| {"current_steps": 100, "total_steps": 3075, "loss": 0.8994, "learning_rate": 1.6233766233766235e-06, "epoch": 0.16260162601626016, "percentage": 3.25, "elapsed_time": "0:17:36", "remaining_time": "8:43:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 200, "total_steps": 3075, "loss": 0.0317, "learning_rate": 3.246753246753247e-06, "epoch": 0.3252032520325203, "percentage": 6.5, "elapsed_time": "0:35:10", "remaining_time": "8:25:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 300, "total_steps": 3075, "loss": 0.0245, "learning_rate": 4.870129870129871e-06, "epoch": 0.4878048780487805, "percentage": 9.76, "elapsed_time": "0:52:41", "remaining_time": "8:07:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 400, "total_steps": 3075, "loss": 0.0222, "learning_rate": 4.986373880811079e-06, "epoch": 0.6504065040650406, "percentage": 13.01, "elapsed_time": "1:10:13", "remaining_time": "7:49:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 3075, "loss": 0.0208, "learning_rate": 4.940833840455932e-06, "epoch": 0.8130081300813008, "percentage": 16.26, "elapsed_time": "1:27:45", "remaining_time": "7:31:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 600, "total_steps": 3075, "loss": 0.02, "learning_rate": 4.863863172170709e-06, "epoch": 0.975609756097561, "percentage": 19.51, "elapsed_time": "1:45:13", "remaining_time": "7:14:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 700, "total_steps": 3075, "loss": 0.0171, "learning_rate": 4.756453027584134e-06, "epoch": 1.1382113821138211, "percentage": 22.76, "elapsed_time": "2:02:42", "remaining_time": "6:56:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 800, "total_steps": 3075, "loss": 0.0173, "learning_rate": 4.619986527593033e-06, "epoch": 1.3008130081300813, "percentage": 26.02, "elapsed_time": "2:20:11", "remaining_time": "6:38:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 900, "total_steps": 3075, "loss": 0.0174, "learning_rate": 4.4562209519085615e-06, "epoch": 1.4634146341463414, "percentage": 29.27, "elapsed_time": "2:37:40", "remaining_time": "6:21:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 3075, "loss": 0.0169, "learning_rate": 4.26726511055776e-06, "epoch": 1.6260162601626016, "percentage": 32.52, "elapsed_time": "2:55:11", "remaining_time": "6:03:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 3075, "eval_loss": 0.017106017097830772, "epoch": 1.6260162601626016, "percentage": 32.52, "elapsed_time": "2:58:22", "remaining_time": "6:10:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1100, "total_steps": 3075, "loss": 0.0159, "learning_rate": 4.055552188727706e-06, "epoch": 1.7886178861788617, "percentage": 35.77, "elapsed_time": "3:17:32", "remaining_time": "5:54:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1200, "total_steps": 3075, "loss": 0.016, "learning_rate": 3.823808414629323e-06, "epoch": 1.951219512195122, "percentage": 39.02, "elapsed_time": "3:35:00", "remaining_time": "5:35:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1300, "total_steps": 3075, "loss": 0.0126, "learning_rate": 3.575017953844908e-06, "epoch": 2.113821138211382, "percentage": 42.28, "elapsed_time": "3:52:28", "remaining_time": "5:17:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1400, "total_steps": 3075, "loss": 0.0119, "learning_rate": 3.3123844822150126e-06, "epoch": 2.2764227642276422, "percentage": 45.53, "elapsed_time": "4:09:56", "remaining_time": "4:59:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1500, "total_steps": 3075, "loss": 0.0118, "learning_rate": 3.0392899320907716e-06, "epoch": 2.4390243902439024, "percentage": 48.78, "elapsed_time": "4:27:24", "remaining_time": "4:40:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1600, "total_steps": 3075, "loss": 0.0117, "learning_rate": 2.759250943176377e-06, "epoch": 2.6016260162601625, "percentage": 52.03, "elapsed_time": "4:44:54", "remaining_time": "4:22:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1700, "total_steps": 3075, "loss": 0.0116, "learning_rate": 2.4758735787443878e-06, "epoch": 2.7642276422764227, "percentage": 55.28, "elapsed_time": "5:02:24", "remaining_time": "4:04:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1800, "total_steps": 3075, "loss": 0.0111, "learning_rate": 2.192806890343352e-06, "epoch": 2.926829268292683, "percentage": 58.54, "elapsed_time": "5:19:52", "remaining_time": "3:46:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1900, "total_steps": 3075, "loss": 0.0077, "learning_rate": 1.9136959289452223e-06, "epoch": 3.089430894308943, "percentage": 61.79, "elapsed_time": "5:37:23", "remaining_time": "3:28:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2000, "total_steps": 3075, "loss": 0.0049, "learning_rate": 1.6421348076082123e-06, "epoch": 3.252032520325203, "percentage": 65.04, "elapsed_time": "5:54:51", "remaining_time": "3:10:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2000, "total_steps": 3075, "eval_loss": 0.023808766156435013, "epoch": 3.252032520325203, "percentage": 65.04, "elapsed_time": "5:58:02", "remaining_time": "3:12:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2100, "total_steps": 3075, "loss": 0.0052, "learning_rate": 1.3816204200673827e-06, "epoch": 3.4146341463414633, "percentage": 68.29, "elapsed_time": "6:17:20", "remaining_time": "2:55:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2200, "total_steps": 3075, "loss": 0.0048, "learning_rate": 1.1355074112188802e-06, "epoch": 3.5772357723577235, "percentage": 71.54, "elapsed_time": "6:34:51", "remaining_time": "2:37:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2300, "total_steps": 3075, "loss": 0.0046, "learning_rate": 9.069649793430869e-07, "epoch": 3.7398373983739837, "percentage": 74.8, "elapsed_time": "6:52:21", "remaining_time": "2:18:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2400, "total_steps": 3075, "loss": 0.0043, "learning_rate": 6.989360663246406e-07, "epoch": 3.902439024390244, "percentage": 78.05, "elapsed_time": "7:09:55", "remaining_time": "2:00:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2500, "total_steps": 3075, "loss": 0.003, "learning_rate": 5.1409946137705e-07, "epoch": 4.065040650406504, "percentage": 81.3, "elapsed_time": "7:27:26", "remaining_time": "1:42:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2600, "total_steps": 3075, "loss": 0.0007, "learning_rate": 3.548353062623949e-07, "epoch": 4.227642276422764, "percentage": 84.55, "elapsed_time": "7:44:58", "remaining_time": "1:24:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2700, "total_steps": 3075, "loss": 0.0007, "learning_rate": 2.231944461955507e-07, "epoch": 4.390243902439025, "percentage": 87.8, "elapsed_time": "8:02:29", "remaining_time": "1:07:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2800, "total_steps": 3075, "loss": 0.0007, "learning_rate": 1.2087202110147994e-07, "epoch": 4.5528455284552845, "percentage": 91.06, "elapsed_time": "8:20:02", "remaining_time": "0:49:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2900, "total_steps": 3075, "loss": 0.0006, "learning_rate": 4.9185637291078724e-08, "epoch": 4.715447154471545, "percentage": 94.31, "elapsed_time": "8:37:34", "remaining_time": "0:31:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3000, "total_steps": 3075, "loss": 0.0005, "learning_rate": 9.058400639009313e-09, "epoch": 4.878048780487805, "percentage": 97.56, "elapsed_time": "8:55:04", "remaining_time": "0:13:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3000, "total_steps": 3075, "eval_loss": 0.03251836076378822, "epoch": 4.878048780487805, "percentage": 97.56, "elapsed_time": "8:58:15", "remaining_time": "0:13:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3075, "total_steps": 3075, "epoch": 5.0, "percentage": 100.0, "elapsed_time": "9:15:24", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |