hp_ablations_qwen_adambeta1_0.9 / trainer_log.jsonl
sedrickkeh's picture
Training in progress, epoch 2
383a365 verified
{"current_steps": 10, "total_steps": 1314, "loss": 0.7556, "learning_rate": 5e-06, "epoch": 0.022792022792022793, "percentage": 0.76, "elapsed_time": "0:08:44", "remaining_time": "19:00:01"}
{"current_steps": 20, "total_steps": 1314, "loss": 0.7003, "learning_rate": 5e-06, "epoch": 0.045584045584045586, "percentage": 1.52, "elapsed_time": "0:17:24", "remaining_time": "18:46:41"}
{"current_steps": 30, "total_steps": 1314, "loss": 0.6796, "learning_rate": 5e-06, "epoch": 0.06837606837606838, "percentage": 2.28, "elapsed_time": "0:26:04", "remaining_time": "18:36:17"}
{"current_steps": 40, "total_steps": 1314, "loss": 0.6794, "learning_rate": 5e-06, "epoch": 0.09116809116809117, "percentage": 3.04, "elapsed_time": "0:34:44", "remaining_time": "18:26:30"}
{"current_steps": 50, "total_steps": 1314, "loss": 0.6755, "learning_rate": 5e-06, "epoch": 0.11396011396011396, "percentage": 3.81, "elapsed_time": "0:43:24", "remaining_time": "18:17:20"}
{"current_steps": 60, "total_steps": 1314, "loss": 0.6551, "learning_rate": 5e-06, "epoch": 0.13675213675213677, "percentage": 4.57, "elapsed_time": "0:52:04", "remaining_time": "18:08:15"}
{"current_steps": 70, "total_steps": 1314, "loss": 0.652, "learning_rate": 5e-06, "epoch": 0.15954415954415954, "percentage": 5.33, "elapsed_time": "1:00:43", "remaining_time": "17:59:15"}
{"current_steps": 80, "total_steps": 1314, "loss": 0.655, "learning_rate": 5e-06, "epoch": 0.18233618233618235, "percentage": 6.09, "elapsed_time": "1:09:23", "remaining_time": "17:50:25"}
{"current_steps": 90, "total_steps": 1314, "loss": 0.6397, "learning_rate": 5e-06, "epoch": 0.20512820512820512, "percentage": 6.85, "elapsed_time": "1:18:03", "remaining_time": "17:41:40"}
{"current_steps": 100, "total_steps": 1314, "loss": 0.641, "learning_rate": 5e-06, "epoch": 0.22792022792022792, "percentage": 7.61, "elapsed_time": "1:26:44", "remaining_time": "17:32:59"}
{"current_steps": 110, "total_steps": 1314, "loss": 0.6323, "learning_rate": 5e-06, "epoch": 0.25071225071225073, "percentage": 8.37, "elapsed_time": "1:35:24", "remaining_time": "17:24:22"}
{"current_steps": 120, "total_steps": 1314, "loss": 0.6406, "learning_rate": 5e-06, "epoch": 0.27350427350427353, "percentage": 9.13, "elapsed_time": "1:44:05", "remaining_time": "17:15:40"}
{"current_steps": 130, "total_steps": 1314, "loss": 0.6416, "learning_rate": 5e-06, "epoch": 0.2962962962962963, "percentage": 9.89, "elapsed_time": "1:52:45", "remaining_time": "17:06:54"}
{"current_steps": 140, "total_steps": 1314, "loss": 0.6437, "learning_rate": 5e-06, "epoch": 0.3190883190883191, "percentage": 10.65, "elapsed_time": "2:01:25", "remaining_time": "16:58:11"}
{"current_steps": 150, "total_steps": 1314, "loss": 0.6362, "learning_rate": 5e-06, "epoch": 0.3418803418803419, "percentage": 11.42, "elapsed_time": "2:10:05", "remaining_time": "16:49:31"}
{"current_steps": 160, "total_steps": 1314, "loss": 0.6414, "learning_rate": 5e-06, "epoch": 0.3646723646723647, "percentage": 12.18, "elapsed_time": "2:18:45", "remaining_time": "16:40:49"}
{"current_steps": 170, "total_steps": 1314, "loss": 0.6332, "learning_rate": 5e-06, "epoch": 0.38746438746438744, "percentage": 12.94, "elapsed_time": "2:27:25", "remaining_time": "16:32:07"}
{"current_steps": 180, "total_steps": 1314, "loss": 0.6387, "learning_rate": 5e-06, "epoch": 0.41025641025641024, "percentage": 13.7, "elapsed_time": "2:36:05", "remaining_time": "16:23:25"}
{"current_steps": 190, "total_steps": 1314, "loss": 0.6313, "learning_rate": 5e-06, "epoch": 0.43304843304843305, "percentage": 14.46, "elapsed_time": "2:44:46", "remaining_time": "16:14:44"}
{"current_steps": 200, "total_steps": 1314, "loss": 0.6394, "learning_rate": 5e-06, "epoch": 0.45584045584045585, "percentage": 15.22, "elapsed_time": "2:53:26", "remaining_time": "16:06:02"}
{"current_steps": 210, "total_steps": 1314, "loss": 0.6357, "learning_rate": 5e-06, "epoch": 0.47863247863247865, "percentage": 15.98, "elapsed_time": "3:02:06", "remaining_time": "15:57:21"}
{"current_steps": 220, "total_steps": 1314, "loss": 0.6394, "learning_rate": 5e-06, "epoch": 0.5014245014245015, "percentage": 16.74, "elapsed_time": "3:10:45", "remaining_time": "15:48:35"}
{"current_steps": 230, "total_steps": 1314, "loss": 0.635, "learning_rate": 5e-06, "epoch": 0.5242165242165242, "percentage": 17.5, "elapsed_time": "3:19:25", "remaining_time": "15:39:53"}
{"current_steps": 240, "total_steps": 1314, "loss": 0.6309, "learning_rate": 5e-06, "epoch": 0.5470085470085471, "percentage": 18.26, "elapsed_time": "3:28:05", "remaining_time": "15:31:11"}
{"current_steps": 250, "total_steps": 1314, "loss": 0.6358, "learning_rate": 5e-06, "epoch": 0.5698005698005698, "percentage": 19.03, "elapsed_time": "3:36:44", "remaining_time": "15:22:28"}
{"current_steps": 260, "total_steps": 1314, "loss": 0.634, "learning_rate": 5e-06, "epoch": 0.5925925925925926, "percentage": 19.79, "elapsed_time": "3:45:24", "remaining_time": "15:13:44"}
{"current_steps": 270, "total_steps": 1314, "loss": 0.6311, "learning_rate": 5e-06, "epoch": 0.6153846153846154, "percentage": 20.55, "elapsed_time": "3:54:03", "remaining_time": "15:05:02"}
{"current_steps": 280, "total_steps": 1314, "loss": 0.6419, "learning_rate": 5e-06, "epoch": 0.6381766381766382, "percentage": 21.31, "elapsed_time": "4:02:44", "remaining_time": "14:56:23"}
{"current_steps": 290, "total_steps": 1314, "loss": 0.6306, "learning_rate": 5e-06, "epoch": 0.6609686609686609, "percentage": 22.07, "elapsed_time": "4:11:25", "remaining_time": "14:47:46"}
{"current_steps": 300, "total_steps": 1314, "loss": 0.6287, "learning_rate": 5e-06, "epoch": 0.6837606837606838, "percentage": 22.83, "elapsed_time": "4:20:06", "remaining_time": "14:39:09"}
{"current_steps": 310, "total_steps": 1314, "loss": 0.6286, "learning_rate": 5e-06, "epoch": 0.7065527065527065, "percentage": 23.59, "elapsed_time": "4:28:47", "remaining_time": "14:30:32"}
{"current_steps": 320, "total_steps": 1314, "loss": 0.6291, "learning_rate": 5e-06, "epoch": 0.7293447293447294, "percentage": 24.35, "elapsed_time": "4:37:28", "remaining_time": "14:21:53"}
{"current_steps": 330, "total_steps": 1314, "loss": 0.6258, "learning_rate": 5e-06, "epoch": 0.7521367521367521, "percentage": 25.11, "elapsed_time": "4:46:08", "remaining_time": "14:13:14"}
{"current_steps": 340, "total_steps": 1314, "loss": 0.6316, "learning_rate": 5e-06, "epoch": 0.7749287749287749, "percentage": 25.88, "elapsed_time": "4:54:49", "remaining_time": "14:04:35"}
{"current_steps": 350, "total_steps": 1314, "loss": 0.6387, "learning_rate": 5e-06, "epoch": 0.7977207977207977, "percentage": 26.64, "elapsed_time": "5:03:29", "remaining_time": "13:55:55"}
{"current_steps": 360, "total_steps": 1314, "loss": 0.6304, "learning_rate": 5e-06, "epoch": 0.8205128205128205, "percentage": 27.4, "elapsed_time": "5:12:10", "remaining_time": "13:47:15"}
{"current_steps": 370, "total_steps": 1314, "loss": 0.6242, "learning_rate": 5e-06, "epoch": 0.8433048433048433, "percentage": 28.16, "elapsed_time": "5:20:51", "remaining_time": "13:38:36"}
{"current_steps": 380, "total_steps": 1314, "loss": 0.6365, "learning_rate": 5e-06, "epoch": 0.8660968660968661, "percentage": 28.92, "elapsed_time": "5:29:30", "remaining_time": "13:29:54"}
{"current_steps": 390, "total_steps": 1314, "loss": 0.635, "learning_rate": 5e-06, "epoch": 0.8888888888888888, "percentage": 29.68, "elapsed_time": "5:38:10", "remaining_time": "13:21:13"}
{"current_steps": 400, "total_steps": 1314, "loss": 0.6382, "learning_rate": 5e-06, "epoch": 0.9116809116809117, "percentage": 30.44, "elapsed_time": "5:46:50", "remaining_time": "13:12:32"}
{"current_steps": 410, "total_steps": 1314, "loss": 0.6345, "learning_rate": 5e-06, "epoch": 0.9344729344729344, "percentage": 31.2, "elapsed_time": "5:55:31", "remaining_time": "13:03:53"}
{"current_steps": 420, "total_steps": 1314, "loss": 0.6214, "learning_rate": 5e-06, "epoch": 0.9572649572649573, "percentage": 31.96, "elapsed_time": "6:04:12", "remaining_time": "12:55:13"}
{"current_steps": 430, "total_steps": 1314, "loss": 0.6344, "learning_rate": 5e-06, "epoch": 0.98005698005698, "percentage": 32.72, "elapsed_time": "6:12:52", "remaining_time": "12:46:33"}
{"current_steps": 438, "total_steps": 1314, "eval_loss": 0.6250635981559753, "epoch": 0.9982905982905983, "percentage": 33.33, "elapsed_time": "6:27:33", "remaining_time": "12:55:06"}
{"current_steps": 440, "total_steps": 1314, "loss": 0.6471, "learning_rate": 5e-06, "epoch": 1.0034188034188034, "percentage": 33.49, "elapsed_time": "6:29:57", "remaining_time": "12:54:35"}
{"current_steps": 450, "total_steps": 1314, "loss": 0.6014, "learning_rate": 5e-06, "epoch": 1.0262108262108263, "percentage": 34.25, "elapsed_time": "6:38:37", "remaining_time": "12:45:22"}
{"current_steps": 460, "total_steps": 1314, "loss": 0.5952, "learning_rate": 5e-06, "epoch": 1.049002849002849, "percentage": 35.01, "elapsed_time": "6:47:18", "remaining_time": "12:36:10"}
{"current_steps": 470, "total_steps": 1314, "loss": 0.596, "learning_rate": 5e-06, "epoch": 1.0717948717948718, "percentage": 35.77, "elapsed_time": "6:55:59", "remaining_time": "12:27:00"}
{"current_steps": 480, "total_steps": 1314, "loss": 0.6035, "learning_rate": 5e-06, "epoch": 1.0945868945868946, "percentage": 36.53, "elapsed_time": "7:04:40", "remaining_time": "12:17:51"}
{"current_steps": 490, "total_steps": 1314, "loss": 0.6086, "learning_rate": 5e-06, "epoch": 1.1173789173789175, "percentage": 37.29, "elapsed_time": "7:13:21", "remaining_time": "12:08:44"}
{"current_steps": 500, "total_steps": 1314, "loss": 0.5984, "learning_rate": 5e-06, "epoch": 1.1401709401709401, "percentage": 38.05, "elapsed_time": "7:22:01", "remaining_time": "11:59:37"}
{"current_steps": 510, "total_steps": 1314, "loss": 0.5963, "learning_rate": 5e-06, "epoch": 1.162962962962963, "percentage": 38.81, "elapsed_time": "7:30:42", "remaining_time": "11:50:31"}
{"current_steps": 520, "total_steps": 1314, "loss": 0.6021, "learning_rate": 5e-06, "epoch": 1.1857549857549858, "percentage": 39.57, "elapsed_time": "7:39:22", "remaining_time": "11:41:25"}
{"current_steps": 530, "total_steps": 1314, "loss": 0.5951, "learning_rate": 5e-06, "epoch": 1.2085470085470085, "percentage": 40.33, "elapsed_time": "7:48:03", "remaining_time": "11:32:22"}
{"current_steps": 540, "total_steps": 1314, "loss": 0.6053, "learning_rate": 5e-06, "epoch": 1.2313390313390313, "percentage": 41.1, "elapsed_time": "7:56:44", "remaining_time": "11:23:19"}
{"current_steps": 550, "total_steps": 1314, "loss": 0.6019, "learning_rate": 5e-06, "epoch": 1.2541310541310542, "percentage": 41.86, "elapsed_time": "8:05:25", "remaining_time": "11:14:18"}
{"current_steps": 560, "total_steps": 1314, "loss": 0.6034, "learning_rate": 5e-06, "epoch": 1.2769230769230768, "percentage": 42.62, "elapsed_time": "8:14:07", "remaining_time": "11:05:17"}
{"current_steps": 570, "total_steps": 1314, "loss": 0.5975, "learning_rate": 5e-06, "epoch": 1.2997150997150997, "percentage": 43.38, "elapsed_time": "8:22:48", "remaining_time": "10:56:17"}
{"current_steps": 580, "total_steps": 1314, "loss": 0.5959, "learning_rate": 5e-06, "epoch": 1.3225071225071225, "percentage": 44.14, "elapsed_time": "8:31:30", "remaining_time": "10:47:18"}
{"current_steps": 590, "total_steps": 1314, "loss": 0.5957, "learning_rate": 5e-06, "epoch": 1.3452991452991454, "percentage": 44.9, "elapsed_time": "8:40:12", "remaining_time": "10:38:20"}
{"current_steps": 600, "total_steps": 1314, "loss": 0.6013, "learning_rate": 5e-06, "epoch": 1.368091168091168, "percentage": 45.66, "elapsed_time": "8:48:53", "remaining_time": "10:29:23"}
{"current_steps": 610, "total_steps": 1314, "loss": 0.6013, "learning_rate": 5e-06, "epoch": 1.390883190883191, "percentage": 46.42, "elapsed_time": "8:57:35", "remaining_time": "10:20:26"}
{"current_steps": 620, "total_steps": 1314, "loss": 0.6039, "learning_rate": 5e-06, "epoch": 1.4136752136752135, "percentage": 47.18, "elapsed_time": "9:06:17", "remaining_time": "10:11:29"}
{"current_steps": 630, "total_steps": 1314, "loss": 0.6019, "learning_rate": 5e-06, "epoch": 1.4364672364672364, "percentage": 47.95, "elapsed_time": "9:14:59", "remaining_time": "10:02:33"}
{"current_steps": 640, "total_steps": 1314, "loss": 0.5968, "learning_rate": 5e-06, "epoch": 1.4592592592592593, "percentage": 48.71, "elapsed_time": "9:23:39", "remaining_time": "9:53:36"}
{"current_steps": 650, "total_steps": 1314, "loss": 0.5947, "learning_rate": 5e-06, "epoch": 1.4820512820512821, "percentage": 49.47, "elapsed_time": "9:32:20", "remaining_time": "9:44:39"}
{"current_steps": 660, "total_steps": 1314, "loss": 0.5944, "learning_rate": 5e-06, "epoch": 1.504843304843305, "percentage": 50.23, "elapsed_time": "9:41:01", "remaining_time": "9:35:44"}
{"current_steps": 670, "total_steps": 1314, "loss": 0.602, "learning_rate": 5e-06, "epoch": 1.5276353276353276, "percentage": 50.99, "elapsed_time": "9:49:41", "remaining_time": "9:26:48"}
{"current_steps": 680, "total_steps": 1314, "loss": 0.5906, "learning_rate": 5e-06, "epoch": 1.5504273504273505, "percentage": 51.75, "elapsed_time": "9:58:22", "remaining_time": "9:17:53"}
{"current_steps": 690, "total_steps": 1314, "loss": 0.5945, "learning_rate": 5e-06, "epoch": 1.573219373219373, "percentage": 52.51, "elapsed_time": "10:07:02", "remaining_time": "9:08:58"}
{"current_steps": 700, "total_steps": 1314, "loss": 0.6101, "learning_rate": 5e-06, "epoch": 1.596011396011396, "percentage": 53.27, "elapsed_time": "10:15:42", "remaining_time": "9:00:04"}
{"current_steps": 710, "total_steps": 1314, "loss": 0.5971, "learning_rate": 5e-06, "epoch": 1.6188034188034188, "percentage": 54.03, "elapsed_time": "10:24:23", "remaining_time": "8:51:10"}
{"current_steps": 720, "total_steps": 1314, "loss": 0.6017, "learning_rate": 5e-06, "epoch": 1.6415954415954417, "percentage": 54.79, "elapsed_time": "10:33:04", "remaining_time": "8:42:16"}
{"current_steps": 730, "total_steps": 1314, "loss": 0.5948, "learning_rate": 5e-06, "epoch": 1.6643874643874645, "percentage": 55.56, "elapsed_time": "10:41:45", "remaining_time": "8:33:24"}
{"current_steps": 740, "total_steps": 1314, "loss": 0.5959, "learning_rate": 5e-06, "epoch": 1.6871794871794872, "percentage": 56.32, "elapsed_time": "10:50:25", "remaining_time": "8:24:31"}
{"current_steps": 750, "total_steps": 1314, "loss": 0.5933, "learning_rate": 5e-06, "epoch": 1.7099715099715098, "percentage": 57.08, "elapsed_time": "10:59:06", "remaining_time": "8:15:39"}
{"current_steps": 760, "total_steps": 1314, "loss": 0.5972, "learning_rate": 5e-06, "epoch": 1.7327635327635327, "percentage": 57.84, "elapsed_time": "11:07:46", "remaining_time": "8:06:46"}
{"current_steps": 770, "total_steps": 1314, "loss": 0.5936, "learning_rate": 5e-06, "epoch": 1.7555555555555555, "percentage": 58.6, "elapsed_time": "11:16:26", "remaining_time": "7:57:54"}
{"current_steps": 780, "total_steps": 1314, "loss": 0.6008, "learning_rate": 5e-06, "epoch": 1.7783475783475784, "percentage": 59.36, "elapsed_time": "11:25:06", "remaining_time": "7:49:02"}
{"current_steps": 790, "total_steps": 1314, "loss": 0.5984, "learning_rate": 5e-06, "epoch": 1.8011396011396013, "percentage": 60.12, "elapsed_time": "11:33:46", "remaining_time": "7:40:10"}
{"current_steps": 800, "total_steps": 1314, "loss": 0.5851, "learning_rate": 5e-06, "epoch": 1.823931623931624, "percentage": 60.88, "elapsed_time": "11:42:26", "remaining_time": "7:31:19"}
{"current_steps": 810, "total_steps": 1314, "loss": 0.593, "learning_rate": 5e-06, "epoch": 1.8467236467236468, "percentage": 61.64, "elapsed_time": "11:51:06", "remaining_time": "7:22:28"}
{"current_steps": 820, "total_steps": 1314, "loss": 0.5908, "learning_rate": 5e-06, "epoch": 1.8695156695156694, "percentage": 62.4, "elapsed_time": "11:59:46", "remaining_time": "7:13:37"}
{"current_steps": 830, "total_steps": 1314, "loss": 0.5946, "learning_rate": 5e-06, "epoch": 1.8923076923076922, "percentage": 63.17, "elapsed_time": "12:08:26", "remaining_time": "7:04:46"}
{"current_steps": 840, "total_steps": 1314, "loss": 0.599, "learning_rate": 5e-06, "epoch": 1.915099715099715, "percentage": 63.93, "elapsed_time": "12:17:06", "remaining_time": "6:55:56"}
{"current_steps": 850, "total_steps": 1314, "loss": 0.5972, "learning_rate": 5e-06, "epoch": 1.937891737891738, "percentage": 64.69, "elapsed_time": "12:25:46", "remaining_time": "6:47:06"}
{"current_steps": 860, "total_steps": 1314, "loss": 0.5985, "learning_rate": 5e-06, "epoch": 1.9606837606837608, "percentage": 65.45, "elapsed_time": "12:34:27", "remaining_time": "6:38:16"}
{"current_steps": 870, "total_steps": 1314, "loss": 0.5959, "learning_rate": 5e-06, "epoch": 1.9834757834757835, "percentage": 66.21, "elapsed_time": "12:43:07", "remaining_time": "6:29:27"}
{"current_steps": 877, "total_steps": 1314, "eval_loss": 0.6187016367912292, "epoch": 1.9994301994301993, "percentage": 66.74, "elapsed_time": "12:56:28", "remaining_time": "6:26:54"}
{"current_steps": 880, "total_steps": 1314, "loss": 0.6153, "learning_rate": 5e-06, "epoch": 2.006837606837607, "percentage": 66.97, "elapsed_time": "13:00:16", "remaining_time": "6:24:48"}
{"current_steps": 890, "total_steps": 1314, "loss": 0.5544, "learning_rate": 5e-06, "epoch": 2.0296296296296297, "percentage": 67.73, "elapsed_time": "13:08:57", "remaining_time": "6:15:51"}
{"current_steps": 900, "total_steps": 1314, "loss": 0.5695, "learning_rate": 5e-06, "epoch": 2.0524216524216525, "percentage": 68.49, "elapsed_time": "13:17:39", "remaining_time": "6:06:55"}
{"current_steps": 910, "total_steps": 1314, "loss": 0.5676, "learning_rate": 5e-06, "epoch": 2.0752136752136754, "percentage": 69.25, "elapsed_time": "13:26:20", "remaining_time": "5:57:58"}
{"current_steps": 920, "total_steps": 1314, "loss": 0.5708, "learning_rate": 5e-06, "epoch": 2.098005698005698, "percentage": 70.02, "elapsed_time": "13:35:02", "remaining_time": "5:49:03"}
{"current_steps": 930, "total_steps": 1314, "loss": 0.5688, "learning_rate": 5e-06, "epoch": 2.1207977207977207, "percentage": 70.78, "elapsed_time": "13:43:44", "remaining_time": "5:40:07"}
{"current_steps": 940, "total_steps": 1314, "loss": 0.5635, "learning_rate": 5e-06, "epoch": 2.1435897435897435, "percentage": 71.54, "elapsed_time": "13:52:26", "remaining_time": "5:31:12"}
{"current_steps": 950, "total_steps": 1314, "loss": 0.5643, "learning_rate": 5e-06, "epoch": 2.1663817663817664, "percentage": 72.3, "elapsed_time": "14:01:07", "remaining_time": "5:22:17"}
{"current_steps": 960, "total_steps": 1314, "loss": 0.5694, "learning_rate": 5e-06, "epoch": 2.1891737891737892, "percentage": 73.06, "elapsed_time": "14:09:49", "remaining_time": "5:13:22"}
{"current_steps": 970, "total_steps": 1314, "loss": 0.5689, "learning_rate": 5e-06, "epoch": 2.211965811965812, "percentage": 73.82, "elapsed_time": "14:18:30", "remaining_time": "5:04:27"}
{"current_steps": 980, "total_steps": 1314, "loss": 0.5585, "learning_rate": 5e-06, "epoch": 2.234757834757835, "percentage": 74.58, "elapsed_time": "14:27:12", "remaining_time": "4:55:33"}
{"current_steps": 990, "total_steps": 1314, "loss": 0.5676, "learning_rate": 5e-06, "epoch": 2.2575498575498574, "percentage": 75.34, "elapsed_time": "14:35:54", "remaining_time": "4:46:39"}
{"current_steps": 1000, "total_steps": 1314, "loss": 0.5679, "learning_rate": 5e-06, "epoch": 2.2803418803418802, "percentage": 76.1, "elapsed_time": "14:44:36", "remaining_time": "4:37:45"}
{"current_steps": 1010, "total_steps": 1314, "loss": 0.5654, "learning_rate": 5e-06, "epoch": 2.303133903133903, "percentage": 76.86, "elapsed_time": "14:53:17", "remaining_time": "4:28:52"}
{"current_steps": 1020, "total_steps": 1314, "loss": 0.5684, "learning_rate": 5e-06, "epoch": 2.325925925925926, "percentage": 77.63, "elapsed_time": "15:01:59", "remaining_time": "4:19:59"}
{"current_steps": 1030, "total_steps": 1314, "loss": 0.5697, "learning_rate": 5e-06, "epoch": 2.348717948717949, "percentage": 78.39, "elapsed_time": "15:10:41", "remaining_time": "4:11:06"}
{"current_steps": 1040, "total_steps": 1314, "loss": 0.5677, "learning_rate": 5e-06, "epoch": 2.3715099715099717, "percentage": 79.15, "elapsed_time": "15:19:23", "remaining_time": "4:02:13"}
{"current_steps": 1050, "total_steps": 1314, "loss": 0.5671, "learning_rate": 5e-06, "epoch": 2.394301994301994, "percentage": 79.91, "elapsed_time": "15:28:05", "remaining_time": "3:53:20"}
{"current_steps": 1060, "total_steps": 1314, "loss": 0.5754, "learning_rate": 5e-06, "epoch": 2.417094017094017, "percentage": 80.67, "elapsed_time": "15:36:47", "remaining_time": "3:44:28"}
{"current_steps": 1070, "total_steps": 1314, "loss": 0.5782, "learning_rate": 5e-06, "epoch": 2.43988603988604, "percentage": 81.43, "elapsed_time": "15:45:28", "remaining_time": "3:35:36"}
{"current_steps": 1080, "total_steps": 1314, "loss": 0.5735, "learning_rate": 5e-06, "epoch": 2.4626780626780627, "percentage": 82.19, "elapsed_time": "15:54:10", "remaining_time": "3:26:44"}
{"current_steps": 1090, "total_steps": 1314, "loss": 0.5711, "learning_rate": 5e-06, "epoch": 2.4854700854700855, "percentage": 82.95, "elapsed_time": "16:02:52", "remaining_time": "3:17:52"}
{"current_steps": 1100, "total_steps": 1314, "loss": 0.5776, "learning_rate": 5e-06, "epoch": 2.5082621082621084, "percentage": 83.71, "elapsed_time": "16:11:34", "remaining_time": "3:09:00"}
{"current_steps": 1110, "total_steps": 1314, "loss": 0.5711, "learning_rate": 5e-06, "epoch": 2.5310541310541312, "percentage": 84.47, "elapsed_time": "16:20:16", "remaining_time": "3:00:09"}
{"current_steps": 1120, "total_steps": 1314, "loss": 0.5692, "learning_rate": 5e-06, "epoch": 2.5538461538461537, "percentage": 85.24, "elapsed_time": "16:28:59", "remaining_time": "2:51:18"}
{"current_steps": 1130, "total_steps": 1314, "loss": 0.5687, "learning_rate": 5e-06, "epoch": 2.5766381766381765, "percentage": 86.0, "elapsed_time": "16:37:41", "remaining_time": "2:42:27"}
{"current_steps": 1140, "total_steps": 1314, "loss": 0.5683, "learning_rate": 5e-06, "epoch": 2.5994301994301994, "percentage": 86.76, "elapsed_time": "16:46:22", "remaining_time": "2:33:36"}
{"current_steps": 1150, "total_steps": 1314, "loss": 0.5686, "learning_rate": 5e-06, "epoch": 2.6222222222222222, "percentage": 87.52, "elapsed_time": "16:55:05", "remaining_time": "2:24:45"}
{"current_steps": 1160, "total_steps": 1314, "loss": 0.5698, "learning_rate": 5e-06, "epoch": 2.645014245014245, "percentage": 88.28, "elapsed_time": "17:03:47", "remaining_time": "2:15:55"}
{"current_steps": 1170, "total_steps": 1314, "loss": 0.5726, "learning_rate": 5e-06, "epoch": 2.667806267806268, "percentage": 89.04, "elapsed_time": "17:12:29", "remaining_time": "2:07:04"}
{"current_steps": 1180, "total_steps": 1314, "loss": 0.5677, "learning_rate": 5e-06, "epoch": 2.690598290598291, "percentage": 89.8, "elapsed_time": "17:21:10", "remaining_time": "1:58:14"}
{"current_steps": 1190, "total_steps": 1314, "loss": 0.5687, "learning_rate": 5e-06, "epoch": 2.7133903133903132, "percentage": 90.56, "elapsed_time": "17:29:52", "remaining_time": "1:49:23"}
{"current_steps": 1200, "total_steps": 1314, "loss": 0.5719, "learning_rate": 5e-06, "epoch": 2.736182336182336, "percentage": 91.32, "elapsed_time": "17:38:34", "remaining_time": "1:40:33"}
{"current_steps": 1210, "total_steps": 1314, "loss": 0.5753, "learning_rate": 5e-06, "epoch": 2.758974358974359, "percentage": 92.09, "elapsed_time": "17:47:15", "remaining_time": "1:31:43"}
{"current_steps": 1220, "total_steps": 1314, "loss": 0.5652, "learning_rate": 5e-06, "epoch": 2.781766381766382, "percentage": 92.85, "elapsed_time": "17:55:57", "remaining_time": "1:22:54"}
{"current_steps": 1230, "total_steps": 1314, "loss": 0.5637, "learning_rate": 5e-06, "epoch": 2.8045584045584047, "percentage": 93.61, "elapsed_time": "18:04:39", "remaining_time": "1:14:04"}
{"current_steps": 1240, "total_steps": 1314, "loss": 0.5784, "learning_rate": 5e-06, "epoch": 2.827350427350427, "percentage": 94.37, "elapsed_time": "18:13:22", "remaining_time": "1:05:14"}
{"current_steps": 1250, "total_steps": 1314, "loss": 0.5751, "learning_rate": 5e-06, "epoch": 2.8501424501424504, "percentage": 95.13, "elapsed_time": "18:22:04", "remaining_time": "0:56:25"}
{"current_steps": 1260, "total_steps": 1314, "loss": 0.5723, "learning_rate": 5e-06, "epoch": 2.872934472934473, "percentage": 95.89, "elapsed_time": "18:30:46", "remaining_time": "0:47:36"}
{"current_steps": 1270, "total_steps": 1314, "loss": 0.5732, "learning_rate": 5e-06, "epoch": 2.8957264957264957, "percentage": 96.65, "elapsed_time": "18:39:27", "remaining_time": "0:38:47"}
{"current_steps": 1280, "total_steps": 1314, "loss": 0.5634, "learning_rate": 5e-06, "epoch": 2.9185185185185185, "percentage": 97.41, "elapsed_time": "18:48:08", "remaining_time": "0:29:57"}
{"current_steps": 1290, "total_steps": 1314, "loss": 0.5636, "learning_rate": 5e-06, "epoch": 2.9413105413105414, "percentage": 98.17, "elapsed_time": "18:56:49", "remaining_time": "0:21:09"}
{"current_steps": 1300, "total_steps": 1314, "loss": 0.5648, "learning_rate": 5e-06, "epoch": 2.9641025641025642, "percentage": 98.93, "elapsed_time": "19:05:30", "remaining_time": "0:12:20"}
{"current_steps": 1310, "total_steps": 1314, "loss": 0.5746, "learning_rate": 5e-06, "epoch": 2.9868945868945866, "percentage": 99.7, "elapsed_time": "19:14:12", "remaining_time": "0:03:31"}
{"current_steps": 1314, "total_steps": 1314, "eval_loss": 0.6186709403991699, "epoch": 2.996011396011396, "percentage": 100.0, "elapsed_time": "19:26:09", "remaining_time": "0:00:00"}
{"current_steps": 1314, "total_steps": 1314, "epoch": 2.996011396011396, "percentage": 100.0, "elapsed_time": "19:27:32", "remaining_time": "0:00:00"}