| {"current_steps": 5, "total_steps": 4274, "loss": 0.7204, "learning_rate": 5.000000000000001e-07, "epoch": 0.0023394549070066676, "percentage": 0.12, "elapsed_time": "0:01:30", "remaining_time": "21:33:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 10, "total_steps": 4274, "loss": 0.7411, "learning_rate": 1.0000000000000002e-06, "epoch": 0.004678909814013335, "percentage": 0.23, "elapsed_time": "0:03:00", "remaining_time": "21:23:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 15, "total_steps": 4274, "loss": 0.7235, "learning_rate": 1.5e-06, "epoch": 0.007018364721020002, "percentage": 0.35, "elapsed_time": "0:04:32", "remaining_time": "21:28:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 20, "total_steps": 4274, "loss": 0.7176, "learning_rate": 2.0000000000000003e-06, "epoch": 0.00935781962802667, "percentage": 0.47, "elapsed_time": "0:06:04", "remaining_time": "21:32:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 25, "total_steps": 4274, "loss": 0.7439, "learning_rate": 2.5e-06, "epoch": 0.011697274535033338, "percentage": 0.58, "elapsed_time": "0:07:37", "remaining_time": "21:34:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 30, "total_steps": 4274, "loss": 0.6894, "learning_rate": 3e-06, "epoch": 0.014036729442040005, "percentage": 0.7, "elapsed_time": "0:09:09", "remaining_time": "21:36:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 35, "total_steps": 4274, "loss": 0.6607, "learning_rate": 3.5000000000000004e-06, "epoch": 0.016376184349046672, "percentage": 0.82, "elapsed_time": "0:10:42", "remaining_time": "21:37:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 40, "total_steps": 4274, "loss": 0.6335, "learning_rate": 4.000000000000001e-06, "epoch": 0.01871563925605334, "percentage": 0.94, "elapsed_time": "0:12:15", "remaining_time": "21:37:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 45, "total_steps": 4274, "loss": 0.5814, "learning_rate": 4.5e-06, "epoch": 0.021055094163060006, "percentage": 1.05, "elapsed_time": "0:13:48", "remaining_time": "21:37:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 50, "total_steps": 4274, "loss": 0.5636, "learning_rate": 5e-06, "epoch": 0.023394549070066675, "percentage": 1.17, "elapsed_time": "0:15:21", "remaining_time": "21:37:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 55, "total_steps": 4274, "loss": 0.5845, "learning_rate": 5.500000000000001e-06, "epoch": 0.02573400397707334, "percentage": 1.29, "elapsed_time": "0:16:54", "remaining_time": "21:37:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 60, "total_steps": 4274, "loss": 0.494, "learning_rate": 6e-06, "epoch": 0.02807345888408001, "percentage": 1.4, "elapsed_time": "0:18:27", "remaining_time": "21:36:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 65, "total_steps": 4274, "loss": 0.4809, "learning_rate": 6.5000000000000004e-06, "epoch": 0.030412913791086675, "percentage": 1.52, "elapsed_time": "0:20:00", "remaining_time": "21:35:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 70, "total_steps": 4274, "loss": 0.446, "learning_rate": 7.000000000000001e-06, "epoch": 0.032752368698093344, "percentage": 1.64, "elapsed_time": "0:21:33", "remaining_time": "21:34:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 75, "total_steps": 4274, "loss": 0.3915, "learning_rate": 7.5e-06, "epoch": 0.03509182360510001, "percentage": 1.75, "elapsed_time": "0:23:05", "remaining_time": "21:32:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 80, "total_steps": 4274, "loss": 0.3696, "learning_rate": 8.000000000000001e-06, "epoch": 0.03743127851210668, "percentage": 1.87, "elapsed_time": "0:24:38", "remaining_time": "21:31:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 85, "total_steps": 4274, "loss": 0.3464, "learning_rate": 8.500000000000002e-06, "epoch": 0.039770733419113344, "percentage": 1.99, "elapsed_time": "0:26:10", "remaining_time": "21:30:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 90, "total_steps": 4274, "loss": 0.3097, "learning_rate": 9e-06, "epoch": 0.04211018832612001, "percentage": 2.11, "elapsed_time": "0:27:43", "remaining_time": "21:28:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 95, "total_steps": 4274, "loss": 0.3111, "learning_rate": 9.5e-06, "epoch": 0.04444964323312668, "percentage": 2.22, "elapsed_time": "0:29:16", "remaining_time": "21:27:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 100, "total_steps": 4274, "loss": 0.2696, "learning_rate": 1e-05, "epoch": 0.04678909814013335, "percentage": 2.34, "elapsed_time": "0:30:48", "remaining_time": "21:26:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 105, "total_steps": 4274, "loss": 0.2461, "learning_rate": 1.05e-05, "epoch": 0.04912855304714002, "percentage": 2.46, "elapsed_time": "0:32:21", "remaining_time": "21:24:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 110, "total_steps": 4274, "loss": 0.2139, "learning_rate": 1.1000000000000001e-05, "epoch": 0.05146800795414668, "percentage": 2.57, "elapsed_time": "0:33:54", "remaining_time": "21:23:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 115, "total_steps": 4274, "loss": 0.2045, "learning_rate": 1.1500000000000002e-05, "epoch": 0.05380746286115335, "percentage": 2.69, "elapsed_time": "0:35:26", "remaining_time": "21:21:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 120, "total_steps": 4274, "loss": 0.1987, "learning_rate": 1.2e-05, "epoch": 0.05614691776816002, "percentage": 2.81, "elapsed_time": "0:36:59", "remaining_time": "21:20:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 125, "total_steps": 4274, "loss": 0.1899, "learning_rate": 1.25e-05, "epoch": 0.05848637267516669, "percentage": 2.92, "elapsed_time": "0:38:32", "remaining_time": "21:19:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 130, "total_steps": 4274, "loss": 0.171, "learning_rate": 1.3000000000000001e-05, "epoch": 0.06082582758217335, "percentage": 3.04, "elapsed_time": "0:40:04", "remaining_time": "21:17:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 135, "total_steps": 4274, "loss": 0.1636, "learning_rate": 1.3500000000000001e-05, "epoch": 0.06316528248918002, "percentage": 3.16, "elapsed_time": "0:41:37", "remaining_time": "21:16:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 140, "total_steps": 4274, "loss": 0.1579, "learning_rate": 1.4000000000000001e-05, "epoch": 0.06550473739618669, "percentage": 3.28, "elapsed_time": "0:43:10", "remaining_time": "21:14:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 145, "total_steps": 4274, "loss": 0.1377, "learning_rate": 1.45e-05, "epoch": 0.06784419230319336, "percentage": 3.39, "elapsed_time": "0:44:43", "remaining_time": "21:13:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 150, "total_steps": 4274, "loss": 0.1325, "learning_rate": 1.5e-05, "epoch": 0.07018364721020003, "percentage": 3.51, "elapsed_time": "0:46:16", "remaining_time": "21:12:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 155, "total_steps": 4274, "loss": 0.1384, "learning_rate": 1.55e-05, "epoch": 0.0725231021172067, "percentage": 3.63, "elapsed_time": "0:47:49", "remaining_time": "21:10:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 160, "total_steps": 4274, "loss": 0.1198, "learning_rate": 1.6000000000000003e-05, "epoch": 0.07486255702421336, "percentage": 3.74, "elapsed_time": "0:49:22", "remaining_time": "21:09:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 165, "total_steps": 4274, "loss": 0.1192, "learning_rate": 1.65e-05, "epoch": 0.07720201193122003, "percentage": 3.86, "elapsed_time": "0:50:55", "remaining_time": "21:07:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 170, "total_steps": 4274, "loss": 0.1208, "learning_rate": 1.7000000000000003e-05, "epoch": 0.07954146683822669, "percentage": 3.98, "elapsed_time": "0:52:27", "remaining_time": "21:06:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 175, "total_steps": 4274, "loss": 0.1099, "learning_rate": 1.75e-05, "epoch": 0.08188092174523336, "percentage": 4.09, "elapsed_time": "0:54:00", "remaining_time": "21:05:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 180, "total_steps": 4274, "loss": 0.104, "learning_rate": 1.8e-05, "epoch": 0.08422037665224003, "percentage": 4.21, "elapsed_time": "0:55:33", "remaining_time": "21:03:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 185, "total_steps": 4274, "loss": 0.1489, "learning_rate": 1.85e-05, "epoch": 0.0865598315592467, "percentage": 4.33, "elapsed_time": "0:57:06", "remaining_time": "21:02:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 190, "total_steps": 4274, "loss": 0.1102, "learning_rate": 1.9e-05, "epoch": 0.08889928646625336, "percentage": 4.45, "elapsed_time": "0:58:39", "remaining_time": "21:00:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 195, "total_steps": 4274, "loss": 0.1172, "learning_rate": 1.9500000000000003e-05, "epoch": 0.09123874137326003, "percentage": 4.56, "elapsed_time": "1:00:11", "remaining_time": "20:59:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 200, "total_steps": 4274, "loss": 0.1046, "learning_rate": 2e-05, "epoch": 0.0935781962802667, "percentage": 4.68, "elapsed_time": "1:01:44", "remaining_time": "20:57:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 205, "total_steps": 4274, "loss": 0.1024, "learning_rate": 2.05e-05, "epoch": 0.09591765118727337, "percentage": 4.8, "elapsed_time": "1:03:17", "remaining_time": "20:56:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 210, "total_steps": 4274, "loss": 0.1117, "learning_rate": 2.1e-05, "epoch": 0.09825710609428004, "percentage": 4.91, "elapsed_time": "1:04:50", "remaining_time": "20:54:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 215, "total_steps": 4274, "loss": 0.1177, "learning_rate": 2.15e-05, "epoch": 0.1005965610012867, "percentage": 5.03, "elapsed_time": "1:06:22", "remaining_time": "20:53:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 220, "total_steps": 4274, "loss": 0.0988, "learning_rate": 2.2000000000000003e-05, "epoch": 0.10293601590829336, "percentage": 5.15, "elapsed_time": "1:07:55", "remaining_time": "20:51:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 225, "total_steps": 4274, "loss": 0.0943, "learning_rate": 2.25e-05, "epoch": 0.10527547081530003, "percentage": 5.26, "elapsed_time": "1:09:28", "remaining_time": "20:50:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 230, "total_steps": 4274, "loss": 0.0989, "learning_rate": 2.3000000000000003e-05, "epoch": 0.1076149257223067, "percentage": 5.38, "elapsed_time": "1:11:01", "remaining_time": "20:48:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 235, "total_steps": 4274, "loss": 0.0953, "learning_rate": 2.35e-05, "epoch": 0.10995438062931337, "percentage": 5.5, "elapsed_time": "1:12:34", "remaining_time": "20:47:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 240, "total_steps": 4274, "loss": 0.0989, "learning_rate": 2.4e-05, "epoch": 0.11229383553632004, "percentage": 5.62, "elapsed_time": "1:14:07", "remaining_time": "20:45:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 245, "total_steps": 4274, "loss": 0.0933, "learning_rate": 2.45e-05, "epoch": 0.11463329044332671, "percentage": 5.73, "elapsed_time": "1:15:39", "remaining_time": "20:44:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 250, "total_steps": 4274, "loss": 0.1041, "learning_rate": 2.5e-05, "epoch": 0.11697274535033338, "percentage": 5.85, "elapsed_time": "1:17:12", "remaining_time": "20:42:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 255, "total_steps": 4274, "loss": 0.0983, "learning_rate": 2.5500000000000003e-05, "epoch": 0.11931220025734005, "percentage": 5.97, "elapsed_time": "1:18:44", "remaining_time": "20:40:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 260, "total_steps": 4274, "loss": 0.1017, "learning_rate": 2.6000000000000002e-05, "epoch": 0.1216516551643467, "percentage": 6.08, "elapsed_time": "1:20:17", "remaining_time": "20:39:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 265, "total_steps": 4274, "loss": 0.1486, "learning_rate": 2.6500000000000004e-05, "epoch": 0.12399111007135337, "percentage": 6.2, "elapsed_time": "1:21:49", "remaining_time": "20:37:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 270, "total_steps": 4274, "loss": 0.0973, "learning_rate": 2.7000000000000002e-05, "epoch": 0.12633056497836004, "percentage": 6.32, "elapsed_time": "1:23:20", "remaining_time": "20:35:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 275, "total_steps": 4274, "loss": 0.0957, "learning_rate": 2.7500000000000004e-05, "epoch": 0.12867001988536672, "percentage": 6.43, "elapsed_time": "1:24:51", "remaining_time": "20:34:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 280, "total_steps": 4274, "loss": 0.1, "learning_rate": 2.8000000000000003e-05, "epoch": 0.13100947479237338, "percentage": 6.55, "elapsed_time": "1:26:24", "remaining_time": "20:32:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 285, "total_steps": 4274, "loss": 0.0878, "learning_rate": 2.8499999999999998e-05, "epoch": 0.13334892969938003, "percentage": 6.67, "elapsed_time": "1:27:56", "remaining_time": "20:30:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 290, "total_steps": 4274, "loss": 0.0881, "learning_rate": 2.9e-05, "epoch": 0.1356883846063867, "percentage": 6.79, "elapsed_time": "1:29:29", "remaining_time": "20:29:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 295, "total_steps": 4274, "loss": 0.1115, "learning_rate": 2.95e-05, "epoch": 0.13802783951339337, "percentage": 6.9, "elapsed_time": "1:31:01", "remaining_time": "20:27:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 300, "total_steps": 4274, "loss": 0.1025, "learning_rate": 3e-05, "epoch": 0.14036729442040005, "percentage": 7.02, "elapsed_time": "1:32:34", "remaining_time": "20:26:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 305, "total_steps": 4274, "loss": 0.0978, "learning_rate": 3.05e-05, "epoch": 0.1427067493274067, "percentage": 7.14, "elapsed_time": "1:34:06", "remaining_time": "20:24:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 310, "total_steps": 4274, "loss": 0.0931, "learning_rate": 3.1e-05, "epoch": 0.1450462042344134, "percentage": 7.25, "elapsed_time": "1:35:38", "remaining_time": "20:22:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 315, "total_steps": 4274, "loss": 0.0947, "learning_rate": 3.15e-05, "epoch": 0.14738565914142004, "percentage": 7.37, "elapsed_time": "1:37:10", "remaining_time": "20:21:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 320, "total_steps": 4274, "loss": 0.0876, "learning_rate": 3.2000000000000005e-05, "epoch": 0.14972511404842673, "percentage": 7.49, "elapsed_time": "1:38:43", "remaining_time": "20:19:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 325, "total_steps": 4274, "loss": 0.0941, "learning_rate": 3.2500000000000004e-05, "epoch": 0.15206456895543338, "percentage": 7.6, "elapsed_time": "1:40:15", "remaining_time": "20:18:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 330, "total_steps": 4274, "loss": 0.0825, "learning_rate": 3.3e-05, "epoch": 0.15440402386244007, "percentage": 7.72, "elapsed_time": "1:41:48", "remaining_time": "20:16:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 335, "total_steps": 4274, "loss": 0.0803, "learning_rate": 3.35e-05, "epoch": 0.15674347876944672, "percentage": 7.84, "elapsed_time": "1:43:20", "remaining_time": "20:15:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 340, "total_steps": 4274, "loss": 0.081, "learning_rate": 3.4000000000000007e-05, "epoch": 0.15908293367645338, "percentage": 7.96, "elapsed_time": "1:44:53", "remaining_time": "20:13:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 345, "total_steps": 4274, "loss": 0.0895, "learning_rate": 3.45e-05, "epoch": 0.16142238858346006, "percentage": 8.07, "elapsed_time": "1:46:25", "remaining_time": "20:12:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 350, "total_steps": 4274, "loss": 0.0882, "learning_rate": 3.5e-05, "epoch": 0.1637618434904667, "percentage": 8.19, "elapsed_time": "1:47:58", "remaining_time": "20:10:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 355, "total_steps": 4274, "loss": 0.0928, "learning_rate": 3.55e-05, "epoch": 0.1661012983974734, "percentage": 8.31, "elapsed_time": "1:49:30", "remaining_time": "20:08:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 360, "total_steps": 4274, "loss": 0.0939, "learning_rate": 3.6e-05, "epoch": 0.16844075330448005, "percentage": 8.42, "elapsed_time": "1:51:01", "remaining_time": "20:07:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 365, "total_steps": 4274, "loss": 0.0842, "learning_rate": 3.65e-05, "epoch": 0.17078020821148673, "percentage": 8.54, "elapsed_time": "1:52:33", "remaining_time": "20:05:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 370, "total_steps": 4274, "loss": 0.0954, "learning_rate": 3.7e-05, "epoch": 0.1731196631184934, "percentage": 8.66, "elapsed_time": "1:54:05", "remaining_time": "20:03:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 375, "total_steps": 4274, "loss": 0.0838, "learning_rate": 3.7500000000000003e-05, "epoch": 0.17545911802550007, "percentage": 8.77, "elapsed_time": "1:55:35", "remaining_time": "20:01:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 380, "total_steps": 4274, "loss": 0.101, "learning_rate": 3.8e-05, "epoch": 0.17779857293250673, "percentage": 8.89, "elapsed_time": "1:57:04", "remaining_time": "19:59:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 385, "total_steps": 4274, "loss": 0.0873, "learning_rate": 3.85e-05, "epoch": 0.18013802783951338, "percentage": 9.01, "elapsed_time": "1:58:36", "remaining_time": "19:58:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 390, "total_steps": 4274, "loss": 0.09, "learning_rate": 3.9000000000000006e-05, "epoch": 0.18247748274652006, "percentage": 9.12, "elapsed_time": "2:00:08", "remaining_time": "19:56:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 395, "total_steps": 4274, "loss": 0.0892, "learning_rate": 3.9500000000000005e-05, "epoch": 0.18481693765352672, "percentage": 9.24, "elapsed_time": "2:01:39", "remaining_time": "19:54:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 400, "total_steps": 4274, "loss": 0.0832, "learning_rate": 4e-05, "epoch": 0.1871563925605334, "percentage": 9.36, "elapsed_time": "2:03:11", "remaining_time": "19:53:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 405, "total_steps": 4274, "loss": 0.093, "learning_rate": 4.05e-05, "epoch": 0.18949584746754006, "percentage": 9.48, "elapsed_time": "2:04:43", "remaining_time": "19:51:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 410, "total_steps": 4274, "loss": 0.0877, "learning_rate": 4.1e-05, "epoch": 0.19183530237454674, "percentage": 9.59, "elapsed_time": "2:06:14", "remaining_time": "19:49:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 415, "total_steps": 4274, "loss": 0.0841, "learning_rate": 4.15e-05, "epoch": 0.1941747572815534, "percentage": 9.71, "elapsed_time": "2:07:46", "remaining_time": "19:48:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 420, "total_steps": 4274, "loss": 0.0923, "learning_rate": 4.2e-05, "epoch": 0.19651421218856008, "percentage": 9.83, "elapsed_time": "2:09:17", "remaining_time": "19:46:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 425, "total_steps": 4274, "loss": 0.081, "learning_rate": 4.25e-05, "epoch": 0.19885366709556673, "percentage": 9.94, "elapsed_time": "2:10:50", "remaining_time": "19:44:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 430, "total_steps": 4274, "loss": 0.0808, "learning_rate": 4.3e-05, "epoch": 0.2011931220025734, "percentage": 10.06, "elapsed_time": "2:12:22", "remaining_time": "19:43:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 435, "total_steps": 4274, "loss": 0.1383, "learning_rate": 4.35e-05, "epoch": 0.20353257690958007, "percentage": 10.18, "elapsed_time": "2:13:54", "remaining_time": "19:41:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 440, "total_steps": 4274, "loss": 0.0836, "learning_rate": 4.4000000000000006e-05, "epoch": 0.20587203181658673, "percentage": 10.29, "elapsed_time": "2:15:26", "remaining_time": "19:40:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 445, "total_steps": 4274, "loss": 0.0796, "learning_rate": 4.4500000000000004e-05, "epoch": 0.2082114867235934, "percentage": 10.41, "elapsed_time": "2:16:57", "remaining_time": "19:38:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 450, "total_steps": 4274, "loss": 0.09, "learning_rate": 4.5e-05, "epoch": 0.21055094163060006, "percentage": 10.53, "elapsed_time": "2:18:29", "remaining_time": "19:36:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 455, "total_steps": 4274, "loss": 0.0928, "learning_rate": 4.55e-05, "epoch": 0.21289039653760675, "percentage": 10.65, "elapsed_time": "2:20:01", "remaining_time": "19:35:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 460, "total_steps": 4274, "loss": 0.0914, "learning_rate": 4.600000000000001e-05, "epoch": 0.2152298514446134, "percentage": 10.76, "elapsed_time": "2:21:34", "remaining_time": "19:33:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 465, "total_steps": 4274, "loss": 0.0978, "learning_rate": 4.6500000000000005e-05, "epoch": 0.21756930635162008, "percentage": 10.88, "elapsed_time": "2:23:06", "remaining_time": "19:32:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 470, "total_steps": 4274, "loss": 0.1109, "learning_rate": 4.7e-05, "epoch": 0.21990876125862674, "percentage": 11.0, "elapsed_time": "2:24:39", "remaining_time": "19:30:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 475, "total_steps": 4274, "loss": 0.0804, "learning_rate": 4.75e-05, "epoch": 0.2222482161656334, "percentage": 11.11, "elapsed_time": "2:26:11", "remaining_time": "19:29:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 480, "total_steps": 4274, "loss": 0.0737, "learning_rate": 4.8e-05, "epoch": 0.22458767107264008, "percentage": 11.23, "elapsed_time": "2:27:43", "remaining_time": "19:27:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 485, "total_steps": 4274, "loss": 0.0801, "learning_rate": 4.85e-05, "epoch": 0.22692712597964673, "percentage": 11.35, "elapsed_time": "2:29:15", "remaining_time": "19:26:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 490, "total_steps": 4274, "loss": 0.0784, "learning_rate": 4.9e-05, "epoch": 0.22926658088665341, "percentage": 11.46, "elapsed_time": "2:30:47", "remaining_time": "19:24:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 495, "total_steps": 4274, "loss": 0.0838, "learning_rate": 4.9500000000000004e-05, "epoch": 0.23160603579366007, "percentage": 11.58, "elapsed_time": "2:32:19", "remaining_time": "19:22:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 4274, "loss": 0.0893, "learning_rate": 5e-05, "epoch": 0.23394549070066675, "percentage": 11.7, "elapsed_time": "2:33:52", "remaining_time": "19:21:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 4274, "eval_loss": 0.09720832854509354, "epoch": 0.23394549070066675, "percentage": 11.7, "elapsed_time": "2:37:47", "remaining_time": "19:51:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 505, "total_steps": 4274, "loss": 0.0818, "learning_rate": 4.999978345640206e-05, "epoch": 0.2362849456076734, "percentage": 11.82, "elapsed_time": "2:39:20", "remaining_time": "19:49:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 510, "total_steps": 4274, "loss": 0.0812, "learning_rate": 4.9999133829359514e-05, "epoch": 0.2386244005146801, "percentage": 11.93, "elapsed_time": "2:40:53", "remaining_time": "19:47:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 515, "total_steps": 4274, "loss": 0.0732, "learning_rate": 4.999805113012618e-05, "epoch": 0.24096385542168675, "percentage": 12.05, "elapsed_time": "2:42:25", "remaining_time": "19:45:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 520, "total_steps": 4274, "loss": 0.0843, "learning_rate": 4.999653537745817e-05, "epoch": 0.2433033103286934, "percentage": 12.17, "elapsed_time": "2:43:58", "remaining_time": "19:43:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 525, "total_steps": 4274, "loss": 0.07, "learning_rate": 4.999458659761363e-05, "epoch": 0.24564276523570008, "percentage": 12.28, "elapsed_time": "2:45:30", "remaining_time": "19:41:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 530, "total_steps": 4274, "loss": 0.0852, "learning_rate": 4.9992204824352213e-05, "epoch": 0.24798222014270674, "percentage": 12.4, "elapsed_time": "2:47:03", "remaining_time": "19:40:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 535, "total_steps": 4274, "loss": 0.0847, "learning_rate": 4.998939009893453e-05, "epoch": 0.2503216750497134, "percentage": 12.52, "elapsed_time": "2:48:35", "remaining_time": "19:38:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 540, "total_steps": 4274, "loss": 0.0727, "learning_rate": 4.9986142470121453e-05, "epoch": 0.2526611299567201, "percentage": 12.63, "elapsed_time": "2:50:08", "remaining_time": "19:36:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 545, "total_steps": 4274, "loss": 0.0893, "learning_rate": 4.9982461994173234e-05, "epoch": 0.25500058486372673, "percentage": 12.75, "elapsed_time": "2:51:40", "remaining_time": "19:34:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 550, "total_steps": 4274, "loss": 0.0764, "learning_rate": 4.9978348734848556e-05, "epoch": 0.25734003977073344, "percentage": 12.87, "elapsed_time": "2:53:13", "remaining_time": "19:32:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 555, "total_steps": 4274, "loss": 0.1747, "learning_rate": 4.9973802763403424e-05, "epoch": 0.2596794946777401, "percentage": 12.99, "elapsed_time": "2:54:45", "remaining_time": "19:31:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 560, "total_steps": 4274, "loss": 0.0808, "learning_rate": 4.996882415858991e-05, "epoch": 0.26201894958474675, "percentage": 13.1, "elapsed_time": "2:56:18", "remaining_time": "19:29:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 565, "total_steps": 4274, "loss": 0.0847, "learning_rate": 4.996341300665481e-05, "epoch": 0.2643584044917534, "percentage": 13.22, "elapsed_time": "2:57:50", "remaining_time": "19:27:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 570, "total_steps": 4274, "loss": 0.0884, "learning_rate": 4.995756940133817e-05, "epoch": 0.26669785939876006, "percentage": 13.34, "elapsed_time": "2:59:22", "remaining_time": "19:25:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 575, "total_steps": 4274, "loss": 0.0784, "learning_rate": 4.995129344387158e-05, "epoch": 0.2690373143057668, "percentage": 13.45, "elapsed_time": "3:00:55", "remaining_time": "19:23:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 580, "total_steps": 4274, "loss": 0.146, "learning_rate": 4.9944585242976546e-05, "epoch": 0.2713767692127734, "percentage": 13.57, "elapsed_time": "3:02:27", "remaining_time": "19:22:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 585, "total_steps": 4274, "loss": 0.0894, "learning_rate": 4.99374449148625e-05, "epoch": 0.2737162241197801, "percentage": 13.69, "elapsed_time": "3:03:58", "remaining_time": "19:20:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 590, "total_steps": 4274, "loss": 0.0821, "learning_rate": 4.992987258322481e-05, "epoch": 0.27605567902678674, "percentage": 13.8, "elapsed_time": "3:05:30", "remaining_time": "19:18:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 595, "total_steps": 4274, "loss": 0.098, "learning_rate": 4.992186837924268e-05, "epoch": 0.27839513393379345, "percentage": 13.92, "elapsed_time": "3:07:03", "remaining_time": "19:16:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 600, "total_steps": 4274, "loss": 0.0776, "learning_rate": 4.9913432441576845e-05, "epoch": 0.2807345888408001, "percentage": 14.04, "elapsed_time": "3:08:35", "remaining_time": "19:14:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 605, "total_steps": 4274, "loss": 0.0862, "learning_rate": 4.9904564916367166e-05, "epoch": 0.28307404374780676, "percentage": 14.16, "elapsed_time": "3:10:07", "remaining_time": "19:13:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 610, "total_steps": 4274, "loss": 0.0921, "learning_rate": 4.989526595723012e-05, "epoch": 0.2854134986548134, "percentage": 14.27, "elapsed_time": "3:11:40", "remaining_time": "19:11:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 615, "total_steps": 4274, "loss": 0.0811, "learning_rate": 4.988553572525609e-05, "epoch": 0.28775295356182007, "percentage": 14.39, "elapsed_time": "3:13:12", "remaining_time": "19:09:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 620, "total_steps": 4274, "loss": 0.0711, "learning_rate": 4.987537438900664e-05, "epoch": 0.2900924084688268, "percentage": 14.51, "elapsed_time": "3:14:45", "remaining_time": "19:07:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 625, "total_steps": 4274, "loss": 0.0802, "learning_rate": 4.986478212451157e-05, "epoch": 0.29243186337583343, "percentage": 14.62, "elapsed_time": "3:16:18", "remaining_time": "19:06:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 630, "total_steps": 4274, "loss": 0.0825, "learning_rate": 4.985375911526582e-05, "epoch": 0.2947713182828401, "percentage": 14.74, "elapsed_time": "3:17:50", "remaining_time": "19:04:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 635, "total_steps": 4274, "loss": 0.0676, "learning_rate": 4.9842305552226365e-05, "epoch": 0.29711077318984674, "percentage": 14.86, "elapsed_time": "3:19:22", "remaining_time": "19:02:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 640, "total_steps": 4274, "loss": 0.0752, "learning_rate": 4.9830421633808876e-05, "epoch": 0.29945022809685345, "percentage": 14.97, "elapsed_time": "3:20:55", "remaining_time": "19:00:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 645, "total_steps": 4274, "loss": 0.0823, "learning_rate": 4.981810756588426e-05, "epoch": 0.3017896830038601, "percentage": 15.09, "elapsed_time": "3:22:27", "remaining_time": "18:59:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 650, "total_steps": 4274, "loss": 0.0795, "learning_rate": 4.980536356177513e-05, "epoch": 0.30412913791086676, "percentage": 15.21, "elapsed_time": "3:24:00", "remaining_time": "18:57:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 655, "total_steps": 4274, "loss": 0.0711, "learning_rate": 4.9792189842252074e-05, "epoch": 0.3064685928178734, "percentage": 15.33, "elapsed_time": "3:25:32", "remaining_time": "18:55:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 660, "total_steps": 4274, "loss": 0.0756, "learning_rate": 4.977858663552987e-05, "epoch": 0.30880804772488013, "percentage": 15.44, "elapsed_time": "3:27:05", "remaining_time": "18:53:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 665, "total_steps": 4274, "loss": 0.0742, "learning_rate": 4.9764554177263507e-05, "epoch": 0.3111475026318868, "percentage": 15.56, "elapsed_time": "3:28:37", "remaining_time": "18:52:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 670, "total_steps": 4274, "loss": 0.0698, "learning_rate": 4.975009271054409e-05, "epoch": 0.31348695753889344, "percentage": 15.68, "elapsed_time": "3:30:10", "remaining_time": "18:50:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 675, "total_steps": 4274, "loss": 0.0858, "learning_rate": 4.973520248589469e-05, "epoch": 0.3158264124459001, "percentage": 15.79, "elapsed_time": "3:31:42", "remaining_time": "18:48:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 680, "total_steps": 4274, "loss": 0.0818, "learning_rate": 4.9719883761265906e-05, "epoch": 0.31816586735290675, "percentage": 15.91, "elapsed_time": "3:33:15", "remaining_time": "18:47:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 685, "total_steps": 4274, "loss": 0.0671, "learning_rate": 4.9704136802031485e-05, "epoch": 0.32050532225991346, "percentage": 16.03, "elapsed_time": "3:34:47", "remaining_time": "18:45:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 690, "total_steps": 4274, "loss": 0.0725, "learning_rate": 4.968796188098369e-05, "epoch": 0.3228447771669201, "percentage": 16.14, "elapsed_time": "3:36:20", "remaining_time": "18:43:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 695, "total_steps": 4274, "loss": 0.0852, "learning_rate": 4.967135927832856e-05, "epoch": 0.32518423207392677, "percentage": 16.26, "elapsed_time": "3:37:52", "remaining_time": "18:41:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 700, "total_steps": 4274, "loss": 0.0754, "learning_rate": 4.9654329281681094e-05, "epoch": 0.3275236869809334, "percentage": 16.38, "elapsed_time": "3:39:24", "remaining_time": "18:40:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 705, "total_steps": 4274, "loss": 0.0961, "learning_rate": 4.9636872186060215e-05, "epoch": 0.32986314188794014, "percentage": 16.5, "elapsed_time": "3:40:57", "remaining_time": "18:38:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 710, "total_steps": 4274, "loss": 0.1093, "learning_rate": 4.961898829388372e-05, "epoch": 0.3322025967949468, "percentage": 16.61, "elapsed_time": "3:42:29", "remaining_time": "18:36:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 715, "total_steps": 4274, "loss": 0.0793, "learning_rate": 4.960067791496299e-05, "epoch": 0.33454205170195345, "percentage": 16.73, "elapsed_time": "3:44:01", "remaining_time": "18:35:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 720, "total_steps": 4274, "loss": 0.0805, "learning_rate": 4.958194136649765e-05, "epoch": 0.3368815066089601, "percentage": 16.85, "elapsed_time": "3:45:33", "remaining_time": "18:33:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 725, "total_steps": 4274, "loss": 0.073, "learning_rate": 4.956277897307008e-05, "epoch": 0.33922096151596676, "percentage": 16.96, "elapsed_time": "3:47:05", "remaining_time": "18:31:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 730, "total_steps": 4274, "loss": 0.0769, "learning_rate": 4.954319106663976e-05, "epoch": 0.34156041642297347, "percentage": 17.08, "elapsed_time": "3:48:37", "remaining_time": "18:29:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 735, "total_steps": 4274, "loss": 0.075, "learning_rate": 4.952317798653755e-05, "epoch": 0.3438998713299801, "percentage": 17.2, "elapsed_time": "3:50:09", "remaining_time": "18:28:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 740, "total_steps": 4274, "loss": 0.0697, "learning_rate": 4.95027400794598e-05, "epoch": 0.3462393262369868, "percentage": 17.31, "elapsed_time": "3:51:40", "remaining_time": "18:26:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 745, "total_steps": 4274, "loss": 0.08, "learning_rate": 4.948187769946234e-05, "epoch": 0.34857878114399343, "percentage": 17.43, "elapsed_time": "3:53:13", "remaining_time": "18:24:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 750, "total_steps": 4274, "loss": 0.0963, "learning_rate": 4.946059120795437e-05, "epoch": 0.35091823605100014, "percentage": 17.55, "elapsed_time": "3:54:45", "remaining_time": "18:23:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 755, "total_steps": 4274, "loss": 0.0913, "learning_rate": 4.943888097369216e-05, "epoch": 0.3532576909580068, "percentage": 17.66, "elapsed_time": "3:56:17", "remaining_time": "18:21:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 760, "total_steps": 4274, "loss": 0.0845, "learning_rate": 4.941674737277268e-05, "epoch": 0.35559714586501345, "percentage": 17.78, "elapsed_time": "3:57:49", "remaining_time": "18:19:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 765, "total_steps": 4274, "loss": 0.0756, "learning_rate": 4.9394190788627106e-05, "epoch": 0.3579366007720201, "percentage": 17.9, "elapsed_time": "3:59:20", "remaining_time": "18:17:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 770, "total_steps": 4274, "loss": 0.0803, "learning_rate": 4.937121161201415e-05, "epoch": 0.36027605567902676, "percentage": 18.02, "elapsed_time": "4:00:51", "remaining_time": "18:16:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 775, "total_steps": 4274, "loss": 0.0823, "learning_rate": 4.9347810241013294e-05, "epoch": 0.3626155105860335, "percentage": 18.13, "elapsed_time": "4:02:23", "remaining_time": "18:14:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 780, "total_steps": 4274, "loss": 0.0703, "learning_rate": 4.932398708101791e-05, "epoch": 0.36495496549304013, "percentage": 18.25, "elapsed_time": "4:03:55", "remaining_time": "18:12:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 785, "total_steps": 4274, "loss": 0.084, "learning_rate": 4.92997425447282e-05, "epoch": 0.3672944204000468, "percentage": 18.37, "elapsed_time": "4:05:27", "remaining_time": "18:10:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 790, "total_steps": 4274, "loss": 0.0868, "learning_rate": 4.927507705214412e-05, "epoch": 0.36963387530705344, "percentage": 18.48, "elapsed_time": "4:06:59", "remaining_time": "18:09:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 795, "total_steps": 4274, "loss": 0.0754, "learning_rate": 4.924999103055802e-05, "epoch": 0.37197333021406015, "percentage": 18.6, "elapsed_time": "4:08:32", "remaining_time": "18:07:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 800, "total_steps": 4274, "loss": 0.0803, "learning_rate": 4.922448491454729e-05, "epoch": 0.3743127851210668, "percentage": 18.72, "elapsed_time": "4:10:04", "remaining_time": "18:05:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 805, "total_steps": 4274, "loss": 0.0751, "learning_rate": 4.9198559145966824e-05, "epoch": 0.37665224002807346, "percentage": 18.83, "elapsed_time": "4:11:36", "remaining_time": "18:04:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 810, "total_steps": 4274, "loss": 0.0798, "learning_rate": 4.917221417394134e-05, "epoch": 0.3789916949350801, "percentage": 18.95, "elapsed_time": "4:13:05", "remaining_time": "18:02:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 815, "total_steps": 4274, "loss": 0.0776, "learning_rate": 4.914545045485767e-05, "epoch": 0.38133114984208677, "percentage": 19.07, "elapsed_time": "4:14:38", "remaining_time": "18:00:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 820, "total_steps": 4274, "loss": 0.0767, "learning_rate": 4.911826845235676e-05, "epoch": 0.3836706047490935, "percentage": 19.19, "elapsed_time": "4:16:10", "remaining_time": "17:59:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 825, "total_steps": 4274, "loss": 0.0752, "learning_rate": 4.9090668637325696e-05, "epoch": 0.38601005965610014, "percentage": 19.3, "elapsed_time": "4:17:42", "remaining_time": "17:57:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 830, "total_steps": 4274, "loss": 0.0722, "learning_rate": 4.906265148788954e-05, "epoch": 0.3883495145631068, "percentage": 19.42, "elapsed_time": "4:19:14", "remaining_time": "17:55:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 835, "total_steps": 4274, "loss": 0.0764, "learning_rate": 4.9034217489403045e-05, "epoch": 0.39068896947011345, "percentage": 19.54, "elapsed_time": "4:20:46", "remaining_time": "17:53:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 840, "total_steps": 4274, "loss": 0.0767, "learning_rate": 4.9005367134442235e-05, "epoch": 0.39302842437712016, "percentage": 19.65, "elapsed_time": "4:22:17", "remaining_time": "17:52:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 845, "total_steps": 4274, "loss": 0.0815, "learning_rate": 4.8976100922795884e-05, "epoch": 0.3953678792841268, "percentage": 19.77, "elapsed_time": "4:23:50", "remaining_time": "17:50:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 850, "total_steps": 4274, "loss": 0.0935, "learning_rate": 4.894641936145686e-05, "epoch": 0.39770733419113347, "percentage": 19.89, "elapsed_time": "4:25:22", "remaining_time": "17:49:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 855, "total_steps": 4274, "loss": 0.0767, "learning_rate": 4.8916322964613316e-05, "epoch": 0.4000467890981401, "percentage": 20.0, "elapsed_time": "4:26:53", "remaining_time": "17:47:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 860, "total_steps": 4274, "loss": 0.0771, "learning_rate": 4.888581225363982e-05, "epoch": 0.4023862440051468, "percentage": 20.12, "elapsed_time": "4:28:26", "remaining_time": "17:45:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 865, "total_steps": 4274, "loss": 0.0798, "learning_rate": 4.885488775708831e-05, "epoch": 0.4047256989121535, "percentage": 20.24, "elapsed_time": "4:29:58", "remaining_time": "17:44:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 870, "total_steps": 4274, "loss": 0.0723, "learning_rate": 4.882355001067892e-05, "epoch": 0.40706515381916014, "percentage": 20.36, "elapsed_time": "4:31:31", "remaining_time": "17:42:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 875, "total_steps": 4274, "loss": 0.0777, "learning_rate": 4.879179955729072e-05, "epoch": 0.4094046087261668, "percentage": 20.47, "elapsed_time": "4:33:04", "remaining_time": "17:40:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 880, "total_steps": 4274, "loss": 0.0777, "learning_rate": 4.87596369469523e-05, "epoch": 0.41174406363317345, "percentage": 20.59, "elapsed_time": "4:34:36", "remaining_time": "17:39:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 885, "total_steps": 4274, "loss": 0.0648, "learning_rate": 4.872706273683225e-05, "epoch": 0.41408351854018016, "percentage": 20.71, "elapsed_time": "4:36:08", "remaining_time": "17:37:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 890, "total_steps": 4274, "loss": 0.0688, "learning_rate": 4.869407749122951e-05, "epoch": 0.4164229734471868, "percentage": 20.82, "elapsed_time": "4:37:41", "remaining_time": "17:35:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 895, "total_steps": 4274, "loss": 0.0715, "learning_rate": 4.8660681781563576e-05, "epoch": 0.41876242835419347, "percentage": 20.94, "elapsed_time": "4:39:13", "remaining_time": "17:34:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 900, "total_steps": 4274, "loss": 0.0651, "learning_rate": 4.8626876186364624e-05, "epoch": 0.4211018832612001, "percentage": 21.06, "elapsed_time": "4:40:46", "remaining_time": "17:32:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 905, "total_steps": 4274, "loss": 0.0754, "learning_rate": 4.859266129126345e-05, "epoch": 0.4234413381682068, "percentage": 21.17, "elapsed_time": "4:42:18", "remaining_time": "17:30:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 910, "total_steps": 4274, "loss": 0.1229, "learning_rate": 4.85580376889814e-05, "epoch": 0.4257807930752135, "percentage": 21.29, "elapsed_time": "4:43:51", "remaining_time": "17:29:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 915, "total_steps": 4274, "loss": 0.0755, "learning_rate": 4.8523005979320015e-05, "epoch": 0.42812024798222015, "percentage": 21.41, "elapsed_time": "4:45:22", "remaining_time": "17:27:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 920, "total_steps": 4274, "loss": 0.0848, "learning_rate": 4.848756676915069e-05, "epoch": 0.4304597028892268, "percentage": 21.53, "elapsed_time": "4:46:55", "remaining_time": "17:26:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 925, "total_steps": 4274, "loss": 0.0764, "learning_rate": 4.845172067240415e-05, "epoch": 0.43279915779623346, "percentage": 21.64, "elapsed_time": "4:48:27", "remaining_time": "17:24:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 930, "total_steps": 4274, "loss": 0.0821, "learning_rate": 4.841546831005982e-05, "epoch": 0.43513861270324017, "percentage": 21.76, "elapsed_time": "4:50:00", "remaining_time": "17:22:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 935, "total_steps": 4274, "loss": 0.0755, "learning_rate": 4.837881031013506e-05, "epoch": 0.4374780676102468, "percentage": 21.88, "elapsed_time": "4:51:33", "remaining_time": "17:21:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 940, "total_steps": 4274, "loss": 0.0695, "learning_rate": 4.8341747307674276e-05, "epoch": 0.4398175225172535, "percentage": 21.99, "elapsed_time": "4:53:05", "remaining_time": "17:19:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 945, "total_steps": 4274, "loss": 0.086, "learning_rate": 4.8304279944737954e-05, "epoch": 0.44215697742426013, "percentage": 22.11, "elapsed_time": "4:54:38", "remaining_time": "17:17:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 950, "total_steps": 4274, "loss": 0.0797, "learning_rate": 4.8266408870391484e-05, "epoch": 0.4444964323312668, "percentage": 22.23, "elapsed_time": "4:56:10", "remaining_time": "17:16:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 955, "total_steps": 4274, "loss": 0.084, "learning_rate": 4.822813474069398e-05, "epoch": 0.4468358872382735, "percentage": 22.34, "elapsed_time": "4:57:43", "remaining_time": "17:14:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 960, "total_steps": 4274, "loss": 0.0754, "learning_rate": 4.8189458218686846e-05, "epoch": 0.44917534214528015, "percentage": 22.46, "elapsed_time": "4:59:15", "remaining_time": "17:13:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 965, "total_steps": 4274, "loss": 0.0762, "learning_rate": 4.815037997438234e-05, "epoch": 0.4515147970522868, "percentage": 22.58, "elapsed_time": "5:00:47", "remaining_time": "17:11:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 970, "total_steps": 4274, "loss": 0.0686, "learning_rate": 4.811090068475197e-05, "epoch": 0.45385425195929346, "percentage": 22.7, "elapsed_time": "5:02:20", "remaining_time": "17:09:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 975, "total_steps": 4274, "loss": 0.0818, "learning_rate": 4.807102103371472e-05, "epoch": 0.4561937068663002, "percentage": 22.81, "elapsed_time": "5:03:51", "remaining_time": "17:08:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 980, "total_steps": 4274, "loss": 0.0693, "learning_rate": 4.803074171212524e-05, "epoch": 0.45853316177330683, "percentage": 22.93, "elapsed_time": "5:05:23", "remaining_time": "17:06:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 985, "total_steps": 4274, "loss": 0.0642, "learning_rate": 4.799006341776185e-05, "epoch": 0.4608726166803135, "percentage": 23.05, "elapsed_time": "5:06:56", "remaining_time": "17:04:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 990, "total_steps": 4274, "loss": 0.0759, "learning_rate": 4.7948986855314523e-05, "epoch": 0.46321207158732014, "percentage": 23.16, "elapsed_time": "5:08:28", "remaining_time": "17:03:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 995, "total_steps": 4274, "loss": 0.0761, "learning_rate": 4.790751273637256e-05, "epoch": 0.4655515264943268, "percentage": 23.28, "elapsed_time": "5:10:01", "remaining_time": "17:01:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 4274, "loss": 0.0753, "learning_rate": 4.7865641779412366e-05, "epoch": 0.4678909814013335, "percentage": 23.4, "elapsed_time": "5:11:33", "remaining_time": "17:00:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 4274, "eval_loss": 0.08787659555673599, "epoch": 0.4678909814013335, "percentage": 23.4, "elapsed_time": "5:15:30", "remaining_time": "17:12:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1005, "total_steps": 4274, "loss": 0.0731, "learning_rate": 4.7823374709784955e-05, "epoch": 0.47023043630834016, "percentage": 23.51, "elapsed_time": "5:17:03", "remaining_time": "17:11:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1010, "total_steps": 4274, "loss": 0.082, "learning_rate": 4.77807122597034e-05, "epoch": 0.4725698912153468, "percentage": 23.63, "elapsed_time": "5:18:36", "remaining_time": "17:09:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1015, "total_steps": 4274, "loss": 0.0775, "learning_rate": 4.7737655168230114e-05, "epoch": 0.47490934612235347, "percentage": 23.75, "elapsed_time": "5:20:08", "remaining_time": "17:07:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1020, "total_steps": 4274, "loss": 0.0805, "learning_rate": 4.769420418126412e-05, "epoch": 0.4772488010293602, "percentage": 23.87, "elapsed_time": "5:21:41", "remaining_time": "17:06:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1025, "total_steps": 4274, "loss": 0.0754, "learning_rate": 4.7650360051528054e-05, "epoch": 0.47958825593636684, "percentage": 23.98, "elapsed_time": "5:23:13", "remaining_time": "17:04:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1030, "total_steps": 4274, "loss": 0.068, "learning_rate": 4.760612353855517e-05, "epoch": 0.4819277108433735, "percentage": 24.1, "elapsed_time": "5:24:45", "remaining_time": "17:02:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1035, "total_steps": 4274, "loss": 0.073, "learning_rate": 4.7561495408676146e-05, "epoch": 0.48426716575038015, "percentage": 24.22, "elapsed_time": "5:26:18", "remaining_time": "17:01:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1040, "total_steps": 4274, "loss": 0.074, "learning_rate": 4.7516476435005865e-05, "epoch": 0.4866066206573868, "percentage": 24.33, "elapsed_time": "5:27:50", "remaining_time": "16:59:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1045, "total_steps": 4274, "loss": 0.0734, "learning_rate": 4.7471067397429956e-05, "epoch": 0.4889460755643935, "percentage": 24.45, "elapsed_time": "5:29:23", "remaining_time": "16:57:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1050, "total_steps": 4274, "loss": 0.0763, "learning_rate": 4.742526908259134e-05, "epoch": 0.49128553047140017, "percentage": 24.57, "elapsed_time": "5:30:55", "remaining_time": "16:56:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1055, "total_steps": 4274, "loss": 0.0703, "learning_rate": 4.7379082283876566e-05, "epoch": 0.4936249853784068, "percentage": 24.68, "elapsed_time": "5:32:28", "remaining_time": "16:54:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1060, "total_steps": 4274, "loss": 0.0863, "learning_rate": 4.733250780140206e-05, "epoch": 0.4959644402854135, "percentage": 24.8, "elapsed_time": "5:34:01", "remaining_time": "16:52:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1065, "total_steps": 4274, "loss": 0.1042, "learning_rate": 4.728554644200034e-05, "epoch": 0.4983038951924202, "percentage": 24.92, "elapsed_time": "5:35:33", "remaining_time": "16:51:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1070, "total_steps": 4274, "loss": 0.0743, "learning_rate": 4.723819901920591e-05, "epoch": 0.5006433500994268, "percentage": 25.04, "elapsed_time": "5:37:05", "remaining_time": "16:49:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1075, "total_steps": 4274, "loss": 0.0707, "learning_rate": 4.719046635324129e-05, "epoch": 0.5029828050064336, "percentage": 25.15, "elapsed_time": "5:38:38", "remaining_time": "16:47:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1080, "total_steps": 4274, "loss": 0.0738, "learning_rate": 4.7142349271002735e-05, "epoch": 0.5053222599134402, "percentage": 25.27, "elapsed_time": "5:40:10", "remaining_time": "16:46:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1085, "total_steps": 4274, "loss": 0.0724, "learning_rate": 4.709384860604593e-05, "epoch": 0.5076617148204469, "percentage": 25.39, "elapsed_time": "5:41:42", "remaining_time": "16:44:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1090, "total_steps": 4274, "loss": 0.0738, "learning_rate": 4.704496519857156e-05, "epoch": 0.5100011697274535, "percentage": 25.5, "elapsed_time": "5:43:14", "remaining_time": "16:42:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1095, "total_steps": 4274, "loss": 0.0699, "learning_rate": 4.699569989541074e-05, "epoch": 0.5123406246344602, "percentage": 25.62, "elapsed_time": "5:44:46", "remaining_time": "16:40:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1100, "total_steps": 4274, "loss": 0.072, "learning_rate": 4.694605355001034e-05, "epoch": 0.5146800795414669, "percentage": 25.74, "elapsed_time": "5:46:18", "remaining_time": "16:39:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1105, "total_steps": 4274, "loss": 0.0761, "learning_rate": 4.689602702241823e-05, "epoch": 0.5170195344484735, "percentage": 25.85, "elapsed_time": "5:47:50", "remaining_time": "16:37:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1110, "total_steps": 4274, "loss": 0.0793, "learning_rate": 4.684562117926835e-05, "epoch": 0.5193589893554802, "percentage": 25.97, "elapsed_time": "5:49:23", "remaining_time": "16:35:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1115, "total_steps": 4274, "loss": 0.0731, "learning_rate": 4.679483689376571e-05, "epoch": 0.5216984442624868, "percentage": 26.09, "elapsed_time": "5:50:55", "remaining_time": "16:34:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1120, "total_steps": 4274, "loss": 0.0795, "learning_rate": 4.674367504567127e-05, "epoch": 0.5240378991694935, "percentage": 26.2, "elapsed_time": "5:52:28", "remaining_time": "16:32:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1125, "total_steps": 4274, "loss": 0.0716, "learning_rate": 4.669213652128667e-05, "epoch": 0.5263773540765002, "percentage": 26.32, "elapsed_time": "5:54:00", "remaining_time": "16:30:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1130, "total_steps": 4274, "loss": 0.0717, "learning_rate": 4.664022221343892e-05, "epoch": 0.5287168089835068, "percentage": 26.44, "elapsed_time": "5:55:33", "remaining_time": "16:29:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1135, "total_steps": 4274, "loss": 0.0756, "learning_rate": 4.658793302146489e-05, "epoch": 0.5310562638905135, "percentage": 26.56, "elapsed_time": "5:57:05", "remaining_time": "16:27:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1140, "total_steps": 4274, "loss": 0.0789, "learning_rate": 4.653526985119577e-05, "epoch": 0.5333957187975201, "percentage": 26.67, "elapsed_time": "5:58:38", "remaining_time": "16:25:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1145, "total_steps": 4274, "loss": 0.0823, "learning_rate": 4.648223361494135e-05, "epoch": 0.5357351737045268, "percentage": 26.79, "elapsed_time": "6:00:10", "remaining_time": "16:24:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1150, "total_steps": 4274, "loss": 0.0804, "learning_rate": 4.642882523147422e-05, "epoch": 0.5380746286115335, "percentage": 26.91, "elapsed_time": "6:01:42", "remaining_time": "16:22:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1155, "total_steps": 4274, "loss": 0.0699, "learning_rate": 4.637504562601386e-05, "epoch": 0.5404140835185401, "percentage": 27.02, "elapsed_time": "6:03:13", "remaining_time": "16:20:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1160, "total_steps": 4274, "loss": 0.0717, "learning_rate": 4.6320895730210616e-05, "epoch": 0.5427535384255469, "percentage": 27.14, "elapsed_time": "6:04:46", "remaining_time": "16:19:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1165, "total_steps": 4274, "loss": 0.0816, "learning_rate": 4.626637648212955e-05, "epoch": 0.5450929933325536, "percentage": 27.26, "elapsed_time": "6:06:18", "remaining_time": "16:17:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1170, "total_steps": 4274, "loss": 0.0766, "learning_rate": 4.6211488826234186e-05, "epoch": 0.5474324482395602, "percentage": 27.37, "elapsed_time": "6:07:51", "remaining_time": "16:15:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1175, "total_steps": 4274, "loss": 0.0728, "learning_rate": 4.615623371337016e-05, "epoch": 0.5497719031465669, "percentage": 27.49, "elapsed_time": "6:09:23", "remaining_time": "16:14:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1180, "total_steps": 4274, "loss": 0.073, "learning_rate": 4.6100612100748765e-05, "epoch": 0.5521113580535735, "percentage": 27.61, "elapsed_time": "6:10:54", "remaining_time": "16:12:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1185, "total_steps": 4274, "loss": 0.0803, "learning_rate": 4.604462495193031e-05, "epoch": 0.5544508129605802, "percentage": 27.73, "elapsed_time": "6:12:26", "remaining_time": "16:10:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1190, "total_steps": 4274, "loss": 0.0728, "learning_rate": 4.59882732368075e-05, "epoch": 0.5567902678675869, "percentage": 27.84, "elapsed_time": "6:13:58", "remaining_time": "16:09:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1195, "total_steps": 4274, "loss": 0.0697, "learning_rate": 4.593155793158859e-05, "epoch": 0.5591297227745935, "percentage": 27.96, "elapsed_time": "6:15:29", "remaining_time": "16:07:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1200, "total_steps": 4274, "loss": 0.0771, "learning_rate": 4.5874480018780456e-05, "epoch": 0.5614691776816002, "percentage": 28.08, "elapsed_time": "6:17:01", "remaining_time": "16:05:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1205, "total_steps": 4274, "loss": 0.0722, "learning_rate": 4.581704048717166e-05, "epoch": 0.5638086325886068, "percentage": 28.19, "elapsed_time": "6:18:33", "remaining_time": "16:04:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1210, "total_steps": 4274, "loss": 0.0739, "learning_rate": 4.57592403318152e-05, "epoch": 0.5661480874956135, "percentage": 28.31, "elapsed_time": "6:20:06", "remaining_time": "16:02:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1215, "total_steps": 4274, "loss": 0.076, "learning_rate": 4.570108055401138e-05, "epoch": 0.5684875424026202, "percentage": 28.43, "elapsed_time": "6:21:38", "remaining_time": "16:00:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1220, "total_steps": 4274, "loss": 0.0708, "learning_rate": 4.5642562161290406e-05, "epoch": 0.5708269973096268, "percentage": 28.54, "elapsed_time": "6:23:10", "remaining_time": "15:59:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1225, "total_steps": 4274, "loss": 0.0775, "learning_rate": 4.558368616739493e-05, "epoch": 0.5731664522166335, "percentage": 28.66, "elapsed_time": "6:24:41", "remaining_time": "15:57:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1230, "total_steps": 4274, "loss": 0.0758, "learning_rate": 4.552445359226252e-05, "epoch": 0.5755059071236401, "percentage": 28.78, "elapsed_time": "6:26:14", "remaining_time": "15:55:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1235, "total_steps": 4274, "loss": 0.0648, "learning_rate": 4.546486546200798e-05, "epoch": 0.5778453620306468, "percentage": 28.9, "elapsed_time": "6:27:46", "remaining_time": "15:54:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1240, "total_steps": 4274, "loss": 0.0684, "learning_rate": 4.540492280890555e-05, "epoch": 0.5801848169376536, "percentage": 29.01, "elapsed_time": "6:29:19", "remaining_time": "15:52:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1245, "total_steps": 4274, "loss": 0.0746, "learning_rate": 4.534462667137105e-05, "epoch": 0.5825242718446602, "percentage": 29.13, "elapsed_time": "6:30:50", "remaining_time": "15:50:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1250, "total_steps": 4274, "loss": 0.0771, "learning_rate": 4.528397809394389e-05, "epoch": 0.5848637267516669, "percentage": 29.25, "elapsed_time": "6:32:22", "remaining_time": "15:49:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1255, "total_steps": 4274, "loss": 0.0654, "learning_rate": 4.522297812726897e-05, "epoch": 0.5872031816586736, "percentage": 29.36, "elapsed_time": "6:33:54", "remaining_time": "15:47:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1260, "total_steps": 4274, "loss": 0.0759, "learning_rate": 4.516162782807845e-05, "epoch": 0.5895426365656802, "percentage": 29.48, "elapsed_time": "6:35:26", "remaining_time": "15:45:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1265, "total_steps": 4274, "loss": 0.0666, "learning_rate": 4.5099928259173516e-05, "epoch": 0.5918820914726869, "percentage": 29.6, "elapsed_time": "6:36:58", "remaining_time": "15:44:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1270, "total_steps": 4274, "loss": 0.0677, "learning_rate": 4.503788048940589e-05, "epoch": 0.5942215463796935, "percentage": 29.71, "elapsed_time": "6:38:31", "remaining_time": "15:42:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1275, "total_steps": 4274, "loss": 0.0697, "learning_rate": 4.497548559365935e-05, "epoch": 0.5965610012867002, "percentage": 29.83, "elapsed_time": "6:40:03", "remaining_time": "15:40:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1280, "total_steps": 4274, "loss": 0.072, "learning_rate": 4.4912744652831116e-05, "epoch": 0.5989004561937069, "percentage": 29.95, "elapsed_time": "6:41:35", "remaining_time": "15:39:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1285, "total_steps": 4274, "loss": 0.068, "learning_rate": 4.4849658753813126e-05, "epoch": 0.6012399111007135, "percentage": 30.07, "elapsed_time": "6:43:07", "remaining_time": "15:37:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1290, "total_steps": 4274, "loss": 0.0744, "learning_rate": 4.4786228989473164e-05, "epoch": 0.6035793660077202, "percentage": 30.18, "elapsed_time": "6:44:40", "remaining_time": "15:36:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1295, "total_steps": 4274, "loss": 0.0745, "learning_rate": 4.4722456458635995e-05, "epoch": 0.6059188209147268, "percentage": 30.3, "elapsed_time": "6:46:12", "remaining_time": "15:34:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1300, "total_steps": 4274, "loss": 0.0769, "learning_rate": 4.465834226606428e-05, "epoch": 0.6082582758217335, "percentage": 30.42, "elapsed_time": "6:47:45", "remaining_time": "15:32:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1305, "total_steps": 4274, "loss": 0.0809, "learning_rate": 4.459388752243945e-05, "epoch": 0.6105977307287402, "percentage": 30.53, "elapsed_time": "6:49:17", "remaining_time": "15:31:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1310, "total_steps": 4274, "loss": 0.08, "learning_rate": 4.452909334434247e-05, "epoch": 0.6129371856357468, "percentage": 30.65, "elapsed_time": "6:50:49", "remaining_time": "15:29:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1315, "total_steps": 4274, "loss": 0.0668, "learning_rate": 4.4463960854234506e-05, "epoch": 0.6152766405427536, "percentage": 30.77, "elapsed_time": "6:52:22", "remaining_time": "15:27:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1320, "total_steps": 4274, "loss": 0.0775, "learning_rate": 4.4398491180437446e-05, "epoch": 0.6176160954497603, "percentage": 30.88, "elapsed_time": "6:53:54", "remaining_time": "15:26:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1325, "total_steps": 4274, "loss": 0.0714, "learning_rate": 4.43326854571144e-05, "epoch": 0.6199555503567669, "percentage": 31.0, "elapsed_time": "6:55:27", "remaining_time": "15:24:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1330, "total_steps": 4274, "loss": 0.0789, "learning_rate": 4.426654482425e-05, "epoch": 0.6222950052637736, "percentage": 31.12, "elapsed_time": "6:56:59", "remaining_time": "15:23:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1335, "total_steps": 4274, "loss": 0.0641, "learning_rate": 4.420007042763071e-05, "epoch": 0.6246344601707802, "percentage": 31.24, "elapsed_time": "6:58:31", "remaining_time": "15:21:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1340, "total_steps": 4274, "loss": 0.0665, "learning_rate": 4.413326341882492e-05, "epoch": 0.6269739150777869, "percentage": 31.35, "elapsed_time": "7:00:04", "remaining_time": "15:19:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1345, "total_steps": 4274, "loss": 0.0784, "learning_rate": 4.4066124955163046e-05, "epoch": 0.6293133699847936, "percentage": 31.47, "elapsed_time": "7:01:36", "remaining_time": "15:18:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1350, "total_steps": 4274, "loss": 0.0675, "learning_rate": 4.3998656199717435e-05, "epoch": 0.6316528248918002, "percentage": 31.59, "elapsed_time": "7:03:09", "remaining_time": "15:16:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1355, "total_steps": 4274, "loss": 0.0711, "learning_rate": 4.393085832128226e-05, "epoch": 0.6339922797988069, "percentage": 31.7, "elapsed_time": "7:04:40", "remaining_time": "15:14:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1360, "total_steps": 4274, "loss": 0.0766, "learning_rate": 4.386273249435324e-05, "epoch": 0.6363317347058135, "percentage": 31.82, "elapsed_time": "7:06:10", "remaining_time": "15:13:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1365, "total_steps": 4274, "loss": 0.0746, "learning_rate": 4.3794279899107304e-05, "epoch": 0.6386711896128202, "percentage": 31.94, "elapsed_time": "7:07:42", "remaining_time": "15:11:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1370, "total_steps": 4274, "loss": 0.0638, "learning_rate": 4.3725501721382165e-05, "epoch": 0.6410106445198269, "percentage": 32.05, "elapsed_time": "7:09:15", "remaining_time": "15:09:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1375, "total_steps": 4274, "loss": 0.0698, "learning_rate": 4.3656399152655735e-05, "epoch": 0.6433500994268335, "percentage": 32.17, "elapsed_time": "7:10:47", "remaining_time": "15:08:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1380, "total_steps": 4274, "loss": 0.0952, "learning_rate": 4.358697339002553e-05, "epoch": 0.6456895543338402, "percentage": 32.29, "elapsed_time": "7:12:20", "remaining_time": "15:06:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1385, "total_steps": 4274, "loss": 0.0728, "learning_rate": 4.3517225636187906e-05, "epoch": 0.6480290092408468, "percentage": 32.41, "elapsed_time": "7:13:52", "remaining_time": "15:05:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1390, "total_steps": 4274, "loss": 0.0721, "learning_rate": 4.344715709941722e-05, "epoch": 0.6503684641478535, "percentage": 32.52, "elapsed_time": "7:15:23", "remaining_time": "15:03:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1395, "total_steps": 4274, "loss": 0.072, "learning_rate": 4.337676899354493e-05, "epoch": 0.6527079190548603, "percentage": 32.64, "elapsed_time": "7:16:56", "remaining_time": "15:01:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1400, "total_steps": 4274, "loss": 0.072, "learning_rate": 4.330606253793851e-05, "epoch": 0.6550473739618669, "percentage": 32.76, "elapsed_time": "7:18:28", "remaining_time": "15:00:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1405, "total_steps": 4274, "loss": 0.0747, "learning_rate": 4.32350389574804e-05, "epoch": 0.6573868288688736, "percentage": 32.87, "elapsed_time": "7:19:59", "remaining_time": "14:58:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1410, "total_steps": 4274, "loss": 0.0698, "learning_rate": 4.316369948254674e-05, "epoch": 0.6597262837758803, "percentage": 32.99, "elapsed_time": "7:21:32", "remaining_time": "14:56:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1415, "total_steps": 4274, "loss": 0.0683, "learning_rate": 4.3092045348986034e-05, "epoch": 0.6620657386828869, "percentage": 33.11, "elapsed_time": "7:23:04", "remaining_time": "14:55:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1420, "total_steps": 4274, "loss": 0.0672, "learning_rate": 4.302007779809781e-05, "epoch": 0.6644051935898936, "percentage": 33.22, "elapsed_time": "7:24:35", "remaining_time": "14:53:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1425, "total_steps": 4274, "loss": 0.0683, "learning_rate": 4.294779807661105e-05, "epoch": 0.6667446484969002, "percentage": 33.34, "elapsed_time": "7:26:07", "remaining_time": "14:51:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1430, "total_steps": 4274, "loss": 0.0728, "learning_rate": 4.287520743666263e-05, "epoch": 0.6690841034039069, "percentage": 33.46, "elapsed_time": "7:27:40", "remaining_time": "14:50:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1435, "total_steps": 4274, "loss": 0.077, "learning_rate": 4.280230713577564e-05, "epoch": 0.6714235583109136, "percentage": 33.58, "elapsed_time": "7:29:12", "remaining_time": "14:48:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1440, "total_steps": 4274, "loss": 0.0739, "learning_rate": 4.2729098436837536e-05, "epoch": 0.6737630132179202, "percentage": 33.69, "elapsed_time": "7:30:44", "remaining_time": "14:47:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1445, "total_steps": 4274, "loss": 0.079, "learning_rate": 4.2655582608078315e-05, "epoch": 0.6761024681249269, "percentage": 33.81, "elapsed_time": "7:32:17", "remaining_time": "14:45:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1450, "total_steps": 4274, "loss": 0.0653, "learning_rate": 4.2581760923048554e-05, "epoch": 0.6784419230319335, "percentage": 33.93, "elapsed_time": "7:33:49", "remaining_time": "14:43:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1455, "total_steps": 4274, "loss": 0.0793, "learning_rate": 4.2507634660597315e-05, "epoch": 0.6807813779389402, "percentage": 34.04, "elapsed_time": "7:35:20", "remaining_time": "14:42:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1460, "total_steps": 4274, "loss": 0.0689, "learning_rate": 4.243320510485001e-05, "epoch": 0.6831208328459469, "percentage": 34.16, "elapsed_time": "7:36:52", "remaining_time": "14:40:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1465, "total_steps": 4274, "loss": 0.0726, "learning_rate": 4.235847354518614e-05, "epoch": 0.6854602877529535, "percentage": 34.28, "elapsed_time": "7:38:25", "remaining_time": "14:38:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1470, "total_steps": 4274, "loss": 0.0691, "learning_rate": 4.228344127621695e-05, "epoch": 0.6877997426599602, "percentage": 34.39, "elapsed_time": "7:39:58", "remaining_time": "14:37:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1475, "total_steps": 4274, "loss": 0.0684, "learning_rate": 4.220810959776307e-05, "epoch": 0.6901391975669668, "percentage": 34.51, "elapsed_time": "7:41:30", "remaining_time": "14:35:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1480, "total_steps": 4274, "loss": 0.0624, "learning_rate": 4.213247981483189e-05, "epoch": 0.6924786524739736, "percentage": 34.63, "elapsed_time": "7:43:03", "remaining_time": "14:34:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1485, "total_steps": 4274, "loss": 0.0751, "learning_rate": 4.205655323759505e-05, "epoch": 0.6948181073809803, "percentage": 34.74, "elapsed_time": "7:44:35", "remaining_time": "14:32:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1490, "total_steps": 4274, "loss": 0.0704, "learning_rate": 4.1980331181365685e-05, "epoch": 0.6971575622879869, "percentage": 34.86, "elapsed_time": "7:46:08", "remaining_time": "14:30:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1495, "total_steps": 4274, "loss": 0.0717, "learning_rate": 4.190381496657565e-05, "epoch": 0.6994970171949936, "percentage": 34.98, "elapsed_time": "7:47:40", "remaining_time": "14:29:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1500, "total_steps": 4274, "loss": 0.071, "learning_rate": 4.182700591875267e-05, "epoch": 0.7018364721020003, "percentage": 35.1, "elapsed_time": "7:49:13", "remaining_time": "14:27:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1500, "total_steps": 4274, "eval_loss": 0.0841016098856926, "epoch": 0.7018364721020003, "percentage": 35.1, "elapsed_time": "7:53:10", "remaining_time": "14:35:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1505, "total_steps": 4274, "loss": 0.0733, "learning_rate": 4.1749905368497345e-05, "epoch": 0.7041759270090069, "percentage": 35.21, "elapsed_time": "7:54:43", "remaining_time": "14:33:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1510, "total_steps": 4274, "loss": 0.0646, "learning_rate": 4.1672514651460124e-05, "epoch": 0.7065153819160136, "percentage": 35.33, "elapsed_time": "7:56:15", "remaining_time": "14:31:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1515, "total_steps": 4274, "loss": 0.0623, "learning_rate": 4.159483510831814e-05, "epoch": 0.7088548368230202, "percentage": 35.45, "elapsed_time": "7:57:48", "remaining_time": "14:30:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1520, "total_steps": 4274, "loss": 0.0638, "learning_rate": 4.151686808475204e-05, "epoch": 0.7111942917300269, "percentage": 35.56, "elapsed_time": "7:59:20", "remaining_time": "14:28:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1525, "total_steps": 4274, "loss": 0.0714, "learning_rate": 4.143861493142258e-05, "epoch": 0.7135337466370336, "percentage": 35.68, "elapsed_time": "8:00:53", "remaining_time": "14:26:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1530, "total_steps": 4274, "loss": 0.0732, "learning_rate": 4.136007700394733e-05, "epoch": 0.7158732015440402, "percentage": 35.8, "elapsed_time": "8:02:25", "remaining_time": "14:25:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1535, "total_steps": 4274, "loss": 0.0705, "learning_rate": 4.128125566287711e-05, "epoch": 0.7182126564510469, "percentage": 35.91, "elapsed_time": "8:03:58", "remaining_time": "14:23:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1540, "total_steps": 4274, "loss": 0.0722, "learning_rate": 4.120215227367247e-05, "epoch": 0.7205521113580535, "percentage": 36.03, "elapsed_time": "8:05:30", "remaining_time": "14:21:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1545, "total_steps": 4274, "loss": 0.0654, "learning_rate": 4.112276820668001e-05, "epoch": 0.7228915662650602, "percentage": 36.15, "elapsed_time": "8:07:03", "remaining_time": "14:20:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1550, "total_steps": 4274, "loss": 0.0664, "learning_rate": 4.104310483710864e-05, "epoch": 0.725231021172067, "percentage": 36.27, "elapsed_time": "8:08:35", "remaining_time": "14:18:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1555, "total_steps": 4274, "loss": 0.0647, "learning_rate": 4.096316354500578e-05, "epoch": 0.7275704760790735, "percentage": 36.38, "elapsed_time": "8:10:07", "remaining_time": "14:17:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1560, "total_steps": 4274, "loss": 0.0647, "learning_rate": 4.0882945715233426e-05, "epoch": 0.7299099309860803, "percentage": 36.5, "elapsed_time": "8:11:40", "remaining_time": "14:15:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1565, "total_steps": 4274, "loss": 0.0681, "learning_rate": 4.080245273744419e-05, "epoch": 0.7322493858930869, "percentage": 36.62, "elapsed_time": "8:13:13", "remaining_time": "14:13:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1570, "total_steps": 4274, "loss": 0.0682, "learning_rate": 4.072168600605718e-05, "epoch": 0.7345888408000936, "percentage": 36.73, "elapsed_time": "8:14:45", "remaining_time": "14:12:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1575, "total_steps": 4274, "loss": 0.0626, "learning_rate": 4.064064692023389e-05, "epoch": 0.7369282957071003, "percentage": 36.85, "elapsed_time": "8:16:18", "remaining_time": "14:10:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1580, "total_steps": 4274, "loss": 0.1316, "learning_rate": 4.055933688385394e-05, "epoch": 0.7392677506141069, "percentage": 36.97, "elapsed_time": "8:17:50", "remaining_time": "14:08:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1585, "total_steps": 4274, "loss": 0.08, "learning_rate": 4.047775730549075e-05, "epoch": 0.7416072055211136, "percentage": 37.08, "elapsed_time": "8:19:23", "remaining_time": "14:07:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1590, "total_steps": 4274, "loss": 0.0722, "learning_rate": 4.039590959838715e-05, "epoch": 0.7439466604281203, "percentage": 37.2, "elapsed_time": "8:20:55", "remaining_time": "14:05:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1595, "total_steps": 4274, "loss": 0.0708, "learning_rate": 4.031379518043091e-05, "epoch": 0.7462861153351269, "percentage": 37.32, "elapsed_time": "8:22:28", "remaining_time": "14:03:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1600, "total_steps": 4274, "loss": 0.063, "learning_rate": 4.023141547413015e-05, "epoch": 0.7486255702421336, "percentage": 37.44, "elapsed_time": "8:24:00", "remaining_time": "14:02:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1605, "total_steps": 4274, "loss": 0.0766, "learning_rate": 4.0148771906588706e-05, "epoch": 0.7509650251491402, "percentage": 37.55, "elapsed_time": "8:25:33", "remaining_time": "14:00:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1610, "total_steps": 4274, "loss": 0.0663, "learning_rate": 4.0065865909481417e-05, "epoch": 0.7533044800561469, "percentage": 37.67, "elapsed_time": "8:27:05", "remaining_time": "13:59:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1615, "total_steps": 4274, "loss": 0.0804, "learning_rate": 3.9982698919029305e-05, "epoch": 0.7556439349631536, "percentage": 37.79, "elapsed_time": "8:28:37", "remaining_time": "13:57:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1620, "total_steps": 4274, "loss": 0.076, "learning_rate": 3.9899272375974726e-05, "epoch": 0.7579833898701602, "percentage": 37.9, "elapsed_time": "8:30:09", "remaining_time": "13:55:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1625, "total_steps": 4274, "loss": 0.0697, "learning_rate": 3.981558772555638e-05, "epoch": 0.7603228447771669, "percentage": 38.02, "elapsed_time": "8:31:42", "remaining_time": "13:54:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1630, "total_steps": 4274, "loss": 0.0752, "learning_rate": 3.97316464174843e-05, "epoch": 0.7626622996841735, "percentage": 38.14, "elapsed_time": "8:33:14", "remaining_time": "13:52:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1635, "total_steps": 4274, "loss": 0.0827, "learning_rate": 3.964744990591471e-05, "epoch": 0.7650017545911802, "percentage": 38.25, "elapsed_time": "8:34:46", "remaining_time": "13:50:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1640, "total_steps": 4274, "loss": 0.0716, "learning_rate": 3.956299964942485e-05, "epoch": 0.767341209498187, "percentage": 38.37, "elapsed_time": "8:36:19", "remaining_time": "13:49:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1645, "total_steps": 4274, "loss": 0.0763, "learning_rate": 3.947829711098772e-05, "epoch": 0.7696806644051936, "percentage": 38.49, "elapsed_time": "8:37:51", "remaining_time": "13:47:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1650, "total_steps": 4274, "loss": 0.0662, "learning_rate": 3.9393343757946696e-05, "epoch": 0.7720201193122003, "percentage": 38.61, "elapsed_time": "8:39:24", "remaining_time": "13:46:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1655, "total_steps": 4274, "loss": 0.0721, "learning_rate": 3.930814106199018e-05, "epoch": 0.7743595742192069, "percentage": 38.72, "elapsed_time": "8:40:56", "remaining_time": "13:44:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1660, "total_steps": 4274, "loss": 0.0755, "learning_rate": 3.922269049912602e-05, "epoch": 0.7766990291262136, "percentage": 38.84, "elapsed_time": "8:42:29", "remaining_time": "13:42:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1665, "total_steps": 4274, "loss": 0.0721, "learning_rate": 3.9136993549656006e-05, "epoch": 0.7790384840332203, "percentage": 38.96, "elapsed_time": "8:44:02", "remaining_time": "13:41:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1670, "total_steps": 4274, "loss": 0.0638, "learning_rate": 3.905105169815021e-05, "epoch": 0.7813779389402269, "percentage": 39.07, "elapsed_time": "8:45:34", "remaining_time": "13:39:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1675, "total_steps": 4274, "loss": 0.0779, "learning_rate": 3.896486643342124e-05, "epoch": 0.7837173938472336, "percentage": 39.19, "elapsed_time": "8:47:07", "remaining_time": "13:37:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1680, "total_steps": 4274, "loss": 0.0715, "learning_rate": 3.887843924849849e-05, "epoch": 0.7860568487542403, "percentage": 39.31, "elapsed_time": "8:48:39", "remaining_time": "13:36:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1685, "total_steps": 4274, "loss": 0.0719, "learning_rate": 3.879177164060225e-05, "epoch": 0.7883963036612469, "percentage": 39.42, "elapsed_time": "8:50:12", "remaining_time": "13:34:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1690, "total_steps": 4274, "loss": 0.0661, "learning_rate": 3.8704865111117746e-05, "epoch": 0.7907357585682536, "percentage": 39.54, "elapsed_time": "8:51:44", "remaining_time": "13:33:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1695, "total_steps": 4274, "loss": 0.0715, "learning_rate": 3.861772116556921e-05, "epoch": 0.7930752134752602, "percentage": 39.66, "elapsed_time": "8:53:17", "remaining_time": "13:31:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1700, "total_steps": 4274, "loss": 0.0726, "learning_rate": 3.853034131359371e-05, "epoch": 0.7954146683822669, "percentage": 39.78, "elapsed_time": "8:54:49", "remaining_time": "13:29:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1705, "total_steps": 4274, "loss": 0.0736, "learning_rate": 3.8442727068915066e-05, "epoch": 0.7977541232892736, "percentage": 39.89, "elapsed_time": "8:56:22", "remaining_time": "13:28:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1710, "total_steps": 4274, "loss": 0.0641, "learning_rate": 3.8354879949317546e-05, "epoch": 0.8000935781962802, "percentage": 40.01, "elapsed_time": "8:57:54", "remaining_time": "13:26:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1715, "total_steps": 4274, "loss": 0.07, "learning_rate": 3.8266801476619694e-05, "epoch": 0.802433033103287, "percentage": 40.13, "elapsed_time": "8:59:27", "remaining_time": "13:24:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1720, "total_steps": 4274, "loss": 0.069, "learning_rate": 3.817849317664784e-05, "epoch": 0.8047724880102936, "percentage": 40.24, "elapsed_time": "9:00:59", "remaining_time": "13:23:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1725, "total_steps": 4274, "loss": 0.0686, "learning_rate": 3.808995657920975e-05, "epoch": 0.8071119429173003, "percentage": 40.36, "elapsed_time": "9:02:32", "remaining_time": "13:21:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1730, "total_steps": 4274, "loss": 0.0643, "learning_rate": 3.800119321806809e-05, "epoch": 0.809451397824307, "percentage": 40.48, "elapsed_time": "9:04:04", "remaining_time": "13:20:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1735, "total_steps": 4274, "loss": 0.0846, "learning_rate": 3.791220463091387e-05, "epoch": 0.8117908527313136, "percentage": 40.59, "elapsed_time": "9:05:37", "remaining_time": "13:18:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1740, "total_steps": 4274, "loss": 0.0697, "learning_rate": 3.7822992359339794e-05, "epoch": 0.8141303076383203, "percentage": 40.71, "elapsed_time": "9:07:09", "remaining_time": "13:16:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1745, "total_steps": 4274, "loss": 0.0679, "learning_rate": 3.7733557948813575e-05, "epoch": 0.816469762545327, "percentage": 40.83, "elapsed_time": "9:08:42", "remaining_time": "13:15:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1750, "total_steps": 4274, "loss": 0.0779, "learning_rate": 3.764390294865112e-05, "epoch": 0.8188092174523336, "percentage": 40.95, "elapsed_time": "9:10:14", "remaining_time": "13:13:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1755, "total_steps": 4274, "loss": 0.141, "learning_rate": 3.755402891198974e-05, "epoch": 0.8211486723593403, "percentage": 41.06, "elapsed_time": "9:11:45", "remaining_time": "13:11:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1760, "total_steps": 4274, "loss": 0.0675, "learning_rate": 3.746393739576121e-05, "epoch": 0.8234881272663469, "percentage": 41.18, "elapsed_time": "9:13:17", "remaining_time": "13:10:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1765, "total_steps": 4274, "loss": 0.0708, "learning_rate": 3.737362996066483e-05, "epoch": 0.8258275821733536, "percentage": 41.3, "elapsed_time": "9:14:49", "remaining_time": "13:08:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1770, "total_steps": 4274, "loss": 0.0675, "learning_rate": 3.728310817114034e-05, "epoch": 0.8281670370803603, "percentage": 41.41, "elapsed_time": "9:16:22", "remaining_time": "13:07:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1775, "total_steps": 4274, "loss": 0.0758, "learning_rate": 3.719237359534087e-05, "epoch": 0.8305064919873669, "percentage": 41.53, "elapsed_time": "9:17:54", "remaining_time": "13:05:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1780, "total_steps": 4274, "loss": 0.0728, "learning_rate": 3.710142780510573e-05, "epoch": 0.8328459468943736, "percentage": 41.65, "elapsed_time": "9:19:27", "remaining_time": "13:03:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1785, "total_steps": 4274, "loss": 0.0651, "learning_rate": 3.7010272375933216e-05, "epoch": 0.8351854018013802, "percentage": 41.76, "elapsed_time": "9:20:59", "remaining_time": "13:02:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1790, "total_steps": 4274, "loss": 0.0659, "learning_rate": 3.691890888695329e-05, "epoch": 0.8375248567083869, "percentage": 41.88, "elapsed_time": "9:22:32", "remaining_time": "13:00:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1795, "total_steps": 4274, "loss": 0.065, "learning_rate": 3.6827338920900254e-05, "epoch": 0.8398643116153937, "percentage": 42.0, "elapsed_time": "9:24:04", "remaining_time": "12:59:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1800, "total_steps": 4274, "loss": 0.0748, "learning_rate": 3.6735564064085296e-05, "epoch": 0.8422037665224003, "percentage": 42.12, "elapsed_time": "9:25:34", "remaining_time": "12:57:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1805, "total_steps": 4274, "loss": 0.0857, "learning_rate": 3.664358590636903e-05, "epoch": 0.844543221429407, "percentage": 42.23, "elapsed_time": "9:27:05", "remaining_time": "12:55:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1810, "total_steps": 4274, "loss": 0.0721, "learning_rate": 3.655140604113395e-05, "epoch": 0.8468826763364136, "percentage": 42.35, "elapsed_time": "9:28:37", "remaining_time": "12:54:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1815, "total_steps": 4274, "loss": 0.0653, "learning_rate": 3.645902606525683e-05, "epoch": 0.8492221312434203, "percentage": 42.47, "elapsed_time": "9:30:10", "remaining_time": "12:52:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1820, "total_steps": 4274, "loss": 0.0752, "learning_rate": 3.636644757908107e-05, "epoch": 0.851561586150427, "percentage": 42.58, "elapsed_time": "9:31:42", "remaining_time": "12:50:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1825, "total_steps": 4274, "loss": 0.0692, "learning_rate": 3.627367218638893e-05, "epoch": 0.8539010410574336, "percentage": 42.7, "elapsed_time": "9:33:15", "remaining_time": "12:49:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1830, "total_steps": 4274, "loss": 0.0706, "learning_rate": 3.618070149437381e-05, "epoch": 0.8562404959644403, "percentage": 42.82, "elapsed_time": "9:34:47", "remaining_time": "12:47:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1835, "total_steps": 4274, "loss": 0.0612, "learning_rate": 3.6087537113612364e-05, "epoch": 0.858579950871447, "percentage": 42.93, "elapsed_time": "9:36:20", "remaining_time": "12:46:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1840, "total_steps": 4274, "loss": 0.0724, "learning_rate": 3.59941806580366e-05, "epoch": 0.8609194057784536, "percentage": 43.05, "elapsed_time": "9:37:52", "remaining_time": "12:44:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1845, "total_steps": 4274, "loss": 0.0851, "learning_rate": 3.590063374490595e-05, "epoch": 0.8632588606854603, "percentage": 43.17, "elapsed_time": "9:39:24", "remaining_time": "12:42:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1850, "total_steps": 4274, "loss": 0.0669, "learning_rate": 3.580689799477921e-05, "epoch": 0.8655983155924669, "percentage": 43.28, "elapsed_time": "9:40:53", "remaining_time": "12:41:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1855, "total_steps": 4274, "loss": 0.0716, "learning_rate": 3.5712975031486525e-05, "epoch": 0.8679377704994736, "percentage": 43.4, "elapsed_time": "9:42:25", "remaining_time": "12:39:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1860, "total_steps": 4274, "loss": 0.0748, "learning_rate": 3.56188664821012e-05, "epoch": 0.8702772254064803, "percentage": 43.52, "elapsed_time": "9:43:56", "remaining_time": "12:37:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1865, "total_steps": 4274, "loss": 0.079, "learning_rate": 3.5524573976911546e-05, "epoch": 0.8726166803134869, "percentage": 43.64, "elapsed_time": "9:45:29", "remaining_time": "12:36:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1870, "total_steps": 4274, "loss": 0.0747, "learning_rate": 3.543009914939261e-05, "epoch": 0.8749561352204936, "percentage": 43.75, "elapsed_time": "9:47:01", "remaining_time": "12:34:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1875, "total_steps": 4274, "loss": 0.0657, "learning_rate": 3.533544363617796e-05, "epoch": 0.8772955901275002, "percentage": 43.87, "elapsed_time": "9:48:33", "remaining_time": "12:33:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1880, "total_steps": 4274, "loss": 0.0642, "learning_rate": 3.5240609077031196e-05, "epoch": 0.879635045034507, "percentage": 43.99, "elapsed_time": "9:50:05", "remaining_time": "12:31:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1885, "total_steps": 4274, "loss": 0.0608, "learning_rate": 3.514559711481765e-05, "epoch": 0.8819744999415137, "percentage": 44.1, "elapsed_time": "9:51:37", "remaining_time": "12:29:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1890, "total_steps": 4274, "loss": 0.077, "learning_rate": 3.505040939547591e-05, "epoch": 0.8843139548485203, "percentage": 44.22, "elapsed_time": "9:53:09", "remaining_time": "12:28:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1895, "total_steps": 4274, "loss": 0.0788, "learning_rate": 3.495504756798926e-05, "epoch": 0.886653409755527, "percentage": 44.34, "elapsed_time": "9:54:41", "remaining_time": "12:26:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1900, "total_steps": 4274, "loss": 0.0764, "learning_rate": 3.485951328435716e-05, "epoch": 0.8889928646625336, "percentage": 44.45, "elapsed_time": "9:56:12", "remaining_time": "12:24:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1905, "total_steps": 4274, "loss": 0.0722, "learning_rate": 3.476380819956662e-05, "epoch": 0.8913323195695403, "percentage": 44.57, "elapsed_time": "9:57:42", "remaining_time": "12:23:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1910, "total_steps": 4274, "loss": 0.0694, "learning_rate": 3.466793397156351e-05, "epoch": 0.893671774476547, "percentage": 44.69, "elapsed_time": "9:59:12", "remaining_time": "12:21:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1915, "total_steps": 4274, "loss": 0.0652, "learning_rate": 3.457189226122384e-05, "epoch": 0.8960112293835536, "percentage": 44.81, "elapsed_time": "10:00:44", "remaining_time": "12:20:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1920, "total_steps": 4274, "loss": 0.0763, "learning_rate": 3.4475684732325024e-05, "epoch": 0.8983506842905603, "percentage": 44.92, "elapsed_time": "10:02:16", "remaining_time": "12:18:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1925, "total_steps": 4274, "loss": 0.0739, "learning_rate": 3.437931305151701e-05, "epoch": 0.900690139197567, "percentage": 45.04, "elapsed_time": "10:03:48", "remaining_time": "12:16:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1930, "total_steps": 4274, "loss": 0.0839, "learning_rate": 3.428277888829344e-05, "epoch": 0.9030295941045736, "percentage": 45.16, "elapsed_time": "10:05:20", "remaining_time": "12:15:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1935, "total_steps": 4274, "loss": 0.0695, "learning_rate": 3.418608391496273e-05, "epoch": 0.9053690490115803, "percentage": 45.27, "elapsed_time": "10:06:52", "remaining_time": "12:13:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1940, "total_steps": 4274, "loss": 0.0681, "learning_rate": 3.408922980661905e-05, "epoch": 0.9077085039185869, "percentage": 45.39, "elapsed_time": "10:08:23", "remaining_time": "12:11:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1945, "total_steps": 4274, "loss": 0.0694, "learning_rate": 3.39922182411134e-05, "epoch": 0.9100479588255936, "percentage": 45.51, "elapsed_time": "10:09:55", "remaining_time": "12:10:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1950, "total_steps": 4274, "loss": 0.1077, "learning_rate": 3.3895050899024416e-05, "epoch": 0.9123874137326003, "percentage": 45.62, "elapsed_time": "10:11:27", "remaining_time": "12:08:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1955, "total_steps": 4274, "loss": 0.0604, "learning_rate": 3.379772946362939e-05, "epoch": 0.914726868639607, "percentage": 45.74, "elapsed_time": "10:12:59", "remaining_time": "12:07:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1960, "total_steps": 4274, "loss": 0.061, "learning_rate": 3.3700255620875025e-05, "epoch": 0.9170663235466137, "percentage": 45.86, "elapsed_time": "10:14:31", "remaining_time": "12:05:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1965, "total_steps": 4274, "loss": 0.0714, "learning_rate": 3.3602631059348245e-05, "epoch": 0.9194057784536203, "percentage": 45.98, "elapsed_time": "10:16:04", "remaining_time": "12:03:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1970, "total_steps": 4274, "loss": 0.0724, "learning_rate": 3.350485747024695e-05, "epoch": 0.921745233360627, "percentage": 46.09, "elapsed_time": "10:17:36", "remaining_time": "12:02:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1975, "total_steps": 4274, "loss": 0.064, "learning_rate": 3.340693654735073e-05, "epoch": 0.9240846882676337, "percentage": 46.21, "elapsed_time": "10:19:08", "remaining_time": "12:00:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1980, "total_steps": 4274, "loss": 0.0639, "learning_rate": 3.330886998699149e-05, "epoch": 0.9264241431746403, "percentage": 46.33, "elapsed_time": "10:20:41", "remaining_time": "11:59:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1985, "total_steps": 4274, "loss": 0.0667, "learning_rate": 3.321065948802411e-05, "epoch": 0.928763598081647, "percentage": 46.44, "elapsed_time": "10:22:13", "remaining_time": "11:57:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1990, "total_steps": 4274, "loss": 0.0634, "learning_rate": 3.311230675179697e-05, "epoch": 0.9311030529886536, "percentage": 46.56, "elapsed_time": "10:23:45", "remaining_time": "11:55:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1995, "total_steps": 4274, "loss": 0.0632, "learning_rate": 3.301381348212249e-05, "epoch": 0.9334425078956603, "percentage": 46.68, "elapsed_time": "10:25:17", "remaining_time": "11:54:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2000, "total_steps": 4274, "loss": 0.0678, "learning_rate": 3.291518138524764e-05, "epoch": 0.935781962802667, "percentage": 46.79, "elapsed_time": "10:26:49", "remaining_time": "11:52:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2000, "total_steps": 4274, "eval_loss": 0.08131425082683563, "epoch": 0.935781962802667, "percentage": 46.79, "elapsed_time": "10:30:41", "remaining_time": "11:57:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2005, "total_steps": 4274, "loss": 0.1141, "learning_rate": 3.281641216982435e-05, "epoch": 0.9381214177096736, "percentage": 46.91, "elapsed_time": "10:32:13", "remaining_time": "11:55:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2010, "total_steps": 4274, "loss": 0.0658, "learning_rate": 3.271750754687991e-05, "epoch": 0.9404608726166803, "percentage": 47.03, "elapsed_time": "10:33:43", "remaining_time": "11:53:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2015, "total_steps": 4274, "loss": 0.0727, "learning_rate": 3.261846922978736e-05, "epoch": 0.942800327523687, "percentage": 47.15, "elapsed_time": "10:35:15", "remaining_time": "11:52:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2020, "total_steps": 4274, "loss": 0.0683, "learning_rate": 3.2519298934235796e-05, "epoch": 0.9451397824306936, "percentage": 47.26, "elapsed_time": "10:36:47", "remaining_time": "11:50:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2025, "total_steps": 4274, "loss": 0.0663, "learning_rate": 3.24199983782006e-05, "epoch": 0.9474792373377003, "percentage": 47.38, "elapsed_time": "10:38:16", "remaining_time": "11:48:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2030, "total_steps": 4274, "loss": 0.0743, "learning_rate": 3.232056928191376e-05, "epoch": 0.9498186922447069, "percentage": 47.5, "elapsed_time": "10:39:48", "remaining_time": "11:47:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2035, "total_steps": 4274, "loss": 0.0994, "learning_rate": 3.2221013367834016e-05, "epoch": 0.9521581471517137, "percentage": 47.61, "elapsed_time": "10:41:21", "remaining_time": "11:45:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2040, "total_steps": 4274, "loss": 0.0835, "learning_rate": 3.212133236061704e-05, "epoch": 0.9544976020587204, "percentage": 47.73, "elapsed_time": "10:42:53", "remaining_time": "11:44:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2045, "total_steps": 4274, "loss": 0.0843, "learning_rate": 3.202152798708554e-05, "epoch": 0.956837056965727, "percentage": 47.85, "elapsed_time": "10:44:25", "remaining_time": "11:42:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2050, "total_steps": 4274, "loss": 0.0744, "learning_rate": 3.1921601976199365e-05, "epoch": 0.9591765118727337, "percentage": 47.96, "elapsed_time": "10:45:57", "remaining_time": "11:40:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2055, "total_steps": 4274, "loss": 0.0654, "learning_rate": 3.182155605902556e-05, "epoch": 0.9615159667797403, "percentage": 48.08, "elapsed_time": "10:47:30", "remaining_time": "11:39:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2060, "total_steps": 4274, "loss": 0.0665, "learning_rate": 3.1721391968708335e-05, "epoch": 0.963855421686747, "percentage": 48.2, "elapsed_time": "10:49:02", "remaining_time": "11:37:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2065, "total_steps": 4274, "loss": 0.072, "learning_rate": 3.162111144043911e-05, "epoch": 0.9661948765937537, "percentage": 48.32, "elapsed_time": "10:50:34", "remaining_time": "11:35:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2070, "total_steps": 4274, "loss": 0.0754, "learning_rate": 3.1520716211426385e-05, "epoch": 0.9685343315007603, "percentage": 48.43, "elapsed_time": "10:52:06", "remaining_time": "11:34:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2075, "total_steps": 4274, "loss": 0.0739, "learning_rate": 3.142020802086569e-05, "epoch": 0.970873786407767, "percentage": 48.55, "elapsed_time": "10:53:37", "remaining_time": "11:32:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2080, "total_steps": 4274, "loss": 0.0617, "learning_rate": 3.131958860990945e-05, "epoch": 0.9732132413147736, "percentage": 48.67, "elapsed_time": "10:55:08", "remaining_time": "11:31:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2085, "total_steps": 4274, "loss": 0.0785, "learning_rate": 3.121885972163681e-05, "epoch": 0.9755526962217803, "percentage": 48.78, "elapsed_time": "10:56:40", "remaining_time": "11:29:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2090, "total_steps": 4274, "loss": 0.0675, "learning_rate": 3.111802310102341e-05, "epoch": 0.977892151128787, "percentage": 48.9, "elapsed_time": "10:58:13", "remaining_time": "11:27:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2095, "total_steps": 4274, "loss": 0.0694, "learning_rate": 3.101708049491125e-05, "epoch": 0.9802316060357936, "percentage": 49.02, "elapsed_time": "10:59:45", "remaining_time": "11:26:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2100, "total_steps": 4274, "loss": 0.0743, "learning_rate": 3.091603365197833e-05, "epoch": 0.9825710609428003, "percentage": 49.13, "elapsed_time": "11:01:18", "remaining_time": "11:24:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2105, "total_steps": 4274, "loss": 0.0664, "learning_rate": 3.081488432270841e-05, "epoch": 0.984910515849807, "percentage": 49.25, "elapsed_time": "11:02:50", "remaining_time": "11:23:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2110, "total_steps": 4274, "loss": 0.0762, "learning_rate": 3.071363425936066e-05, "epoch": 0.9872499707568136, "percentage": 49.37, "elapsed_time": "11:04:23", "remaining_time": "11:21:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2115, "total_steps": 4274, "loss": 0.0636, "learning_rate": 3.061228521593931e-05, "epoch": 0.9895894256638204, "percentage": 49.49, "elapsed_time": "11:05:55", "remaining_time": "11:19:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2120, "total_steps": 4274, "loss": 0.0836, "learning_rate": 3.0510838948163308e-05, "epoch": 0.991928880570827, "percentage": 49.6, "elapsed_time": "11:07:28", "remaining_time": "11:18:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2125, "total_steps": 4274, "loss": 0.0723, "learning_rate": 3.0409297213435818e-05, "epoch": 0.9942683354778337, "percentage": 49.72, "elapsed_time": "11:09:00", "remaining_time": "11:16:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2130, "total_steps": 4274, "loss": 0.0766, "learning_rate": 3.030766177081385e-05, "epoch": 0.9966077903848404, "percentage": 49.84, "elapsed_time": "11:10:32", "remaining_time": "11:14:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2135, "total_steps": 4274, "loss": 0.0619, "learning_rate": 3.020593438097776e-05, "epoch": 0.998947245291847, "percentage": 49.95, "elapsed_time": "11:12:04", "remaining_time": "11:13:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2140, "total_steps": 4274, "loss": 0.0641, "learning_rate": 3.010411680620075e-05, "epoch": 1.0012867001988537, "percentage": 50.07, "elapsed_time": "11:13:36", "remaining_time": "11:11:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2145, "total_steps": 4274, "loss": 0.0735, "learning_rate": 3.0002210810318338e-05, "epoch": 1.0036261551058603, "percentage": 50.19, "elapsed_time": "11:15:08", "remaining_time": "11:10:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2150, "total_steps": 4274, "loss": 0.0576, "learning_rate": 2.9900218158697797e-05, "epoch": 1.005965610012867, "percentage": 50.3, "elapsed_time": "11:16:40", "remaining_time": "11:08:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2155, "total_steps": 4274, "loss": 0.0721, "learning_rate": 2.9798140618207598e-05, "epoch": 1.0083050649198737, "percentage": 50.42, "elapsed_time": "11:18:12", "remaining_time": "11:06:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2160, "total_steps": 4274, "loss": 0.0642, "learning_rate": 2.9695979957186764e-05, "epoch": 1.0106445198268803, "percentage": 50.54, "elapsed_time": "11:19:44", "remaining_time": "11:05:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2165, "total_steps": 4274, "loss": 0.0751, "learning_rate": 2.9593737945414264e-05, "epoch": 1.012983974733887, "percentage": 50.66, "elapsed_time": "11:21:17", "remaining_time": "11:03:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2170, "total_steps": 4274, "loss": 0.0619, "learning_rate": 2.9491416354078343e-05, "epoch": 1.0153234296408937, "percentage": 50.77, "elapsed_time": "11:22:48", "remaining_time": "11:02:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2175, "total_steps": 4274, "loss": 0.0719, "learning_rate": 2.938901695574585e-05, "epoch": 1.0176628845479003, "percentage": 50.89, "elapsed_time": "11:24:20", "remaining_time": "11:00:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2180, "total_steps": 4274, "loss": 0.0597, "learning_rate": 2.9286541524331525e-05, "epoch": 1.020002339454907, "percentage": 51.01, "elapsed_time": "11:25:52", "remaining_time": "10:58:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2185, "total_steps": 4274, "loss": 0.0585, "learning_rate": 2.9183991835067237e-05, "epoch": 1.0223417943619137, "percentage": 51.12, "elapsed_time": "11:27:24", "remaining_time": "10:57:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2190, "total_steps": 4274, "loss": 0.0712, "learning_rate": 2.9081369664471293e-05, "epoch": 1.0246812492689203, "percentage": 51.24, "elapsed_time": "11:28:57", "remaining_time": "10:55:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2195, "total_steps": 4274, "loss": 0.0631, "learning_rate": 2.8978676790317605e-05, "epoch": 1.027020704175927, "percentage": 51.36, "elapsed_time": "11:30:29", "remaining_time": "10:54:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2200, "total_steps": 4274, "loss": 0.0611, "learning_rate": 2.8875914991604948e-05, "epoch": 1.0293601590829338, "percentage": 51.47, "elapsed_time": "11:32:01", "remaining_time": "10:52:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2205, "total_steps": 4274, "loss": 0.1481, "learning_rate": 2.8773086048526076e-05, "epoch": 1.0316996139899404, "percentage": 51.59, "elapsed_time": "11:33:33", "remaining_time": "10:50:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2210, "total_steps": 4274, "loss": 0.0623, "learning_rate": 2.8670191742436946e-05, "epoch": 1.034039068896947, "percentage": 51.71, "elapsed_time": "11:35:05", "remaining_time": "10:49:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2215, "total_steps": 4274, "loss": 0.0599, "learning_rate": 2.8567233855825813e-05, "epoch": 1.0363785238039536, "percentage": 51.82, "elapsed_time": "11:36:38", "remaining_time": "10:47:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2220, "total_steps": 4274, "loss": 0.0593, "learning_rate": 2.8464214172282367e-05, "epoch": 1.0387179787109604, "percentage": 51.94, "elapsed_time": "11:38:10", "remaining_time": "10:45:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2225, "total_steps": 4274, "loss": 0.0643, "learning_rate": 2.8361134476466843e-05, "epoch": 1.041057433617967, "percentage": 52.06, "elapsed_time": "11:39:42", "remaining_time": "10:44:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2230, "total_steps": 4274, "loss": 0.0582, "learning_rate": 2.8257996554079113e-05, "epoch": 1.0433968885249736, "percentage": 52.18, "elapsed_time": "11:41:14", "remaining_time": "10:42:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2235, "total_steps": 4274, "loss": 0.0577, "learning_rate": 2.8154802191827706e-05, "epoch": 1.0457363434319804, "percentage": 52.29, "elapsed_time": "11:42:46", "remaining_time": "10:41:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2240, "total_steps": 4274, "loss": 0.0618, "learning_rate": 2.805155317739891e-05, "epoch": 1.048075798338987, "percentage": 52.41, "elapsed_time": "11:44:16", "remaining_time": "10:39:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2245, "total_steps": 4274, "loss": 0.067, "learning_rate": 2.7948251299425758e-05, "epoch": 1.0504152532459936, "percentage": 52.53, "elapsed_time": "11:45:48", "remaining_time": "10:37:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2250, "total_steps": 4274, "loss": 0.0626, "learning_rate": 2.784489834745709e-05, "epoch": 1.0527547081530004, "percentage": 52.64, "elapsed_time": "11:47:20", "remaining_time": "10:36:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2255, "total_steps": 4274, "loss": 0.0627, "learning_rate": 2.77414961119265e-05, "epoch": 1.055094163060007, "percentage": 52.76, "elapsed_time": "11:48:53", "remaining_time": "10:34:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2260, "total_steps": 4274, "loss": 0.0653, "learning_rate": 2.763804638412137e-05, "epoch": 1.0574336179670136, "percentage": 52.88, "elapsed_time": "11:50:25", "remaining_time": "10:33:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2265, "total_steps": 4274, "loss": 0.0667, "learning_rate": 2.753455095615179e-05, "epoch": 1.0597730728740204, "percentage": 52.99, "elapsed_time": "11:51:58", "remaining_time": "10:31:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2270, "total_steps": 4274, "loss": 0.0653, "learning_rate": 2.7431011620919554e-05, "epoch": 1.062112527781027, "percentage": 53.11, "elapsed_time": "11:53:30", "remaining_time": "10:29:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2275, "total_steps": 4274, "loss": 0.0605, "learning_rate": 2.7327430172087077e-05, "epoch": 1.0644519826880336, "percentage": 53.23, "elapsed_time": "11:55:01", "remaining_time": "10:28:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2280, "total_steps": 4274, "loss": 0.0595, "learning_rate": 2.722380840404632e-05, "epoch": 1.0667914375950405, "percentage": 53.35, "elapsed_time": "11:56:34", "remaining_time": "10:26:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2285, "total_steps": 4274, "loss": 0.0706, "learning_rate": 2.7120148111887732e-05, "epoch": 1.069130892502047, "percentage": 53.46, "elapsed_time": "11:58:06", "remaining_time": "10:25:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2290, "total_steps": 4274, "loss": 0.0646, "learning_rate": 2.7016451091369116e-05, "epoch": 1.0714703474090537, "percentage": 53.58, "elapsed_time": "11:59:39", "remaining_time": "10:23:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2295, "total_steps": 4274, "loss": 0.0612, "learning_rate": 2.6912719138884553e-05, "epoch": 1.0738098023160603, "percentage": 53.7, "elapsed_time": "12:01:11", "remaining_time": "10:21:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2300, "total_steps": 4274, "loss": 0.0657, "learning_rate": 2.6808954051433248e-05, "epoch": 1.076149257223067, "percentage": 53.81, "elapsed_time": "12:02:44", "remaining_time": "10:20:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2305, "total_steps": 4274, "loss": 0.0649, "learning_rate": 2.670515762658844e-05, "epoch": 1.0784887121300737, "percentage": 53.93, "elapsed_time": "12:04:16", "remaining_time": "10:18:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2310, "total_steps": 4274, "loss": 0.0523, "learning_rate": 2.6601331662466227e-05, "epoch": 1.0808281670370803, "percentage": 54.05, "elapsed_time": "12:05:48", "remaining_time": "10:17:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2315, "total_steps": 4274, "loss": 0.0565, "learning_rate": 2.6497477957694443e-05, "epoch": 1.0831676219440871, "percentage": 54.16, "elapsed_time": "12:07:21", "remaining_time": "10:15:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2320, "total_steps": 4274, "loss": 0.0629, "learning_rate": 2.6393598311381473e-05, "epoch": 1.0855070768510937, "percentage": 54.28, "elapsed_time": "12:08:53", "remaining_time": "10:13:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2325, "total_steps": 4274, "loss": 0.0685, "learning_rate": 2.6289694523085112e-05, "epoch": 1.0878465317581003, "percentage": 54.4, "elapsed_time": "12:10:25", "remaining_time": "10:12:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2330, "total_steps": 4274, "loss": 0.056, "learning_rate": 2.6185768392781366e-05, "epoch": 1.0901859866651071, "percentage": 54.52, "elapsed_time": "12:11:57", "remaining_time": "10:10:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2335, "total_steps": 4274, "loss": 0.0569, "learning_rate": 2.6081821720833287e-05, "epoch": 1.0925254415721137, "percentage": 54.63, "elapsed_time": "12:13:29", "remaining_time": "10:09:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2340, "total_steps": 4274, "loss": 0.0622, "learning_rate": 2.5977856307959798e-05, "epoch": 1.0948648964791203, "percentage": 54.75, "elapsed_time": "12:15:01", "remaining_time": "10:07:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2345, "total_steps": 4274, "loss": 0.0574, "learning_rate": 2.5873873955204448e-05, "epoch": 1.097204351386127, "percentage": 54.87, "elapsed_time": "12:16:33", "remaining_time": "10:05:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2350, "total_steps": 4274, "loss": 0.0575, "learning_rate": 2.5769876463904265e-05, "epoch": 1.0995438062931338, "percentage": 54.98, "elapsed_time": "12:18:04", "remaining_time": "10:04:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2355, "total_steps": 4274, "loss": 0.063, "learning_rate": 2.5665865635658527e-05, "epoch": 1.1018832612001404, "percentage": 55.1, "elapsed_time": "12:19:36", "remaining_time": "10:02:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2360, "total_steps": 4274, "loss": 0.0625, "learning_rate": 2.5561843272297536e-05, "epoch": 1.104222716107147, "percentage": 55.22, "elapsed_time": "12:21:08", "remaining_time": "10:01:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2365, "total_steps": 4274, "loss": 0.0539, "learning_rate": 2.5457811175851465e-05, "epoch": 1.1065621710141538, "percentage": 55.33, "elapsed_time": "12:22:40", "remaining_time": "9:59:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2370, "total_steps": 4274, "loss": 0.0639, "learning_rate": 2.5353771148519057e-05, "epoch": 1.1089016259211604, "percentage": 55.45, "elapsed_time": "12:24:13", "remaining_time": "9:57:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2375, "total_steps": 4274, "loss": 0.0749, "learning_rate": 2.524972499263646e-05, "epoch": 1.111241080828167, "percentage": 55.57, "elapsed_time": "12:25:45", "remaining_time": "9:56:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2380, "total_steps": 4274, "loss": 0.0632, "learning_rate": 2.514567451064599e-05, "epoch": 1.1135805357351738, "percentage": 55.69, "elapsed_time": "12:27:18", "remaining_time": "9:54:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2385, "total_steps": 4274, "loss": 0.0578, "learning_rate": 2.50416215050649e-05, "epoch": 1.1159199906421804, "percentage": 55.8, "elapsed_time": "12:28:50", "remaining_time": "9:53:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2390, "total_steps": 4274, "loss": 0.0603, "learning_rate": 2.4937567778454188e-05, "epoch": 1.118259445549187, "percentage": 55.92, "elapsed_time": "12:30:23", "remaining_time": "9:51:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2395, "total_steps": 4274, "loss": 0.0577, "learning_rate": 2.4833515133387296e-05, "epoch": 1.1205989004561938, "percentage": 56.04, "elapsed_time": "12:31:55", "remaining_time": "9:49:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2400, "total_steps": 4274, "loss": 0.0715, "learning_rate": 2.4729465372418972e-05, "epoch": 1.1229383553632004, "percentage": 56.15, "elapsed_time": "12:33:28", "remaining_time": "9:48:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2405, "total_steps": 4274, "loss": 0.0696, "learning_rate": 2.4625420298053968e-05, "epoch": 1.125277810270207, "percentage": 56.27, "elapsed_time": "12:35:00", "remaining_time": "9:46:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2410, "total_steps": 4274, "loss": 0.0631, "learning_rate": 2.4521381712715884e-05, "epoch": 1.1276172651772136, "percentage": 56.39, "elapsed_time": "12:36:32", "remaining_time": "9:45:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2415, "total_steps": 4274, "loss": 0.0584, "learning_rate": 2.441735141871587e-05, "epoch": 1.1299567200842204, "percentage": 56.5, "elapsed_time": "12:38:05", "remaining_time": "9:43:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2420, "total_steps": 4274, "loss": 0.0658, "learning_rate": 2.4313331218221476e-05, "epoch": 1.132296174991227, "percentage": 56.62, "elapsed_time": "12:39:37", "remaining_time": "9:41:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2425, "total_steps": 4274, "loss": 0.0636, "learning_rate": 2.420932291322536e-05, "epoch": 1.1346356298982336, "percentage": 56.74, "elapsed_time": "12:41:09", "remaining_time": "9:40:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2430, "total_steps": 4274, "loss": 0.0687, "learning_rate": 2.410532830551415e-05, "epoch": 1.1369750848052405, "percentage": 56.86, "elapsed_time": "12:42:41", "remaining_time": "9:38:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2435, "total_steps": 4274, "loss": 0.065, "learning_rate": 2.4001349196637144e-05, "epoch": 1.139314539712247, "percentage": 56.97, "elapsed_time": "12:44:14", "remaining_time": "9:37:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2440, "total_steps": 4274, "loss": 0.0597, "learning_rate": 2.3897387387875188e-05, "epoch": 1.1416539946192537, "percentage": 57.09, "elapsed_time": "12:45:47", "remaining_time": "9:35:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2445, "total_steps": 4274, "loss": 0.0617, "learning_rate": 2.3793444680209397e-05, "epoch": 1.1439934495262605, "percentage": 57.21, "elapsed_time": "12:47:19", "remaining_time": "9:34:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2450, "total_steps": 4274, "loss": 0.0646, "learning_rate": 2.3689522874290028e-05, "epoch": 1.146332904433267, "percentage": 57.32, "elapsed_time": "12:48:51", "remaining_time": "9:32:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2455, "total_steps": 4274, "loss": 0.0734, "learning_rate": 2.358562377040519e-05, "epoch": 1.1486723593402737, "percentage": 57.44, "elapsed_time": "12:50:24", "remaining_time": "9:30:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2460, "total_steps": 4274, "loss": 0.0599, "learning_rate": 2.3481749168449774e-05, "epoch": 1.1510118142472803, "percentage": 57.56, "elapsed_time": "12:51:56", "remaining_time": "9:29:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2465, "total_steps": 4274, "loss": 0.0587, "learning_rate": 2.3377900867894158e-05, "epoch": 1.153351269154287, "percentage": 57.67, "elapsed_time": "12:53:29", "remaining_time": "9:27:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2470, "total_steps": 4274, "loss": 0.0713, "learning_rate": 2.3274080667753128e-05, "epoch": 1.1556907240612937, "percentage": 57.79, "elapsed_time": "12:55:01", "remaining_time": "9:26:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2475, "total_steps": 4274, "loss": 0.0681, "learning_rate": 2.3170290366554653e-05, "epoch": 1.1580301789683003, "percentage": 57.91, "elapsed_time": "12:56:33", "remaining_time": "9:24:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2480, "total_steps": 4274, "loss": 0.0598, "learning_rate": 2.3066531762308766e-05, "epoch": 1.1603696338753071, "percentage": 58.03, "elapsed_time": "12:58:06", "remaining_time": "9:22:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2485, "total_steps": 4274, "loss": 0.0553, "learning_rate": 2.2962806652476363e-05, "epoch": 1.1627090887823137, "percentage": 58.14, "elapsed_time": "12:59:38", "remaining_time": "9:21:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2490, "total_steps": 4274, "loss": 0.0694, "learning_rate": 2.2859116833938146e-05, "epoch": 1.1650485436893203, "percentage": 58.26, "elapsed_time": "13:01:10", "remaining_time": "9:19:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2495, "total_steps": 4274, "loss": 0.0576, "learning_rate": 2.2755464102963408e-05, "epoch": 1.1673879985963271, "percentage": 58.38, "elapsed_time": "13:02:43", "remaining_time": "9:18:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2500, "total_steps": 4274, "loss": 0.0697, "learning_rate": 2.2651850255178974e-05, "epoch": 1.1697274535033337, "percentage": 58.49, "elapsed_time": "13:04:15", "remaining_time": "9:16:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2500, "total_steps": 4274, "eval_loss": 0.07991591095924377, "epoch": 1.1697274535033337, "percentage": 58.49, "elapsed_time": "13:08:12", "remaining_time": "9:19:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2505, "total_steps": 4274, "loss": 0.0669, "learning_rate": 2.2548277085538092e-05, "epoch": 1.1720669084103403, "percentage": 58.61, "elapsed_time": "13:09:45", "remaining_time": "9:17:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2510, "total_steps": 4274, "loss": 0.0853, "learning_rate": 2.2444746388289283e-05, "epoch": 1.1744063633173472, "percentage": 58.73, "elapsed_time": "13:11:18", "remaining_time": "9:16:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2515, "total_steps": 4274, "loss": 0.0595, "learning_rate": 2.2341259956945342e-05, "epoch": 1.1767458182243538, "percentage": 58.84, "elapsed_time": "13:12:50", "remaining_time": "9:14:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2520, "total_steps": 4274, "loss": 0.0544, "learning_rate": 2.2237819584252188e-05, "epoch": 1.1790852731313604, "percentage": 58.96, "elapsed_time": "13:14:22", "remaining_time": "9:12:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2525, "total_steps": 4274, "loss": 0.0652, "learning_rate": 2.2134427062157866e-05, "epoch": 1.1814247280383672, "percentage": 59.08, "elapsed_time": "13:15:54", "remaining_time": "9:11:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2530, "total_steps": 4274, "loss": 0.0637, "learning_rate": 2.2031084181781476e-05, "epoch": 1.1837641829453738, "percentage": 59.2, "elapsed_time": "13:17:26", "remaining_time": "9:09:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2535, "total_steps": 4274, "loss": 0.079, "learning_rate": 2.192779273338215e-05, "epoch": 1.1861036378523804, "percentage": 59.31, "elapsed_time": "13:18:58", "remaining_time": "9:08:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2540, "total_steps": 4274, "loss": 0.0569, "learning_rate": 2.182455450632803e-05, "epoch": 1.188443092759387, "percentage": 59.43, "elapsed_time": "13:20:30", "remaining_time": "9:06:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2545, "total_steps": 4274, "loss": 0.0569, "learning_rate": 2.17213712890653e-05, "epoch": 1.1907825476663938, "percentage": 59.55, "elapsed_time": "13:22:03", "remaining_time": "9:04:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2550, "total_steps": 4274, "loss": 0.0651, "learning_rate": 2.1618244869087157e-05, "epoch": 1.1931220025734004, "percentage": 59.66, "elapsed_time": "13:23:35", "remaining_time": "9:03:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2555, "total_steps": 4274, "loss": 0.0693, "learning_rate": 2.151517703290289e-05, "epoch": 1.195461457480407, "percentage": 59.78, "elapsed_time": "13:25:07", "remaining_time": "9:01:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2560, "total_steps": 4274, "loss": 0.0598, "learning_rate": 2.1412169566006897e-05, "epoch": 1.1978009123874138, "percentage": 59.9, "elapsed_time": "13:26:39", "remaining_time": "9:00:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2565, "total_steps": 4274, "loss": 0.0593, "learning_rate": 2.1309224252847793e-05, "epoch": 1.2001403672944204, "percentage": 60.01, "elapsed_time": "13:28:12", "remaining_time": "8:58:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2570, "total_steps": 4274, "loss": 0.0604, "learning_rate": 2.1206342876797436e-05, "epoch": 1.202479822201427, "percentage": 60.13, "elapsed_time": "13:29:44", "remaining_time": "8:56:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2575, "total_steps": 4274, "loss": 0.063, "learning_rate": 2.110352722012011e-05, "epoch": 1.2048192771084336, "percentage": 60.25, "elapsed_time": "13:31:16", "remaining_time": "8:55:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2580, "total_steps": 4274, "loss": 0.0568, "learning_rate": 2.1000779063941585e-05, "epoch": 1.2071587320154404, "percentage": 60.36, "elapsed_time": "13:32:48", "remaining_time": "8:53:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2585, "total_steps": 4274, "loss": 0.0634, "learning_rate": 2.0898100188218313e-05, "epoch": 1.209498186922447, "percentage": 60.48, "elapsed_time": "13:34:20", "remaining_time": "8:52:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2590, "total_steps": 4274, "loss": 0.0597, "learning_rate": 2.0795492371706527e-05, "epoch": 1.2118376418294536, "percentage": 60.6, "elapsed_time": "13:35:52", "remaining_time": "8:50:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2595, "total_steps": 4274, "loss": 0.0654, "learning_rate": 2.0692957391931485e-05, "epoch": 1.2141770967364605, "percentage": 60.72, "elapsed_time": "13:37:24", "remaining_time": "8:48:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2600, "total_steps": 4274, "loss": 0.0591, "learning_rate": 2.0590497025156676e-05, "epoch": 1.216516551643467, "percentage": 60.83, "elapsed_time": "13:38:56", "remaining_time": "8:47:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2605, "total_steps": 4274, "loss": 0.0619, "learning_rate": 2.0488113046353004e-05, "epoch": 1.2188560065504737, "percentage": 60.95, "elapsed_time": "13:40:27", "remaining_time": "8:45:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2610, "total_steps": 4274, "loss": 0.0653, "learning_rate": 2.0385807229168104e-05, "epoch": 1.2211954614574805, "percentage": 61.07, "elapsed_time": "13:41:59", "remaining_time": "8:44:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2615, "total_steps": 4274, "loss": 0.0647, "learning_rate": 2.028358134589553e-05, "epoch": 1.223534916364487, "percentage": 61.18, "elapsed_time": "13:43:32", "remaining_time": "8:42:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2620, "total_steps": 4274, "loss": 0.0632, "learning_rate": 2.0181437167444144e-05, "epoch": 1.2258743712714937, "percentage": 61.3, "elapsed_time": "13:45:04", "remaining_time": "8:40:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2625, "total_steps": 4274, "loss": 0.0619, "learning_rate": 2.0079376463307368e-05, "epoch": 1.2282138261785005, "percentage": 61.42, "elapsed_time": "13:46:37", "remaining_time": "8:39:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2630, "total_steps": 4274, "loss": 0.0588, "learning_rate": 1.9977401001532576e-05, "epoch": 1.230553281085507, "percentage": 61.53, "elapsed_time": "13:48:09", "remaining_time": "8:37:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2635, "total_steps": 4274, "loss": 0.0591, "learning_rate": 1.9875512548690433e-05, "epoch": 1.2328927359925137, "percentage": 61.65, "elapsed_time": "13:49:40", "remaining_time": "8:36:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2640, "total_steps": 4274, "loss": 0.0708, "learning_rate": 1.9773712869844323e-05, "epoch": 1.2352321908995205, "percentage": 61.77, "elapsed_time": "13:51:10", "remaining_time": "8:34:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2645, "total_steps": 4274, "loss": 0.0609, "learning_rate": 1.9672003728519728e-05, "epoch": 1.2375716458065271, "percentage": 61.89, "elapsed_time": "13:52:42", "remaining_time": "8:32:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2650, "total_steps": 4274, "loss": 0.0642, "learning_rate": 1.957038688667374e-05, "epoch": 1.2399111007135337, "percentage": 62.0, "elapsed_time": "13:54:15", "remaining_time": "8:31:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2655, "total_steps": 4274, "loss": 0.0638, "learning_rate": 1.9468864104664464e-05, "epoch": 1.2422505556205403, "percentage": 62.12, "elapsed_time": "13:55:48", "remaining_time": "8:29:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2660, "total_steps": 4274, "loss": 0.0643, "learning_rate": 1.9367437141220594e-05, "epoch": 1.2445900105275471, "percentage": 62.24, "elapsed_time": "13:57:20", "remaining_time": "8:28:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2665, "total_steps": 4274, "loss": 0.0611, "learning_rate": 1.9266107753410884e-05, "epoch": 1.2469294654345537, "percentage": 62.35, "elapsed_time": "13:58:52", "remaining_time": "8:26:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2670, "total_steps": 4274, "loss": 0.055, "learning_rate": 1.916487769661377e-05, "epoch": 1.2492689203415603, "percentage": 62.47, "elapsed_time": "14:00:25", "remaining_time": "8:24:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2675, "total_steps": 4274, "loss": 0.0518, "learning_rate": 1.9063748724486887e-05, "epoch": 1.2516083752485672, "percentage": 62.59, "elapsed_time": "14:01:57", "remaining_time": "8:23:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2680, "total_steps": 4274, "loss": 0.0618, "learning_rate": 1.896272258893677e-05, "epoch": 1.2539478301555738, "percentage": 62.7, "elapsed_time": "14:03:30", "remaining_time": "8:21:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2685, "total_steps": 4274, "loss": 0.0553, "learning_rate": 1.8861801040088445e-05, "epoch": 1.2562872850625804, "percentage": 62.82, "elapsed_time": "14:05:02", "remaining_time": "8:20:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2690, "total_steps": 4274, "loss": 0.063, "learning_rate": 1.8760985826255125e-05, "epoch": 1.258626739969587, "percentage": 62.94, "elapsed_time": "14:06:35", "remaining_time": "8:18:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2695, "total_steps": 4274, "loss": 0.0645, "learning_rate": 1.8660278693907962e-05, "epoch": 1.2609661948765938, "percentage": 63.06, "elapsed_time": "14:08:07", "remaining_time": "8:16:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2700, "total_steps": 4274, "loss": 0.0579, "learning_rate": 1.8559681387645715e-05, "epoch": 1.2633056497836004, "percentage": 63.17, "elapsed_time": "14:09:40", "remaining_time": "8:15:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2705, "total_steps": 4274, "loss": 0.0629, "learning_rate": 1.8459195650164612e-05, "epoch": 1.265645104690607, "percentage": 63.29, "elapsed_time": "14:11:12", "remaining_time": "8:13:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2710, "total_steps": 4274, "loss": 0.0627, "learning_rate": 1.8358823222228097e-05, "epoch": 1.2679845595976138, "percentage": 63.41, "elapsed_time": "14:12:45", "remaining_time": "8:12:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2715, "total_steps": 4274, "loss": 0.0582, "learning_rate": 1.8258565842636703e-05, "epoch": 1.2703240145046204, "percentage": 63.52, "elapsed_time": "14:14:17", "remaining_time": "8:10:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2720, "total_steps": 4274, "loss": 0.0606, "learning_rate": 1.815842524819793e-05, "epoch": 1.272663469411627, "percentage": 63.64, "elapsed_time": "14:15:50", "remaining_time": "8:08:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2725, "total_steps": 4274, "loss": 0.0967, "learning_rate": 1.8058403173696152e-05, "epoch": 1.2750029243186338, "percentage": 63.76, "elapsed_time": "14:17:22", "remaining_time": "8:07:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2730, "total_steps": 4274, "loss": 0.0585, "learning_rate": 1.7958501351862545e-05, "epoch": 1.2773423792256404, "percentage": 63.87, "elapsed_time": "14:18:54", "remaining_time": "8:05:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2735, "total_steps": 4274, "loss": 0.0664, "learning_rate": 1.7858721513345117e-05, "epoch": 1.279681834132647, "percentage": 63.99, "elapsed_time": "14:20:27", "remaining_time": "8:04:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2740, "total_steps": 4274, "loss": 0.0548, "learning_rate": 1.7759065386678674e-05, "epoch": 1.2820212890396538, "percentage": 64.11, "elapsed_time": "14:21:59", "remaining_time": "8:02:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2745, "total_steps": 4274, "loss": 0.0577, "learning_rate": 1.765953469825494e-05, "epoch": 1.2843607439466604, "percentage": 64.23, "elapsed_time": "14:23:32", "remaining_time": "8:01:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2750, "total_steps": 4274, "loss": 0.0632, "learning_rate": 1.7560131172292556e-05, "epoch": 1.286700198853667, "percentage": 64.34, "elapsed_time": "14:25:04", "remaining_time": "7:59:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2755, "total_steps": 4274, "loss": 0.0592, "learning_rate": 1.7460856530807315e-05, "epoch": 1.2890396537606739, "percentage": 64.46, "elapsed_time": "14:26:37", "remaining_time": "7:57:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2760, "total_steps": 4274, "loss": 0.057, "learning_rate": 1.7361712493582242e-05, "epoch": 1.2913791086676805, "percentage": 64.58, "elapsed_time": "14:28:09", "remaining_time": "7:56:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2765, "total_steps": 4274, "loss": 0.0603, "learning_rate": 1.726270077813788e-05, "epoch": 1.293718563574687, "percentage": 64.69, "elapsed_time": "14:29:42", "remaining_time": "7:54:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2770, "total_steps": 4274, "loss": 0.0625, "learning_rate": 1.7163823099702466e-05, "epoch": 1.2960580184816939, "percentage": 64.81, "elapsed_time": "14:31:14", "remaining_time": "7:53:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2775, "total_steps": 4274, "loss": 0.0585, "learning_rate": 1.7065081171182264e-05, "epoch": 1.2983974733887005, "percentage": 64.93, "elapsed_time": "14:32:46", "remaining_time": "7:51:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2780, "total_steps": 4274, "loss": 0.0569, "learning_rate": 1.696647670313186e-05, "epoch": 1.300736928295707, "percentage": 65.04, "elapsed_time": "14:34:19", "remaining_time": "7:49:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2785, "total_steps": 4274, "loss": 0.0616, "learning_rate": 1.6868011403724582e-05, "epoch": 1.303076383202714, "percentage": 65.16, "elapsed_time": "14:35:51", "remaining_time": "7:48:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2790, "total_steps": 4274, "loss": 0.063, "learning_rate": 1.676968697872282e-05, "epoch": 1.3054158381097205, "percentage": 65.28, "elapsed_time": "14:37:24", "remaining_time": "7:46:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2795, "total_steps": 4274, "loss": 0.0517, "learning_rate": 1.667150513144856e-05, "epoch": 1.307755293016727, "percentage": 65.4, "elapsed_time": "14:38:56", "remaining_time": "7:45:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2800, "total_steps": 4274, "loss": 0.0614, "learning_rate": 1.657346756275386e-05, "epoch": 1.3100947479237337, "percentage": 65.51, "elapsed_time": "14:40:28", "remaining_time": "7:43:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2805, "total_steps": 4274, "loss": 0.0593, "learning_rate": 1.6475575970991313e-05, "epoch": 1.3124342028307403, "percentage": 65.63, "elapsed_time": "14:41:59", "remaining_time": "7:41:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2810, "total_steps": 4274, "loss": 0.0558, "learning_rate": 1.6377832051984755e-05, "epoch": 1.3147736577377471, "percentage": 65.75, "elapsed_time": "14:43:32", "remaining_time": "7:40:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2815, "total_steps": 4274, "loss": 0.0544, "learning_rate": 1.628023749899975e-05, "epoch": 1.3171131126447537, "percentage": 65.86, "elapsed_time": "14:45:04", "remaining_time": "7:38:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2820, "total_steps": 4274, "loss": 0.0635, "learning_rate": 1.618279400271436e-05, "epoch": 1.3194525675517603, "percentage": 65.98, "elapsed_time": "14:46:36", "remaining_time": "7:37:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2825, "total_steps": 4274, "loss": 0.0559, "learning_rate": 1.608550325118981e-05, "epoch": 1.3217920224587671, "percentage": 66.1, "elapsed_time": "14:48:09", "remaining_time": "7:35:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2830, "total_steps": 4274, "loss": 0.07, "learning_rate": 1.5988366929841254e-05, "epoch": 1.3241314773657737, "percentage": 66.21, "elapsed_time": "14:49:41", "remaining_time": "7:33:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2835, "total_steps": 4274, "loss": 0.0718, "learning_rate": 1.589138672140856e-05, "epoch": 1.3264709322727803, "percentage": 66.33, "elapsed_time": "14:51:13", "remaining_time": "7:32:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2840, "total_steps": 4274, "loss": 0.0566, "learning_rate": 1.57945643059272e-05, "epoch": 1.3288103871797872, "percentage": 66.45, "elapsed_time": "14:52:45", "remaining_time": "7:30:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2845, "total_steps": 4274, "loss": 0.0628, "learning_rate": 1.56979013606991e-05, "epoch": 1.3311498420867938, "percentage": 66.57, "elapsed_time": "14:54:18", "remaining_time": "7:29:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2850, "total_steps": 4274, "loss": 0.058, "learning_rate": 1.560139956026363e-05, "epoch": 1.3334892969938004, "percentage": 66.68, "elapsed_time": "14:55:50", "remaining_time": "7:27:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2855, "total_steps": 4274, "loss": 0.059, "learning_rate": 1.5505060576368545e-05, "epoch": 1.3358287519008072, "percentage": 66.8, "elapsed_time": "14:57:22", "remaining_time": "7:26:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2860, "total_steps": 4274, "loss": 0.066, "learning_rate": 1.5408886077941074e-05, "epoch": 1.3381682068078138, "percentage": 66.92, "elapsed_time": "14:58:54", "remaining_time": "7:24:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2865, "total_steps": 4274, "loss": 0.0602, "learning_rate": 1.531287773105895e-05, "epoch": 1.3405076617148204, "percentage": 67.03, "elapsed_time": "15:00:26", "remaining_time": "7:22:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2870, "total_steps": 4274, "loss": 0.0633, "learning_rate": 1.5217037198921625e-05, "epoch": 1.3428471166218272, "percentage": 67.15, "elapsed_time": "15:01:59", "remaining_time": "7:21:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2875, "total_steps": 4274, "loss": 0.0694, "learning_rate": 1.512136614182137e-05, "epoch": 1.3451865715288338, "percentage": 67.27, "elapsed_time": "15:03:31", "remaining_time": "7:19:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2880, "total_steps": 4274, "loss": 0.0531, "learning_rate": 1.5025866217114592e-05, "epoch": 1.3475260264358404, "percentage": 67.38, "elapsed_time": "15:05:03", "remaining_time": "7:18:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2885, "total_steps": 4274, "loss": 0.0648, "learning_rate": 1.4930539079193078e-05, "epoch": 1.3498654813428472, "percentage": 67.5, "elapsed_time": "15:06:35", "remaining_time": "7:16:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2890, "total_steps": 4274, "loss": 0.0684, "learning_rate": 1.483538637945533e-05, "epoch": 1.3522049362498538, "percentage": 67.62, "elapsed_time": "15:08:08", "remaining_time": "7:14:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2895, "total_steps": 4274, "loss": 0.055, "learning_rate": 1.474040976627799e-05, "epoch": 1.3545443911568604, "percentage": 67.74, "elapsed_time": "15:09:40", "remaining_time": "7:13:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2900, "total_steps": 4274, "loss": 0.1258, "learning_rate": 1.4645610884987265e-05, "epoch": 1.3568838460638672, "percentage": 67.85, "elapsed_time": "15:11:12", "remaining_time": "7:11:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2905, "total_steps": 4274, "loss": 0.0607, "learning_rate": 1.4550991377830426e-05, "epoch": 1.3592233009708738, "percentage": 67.97, "elapsed_time": "15:12:44", "remaining_time": "7:10:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2910, "total_steps": 4274, "loss": 0.0618, "learning_rate": 1.4456552883947333e-05, "epoch": 1.3615627558778804, "percentage": 68.09, "elapsed_time": "15:14:16", "remaining_time": "7:08:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2915, "total_steps": 4274, "loss": 0.0648, "learning_rate": 1.4362297039342098e-05, "epoch": 1.363902210784887, "percentage": 68.2, "elapsed_time": "15:15:48", "remaining_time": "7:06:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2920, "total_steps": 4274, "loss": 0.0681, "learning_rate": 1.4268225476854704e-05, "epoch": 1.3662416656918939, "percentage": 68.32, "elapsed_time": "15:17:21", "remaining_time": "7:05:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2925, "total_steps": 4274, "loss": 0.0587, "learning_rate": 1.4174339826132727e-05, "epoch": 1.3685811205989005, "percentage": 68.44, "elapsed_time": "15:18:54", "remaining_time": "7:03:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2930, "total_steps": 4274, "loss": 0.0669, "learning_rate": 1.4080641713603072e-05, "epoch": 1.370920575505907, "percentage": 68.55, "elapsed_time": "15:20:26", "remaining_time": "7:02:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2935, "total_steps": 4274, "loss": 0.0601, "learning_rate": 1.3987132762443873e-05, "epoch": 1.3732600304129137, "percentage": 68.67, "elapsed_time": "15:21:58", "remaining_time": "7:00:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2940, "total_steps": 4274, "loss": 0.0599, "learning_rate": 1.3893814592556287e-05, "epoch": 1.3755994853199205, "percentage": 68.79, "elapsed_time": "15:23:30", "remaining_time": "6:59:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2945, "total_steps": 4274, "loss": 0.0539, "learning_rate": 1.3800688820536517e-05, "epoch": 1.377938940226927, "percentage": 68.91, "elapsed_time": "15:25:02", "remaining_time": "6:57:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2950, "total_steps": 4274, "loss": 0.0671, "learning_rate": 1.3707757059647725e-05, "epoch": 1.3802783951339337, "percentage": 69.02, "elapsed_time": "15:26:34", "remaining_time": "6:55:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2955, "total_steps": 4274, "loss": 0.0586, "learning_rate": 1.3615020919792151e-05, "epoch": 1.3826178500409405, "percentage": 69.14, "elapsed_time": "15:28:07", "remaining_time": "6:54:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2960, "total_steps": 4274, "loss": 0.0662, "learning_rate": 1.3522482007483172e-05, "epoch": 1.384957304947947, "percentage": 69.26, "elapsed_time": "15:29:39", "remaining_time": "6:52:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2965, "total_steps": 4274, "loss": 0.0637, "learning_rate": 1.3430141925817532e-05, "epoch": 1.3872967598549537, "percentage": 69.37, "elapsed_time": "15:31:11", "remaining_time": "6:51:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2970, "total_steps": 4274, "loss": 0.067, "learning_rate": 1.333800227444749e-05, "epoch": 1.3896362147619605, "percentage": 69.49, "elapsed_time": "15:32:43", "remaining_time": "6:49:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2975, "total_steps": 4274, "loss": 0.0618, "learning_rate": 1.324606464955318e-05, "epoch": 1.3919756696689671, "percentage": 69.61, "elapsed_time": "15:34:11", "remaining_time": "6:47:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2980, "total_steps": 4274, "loss": 0.0564, "learning_rate": 1.3154330643814938e-05, "epoch": 1.3943151245759737, "percentage": 69.72, "elapsed_time": "15:35:43", "remaining_time": "6:46:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2985, "total_steps": 4274, "loss": 0.0592, "learning_rate": 1.30628018463857e-05, "epoch": 1.3966545794829806, "percentage": 69.84, "elapsed_time": "15:37:15", "remaining_time": "6:44:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2990, "total_steps": 4274, "loss": 0.0533, "learning_rate": 1.2971479842863465e-05, "epoch": 1.3989940343899872, "percentage": 69.96, "elapsed_time": "15:38:48", "remaining_time": "6:43:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2995, "total_steps": 4274, "loss": 0.0629, "learning_rate": 1.2880366215263845e-05, "epoch": 1.4013334892969938, "percentage": 70.07, "elapsed_time": "15:40:20", "remaining_time": "6:41:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3000, "total_steps": 4274, "loss": 0.0553, "learning_rate": 1.2789462541992676e-05, "epoch": 1.4036729442040006, "percentage": 70.19, "elapsed_time": "15:41:51", "remaining_time": "6:39:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3000, "total_steps": 4274, "eval_loss": 0.07826597988605499, "epoch": 1.4036729442040006, "percentage": 70.19, "elapsed_time": "15:45:45", "remaining_time": "6:41:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3005, "total_steps": 4274, "loss": 0.13, "learning_rate": 1.2698770397818616e-05, "epoch": 1.4060123991110072, "percentage": 70.31, "elapsed_time": "15:47:18", "remaining_time": "6:40:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3010, "total_steps": 4274, "loss": 0.063, "learning_rate": 1.2608291353845941e-05, "epoch": 1.4083518540180138, "percentage": 70.43, "elapsed_time": "15:48:49", "remaining_time": "6:38:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3015, "total_steps": 4274, "loss": 0.0598, "learning_rate": 1.2518026977487252e-05, "epoch": 1.4106913089250206, "percentage": 70.54, "elapsed_time": "15:50:21", "remaining_time": "6:36:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3020, "total_steps": 4274, "loss": 0.063, "learning_rate": 1.242797883243638e-05, "epoch": 1.4130307638320272, "percentage": 70.66, "elapsed_time": "15:51:53", "remaining_time": "6:35:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3025, "total_steps": 4274, "loss": 0.0691, "learning_rate": 1.233814847864127e-05, "epoch": 1.4153702187390338, "percentage": 70.78, "elapsed_time": "15:53:25", "remaining_time": "6:33:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3030, "total_steps": 4274, "loss": 0.0628, "learning_rate": 1.224853747227698e-05, "epoch": 1.4177096736460404, "percentage": 70.89, "elapsed_time": "15:54:58", "remaining_time": "6:32:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3035, "total_steps": 4274, "loss": 0.0646, "learning_rate": 1.2159147365718664e-05, "epoch": 1.4200491285530472, "percentage": 71.01, "elapsed_time": "15:56:31", "remaining_time": "6:30:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3040, "total_steps": 4274, "loss": 0.0542, "learning_rate": 1.2069979707514766e-05, "epoch": 1.4223885834600538, "percentage": 71.13, "elapsed_time": "15:58:03", "remaining_time": "6:28:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3045, "total_steps": 4274, "loss": 0.059, "learning_rate": 1.1981036042360109e-05, "epoch": 1.4247280383670604, "percentage": 71.24, "elapsed_time": "15:59:35", "remaining_time": "6:27:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3050, "total_steps": 4274, "loss": 0.0575, "learning_rate": 1.1892317911069212e-05, "epoch": 1.427067493274067, "percentage": 71.36, "elapsed_time": "16:01:08", "remaining_time": "6:25:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3055, "total_steps": 4274, "loss": 0.0604, "learning_rate": 1.1803826850549526e-05, "epoch": 1.4294069481810738, "percentage": 71.48, "elapsed_time": "16:02:40", "remaining_time": "6:24:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3060, "total_steps": 4274, "loss": 0.0659, "learning_rate": 1.171556439377488e-05, "epoch": 1.4317464030880804, "percentage": 71.6, "elapsed_time": "16:04:13", "remaining_time": "6:22:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3065, "total_steps": 4274, "loss": 0.0605, "learning_rate": 1.1627532069758851e-05, "epoch": 1.434085857995087, "percentage": 71.71, "elapsed_time": "16:05:45", "remaining_time": "6:20:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3070, "total_steps": 4274, "loss": 0.0754, "learning_rate": 1.1539731403528353e-05, "epoch": 1.4364253129020939, "percentage": 71.83, "elapsed_time": "16:07:17", "remaining_time": "6:19:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3075, "total_steps": 4274, "loss": 0.0689, "learning_rate": 1.1452163916097141e-05, "epoch": 1.4387647678091005, "percentage": 71.95, "elapsed_time": "16:08:49", "remaining_time": "6:17:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3080, "total_steps": 4274, "loss": 0.0669, "learning_rate": 1.1364831124439526e-05, "epoch": 1.441104222716107, "percentage": 72.06, "elapsed_time": "16:10:21", "remaining_time": "6:16:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3085, "total_steps": 4274, "loss": 0.0596, "learning_rate": 1.1277734541464072e-05, "epoch": 1.4434436776231139, "percentage": 72.18, "elapsed_time": "16:11:54", "remaining_time": "6:14:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3090, "total_steps": 4274, "loss": 0.057, "learning_rate": 1.1190875675987356e-05, "epoch": 1.4457831325301205, "percentage": 72.3, "elapsed_time": "16:13:26", "remaining_time": "6:12:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3095, "total_steps": 4274, "loss": 0.0671, "learning_rate": 1.1104256032707894e-05, "epoch": 1.448122587437127, "percentage": 72.41, "elapsed_time": "16:14:59", "remaining_time": "6:11:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3100, "total_steps": 4274, "loss": 0.0583, "learning_rate": 1.1017877112180009e-05, "epoch": 1.450462042344134, "percentage": 72.53, "elapsed_time": "16:16:31", "remaining_time": "6:09:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3105, "total_steps": 4274, "loss": 0.0617, "learning_rate": 1.0931740410787895e-05, "epoch": 1.4528014972511405, "percentage": 72.65, "elapsed_time": "16:18:03", "remaining_time": "6:08:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3110, "total_steps": 4274, "loss": 0.0619, "learning_rate": 1.0845847420719632e-05, "epoch": 1.455140952158147, "percentage": 72.77, "elapsed_time": "16:19:36", "remaining_time": "6:06:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3115, "total_steps": 4274, "loss": 0.0572, "learning_rate": 1.076019962994141e-05, "epoch": 1.457480407065154, "percentage": 72.88, "elapsed_time": "16:21:08", "remaining_time": "6:05:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3120, "total_steps": 4274, "loss": 0.0632, "learning_rate": 1.0674798522171668e-05, "epoch": 1.4598198619721605, "percentage": 73.0, "elapsed_time": "16:22:40", "remaining_time": "6:03:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3125, "total_steps": 4274, "loss": 0.055, "learning_rate": 1.0589645576855477e-05, "epoch": 1.4621593168791671, "percentage": 73.12, "elapsed_time": "16:24:13", "remaining_time": "6:01:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3130, "total_steps": 4274, "loss": 0.0631, "learning_rate": 1.0504742269138835e-05, "epoch": 1.464498771786174, "percentage": 73.23, "elapsed_time": "16:25:45", "remaining_time": "6:00:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3135, "total_steps": 4274, "loss": 0.0557, "learning_rate": 1.0420090069843167e-05, "epoch": 1.4668382266931805, "percentage": 73.35, "elapsed_time": "16:27:17", "remaining_time": "5:58:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3140, "total_steps": 4274, "loss": 0.0623, "learning_rate": 1.0335690445439817e-05, "epoch": 1.4691776816001871, "percentage": 73.47, "elapsed_time": "16:28:49", "remaining_time": "5:57:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3145, "total_steps": 4274, "loss": 0.056, "learning_rate": 1.0251544858024662e-05, "epoch": 1.471517136507194, "percentage": 73.58, "elapsed_time": "16:30:20", "remaining_time": "5:55:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3150, "total_steps": 4274, "loss": 0.0581, "learning_rate": 1.0167654765292742e-05, "epoch": 1.4738565914142006, "percentage": 73.7, "elapsed_time": "16:31:53", "remaining_time": "5:53:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3155, "total_steps": 4274, "loss": 0.0619, "learning_rate": 1.0084021620513079e-05, "epoch": 1.4761960463212072, "percentage": 73.82, "elapsed_time": "16:33:23", "remaining_time": "5:52:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3160, "total_steps": 4274, "loss": 0.0784, "learning_rate": 1.0000646872503419e-05, "epoch": 1.4785355012282138, "percentage": 73.94, "elapsed_time": "16:34:55", "remaining_time": "5:50:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3165, "total_steps": 4274, "loss": 0.062, "learning_rate": 9.917531965605211e-06, "epoch": 1.4808749561352204, "percentage": 74.05, "elapsed_time": "16:36:27", "remaining_time": "5:49:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3170, "total_steps": 4274, "loss": 0.057, "learning_rate": 9.83467833965852e-06, "epoch": 1.4832144110422272, "percentage": 74.17, "elapsed_time": "16:38:00", "remaining_time": "5:47:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3175, "total_steps": 4274, "loss": 0.0674, "learning_rate": 9.752087429977143e-06, "epoch": 1.4855538659492338, "percentage": 74.29, "elapsed_time": "16:39:32", "remaining_time": "5:45:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3180, "total_steps": 4274, "loss": 0.0842, "learning_rate": 9.669760667323685e-06, "epoch": 1.4878933208562404, "percentage": 74.4, "elapsed_time": "16:41:04", "remaining_time": "5:44:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3185, "total_steps": 4274, "loss": 0.0736, "learning_rate": 9.58769947788482e-06, "epoch": 1.4902327757632472, "percentage": 74.52, "elapsed_time": "16:42:37", "remaining_time": "5:42:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3190, "total_steps": 4274, "loss": 0.0638, "learning_rate": 9.505905283246578e-06, "epoch": 1.4925722306702538, "percentage": 74.64, "elapsed_time": "16:44:09", "remaining_time": "5:41:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3195, "total_steps": 4274, "loss": 0.0672, "learning_rate": 9.424379500369674e-06, "epoch": 1.4949116855772604, "percentage": 74.75, "elapsed_time": "16:45:41", "remaining_time": "5:39:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3200, "total_steps": 4274, "loss": 0.058, "learning_rate": 9.343123541565035e-06, "epoch": 1.4972511404842672, "percentage": 74.87, "elapsed_time": "16:47:13", "remaining_time": "5:38:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3205, "total_steps": 4274, "loss": 0.0588, "learning_rate": 9.26213881446926e-06, "epoch": 1.4995905953912738, "percentage": 74.99, "elapsed_time": "16:48:46", "remaining_time": "5:36:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3210, "total_steps": 4274, "loss": 0.0626, "learning_rate": 9.181426722020297e-06, "epoch": 1.5019300502982804, "percentage": 75.11, "elapsed_time": "16:50:18", "remaining_time": "5:34:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3215, "total_steps": 4274, "loss": 0.0824, "learning_rate": 9.100988662433082e-06, "epoch": 1.5042695052052872, "percentage": 75.22, "elapsed_time": "16:51:50", "remaining_time": "5:33:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3220, "total_steps": 4274, "loss": 0.0621, "learning_rate": 9.020826029175384e-06, "epoch": 1.5066089601122938, "percentage": 75.34, "elapsed_time": "16:53:23", "remaining_time": "5:31:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3225, "total_steps": 4274, "loss": 0.0632, "learning_rate": 8.94094021094358e-06, "epoch": 1.5089484150193004, "percentage": 75.46, "elapsed_time": "16:54:55", "remaining_time": "5:30:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3230, "total_steps": 4274, "loss": 0.0638, "learning_rate": 8.86133259163869e-06, "epoch": 1.5112878699263073, "percentage": 75.57, "elapsed_time": "16:56:27", "remaining_time": "5:28:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3235, "total_steps": 4274, "loss": 0.0586, "learning_rate": 8.782004550342323e-06, "epoch": 1.5136273248333139, "percentage": 75.69, "elapsed_time": "16:57:59", "remaining_time": "5:26:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3240, "total_steps": 4274, "loss": 0.0632, "learning_rate": 8.702957461292846e-06, "epoch": 1.5159667797403205, "percentage": 75.81, "elapsed_time": "16:59:32", "remaining_time": "5:25:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3245, "total_steps": 4274, "loss": 0.0578, "learning_rate": 8.624192693861544e-06, "epoch": 1.5183062346473273, "percentage": 75.92, "elapsed_time": "17:01:04", "remaining_time": "5:23:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3250, "total_steps": 4274, "loss": 0.063, "learning_rate": 8.545711612528915e-06, "epoch": 1.5206456895543339, "percentage": 76.04, "elapsed_time": "17:02:37", "remaining_time": "5:22:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3255, "total_steps": 4274, "loss": 0.0657, "learning_rate": 8.467515576861002e-06, "epoch": 1.5229851444613405, "percentage": 76.16, "elapsed_time": "17:04:09", "remaining_time": "5:20:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3260, "total_steps": 4274, "loss": 0.1184, "learning_rate": 8.389605941485893e-06, "epoch": 1.5253245993683473, "percentage": 76.28, "elapsed_time": "17:05:42", "remaining_time": "5:19:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3265, "total_steps": 4274, "loss": 0.055, "learning_rate": 8.31198405607019e-06, "epoch": 1.5276640542753537, "percentage": 76.39, "elapsed_time": "17:07:14", "remaining_time": "5:17:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3270, "total_steps": 4274, "loss": 0.0669, "learning_rate": 8.234651265295702e-06, "epoch": 1.5300035091823605, "percentage": 76.51, "elapsed_time": "17:08:46", "remaining_time": "5:15:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3275, "total_steps": 4274, "loss": 0.0577, "learning_rate": 8.15760890883607e-06, "epoch": 1.5323429640893673, "percentage": 76.63, "elapsed_time": "17:10:19", "remaining_time": "5:14:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3280, "total_steps": 4274, "loss": 0.063, "learning_rate": 8.080858321333633e-06, "epoch": 1.5346824189963737, "percentage": 76.74, "elapsed_time": "17:11:52", "remaining_time": "5:12:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3285, "total_steps": 4274, "loss": 0.0617, "learning_rate": 8.004400832376258e-06, "epoch": 1.5370218739033805, "percentage": 76.86, "elapsed_time": "17:13:24", "remaining_time": "5:11:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3290, "total_steps": 4274, "loss": 0.0668, "learning_rate": 7.928237766474314e-06, "epoch": 1.5393613288103873, "percentage": 76.98, "elapsed_time": "17:14:56", "remaining_time": "5:09:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3295, "total_steps": 4274, "loss": 0.0604, "learning_rate": 7.852370443037751e-06, "epoch": 1.5417007837173937, "percentage": 77.09, "elapsed_time": "17:16:29", "remaining_time": "5:07:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3300, "total_steps": 4274, "loss": 0.0852, "learning_rate": 7.776800176353227e-06, "epoch": 1.5440402386244005, "percentage": 77.21, "elapsed_time": "17:18:01", "remaining_time": "5:06:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3305, "total_steps": 4274, "loss": 0.0571, "learning_rate": 7.701528275561348e-06, "epoch": 1.5463796935314071, "percentage": 77.33, "elapsed_time": "17:19:34", "remaining_time": "5:04:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3310, "total_steps": 4274, "loss": 0.0606, "learning_rate": 7.626556044633951e-06, "epoch": 1.5487191484384137, "percentage": 77.45, "elapsed_time": "17:21:06", "remaining_time": "5:03:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3315, "total_steps": 4274, "loss": 0.0582, "learning_rate": 7.551884782351587e-06, "epoch": 1.5510586033454206, "percentage": 77.56, "elapsed_time": "17:22:38", "remaining_time": "5:01:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3320, "total_steps": 4274, "loss": 0.0656, "learning_rate": 7.477515782280939e-06, "epoch": 1.5533980582524272, "percentage": 77.68, "elapsed_time": "17:24:11", "remaining_time": "5:00:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3325, "total_steps": 4274, "loss": 0.0685, "learning_rate": 7.403450332752493e-06, "epoch": 1.5557375131594338, "percentage": 77.8, "elapsed_time": "17:25:43", "remaining_time": "4:58:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3330, "total_steps": 4274, "loss": 0.06, "learning_rate": 7.329689716838148e-06, "epoch": 1.5580769680664406, "percentage": 77.91, "elapsed_time": "17:27:16", "remaining_time": "4:56:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3335, "total_steps": 4274, "loss": 0.0537, "learning_rate": 7.256235212329055e-06, "epoch": 1.5604164229734472, "percentage": 78.03, "elapsed_time": "17:28:48", "remaining_time": "4:55:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3340, "total_steps": 4274, "loss": 0.0676, "learning_rate": 7.183088091713405e-06, "epoch": 1.5627558778804538, "percentage": 78.15, "elapsed_time": "17:30:19", "remaining_time": "4:53:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3345, "total_steps": 4274, "loss": 0.0596, "learning_rate": 7.110249622154469e-06, "epoch": 1.5650953327874606, "percentage": 78.26, "elapsed_time": "17:31:51", "remaining_time": "4:52:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3350, "total_steps": 4274, "loss": 0.0504, "learning_rate": 7.037721065468591e-06, "epoch": 1.5674347876944672, "percentage": 78.38, "elapsed_time": "17:33:22", "remaining_time": "4:50:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3355, "total_steps": 4274, "loss": 0.0688, "learning_rate": 6.96550367810333e-06, "epoch": 1.5697742426014738, "percentage": 78.5, "elapsed_time": "17:34:55", "remaining_time": "4:48:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3360, "total_steps": 4274, "loss": 0.0545, "learning_rate": 6.89359871111572e-06, "epoch": 1.5721136975084806, "percentage": 78.61, "elapsed_time": "17:36:27", "remaining_time": "4:47:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3365, "total_steps": 4274, "loss": 0.0624, "learning_rate": 6.822007410150591e-06, "epoch": 1.5744531524154872, "percentage": 78.73, "elapsed_time": "17:38:00", "remaining_time": "4:45:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3370, "total_steps": 4274, "loss": 0.0659, "learning_rate": 6.750731015418959e-06, "epoch": 1.5767926073224938, "percentage": 78.85, "elapsed_time": "17:39:32", "remaining_time": "4:44:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3375, "total_steps": 4274, "loss": 0.062, "learning_rate": 6.679770761676596e-06, "epoch": 1.5791320622295006, "percentage": 78.97, "elapsed_time": "17:41:05", "remaining_time": "4:42:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3380, "total_steps": 4274, "loss": 0.0522, "learning_rate": 6.609127878202581e-06, "epoch": 1.5814715171365072, "percentage": 79.08, "elapsed_time": "17:42:37", "remaining_time": "4:41:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3385, "total_steps": 4274, "loss": 0.0588, "learning_rate": 6.538803588778064e-06, "epoch": 1.5838109720435138, "percentage": 79.2, "elapsed_time": "17:44:10", "remaining_time": "4:39:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3390, "total_steps": 4274, "loss": 0.0641, "learning_rate": 6.468799111665003e-06, "epoch": 1.5861504269505207, "percentage": 79.32, "elapsed_time": "17:45:42", "remaining_time": "4:37:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3395, "total_steps": 4274, "loss": 0.0668, "learning_rate": 6.399115659585103e-06, "epoch": 1.588489881857527, "percentage": 79.43, "elapsed_time": "17:47:15", "remaining_time": "4:36:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3400, "total_steps": 4274, "loss": 0.0531, "learning_rate": 6.329754439698804e-06, "epoch": 1.5908293367645339, "percentage": 79.55, "elapsed_time": "17:48:47", "remaining_time": "4:34:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3405, "total_steps": 4274, "loss": 0.0992, "learning_rate": 6.260716653584359e-06, "epoch": 1.5931687916715407, "percentage": 79.67, "elapsed_time": "17:50:20", "remaining_time": "4:33:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3410, "total_steps": 4274, "loss": 0.0574, "learning_rate": 6.192003497217016e-06, "epoch": 1.595508246578547, "percentage": 79.78, "elapsed_time": "17:51:52", "remaining_time": "4:31:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3415, "total_steps": 4274, "loss": 0.0616, "learning_rate": 6.123616160948295e-06, "epoch": 1.5978477014855539, "percentage": 79.9, "elapsed_time": "17:53:24", "remaining_time": "4:30:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3420, "total_steps": 4274, "loss": 0.0584, "learning_rate": 6.055555829485396e-06, "epoch": 1.6001871563925605, "percentage": 80.02, "elapsed_time": "17:54:57", "remaining_time": "4:28:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3425, "total_steps": 4274, "loss": 0.0557, "learning_rate": 5.98782368187063e-06, "epoch": 1.602526611299567, "percentage": 80.14, "elapsed_time": "17:56:29", "remaining_time": "4:26:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3430, "total_steps": 4274, "loss": 0.0618, "learning_rate": 5.920420891461046e-06, "epoch": 1.604866066206574, "percentage": 80.25, "elapsed_time": "17:58:02", "remaining_time": "4:25:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3435, "total_steps": 4274, "loss": 0.0623, "learning_rate": 5.853348625908048e-06, "epoch": 1.6072055211135805, "percentage": 80.37, "elapsed_time": "17:59:34", "remaining_time": "4:23:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3440, "total_steps": 4274, "loss": 0.0564, "learning_rate": 5.786608047137229e-06, "epoch": 1.609544976020587, "percentage": 80.49, "elapsed_time": "18:01:07", "remaining_time": "4:22:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3445, "total_steps": 4274, "loss": 0.0595, "learning_rate": 5.720200311328178e-06, "epoch": 1.611884430927594, "percentage": 80.6, "elapsed_time": "18:02:39", "remaining_time": "4:20:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3450, "total_steps": 4274, "loss": 0.0595, "learning_rate": 5.654126568894513e-06, "epoch": 1.6142238858346005, "percentage": 80.72, "elapsed_time": "18:04:11", "remaining_time": "4:18:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3455, "total_steps": 4274, "loss": 0.0606, "learning_rate": 5.5883879644638925e-06, "epoch": 1.6165633407416071, "percentage": 80.84, "elapsed_time": "18:05:44", "remaining_time": "4:17:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3460, "total_steps": 4274, "loss": 0.058, "learning_rate": 5.522985636858239e-06, "epoch": 1.618902795648614, "percentage": 80.95, "elapsed_time": "18:07:16", "remaining_time": "4:15:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3465, "total_steps": 4274, "loss": 0.0578, "learning_rate": 5.4579207190739775e-06, "epoch": 1.6212422505556205, "percentage": 81.07, "elapsed_time": "18:08:49", "remaining_time": "4:14:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3470, "total_steps": 4274, "loss": 0.0651, "learning_rate": 5.393194338262428e-06, "epoch": 1.6235817054626271, "percentage": 81.19, "elapsed_time": "18:10:21", "remaining_time": "4:12:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3475, "total_steps": 4274, "loss": 0.0586, "learning_rate": 5.328807615710246e-06, "epoch": 1.625921160369634, "percentage": 81.31, "elapsed_time": "18:11:53", "remaining_time": "4:11:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3480, "total_steps": 4274, "loss": 0.059, "learning_rate": 5.264761666820054e-06, "epoch": 1.6282606152766406, "percentage": 81.42, "elapsed_time": "18:13:25", "remaining_time": "4:09:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3485, "total_steps": 4274, "loss": 0.0574, "learning_rate": 5.201057601091056e-06, "epoch": 1.6306000701836472, "percentage": 81.54, "elapsed_time": "18:14:57", "remaining_time": "4:07:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3490, "total_steps": 4274, "loss": 0.0577, "learning_rate": 5.137696522099858e-06, "epoch": 1.632939525090654, "percentage": 81.66, "elapsed_time": "18:16:29", "remaining_time": "4:06:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3495, "total_steps": 4274, "loss": 0.055, "learning_rate": 5.074679527481349e-06, "epoch": 1.6352789799976606, "percentage": 81.77, "elapsed_time": "18:18:02", "remaining_time": "4:04:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3500, "total_steps": 4274, "loss": 0.057, "learning_rate": 5.012007708909661e-06, "epoch": 1.6376184349046672, "percentage": 81.89, "elapsed_time": "18:19:34", "remaining_time": "4:03:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3500, "total_steps": 4274, "eval_loss": 0.07722621411085129, "epoch": 1.6376184349046672, "percentage": 81.89, "elapsed_time": "18:23:30", "remaining_time": "4:04:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3505, "total_steps": 4274, "loss": 0.0658, "learning_rate": 4.949682152079291e-06, "epoch": 1.639957889811674, "percentage": 82.01, "elapsed_time": "18:25:03", "remaining_time": "4:02:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3510, "total_steps": 4274, "loss": 0.0549, "learning_rate": 4.887703936686252e-06, "epoch": 1.6422973447186804, "percentage": 82.12, "elapsed_time": "18:26:35", "remaining_time": "4:00:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3515, "total_steps": 4274, "loss": 0.0665, "learning_rate": 4.826074136409411e-06, "epoch": 1.6446367996256872, "percentage": 82.24, "elapsed_time": "18:28:07", "remaining_time": "3:59:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3520, "total_steps": 4274, "loss": 0.0557, "learning_rate": 4.764793818891866e-06, "epoch": 1.646976254532694, "percentage": 82.36, "elapsed_time": "18:29:40", "remaining_time": "3:57:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3525, "total_steps": 4274, "loss": 0.0609, "learning_rate": 4.703864045722453e-06, "epoch": 1.6493157094397004, "percentage": 82.48, "elapsed_time": "18:31:12", "remaining_time": "3:56:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3530, "total_steps": 4274, "loss": 0.0602, "learning_rate": 4.643285872417346e-06, "epoch": 1.6516551643467072, "percentage": 82.59, "elapsed_time": "18:32:45", "remaining_time": "3:54:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3535, "total_steps": 4274, "loss": 0.0529, "learning_rate": 4.583060348401808e-06, "epoch": 1.653994619253714, "percentage": 82.71, "elapsed_time": "18:34:17", "remaining_time": "3:52:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3540, "total_steps": 4274, "loss": 0.0598, "learning_rate": 4.523188516991958e-06, "epoch": 1.6563340741607204, "percentage": 82.83, "elapsed_time": "18:35:50", "remaining_time": "3:51:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3545, "total_steps": 4274, "loss": 0.058, "learning_rate": 4.4636714153767496e-06, "epoch": 1.6586735290677272, "percentage": 82.94, "elapsed_time": "18:37:22", "remaining_time": "3:49:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3550, "total_steps": 4274, "loss": 0.061, "learning_rate": 4.40451007459996e-06, "epoch": 1.6610129839747338, "percentage": 83.06, "elapsed_time": "18:38:55", "remaining_time": "3:48:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3555, "total_steps": 4274, "loss": 0.0557, "learning_rate": 4.3457055195423645e-06, "epoch": 1.6633524388817404, "percentage": 83.18, "elapsed_time": "18:40:27", "remaining_time": "3:46:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3560, "total_steps": 4274, "loss": 0.055, "learning_rate": 4.2872587689039484e-06, "epoch": 1.6656918937887473, "percentage": 83.29, "elapsed_time": "18:42:00", "remaining_time": "3:45:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3565, "total_steps": 4274, "loss": 0.0573, "learning_rate": 4.229170835186292e-06, "epoch": 1.6680313486957539, "percentage": 83.41, "elapsed_time": "18:43:32", "remaining_time": "3:43:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3570, "total_steps": 4274, "loss": 0.0599, "learning_rate": 4.171442724675004e-06, "epoch": 1.6703708036027605, "percentage": 83.53, "elapsed_time": "18:45:05", "remaining_time": "3:41:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3575, "total_steps": 4274, "loss": 0.0663, "learning_rate": 4.114075437422313e-06, "epoch": 1.6727102585097673, "percentage": 83.65, "elapsed_time": "18:46:37", "remaining_time": "3:40:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3580, "total_steps": 4274, "loss": 0.0759, "learning_rate": 4.057069967229718e-06, "epoch": 1.6750497134167739, "percentage": 83.76, "elapsed_time": "18:48:09", "remaining_time": "3:38:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3585, "total_steps": 4274, "loss": 0.0576, "learning_rate": 4.00042730163078e-06, "epoch": 1.6773891683237805, "percentage": 83.88, "elapsed_time": "18:49:42", "remaining_time": "3:37:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3590, "total_steps": 4274, "loss": 0.0579, "learning_rate": 3.944148421874042e-06, "epoch": 1.6797286232307873, "percentage": 84.0, "elapsed_time": "18:51:14", "remaining_time": "3:35:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3595, "total_steps": 4274, "loss": 0.0625, "learning_rate": 3.888234302905977e-06, "epoch": 1.682068078137794, "percentage": 84.11, "elapsed_time": "18:52:47", "remaining_time": "3:33:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3600, "total_steps": 4274, "loss": 0.0551, "learning_rate": 3.832685913354159e-06, "epoch": 1.6844075330448005, "percentage": 84.23, "elapsed_time": "18:54:19", "remaining_time": "3:32:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3605, "total_steps": 4274, "loss": 0.0643, "learning_rate": 3.777504215510427e-06, "epoch": 1.6867469879518073, "percentage": 84.35, "elapsed_time": "18:55:51", "remaining_time": "3:30:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3610, "total_steps": 4274, "loss": 0.0616, "learning_rate": 3.7226901653142634e-06, "epoch": 1.689086442858814, "percentage": 84.46, "elapsed_time": "18:57:24", "remaining_time": "3:29:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3615, "total_steps": 4274, "loss": 0.0688, "learning_rate": 3.6682447123361885e-06, "epoch": 1.6914258977658205, "percentage": 84.58, "elapsed_time": "18:58:56", "remaining_time": "3:27:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3620, "total_steps": 4274, "loss": 0.062, "learning_rate": 3.61416879976135e-06, "epoch": 1.6937653526728274, "percentage": 84.7, "elapsed_time": "19:00:29", "remaining_time": "3:26:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3625, "total_steps": 4274, "loss": 0.0552, "learning_rate": 3.560463364373165e-06, "epoch": 1.696104807579834, "percentage": 84.82, "elapsed_time": "19:02:01", "remaining_time": "3:24:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3630, "total_steps": 4274, "loss": 0.0694, "learning_rate": 3.5071293365370904e-06, "epoch": 1.6984442624868406, "percentage": 84.93, "elapsed_time": "19:03:33", "remaining_time": "3:22:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3635, "total_steps": 4274, "loss": 0.0562, "learning_rate": 3.4541676401845027e-06, "epoch": 1.7007837173938474, "percentage": 85.05, "elapsed_time": "19:05:06", "remaining_time": "3:21:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3640, "total_steps": 4274, "loss": 0.0596, "learning_rate": 3.4015791927967083e-06, "epoch": 1.7031231723008537, "percentage": 85.17, "elapsed_time": "19:06:38", "remaining_time": "3:19:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3645, "total_steps": 4274, "loss": 0.0797, "learning_rate": 3.3493649053890326e-06, "epoch": 1.7054626272078606, "percentage": 85.28, "elapsed_time": "19:08:08", "remaining_time": "3:18:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3650, "total_steps": 4274, "loss": 0.0594, "learning_rate": 3.2975256824950564e-06, "epoch": 1.7078020821148674, "percentage": 85.4, "elapsed_time": "19:09:41", "remaining_time": "3:16:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3655, "total_steps": 4274, "loss": 0.0563, "learning_rate": 3.2460624221509137e-06, "epoch": 1.7101415370218738, "percentage": 85.52, "elapsed_time": "19:11:13", "remaining_time": "3:14:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3660, "total_steps": 4274, "loss": 0.0614, "learning_rate": 3.1949760158797833e-06, "epoch": 1.7124809919288806, "percentage": 85.63, "elapsed_time": "19:12:46", "remaining_time": "3:13:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3665, "total_steps": 4274, "loss": 0.0648, "learning_rate": 3.1442673486763912e-06, "epoch": 1.7148204468358872, "percentage": 85.75, "elapsed_time": "19:14:18", "remaining_time": "3:11:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3670, "total_steps": 4274, "loss": 0.0593, "learning_rate": 3.0939372989917254e-06, "epoch": 1.7171599017428938, "percentage": 85.87, "elapsed_time": "19:15:51", "remaining_time": "3:10:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3675, "total_steps": 4274, "loss": 0.0588, "learning_rate": 3.04398673871778e-06, "epoch": 1.7194993566499006, "percentage": 85.99, "elapsed_time": "19:17:23", "remaining_time": "3:08:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3680, "total_steps": 4274, "loss": 0.0567, "learning_rate": 2.9944165331724866e-06, "epoch": 1.7218388115569072, "percentage": 86.1, "elapsed_time": "19:18:56", "remaining_time": "3:07:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3685, "total_steps": 4274, "loss": 0.0548, "learning_rate": 2.945227541084697e-06, "epoch": 1.7241782664639138, "percentage": 86.22, "elapsed_time": "19:20:29", "remaining_time": "3:05:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3690, "total_steps": 4274, "loss": 0.0515, "learning_rate": 2.896420614579312e-06, "epoch": 1.7265177213709206, "percentage": 86.34, "elapsed_time": "19:22:01", "remaining_time": "3:03:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3695, "total_steps": 4274, "loss": 0.0575, "learning_rate": 2.847996599162539e-06, "epoch": 1.7288571762779272, "percentage": 86.45, "elapsed_time": "19:23:34", "remaining_time": "3:02:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3700, "total_steps": 4274, "loss": 0.064, "learning_rate": 2.799956333707207e-06, "epoch": 1.7311966311849338, "percentage": 86.57, "elapsed_time": "19:25:06", "remaining_time": "3:00:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3705, "total_steps": 4274, "loss": 0.0572, "learning_rate": 2.752300650438283e-06, "epoch": 1.7335360860919407, "percentage": 86.69, "elapsed_time": "19:26:39", "remaining_time": "2:59:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3710, "total_steps": 4274, "loss": 0.0676, "learning_rate": 2.7050303749184046e-06, "epoch": 1.7358755409989473, "percentage": 86.8, "elapsed_time": "19:28:11", "remaining_time": "2:57:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3715, "total_steps": 4274, "loss": 0.0646, "learning_rate": 2.6581463260336252e-06, "epoch": 1.7382149959059539, "percentage": 86.92, "elapsed_time": "19:29:43", "remaining_time": "2:56:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3720, "total_steps": 4274, "loss": 0.0499, "learning_rate": 2.6116493159791866e-06, "epoch": 1.7405544508129607, "percentage": 87.04, "elapsed_time": "19:31:16", "remaining_time": "2:54:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3725, "total_steps": 4274, "loss": 0.0668, "learning_rate": 2.5655401502454795e-06, "epoch": 1.7428939057199673, "percentage": 87.15, "elapsed_time": "19:32:48", "remaining_time": "2:52:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3730, "total_steps": 4274, "loss": 0.0579, "learning_rate": 2.5198196276040782e-06, "epoch": 1.7452333606269739, "percentage": 87.27, "elapsed_time": "19:34:21", "remaining_time": "2:51:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3735, "total_steps": 4274, "loss": 0.0642, "learning_rate": 2.4744885400939026e-06, "epoch": 1.7475728155339807, "percentage": 87.39, "elapsed_time": "19:35:53", "remaining_time": "2:49:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3740, "total_steps": 4274, "loss": 0.0561, "learning_rate": 2.42954767300749e-06, "epoch": 1.7499122704409873, "percentage": 87.51, "elapsed_time": "19:37:26", "remaining_time": "2:48:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3745, "total_steps": 4274, "loss": 0.0545, "learning_rate": 2.384997804877412e-06, "epoch": 1.752251725347994, "percentage": 87.62, "elapsed_time": "19:38:58", "remaining_time": "2:46:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3750, "total_steps": 4274, "loss": 0.0556, "learning_rate": 2.340839707462755e-06, "epoch": 1.7545911802550007, "percentage": 87.74, "elapsed_time": "19:40:31", "remaining_time": "2:44:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3755, "total_steps": 4274, "loss": 0.0587, "learning_rate": 2.2970741457357973e-06, "epoch": 1.756930635162007, "percentage": 87.86, "elapsed_time": "19:42:03", "remaining_time": "2:43:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3760, "total_steps": 4274, "loss": 0.0729, "learning_rate": 2.2537018778687026e-06, "epoch": 1.759270090069014, "percentage": 87.97, "elapsed_time": "19:43:36", "remaining_time": "2:41:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3765, "total_steps": 4274, "loss": 0.0619, "learning_rate": 2.210723655220434e-06, "epoch": 1.7616095449760207, "percentage": 88.09, "elapsed_time": "19:45:08", "remaining_time": "2:40:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3770, "total_steps": 4274, "loss": 0.0619, "learning_rate": 2.168140222323703e-06, "epoch": 1.7639489998830271, "percentage": 88.21, "elapsed_time": "19:46:40", "remaining_time": "2:38:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3775, "total_steps": 4274, "loss": 0.0672, "learning_rate": 2.1259523168720957e-06, "epoch": 1.766288454790034, "percentage": 88.32, "elapsed_time": "19:48:13", "remaining_time": "2:37:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3780, "total_steps": 4274, "loss": 0.056, "learning_rate": 2.0841606697072745e-06, "epoch": 1.7686279096970405, "percentage": 88.44, "elapsed_time": "19:49:45", "remaining_time": "2:35:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3785, "total_steps": 4274, "loss": 0.1111, "learning_rate": 2.042766004806332e-06, "epoch": 1.7709673646040471, "percentage": 88.56, "elapsed_time": "19:51:18", "remaining_time": "2:33:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3790, "total_steps": 4274, "loss": 0.0606, "learning_rate": 2.001769039269247e-06, "epoch": 1.773306819511054, "percentage": 88.68, "elapsed_time": "19:52:50", "remaining_time": "2:32:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3795, "total_steps": 4274, "loss": 0.0702, "learning_rate": 1.9611704833064486e-06, "epoch": 1.7756462744180606, "percentage": 88.79, "elapsed_time": "19:54:23", "remaining_time": "2:30:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3800, "total_steps": 4274, "loss": 0.0586, "learning_rate": 1.9209710402265295e-06, "epoch": 1.7779857293250672, "percentage": 88.91, "elapsed_time": "19:55:55", "remaining_time": "2:29:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3805, "total_steps": 4274, "loss": 0.059, "learning_rate": 1.8811714064240481e-06, "epoch": 1.780325184232074, "percentage": 89.03, "elapsed_time": "19:57:27", "remaining_time": "2:27:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3810, "total_steps": 4274, "loss": 0.0543, "learning_rate": 1.841772271367484e-06, "epoch": 1.7826646391390806, "percentage": 89.14, "elapsed_time": "19:59:00", "remaining_time": "2:26:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3815, "total_steps": 4274, "loss": 0.0507, "learning_rate": 1.8027743175872664e-06, "epoch": 1.7850040940460872, "percentage": 89.26, "elapsed_time": "20:00:32", "remaining_time": "2:24:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3820, "total_steps": 4274, "loss": 0.0601, "learning_rate": 1.7641782206639813e-06, "epoch": 1.787343548953094, "percentage": 89.38, "elapsed_time": "20:02:04", "remaining_time": "2:22:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3825, "total_steps": 4274, "loss": 0.0604, "learning_rate": 1.7259846492166359e-06, "epoch": 1.7896830038601006, "percentage": 89.49, "elapsed_time": "20:03:36", "remaining_time": "2:21:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3830, "total_steps": 4274, "loss": 0.0547, "learning_rate": 1.6881942648911076e-06, "epoch": 1.7920224587671072, "percentage": 89.61, "elapsed_time": "20:05:08", "remaining_time": "2:19:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3835, "total_steps": 4274, "loss": 0.0605, "learning_rate": 1.6508077223486546e-06, "epoch": 1.794361913674114, "percentage": 89.73, "elapsed_time": "20:06:39", "remaining_time": "2:18:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3840, "total_steps": 4274, "loss": 0.0553, "learning_rate": 1.61382566925459e-06, "epoch": 1.7967013685811206, "percentage": 89.85, "elapsed_time": "20:08:12", "remaining_time": "2:16:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3845, "total_steps": 4274, "loss": 0.0654, "learning_rate": 1.5772487462670681e-06, "epoch": 1.7990408234881272, "percentage": 89.96, "elapsed_time": "20:09:44", "remaining_time": "2:14:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3850, "total_steps": 4274, "loss": 0.0558, "learning_rate": 1.5410775870259686e-06, "epoch": 1.801380278395134, "percentage": 90.08, "elapsed_time": "20:11:16", "remaining_time": "2:13:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3855, "total_steps": 4274, "loss": 0.06, "learning_rate": 1.5053128181419184e-06, "epoch": 1.8037197333021406, "percentage": 90.2, "elapsed_time": "20:12:47", "remaining_time": "2:11:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3860, "total_steps": 4274, "loss": 0.0645, "learning_rate": 1.469955059185471e-06, "epoch": 1.8060591882091472, "percentage": 90.31, "elapsed_time": "20:14:19", "remaining_time": "2:10:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3865, "total_steps": 4274, "loss": 0.0631, "learning_rate": 1.4350049226763224e-06, "epoch": 1.808398643116154, "percentage": 90.43, "elapsed_time": "20:15:52", "remaining_time": "2:08:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3870, "total_steps": 4274, "loss": 0.0639, "learning_rate": 1.400463014072742e-06, "epoch": 1.8107380980231607, "percentage": 90.55, "elapsed_time": "20:17:24", "remaining_time": "2:07:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3875, "total_steps": 4274, "loss": 0.0631, "learning_rate": 1.3663299317610595e-06, "epoch": 1.8130775529301673, "percentage": 90.66, "elapsed_time": "20:18:56", "remaining_time": "2:05:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3880, "total_steps": 4274, "loss": 0.0693, "learning_rate": 1.3326062670453194e-06, "epoch": 1.815417007837174, "percentage": 90.78, "elapsed_time": "20:20:27", "remaining_time": "2:03:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3885, "total_steps": 4274, "loss": 0.059, "learning_rate": 1.2992926041370064e-06, "epoch": 1.8177564627441805, "percentage": 90.9, "elapsed_time": "20:21:58", "remaining_time": "2:02:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3890, "total_steps": 4274, "loss": 0.058, "learning_rate": 1.2663895201449588e-06, "epoch": 1.8200959176511873, "percentage": 91.02, "elapsed_time": "20:23:30", "remaining_time": "2:00:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3895, "total_steps": 4274, "loss": 0.0562, "learning_rate": 1.2338975850653579e-06, "epoch": 1.822435372558194, "percentage": 91.13, "elapsed_time": "20:25:02", "remaining_time": "1:59:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3900, "total_steps": 4274, "loss": 0.0555, "learning_rate": 1.201817361771837e-06, "epoch": 1.8247748274652005, "percentage": 91.25, "elapsed_time": "20:26:35", "remaining_time": "1:57:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3905, "total_steps": 4274, "loss": 0.0635, "learning_rate": 1.1701494060057606e-06, "epoch": 1.8271142823722073, "percentage": 91.37, "elapsed_time": "20:28:07", "remaining_time": "1:56:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3910, "total_steps": 4274, "loss": 0.0628, "learning_rate": 1.138894266366572e-06, "epoch": 1.829453737279214, "percentage": 91.48, "elapsed_time": "20:29:40", "remaining_time": "1:54:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3915, "total_steps": 4274, "loss": 0.0569, "learning_rate": 1.108052484302302e-06, "epoch": 1.8317931921862205, "percentage": 91.6, "elapsed_time": "20:31:12", "remaining_time": "1:52:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3920, "total_steps": 4274, "loss": 0.0558, "learning_rate": 1.0776245941001878e-06, "epoch": 1.8341326470932273, "percentage": 91.72, "elapsed_time": "20:32:44", "remaining_time": "1:51:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3925, "total_steps": 4274, "loss": 0.0571, "learning_rate": 1.0476111228774178e-06, "epoch": 1.836472102000234, "percentage": 91.83, "elapsed_time": "20:34:16", "remaining_time": "1:49:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3930, "total_steps": 4274, "loss": 0.054, "learning_rate": 1.01801259057199e-06, "epoch": 1.8388115569072405, "percentage": 91.95, "elapsed_time": "20:35:48", "remaining_time": "1:48:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3935, "total_steps": 4274, "loss": 0.0611, "learning_rate": 9.888295099337252e-07, "epoch": 1.8411510118142473, "percentage": 92.07, "elapsed_time": "20:37:20", "remaining_time": "1:46:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3940, "total_steps": 4274, "loss": 0.065, "learning_rate": 9.60062386515359e-07, "epoch": 1.843490466721254, "percentage": 92.19, "elapsed_time": "20:38:52", "remaining_time": "1:45:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3945, "total_steps": 4274, "loss": 0.0591, "learning_rate": 9.317117186638108e-07, "epoch": 1.8458299216282605, "percentage": 92.3, "elapsed_time": "20:40:25", "remaining_time": "1:43:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3950, "total_steps": 4274, "loss": 0.0567, "learning_rate": 9.037779975115235e-07, "epoch": 1.8481693765352674, "percentage": 92.42, "elapsed_time": "20:41:57", "remaining_time": "1:41:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3955, "total_steps": 4274, "loss": 0.0613, "learning_rate": 8.762617069679846e-07, "epoch": 1.850508831442274, "percentage": 92.54, "elapsed_time": "20:43:30", "remaining_time": "1:40:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3960, "total_steps": 4274, "loss": 0.0593, "learning_rate": 8.491633237113078e-07, "epoch": 1.8528482863492806, "percentage": 92.65, "elapsed_time": "20:45:02", "remaining_time": "1:38:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3965, "total_steps": 4274, "loss": 0.0607, "learning_rate": 8.224833171800173e-07, "epoch": 1.8551877412562874, "percentage": 92.77, "elapsed_time": "20:46:34", "remaining_time": "1:37:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3970, "total_steps": 4274, "loss": 0.0625, "learning_rate": 7.962221495648708e-07, "epoch": 1.857527196163294, "percentage": 92.89, "elapsed_time": "20:48:07", "remaining_time": "1:35:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3975, "total_steps": 4274, "loss": 0.0629, "learning_rate": 7.703802758008943e-07, "epoch": 1.8598666510703006, "percentage": 93.0, "elapsed_time": "20:49:39", "remaining_time": "1:33:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3980, "total_steps": 4274, "loss": 0.0592, "learning_rate": 7.44958143559471e-07, "epoch": 1.8622061059773074, "percentage": 93.12, "elapsed_time": "20:51:11", "remaining_time": "1:32:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3985, "total_steps": 4274, "loss": 0.0606, "learning_rate": 7.199561932405952e-07, "epoch": 1.864545560884314, "percentage": 93.24, "elapsed_time": "20:52:44", "remaining_time": "1:30:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3990, "total_steps": 4274, "loss": 0.0687, "learning_rate": 6.953748579652558e-07, "epoch": 1.8668850157913206, "percentage": 93.36, "elapsed_time": "20:54:16", "remaining_time": "1:29:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 3995, "total_steps": 4274, "loss": 0.0681, "learning_rate": 6.712145635679096e-07, "epoch": 1.8692244706983274, "percentage": 93.47, "elapsed_time": "20:55:48", "remaining_time": "1:27:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4000, "total_steps": 4274, "loss": 0.0599, "learning_rate": 6.474757285891253e-07, "epoch": 1.8715639256053338, "percentage": 93.59, "elapsed_time": "20:57:20", "remaining_time": "1:26:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4000, "total_steps": 4274, "eval_loss": 0.07669652253389359, "epoch": 1.8715639256053338, "percentage": 93.59, "elapsed_time": "21:01:14", "remaining_time": "1:26:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4005, "total_steps": 4274, "loss": 0.0624, "learning_rate": 6.241587642683206e-07, "epoch": 1.8739033805123406, "percentage": 93.71, "elapsed_time": "21:02:46", "remaining_time": "1:24:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4010, "total_steps": 4274, "loss": 0.0526, "learning_rate": 6.012640745366477e-07, "epoch": 1.8762428354193474, "percentage": 93.82, "elapsed_time": "21:04:18", "remaining_time": "1:23:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4015, "total_steps": 4274, "loss": 0.058, "learning_rate": 5.78792056009983e-07, "epoch": 1.8785822903263538, "percentage": 93.94, "elapsed_time": "21:05:50", "remaining_time": "1:21:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4020, "total_steps": 4274, "loss": 0.1072, "learning_rate": 5.567430979820654e-07, "epoch": 1.8809217452333606, "percentage": 94.06, "elapsed_time": "21:07:22", "remaining_time": "1:20:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4025, "total_steps": 4274, "loss": 0.0527, "learning_rate": 5.35117582417749e-07, "epoch": 1.8832612001403672, "percentage": 94.17, "elapsed_time": "21:08:53", "remaining_time": "1:18:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4030, "total_steps": 4274, "loss": 0.1046, "learning_rate": 5.139158839463948e-07, "epoch": 1.8856006550473738, "percentage": 94.29, "elapsed_time": "21:10:24", "remaining_time": "1:16:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4035, "total_steps": 4274, "loss": 0.0654, "learning_rate": 4.931383698553643e-07, "epoch": 1.8879401099543807, "percentage": 94.41, "elapsed_time": "21:11:55", "remaining_time": "1:15:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4040, "total_steps": 4274, "loss": 0.0818, "learning_rate": 4.727854000836723e-07, "epoch": 1.8902795648613873, "percentage": 94.53, "elapsed_time": "21:13:27", "remaining_time": "1:13:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4045, "total_steps": 4274, "loss": 0.0562, "learning_rate": 4.528573272157416e-07, "epoch": 1.8926190197683939, "percentage": 94.64, "elapsed_time": "21:14:59", "remaining_time": "1:12:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4050, "total_steps": 4274, "loss": 0.0563, "learning_rate": 4.333544964753022e-07, "epoch": 1.8949584746754007, "percentage": 94.76, "elapsed_time": "21:16:32", "remaining_time": "1:10:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4055, "total_steps": 4274, "loss": 0.0677, "learning_rate": 4.1427724571940217e-07, "epoch": 1.8972979295824073, "percentage": 94.88, "elapsed_time": "21:18:04", "remaining_time": "1:09:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4060, "total_steps": 4274, "loss": 0.0651, "learning_rate": 3.9562590543256175e-07, "epoch": 1.8996373844894139, "percentage": 94.99, "elapsed_time": "21:19:36", "remaining_time": "1:07:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4065, "total_steps": 4274, "loss": 0.0555, "learning_rate": 3.7740079872105317e-07, "epoch": 1.9019768393964207, "percentage": 95.11, "elapsed_time": "21:21:07", "remaining_time": "1:05:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4070, "total_steps": 4274, "loss": 0.0634, "learning_rate": 3.5960224130728857e-07, "epoch": 1.9043162943034273, "percentage": 95.23, "elapsed_time": "21:22:39", "remaining_time": "1:04:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4075, "total_steps": 4274, "loss": 0.0564, "learning_rate": 3.422305415243576e-07, "epoch": 1.906655749210434, "percentage": 95.34, "elapsed_time": "21:24:11", "remaining_time": "1:02:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4080, "total_steps": 4274, "loss": 0.0586, "learning_rate": 3.2528600031069557e-07, "epoch": 1.9089952041174407, "percentage": 95.46, "elapsed_time": "21:25:43", "remaining_time": "1:01:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4085, "total_steps": 4274, "loss": 0.0564, "learning_rate": 3.087689112048542e-07, "epoch": 1.9113346590244473, "percentage": 95.58, "elapsed_time": "21:27:16", "remaining_time": "0:59:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4090, "total_steps": 4274, "loss": 0.0609, "learning_rate": 2.926795603404198e-07, "epoch": 1.913674113931454, "percentage": 95.69, "elapsed_time": "21:28:48", "remaining_time": "0:57:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4095, "total_steps": 4274, "loss": 0.0557, "learning_rate": 2.7701822644107536e-07, "epoch": 1.9160135688384607, "percentage": 95.81, "elapsed_time": "21:30:20", "remaining_time": "0:56:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4100, "total_steps": 4274, "loss": 0.0565, "learning_rate": 2.6178518081574064e-07, "epoch": 1.9183530237454673, "percentage": 95.93, "elapsed_time": "21:31:51", "remaining_time": "0:54:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4105, "total_steps": 4274, "loss": 0.0609, "learning_rate": 2.46980687353901e-07, "epoch": 1.920692478652474, "percentage": 96.05, "elapsed_time": "21:33:24", "remaining_time": "0:53:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4110, "total_steps": 4274, "loss": 0.0618, "learning_rate": 2.3260500252101636e-07, "epoch": 1.9230319335594808, "percentage": 96.16, "elapsed_time": "21:34:56", "remaining_time": "0:51:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4115, "total_steps": 4274, "loss": 0.0516, "learning_rate": 2.1865837535409162e-07, "epoch": 1.9253713884664871, "percentage": 96.28, "elapsed_time": "21:36:28", "remaining_time": "0:50:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4120, "total_steps": 4274, "loss": 0.0646, "learning_rate": 2.0514104745734674e-07, "epoch": 1.927710843373494, "percentage": 96.4, "elapsed_time": "21:38:00", "remaining_time": "0:48:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4125, "total_steps": 4274, "loss": 0.072, "learning_rate": 1.9205325299805622e-07, "epoch": 1.9300502982805008, "percentage": 96.51, "elapsed_time": "21:39:32", "remaining_time": "0:46:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4130, "total_steps": 4274, "loss": 0.0584, "learning_rate": 1.7939521870245779e-07, "epoch": 1.9323897531875072, "percentage": 96.63, "elapsed_time": "21:41:04", "remaining_time": "0:45:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4135, "total_steps": 4274, "loss": 0.0576, "learning_rate": 1.6716716385186126e-07, "epoch": 1.934729208094514, "percentage": 96.75, "elapsed_time": "21:42:37", "remaining_time": "0:43:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4140, "total_steps": 4274, "loss": 0.1097, "learning_rate": 1.5536930027882368e-07, "epoch": 1.9370686630015208, "percentage": 96.86, "elapsed_time": "21:44:09", "remaining_time": "0:42:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4145, "total_steps": 4274, "loss": 0.0637, "learning_rate": 1.4400183236349397e-07, "epoch": 1.9394081179085272, "percentage": 96.98, "elapsed_time": "21:45:41", "remaining_time": "0:40:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4150, "total_steps": 4274, "loss": 0.0683, "learning_rate": 1.330649570300574e-07, "epoch": 1.941747572815534, "percentage": 97.1, "elapsed_time": "21:47:14", "remaining_time": "0:39:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4155, "total_steps": 4274, "loss": 0.0539, "learning_rate": 1.2255886374334946e-07, "epoch": 1.9440870277225406, "percentage": 97.22, "elapsed_time": "21:48:46", "remaining_time": "0:37:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4160, "total_steps": 4274, "loss": 0.0618, "learning_rate": 1.1248373450554462e-07, "epoch": 1.9464264826295472, "percentage": 97.33, "elapsed_time": "21:50:18", "remaining_time": "0:35:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4165, "total_steps": 4274, "loss": 0.06, "learning_rate": 1.0283974385301986e-07, "epoch": 1.948765937536554, "percentage": 97.45, "elapsed_time": "21:51:51", "remaining_time": "0:34:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4170, "total_steps": 4274, "loss": 0.0574, "learning_rate": 9.3627058853335e-08, "epoch": 1.9511053924435606, "percentage": 97.57, "elapsed_time": "21:53:23", "remaining_time": "0:32:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4175, "total_steps": 4274, "loss": 0.0595, "learning_rate": 8.484583910232379e-08, "epoch": 1.9534448473505672, "percentage": 97.68, "elapsed_time": "21:54:56", "remaining_time": "0:31:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4180, "total_steps": 4274, "loss": 0.0605, "learning_rate": 7.649623672134065e-08, "epoch": 1.955784302257574, "percentage": 97.8, "elapsed_time": "21:56:28", "remaining_time": "0:29:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4185, "total_steps": 4274, "loss": 0.0619, "learning_rate": 6.857839635462104e-08, "epoch": 1.9581237571645806, "percentage": 97.92, "elapsed_time": "21:58:00", "remaining_time": "0:28:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4190, "total_steps": 4274, "loss": 0.0623, "learning_rate": 6.109245516677242e-08, "epoch": 1.9604632120715872, "percentage": 98.03, "elapsed_time": "21:59:31", "remaining_time": "0:26:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4195, "total_steps": 4274, "loss": 0.0573, "learning_rate": 5.403854284040943e-08, "epoch": 1.962802666978594, "percentage": 98.15, "elapsed_time": "22:01:02", "remaining_time": "0:24:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4200, "total_steps": 4274, "loss": 0.0646, "learning_rate": 4.741678157389739e-08, "epoch": 1.9651421218856007, "percentage": 98.27, "elapsed_time": "22:02:34", "remaining_time": "0:23:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4205, "total_steps": 4274, "loss": 0.0602, "learning_rate": 4.122728607923731e-08, "epoch": 1.9674815767926073, "percentage": 98.39, "elapsed_time": "22:04:06", "remaining_time": "0:21:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4210, "total_steps": 4274, "loss": 0.0947, "learning_rate": 3.5470163580073046e-08, "epoch": 1.969821031699614, "percentage": 98.5, "elapsed_time": "22:05:38", "remaining_time": "0:20:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4215, "total_steps": 4274, "loss": 0.0566, "learning_rate": 3.014551380985386e-08, "epoch": 1.9721604866066207, "percentage": 98.62, "elapsed_time": "22:07:11", "remaining_time": "0:18:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4220, "total_steps": 4274, "loss": 0.0612, "learning_rate": 2.525342901008032e-08, "epoch": 1.9744999415136273, "percentage": 98.74, "elapsed_time": "22:08:43", "remaining_time": "0:17:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4225, "total_steps": 4274, "loss": 0.0652, "learning_rate": 2.0793993928724942e-08, "epoch": 1.976839396420634, "percentage": 98.85, "elapsed_time": "22:10:15", "remaining_time": "0:15:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4230, "total_steps": 4274, "loss": 0.0576, "learning_rate": 1.676728581876119e-08, "epoch": 1.9791788513276407, "percentage": 98.97, "elapsed_time": "22:11:48", "remaining_time": "0:13:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4235, "total_steps": 4274, "loss": 0.0639, "learning_rate": 1.3173374436811769e-08, "epoch": 1.9815183062346473, "percentage": 99.09, "elapsed_time": "22:13:20", "remaining_time": "0:12:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4240, "total_steps": 4274, "loss": 0.0503, "learning_rate": 1.0012322041960676e-08, "epoch": 1.9838577611416541, "percentage": 99.2, "elapsed_time": "22:14:51", "remaining_time": "0:10:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4245, "total_steps": 4274, "loss": 0.0557, "learning_rate": 7.284183394656863e-09, "epoch": 1.9861972160486605, "percentage": 99.32, "elapsed_time": "22:16:21", "remaining_time": "0:09:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4250, "total_steps": 4274, "loss": 0.0591, "learning_rate": 4.98900575578165e-09, "epoch": 1.9885366709556673, "percentage": 99.44, "elapsed_time": "22:17:54", "remaining_time": "0:07:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4255, "total_steps": 4274, "loss": 0.0607, "learning_rate": 3.126828885816058e-09, "epoch": 1.9908761258626742, "percentage": 99.56, "elapsed_time": "22:19:26", "remaining_time": "0:05:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4260, "total_steps": 4274, "loss": 0.059, "learning_rate": 1.6976850441552394e-09, "epoch": 1.9932155807696805, "percentage": 99.67, "elapsed_time": "22:20:58", "remaining_time": "0:04:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4265, "total_steps": 4274, "loss": 0.0575, "learning_rate": 7.015989885589269e-10, "epoch": 1.9955550356766873, "percentage": 99.79, "elapsed_time": "22:22:30", "remaining_time": "0:02:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4270, "total_steps": 4274, "loss": 0.0704, "learning_rate": 1.3858797471011376e-10, "epoch": 1.997894490583694, "percentage": 99.91, "elapsed_time": "22:24:02", "remaining_time": "0:01:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 4274, "total_steps": 4274, "epoch": 1.9997660545092995, "percentage": 100.0, "elapsed_time": "22:25:16", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |