| {"current_steps": 5, "total_steps": 870, "loss": 1.2873, "learning_rate": 5.747126436781609e-07, "epoch": 0.017241379310344827, "percentage": 0.57, "elapsed_time": "0:01:05", "remaining_time": "3:08:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 10, "total_steps": 870, "loss": 1.3464, "learning_rate": 1.1494252873563219e-06, "epoch": 0.034482758620689655, "percentage": 1.15, "elapsed_time": "0:02:02", "remaining_time": "2:55:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 15, "total_steps": 870, "loss": 1.2482, "learning_rate": 1.724137931034483e-06, "epoch": 0.05172413793103448, "percentage": 1.72, "elapsed_time": "0:02:57", "remaining_time": "2:48:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 20, "total_steps": 870, "loss": 1.2549, "learning_rate": 2.2988505747126437e-06, "epoch": 0.06896551724137931, "percentage": 2.3, "elapsed_time": "0:03:57", "remaining_time": "2:48:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 25, "total_steps": 870, "loss": 1.1964, "learning_rate": 2.8735632183908046e-06, "epoch": 0.08620689655172414, "percentage": 2.87, "elapsed_time": "0:04:58", "remaining_time": "2:48:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 30, "total_steps": 870, "loss": 1.1774, "learning_rate": 3.448275862068966e-06, "epoch": 0.10344827586206896, "percentage": 3.45, "elapsed_time": "0:05:54", "remaining_time": "2:45:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 35, "total_steps": 870, "loss": 1.1393, "learning_rate": 4.022988505747127e-06, "epoch": 0.1206896551724138, "percentage": 4.02, "elapsed_time": "0:06:48", "remaining_time": "2:42:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 40, "total_steps": 870, "loss": 1.1998, "learning_rate": 4.5977011494252875e-06, "epoch": 0.13793103448275862, "percentage": 4.6, "elapsed_time": "0:07:42", "remaining_time": "2:39:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 45, "total_steps": 870, "loss": 1.205, "learning_rate": 5.172413793103449e-06, "epoch": 0.15517241379310345, "percentage": 5.17, "elapsed_time": "0:08:38", "remaining_time": "2:38:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 50, "total_steps": 870, "loss": 1.1961, "learning_rate": 5.747126436781609e-06, "epoch": 0.1724137931034483, "percentage": 5.75, "elapsed_time": "0:09:32", "remaining_time": "2:36:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 55, "total_steps": 870, "loss": 1.1906, "learning_rate": 6.321839080459771e-06, "epoch": 0.1896551724137931, "percentage": 6.32, "elapsed_time": "0:10:27", "remaining_time": "2:34:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 60, "total_steps": 870, "loss": 1.1844, "learning_rate": 6.896551724137932e-06, "epoch": 0.20689655172413793, "percentage": 6.9, "elapsed_time": "0:11:26", "remaining_time": "2:34:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 65, "total_steps": 870, "loss": 1.1145, "learning_rate": 7.4712643678160925e-06, "epoch": 0.22413793103448276, "percentage": 7.47, "elapsed_time": "0:12:27", "remaining_time": "2:34:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 70, "total_steps": 870, "loss": 1.1685, "learning_rate": 8.045977011494253e-06, "epoch": 0.2413793103448276, "percentage": 8.05, "elapsed_time": "0:13:23", "remaining_time": "2:32:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 75, "total_steps": 870, "loss": 1.2057, "learning_rate": 8.620689655172414e-06, "epoch": 0.25862068965517243, "percentage": 8.62, "elapsed_time": "0:14:21", "remaining_time": "2:32:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 80, "total_steps": 870, "loss": 1.1636, "learning_rate": 9.195402298850575e-06, "epoch": 0.27586206896551724, "percentage": 9.2, "elapsed_time": "0:15:24", "remaining_time": "2:32:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 85, "total_steps": 870, "loss": 1.1612, "learning_rate": 9.770114942528738e-06, "epoch": 0.29310344827586204, "percentage": 9.77, "elapsed_time": "0:16:25", "remaining_time": "2:31:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 90, "total_steps": 870, "loss": 1.1645, "learning_rate": 9.999637795788383e-06, "epoch": 0.3103448275862069, "percentage": 10.34, "elapsed_time": "0:17:23", "remaining_time": "2:30:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 95, "total_steps": 870, "loss": 1.1547, "learning_rate": 9.997424515642709e-06, "epoch": 0.3275862068965517, "percentage": 10.92, "elapsed_time": "0:18:20", "remaining_time": "2:29:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 100, "total_steps": 870, "loss": 1.1738, "learning_rate": 9.993200069547117e-06, "epoch": 0.3448275862068966, "percentage": 11.49, "elapsed_time": "0:19:20", "remaining_time": "2:28:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 105, "total_steps": 870, "loss": 1.1797, "learning_rate": 9.986966157589751e-06, "epoch": 0.3620689655172414, "percentage": 12.07, "elapsed_time": "0:20:17", "remaining_time": "2:27:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 110, "total_steps": 870, "loss": 1.2318, "learning_rate": 9.978725288549161e-06, "epoch": 0.3793103448275862, "percentage": 12.64, "elapsed_time": "0:21:17", "remaining_time": "2:27:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 115, "total_steps": 870, "loss": 1.205, "learning_rate": 9.968480778884693e-06, "epoch": 0.39655172413793105, "percentage": 13.22, "elapsed_time": "0:22:13", "remaining_time": "2:25:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 120, "total_steps": 870, "loss": 1.2033, "learning_rate": 9.95623675140179e-06, "epoch": 0.41379310344827586, "percentage": 13.79, "elapsed_time": "0:23:11", "remaining_time": "2:24:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 125, "total_steps": 870, "loss": 1.1718, "learning_rate": 9.941998133592825e-06, "epoch": 0.43103448275862066, "percentage": 14.37, "elapsed_time": "0:24:14", "remaining_time": "2:24:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 130, "total_steps": 870, "loss": 1.1035, "learning_rate": 9.925770655654061e-06, "epoch": 0.4482758620689655, "percentage": 14.94, "elapsed_time": "0:25:12", "remaining_time": "2:23:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 135, "total_steps": 870, "loss": 1.1374, "learning_rate": 9.907560848179607e-06, "epoch": 0.46551724137931033, "percentage": 15.52, "elapsed_time": "0:26:08", "remaining_time": "2:22:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 140, "total_steps": 870, "loss": 1.2123, "learning_rate": 9.887376039533227e-06, "epoch": 0.4827586206896552, "percentage": 16.09, "elapsed_time": "0:27:05", "remaining_time": "2:21:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 145, "total_steps": 870, "loss": 1.2104, "learning_rate": 9.86522435289912e-06, "epoch": 0.5, "percentage": 16.67, "elapsed_time": "0:27:58", "remaining_time": "2:19:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 150, "total_steps": 870, "loss": 1.1937, "learning_rate": 9.841114703012817e-06, "epoch": 0.5172413793103449, "percentage": 17.24, "elapsed_time": "0:28:55", "remaining_time": "2:18:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 155, "total_steps": 870, "loss": 1.2122, "learning_rate": 9.815056792573531e-06, "epoch": 0.5344827586206896, "percentage": 17.82, "elapsed_time": "0:29:56", "remaining_time": "2:18:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 160, "total_steps": 870, "loss": 1.2088, "learning_rate": 9.7870611083394e-06, "epoch": 0.5517241379310345, "percentage": 18.39, "elapsed_time": "0:30:55", "remaining_time": "2:17:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 165, "total_steps": 870, "loss": 1.1936, "learning_rate": 9.757138916907184e-06, "epoch": 0.5689655172413793, "percentage": 18.97, "elapsed_time": "0:31:57", "remaining_time": "2:16:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 170, "total_steps": 870, "loss": 1.1571, "learning_rate": 9.725302260178145e-06, "epoch": 0.5862068965517241, "percentage": 19.54, "elapsed_time": "0:32:50", "remaining_time": "2:15:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 175, "total_steps": 870, "loss": 1.1808, "learning_rate": 9.69156395051188e-06, "epoch": 0.603448275862069, "percentage": 20.11, "elapsed_time": "0:33:46", "remaining_time": "2:14:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 180, "total_steps": 870, "loss": 1.1737, "learning_rate": 9.655937565570124e-06, "epoch": 0.6206896551724138, "percentage": 20.69, "elapsed_time": "0:34:44", "remaining_time": "2:13:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 185, "total_steps": 870, "loss": 1.1372, "learning_rate": 9.618437442852539e-06, "epoch": 0.6379310344827587, "percentage": 21.26, "elapsed_time": "0:35:45", "remaining_time": "2:12:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 190, "total_steps": 870, "loss": 1.188, "learning_rate": 9.579078673926729e-06, "epoch": 0.6551724137931034, "percentage": 21.84, "elapsed_time": "0:36:37", "remaining_time": "2:11:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 195, "total_steps": 870, "loss": 1.1737, "learning_rate": 9.537877098354787e-06, "epoch": 0.6724137931034483, "percentage": 22.41, "elapsed_time": "0:37:35", "remaining_time": "2:10:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 200, "total_steps": 870, "loss": 1.1371, "learning_rate": 9.494849297318795e-06, "epoch": 0.6896551724137931, "percentage": 22.99, "elapsed_time": "0:38:31", "remaining_time": "2:09:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 205, "total_steps": 870, "loss": 1.1921, "learning_rate": 9.450012586947912e-06, "epoch": 0.7068965517241379, "percentage": 23.56, "elapsed_time": "0:41:31", "remaining_time": "2:14:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 210, "total_steps": 870, "loss": 1.1577, "learning_rate": 9.40338501134964e-06, "epoch": 0.7241379310344828, "percentage": 24.14, "elapsed_time": "0:42:27", "remaining_time": "2:13:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 215, "total_steps": 870, "loss": 1.0958, "learning_rate": 9.354985335348155e-06, "epoch": 0.7413793103448276, "percentage": 24.71, "elapsed_time": "0:43:21", "remaining_time": "2:12:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 220, "total_steps": 870, "loss": 1.1759, "learning_rate": 9.30483303693258e-06, "epoch": 0.7586206896551724, "percentage": 25.29, "elapsed_time": "0:44:22", "remaining_time": "2:11:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 225, "total_steps": 870, "loss": 1.187, "learning_rate": 9.252948299418255e-06, "epoch": 0.7758620689655172, "percentage": 25.86, "elapsed_time": "0:45:16", "remaining_time": "2:09:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 230, "total_steps": 870, "loss": 1.1741, "learning_rate": 9.199352003324151e-06, "epoch": 0.7931034482758621, "percentage": 26.44, "elapsed_time": "0:46:14", "remaining_time": "2:08:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 235, "total_steps": 870, "loss": 1.14, "learning_rate": 9.144065717969707e-06, "epoch": 0.8103448275862069, "percentage": 27.01, "elapsed_time": "0:47:07", "remaining_time": "2:07:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 240, "total_steps": 870, "loss": 1.2098, "learning_rate": 9.08711169279446e-06, "epoch": 0.8275862068965517, "percentage": 27.59, "elapsed_time": "0:48:04", "remaining_time": "2:06:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 245, "total_steps": 870, "loss": 1.1585, "learning_rate": 9.028512848403971e-06, "epoch": 0.8448275862068966, "percentage": 28.16, "elapsed_time": "0:48:59", "remaining_time": "2:04:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 250, "total_steps": 870, "loss": 1.199, "learning_rate": 8.968292767345646e-06, "epoch": 0.8620689655172413, "percentage": 28.74, "elapsed_time": "0:49:56", "remaining_time": "2:03:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 255, "total_steps": 870, "loss": 1.1734, "learning_rate": 8.90647568461816e-06, "epoch": 0.8793103448275862, "percentage": 29.31, "elapsed_time": "0:50:51", "remaining_time": "2:02:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 260, "total_steps": 870, "loss": 1.1296, "learning_rate": 8.843086477918317e-06, "epoch": 0.896551724137931, "percentage": 29.89, "elapsed_time": "0:51:50", "remaining_time": "2:01:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 265, "total_steps": 870, "loss": 1.1756, "learning_rate": 8.778150657629258e-06, "epoch": 0.9137931034482759, "percentage": 30.46, "elapsed_time": "0:52:52", "remaining_time": "2:00:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 270, "total_steps": 870, "loss": 1.159, "learning_rate": 8.71169435655405e-06, "epoch": 0.9310344827586207, "percentage": 31.03, "elapsed_time": "0:53:47", "remaining_time": "1:59:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 275, "total_steps": 870, "loss": 1.1451, "learning_rate": 8.643744319398781e-06, "epoch": 0.9482758620689655, "percentage": 31.61, "elapsed_time": "0:54:43", "remaining_time": "1:58:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 280, "total_steps": 870, "loss": 1.1701, "learning_rate": 8.574327892009415e-06, "epoch": 0.9655172413793104, "percentage": 32.18, "elapsed_time": "0:55:38", "remaining_time": "1:57:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 285, "total_steps": 870, "loss": 1.175, "learning_rate": 8.503473010366713e-06, "epoch": 0.9827586206896551, "percentage": 32.76, "elapsed_time": "0:56:32", "remaining_time": "1:56:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 290, "total_steps": 870, "loss": 1.1198, "learning_rate": 8.43120818934367e-06, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "0:57:29", "remaining_time": "1:54:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 295, "total_steps": 870, "loss": 0.7509, "learning_rate": 8.357562511229961e-06, "epoch": 1.0172413793103448, "percentage": 33.91, "elapsed_time": "0:58:31", "remaining_time": "1:54:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 300, "total_steps": 870, "loss": 0.6958, "learning_rate": 8.282565614028068e-06, "epoch": 1.0344827586206897, "percentage": 34.48, "elapsed_time": "0:59:27", "remaining_time": "1:52:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 305, "total_steps": 870, "loss": 0.6641, "learning_rate": 8.206247679525736e-06, "epoch": 1.0517241379310345, "percentage": 35.06, "elapsed_time": "1:00:24", "remaining_time": "1:51:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 310, "total_steps": 870, "loss": 0.6752, "learning_rate": 8.1286394211496e-06, "epoch": 1.0689655172413792, "percentage": 35.63, "elapsed_time": "1:01:19", "remaining_time": "1:50:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 315, "total_steps": 870, "loss": 0.6926, "learning_rate": 8.049772071604864e-06, "epoch": 1.0862068965517242, "percentage": 36.21, "elapsed_time": "1:02:10", "remaining_time": "1:49:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 320, "total_steps": 870, "loss": 0.6654, "learning_rate": 7.969677370306e-06, "epoch": 1.103448275862069, "percentage": 36.78, "elapsed_time": "1:03:08", "remaining_time": "1:48:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 325, "total_steps": 870, "loss": 0.6767, "learning_rate": 7.888387550603505e-06, "epoch": 1.1206896551724137, "percentage": 37.36, "elapsed_time": "1:04:16", "remaining_time": "1:47:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 330, "total_steps": 870, "loss": 0.6968, "learning_rate": 7.805935326811913e-06, "epoch": 1.1379310344827587, "percentage": 37.93, "elapsed_time": "1:05:14", "remaining_time": "1:46:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 335, "total_steps": 870, "loss": 0.6625, "learning_rate": 7.722353881044223e-06, "epoch": 1.1551724137931034, "percentage": 38.51, "elapsed_time": "1:06:21", "remaining_time": "1:45:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 340, "total_steps": 870, "loss": 0.7104, "learning_rate": 7.637676849858077e-06, "epoch": 1.1724137931034484, "percentage": 39.08, "elapsed_time": "1:07:24", "remaining_time": "1:45:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 345, "total_steps": 870, "loss": 0.6944, "learning_rate": 7.551938310719043e-06, "epoch": 1.1896551724137931, "percentage": 39.66, "elapsed_time": "1:08:28", "remaining_time": "1:44:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 350, "total_steps": 870, "loss": 0.6516, "learning_rate": 7.465172768286463e-06, "epoch": 1.206896551724138, "percentage": 40.23, "elapsed_time": "1:09:28", "remaining_time": "1:43:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 355, "total_steps": 870, "loss": 0.6446, "learning_rate": 7.377415140527388e-06, "epoch": 1.2241379310344827, "percentage": 40.8, "elapsed_time": "1:10:26", "remaining_time": "1:42:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 360, "total_steps": 870, "loss": 0.679, "learning_rate": 7.288700744664167e-06, "epoch": 1.2413793103448276, "percentage": 41.38, "elapsed_time": "1:11:26", "remaining_time": "1:41:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 365, "total_steps": 870, "loss": 0.7054, "learning_rate": 7.199065282961372e-06, "epoch": 1.2586206896551724, "percentage": 41.95, "elapsed_time": "1:12:24", "remaining_time": "1:40:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 370, "total_steps": 870, "loss": 0.6966, "learning_rate": 7.1085448283577556e-06, "epoch": 1.2758620689655173, "percentage": 42.53, "elapsed_time": "1:13:21", "remaining_time": "1:39:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 375, "total_steps": 870, "loss": 0.678, "learning_rate": 7.017175809949044e-06, "epoch": 1.293103448275862, "percentage": 43.1, "elapsed_time": "1:14:19", "remaining_time": "1:38:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 380, "total_steps": 870, "loss": 0.6925, "learning_rate": 6.924994998327395e-06, "epoch": 1.3103448275862069, "percentage": 43.68, "elapsed_time": "1:15:20", "remaining_time": "1:37:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 385, "total_steps": 870, "loss": 0.7251, "learning_rate": 6.832039490783422e-06, "epoch": 1.3275862068965516, "percentage": 44.25, "elapsed_time": "1:16:19", "remaining_time": "1:36:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 390, "total_steps": 870, "loss": 0.6822, "learning_rate": 6.7383466963767386e-06, "epoch": 1.3448275862068966, "percentage": 44.83, "elapsed_time": "1:17:21", "remaining_time": "1:35:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 395, "total_steps": 870, "loss": 0.6862, "learning_rate": 6.643954320881045e-06, "epoch": 1.3620689655172413, "percentage": 45.4, "elapsed_time": "1:18:19", "remaining_time": "1:34:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 400, "total_steps": 870, "loss": 0.6842, "learning_rate": 6.548900351609794e-06, "epoch": 1.3793103448275863, "percentage": 45.98, "elapsed_time": "1:19:14", "remaining_time": "1:33:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 405, "total_steps": 870, "loss": 0.6621, "learning_rate": 6.453223042128556e-06, "epoch": 1.396551724137931, "percentage": 46.55, "elapsed_time": "1:22:10", "remaining_time": "1:34:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 410, "total_steps": 870, "loss": 0.6398, "learning_rate": 6.3569608968602415e-06, "epoch": 1.4137931034482758, "percentage": 47.13, "elapsed_time": "1:23:08", "remaining_time": "1:33:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 415, "total_steps": 870, "loss": 0.6625, "learning_rate": 6.260152655589358e-06, "epoch": 1.4310344827586206, "percentage": 47.7, "elapsed_time": "1:24:04", "remaining_time": "1:32:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 420, "total_steps": 870, "loss": 0.7112, "learning_rate": 6.162837277871553e-06, "epoch": 1.4482758620689655, "percentage": 48.28, "elapsed_time": "1:25:02", "remaining_time": "1:31:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 425, "total_steps": 870, "loss": 0.6804, "learning_rate": 6.0650539273547145e-06, "epoch": 1.4655172413793103, "percentage": 48.85, "elapsed_time": "1:25:59", "remaining_time": "1:30:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 430, "total_steps": 870, "loss": 0.7072, "learning_rate": 5.966841956017928e-06, "epoch": 1.4827586206896552, "percentage": 49.43, "elapsed_time": "1:27:00", "remaining_time": "1:29:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 435, "total_steps": 870, "loss": 0.6718, "learning_rate": 5.8682408883346535e-06, "epoch": 1.5, "percentage": 50.0, "elapsed_time": "1:27:57", "remaining_time": "1:27:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 440, "total_steps": 870, "loss": 0.7114, "learning_rate": 5.769290405366469e-06, "epoch": 1.5172413793103448, "percentage": 50.57, "elapsed_time": "1:28:52", "remaining_time": "1:26:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 445, "total_steps": 870, "loss": 0.6678, "learning_rate": 5.670030328793812e-06, "epoch": 1.5344827586206895, "percentage": 51.15, "elapsed_time": "1:29:50", "remaining_time": "1:25:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 450, "total_steps": 870, "loss": 0.6433, "learning_rate": 5.570500604890124e-06, "epoch": 1.5517241379310345, "percentage": 51.72, "elapsed_time": "1:30:46", "remaining_time": "1:24:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 455, "total_steps": 870, "loss": 0.6791, "learning_rate": 5.470741288445844e-06, "epoch": 1.5689655172413794, "percentage": 52.3, "elapsed_time": "1:31:41", "remaining_time": "1:23:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 460, "total_steps": 870, "loss": 0.6389, "learning_rate": 5.370792526648747e-06, "epoch": 1.5862068965517242, "percentage": 52.87, "elapsed_time": "1:32:36", "remaining_time": "1:22:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 465, "total_steps": 870, "loss": 0.6676, "learning_rate": 5.270694542927089e-06, "epoch": 1.603448275862069, "percentage": 53.45, "elapsed_time": "1:33:31", "remaining_time": "1:21:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 470, "total_steps": 870, "loss": 0.6483, "learning_rate": 5.170487620762066e-06, "epoch": 1.6206896551724137, "percentage": 54.02, "elapsed_time": "1:34:30", "remaining_time": "1:20:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 475, "total_steps": 870, "loss": 0.7341, "learning_rate": 5.070212087476116e-06, "epoch": 1.6379310344827587, "percentage": 54.6, "elapsed_time": "1:35:24", "remaining_time": "1:19:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 480, "total_steps": 870, "loss": 0.6957, "learning_rate": 4.9699082980035735e-06, "epoch": 1.6551724137931034, "percentage": 55.17, "elapsed_time": "1:36:24", "remaining_time": "1:18:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 485, "total_steps": 870, "loss": 0.6611, "learning_rate": 4.869616618650201e-06, "epoch": 1.6724137931034484, "percentage": 55.75, "elapsed_time": "1:37:23", "remaining_time": "1:17:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 490, "total_steps": 870, "loss": 0.6827, "learning_rate": 4.769377410848162e-06, "epoch": 1.6896551724137931, "percentage": 56.32, "elapsed_time": "1:38:20", "remaining_time": "1:16:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 495, "total_steps": 870, "loss": 0.7234, "learning_rate": 4.6692310149129425e-06, "epoch": 1.706896551724138, "percentage": 56.9, "elapsed_time": "1:39:15", "remaining_time": "1:15:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 870, "loss": 0.6543, "learning_rate": 4.569217733808774e-06, "epoch": 1.7241379310344827, "percentage": 57.47, "elapsed_time": "1:40:12", "remaining_time": "1:14:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 505, "total_steps": 870, "loss": 0.6761, "learning_rate": 4.4693778169290934e-06, "epoch": 1.7413793103448276, "percentage": 58.05, "elapsed_time": "1:41:12", "remaining_time": "1:13:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 510, "total_steps": 870, "loss": 0.6471, "learning_rate": 4.369751443898554e-06, "epoch": 1.7586206896551724, "percentage": 58.62, "elapsed_time": "1:42:13", "remaining_time": "1:12:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 515, "total_steps": 870, "loss": 0.695, "learning_rate": 4.2703787084031175e-06, "epoch": 1.7758620689655173, "percentage": 59.2, "elapsed_time": "1:43:07", "remaining_time": "1:11:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 520, "total_steps": 870, "loss": 0.6512, "learning_rate": 4.171299602054736e-06, "epoch": 1.793103448275862, "percentage": 59.77, "elapsed_time": "1:44:04", "remaining_time": "1:10:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 525, "total_steps": 870, "loss": 0.6586, "learning_rate": 4.072553998297103e-06, "epoch": 1.8103448275862069, "percentage": 60.34, "elapsed_time": "1:45:03", "remaining_time": "1:09:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 530, "total_steps": 870, "loss": 0.6861, "learning_rate": 3.974181636358963e-06, "epoch": 1.8275862068965516, "percentage": 60.92, "elapsed_time": "1:45:57", "remaining_time": "1:07:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 535, "total_steps": 870, "loss": 0.6821, "learning_rate": 3.87622210526145e-06, "epoch": 1.8448275862068966, "percentage": 61.49, "elapsed_time": "1:46:53", "remaining_time": "1:06:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 540, "total_steps": 870, "loss": 0.6854, "learning_rate": 3.7787148278858453e-06, "epoch": 1.8620689655172413, "percentage": 62.07, "elapsed_time": "1:47:49", "remaining_time": "1:05:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 545, "total_steps": 870, "loss": 0.6489, "learning_rate": 3.6816990451082297e-06, "epoch": 1.8793103448275863, "percentage": 62.64, "elapsed_time": "1:48:44", "remaining_time": "1:04:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 550, "total_steps": 870, "loss": 0.7257, "learning_rate": 3.5852138000073566e-06, "epoch": 1.896551724137931, "percentage": 63.22, "elapsed_time": "1:49:41", "remaining_time": "1:03:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 555, "total_steps": 870, "loss": 0.6828, "learning_rate": 3.489297922152136e-06, "epoch": 1.9137931034482758, "percentage": 63.79, "elapsed_time": "1:50:37", "remaining_time": "1:02:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 560, "total_steps": 870, "loss": 0.6568, "learning_rate": 3.3939900119750543e-06, "epoch": 1.9310344827586206, "percentage": 64.37, "elapsed_time": "1:51:34", "remaining_time": "1:01:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 565, "total_steps": 870, "loss": 0.6964, "learning_rate": 3.299328425237781e-06, "epoch": 1.9482758620689655, "percentage": 64.94, "elapsed_time": "1:52:30", "remaining_time": "1:00:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 570, "total_steps": 870, "loss": 0.6066, "learning_rate": 3.205351257595272e-06, "epoch": 1.9655172413793105, "percentage": 65.52, "elapsed_time": "1:53:27", "remaining_time": "0:59:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 575, "total_steps": 870, "loss": 0.6529, "learning_rate": 3.11209632926453e-06, "epoch": 1.9827586206896552, "percentage": 66.09, "elapsed_time": "1:54:21", "remaining_time": "0:58:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 580, "total_steps": 870, "loss": 0.6351, "learning_rate": 3.019601169804216e-06, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "1:55:20", "remaining_time": "0:57:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 585, "total_steps": 870, "loss": 0.3206, "learning_rate": 2.927903003011241e-06, "epoch": 2.0172413793103448, "percentage": 67.24, "elapsed_time": "1:56:18", "remaining_time": "0:56:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 590, "total_steps": 870, "loss": 0.2667, "learning_rate": 2.837038731940397e-06, "epoch": 2.0344827586206895, "percentage": 67.82, "elapsed_time": "1:57:15", "remaining_time": "0:55:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 595, "total_steps": 870, "loss": 0.2633, "learning_rate": 2.7470449240530785e-06, "epoch": 2.0517241379310347, "percentage": 68.39, "elapsed_time": "1:58:16", "remaining_time": "0:54:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 600, "total_steps": 870, "loss": 0.2882, "learning_rate": 2.65795779650105e-06, "epoch": 2.0689655172413794, "percentage": 68.97, "elapsed_time": "1:59:12", "remaining_time": "0:53:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 405, "total_steps": 870, "loss": 0.6621, "learning_rate": 6.453223042128556e-06, "epoch": 1.396551724137931, "percentage": 46.55, "elapsed_time": "0:01:12", "remaining_time": "0:01:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 410, "total_steps": 870, "loss": 0.6399, "learning_rate": 6.3569608968602415e-06, "epoch": 1.4137931034482758, "percentage": 47.13, "elapsed_time": "0:02:10", "remaining_time": "0:02:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 415, "total_steps": 870, "loss": 0.6625, "learning_rate": 6.260152655589358e-06, "epoch": 1.4310344827586206, "percentage": 47.7, "elapsed_time": "0:03:06", "remaining_time": "0:03:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 420, "total_steps": 870, "loss": 0.7111, "learning_rate": 6.162837277871553e-06, "epoch": 1.4482758620689655, "percentage": 48.28, "elapsed_time": "0:04:02", "remaining_time": "0:04:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 425, "total_steps": 870, "loss": 0.6805, "learning_rate": 6.0650539273547145e-06, "epoch": 1.4655172413793103, "percentage": 48.85, "elapsed_time": "0:04:59", "remaining_time": "0:05:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 430, "total_steps": 870, "loss": 0.7073, "learning_rate": 5.966841956017928e-06, "epoch": 1.4827586206896552, "percentage": 49.43, "elapsed_time": "0:05:59", "remaining_time": "0:06:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 435, "total_steps": 870, "loss": 0.6719, "learning_rate": 5.8682408883346535e-06, "epoch": 1.5, "percentage": 50.0, "elapsed_time": "0:06:55", "remaining_time": "0:06:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 440, "total_steps": 870, "loss": 0.7115, "learning_rate": 5.769290405366469e-06, "epoch": 1.5172413793103448, "percentage": 50.57, "elapsed_time": "0:07:50", "remaining_time": "0:07:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 445, "total_steps": 870, "loss": 0.6679, "learning_rate": 5.670030328793812e-06, "epoch": 1.5344827586206895, "percentage": 51.15, "elapsed_time": "0:08:49", "remaining_time": "0:08:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 450, "total_steps": 870, "loss": 0.6434, "learning_rate": 5.570500604890124e-06, "epoch": 1.5517241379310345, "percentage": 51.72, "elapsed_time": "0:09:44", "remaining_time": "0:09:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 455, "total_steps": 870, "loss": 0.6791, "learning_rate": 5.470741288445844e-06, "epoch": 1.5689655172413794, "percentage": 52.3, "elapsed_time": "0:10:39", "remaining_time": "0:09:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 460, "total_steps": 870, "loss": 0.6392, "learning_rate": 5.370792526648747e-06, "epoch": 1.5862068965517242, "percentage": 52.87, "elapsed_time": "0:11:34", "remaining_time": "0:10:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 465, "total_steps": 870, "loss": 0.6675, "learning_rate": 5.270694542927089e-06, "epoch": 1.603448275862069, "percentage": 53.45, "elapsed_time": "0:12:28", "remaining_time": "0:10:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 470, "total_steps": 870, "loss": 0.6485, "learning_rate": 5.170487620762066e-06, "epoch": 1.6206896551724137, "percentage": 54.02, "elapsed_time": "0:13:28", "remaining_time": "0:11:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 475, "total_steps": 870, "loss": 0.734, "learning_rate": 5.070212087476116e-06, "epoch": 1.6379310344827587, "percentage": 54.6, "elapsed_time": "0:14:21", "remaining_time": "0:11:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 480, "total_steps": 870, "loss": 0.696, "learning_rate": 4.9699082980035735e-06, "epoch": 1.6551724137931034, "percentage": 55.17, "elapsed_time": "0:15:20", "remaining_time": "0:12:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 485, "total_steps": 870, "loss": 0.6611, "learning_rate": 4.869616618650201e-06, "epoch": 1.6724137931034484, "percentage": 55.75, "elapsed_time": "0:16:19", "remaining_time": "0:12:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 490, "total_steps": 870, "loss": 0.6828, "learning_rate": 4.769377410848162e-06, "epoch": 1.6896551724137931, "percentage": 56.32, "elapsed_time": "0:17:16", "remaining_time": "0:13:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 495, "total_steps": 870, "loss": 0.7233, "learning_rate": 4.6692310149129425e-06, "epoch": 1.706896551724138, "percentage": 56.9, "elapsed_time": "0:18:12", "remaining_time": "0:13:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 870, "loss": 0.6543, "learning_rate": 4.569217733808774e-06, "epoch": 1.7241379310344827, "percentage": 57.47, "elapsed_time": "0:19:08", "remaining_time": "0:14:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 505, "total_steps": 870, "loss": 0.6763, "learning_rate": 4.4693778169290934e-06, "epoch": 1.7413793103448276, "percentage": 58.05, "elapsed_time": "0:20:07", "remaining_time": "0:14:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 510, "total_steps": 870, "loss": 0.6468, "learning_rate": 4.369751443898554e-06, "epoch": 1.7586206896551724, "percentage": 58.62, "elapsed_time": "0:21:09", "remaining_time": "0:14:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 515, "total_steps": 870, "loss": 0.6951, "learning_rate": 4.2703787084031175e-06, "epoch": 1.7758620689655173, "percentage": 59.2, "elapsed_time": "0:22:03", "remaining_time": "0:15:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 520, "total_steps": 870, "loss": 0.651, "learning_rate": 4.171299602054736e-06, "epoch": 1.793103448275862, "percentage": 59.77, "elapsed_time": "0:22:59", "remaining_time": "0:15:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 525, "total_steps": 870, "loss": 0.6584, "learning_rate": 4.072553998297103e-06, "epoch": 1.8103448275862069, "percentage": 60.34, "elapsed_time": "0:23:56", "remaining_time": "0:15:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 530, "total_steps": 870, "loss": 0.6862, "learning_rate": 3.974181636358963e-06, "epoch": 1.8275862068965516, "percentage": 60.92, "elapsed_time": "0:24:50", "remaining_time": "0:15:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 535, "total_steps": 870, "loss": 0.6821, "learning_rate": 3.87622210526145e-06, "epoch": 1.8448275862068966, "percentage": 61.49, "elapsed_time": "0:25:46", "remaining_time": "0:16:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 540, "total_steps": 870, "loss": 0.6854, "learning_rate": 3.7787148278858453e-06, "epoch": 1.8620689655172413, "percentage": 62.07, "elapsed_time": "0:26:42", "remaining_time": "0:16:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 545, "total_steps": 870, "loss": 0.6488, "learning_rate": 3.6816990451082297e-06, "epoch": 1.8793103448275863, "percentage": 62.64, "elapsed_time": "0:27:37", "remaining_time": "0:16:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 550, "total_steps": 870, "loss": 0.7261, "learning_rate": 3.5852138000073566e-06, "epoch": 1.896551724137931, "percentage": 63.22, "elapsed_time": "0:28:33", "remaining_time": "0:16:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 555, "total_steps": 870, "loss": 0.6829, "learning_rate": 3.489297922152136e-06, "epoch": 1.9137931034482758, "percentage": 63.79, "elapsed_time": "0:29:29", "remaining_time": "0:16:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 560, "total_steps": 870, "loss": 0.6567, "learning_rate": 3.3939900119750543e-06, "epoch": 1.9310344827586206, "percentage": 64.37, "elapsed_time": "0:30:26", "remaining_time": "0:16:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 565, "total_steps": 870, "loss": 0.6965, "learning_rate": 3.299328425237781e-06, "epoch": 1.9482758620689655, "percentage": 64.94, "elapsed_time": "0:31:22", "remaining_time": "0:16:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 570, "total_steps": 870, "loss": 0.6066, "learning_rate": 3.205351257595272e-06, "epoch": 1.9655172413793105, "percentage": 65.52, "elapsed_time": "0:32:18", "remaining_time": "0:17:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 575, "total_steps": 870, "loss": 0.6526, "learning_rate": 3.11209632926453e-06, "epoch": 1.9827586206896552, "percentage": 66.09, "elapsed_time": "0:33:13", "remaining_time": "0:17:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 580, "total_steps": 870, "loss": 0.6352, "learning_rate": 3.019601169804216e-06, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "0:34:10", "remaining_time": "0:17:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 585, "total_steps": 870, "loss": 0.321, "learning_rate": 2.927903003011241e-06, "epoch": 2.0172413793103448, "percentage": 67.24, "elapsed_time": "0:35:09", "remaining_time": "0:17:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 590, "total_steps": 870, "loss": 0.2667, "learning_rate": 2.837038731940397e-06, "epoch": 2.0344827586206895, "percentage": 67.82, "elapsed_time": "0:36:06", "remaining_time": "0:17:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 595, "total_steps": 870, "loss": 0.2632, "learning_rate": 2.7470449240530785e-06, "epoch": 2.0517241379310347, "percentage": 68.39, "elapsed_time": "0:37:06", "remaining_time": "0:17:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 600, "total_steps": 870, "loss": 0.288, "learning_rate": 2.65795779650105e-06, "epoch": 2.0689655172413794, "percentage": 68.97, "elapsed_time": "0:38:03", "remaining_time": "0:17:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 605, "total_steps": 870, "loss": 0.2973, "learning_rate": 2.569813201551205e-06, "epoch": 2.086206896551724, "percentage": 69.54, "elapsed_time": "0:40:59", "remaining_time": "0:17:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 610, "total_steps": 870, "loss": 0.2813, "learning_rate": 2.4826466121571575e-06, "epoch": 2.103448275862069, "percentage": 70.11, "elapsed_time": "0:41:54", "remaining_time": "0:17:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 615, "total_steps": 870, "loss": 0.265, "learning_rate": 2.396493107683488e-06, "epoch": 2.1206896551724137, "percentage": 70.69, "elapsed_time": "0:42:51", "remaining_time": "0:17:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 620, "total_steps": 870, "loss": 0.258, "learning_rate": 2.311387359788395e-06, "epoch": 2.1379310344827585, "percentage": 71.26, "elapsed_time": "0:43:55", "remaining_time": "0:17:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 625, "total_steps": 870, "loss": 0.2817, "learning_rate": 2.2273636184704074e-06, "epoch": 2.1551724137931036, "percentage": 71.84, "elapsed_time": "0:44:51", "remaining_time": "0:17:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 630, "total_steps": 870, "loss": 0.2595, "learning_rate": 2.1444556982847996e-06, "epoch": 2.1724137931034484, "percentage": 72.41, "elapsed_time": "0:45:43", "remaining_time": "0:17:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 635, "total_steps": 870, "loss": 0.2805, "learning_rate": 2.06269696473525e-06, "epoch": 2.189655172413793, "percentage": 72.99, "elapsed_time": "0:46:40", "remaining_time": "0:17:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 640, "total_steps": 870, "loss": 0.2689, "learning_rate": 1.982120320846208e-06, "epoch": 2.206896551724138, "percentage": 73.56, "elapsed_time": "0:47:41", "remaining_time": "0:17:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 645, "total_steps": 870, "loss": 0.2696, "learning_rate": 1.9027581939213852e-06, "epoch": 2.2241379310344827, "percentage": 74.14, "elapsed_time": "0:48:40", "remaining_time": "0:16:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 650, "total_steps": 870, "loss": 0.2727, "learning_rate": 1.8246425224936986e-06, "epoch": 2.2413793103448274, "percentage": 74.71, "elapsed_time": "0:49:38", "remaining_time": "0:16:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 655, "total_steps": 870, "loss": 0.2664, "learning_rate": 1.747804743471907e-06, "epoch": 2.2586206896551726, "percentage": 75.29, "elapsed_time": "0:50:35", "remaining_time": "0:16:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 660, "total_steps": 870, "loss": 0.2969, "learning_rate": 1.6722757794891287e-06, "epoch": 2.2758620689655173, "percentage": 75.86, "elapsed_time": "0:51:31", "remaining_time": "0:16:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 665, "total_steps": 870, "loss": 0.2547, "learning_rate": 1.598086026458322e-06, "epoch": 2.293103448275862, "percentage": 76.44, "elapsed_time": "0:52:26", "remaining_time": "0:16:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 670, "total_steps": 870, "loss": 0.3067, "learning_rate": 1.52526534133974e-06, "epoch": 2.310344827586207, "percentage": 77.01, "elapsed_time": "0:53:24", "remaining_time": "0:15:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 675, "total_steps": 870, "loss": 0.2798, "learning_rate": 1.4538430301252783e-06, "epoch": 2.3275862068965516, "percentage": 77.59, "elapsed_time": "0:54:18", "remaining_time": "0:15:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 680, "total_steps": 870, "loss": 0.2911, "learning_rate": 1.3838478360445616e-06, "epoch": 2.344827586206897, "percentage": 78.16, "elapsed_time": "0:55:14", "remaining_time": "0:15:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 685, "total_steps": 870, "loss": 0.2916, "learning_rate": 1.3153079279975011e-06, "epoch": 2.3620689655172415, "percentage": 78.74, "elapsed_time": "0:56:07", "remaining_time": "0:15:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 690, "total_steps": 870, "loss": 0.2583, "learning_rate": 1.2482508892179884e-06, "epoch": 2.3793103448275863, "percentage": 79.31, "elapsed_time": "0:57:04", "remaining_time": "0:14:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 695, "total_steps": 870, "loss": 0.2769, "learning_rate": 1.1827037061732877e-06, "epoch": 2.396551724137931, "percentage": 79.89, "elapsed_time": "0:57:56", "remaining_time": "0:14:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 700, "total_steps": 870, "loss": 0.2567, "learning_rate": 1.1186927577035867e-06, "epoch": 2.413793103448276, "percentage": 80.46, "elapsed_time": "0:58:51", "remaining_time": "0:14:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 705, "total_steps": 870, "loss": 0.2668, "learning_rate": 1.0562438044060846e-06, "epoch": 2.4310344827586206, "percentage": 81.03, "elapsed_time": "0:59:51", "remaining_time": "0:14:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 710, "total_steps": 870, "loss": 0.256, "learning_rate": 9.953819782678887e-07, "epoch": 2.4482758620689653, "percentage": 81.61, "elapsed_time": "1:00:47", "remaining_time": "0:13:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 715, "total_steps": 870, "loss": 0.3042, "learning_rate": 9.361317725518749e-07, "epoch": 2.4655172413793105, "percentage": 82.18, "elapsed_time": "1:01:45", "remaining_time": "0:13:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 720, "total_steps": 870, "loss": 0.2751, "learning_rate": 8.785170319396174e-07, "epoch": 2.4827586206896552, "percentage": 82.76, "elapsed_time": "1:02:44", "remaining_time": "0:13:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 725, "total_steps": 870, "loss": 0.2849, "learning_rate": 8.225609429353187e-07, "epoch": 2.5, "percentage": 83.33, "elapsed_time": "1:03:49", "remaining_time": "0:12:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 730, "total_steps": 870, "loss": 0.2763, "learning_rate": 7.682860245346213e-07, "epoch": 2.5172413793103448, "percentage": 83.91, "elapsed_time": "1:04:49", "remaining_time": "0:12:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 735, "total_steps": 870, "loss": 0.2737, "learning_rate": 7.157141191620548e-07, "epoch": 2.5344827586206895, "percentage": 84.48, "elapsed_time": "1:05:51", "remaining_time": "0:12:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 740, "total_steps": 870, "loss": 0.2636, "learning_rate": 6.648663838807562e-07, "epoch": 2.5517241379310347, "percentage": 85.06, "elapsed_time": "1:06:44", "remaining_time": "0:11:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 745, "total_steps": 870, "loss": 0.2942, "learning_rate": 6.157632818780179e-07, "epoch": 2.5689655172413794, "percentage": 85.63, "elapsed_time": "1:07:40", "remaining_time": "0:11:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 750, "total_steps": 870, "loss": 0.2632, "learning_rate": 5.684245742300625e-07, "epoch": 2.586206896551724, "percentage": 86.21, "elapsed_time": "1:08:40", "remaining_time": "0:10:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 755, "total_steps": 870, "loss": 0.2742, "learning_rate": 5.228693119493955e-07, "epoch": 2.603448275862069, "percentage": 86.78, "elapsed_time": "1:09:33", "remaining_time": "0:10:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 760, "total_steps": 870, "loss": 0.2598, "learning_rate": 4.791158283178999e-07, "epoch": 2.6206896551724137, "percentage": 87.36, "elapsed_time": "1:10:28", "remaining_time": "0:10:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 765, "total_steps": 870, "loss": 0.2476, "learning_rate": 4.371817315087845e-07, "epoch": 2.637931034482759, "percentage": 87.93, "elapsed_time": "1:11:24", "remaining_time": "0:09:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 770, "total_steps": 870, "loss": 0.2914, "learning_rate": 3.9708389750034105e-07, "epoch": 2.655172413793103, "percentage": 88.51, "elapsed_time": "1:12:23", "remaining_time": "0:09:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 775, "total_steps": 870, "loss": 0.2648, "learning_rate": 3.5883846328436943e-07, "epoch": 2.6724137931034484, "percentage": 89.08, "elapsed_time": "1:13:23", "remaining_time": "0:08:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 780, "total_steps": 870, "loss": 0.2959, "learning_rate": 3.224608203719953e-07, "epoch": 2.689655172413793, "percentage": 89.66, "elapsed_time": "1:14:22", "remaining_time": "0:08:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 785, "total_steps": 870, "loss": 0.2852, "learning_rate": 2.879656085995042e-07, "epoch": 2.706896551724138, "percentage": 90.23, "elapsed_time": "1:15:23", "remaining_time": "0:08:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 790, "total_steps": 870, "loss": 0.2783, "learning_rate": 2.5536671023668305e-07, "epoch": 2.7241379310344827, "percentage": 90.8, "elapsed_time": "1:16:21", "remaining_time": "0:07:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 795, "total_steps": 870, "loss": 0.2732, "learning_rate": 2.2467724440002336e-07, "epoch": 2.7413793103448274, "percentage": 91.38, "elapsed_time": "1:17:14", "remaining_time": "0:07:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 800, "total_steps": 870, "loss": 0.2724, "learning_rate": 1.9590956177306665e-07, "epoch": 2.7586206896551726, "percentage": 91.95, "elapsed_time": "1:18:13", "remaining_time": "0:06:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 805, "total_steps": 870, "loss": 0.2992, "learning_rate": 1.690752396359857e-07, "epoch": 2.7758620689655173, "percentage": 92.53, "elapsed_time": "1:21:16", "remaining_time": "0:06:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 810, "total_steps": 870, "loss": 0.2684, "learning_rate": 1.4418507720641794e-07, "epoch": 2.793103448275862, "percentage": 93.1, "elapsed_time": "1:22:10", "remaining_time": "0:06:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 815, "total_steps": 870, "loss": 0.2835, "learning_rate": 1.2124909129342332e-07, "epoch": 2.810344827586207, "percentage": 93.68, "elapsed_time": "1:23:07", "remaining_time": "0:05:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 820, "total_steps": 870, "loss": 0.2831, "learning_rate": 1.0027651226631463e-07, "epoch": 2.8275862068965516, "percentage": 94.25, "elapsed_time": "1:24:05", "remaining_time": "0:05:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 825, "total_steps": 870, "loss": 0.2958, "learning_rate": 8.127578033998663e-08, "epoch": 2.844827586206897, "percentage": 94.83, "elapsed_time": "1:25:05", "remaining_time": "0:04:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 830, "total_steps": 870, "loss": 0.2929, "learning_rate": 6.425454217822425e-08, "epoch": 2.862068965517241, "percentage": 95.4, "elapsed_time": "1:26:01", "remaining_time": "0:04:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 835, "total_steps": 870, "loss": 0.2711, "learning_rate": 4.9219647816383666e-08, "epoch": 2.8793103448275863, "percentage": 95.98, "elapsed_time": "1:26:57", "remaining_time": "0:03:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 840, "total_steps": 870, "loss": 0.278, "learning_rate": 3.617714790465576e-08, "epoch": 2.896551724137931, "percentage": 96.55, "elapsed_time": "1:27:54", "remaining_time": "0:03:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 845, "total_steps": 870, "loss": 0.2778, "learning_rate": 2.5132291273042288e-08, "epoch": 2.913793103448276, "percentage": 97.13, "elapsed_time": "1:28:56", "remaining_time": "0:02:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 850, "total_steps": 870, "loss": 0.2846, "learning_rate": 1.608952281901055e-08, "epoch": 2.9310344827586206, "percentage": 97.7, "elapsed_time": "1:29:55", "remaining_time": "0:02:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 855, "total_steps": 870, "loss": 0.284, "learning_rate": 9.052481718690998e-09, "epoch": 2.9482758620689653, "percentage": 98.28, "elapsed_time": "1:30:51", "remaining_time": "0:01:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 860, "total_steps": 870, "loss": 0.2666, "learning_rate": 4.0239999623226115e-09, "epoch": 2.9655172413793105, "percentage": 98.85, "elapsed_time": "1:31:46", "remaining_time": "0:01:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 865, "total_steps": 870, "loss": 0.2556, "learning_rate": 1.006101214545696e-09, "epoch": 2.9827586206896552, "percentage": 99.43, "elapsed_time": "1:32:47", "remaining_time": "0:00:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 870, "total_steps": 870, "loss": 0.2884, "learning_rate": 0.0, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "1:33:41", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 870, "total_steps": 870, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "1:35:46", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |