| {"current_steps": 5, "total_steps": 156, "loss": 0.678, "learning_rate": 2.9924022525939684e-05, "epoch": 0.032, "percentage": 3.21, "elapsed_time": "0:00:56", "remaining_time": "0:28:41", "throughput": 2874.51, "total_tokens": 163840} | |
| {"current_steps": 10, "total_steps": 156, "loss": 0.6551, "learning_rate": 2.9696859780634016e-05, "epoch": 0.064, "percentage": 6.41, "elapsed_time": "0:01:53", "remaining_time": "0:27:39", "throughput": 2882.14, "total_tokens": 327680} | |
| {"current_steps": 15, "total_steps": 156, "loss": 0.6372, "learning_rate": 2.9320812997628184e-05, "epoch": 0.096, "percentage": 9.62, "elapsed_time": "0:02:50", "remaining_time": "0:26:40", "throughput": 2885.94, "total_tokens": 491520} | |
| {"current_steps": 20, "total_steps": 156, "loss": 0.6201, "learning_rate": 2.8799691654882365e-05, "epoch": 0.128, "percentage": 12.82, "elapsed_time": "0:03:47", "remaining_time": "0:25:43", "throughput": 2886.5, "total_tokens": 655360} | |
| {"current_steps": 25, "total_steps": 156, "loss": 0.5965, "learning_rate": 2.8138774883503317e-05, "epoch": 0.16, "percentage": 16.03, "elapsed_time": "0:04:43", "remaining_time": "0:24:45", "throughput": 2889.67, "total_tokens": 819200} | |
| {"current_steps": 30, "total_steps": 156, "loss": 0.5959, "learning_rate": 2.7344757988404845e-05, "epoch": 0.192, "percentage": 19.23, "elapsed_time": "0:05:40", "remaining_time": "0:23:48", "throughput": 2889.27, "total_tokens": 983040} | |
| {"current_steps": 35, "total_steps": 156, "loss": 0.6006, "learning_rate": 2.6425684622660387e-05, "epoch": 0.224, "percentage": 22.44, "elapsed_time": "0:06:36", "remaining_time": "0:22:51", "throughput": 2891.08, "total_tokens": 1146880} | |
| {"current_steps": 40, "total_steps": 156, "loss": 0.594, "learning_rate": 2.5390865302643993e-05, "epoch": 0.256, "percentage": 25.64, "elapsed_time": "0:07:33", "remaining_time": "0:21:54", "throughput": 2891.9, "total_tokens": 1310720} | |
| {"current_steps": 45, "total_steps": 156, "loss": 0.5825, "learning_rate": 2.425078308942815e-05, "epoch": 0.288, "percentage": 28.85, "elapsed_time": "0:08:29", "remaining_time": "0:20:57", "throughput": 2892.01, "total_tokens": 1474560} | |
| {"current_steps": 50, "total_steps": 156, "loss": 0.5871, "learning_rate": 2.3016987391917016e-05, "epoch": 0.32, "percentage": 32.05, "elapsed_time": "0:09:26", "remaining_time": "0:20:01", "throughput": 2891.34, "total_tokens": 1638400} | |
| {"current_steps": 55, "total_steps": 156, "loss": 0.5771, "learning_rate": 2.1701976967524388e-05, "epoch": 0.352, "percentage": 35.26, "elapsed_time": "0:10:23", "remaining_time": "0:19:04", "throughput": 2890.51, "total_tokens": 1802240} | |
| {"current_steps": 60, "total_steps": 156, "loss": 0.5544, "learning_rate": 2.0319073305638035e-05, "epoch": 0.384, "percentage": 38.46, "elapsed_time": "0:11:19", "remaining_time": "0:18:07", "throughput": 2891.57, "total_tokens": 1966080} | |
| {"current_steps": 65, "total_steps": 156, "loss": 0.5768, "learning_rate": 1.888228567653781e-05, "epoch": 0.416, "percentage": 41.67, "elapsed_time": "0:12:16", "remaining_time": "0:17:11", "throughput": 2892.08, "total_tokens": 2129920} | |
| {"current_steps": 70, "total_steps": 156, "loss": 0.5534, "learning_rate": 1.7406169212866405e-05, "epoch": 0.448, "percentage": 44.87, "elapsed_time": "0:13:13", "remaining_time": "0:16:14", "throughput": 2891.55, "total_tokens": 2293760} | |
| {"current_steps": 75, "total_steps": 156, "loss": 0.5597, "learning_rate": 1.5905677461334292e-05, "epoch": 0.48, "percentage": 48.08, "elapsed_time": "0:14:09", "remaining_time": "0:15:17", "throughput": 2891.32, "total_tokens": 2457600} | |
| {"current_steps": 80, "total_steps": 156, "loss": 0.571, "learning_rate": 1.4396010898358778e-05, "epoch": 0.512, "percentage": 51.28, "elapsed_time": "0:15:06", "remaining_time": "0:14:21", "throughput": 2891.43, "total_tokens": 2621440} | |
| {"current_steps": 85, "total_steps": 156, "loss": 0.5572, "learning_rate": 1.2892462944223613e-05, "epoch": 0.544, "percentage": 54.49, "elapsed_time": "0:16:03", "remaining_time": "0:13:24", "throughput": 2891.18, "total_tokens": 2785280} | |
| {"current_steps": 90, "total_steps": 156, "loss": 0.5686, "learning_rate": 1.1410265035686639e-05, "epoch": 0.576, "percentage": 57.69, "elapsed_time": "0:16:59", "remaining_time": "0:12:27", "throughput": 2892.41, "total_tokens": 2949120} | |
| {"current_steps": 95, "total_steps": 156, "loss": 0.5525, "learning_rate": 9.964432326500933e-06, "epoch": 0.608, "percentage": 60.9, "elapsed_time": "0:17:56", "remaining_time": "0:11:31", "throughput": 2892.02, "total_tokens": 3112960} | |
| {"current_steps": 100, "total_steps": 156, "loss": 0.5594, "learning_rate": 8.569611578954186e-06, "epoch": 0.64, "percentage": 64.1, "elapsed_time": "0:18:53", "remaining_time": "0:10:34", "throughput": 2891.69, "total_tokens": 3276800} | |
| {"current_steps": 105, "total_steps": 156, "loss": 0.5709, "learning_rate": 7.239932787335147e-06, "epoch": 0.672, "percentage": 67.31, "elapsed_time": "0:19:50", "remaining_time": "0:09:38", "throughput": 2890.41, "total_tokens": 3440640} | |
| {"current_steps": 110, "total_steps": 156, "loss": 0.5555, "learning_rate": 5.988866036430314e-06, "epoch": 0.704, "percentage": 70.51, "elapsed_time": "0:20:46", "remaining_time": "0:08:41", "throughput": 2890.93, "total_tokens": 3604480} | |
| {"current_steps": 115, "total_steps": 156, "loss": 0.5597, "learning_rate": 4.829085045121636e-06, "epoch": 0.736, "percentage": 73.72, "elapsed_time": "0:21:43", "remaining_time": "0:07:44", "throughput": 2891.4, "total_tokens": 3768320} | |
| {"current_steps": 120, "total_steps": 156, "loss": 0.5432, "learning_rate": 3.772338777433482e-06, "epoch": 0.768, "percentage": 76.92, "elapsed_time": "0:22:40", "remaining_time": "0:06:48", "throughput": 2890.92, "total_tokens": 3932160} | |
| {"current_steps": 125, "total_steps": 156, "loss": 0.5363, "learning_rate": 2.829332421651404e-06, "epoch": 0.8, "percentage": 80.13, "elapsed_time": "0:23:36", "remaining_time": "0:05:51", "throughput": 2891.26, "total_tokens": 4096000} | |
| {"current_steps": 130, "total_steps": 156, "loss": 0.5582, "learning_rate": 2.0096189432334194e-06, "epoch": 0.832, "percentage": 83.33, "elapsed_time": "0:24:33", "remaining_time": "0:04:54", "throughput": 2891.48, "total_tokens": 4259840} | |
| {"current_steps": 135, "total_steps": 156, "loss": 0.5539, "learning_rate": 1.321502310118649e-06, "epoch": 0.864, "percentage": 86.54, "elapsed_time": "0:25:29", "remaining_time": "0:03:57", "throughput": 2891.37, "total_tokens": 4423680} | |
| {"current_steps": 140, "total_steps": 156, "loss": 0.5529, "learning_rate": 7.719533707928178e-07, "epoch": 0.896, "percentage": 89.74, "elapsed_time": "0:26:26", "remaining_time": "0:03:01", "throughput": 2891.6, "total_tokens": 4587520} | |
| {"current_steps": 145, "total_steps": 156, "loss": 0.5584, "learning_rate": 3.665392372935922e-07, "epoch": 0.928, "percentage": 92.95, "elapsed_time": "0:27:23", "remaining_time": "0:02:04", "throughput": 2891.86, "total_tokens": 4751360} | |
| {"current_steps": 150, "total_steps": 156, "loss": 0.5503, "learning_rate": 1.0936688852919042e-07, "epoch": 0.96, "percentage": 96.15, "elapsed_time": "0:28:19", "remaining_time": "0:01:07", "throughput": 2891.99, "total_tokens": 4915200} | |
| {"current_steps": 155, "total_steps": 156, "loss": 0.5549, "learning_rate": 3.0415652272480776e-09, "epoch": 0.992, "percentage": 99.36, "elapsed_time": "0:29:16", "remaining_time": "0:00:11", "throughput": 2892.02, "total_tokens": 5079040} | |
| {"current_steps": 156, "total_steps": 156, "epoch": 0.9984, "percentage": 100.0, "elapsed_time": "0:29:28", "remaining_time": "0:00:00", "throughput": 2890.82, "total_tokens": 5111808} | |