| {"current_steps": 5, "total_steps": 1800, "loss": 0.4212, "learning_rate": 4.999904807660428e-05, "epoch": 0.008333333333333333, "percentage": 0.28, "elapsed_time": "0:01:34", "remaining_time": "9:25:43", "throughput": "317.96", "total_tokens": 30064} | |
| {"current_steps": 10, "total_steps": 1800, "loss": 0.2803, "learning_rate": 4.9996192378909786e-05, "epoch": 0.016666666666666666, "percentage": 0.56, "elapsed_time": "0:03:06", "remaining_time": "9:16:17", "throughput": "307.70", "total_tokens": 57376} | |
| {"current_steps": 15, "total_steps": 1800, "loss": 0.1484, "learning_rate": 4.999143312438893e-05, "epoch": 0.025, "percentage": 0.83, "elapsed_time": "0:04:41", "remaining_time": "9:18:46", "throughput": "305.82", "total_tokens": 86160} | |
| {"current_steps": 20, "total_steps": 1800, "loss": 0.1151, "learning_rate": 4.99847706754774e-05, "epoch": 0.03333333333333333, "percentage": 1.11, "elapsed_time": "0:06:11", "remaining_time": "9:11:09", "throughput": "297.64", "total_tokens": 110592} | |
| {"current_steps": 25, "total_steps": 1800, "loss": 0.0751, "learning_rate": 4.997620553954645e-05, "epoch": 0.041666666666666664, "percentage": 1.39, "elapsed_time": "0:07:46", "remaining_time": "9:12:11", "throughput": "297.62", "total_tokens": 138880} | |
| {"current_steps": 30, "total_steps": 1800, "loss": 0.0471, "learning_rate": 4.996573836886435e-05, "epoch": 0.05, "percentage": 1.67, "elapsed_time": "0:09:24", "remaining_time": "9:15:14", "throughput": "300.36", "total_tokens": 169600} | |
| {"current_steps": 35, "total_steps": 1800, "loss": 0.0368, "learning_rate": 4.9953369960546676e-05, "epoch": 0.058333333333333334, "percentage": 1.94, "elapsed_time": "0:11:04", "remaining_time": "9:18:40", "throughput": "303.39", "total_tokens": 201664} | |
| {"current_steps": 40, "total_steps": 1800, "loss": 0.0311, "learning_rate": 4.993910125649561e-05, "epoch": 0.06666666666666667, "percentage": 2.22, "elapsed_time": "0:12:39", "remaining_time": "9:17:14", "throughput": "302.60", "total_tokens": 229936} | |
| {"current_steps": 45, "total_steps": 1800, "loss": 0.0262, "learning_rate": 4.99229333433282e-05, "epoch": 0.075, "percentage": 2.5, "elapsed_time": "0:14:15", "remaining_time": "9:16:17", "throughput": "302.73", "total_tokens": 259088} | |
| {"current_steps": 50, "total_steps": 1800, "loss": 0.0244, "learning_rate": 4.990486745229364e-05, "epoch": 0.08333333333333333, "percentage": 2.78, "elapsed_time": "0:15:53", "remaining_time": "9:16:27", "throughput": "303.83", "total_tokens": 289824} | |
| {"current_steps": 55, "total_steps": 1800, "loss": 0.0266, "learning_rate": 4.988490495917947e-05, "epoch": 0.09166666666666666, "percentage": 3.06, "elapsed_time": "0:17:28", "remaining_time": "9:14:23", "throughput": "303.09", "total_tokens": 317760} | |
| {"current_steps": 60, "total_steps": 1800, "loss": 0.0283, "learning_rate": 4.9867570427929354e-05, "epoch": 0.1, "percentage": 3.33, "elapsed_time": "0:19:03", "remaining_time": "9:12:47", "throughput": "300.01", "total_tokens": 343120} | |
| {"current_steps": 65, "total_steps": 1800, "loss": 0.0265, "learning_rate": 4.984419797901491e-05, "epoch": 0.10833333333333334, "percentage": 3.61, "elapsed_time": "0:20:38", "remaining_time": "9:11:06", "throughput": "299.96", "total_tokens": 371600} | |
| {"current_steps": 70, "total_steps": 1800, "loss": 0.0235, "learning_rate": 4.981893354823614e-05, "epoch": 0.11666666666666667, "percentage": 3.89, "elapsed_time": "0:22:11", "remaining_time": "9:08:34", "throughput": "299.39", "total_tokens": 398720} | |
| {"current_steps": 75, "total_steps": 1800, "loss": 0.0229, "learning_rate": 4.979177905957726e-05, "epoch": 0.125, "percentage": 4.17, "elapsed_time": "0:23:45", "remaining_time": "9:06:24", "throughput": "299.15", "total_tokens": 426416} | |
| {"current_steps": 80, "total_steps": 1800, "loss": 0.0238, "learning_rate": 4.976273658095772e-05, "epoch": 0.13333333333333333, "percentage": 4.44, "elapsed_time": "0:25:18", "remaining_time": "9:04:05", "throughput": "298.31", "total_tokens": 452944} | |
| {"current_steps": 85, "total_steps": 1800, "loss": 0.0208, "learning_rate": 4.9731808324074717e-05, "epoch": 0.14166666666666666, "percentage": 4.72, "elapsed_time": "0:26:52", "remaining_time": "9:02:13", "throughput": "298.24", "total_tokens": 480896} | |
| {"current_steps": 90, "total_steps": 1800, "loss": 0.0191, "learning_rate": 4.969899664423473e-05, "epoch": 0.15, "percentage": 5.0, "elapsed_time": "0:28:28", "remaining_time": "9:01:00", "throughput": "298.57", "total_tokens": 510096} | |
| {"current_steps": 95, "total_steps": 1800, "loss": 0.0205, "learning_rate": 4.966430404017424e-05, "epoch": 0.15833333333333333, "percentage": 5.28, "elapsed_time": "0:30:00", "remaining_time": "8:58:40", "throughput": "298.01", "total_tokens": 536672} | |
| {"current_steps": 100, "total_steps": 1800, "loss": 0.0193, "learning_rate": 4.962773315386935e-05, "epoch": 0.16666666666666666, "percentage": 5.56, "elapsed_time": "0:31:37", "remaining_time": "8:57:37", "throughput": "298.56", "total_tokens": 566528} | |
| {"current_steps": 105, "total_steps": 1800, "loss": 0.024, "learning_rate": 4.9589286770334654e-05, "epoch": 0.175, "percentage": 5.83, "elapsed_time": "0:33:11", "remaining_time": "8:55:41", "throughput": "298.12", "total_tokens": 593584} | |
| {"current_steps": 110, "total_steps": 1800, "loss": 0.0206, "learning_rate": 4.954896781741109e-05, "epoch": 0.18333333333333332, "percentage": 6.11, "elapsed_time": "0:34:44", "remaining_time": "8:53:47", "throughput": "297.89", "total_tokens": 620992} | |
| {"current_steps": 115, "total_steps": 1800, "loss": 0.0188, "learning_rate": 4.9506779365543046e-05, "epoch": 0.19166666666666668, "percentage": 6.39, "elapsed_time": "0:36:17", "remaining_time": "8:51:38", "throughput": "297.43", "total_tokens": 647504} | |
| {"current_steps": 120, "total_steps": 1800, "loss": 0.0183, "learning_rate": 4.9462724627544466e-05, "epoch": 0.2, "percentage": 6.67, "elapsed_time": "0:37:52", "remaining_time": "8:50:12", "throughput": "297.64", "total_tokens": 676336} | |
| {"current_steps": 125, "total_steps": 1800, "loss": 0.0189, "learning_rate": 4.94168069583542e-05, "epoch": 0.20833333333333334, "percentage": 6.94, "elapsed_time": "0:39:21", "remaining_time": "8:47:22", "throughput": "297.06", "total_tokens": 701488} | |
| {"current_steps": 130, "total_steps": 1800, "loss": 0.0213, "learning_rate": 4.936902985478055e-05, "epoch": 0.21666666666666667, "percentage": 7.22, "elapsed_time": "0:40:57", "remaining_time": "8:46:08", "throughput": "297.53", "total_tokens": 731152} | |
| {"current_steps": 135, "total_steps": 1800, "loss": 0.0214, "learning_rate": 4.931939695523492e-05, "epoch": 0.225, "percentage": 7.5, "elapsed_time": "0:42:29", "remaining_time": "8:44:04", "throughput": "297.22", "total_tokens": 757792} | |
| {"current_steps": 140, "total_steps": 1800, "loss": 0.0214, "learning_rate": 4.926791203945477e-05, "epoch": 0.23333333333333334, "percentage": 7.78, "elapsed_time": "0:44:04", "remaining_time": "8:42:39", "throughput": "297.22", "total_tokens": 786064} | |
| {"current_steps": 145, "total_steps": 1800, "loss": 0.0203, "learning_rate": 4.9214579028215776e-05, "epoch": 0.24166666666666667, "percentage": 8.06, "elapsed_time": "0:45:39", "remaining_time": "8:41:09", "throughput": "297.36", "total_tokens": 814640} | |
| {"current_steps": 150, "total_steps": 1800, "loss": 0.0194, "learning_rate": 4.915940198303324e-05, "epoch": 0.25, "percentage": 8.33, "elapsed_time": "0:47:13", "remaining_time": "8:39:32", "throughput": "297.35", "total_tokens": 842656} | |
| {"current_steps": 155, "total_steps": 1800, "loss": 0.0158, "learning_rate": 4.910238510585276e-05, "epoch": 0.25833333333333336, "percentage": 8.61, "elapsed_time": "0:48:44", "remaining_time": "8:37:19", "throughput": "296.96", "total_tokens": 868512} | |
| {"current_steps": 160, "total_steps": 1800, "loss": 0.0172, "learning_rate": 4.9043532738730284e-05, "epoch": 0.26666666666666666, "percentage": 8.89, "elapsed_time": "0:50:18", "remaining_time": "8:35:43", "throughput": "296.94", "total_tokens": 896416} | |
| {"current_steps": 165, "total_steps": 1800, "loss": 0.0193, "learning_rate": 4.898284936350144e-05, "epoch": 0.275, "percentage": 9.17, "elapsed_time": "0:51:52", "remaining_time": "8:34:06", "throughput": "297.06", "total_tokens": 924720} | |
| {"current_steps": 170, "total_steps": 1800, "loss": 0.0189, "learning_rate": 4.892033960144019e-05, "epoch": 0.2833333333333333, "percentage": 9.44, "elapsed_time": "0:53:30", "remaining_time": "8:33:05", "throughput": "297.62", "total_tokens": 955584} | |
| {"current_steps": 175, "total_steps": 1800, "loss": 0.019, "learning_rate": 4.8856008212906925e-05, "epoch": 0.2916666666666667, "percentage": 9.72, "elapsed_time": "0:55:06", "remaining_time": "8:31:43", "throughput": "297.91", "total_tokens": 985072} | |
| {"current_steps": 180, "total_steps": 1800, "loss": 0.0168, "learning_rate": 4.878986009698596e-05, "epoch": 0.3, "percentage": 10.0, "elapsed_time": "0:56:40", "remaining_time": "8:30:03", "throughput": "297.99", "total_tokens": 1013296} | |
| {"current_steps": 185, "total_steps": 1800, "loss": 0.0194, "learning_rate": 4.8721900291112415e-05, "epoch": 0.30833333333333335, "percentage": 10.28, "elapsed_time": "0:58:17", "remaining_time": "8:28:53", "throughput": "298.39", "total_tokens": 1043664} | |
| {"current_steps": 190, "total_steps": 1800, "loss": 0.0155, "learning_rate": 4.8652133970688636e-05, "epoch": 0.31666666666666665, "percentage": 10.56, "elapsed_time": "0:59:50", "remaining_time": "8:27:07", "throughput": "298.24", "total_tokens": 1070912} | |
| {"current_steps": 195, "total_steps": 1800, "loss": 0.0193, "learning_rate": 4.858056644869002e-05, "epoch": 0.325, "percentage": 10.83, "elapsed_time": "1:01:22", "remaining_time": "8:25:12", "throughput": "298.10", "total_tokens": 1097856} | |
| {"current_steps": 200, "total_steps": 1800, "loss": 0.0164, "learning_rate": 4.850720317526047e-05, "epoch": 0.3333333333333333, "percentage": 11.11, "elapsed_time": "1:02:57", "remaining_time": "8:23:39", "throughput": "298.12", "total_tokens": 1126128} | |
| {"current_steps": 205, "total_steps": 1800, "loss": 0.014, "learning_rate": 4.843204973729729e-05, "epoch": 0.3416666666666667, "percentage": 11.39, "elapsed_time": "1:04:35", "remaining_time": "8:22:36", "throughput": "298.66", "total_tokens": 1157584} | |
| {"current_steps": 210, "total_steps": 1800, "loss": 0.0154, "learning_rate": 4.835511185802574e-05, "epoch": 0.35, "percentage": 11.67, "elapsed_time": "1:06:08", "remaining_time": "8:20:45", "throughput": "298.57", "total_tokens": 1184800} | |
| {"current_steps": 215, "total_steps": 1800, "loss": 0.0147, "learning_rate": 4.827639539656321e-05, "epoch": 0.35833333333333334, "percentage": 11.94, "elapsed_time": "1:07:42", "remaining_time": "8:19:07", "throughput": "298.63", "total_tokens": 1213104} | |
| {"current_steps": 220, "total_steps": 1800, "loss": 0.0185, "learning_rate": 4.8195906347473e-05, "epoch": 0.36666666666666664, "percentage": 12.22, "elapsed_time": "1:09:17", "remaining_time": "8:17:35", "throughput": "298.72", "total_tokens": 1241808} | |
| {"current_steps": 225, "total_steps": 1800, "loss": 0.016, "learning_rate": 4.8113650840307834e-05, "epoch": 0.375, "percentage": 12.5, "elapsed_time": "1:10:51", "remaining_time": "8:16:01", "throughput": "298.74", "total_tokens": 1270160} | |
| {"current_steps": 230, "total_steps": 1800, "loss": 0.0168, "learning_rate": 4.802963513914304e-05, "epoch": 0.38333333333333336, "percentage": 12.78, "elapsed_time": "1:12:24", "remaining_time": "8:14:14", "throughput": "298.65", "total_tokens": 1297456} | |
| {"current_steps": 235, "total_steps": 1800, "loss": 0.0154, "learning_rate": 4.794386564209953e-05, "epoch": 0.39166666666666666, "percentage": 13.06, "elapsed_time": "1:13:55", "remaining_time": "8:12:16", "throughput": "298.39", "total_tokens": 1323392} | |
| {"current_steps": 240, "total_steps": 1800, "loss": 0.0149, "learning_rate": 4.7856348880856595e-05, "epoch": 0.4, "percentage": 13.33, "elapsed_time": "1:15:38", "remaining_time": "8:11:43", "throughput": "299.32", "total_tokens": 1358576} | |
| {"current_steps": 242, "total_steps": 1800, "epoch": 0.4033333333333333, "percentage": 13.44, "elapsed_time": "1:16:23", "remaining_time": "8:11:46", "throughput": "299.27", "total_tokens": 1371584} | |