| {"current_steps": 5, "total_steps": 124, "loss": 5.4476, "learning_rate": 9.959935885253715e-06, "epoch": 0.08, "percentage": 4.03, "elapsed_time": "0:01:00", "remaining_time": "0:24:03", "throughput": 1361.14, "total_tokens": 82560} | |
| {"current_steps": 10, "total_steps": 124, "loss": 5.3325, "learning_rate": 9.840385594331022e-06, "epoch": 0.16, "percentage": 8.06, "elapsed_time": "0:02:00", "remaining_time": "0:22:50", "throughput": 1373.06, "total_tokens": 165120} | |
| {"current_steps": 15, "total_steps": 124, "loss": 5.0431, "learning_rate": 9.643264997861312e-06, "epoch": 0.24, "percentage": 12.1, "elapsed_time": "0:02:59", "remaining_time": "0:21:46", "throughput": 1377.8, "total_tokens": 247680} | |
| {"current_steps": 20, "total_steps": 124, "loss": 4.4316, "learning_rate": 9.371733080722911e-06, "epoch": 0.32, "percentage": 16.13, "elapsed_time": "0:03:59", "remaining_time": "0:20:44", "throughput": 1380.19, "total_tokens": 330240} | |
| {"current_steps": 25, "total_steps": 124, "loss": 3.253, "learning_rate": 9.030141317270026e-06, "epoch": 0.4, "percentage": 20.16, "elapsed_time": "0:04:58", "remaining_time": "0:19:43", "throughput": 1381.65, "total_tokens": 412800} | |
| {"current_steps": 30, "total_steps": 124, "loss": 3.1498, "learning_rate": 8.6239639361456e-06, "epoch": 0.48, "percentage": 24.19, "elapsed_time": "0:05:58", "remaining_time": "0:18:42", "throughput": 1382.46, "total_tokens": 495360} | |
| {"current_steps": 35, "total_steps": 124, "loss": 2.2606, "learning_rate": 8.15971019223152e-06, "epoch": 0.56, "percentage": 28.23, "elapsed_time": "0:06:57", "remaining_time": "0:17:42", "throughput": 1383.07, "total_tokens": 577920} | |
| {"current_steps": 40, "total_steps": 124, "loss": 1.9825, "learning_rate": 7.644820051634813e-06, "epoch": 0.64, "percentage": 32.26, "elapsed_time": "0:07:57", "remaining_time": "0:16:42", "throughput": 1383.63, "total_tokens": 660480} | |
| {"current_steps": 45, "total_steps": 124, "loss": 1.4754, "learning_rate": 7.087544961425317e-06, "epoch": 0.72, "percentage": 36.29, "elapsed_time": "0:08:56", "remaining_time": "0:15:42", "throughput": 1383.97, "total_tokens": 743040} | |
| {"current_steps": 50, "total_steps": 124, "loss": 0.9276, "learning_rate": 6.496815614866792e-06, "epoch": 0.8, "percentage": 40.32, "elapsed_time": "0:09:56", "remaining_time": "0:14:42", "throughput": 1384.16, "total_tokens": 825600} | |
| {"current_steps": 55, "total_steps": 124, "loss": 0.3203, "learning_rate": 5.882098831289044e-06, "epoch": 0.88, "percentage": 44.35, "elapsed_time": "0:10:55", "remaining_time": "0:13:42", "throughput": 1384.43, "total_tokens": 908160} | |
| {"current_steps": 60, "total_steps": 124, "loss": 0.3668, "learning_rate": 5.253245844193564e-06, "epoch": 0.96, "percentage": 48.39, "elapsed_time": "0:11:55", "remaining_time": "0:12:43", "throughput": 1384.62, "total_tokens": 990720} | |
| {"current_steps": 65, "total_steps": 124, "loss": 0.1251, "learning_rate": 4.62033442887377e-06, "epoch": 1.04, "percentage": 52.42, "elapsed_time": "0:12:55", "remaining_time": "0:11:43", "throughput": 1384.74, "total_tokens": 1073280} | |
| {"current_steps": 70, "total_steps": 124, "loss": 0.0937, "learning_rate": 3.993507399556699e-06, "epoch": 1.12, "percentage": 56.45, "elapsed_time": "0:13:54", "remaining_time": "0:10:43", "throughput": 1385.01, "total_tokens": 1155840} | |
| {"current_steps": 75, "total_steps": 124, "loss": 0.1095, "learning_rate": 3.3828100642538097e-06, "epoch": 1.2, "percentage": 60.48, "elapsed_time": "0:14:54", "remaining_time": "0:09:44", "throughput": 1385.13, "total_tokens": 1238400} | |
| {"current_steps": 80, "total_steps": 124, "loss": 0.0431, "learning_rate": 2.7980292422118282e-06, "epoch": 1.28, "percentage": 64.52, "elapsed_time": "0:15:53", "remaining_time": "0:08:44", "throughput": 1385.13, "total_tokens": 1320960} | |
| {"current_steps": 85, "total_steps": 124, "loss": 0.0373, "learning_rate": 2.2485364238130435e-06, "epoch": 1.3599999999999999, "percentage": 68.55, "elapsed_time": "0:16:53", "remaining_time": "0:07:44", "throughput": 1385.2, "total_tokens": 1403520} | |
| {"current_steps": 90, "total_steps": 124, "loss": 0.0501, "learning_rate": 1.74313758638889e-06, "epoch": 1.44, "percentage": 72.58, "elapsed_time": "0:17:52", "remaining_time": "0:06:45", "throughput": 1385.36, "total_tokens": 1486080} | |
| {"current_steps": 95, "total_steps": 124, "loss": 0.0309, "learning_rate": 1.2899320727454472e-06, "epoch": 1.52, "percentage": 76.61, "elapsed_time": "0:18:52", "remaining_time": "0:05:45", "throughput": 1385.43, "total_tokens": 1568640} | |
| {"current_steps": 100, "total_steps": 124, "loss": 0.03, "learning_rate": 8.961827939636198e-07, "epoch": 1.6, "percentage": 80.65, "elapsed_time": "0:19:51", "remaining_time": "0:04:46", "throughput": 1385.56, "total_tokens": 1651200} | |
| {"current_steps": 105, "total_steps": 124, "loss": 0.4213, "learning_rate": 5.681998365579594e-07, "epoch": 1.6800000000000002, "percentage": 84.68, "elapsed_time": "0:20:52", "remaining_time": "0:03:46", "throughput": 1384.69, "total_tokens": 1733760} | |
| {"current_steps": 110, "total_steps": 124, "loss": 0.0281, "learning_rate": 3.112393392645985e-07, "epoch": 1.76, "percentage": 88.71, "elapsed_time": "0:21:51", "remaining_time": "0:02:46", "throughput": 1384.94, "total_tokens": 1816320} | |
| {"current_steps": 115, "total_steps": 124, "loss": 0.0125, "learning_rate": 1.2941926002306536e-07, "epoch": 1.8399999999999999, "percentage": 92.74, "elapsed_time": "0:22:50", "remaining_time": "0:01:47", "throughput": 1385.14, "total_tokens": 1898880} | |
| {"current_steps": 120, "total_steps": 124, "loss": 0.0098, "learning_rate": 2.5653383040524228e-08, "epoch": 1.92, "percentage": 96.77, "elapsed_time": "0:23:50", "remaining_time": "0:00:47", "throughput": 1385.24, "total_tokens": 1981440} | |
| {"current_steps": 124, "total_steps": 124, "epoch": 1.984, "percentage": 100.0, "elapsed_time": "0:24:38", "remaining_time": "0:00:00", "throughput": 1384.39, "total_tokens": 2047488} | |