hp_ablations_gemma_lr8e-6 / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 2
f45545e verified
{"current_steps": 10, "total_steps": 1329, "loss": 0.762, "learning_rate": 8e-06, "epoch": 0.022566995768688293, "percentage": 0.75, "elapsed_time": "0:14:46", "remaining_time": "1 day, 8:27:59"}
{"current_steps": 20, "total_steps": 1329, "loss": 0.6779, "learning_rate": 8e-06, "epoch": 0.045133991537376586, "percentage": 1.5, "elapsed_time": "0:29:27", "remaining_time": "1 day, 8:08:24"}
{"current_steps": 30, "total_steps": 1329, "loss": 0.6608, "learning_rate": 8e-06, "epoch": 0.06770098730606489, "percentage": 2.26, "elapsed_time": "0:44:10", "remaining_time": "1 day, 7:52:59"}
{"current_steps": 40, "total_steps": 1329, "loss": 0.6386, "learning_rate": 8e-06, "epoch": 0.09026798307475317, "percentage": 3.01, "elapsed_time": "0:58:53", "remaining_time": "1 day, 7:37:57"}
{"current_steps": 50, "total_steps": 1329, "loss": 0.6327, "learning_rate": 8e-06, "epoch": 0.11283497884344147, "percentage": 3.76, "elapsed_time": "1:13:36", "remaining_time": "1 day, 7:22:55"}
{"current_steps": 60, "total_steps": 1329, "loss": 0.6322, "learning_rate": 8e-06, "epoch": 0.13540197461212977, "percentage": 4.51, "elapsed_time": "1:28:20", "remaining_time": "1 day, 7:08:15"}
{"current_steps": 70, "total_steps": 1329, "loss": 0.6278, "learning_rate": 8e-06, "epoch": 0.15796897038081806, "percentage": 5.27, "elapsed_time": "1:43:02", "remaining_time": "1 day, 6:53:19"}
{"current_steps": 80, "total_steps": 1329, "loss": 0.6298, "learning_rate": 8e-06, "epoch": 0.18053596614950634, "percentage": 6.02, "elapsed_time": "1:57:44", "remaining_time": "1 day, 6:38:08"}
{"current_steps": 90, "total_steps": 1329, "loss": 0.6275, "learning_rate": 8e-06, "epoch": 0.20310296191819463, "percentage": 6.77, "elapsed_time": "2:12:25", "remaining_time": "1 day, 6:23:09"}
{"current_steps": 100, "total_steps": 1329, "loss": 0.6268, "learning_rate": 8e-06, "epoch": 0.22566995768688294, "percentage": 7.52, "elapsed_time": "2:27:07", "remaining_time": "1 day, 6:08:13"}
{"current_steps": 110, "total_steps": 1329, "loss": 0.6174, "learning_rate": 8e-06, "epoch": 0.24823695345557123, "percentage": 8.28, "elapsed_time": "2:41:49", "remaining_time": "1 day, 5:53:23"}
{"current_steps": 120, "total_steps": 1329, "loss": 0.617, "learning_rate": 8e-06, "epoch": 0.27080394922425954, "percentage": 9.03, "elapsed_time": "2:56:32", "remaining_time": "1 day, 5:38:37"}
{"current_steps": 130, "total_steps": 1329, "loss": 0.6182, "learning_rate": 8e-06, "epoch": 0.2933709449929478, "percentage": 9.78, "elapsed_time": "3:11:14", "remaining_time": "1 day, 5:23:49"}
{"current_steps": 140, "total_steps": 1329, "loss": 0.6133, "learning_rate": 8e-06, "epoch": 0.3159379407616361, "percentage": 10.53, "elapsed_time": "3:25:56", "remaining_time": "1 day, 5:09:03"}
{"current_steps": 150, "total_steps": 1329, "loss": 0.6138, "learning_rate": 8e-06, "epoch": 0.3385049365303244, "percentage": 11.29, "elapsed_time": "3:40:38", "remaining_time": "1 day, 4:54:16"}
{"current_steps": 160, "total_steps": 1329, "loss": 0.6159, "learning_rate": 8e-06, "epoch": 0.3610719322990127, "percentage": 12.04, "elapsed_time": "3:55:20", "remaining_time": "1 day, 4:39:29"}
{"current_steps": 170, "total_steps": 1329, "loss": 0.6119, "learning_rate": 8e-06, "epoch": 0.383638928067701, "percentage": 12.79, "elapsed_time": "4:10:02", "remaining_time": "1 day, 4:24:41"}
{"current_steps": 180, "total_steps": 1329, "loss": 0.609, "learning_rate": 8e-06, "epoch": 0.40620592383638926, "percentage": 13.54, "elapsed_time": "4:24:44", "remaining_time": "1 day, 4:09:53"}
{"current_steps": 190, "total_steps": 1329, "loss": 0.6034, "learning_rate": 8e-06, "epoch": 0.4287729196050776, "percentage": 14.3, "elapsed_time": "4:39:24", "remaining_time": "1 day, 3:55:01"}
{"current_steps": 200, "total_steps": 1329, "loss": 0.6102, "learning_rate": 8e-06, "epoch": 0.4513399153737659, "percentage": 15.05, "elapsed_time": "4:54:05", "remaining_time": "1 day, 3:40:10"}
{"current_steps": 210, "total_steps": 1329, "loss": 0.6067, "learning_rate": 8e-06, "epoch": 0.47390691114245415, "percentage": 15.8, "elapsed_time": "5:08:46", "remaining_time": "1 day, 3:25:21"}
{"current_steps": 220, "total_steps": 1329, "loss": 0.5983, "learning_rate": 8e-06, "epoch": 0.49647390691114246, "percentage": 16.55, "elapsed_time": "5:23:28", "remaining_time": "1 day, 3:10:34"}
{"current_steps": 230, "total_steps": 1329, "loss": 0.6073, "learning_rate": 8e-06, "epoch": 0.5190409026798307, "percentage": 17.31, "elapsed_time": "5:38:09", "remaining_time": "1 day, 2:55:48"}
{"current_steps": 240, "total_steps": 1329, "loss": 0.6061, "learning_rate": 8e-06, "epoch": 0.5416078984485191, "percentage": 18.06, "elapsed_time": "5:52:51", "remaining_time": "1 day, 2:41:03"}
{"current_steps": 250, "total_steps": 1329, "loss": 0.6001, "learning_rate": 8e-06, "epoch": 0.5641748942172073, "percentage": 18.81, "elapsed_time": "6:07:33", "remaining_time": "1 day, 2:26:22"}
{"current_steps": 260, "total_steps": 1329, "loss": 0.6026, "learning_rate": 8e-06, "epoch": 0.5867418899858956, "percentage": 19.56, "elapsed_time": "6:22:15", "remaining_time": "1 day, 2:11:40"}
{"current_steps": 270, "total_steps": 1329, "loss": 0.5981, "learning_rate": 8e-06, "epoch": 0.609308885754584, "percentage": 20.32, "elapsed_time": "6:36:57", "remaining_time": "1 day, 1:56:58"}
{"current_steps": 280, "total_steps": 1329, "loss": 0.6048, "learning_rate": 8e-06, "epoch": 0.6318758815232722, "percentage": 21.07, "elapsed_time": "6:51:39", "remaining_time": "1 day, 1:42:13"}
{"current_steps": 290, "total_steps": 1329, "loss": 0.5983, "learning_rate": 8e-06, "epoch": 0.6544428772919605, "percentage": 21.82, "elapsed_time": "7:06:20", "remaining_time": "1 day, 1:27:29"}
{"current_steps": 300, "total_steps": 1329, "loss": 0.6045, "learning_rate": 8e-06, "epoch": 0.6770098730606487, "percentage": 22.57, "elapsed_time": "7:21:02", "remaining_time": "1 day, 1:12:47"}
{"current_steps": 310, "total_steps": 1329, "loss": 0.5977, "learning_rate": 8e-06, "epoch": 0.6995768688293371, "percentage": 23.33, "elapsed_time": "7:35:44", "remaining_time": "1 day, 0:58:04"}
{"current_steps": 320, "total_steps": 1329, "loss": 0.6009, "learning_rate": 8e-06, "epoch": 0.7221438645980254, "percentage": 24.08, "elapsed_time": "7:50:26", "remaining_time": "1 day, 0:43:21"}
{"current_steps": 330, "total_steps": 1329, "loss": 0.5966, "learning_rate": 8e-06, "epoch": 0.7447108603667136, "percentage": 24.83, "elapsed_time": "8:05:08", "remaining_time": "1 day, 0:28:40"}
{"current_steps": 340, "total_steps": 1329, "loss": 0.5961, "learning_rate": 8e-06, "epoch": 0.767277856135402, "percentage": 25.58, "elapsed_time": "8:19:51", "remaining_time": "1 day, 0:13:59"}
{"current_steps": 350, "total_steps": 1329, "loss": 0.5974, "learning_rate": 8e-06, "epoch": 0.7898448519040903, "percentage": 26.34, "elapsed_time": "8:34:32", "remaining_time": "23:59:16"}
{"current_steps": 360, "total_steps": 1329, "loss": 0.5965, "learning_rate": 8e-06, "epoch": 0.8124118476727785, "percentage": 27.09, "elapsed_time": "8:49:15", "remaining_time": "23:44:34"}
{"current_steps": 370, "total_steps": 1329, "loss": 0.6035, "learning_rate": 8e-06, "epoch": 0.8349788434414669, "percentage": 27.84, "elapsed_time": "9:03:58", "remaining_time": "23:29:54"}
{"current_steps": 380, "total_steps": 1329, "loss": 0.6004, "learning_rate": 8e-06, "epoch": 0.8575458392101551, "percentage": 28.59, "elapsed_time": "9:18:40", "remaining_time": "23:15:11"}
{"current_steps": 390, "total_steps": 1329, "loss": 0.5943, "learning_rate": 8e-06, "epoch": 0.8801128349788434, "percentage": 29.35, "elapsed_time": "9:33:22", "remaining_time": "23:00:29"}
{"current_steps": 400, "total_steps": 1329, "loss": 0.5967, "learning_rate": 8e-06, "epoch": 0.9026798307475318, "percentage": 30.1, "elapsed_time": "9:48:03", "remaining_time": "22:45:46"}
{"current_steps": 410, "total_steps": 1329, "loss": 0.5946, "learning_rate": 8e-06, "epoch": 0.92524682651622, "percentage": 30.85, "elapsed_time": "10:02:45", "remaining_time": "22:31:04"}
{"current_steps": 420, "total_steps": 1329, "loss": 0.5952, "learning_rate": 8e-06, "epoch": 0.9478138222849083, "percentage": 31.6, "elapsed_time": "10:17:28", "remaining_time": "22:16:22"}
{"current_steps": 430, "total_steps": 1329, "loss": 0.5895, "learning_rate": 8e-06, "epoch": 0.9703808180535967, "percentage": 32.36, "elapsed_time": "10:32:10", "remaining_time": "22:01:41"}
{"current_steps": 440, "total_steps": 1329, "loss": 0.5945, "learning_rate": 8e-06, "epoch": 0.9929478138222849, "percentage": 33.11, "elapsed_time": "10:46:53", "remaining_time": "21:47:00"}
{"current_steps": 443, "total_steps": 1329, "eval_loss": 0.592284083366394, "epoch": 0.9997179125528914, "percentage": 33.33, "elapsed_time": "11:02:51", "remaining_time": "22:05:43"}
{"current_steps": 450, "total_steps": 1329, "loss": 0.5895, "learning_rate": 8e-06, "epoch": 1.0155148095909732, "percentage": 33.86, "elapsed_time": "11:14:15", "remaining_time": "21:57:03"}
{"current_steps": 460, "total_steps": 1329, "loss": 0.5113, "learning_rate": 8e-06, "epoch": 1.0380818053596614, "percentage": 34.61, "elapsed_time": "11:28:58", "remaining_time": "21:41:33"}
{"current_steps": 470, "total_steps": 1329, "loss": 0.5247, "learning_rate": 8e-06, "epoch": 1.0606488011283497, "percentage": 35.36, "elapsed_time": "11:43:40", "remaining_time": "21:26:05"}
{"current_steps": 480, "total_steps": 1329, "loss": 0.5133, "learning_rate": 8e-06, "epoch": 1.0832157968970382, "percentage": 36.12, "elapsed_time": "11:58:22", "remaining_time": "21:10:38"}
{"current_steps": 490, "total_steps": 1329, "loss": 0.5231, "learning_rate": 8e-06, "epoch": 1.1057827926657264, "percentage": 36.87, "elapsed_time": "12:13:05", "remaining_time": "20:55:13"}
{"current_steps": 500, "total_steps": 1329, "loss": 0.522, "learning_rate": 8e-06, "epoch": 1.1283497884344147, "percentage": 37.62, "elapsed_time": "12:27:47", "remaining_time": "20:39:50"}
{"current_steps": 510, "total_steps": 1329, "loss": 0.5236, "learning_rate": 8e-06, "epoch": 1.150916784203103, "percentage": 38.37, "elapsed_time": "12:42:30", "remaining_time": "20:24:29"}
{"current_steps": 520, "total_steps": 1329, "loss": 0.5166, "learning_rate": 8e-06, "epoch": 1.1734837799717912, "percentage": 39.13, "elapsed_time": "12:57:13", "remaining_time": "20:09:10"}
{"current_steps": 530, "total_steps": 1329, "loss": 0.5229, "learning_rate": 8e-06, "epoch": 1.1960507757404795, "percentage": 39.88, "elapsed_time": "13:11:56", "remaining_time": "19:53:53"}
{"current_steps": 540, "total_steps": 1329, "loss": 0.5207, "learning_rate": 8e-06, "epoch": 1.2186177715091677, "percentage": 40.63, "elapsed_time": "13:26:39", "remaining_time": "19:38:36"}
{"current_steps": 550, "total_steps": 1329, "loss": 0.5256, "learning_rate": 8e-06, "epoch": 1.2411847672778562, "percentage": 41.38, "elapsed_time": "13:41:22", "remaining_time": "19:23:21"}
{"current_steps": 560, "total_steps": 1329, "loss": 0.5274, "learning_rate": 8e-06, "epoch": 1.2637517630465445, "percentage": 42.14, "elapsed_time": "13:56:05", "remaining_time": "19:08:08"}
{"current_steps": 570, "total_steps": 1329, "loss": 0.5251, "learning_rate": 8e-06, "epoch": 1.2863187588152327, "percentage": 42.89, "elapsed_time": "14:10:48", "remaining_time": "18:52:55"}
{"current_steps": 580, "total_steps": 1329, "loss": 0.5253, "learning_rate": 8e-06, "epoch": 1.308885754583921, "percentage": 43.64, "elapsed_time": "14:25:31", "remaining_time": "18:37:43"}
{"current_steps": 590, "total_steps": 1329, "loss": 0.5255, "learning_rate": 8e-06, "epoch": 1.3314527503526092, "percentage": 44.39, "elapsed_time": "14:40:13", "remaining_time": "18:22:30"}
{"current_steps": 600, "total_steps": 1329, "loss": 0.5308, "learning_rate": 8e-06, "epoch": 1.3540197461212977, "percentage": 45.15, "elapsed_time": "14:54:55", "remaining_time": "18:07:19"}
{"current_steps": 610, "total_steps": 1329, "loss": 0.5238, "learning_rate": 8e-06, "epoch": 1.376586741889986, "percentage": 45.9, "elapsed_time": "15:09:37", "remaining_time": "17:52:09"}
{"current_steps": 620, "total_steps": 1329, "loss": 0.5282, "learning_rate": 8e-06, "epoch": 1.3991537376586742, "percentage": 46.65, "elapsed_time": "15:24:19", "remaining_time": "17:37:00"}
{"current_steps": 630, "total_steps": 1329, "loss": 0.5265, "learning_rate": 8e-06, "epoch": 1.4217207334273625, "percentage": 47.4, "elapsed_time": "15:39:01", "remaining_time": "17:21:52"}
{"current_steps": 640, "total_steps": 1329, "loss": 0.5265, "learning_rate": 8e-06, "epoch": 1.4442877291960508, "percentage": 48.16, "elapsed_time": "15:53:44", "remaining_time": "17:06:45"}
{"current_steps": 650, "total_steps": 1329, "loss": 0.5271, "learning_rate": 8e-06, "epoch": 1.466854724964739, "percentage": 48.91, "elapsed_time": "16:08:26", "remaining_time": "16:51:39"}
{"current_steps": 660, "total_steps": 1329, "loss": 0.5313, "learning_rate": 8e-06, "epoch": 1.4894217207334273, "percentage": 49.66, "elapsed_time": "16:23:08", "remaining_time": "16:36:33"}
{"current_steps": 670, "total_steps": 1329, "loss": 0.526, "learning_rate": 8e-06, "epoch": 1.5119887165021155, "percentage": 50.41, "elapsed_time": "16:37:50", "remaining_time": "16:21:27"}
{"current_steps": 680, "total_steps": 1329, "loss": 0.5207, "learning_rate": 8e-06, "epoch": 1.5345557122708038, "percentage": 51.17, "elapsed_time": "16:52:32", "remaining_time": "16:06:23"}
{"current_steps": 690, "total_steps": 1329, "loss": 0.5227, "learning_rate": 8e-06, "epoch": 1.5571227080394923, "percentage": 51.92, "elapsed_time": "17:07:14", "remaining_time": "15:51:19"}
{"current_steps": 700, "total_steps": 1329, "loss": 0.5271, "learning_rate": 8e-06, "epoch": 1.5796897038081805, "percentage": 52.67, "elapsed_time": "17:21:56", "remaining_time": "15:36:15"}
{"current_steps": 710, "total_steps": 1329, "loss": 0.5299, "learning_rate": 8e-06, "epoch": 1.6022566995768688, "percentage": 53.42, "elapsed_time": "17:36:39", "remaining_time": "15:21:13"}
{"current_steps": 720, "total_steps": 1329, "loss": 0.5378, "learning_rate": 8e-06, "epoch": 1.6248236953455573, "percentage": 54.18, "elapsed_time": "17:51:21", "remaining_time": "15:06:11"}
{"current_steps": 730, "total_steps": 1329, "loss": 0.5217, "learning_rate": 8e-06, "epoch": 1.6473906911142455, "percentage": 54.93, "elapsed_time": "18:06:03", "remaining_time": "14:51:09"}
{"current_steps": 740, "total_steps": 1329, "loss": 0.5371, "learning_rate": 8e-06, "epoch": 1.6699576868829338, "percentage": 55.68, "elapsed_time": "18:20:45", "remaining_time": "14:36:08"}
{"current_steps": 750, "total_steps": 1329, "loss": 0.5289, "learning_rate": 8e-06, "epoch": 1.692524682651622, "percentage": 56.43, "elapsed_time": "18:35:28", "remaining_time": "14:21:08"}
{"current_steps": 760, "total_steps": 1329, "loss": 0.5279, "learning_rate": 8e-06, "epoch": 1.7150916784203103, "percentage": 57.19, "elapsed_time": "18:50:10", "remaining_time": "14:06:08"}
{"current_steps": 770, "total_steps": 1329, "loss": 0.5288, "learning_rate": 8e-06, "epoch": 1.7376586741889986, "percentage": 57.94, "elapsed_time": "19:04:52", "remaining_time": "13:51:09"}
{"current_steps": 780, "total_steps": 1329, "loss": 0.5247, "learning_rate": 8e-06, "epoch": 1.7602256699576868, "percentage": 58.69, "elapsed_time": "19:19:35", "remaining_time": "13:36:10"}
{"current_steps": 790, "total_steps": 1329, "loss": 0.5283, "learning_rate": 8e-06, "epoch": 1.782792665726375, "percentage": 59.44, "elapsed_time": "19:34:17", "remaining_time": "13:21:11"}
{"current_steps": 800, "total_steps": 1329, "loss": 0.5237, "learning_rate": 8e-06, "epoch": 1.8053596614950633, "percentage": 60.2, "elapsed_time": "19:48:59", "remaining_time": "13:06:13"}
{"current_steps": 810, "total_steps": 1329, "loss": 0.5342, "learning_rate": 8e-06, "epoch": 1.8279266572637518, "percentage": 60.95, "elapsed_time": "20:03:42", "remaining_time": "12:51:15"}
{"current_steps": 820, "total_steps": 1329, "loss": 0.5219, "learning_rate": 8e-06, "epoch": 1.85049365303244, "percentage": 61.7, "elapsed_time": "20:18:24", "remaining_time": "12:36:18"}
{"current_steps": 830, "total_steps": 1329, "loss": 0.5336, "learning_rate": 8e-06, "epoch": 1.8730606488011283, "percentage": 62.45, "elapsed_time": "20:33:07", "remaining_time": "12:21:21"}
{"current_steps": 840, "total_steps": 1329, "loss": 0.5339, "learning_rate": 8e-06, "epoch": 1.8956276445698168, "percentage": 63.21, "elapsed_time": "20:47:50", "remaining_time": "12:06:25"}
{"current_steps": 850, "total_steps": 1329, "loss": 0.5224, "learning_rate": 8e-06, "epoch": 1.918194640338505, "percentage": 63.96, "elapsed_time": "21:02:33", "remaining_time": "11:51:29"}
{"current_steps": 860, "total_steps": 1329, "loss": 0.5275, "learning_rate": 8e-06, "epoch": 1.9407616361071933, "percentage": 64.71, "elapsed_time": "21:17:16", "remaining_time": "11:36:33"}
{"current_steps": 870, "total_steps": 1329, "loss": 0.5337, "learning_rate": 8e-06, "epoch": 1.9633286318758816, "percentage": 65.46, "elapsed_time": "21:31:59", "remaining_time": "11:21:38"}
{"current_steps": 880, "total_steps": 1329, "loss": 0.5319, "learning_rate": 8e-06, "epoch": 1.9858956276445698, "percentage": 66.22, "elapsed_time": "21:46:41", "remaining_time": "11:06:42"}
{"current_steps": 886, "total_steps": 1329, "eval_loss": 0.5926612615585327, "epoch": 1.9994358251057829, "percentage": 66.67, "elapsed_time": "22:07:15", "remaining_time": "11:03:37"}
{"current_steps": 890, "total_steps": 1329, "loss": 0.5456, "learning_rate": 8e-06, "epoch": 2.008462623413258, "percentage": 66.97, "elapsed_time": "22:14:10", "remaining_time": "10:58:05"}
{"current_steps": 900, "total_steps": 1329, "loss": 0.4427, "learning_rate": 8e-06, "epoch": 2.0310296191819464, "percentage": 67.72, "elapsed_time": "22:28:54", "remaining_time": "10:42:58"}
{"current_steps": 910, "total_steps": 1329, "loss": 0.4433, "learning_rate": 8e-06, "epoch": 2.0535966149506346, "percentage": 68.47, "elapsed_time": "22:43:37", "remaining_time": "10:27:52"}
{"current_steps": 920, "total_steps": 1329, "loss": 0.436, "learning_rate": 8e-06, "epoch": 2.076163610719323, "percentage": 69.22, "elapsed_time": "22:58:21", "remaining_time": "10:12:46"}
{"current_steps": 930, "total_steps": 1329, "loss": 0.4431, "learning_rate": 8e-06, "epoch": 2.098730606488011, "percentage": 69.98, "elapsed_time": "23:13:06", "remaining_time": "9:57:41"}
{"current_steps": 940, "total_steps": 1329, "loss": 0.4477, "learning_rate": 8e-06, "epoch": 2.1212976022566994, "percentage": 70.73, "elapsed_time": "23:27:50", "remaining_time": "9:42:36"}
{"current_steps": 950, "total_steps": 1329, "loss": 0.4495, "learning_rate": 8e-06, "epoch": 2.143864598025388, "percentage": 71.48, "elapsed_time": "23:42:33", "remaining_time": "9:27:31"}
{"current_steps": 960, "total_steps": 1329, "loss": 0.446, "learning_rate": 8e-06, "epoch": 2.1664315937940763, "percentage": 72.23, "elapsed_time": "23:57:17", "remaining_time": "9:12:27"}
{"current_steps": 970, "total_steps": 1329, "loss": 0.4452, "learning_rate": 8e-06, "epoch": 2.1889985895627646, "percentage": 72.99, "elapsed_time": "1 day, 0:12:01", "remaining_time": "8:57:23"}
{"current_steps": 980, "total_steps": 1329, "loss": 0.4411, "learning_rate": 8e-06, "epoch": 2.211565585331453, "percentage": 73.74, "elapsed_time": "1 day, 0:26:45", "remaining_time": "8:42:20"}
{"current_steps": 990, "total_steps": 1329, "loss": 0.4412, "learning_rate": 8e-06, "epoch": 2.234132581100141, "percentage": 74.49, "elapsed_time": "1 day, 0:41:29", "remaining_time": "8:27:17"}
{"current_steps": 1000, "total_steps": 1329, "loss": 0.4491, "learning_rate": 8e-06, "epoch": 2.2566995768688294, "percentage": 75.24, "elapsed_time": "1 day, 0:56:13", "remaining_time": "8:12:15"}
{"current_steps": 1010, "total_steps": 1329, "loss": 0.4552, "learning_rate": 8e-06, "epoch": 2.2792665726375176, "percentage": 76.0, "elapsed_time": "1 day, 1:10:57", "remaining_time": "7:57:13"}
{"current_steps": 1020, "total_steps": 1329, "loss": 0.451, "learning_rate": 8e-06, "epoch": 2.301833568406206, "percentage": 76.75, "elapsed_time": "1 day, 1:25:42", "remaining_time": "7:42:11"}
{"current_steps": 1030, "total_steps": 1329, "loss": 0.455, "learning_rate": 8e-06, "epoch": 2.324400564174894, "percentage": 77.5, "elapsed_time": "1 day, 1:40:25", "remaining_time": "7:27:10"}
{"current_steps": 1040, "total_steps": 1329, "loss": 0.4537, "learning_rate": 8e-06, "epoch": 2.3469675599435824, "percentage": 78.25, "elapsed_time": "1 day, 1:55:07", "remaining_time": "7:12:08"}
{"current_steps": 1050, "total_steps": 1329, "loss": 0.4518, "learning_rate": 8e-06, "epoch": 2.3695345557122707, "percentage": 79.01, "elapsed_time": "1 day, 2:09:51", "remaining_time": "6:57:07"}
{"current_steps": 1060, "total_steps": 1329, "loss": 0.4513, "learning_rate": 8e-06, "epoch": 2.392101551480959, "percentage": 79.76, "elapsed_time": "1 day, 2:24:34", "remaining_time": "6:42:07"}
{"current_steps": 1070, "total_steps": 1329, "loss": 0.4528, "learning_rate": 8e-06, "epoch": 2.414668547249647, "percentage": 80.51, "elapsed_time": "1 day, 2:39:17", "remaining_time": "6:27:07"}
{"current_steps": 1080, "total_steps": 1329, "loss": 0.4555, "learning_rate": 8e-06, "epoch": 2.4372355430183354, "percentage": 81.26, "elapsed_time": "1 day, 2:54:00", "remaining_time": "6:12:07"}
{"current_steps": 1090, "total_steps": 1329, "loss": 0.4543, "learning_rate": 8e-06, "epoch": 2.459802538787024, "percentage": 82.02, "elapsed_time": "1 day, 3:08:43", "remaining_time": "5:57:07"}
{"current_steps": 1100, "total_steps": 1329, "loss": 0.458, "learning_rate": 8e-06, "epoch": 2.4823695345557124, "percentage": 82.77, "elapsed_time": "1 day, 3:23:26", "remaining_time": "5:42:08"}
{"current_steps": 1110, "total_steps": 1329, "loss": 0.4607, "learning_rate": 8e-06, "epoch": 2.5049365303244007, "percentage": 83.52, "elapsed_time": "1 day, 3:38:09", "remaining_time": "5:27:09"}
{"current_steps": 1120, "total_steps": 1329, "loss": 0.4526, "learning_rate": 8e-06, "epoch": 2.527503526093089, "percentage": 84.27, "elapsed_time": "1 day, 3:52:53", "remaining_time": "5:12:10"}
{"current_steps": 1130, "total_steps": 1329, "loss": 0.4574, "learning_rate": 8e-06, "epoch": 2.550070521861777, "percentage": 85.03, "elapsed_time": "1 day, 4:07:36", "remaining_time": "4:57:11"}
{"current_steps": 1140, "total_steps": 1329, "loss": 0.4558, "learning_rate": 8e-06, "epoch": 2.5726375176304654, "percentage": 85.78, "elapsed_time": "1 day, 4:22:19", "remaining_time": "4:42:13"}
{"current_steps": 1150, "total_steps": 1329, "loss": 0.4565, "learning_rate": 8e-06, "epoch": 2.5952045133991537, "percentage": 86.53, "elapsed_time": "1 day, 4:37:02", "remaining_time": "4:27:15"}
{"current_steps": 1160, "total_steps": 1329, "loss": 0.4602, "learning_rate": 8e-06, "epoch": 2.617771509167842, "percentage": 87.28, "elapsed_time": "1 day, 4:51:45", "remaining_time": "4:12:18"}
{"current_steps": 1170, "total_steps": 1329, "loss": 0.4503, "learning_rate": 8e-06, "epoch": 2.64033850493653, "percentage": 88.04, "elapsed_time": "1 day, 5:06:28", "remaining_time": "3:57:20"}
{"current_steps": 1180, "total_steps": 1329, "loss": 0.4604, "learning_rate": 8e-06, "epoch": 2.6629055007052185, "percentage": 88.79, "elapsed_time": "1 day, 5:21:12", "remaining_time": "3:42:23"}
{"current_steps": 1190, "total_steps": 1329, "loss": 0.4611, "learning_rate": 8e-06, "epoch": 2.685472496473907, "percentage": 89.54, "elapsed_time": "1 day, 5:35:55", "remaining_time": "3:27:26"}
{"current_steps": 1200, "total_steps": 1329, "loss": 0.4602, "learning_rate": 8e-06, "epoch": 2.7080394922425954, "percentage": 90.29, "elapsed_time": "1 day, 5:50:39", "remaining_time": "3:12:29"}
{"current_steps": 1210, "total_steps": 1329, "loss": 0.4615, "learning_rate": 8e-06, "epoch": 2.7306064880112837, "percentage": 91.05, "elapsed_time": "1 day, 6:05:23", "remaining_time": "2:57:33"}
{"current_steps": 1220, "total_steps": 1329, "loss": 0.4588, "learning_rate": 8e-06, "epoch": 2.753173483779972, "percentage": 91.8, "elapsed_time": "1 day, 6:20:06", "remaining_time": "2:42:36"}
{"current_steps": 1230, "total_steps": 1329, "loss": 0.4652, "learning_rate": 8e-06, "epoch": 2.77574047954866, "percentage": 92.55, "elapsed_time": "1 day, 6:34:49", "remaining_time": "2:27:40"}
{"current_steps": 1240, "total_steps": 1329, "loss": 0.4633, "learning_rate": 8e-06, "epoch": 2.7983074753173485, "percentage": 93.3, "elapsed_time": "1 day, 6:49:32", "remaining_time": "2:12:44"}
{"current_steps": 1250, "total_steps": 1329, "loss": 0.4637, "learning_rate": 8e-06, "epoch": 2.8208744710860367, "percentage": 94.06, "elapsed_time": "1 day, 7:04:15", "remaining_time": "1:57:49"}
{"current_steps": 1260, "total_steps": 1329, "loss": 0.4654, "learning_rate": 8e-06, "epoch": 2.843441466854725, "percentage": 94.81, "elapsed_time": "1 day, 7:18:58", "remaining_time": "1:42:53"}
{"current_steps": 1270, "total_steps": 1329, "loss": 0.4629, "learning_rate": 8e-06, "epoch": 2.8660084626234132, "percentage": 95.56, "elapsed_time": "1 day, 7:33:41", "remaining_time": "1:27:58"}
{"current_steps": 1280, "total_steps": 1329, "loss": 0.4613, "learning_rate": 8e-06, "epoch": 2.8885754583921015, "percentage": 96.31, "elapsed_time": "1 day, 7:48:24", "remaining_time": "1:13:03"}
{"current_steps": 1290, "total_steps": 1329, "loss": 0.4645, "learning_rate": 8e-06, "epoch": 2.9111424541607898, "percentage": 97.07, "elapsed_time": "1 day, 8:03:07", "remaining_time": "0:58:08"}
{"current_steps": 1300, "total_steps": 1329, "loss": 0.4607, "learning_rate": 8e-06, "epoch": 2.933709449929478, "percentage": 97.82, "elapsed_time": "1 day, 8:17:51", "remaining_time": "0:43:13"}
{"current_steps": 1310, "total_steps": 1329, "loss": 0.4645, "learning_rate": 8e-06, "epoch": 2.9562764456981663, "percentage": 98.57, "elapsed_time": "1 day, 8:32:34", "remaining_time": "0:28:19"}
{"current_steps": 1320, "total_steps": 1329, "loss": 0.4618, "learning_rate": 8e-06, "epoch": 2.9788434414668545, "percentage": 99.32, "elapsed_time": "1 day, 8:47:16", "remaining_time": "0:13:24"}
{"current_steps": 1329, "total_steps": 1329, "eval_loss": 0.620595395565033, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:13:20", "remaining_time": "0:00:00"}
{"current_steps": 1329, "total_steps": 1329, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:15:01", "remaining_time": "0:00:00"}