 6.491634692845781e-06, "epoch": 1.392, "percentage": 46.47, "elapsed_time": "4:29:57", "remaining_time": "5:10:55"}
{"current_steps": 146, "total_steps": 312, "loss": 0.538, "lr": 6.437996637160086e-06, "epoch": 1.4016, "percentage": 46.79, "elapsed_time": "4:31:41", "remaining_time": "5:08:54"}
{"current_steps": 147, "total_steps": 312, "loss": 0.547, "lr": 6.384177557124247e-06, "epoch": 1.4112, "percentage": 47.12, "elapsed_time": "4:33:34", "remaining_time": "5:07:04"}
{"current_steps": 148, "total_steps": 312, "loss": 0.5622, "lr": 6.330184227833376e-06, "epoch": 1.4208, "percentage": 47.44, "elapsed_time": "4:35:21", "remaining_time": "5:05:08"}
{"current_steps": 149, "total_steps": 312, "loss": 0.5797, "lr": 6.276023446318214e-06, "epoch": 1.4304000000000001, "percentage": 47.76, "elapsed_time": "4:37:13", "remaining_time": "5:03:16"}
{"current_steps": 150, "total_steps": 312, "loss": 0.5779, "lr": 6.2217020306894705e-06, "epoch": 1.44, "percentage": 48.08, "elapsed_time": "4:38:56", "remaining_time": "5:01:15"}
{"current_steps": 151, "total_steps": 312, "loss": 0.5671, "lr": 6.1672268192795285e-06, "epoch": 1.4496, "percentage": 48.4, "elapsed_time": "4:40:50", "remaining_time": "4:59:26"}
{"current_steps": 152, "total_steps": 312, "loss": 0.5872, "lr": 6.112604669781572e-06, "epoch": 1.4592, "percentage": 48.72, "elapsed_time": "4:42:37", "remaining_time": "4:57:29"}
{"current_steps": 153, "total_steps": 312, "loss": 0.5523, "lr": 6.057842458386315e-06, "epoch": 1.4687999999999999, "percentage": 49.04, "elapsed_time": "4:44:26", "remaining_time": "4:55:35"}
{"current_steps": 154, "total_steps": 312, "loss": 0.5988, "lr": 6.002947078916365e-06, "epoch": 1.4784, "percentage": 49.36, "elapsed_time": "4:46:21", "remaining_time": "4:53:47"}
{"current_steps": 155, "total_steps": 312, "loss": 0.5751, "lr": 5.947925441958393e-06, "epoch": 1.488, "percentage": 49.68, "elapsed_time": "4:48:23", "remaining_time": "4:52:06"}
{"current_steps": 156, "total_steps": 312, "loss": 0.5724, "lr": 5.892784473993184e-06, "epoch": 1.4976, "percentage": 50.0, "elapsed_time": "4:50:27", "remaining_time": "4:50:27"}
{"current_steps": 157, "total_steps": 312, "loss": 0.579, "lr": 5.837531116523683e-06, "epoch": 1.5072, "percentage": 50.32, "elapsed_time": "4:52:29", "remaining_time": "4:48:46"}
{"current_steps": 158, "total_steps": 312, "loss": 0.5566, "lr": 5.782172325201155e-06, "epoch": 1.5168, "percentage": 50.64, "elapsed_time": "4:54:27", "remaining_time": "4:47:00"}
{"current_steps": 159, "total_steps": 312, "loss": 0.5567, "lr": 5.726715068949564e-06, "epoch": 1.5264, "percentage": 50.96, "elapsed_time": "4:56:19", "remaining_time": "4:45:08"}
{"current_steps": 160, "total_steps": 312, "loss": 0.5581, "lr": 5.671166329088278e-06, "epoch": 1.536, "percentage": 51.28, "elapsed_time": "4:58:23", "remaining_time": "4:43:28"}
{"current_steps": 161, "total_steps": 312, "loss": 0.5532, "lr": 5.615533098453215e-06, "epoch": 1.5455999999999999, "percentage": 51.6, "elapsed_time": "5:00:13", "remaining_time": "4:41:35"}
{"current_steps": 162, "total_steps": 312, "loss": 0.5654, "lr": 5.559822380516539e-06, "epoch": 1.5552000000000001, "percentage": 51.92, "elapsed_time": "5:02:02", "remaining_time": "4:39:39"}
{"current_steps": 163, "total_steps": 312, "loss": 0.5603, "lr": 5.504041188505022e-06, "epoch": 1.5648, "percentage": 52.24, "elapsed_time": "5:03:41", "remaining_time": "4:37:36"}
{"current_steps": 164, "total_steps": 312, "loss": 0.5582, "lr": 5.448196544517168e-06, "epoch": 1.5744, "percentage": 52.56, "elapsed_time": "5:05:34", "remaining_time": "4:35:45"}
{"current_steps": 165, "total_steps": 312, "loss": 0.5717, "lr": 5.392295478639226e-06, "epoch": 1.584, "percentage": 52.88, "elapsed_time": "5:07:33", "remaining_time": "4:34:00"}
{"current_steps": 166, "total_steps": 312, "loss": 0.5595, "lr": 5.336345028060199e-06, "epoch": 1.5936, "percentage": 53.21, "elapsed_time": "5:09:17", "remaining_time": "4:32:01"}
{"current_steps": 167, "total_steps": 312, "loss": 0.5656, "lr": 5.2803522361859596e-06, "epoch": 1.6032, "percentage": 53.53, "elapsed_time": "5:11:13", "remaining_time": "4:30:13"}
{"current_steps": 168, "total_steps": 312, "loss": 0.5485, "lr": 5.224324151752575e-06, "epoch": 1.6128, "percentage": 53.85, "elapsed_time": "5:13:05", "remaining_time": "4:28:22"}
{"current_steps": 169, "total_steps": 312, "loss": 0.544, "lr": 5.168267827938971e-06, "epoch": 1.6223999999999998, "percentage": 54.17, "elapsed_time": "5:14:57", "remaining_time": "4:26:30"}
{"current_steps": 170, "total_steps": 312, "loss": 0.5975, "lr": 5.112190321479026e-06, "epoch": 1.6320000000000001, "percentage": 54.49, "elapsed_time": "5:16:50", "remaining_time": "4:24:39"}
{"current_steps": 171, "total_steps": 312, "loss": 0.5347, "lr": 5.05609869177323e-06, "epoch": 1.6416, "percentage": 54.81, "elapsed_time": "5:18:29", "remaining_time": "4:22:37"}
{"current_steps": 172, "total_steps": 312, "loss": 0.5404, "lr": 5e-06, "epoch": 1.6512, "percentage": 55.13, "elapsed_time": "5:20:15", "remaining_time": "4:20:40"}
{"current_steps": 173, "total_steps": 312, "loss": 0.5913, "lr": 4.943901308226771e-06, "epoch": 1.6608, "percentage": 55.45, "elapsed_time": "5:22:09", "remaining_time": "4:18:50"}
{"current_steps": 174, "total_steps": 312, "loss": 0.5423, "lr": 4.887809678520976e-06, "epoch": 1.6703999999999999, "percentage": 55.77, "elapsed_time": "5:24:00", "remaining_time": "4:16:58"}
{"current_steps": 175, "total_steps": 312, "loss": 0.5624, "lr": 4.831732172061032e-06, "epoch": 1.6800000000000002, "percentage": 56.09, "elapsed_time": "5:26:02", "remaining_time": "4:15:14"}
{"current_steps": 176, "total_steps": 312, "loss": 0.5671, "lr": 4.775675848247427e-06, "epoch": 1.6896, "percentage": 56.41, "elapsed_time": "5:27:48", "remaining_time": "4:13:18"}
{"current_steps": 177, "total_steps": 312, "loss": 0.5655, "lr": 4.719647763814041e-06, "epoch": 1.6992, "percentage": 56.73, "elapsed_time": "5:29:54", "remaining_time": "4:11:37"}
{"current_steps": 178, "total_steps": 312, "loss": 0.5643, "lr": 4.663654971939802e-06, "epoch": 1.7088, "percentage": 57.05, "elapsed_time": "5:31:40", "remaining_time": "4:09:41"}
{"current_steps": 179, "total_steps": 312, "loss": 0.5532, "lr": 4.6077045213607765e-06, "epoch": 1.7184, "percentage": 57.37, "elapsed_time": "5:33:31", "remaining_time": "4:07:49"}
{"current_steps": 180, "total_steps": 312, "loss": 0.6003, "lr": 4.551803455482833e-06, "epoch": 1.728, "percentage": 57.69, "elapsed_time": "5:35:40", "remaining_time": "4:06:09"}
{"current_steps": 181, "total_steps": 312, "loss": 0.5647, "lr": 4.4959588114949785e-06, "epoch": 1.7376, "percentage": 58.01, "elapsed_time": "5:37:33", "remaining_time": "4:04:18"}
{"current_steps": 182, "total_steps": 312, "loss": 0.5233, "lr": 4.4401776194834615e-06, "epoch": 1.7471999999999999, "percentage": 58.33, "elapsed_time": "5:39:20", "remaining_time": "4:02:23"}
{"current_steps": 183, "total_steps": 312, "loss": 0.5912, "lr": 4.384466901546786e-06, "epoch": 1.7568000000000001, "percentage": 58.65, "elapsed_time": "5:41:10", "remaining_time": "4:00:30"}
{"current_steps": 184, "total_steps": 312, "loss": 0.554, "lr": 4.3288336709117246e-06, "epoch": 1.7664, "percentage": 58.97, "elapsed_time": "5:42:57", "remaining_time": "3:58:35"}
{"current_steps": 185, "total_steps": 312, "loss": 0.5463, "lr": 4.273284931050438e-06, "epoch": 1.776, "percentage": 59.29, "elapsed_time": "5:44:58", "remaining_time": "3:56:49"}
{"current_steps": 186, "total_steps": 312, "loss": 0.5844, "lr": 4.217827674798845e-06, "epoch": 1.7856, "percentage": 59.62, "elapsed_time": "5:46:48", "remaining_time": "3:54:55"}
{"current_steps": 187, "total_steps": 312, "loss": 0.5171, "lr": 4.162468883476319e-06, "epoch": 1.7952, "percentage": 59.94, "elapsed_time": "5:48:36", "remaining_time": "3:53:01"}
{"current_steps": 188, "total_steps": 312, "loss": 0.6142, "lr": 4.107215526006818e-06, "epoch": 1.8048, "percentage": 60.26, "elapsed_time": "5:50:25", "remaining_time": "3:51:07"}
{"current_steps": 189, "total_steps": 312, "loss": 0.5573, "lr": 4.052074558041608e-06, "epoch": 1.8144, "percentage": 60.58, "elapsed_time": "5:52:09", "remaining_time": "3:49:11"}
{"current_steps": 190, "total_steps": 312, "loss": 0.5657, "lr": 3.997052921083637e-06, "epoch": 1.8239999999999998, "percentage": 60.9, "elapsed_time": "5:53:48", "remaining_time": "3:47:10"}
{"current_steps": 191, "total_steps": 312, "loss": 0.5581, "lr": 3.9421575416136866e-06, "epoch": 1.8336000000000001, "percentage": 61.22, "elapsed_time": "5:55:51", "remaining_time": "3:45:26"}
{"current_steps": 192, "total_steps": 312, "loss": 0.5527, "lr": 3.887395330218429e-06, "epoch": 1.8432, "percentage": 61.54, "elapsed_time": "5:57:39", "remaining_time": "3:43:32"}
{"current_steps": 193, "total_steps": 312, "loss": 0.5836, "lr": 3.832773180720475e-06, "epoch": 1.8528, "percentage": 61.86, "elapsed_time": "5:59:26", "remaining_time": "3:41:37"}
{"current_steps": 194, "total_steps": 312, "loss": 0.5439, "lr": 3.778297969310529e-06, "epoch": 1.8624, "percentage": 62.18, "elapsed_time": "6:01:19", "remaining_time": "3:39:46"}
{"current_steps": 195, "total_steps": 312, "loss": 0.5474, "lr": 3.723976553681787e-06, "epoch": 1.8719999999999999, "percentage": 62.5, "elapsed_time": "6:03:09", "remaining_time": "3:37:53"}
{"current_steps": 196, "total_steps": 312, "loss": 0.5936, "lr": 3.669815772166625e-06, "epoch": 1.8816000000000002, "percentage": 62.82, "elapsed_time": "6:04:52", "remaining_time": "3:35:57"}
{"current_steps": 197, "total_steps": 312, "loss": 0.5989, "lr": 3.6158224428757538e-06, "epoch": 1.8912, "percentage": 63.14, "elapsed_time": "6:06:50", "remaining_time": "3:34:08"}
{"current_steps": 198, "total_steps": 312, "loss": 0.5479, "lr": 3.562003362839914e-06, "epoch": 1.9008, "percentage": 63.46, "elapsed_time": "6:08:47", "remaining_time": "3:32:20"}
{"current_steps": 199, "total_steps": 312, "loss": 0.5527, "lr": 3.50836530715422e-06, "epoch": 1.9104, "percentage": 63.78, "elapsed_time": "6:10:49", "remaining_time": "3:30:34"}
{"current_steps": 200, "total_steps": 312, "loss": 0.5538, "lr": 3.4549150281252635e-06, "epoch": 1.92, "percentage": 64.1, "elapsed_time": "6:12:32", "remaining_time": "3:28:37"}
{"current_steps": 201, "total_steps": 312, "loss": 0.5691, "lr": 3.4016592544210937e-06, "epoch": 1.9296, "percentage": 64.42, "elapsed_time": "6:14:17", "remaining_time": "3:26:41"}
{"current_steps": 202, "total_steps": 312, "loss": 0.5462, "lr": 3.3486046902241663e-06, "epoch": 1.9392, "percentage": 64.74, "elapsed_time": "6:16:16", "remaining_time": "3:24:54"}
{"current_steps": 203, "total_steps": 312, "loss": 0.5764, "lr": 3.295758014387375e-06, "epoch": 1.9487999999999999, "percentage": 65.06, "elapsed_time": "6:18:10", "remaining_time": "3:23:03"}
{"current_steps": 204, "total_steps": 312, "loss": 0.536, "lr": 3.2431258795932863e-06, "epoch": 1.9584000000000001, "percentage": 65.38, "elapsed_time": "6:20:08", "remaining_time": "3:21:14"}
{"current_steps": 205, "total_steps": 312, "loss": 0.5945, "lr": 3.1907149115166403e-06, "epoch": 1.968, "percentage": 65.71, "elapsed_time": "6:21:52", "remaining_time": "3:19:19"}
{"current_steps": 206, "total_steps": 312, "loss": 0.56, "lr": 3.1385317079902743e-06, "epoch": 1.9776, "percentage": 66.03, "elapsed_time": "6:23:44", "remaining_time": "3:17:27"}
{"current_steps": 207, "total_steps": 312, "loss": 0.5542, "lr": 3.0865828381745515e-06, "epoch": 1.9872, "percentage": 66.35, "elapsed_time": "6:25:33", "remaining_time": "3:15:34"}
{"current_steps": 208, "total_steps": 312, "loss": 0.5955, "lr": 3.0348748417303826e-06, "epoch": 1.9968, "percentage": 66.67, "elapsed_time": "6:27:15", "remaining_time": "3:13:37"}
{"current_steps": 209, "total_steps": 312, "loss": 1.0655, "lr": 2.9834142279959754e-06, "epoch": 2.0064, "percentage": 66.99, "elapsed_time": "6:30:11", "remaining_time": "3:12:17"}
{"current_steps": 210, "total_steps": 312, "loss": 0.5323, "lr": 2.932207475167398e-06, "epoch": 2.016, "percentage": 67.31, "elapsed_time": "6:32:01", "remaining_time": "3:10:24"}
{"current_steps": 211, "total_steps": 312, "loss": 0.5364, "lr": 2.8812610294830568e-06, "epoch": 2.0256, "percentage": 67.63, "elapsed_time": "6:33:44", "remaining_time": "3:08:28"}
{"current_steps": 212, "total_steps": 312, "loss": 0.5504, "lr": 2.83058130441221e-06, "epoch": 2.0352, "percentage": 67.95, "elapsed_time": "6:35:33", "remaining_time": "3:06:35"}
{"current_steps": 213, "total_steps": 312, "loss": 0.4886, "lr": 2.7801746798475905e-06, "epoch": 2.0448, "percentage": 68.27, "elapsed_time": "6:37:25", "remaining_time": "3:04:43"}
{"current_steps": 214, "total_steps": 312, "loss": 0.563, "lr": 2.7300475013022666e-06, "epoch": 2.0544, "percentage": 68.59, "elapsed_time": "6:39:24", "remaining_time": "3:02:54"}
{"current_steps": 215, "total_steps": 312, "loss": 0.518, "lr": 2.6802060791108304e-06, "epoch": 2.064, "percentage": 68.91, "elapsed_time": "6:41:19", "remaining_time": "3:01:03"}
{"current_steps": 216, "total_steps": 312, "loss": 0.465, "lr": 2.6306566876350072e-06, "epoch": 2.0736, "percentage": 69.23, "elapsed_time": "6:43:11", "remaining_time": "2:59:11"}
{"current_steps": 217, "total_steps": 312, "loss": 0.5807, "lr": 2.5814055644738013e-06, "epoch": 2.0832, "percentage": 69.55, "elapsed_time": "6:45:16", "remaining_time": "2:57:25"}
{"current_steps": 218, "total_steps": 312, "loss": 0.5353, "lr": 2.532458909678266e-06, "epoch": 2.0928, "percentage": 69.87, "elapsed_time": "6:47:03", "remaining_time": "2:55:31"}
{"current_steps": 219, "total_steps": 312, "loss": 0.5228, "lr": 2.483822884971e-06, "epoch": 2.1024, "percentage": 70.19, "elapsed_time": "6:49:06", "remaining_time": "2:53:43"}
{"current_steps": 220, "total_steps": 312, "loss": 0.5144, "lr": 2.43550361297047e-06, "epoch": 2.112, "percentage": 70.51, "elapsed_time": "6:51:09", "remaining_time": "2:51:56"}
{"current_steps": 221, "total_steps": 312, "loss": 0.5289, "lr": 2.387507176420256e-06, "epoch": 2.1216, "percentage": 70.83, "elapsed_time": "6:53:10", "remaining_time": "2:50:07"}
{"current_steps": 222, "total_steps": 312, "loss": 0.5164, "lr": 2.339839617423318e-06, "epoch": 2.1312, "percentage": 71.15, "elapsed_time": "6:54:50", "remaining_time": "2:48:10"}
{"current_steps": 223, "total_steps": 312, "loss": 0.5183, "lr": 2.2925069366813718e-06, "epoch": 2.1408, "percentage": 71.47, "elapsed_time": "6:56:58", "remaining_time": "2:46:25"}
{"current_steps": 224, "total_steps": 312, "loss": 0.5208, "lr": 2.245515092739488e-06, "epoch": 2.1504, "percentage": 71.79, "elapsed_time": "6:58:56", "remaining_time": "2:44:34"}
{"current_steps": 225, "total_steps": 312, "loss": 0.5363, "lr": 2.1988700012359865e-06, "epoch": 2.16, "percentage": 72.12, "elapsed_time": "7:00:52", "remaining_time": "2:42:44"}
{"current_steps": 226, "total_steps": 312, "loss": 0.5223, "lr": 2.1525775341577404e-06, "epoch": 2.1696, "percentage": 72.44, "elapsed_time": "7:02:41", "remaining_time": "2:40:50"}
{"current_steps": 227, "total_steps": 312, "loss": 0.5162, "lr": 2.1066435191009717e-06, "epoch": 2.1792, "percentage": 72.76, "elapsed_time": "7:04:36", "remaining_time": "2:38:59"}
{"current_steps": 228, "total_steps": 312, "loss": 0.5156, "lr": 2.061073738537635e-06, "epoch": 2.1888, "percentage": 73.08, "elapsed_time": "7:06:13", "remaining_time": "2:37:01"}
{"current_steps": 229, "total_steps": 312, "loss": 0.5397, "lr": 2.0158739290874822e-06, "epoch": 2.1984, "percentage": 73.4, "elapsed_time": "7:08:25", "remaining_time": "2:35:16"}
{"current_steps": 230, "total_steps": 312, "loss": 0.5342, "lr": 1.971049780795901e-06, "epoch": 2.208, "percentage": 73.72, "elapsed_time": "7:10:27", "remaining_time": "2:33:27"}
{"current_steps": 231, "total_steps": 312, "loss": 0.5118, "lr": 1.9266069364176144e-06, "epoch": 2.2176, "percentage": 74.04, "elapsed_time": "7:12:07", "remaining_time": "2:31:31"}
{"current_steps": 232, "total_steps": 312, "loss": 0.5245, "lr": 1.8825509907063328e-06, "epoch": 2.2272, "percentage": 74.36, "elapsed_time": "7:13:50", "remaining_time": "2:29:36"}
{"current_steps": 233, "total_steps": 312, "loss": 0.569, "lr": 1.838887489710452e-06, "epoch": 2.2368, "percentage": 74.68, "elapsed_time": "7:15:37", "remaining_time": "2:27:42"}
{"current_steps": 234, "total_steps": 312, "loss": 0.494, "lr": 1.7956219300748796e-06, "epoch": 2.2464, "percentage": 75.0, "elapsed_time": "7:17:23", "remaining_time": "2:25:47"}
{"current_steps": 235, "total_steps": 312, "loss": 0.5668, "lr": 1.7527597583490825e-06, "epoch": 2.2560000000000002, "percentage": 75.32, "elapsed_time": "7:19:26", "remaining_time": "2:23:59"}
{"current_steps": 236, "total_steps": 312, "loss": 0.5039, "lr": 1.7103063703014372e-06, "epoch": 2.2656, "percentage": 75.64, "elapsed_time": "7:21:18", "remaining_time": "2:22:06"}
{"current_steps": 237, "total_steps": 312, "loss": 0.5206, "lr": 1.6682671102399806e-06, "epoch": 2.2752, "percentage": 75.96, "elapsed_time": "7:23:16", "remaining_time": "2:20:16"}
{"current_steps": 238, "total_steps": 312, "loss": 0.5307, "lr": 1.6266472703396286e-06, "epoch": 2.2848, "percentage": 76.28, "elapsed_time": "7:25:05", "remaining_time": "2:18:23"}
{"current_steps": 239, "total_steps": 312, "loss": 0.4928, "lr": 1.5854520899759656e-06, "epoch": 2.2944, "percentage": 76.6, "elapsed_time": "7:27:11", "remaining_time": "2:16:35"}
{"current_steps": 240, "total_steps": 312, "loss": 0.565, "lr": 1.544686755065677e-06, "epoch": 2.304, "percentage": 76.92, "elapsed_time": "7:29:11", "remaining_time": "2:14:45"}
{"current_steps": 241, "total_steps": 312, "loss": 0.5175, "lr": 1.5043563974137132e-06, "epoch": 2.3136, "percentage": 77.24, "elapsed_time": "7:31:02", "remaining_time": "2:12:52"}
{"current_steps": 242, "total_steps": 312, "loss": 0.5576, "lr": 1.4644660940672628e-06, "epoch": 2.3232, "percentage": 77.56, "elapsed_time": "7:33:02", "remaining_time": "2:11:02"}
{"current_steps": 243, "total_steps": 312, "loss": 0.4878, "lr": 1.4250208666766235e-06, "epoch": 2.3327999999999998, "percentage": 77.88, "elapsed_time": "7:34:47", "remaining_time": "2:09:08"}
{"current_steps": 244, "total_steps": 312, "loss": 0.474, "lr": 1.3860256808630429e-06, "epoch": 2.3424, "percentage": 78.21, "elapsed_time": "7:36:43", "remaining_time": "2:07:17"}
{"current_steps": 245, "total_steps": 312, "loss": 0.5516, "lr": 1.3474854455936126e-06, "epoch": 2.352, "percentage": 78.53, "elapsed_time": "7:38:28", "remaining_time": "2:05:22"}
{"current_steps": 246, "total_steps": 312, "loss": 0.5117, "lr": 1.3094050125632973e-06, "epoch": 2.3616, "percentage": 78.85, "elapsed_time": "7:40:10", "remaining_time": "2:03:27"}
{"current_steps": 247, "total_steps": 312, "loss": 0.5253, "lr": 1.2717891755841722e-06, "epoch": 2.3712, "percentage": 79.17, "elapsed_time": "7:42:14", "remaining_time": "2:01:38"}
{"current_steps": 248, "total_steps": 312, "loss": 0.4987, "lr": 1.234642669981946e-06, "epoch": 2.3808, "percentage": 79.49, "elapsed_time": "7:43:47", "remaining_time": "1:59:41"}
{"current_steps": 249, "total_steps": 312, "loss": 0.5615, "lr": 1.1979701719998454e-06, "epoch": 2.3904, "percentage": 79.81, "elapsed_time": "7:45:49", "remaining_time": "1:57:51"}
{"current_steps": 250, "total_steps": 312, "loss": 0.5138, "lr": 1.1617762982099446e-06, "epoch": 2.4, "percentage": 80.13, "elapsed_time": "7:47:47", "remaining_time": "1:56:00"}
{"current_steps": 251, "total_steps": 312, "loss": 0.529, "lr": 1.1260656049319957e-06, "epoch": 2.4096, "percentage": 80.45, "elapsed_time": "7:49:44", "remaining_time": "1:54:09"}
{"current_steps": 252, "total_steps": 312, "loss": 0.5435, "lr": 1.0908425876598512e-06, "epoch": 2.4192, "percentage": 80.77, "elapsed_time": "7:51:32", "remaining_time": "1:52:16"}
{"current_steps": 253, "total_steps": 312, "loss": 0.53, "lr": 1.0561116804955451e-06, "epoch": 2.4288, "percentage": 81.09, "elapsed_time": "7:53:10", "remaining_time": "1:50:20"}
{"current_steps": 254, "total_steps": 312, "loss": 0.5087, "lr": 1.0218772555910955e-06, "epoch": 2.4384, "percentage": 81.41, "elapsed_time": "7:54:56", "remaining_time": "1:48:27"}
{"current_steps": 255, "total_steps": 312, "loss": 0.5394, "lr": 9.881436225981107e-07, "epoch": 2.448, "percentage": 81.73, "elapsed_time": "7:56:51", "remaining_time": "1:46:35"}
{"current_steps": 256, "total_steps": 312, "loss": 0.4832, "lr": 9.549150281252633e-07, "epoch": 2.4576000000000002, "percentage": 82.05, "elapsed_time": "7:58:42", "remaining_time": "1:44:43"}
{"current_steps": 257, "total_steps": 312, "loss": 0.5408, "lr": 9.221956552036992e-07, "epoch": 2.4672, "percentage": 82.37, "elapsed_time": "8:00:25", "remaining_time": "1:42:48"}
{"current_steps": 258, "total_steps": 312, "loss": 0.5107, "lr": 8.899896227604509e-07, "epoch": 2.4768, "percentage": 82.69, "elapsed_time": "8:02:11", "remaining_time": "1:40:55"}
{"current_steps": 259, "total_steps": 312, "loss": 0.5341, "lr": 8.58300985099918e-07, "epoch": 2.4864, "percentage": 83.01, "elapsed_time": "8:04:20", "remaining_time": "1:39:06"}
{"current_steps": 260, "total_steps": 312, "loss": 0.5153, "lr": 8.271337313934869e-07, "epoch": 2.496, "percentage": 83.33, "elapsed_time": "8:06:16", "remaining_time": "1:37:15"}
{"current_steps": 261, "total_steps": 312, "loss": 0.5345, "lr": 7.964917851773496e-07, "epoch": 2.5056000000000003, "percentage": 83.65, "elapsed_time": "8:07:59", "remaining_time": "1:35:21"}
{"current_steps": 262, "total_steps": 312, "loss": 0.5145, "lr": 7.663790038585794e-07, "epoch": 2.5152, "percentage": 83.97, "elapsed_time": "8:09:50", "remaining_time": "1:33:28"}
{"current_steps": 263, "total_steps": 312, "loss": 0.5243, "lr": 7.367991782295392e-07, "epoch": 2.5248, "percentage": 84.29, "elapsed_time": "8:11:35", "remaining_time": "1:31:35"}
{"current_steps": 264, "total_steps": 312, "loss": 0.5349, "lr": 7.077560319906696e-07, "epoch": 2.5343999999999998, "percentage": 84.62, "elapsed_time": "8:13:22", "remaining_time": "1:29:42"}
{"current_steps": 265, "total_steps": 312, "loss": 0.5619, "lr": 6.792532212817271e-07, "epoch": 2.544, "percentage": 84.94, "elapsed_time": "8:15:11", "remaining_time": "1:27:49"}
{"current_steps": 266, "total_steps": 312, "loss": 0.4899, "lr": 6.512943342215234e-07, "epoch": 2.5536, "percentage": 85.26, "elapsed_time": "8:16:57", "remaining_time": "1:25:56"}
{"current_steps": 267, "total_steps": 312, "loss": 0.5072, "lr": 6.238828904562316e-07, "epoch": 2.5632, "percentage": 85.58, "elapsed_time": "8:18:44", "remaining_time": "1:24:03"}
{"current_steps": 268, "total_steps": 312, "loss": 0.5194, "lr": 5.9702234071631e-07, "epoch": 2.5728, "percentage": 85.9, "elapsed_time": "8:20:19", "remaining_time": "1:22:08"}
{"current_steps": 269, "total_steps": 312, "loss": 0.5599, "lr": 5.707160663821009e-07, "epoch": 2.5824, "percentage": 86.22, "elapsed_time": "8:22:07", "remaining_time": "1:20:15"}
{"current_steps": 270, "total_steps": 312, "loss": 0.5238, "lr": 5.449673790581611e-07, "epoch": 2.592, "percentage": 86.54, "elapsed_time": "8:24:02", "remaining_time": "1:18:24"}
{"current_steps": 271, "total_steps": 312, "loss": 0.5116, "lr": 5.197795201563744e-07, "epoch": 2.6016, "percentage": 86.86, "elapsed_time": "8:25:45", "remaining_time": "1:16:30"}
{"current_steps": 272, "total_steps": 312, "loss": 0.5346, "lr": 4.951556604879049e-07, "epoch": 2.6112, "percentage": 87.18, "elapsed_time": "8:27:38", "remaining_time": "1:14:39"}
{"current_steps": 273, "total_steps": 312, "loss": 0.539, "lr": 4.710988998640298e-07, "epoch": 2.6208, "percentage": 87.5, "elapsed_time": "8:29:38", "remaining_time": "1:12:48"}
{"current_steps": 274, "total_steps": 312, "loss": 0.5431, "lr": 4.4761226670592074e-07, "epoch": 2.6304, "percentage": 87.82, "elapsed_time": "8:31:41", "remaining_time": "1:10:57"}
{"current_steps": 275, "total_steps": 312, "loss": 0.5157, "lr": 4.2469871766340096e-07, "epoch": 2.64, "percentage": 88.14, "elapsed_time": "8:33:28", "remaining_time": "1:09:05"}
{"current_steps": 276, "total_steps": 312, "loss": 0.5274, "lr": 4.0236113724274716e-07, "epoch": 2.6496, "percentage": 88.46, "elapsed_time": "8:35:10", "remaining_time": "1:07:11"}
{"current_steps": 277, "total_steps": 312, "loss": 0.5359, "lr": 3.8060233744356634e-07, "epoch": 2.6592000000000002, "percentage": 88.78, "elapsed_time": "8:37:07", "remaining_time": "1:05:20"}
{"current_steps": 278, "total_steps": 312, "loss": 0.4746, "lr": 3.5942505740480583e-07, "epoch": 2.6688, "percentage": 89.1, "elapsed_time": "8:38:43", "remaining_time": "1:03:26"}
{"current_steps": 279, "total_steps": 312, "loss": 0.5251, "lr": 3.3883196305992906e-07, "epoch": 2.6784, "percentage": 89.42, "elapsed_time": "8:40:34", "remaining_time": "1:01:34"}
{"current_steps": 280, "total_steps": 312, "loss": 0.5405, "lr": 3.18825646801314e-07, "epoch": 2.6879999999999997, "percentage": 89.74, "elapsed_time": "8:42:32", "remaining_time": "0:59:43"}
{"current_steps": 281, "total_steps": 312, "loss": 0.5057, "lr": 2.9940862715390483e-07, "epoch": 2.6976, "percentage": 90.06, "elapsed_time": "8:44:31", "remaining_time": "0:57:51"}
{"current_steps": 282, "total_steps": 312, "loss": 0.5334, "lr": 2.8058334845816214e-07, "epoch": 2.7072000000000003, "percentage": 90.38, "elapsed_time": "8:46:07", "remaining_time": "0:55:58"}
{"current_steps": 283, "total_steps": 312, "loss": 0.5456, "lr": 2.6235218056235633e-07, "epoch": 2.7168, "percentage": 90.71, "elapsed_time": "8:47:46", "remaining_time": "0:54:05"}
{"current_steps": 284, "total_steps": 312, "loss": 0.5134, "lr": 2.447174185242324e-07, "epoch": 2.7264, "percentage": 91.03, "elapsed_time": "8:49:35", "remaining_time": "0:52:12"}
{"current_steps": 285, "total_steps": 312, "loss": 0.4706, "lr": 2.276812823220964e-07, "epoch": 2.7359999999999998, "percentage": 91.35, "elapsed_time": "8:51:20", "remaining_time": "0:50:20"}
{"current_steps": 286, "total_steps": 312, "loss": 0.5894, "lr": 2.1124591657534776e-07, "epoch": 2.7456, "percentage": 91.67, "elapsed_time": "8:53:15", "remaining_time": "0:48:28"}
{"current_steps": 287, "total_steps": 312, "loss": 0.5377, "lr": 1.9541339027450256e-07, "epoch": 2.7552, "percentage": 91.99, "elapsed_time": "8:55:08", "remaining_time": "0:46:36"}
{"current_steps": 288, "total_steps": 312, "loss": 0.5063, "lr": 1.801856965207338e-07, "epoch": 2.7648, "percentage": 92.31, "elapsed_time": "8:56:57", "remaining_time": "0:44:44"}
{"current_steps": 289, "total_steps": 312, "loss": 0.5533, "lr": 1.6556475227496816e-07, "epoch": 2.7744, "percentage": 92.63, "elapsed_time": "8:58:48", "remaining_time": "0:42:52"}
{"current_steps": 290, "total_steps": 312, "loss": 0.4984, "lr": 1.5155239811656562e-07, "epoch": 2.784, "percentage": 92.95, "elapsed_time": "9:00:40", "remaining_time": "0:41:01"}
{"current_steps": 291, "total_steps": 312, "loss": 0.5247, "lr": 1.3815039801161723e-07, "epoch": 2.7936, "percentage": 93.27, "elapsed_time": "9:02:32", "remaining_time": "0:39:09"}
{"current_steps": 292, "total_steps": 312, "loss": 0.567, "lr": 1.253604390908819e-07, "epoch": 2.8032, "percentage": 93.59, "elapsed_time": "9:04:29", "remaining_time": "0:37:17"}
{"current_steps": 293, "total_steps": 312, "loss": 0.5089, "lr": 1.1318413143740436e-07, "epoch": 2.8128, "percentage": 93.91, "elapsed_time": "9:06:26", "remaining_time": "0:35:26"}
{"current_steps": 294, "total_steps": 312, "loss": 0.5772, "lr": 1.0162300788382263e-07, "epoch": 2.8224, "percentage": 94.23, "elapsed_time": "9:08:26", "remaining_time": "0:33:34"}
{"current_steps": 295, "total_steps": 312, "loss": 0.5099, "lr": 9.0678523819408e-08, "epoch": 2.832, "percentage": 94.55, "elapsed_time": "9:10:23", "remaining_time": "0:31:43"}
{"current_steps": 296, "total_steps": 312, "loss": 0.5103, "lr": 8.035205700685167e-08, "epoch": 2.8416, "percentage": 94.87, "elapsed_time": "9:12:08", "remaining_time": "0:29:50"}
{"current_steps": 297, "total_steps": 312, "loss": 0.506, "lr": 7.064490740882057e-08, "epoch": 2.8512, "percentage": 95.19, "elapsed_time": "9:13:56", "remaining_time": "0:27:58"}
{"current_steps": 298, "total_steps": 312, "loss": 0.5043, "lr": 6.15582970243117e-08, "epoch": 2.8608000000000002, "percentage": 95.51, "elapsed_time": "9:15:34", "remaining_time": "0:26:06"}
{"current_steps": 299, "total_steps": 312, "loss": 0.5602, "lr": 5.3093369734816824e-08, "epoch": 2.8704, "percentage": 95.83, "elapsed_time": "9:17:18", "remaining_time": "0:24:13"}
{"current_steps": 300, "total_steps": 312, "loss": 0.5101, "lr": 4.52511911603265e-08, "epoch": 2.88, "percentage": 96.15, "elapsed_time": "9:19:14", "remaining_time": "0:22:22"}
{"current_steps": 301, "total_steps": 312, "loss": 0.5045, "lr": 3.8032748525179684e-08, "epoch": 2.8895999999999997, "percentage": 96.47, "elapsed_time": "9:20:50", "remaining_time": "0:20:29"}
{"current_steps": 302, "total_steps": 312, "loss": 0.5484, "lr": 3.143895053378698e-08, "epoch": 2.8992, "percentage": 96.79, "elapsed_time": "9:22:39", "remaining_time": "0:18:37"}
{"current_steps": 303, "total_steps": 312, "loss": 0.5411, "lr": 2.547062725623828e-08, "epoch": 2.9088000000000003, "percentage": 97.12, "elapsed_time": "9:24:41", "remaining_time": "0:16:46"}
{"current_steps": 304, "total_steps": 312, "loss": 0.5469, "lr": 2.012853002380466e-08, "epoch": 2.9184, "percentage": 97.44, "elapsed_time": "9:26:30", "remaining_time": "0:14:54"}
{"current_steps": 305, "total_steps": 312, "loss": 0.506, "lr": 1.541333133436018e-08, "epoch": 2.928, "percentage": 97.76, "elapsed_time": "9:28:12", "remaining_time": "0:13:02"}
{"current_steps": 306, "total_steps": 312, "loss": 0.5208, "lr": 1.132562476771959e-08, "epoch": 2.9375999999999998, "percentage": 98.08, "elapsed_time": "9:30:00", "remaining_time": "0:11:10"}
{"current_steps": 307, "total_steps": 312, "loss": 0.5208, "lr": 7.865924910916977e-09, "epoch": 2.9472, "percentage": 98.4, "elapsed_time": "9:31:51", "remaining_time": "0:09:18"}
{"current_steps": 308, "total_steps": 312, "loss": 0.5355, "lr": 5.034667293427053e-09, "epoch": 2.9568, "percentage": 98.72, "elapsed_time": "9:33:35", "remaining_time": "0:07:26"}
{"current_steps": 309, "total_steps": 312, "loss": 0.5167, "lr": 2.8322083323334417e-09, "epoch": 2.9664, "percentage": 99.04, "elapsed_time": "9:35:27", "remaining_time": "0:05:35"}
{"current_steps": 310, "total_steps": 312, "loss": 0.5175, "lr": 1.2588252874673469e-09, "epoch": 2.976, "percentage": 99.36, "elapsed_time": "9:37:19", "remaining_time": "0:03:43"}
{"current_steps": 311, "total_steps": 312, "loss": 0.5088, "lr": 3.147162264971471e-10, "epoch": 2.9856, "percentage": 99.68, "elapsed_time": "9:39:15", "remaining_time": "0:01:51"}
{"current_steps": 312, "total_steps": 312, "loss": 0.5469, "lr": 0.0, "epoch": 2.9952, "percentage": 100.0, "elapsed_time": "9:41:09", "remaining_time": "0:00:00"}
{"current_steps": 312, "total_steps": 312, "epoch": 2.9952, "percentage": 100.0, "elapsed_time": "9:43:14", "remaining_time": "0:00:00"}