hp_ablations_gemma_lr2e-6 / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 2
a0dab42 verified
{"current_steps": 10, "total_steps": 1329, "loss": 0.7568, "learning_rate": 2e-06, "epoch": 0.022566995768688293, "percentage": 0.75, "elapsed_time": "0:14:52", "remaining_time": "1 day, 8:42:44"}
{"current_steps": 20, "total_steps": 1329, "loss": 0.6946, "learning_rate": 2e-06, "epoch": 0.045133991537376586, "percentage": 1.5, "elapsed_time": "0:29:40", "remaining_time": "1 day, 8:22:36"}
{"current_steps": 30, "total_steps": 1329, "loss": 0.6768, "learning_rate": 2e-06, "epoch": 0.06770098730606489, "percentage": 2.26, "elapsed_time": "0:44:30", "remaining_time": "1 day, 8:06:58"}
{"current_steps": 40, "total_steps": 1329, "loss": 0.6534, "learning_rate": 2e-06, "epoch": 0.09026798307475317, "percentage": 3.01, "elapsed_time": "0:59:19", "remaining_time": "1 day, 7:51:50"}
{"current_steps": 50, "total_steps": 1329, "loss": 0.6463, "learning_rate": 2e-06, "epoch": 0.11283497884344147, "percentage": 3.76, "elapsed_time": "1:14:07", "remaining_time": "1 day, 7:36:18"}
{"current_steps": 60, "total_steps": 1329, "loss": 0.645, "learning_rate": 2e-06, "epoch": 0.13540197461212977, "percentage": 4.51, "elapsed_time": "1:28:57", "remaining_time": "1 day, 7:21:23"}
{"current_steps": 70, "total_steps": 1329, "loss": 0.6403, "learning_rate": 2e-06, "epoch": 0.15796897038081806, "percentage": 5.27, "elapsed_time": "1:43:47", "remaining_time": "1 day, 7:06:47"}
{"current_steps": 80, "total_steps": 1329, "loss": 0.6415, "learning_rate": 2e-06, "epoch": 0.18053596614950634, "percentage": 6.02, "elapsed_time": "1:58:38", "remaining_time": "1 day, 6:52:10"}
{"current_steps": 90, "total_steps": 1329, "loss": 0.6391, "learning_rate": 2e-06, "epoch": 0.20310296191819463, "percentage": 6.77, "elapsed_time": "2:13:27", "remaining_time": "1 day, 6:37:10"}
{"current_steps": 100, "total_steps": 1329, "loss": 0.6378, "learning_rate": 2e-06, "epoch": 0.22566995768688294, "percentage": 7.52, "elapsed_time": "2:28:16", "remaining_time": "1 day, 6:22:21"}
{"current_steps": 110, "total_steps": 1329, "loss": 0.6284, "learning_rate": 2e-06, "epoch": 0.24823695345557123, "percentage": 8.28, "elapsed_time": "2:43:06", "remaining_time": "1 day, 6:07:32"}
{"current_steps": 120, "total_steps": 1329, "loss": 0.6275, "learning_rate": 2e-06, "epoch": 0.27080394922425954, "percentage": 9.03, "elapsed_time": "2:57:55", "remaining_time": "1 day, 5:52:35"}
{"current_steps": 130, "total_steps": 1329, "loss": 0.6285, "learning_rate": 2e-06, "epoch": 0.2933709449929478, "percentage": 9.78, "elapsed_time": "3:12:42", "remaining_time": "1 day, 5:37:21"}
{"current_steps": 140, "total_steps": 1329, "loss": 0.6233, "learning_rate": 2e-06, "epoch": 0.3159379407616361, "percentage": 10.53, "elapsed_time": "3:27:31", "remaining_time": "1 day, 5:22:29"}
{"current_steps": 150, "total_steps": 1329, "loss": 0.6236, "learning_rate": 2e-06, "epoch": 0.3385049365303244, "percentage": 11.29, "elapsed_time": "3:42:22", "remaining_time": "1 day, 5:07:48"}
{"current_steps": 160, "total_steps": 1329, "loss": 0.6256, "learning_rate": 2e-06, "epoch": 0.3610719322990127, "percentage": 12.04, "elapsed_time": "3:57:11", "remaining_time": "1 day, 4:52:55"}
{"current_steps": 170, "total_steps": 1329, "loss": 0.6215, "learning_rate": 2e-06, "epoch": 0.383638928067701, "percentage": 12.79, "elapsed_time": "4:12:01", "remaining_time": "1 day, 4:38:12"}
{"current_steps": 180, "total_steps": 1329, "loss": 0.6186, "learning_rate": 2e-06, "epoch": 0.40620592383638926, "percentage": 13.54, "elapsed_time": "4:26:50", "remaining_time": "1 day, 4:23:20"}
{"current_steps": 190, "total_steps": 1329, "loss": 0.6128, "learning_rate": 2e-06, "epoch": 0.4287729196050776, "percentage": 14.3, "elapsed_time": "4:41:41", "remaining_time": "1 day, 4:08:37"}
{"current_steps": 200, "total_steps": 1329, "loss": 0.6191, "learning_rate": 2e-06, "epoch": 0.4513399153737659, "percentage": 15.05, "elapsed_time": "4:56:30", "remaining_time": "1 day, 3:53:50"}
{"current_steps": 210, "total_steps": 1329, "loss": 0.6157, "learning_rate": 2e-06, "epoch": 0.47390691114245415, "percentage": 15.8, "elapsed_time": "5:11:20", "remaining_time": "1 day, 3:38:59"}
{"current_steps": 220, "total_steps": 1329, "loss": 0.6071, "learning_rate": 2e-06, "epoch": 0.49647390691114246, "percentage": 16.55, "elapsed_time": "5:26:10", "remaining_time": "1 day, 3:24:14"}
{"current_steps": 230, "total_steps": 1329, "loss": 0.6161, "learning_rate": 2e-06, "epoch": 0.5190409026798307, "percentage": 17.31, "elapsed_time": "5:40:58", "remaining_time": "1 day, 3:09:18"}
{"current_steps": 240, "total_steps": 1329, "loss": 0.6146, "learning_rate": 2e-06, "epoch": 0.5416078984485191, "percentage": 18.06, "elapsed_time": "5:55:48", "remaining_time": "1 day, 2:54:30"}
{"current_steps": 250, "total_steps": 1329, "loss": 0.6088, "learning_rate": 2e-06, "epoch": 0.5641748942172073, "percentage": 18.81, "elapsed_time": "6:10:36", "remaining_time": "1 day, 2:39:34"}
{"current_steps": 260, "total_steps": 1329, "loss": 0.6114, "learning_rate": 2e-06, "epoch": 0.5867418899858956, "percentage": 19.56, "elapsed_time": "6:25:26", "remaining_time": "1 day, 2:24:45"}
{"current_steps": 270, "total_steps": 1329, "loss": 0.6068, "learning_rate": 2e-06, "epoch": 0.609308885754584, "percentage": 20.32, "elapsed_time": "6:40:15", "remaining_time": "1 day, 2:09:53"}
{"current_steps": 280, "total_steps": 1329, "loss": 0.6135, "learning_rate": 2e-06, "epoch": 0.6318758815232722, "percentage": 21.07, "elapsed_time": "6:55:05", "remaining_time": "1 day, 1:55:05"}
{"current_steps": 290, "total_steps": 1329, "loss": 0.6069, "learning_rate": 2e-06, "epoch": 0.6544428772919605, "percentage": 21.82, "elapsed_time": "7:09:55", "remaining_time": "1 day, 1:40:18"}
{"current_steps": 300, "total_steps": 1329, "loss": 0.6129, "learning_rate": 2e-06, "epoch": 0.6770098730606487, "percentage": 22.57, "elapsed_time": "7:24:45", "remaining_time": "1 day, 1:25:31"}
{"current_steps": 310, "total_steps": 1329, "loss": 0.6059, "learning_rate": 2e-06, "epoch": 0.6995768688293371, "percentage": 23.33, "elapsed_time": "7:39:35", "remaining_time": "1 day, 1:10:44"}
{"current_steps": 320, "total_steps": 1329, "loss": 0.6093, "learning_rate": 2e-06, "epoch": 0.7221438645980254, "percentage": 24.08, "elapsed_time": "7:54:25", "remaining_time": "1 day, 0:55:56"}
{"current_steps": 330, "total_steps": 1329, "loss": 0.6048, "learning_rate": 2e-06, "epoch": 0.7447108603667136, "percentage": 24.83, "elapsed_time": "8:09:16", "remaining_time": "1 day, 0:41:08"}
{"current_steps": 340, "total_steps": 1329, "loss": 0.604, "learning_rate": 2e-06, "epoch": 0.767277856135402, "percentage": 25.58, "elapsed_time": "8:24:05", "remaining_time": "1 day, 0:26:18"}
{"current_steps": 350, "total_steps": 1329, "loss": 0.6053, "learning_rate": 2e-06, "epoch": 0.7898448519040903, "percentage": 26.34, "elapsed_time": "8:38:53", "remaining_time": "1 day, 0:11:24"}
{"current_steps": 360, "total_steps": 1329, "loss": 0.6044, "learning_rate": 2e-06, "epoch": 0.8124118476727785, "percentage": 27.09, "elapsed_time": "8:53:42", "remaining_time": "23:56:35"}
{"current_steps": 370, "total_steps": 1329, "loss": 0.6114, "learning_rate": 2e-06, "epoch": 0.8349788434414669, "percentage": 27.84, "elapsed_time": "9:08:31", "remaining_time": "23:41:43"}
{"current_steps": 380, "total_steps": 1329, "loss": 0.6079, "learning_rate": 2e-06, "epoch": 0.8575458392101551, "percentage": 28.59, "elapsed_time": "9:23:20", "remaining_time": "23:26:52"}
{"current_steps": 390, "total_steps": 1329, "loss": 0.6021, "learning_rate": 2e-06, "epoch": 0.8801128349788434, "percentage": 29.35, "elapsed_time": "9:38:09", "remaining_time": "23:12:02"}
{"current_steps": 400, "total_steps": 1329, "loss": 0.6043, "learning_rate": 2e-06, "epoch": 0.9026798307475318, "percentage": 30.1, "elapsed_time": "9:53:00", "remaining_time": "22:57:14"}
{"current_steps": 410, "total_steps": 1329, "loss": 0.6023, "learning_rate": 2e-06, "epoch": 0.92524682651622, "percentage": 30.85, "elapsed_time": "10:07:50", "remaining_time": "22:42:26"}
{"current_steps": 420, "total_steps": 1329, "loss": 0.6028, "learning_rate": 2e-06, "epoch": 0.9478138222849083, "percentage": 31.6, "elapsed_time": "10:22:39", "remaining_time": "22:27:35"}
{"current_steps": 430, "total_steps": 1329, "loss": 0.5968, "learning_rate": 2e-06, "epoch": 0.9703808180535967, "percentage": 32.36, "elapsed_time": "10:37:29", "remaining_time": "22:12:49"}
{"current_steps": 440, "total_steps": 1329, "loss": 0.6015, "learning_rate": 2e-06, "epoch": 0.9929478138222849, "percentage": 33.11, "elapsed_time": "10:52:20", "remaining_time": "21:58:01"}
{"current_steps": 443, "total_steps": 1329, "eval_loss": 0.5996821522712708, "epoch": 0.9997179125528914, "percentage": 33.33, "elapsed_time": "11:08:23", "remaining_time": "22:16:46"}
{"current_steps": 450, "total_steps": 1329, "loss": 0.6277, "learning_rate": 2e-06, "epoch": 1.0155148095909732, "percentage": 33.86, "elapsed_time": "11:19:53", "remaining_time": "22:08:02"}
{"current_steps": 460, "total_steps": 1329, "loss": 0.5623, "learning_rate": 2e-06, "epoch": 1.0380818053596614, "percentage": 34.61, "elapsed_time": "11:34:42", "remaining_time": "21:52:23"}
{"current_steps": 470, "total_steps": 1329, "loss": 0.5779, "learning_rate": 2e-06, "epoch": 1.0606488011283497, "percentage": 35.36, "elapsed_time": "11:49:33", "remaining_time": "21:36:50"}
{"current_steps": 480, "total_steps": 1329, "loss": 0.5646, "learning_rate": 2e-06, "epoch": 1.0832157968970382, "percentage": 36.12, "elapsed_time": "12:04:24", "remaining_time": "21:21:17"}
{"current_steps": 490, "total_steps": 1329, "loss": 0.5756, "learning_rate": 2e-06, "epoch": 1.1057827926657264, "percentage": 36.87, "elapsed_time": "12:19:15", "remaining_time": "21:05:47"}
{"current_steps": 500, "total_steps": 1329, "loss": 0.5737, "learning_rate": 2e-06, "epoch": 1.1283497884344147, "percentage": 37.62, "elapsed_time": "12:34:06", "remaining_time": "20:50:18"}
{"current_steps": 510, "total_steps": 1329, "loss": 0.5738, "learning_rate": 2e-06, "epoch": 1.150916784203103, "percentage": 38.37, "elapsed_time": "12:48:57", "remaining_time": "20:34:51"}
{"current_steps": 520, "total_steps": 1329, "loss": 0.5658, "learning_rate": 2e-06, "epoch": 1.1734837799717912, "percentage": 39.13, "elapsed_time": "13:03:48", "remaining_time": "20:19:24"}
{"current_steps": 530, "total_steps": 1329, "loss": 0.5719, "learning_rate": 2e-06, "epoch": 1.1960507757404795, "percentage": 39.88, "elapsed_time": "13:18:38", "remaining_time": "20:03:58"}
{"current_steps": 540, "total_steps": 1329, "loss": 0.5689, "learning_rate": 2e-06, "epoch": 1.2186177715091677, "percentage": 40.63, "elapsed_time": "13:33:29", "remaining_time": "19:48:36"}
{"current_steps": 550, "total_steps": 1329, "loss": 0.5735, "learning_rate": 2e-06, "epoch": 1.2411847672778562, "percentage": 41.38, "elapsed_time": "13:48:19", "remaining_time": "19:33:12"}
{"current_steps": 560, "total_steps": 1329, "loss": 0.5751, "learning_rate": 2e-06, "epoch": 1.2637517630465445, "percentage": 42.14, "elapsed_time": "14:03:09", "remaining_time": "19:17:50"}
{"current_steps": 570, "total_steps": 1329, "loss": 0.572, "learning_rate": 2e-06, "epoch": 1.2863187588152327, "percentage": 42.89, "elapsed_time": "14:18:00", "remaining_time": "19:02:29"}
{"current_steps": 580, "total_steps": 1329, "loss": 0.5716, "learning_rate": 2e-06, "epoch": 1.308885754583921, "percentage": 43.64, "elapsed_time": "14:32:48", "remaining_time": "18:47:07"}
{"current_steps": 590, "total_steps": 1329, "loss": 0.571, "learning_rate": 2e-06, "epoch": 1.3314527503526092, "percentage": 44.39, "elapsed_time": "14:47:36", "remaining_time": "18:31:46"}
{"current_steps": 600, "total_steps": 1329, "loss": 0.577, "learning_rate": 2e-06, "epoch": 1.3540197461212977, "percentage": 45.15, "elapsed_time": "15:02:27", "remaining_time": "18:16:29"}
{"current_steps": 610, "total_steps": 1329, "loss": 0.5689, "learning_rate": 2e-06, "epoch": 1.376586741889986, "percentage": 45.9, "elapsed_time": "15:17:18", "remaining_time": "18:01:13"}
{"current_steps": 620, "total_steps": 1329, "loss": 0.5728, "learning_rate": 2e-06, "epoch": 1.3991537376586742, "percentage": 46.65, "elapsed_time": "15:32:08", "remaining_time": "17:45:57"}
{"current_steps": 630, "total_steps": 1329, "loss": 0.5706, "learning_rate": 2e-06, "epoch": 1.4217207334273625, "percentage": 47.4, "elapsed_time": "15:46:58", "remaining_time": "17:30:41"}
{"current_steps": 640, "total_steps": 1329, "loss": 0.5703, "learning_rate": 2e-06, "epoch": 1.4442877291960508, "percentage": 48.16, "elapsed_time": "16:01:49", "remaining_time": "17:15:27"}
{"current_steps": 650, "total_steps": 1329, "loss": 0.5709, "learning_rate": 2e-06, "epoch": 1.466854724964739, "percentage": 48.91, "elapsed_time": "16:16:38", "remaining_time": "17:00:12"}
{"current_steps": 660, "total_steps": 1329, "loss": 0.5746, "learning_rate": 2e-06, "epoch": 1.4894217207334273, "percentage": 49.66, "elapsed_time": "16:31:28", "remaining_time": "16:45:00"}
{"current_steps": 670, "total_steps": 1329, "loss": 0.5689, "learning_rate": 2e-06, "epoch": 1.5119887165021155, "percentage": 50.41, "elapsed_time": "16:46:19", "remaining_time": "16:29:48"}
{"current_steps": 680, "total_steps": 1329, "loss": 0.5625, "learning_rate": 2e-06, "epoch": 1.5345557122708038, "percentage": 51.17, "elapsed_time": "17:01:10", "remaining_time": "16:14:36"}
{"current_steps": 690, "total_steps": 1329, "loss": 0.565, "learning_rate": 2e-06, "epoch": 1.5571227080394923, "percentage": 51.92, "elapsed_time": "17:16:01", "remaining_time": "15:59:27"}
{"current_steps": 700, "total_steps": 1329, "loss": 0.5688, "learning_rate": 2e-06, "epoch": 1.5796897038081805, "percentage": 52.67, "elapsed_time": "17:30:53", "remaining_time": "15:44:17"}
{"current_steps": 710, "total_steps": 1329, "loss": 0.5715, "learning_rate": 2e-06, "epoch": 1.6022566995768688, "percentage": 53.42, "elapsed_time": "17:45:43", "remaining_time": "15:29:07"}
{"current_steps": 720, "total_steps": 1329, "loss": 0.5798, "learning_rate": 2e-06, "epoch": 1.6248236953455573, "percentage": 54.18, "elapsed_time": "18:00:33", "remaining_time": "15:13:58"}
{"current_steps": 730, "total_steps": 1329, "loss": 0.5621, "learning_rate": 2e-06, "epoch": 1.6473906911142455, "percentage": 54.93, "elapsed_time": "18:15:23", "remaining_time": "14:58:49"}
{"current_steps": 740, "total_steps": 1329, "loss": 0.5786, "learning_rate": 2e-06, "epoch": 1.6699576868829338, "percentage": 55.68, "elapsed_time": "18:30:15", "remaining_time": "14:43:42"}
{"current_steps": 750, "total_steps": 1329, "loss": 0.5694, "learning_rate": 2e-06, "epoch": 1.692524682651622, "percentage": 56.43, "elapsed_time": "18:45:05", "remaining_time": "14:28:33"}
{"current_steps": 760, "total_steps": 1329, "loss": 0.568, "learning_rate": 2e-06, "epoch": 1.7150916784203103, "percentage": 57.19, "elapsed_time": "18:59:54", "remaining_time": "14:13:25"}
{"current_steps": 770, "total_steps": 1329, "loss": 0.5683, "learning_rate": 2e-06, "epoch": 1.7376586741889986, "percentage": 57.94, "elapsed_time": "19:14:44", "remaining_time": "13:58:18"}
{"current_steps": 780, "total_steps": 1329, "loss": 0.5641, "learning_rate": 2e-06, "epoch": 1.7602256699576868, "percentage": 58.69, "elapsed_time": "19:29:34", "remaining_time": "13:43:11"}
{"current_steps": 790, "total_steps": 1329, "loss": 0.5678, "learning_rate": 2e-06, "epoch": 1.782792665726375, "percentage": 59.44, "elapsed_time": "19:44:24", "remaining_time": "13:28:05"}
{"current_steps": 800, "total_steps": 1329, "loss": 0.5623, "learning_rate": 2e-06, "epoch": 1.8053596614950633, "percentage": 60.2, "elapsed_time": "19:59:14", "remaining_time": "13:12:59"}
{"current_steps": 810, "total_steps": 1329, "loss": 0.5731, "learning_rate": 2e-06, "epoch": 1.8279266572637518, "percentage": 60.95, "elapsed_time": "20:14:03", "remaining_time": "12:57:53"}
{"current_steps": 820, "total_steps": 1329, "loss": 0.56, "learning_rate": 2e-06, "epoch": 1.85049365303244, "percentage": 61.7, "elapsed_time": "20:28:52", "remaining_time": "12:42:47"}
{"current_steps": 830, "total_steps": 1329, "loss": 0.5726, "learning_rate": 2e-06, "epoch": 1.8730606488011283, "percentage": 62.45, "elapsed_time": "20:43:41", "remaining_time": "12:27:42"}
{"current_steps": 840, "total_steps": 1329, "loss": 0.5723, "learning_rate": 2e-06, "epoch": 1.8956276445698168, "percentage": 63.21, "elapsed_time": "20:58:32", "remaining_time": "12:12:38"}
{"current_steps": 850, "total_steps": 1329, "loss": 0.5595, "learning_rate": 2e-06, "epoch": 1.918194640338505, "percentage": 63.96, "elapsed_time": "21:13:22", "remaining_time": "11:57:34"}
{"current_steps": 860, "total_steps": 1329, "loss": 0.5645, "learning_rate": 2e-06, "epoch": 1.9407616361071933, "percentage": 64.71, "elapsed_time": "21:28:11", "remaining_time": "11:42:30"}
{"current_steps": 870, "total_steps": 1329, "loss": 0.5712, "learning_rate": 2e-06, "epoch": 1.9633286318758816, "percentage": 65.46, "elapsed_time": "21:43:02", "remaining_time": "11:27:27"}
{"current_steps": 880, "total_steps": 1329, "loss": 0.5692, "learning_rate": 2e-06, "epoch": 1.9858956276445698, "percentage": 66.22, "elapsed_time": "21:57:53", "remaining_time": "11:12:25"}
{"current_steps": 886, "total_steps": 1329, "eval_loss": 0.5890958905220032, "epoch": 1.9994358251057829, "percentage": 66.67, "elapsed_time": "22:18:35", "remaining_time": "11:09:17"}
{"current_steps": 890, "total_steps": 1329, "loss": 0.6066, "learning_rate": 2e-06, "epoch": 2.008462623413258, "percentage": 66.97, "elapsed_time": "22:25:33", "remaining_time": "11:03:42"}
{"current_steps": 900, "total_steps": 1329, "loss": 0.5346, "learning_rate": 2e-06, "epoch": 2.0310296191819464, "percentage": 67.72, "elapsed_time": "22:40:23", "remaining_time": "10:48:27"}
{"current_steps": 910, "total_steps": 1329, "loss": 0.5375, "learning_rate": 2e-06, "epoch": 2.0535966149506346, "percentage": 68.47, "elapsed_time": "22:55:13", "remaining_time": "10:33:12"}
{"current_steps": 920, "total_steps": 1329, "loss": 0.5296, "learning_rate": 2e-06, "epoch": 2.076163610719323, "percentage": 69.22, "elapsed_time": "23:10:02", "remaining_time": "10:17:57"}
{"current_steps": 930, "total_steps": 1329, "loss": 0.537, "learning_rate": 2e-06, "epoch": 2.098730606488011, "percentage": 69.98, "elapsed_time": "23:24:51", "remaining_time": "10:02:43"}
{"current_steps": 940, "total_steps": 1329, "loss": 0.5406, "learning_rate": 2e-06, "epoch": 2.1212976022566994, "percentage": 70.73, "elapsed_time": "23:39:39", "remaining_time": "9:47:29"}
{"current_steps": 950, "total_steps": 1329, "loss": 0.5421, "learning_rate": 2e-06, "epoch": 2.143864598025388, "percentage": 71.48, "elapsed_time": "23:54:27", "remaining_time": "9:32:16"}
{"current_steps": 960, "total_steps": 1329, "loss": 0.5381, "learning_rate": 2e-06, "epoch": 2.1664315937940763, "percentage": 72.23, "elapsed_time": "1 day, 0:09:16", "remaining_time": "9:17:03"}
{"current_steps": 970, "total_steps": 1329, "loss": 0.5349, "learning_rate": 2e-06, "epoch": 2.1889985895627646, "percentage": 72.99, "elapsed_time": "1 day, 0:24:04", "remaining_time": "9:01:51"}
{"current_steps": 980, "total_steps": 1329, "loss": 0.5289, "learning_rate": 2e-06, "epoch": 2.211565585331453, "percentage": 73.74, "elapsed_time": "1 day, 0:38:53", "remaining_time": "8:46:40"}
{"current_steps": 990, "total_steps": 1329, "loss": 0.5283, "learning_rate": 2e-06, "epoch": 2.234132581100141, "percentage": 74.49, "elapsed_time": "1 day, 0:53:42", "remaining_time": "8:31:29"}
{"current_steps": 1000, "total_steps": 1329, "loss": 0.5384, "learning_rate": 2e-06, "epoch": 2.2566995768688294, "percentage": 75.24, "elapsed_time": "1 day, 1:08:32", "remaining_time": "8:16:18"}
{"current_steps": 1010, "total_steps": 1329, "loss": 0.5449, "learning_rate": 2e-06, "epoch": 2.2792665726375176, "percentage": 76.0, "elapsed_time": "1 day, 1:23:21", "remaining_time": "8:01:08"}
{"current_steps": 1020, "total_steps": 1329, "loss": 0.5378, "learning_rate": 2e-06, "epoch": 2.301833568406206, "percentage": 76.75, "elapsed_time": "1 day, 1:38:10", "remaining_time": "7:45:58"}
{"current_steps": 1030, "total_steps": 1329, "loss": 0.5424, "learning_rate": 2e-06, "epoch": 2.324400564174894, "percentage": 77.5, "elapsed_time": "1 day, 1:52:59", "remaining_time": "7:30:49"}
{"current_steps": 1040, "total_steps": 1329, "loss": 0.5393, "learning_rate": 2e-06, "epoch": 2.3469675599435824, "percentage": 78.25, "elapsed_time": "1 day, 2:07:49", "remaining_time": "7:15:40"}
{"current_steps": 1050, "total_steps": 1329, "loss": 0.5363, "learning_rate": 2e-06, "epoch": 2.3695345557122707, "percentage": 79.01, "elapsed_time": "1 day, 2:22:39", "remaining_time": "7:00:32"}
{"current_steps": 1060, "total_steps": 1329, "loss": 0.5357, "learning_rate": 2e-06, "epoch": 2.392101551480959, "percentage": 79.76, "elapsed_time": "1 day, 2:37:27", "remaining_time": "6:45:23"}
{"current_steps": 1070, "total_steps": 1329, "loss": 0.5361, "learning_rate": 2e-06, "epoch": 2.414668547249647, "percentage": 80.51, "elapsed_time": "1 day, 2:52:16", "remaining_time": "6:30:15"}
{"current_steps": 1080, "total_steps": 1329, "loss": 0.5399, "learning_rate": 2e-06, "epoch": 2.4372355430183354, "percentage": 81.26, "elapsed_time": "1 day, 3:07:05", "remaining_time": "6:15:08"}
{"current_steps": 1090, "total_steps": 1329, "loss": 0.538, "learning_rate": 2e-06, "epoch": 2.459802538787024, "percentage": 82.02, "elapsed_time": "1 day, 3:21:54", "remaining_time": "6:00:00"}
{"current_steps": 1100, "total_steps": 1329, "loss": 0.5405, "learning_rate": 2e-06, "epoch": 2.4823695345557124, "percentage": 82.77, "elapsed_time": "1 day, 3:36:42", "remaining_time": "5:44:53"}
{"current_steps": 1110, "total_steps": 1329, "loss": 0.544, "learning_rate": 2e-06, "epoch": 2.5049365303244007, "percentage": 83.52, "elapsed_time": "1 day, 3:51:32", "remaining_time": "5:29:47"}
{"current_steps": 1120, "total_steps": 1329, "loss": 0.5332, "learning_rate": 2e-06, "epoch": 2.527503526093089, "percentage": 84.27, "elapsed_time": "1 day, 4:06:23", "remaining_time": "5:14:41"}
{"current_steps": 1130, "total_steps": 1329, "loss": 0.5387, "learning_rate": 2e-06, "epoch": 2.550070521861777, "percentage": 85.03, "elapsed_time": "1 day, 4:21:12", "remaining_time": "4:59:35"}
{"current_steps": 1140, "total_steps": 1329, "loss": 0.5366, "learning_rate": 2e-06, "epoch": 2.5726375176304654, "percentage": 85.78, "elapsed_time": "1 day, 4:36:01", "remaining_time": "4:44:30"}
{"current_steps": 1150, "total_steps": 1329, "loss": 0.5375, "learning_rate": 2e-06, "epoch": 2.5952045133991537, "percentage": 86.53, "elapsed_time": "1 day, 4:50:49", "remaining_time": "4:29:24"}
{"current_steps": 1160, "total_steps": 1329, "loss": 0.5411, "learning_rate": 2e-06, "epoch": 2.617771509167842, "percentage": 87.28, "elapsed_time": "1 day, 5:05:40", "remaining_time": "4:14:19"}
{"current_steps": 1170, "total_steps": 1329, "loss": 0.5286, "learning_rate": 2e-06, "epoch": 2.64033850493653, "percentage": 88.04, "elapsed_time": "1 day, 5:20:30", "remaining_time": "3:59:14"}
{"current_steps": 1180, "total_steps": 1329, "loss": 0.5404, "learning_rate": 2e-06, "epoch": 2.6629055007052185, "percentage": 88.79, "elapsed_time": "1 day, 5:35:20", "remaining_time": "3:44:10"}
{"current_steps": 1190, "total_steps": 1329, "loss": 0.541, "learning_rate": 2e-06, "epoch": 2.685472496473907, "percentage": 89.54, "elapsed_time": "1 day, 5:50:07", "remaining_time": "3:29:05"}
{"current_steps": 1200, "total_steps": 1329, "loss": 0.5393, "learning_rate": 2e-06, "epoch": 2.7080394922425954, "percentage": 90.29, "elapsed_time": "1 day, 6:04:58", "remaining_time": "3:14:02"}
{"current_steps": 1210, "total_steps": 1329, "loss": 0.5404, "learning_rate": 2e-06, "epoch": 2.7306064880112837, "percentage": 91.05, "elapsed_time": "1 day, 6:19:48", "remaining_time": "2:58:58"}
{"current_steps": 1220, "total_steps": 1329, "loss": 0.5361, "learning_rate": 2e-06, "epoch": 2.753173483779972, "percentage": 91.8, "elapsed_time": "1 day, 6:34:38", "remaining_time": "2:43:54"}
{"current_steps": 1230, "total_steps": 1329, "loss": 0.5434, "learning_rate": 2e-06, "epoch": 2.77574047954866, "percentage": 92.55, "elapsed_time": "1 day, 6:49:27", "remaining_time": "2:28:51"}
{"current_steps": 1240, "total_steps": 1329, "loss": 0.5417, "learning_rate": 2e-06, "epoch": 2.7983074753173485, "percentage": 93.3, "elapsed_time": "1 day, 7:04:16", "remaining_time": "2:13:48"}
{"current_steps": 1250, "total_steps": 1329, "loss": 0.5413, "learning_rate": 2e-06, "epoch": 2.8208744710860367, "percentage": 94.06, "elapsed_time": "1 day, 7:19:04", "remaining_time": "1:58:45"}
{"current_steps": 1260, "total_steps": 1329, "loss": 0.5431, "learning_rate": 2e-06, "epoch": 2.843441466854725, "percentage": 94.81, "elapsed_time": "1 day, 7:33:53", "remaining_time": "1:43:42"}
{"current_steps": 1270, "total_steps": 1329, "loss": 0.5396, "learning_rate": 2e-06, "epoch": 2.8660084626234132, "percentage": 95.56, "elapsed_time": "1 day, 7:48:40", "remaining_time": "1:28:40"}
{"current_steps": 1280, "total_steps": 1329, "loss": 0.5371, "learning_rate": 2e-06, "epoch": 2.8885754583921015, "percentage": 96.31, "elapsed_time": "1 day, 8:03:29", "remaining_time": "1:13:38"}
{"current_steps": 1290, "total_steps": 1329, "loss": 0.5416, "learning_rate": 2e-06, "epoch": 2.9111424541607898, "percentage": 97.07, "elapsed_time": "1 day, 8:18:20", "remaining_time": "0:58:36"}
{"current_steps": 1300, "total_steps": 1329, "loss": 0.5361, "learning_rate": 2e-06, "epoch": 2.933709449929478, "percentage": 97.82, "elapsed_time": "1 day, 8:33:09", "remaining_time": "0:43:34"}
{"current_steps": 1310, "total_steps": 1329, "loss": 0.5405, "learning_rate": 2e-06, "epoch": 2.9562764456981663, "percentage": 98.57, "elapsed_time": "1 day, 8:47:59", "remaining_time": "0:28:32"}
{"current_steps": 1320, "total_steps": 1329, "loss": 0.5356, "learning_rate": 2e-06, "epoch": 2.9788434414668545, "percentage": 99.32, "elapsed_time": "1 day, 9:02:49", "remaining_time": "0:13:31"}
{"current_steps": 1329, "total_steps": 1329, "eval_loss": 0.5875794291496277, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:29:00", "remaining_time": "0:00:00"}
{"current_steps": 1329, "total_steps": 1329, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:30:41", "remaining_time": "0:00:00"}