hp_ablations_gemma_lr5e-6 / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 2
c55c63c verified
{"current_steps": 10, "total_steps": 1329, "loss": 0.7438, "learning_rate": 5e-06, "epoch": 0.022566995768688293, "percentage": 0.75, "elapsed_time": "0:14:48", "remaining_time": "1 day, 8:34:13"}
{"current_steps": 20, "total_steps": 1329, "loss": 0.6802, "learning_rate": 5e-06, "epoch": 0.045133991537376586, "percentage": 1.5, "elapsed_time": "0:29:33", "remaining_time": "1 day, 8:14:58"}
{"current_steps": 30, "total_steps": 1329, "loss": 0.6629, "learning_rate": 5e-06, "epoch": 0.06770098730606489, "percentage": 2.26, "elapsed_time": "0:44:16", "remaining_time": "1 day, 7:57:08"}
{"current_steps": 40, "total_steps": 1329, "loss": 0.6406, "learning_rate": 5e-06, "epoch": 0.09026798307475317, "percentage": 3.01, "elapsed_time": "0:58:59", "remaining_time": "1 day, 7:40:56"}
{"current_steps": 50, "total_steps": 1329, "loss": 0.634, "learning_rate": 5e-06, "epoch": 0.11283497884344147, "percentage": 3.76, "elapsed_time": "1:13:45", "remaining_time": "1 day, 7:26:32"}
{"current_steps": 60, "total_steps": 1329, "loss": 0.6332, "learning_rate": 5e-06, "epoch": 0.13540197461212977, "percentage": 4.51, "elapsed_time": "1:28:28", "remaining_time": "1 day, 7:11:13"}
{"current_steps": 70, "total_steps": 1329, "loss": 0.629, "learning_rate": 5e-06, "epoch": 0.15796897038081806, "percentage": 5.27, "elapsed_time": "1:43:13", "remaining_time": "1 day, 6:56:26"}
{"current_steps": 80, "total_steps": 1329, "loss": 0.6306, "learning_rate": 5e-06, "epoch": 0.18053596614950634, "percentage": 6.02, "elapsed_time": "1:57:57", "remaining_time": "1 day, 6:41:38"}
{"current_steps": 90, "total_steps": 1329, "loss": 0.6282, "learning_rate": 5e-06, "epoch": 0.20310296191819463, "percentage": 6.77, "elapsed_time": "2:12:41", "remaining_time": "1 day, 6:26:43"}
{"current_steps": 100, "total_steps": 1329, "loss": 0.6272, "learning_rate": 5e-06, "epoch": 0.22566995768688294, "percentage": 7.52, "elapsed_time": "2:27:25", "remaining_time": "1 day, 6:11:50"}
{"current_steps": 110, "total_steps": 1329, "loss": 0.618, "learning_rate": 5e-06, "epoch": 0.24823695345557123, "percentage": 8.28, "elapsed_time": "2:42:08", "remaining_time": "1 day, 5:56:52"}
{"current_steps": 120, "total_steps": 1329, "loss": 0.6173, "learning_rate": 5e-06, "epoch": 0.27080394922425954, "percentage": 9.03, "elapsed_time": "2:56:53", "remaining_time": "1 day, 5:42:06"}
{"current_steps": 130, "total_steps": 1329, "loss": 0.6185, "learning_rate": 5e-06, "epoch": 0.2933709449929478, "percentage": 9.78, "elapsed_time": "3:11:38", "remaining_time": "1 day, 5:27:27"}
{"current_steps": 140, "total_steps": 1329, "loss": 0.6135, "learning_rate": 5e-06, "epoch": 0.3159379407616361, "percentage": 10.53, "elapsed_time": "3:26:22", "remaining_time": "1 day, 5:12:41"}
{"current_steps": 150, "total_steps": 1329, "loss": 0.614, "learning_rate": 5e-06, "epoch": 0.3385049365303244, "percentage": 11.29, "elapsed_time": "3:41:06", "remaining_time": "1 day, 4:57:56"}
{"current_steps": 160, "total_steps": 1329, "loss": 0.6157, "learning_rate": 5e-06, "epoch": 0.3610719322990127, "percentage": 12.04, "elapsed_time": "3:55:51", "remaining_time": "1 day, 4:43:12"}
{"current_steps": 170, "total_steps": 1329, "loss": 0.6118, "learning_rate": 5e-06, "epoch": 0.383638928067701, "percentage": 12.79, "elapsed_time": "4:10:35", "remaining_time": "1 day, 4:28:26"}
{"current_steps": 180, "total_steps": 1329, "loss": 0.6087, "learning_rate": 5e-06, "epoch": 0.40620592383638926, "percentage": 13.54, "elapsed_time": "4:25:20", "remaining_time": "1 day, 4:13:47"}
{"current_steps": 190, "total_steps": 1329, "loss": 0.6032, "learning_rate": 5e-06, "epoch": 0.4287729196050776, "percentage": 14.3, "elapsed_time": "4:40:03", "remaining_time": "1 day, 3:58:54"}
{"current_steps": 200, "total_steps": 1329, "loss": 0.6099, "learning_rate": 5e-06, "epoch": 0.4513399153737659, "percentage": 15.05, "elapsed_time": "4:54:46", "remaining_time": "1 day, 3:44:02"}
{"current_steps": 210, "total_steps": 1329, "loss": 0.6063, "learning_rate": 5e-06, "epoch": 0.47390691114245415, "percentage": 15.8, "elapsed_time": "5:09:31", "remaining_time": "1 day, 3:29:18"}
{"current_steps": 220, "total_steps": 1329, "loss": 0.598, "learning_rate": 5e-06, "epoch": 0.49647390691114246, "percentage": 16.55, "elapsed_time": "5:24:14", "remaining_time": "1 day, 3:14:30"}
{"current_steps": 230, "total_steps": 1329, "loss": 0.6068, "learning_rate": 5e-06, "epoch": 0.5190409026798307, "percentage": 17.31, "elapsed_time": "5:38:58", "remaining_time": "1 day, 2:59:40"}
{"current_steps": 240, "total_steps": 1329, "loss": 0.6057, "learning_rate": 5e-06, "epoch": 0.5416078984485191, "percentage": 18.06, "elapsed_time": "5:53:42", "remaining_time": "1 day, 2:44:57"}
{"current_steps": 250, "total_steps": 1329, "loss": 0.5998, "learning_rate": 5e-06, "epoch": 0.5641748942172073, "percentage": 18.81, "elapsed_time": "6:08:25", "remaining_time": "1 day, 2:30:08"}
{"current_steps": 260, "total_steps": 1329, "loss": 0.6025, "learning_rate": 5e-06, "epoch": 0.5867418899858956, "percentage": 19.56, "elapsed_time": "6:23:09", "remaining_time": "1 day, 2:15:21"}
{"current_steps": 270, "total_steps": 1329, "loss": 0.5979, "learning_rate": 5e-06, "epoch": 0.609308885754584, "percentage": 20.32, "elapsed_time": "6:37:52", "remaining_time": "1 day, 2:00:34"}
{"current_steps": 280, "total_steps": 1329, "loss": 0.6045, "learning_rate": 5e-06, "epoch": 0.6318758815232722, "percentage": 21.07, "elapsed_time": "6:52:36", "remaining_time": "1 day, 1:45:50"}
{"current_steps": 290, "total_steps": 1329, "loss": 0.5979, "learning_rate": 5e-06, "epoch": 0.6544428772919605, "percentage": 21.82, "elapsed_time": "7:07:21", "remaining_time": "1 day, 1:31:08"}
{"current_steps": 300, "total_steps": 1329, "loss": 0.604, "learning_rate": 5e-06, "epoch": 0.6770098730606487, "percentage": 22.57, "elapsed_time": "7:22:05", "remaining_time": "1 day, 1:16:21"}
{"current_steps": 310, "total_steps": 1329, "loss": 0.5972, "learning_rate": 5e-06, "epoch": 0.6995768688293371, "percentage": 23.33, "elapsed_time": "7:36:49", "remaining_time": "1 day, 1:01:37"}
{"current_steps": 320, "total_steps": 1329, "loss": 0.6004, "learning_rate": 5e-06, "epoch": 0.7221438645980254, "percentage": 24.08, "elapsed_time": "7:51:34", "remaining_time": "1 day, 0:46:57"}
{"current_steps": 330, "total_steps": 1329, "loss": 0.596, "learning_rate": 5e-06, "epoch": 0.7447108603667136, "percentage": 24.83, "elapsed_time": "8:06:18", "remaining_time": "1 day, 0:32:12"}
{"current_steps": 340, "total_steps": 1329, "loss": 0.5955, "learning_rate": 5e-06, "epoch": 0.767277856135402, "percentage": 25.58, "elapsed_time": "8:21:03", "remaining_time": "1 day, 0:17:29"}
{"current_steps": 350, "total_steps": 1329, "loss": 0.597, "learning_rate": 5e-06, "epoch": 0.7898448519040903, "percentage": 26.34, "elapsed_time": "8:35:47", "remaining_time": "1 day, 0:02:43"}
{"current_steps": 360, "total_steps": 1329, "loss": 0.5959, "learning_rate": 5e-06, "epoch": 0.8124118476727785, "percentage": 27.09, "elapsed_time": "8:50:30", "remaining_time": "23:47:57"}
{"current_steps": 370, "total_steps": 1329, "loss": 0.603, "learning_rate": 5e-06, "epoch": 0.8349788434414669, "percentage": 27.84, "elapsed_time": "9:05:15", "remaining_time": "23:33:14"}
{"current_steps": 380, "total_steps": 1329, "loss": 0.5997, "learning_rate": 5e-06, "epoch": 0.8575458392101551, "percentage": 28.59, "elapsed_time": "9:20:00", "remaining_time": "23:18:32"}
{"current_steps": 390, "total_steps": 1329, "loss": 0.5937, "learning_rate": 5e-06, "epoch": 0.8801128349788434, "percentage": 29.35, "elapsed_time": "9:34:44", "remaining_time": "23:03:47"}
{"current_steps": 400, "total_steps": 1329, "loss": 0.596, "learning_rate": 5e-06, "epoch": 0.9026798307475318, "percentage": 30.1, "elapsed_time": "9:49:29", "remaining_time": "22:49:04"}
{"current_steps": 410, "total_steps": 1329, "loss": 0.594, "learning_rate": 5e-06, "epoch": 0.92524682651622, "percentage": 30.85, "elapsed_time": "10:04:11", "remaining_time": "22:34:17"}
{"current_steps": 420, "total_steps": 1329, "loss": 0.5946, "learning_rate": 5e-06, "epoch": 0.9478138222849083, "percentage": 31.6, "elapsed_time": "10:18:56", "remaining_time": "22:19:32"}
{"current_steps": 430, "total_steps": 1329, "loss": 0.5886, "learning_rate": 5e-06, "epoch": 0.9703808180535967, "percentage": 32.36, "elapsed_time": "10:33:39", "remaining_time": "22:04:47"}
{"current_steps": 440, "total_steps": 1329, "loss": 0.5934, "learning_rate": 5e-06, "epoch": 0.9929478138222849, "percentage": 33.11, "elapsed_time": "10:48:21", "remaining_time": "21:49:58"}
{"current_steps": 443, "total_steps": 1329, "eval_loss": 0.5912777781486511, "epoch": 0.9997179125528914, "percentage": 33.33, "elapsed_time": "11:04:20", "remaining_time": "22:08:41"}
{"current_steps": 450, "total_steps": 1329, "loss": 0.6008, "learning_rate": 5e-06, "epoch": 1.0155148095909732, "percentage": 33.86, "elapsed_time": "11:15:44", "remaining_time": "21:59:57"}
{"current_steps": 460, "total_steps": 1329, "loss": 0.528, "learning_rate": 5e-06, "epoch": 1.0380818053596614, "percentage": 34.61, "elapsed_time": "11:30:28", "remaining_time": "21:44:24"}
{"current_steps": 470, "total_steps": 1329, "loss": 0.5429, "learning_rate": 5e-06, "epoch": 1.0606488011283497, "percentage": 35.36, "elapsed_time": "11:45:11", "remaining_time": "21:28:50"}
{"current_steps": 480, "total_steps": 1329, "loss": 0.5308, "learning_rate": 5e-06, "epoch": 1.0832157968970382, "percentage": 36.12, "elapsed_time": "11:59:53", "remaining_time": "21:13:19"}
{"current_steps": 490, "total_steps": 1329, "loss": 0.5408, "learning_rate": 5e-06, "epoch": 1.1057827926657264, "percentage": 36.87, "elapsed_time": "12:14:38", "remaining_time": "20:57:53"}
{"current_steps": 500, "total_steps": 1329, "loss": 0.5396, "learning_rate": 5e-06, "epoch": 1.1283497884344147, "percentage": 37.62, "elapsed_time": "12:29:23", "remaining_time": "20:42:28"}
{"current_steps": 510, "total_steps": 1329, "loss": 0.5407, "learning_rate": 5e-06, "epoch": 1.150916784203103, "percentage": 38.37, "elapsed_time": "12:44:06", "remaining_time": "20:27:04"}
{"current_steps": 520, "total_steps": 1329, "loss": 0.5333, "learning_rate": 5e-06, "epoch": 1.1734837799717912, "percentage": 39.13, "elapsed_time": "12:58:48", "remaining_time": "20:11:38"}
{"current_steps": 530, "total_steps": 1329, "loss": 0.5395, "learning_rate": 5e-06, "epoch": 1.1960507757404795, "percentage": 39.88, "elapsed_time": "13:13:31", "remaining_time": "19:56:17"}
{"current_steps": 540, "total_steps": 1329, "loss": 0.5371, "learning_rate": 5e-06, "epoch": 1.2186177715091677, "percentage": 40.63, "elapsed_time": "13:28:15", "remaining_time": "19:40:57"}
{"current_steps": 550, "total_steps": 1329, "loss": 0.5419, "learning_rate": 5e-06, "epoch": 1.2411847672778562, "percentage": 41.38, "elapsed_time": "13:42:58", "remaining_time": "19:25:37"}
{"current_steps": 560, "total_steps": 1329, "loss": 0.5438, "learning_rate": 5e-06, "epoch": 1.2637517630465445, "percentage": 42.14, "elapsed_time": "13:57:41", "remaining_time": "19:10:20"}
{"current_steps": 570, "total_steps": 1329, "loss": 0.541, "learning_rate": 5e-06, "epoch": 1.2863187588152327, "percentage": 42.89, "elapsed_time": "14:12:26", "remaining_time": "18:55:05"}
{"current_steps": 580, "total_steps": 1329, "loss": 0.5409, "learning_rate": 5e-06, "epoch": 1.308885754583921, "percentage": 43.64, "elapsed_time": "14:27:08", "remaining_time": "18:39:49"}
{"current_steps": 590, "total_steps": 1329, "loss": 0.5411, "learning_rate": 5e-06, "epoch": 1.3314527503526092, "percentage": 44.39, "elapsed_time": "14:41:53", "remaining_time": "18:24:36"}
{"current_steps": 600, "total_steps": 1329, "loss": 0.5463, "learning_rate": 5e-06, "epoch": 1.3540197461212977, "percentage": 45.15, "elapsed_time": "14:56:37", "remaining_time": "18:09:23"}
{"current_steps": 610, "total_steps": 1329, "loss": 0.5391, "learning_rate": 5e-06, "epoch": 1.376586741889986, "percentage": 45.9, "elapsed_time": "15:11:20", "remaining_time": "17:54:11"}
{"current_steps": 620, "total_steps": 1329, "loss": 0.5433, "learning_rate": 5e-06, "epoch": 1.3991537376586742, "percentage": 46.65, "elapsed_time": "15:26:04", "remaining_time": "17:39:00"}
{"current_steps": 630, "total_steps": 1329, "loss": 0.5413, "learning_rate": 5e-06, "epoch": 1.4217207334273625, "percentage": 47.4, "elapsed_time": "15:40:48", "remaining_time": "17:23:51"}
{"current_steps": 640, "total_steps": 1329, "loss": 0.5409, "learning_rate": 5e-06, "epoch": 1.4442877291960508, "percentage": 48.16, "elapsed_time": "15:55:32", "remaining_time": "17:08:42"}
{"current_steps": 650, "total_steps": 1329, "loss": 0.5418, "learning_rate": 5e-06, "epoch": 1.466854724964739, "percentage": 48.91, "elapsed_time": "16:10:15", "remaining_time": "16:53:32"}
{"current_steps": 660, "total_steps": 1329, "loss": 0.5456, "learning_rate": 5e-06, "epoch": 1.4894217207334273, "percentage": 49.66, "elapsed_time": "16:24:59", "remaining_time": "16:38:25"}
{"current_steps": 670, "total_steps": 1329, "loss": 0.5403, "learning_rate": 5e-06, "epoch": 1.5119887165021155, "percentage": 50.41, "elapsed_time": "16:39:43", "remaining_time": "16:23:19"}
{"current_steps": 680, "total_steps": 1329, "loss": 0.5344, "learning_rate": 5e-06, "epoch": 1.5345557122708038, "percentage": 51.17, "elapsed_time": "16:54:27", "remaining_time": "16:08:12"}
{"current_steps": 690, "total_steps": 1329, "loss": 0.5366, "learning_rate": 5e-06, "epoch": 1.5571227080394923, "percentage": 51.92, "elapsed_time": "17:09:11", "remaining_time": "15:53:06"}
{"current_steps": 700, "total_steps": 1329, "loss": 0.5408, "learning_rate": 5e-06, "epoch": 1.5796897038081805, "percentage": 52.67, "elapsed_time": "17:23:55", "remaining_time": "15:38:02"}
{"current_steps": 710, "total_steps": 1329, "loss": 0.5435, "learning_rate": 5e-06, "epoch": 1.6022566995768688, "percentage": 53.42, "elapsed_time": "17:38:39", "remaining_time": "15:22:58"}
{"current_steps": 720, "total_steps": 1329, "loss": 0.5518, "learning_rate": 5e-06, "epoch": 1.6248236953455573, "percentage": 54.18, "elapsed_time": "17:53:23", "remaining_time": "15:07:54"}
{"current_steps": 730, "total_steps": 1329, "loss": 0.535, "learning_rate": 5e-06, "epoch": 1.6473906911142455, "percentage": 54.93, "elapsed_time": "18:08:06", "remaining_time": "14:52:50"}
{"current_steps": 740, "total_steps": 1329, "loss": 0.5508, "learning_rate": 5e-06, "epoch": 1.6699576868829338, "percentage": 55.68, "elapsed_time": "18:22:50", "remaining_time": "14:37:47"}
{"current_steps": 750, "total_steps": 1329, "loss": 0.5422, "learning_rate": 5e-06, "epoch": 1.692524682651622, "percentage": 56.43, "elapsed_time": "18:37:33", "remaining_time": "14:22:45"}
{"current_steps": 760, "total_steps": 1329, "loss": 0.541, "learning_rate": 5e-06, "epoch": 1.7150916784203103, "percentage": 57.19, "elapsed_time": "18:52:18", "remaining_time": "14:07:44"}
{"current_steps": 770, "total_steps": 1329, "loss": 0.5417, "learning_rate": 5e-06, "epoch": 1.7376586741889986, "percentage": 57.94, "elapsed_time": "19:07:02", "remaining_time": "13:52:43"}
{"current_steps": 780, "total_steps": 1329, "loss": 0.5375, "learning_rate": 5e-06, "epoch": 1.7602256699576868, "percentage": 58.69, "elapsed_time": "19:21:45", "remaining_time": "13:37:42"}
{"current_steps": 790, "total_steps": 1329, "loss": 0.541, "learning_rate": 5e-06, "epoch": 1.782792665726375, "percentage": 59.44, "elapsed_time": "19:36:30", "remaining_time": "13:22:42"}
{"current_steps": 800, "total_steps": 1329, "loss": 0.536, "learning_rate": 5e-06, "epoch": 1.8053596614950633, "percentage": 60.2, "elapsed_time": "19:51:14", "remaining_time": "13:07:42"}
{"current_steps": 810, "total_steps": 1329, "loss": 0.5468, "learning_rate": 5e-06, "epoch": 1.8279266572637518, "percentage": 60.95, "elapsed_time": "20:05:57", "remaining_time": "12:52:42"}
{"current_steps": 820, "total_steps": 1329, "loss": 0.5341, "learning_rate": 5e-06, "epoch": 1.85049365303244, "percentage": 61.7, "elapsed_time": "20:20:41", "remaining_time": "12:37:43"}
{"current_steps": 830, "total_steps": 1329, "loss": 0.5462, "learning_rate": 5e-06, "epoch": 1.8730606488011283, "percentage": 62.45, "elapsed_time": "20:35:24", "remaining_time": "12:22:43"}
{"current_steps": 840, "total_steps": 1329, "loss": 0.5461, "learning_rate": 5e-06, "epoch": 1.8956276445698168, "percentage": 63.21, "elapsed_time": "20:50:07", "remaining_time": "12:07:44"}
{"current_steps": 850, "total_steps": 1329, "loss": 0.534, "learning_rate": 5e-06, "epoch": 1.918194640338505, "percentage": 63.96, "elapsed_time": "21:04:50", "remaining_time": "11:52:46"}
{"current_steps": 860, "total_steps": 1329, "loss": 0.539, "learning_rate": 5e-06, "epoch": 1.9407616361071933, "percentage": 64.71, "elapsed_time": "21:19:33", "remaining_time": "11:37:48"}
{"current_steps": 870, "total_steps": 1329, "loss": 0.5456, "learning_rate": 5e-06, "epoch": 1.9633286318758816, "percentage": 65.46, "elapsed_time": "21:34:17", "remaining_time": "11:22:51"}
{"current_steps": 880, "total_steps": 1329, "loss": 0.5439, "learning_rate": 5e-06, "epoch": 1.9858956276445698, "percentage": 66.22, "elapsed_time": "21:49:02", "remaining_time": "11:07:54"}
{"current_steps": 886, "total_steps": 1329, "eval_loss": 0.5872690677642822, "epoch": 1.9994358251057829, "percentage": 66.67, "elapsed_time": "22:09:31", "remaining_time": "11:04:45"}
{"current_steps": 890, "total_steps": 1329, "loss": 0.5671, "learning_rate": 5e-06, "epoch": 2.008462623413258, "percentage": 66.97, "elapsed_time": "22:16:25", "remaining_time": "10:59:12"}
{"current_steps": 900, "total_steps": 1329, "loss": 0.4778, "learning_rate": 5e-06, "epoch": 2.0310296191819464, "percentage": 67.72, "elapsed_time": "22:31:09", "remaining_time": "10:44:03"}
{"current_steps": 910, "total_steps": 1329, "loss": 0.4796, "learning_rate": 5e-06, "epoch": 2.0535966149506346, "percentage": 68.47, "elapsed_time": "22:45:53", "remaining_time": "10:28:54"}
{"current_steps": 920, "total_steps": 1329, "loss": 0.4721, "learning_rate": 5e-06, "epoch": 2.076163610719323, "percentage": 69.22, "elapsed_time": "23:00:37", "remaining_time": "10:13:46"}
{"current_steps": 930, "total_steps": 1329, "loss": 0.4795, "learning_rate": 5e-06, "epoch": 2.098730606488011, "percentage": 69.98, "elapsed_time": "23:15:20", "remaining_time": "9:58:38"}
{"current_steps": 940, "total_steps": 1329, "loss": 0.4837, "learning_rate": 5e-06, "epoch": 2.1212976022566994, "percentage": 70.73, "elapsed_time": "23:30:04", "remaining_time": "9:43:32"}
{"current_steps": 950, "total_steps": 1329, "loss": 0.4853, "learning_rate": 5e-06, "epoch": 2.143864598025388, "percentage": 71.48, "elapsed_time": "23:44:48", "remaining_time": "9:28:25"}
{"current_steps": 960, "total_steps": 1329, "loss": 0.4816, "learning_rate": 5e-06, "epoch": 2.1664315937940763, "percentage": 72.23, "elapsed_time": "23:59:31", "remaining_time": "9:13:18"}
{"current_steps": 970, "total_steps": 1329, "loss": 0.4798, "learning_rate": 5e-06, "epoch": 2.1889985895627646, "percentage": 72.99, "elapsed_time": "1 day, 0:14:14", "remaining_time": "8:58:13"}
{"current_steps": 980, "total_steps": 1329, "loss": 0.4748, "learning_rate": 5e-06, "epoch": 2.211565585331453, "percentage": 73.74, "elapsed_time": "1 day, 0:28:57", "remaining_time": "8:43:07"}
{"current_steps": 990, "total_steps": 1329, "loss": 0.4747, "learning_rate": 5e-06, "epoch": 2.234132581100141, "percentage": 74.49, "elapsed_time": "1 day, 0:43:41", "remaining_time": "8:28:02"}
{"current_steps": 1000, "total_steps": 1329, "loss": 0.4836, "learning_rate": 5e-06, "epoch": 2.2566995768688294, "percentage": 75.24, "elapsed_time": "1 day, 0:58:23", "remaining_time": "8:12:58"}
{"current_steps": 1010, "total_steps": 1329, "loss": 0.4899, "learning_rate": 5e-06, "epoch": 2.2792665726375176, "percentage": 76.0, "elapsed_time": "1 day, 1:13:07", "remaining_time": "7:57:54"}
{"current_steps": 1020, "total_steps": 1329, "loss": 0.4843, "learning_rate": 5e-06, "epoch": 2.301833568406206, "percentage": 76.75, "elapsed_time": "1 day, 1:27:51", "remaining_time": "7:42:51"}
{"current_steps": 1030, "total_steps": 1329, "loss": 0.4887, "learning_rate": 5e-06, "epoch": 2.324400564174894, "percentage": 77.5, "elapsed_time": "1 day, 1:42:35", "remaining_time": "7:27:48"}
{"current_steps": 1040, "total_steps": 1329, "loss": 0.4865, "learning_rate": 5e-06, "epoch": 2.3469675599435824, "percentage": 78.25, "elapsed_time": "1 day, 1:57:18", "remaining_time": "7:12:45"}
{"current_steps": 1050, "total_steps": 1329, "loss": 0.4844, "learning_rate": 5e-06, "epoch": 2.3695345557122707, "percentage": 79.01, "elapsed_time": "1 day, 2:12:01", "remaining_time": "6:57:42"}
{"current_steps": 1060, "total_steps": 1329, "loss": 0.4837, "learning_rate": 5e-06, "epoch": 2.392101551480959, "percentage": 79.76, "elapsed_time": "1 day, 2:26:46", "remaining_time": "6:42:40"}
{"current_steps": 1070, "total_steps": 1329, "loss": 0.4849, "learning_rate": 5e-06, "epoch": 2.414668547249647, "percentage": 80.51, "elapsed_time": "1 day, 2:41:30", "remaining_time": "6:27:39"}
{"current_steps": 1080, "total_steps": 1329, "loss": 0.4883, "learning_rate": 5e-06, "epoch": 2.4372355430183354, "percentage": 81.26, "elapsed_time": "1 day, 2:56:12", "remaining_time": "6:12:37"}
{"current_steps": 1090, "total_steps": 1329, "loss": 0.4866, "learning_rate": 5e-06, "epoch": 2.459802538787024, "percentage": 82.02, "elapsed_time": "1 day, 3:10:56", "remaining_time": "5:57:36"}
{"current_steps": 1100, "total_steps": 1329, "loss": 0.4897, "learning_rate": 5e-06, "epoch": 2.4823695345557124, "percentage": 82.77, "elapsed_time": "1 day, 3:25:39", "remaining_time": "5:42:35"}
{"current_steps": 1110, "total_steps": 1329, "loss": 0.4928, "learning_rate": 5e-06, "epoch": 2.5049365303244007, "percentage": 83.52, "elapsed_time": "1 day, 3:40:23", "remaining_time": "5:27:35"}
{"current_steps": 1120, "total_steps": 1329, "loss": 0.4834, "learning_rate": 5e-06, "epoch": 2.527503526093089, "percentage": 84.27, "elapsed_time": "1 day, 3:55:08", "remaining_time": "5:12:35"}
{"current_steps": 1130, "total_steps": 1329, "loss": 0.4884, "learning_rate": 5e-06, "epoch": 2.550070521861777, "percentage": 85.03, "elapsed_time": "1 day, 4:09:51", "remaining_time": "4:57:35"}
{"current_steps": 1140, "total_steps": 1329, "loss": 0.4867, "learning_rate": 5e-06, "epoch": 2.5726375176304654, "percentage": 85.78, "elapsed_time": "1 day, 4:24:35", "remaining_time": "4:42:36"}
{"current_steps": 1150, "total_steps": 1329, "loss": 0.4875, "learning_rate": 5e-06, "epoch": 2.5952045133991537, "percentage": 86.53, "elapsed_time": "1 day, 4:39:20", "remaining_time": "4:27:37"}
{"current_steps": 1160, "total_steps": 1329, "loss": 0.4911, "learning_rate": 5e-06, "epoch": 2.617771509167842, "percentage": 87.28, "elapsed_time": "1 day, 4:54:05", "remaining_time": "4:12:38"}
{"current_steps": 1170, "total_steps": 1329, "loss": 0.4803, "learning_rate": 5e-06, "epoch": 2.64033850493653, "percentage": 88.04, "elapsed_time": "1 day, 5:08:48", "remaining_time": "3:57:39"}
{"current_steps": 1180, "total_steps": 1329, "loss": 0.491, "learning_rate": 5e-06, "epoch": 2.6629055007052185, "percentage": 88.79, "elapsed_time": "1 day, 5:23:34", "remaining_time": "3:42:41"}
{"current_steps": 1190, "total_steps": 1329, "loss": 0.4917, "learning_rate": 5e-06, "epoch": 2.685472496473907, "percentage": 89.54, "elapsed_time": "1 day, 5:38:19", "remaining_time": "3:27:43"}
{"current_steps": 1200, "total_steps": 1329, "loss": 0.4903, "learning_rate": 5e-06, "epoch": 2.7080394922425954, "percentage": 90.29, "elapsed_time": "1 day, 5:53:04", "remaining_time": "3:12:45"}
{"current_steps": 1210, "total_steps": 1329, "loss": 0.4916, "learning_rate": 5e-06, "epoch": 2.7306064880112837, "percentage": 91.05, "elapsed_time": "1 day, 6:07:49", "remaining_time": "2:57:47"}
{"current_steps": 1220, "total_steps": 1329, "loss": 0.4882, "learning_rate": 5e-06, "epoch": 2.753173483779972, "percentage": 91.8, "elapsed_time": "1 day, 6:22:35", "remaining_time": "2:42:50"}
{"current_steps": 1230, "total_steps": 1329, "loss": 0.4947, "learning_rate": 5e-06, "epoch": 2.77574047954866, "percentage": 92.55, "elapsed_time": "1 day, 6:37:20", "remaining_time": "2:27:53"}
{"current_steps": 1240, "total_steps": 1329, "loss": 0.4929, "learning_rate": 5e-06, "epoch": 2.7983074753173485, "percentage": 93.3, "elapsed_time": "1 day, 6:52:04", "remaining_time": "2:12:55"}
{"current_steps": 1250, "total_steps": 1329, "loss": 0.4931, "learning_rate": 5e-06, "epoch": 2.8208744710860367, "percentage": 94.06, "elapsed_time": "1 day, 7:06:49", "remaining_time": "1:57:59"}
{"current_steps": 1260, "total_steps": 1329, "loss": 0.4949, "learning_rate": 5e-06, "epoch": 2.843441466854725, "percentage": 94.81, "elapsed_time": "1 day, 7:21:34", "remaining_time": "1:43:02"}
{"current_steps": 1270, "total_steps": 1329, "loss": 0.4918, "learning_rate": 5e-06, "epoch": 2.8660084626234132, "percentage": 95.56, "elapsed_time": "1 day, 7:36:19", "remaining_time": "1:28:05"}
{"current_steps": 1280, "total_steps": 1329, "loss": 0.4899, "learning_rate": 5e-06, "epoch": 2.8885754583921015, "percentage": 96.31, "elapsed_time": "1 day, 7:51:05", "remaining_time": "1:13:09"}
{"current_steps": 1290, "total_steps": 1329, "loss": 0.4937, "learning_rate": 5e-06, "epoch": 2.9111424541607898, "percentage": 97.07, "elapsed_time": "1 day, 8:05:49", "remaining_time": "0:58:13"}
{"current_steps": 1300, "total_steps": 1329, "loss": 0.4893, "learning_rate": 5e-06, "epoch": 2.933709449929478, "percentage": 97.82, "elapsed_time": "1 day, 8:20:35", "remaining_time": "0:43:17"}
{"current_steps": 1310, "total_steps": 1329, "loss": 0.4932, "learning_rate": 5e-06, "epoch": 2.9562764456981663, "percentage": 98.57, "elapsed_time": "1 day, 8:35:19", "remaining_time": "0:28:21"}
{"current_steps": 1320, "total_steps": 1329, "loss": 0.4896, "learning_rate": 5e-06, "epoch": 2.9788434414668545, "percentage": 99.32, "elapsed_time": "1 day, 8:50:04", "remaining_time": "0:13:25"}
{"current_steps": 1329, "total_steps": 1329, "eval_loss": 0.6003954410552979, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:16:05", "remaining_time": "0:00:00"}
{"current_steps": 1329, "total_steps": 1329, "epoch": 2.9991537376586743, "percentage": 100.0, "elapsed_time": "1 day, 9:17:44", "remaining_time": "0:00:00"}