OH_original_wo_slimorca_550k / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 3
e0ed9f8 verified
{"current_steps": 10, "total_steps": 591, "loss": 0.7058, "learning_rate": 5e-06, "epoch": 0.050761421319796954, "percentage": 1.69, "elapsed_time": "0:02:57", "remaining_time": "2:51:43"}
{"current_steps": 20, "total_steps": 591, "loss": 0.6396, "learning_rate": 5e-06, "epoch": 0.10152284263959391, "percentage": 3.38, "elapsed_time": "0:05:48", "remaining_time": "2:45:53"}
{"current_steps": 30, "total_steps": 591, "loss": 0.6263, "learning_rate": 5e-06, "epoch": 0.15228426395939088, "percentage": 5.08, "elapsed_time": "0:08:39", "remaining_time": "2:42:02"}
{"current_steps": 40, "total_steps": 591, "loss": 0.6163, "learning_rate": 5e-06, "epoch": 0.20304568527918782, "percentage": 6.77, "elapsed_time": "0:11:31", "remaining_time": "2:38:48"}
{"current_steps": 50, "total_steps": 591, "loss": 0.5983, "learning_rate": 5e-06, "epoch": 0.25380710659898476, "percentage": 8.46, "elapsed_time": "0:14:23", "remaining_time": "2:35:45"}
{"current_steps": 60, "total_steps": 591, "loss": 0.5841, "learning_rate": 5e-06, "epoch": 0.30456852791878175, "percentage": 10.15, "elapsed_time": "0:17:15", "remaining_time": "2:32:46"}
{"current_steps": 70, "total_steps": 591, "loss": 0.5851, "learning_rate": 5e-06, "epoch": 0.3553299492385787, "percentage": 11.84, "elapsed_time": "0:20:06", "remaining_time": "2:29:37"}
{"current_steps": 80, "total_steps": 591, "loss": 0.5743, "learning_rate": 5e-06, "epoch": 0.40609137055837563, "percentage": 13.54, "elapsed_time": "0:22:57", "remaining_time": "2:26:40"}
{"current_steps": 90, "total_steps": 591, "loss": 0.5631, "learning_rate": 5e-06, "epoch": 0.45685279187817257, "percentage": 15.23, "elapsed_time": "0:25:49", "remaining_time": "2:23:44"}
{"current_steps": 100, "total_steps": 591, "loss": 0.5621, "learning_rate": 5e-06, "epoch": 0.5076142131979695, "percentage": 16.92, "elapsed_time": "0:28:41", "remaining_time": "2:20:54"}
{"current_steps": 110, "total_steps": 591, "loss": 0.5626, "learning_rate": 5e-06, "epoch": 0.5583756345177665, "percentage": 18.61, "elapsed_time": "0:31:33", "remaining_time": "2:18:01"}
{"current_steps": 120, "total_steps": 591, "loss": 0.5576, "learning_rate": 5e-06, "epoch": 0.6091370558375635, "percentage": 20.3, "elapsed_time": "0:34:24", "remaining_time": "2:15:04"}
{"current_steps": 130, "total_steps": 591, "loss": 0.5634, "learning_rate": 5e-06, "epoch": 0.6598984771573604, "percentage": 22.0, "elapsed_time": "0:37:15", "remaining_time": "2:12:05"}
{"current_steps": 140, "total_steps": 591, "loss": 0.5556, "learning_rate": 5e-06, "epoch": 0.7106598984771574, "percentage": 23.69, "elapsed_time": "0:40:06", "remaining_time": "2:09:12"}
{"current_steps": 150, "total_steps": 591, "loss": 0.5485, "learning_rate": 5e-06, "epoch": 0.7614213197969543, "percentage": 25.38, "elapsed_time": "0:42:59", "remaining_time": "2:06:22"}
{"current_steps": 160, "total_steps": 591, "loss": 0.5555, "learning_rate": 5e-06, "epoch": 0.8121827411167513, "percentage": 27.07, "elapsed_time": "0:45:51", "remaining_time": "2:03:32"}
{"current_steps": 170, "total_steps": 591, "loss": 0.5534, "learning_rate": 5e-06, "epoch": 0.8629441624365483, "percentage": 28.76, "elapsed_time": "0:48:44", "remaining_time": "2:00:42"}
{"current_steps": 180, "total_steps": 591, "loss": 0.5504, "learning_rate": 5e-06, "epoch": 0.9137055837563451, "percentage": 30.46, "elapsed_time": "0:51:37", "remaining_time": "1:57:51"}
{"current_steps": 190, "total_steps": 591, "loss": 0.5428, "learning_rate": 5e-06, "epoch": 0.9644670050761421, "percentage": 32.15, "elapsed_time": "0:54:27", "remaining_time": "1:54:56"}
{"current_steps": 197, "total_steps": 591, "eval_loss": 0.545667827129364, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "0:57:36", "remaining_time": "1:55:12"}
{"current_steps": 200, "total_steps": 591, "loss": 0.538, "learning_rate": 5e-06, "epoch": 1.015228426395939, "percentage": 33.84, "elapsed_time": "0:59:26", "remaining_time": "1:56:12"}
{"current_steps": 210, "total_steps": 591, "loss": 0.5005, "learning_rate": 5e-06, "epoch": 1.0659898477157361, "percentage": 35.53, "elapsed_time": "1:02:18", "remaining_time": "1:53:01"}
{"current_steps": 220, "total_steps": 591, "loss": 0.5116, "learning_rate": 5e-06, "epoch": 1.116751269035533, "percentage": 37.23, "elapsed_time": "1:05:07", "remaining_time": "1:49:50"}
{"current_steps": 230, "total_steps": 591, "loss": 0.5039, "learning_rate": 5e-06, "epoch": 1.16751269035533, "percentage": 38.92, "elapsed_time": "1:07:58", "remaining_time": "1:46:41"}
{"current_steps": 240, "total_steps": 591, "loss": 0.4979, "learning_rate": 5e-06, "epoch": 1.218274111675127, "percentage": 40.61, "elapsed_time": "1:10:49", "remaining_time": "1:43:34"}
{"current_steps": 250, "total_steps": 591, "loss": 0.4977, "learning_rate": 5e-06, "epoch": 1.2690355329949239, "percentage": 42.3, "elapsed_time": "1:13:40", "remaining_time": "1:40:29"}
{"current_steps": 260, "total_steps": 591, "loss": 0.5015, "learning_rate": 5e-06, "epoch": 1.3197969543147208, "percentage": 43.99, "elapsed_time": "1:16:31", "remaining_time": "1:37:25"}
{"current_steps": 270, "total_steps": 591, "loss": 0.5052, "learning_rate": 5e-06, "epoch": 1.3705583756345177, "percentage": 45.69, "elapsed_time": "1:19:21", "remaining_time": "1:34:20"}
{"current_steps": 280, "total_steps": 591, "loss": 0.4974, "learning_rate": 5e-06, "epoch": 1.4213197969543148, "percentage": 47.38, "elapsed_time": "1:22:11", "remaining_time": "1:31:18"}
{"current_steps": 290, "total_steps": 591, "loss": 0.5044, "learning_rate": 5e-06, "epoch": 1.4720812182741116, "percentage": 49.07, "elapsed_time": "1:25:02", "remaining_time": "1:28:15"}
{"current_steps": 300, "total_steps": 591, "loss": 0.5134, "learning_rate": 5e-06, "epoch": 1.5228426395939088, "percentage": 50.76, "elapsed_time": "1:27:53", "remaining_time": "1:25:15"}
{"current_steps": 310, "total_steps": 591, "loss": 0.5022, "learning_rate": 5e-06, "epoch": 1.5736040609137056, "percentage": 52.45, "elapsed_time": "1:30:44", "remaining_time": "1:22:14"}
{"current_steps": 320, "total_steps": 591, "loss": 0.5001, "learning_rate": 5e-06, "epoch": 1.6243654822335025, "percentage": 54.15, "elapsed_time": "1:33:33", "remaining_time": "1:19:13"}
{"current_steps": 330, "total_steps": 591, "loss": 0.504, "learning_rate": 5e-06, "epoch": 1.6751269035532994, "percentage": 55.84, "elapsed_time": "1:36:23", "remaining_time": "1:16:14"}
{"current_steps": 340, "total_steps": 591, "loss": 0.5008, "learning_rate": 5e-06, "epoch": 1.7258883248730963, "percentage": 57.53, "elapsed_time": "1:39:13", "remaining_time": "1:13:15"}
{"current_steps": 350, "total_steps": 591, "loss": 0.5058, "learning_rate": 5e-06, "epoch": 1.7766497461928934, "percentage": 59.22, "elapsed_time": "1:42:04", "remaining_time": "1:10:17"}
{"current_steps": 360, "total_steps": 591, "loss": 0.5013, "learning_rate": 5e-06, "epoch": 1.8274111675126905, "percentage": 60.91, "elapsed_time": "1:44:55", "remaining_time": "1:07:19"}
{"current_steps": 370, "total_steps": 591, "loss": 0.505, "learning_rate": 5e-06, "epoch": 1.8781725888324874, "percentage": 62.61, "elapsed_time": "1:47:45", "remaining_time": "1:04:22"}
{"current_steps": 380, "total_steps": 591, "loss": 0.5029, "learning_rate": 5e-06, "epoch": 1.9289340101522843, "percentage": 64.3, "elapsed_time": "1:50:35", "remaining_time": "1:01:24"}
{"current_steps": 390, "total_steps": 591, "loss": 0.4991, "learning_rate": 5e-06, "epoch": 1.9796954314720812, "percentage": 65.99, "elapsed_time": "1:53:26", "remaining_time": "0:58:27"}
{"current_steps": 394, "total_steps": 591, "eval_loss": 0.5367357730865479, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "1:55:43", "remaining_time": "0:57:51"}
{"current_steps": 400, "total_steps": 591, "loss": 0.4785, "learning_rate": 5e-06, "epoch": 2.030456852791878, "percentage": 67.68, "elapsed_time": "1:58:31", "remaining_time": "0:56:35"}
{"current_steps": 410, "total_steps": 591, "loss": 0.4563, "learning_rate": 5e-06, "epoch": 2.081218274111675, "percentage": 69.37, "elapsed_time": "2:01:20", "remaining_time": "0:53:34"}
{"current_steps": 420, "total_steps": 591, "loss": 0.4564, "learning_rate": 5e-06, "epoch": 2.1319796954314723, "percentage": 71.07, "elapsed_time": "2:04:09", "remaining_time": "0:50:33"}
{"current_steps": 430, "total_steps": 591, "loss": 0.4559, "learning_rate": 5e-06, "epoch": 2.182741116751269, "percentage": 72.76, "elapsed_time": "2:06:59", "remaining_time": "0:47:32"}
{"current_steps": 440, "total_steps": 591, "loss": 0.454, "learning_rate": 5e-06, "epoch": 2.233502538071066, "percentage": 74.45, "elapsed_time": "2:09:47", "remaining_time": "0:44:32"}
{"current_steps": 450, "total_steps": 591, "loss": 0.4584, "learning_rate": 5e-06, "epoch": 2.284263959390863, "percentage": 76.14, "elapsed_time": "2:12:38", "remaining_time": "0:41:33"}
{"current_steps": 460, "total_steps": 591, "loss": 0.4603, "learning_rate": 5e-06, "epoch": 2.33502538071066, "percentage": 77.83, "elapsed_time": "2:15:29", "remaining_time": "0:38:35"}
{"current_steps": 470, "total_steps": 591, "loss": 0.4586, "learning_rate": 5e-06, "epoch": 2.3857868020304567, "percentage": 79.53, "elapsed_time": "2:18:19", "remaining_time": "0:35:36"}
{"current_steps": 480, "total_steps": 591, "loss": 0.4547, "learning_rate": 5e-06, "epoch": 2.436548223350254, "percentage": 81.22, "elapsed_time": "2:21:09", "remaining_time": "0:32:38"}
{"current_steps": 490, "total_steps": 591, "loss": 0.4553, "learning_rate": 5e-06, "epoch": 2.487309644670051, "percentage": 82.91, "elapsed_time": "2:24:00", "remaining_time": "0:29:41"}
{"current_steps": 500, "total_steps": 591, "loss": 0.4552, "learning_rate": 5e-06, "epoch": 2.5380710659898478, "percentage": 84.6, "elapsed_time": "2:26:51", "remaining_time": "0:26:43"}
{"current_steps": 510, "total_steps": 591, "loss": 0.4495, "learning_rate": 5e-06, "epoch": 2.5888324873096447, "percentage": 86.29, "elapsed_time": "2:29:40", "remaining_time": "0:23:46"}
{"current_steps": 520, "total_steps": 591, "loss": 0.4601, "learning_rate": 5e-06, "epoch": 2.6395939086294415, "percentage": 87.99, "elapsed_time": "2:32:31", "remaining_time": "0:20:49"}
{"current_steps": 530, "total_steps": 591, "loss": 0.4639, "learning_rate": 5e-06, "epoch": 2.6903553299492384, "percentage": 89.68, "elapsed_time": "2:35:21", "remaining_time": "0:17:52"}
{"current_steps": 540, "total_steps": 591, "loss": 0.4561, "learning_rate": 5e-06, "epoch": 2.7411167512690353, "percentage": 91.37, "elapsed_time": "2:38:10", "remaining_time": "0:14:56"}
{"current_steps": 550, "total_steps": 591, "loss": 0.458, "learning_rate": 5e-06, "epoch": 2.7918781725888326, "percentage": 93.06, "elapsed_time": "2:40:59", "remaining_time": "0:12:00"}
{"current_steps": 560, "total_steps": 591, "loss": 0.4582, "learning_rate": 5e-06, "epoch": 2.8426395939086295, "percentage": 94.75, "elapsed_time": "2:43:47", "remaining_time": "0:09:04"}
{"current_steps": 570, "total_steps": 591, "loss": 0.4535, "learning_rate": 5e-06, "epoch": 2.8934010152284264, "percentage": 96.45, "elapsed_time": "2:46:36", "remaining_time": "0:06:08"}
{"current_steps": 580, "total_steps": 591, "loss": 0.4609, "learning_rate": 5e-06, "epoch": 2.9441624365482233, "percentage": 98.14, "elapsed_time": "2:49:25", "remaining_time": "0:03:12"}
{"current_steps": 590, "total_steps": 591, "loss": 0.4636, "learning_rate": 5e-06, "epoch": 2.99492385786802, "percentage": 99.83, "elapsed_time": "2:52:15", "remaining_time": "0:00:17"}
{"current_steps": 591, "total_steps": 591, "eval_loss": 0.5386124849319458, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "2:54:45", "remaining_time": "0:00:00"}
{"current_steps": 591, "total_steps": 591, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "2:55:42", "remaining_time": "0:00:00"}