hf-reset
Reset repository without checkpoints directories
5d89611
{"current_steps": 5, "total_steps": 3750, "loss": 0.704, "lr": 4.266666666666667e-07, "epoch": 0.008, "percentage": 0.13, "elapsed_time": "0:00:24", "remaining_time": "5:00:09"}
{"current_steps": 10, "total_steps": 3750, "loss": 0.7201, "lr": 9.600000000000001e-07, "epoch": 0.016, "percentage": 0.27, "elapsed_time": "0:00:36", "remaining_time": "3:49:51"}
{"current_steps": 15, "total_steps": 3750, "loss": 0.7231, "lr": 1.4933333333333336e-06, "epoch": 0.024, "percentage": 0.4, "elapsed_time": "0:00:50", "remaining_time": "3:28:03"}
{"current_steps": 20, "total_steps": 3750, "loss": 0.7279, "lr": 2.0266666666666666e-06, "epoch": 0.032, "percentage": 0.53, "elapsed_time": "0:01:02", "remaining_time": "3:13:49"}
{"current_steps": 25, "total_steps": 3750, "loss": 0.6921, "lr": 2.56e-06, "epoch": 0.04, "percentage": 0.67, "elapsed_time": "0:01:15", "remaining_time": "3:08:37"}
{"current_steps": 30, "total_steps": 3750, "loss": 0.7105, "lr": 3.093333333333334e-06, "epoch": 0.048, "percentage": 0.8, "elapsed_time": "0:01:29", "remaining_time": "3:04:09"}
{"current_steps": 35, "total_steps": 3750, "loss": 0.7183, "lr": 3.6266666666666674e-06, "epoch": 0.056, "percentage": 0.93, "elapsed_time": "0:01:41", "remaining_time": "3:00:03"}
{"current_steps": 40, "total_steps": 3750, "loss": 0.7025, "lr": 4.16e-06, "epoch": 0.064, "percentage": 1.07, "elapsed_time": "0:01:54", "remaining_time": "2:56:45"}
{"current_steps": 45, "total_steps": 3750, "loss": 0.6852, "lr": 4.693333333333334e-06, "epoch": 0.072, "percentage": 1.2, "elapsed_time": "0:02:06", "remaining_time": "2:54:05"}
{"current_steps": 50, "total_steps": 3750, "loss": 0.6848, "lr": 5.226666666666667e-06, "epoch": 0.08, "percentage": 1.33, "elapsed_time": "0:02:18", "remaining_time": "2:51:07"}
{"current_steps": 55, "total_steps": 3750, "loss": 0.6499, "lr": 5.76e-06, "epoch": 0.088, "percentage": 1.47, "elapsed_time": "0:02:31", "remaining_time": "2:49:34"}
{"current_steps": 60, "total_steps": 3750, "loss": 0.6849, "lr": 6.293333333333334e-06, "epoch": 0.096, "percentage": 1.6, "elapsed_time": "0:02:43", "remaining_time": "2:47:16"}
{"current_steps": 65, "total_steps": 3750, "loss": 0.6651, "lr": 6.826666666666667e-06, "epoch": 0.104, "percentage": 1.73, "elapsed_time": "0:02:55", "remaining_time": "2:46:14"}
{"current_steps": 70, "total_steps": 3750, "loss": 0.6895, "lr": 7.360000000000001e-06, "epoch": 0.112, "percentage": 1.87, "elapsed_time": "0:03:09", "remaining_time": "2:45:39"}
{"current_steps": 75, "total_steps": 3750, "loss": 0.6592, "lr": 7.893333333333335e-06, "epoch": 0.12, "percentage": 2.0, "elapsed_time": "0:03:22", "remaining_time": "2:45:22"}
{"current_steps": 80, "total_steps": 3750, "loss": 0.6613, "lr": 8.426666666666667e-06, "epoch": 0.128, "percentage": 2.13, "elapsed_time": "0:03:35", "remaining_time": "2:44:26"}
{"current_steps": 85, "total_steps": 3750, "loss": 0.6649, "lr": 8.96e-06, "epoch": 0.136, "percentage": 2.27, "elapsed_time": "0:03:47", "remaining_time": "2:43:13"}
{"current_steps": 90, "total_steps": 3750, "loss": 0.662, "lr": 9.493333333333334e-06, "epoch": 0.144, "percentage": 2.4, "elapsed_time": "0:03:59", "remaining_time": "2:42:13"}
{"current_steps": 95, "total_steps": 3750, "loss": 0.6341, "lr": 1.0026666666666667e-05, "epoch": 0.152, "percentage": 2.53, "elapsed_time": "0:04:11", "remaining_time": "2:41:19"}
{"current_steps": 100, "total_steps": 3750, "loss": 0.6417, "lr": 1.056e-05, "epoch": 0.16, "percentage": 2.67, "elapsed_time": "0:04:24", "remaining_time": "2:40:38"}
{"current_steps": 105, "total_steps": 3750, "loss": 0.6585, "lr": 1.1093333333333334e-05, "epoch": 0.168, "percentage": 2.8, "elapsed_time": "0:05:40", "remaining_time": "3:17:15"}
{"current_steps": 110, "total_steps": 3750, "loss": 0.6296, "lr": 1.1626666666666668e-05, "epoch": 0.176, "percentage": 2.93, "elapsed_time": "0:05:53", "remaining_time": "3:15:11"}
{"current_steps": 115, "total_steps": 3750, "loss": 0.6573, "lr": 1.216e-05, "epoch": 0.184, "percentage": 3.07, "elapsed_time": "0:06:06", "remaining_time": "3:12:56"}
{"current_steps": 120, "total_steps": 3750, "loss": 0.6386, "lr": 1.2693333333333336e-05, "epoch": 0.192, "percentage": 3.2, "elapsed_time": "0:06:18", "remaining_time": "3:11:02"}
{"current_steps": 125, "total_steps": 3750, "loss": 0.646, "lr": 1.3226666666666668e-05, "epoch": 0.2, "percentage": 3.33, "elapsed_time": "0:06:32", "remaining_time": "3:09:29"}
{"current_steps": 130, "total_steps": 3750, "loss": 0.6219, "lr": 1.376e-05, "epoch": 0.208, "percentage": 3.47, "elapsed_time": "0:06:44", "remaining_time": "3:07:48"}
{"current_steps": 135, "total_steps": 3750, "loss": 0.6543, "lr": 1.4293333333333334e-05, "epoch": 0.216, "percentage": 3.6, "elapsed_time": "0:06:57", "remaining_time": "3:06:06"}
{"current_steps": 140, "total_steps": 3750, "loss": 0.6111, "lr": 1.4826666666666666e-05, "epoch": 0.224, "percentage": 3.73, "elapsed_time": "0:07:08", "remaining_time": "3:04:07"}
{"current_steps": 145, "total_steps": 3750, "loss": 0.661, "lr": 1.5360000000000002e-05, "epoch": 0.232, "percentage": 3.87, "elapsed_time": "0:07:21", "remaining_time": "3:03:01"}
{"current_steps": 150, "total_steps": 3750, "loss": 0.6406, "lr": 1.5893333333333333e-05, "epoch": 0.24, "percentage": 4.0, "elapsed_time": "0:07:33", "remaining_time": "3:01:23"}
{"current_steps": 155, "total_steps": 3750, "loss": 0.62, "lr": 1.642666666666667e-05, "epoch": 0.248, "percentage": 4.13, "elapsed_time": "0:07:45", "remaining_time": "2:59:55"}
{"current_steps": 160, "total_steps": 3750, "loss": 0.6316, "lr": 1.696e-05, "epoch": 0.256, "percentage": 4.27, "elapsed_time": "0:07:58", "remaining_time": "2:58:50"}
{"current_steps": 165, "total_steps": 3750, "loss": 0.6345, "lr": 1.7493333333333334e-05, "epoch": 0.264, "percentage": 4.4, "elapsed_time": "0:08:10", "remaining_time": "2:57:30"}
{"current_steps": 170, "total_steps": 3750, "loss": 0.6264, "lr": 1.8026666666666668e-05, "epoch": 0.272, "percentage": 4.53, "elapsed_time": "0:08:22", "remaining_time": "2:56:25"}
{"current_steps": 175, "total_steps": 3750, "loss": 0.6292, "lr": 1.8560000000000002e-05, "epoch": 0.28, "percentage": 4.67, "elapsed_time": "0:08:34", "remaining_time": "2:55:16"}
{"current_steps": 180, "total_steps": 3750, "loss": 0.5864, "lr": 1.9093333333333336e-05, "epoch": 0.288, "percentage": 4.8, "elapsed_time": "0:08:47", "remaining_time": "2:54:15"}
{"current_steps": 185, "total_steps": 3750, "loss": 0.6134, "lr": 1.9626666666666666e-05, "epoch": 0.296, "percentage": 4.93, "elapsed_time": "0:08:58", "remaining_time": "2:53:03"}
{"current_steps": 190, "total_steps": 3750, "loss": 0.6279, "lr": 2.016e-05, "epoch": 0.304, "percentage": 5.07, "elapsed_time": "0:09:11", "remaining_time": "2:52:11"}
{"current_steps": 195, "total_steps": 3750, "loss": 0.5989, "lr": 2.0693333333333334e-05, "epoch": 0.312, "percentage": 5.2, "elapsed_time": "0:09:23", "remaining_time": "2:51:16"}
{"current_steps": 200, "total_steps": 3750, "loss": 0.641, "lr": 2.1226666666666668e-05, "epoch": 0.32, "percentage": 5.33, "elapsed_time": "0:09:36", "remaining_time": "2:50:28"}
{"current_steps": 205, "total_steps": 3750, "loss": 0.6116, "lr": 2.1760000000000002e-05, "epoch": 0.328, "percentage": 5.47, "elapsed_time": "0:10:47", "remaining_time": "3:06:41"}
{"current_steps": 210, "total_steps": 3750, "loss": 0.6064, "lr": 2.2293333333333336e-05, "epoch": 0.336, "percentage": 5.6, "elapsed_time": "0:10:59", "remaining_time": "3:05:24"}
{"current_steps": 215, "total_steps": 3750, "loss": 0.6308, "lr": 2.282666666666667e-05, "epoch": 0.344, "percentage": 5.73, "elapsed_time": "0:11:12", "remaining_time": "3:04:14"}
{"current_steps": 220, "total_steps": 3750, "loss": 0.6001, "lr": 2.336e-05, "epoch": 0.352, "percentage": 5.87, "elapsed_time": "0:11:24", "remaining_time": "3:02:56"}
{"current_steps": 225, "total_steps": 3750, "loss": 0.589, "lr": 2.3893333333333337e-05, "epoch": 0.36, "percentage": 6.0, "elapsed_time": "0:11:36", "remaining_time": "3:01:58"}
{"current_steps": 230, "total_steps": 3750, "loss": 0.6268, "lr": 2.442666666666667e-05, "epoch": 0.368, "percentage": 6.13, "elapsed_time": "0:11:49", "remaining_time": "3:00:52"}
{"current_steps": 235, "total_steps": 3750, "loss": 0.5895, "lr": 2.496e-05, "epoch": 0.376, "percentage": 6.27, "elapsed_time": "0:12:00", "remaining_time": "2:59:40"}
{"current_steps": 240, "total_steps": 3750, "loss": 0.5914, "lr": 2.5493333333333335e-05, "epoch": 0.384, "percentage": 6.4, "elapsed_time": "0:12:13", "remaining_time": "2:58:41"}
{"current_steps": 245, "total_steps": 3750, "loss": 0.6168, "lr": 2.6026666666666666e-05, "epoch": 0.392, "percentage": 6.53, "elapsed_time": "0:12:25", "remaining_time": "2:57:46"}
{"current_steps": 250, "total_steps": 3750, "loss": 0.6405, "lr": 2.6560000000000003e-05, "epoch": 0.4, "percentage": 6.67, "elapsed_time": "0:12:37", "remaining_time": "2:56:48"}
{"current_steps": 255, "total_steps": 3750, "loss": 0.6121, "lr": 2.7093333333333337e-05, "epoch": 0.408, "percentage": 6.8, "elapsed_time": "0:12:49", "remaining_time": "2:55:49"}
{"current_steps": 260, "total_steps": 3750, "loss": 0.6419, "lr": 2.7626666666666668e-05, "epoch": 0.416, "percentage": 6.93, "elapsed_time": "0:13:01", "remaining_time": "2:54:54"}
{"current_steps": 265, "total_steps": 3750, "loss": 0.6228, "lr": 2.816e-05, "epoch": 0.424, "percentage": 7.07, "elapsed_time": "0:13:14", "remaining_time": "2:54:12"}
{"current_steps": 270, "total_steps": 3750, "loss": 0.5771, "lr": 2.869333333333334e-05, "epoch": 0.432, "percentage": 7.2, "elapsed_time": "0:13:27", "remaining_time": "2:53:25"}
{"current_steps": 275, "total_steps": 3750, "loss": 0.6036, "lr": 2.922666666666667e-05, "epoch": 0.44, "percentage": 7.33, "elapsed_time": "0:13:39", "remaining_time": "2:52:40"}
{"current_steps": 280, "total_steps": 3750, "loss": 0.6103, "lr": 2.9760000000000003e-05, "epoch": 0.448, "percentage": 7.47, "elapsed_time": "0:13:51", "remaining_time": "2:51:47"}
{"current_steps": 285, "total_steps": 3750, "loss": 0.6051, "lr": 3.0293333333333334e-05, "epoch": 0.456, "percentage": 7.6, "elapsed_time": "0:14:03", "remaining_time": "2:51:00"}
{"current_steps": 290, "total_steps": 3750, "loss": 0.5792, "lr": 3.0826666666666674e-05, "epoch": 0.464, "percentage": 7.73, "elapsed_time": "0:14:15", "remaining_time": "2:50:06"}
{"current_steps": 295, "total_steps": 3750, "loss": 0.6019, "lr": 3.1360000000000005e-05, "epoch": 0.472, "percentage": 7.87, "elapsed_time": "0:14:27", "remaining_time": "2:49:17"}
{"current_steps": 300, "total_steps": 3750, "loss": 0.6129, "lr": 3.1893333333333335e-05, "epoch": 0.48, "percentage": 8.0, "elapsed_time": "0:14:39", "remaining_time": "2:48:36"}
{"current_steps": 305, "total_steps": 3750, "loss": 0.6034, "lr": 3.2426666666666666e-05, "epoch": 0.488, "percentage": 8.13, "elapsed_time": "0:15:57", "remaining_time": "3:00:15"}
{"current_steps": 310, "total_steps": 3750, "loss": 0.5785, "lr": 3.296e-05, "epoch": 0.496, "percentage": 8.27, "elapsed_time": "0:16:09", "remaining_time": "2:59:20"}
{"current_steps": 315, "total_steps": 3750, "loss": 0.5833, "lr": 3.349333333333334e-05, "epoch": 0.504, "percentage": 8.4, "elapsed_time": "0:16:21", "remaining_time": "2:58:27"}
{"current_steps": 320, "total_steps": 3750, "loss": 0.6113, "lr": 3.402666666666667e-05, "epoch": 0.512, "percentage": 8.53, "elapsed_time": "0:16:34", "remaining_time": "2:57:34"}
{"current_steps": 325, "total_steps": 3750, "loss": 0.5893, "lr": 3.456e-05, "epoch": 0.52, "percentage": 8.67, "elapsed_time": "0:16:46", "remaining_time": "2:56:49"}
{"current_steps": 330, "total_steps": 3750, "loss": 0.6266, "lr": 3.509333333333333e-05, "epoch": 0.528, "percentage": 8.8, "elapsed_time": "0:16:58", "remaining_time": "2:55:50"}
{"current_steps": 335, "total_steps": 3750, "loss": 0.5987, "lr": 3.562666666666667e-05, "epoch": 0.536, "percentage": 8.93, "elapsed_time": "0:17:10", "remaining_time": "2:55:03"}
{"current_steps": 340, "total_steps": 3750, "loss": 0.5821, "lr": 3.6160000000000006e-05, "epoch": 0.544, "percentage": 9.07, "elapsed_time": "0:17:22", "remaining_time": "2:54:17"}
{"current_steps": 345, "total_steps": 3750, "loss": 0.5946, "lr": 3.669333333333334e-05, "epoch": 0.552, "percentage": 9.2, "elapsed_time": "0:17:34", "remaining_time": "2:53:31"}
{"current_steps": 350, "total_steps": 3750, "loss": 0.6028, "lr": 3.722666666666667e-05, "epoch": 0.56, "percentage": 9.33, "elapsed_time": "0:17:46", "remaining_time": "2:52:38"}
{"current_steps": 355, "total_steps": 3750, "loss": 0.6086, "lr": 3.7760000000000004e-05, "epoch": 0.568, "percentage": 9.47, "elapsed_time": "0:17:57", "remaining_time": "2:51:48"}
{"current_steps": 360, "total_steps": 3750, "loss": 0.594, "lr": 3.8293333333333335e-05, "epoch": 0.576, "percentage": 9.6, "elapsed_time": "0:18:10", "remaining_time": "2:51:05"}
{"current_steps": 365, "total_steps": 3750, "loss": 0.6019, "lr": 3.882666666666667e-05, "epoch": 0.584, "percentage": 9.73, "elapsed_time": "0:18:23", "remaining_time": "2:50:33"}
{"current_steps": 370, "total_steps": 3750, "loss": 0.6106, "lr": 3.936e-05, "epoch": 0.592, "percentage": 9.87, "elapsed_time": "0:18:34", "remaining_time": "2:49:45"}
{"current_steps": 375, "total_steps": 3750, "loss": 0.5994, "lr": 3.989333333333333e-05, "epoch": 0.6, "percentage": 10.0, "elapsed_time": "0:18:47", "remaining_time": "2:49:03"}
{"current_steps": 380, "total_steps": 3750, "loss": 0.6135, "lr": 3.9999861365387784e-05, "epoch": 0.608, "percentage": 10.13, "elapsed_time": "0:18:58", "remaining_time": "2:48:14"}
{"current_steps": 385, "total_steps": 3750, "loss": 0.581, "lr": 3.9999298165569614e-05, "epoch": 0.616, "percentage": 10.27, "elapsed_time": "0:19:10", "remaining_time": "2:47:35"}
{"current_steps": 390, "total_steps": 3750, "loss": 0.578, "lr": 3.999830174807269e-05, "epoch": 0.624, "percentage": 10.4, "elapsed_time": "0:19:22", "remaining_time": "2:46:53"}
{"current_steps": 395, "total_steps": 3750, "loss": 0.5649, "lr": 3.9996872134481036e-05, "epoch": 0.632, "percentage": 10.53, "elapsed_time": "0:19:34", "remaining_time": "2:46:19"}
{"current_steps": 400, "total_steps": 3750, "loss": 0.5994, "lr": 3.999500935576245e-05, "epoch": 0.64, "percentage": 10.67, "elapsed_time": "0:19:46", "remaining_time": "2:45:38"}
{"current_steps": 405, "total_steps": 3750, "loss": 0.5986, "lr": 3.999271345226776e-05, "epoch": 0.648, "percentage": 10.8, "elapsed_time": "0:20:58", "remaining_time": "2:53:16"}
{"current_steps": 410, "total_steps": 3750, "loss": 0.6028, "lr": 3.9989984473730035e-05, "epoch": 0.656, "percentage": 10.93, "elapsed_time": "0:21:10", "remaining_time": "2:52:30"}
{"current_steps": 415, "total_steps": 3750, "loss": 0.5702, "lr": 3.998682247926343e-05, "epoch": 0.664, "percentage": 11.07, "elapsed_time": "0:21:23", "remaining_time": "2:51:52"}
{"current_steps": 420, "total_steps": 3750, "loss": 0.5845, "lr": 3.998322753736193e-05, "epoch": 0.672, "percentage": 11.2, "elapsed_time": "0:21:35", "remaining_time": "2:51:07"}
{"current_steps": 425, "total_steps": 3750, "loss": 0.5944, "lr": 3.99791997258979e-05, "epoch": 0.68, "percentage": 11.33, "elapsed_time": "0:21:46", "remaining_time": "2:50:18"}
{"current_steps": 430, "total_steps": 3750, "loss": 0.5797, "lr": 3.997473913212036e-05, "epoch": 0.688, "percentage": 11.47, "elapsed_time": "0:21:57", "remaining_time": "2:49:33"}
{"current_steps": 435, "total_steps": 3750, "loss": 0.594, "lr": 3.9969845852653087e-05, "epoch": 0.696, "percentage": 11.6, "elapsed_time": "0:22:09", "remaining_time": "2:48:50"}
{"current_steps": 440, "total_steps": 3750, "loss": 0.5528, "lr": 3.996451999349258e-05, "epoch": 0.704, "percentage": 11.73, "elapsed_time": "0:22:21", "remaining_time": "2:48:08"}
{"current_steps": 445, "total_steps": 3750, "loss": 0.5877, "lr": 3.995876167000569e-05, "epoch": 0.712, "percentage": 11.87, "elapsed_time": "0:22:33", "remaining_time": "2:47:30"}
{"current_steps": 450, "total_steps": 3750, "loss": 0.5711, "lr": 3.9952571006927186e-05, "epoch": 0.72, "percentage": 12.0, "elapsed_time": "0:22:45", "remaining_time": "2:46:50"}
{"current_steps": 455, "total_steps": 3750, "loss": 0.5872, "lr": 3.9945948138356995e-05, "epoch": 0.728, "percentage": 12.13, "elapsed_time": "0:22:57", "remaining_time": "2:46:15"}
{"current_steps": 460, "total_steps": 3750, "loss": 0.5864, "lr": 3.993889320775735e-05, "epoch": 0.736, "percentage": 12.27, "elapsed_time": "0:23:09", "remaining_time": "2:45:39"}
{"current_steps": 465, "total_steps": 3750, "loss": 0.5857, "lr": 3.9931406367949627e-05, "epoch": 0.744, "percentage": 12.4, "elapsed_time": "0:23:21", "remaining_time": "2:45:00"}
{"current_steps": 470, "total_steps": 3750, "loss": 0.5664, "lr": 3.9923487781111106e-05, "epoch": 0.752, "percentage": 12.53, "elapsed_time": "0:23:33", "remaining_time": "2:44:26"}
{"current_steps": 475, "total_steps": 3750, "loss": 0.5902, "lr": 3.9915137618771386e-05, "epoch": 0.76, "percentage": 12.67, "elapsed_time": "0:23:46", "remaining_time": "2:43:52"}
{"current_steps": 480, "total_steps": 3750, "loss": 0.571, "lr": 3.9906356061808713e-05, "epoch": 0.768, "percentage": 12.8, "elapsed_time": "0:23:58", "remaining_time": "2:43:17"}
{"current_steps": 485, "total_steps": 3750, "loss": 0.5887, "lr": 3.9897143300446055e-05, "epoch": 0.776, "percentage": 12.93, "elapsed_time": "0:24:10", "remaining_time": "2:42:43"}
{"current_steps": 490, "total_steps": 3750, "loss": 0.5958, "lr": 3.988749953424696e-05, "epoch": 0.784, "percentage": 13.07, "elapsed_time": "0:24:22", "remaining_time": "2:42:11"}
{"current_steps": 495, "total_steps": 3750, "loss": 0.5836, "lr": 3.9877424972111264e-05, "epoch": 0.792, "percentage": 13.2, "elapsed_time": "0:24:34", "remaining_time": "2:41:34"}
{"current_steps": 500, "total_steps": 3750, "loss": 0.5627, "lr": 3.9866919832270554e-05, "epoch": 0.8, "percentage": 13.33, "elapsed_time": "0:24:46", "remaining_time": "2:41:03"}
{"current_steps": 505, "total_steps": 3750, "loss": 0.5529, "lr": 3.9855984342283414e-05, "epoch": 0.808, "percentage": 13.47, "elapsed_time": "0:25:58", "remaining_time": "2:46:57"}
{"current_steps": 510, "total_steps": 3750, "loss": 0.596, "lr": 3.9844618739030545e-05, "epoch": 0.816, "percentage": 13.6, "elapsed_time": "0:26:10", "remaining_time": "2:46:18"}
{"current_steps": 515, "total_steps": 3750, "loss": 0.6055, "lr": 3.98328232687096e-05, "epoch": 0.824, "percentage": 13.73, "elapsed_time": "0:26:22", "remaining_time": "2:45:41"}
{"current_steps": 520, "total_steps": 3750, "loss": 0.5943, "lr": 3.982059818682986e-05, "epoch": 0.832, "percentage": 13.87, "elapsed_time": "0:26:35", "remaining_time": "2:45:08"}
{"current_steps": 525, "total_steps": 3750, "loss": 0.5474, "lr": 3.980794375820669e-05, "epoch": 0.84, "percentage": 14.0, "elapsed_time": "0:26:47", "remaining_time": "2:44:34"}
{"current_steps": 530, "total_steps": 3750, "loss": 0.5945, "lr": 3.9794860256955825e-05, "epoch": 0.848, "percentage": 14.13, "elapsed_time": "0:26:59", "remaining_time": "2:43:58"}
{"current_steps": 535, "total_steps": 3750, "loss": 0.611, "lr": 3.9781347966487415e-05, "epoch": 0.856, "percentage": 14.27, "elapsed_time": "0:27:10", "remaining_time": "2:43:19"}
{"current_steps": 540, "total_steps": 3750, "loss": 0.5775, "lr": 3.9767407179499875e-05, "epoch": 0.864, "percentage": 14.4, "elapsed_time": "0:27:23", "remaining_time": "2:42:50"}
{"current_steps": 545, "total_steps": 3750, "loss": 0.5458, "lr": 3.975303819797358e-05, "epoch": 0.872, "percentage": 14.53, "elapsed_time": "0:27:35", "remaining_time": "2:42:15"}
{"current_steps": 550, "total_steps": 3750, "loss": 0.559, "lr": 3.973824133316431e-05, "epoch": 0.88, "percentage": 14.67, "elapsed_time": "0:27:47", "remaining_time": "2:41:39"}
{"current_steps": 555, "total_steps": 3750, "loss": 0.5543, "lr": 3.972301690559645e-05, "epoch": 0.888, "percentage": 14.8, "elapsed_time": "0:27:58", "remaining_time": "2:41:04"}
{"current_steps": 560, "total_steps": 3750, "loss": 0.5807, "lr": 3.970736524505615e-05, "epoch": 0.896, "percentage": 14.93, "elapsed_time": "0:28:10", "remaining_time": "2:40:29"}
{"current_steps": 565, "total_steps": 3750, "loss": 0.5669, "lr": 3.969128669058411e-05, "epoch": 0.904, "percentage": 15.07, "elapsed_time": "0:28:22", "remaining_time": "2:39:55"}
{"current_steps": 570, "total_steps": 3750, "loss": 0.5709, "lr": 3.9674781590468256e-05, "epoch": 0.912, "percentage": 15.2, "elapsed_time": "0:28:33", "remaining_time": "2:39:21"}
{"current_steps": 575, "total_steps": 3750, "loss": 0.5577, "lr": 3.9657850302236184e-05, "epoch": 0.92, "percentage": 15.33, "elapsed_time": "0:28:45", "remaining_time": "2:38:49"}
{"current_steps": 580, "total_steps": 3750, "loss": 0.5669, "lr": 3.964049319264744e-05, "epoch": 0.928, "percentage": 15.47, "elapsed_time": "0:28:57", "remaining_time": "2:38:17"}
{"current_steps": 585, "total_steps": 3750, "loss": 0.5691, "lr": 3.962271063768555e-05, "epoch": 0.936, "percentage": 15.6, "elapsed_time": "0:29:10", "remaining_time": "2:37:48"}
{"current_steps": 590, "total_steps": 3750, "loss": 0.5635, "lr": 3.960450302254989e-05, "epoch": 0.944, "percentage": 15.73, "elapsed_time": "0:29:22", "remaining_time": "2:37:18"}
{"current_steps": 595, "total_steps": 3750, "loss": 0.555, "lr": 3.958587074164735e-05, "epoch": 0.952, "percentage": 15.87, "elapsed_time": "0:29:34", "remaining_time": "2:36:49"}
{"current_steps": 600, "total_steps": 3750, "loss": 0.5728, "lr": 3.956681419858376e-05, "epoch": 0.96, "percentage": 16.0, "elapsed_time": "0:29:46", "remaining_time": "2:36:16"}
{"current_steps": 605, "total_steps": 3750, "loss": 0.5537, "lr": 3.954733380615516e-05, "epoch": 0.968, "percentage": 16.13, "elapsed_time": "0:31:03", "remaining_time": "2:41:25"}
{"current_steps": 610, "total_steps": 3750, "loss": 0.5597, "lr": 3.95274299863389e-05, "epoch": 0.976, "percentage": 16.27, "elapsed_time": "0:31:15", "remaining_time": "2:40:53"}
{"current_steps": 615, "total_steps": 3750, "loss": 0.5888, "lr": 3.950710317028443e-05, "epoch": 0.984, "percentage": 16.4, "elapsed_time": "0:31:27", "remaining_time": "2:40:23"}
{"current_steps": 620, "total_steps": 3750, "loss": 0.5506, "lr": 3.9486353798303996e-05, "epoch": 0.992, "percentage": 16.53, "elapsed_time": "0:31:40", "remaining_time": "2:39:53"}
{"current_steps": 625, "total_steps": 3750, "loss": 0.5644, "lr": 3.946518231986313e-05, "epoch": 1.0, "percentage": 16.67, "elapsed_time": "0:31:51", "remaining_time": "2:39:19"}
{"current_steps": 630, "total_steps": 3750, "loss": 0.5348, "lr": 3.9443589193570847e-05, "epoch": 1.008, "percentage": 16.8, "elapsed_time": "0:32:03", "remaining_time": "2:38:46"}
{"current_steps": 635, "total_steps": 3750, "loss": 0.5286, "lr": 3.942157488716976e-05, "epoch": 1.016, "percentage": 16.93, "elapsed_time": "0:32:15", "remaining_time": "2:38:14"}
{"current_steps": 640, "total_steps": 3750, "loss": 0.5288, "lr": 3.939913987752595e-05, "epoch": 1.024, "percentage": 17.07, "elapsed_time": "0:32:27", "remaining_time": "2:37:44"}
{"current_steps": 645, "total_steps": 3750, "loss": 0.5258, "lr": 3.9376284650618605e-05, "epoch": 1.032, "percentage": 17.2, "elapsed_time": "0:32:40", "remaining_time": "2:37:16"}
{"current_steps": 650, "total_steps": 3750, "loss": 0.5432, "lr": 3.935300970152952e-05, "epoch": 1.04, "percentage": 17.33, "elapsed_time": "0:32:51", "remaining_time": "2:36:40"}
{"current_steps": 655, "total_steps": 3750, "loss": 0.541, "lr": 3.932931553443235e-05, "epoch": 1.048, "percentage": 17.47, "elapsed_time": "0:33:02", "remaining_time": "2:36:09"}
{"current_steps": 660, "total_steps": 3750, "loss": 0.5454, "lr": 3.930520266258173e-05, "epoch": 1.056, "percentage": 17.6, "elapsed_time": "0:33:14", "remaining_time": "2:35:36"}
{"current_steps": 665, "total_steps": 3750, "loss": 0.5446, "lr": 3.928067160830208e-05, "epoch": 1.064, "percentage": 17.73, "elapsed_time": "0:33:26", "remaining_time": "2:35:08"}
{"current_steps": 670, "total_steps": 3750, "loss": 0.5587, "lr": 3.925572290297638e-05, "epoch": 1.072, "percentage": 17.87, "elapsed_time": "0:33:38", "remaining_time": "2:34:39"}
{"current_steps": 675, "total_steps": 3750, "loss": 0.5189, "lr": 3.9230357087034606e-05, "epoch": 1.08, "percentage": 18.0, "elapsed_time": "0:33:50", "remaining_time": "2:34:11"}
{"current_steps": 680, "total_steps": 3750, "loss": 0.5376, "lr": 3.9204574709942036e-05, "epoch": 1.088, "percentage": 18.13, "elapsed_time": "0:34:02", "remaining_time": "2:33:41"}
{"current_steps": 685, "total_steps": 3750, "loss": 0.544, "lr": 3.917837633018734e-05, "epoch": 1.096, "percentage": 18.27, "elapsed_time": "0:34:13", "remaining_time": "2:33:09"}
{"current_steps": 690, "total_steps": 3750, "loss": 0.5373, "lr": 3.915176251527051e-05, "epoch": 1.104, "percentage": 18.4, "elapsed_time": "0:34:25", "remaining_time": "2:32:39"}
{"current_steps": 695, "total_steps": 3750, "loss": 0.5227, "lr": 3.912473384169051e-05, "epoch": 1.112, "percentage": 18.53, "elapsed_time": "0:34:37", "remaining_time": "2:32:12"}
{"current_steps": 700, "total_steps": 3750, "loss": 0.5205, "lr": 3.9097290894932866e-05, "epoch": 1.12, "percentage": 18.67, "elapsed_time": "0:34:49", "remaining_time": "2:31:45"}
{"current_steps": 705, "total_steps": 3750, "loss": 0.5128, "lr": 3.906943426945691e-05, "epoch": 1.1280000000000001, "percentage": 18.8, "elapsed_time": "0:35:59", "remaining_time": "2:35:28"}
{"current_steps": 710, "total_steps": 3750, "loss": 0.5301, "lr": 3.9041164568682955e-05, "epoch": 1.1360000000000001, "percentage": 18.93, "elapsed_time": "0:36:11", "remaining_time": "2:34:58"}
{"current_steps": 715, "total_steps": 3750, "loss": 0.5582, "lr": 3.90124824049792e-05, "epoch": 1.144, "percentage": 19.07, "elapsed_time": "0:36:23", "remaining_time": "2:34:27"}
{"current_steps": 720, "total_steps": 3750, "loss": 0.518, "lr": 3.8983388399648465e-05, "epoch": 1.152, "percentage": 19.2, "elapsed_time": "0:36:35", "remaining_time": "2:33:57"}
{"current_steps": 725, "total_steps": 3750, "loss": 0.5016, "lr": 3.895388318291474e-05, "epoch": 1.16, "percentage": 19.33, "elapsed_time": "0:36:47", "remaining_time": "2:33:30"}
{"current_steps": 730, "total_steps": 3750, "loss": 0.5194, "lr": 3.892396739390952e-05, "epoch": 1.168, "percentage": 19.47, "elapsed_time": "0:37:00", "remaining_time": "2:33:05"}
{"current_steps": 735, "total_steps": 3750, "loss": 0.5392, "lr": 3.8893641680657986e-05, "epoch": 1.176, "percentage": 19.6, "elapsed_time": "0:37:12", "remaining_time": "2:32:36"}
{"current_steps": 740, "total_steps": 3750, "loss": 0.515, "lr": 3.886290670006495e-05, "epoch": 1.184, "percentage": 19.73, "elapsed_time": "0:37:24", "remaining_time": "2:32:08"}
{"current_steps": 745, "total_steps": 3750, "loss": 0.5526, "lr": 3.8831763117900605e-05, "epoch": 1.192, "percentage": 19.87, "elapsed_time": "0:37:35", "remaining_time": "2:31:37"}
{"current_steps": 750, "total_steps": 3750, "loss": 0.52, "lr": 3.8800211608786166e-05, "epoch": 1.2, "percentage": 20.0, "elapsed_time": "0:37:47", "remaining_time": "2:31:11"}
{"current_steps": 755, "total_steps": 3750, "loss": 0.5148, "lr": 3.876825285617918e-05, "epoch": 1.208, "percentage": 20.13, "elapsed_time": "0:37:59", "remaining_time": "2:30:41"}
{"current_steps": 760, "total_steps": 3750, "loss": 0.5473, "lr": 3.873588755235876e-05, "epoch": 1.216, "percentage": 20.27, "elapsed_time": "0:38:11", "remaining_time": "2:30:16"}
{"current_steps": 765, "total_steps": 3750, "loss": 0.5225, "lr": 3.870311639841062e-05, "epoch": 1.224, "percentage": 20.4, "elapsed_time": "0:38:23", "remaining_time": "2:29:48"}
{"current_steps": 770, "total_steps": 3750, "loss": 0.5488, "lr": 3.866994010421182e-05, "epoch": 1.232, "percentage": 20.53, "elapsed_time": "0:38:35", "remaining_time": "2:29:22"}
{"current_steps": 775, "total_steps": 3750, "loss": 0.5377, "lr": 3.863635938841545e-05, "epoch": 1.24, "percentage": 20.67, "elapsed_time": "0:38:47", "remaining_time": "2:28:52"}
{"current_steps": 780, "total_steps": 3750, "loss": 0.5146, "lr": 3.8602374978435015e-05, "epoch": 1.248, "percentage": 20.8, "elapsed_time": "0:38:59", "remaining_time": "2:28:26"}
{"current_steps": 785, "total_steps": 3750, "loss": 0.5492, "lr": 3.8567987610428705e-05, "epoch": 1.256, "percentage": 20.93, "elapsed_time": "0:39:11", "remaining_time": "2:28:00"}
{"current_steps": 790, "total_steps": 3750, "loss": 0.5305, "lr": 3.853319802928345e-05, "epoch": 1.264, "percentage": 21.07, "elapsed_time": "0:39:23", "remaining_time": "2:27:34"}
{"current_steps": 795, "total_steps": 3750, "loss": 0.5529, "lr": 3.849800698859877e-05, "epoch": 1.272, "percentage": 21.2, "elapsed_time": "0:39:34", "remaining_time": "2:27:07"}
{"current_steps": 800, "total_steps": 3750, "loss": 0.5639, "lr": 3.846241525067047e-05, "epoch": 1.28, "percentage": 21.33, "elapsed_time": "0:39:47", "remaining_time": "2:26:42"}
{"current_steps": 805, "total_steps": 3750, "loss": 0.5431, "lr": 3.842642358647411e-05, "epoch": 1.288, "percentage": 21.47, "elapsed_time": "0:41:01", "remaining_time": "2:30:05"}
{"current_steps": 810, "total_steps": 3750, "loss": 0.5055, "lr": 3.839003277564831e-05, "epoch": 1.296, "percentage": 21.6, "elapsed_time": "0:41:13", "remaining_time": "2:29:36"}
{"current_steps": 815, "total_steps": 3750, "loss": 0.5471, "lr": 3.835324360647785e-05, "epoch": 1.304, "percentage": 21.73, "elapsed_time": "0:41:25", "remaining_time": "2:29:09"}
{"current_steps": 820, "total_steps": 3750, "loss": 0.5337, "lr": 3.831605687587663e-05, "epoch": 1.312, "percentage": 21.87, "elapsed_time": "0:41:36", "remaining_time": "2:28:41"}
{"current_steps": 825, "total_steps": 3750, "loss": 0.5486, "lr": 3.827847338937037e-05, "epoch": 1.32, "percentage": 22.0, "elapsed_time": "0:41:49", "remaining_time": "2:28:15"}
{"current_steps": 830, "total_steps": 3750, "loss": 0.5417, "lr": 3.824049396107918e-05, "epoch": 1.328, "percentage": 22.13, "elapsed_time": "0:42:00", "remaining_time": "2:27:48"}
{"current_steps": 835, "total_steps": 3750, "loss": 0.5434, "lr": 3.8202119413699914e-05, "epoch": 1.336, "percentage": 22.27, "elapsed_time": "0:42:13", "remaining_time": "2:27:25"}
{"current_steps": 840, "total_steps": 3750, "loss": 0.5252, "lr": 3.8163350578488366e-05, "epoch": 1.3439999999999999, "percentage": 22.4, "elapsed_time": "0:42:25", "remaining_time": "2:26:57"}
{"current_steps": 845, "total_steps": 3750, "loss": 0.5255, "lr": 3.812418829524124e-05, "epoch": 1.3519999999999999, "percentage": 22.53, "elapsed_time": "0:42:37", "remaining_time": "2:26:32"}
{"current_steps": 850, "total_steps": 3750, "loss": 0.5277, "lr": 3.8084633412277974e-05, "epoch": 1.3599999999999999, "percentage": 22.67, "elapsed_time": "0:42:50", "remaining_time": "2:26:08"}
{"current_steps": 855, "total_steps": 3750, "loss": 0.5455, "lr": 3.804468678642238e-05, "epoch": 1.3679999999999999, "percentage": 22.8, "elapsed_time": "0:43:01", "remaining_time": "2:25:41"}
{"current_steps": 860, "total_steps": 3750, "loss": 0.5099, "lr": 3.800434928298403e-05, "epoch": 1.376, "percentage": 22.93, "elapsed_time": "0:43:13", "remaining_time": "2:25:14"}
{"current_steps": 865, "total_steps": 3750, "loss": 0.5291, "lr": 3.796362177573957e-05, "epoch": 1.384, "percentage": 23.07, "elapsed_time": "0:43:25", "remaining_time": "2:24:49"}
{"current_steps": 870, "total_steps": 3750, "loss": 0.5276, "lr": 3.792250514691378e-05, "epoch": 1.392, "percentage": 23.2, "elapsed_time": "0:43:37", "remaining_time": "2:24:24"}
{"current_steps": 875, "total_steps": 3750, "loss": 0.5434, "lr": 3.788100028716043e-05, "epoch": 1.4, "percentage": 23.33, "elapsed_time": "0:43:48", "remaining_time": "2:23:57"}
{"current_steps": 880, "total_steps": 3750, "loss": 0.5263, "lr": 3.7839108095543016e-05, "epoch": 1.408, "percentage": 23.47, "elapsed_time": "0:44:00", "remaining_time": "2:23:32"}
{"current_steps": 885, "total_steps": 3750, "loss": 0.5275, "lr": 3.7796829479515295e-05, "epoch": 1.416, "percentage": 23.6, "elapsed_time": "0:44:12", "remaining_time": "2:23:05"}
{"current_steps": 890, "total_steps": 3750, "loss": 0.5507, "lr": 3.775416535490159e-05, "epoch": 1.424, "percentage": 23.73, "elapsed_time": "0:44:23", "remaining_time": "2:22:39"}
{"current_steps": 895, "total_steps": 3750, "loss": 0.5235, "lr": 3.7711116645876984e-05, "epoch": 1.432, "percentage": 23.87, "elapsed_time": "0:44:35", "remaining_time": "2:22:16"}
{"current_steps": 900, "total_steps": 3750, "loss": 0.5232, "lr": 3.7667684284947286e-05, "epoch": 1.44, "percentage": 24.0, "elapsed_time": "0:44:47", "remaining_time": "2:21:49"}
{"current_steps": 905, "total_steps": 3750, "loss": 0.5287, "lr": 3.762386921292885e-05, "epoch": 1.448, "percentage": 24.13, "elapsed_time": "0:45:31", "remaining_time": "2:23:05"}
{"current_steps": 910, "total_steps": 3750, "loss": 0.5258, "lr": 3.757967237892818e-05, "epoch": 1.456, "percentage": 24.27, "elapsed_time": "0:45:43", "remaining_time": "2:22:43"}
{"current_steps": 915, "total_steps": 3750, "loss": 0.5217, "lr": 3.7535094740321334e-05, "epoch": 1.464, "percentage": 24.4, "elapsed_time": "0:45:55", "remaining_time": "2:22:16"}
{"current_steps": 920, "total_steps": 3750, "loss": 0.5334, "lr": 3.749013726273328e-05, "epoch": 1.472, "percentage": 24.53, "elapsed_time": "0:46:07", "remaining_time": "2:21:51"}
{"current_steps": 925, "total_steps": 3750, "loss": 0.5069, "lr": 3.7444800920016875e-05, "epoch": 1.48, "percentage": 24.67, "elapsed_time": "0:46:19", "remaining_time": "2:21:28"}
{"current_steps": 930, "total_steps": 3750, "loss": 0.5378, "lr": 3.7399086694231864e-05, "epoch": 1.488, "percentage": 24.8, "elapsed_time": "0:46:30", "remaining_time": "2:21:01"}
{"current_steps": 935, "total_steps": 3750, "loss": 0.5254, "lr": 3.735299557562352e-05, "epoch": 1.496, "percentage": 24.93, "elapsed_time": "0:46:42", "remaining_time": "2:20:38"}
{"current_steps": 940, "total_steps": 3750, "loss": 0.5322, "lr": 3.7306528562601245e-05, "epoch": 1.504, "percentage": 25.07, "elapsed_time": "0:46:54", "remaining_time": "2:20:13"}
{"current_steps": 945, "total_steps": 3750, "loss": 0.5167, "lr": 3.7259686661716945e-05, "epoch": 1.512, "percentage": 25.2, "elapsed_time": "0:47:06", "remaining_time": "2:19:49"}
{"current_steps": 950, "total_steps": 3750, "loss": 0.5379, "lr": 3.7212470887643204e-05, "epoch": 1.52, "percentage": 25.33, "elapsed_time": "0:47:18", "remaining_time": "2:19:25"}
{"current_steps": 955, "total_steps": 3750, "loss": 0.5293, "lr": 3.7164882263151315e-05, "epoch": 1.528, "percentage": 25.47, "elapsed_time": "0:47:29", "remaining_time": "2:18:59"}
{"current_steps": 960, "total_steps": 3750, "loss": 0.549, "lr": 3.711692181908913e-05, "epoch": 1.536, "percentage": 25.6, "elapsed_time": "0:47:41", "remaining_time": "2:18:35"}
{"current_steps": 965, "total_steps": 3750, "loss": 0.5308, "lr": 3.706859059435871e-05, "epoch": 1.544, "percentage": 25.73, "elapsed_time": "0:47:53", "remaining_time": "2:18:13"}
{"current_steps": 970, "total_steps": 3750, "loss": 0.5316, "lr": 3.701988963589384e-05, "epoch": 1.552, "percentage": 25.87, "elapsed_time": "0:48:05", "remaining_time": "2:17:50"}
{"current_steps": 975, "total_steps": 3750, "loss": 0.5423, "lr": 3.697081999863736e-05, "epoch": 1.56, "percentage": 26.0, "elapsed_time": "0:48:17", "remaining_time": "2:17:26"}
{"current_steps": 980, "total_steps": 3750, "loss": 0.5376, "lr": 3.692138274551828e-05, "epoch": 1.568, "percentage": 26.13, "elapsed_time": "0:48:29", "remaining_time": "2:17:03"}
{"current_steps": 985, "total_steps": 3750, "loss": 0.529, "lr": 3.687157894742878e-05, "epoch": 1.576, "percentage": 26.27, "elapsed_time": "0:48:41", "remaining_time": "2:16:40"}
{"current_steps": 990, "total_steps": 3750, "loss": 0.5246, "lr": 3.682140968320101e-05, "epoch": 1.584, "percentage": 26.4, "elapsed_time": "0:48:52", "remaining_time": "2:16:14"}
{"current_steps": 995, "total_steps": 3750, "loss": 0.5529, "lr": 3.6770876039583725e-05, "epoch": 1.592, "percentage": 26.53, "elapsed_time": "0:49:04", "remaining_time": "2:15:52"}
{"current_steps": 1000, "total_steps": 3750, "loss": 0.5314, "lr": 3.671997911121871e-05, "epoch": 1.6, "percentage": 26.67, "elapsed_time": "0:49:16", "remaining_time": "2:15:30"}
{"current_steps": 1005, "total_steps": 3750, "loss": 0.5154, "lr": 3.6668720000617126e-05, "epoch": 1.608, "percentage": 26.8, "elapsed_time": "0:50:29", "remaining_time": "2:17:55"}
{"current_steps": 1010, "total_steps": 3750, "loss": 0.5425, "lr": 3.661709981813558e-05, "epoch": 1.616, "percentage": 26.93, "elapsed_time": "0:50:41", "remaining_time": "2:17:30"}
{"current_steps": 1015, "total_steps": 3750, "loss": 0.5444, "lr": 3.6565119681952086e-05, "epoch": 1.624, "percentage": 27.07, "elapsed_time": "0:50:53", "remaining_time": "2:17:07"}
{"current_steps": 1020, "total_steps": 3750, "loss": 0.531, "lr": 3.651278071804186e-05, "epoch": 1.6320000000000001, "percentage": 27.2, "elapsed_time": "0:51:05", "remaining_time": "2:16:45"}
{"current_steps": 1025, "total_steps": 3750, "loss": 0.5404, "lr": 3.646008406015291e-05, "epoch": 1.6400000000000001, "percentage": 27.33, "elapsed_time": "0:51:17", "remaining_time": "2:16:22"}
{"current_steps": 1030, "total_steps": 3750, "loss": 0.5297, "lr": 3.6407030849781475e-05, "epoch": 1.6480000000000001, "percentage": 27.47, "elapsed_time": "0:51:28", "remaining_time": "2:15:57"}
{"current_steps": 1035, "total_steps": 3750, "loss": 0.5129, "lr": 3.635362223614733e-05, "epoch": 1.6560000000000001, "percentage": 27.6, "elapsed_time": "0:51:40", "remaining_time": "2:15:33"}
{"current_steps": 1040, "total_steps": 3750, "loss": 0.5153, "lr": 3.629985937616884e-05, "epoch": 1.6640000000000001, "percentage": 27.73, "elapsed_time": "0:51:52", "remaining_time": "2:15:10"}
{"current_steps": 1045, "total_steps": 3750, "loss": 0.5295, "lr": 3.624574343443794e-05, "epoch": 1.6720000000000002, "percentage": 27.87, "elapsed_time": "0:52:04", "remaining_time": "2:14:46"}
{"current_steps": 1050, "total_steps": 3750, "loss": 0.5414, "lr": 3.619127558319492e-05, "epoch": 1.6800000000000002, "percentage": 28.0, "elapsed_time": "0:52:15", "remaining_time": "2:14:22"}
{"current_steps": 1055, "total_steps": 3750, "loss": 0.5298, "lr": 3.613645700230298e-05, "epoch": 1.688, "percentage": 28.13, "elapsed_time": "0:52:27", "remaining_time": "2:14:00"}
{"current_steps": 1060, "total_steps": 3750, "loss": 0.5319, "lr": 3.6081288879222696e-05, "epoch": 1.696, "percentage": 28.27, "elapsed_time": "0:52:40", "remaining_time": "2:13:41"}
{"current_steps": 1065, "total_steps": 3750, "loss": 0.5248, "lr": 3.602577240898633e-05, "epoch": 1.704, "percentage": 28.4, "elapsed_time": "0:52:52", "remaining_time": "2:13:19"}
{"current_steps": 1070, "total_steps": 3750, "loss": 0.54, "lr": 3.596990879417188e-05, "epoch": 1.712, "percentage": 28.53, "elapsed_time": "0:53:04", "remaining_time": "2:12:55"}
{"current_steps": 1075, "total_steps": 3750, "loss": 0.5172, "lr": 3.591369924487711e-05, "epoch": 1.72, "percentage": 28.67, "elapsed_time": "0:53:16", "remaining_time": "2:12:33"}
{"current_steps": 1080, "total_steps": 3750, "loss": 0.5289, "lr": 3.585714497869326e-05, "epoch": 1.728, "percentage": 28.8, "elapsed_time": "0:53:28", "remaining_time": "2:12:12"}
{"current_steps": 1085, "total_steps": 3750, "loss": 0.5174, "lr": 3.580024722067872e-05, "epoch": 1.736, "percentage": 28.93, "elapsed_time": "0:53:40", "remaining_time": "2:11:51"}
{"current_steps": 1090, "total_steps": 3750, "loss": 0.5303, "lr": 3.574300720333247e-05, "epoch": 1.744, "percentage": 29.07, "elapsed_time": "0:53:52", "remaining_time": "2:11:28"}
{"current_steps": 1095, "total_steps": 3750, "loss": 0.531, "lr": 3.568542616656739e-05, "epoch": 1.752, "percentage": 29.2, "elapsed_time": "0:54:04", "remaining_time": "2:11:06"}
{"current_steps": 1100, "total_steps": 3750, "loss": 0.5449, "lr": 3.5627505357683404e-05, "epoch": 1.76, "percentage": 29.33, "elapsed_time": "0:54:16", "remaining_time": "2:10:44"}
{"current_steps": 1105, "total_steps": 3750, "loss": 0.5432, "lr": 3.5569246031340474e-05, "epoch": 1.768, "percentage": 29.47, "elapsed_time": "0:55:30", "remaining_time": "2:12:52"}
{"current_steps": 1110, "total_steps": 3750, "loss": 0.5102, "lr": 3.5510649449531375e-05, "epoch": 1.776, "percentage": 29.6, "elapsed_time": "0:55:42", "remaining_time": "2:12:30"}
{"current_steps": 1115, "total_steps": 3750, "loss": 0.5264, "lr": 3.545171688155441e-05, "epoch": 1.784, "percentage": 29.73, "elapsed_time": "0:55:54", "remaining_time": "2:12:07"}
{"current_steps": 1120, "total_steps": 3750, "loss": 0.5244, "lr": 3.5392449603985894e-05, "epoch": 1.792, "percentage": 29.87, "elapsed_time": "0:56:06", "remaining_time": "2:11:44"}
{"current_steps": 1125, "total_steps": 3750, "loss": 0.519, "lr": 3.53328489006525e-05, "epoch": 1.8, "percentage": 30.0, "elapsed_time": "0:56:17", "remaining_time": "2:11:21"}
{"current_steps": 1130, "total_steps": 3750, "loss": 0.517, "lr": 3.527291606260345e-05, "epoch": 1.808, "percentage": 30.13, "elapsed_time": "0:56:29", "remaining_time": "2:10:58"}
{"current_steps": 1135, "total_steps": 3750, "loss": 0.5349, "lr": 3.521265238808255e-05, "epoch": 1.8159999999999998, "percentage": 30.27, "elapsed_time": "0:56:41", "remaining_time": "2:10:36"}
{"current_steps": 1140, "total_steps": 3750, "loss": 0.5199, "lr": 3.515205918250007e-05, "epoch": 1.8239999999999998, "percentage": 30.4, "elapsed_time": "0:56:52", "remaining_time": "2:10:13"}
{"current_steps": 1145, "total_steps": 3750, "loss": 0.5449, "lr": 3.5091137758404456e-05, "epoch": 1.8319999999999999, "percentage": 30.53, "elapsed_time": "0:57:04", "remaining_time": "2:09:51"}
{"current_steps": 1150, "total_steps": 3750, "loss": 0.5134, "lr": 3.5029889435453924e-05, "epoch": 1.8399999999999999, "percentage": 30.67, "elapsed_time": "0:57:16", "remaining_time": "2:09:29"}
{"current_steps": 1155, "total_steps": 3750, "loss": 0.5158, "lr": 3.496831554038784e-05, "epoch": 1.8479999999999999, "percentage": 30.8, "elapsed_time": "0:57:28", "remaining_time": "2:09:08"}
{"current_steps": 1160, "total_steps": 3750, "loss": 0.5183, "lr": 3.490641740699801e-05, "epoch": 1.8559999999999999, "percentage": 30.93, "elapsed_time": "0:57:40", "remaining_time": "2:08:47"}
{"current_steps": 1165, "total_steps": 3750, "loss": 0.5232, "lr": 3.484419637609977e-05, "epoch": 1.8639999999999999, "percentage": 31.07, "elapsed_time": "0:57:52", "remaining_time": "2:08:25"}
{"current_steps": 1170, "total_steps": 3750, "loss": 0.5153, "lr": 3.478165379550292e-05, "epoch": 1.8719999999999999, "percentage": 31.2, "elapsed_time": "0:58:04", "remaining_time": "2:08:04"}
{"current_steps": 1175, "total_steps": 3750, "loss": 0.5516, "lr": 3.471879101998262e-05, "epoch": 1.88, "percentage": 31.33, "elapsed_time": "0:58:16", "remaining_time": "2:07:42"}
{"current_steps": 1180, "total_steps": 3750, "loss": 0.5333, "lr": 3.465560941124992e-05, "epoch": 1.888, "percentage": 31.47, "elapsed_time": "0:58:27", "remaining_time": "2:07:20"}
{"current_steps": 1185, "total_steps": 3750, "loss": 0.5241, "lr": 3.459211033792233e-05, "epoch": 1.896, "percentage": 31.6, "elapsed_time": "0:58:39", "remaining_time": "2:06:57"}
{"current_steps": 1190, "total_steps": 3750, "loss": 0.5196, "lr": 3.4528295175494194e-05, "epoch": 1.904, "percentage": 31.73, "elapsed_time": "0:58:50", "remaining_time": "2:06:35"}
{"current_steps": 1195, "total_steps": 3750, "loss": 0.5284, "lr": 3.4464165306306845e-05, "epoch": 1.912, "percentage": 31.87, "elapsed_time": "0:59:03", "remaining_time": "2:06:15"}
{"current_steps": 1200, "total_steps": 3750, "loss": 0.5203, "lr": 3.4399722119518675e-05, "epoch": 1.92, "percentage": 32.0, "elapsed_time": "0:59:14", "remaining_time": "2:05:53"}
{"current_steps": 1205, "total_steps": 3750, "loss": 0.5022, "lr": 3.433496701107506e-05, "epoch": 1.928, "percentage": 32.13, "elapsed_time": "1:00:28", "remaining_time": "2:07:44"}
{"current_steps": 1210, "total_steps": 3750, "loss": 0.5159, "lr": 3.426990138367813e-05, "epoch": 1.936, "percentage": 32.27, "elapsed_time": "1:00:40", "remaining_time": "2:07:21"}
{"current_steps": 1215, "total_steps": 3750, "loss": 0.537, "lr": 3.420452664675633e-05, "epoch": 1.944, "percentage": 32.4, "elapsed_time": "1:00:51", "remaining_time": "2:06:58"}
{"current_steps": 1220, "total_steps": 3750, "loss": 0.5262, "lr": 3.4138844216433946e-05, "epoch": 1.952, "percentage": 32.53, "elapsed_time": "1:01:03", "remaining_time": "2:06:36"}
{"current_steps": 1225, "total_steps": 3750, "loss": 0.504, "lr": 3.407285551550041e-05, "epoch": 1.96, "percentage": 32.67, "elapsed_time": "1:01:15", "remaining_time": "2:06:16"}
{"current_steps": 1230, "total_steps": 3750, "loss": 0.533, "lr": 3.4006561973379466e-05, "epoch": 1.968, "percentage": 32.8, "elapsed_time": "1:01:27", "remaining_time": "2:05:55"}
{"current_steps": 1235, "total_steps": 3750, "loss": 0.5059, "lr": 3.3939965026098245e-05, "epoch": 1.976, "percentage": 32.93, "elapsed_time": "1:01:40", "remaining_time": "2:05:36"}
{"current_steps": 1240, "total_steps": 3750, "loss": 0.5043, "lr": 3.38730661162561e-05, "epoch": 1.984, "percentage": 33.07, "elapsed_time": "1:01:52", "remaining_time": "2:05:15"}
{"current_steps": 1245, "total_steps": 3750, "loss": 0.5099, "lr": 3.3805866692993414e-05, "epoch": 1.992, "percentage": 33.2, "elapsed_time": "1:02:05", "remaining_time": "2:04:56"}
{"current_steps": 1250, "total_steps": 3750, "loss": 0.5193, "lr": 3.373836821196018e-05, "epoch": 2.0, "percentage": 33.33, "elapsed_time": "1:02:17", "remaining_time": "2:04:34"}
{"current_steps": 1255, "total_steps": 3750, "loss": 0.5071, "lr": 3.3670572135284456e-05, "epoch": 2.008, "percentage": 33.47, "elapsed_time": "1:02:30", "remaining_time": "2:04:15"}
{"current_steps": 1260, "total_steps": 3750, "loss": 0.4741, "lr": 3.360247993154073e-05, "epoch": 2.016, "percentage": 33.6, "elapsed_time": "1:02:41", "remaining_time": "2:03:53"}
{"current_steps": 1265, "total_steps": 3750, "loss": 0.5018, "lr": 3.35340930757181e-05, "epoch": 2.024, "percentage": 33.73, "elapsed_time": "1:02:52", "remaining_time": "2:03:31"}
{"current_steps": 1270, "total_steps": 3750, "loss": 0.4722, "lr": 3.3465413049188276e-05, "epoch": 2.032, "percentage": 33.87, "elapsed_time": "1:03:05", "remaining_time": "2:03:11"}
{"current_steps": 1275, "total_steps": 3750, "loss": 0.4787, "lr": 3.3396441339673564e-05, "epoch": 2.04, "percentage": 34.0, "elapsed_time": "1:03:16", "remaining_time": "2:02:49"}
{"current_steps": 1280, "total_steps": 3750, "loss": 0.4792, "lr": 3.3327179441214574e-05, "epoch": 2.048, "percentage": 34.13, "elapsed_time": "1:03:27", "remaining_time": "2:02:27"}
{"current_steps": 1285, "total_steps": 3750, "loss": 0.4796, "lr": 3.325762885413791e-05, "epoch": 2.056, "percentage": 34.27, "elapsed_time": "1:03:39", "remaining_time": "2:02:06"}
{"current_steps": 1290, "total_steps": 3750, "loss": 0.4758, "lr": 3.318779108502362e-05, "epoch": 2.064, "percentage": 34.4, "elapsed_time": "1:03:51", "remaining_time": "2:01:45"}
{"current_steps": 1295, "total_steps": 3750, "loss": 0.4661, "lr": 3.3117667646672616e-05, "epoch": 2.072, "percentage": 34.53, "elapsed_time": "1:04:03", "remaining_time": "2:01:27"}
{"current_steps": 1300, "total_steps": 3750, "loss": 0.4497, "lr": 3.304726005807386e-05, "epoch": 2.08, "percentage": 34.67, "elapsed_time": "1:04:15", "remaining_time": "2:01:06"}
{"current_steps": 1305, "total_steps": 3750, "loss": 0.4882, "lr": 3.297656984437148e-05, "epoch": 2.088, "percentage": 34.8, "elapsed_time": "1:05:02", "remaining_time": "2:01:51"}
{"current_steps": 1310, "total_steps": 3750, "loss": 0.4864, "lr": 3.2905598536831715e-05, "epoch": 2.096, "percentage": 34.93, "elapsed_time": "1:05:14", "remaining_time": "2:01:30"}
{"current_steps": 1315, "total_steps": 3750, "loss": 0.4716, "lr": 3.2834347672809776e-05, "epoch": 2.104, "percentage": 35.07, "elapsed_time": "1:05:26", "remaining_time": "2:01:10"}
{"current_steps": 1320, "total_steps": 3750, "loss": 0.4644, "lr": 3.276281879571651e-05, "epoch": 2.112, "percentage": 35.2, "elapsed_time": "1:05:38", "remaining_time": "2:00:50"}
{"current_steps": 1325, "total_steps": 3750, "loss": 0.4894, "lr": 3.2691013454985006e-05, "epoch": 2.12, "percentage": 35.33, "elapsed_time": "1:05:50", "remaining_time": "2:00:29"}
{"current_steps": 1330, "total_steps": 3750, "loss": 0.4633, "lr": 3.2618933206036994e-05, "epoch": 2.128, "percentage": 35.47, "elapsed_time": "1:06:01", "remaining_time": "2:00:08"}
{"current_steps": 1335, "total_steps": 3750, "loss": 0.4743, "lr": 3.2546579610249177e-05, "epoch": 2.136, "percentage": 35.6, "elapsed_time": "1:06:13", "remaining_time": "1:59:47"}
{"current_steps": 1340, "total_steps": 3750, "loss": 0.4832, "lr": 3.2473954234919386e-05, "epoch": 2.144, "percentage": 35.73, "elapsed_time": "1:06:25", "remaining_time": "1:59:28"}
{"current_steps": 1345, "total_steps": 3750, "loss": 0.4816, "lr": 3.240105865323266e-05, "epoch": 2.152, "percentage": 35.87, "elapsed_time": "1:06:37", "remaining_time": "1:59:08"}
{"current_steps": 1350, "total_steps": 3750, "loss": 0.4814, "lr": 3.232789444422714e-05, "epoch": 2.16, "percentage": 36.0, "elapsed_time": "1:06:49", "remaining_time": "1:58:47"}
{"current_steps": 1355, "total_steps": 3750, "loss": 0.4612, "lr": 3.225446319275988e-05, "epoch": 2.168, "percentage": 36.13, "elapsed_time": "1:07:00", "remaining_time": "1:58:27"}
{"current_steps": 1360, "total_steps": 3750, "loss": 0.4596, "lr": 3.218076648947251e-05, "epoch": 2.176, "percentage": 36.27, "elapsed_time": "1:07:11", "remaining_time": "1:58:05"}
{"current_steps": 1365, "total_steps": 3750, "loss": 0.4719, "lr": 3.2106805930756804e-05, "epoch": 2.184, "percentage": 36.4, "elapsed_time": "1:07:24", "remaining_time": "1:57:46"}
{"current_steps": 1370, "total_steps": 3750, "loss": 0.466, "lr": 3.2032583118720045e-05, "epoch": 2.192, "percentage": 36.53, "elapsed_time": "1:07:35", "remaining_time": "1:57:25"}
{"current_steps": 1375, "total_steps": 3750, "loss": 0.4735, "lr": 3.195809966115038e-05, "epoch": 2.2, "percentage": 36.67, "elapsed_time": "1:07:47", "remaining_time": "1:57:04"}
{"current_steps": 1380, "total_steps": 3750, "loss": 0.4772, "lr": 3.188335717148195e-05, "epoch": 2.208, "percentage": 36.8, "elapsed_time": "1:07:59", "remaining_time": "1:56:46"}
{"current_steps": 1385, "total_steps": 3750, "loss": 0.466, "lr": 3.1808357268759964e-05, "epoch": 2.216, "percentage": 36.93, "elapsed_time": "1:08:11", "remaining_time": "1:56:26"}
{"current_steps": 1390, "total_steps": 3750, "loss": 0.4829, "lr": 3.173310157760563e-05, "epoch": 2.224, "percentage": 37.07, "elapsed_time": "1:08:23", "remaining_time": "1:56:06"}
{"current_steps": 1395, "total_steps": 3750, "loss": 0.4649, "lr": 3.165759172818093e-05, "epoch": 2.232, "percentage": 37.2, "elapsed_time": "1:08:34", "remaining_time": "1:55:46"}
{"current_steps": 1400, "total_steps": 3750, "loss": 0.473, "lr": 3.158182935615336e-05, "epoch": 2.24, "percentage": 37.33, "elapsed_time": "1:08:46", "remaining_time": "1:55:26"}
{"current_steps": 1405, "total_steps": 3750, "loss": 0.4748, "lr": 3.150581610266046e-05, "epoch": 2.248, "percentage": 37.47, "elapsed_time": "1:10:07", "remaining_time": "1:57:02"}
{"current_steps": 1410, "total_steps": 3750, "loss": 0.4536, "lr": 3.1429553614274256e-05, "epoch": 2.2560000000000002, "percentage": 37.6, "elapsed_time": "1:10:19", "remaining_time": "1:56:42"}
{"current_steps": 1415, "total_steps": 3750, "loss": 0.4718, "lr": 3.1353043542965636e-05, "epoch": 2.2640000000000002, "percentage": 37.73, "elapsed_time": "1:10:31", "remaining_time": "1:56:22"}
{"current_steps": 1420, "total_steps": 3750, "loss": 0.4651, "lr": 3.1276287546068536e-05, "epoch": 2.2720000000000002, "percentage": 37.87, "elapsed_time": "1:10:42", "remaining_time": "1:56:01"}
{"current_steps": 1425, "total_steps": 3750, "loss": 0.4517, "lr": 3.1199287286244047e-05, "epoch": 2.2800000000000002, "percentage": 38.0, "elapsed_time": "1:10:55", "remaining_time": "1:55:42"}
{"current_steps": 1430, "total_steps": 3750, "loss": 0.4751, "lr": 3.112204443144438e-05, "epoch": 2.288, "percentage": 38.13, "elapsed_time": "1:11:06", "remaining_time": "1:55:22"}
{"current_steps": 1435, "total_steps": 3750, "loss": 0.4624, "lr": 3.1044560654876775e-05, "epoch": 2.296, "percentage": 38.27, "elapsed_time": "1:11:19", "remaining_time": "1:55:03"}
{"current_steps": 1440, "total_steps": 3750, "loss": 0.4665, "lr": 3.0966837634967215e-05, "epoch": 2.304, "percentage": 38.4, "elapsed_time": "1:11:31", "remaining_time": "1:54:44"}
{"current_steps": 1445, "total_steps": 3750, "loss": 0.4895, "lr": 3.088887705532409e-05, "epoch": 2.312, "percentage": 38.53, "elapsed_time": "1:11:43", "remaining_time": "1:54:24"}
{"current_steps": 1450, "total_steps": 3750, "loss": 0.4725, "lr": 3.081068060470174e-05, "epoch": 2.32, "percentage": 38.67, "elapsed_time": "1:11:55", "remaining_time": "1:54:05"}
{"current_steps": 1455, "total_steps": 3750, "loss": 0.4604, "lr": 3.073224997696385e-05, "epoch": 2.328, "percentage": 38.8, "elapsed_time": "1:12:07", "remaining_time": "1:53:45"}
{"current_steps": 1460, "total_steps": 3750, "loss": 0.4486, "lr": 3.065358687104675e-05, "epoch": 2.336, "percentage": 38.93, "elapsed_time": "1:12:18", "remaining_time": "1:53:25"}
{"current_steps": 1465, "total_steps": 3750, "loss": 0.4844, "lr": 3.057469299092264e-05, "epoch": 2.344, "percentage": 39.07, "elapsed_time": "1:12:30", "remaining_time": "1:53:05"}
{"current_steps": 1470, "total_steps": 3750, "loss": 0.4423, "lr": 3.0495570045562686e-05, "epoch": 2.352, "percentage": 39.2, "elapsed_time": "1:12:42", "remaining_time": "1:52:46"}
{"current_steps": 1475, "total_steps": 3750, "loss": 0.47, "lr": 3.041621974889996e-05, "epoch": 2.36, "percentage": 39.33, "elapsed_time": "1:12:53", "remaining_time": "1:52:26"}
{"current_steps": 1480, "total_steps": 3750, "loss": 0.4645, "lr": 3.0336643819792342e-05, "epoch": 2.368, "percentage": 39.47, "elapsed_time": "1:13:05", "remaining_time": "1:52:05"}
{"current_steps": 1485, "total_steps": 3750, "loss": 0.485, "lr": 3.0256843981985295e-05, "epoch": 2.376, "percentage": 39.6, "elapsed_time": "1:13:17", "remaining_time": "1:51:47"}
{"current_steps": 1490, "total_steps": 3750, "loss": 0.4919, "lr": 3.0176821964074503e-05, "epoch": 2.384, "percentage": 39.73, "elapsed_time": "1:13:29", "remaining_time": "1:51:28"}
{"current_steps": 1495, "total_steps": 3750, "loss": 0.477, "lr": 3.009657949946844e-05, "epoch": 2.392, "percentage": 39.87, "elapsed_time": "1:13:41", "remaining_time": "1:51:09"}
{"current_steps": 1500, "total_steps": 3750, "loss": 0.4914, "lr": 3.00161183263508e-05, "epoch": 2.4, "percentage": 40.0, "elapsed_time": "1:13:53", "remaining_time": "1:50:49"}
{"current_steps": 1505, "total_steps": 3750, "loss": 0.4599, "lr": 2.993544018764289e-05, "epoch": 2.408, "percentage": 40.13, "elapsed_time": "1:15:07", "remaining_time": "1:52:03"}
{"current_steps": 1510, "total_steps": 3750, "loss": 0.4859, "lr": 2.9854546830965833e-05, "epoch": 2.416, "percentage": 40.27, "elapsed_time": "1:15:19", "remaining_time": "1:51:44"}
{"current_steps": 1515, "total_steps": 3750, "loss": 0.4789, "lr": 2.9773440008602736e-05, "epoch": 2.424, "percentage": 40.4, "elapsed_time": "1:15:31", "remaining_time": "1:51:25"}
{"current_steps": 1520, "total_steps": 3750, "loss": 0.4723, "lr": 2.96921214774607e-05, "epoch": 2.432, "percentage": 40.53, "elapsed_time": "1:15:43", "remaining_time": "1:51:05"}
{"current_steps": 1525, "total_steps": 3750, "loss": 0.4724, "lr": 2.9610592999032815e-05, "epoch": 2.44, "percentage": 40.67, "elapsed_time": "1:15:55", "remaining_time": "1:50:46"}
{"current_steps": 1530, "total_steps": 3750, "loss": 0.4863, "lr": 2.9528856339359973e-05, "epoch": 2.448, "percentage": 40.8, "elapsed_time": "1:16:07", "remaining_time": "1:50:27"}
{"current_steps": 1535, "total_steps": 3750, "loss": 0.4675, "lr": 2.9446913268992588e-05, "epoch": 2.456, "percentage": 40.93, "elapsed_time": "1:16:18", "remaining_time": "1:50:07"}
{"current_steps": 1540, "total_steps": 3750, "loss": 0.474, "lr": 2.936476556295229e-05, "epoch": 2.464, "percentage": 41.07, "elapsed_time": "1:16:30", "remaining_time": "1:49:48"}
{"current_steps": 1545, "total_steps": 3750, "loss": 0.4866, "lr": 2.928241500069346e-05, "epoch": 2.472, "percentage": 41.2, "elapsed_time": "1:16:42", "remaining_time": "1:49:28"}
{"current_steps": 1550, "total_steps": 3750, "loss": 0.4745, "lr": 2.9199863366064655e-05, "epoch": 2.48, "percentage": 41.33, "elapsed_time": "1:16:54", "remaining_time": "1:49:10"}
{"current_steps": 1555, "total_steps": 3750, "loss": 0.4921, "lr": 2.9117112447270007e-05, "epoch": 2.488, "percentage": 41.47, "elapsed_time": "1:17:06", "remaining_time": "1:48:50"}
{"current_steps": 1560, "total_steps": 3750, "loss": 0.4813, "lr": 2.9034164036830462e-05, "epoch": 2.496, "percentage": 41.6, "elapsed_time": "1:17:18", "remaining_time": "1:48:31"}
{"current_steps": 1565, "total_steps": 3750, "loss": 0.4923, "lr": 2.8951019931544975e-05, "epoch": 2.504, "percentage": 41.73, "elapsed_time": "1:17:30", "remaining_time": "1:48:12"}
{"current_steps": 1570, "total_steps": 3750, "loss": 0.4612, "lr": 2.8867681932451544e-05, "epoch": 2.512, "percentage": 41.87, "elapsed_time": "1:17:42", "remaining_time": "1:47:54"}
{"current_steps": 1575, "total_steps": 3750, "loss": 0.4875, "lr": 2.8784151844788267e-05, "epoch": 2.52, "percentage": 42.0, "elapsed_time": "1:17:54", "remaining_time": "1:47:35"}
{"current_steps": 1580, "total_steps": 3750, "loss": 0.4657, "lr": 2.8700431477954155e-05, "epoch": 2.528, "percentage": 42.13, "elapsed_time": "1:18:06", "remaining_time": "1:47:16"}
{"current_steps": 1585, "total_steps": 3750, "loss": 0.4815, "lr": 2.8616522645470012e-05, "epoch": 2.536, "percentage": 42.27, "elapsed_time": "1:18:19", "remaining_time": "1:46:58"}
{"current_steps": 1590, "total_steps": 3750, "loss": 0.4572, "lr": 2.8532427164939086e-05, "epoch": 2.544, "percentage": 42.4, "elapsed_time": "1:18:31", "remaining_time": "1:46:39"}
{"current_steps": 1595, "total_steps": 3750, "loss": 0.4655, "lr": 2.844814685800776e-05, "epoch": 2.552, "percentage": 42.53, "elapsed_time": "1:18:42", "remaining_time": "1:46:21"}
{"current_steps": 1600, "total_steps": 3750, "loss": 0.4663, "lr": 2.8363683550326028e-05, "epoch": 2.56, "percentage": 42.67, "elapsed_time": "1:18:54", "remaining_time": "1:46:02"}
{"current_steps": 1605, "total_steps": 3750, "loss": 0.4714, "lr": 2.8279039071508024e-05, "epoch": 2.568, "percentage": 42.8, "elapsed_time": "1:20:07", "remaining_time": "1:47:04"}
{"current_steps": 1610, "total_steps": 3750, "loss": 0.4817, "lr": 2.81942152550923e-05, "epoch": 2.576, "percentage": 42.93, "elapsed_time": "1:20:19", "remaining_time": "1:46:45"}
{"current_steps": 1615, "total_steps": 3750, "loss": 0.4717, "lr": 2.810921393850219e-05, "epoch": 2.584, "percentage": 43.07, "elapsed_time": "1:20:31", "remaining_time": "1:46:26"}
{"current_steps": 1620, "total_steps": 3750, "loss": 0.486, "lr": 2.802403696300595e-05, "epoch": 2.592, "percentage": 43.2, "elapsed_time": "1:20:43", "remaining_time": "1:46:07"}
{"current_steps": 1625, "total_steps": 3750, "loss": 0.4758, "lr": 2.7938686173676915e-05, "epoch": 2.6, "percentage": 43.33, "elapsed_time": "1:20:55", "remaining_time": "1:45:49"}
{"current_steps": 1630, "total_steps": 3750, "loss": 0.4815, "lr": 2.7853163419353505e-05, "epoch": 2.608, "percentage": 43.47, "elapsed_time": "1:21:06", "remaining_time": "1:45:29"}
{"current_steps": 1635, "total_steps": 3750, "loss": 0.4885, "lr": 2.776747055259918e-05, "epoch": 2.616, "percentage": 43.6, "elapsed_time": "1:21:19", "remaining_time": "1:45:11"}
{"current_steps": 1640, "total_steps": 3750, "loss": 0.4749, "lr": 2.768160942966233e-05, "epoch": 2.624, "percentage": 43.73, "elapsed_time": "1:21:30", "remaining_time": "1:44:52"}
{"current_steps": 1645, "total_steps": 3750, "loss": 0.4977, "lr": 2.759558191043603e-05, "epoch": 2.632, "percentage": 43.87, "elapsed_time": "1:21:41", "remaining_time": "1:44:32"}
{"current_steps": 1650, "total_steps": 3750, "loss": 0.479, "lr": 2.7509389858417783e-05, "epoch": 2.64, "percentage": 44.0, "elapsed_time": "1:21:53", "remaining_time": "1:44:12"}
{"current_steps": 1655, "total_steps": 3750, "loss": 0.4603, "lr": 2.7423035140669147e-05, "epoch": 2.648, "percentage": 44.13, "elapsed_time": "1:22:04", "remaining_time": "1:43:53"}
{"current_steps": 1660, "total_steps": 3750, "loss": 0.4498, "lr": 2.7336519627775288e-05, "epoch": 2.656, "percentage": 44.27, "elapsed_time": "1:22:15", "remaining_time": "1:43:34"}
{"current_steps": 1665, "total_steps": 3750, "loss": 0.4847, "lr": 2.724984519380444e-05, "epoch": 2.664, "percentage": 44.4, "elapsed_time": "1:22:27", "remaining_time": "1:43:15"}
{"current_steps": 1670, "total_steps": 3750, "loss": 0.4609, "lr": 2.7163013716267353e-05, "epoch": 2.672, "percentage": 44.53, "elapsed_time": "1:22:39", "remaining_time": "1:42:57"}
{"current_steps": 1675, "total_steps": 3750, "loss": 0.4769, "lr": 2.707602707607659e-05, "epoch": 2.68, "percentage": 44.67, "elapsed_time": "1:22:51", "remaining_time": "1:42:38"}
{"current_steps": 1680, "total_steps": 3750, "loss": 0.4931, "lr": 2.6988887157505786e-05, "epoch": 2.6879999999999997, "percentage": 44.8, "elapsed_time": "1:23:02", "remaining_time": "1:42:19"}
{"current_steps": 1685, "total_steps": 3750, "loss": 0.4871, "lr": 2.6901595848148842e-05, "epoch": 2.6959999999999997, "percentage": 44.93, "elapsed_time": "1:23:14", "remaining_time": "1:42:01"}
{"current_steps": 1690, "total_steps": 3750, "loss": 0.4844, "lr": 2.681415503887904e-05, "epoch": 2.7039999999999997, "percentage": 45.07, "elapsed_time": "1:23:27", "remaining_time": "1:41:43"}
{"current_steps": 1695, "total_steps": 3750, "loss": 0.4862, "lr": 2.672656662380805e-05, "epoch": 2.7119999999999997, "percentage": 45.2, "elapsed_time": "1:23:39", "remaining_time": "1:41:25"}
{"current_steps": 1700, "total_steps": 3750, "loss": 0.4784, "lr": 2.6638832500244967e-05, "epoch": 2.7199999999999998, "percentage": 45.33, "elapsed_time": "1:23:51", "remaining_time": "1:41:07"}
{"current_steps": 1705, "total_steps": 3750, "loss": 0.4762, "lr": 2.655095456865514e-05, "epoch": 2.7279999999999998, "percentage": 45.47, "elapsed_time": "1:25:06", "remaining_time": "1:42:04"}
{"current_steps": 1710, "total_steps": 3750, "loss": 0.47, "lr": 2.6462934732619047e-05, "epoch": 2.7359999999999998, "percentage": 45.6, "elapsed_time": "1:25:18", "remaining_time": "1:41:46"}
{"current_steps": 1715, "total_steps": 3750, "loss": 0.4754, "lr": 2.6374774898791047e-05, "epoch": 2.7439999999999998, "percentage": 45.73, "elapsed_time": "1:25:29", "remaining_time": "1:41:27"}
{"current_steps": 1720, "total_steps": 3750, "loss": 0.4901, "lr": 2.6286476976858084e-05, "epoch": 2.752, "percentage": 45.87, "elapsed_time": "1:25:41", "remaining_time": "1:41:08"}
{"current_steps": 1725, "total_steps": 3750, "loss": 0.491, "lr": 2.619804287949831e-05, "epoch": 2.76, "percentage": 46.0, "elapsed_time": "1:25:52", "remaining_time": "1:40:49"}
{"current_steps": 1730, "total_steps": 3750, "loss": 0.4558, "lr": 2.6109474522339676e-05, "epoch": 2.768, "percentage": 46.13, "elapsed_time": "1:26:05", "remaining_time": "1:40:31"}
{"current_steps": 1735, "total_steps": 3750, "loss": 0.4892, "lr": 2.6020773823918414e-05, "epoch": 2.776, "percentage": 46.27, "elapsed_time": "1:26:17", "remaining_time": "1:40:12"}
{"current_steps": 1740, "total_steps": 3750, "loss": 0.4632, "lr": 2.5931942705637473e-05, "epoch": 2.784, "percentage": 46.4, "elapsed_time": "1:26:28", "remaining_time": "1:39:53"}
{"current_steps": 1745, "total_steps": 3750, "loss": 0.4562, "lr": 2.5842983091724923e-05, "epoch": 2.792, "percentage": 46.53, "elapsed_time": "1:26:40", "remaining_time": "1:39:35"}
{"current_steps": 1750, "total_steps": 3750, "loss": 0.4752, "lr": 2.575389690919226e-05, "epoch": 2.8, "percentage": 46.67, "elapsed_time": "1:26:51", "remaining_time": "1:39:15"}
{"current_steps": 1755, "total_steps": 3750, "loss": 0.4647, "lr": 2.5664686087792658e-05, "epoch": 2.808, "percentage": 46.8, "elapsed_time": "1:27:03", "remaining_time": "1:38:57"}
{"current_steps": 1760, "total_steps": 3750, "loss": 0.4813, "lr": 2.5575352559979188e-05, "epoch": 2.816, "percentage": 46.93, "elapsed_time": "1:27:15", "remaining_time": "1:38:39"}
{"current_steps": 1765, "total_steps": 3750, "loss": 0.4616, "lr": 2.5485898260862936e-05, "epoch": 2.824, "percentage": 47.07, "elapsed_time": "1:27:26", "remaining_time": "1:38:20"}
{"current_steps": 1770, "total_steps": 3750, "loss": 0.4752, "lr": 2.5396325128171072e-05, "epoch": 2.832, "percentage": 47.2, "elapsed_time": "1:27:38", "remaining_time": "1:38:02"}
{"current_steps": 1775, "total_steps": 3750, "loss": 0.4838, "lr": 2.5306635102204942e-05, "epoch": 2.84, "percentage": 47.33, "elapsed_time": "1:27:50", "remaining_time": "1:37:44"}
{"current_steps": 1780, "total_steps": 3750, "loss": 0.4474, "lr": 2.5216830125797943e-05, "epoch": 2.848, "percentage": 47.47, "elapsed_time": "1:28:02", "remaining_time": "1:37:26"}
{"current_steps": 1785, "total_steps": 3750, "loss": 0.4633, "lr": 2.5126912144273517e-05, "epoch": 2.856, "percentage": 47.6, "elapsed_time": "1:28:14", "remaining_time": "1:37:08"}
{"current_steps": 1790, "total_steps": 3750, "loss": 0.4821, "lr": 2.5036883105402985e-05, "epoch": 2.864, "percentage": 47.73, "elapsed_time": "1:28:26", "remaining_time": "1:36:50"}
{"current_steps": 1795, "total_steps": 3750, "loss": 0.4825, "lr": 2.4946744959363343e-05, "epoch": 2.872, "percentage": 47.87, "elapsed_time": "1:28:38", "remaining_time": "1:36:32"}
{"current_steps": 1800, "total_steps": 3750, "loss": 0.4818, "lr": 2.4856499658695018e-05, "epoch": 2.88, "percentage": 48.0, "elapsed_time": "1:28:50", "remaining_time": "1:36:14"}
{"current_steps": 1805, "total_steps": 3750, "loss": 0.4941, "lr": 2.4766149158259603e-05, "epoch": 2.888, "percentage": 48.13, "elapsed_time": "1:30:00", "remaining_time": "1:36:59"}
{"current_steps": 1810, "total_steps": 3750, "loss": 0.4665, "lr": 2.4675695415197476e-05, "epoch": 2.896, "percentage": 48.27, "elapsed_time": "1:30:11", "remaining_time": "1:36:40"}
{"current_steps": 1815, "total_steps": 3750, "loss": 0.4867, "lr": 2.458514038888543e-05, "epoch": 2.904, "percentage": 48.4, "elapsed_time": "1:30:22", "remaining_time": "1:36:21"}
{"current_steps": 1820, "total_steps": 3750, "loss": 0.4787, "lr": 2.4494486040894208e-05, "epoch": 2.912, "percentage": 48.53, "elapsed_time": "1:30:34", "remaining_time": "1:36:03"}
{"current_steps": 1825, "total_steps": 3750, "loss": 0.4904, "lr": 2.440373433494603e-05, "epoch": 2.92, "percentage": 48.67, "elapsed_time": "1:30:46", "remaining_time": "1:35:44"}
{"current_steps": 1830, "total_steps": 3750, "loss": 0.4928, "lr": 2.4312887236872066e-05, "epoch": 2.928, "percentage": 48.8, "elapsed_time": "1:30:57", "remaining_time": "1:35:26"}
{"current_steps": 1835, "total_steps": 3750, "loss": 0.5064, "lr": 2.4221946714569803e-05, "epoch": 2.936, "percentage": 48.93, "elapsed_time": "1:31:09", "remaining_time": "1:35:07"}
{"current_steps": 1840, "total_steps": 3750, "loss": 0.4664, "lr": 2.4130914737960472e-05, "epoch": 2.944, "percentage": 49.07, "elapsed_time": "1:31:20", "remaining_time": "1:34:49"}
{"current_steps": 1845, "total_steps": 3750, "loss": 0.4479, "lr": 2.4039793278946358e-05, "epoch": 2.952, "percentage": 49.2, "elapsed_time": "1:31:32", "remaining_time": "1:34:31"}
{"current_steps": 1850, "total_steps": 3750, "loss": 0.4816, "lr": 2.394858431136806e-05, "epoch": 2.96, "percentage": 49.33, "elapsed_time": "1:31:43", "remaining_time": "1:34:12"}
{"current_steps": 1855, "total_steps": 3750, "loss": 0.4802, "lr": 2.385728981096178e-05, "epoch": 2.968, "percentage": 49.47, "elapsed_time": "1:31:55", "remaining_time": "1:33:53"}
{"current_steps": 1860, "total_steps": 3750, "loss": 0.4893, "lr": 2.3765911755316503e-05, "epoch": 2.976, "percentage": 49.6, "elapsed_time": "1:32:06", "remaining_time": "1:33:35"}
{"current_steps": 1865, "total_steps": 3750, "loss": 0.467, "lr": 2.3674452123831125e-05, "epoch": 2.984, "percentage": 49.73, "elapsed_time": "1:32:18", "remaining_time": "1:33:18"}
{"current_steps": 1870, "total_steps": 3750, "loss": 0.4914, "lr": 2.358291289767165e-05, "epoch": 2.992, "percentage": 49.87, "elapsed_time": "1:32:30", "remaining_time": "1:32:59"}
{"current_steps": 1875, "total_steps": 3750, "loss": 0.4783, "lr": 2.3491296059728202e-05, "epoch": 3.0, "percentage": 50.0, "elapsed_time": "1:32:42", "remaining_time": "1:32:42"}
{"current_steps": 1880, "total_steps": 3750, "loss": 0.4038, "lr": 2.339960359457212e-05, "epoch": 3.008, "percentage": 50.13, "elapsed_time": "1:32:54", "remaining_time": "1:32:25"}
{"current_steps": 1885, "total_steps": 3750, "loss": 0.4146, "lr": 2.3307837488412955e-05, "epoch": 3.016, "percentage": 50.27, "elapsed_time": "1:33:07", "remaining_time": "1:32:07"}
{"current_steps": 1890, "total_steps": 3750, "loss": 0.4308, "lr": 2.3215999729055437e-05, "epoch": 3.024, "percentage": 50.4, "elapsed_time": "1:33:19", "remaining_time": "1:31:50"}
{"current_steps": 1895, "total_steps": 3750, "loss": 0.4104, "lr": 2.312409230585641e-05, "epoch": 3.032, "percentage": 50.53, "elapsed_time": "1:33:31", "remaining_time": "1:31:32"}
{"current_steps": 1900, "total_steps": 3750, "loss": 0.4465, "lr": 2.3032117209681782e-05, "epoch": 3.04, "percentage": 50.67, "elapsed_time": "1:33:43", "remaining_time": "1:31:15"}
{"current_steps": 1905, "total_steps": 3750, "loss": 0.4126, "lr": 2.2940076432863335e-05, "epoch": 3.048, "percentage": 50.8, "elapsed_time": "1:34:57", "remaining_time": "1:31:58"}
{"current_steps": 1910, "total_steps": 3750, "loss": 0.4584, "lr": 2.2847971969155626e-05, "epoch": 3.056, "percentage": 50.93, "elapsed_time": "1:35:09", "remaining_time": "1:31:40"}
{"current_steps": 1915, "total_steps": 3750, "loss": 0.4231, "lr": 2.275580581369276e-05, "epoch": 3.064, "percentage": 51.07, "elapsed_time": "1:35:22", "remaining_time": "1:31:23"}
{"current_steps": 1920, "total_steps": 3750, "loss": 0.4335, "lr": 2.2663579962945205e-05, "epoch": 3.072, "percentage": 51.2, "elapsed_time": "1:35:33", "remaining_time": "1:31:04"}
{"current_steps": 1925, "total_steps": 3750, "loss": 0.4196, "lr": 2.2571296414676503e-05, "epoch": 3.08, "percentage": 51.33, "elapsed_time": "1:35:45", "remaining_time": "1:30:47"}
{"current_steps": 1930, "total_steps": 3750, "loss": 0.4236, "lr": 2.2478957167900038e-05, "epoch": 3.088, "percentage": 51.47, "elapsed_time": "1:35:57", "remaining_time": "1:30:29"}
{"current_steps": 1935, "total_steps": 3750, "loss": 0.4177, "lr": 2.23865642228357e-05, "epoch": 3.096, "percentage": 51.6, "elapsed_time": "1:36:09", "remaining_time": "1:30:11"}
{"current_steps": 1940, "total_steps": 3750, "loss": 0.4383, "lr": 2.2294119580866592e-05, "epoch": 3.104, "percentage": 51.73, "elapsed_time": "1:36:20", "remaining_time": "1:29:53"}
{"current_steps": 1945, "total_steps": 3750, "loss": 0.418, "lr": 2.2201625244495646e-05, "epoch": 3.112, "percentage": 51.87, "elapsed_time": "1:36:32", "remaining_time": "1:29:35"}
{"current_steps": 1950, "total_steps": 3750, "loss": 0.4467, "lr": 2.2109083217302242e-05, "epoch": 3.12, "percentage": 52.0, "elapsed_time": "1:36:44", "remaining_time": "1:29:17"}
{"current_steps": 1955, "total_steps": 3750, "loss": 0.42, "lr": 2.201649550389885e-05, "epoch": 3.128, "percentage": 52.13, "elapsed_time": "1:36:55", "remaining_time": "1:28:59"}
{"current_steps": 1960, "total_steps": 3750, "loss": 0.4467, "lr": 2.1923864109887556e-05, "epoch": 3.136, "percentage": 52.27, "elapsed_time": "1:37:07", "remaining_time": "1:28:41"}
{"current_steps": 1965, "total_steps": 3750, "loss": 0.4113, "lr": 2.1831191041816652e-05, "epoch": 3.144, "percentage": 52.4, "elapsed_time": "1:37:18", "remaining_time": "1:28:24"}
{"current_steps": 1970, "total_steps": 3750, "loss": 0.4299, "lr": 2.173847830713715e-05, "epoch": 3.152, "percentage": 52.53, "elapsed_time": "1:37:30", "remaining_time": "1:28:06"}
{"current_steps": 1975, "total_steps": 3750, "loss": 0.4111, "lr": 2.1645727914159315e-05, "epoch": 3.16, "percentage": 52.67, "elapsed_time": "1:37:41", "remaining_time": "1:27:48"}
{"current_steps": 1980, "total_steps": 3750, "loss": 0.4379, "lr": 2.1552941872009144e-05, "epoch": 3.168, "percentage": 52.8, "elapsed_time": "1:37:53", "remaining_time": "1:27:30"}
{"current_steps": 1985, "total_steps": 3750, "loss": 0.4421, "lr": 2.1460122190584868e-05, "epoch": 3.176, "percentage": 52.93, "elapsed_time": "1:38:05", "remaining_time": "1:27:12"}
{"current_steps": 1990, "total_steps": 3750, "loss": 0.4159, "lr": 2.1367270880513377e-05, "epoch": 3.184, "percentage": 53.07, "elapsed_time": "1:38:17", "remaining_time": "1:26:55"}
{"current_steps": 1995, "total_steps": 3750, "loss": 0.4265, "lr": 2.127438995310671e-05, "epoch": 3.192, "percentage": 53.2, "elapsed_time": "1:38:29", "remaining_time": "1:26:38"}
{"current_steps": 2000, "total_steps": 3750, "loss": 0.4428, "lr": 2.118148142031846e-05, "epoch": 3.2, "percentage": 53.33, "elapsed_time": "1:38:41", "remaining_time": "1:26:21"}
{"current_steps": 2005, "total_steps": 3750, "loss": 0.4221, "lr": 2.1088547294700182e-05, "epoch": 3.208, "percentage": 53.47, "elapsed_time": "1:40:07", "remaining_time": "1:27:08"}
{"current_steps": 2010, "total_steps": 3750, "loss": 0.4255, "lr": 2.0995589589357846e-05, "epoch": 3.216, "percentage": 53.6, "elapsed_time": "1:40:18", "remaining_time": "1:26:50"}
{"current_steps": 2015, "total_steps": 3750, "loss": 0.4305, "lr": 2.0902610317908175e-05, "epoch": 3.224, "percentage": 53.73, "elapsed_time": "1:40:30", "remaining_time": "1:26:32"}
{"current_steps": 2020, "total_steps": 3750, "loss": 0.4256, "lr": 2.080961149443505e-05, "epoch": 3.232, "percentage": 53.87, "elapsed_time": "1:40:42", "remaining_time": "1:26:15"}
{"current_steps": 2025, "total_steps": 3750, "loss": 0.4013, "lr": 2.071659513344589e-05, "epoch": 3.24, "percentage": 54.0, "elapsed_time": "1:40:54", "remaining_time": "1:25:57"}
{"current_steps": 2030, "total_steps": 3750, "loss": 0.4306, "lr": 2.0623563249828e-05, "epoch": 3.248, "percentage": 54.13, "elapsed_time": "1:41:06", "remaining_time": "1:25:40"}
{"current_steps": 2035, "total_steps": 3750, "loss": 0.4129, "lr": 2.053051785880492e-05, "epoch": 3.2560000000000002, "percentage": 54.27, "elapsed_time": "1:41:18", "remaining_time": "1:25:22"}
{"current_steps": 2040, "total_steps": 3750, "loss": 0.4432, "lr": 2.0437460975892814e-05, "epoch": 3.2640000000000002, "percentage": 54.4, "elapsed_time": "1:41:31", "remaining_time": "1:25:05"}
{"current_steps": 2045, "total_steps": 3750, "loss": 0.4222, "lr": 2.0344394616856736e-05, "epoch": 3.2720000000000002, "percentage": 54.53, "elapsed_time": "1:41:42", "remaining_time": "1:24:47"}
{"current_steps": 2050, "total_steps": 3750, "loss": 0.4177, "lr": 2.0251320797667056e-05, "epoch": 3.2800000000000002, "percentage": 54.67, "elapsed_time": "1:41:54", "remaining_time": "1:24:30"}
{"current_steps": 2055, "total_steps": 3750, "loss": 0.424, "lr": 2.01582415344557e-05, "epoch": 3.288, "percentage": 54.8, "elapsed_time": "1:42:06", "remaining_time": "1:24:12"}
{"current_steps": 2060, "total_steps": 3750, "loss": 0.4053, "lr": 2.006515884347255e-05, "epoch": 3.296, "percentage": 54.93, "elapsed_time": "1:42:18", "remaining_time": "1:23:55"}
{"current_steps": 2065, "total_steps": 3750, "loss": 0.4265, "lr": 1.9972074741041712e-05, "epoch": 3.304, "percentage": 55.07, "elapsed_time": "1:42:30", "remaining_time": "1:23:38"}
{"current_steps": 2070, "total_steps": 3750, "loss": 0.4469, "lr": 1.9878991243517913e-05, "epoch": 3.312, "percentage": 55.2, "elapsed_time": "1:42:42", "remaining_time": "1:23:21"}
{"current_steps": 2075, "total_steps": 3750, "loss": 0.433, "lr": 1.9785910367242712e-05, "epoch": 3.32, "percentage": 55.33, "elapsed_time": "1:42:54", "remaining_time": "1:23:04"}
{"current_steps": 2080, "total_steps": 3750, "loss": 0.4137, "lr": 1.969283412850094e-05, "epoch": 3.328, "percentage": 55.47, "elapsed_time": "1:43:06", "remaining_time": "1:22:46"}
{"current_steps": 2085, "total_steps": 3750, "loss": 0.4228, "lr": 1.959976454347696e-05, "epoch": 3.336, "percentage": 55.6, "elapsed_time": "1:43:18", "remaining_time": "1:22:29"}
{"current_steps": 2090, "total_steps": 3750, "loss": 0.4163, "lr": 1.950670362821098e-05, "epoch": 3.344, "percentage": 55.73, "elapsed_time": "1:43:30", "remaining_time": "1:22:12"}
{"current_steps": 2095, "total_steps": 3750, "loss": 0.4237, "lr": 1.9413653398555437e-05, "epoch": 3.352, "percentage": 55.87, "elapsed_time": "1:43:42", "remaining_time": "1:21:55"}
{"current_steps": 2100, "total_steps": 3750, "loss": 0.4188, "lr": 1.9320615870131282e-05, "epoch": 3.36, "percentage": 56.0, "elapsed_time": "1:43:54", "remaining_time": "1:21:38"}
{"current_steps": 2105, "total_steps": 3750, "loss": 0.4199, "lr": 1.9227593058284343e-05, "epoch": 3.368, "percentage": 56.13, "elapsed_time": "1:45:23", "remaining_time": "1:22:21"}
{"current_steps": 2110, "total_steps": 3750, "loss": 0.411, "lr": 1.9134586978041663e-05, "epoch": 3.376, "percentage": 56.27, "elapsed_time": "1:45:35", "remaining_time": "1:22:04"}
{"current_steps": 2115, "total_steps": 3750, "loss": 0.4368, "lr": 1.9041599644067846e-05, "epoch": 3.384, "percentage": 56.4, "elapsed_time": "1:45:47", "remaining_time": "1:21:46"}
{"current_steps": 2120, "total_steps": 3750, "loss": 0.4266, "lr": 1.8948633070621433e-05, "epoch": 3.392, "percentage": 56.53, "elapsed_time": "1:45:59", "remaining_time": "1:21:29"}
{"current_steps": 2125, "total_steps": 3750, "loss": 0.43, "lr": 1.885568927151124e-05, "epoch": 3.4, "percentage": 56.67, "elapsed_time": "1:46:10", "remaining_time": "1:21:11"}
{"current_steps": 2130, "total_steps": 3750, "loss": 0.4438, "lr": 1.8762770260052773e-05, "epoch": 3.408, "percentage": 56.8, "elapsed_time": "1:46:22", "remaining_time": "1:20:54"}
{"current_steps": 2135, "total_steps": 3750, "loss": 0.4433, "lr": 1.8669878049024575e-05, "epoch": 3.416, "percentage": 56.93, "elapsed_time": "1:46:34", "remaining_time": "1:20:37"}
{"current_steps": 2140, "total_steps": 3750, "loss": 0.4285, "lr": 1.857701465062467e-05, "epoch": 3.424, "percentage": 57.07, "elapsed_time": "1:46:46", "remaining_time": "1:20:19"}
{"current_steps": 2145, "total_steps": 3750, "loss": 0.4422, "lr": 1.848418207642693e-05, "epoch": 3.432, "percentage": 57.2, "elapsed_time": "1:46:58", "remaining_time": "1:20:02"}
{"current_steps": 2150, "total_steps": 3750, "loss": 0.4362, "lr": 1.8391382337337548e-05, "epoch": 3.44, "percentage": 57.33, "elapsed_time": "1:47:09", "remaining_time": "1:19:45"}
{"current_steps": 2155, "total_steps": 3750, "loss": 0.4092, "lr": 1.829861744355144e-05, "epoch": 3.448, "percentage": 57.47, "elapsed_time": "1:47:21", "remaining_time": "1:19:27"}
{"current_steps": 2160, "total_steps": 3750, "loss": 0.4367, "lr": 1.820588940450872e-05, "epoch": 3.456, "percentage": 57.6, "elapsed_time": "1:47:33", "remaining_time": "1:19:10"}
{"current_steps": 2165, "total_steps": 3750, "loss": 0.4345, "lr": 1.8113200228851163e-05, "epoch": 3.464, "percentage": 57.73, "elapsed_time": "1:47:45", "remaining_time": "1:18:53"}
{"current_steps": 2170, "total_steps": 3750, "loss": 0.4156, "lr": 1.80205519243787e-05, "epoch": 3.472, "percentage": 57.87, "elapsed_time": "1:47:57", "remaining_time": "1:18:36"}
{"current_steps": 2175, "total_steps": 3750, "loss": 0.4466, "lr": 1.7927946498005934e-05, "epoch": 3.48, "percentage": 58.0, "elapsed_time": "1:48:10", "remaining_time": "1:18:19"}
{"current_steps": 2180, "total_steps": 3750, "loss": 0.4498, "lr": 1.7835385955718653e-05, "epoch": 3.488, "percentage": 58.13, "elapsed_time": "1:48:22", "remaining_time": "1:18:02"}
{"current_steps": 2185, "total_steps": 3750, "loss": 0.4232, "lr": 1.7742872302530366e-05, "epoch": 3.496, "percentage": 58.27, "elapsed_time": "1:48:34", "remaining_time": "1:17:46"}
{"current_steps": 2190, "total_steps": 3750, "loss": 0.4309, "lr": 1.765040754243892e-05, "epoch": 3.504, "percentage": 58.4, "elapsed_time": "1:48:47", "remaining_time": "1:17:29"}
{"current_steps": 2195, "total_steps": 3750, "loss": 0.4227, "lr": 1.755799367838302e-05, "epoch": 3.512, "percentage": 58.53, "elapsed_time": "1:48:59", "remaining_time": "1:17:12"}
{"current_steps": 2200, "total_steps": 3750, "loss": 0.4217, "lr": 1.746563271219891e-05, "epoch": 3.52, "percentage": 58.67, "elapsed_time": "1:49:11", "remaining_time": "1:16:56"}
{"current_steps": 2205, "total_steps": 3750, "loss": 0.4377, "lr": 1.7373326644576965e-05, "epoch": 3.528, "percentage": 58.8, "elapsed_time": "1:50:24", "remaining_time": "1:17:21"}
{"current_steps": 2210, "total_steps": 3750, "loss": 0.4391, "lr": 1.728107747501836e-05, "epoch": 3.536, "percentage": 58.93, "elapsed_time": "1:50:35", "remaining_time": "1:17:03"}
{"current_steps": 2215, "total_steps": 3750, "loss": 0.4245, "lr": 1.7188887201791785e-05, "epoch": 3.544, "percentage": 59.07, "elapsed_time": "1:50:47", "remaining_time": "1:16:46"}
{"current_steps": 2220, "total_steps": 3750, "loss": 0.4286, "lr": 1.7096757821890117e-05, "epoch": 3.552, "percentage": 59.2, "elapsed_time": "1:50:59", "remaining_time": "1:16:29"}
{"current_steps": 2225, "total_steps": 3750, "loss": 0.4411, "lr": 1.7004691330987196e-05, "epoch": 3.56, "percentage": 59.33, "elapsed_time": "1:51:11", "remaining_time": "1:16:12"}
{"current_steps": 2230, "total_steps": 3750, "loss": 0.4127, "lr": 1.691268972339458e-05, "epoch": 3.568, "percentage": 59.47, "elapsed_time": "1:51:23", "remaining_time": "1:15:55"}
{"current_steps": 2235, "total_steps": 3750, "loss": 0.4371, "lr": 1.6820754992018344e-05, "epoch": 3.576, "percentage": 59.6, "elapsed_time": "1:51:36", "remaining_time": "1:15:39"}
{"current_steps": 2240, "total_steps": 3750, "loss": 0.4243, "lr": 1.6728889128315932e-05, "epoch": 3.584, "percentage": 59.73, "elapsed_time": "1:51:48", "remaining_time": "1:15:22"}
{"current_steps": 2245, "total_steps": 3750, "loss": 0.4373, "lr": 1.663709412225297e-05, "epoch": 3.592, "percentage": 59.87, "elapsed_time": "1:52:00", "remaining_time": "1:15:05"}
{"current_steps": 2250, "total_steps": 3750, "loss": 0.4274, "lr": 1.654537196226022e-05, "epoch": 3.6, "percentage": 60.0, "elapsed_time": "1:52:11", "remaining_time": "1:14:47"}
{"current_steps": 2255, "total_steps": 3750, "loss": 0.4293, "lr": 1.6453724635190455e-05, "epoch": 3.608, "percentage": 60.13, "elapsed_time": "1:52:23", "remaining_time": "1:14:30"}
{"current_steps": 2260, "total_steps": 3750, "loss": 0.4193, "lr": 1.6362154126275467e-05, "epoch": 3.616, "percentage": 60.27, "elapsed_time": "1:52:34", "remaining_time": "1:14:13"}
{"current_steps": 2265, "total_steps": 3750, "loss": 0.426, "lr": 1.6270662419083018e-05, "epoch": 3.624, "percentage": 60.4, "elapsed_time": "1:52:47", "remaining_time": "1:13:56"}
{"current_steps": 2270, "total_steps": 3750, "loss": 0.4208, "lr": 1.617925149547391e-05, "epoch": 3.632, "percentage": 60.53, "elapsed_time": "1:52:58", "remaining_time": "1:13:39"}
{"current_steps": 2275, "total_steps": 3750, "loss": 0.4113, "lr": 1.608792333555904e-05, "epoch": 3.64, "percentage": 60.67, "elapsed_time": "1:53:11", "remaining_time": "1:13:23"}
{"current_steps": 2280, "total_steps": 3750, "loss": 0.4308, "lr": 1.5996679917656492e-05, "epoch": 3.648, "percentage": 60.8, "elapsed_time": "1:53:22", "remaining_time": "1:13:05"}
{"current_steps": 2285, "total_steps": 3750, "loss": 0.441, "lr": 1.5905523218248723e-05, "epoch": 3.656, "percentage": 60.93, "elapsed_time": "1:53:34", "remaining_time": "1:12:48"}
{"current_steps": 2290, "total_steps": 3750, "loss": 0.3989, "lr": 1.5814455211939698e-05, "epoch": 3.664, "percentage": 61.07, "elapsed_time": "1:53:46", "remaining_time": "1:12:32"}
{"current_steps": 2295, "total_steps": 3750, "loss": 0.4348, "lr": 1.5723477871412168e-05, "epoch": 3.672, "percentage": 61.2, "elapsed_time": "1:53:57", "remaining_time": "1:12:15"}
{"current_steps": 2300, "total_steps": 3750, "loss": 0.4451, "lr": 1.56325931673849e-05, "epoch": 3.68, "percentage": 61.33, "elapsed_time": "1:54:09", "remaining_time": "1:11:58"}
{"current_steps": 2305, "total_steps": 3750, "loss": 0.4361, "lr": 1.5541803068569993e-05, "epoch": 3.6879999999999997, "percentage": 61.47, "elapsed_time": "1:55:25", "remaining_time": "1:12:21"}
{"current_steps": 2310, "total_steps": 3750, "loss": 0.409, "lr": 1.5451109541630275e-05, "epoch": 3.6959999999999997, "percentage": 61.6, "elapsed_time": "1:55:37", "remaining_time": "1:12:04"}
{"current_steps": 2315, "total_steps": 3750, "loss": 0.4197, "lr": 1.536051455113663e-05, "epoch": 3.7039999999999997, "percentage": 61.73, "elapsed_time": "1:55:48", "remaining_time": "1:11:47"}
{"current_steps": 2320, "total_steps": 3750, "loss": 0.4309, "lr": 1.527002005952551e-05, "epoch": 3.7119999999999997, "percentage": 61.87, "elapsed_time": "1:55:59", "remaining_time": "1:11:29"}
{"current_steps": 2325, "total_steps": 3750, "loss": 0.4342, "lr": 1.5179628027056373e-05, "epoch": 3.7199999999999998, "percentage": 62.0, "elapsed_time": "1:56:10", "remaining_time": "1:11:12"}
{"current_steps": 2330, "total_steps": 3750, "loss": 0.4339, "lr": 1.5089340411769257e-05, "epoch": 3.7279999999999998, "percentage": 62.13, "elapsed_time": "1:56:22", "remaining_time": "1:10:55"}
{"current_steps": 2335, "total_steps": 3750, "loss": 0.4105, "lr": 1.499915916944236e-05, "epoch": 3.7359999999999998, "percentage": 62.27, "elapsed_time": "1:56:34", "remaining_time": "1:10:38"}
{"current_steps": 2340, "total_steps": 3750, "loss": 0.4387, "lr": 1.490908625354964e-05, "epoch": 3.7439999999999998, "percentage": 62.4, "elapsed_time": "1:56:46", "remaining_time": "1:10:21"}
{"current_steps": 2345, "total_steps": 3750, "loss": 0.4325, "lr": 1.4819123615218556e-05, "epoch": 3.752, "percentage": 62.53, "elapsed_time": "1:56:57", "remaining_time": "1:10:04"}
{"current_steps": 2350, "total_steps": 3750, "loss": 0.4152, "lr": 1.472927320318775e-05, "epoch": 3.76, "percentage": 62.67, "elapsed_time": "1:57:09", "remaining_time": "1:09:47"}
{"current_steps": 2355, "total_steps": 3750, "loss": 0.4434, "lr": 1.4639536963764878e-05, "epoch": 3.768, "percentage": 62.8, "elapsed_time": "1:57:20", "remaining_time": "1:09:30"}
{"current_steps": 2360, "total_steps": 3750, "loss": 0.4395, "lr": 1.4549916840784409e-05, "epoch": 3.776, "percentage": 62.93, "elapsed_time": "1:57:31", "remaining_time": "1:09:13"}
{"current_steps": 2365, "total_steps": 3750, "loss": 0.4161, "lr": 1.4460414775565555e-05, "epoch": 3.784, "percentage": 63.07, "elapsed_time": "1:57:43", "remaining_time": "1:08:56"}
{"current_steps": 2370, "total_steps": 3750, "loss": 0.407, "lr": 1.43710327068702e-05, "epoch": 3.792, "percentage": 63.2, "elapsed_time": "1:57:55", "remaining_time": "1:08:39"}
{"current_steps": 2375, "total_steps": 3750, "loss": 0.4124, "lr": 1.4281772570860897e-05, "epoch": 3.8, "percentage": 63.33, "elapsed_time": "1:58:07", "remaining_time": "1:08:23"}
{"current_steps": 2380, "total_steps": 3750, "loss": 0.4187, "lr": 1.4192636301058952e-05, "epoch": 3.808, "percentage": 63.47, "elapsed_time": "1:58:19", "remaining_time": "1:08:06"}
{"current_steps": 2385, "total_steps": 3750, "loss": 0.441, "lr": 1.4103625828302508e-05, "epoch": 3.816, "percentage": 63.6, "elapsed_time": "1:58:31", "remaining_time": "1:07:50"}
{"current_steps": 2390, "total_steps": 3750, "loss": 0.4277, "lr": 1.4014743080704743e-05, "epoch": 3.824, "percentage": 63.73, "elapsed_time": "1:58:42", "remaining_time": "1:07:32"}
{"current_steps": 2395, "total_steps": 3750, "loss": 0.4364, "lr": 1.3925989983612118e-05, "epoch": 3.832, "percentage": 63.87, "elapsed_time": "1:58:53", "remaining_time": "1:07:15"}
{"current_steps": 2400, "total_steps": 3750, "loss": 0.4145, "lr": 1.383736845956261e-05, "epoch": 3.84, "percentage": 64.0, "elapsed_time": "1:59:05", "remaining_time": "1:06:59"}
{"current_steps": 2405, "total_steps": 3750, "loss": 0.4145, "lr": 1.3748880428244154e-05, "epoch": 3.848, "percentage": 64.13, "elapsed_time": "2:00:20", "remaining_time": "1:07:17"}
{"current_steps": 2410, "total_steps": 3750, "loss": 0.4291, "lr": 1.3660527806452965e-05, "epoch": 3.856, "percentage": 64.27, "elapsed_time": "2:00:32", "remaining_time": "1:07:01"}
{"current_steps": 2415, "total_steps": 3750, "loss": 0.4544, "lr": 1.3572312508052118e-05, "epoch": 3.864, "percentage": 64.4, "elapsed_time": "2:00:42", "remaining_time": "1:06:43"}
{"current_steps": 2420, "total_steps": 3750, "loss": 0.416, "lr": 1.3484236443929982e-05, "epoch": 3.872, "percentage": 64.53, "elapsed_time": "2:00:54", "remaining_time": "1:06:26"}
{"current_steps": 2425, "total_steps": 3750, "loss": 0.4331, "lr": 1.3396301521958926e-05, "epoch": 3.88, "percentage": 64.67, "elapsed_time": "2:01:05", "remaining_time": "1:06:10"}
{"current_steps": 2430, "total_steps": 3750, "loss": 0.4116, "lr": 1.3308509646953934e-05, "epoch": 3.888, "percentage": 64.8, "elapsed_time": "2:01:17", "remaining_time": "1:05:53"}
{"current_steps": 2435, "total_steps": 3750, "loss": 0.424, "lr": 1.3220862720631349e-05, "epoch": 3.896, "percentage": 64.93, "elapsed_time": "2:01:29", "remaining_time": "1:05:36"}
{"current_steps": 2440, "total_steps": 3750, "loss": 0.4395, "lr": 1.3133362641567697e-05, "epoch": 3.904, "percentage": 65.07, "elapsed_time": "2:01:41", "remaining_time": "1:05:19"}
{"current_steps": 2445, "total_steps": 3750, "loss": 0.4358, "lr": 1.3046011305158546e-05, "epoch": 3.912, "percentage": 65.2, "elapsed_time": "2:01:53", "remaining_time": "1:05:03"}
{"current_steps": 2450, "total_steps": 3750, "loss": 0.4278, "lr": 1.2958810603577456e-05, "epoch": 3.92, "percentage": 65.33, "elapsed_time": "2:02:04", "remaining_time": "1:04:46"}
{"current_steps": 2455, "total_steps": 3750, "loss": 0.416, "lr": 1.2871762425734989e-05, "epoch": 3.928, "percentage": 65.47, "elapsed_time": "2:02:16", "remaining_time": "1:04:29"}
{"current_steps": 2460, "total_steps": 3750, "loss": 0.4115, "lr": 1.278486865723779e-05, "epoch": 3.936, "percentage": 65.6, "elapsed_time": "2:02:27", "remaining_time": "1:04:12"}
{"current_steps": 2465, "total_steps": 3750, "loss": 0.4284, "lr": 1.269813118034775e-05, "epoch": 3.944, "percentage": 65.73, "elapsed_time": "2:02:39", "remaining_time": "1:03:56"}
{"current_steps": 2470, "total_steps": 3750, "loss": 0.4416, "lr": 1.2611551873941213e-05, "epoch": 3.952, "percentage": 65.87, "elapsed_time": "2:02:51", "remaining_time": "1:03:40"}
{"current_steps": 2475, "total_steps": 3750, "loss": 0.4294, "lr": 1.2525132613468309e-05, "epoch": 3.96, "percentage": 66.0, "elapsed_time": "2:03:04", "remaining_time": "1:03:23"}
{"current_steps": 2480, "total_steps": 3750, "loss": 0.4379, "lr": 1.2438875270912294e-05, "epoch": 3.968, "percentage": 66.13, "elapsed_time": "2:03:15", "remaining_time": "1:03:06"}
{"current_steps": 2485, "total_steps": 3750, "loss": 0.407, "lr": 1.2352781714749016e-05, "epoch": 3.976, "percentage": 66.27, "elapsed_time": "2:03:26", "remaining_time": "1:02:50"}
{"current_steps": 2490, "total_steps": 3750, "loss": 0.4326, "lr": 1.2266853809906469e-05, "epoch": 3.984, "percentage": 66.4, "elapsed_time": "2:03:39", "remaining_time": "1:02:34"}
{"current_steps": 2495, "total_steps": 3750, "loss": 0.422, "lr": 1.2181093417724317e-05, "epoch": 3.992, "percentage": 66.53, "elapsed_time": "2:03:51", "remaining_time": "1:02:17"}
{"current_steps": 2500, "total_steps": 3750, "loss": 0.4068, "lr": 1.2095502395913676e-05, "epoch": 4.0, "percentage": 66.67, "elapsed_time": "2:04:02", "remaining_time": "1:02:01"}
{"current_steps": 2505, "total_steps": 3750, "loss": 0.3857, "lr": 1.2010082598516775e-05, "epoch": 4.008, "percentage": 66.8, "elapsed_time": "2:05:16", "remaining_time": "1:02:15"}
{"current_steps": 2510, "total_steps": 3750, "loss": 0.3836, "lr": 1.1924835875866884e-05, "epoch": 4.016, "percentage": 66.93, "elapsed_time": "2:05:27", "remaining_time": "1:01:58"}
{"current_steps": 2515, "total_steps": 3750, "loss": 0.3697, "lr": 1.1839764074548145e-05, "epoch": 4.024, "percentage": 67.07, "elapsed_time": "2:05:39", "remaining_time": "1:01:42"}
{"current_steps": 2520, "total_steps": 3750, "loss": 0.3797, "lr": 1.1754869037355659e-05, "epoch": 4.032, "percentage": 67.2, "elapsed_time": "2:05:50", "remaining_time": "1:01:25"}
{"current_steps": 2525, "total_steps": 3750, "loss": 0.3834, "lr": 1.1670152603255504e-05, "epoch": 4.04, "percentage": 67.33, "elapsed_time": "2:06:02", "remaining_time": "1:01:09"}
{"current_steps": 2530, "total_steps": 3750, "loss": 0.3829, "lr": 1.1585616607344909e-05, "epoch": 4.048, "percentage": 67.47, "elapsed_time": "2:06:14", "remaining_time": "1:00:52"}
{"current_steps": 2535, "total_steps": 3750, "loss": 0.393, "lr": 1.1501262880812547e-05, "epoch": 4.056, "percentage": 67.6, "elapsed_time": "2:06:25", "remaining_time": "1:00:35"}
{"current_steps": 2540, "total_steps": 3750, "loss": 0.3786, "lr": 1.141709325089881e-05, "epoch": 4.064, "percentage": 67.73, "elapsed_time": "2:06:37", "remaining_time": "1:00:19"}
{"current_steps": 2545, "total_steps": 3750, "loss": 0.3874, "lr": 1.1333109540856257e-05, "epoch": 4.072, "percentage": 67.87, "elapsed_time": "2:06:48", "remaining_time": "1:00:02"}
{"current_steps": 2550, "total_steps": 3750, "loss": 0.4088, "lr": 1.1249313569910143e-05, "epoch": 4.08, "percentage": 68.0, "elapsed_time": "2:07:00", "remaining_time": "0:59:45"}
{"current_steps": 2555, "total_steps": 3750, "loss": 0.3787, "lr": 1.1165707153218942e-05, "epoch": 4.088, "percentage": 68.13, "elapsed_time": "2:07:11", "remaining_time": "0:59:29"}
{"current_steps": 2560, "total_steps": 3750, "loss": 0.3866, "lr": 1.1082292101835121e-05, "epoch": 4.096, "percentage": 68.27, "elapsed_time": "2:07:23", "remaining_time": "0:59:13"}
{"current_steps": 2565, "total_steps": 3750, "loss": 0.3894, "lr": 1.099907022266582e-05, "epoch": 4.104, "percentage": 68.4, "elapsed_time": "2:07:35", "remaining_time": "0:58:56"}
{"current_steps": 2570, "total_steps": 3750, "loss": 0.3778, "lr": 1.0916043318433767e-05, "epoch": 4.112, "percentage": 68.53, "elapsed_time": "2:07:47", "remaining_time": "0:58:40"}
{"current_steps": 2575, "total_steps": 3750, "loss": 0.3926, "lr": 1.0833213187638203e-05, "epoch": 4.12, "percentage": 68.67, "elapsed_time": "2:07:58", "remaining_time": "0:58:23"}
{"current_steps": 2580, "total_steps": 3750, "loss": 0.3861, "lr": 1.0750581624515957e-05, "epoch": 4.128, "percentage": 68.8, "elapsed_time": "2:08:10", "remaining_time": "0:58:07"}
{"current_steps": 2585, "total_steps": 3750, "loss": 0.3925, "lr": 1.0668150419002527e-05, "epoch": 4.136, "percentage": 68.93, "elapsed_time": "2:08:22", "remaining_time": "0:57:51"}
{"current_steps": 2590, "total_steps": 3750, "loss": 0.3887, "lr": 1.0585921356693349e-05, "epoch": 4.144, "percentage": 69.07, "elapsed_time": "2:08:34", "remaining_time": "0:57:35"}
{"current_steps": 2595, "total_steps": 3750, "loss": 0.3727, "lr": 1.0503896218805112e-05, "epoch": 4.152, "percentage": 69.2, "elapsed_time": "2:08:47", "remaining_time": "0:57:19"}
{"current_steps": 2600, "total_steps": 3750, "loss": 0.3704, "lr": 1.0422076782137155e-05, "epoch": 4.16, "percentage": 69.33, "elapsed_time": "2:08:59", "remaining_time": "0:57:03"}
{"current_steps": 2605, "total_steps": 3750, "loss": 0.3774, "lr": 1.0340464819032991e-05, "epoch": 4.168, "percentage": 69.47, "elapsed_time": "2:10:32", "remaining_time": "0:57:22"}
{"current_steps": 2610, "total_steps": 3750, "loss": 0.3901, "lr": 1.0259062097341911e-05, "epoch": 4.176, "percentage": 69.6, "elapsed_time": "2:10:44", "remaining_time": "0:57:06"}
{"current_steps": 2615, "total_steps": 3750, "loss": 0.3889, "lr": 1.017787038038071e-05, "epoch": 4.184, "percentage": 69.73, "elapsed_time": "2:10:55", "remaining_time": "0:56:49"}
{"current_steps": 2620, "total_steps": 3750, "loss": 0.3867, "lr": 1.0096891426895476e-05, "epoch": 4.192, "percentage": 69.87, "elapsed_time": "2:11:07", "remaining_time": "0:56:33"}
{"current_steps": 2625, "total_steps": 3750, "loss": 0.384, "lr": 1.0016126991023447e-05, "epoch": 4.2, "percentage": 70.0, "elapsed_time": "2:11:18", "remaining_time": "0:56:16"}
{"current_steps": 2630, "total_steps": 3750, "loss": 0.3669, "lr": 9.935578822255113e-06, "epoch": 4.208, "percentage": 70.13, "elapsed_time": "2:11:30", "remaining_time": "0:56:00"}
{"current_steps": 2635, "total_steps": 3750, "loss": 0.4026, "lr": 9.855248665396218e-06, "epoch": 4.216, "percentage": 70.27, "elapsed_time": "2:11:43", "remaining_time": "0:55:44"}
{"current_steps": 2640, "total_steps": 3750, "loss": 0.3645, "lr": 9.775138260530046e-06, "epoch": 4.224, "percentage": 70.4, "elapsed_time": "2:11:54", "remaining_time": "0:55:27"}
{"current_steps": 2645, "total_steps": 3750, "loss": 0.3854, "lr": 9.695249342979667e-06, "epoch": 4.232, "percentage": 70.53, "elapsed_time": "2:12:06", "remaining_time": "0:55:11"}
{"current_steps": 2650, "total_steps": 3750, "loss": 0.3757, "lr": 9.615583643270371e-06, "epoch": 4.24, "percentage": 70.67, "elapsed_time": "2:12:19", "remaining_time": "0:54:55"}
{"current_steps": 2655, "total_steps": 3750, "loss": 0.3994, "lr": 9.536142887092208e-06, "epoch": 4.248, "percentage": 70.8, "elapsed_time": "2:12:30", "remaining_time": "0:54:38"}
{"current_steps": 2660, "total_steps": 3750, "loss": 0.3741, "lr": 9.456928795262552e-06, "epoch": 4.256, "percentage": 70.93, "elapsed_time": "2:12:41", "remaining_time": "0:54:22"}
{"current_steps": 2665, "total_steps": 3750, "loss": 0.3876, "lr": 9.377943083688873e-06, "epoch": 4.264, "percentage": 71.07, "elapsed_time": "2:12:54", "remaining_time": "0:54:06"}
{"current_steps": 2670, "total_steps": 3750, "loss": 0.3717, "lr": 9.29918746333153e-06, "epoch": 4.272, "percentage": 71.2, "elapsed_time": "2:13:06", "remaining_time": "0:53:50"}
{"current_steps": 2675, "total_steps": 3750, "loss": 0.3804, "lr": 9.220663640166756e-06, "epoch": 4.28, "percentage": 71.33, "elapsed_time": "2:13:18", "remaining_time": "0:53:34"}
{"current_steps": 2680, "total_steps": 3750, "loss": 0.3823, "lr": 9.142373315149655e-06, "epoch": 4.288, "percentage": 71.47, "elapsed_time": "2:13:29", "remaining_time": "0:53:17"}
{"current_steps": 2685, "total_steps": 3750, "loss": 0.3767, "lr": 9.064318184177373e-06, "epoch": 4.296, "percentage": 71.6, "elapsed_time": "2:13:41", "remaining_time": "0:53:01"}
{"current_steps": 2690, "total_steps": 3750, "loss": 0.3871, "lr": 8.986499938052396e-06, "epoch": 4.304, "percentage": 71.73, "elapsed_time": "2:13:53", "remaining_time": "0:52:45"}
{"current_steps": 2695, "total_steps": 3750, "loss": 0.3962, "lr": 8.908920262445859e-06, "epoch": 4.312, "percentage": 71.87, "elapsed_time": "2:14:05", "remaining_time": "0:52:29"}
{"current_steps": 2700, "total_steps": 3750, "loss": 0.3902, "lr": 8.831580837861082e-06, "epoch": 4.32, "percentage": 72.0, "elapsed_time": "2:14:16", "remaining_time": "0:52:13"}
{"current_steps": 2705, "total_steps": 3750, "loss": 0.4052, "lr": 8.754483339597166e-06, "epoch": 4.328, "percentage": 72.13, "elapsed_time": "2:15:36", "remaining_time": "0:52:23"}
{"current_steps": 2710, "total_steps": 3750, "loss": 0.3951, "lr": 8.677629437712665e-06, "epoch": 4.336, "percentage": 72.27, "elapsed_time": "2:15:48", "remaining_time": "0:52:06"}
{"current_steps": 2715, "total_steps": 3750, "loss": 0.3825, "lr": 8.601020796989467e-06, "epoch": 4.344, "percentage": 72.4, "elapsed_time": "2:15:59", "remaining_time": "0:51:50"}
{"current_steps": 2720, "total_steps": 3750, "loss": 0.3893, "lr": 8.524659076896656e-06, "epoch": 4.352, "percentage": 72.53, "elapsed_time": "2:16:11", "remaining_time": "0:51:34"}
{"current_steps": 2725, "total_steps": 3750, "loss": 0.3809, "lr": 8.448545931554652e-06, "epoch": 4.36, "percentage": 72.67, "elapsed_time": "2:16:23", "remaining_time": "0:51:18"}
{"current_steps": 2730, "total_steps": 3750, "loss": 0.4045, "lr": 8.372683009699307e-06, "epoch": 4.368, "percentage": 72.8, "elapsed_time": "2:16:34", "remaining_time": "0:51:01"}
{"current_steps": 2735, "total_steps": 3750, "loss": 0.3735, "lr": 8.297071954646248e-06, "epoch": 4.376, "percentage": 72.93, "elapsed_time": "2:16:46", "remaining_time": "0:50:45"}
{"current_steps": 2740, "total_steps": 3750, "loss": 0.3843, "lr": 8.22171440425523e-06, "epoch": 4.384, "percentage": 73.07, "elapsed_time": "2:16:58", "remaining_time": "0:50:29"}
{"current_steps": 2745, "total_steps": 3750, "loss": 0.3739, "lr": 8.146611990894683e-06, "epoch": 4.392, "percentage": 73.2, "elapsed_time": "2:17:10", "remaining_time": "0:50:13"}
{"current_steps": 2750, "total_steps": 3750, "loss": 0.3942, "lr": 8.071766341406363e-06, "epoch": 4.4, "percentage": 73.33, "elapsed_time": "2:17:21", "remaining_time": "0:49:56"}
{"current_steps": 2755, "total_steps": 3750, "loss": 0.3892, "lr": 7.997179077070092e-06, "epoch": 4.408, "percentage": 73.47, "elapsed_time": "2:17:33", "remaining_time": "0:49:40"}
{"current_steps": 2760, "total_steps": 3750, "loss": 0.4054, "lr": 7.92285181356864e-06, "epoch": 4.416, "percentage": 73.6, "elapsed_time": "2:17:45", "remaining_time": "0:49:24"}
{"current_steps": 2765, "total_steps": 3750, "loss": 0.3891, "lr": 7.848786160952726e-06, "epoch": 4.424, "percentage": 73.73, "elapsed_time": "2:17:57", "remaining_time": "0:49:08"}
{"current_steps": 2770, "total_steps": 3750, "loss": 0.3946, "lr": 7.77498372360617e-06, "epoch": 4.432, "percentage": 73.87, "elapsed_time": "2:18:09", "remaining_time": "0:48:52"}
{"current_steps": 2775, "total_steps": 3750, "loss": 0.4016, "lr": 7.701446100211095e-06, "epoch": 4.44, "percentage": 74.0, "elapsed_time": "2:18:21", "remaining_time": "0:48:36"}
{"current_steps": 2780, "total_steps": 3750, "loss": 0.3759, "lr": 7.628174883713322e-06, "epoch": 4.448, "percentage": 74.13, "elapsed_time": "2:18:33", "remaining_time": "0:48:20"}
{"current_steps": 2785, "total_steps": 3750, "loss": 0.4006, "lr": 7.555171661287875e-06, "epoch": 4.456, "percentage": 74.27, "elapsed_time": "2:18:45", "remaining_time": "0:48:04"}
{"current_steps": 2790, "total_steps": 3750, "loss": 0.3809, "lr": 7.482438014304567e-06, "epoch": 4.464, "percentage": 74.4, "elapsed_time": "2:18:57", "remaining_time": "0:47:48"}
{"current_steps": 2795, "total_steps": 3750, "loss": 0.3965, "lr": 7.4099755182937685e-06, "epoch": 4.4719999999999995, "percentage": 74.53, "elapsed_time": "2:19:09", "remaining_time": "0:47:32"}
{"current_steps": 2800, "total_steps": 3750, "loss": 0.3825, "lr": 7.337785742912289e-06, "epoch": 4.48, "percentage": 74.67, "elapsed_time": "2:19:21", "remaining_time": "0:47:16"}
{"current_steps": 2805, "total_steps": 3750, "loss": 0.3946, "lr": 7.265870251909335e-06, "epoch": 4.4879999999999995, "percentage": 74.8, "elapsed_time": "2:20:35", "remaining_time": "0:47:21"}
{"current_steps": 2810, "total_steps": 3750, "loss": 0.3899, "lr": 7.194230603092697e-06, "epoch": 4.496, "percentage": 74.93, "elapsed_time": "2:20:46", "remaining_time": "0:47:05"}
{"current_steps": 2815, "total_steps": 3750, "loss": 0.3958, "lr": 7.122868348294927e-06, "epoch": 4.504, "percentage": 75.07, "elapsed_time": "2:20:57", "remaining_time": "0:46:49"}
{"current_steps": 2820, "total_steps": 3750, "loss": 0.3969, "lr": 7.051785033339804e-06, "epoch": 4.5120000000000005, "percentage": 75.2, "elapsed_time": "2:21:08", "remaining_time": "0:46:32"}
{"current_steps": 2825, "total_steps": 3750, "loss": 0.3803, "lr": 6.980982198008785e-06, "epoch": 4.52, "percentage": 75.33, "elapsed_time": "2:21:20", "remaining_time": "0:46:16"}
{"current_steps": 2830, "total_steps": 3750, "loss": 0.3868, "lr": 6.910461376007704e-06, "epoch": 4.5280000000000005, "percentage": 75.47, "elapsed_time": "2:21:32", "remaining_time": "0:46:00"}
{"current_steps": 2835, "total_steps": 3750, "loss": 0.4073, "lr": 6.840224094933501e-06, "epoch": 4.536, "percentage": 75.6, "elapsed_time": "2:21:44", "remaining_time": "0:45:44"}
{"current_steps": 2840, "total_steps": 3750, "loss": 0.3798, "lr": 6.7702718762411505e-06, "epoch": 4.5440000000000005, "percentage": 75.73, "elapsed_time": "2:21:56", "remaining_time": "0:45:28"}
{"current_steps": 2845, "total_steps": 3750, "loss": 0.3632, "lr": 6.700606235210731e-06, "epoch": 4.552, "percentage": 75.87, "elapsed_time": "2:22:08", "remaining_time": "0:45:12"}
{"current_steps": 2850, "total_steps": 3750, "loss": 0.3943, "lr": 6.631228680914558e-06, "epoch": 4.5600000000000005, "percentage": 76.0, "elapsed_time": "2:22:19", "remaining_time": "0:44:56"}
{"current_steps": 2855, "total_steps": 3750, "loss": 0.3778, "lr": 6.562140716184515e-06, "epoch": 4.568, "percentage": 76.13, "elapsed_time": "2:22:30", "remaining_time": "0:44:40"}
{"current_steps": 2860, "total_steps": 3750, "loss": 0.3888, "lr": 6.493343837579511e-06, "epoch": 4.576, "percentage": 76.27, "elapsed_time": "2:22:42", "remaining_time": "0:44:24"}
{"current_steps": 2865, "total_steps": 3750, "loss": 0.3947, "lr": 6.424839535353045e-06, "epoch": 4.584, "percentage": 76.4, "elapsed_time": "2:22:54", "remaining_time": "0:44:08"}
{"current_steps": 2870, "total_steps": 3750, "loss": 0.3907, "lr": 6.356629293420926e-06, "epoch": 4.592, "percentage": 76.53, "elapsed_time": "2:23:06", "remaining_time": "0:43:52"}
{"current_steps": 2875, "total_steps": 3750, "loss": 0.4043, "lr": 6.28871458932913e-06, "epoch": 4.6, "percentage": 76.67, "elapsed_time": "2:23:18", "remaining_time": "0:43:36"}
{"current_steps": 2880, "total_steps": 3750, "loss": 0.3974, "lr": 6.2210968942218206e-06, "epoch": 4.608, "percentage": 76.8, "elapsed_time": "2:23:29", "remaining_time": "0:43:20"}
{"current_steps": 2885, "total_steps": 3750, "loss": 0.4084, "lr": 6.153777672809438e-06, "epoch": 4.616, "percentage": 76.93, "elapsed_time": "2:23:41", "remaining_time": "0:43:04"}
{"current_steps": 2890, "total_steps": 3750, "loss": 0.3918, "lr": 6.086758383336984e-06, "epoch": 4.624, "percentage": 77.07, "elapsed_time": "2:23:53", "remaining_time": "0:42:49"}
{"current_steps": 2895, "total_steps": 3750, "loss": 0.3943, "lr": 6.0200404775524715e-06, "epoch": 4.632, "percentage": 77.2, "elapsed_time": "2:24:04", "remaining_time": "0:42:33"}
{"current_steps": 2900, "total_steps": 3750, "loss": 0.3851, "lr": 5.9536254006754155e-06, "epoch": 4.64, "percentage": 77.33, "elapsed_time": "2:24:17", "remaining_time": "0:42:17"}
{"current_steps": 2905, "total_steps": 3750, "loss": 0.3777, "lr": 5.887514591365593e-06, "epoch": 4.648, "percentage": 77.47, "elapsed_time": "2:25:35", "remaining_time": "0:42:20"}
{"current_steps": 2910, "total_steps": 3750, "loss": 0.394, "lr": 5.821709481691798e-06, "epoch": 4.656, "percentage": 77.6, "elapsed_time": "2:25:46", "remaining_time": "0:42:04"}
{"current_steps": 2915, "total_steps": 3750, "loss": 0.405, "lr": 5.75621149710091e-06, "epoch": 4.664, "percentage": 77.73, "elapsed_time": "2:25:58", "remaining_time": "0:41:48"}
{"current_steps": 2920, "total_steps": 3750, "loss": 0.3899, "lr": 5.691022056386961e-06, "epoch": 4.672, "percentage": 77.87, "elapsed_time": "2:26:10", "remaining_time": "0:41:33"}
{"current_steps": 2925, "total_steps": 3750, "loss": 0.3919, "lr": 5.6261425716604136e-06, "epoch": 4.68, "percentage": 78.0, "elapsed_time": "2:26:22", "remaining_time": "0:41:17"}
{"current_steps": 2930, "total_steps": 3750, "loss": 0.3954, "lr": 5.56157444831757e-06, "epoch": 4.688, "percentage": 78.13, "elapsed_time": "2:26:34", "remaining_time": "0:41:01"}
{"current_steps": 2935, "total_steps": 3750, "loss": 0.3752, "lr": 5.4973190850101334e-06, "epoch": 4.696, "percentage": 78.27, "elapsed_time": "2:26:45", "remaining_time": "0:40:45"}
{"current_steps": 2940, "total_steps": 3750, "loss": 0.3892, "lr": 5.433377873614925e-06, "epoch": 4.704, "percentage": 78.4, "elapsed_time": "2:26:57", "remaining_time": "0:40:29"}
{"current_steps": 2945, "total_steps": 3750, "loss": 0.391, "lr": 5.369752199203702e-06, "epoch": 4.712, "percentage": 78.53, "elapsed_time": "2:27:09", "remaining_time": "0:40:13"}
{"current_steps": 2950, "total_steps": 3750, "loss": 0.3853, "lr": 5.306443440013171e-06, "epoch": 4.72, "percentage": 78.67, "elapsed_time": "2:27:21", "remaining_time": "0:39:57"}
{"current_steps": 2955, "total_steps": 3750, "loss": 0.3575, "lr": 5.243452967415155e-06, "epoch": 4.728, "percentage": 78.8, "elapsed_time": "2:27:33", "remaining_time": "0:39:41"}
{"current_steps": 2960, "total_steps": 3750, "loss": 0.4013, "lr": 5.180782145886846e-06, "epoch": 4.736, "percentage": 78.93, "elapsed_time": "2:27:45", "remaining_time": "0:39:26"}
{"current_steps": 2965, "total_steps": 3750, "loss": 0.4111, "lr": 5.118432332981273e-06, "epoch": 4.744, "percentage": 79.07, "elapsed_time": "2:27:57", "remaining_time": "0:39:10"}
{"current_steps": 2970, "total_steps": 3750, "loss": 0.3768, "lr": 5.056404879297887e-06, "epoch": 4.752, "percentage": 79.2, "elapsed_time": "2:28:08", "remaining_time": "0:38:54"}
{"current_steps": 2975, "total_steps": 3750, "loss": 0.3946, "lr": 4.994701128453325e-06, "epoch": 4.76, "percentage": 79.33, "elapsed_time": "2:28:20", "remaining_time": "0:38:38"}
{"current_steps": 2980, "total_steps": 3750, "loss": 0.3851, "lr": 4.933322417052269e-06, "epoch": 4.768, "percentage": 79.47, "elapsed_time": "2:28:32", "remaining_time": "0:38:22"}
{"current_steps": 2985, "total_steps": 3750, "loss": 0.3817, "lr": 4.8722700746585135e-06, "epoch": 4.776, "percentage": 79.6, "elapsed_time": "2:28:43", "remaining_time": "0:38:06"}
{"current_steps": 2990, "total_steps": 3750, "loss": 0.3941, "lr": 4.811545423766184e-06, "epoch": 4.784, "percentage": 79.73, "elapsed_time": "2:28:55", "remaining_time": "0:37:51"}
{"current_steps": 2995, "total_steps": 3750, "loss": 0.3852, "lr": 4.75114977977104e-06, "epoch": 4.792, "percentage": 79.87, "elapsed_time": "2:29:06", "remaining_time": "0:37:35"}
{"current_steps": 3000, "total_steps": 3750, "loss": 0.373, "lr": 4.691084450942047e-06, "epoch": 4.8, "percentage": 80.0, "elapsed_time": "2:29:18", "remaining_time": "0:37:19"}
{"current_steps": 3005, "total_steps": 3750, "loss": 0.3955, "lr": 4.631350738392955e-06, "epoch": 4.808, "percentage": 80.13, "elapsed_time": "2:30:37", "remaining_time": "0:37:20"}
{"current_steps": 3010, "total_steps": 3750, "loss": 0.3794, "lr": 4.571949936054197e-06, "epoch": 4.816, "percentage": 80.27, "elapsed_time": "2:30:49", "remaining_time": "0:37:04"}
{"current_steps": 3015, "total_steps": 3750, "loss": 0.3878, "lr": 4.512883330644815e-06, "epoch": 4.824, "percentage": 80.4, "elapsed_time": "2:31:01", "remaining_time": "0:36:48"}
{"current_steps": 3020, "total_steps": 3750, "loss": 0.4219, "lr": 4.454152201644591e-06, "epoch": 4.832, "percentage": 80.53, "elapsed_time": "2:31:12", "remaining_time": "0:36:33"}
{"current_steps": 3025, "total_steps": 3750, "loss": 0.3975, "lr": 4.395757821266333e-06, "epoch": 4.84, "percentage": 80.67, "elapsed_time": "2:31:24", "remaining_time": "0:36:17"}
{"current_steps": 3030, "total_steps": 3750, "loss": 0.3716, "lr": 4.337701454428322e-06, "epoch": 4.848, "percentage": 80.8, "elapsed_time": "2:31:36", "remaining_time": "0:36:01"}
{"current_steps": 3035, "total_steps": 3750, "loss": 0.4145, "lr": 4.279984358726925e-06, "epoch": 4.856, "percentage": 80.93, "elapsed_time": "2:31:48", "remaining_time": "0:35:45"}
{"current_steps": 3040, "total_steps": 3750, "loss": 0.3811, "lr": 4.2226077844093205e-06, "epoch": 4.864, "percentage": 81.07, "elapsed_time": "2:31:59", "remaining_time": "0:35:29"}
{"current_steps": 3045, "total_steps": 3750, "loss": 0.3861, "lr": 4.165572974346435e-06, "epoch": 4.872, "percentage": 81.2, "elapsed_time": "2:32:10", "remaining_time": "0:35:14"}
{"current_steps": 3050, "total_steps": 3750, "loss": 0.3795, "lr": 4.108881164006033e-06, "epoch": 4.88, "percentage": 81.33, "elapsed_time": "2:32:22", "remaining_time": "0:34:58"}
{"current_steps": 3055, "total_steps": 3750, "loss": 0.3925, "lr": 4.05253358142593e-06, "epoch": 4.888, "percentage": 81.47, "elapsed_time": "2:32:35", "remaining_time": "0:34:42"}
{"current_steps": 3060, "total_steps": 3750, "loss": 0.3874, "lr": 3.9965314471874035e-06, "epoch": 4.896, "percentage": 81.6, "elapsed_time": "2:32:47", "remaining_time": "0:34:27"}