trainer_log.jsonl

{"current_steps": 10, "total_steps": 1023, "loss": 0.8861, "learning_rate": 9.615384615384617e-07, "epoch": 0.029261155815654718, "percentage": 0.98, "elapsed_time": "0:08:50", "remaining_time": "14:55:46"}
{"current_steps": 20, "total_steps": 1023, "loss": 0.8326, "learning_rate": 1.9230769230769234e-06, "epoch": 0.058522311631309436, "percentage": 1.96, "elapsed_time": "0:17:36", "remaining_time": "14:42:47"}
{"current_steps": 30, "total_steps": 1023, "loss": 0.7582, "learning_rate": 2.8846153846153845e-06, "epoch": 0.08778346744696415, "percentage": 2.93, "elapsed_time": "0:26:20", "remaining_time": "14:31:42"}
{"current_steps": 40, "total_steps": 1023, "loss": 0.7319, "learning_rate": 3.846153846153847e-06, "epoch": 0.11704462326261887, "percentage": 3.91, "elapsed_time": "0:35:03", "remaining_time": "14:21:40"}
{"current_steps": 50, "total_steps": 1023, "loss": 0.715, "learning_rate": 4.807692307692308e-06, "epoch": 0.14630577907827358, "percentage": 4.89, "elapsed_time": "0:43:47", "remaining_time": "14:12:09"}
{"current_steps": 60, "total_steps": 1023, "loss": 0.6993, "learning_rate": 4.999162611434756e-06, "epoch": 0.1755669348939283, "percentage": 5.87, "elapsed_time": "0:52:32", "remaining_time": "14:03:22"}
{"current_steps": 70, "total_steps": 1023, "loss": 0.6873, "learning_rate": 4.995761681803464e-06, "epoch": 0.20482809070958302, "percentage": 6.84, "elapsed_time": "1:01:18", "remaining_time": "13:54:40"}
{"current_steps": 80, "total_steps": 1023, "loss": 0.6767, "learning_rate": 4.989748431344101e-06, "epoch": 0.23408924652523774, "percentage": 7.82, "elapsed_time": "1:10:04", "remaining_time": "13:45:54"}
{"current_steps": 90, "total_steps": 1023, "loss": 0.669, "learning_rate": 4.981129154142974e-06, "epoch": 0.26335040234089246, "percentage": 8.8, "elapsed_time": "1:18:49", "remaining_time": "13:37:04"}
{"current_steps": 100, "total_steps": 1023, "loss": 0.6592, "learning_rate": 4.969912872021965e-06, "epoch": 0.29261155815654716, "percentage": 9.78, "elapsed_time": "1:27:33", "remaining_time": "13:28:14"}
{"current_steps": 110, "total_steps": 1023, "loss": 0.667, "learning_rate": 4.956111325095368e-06, "epoch": 0.3218727139722019, "percentage": 10.75, "elapsed_time": "1:36:19", "remaining_time": "13:19:27"}
{"current_steps": 120, "total_steps": 1023, "loss": 0.6714, "learning_rate": 4.9397389594814805e-06, "epoch": 0.3511338697878566, "percentage": 11.73, "elapsed_time": "1:45:04", "remaining_time": "13:10:39"}
{"current_steps": 130, "total_steps": 1023, "loss": 0.6549, "learning_rate": 4.920812912181809e-06, "epoch": 0.38039502560351135, "percentage": 12.71, "elapsed_time": "1:53:49", "remaining_time": "13:01:50"}
{"current_steps": 140, "total_steps": 1023, "loss": 0.6553, "learning_rate": 4.899352993143717e-06, "epoch": 0.40965618141916604, "percentage": 13.69, "elapsed_time": "2:02:33", "remaining_time": "12:52:58"}
{"current_steps": 150, "total_steps": 1023, "loss": 0.6504, "learning_rate": 4.875381664525304e-06, "epoch": 0.4389173372348208, "percentage": 14.66, "elapsed_time": "2:11:18", "remaining_time": "12:44:12"}
{"current_steps": 160, "total_steps": 1023, "loss": 0.6521, "learning_rate": 4.848924017184196e-06, "epoch": 0.4681784930504755, "percentage": 15.64, "elapsed_time": "2:20:02", "remaining_time": "12:35:21"}
{"current_steps": 170, "total_steps": 1023, "loss": 0.6571, "learning_rate": 4.820007744414875e-06, "epoch": 0.49743964886613024, "percentage": 16.62, "elapsed_time": "2:28:47", "remaining_time": "12:26:36"}
{"current_steps": 180, "total_steps": 1023, "loss": 0.6566, "learning_rate": 4.788663112962028e-06, "epoch": 0.5267008046817849, "percentage": 17.6, "elapsed_time": "2:37:32", "remaining_time": "12:17:49"}
{"current_steps": 190, "total_steps": 1023, "loss": 0.645, "learning_rate": 4.7549229313402636e-06, "epoch": 0.5559619604974396, "percentage": 18.57, "elapsed_time": "2:46:17", "remaining_time": "12:09:03"}
{"current_steps": 200, "total_steps": 1023, "loss": 0.6461, "learning_rate": 4.71882251549334e-06, "epoch": 0.5852231163130943, "percentage": 19.55, "elapsed_time": "2:55:02", "remaining_time": "12:00:17"}
{"current_steps": 210, "total_steps": 1023, "loss": 0.6526, "learning_rate": 4.680399651828873e-06, "epoch": 0.6144842721287491, "percentage": 20.53, "elapsed_time": "3:03:47", "remaining_time": "11:51:33"}
{"current_steps": 220, "total_steps": 1023, "loss": 0.6404, "learning_rate": 4.639694557667194e-06, "epoch": 0.6437454279444038, "percentage": 21.51, "elapsed_time": "3:12:32", "remaining_time": "11:42:46"}
{"current_steps": 230, "total_steps": 1023, "loss": 0.6403, "learning_rate": 4.596749839145763e-06, "epoch": 0.6730065837600585, "percentage": 22.48, "elapsed_time": "3:21:16", "remaining_time": "11:33:57"}
{"current_steps": 240, "total_steps": 1023, "loss": 0.6472, "learning_rate": 4.551610446623209e-06, "epoch": 0.7022677395757132, "percentage": 23.46, "elapsed_time": "3:30:00", "remaining_time": "11:25:10"}
{"current_steps": 250, "total_steps": 1023, "loss": 0.6455, "learning_rate": 4.504323627629662e-06, "epoch": 0.731528895391368, "percentage": 24.44, "elapsed_time": "3:38:45", "remaining_time": "11:16:22"}
{"current_steps": 260, "total_steps": 1023, "loss": 0.6481, "learning_rate": 4.454938877412623e-06, "epoch": 0.7607900512070227, "percentage": 25.42, "elapsed_time": "3:47:28", "remaining_time": "11:07:34"}
{"current_steps": 270, "total_steps": 1023, "loss": 0.6503, "learning_rate": 4.4035078871301565e-06, "epoch": 0.7900512070226774, "percentage": 26.39, "elapsed_time": "3:56:13", "remaining_time": "10:58:47"}
{"current_steps": 280, "total_steps": 1023, "loss": 0.6491, "learning_rate": 4.350084489745601e-06, "epoch": 0.8193123628383321, "percentage": 27.37, "elapsed_time": "4:04:57", "remaining_time": "10:50:01"}
{"current_steps": 290, "total_steps": 1023, "loss": 0.6413, "learning_rate": 4.294724603680461e-06, "epoch": 0.8485735186539868, "percentage": 28.35, "elapsed_time": "4:13:42", "remaining_time": "10:41:15"}
{"current_steps": 300, "total_steps": 1023, "loss": 0.6331, "learning_rate": 4.23748617428443e-06, "epoch": 0.8778346744696416, "percentage": 29.33, "elapsed_time": "4:22:26", "remaining_time": "10:32:28"}
{"current_steps": 310, "total_steps": 1023, "loss": 0.6463, "learning_rate": 4.178429113183835e-06, "epoch": 0.9070958302852963, "percentage": 30.3, "elapsed_time": "4:31:10", "remaining_time": "10:23:42"}
{"current_steps": 320, "total_steps": 1023, "loss": 0.6381, "learning_rate": 4.117615235571962e-06, "epoch": 0.936356986100951, "percentage": 31.28, "elapsed_time": "4:39:55", "remaining_time": "10:14:57"}
{"current_steps": 330, "total_steps": 1023, "loss": 0.6414, "learning_rate": 4.05510819550692e-06, "epoch": 0.9656181419166057, "percentage": 32.26, "elapsed_time": "4:48:39", "remaining_time": "10:06:10"}
{"current_steps": 340, "total_steps": 1023, "loss": 0.6374, "learning_rate": 3.990973419284758e-06, "epoch": 0.9948792977322605, "percentage": 33.24, "elapsed_time": "4:57:23", "remaining_time": "9:57:23"}
{"current_steps": 341, "total_steps": 1023, "eval_loss": 0.641979992389679, "epoch": 0.9978054133138259, "percentage": 33.33, "elapsed_time": "5:04:21", "remaining_time": "10:08:42"}