| {"current_steps": 5, "total_steps": 480, "loss": 2.3536, "learning_rate": 5.208333333333334e-06, "epoch": 0.03125, "percentage": 1.04, "elapsed_time": "0:01:36", "remaining_time": "2:32:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 10, "total_steps": 480, "loss": 2.3694, "learning_rate": 1.0416666666666668e-05, "epoch": 0.0625, "percentage": 2.08, "elapsed_time": "0:03:06", "remaining_time": "2:25:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 15, "total_steps": 480, "loss": 2.3358, "learning_rate": 1.5625e-05, "epoch": 0.09375, "percentage": 3.12, "elapsed_time": "0:04:35", "remaining_time": "2:22:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 20, "total_steps": 480, "loss": 2.3675, "learning_rate": 2.0833333333333336e-05, "epoch": 0.125, "percentage": 4.17, "elapsed_time": "0:06:05", "remaining_time": "2:19:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 25, "total_steps": 480, "loss": 2.339, "learning_rate": 2.604166666666667e-05, "epoch": 0.15625, "percentage": 5.21, "elapsed_time": "0:07:34", "remaining_time": "2:17:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 30, "total_steps": 480, "loss": 2.3121, "learning_rate": 3.125e-05, "epoch": 0.1875, "percentage": 6.25, "elapsed_time": "0:09:04", "remaining_time": "2:16:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 35, "total_steps": 480, "loss": 2.3198, "learning_rate": 3.6458333333333336e-05, "epoch": 0.21875, "percentage": 7.29, "elapsed_time": "0:10:33", "remaining_time": "2:14:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 40, "total_steps": 480, "loss": 2.3085, "learning_rate": 4.166666666666667e-05, "epoch": 0.25, "percentage": 8.33, "elapsed_time": "0:12:02", "remaining_time": "2:12:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 45, "total_steps": 480, "loss": 2.3422, "learning_rate": 4.6875e-05, "epoch": 0.28125, "percentage": 9.38, "elapsed_time": "0:13:31", "remaining_time": "2:10:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 50, "total_steps": 480, "loss": 2.2933, "learning_rate": 4.976851851851852e-05, "epoch": 0.3125, "percentage": 10.42, "elapsed_time": "0:15:00", "remaining_time": "2:09:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 55, "total_steps": 480, "loss": 2.278, "learning_rate": 4.9189814814814815e-05, "epoch": 0.34375, "percentage": 11.46, "elapsed_time": "0:16:29", "remaining_time": "2:07:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 60, "total_steps": 480, "loss": 2.2502, "learning_rate": 4.8611111111111115e-05, "epoch": 0.375, "percentage": 12.5, "elapsed_time": "0:17:58", "remaining_time": "2:05:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 65, "total_steps": 480, "loss": 2.3312, "learning_rate": 4.803240740740741e-05, "epoch": 0.40625, "percentage": 13.54, "elapsed_time": "0:19:26", "remaining_time": "2:04:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 70, "total_steps": 480, "loss": 2.2619, "learning_rate": 4.745370370370371e-05, "epoch": 0.4375, "percentage": 14.58, "elapsed_time": "0:20:55", "remaining_time": "2:02:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 75, "total_steps": 480, "loss": 2.2527, "learning_rate": 4.6875e-05, "epoch": 0.46875, "percentage": 15.62, "elapsed_time": "0:22:24", "remaining_time": "2:01:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 80, "total_steps": 480, "loss": 2.3142, "learning_rate": 4.62962962962963e-05, "epoch": 0.5, "percentage": 16.67, "elapsed_time": "0:23:53", "remaining_time": "1:59:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 85, "total_steps": 480, "loss": 2.2887, "learning_rate": 4.5717592592592594e-05, "epoch": 0.53125, "percentage": 17.71, "elapsed_time": "0:25:22", "remaining_time": "1:57:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 90, "total_steps": 480, "loss": 2.2449, "learning_rate": 4.5138888888888894e-05, "epoch": 0.5625, "percentage": 18.75, "elapsed_time": "0:26:50", "remaining_time": "1:56:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 95, "total_steps": 480, "loss": 2.2605, "learning_rate": 4.456018518518519e-05, "epoch": 0.59375, "percentage": 19.79, "elapsed_time": "0:28:19", "remaining_time": "1:54:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 100, "total_steps": 480, "loss": 2.2888, "learning_rate": 4.3981481481481486e-05, "epoch": 0.625, "percentage": 20.83, "elapsed_time": "0:29:48", "remaining_time": "1:53:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 105, "total_steps": 480, "loss": 2.3326, "learning_rate": 4.340277777777778e-05, "epoch": 0.65625, "percentage": 21.88, "elapsed_time": "0:31:17", "remaining_time": "1:51:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 110, "total_steps": 480, "loss": 2.2427, "learning_rate": 4.282407407407408e-05, "epoch": 0.6875, "percentage": 22.92, "elapsed_time": "0:32:46", "remaining_time": "1:50:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 115, "total_steps": 480, "loss": 2.2691, "learning_rate": 4.224537037037037e-05, "epoch": 0.71875, "percentage": 23.96, "elapsed_time": "0:34:15", "remaining_time": "1:48:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 120, "total_steps": 480, "loss": 2.274, "learning_rate": 4.166666666666667e-05, "epoch": 0.75, "percentage": 25.0, "elapsed_time": "0:35:44", "remaining_time": "1:47:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 125, "total_steps": 480, "loss": 2.2769, "learning_rate": 4.1087962962962965e-05, "epoch": 0.78125, "percentage": 26.04, "elapsed_time": "0:37:12", "remaining_time": "1:45:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 130, "total_steps": 480, "loss": 2.1991, "learning_rate": 4.0509259259259265e-05, "epoch": 0.8125, "percentage": 27.08, "elapsed_time": "0:38:41", "remaining_time": "1:44:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 135, "total_steps": 480, "loss": 2.2433, "learning_rate": 3.993055555555556e-05, "epoch": 0.84375, "percentage": 28.12, "elapsed_time": "0:40:10", "remaining_time": "1:42:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 140, "total_steps": 480, "loss": 2.2614, "learning_rate": 3.935185185185186e-05, "epoch": 0.875, "percentage": 29.17, "elapsed_time": "0:41:39", "remaining_time": "1:41:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 145, "total_steps": 480, "loss": 2.2938, "learning_rate": 3.877314814814815e-05, "epoch": 0.90625, "percentage": 30.21, "elapsed_time": "0:43:07", "remaining_time": "1:39:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 150, "total_steps": 480, "loss": 2.2696, "learning_rate": 3.8194444444444444e-05, "epoch": 0.9375, "percentage": 31.25, "elapsed_time": "0:44:36", "remaining_time": "1:38:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 155, "total_steps": 480, "loss": 2.2715, "learning_rate": 3.7615740740740744e-05, "epoch": 0.96875, "percentage": 32.29, "elapsed_time": "0:46:05", "remaining_time": "1:36:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 160, "total_steps": 480, "loss": 2.2377, "learning_rate": 3.7037037037037037e-05, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "0:47:31", "remaining_time": "1:35:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 165, "total_steps": 480, "loss": 2.0091, "learning_rate": 3.6458333333333336e-05, "epoch": 1.03125, "percentage": 34.38, "elapsed_time": "0:49:00", "remaining_time": "1:33:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 170, "total_steps": 480, "loss": 1.9842, "learning_rate": 3.587962962962963e-05, "epoch": 1.0625, "percentage": 35.42, "elapsed_time": "0:50:29", "remaining_time": "1:32:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 175, "total_steps": 480, "loss": 1.9515, "learning_rate": 3.530092592592593e-05, "epoch": 1.09375, "percentage": 36.46, "elapsed_time": "0:51:57", "remaining_time": "1:30:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 180, "total_steps": 480, "loss": 1.9648, "learning_rate": 3.472222222222222e-05, "epoch": 1.125, "percentage": 37.5, "elapsed_time": "0:53:26", "remaining_time": "1:29:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 185, "total_steps": 480, "loss": 1.9811, "learning_rate": 3.414351851851852e-05, "epoch": 1.15625, "percentage": 38.54, "elapsed_time": "0:54:55", "remaining_time": "1:27:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 190, "total_steps": 480, "loss": 1.9638, "learning_rate": 3.3564814814814815e-05, "epoch": 1.1875, "percentage": 39.58, "elapsed_time": "0:56:24", "remaining_time": "1:26:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 195, "total_steps": 480, "loss": 1.9788, "learning_rate": 3.2986111111111115e-05, "epoch": 1.21875, "percentage": 40.62, "elapsed_time": "0:57:52", "remaining_time": "1:24:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 200, "total_steps": 480, "loss": 2.014, "learning_rate": 3.240740740740741e-05, "epoch": 1.25, "percentage": 41.67, "elapsed_time": "0:59:21", "remaining_time": "1:23:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 205, "total_steps": 480, "loss": 1.9551, "learning_rate": 3.182870370370371e-05, "epoch": 1.28125, "percentage": 42.71, "elapsed_time": "1:00:50", "remaining_time": "1:21:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 210, "total_steps": 480, "loss": 1.995, "learning_rate": 3.125e-05, "epoch": 1.3125, "percentage": 43.75, "elapsed_time": "1:02:18", "remaining_time": "1:20:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 215, "total_steps": 480, "loss": 1.9717, "learning_rate": 3.06712962962963e-05, "epoch": 1.34375, "percentage": 44.79, "elapsed_time": "1:03:47", "remaining_time": "1:18:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 220, "total_steps": 480, "loss": 2.0301, "learning_rate": 3.0092592592592593e-05, "epoch": 1.375, "percentage": 45.83, "elapsed_time": "1:05:19", "remaining_time": "1:17:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 225, "total_steps": 480, "loss": 1.9311, "learning_rate": 2.951388888888889e-05, "epoch": 1.40625, "percentage": 46.88, "elapsed_time": "1:06:48", "remaining_time": "1:15:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 230, "total_steps": 480, "loss": 2.0162, "learning_rate": 2.8935185185185186e-05, "epoch": 1.4375, "percentage": 47.92, "elapsed_time": "1:08:17", "remaining_time": "1:14:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 235, "total_steps": 480, "loss": 1.9227, "learning_rate": 2.8356481481481483e-05, "epoch": 1.46875, "percentage": 48.96, "elapsed_time": "1:09:46", "remaining_time": "1:12:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 240, "total_steps": 480, "loss": 1.9816, "learning_rate": 2.777777777777778e-05, "epoch": 1.5, "percentage": 50.0, "elapsed_time": "1:11:14", "remaining_time": "1:11:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 245, "total_steps": 480, "loss": 1.9684, "learning_rate": 2.7199074074074076e-05, "epoch": 1.53125, "percentage": 51.04, "elapsed_time": "1:12:43", "remaining_time": "1:09:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 250, "total_steps": 480, "loss": 1.9591, "learning_rate": 2.6620370370370372e-05, "epoch": 1.5625, "percentage": 52.08, "elapsed_time": "1:14:12", "remaining_time": "1:08:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 255, "total_steps": 480, "loss": 1.997, "learning_rate": 2.604166666666667e-05, "epoch": 1.59375, "percentage": 53.12, "elapsed_time": "1:15:41", "remaining_time": "1:06:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 260, "total_steps": 480, "loss": 1.9312, "learning_rate": 2.5462962962962965e-05, "epoch": 1.625, "percentage": 54.17, "elapsed_time": "1:17:09", "remaining_time": "1:05:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 265, "total_steps": 480, "loss": 1.9073, "learning_rate": 2.488425925925926e-05, "epoch": 1.65625, "percentage": 55.21, "elapsed_time": "1:18:38", "remaining_time": "1:03:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 270, "total_steps": 480, "loss": 1.9367, "learning_rate": 2.4305555555555558e-05, "epoch": 1.6875, "percentage": 56.25, "elapsed_time": "1:20:07", "remaining_time": "1:02:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 275, "total_steps": 480, "loss": 1.9933, "learning_rate": 2.3726851851851854e-05, "epoch": 1.71875, "percentage": 57.29, "elapsed_time": "1:21:36", "remaining_time": "1:00:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 280, "total_steps": 480, "loss": 1.9843, "learning_rate": 2.314814814814815e-05, "epoch": 1.75, "percentage": 58.33, "elapsed_time": "1:23:04", "remaining_time": "0:59:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 285, "total_steps": 480, "loss": 1.9439, "learning_rate": 2.2569444444444447e-05, "epoch": 1.78125, "percentage": 59.38, "elapsed_time": "1:24:33", "remaining_time": "0:57:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 290, "total_steps": 480, "loss": 1.9441, "learning_rate": 2.1990740740740743e-05, "epoch": 1.8125, "percentage": 60.42, "elapsed_time": "1:26:02", "remaining_time": "0:56:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 295, "total_steps": 480, "loss": 1.9521, "learning_rate": 2.141203703703704e-05, "epoch": 1.84375, "percentage": 61.46, "elapsed_time": "1:27:30", "remaining_time": "0:54:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 300, "total_steps": 480, "loss": 1.9585, "learning_rate": 2.0833333333333336e-05, "epoch": 1.875, "percentage": 62.5, "elapsed_time": "1:28:59", "remaining_time": "0:53:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 305, "total_steps": 480, "loss": 1.9464, "learning_rate": 2.0254629629629632e-05, "epoch": 1.90625, "percentage": 63.54, "elapsed_time": "1:30:28", "remaining_time": "0:51:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 310, "total_steps": 480, "loss": 1.8949, "learning_rate": 1.967592592592593e-05, "epoch": 1.9375, "percentage": 64.58, "elapsed_time": "1:31:56", "remaining_time": "0:50:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 315, "total_steps": 480, "loss": 1.9655, "learning_rate": 1.9097222222222222e-05, "epoch": 1.96875, "percentage": 65.62, "elapsed_time": "1:33:25", "remaining_time": "0:48:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 320, "total_steps": 480, "loss": 1.9846, "learning_rate": 1.8518518518518518e-05, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "1:34:52", "remaining_time": "0:47:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 325, "total_steps": 480, "loss": 1.7054, "learning_rate": 1.7939814814814815e-05, "epoch": 2.03125, "percentage": 67.71, "elapsed_time": "1:36:21", "remaining_time": "0:45:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 330, "total_steps": 480, "loss": 1.7147, "learning_rate": 1.736111111111111e-05, "epoch": 2.0625, "percentage": 68.75, "elapsed_time": "1:37:50", "remaining_time": "0:44:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 335, "total_steps": 480, "loss": 1.7225, "learning_rate": 1.6782407407407408e-05, "epoch": 2.09375, "percentage": 69.79, "elapsed_time": "1:39:19", "remaining_time": "0:42:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 340, "total_steps": 480, "loss": 1.7715, "learning_rate": 1.6203703703703704e-05, "epoch": 2.125, "percentage": 70.83, "elapsed_time": "1:40:48", "remaining_time": "0:41:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 345, "total_steps": 480, "loss": 1.6884, "learning_rate": 1.5625e-05, "epoch": 2.15625, "percentage": 71.88, "elapsed_time": "1:42:16", "remaining_time": "0:40:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 350, "total_steps": 480, "loss": 1.7029, "learning_rate": 1.5046296296296297e-05, "epoch": 2.1875, "percentage": 72.92, "elapsed_time": "1:43:45", "remaining_time": "0:38:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 355, "total_steps": 480, "loss": 1.6946, "learning_rate": 1.4467592592592593e-05, "epoch": 2.21875, "percentage": 73.96, "elapsed_time": "1:45:14", "remaining_time": "0:37:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 360, "total_steps": 480, "loss": 1.7329, "learning_rate": 1.388888888888889e-05, "epoch": 2.25, "percentage": 75.0, "elapsed_time": "1:46:42", "remaining_time": "0:35:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 365, "total_steps": 480, "loss": 1.7032, "learning_rate": 1.3310185185185186e-05, "epoch": 2.28125, "percentage": 76.04, "elapsed_time": "1:48:11", "remaining_time": "0:34:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 370, "total_steps": 480, "loss": 1.7113, "learning_rate": 1.2731481481481482e-05, "epoch": 2.3125, "percentage": 77.08, "elapsed_time": "1:49:40", "remaining_time": "0:32:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 375, "total_steps": 480, "loss": 1.6568, "learning_rate": 1.2152777777777779e-05, "epoch": 2.34375, "percentage": 78.12, "elapsed_time": "1:51:09", "remaining_time": "0:31:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 380, "total_steps": 480, "loss": 1.7525, "learning_rate": 1.1574074074074075e-05, "epoch": 2.375, "percentage": 79.17, "elapsed_time": "1:52:37", "remaining_time": "0:29:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 385, "total_steps": 480, "loss": 1.6914, "learning_rate": 1.0995370370370372e-05, "epoch": 2.40625, "percentage": 80.21, "elapsed_time": "1:54:06", "remaining_time": "0:28:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 390, "total_steps": 480, "loss": 1.6791, "learning_rate": 1.0416666666666668e-05, "epoch": 2.4375, "percentage": 81.25, "elapsed_time": "1:55:35", "remaining_time": "0:26:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 395, "total_steps": 480, "loss": 1.7092, "learning_rate": 9.837962962962964e-06, "epoch": 2.46875, "percentage": 82.29, "elapsed_time": "1:57:03", "remaining_time": "0:25:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 400, "total_steps": 480, "loss": 1.6663, "learning_rate": 9.259259259259259e-06, "epoch": 2.5, "percentage": 83.33, "elapsed_time": "1:58:32", "remaining_time": "0:23:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 405, "total_steps": 480, "loss": 1.6959, "learning_rate": 8.680555555555556e-06, "epoch": 2.53125, "percentage": 84.38, "elapsed_time": "2:00:01", "remaining_time": "0:22:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 410, "total_steps": 480, "loss": 1.6847, "learning_rate": 8.101851851851852e-06, "epoch": 2.5625, "percentage": 85.42, "elapsed_time": "2:01:29", "remaining_time": "0:20:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 415, "total_steps": 480, "loss": 1.6527, "learning_rate": 7.523148148148148e-06, "epoch": 2.59375, "percentage": 86.46, "elapsed_time": "2:02:58", "remaining_time": "0:19:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 420, "total_steps": 480, "loss": 1.6578, "learning_rate": 6.944444444444445e-06, "epoch": 2.625, "percentage": 87.5, "elapsed_time": "2:04:27", "remaining_time": "0:17:46", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 425, "total_steps": 480, "loss": 1.7115, "learning_rate": 6.365740740740741e-06, "epoch": 2.65625, "percentage": 88.54, "elapsed_time": "2:05:56", "remaining_time": "0:16:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 430, "total_steps": 480, "loss": 1.7117, "learning_rate": 5.787037037037038e-06, "epoch": 2.6875, "percentage": 89.58, "elapsed_time": "2:07:24", "remaining_time": "0:14:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 435, "total_steps": 480, "loss": 1.7297, "learning_rate": 5.208333333333334e-06, "epoch": 2.71875, "percentage": 90.62, "elapsed_time": "2:08:53", "remaining_time": "0:13:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 440, "total_steps": 480, "loss": 1.7166, "learning_rate": 4.6296296296296296e-06, "epoch": 2.75, "percentage": 91.67, "elapsed_time": "2:10:22", "remaining_time": "0:11:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 445, "total_steps": 480, "loss": 1.6857, "learning_rate": 4.050925925925926e-06, "epoch": 2.78125, "percentage": 92.71, "elapsed_time": "2:11:51", "remaining_time": "0:10:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 450, "total_steps": 480, "loss": 1.6657, "learning_rate": 3.4722222222222224e-06, "epoch": 2.8125, "percentage": 93.75, "elapsed_time": "2:13:20", "remaining_time": "0:08:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 455, "total_steps": 480, "loss": 1.708, "learning_rate": 2.893518518518519e-06, "epoch": 2.84375, "percentage": 94.79, "elapsed_time": "2:14:49", "remaining_time": "0:07:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 460, "total_steps": 480, "loss": 1.7472, "learning_rate": 2.3148148148148148e-06, "epoch": 2.875, "percentage": 95.83, "elapsed_time": "2:16:17", "remaining_time": "0:05:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 465, "total_steps": 480, "loss": 1.6997, "learning_rate": 1.7361111111111112e-06, "epoch": 2.90625, "percentage": 96.88, "elapsed_time": "2:17:46", "remaining_time": "0:04:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 470, "total_steps": 480, "loss": 1.7339, "learning_rate": 1.1574074074074074e-06, "epoch": 2.9375, "percentage": 97.92, "elapsed_time": "2:19:15", "remaining_time": "0:02:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 475, "total_steps": 480, "loss": 1.7217, "learning_rate": 5.787037037037037e-07, "epoch": 2.96875, "percentage": 98.96, "elapsed_time": "2:20:44", "remaining_time": "0:01:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 480, "total_steps": 480, "loss": 1.6548, "learning_rate": 0.0, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "2:22:10", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 480, "total_steps": 480, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "2:23:16", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |