hp_ablations_gemma_lr1e-5 / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 2
a86f292 verified
{"current_steps": 10, "total_steps": 1329, "loss": 0.7777, "learning_rate": 1e-05, "epoch": 0.022566995768688293, "percentage": 0.75, "elapsed_time": "0:14:48", "remaining_time": "1 day, 8:32:14"}
{"current_steps": 20, "total_steps": 1329, "loss": 0.6809, "learning_rate": 1e-05, "epoch": 0.045133991537376586, "percentage": 1.5, "elapsed_time": "0:29:32", "remaining_time": "1 day, 8:13:28"}
{"current_steps": 30, "total_steps": 1329, "loss": 0.6629, "learning_rate": 1e-05, "epoch": 0.06770098730606489, "percentage": 2.26, "elapsed_time": "0:44:16", "remaining_time": "1 day, 7:56:55"}
{"current_steps": 40, "total_steps": 1329, "loss": 0.64, "learning_rate": 1e-05, "epoch": 0.09026798307475317, "percentage": 3.01, "elapsed_time": "0:59:00", "remaining_time": "1 day, 7:41:28"}
{"current_steps": 50, "total_steps": 1329, "loss": 0.6341, "learning_rate": 1e-05, "epoch": 0.11283497884344147, "percentage": 3.76, "elapsed_time": "1:13:44", "remaining_time": "1 day, 7:26:23"}
{"current_steps": 60, "total_steps": 1329, "loss": 0.6334, "learning_rate": 1e-05, "epoch": 0.13540197461212977, "percentage": 4.51, "elapsed_time": "1:28:28", "remaining_time": "1 day, 7:11:23"}
{"current_steps": 70, "total_steps": 1329, "loss": 0.6289, "learning_rate": 1e-05, "epoch": 0.15796897038081806, "percentage": 5.27, "elapsed_time": "1:43:13", "remaining_time": "1 day, 6:56:34"}
{"current_steps": 80, "total_steps": 1329, "loss": 0.6309, "learning_rate": 1e-05, "epoch": 0.18053596614950634, "percentage": 6.02, "elapsed_time": "1:57:57", "remaining_time": "1 day, 6:41:34"}
{"current_steps": 90, "total_steps": 1329, "loss": 0.6284, "learning_rate": 1e-05, "epoch": 0.20310296191819463, "percentage": 6.77, "elapsed_time": "2:12:41", "remaining_time": "1 day, 6:26:48"}
{"current_steps": 100, "total_steps": 1329, "loss": 0.6278, "learning_rate": 1e-05, "epoch": 0.22566995768688294, "percentage": 7.52, "elapsed_time": "2:27:26", "remaining_time": "1 day, 6:12:05"}
{"current_steps": 110, "total_steps": 1329, "loss": 0.6186, "learning_rate": 1e-05, "epoch": 0.24823695345557123, "percentage": 8.28, "elapsed_time": "2:42:11", "remaining_time": "1 day, 5:57:19"}
{"current_steps": 120, "total_steps": 1329, "loss": 0.6183, "learning_rate": 1e-05, "epoch": 0.27080394922425954, "percentage": 9.03, "elapsed_time": "2:56:55", "remaining_time": "1 day, 5:42:30"}
{"current_steps": 130, "total_steps": 1329, "loss": 0.6195, "learning_rate": 1e-05, "epoch": 0.2933709449929478, "percentage": 9.78, "elapsed_time": "3:11:39", "remaining_time": "1 day, 5:27:42"}
{"current_steps": 140, "total_steps": 1329, "loss": 0.6145, "learning_rate": 1e-05, "epoch": 0.3159379407616361, "percentage": 10.53, "elapsed_time": "3:26:23", "remaining_time": "1 day, 5:12:55"}
{"current_steps": 150, "total_steps": 1329, "loss": 0.6151, "learning_rate": 1e-05, "epoch": 0.3385049365303244, "percentage": 11.29, "elapsed_time": "3:41:08", "remaining_time": "1 day, 4:58:09"}
{"current_steps": 160, "total_steps": 1329, "loss": 0.617, "learning_rate": 1e-05, "epoch": 0.3610719322990127, "percentage": 12.04, "elapsed_time": "3:55:52", "remaining_time": "1 day, 4:43:22"}
{"current_steps": 170, "total_steps": 1329, "loss": 0.613, "learning_rate": 1e-05, "epoch": 0.383638928067701, "percentage": 12.79, "elapsed_time": "4:10:36", "remaining_time": "1 day, 4:28:31"}
{"current_steps": 180, "total_steps": 1329, "loss": 0.6102, "learning_rate": 1e-05, "epoch": 0.40620592383638926, "percentage": 13.54, "elapsed_time": "4:25:20", "remaining_time": "1 day, 4:13:44"}
{"current_steps": 190, "total_steps": 1329, "loss": 0.6046, "learning_rate": 1e-05, "epoch": 0.4287729196050776, "percentage": 14.3, "elapsed_time": "4:40:04", "remaining_time": "1 day, 3:58:56"}
{"current_steps": 200, "total_steps": 1329, "loss": 0.6113, "learning_rate": 1e-05, "epoch": 0.4513399153737659, "percentage": 15.05, "elapsed_time": "4:54:48", "remaining_time": "1 day, 3:44:10"}
{"current_steps": 210, "total_steps": 1329, "loss": 0.6078, "learning_rate": 1e-05, "epoch": 0.47390691114245415, "percentage": 15.8, "elapsed_time": "5:09:33", "remaining_time": "1 day, 3:29:29"}
{"current_steps": 220, "total_steps": 1329, "loss": 0.5996, "learning_rate": 1e-05, "epoch": 0.49647390691114246, "percentage": 16.55, "elapsed_time": "5:24:17", "remaining_time": "1 day, 3:14:42"}
{"current_steps": 230, "total_steps": 1329, "loss": 0.6085, "learning_rate": 1e-05, "epoch": 0.5190409026798307, "percentage": 17.31, "elapsed_time": "5:39:01", "remaining_time": "1 day, 2:59:55"}
{"current_steps": 240, "total_steps": 1329, "loss": 0.6073, "learning_rate": 1e-05, "epoch": 0.5416078984485191, "percentage": 18.06, "elapsed_time": "5:53:45", "remaining_time": "1 day, 2:45:09"}
{"current_steps": 250, "total_steps": 1329, "loss": 0.6014, "learning_rate": 1e-05, "epoch": 0.5641748942172073, "percentage": 18.81, "elapsed_time": "6:08:29", "remaining_time": "1 day, 2:30:25"}
{"current_steps": 260, "total_steps": 1329, "loss": 0.604, "learning_rate": 1e-05, "epoch": 0.5867418899858956, "percentage": 19.56, "elapsed_time": "6:23:13", "remaining_time": "1 day, 2:15:40"}
{"current_steps": 270, "total_steps": 1329, "loss": 0.5994, "learning_rate": 1e-05, "epoch": 0.609308885754584, "percentage": 20.32, "elapsed_time": "6:37:58", "remaining_time": "1 day, 2:00:55"}
{"current_steps": 280, "total_steps": 1329, "loss": 0.6062, "learning_rate": 1e-05, "epoch": 0.6318758815232722, "percentage": 21.07, "elapsed_time": "6:52:42", "remaining_time": "1 day, 1:46:10"}
{"current_steps": 290, "total_steps": 1329, "loss": 0.5997, "learning_rate": 1e-05, "epoch": 0.6544428772919605, "percentage": 21.82, "elapsed_time": "7:07:26", "remaining_time": "1 day, 1:31:26"}
{"current_steps": 300, "total_steps": 1329, "loss": 0.6057, "learning_rate": 1e-05, "epoch": 0.6770098730606487, "percentage": 22.57, "elapsed_time": "7:22:11", "remaining_time": "1 day, 1:16:41"}
{"current_steps": 310, "total_steps": 1329, "loss": 0.5989, "learning_rate": 1e-05, "epoch": 0.6995768688293371, "percentage": 23.33, "elapsed_time": "7:36:55", "remaining_time": "1 day, 1:01:57"}
{"current_steps": 320, "total_steps": 1329, "loss": 0.6024, "learning_rate": 1e-05, "epoch": 0.7221438645980254, "percentage": 24.08, "elapsed_time": "7:51:40", "remaining_time": "1 day, 0:47:13"}
{"current_steps": 330, "total_steps": 1329, "loss": 0.5981, "learning_rate": 1e-05, "epoch": 0.7447108603667136, "percentage": 24.83, "elapsed_time": "8:06:24", "remaining_time": "1 day, 0:32:29"}
{"current_steps": 340, "total_steps": 1329, "loss": 0.5975, "learning_rate": 1e-05, "epoch": 0.767277856135402, "percentage": 25.58, "elapsed_time": "8:21:09", "remaining_time": "1 day, 0:17:46"}
{"current_steps": 350, "total_steps": 1329, "loss": 0.5988, "learning_rate": 1e-05, "epoch": 0.7898448519040903, "percentage": 26.34, "elapsed_time": "8:35:53", "remaining_time": "1 day, 0:03:02"}
{"current_steps": 360, "total_steps": 1329, "loss": 0.598, "learning_rate": 1e-05, "epoch": 0.8124118476727785, "percentage": 27.09, "elapsed_time": "8:50:38", "remaining_time": "23:48:18"}
{"current_steps": 370, "total_steps": 1329, "loss": 0.6051, "learning_rate": 1e-05, "epoch": 0.8349788434414669, "percentage": 27.84, "elapsed_time": "9:05:23", "remaining_time": "23:33:35"}
{"current_steps": 380, "total_steps": 1329, "loss": 0.6019, "learning_rate": 1e-05, "epoch": 0.8575458392101551, "percentage": 28.59, "elapsed_time": "9:20:08", "remaining_time": "23:18:52"}
{"current_steps": 390, "total_steps": 1329, "loss": 0.5957, "learning_rate": 1e-05, "epoch": 0.8801128349788434, "percentage": 29.35, "elapsed_time": "9:34:52", "remaining_time": "23:04:08"}
{"current_steps": 400, "total_steps": 1329, "loss": 0.5981, "learning_rate": 1e-05, "epoch": 0.9026798307475318, "percentage": 30.1, "elapsed_time": "9:49:37", "remaining_time": "22:49:24"}
{"current_steps": 410, "total_steps": 1329, "loss": 0.596, "learning_rate": 1e-05, "epoch": 0.92524682651622, "percentage": 30.85, "elapsed_time": "10:04:22", "remaining_time": "22:34:40"}
{"current_steps": 420, "total_steps": 1329, "loss": 0.5966, "learning_rate": 1e-05, "epoch": 0.9478138222849083, "percentage": 31.6, "elapsed_time": "10:19:06", "remaining_time": "22:19:55"}
{"current_steps": 430, "total_steps": 1329, "loss": 0.591, "learning_rate": 1e-05, "epoch": 0.9703808180535967, "percentage": 32.36, "elapsed_time": "10:33:50", "remaining_time": "22:05:11"}
{"current_steps": 440, "total_steps": 1329, "loss": 0.5956, "learning_rate": 1e-05, "epoch": 0.9929478138222849, "percentage": 33.11, "elapsed_time": "10:48:35", "remaining_time": "21:50:26"}
{"current_steps": 443, "total_steps": 1329, "eval_loss": 0.5935375690460205, "epoch": 0.9997179125528914, "percentage": 33.33, "elapsed_time": "11:04:35", "remaining_time": "22:09:11"}
{"current_steps": 450, "total_steps": 1329, "loss": 0.584, "learning_rate": 1e-05, "epoch": 1.0155148095909732, "percentage": 33.86, "elapsed_time": "11:16:00", "remaining_time": "22:00:28"}
{"current_steps": 460, "total_steps": 1329, "loss": 0.5023, "learning_rate": 1e-05, "epoch": 1.0380818053596614, "percentage": 34.61, "elapsed_time": "11:30:46", "remaining_time": "21:44:57"}
{"current_steps": 470, "total_steps": 1329, "loss": 0.5159, "learning_rate": 1e-05, "epoch": 1.0606488011283497, "percentage": 35.36, "elapsed_time": "11:45:32", "remaining_time": "21:29:29"}
{"current_steps": 480, "total_steps": 1329, "loss": 0.5047, "learning_rate": 1e-05, "epoch": 1.0832157968970382, "percentage": 36.12, "elapsed_time": "12:00:18", "remaining_time": "21:14:02"}
{"current_steps": 490, "total_steps": 1329, "loss": 0.5144, "learning_rate": 1e-05, "epoch": 1.1057827926657264, "percentage": 36.87, "elapsed_time": "12:15:03", "remaining_time": "20:58:36"}
{"current_steps": 500, "total_steps": 1329, "loss": 0.5134, "learning_rate": 1e-05, "epoch": 1.1283497884344147, "percentage": 37.62, "elapsed_time": "12:29:49", "remaining_time": "20:43:12"}
{"current_steps": 510, "total_steps": 1329, "loss": 0.5156, "learning_rate": 1e-05, "epoch": 1.150916784203103, "percentage": 38.37, "elapsed_time": "12:44:35", "remaining_time": "20:27:50"}
{"current_steps": 520, "total_steps": 1329, "loss": 0.5087, "learning_rate": 1e-05, "epoch": 1.1734837799717912, "percentage": 39.13, "elapsed_time": "12:59:21", "remaining_time": "20:12:30"}
{"current_steps": 530, "total_steps": 1329, "loss": 0.5151, "learning_rate": 1e-05, "epoch": 1.1960507757404795, "percentage": 39.88, "elapsed_time": "13:14:07", "remaining_time": "19:57:10"}
{"current_steps": 540, "total_steps": 1329, "loss": 0.5131, "learning_rate": 1e-05, "epoch": 1.2186177715091677, "percentage": 40.63, "elapsed_time": "13:28:52", "remaining_time": "19:41:51"}
{"current_steps": 550, "total_steps": 1329, "loss": 0.5183, "learning_rate": 1e-05, "epoch": 1.2411847672778562, "percentage": 41.38, "elapsed_time": "13:43:38", "remaining_time": "19:26:34"}
{"current_steps": 560, "total_steps": 1329, "loss": 0.5202, "learning_rate": 1e-05, "epoch": 1.2637517630465445, "percentage": 42.14, "elapsed_time": "13:58:23", "remaining_time": "19:11:17"}
{"current_steps": 570, "total_steps": 1329, "loss": 0.5176, "learning_rate": 1e-05, "epoch": 1.2863187588152327, "percentage": 42.89, "elapsed_time": "14:13:09", "remaining_time": "18:56:02"}
{"current_steps": 580, "total_steps": 1329, "loss": 0.5183, "learning_rate": 1e-05, "epoch": 1.308885754583921, "percentage": 43.64, "elapsed_time": "14:27:54", "remaining_time": "18:40:47"}
{"current_steps": 590, "total_steps": 1329, "loss": 0.5188, "learning_rate": 1e-05, "epoch": 1.3314527503526092, "percentage": 44.39, "elapsed_time": "14:42:39", "remaining_time": "18:25:34"}
{"current_steps": 600, "total_steps": 1329, "loss": 0.5239, "learning_rate": 1e-05, "epoch": 1.3540197461212977, "percentage": 45.15, "elapsed_time": "14:57:24", "remaining_time": "18:10:21"}
{"current_steps": 610, "total_steps": 1329, "loss": 0.517, "learning_rate": 1e-05, "epoch": 1.376586741889986, "percentage": 45.9, "elapsed_time": "15:12:10", "remaining_time": "17:55:10"}
{"current_steps": 620, "total_steps": 1329, "loss": 0.5216, "learning_rate": 1e-05, "epoch": 1.3991537376586742, "percentage": 46.65, "elapsed_time": "15:26:56", "remaining_time": "17:39:59"}
{"current_steps": 630, "total_steps": 1329, "loss": 0.5198, "learning_rate": 1e-05, "epoch": 1.4217207334273625, "percentage": 47.4, "elapsed_time": "15:41:41", "remaining_time": "17:24:50"}
{"current_steps": 640, "total_steps": 1329, "loss": 0.5199, "learning_rate": 1e-05, "epoch": 1.4442877291960508, "percentage": 48.16, "elapsed_time": "15:56:27", "remaining_time": "17:09:41"}
{"current_steps": 650, "total_steps": 1329, "loss": 0.5207, "learning_rate": 1e-05, "epoch": 1.466854724964739, "percentage": 48.91, "elapsed_time": "16:11:13", "remaining_time": "16:54:33"}
{"current_steps": 660, "total_steps": 1329, "loss": 0.5249, "learning_rate": 1e-05, "epoch": 1.4894217207334273, "percentage": 49.66, "elapsed_time": "16:25:58", "remaining_time": "16:39:24"}
{"current_steps": 670, "total_steps": 1329, "loss": 0.5197, "learning_rate": 1e-05, "epoch": 1.5119887165021155, "percentage": 50.41, "elapsed_time": "16:40:43", "remaining_time": "16:24:17"}
{"current_steps": 680, "total_steps": 1329, "loss": 0.5148, "learning_rate": 1e-05, "epoch": 1.5345557122708038, "percentage": 51.17, "elapsed_time": "16:55:28", "remaining_time": "16:09:10"}
{"current_steps": 690, "total_steps": 1329, "loss": 0.5168, "learning_rate": 1e-05, "epoch": 1.5571227080394923, "percentage": 51.92, "elapsed_time": "17:10:13", "remaining_time": "15:54:05"}
{"current_steps": 700, "total_steps": 1329, "loss": 0.5213, "learning_rate": 1e-05, "epoch": 1.5796897038081805, "percentage": 52.67, "elapsed_time": "17:24:59", "remaining_time": "15:38:59"}
{"current_steps": 710, "total_steps": 1329, "loss": 0.524, "learning_rate": 1e-05, "epoch": 1.6022566995768688, "percentage": 53.42, "elapsed_time": "17:39:44", "remaining_time": "15:23:55"}
{"current_steps": 720, "total_steps": 1329, "loss": 0.5317, "learning_rate": 1e-05, "epoch": 1.6248236953455573, "percentage": 54.18, "elapsed_time": "17:54:29", "remaining_time": "15:08:50"}
{"current_steps": 730, "total_steps": 1329, "loss": 0.5158, "learning_rate": 1e-05, "epoch": 1.6473906911142455, "percentage": 54.93, "elapsed_time": "18:09:14", "remaining_time": "14:53:46"}
{"current_steps": 740, "total_steps": 1329, "loss": 0.5311, "learning_rate": 1e-05, "epoch": 1.6699576868829338, "percentage": 55.68, "elapsed_time": "18:24:00", "remaining_time": "14:38:43"}
{"current_steps": 750, "total_steps": 1329, "loss": 0.5228, "learning_rate": 1e-05, "epoch": 1.692524682651622, "percentage": 56.43, "elapsed_time": "18:38:45", "remaining_time": "14:23:40"}
{"current_steps": 760, "total_steps": 1329, "loss": 0.5222, "learning_rate": 1e-05, "epoch": 1.7150916784203103, "percentage": 57.19, "elapsed_time": "18:53:30", "remaining_time": "14:08:38"}
{"current_steps": 770, "total_steps": 1329, "loss": 0.5231, "learning_rate": 1e-05, "epoch": 1.7376586741889986, "percentage": 57.94, "elapsed_time": "19:08:15", "remaining_time": "13:53:36"}
{"current_steps": 780, "total_steps": 1329, "loss": 0.5189, "learning_rate": 1e-05, "epoch": 1.7602256699576868, "percentage": 58.69, "elapsed_time": "19:23:00", "remaining_time": "13:38:34"}
{"current_steps": 790, "total_steps": 1329, "loss": 0.5224, "learning_rate": 1e-05, "epoch": 1.782792665726375, "percentage": 59.44, "elapsed_time": "19:37:45", "remaining_time": "13:23:33"}
{"current_steps": 800, "total_steps": 1329, "loss": 0.518, "learning_rate": 1e-05, "epoch": 1.8053596614950633, "percentage": 60.2, "elapsed_time": "19:52:30", "remaining_time": "13:08:33"}
{"current_steps": 810, "total_steps": 1329, "loss": 0.5287, "learning_rate": 1e-05, "epoch": 1.8279266572637518, "percentage": 60.95, "elapsed_time": "20:07:16", "remaining_time": "12:53:32"}
{"current_steps": 820, "total_steps": 1329, "loss": 0.5166, "learning_rate": 1e-05, "epoch": 1.85049365303244, "percentage": 61.7, "elapsed_time": "20:22:02", "remaining_time": "12:38:33"}
{"current_steps": 830, "total_steps": 1329, "loss": 0.5281, "learning_rate": 1e-05, "epoch": 1.8730606488011283, "percentage": 62.45, "elapsed_time": "20:36:47", "remaining_time": "12:23:33"}
{"current_steps": 840, "total_steps": 1329, "loss": 0.5285, "learning_rate": 1e-05, "epoch": 1.8956276445698168, "percentage": 63.21, "elapsed_time": "20:51:33", "remaining_time": "12:08:35"}
{"current_steps": 850, "total_steps": 1329, "loss": 0.517, "learning_rate": 1e-05, "epoch": 1.918194640338505, "percentage": 63.96, "elapsed_time": "21:06:18", "remaining_time": "11:53:36"}
{"current_steps": 860, "total_steps": 1329, "loss": 0.522, "learning_rate": 1e-05, "epoch": 1.9407616361071933, "percentage": 64.71, "elapsed_time": "21:21:03", "remaining_time": "11:38:37"}
{"current_steps": 870, "total_steps": 1329, "loss": 0.5283, "learning_rate": 1e-05, "epoch": 1.9633286318758816, "percentage": 65.46, "elapsed_time": "21:35:48", "remaining_time": "11:23:39"}
{"current_steps": 880, "total_steps": 1329, "loss": 0.5266, "learning_rate": 1e-05, "epoch": 1.9858956276445698, "percentage": 66.22, "elapsed_time": "21:50:34", "remaining_time": "11:08:41"}
{"current_steps": 886, "total_steps": 1329, "eval_loss": 0.5968770980834961, "epoch": 1.9994358251057829, "percentage": 66.67, "elapsed_time": "22:11:10", "remaining_time": "11:05:35"}
{"current_steps": 890, "total_steps": 1329, "loss": 0.5349, "learning_rate": 1e-05, "epoch": 2.008462623413258, "percentage": 66.97, "elapsed_time": "22:18:07", "remaining_time": "11:00:02"}
{"current_steps": 900, "total_steps": 1329, "loss": 0.424, "learning_rate": 1e-05, "epoch": 2.0310296191819464, "percentage": 67.72, "elapsed_time": "22:32:52", "remaining_time": "10:44:52"}
{"current_steps": 910, "total_steps": 1329, "loss": 0.4237, "learning_rate": 1e-05, "epoch": 2.0535966149506346, "percentage": 68.47, "elapsed_time": "22:47:37", "remaining_time": "10:29:42"}
{"current_steps": 920, "total_steps": 1329, "loss": 0.4165, "learning_rate": 1e-05, "epoch": 2.076163610719323, "percentage": 69.22, "elapsed_time": "23:02:22", "remaining_time": "10:14:33"}
{"current_steps": 930, "total_steps": 1329, "loss": 0.4236, "learning_rate": 1e-05, "epoch": 2.098730606488011, "percentage": 69.98, "elapsed_time": "23:17:07", "remaining_time": "9:59:24"}
{"current_steps": 940, "total_steps": 1329, "loss": 0.4282, "learning_rate": 1e-05, "epoch": 2.1212976022566994, "percentage": 70.73, "elapsed_time": "23:31:52", "remaining_time": "9:44:16"}
{"current_steps": 950, "total_steps": 1329, "loss": 0.4303, "learning_rate": 1e-05, "epoch": 2.143864598025388, "percentage": 71.48, "elapsed_time": "23:46:37", "remaining_time": "9:29:08"}
{"current_steps": 960, "total_steps": 1329, "loss": 0.4271, "learning_rate": 1e-05, "epoch": 2.1664315937940763, "percentage": 72.23, "elapsed_time": "1 day, 0:01:22", "remaining_time": "9:14:01"}
{"current_steps": 970, "total_steps": 1329, "loss": 0.4268, "learning_rate": 1e-05, "epoch": 2.1889985895627646, "percentage": 72.99, "elapsed_time": "1 day, 0:16:07", "remaining_time": "8:58:55"}
{"current_steps": 980, "total_steps": 1329, "loss": 0.4228, "learning_rate": 1e-05, "epoch": 2.211565585331453, "percentage": 73.74, "elapsed_time": "1 day, 0:30:53", "remaining_time": "8:43:49"}
{"current_steps": 990, "total_steps": 1329, "loss": 0.4235, "learning_rate": 1e-05, "epoch": 2.234132581100141, "percentage": 74.49, "elapsed_time": "1 day, 0:45:38", "remaining_time": "8:28:43"}
{"current_steps": 1000, "total_steps": 1329, "loss": 0.4311, "learning_rate": 1e-05, "epoch": 2.2566995768688294, "percentage": 75.24, "elapsed_time": "1 day, 1:00:23", "remaining_time": "8:13:37"}
{"current_steps": 1010, "total_steps": 1329, "loss": 0.4372, "learning_rate": 1e-05, "epoch": 2.2792665726375176, "percentage": 76.0, "elapsed_time": "1 day, 1:15:09", "remaining_time": "7:58:32"}
{"current_steps": 1020, "total_steps": 1329, "loss": 0.4334, "learning_rate": 1e-05, "epoch": 2.301833568406206, "percentage": 76.75, "elapsed_time": "1 day, 1:29:54", "remaining_time": "7:43:28"}
{"current_steps": 1030, "total_steps": 1329, "loss": 0.4375, "learning_rate": 1e-05, "epoch": 2.324400564174894, "percentage": 77.5, "elapsed_time": "1 day, 1:44:39", "remaining_time": "7:28:24"}
{"current_steps": 1040, "total_steps": 1329, "loss": 0.4363, "learning_rate": 1e-05, "epoch": 2.3469675599435824, "percentage": 78.25, "elapsed_time": "1 day, 1:59:24", "remaining_time": "7:13:20"}
{"current_steps": 1050, "total_steps": 1329, "loss": 0.4348, "learning_rate": 1e-05, "epoch": 2.3695345557122707, "percentage": 79.01, "elapsed_time": "1 day, 2:14:09", "remaining_time": "6:58:16"}
{"current_steps": 1060, "total_steps": 1329, "loss": 0.4342, "learning_rate": 1e-05, "epoch": 2.392101551480959, "percentage": 79.76, "elapsed_time": "1 day, 2:28:55", "remaining_time": "6:43:13"}
{"current_steps": 1070, "total_steps": 1329, "loss": 0.4358, "learning_rate": 1e-05, "epoch": 2.414668547249647, "percentage": 80.51, "elapsed_time": "1 day, 2:43:40", "remaining_time": "6:28:10"}
{"current_steps": 1080, "total_steps": 1329, "loss": 0.4385, "learning_rate": 1e-05, "epoch": 2.4372355430183354, "percentage": 81.26, "elapsed_time": "1 day, 2:58:25", "remaining_time": "6:13:08"}
{"current_steps": 1090, "total_steps": 1329, "loss": 0.4376, "learning_rate": 1e-05, "epoch": 2.459802538787024, "percentage": 82.02, "elapsed_time": "1 day, 3:13:11", "remaining_time": "5:58:06"}
{"current_steps": 1100, "total_steps": 1329, "loss": 0.4416, "learning_rate": 1e-05, "epoch": 2.4823695345557124, "percentage": 82.77, "elapsed_time": "1 day, 3:27:56", "remaining_time": "5:43:04"}
{"current_steps": 1110, "total_steps": 1329, "loss": 0.4442, "learning_rate": 1e-05, "epoch": 2.5049365303244007, "percentage": 83.52, "elapsed_time": "1 day, 3:42:41", "remaining_time": "5:28:02"}
{"current_steps": 1120, "total_steps": 1329, "loss": 0.4367, "learning_rate": 1e-05, "epoch": 2.527503526093089, "percentage": 84.27, "elapsed_time": "1 day, 3:57:26", "remaining_time": "5:13:01"}
{"current_steps": 1130, "total_steps": 1329, "loss": 0.4413, "learning_rate": 1e-05, "epoch": 2.550070521861777, "percentage": 85.03, "elapsed_time": "1 day, 4:12:11", "remaining_time": "4:58:00"}
{"current_steps": 1140, "total_steps": 1329, "loss": 0.4398, "learning_rate": 1e-05, "epoch": 2.5726375176304654, "percentage": 85.78, "elapsed_time": "1 day, 4:26:57", "remaining_time": "4:42:59"}
{"current_steps": 1150, "total_steps": 1329, "loss": 0.4405, "learning_rate": 1e-05, "epoch": 2.5952045133991537, "percentage": 86.53, "elapsed_time": "1 day, 4:41:42", "remaining_time": "4:27:59"}
{"current_steps": 1160, "total_steps": 1329, "loss": 0.4442, "learning_rate": 1e-05, "epoch": 2.617771509167842, "percentage": 87.28, "elapsed_time": "1 day, 4:56:27", "remaining_time": "4:12:59"}
{"current_steps": 1170, "total_steps": 1329, "loss": 0.435, "learning_rate": 1e-05, "epoch": 2.64033850493653, "percentage": 88.04, "elapsed_time": "1 day, 5:11:12", "remaining_time": "3:57:59"}
{"current_steps": 1180, "total_steps": 1329, "loss": 0.4449, "learning_rate": 1e-05, "epoch": 2.6629055007052185, "percentage": 88.79, "elapsed_time": "1 day, 5:25:58", "remaining_time": "3:42:59"}
{"current_steps": 1190, "total_steps": 1329, "loss": 0.4456, "learning_rate": 1e-05, "epoch": 2.685472496473907, "percentage": 89.54, "elapsed_time": "1 day, 5:40:43", "remaining_time": "3:27:59"}
{"current_steps": 1200, "total_steps": 1329, "loss": 0.4447, "learning_rate": 1e-05, "epoch": 2.7080394922425954, "percentage": 90.29, "elapsed_time": "1 day, 5:55:27", "remaining_time": "3:13:00"}
{"current_steps": 1210, "total_steps": 1329, "loss": 0.4459, "learning_rate": 1e-05, "epoch": 2.7306064880112837, "percentage": 91.05, "elapsed_time": "1 day, 6:10:12", "remaining_time": "2:58:01"}
{"current_steps": 1220, "total_steps": 1329, "loss": 0.4434, "learning_rate": 1e-05, "epoch": 2.753173483779972, "percentage": 91.8, "elapsed_time": "1 day, 6:24:57", "remaining_time": "2:43:02"}
{"current_steps": 1230, "total_steps": 1329, "loss": 0.4496, "learning_rate": 1e-05, "epoch": 2.77574047954866, "percentage": 92.55, "elapsed_time": "1 day, 6:39:42", "remaining_time": "2:28:04"}
{"current_steps": 1240, "total_steps": 1329, "loss": 0.4476, "learning_rate": 1e-05, "epoch": 2.7983074753173485, "percentage": 93.3, "elapsed_time": "1 day, 6:54:27", "remaining_time": "2:13:06"}
{"current_steps": 1250, "total_steps": 1329, "loss": 0.4485, "learning_rate": 1e-05, "epoch": 2.8208744710860367, "percentage": 94.06, "elapsed_time": "1 day, 7:09:12", "remaining_time": "1:58:08"}
{"current_steps": 1260, "total_steps": 1329, "loss": 0.4503, "learning_rate": 1e-05, "epoch": 2.843441466854725, "percentage": 94.81, "elapsed_time": "1 day, 7:23:56", "remaining_time": "1:43:10"}
{"current_steps": 1270, "total_steps": 1329, "loss": 0.448, "learning_rate": 1e-05, "epoch": 2.8660084626234132, "percentage": 95.56, "elapsed_time": "1 day, 7:38:41", "remaining_time": "1:28:12"}
{"current_steps": 1280, "total_steps": 1329, "loss": 0.4467, "learning_rate": 1e-05, "epoch": 2.8885754583921015, "percentage": 96.31, "elapsed_time": "1 day, 7:53:26", "remaining_time": "1:13:14"}
{"current_steps": 1290, "total_steps": 1329, "loss": 0.4495, "learning_rate": 1e-05, "epoch": 2.9111424541607898, "percentage": 97.07, "elapsed_time": "1 day, 8:08:11", "remaining_time": "0:58:17"}
{"current_steps": 1300, "total_steps": 1329, "loss": 0.446, "learning_rate": 1e-05, "epoch": 2.933709449929478, "percentage": 97.82, "elapsed_time": "1 day, 8:22:56", "remaining_time": "0:43:20"}
{"current_steps": 1310, "total_steps": 1329, "loss": 0.4494, "learning_rate": 1e-05, "epoch": 2.9562764456981663, "percentage": 98.57, "elapsed_time": "1 day, 8:37:41", "remaining_time": "0:28:23"}
{"current_steps": 1320, "total_steps": 1329, "loss": 0.4473, "learning_rate": 1e-05, "epoch": 2.9788434414668545, "percentage": 99.32, "elapsed_time": "1 day, 8:52:26", "remaining_time": "0:13:26"}
{"current_steps": 1329, "total_steps": 1329, "eval_loss": 0.6338749527931213, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:18:32", "remaining_time": "0:00:00"}
{"current_steps": 1329, "total_steps": 1329, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:20:12", "remaining_time": "0:00:00"}