chchen's picture
Training in progress, step 3275
2d3cf2c verified
{"current_steps": 10, "total_steps": 3275, "loss": 1.6278, "lr": 3.0487804878048782e-06, "epoch": 0.015265718919950386, "percentage": 0.31, "elapsed_time": "0:01:18", "remaining_time": "7:07:16"}
{"current_steps": 20, "total_steps": 3275, "loss": 1.5554, "lr": 6.0975609756097564e-06, "epoch": 0.030531437839900772, "percentage": 0.61, "elapsed_time": "0:02:35", "remaining_time": "7:02:28"}
{"current_steps": 30, "total_steps": 3275, "loss": 1.3738, "lr": 9.146341463414634e-06, "epoch": 0.04579715675985116, "percentage": 0.92, "elapsed_time": "0:03:52", "remaining_time": "6:59:48"}
{"current_steps": 40, "total_steps": 3275, "loss": 1.0174, "lr": 1.2195121951219513e-05, "epoch": 0.061062875679801544, "percentage": 1.22, "elapsed_time": "0:05:10", "remaining_time": "6:58:07"}
{"current_steps": 50, "total_steps": 3275, "loss": 0.5681, "lr": 1.524390243902439e-05, "epoch": 0.07632859459975193, "percentage": 1.53, "elapsed_time": "0:06:27", "remaining_time": "6:56:54"}
{"current_steps": 50, "total_steps": 3275, "eval_loss": 0.3813890814781189, "epoch": 0.07632859459975193, "percentage": 1.53, "elapsed_time": "0:09:06", "remaining_time": "9:47:41"}
{"current_steps": 60, "total_steps": 3275, "loss": 0.3362, "lr": 1.8292682926829268e-05, "epoch": 0.09159431351970232, "percentage": 1.83, "elapsed_time": "0:10:29", "remaining_time": "9:22:25"}
{"current_steps": 70, "total_steps": 3275, "loss": 0.246, "lr": 2.134146341463415e-05, "epoch": 0.10686003243965271, "percentage": 2.14, "elapsed_time": "0:11:47", "remaining_time": "8:59:42"}
{"current_steps": 80, "total_steps": 3275, "loss": 0.1841, "lr": 2.4390243902439026e-05, "epoch": 0.12212575135960309, "percentage": 2.44, "elapsed_time": "0:13:04", "remaining_time": "8:41:57"}
{"current_steps": 90, "total_steps": 3275, "loss": 0.115, "lr": 2.7439024390243906e-05, "epoch": 0.13739147027955348, "percentage": 2.75, "elapsed_time": "0:14:21", "remaining_time": "8:27:53"}
{"current_steps": 100, "total_steps": 3275, "loss": 0.122, "lr": 3.048780487804878e-05, "epoch": 0.15265718919950386, "percentage": 3.05, "elapsed_time": "0:15:38", "remaining_time": "8:16:44"}
{"current_steps": 100, "total_steps": 3275, "eval_loss": 0.07960689812898636, "epoch": 0.15265718919950386, "percentage": 3.05, "elapsed_time": "0:18:17", "remaining_time": "9:40:42"}
{"current_steps": 110, "total_steps": 3275, "loss": 0.0802, "lr": 3.353658536585366e-05, "epoch": 0.16792290811945426, "percentage": 3.36, "elapsed_time": "0:19:38", "remaining_time": "9:25:03"}
{"current_steps": 120, "total_steps": 3275, "loss": 0.0705, "lr": 3.6585365853658535e-05, "epoch": 0.18318862703940464, "percentage": 3.66, "elapsed_time": "0:20:55", "remaining_time": "9:09:59"}
{"current_steps": 130, "total_steps": 3275, "loss": 0.08, "lr": 3.9634146341463416e-05, "epoch": 0.19845434595935502, "percentage": 3.97, "elapsed_time": "0:22:12", "remaining_time": "8:57:13"}
{"current_steps": 140, "total_steps": 3275, "loss": 0.0655, "lr": 4.26829268292683e-05, "epoch": 0.21372006487930542, "percentage": 4.27, "elapsed_time": "0:23:29", "remaining_time": "8:46:04"}
{"current_steps": 150, "total_steps": 3275, "loss": 0.0541, "lr": 4.573170731707318e-05, "epoch": 0.2289857837992558, "percentage": 4.58, "elapsed_time": "0:24:47", "remaining_time": "8:36:21"}
{"current_steps": 150, "total_steps": 3275, "eval_loss": 0.059219248592853546, "epoch": 0.2289857837992558, "percentage": 4.58, "elapsed_time": "0:27:25", "remaining_time": "9:31:26"}
{"current_steps": 160, "total_steps": 3275, "loss": 0.066, "lr": 4.878048780487805e-05, "epoch": 0.24425150271920618, "percentage": 4.89, "elapsed_time": "0:28:47", "remaining_time": "9:20:36"}
{"current_steps": 170, "total_steps": 3275, "loss": 0.0751, "lr": 5.182926829268293e-05, "epoch": 0.2595172216391566, "percentage": 5.19, "elapsed_time": "0:30:05", "remaining_time": "9:09:30"}
{"current_steps": 180, "total_steps": 3275, "loss": 0.0772, "lr": 5.487804878048781e-05, "epoch": 0.27478294055910696, "percentage": 5.5, "elapsed_time": "0:31:22", "remaining_time": "8:59:26"}
{"current_steps": 190, "total_steps": 3275, "loss": 0.0687, "lr": 5.792682926829268e-05, "epoch": 0.29004865947905734, "percentage": 5.8, "elapsed_time": "0:32:39", "remaining_time": "8:50:14"}
{"current_steps": 200, "total_steps": 3275, "loss": 0.0502, "lr": 6.097560975609756e-05, "epoch": 0.3053143783990077, "percentage": 6.11, "elapsed_time": "0:33:56", "remaining_time": "8:41:51"}
{"current_steps": 200, "total_steps": 3275, "eval_loss": 0.05221959203481674, "epoch": 0.3053143783990077, "percentage": 6.11, "elapsed_time": "0:36:35", "remaining_time": "9:22:30"}
{"current_steps": 210, "total_steps": 3275, "loss": 0.0572, "lr": 6.402439024390244e-05, "epoch": 0.3205800973189581, "percentage": 6.41, "elapsed_time": "0:37:56", "remaining_time": "9:13:51"}
{"current_steps": 220, "total_steps": 3275, "loss": 0.0493, "lr": 6.707317073170732e-05, "epoch": 0.3358458162389085, "percentage": 6.72, "elapsed_time": "0:39:14", "remaining_time": "9:04:51"}
{"current_steps": 230, "total_steps": 3275, "loss": 0.0421, "lr": 7.012195121951219e-05, "epoch": 0.3511115351588589, "percentage": 7.02, "elapsed_time": "0:40:31", "remaining_time": "8:56:27"}
{"current_steps": 240, "total_steps": 3275, "loss": 0.0603, "lr": 7.317073170731707e-05, "epoch": 0.3663772540788093, "percentage": 7.33, "elapsed_time": "0:41:48", "remaining_time": "8:48:39"}
{"current_steps": 250, "total_steps": 3275, "loss": 0.0666, "lr": 7.621951219512195e-05, "epoch": 0.38164297299875966, "percentage": 7.63, "elapsed_time": "0:43:05", "remaining_time": "8:41:23"}
{"current_steps": 250, "total_steps": 3275, "eval_loss": 0.05392138659954071, "epoch": 0.38164297299875966, "percentage": 7.63, "elapsed_time": "0:45:44", "remaining_time": "9:13:22"}
{"current_steps": 260, "total_steps": 3275, "loss": 0.0653, "lr": 7.926829268292683e-05, "epoch": 0.39690869191871003, "percentage": 7.94, "elapsed_time": "0:47:05", "remaining_time": "9:06:07"}
{"current_steps": 270, "total_steps": 3275, "loss": 0.0563, "lr": 8.231707317073171e-05, "epoch": 0.4121744108386604, "percentage": 8.24, "elapsed_time": "0:48:22", "remaining_time": "8:58:25"}
{"current_steps": 280, "total_steps": 3275, "loss": 0.0576, "lr": 8.53658536585366e-05, "epoch": 0.42744012975861084, "percentage": 8.55, "elapsed_time": "0:49:39", "remaining_time": "8:51:14"}
{"current_steps": 290, "total_steps": 3275, "loss": 0.0602, "lr": 8.841463414634147e-05, "epoch": 0.4427058486785612, "percentage": 8.85, "elapsed_time": "0:50:56", "remaining_time": "8:44:23"}
{"current_steps": 300, "total_steps": 3275, "loss": 0.046, "lr": 9.146341463414635e-05, "epoch": 0.4579715675985116, "percentage": 9.16, "elapsed_time": "0:52:14", "remaining_time": "8:38:00"}
{"current_steps": 300, "total_steps": 3275, "eval_loss": 0.049273185431957245, "epoch": 0.4579715675985116, "percentage": 9.16, "elapsed_time": "0:54:52", "remaining_time": "9:04:11"}
{"current_steps": 310, "total_steps": 3275, "loss": 0.0649, "lr": 9.451219512195122e-05, "epoch": 0.473237286518462, "percentage": 9.47, "elapsed_time": "0:56:14", "remaining_time": "8:57:59"}
{"current_steps": 320, "total_steps": 3275, "loss": 0.0436, "lr": 9.75609756097561e-05, "epoch": 0.48850300543841235, "percentage": 9.77, "elapsed_time": "0:57:31", "remaining_time": "8:51:16"}
{"current_steps": 330, "total_steps": 3275, "loss": 0.0507, "lr": 9.999988635788465e-05, "epoch": 0.5037687243583627, "percentage": 10.08, "elapsed_time": "0:58:48", "remaining_time": "8:44:53"}
{"current_steps": 340, "total_steps": 3275, "loss": 0.0449, "lr": 9.999590893808788e-05, "epoch": 0.5190344432783132, "percentage": 10.38, "elapsed_time": "1:00:06", "remaining_time": "8:38:50"}
{"current_steps": 350, "total_steps": 3275, "loss": 0.0458, "lr": 9.998624992909386e-05, "epoch": 0.5343001621982635, "percentage": 10.69, "elapsed_time": "1:01:23", "remaining_time": "8:33:02"}
{"current_steps": 350, "total_steps": 3275, "eval_loss": 0.05271424353122711, "epoch": 0.5343001621982635, "percentage": 10.69, "elapsed_time": "1:04:01", "remaining_time": "8:55:06"}
{"current_steps": 360, "total_steps": 3275, "loss": 0.0488, "lr": 9.997091042856284e-05, "epoch": 0.5495658811182139, "percentage": 10.99, "elapsed_time": "1:05:22", "remaining_time": "8:49:23"}
{"current_steps": 370, "total_steps": 3275, "loss": 0.0646, "lr": 9.994989217969224e-05, "epoch": 0.5648316000381643, "percentage": 11.3, "elapsed_time": "1:06:39", "remaining_time": "8:43:21"}
{"current_steps": 380, "total_steps": 3275, "loss": 0.0588, "lr": 9.992319757101863e-05, "epoch": 0.5800973189581147, "percentage": 11.6, "elapsed_time": "1:07:56", "remaining_time": "8:37:34"}
{"current_steps": 390, "total_steps": 3275, "loss": 0.0586, "lr": 9.98908296361462e-05, "epoch": 0.5953630378780651, "percentage": 11.91, "elapsed_time": "1:09:12", "remaining_time": "8:32:01"}
{"current_steps": 400, "total_steps": 3275, "loss": 0.0448, "lr": 9.98527920534021e-05, "epoch": 0.6106287567980154, "percentage": 12.21, "elapsed_time": "1:10:29", "remaining_time": "8:26:41"}
{"current_steps": 400, "total_steps": 3275, "eval_loss": 0.048783060163259506, "epoch": 0.6106287567980154, "percentage": 12.21, "elapsed_time": "1:13:07", "remaining_time": "8:45:35"}
{"current_steps": 410, "total_steps": 3275, "loss": 0.054, "lr": 9.980908914541844e-05, "epoch": 0.6258944757179659, "percentage": 12.52, "elapsed_time": "1:14:29", "remaining_time": "8:40:28"}
{"current_steps": 420, "total_steps": 3275, "loss": 0.0506, "lr": 9.975972587864095e-05, "epoch": 0.6411601946379162, "percentage": 12.82, "elapsed_time": "1:15:46", "remaining_time": "8:35:02"}
{"current_steps": 430, "total_steps": 3275, "loss": 0.0369, "lr": 9.970470786276467e-05, "epoch": 0.6564259135578666, "percentage": 13.13, "elapsed_time": "1:17:03", "remaining_time": "8:29:47"}
{"current_steps": 440, "total_steps": 3275, "loss": 0.0527, "lr": 9.964404135009648e-05, "epoch": 0.671691632477817, "percentage": 13.44, "elapsed_time": "1:18:20", "remaining_time": "8:24:43"}
{"current_steps": 450, "total_steps": 3275, "loss": 0.0567, "lr": 9.957773323484454e-05, "epoch": 0.6869573513977674, "percentage": 13.74, "elapsed_time": "1:19:36", "remaining_time": "8:19:47"}
{"current_steps": 450, "total_steps": 3275, "eval_loss": 0.04618264362215996, "epoch": 0.6869573513977674, "percentage": 13.74, "elapsed_time": "1:22:14", "remaining_time": "8:36:17"}
{"current_steps": 460, "total_steps": 3275, "loss": 0.04, "lr": 9.950579105233483e-05, "epoch": 0.7022230703177178, "percentage": 14.05, "elapsed_time": "1:23:36", "remaining_time": "8:31:36"}
{"current_steps": 470, "total_steps": 3275, "loss": 0.0383, "lr": 9.94282229781548e-05, "epoch": 0.7174887892376681, "percentage": 14.35, "elapsed_time": "1:24:52", "remaining_time": "8:26:34"}
{"current_steps": 480, "total_steps": 3275, "loss": 0.0495, "lr": 9.934503782722438e-05, "epoch": 0.7327545081576186, "percentage": 14.66, "elapsed_time": "1:26:09", "remaining_time": "8:21:39"}
{"current_steps": 490, "total_steps": 3275, "loss": 0.0379, "lr": 9.925624505279411e-05, "epoch": 0.748020227077569, "percentage": 14.96, "elapsed_time": "1:27:25", "remaining_time": "8:16:56"}
{"current_steps": 500, "total_steps": 3275, "loss": 0.0358, "lr": 9.916185474537098e-05, "epoch": 0.7632859459975193, "percentage": 15.27, "elapsed_time": "1:28:42", "remaining_time": "8:12:22"}
{"current_steps": 500, "total_steps": 3275, "eval_loss": 0.04097205400466919, "epoch": 0.7632859459975193, "percentage": 15.27, "elapsed_time": "1:31:20", "remaining_time": "8:26:56"}
{"current_steps": 510, "total_steps": 3275, "loss": 0.0508, "lr": 9.906187763157168e-05, "epoch": 0.7785516649174697, "percentage": 15.57, "elapsed_time": "1:32:41", "remaining_time": "8:22:33"}
{"current_steps": 520, "total_steps": 3275, "loss": 0.05, "lr": 9.895632507290362e-05, "epoch": 0.7938173838374201, "percentage": 15.88, "elapsed_time": "1:33:58", "remaining_time": "8:17:54"}
{"current_steps": 530, "total_steps": 3275, "loss": 0.0376, "lr": 9.884520906447379e-05, "epoch": 0.8090831027573705, "percentage": 16.18, "elapsed_time": "1:35:15", "remaining_time": "8:13:20"}
{"current_steps": 540, "total_steps": 3275, "loss": 0.0482, "lr": 9.872854223362562e-05, "epoch": 0.8243488216773208, "percentage": 16.49, "elapsed_time": "1:36:31", "remaining_time": "8:08:54"}
{"current_steps": 550, "total_steps": 3275, "loss": 0.0445, "lr": 9.860633783850406e-05, "epoch": 0.8396145405972713, "percentage": 16.79, "elapsed_time": "1:37:48", "remaining_time": "8:04:36"}
{"current_steps": 550, "total_steps": 3275, "eval_loss": 0.04072929546236992, "epoch": 0.8396145405972713, "percentage": 16.79, "elapsed_time": "1:40:26", "remaining_time": "8:17:36"}
{"current_steps": 560, "total_steps": 3275, "loss": 0.0501, "lr": 9.847860976654879e-05, "epoch": 0.8548802595172217, "percentage": 17.1, "elapsed_time": "1:41:48", "remaining_time": "8:13:34"}
{"current_steps": 570, "total_steps": 3275, "loss": 0.0401, "lr": 9.834537253291616e-05, "epoch": 0.870145978437172, "percentage": 17.4, "elapsed_time": "1:43:05", "remaining_time": "8:09:12"}
{"current_steps": 580, "total_steps": 3275, "loss": 0.0582, "lr": 9.820664127882957e-05, "epoch": 0.8854116973571224, "percentage": 17.71, "elapsed_time": "1:44:21", "remaining_time": "8:04:55"}
{"current_steps": 590, "total_steps": 3275, "loss": 0.0494, "lr": 9.806243176985888e-05, "epoch": 0.9006774162770728, "percentage": 18.02, "elapsed_time": "1:45:38", "remaining_time": "8:00:44"}
{"current_steps": 600, "total_steps": 3275, "loss": 0.0462, "lr": 9.791276039412875e-05, "epoch": 0.9159431351970232, "percentage": 18.32, "elapsed_time": "1:46:55", "remaining_time": "7:56:41"}
{"current_steps": 600, "total_steps": 3275, "eval_loss": 0.040742386132478714, "epoch": 0.9159431351970232, "percentage": 18.32, "elapsed_time": "1:49:32", "remaining_time": "8:08:22"}
{"current_steps": 610, "total_steps": 3275, "loss": 0.0595, "lr": 9.775764416045628e-05, "epoch": 0.9312088541169735, "percentage": 18.63, "elapsed_time": "1:50:53", "remaining_time": "8:04:29"}
{"current_steps": 620, "total_steps": 3275, "loss": 0.0463, "lr": 9.759710069641814e-05, "epoch": 0.946474573036924, "percentage": 18.93, "elapsed_time": "1:52:10", "remaining_time": "8:00:22"}
{"current_steps": 630, "total_steps": 3275, "loss": 0.033, "lr": 9.743114824634734e-05, "epoch": 0.9617402919568744, "percentage": 19.24, "elapsed_time": "1:53:27", "remaining_time": "7:56:21"}
{"current_steps": 640, "total_steps": 3275, "loss": 0.0497, "lr": 9.725980566925989e-05, "epoch": 0.9770060108768247, "percentage": 19.54, "elapsed_time": "1:54:44", "remaining_time": "7:52:23"}
{"current_steps": 650, "total_steps": 3275, "loss": 0.0363, "lr": 9.708309243671165e-05, "epoch": 0.9922717297967751, "percentage": 19.85, "elapsed_time": "1:56:00", "remaining_time": "7:48:31"}
{"current_steps": 650, "total_steps": 3275, "eval_loss": 0.040968675166368484, "epoch": 0.9922717297967751, "percentage": 19.85, "elapsed_time": "1:58:38", "remaining_time": "7:59:07"}
{"current_steps": 660, "total_steps": 3275, "loss": 0.0343, "lr": 9.690102863058563e-05, "epoch": 1.0075374487167255, "percentage": 20.15, "elapsed_time": "1:59:59", "remaining_time": "7:55:26"}
{"current_steps": 670, "total_steps": 3275, "loss": 0.0333, "lr": 9.67136349408098e-05, "epoch": 1.0228031676366758, "percentage": 20.46, "elapsed_time": "2:01:16", "remaining_time": "7:51:31"}
{"current_steps": 680, "total_steps": 3275, "loss": 0.0401, "lr": 9.652093266300583e-05, "epoch": 1.0380688865566263, "percentage": 20.76, "elapsed_time": "2:02:33", "remaining_time": "7:47:41"}
{"current_steps": 690, "total_steps": 3275, "loss": 0.0271, "lr": 9.632294369606916e-05, "epoch": 1.0533346054765766, "percentage": 21.07, "elapsed_time": "2:03:49", "remaining_time": "7:43:55"}
{"current_steps": 700, "total_steps": 3275, "loss": 0.0343, "lr": 9.61196905396802e-05, "epoch": 1.068600324396527, "percentage": 21.37, "elapsed_time": "2:05:06", "remaining_time": "7:40:14"}
{"current_steps": 700, "total_steps": 3275, "eval_loss": 0.03695390745997429, "epoch": 1.068600324396527, "percentage": 21.37, "elapsed_time": "2:07:44", "remaining_time": "7:49:53"}
{"current_steps": 710, "total_steps": 3275, "loss": 0.0442, "lr": 9.591119629174764e-05, "epoch": 1.0838660433164775, "percentage": 21.68, "elapsed_time": "2:09:06", "remaining_time": "7:46:24"}
{"current_steps": 720, "total_steps": 3275, "loss": 0.0288, "lr": 9.569748464578343e-05, "epoch": 1.0991317622364278, "percentage": 21.98, "elapsed_time": "2:10:22", "remaining_time": "7:42:39"}
{"current_steps": 730, "total_steps": 3275, "loss": 0.0273, "lr": 9.54785798882103e-05, "epoch": 1.1143974811563782, "percentage": 22.29, "elapsed_time": "2:11:39", "remaining_time": "7:38:59"}
{"current_steps": 740, "total_steps": 3275, "loss": 0.0378, "lr": 9.525450689560181e-05, "epoch": 1.1296632000763287, "percentage": 22.6, "elapsed_time": "2:12:56", "remaining_time": "7:35:23"}
{"current_steps": 750, "total_steps": 3275, "loss": 0.0413, "lr": 9.502529113185532e-05, "epoch": 1.144928918996279, "percentage": 22.9, "elapsed_time": "2:14:12", "remaining_time": "7:31:51"}
{"current_steps": 750, "total_steps": 3275, "eval_loss": 0.03775772452354431, "epoch": 1.144928918996279, "percentage": 22.9, "elapsed_time": "2:16:50", "remaining_time": "7:40:41"}
{"current_steps": 760, "total_steps": 3275, "loss": 0.0308, "lr": 9.479095864529828e-05, "epoch": 1.1601946379162293, "percentage": 23.21, "elapsed_time": "2:18:12", "remaining_time": "7:37:21"}
{"current_steps": 770, "total_steps": 3275, "loss": 0.0316, "lr": 9.455153606572806e-05, "epoch": 1.1754603568361797, "percentage": 23.51, "elapsed_time": "2:19:29", "remaining_time": "7:33:47"}
{"current_steps": 780, "total_steps": 3275, "loss": 0.0348, "lr": 9.430705060138569e-05, "epoch": 1.1907260757561302, "percentage": 23.82, "elapsed_time": "2:20:45", "remaining_time": "7:30:15"}
{"current_steps": 790, "total_steps": 3275, "loss": 0.0277, "lr": 9.405753003586395e-05, "epoch": 1.2059917946760805, "percentage": 24.12, "elapsed_time": "2:22:02", "remaining_time": "7:26:49"}
{"current_steps": 800, "total_steps": 3275, "loss": 0.0322, "lr": 9.38030027249499e-05, "epoch": 1.2212575135960309, "percentage": 24.43, "elapsed_time": "2:23:20", "remaining_time": "7:23:26"}
{"current_steps": 800, "total_steps": 3275, "eval_loss": 0.03979082778096199, "epoch": 1.2212575135960309, "percentage": 24.43, "elapsed_time": "2:25:57", "remaining_time": "7:31:33"}
{"current_steps": 810, "total_steps": 3275, "loss": 0.0372, "lr": 9.354349759340263e-05, "epoch": 1.2365232325159814, "percentage": 24.73, "elapsed_time": "2:27:19", "remaining_time": "7:28:21"}
{"current_steps": 820, "total_steps": 3275, "loss": 0.0338, "lr": 9.327904413166615e-05, "epoch": 1.2517889514359317, "percentage": 25.04, "elapsed_time": "2:28:36", "remaining_time": "7:24:55"}
{"current_steps": 830, "total_steps": 3275, "loss": 0.0405, "lr": 9.300967239251798e-05, "epoch": 1.267054670355882, "percentage": 25.34, "elapsed_time": "2:29:53", "remaining_time": "7:21:32"}
{"current_steps": 840, "total_steps": 3275, "loss": 0.0348, "lr": 9.27354129876541e-05, "epoch": 1.2823203892758324, "percentage": 25.65, "elapsed_time": "2:31:10", "remaining_time": "7:18:13"}
{"current_steps": 850, "total_steps": 3275, "loss": 0.0342, "lr": 9.245629708421008e-05, "epoch": 1.297586108195783, "percentage": 25.95, "elapsed_time": "2:32:27", "remaining_time": "7:14:56"}
{"current_steps": 850, "total_steps": 3275, "eval_loss": 0.03846587613224983, "epoch": 1.297586108195783, "percentage": 25.95, "elapsed_time": "2:35:04", "remaining_time": "7:22:26"}
{"current_steps": 860, "total_steps": 3275, "loss": 0.0379, "lr": 9.217235640121926e-05, "epoch": 1.3128518271157332, "percentage": 26.26, "elapsed_time": "2:36:27", "remaining_time": "7:19:20"}
{"current_steps": 870, "total_steps": 3275, "loss": 0.0425, "lr": 9.188362320600812e-05, "epoch": 1.3281175460356835, "percentage": 26.56, "elapsed_time": "2:37:44", "remaining_time": "7:16:02"}
{"current_steps": 880, "total_steps": 3275, "loss": 0.0324, "lr": 9.159013031052943e-05, "epoch": 1.343383264955634, "percentage": 26.87, "elapsed_time": "2:39:01", "remaining_time": "7:12:47"}
{"current_steps": 890, "total_steps": 3275, "loss": 0.0342, "lr": 9.129191106763346e-05, "epoch": 1.3586489838755844, "percentage": 27.18, "elapsed_time": "2:40:18", "remaining_time": "7:09:34"}
{"current_steps": 900, "total_steps": 3275, "loss": 0.0337, "lr": 9.098899936727771e-05, "epoch": 1.3739147027955347, "percentage": 27.48, "elapsed_time": "2:41:34", "remaining_time": "7:06:23"}
{"current_steps": 900, "total_steps": 3275, "eval_loss": 0.043599650263786316, "epoch": 1.3739147027955347, "percentage": 27.48, "elapsed_time": "2:44:12", "remaining_time": "7:13:19"}
{"current_steps": 910, "total_steps": 3275, "loss": 0.029, "lr": 9.068142963267558e-05, "epoch": 1.3891804217154853, "percentage": 27.79, "elapsed_time": "2:45:34", "remaining_time": "7:10:17"}
{"current_steps": 920, "total_steps": 3275, "loss": 0.0337, "lr": 9.036923681638463e-05, "epoch": 1.4044461406354356, "percentage": 28.09, "elapsed_time": "2:46:50", "remaining_time": "7:07:05"}
{"current_steps": 930, "total_steps": 3275, "loss": 0.0269, "lr": 9.00524563963343e-05, "epoch": 1.419711859555386, "percentage": 28.4, "elapsed_time": "2:48:07", "remaining_time": "7:03:56"}
{"current_steps": 940, "total_steps": 3275, "loss": 0.0372, "lr": 8.973112437179436e-05, "epoch": 1.4349775784753362, "percentage": 28.7, "elapsed_time": "2:49:25", "remaining_time": "7:00:50"}
{"current_steps": 950, "total_steps": 3275, "loss": 0.0295, "lr": 8.940527725928383e-05, "epoch": 1.4502432973952868, "percentage": 29.01, "elapsed_time": "2:50:42", "remaining_time": "6:57:46"}
{"current_steps": 950, "total_steps": 3275, "eval_loss": 0.037314947694540024, "epoch": 1.4502432973952868, "percentage": 29.01, "elapsed_time": "2:53:19", "remaining_time": "7:04:12"}
{"current_steps": 960, "total_steps": 3275, "loss": 0.0324, "lr": 8.90749520884212e-05, "epoch": 1.4655090163152371, "percentage": 29.31, "elapsed_time": "2:54:41", "remaining_time": "7:01:15"}
{"current_steps": 970, "total_steps": 3275, "loss": 0.0357, "lr": 8.874018639771637e-05, "epoch": 1.4807747352351874, "percentage": 29.62, "elapsed_time": "2:55:58", "remaining_time": "6:58:09"}
{"current_steps": 980, "total_steps": 3275, "loss": 0.0276, "lr": 8.840101823030471e-05, "epoch": 1.4960404541551378, "percentage": 29.92, "elapsed_time": "2:57:15", "remaining_time": "6:55:06"}
{"current_steps": 990, "total_steps": 3275, "loss": 0.0381, "lr": 8.805748612962382e-05, "epoch": 1.5113061730750883, "percentage": 30.23, "elapsed_time": "2:58:32", "remaining_time": "6:52:04"}
{"current_steps": 1000, "total_steps": 3275, "loss": 0.0267, "lr": 8.77096291350334e-05, "epoch": 1.5265718919950386, "percentage": 30.53, "elapsed_time": "2:59:48", "remaining_time": "6:49:04"}
{"current_steps": 1000, "total_steps": 3275, "eval_loss": 0.03864908590912819, "epoch": 1.5265718919950386, "percentage": 30.53, "elapsed_time": "3:02:27", "remaining_time": "6:55:04"}
{"current_steps": 1010, "total_steps": 3275, "loss": 0.0346, "lr": 8.735748677737874e-05, "epoch": 1.541837610914989, "percentage": 30.84, "elapsed_time": "3:03:51", "remaining_time": "6:52:19"}
{"current_steps": 1020, "total_steps": 3275, "loss": 0.0296, "lr": 8.700109907449845e-05, "epoch": 1.5571033298349395, "percentage": 31.15, "elapsed_time": "3:05:13", "remaining_time": "6:49:29"}
{"current_steps": 1030, "total_steps": 3275, "loss": 0.0311, "lr": 8.66405065266768e-05, "epoch": 1.5723690487548898, "percentage": 31.45, "elapsed_time": "3:06:35", "remaining_time": "6:46:42"}
{"current_steps": 1040, "total_steps": 3275, "loss": 0.0281, "lr": 8.627575011204115e-05, "epoch": 1.5876347676748401, "percentage": 31.76, "elapsed_time": "3:07:58", "remaining_time": "6:43:58"}
{"current_steps": 1050, "total_steps": 3275, "loss": 0.0287, "lr": 8.590687128190516e-05, "epoch": 1.6029004865947907, "percentage": 32.06, "elapsed_time": "3:09:22", "remaining_time": "6:41:17"}
{"current_steps": 1050, "total_steps": 3275, "eval_loss": 0.03797018155455589, "epoch": 1.6029004865947907, "percentage": 32.06, "elapsed_time": "3:12:16", "remaining_time": "6:47:25"}
{"current_steps": 1060, "total_steps": 3275, "loss": 0.0273, "lr": 8.553391195605833e-05, "epoch": 1.618166205514741, "percentage": 32.37, "elapsed_time": "3:13:44", "remaining_time": "6:44:50"}
{"current_steps": 1070, "total_steps": 3275, "loss": 0.0469, "lr": 8.515691451800205e-05, "epoch": 1.6334319244346913, "percentage": 32.67, "elapsed_time": "3:15:08", "remaining_time": "6:42:07"}
{"current_steps": 1080, "total_steps": 3275, "loss": 0.0291, "lr": 8.477592181013316e-05, "epoch": 1.6486976433546419, "percentage": 32.98, "elapsed_time": "3:16:32", "remaining_time": "6:39:26"}
{"current_steps": 1090, "total_steps": 3275, "loss": 0.0297, "lr": 8.439097712887531e-05, "epoch": 1.663963362274592, "percentage": 33.28, "elapsed_time": "3:17:56", "remaining_time": "6:36:46"}
{"current_steps": 1100, "total_steps": 3275, "loss": 0.0504, "lr": 8.400212421975865e-05, "epoch": 1.6792290811945425, "percentage": 33.59, "elapsed_time": "3:19:20", "remaining_time": "6:34:08"}
{"current_steps": 1100, "total_steps": 3275, "eval_loss": 0.03881072625517845, "epoch": 1.6792290811945425, "percentage": 33.59, "elapsed_time": "3:22:15", "remaining_time": "6:39:55"}
{"current_steps": 1110, "total_steps": 3275, "loss": 0.0418, "lr": 8.360940727244859e-05, "epoch": 1.694494800114493, "percentage": 33.89, "elapsed_time": "3:23:42", "remaining_time": "6:37:20"}
{"current_steps": 1120, "total_steps": 3275, "loss": 0.0285, "lr": 8.321287091572403e-05, "epoch": 1.7097605190344431, "percentage": 34.2, "elapsed_time": "3:25:07", "remaining_time": "6:34:40"}
{"current_steps": 1130, "total_steps": 3275, "loss": 0.0299, "lr": 8.281256021240566e-05, "epoch": 1.7250262379543937, "percentage": 34.5, "elapsed_time": "3:26:31", "remaining_time": "6:32:02"}
{"current_steps": 1140, "total_steps": 3275, "loss": 0.0384, "lr": 8.240852065423506e-05, "epoch": 1.740291956874344, "percentage": 34.81, "elapsed_time": "3:27:56", "remaining_time": "6:29:25"}
{"current_steps": 1150, "total_steps": 3275, "loss": 0.0317, "lr": 8.20007981567048e-05, "epoch": 1.7555576757942943, "percentage": 35.11, "elapsed_time": "3:29:20", "remaining_time": "6:26:50"}
{"current_steps": 1150, "total_steps": 3275, "eval_loss": 0.03912714123725891, "epoch": 1.7555576757942943, "percentage": 35.11, "elapsed_time": "3:32:17", "remaining_time": "6:32:16"}
{"current_steps": 1160, "total_steps": 3275, "loss": 0.0384, "lr": 8.158943905384082e-05, "epoch": 1.7708233947142449, "percentage": 35.42, "elapsed_time": "3:33:45", "remaining_time": "6:29:44"}
{"current_steps": 1170, "total_steps": 3275, "loss": 0.0305, "lr": 8.117449009293668e-05, "epoch": 1.7860891136341952, "percentage": 35.73, "elapsed_time": "3:35:10", "remaining_time": "6:27:07"}
{"current_steps": 1180, "total_steps": 3275, "loss": 0.0306, "lr": 8.075599842924139e-05, "epoch": 1.8013548325541455, "percentage": 36.03, "elapsed_time": "3:36:34", "remaining_time": "6:24:30"}
{"current_steps": 1190, "total_steps": 3275, "loss": 0.0361, "lr": 8.033401162060049e-05, "epoch": 1.816620551474096, "percentage": 36.34, "elapsed_time": "3:37:59", "remaining_time": "6:21:56"}
{"current_steps": 1200, "total_steps": 3275, "loss": 0.0448, "lr": 7.990857762205157e-05, "epoch": 1.8318862703940464, "percentage": 36.64, "elapsed_time": "3:39:24", "remaining_time": "6:19:23"}
{"current_steps": 1200, "total_steps": 3275, "eval_loss": 0.03655826300382614, "epoch": 1.8318862703940464, "percentage": 36.64, "elapsed_time": "3:42:20", "remaining_time": "6:24:28"}
{"current_steps": 1210, "total_steps": 3275, "loss": 0.0378, "lr": 7.947974478037468e-05, "epoch": 1.8471519893139967, "percentage": 36.95, "elapsed_time": "3:43:50", "remaining_time": "6:22:00"}
{"current_steps": 1220, "total_steps": 3275, "loss": 0.0309, "lr": 7.904756182859797e-05, "epoch": 1.8624177082339473, "percentage": 37.25, "elapsed_time": "3:45:15", "remaining_time": "6:19:25"}
{"current_steps": 1230, "total_steps": 3275, "loss": 0.0318, "lr": 7.861207788045984e-05, "epoch": 1.8776834271538976, "percentage": 37.56, "elapsed_time": "3:46:39", "remaining_time": "6:16:51"}
{"current_steps": 1240, "total_steps": 3275, "loss": 0.0322, "lr": 7.817334242482738e-05, "epoch": 1.892949146073848, "percentage": 37.86, "elapsed_time": "3:48:04", "remaining_time": "6:14:18"}
{"current_steps": 1250, "total_steps": 3275, "loss": 0.0278, "lr": 7.773140532007262e-05, "epoch": 1.9082148649937984, "percentage": 38.17, "elapsed_time": "3:49:30", "remaining_time": "6:11:47"}
{"current_steps": 1250, "total_steps": 3275, "eval_loss": 0.036195144057273865, "epoch": 1.9082148649937984, "percentage": 38.17, "elapsed_time": "3:52:27", "remaining_time": "6:16:34"}
{"current_steps": 1260, "total_steps": 3275, "loss": 0.0294, "lr": 7.728631678840638e-05, "epoch": 1.9234805839137485, "percentage": 38.47, "elapsed_time": "3:53:56", "remaining_time": "6:14:06"}
{"current_steps": 1270, "total_steps": 3275, "loss": 0.0319, "lr": 7.683812741017112e-05, "epoch": 1.938746302833699, "percentage": 38.78, "elapsed_time": "3:55:21", "remaining_time": "6:11:33"}
{"current_steps": 1280, "total_steps": 3275, "loss": 0.0388, "lr": 7.638688811809274e-05, "epoch": 1.9540120217536494, "percentage": 39.08, "elapsed_time": "3:56:46", "remaining_time": "6:09:02"}
{"current_steps": 1290, "total_steps": 3275, "loss": 0.0267, "lr": 7.593265019149275e-05, "epoch": 1.9692777406735997, "percentage": 39.39, "elapsed_time": "3:58:09", "remaining_time": "6:06:28"}
{"current_steps": 1300, "total_steps": 3275, "loss": 0.0347, "lr": 7.547546525046073e-05, "epoch": 1.9845434595935503, "percentage": 39.69, "elapsed_time": "3:59:32", "remaining_time": "6:03:54"}
{"current_steps": 1300, "total_steps": 3275, "eval_loss": 0.03436192870140076, "epoch": 1.9845434595935503, "percentage": 39.69, "elapsed_time": "4:02:28", "remaining_time": "6:08:21"}
{"current_steps": 1310, "total_steps": 3275, "loss": 0.0353, "lr": 7.501538524998812e-05, "epoch": 1.9998091785135006, "percentage": 40.0, "elapsed_time": "4:03:57", "remaining_time": "6:05:56"}
{"current_steps": 1320, "total_steps": 3275, "loss": 0.0161, "lr": 7.455246247406406e-05, "epoch": 2.015074897433451, "percentage": 40.31, "elapsed_time": "4:05:22", "remaining_time": "6:03:24"}
{"current_steps": 1330, "total_steps": 3275, "loss": 0.0256, "lr": 7.408674952973382e-05, "epoch": 2.0303406163534015, "percentage": 40.61, "elapsed_time": "4:06:47", "remaining_time": "6:00:53"}
{"current_steps": 1340, "total_steps": 3275, "loss": 0.0235, "lr": 7.361829934112036e-05, "epoch": 2.0456063352733516, "percentage": 40.92, "elapsed_time": "4:08:12", "remaining_time": "5:58:24"}
{"current_steps": 1350, "total_steps": 3275, "loss": 0.0201, "lr": 7.314716514341006e-05, "epoch": 2.060872054193302, "percentage": 41.22, "elapsed_time": "4:09:37", "remaining_time": "5:55:57"}
{"current_steps": 1350, "total_steps": 3275, "eval_loss": 0.035528797656297684, "epoch": 2.060872054193302, "percentage": 41.22, "elapsed_time": "4:12:35", "remaining_time": "6:00:10"}
{"current_steps": 1360, "total_steps": 3275, "loss": 0.0164, "lr": 7.267340047680305e-05, "epoch": 2.0761377731132526, "percentage": 41.53, "elapsed_time": "4:14:04", "remaining_time": "5:57:45"}
{"current_steps": 1370, "total_steps": 3275, "loss": 0.0182, "lr": 7.21970591804287e-05, "epoch": 2.0914034920332027, "percentage": 41.83, "elapsed_time": "4:15:29", "remaining_time": "5:55:15"}
{"current_steps": 1380, "total_steps": 3275, "loss": 0.0254, "lr": 7.171819538622747e-05, "epoch": 2.1066692109531533, "percentage": 42.14, "elapsed_time": "4:16:54", "remaining_time": "5:52:46"}
{"current_steps": 1390, "total_steps": 3275, "loss": 0.0231, "lr": 7.123686351279914e-05, "epoch": 2.121934929873104, "percentage": 42.44, "elapsed_time": "4:18:19", "remaining_time": "5:50:18"}
{"current_steps": 1400, "total_steps": 3275, "loss": 0.0238, "lr": 7.07531182592187e-05, "epoch": 2.137200648793054, "percentage": 42.75, "elapsed_time": "4:19:44", "remaining_time": "5:47:52"}
{"current_steps": 1400, "total_steps": 3275, "eval_loss": 0.03574272617697716, "epoch": 2.137200648793054, "percentage": 42.75, "elapsed_time": "4:22:42", "remaining_time": "5:51:50"}
{"current_steps": 1410, "total_steps": 3275, "loss": 0.0315, "lr": 7.026701459882026e-05, "epoch": 2.1524663677130045, "percentage": 43.05, "elapsed_time": "4:24:12", "remaining_time": "5:49:27"}
{"current_steps": 1420, "total_steps": 3275, "loss": 0.0227, "lr": 6.977860777294988e-05, "epoch": 2.167732086632955, "percentage": 43.36, "elapsed_time": "4:25:37", "remaining_time": "5:46:59"}
{"current_steps": 1430, "total_steps": 3275, "loss": 0.0175, "lr": 6.92879532846878e-05, "epoch": 2.182997805552905, "percentage": 43.66, "elapsed_time": "4:27:02", "remaining_time": "5:44:32"}
{"current_steps": 1440, "total_steps": 3275, "loss": 0.0187, "lr": 6.879510689254104e-05, "epoch": 2.1982635244728557, "percentage": 43.97, "elapsed_time": "4:28:27", "remaining_time": "5:42:05"}
{"current_steps": 1450, "total_steps": 3275, "loss": 0.0299, "lr": 6.830012460410697e-05, "epoch": 2.213529243392806, "percentage": 44.27, "elapsed_time": "4:29:52", "remaining_time": "5:39:40"}
{"current_steps": 1450, "total_steps": 3275, "eval_loss": 0.037096813321113586, "epoch": 2.213529243392806, "percentage": 44.27, "elapsed_time": "4:32:49", "remaining_time": "5:43:22"}
{"current_steps": 1460, "total_steps": 3275, "loss": 0.017, "lr": 6.780306266970851e-05, "epoch": 2.2287949623127563, "percentage": 44.58, "elapsed_time": "4:34:17", "remaining_time": "5:40:58"}
{"current_steps": 1470, "total_steps": 3275, "loss": 0.0209, "lr": 6.73039775760018e-05, "epoch": 2.244060681232707, "percentage": 44.89, "elapsed_time": "4:35:41", "remaining_time": "5:38:31"}
{"current_steps": 1480, "total_steps": 3275, "loss": 0.0201, "lr": 6.680292603955702e-05, "epoch": 2.2593264001526574, "percentage": 45.19, "elapsed_time": "4:37:06", "remaining_time": "5:36:05"}
{"current_steps": 1490, "total_steps": 3275, "loss": 0.022, "lr": 6.629996500041299e-05, "epoch": 2.2745921190726075, "percentage": 45.5, "elapsed_time": "4:38:30", "remaining_time": "5:33:38"}
{"current_steps": 1500, "total_steps": 3275, "loss": 0.0155, "lr": 6.579515161560649e-05, "epoch": 2.289857837992558, "percentage": 45.8, "elapsed_time": "4:39:54", "remaining_time": "5:31:13"}
{"current_steps": 1500, "total_steps": 3275, "eval_loss": 0.038406338542699814, "epoch": 2.289857837992558, "percentage": 45.8, "elapsed_time": "4:42:50", "remaining_time": "5:34:41"}
{"current_steps": 1510, "total_steps": 3275, "loss": 0.0268, "lr": 6.528854325267692e-05, "epoch": 2.305123556912508, "percentage": 46.11, "elapsed_time": "4:44:18", "remaining_time": "5:32:19"}
{"current_steps": 1520, "total_steps": 3275, "loss": 0.0213, "lr": 6.478019748314686e-05, "epoch": 2.3203892758324587, "percentage": 46.41, "elapsed_time": "4:45:43", "remaining_time": "5:29:53"}
{"current_steps": 1530, "total_steps": 3275, "loss": 0.0238, "lr": 6.42701720759797e-05, "epoch": 2.3356549947524092, "percentage": 46.72, "elapsed_time": "4:47:06", "remaining_time": "5:27:27"}
{"current_steps": 1540, "total_steps": 3275, "loss": 0.019, "lr": 6.375852499101467e-05, "epoch": 2.3509207136723593, "percentage": 47.02, "elapsed_time": "4:48:30", "remaining_time": "5:25:02"}
{"current_steps": 1550, "total_steps": 3275, "loss": 0.0157, "lr": 6.324531437238019e-05, "epoch": 2.36618643259231, "percentage": 47.33, "elapsed_time": "4:49:54", "remaining_time": "5:22:38"}
{"current_steps": 1550, "total_steps": 3275, "eval_loss": 0.03914285823702812, "epoch": 2.36618643259231, "percentage": 47.33, "elapsed_time": "4:52:49", "remaining_time": "5:25:53"}
{"current_steps": 1560, "total_steps": 3275, "loss": 0.025, "lr": 6.273059854188636e-05, "epoch": 2.3814521515122604, "percentage": 47.63, "elapsed_time": "4:54:16", "remaining_time": "5:23:30"}
{"current_steps": 1570, "total_steps": 3275, "loss": 0.0242, "lr": 6.221443599239721e-05, "epoch": 2.3967178704322105, "percentage": 47.94, "elapsed_time": "4:55:38", "remaining_time": "5:21:03"}
{"current_steps": 1580, "total_steps": 3275, "loss": 0.028, "lr": 6.169688538118342e-05, "epoch": 2.411983589352161, "percentage": 48.24, "elapsed_time": "4:57:00", "remaining_time": "5:18:37"}
{"current_steps": 1590, "total_steps": 3275, "loss": 0.0232, "lr": 6.117800552325655e-05, "epoch": 2.4272493082721116, "percentage": 48.55, "elapsed_time": "4:58:24", "remaining_time": "5:16:13"}
{"current_steps": 1600, "total_steps": 3275, "loss": 0.0222, "lr": 6.06578553846852e-05, "epoch": 2.4425150271920617, "percentage": 48.85, "elapsed_time": "4:59:48", "remaining_time": "5:13:51"}
{"current_steps": 1600, "total_steps": 3275, "eval_loss": 0.03701253607869148, "epoch": 2.4425150271920617, "percentage": 48.85, "elapsed_time": "5:02:43", "remaining_time": "5:16:54"}
{"current_steps": 1610, "total_steps": 3275, "loss": 0.0228, "lr": 6.013649407589401e-05, "epoch": 2.4577807461120122, "percentage": 49.16, "elapsed_time": "5:04:07", "remaining_time": "5:14:30"}
{"current_steps": 1620, "total_steps": 3275, "loss": 0.0246, "lr": 5.961398084494634e-05, "epoch": 2.473046465031963, "percentage": 49.47, "elapsed_time": "5:05:29", "remaining_time": "5:12:05"}
{"current_steps": 1630, "total_steps": 3275, "loss": 0.0283, "lr": 5.909037507081121e-05, "epoch": 2.488312183951913, "percentage": 49.77, "elapsed_time": "5:06:53", "remaining_time": "5:09:43"}
{"current_steps": 1640, "total_steps": 3275, "loss": 0.029, "lr": 5.8565736256615434e-05, "epoch": 2.5035779028718634, "percentage": 50.08, "elapsed_time": "5:08:18", "remaining_time": "5:07:21"}
{"current_steps": 1650, "total_steps": 3275, "loss": 0.0245, "lr": 5.8040124022881625e-05, "epoch": 2.518843621791814, "percentage": 50.38, "elapsed_time": "5:09:42", "remaining_time": "5:05:00"}
{"current_steps": 1650, "total_steps": 3275, "eval_loss": 0.035995036363601685, "epoch": 2.518843621791814, "percentage": 50.38, "elapsed_time": "5:12:32", "remaining_time": "5:07:48"}
{"current_steps": 1660, "total_steps": 3275, "loss": 0.0173, "lr": 5.751359810075284e-05, "epoch": 2.534109340711764, "percentage": 50.69, "elapsed_time": "5:13:55", "remaining_time": "5:05:24"}
{"current_steps": 1670, "total_steps": 3275, "loss": 0.0125, "lr": 5.6986218325204676e-05, "epoch": 2.5493750596317146, "percentage": 50.99, "elapsed_time": "5:15:12", "remaining_time": "5:02:56"}
{"current_steps": 1680, "total_steps": 3275, "loss": 0.0158, "lr": 5.645804462824556e-05, "epoch": 2.5646407785516647, "percentage": 51.3, "elapsed_time": "5:16:29", "remaining_time": "5:00:29"}
{"current_steps": 1690, "total_steps": 3275, "loss": 0.0247, "lr": 5.5929137032106005e-05, "epoch": 2.5799064974716153, "percentage": 51.6, "elapsed_time": "5:17:46", "remaining_time": "4:58:02"}
{"current_steps": 1700, "total_steps": 3275, "loss": 0.0206, "lr": 5.53995556424176e-05, "epoch": 2.595172216391566, "percentage": 51.91, "elapsed_time": "5:19:03", "remaining_time": "4:55:36"}
{"current_steps": 1700, "total_steps": 3275, "eval_loss": 0.03758452460169792, "epoch": 2.595172216391566, "percentage": 51.91, "elapsed_time": "5:21:42", "remaining_time": "4:58:02"}
{"current_steps": 1710, "total_steps": 3275, "loss": 0.0175, "lr": 5.4869360641382615e-05, "epoch": 2.610437935311516, "percentage": 52.21, "elapsed_time": "5:23:03", "remaining_time": "4:55:39"}
{"current_steps": 1720, "total_steps": 3275, "loss": 0.0225, "lr": 5.433861228093471e-05, "epoch": 2.6257036542314665, "percentage": 52.52, "elapsed_time": "5:24:20", "remaining_time": "4:53:13"}
{"current_steps": 1730, "total_steps": 3275, "loss": 0.0179, "lr": 5.380737087589197e-05, "epoch": 2.640969373151417, "percentage": 52.82, "elapsed_time": "5:25:36", "remaining_time": "4:50:47"}
{"current_steps": 1740, "total_steps": 3275, "loss": 0.032, "lr": 5.327569679710256e-05, "epoch": 2.656235092071367, "percentage": 53.13, "elapsed_time": "5:26:53", "remaining_time": "4:48:23"}
{"current_steps": 1750, "total_steps": 3275, "loss": 0.0198, "lr": 5.274365046458416e-05, "epoch": 2.6715008109913176, "percentage": 53.44, "elapsed_time": "5:28:10", "remaining_time": "4:45:59"}
{"current_steps": 1750, "total_steps": 3275, "eval_loss": 0.036341313272714615, "epoch": 2.6715008109913176, "percentage": 53.44, "elapsed_time": "5:30:48", "remaining_time": "4:48:16"}
{"current_steps": 1760, "total_steps": 3275, "loss": 0.0256, "lr": 5.2211292340657804e-05, "epoch": 2.686766529911268, "percentage": 53.74, "elapsed_time": "5:32:09", "remaining_time": "4:45:55"}
{"current_steps": 1770, "total_steps": 3275, "loss": 0.0274, "lr": 5.167868292307678e-05, "epoch": 2.7020322488312183, "percentage": 54.05, "elapsed_time": "5:33:26", "remaining_time": "4:43:31"}
{"current_steps": 1780, "total_steps": 3275, "loss": 0.0184, "lr": 5.114588273815173e-05, "epoch": 2.717297967751169, "percentage": 54.35, "elapsed_time": "5:34:43", "remaining_time": "4:41:07"}
{"current_steps": 1790, "total_steps": 3275, "loss": 0.0272, "lr": 5.061295233387223e-05, "epoch": 2.732563686671119, "percentage": 54.66, "elapsed_time": "5:36:00", "remaining_time": "4:38:45"}
{"current_steps": 1800, "total_steps": 3275, "loss": 0.0209, "lr": 5.007995227302617e-05, "epoch": 2.7478294055910695, "percentage": 54.96, "elapsed_time": "5:37:17", "remaining_time": "4:36:23"}
{"current_steps": 1800, "total_steps": 3275, "eval_loss": 0.037039320915937424, "epoch": 2.7478294055910695, "percentage": 54.96, "elapsed_time": "5:39:55", "remaining_time": "4:38:33"}
{"current_steps": 1810, "total_steps": 3275, "loss": 0.022, "lr": 4.954694312631729e-05, "epoch": 2.76309512451102, "percentage": 55.27, "elapsed_time": "5:41:17", "remaining_time": "4:36:14"}
{"current_steps": 1820, "total_steps": 3275, "loss": 0.0205, "lr": 4.901398546548181e-05, "epoch": 2.7783608434309706, "percentage": 55.57, "elapsed_time": "5:42:34", "remaining_time": "4:33:51"}
{"current_steps": 1830, "total_steps": 3275, "loss": 0.0213, "lr": 4.848113985640513e-05, "epoch": 2.7936265623509207, "percentage": 55.88, "elapsed_time": "5:43:50", "remaining_time": "4:31:30"}
{"current_steps": 1840, "total_steps": 3275, "loss": 0.0231, "lr": 4.794846685223884e-05, "epoch": 2.808892281270871, "percentage": 56.18, "elapsed_time": "5:45:07", "remaining_time": "4:29:09"}
{"current_steps": 1850, "total_steps": 3275, "loss": 0.026, "lr": 4.741602698651966e-05, "epoch": 2.8241580001908213, "percentage": 56.49, "elapsed_time": "5:46:24", "remaining_time": "4:26:49"}
{"current_steps": 1850, "total_steps": 3275, "eval_loss": 0.036198943853378296, "epoch": 2.8241580001908213, "percentage": 56.49, "elapsed_time": "5:49:02", "remaining_time": "4:28:51"}
{"current_steps": 1860, "total_steps": 3275, "loss": 0.0276, "lr": 4.6883880766290086e-05, "epoch": 2.839423719110772, "percentage": 56.79, "elapsed_time": "5:50:23", "remaining_time": "4:26:34"}
{"current_steps": 1870, "total_steps": 3275, "loss": 0.0188, "lr": 4.635208866522251e-05, "epoch": 2.8546894380307224, "percentage": 57.1, "elapsed_time": "5:51:40", "remaining_time": "4:24:13"}
{"current_steps": 1880, "total_steps": 3275, "loss": 0.0193, "lr": 4.5820711116746785e-05, "epoch": 2.8699551569506725, "percentage": 57.4, "elapsed_time": "5:52:57", "remaining_time": "4:21:54"}
{"current_steps": 1890, "total_steps": 3275, "loss": 0.0237, "lr": 4.528980850718255e-05, "epoch": 2.885220875870623, "percentage": 57.71, "elapsed_time": "5:54:14", "remaining_time": "4:19:35"}
{"current_steps": 1900, "total_steps": 3275, "loss": 0.0197, "lr": 4.475944116887695e-05, "epoch": 2.9004865947905736, "percentage": 58.02, "elapsed_time": "5:55:31", "remaining_time": "4:17:17"}
{"current_steps": 1900, "total_steps": 3275, "eval_loss": 0.03581969812512398, "epoch": 2.9004865947905736, "percentage": 58.02, "elapsed_time": "5:58:09", "remaining_time": "4:19:11"}
{"current_steps": 1910, "total_steps": 3275, "loss": 0.0278, "lr": 4.4229669373348226e-05, "epoch": 2.9157523137105237, "percentage": 58.32, "elapsed_time": "5:59:31", "remaining_time": "4:16:56"}
{"current_steps": 1920, "total_steps": 3275, "loss": 0.0216, "lr": 4.3700553324436575e-05, "epoch": 2.9310180326304742, "percentage": 58.63, "elapsed_time": "6:00:48", "remaining_time": "4:14:37"}
{"current_steps": 1930, "total_steps": 3275, "loss": 0.0277, "lr": 4.317215315146238e-05, "epoch": 2.9462837515504248, "percentage": 58.93, "elapsed_time": "6:02:05", "remaining_time": "4:12:20"}
{"current_steps": 1940, "total_steps": 3275, "loss": 0.0242, "lr": 4.264452890239315e-05, "epoch": 2.961549470470375, "percentage": 59.24, "elapsed_time": "6:03:22", "remaining_time": "4:10:03"}
{"current_steps": 1950, "total_steps": 3275, "loss": 0.0291, "lr": 4.211774053701952e-05, "epoch": 2.9768151893903254, "percentage": 59.54, "elapsed_time": "6:04:39", "remaining_time": "4:07:46"}
{"current_steps": 1950, "total_steps": 3275, "eval_loss": 0.03554380312561989, "epoch": 2.9768151893903254, "percentage": 59.54, "elapsed_time": "6:07:17", "remaining_time": "4:09:34"}
{"current_steps": 1960, "total_steps": 3275, "loss": 0.0155, "lr": 4.159184792014145e-05, "epoch": 2.9920809083102755, "percentage": 59.85, "elapsed_time": "6:08:38", "remaining_time": "4:07:20"}
{"current_steps": 1970, "total_steps": 3275, "loss": 0.0118, "lr": 4.1066910814765016e-05, "epoch": 3.007346627230226, "percentage": 60.15, "elapsed_time": "6:09:55", "remaining_time": "4:05:03"}
{"current_steps": 1980, "total_steps": 3275, "loss": 0.0148, "lr": 4.0542988875310995e-05, "epoch": 3.0226123461501766, "percentage": 60.46, "elapsed_time": "6:11:12", "remaining_time": "4:02:47"}
{"current_steps": 1990, "total_steps": 3275, "loss": 0.0136, "lr": 4.002014164083552e-05, "epoch": 3.0378780650701267, "percentage": 60.76, "elapsed_time": "6:12:29", "remaining_time": "4:00:31"}
{"current_steps": 2000, "total_steps": 3275, "loss": 0.0091, "lr": 3.9498428528264204e-05, "epoch": 3.0531437839900772, "percentage": 61.07, "elapsed_time": "6:13:46", "remaining_time": "3:58:17"}
{"current_steps": 2000, "total_steps": 3275, "eval_loss": 0.04159301891922951, "epoch": 3.0531437839900772, "percentage": 61.07, "elapsed_time": "6:16:25", "remaining_time": "3:59:58"}
{"current_steps": 2010, "total_steps": 3275, "loss": 0.0092, "lr": 3.89779088256397e-05, "epoch": 3.068409502910028, "percentage": 61.37, "elapsed_time": "6:17:46", "remaining_time": "3:57:45"}
{"current_steps": 2020, "total_steps": 3275, "loss": 0.0097, "lr": 3.845864168538437e-05, "epoch": 3.083675221829978, "percentage": 61.68, "elapsed_time": "6:19:04", "remaining_time": "3:55:30"}
{"current_steps": 2030, "total_steps": 3275, "loss": 0.0117, "lr": 3.794068611757794e-05, "epoch": 3.0989409407499284, "percentage": 61.98, "elapsed_time": "6:20:20", "remaining_time": "3:53:16"}
{"current_steps": 2040, "total_steps": 3275, "loss": 0.0157, "lr": 3.7424100983251695e-05, "epoch": 3.114206659669879, "percentage": 62.29, "elapsed_time": "6:21:38", "remaining_time": "3:51:02"}
{"current_steps": 2050, "total_steps": 3275, "loss": 0.0132, "lr": 3.6908944987699345e-05, "epoch": 3.129472378589829, "percentage": 62.6, "elapsed_time": "6:22:55", "remaining_time": "3:48:49"}
{"current_steps": 2050, "total_steps": 3275, "eval_loss": 0.042059410363435745, "epoch": 3.129472378589829, "percentage": 62.6, "elapsed_time": "6:25:33", "remaining_time": "3:50:23"}
{"current_steps": 2060, "total_steps": 3275, "loss": 0.0087, "lr": 3.639527667380571e-05, "epoch": 3.1447380975097796, "percentage": 62.9, "elapsed_time": "6:26:55", "remaining_time": "3:48:12"}
{"current_steps": 2070, "total_steps": 3275, "loss": 0.0086, "lr": 3.5883154415393885e-05, "epoch": 3.16000381642973, "percentage": 63.21, "elapsed_time": "6:28:12", "remaining_time": "3:45:59"}
{"current_steps": 2080, "total_steps": 3275, "loss": 0.0111, "lr": 3.537263641059152e-05, "epoch": 3.1752695353496803, "percentage": 63.51, "elapsed_time": "6:29:29", "remaining_time": "3:43:46"}
{"current_steps": 2090, "total_steps": 3275, "loss": 0.01, "lr": 3.486378067521718e-05, "epoch": 3.190535254269631, "percentage": 63.82, "elapsed_time": "6:30:46", "remaining_time": "3:41:33"}
{"current_steps": 2100, "total_steps": 3275, "loss": 0.0115, "lr": 3.435664503618732e-05, "epoch": 3.2058009731895813, "percentage": 64.12, "elapsed_time": "6:32:03", "remaining_time": "3:39:22"}
{"current_steps": 2100, "total_steps": 3275, "eval_loss": 0.04429946094751358, "epoch": 3.2058009731895813, "percentage": 64.12, "elapsed_time": "6:34:42", "remaining_time": "3:40:50"}
{"current_steps": 2110, "total_steps": 3275, "loss": 0.0108, "lr": 3.3851287124944756e-05, "epoch": 3.2210666921095314, "percentage": 64.43, "elapsed_time": "6:36:03", "remaining_time": "3:38:40"}
{"current_steps": 2120, "total_steps": 3275, "loss": 0.0086, "lr": 3.334776437090944e-05, "epoch": 3.236332411029482, "percentage": 64.73, "elapsed_time": "6:37:21", "remaining_time": "3:36:28"}
{"current_steps": 2130, "total_steps": 3275, "loss": 0.0101, "lr": 3.2846133994952046e-05, "epoch": 3.251598129949432, "percentage": 65.04, "elapsed_time": "6:38:38", "remaining_time": "3:34:17"}
{"current_steps": 2140, "total_steps": 3275, "loss": 0.0061, "lr": 3.234645300289136e-05, "epoch": 3.2668638488693826, "percentage": 65.34, "elapsed_time": "6:39:55", "remaining_time": "3:32:06"}
{"current_steps": 2150, "total_steps": 3275, "loss": 0.0131, "lr": 3.1848778179016075e-05, "epoch": 3.282129567789333, "percentage": 65.65, "elapsed_time": "6:41:12", "remaining_time": "3:29:55"}
{"current_steps": 2150, "total_steps": 3275, "eval_loss": 0.04585850238800049, "epoch": 3.282129567789333, "percentage": 65.65, "elapsed_time": "6:43:50", "remaining_time": "3:31:18"}
{"current_steps": 2160, "total_steps": 3275, "loss": 0.0127, "lr": 3.135316607963176e-05, "epoch": 3.2973952867092833, "percentage": 65.95, "elapsed_time": "6:45:11", "remaining_time": "3:29:09"}
{"current_steps": 2170, "total_steps": 3275, "loss": 0.0108, "lr": 3.085967302663375e-05, "epoch": 3.312661005629234, "percentage": 66.26, "elapsed_time": "6:46:28", "remaining_time": "3:26:59"}
{"current_steps": 2180, "total_steps": 3275, "loss": 0.0168, "lr": 3.0368355101106615e-05, "epoch": 3.3279267245491844, "percentage": 66.56, "elapsed_time": "6:47:45", "remaining_time": "3:24:48"}
{"current_steps": 2190, "total_steps": 3275, "loss": 0.0095, "lr": 2.987926813695116e-05, "epoch": 3.3431924434691345, "percentage": 66.87, "elapsed_time": "6:49:02", "remaining_time": "3:22:39"}
{"current_steps": 2200, "total_steps": 3275, "loss": 0.0132, "lr": 2.939246771453924e-05, "epoch": 3.358458162389085, "percentage": 67.18, "elapsed_time": "6:50:19", "remaining_time": "3:20:30"}
{"current_steps": 2200, "total_steps": 3275, "eval_loss": 0.04094408452510834, "epoch": 3.358458162389085, "percentage": 67.18, "elapsed_time": "6:52:58", "remaining_time": "3:21:47"}
{"current_steps": 2210, "total_steps": 3275, "loss": 0.0136, "lr": 2.890800915439772e-05, "epoch": 3.3737238813090356, "percentage": 67.48, "elapsed_time": "6:54:19", "remaining_time": "3:19:39"}
{"current_steps": 2220, "total_steps": 3275, "loss": 0.0094, "lr": 2.842594751092159e-05, "epoch": 3.3889896002289857, "percentage": 67.79, "elapsed_time": "6:55:36", "remaining_time": "3:17:30"}
{"current_steps": 2230, "total_steps": 3275, "loss": 0.0111, "lr": 2.794633756611776e-05, "epoch": 3.404255319148936, "percentage": 68.09, "elapsed_time": "6:56:53", "remaining_time": "3:15:21"}
{"current_steps": 2240, "total_steps": 3275, "loss": 0.0106, "lr": 2.7469233823379347e-05, "epoch": 3.4195210380688867, "percentage": 68.4, "elapsed_time": "6:58:10", "remaining_time": "3:13:13"}
{"current_steps": 2250, "total_steps": 3275, "loss": 0.0077, "lr": 2.6994690501292032e-05, "epoch": 3.434786756988837, "percentage": 68.7, "elapsed_time": "6:59:27", "remaining_time": "3:11:05"}
{"current_steps": 2250, "total_steps": 3275, "eval_loss": 0.04445834830403328, "epoch": 3.434786756988837, "percentage": 68.7, "elapsed_time": "7:02:06", "remaining_time": "3:12:17"}
{"current_steps": 2260, "total_steps": 3275, "loss": 0.0073, "lr": 2.652276152747246e-05, "epoch": 3.4500524759087874, "percentage": 69.01, "elapsed_time": "7:03:27", "remaining_time": "3:10:11"}
{"current_steps": 2270, "total_steps": 3275, "loss": 0.0058, "lr": 2.6053500532439968e-05, "epoch": 3.465318194828738, "percentage": 69.31, "elapsed_time": "7:04:45", "remaining_time": "3:08:03"}
{"current_steps": 2280, "total_steps": 3275, "loss": 0.0135, "lr": 2.5586960843521824e-05, "epoch": 3.480583913748688, "percentage": 69.62, "elapsed_time": "7:06:02", "remaining_time": "3:05:55"}
{"current_steps": 2290, "total_steps": 3275, "loss": 0.0137, "lr": 2.5123195478793217e-05, "epoch": 3.4958496326686386, "percentage": 69.92, "elapsed_time": "7:07:19", "remaining_time": "3:03:48"}
{"current_steps": 2300, "total_steps": 3275, "loss": 0.0156, "lr": 2.466225714105202e-05, "epoch": 3.5111153515885887, "percentage": 70.23, "elapsed_time": "7:08:36", "remaining_time": "3:01:41"}
{"current_steps": 2300, "total_steps": 3275, "eval_loss": 0.04438282549381256, "epoch": 3.5111153515885887, "percentage": 70.23, "elapsed_time": "7:11:14", "remaining_time": "3:02:48"}
{"current_steps": 2310, "total_steps": 3275, "loss": 0.014, "lr": 2.420419821182982e-05, "epoch": 3.526381070508539, "percentage": 70.53, "elapsed_time": "7:12:36", "remaining_time": "3:00:43"}
{"current_steps": 2320, "total_steps": 3275, "loss": 0.0097, "lr": 2.3749070745438996e-05, "epoch": 3.5416467894284898, "percentage": 70.84, "elapsed_time": "7:13:53", "remaining_time": "2:58:36"}
{"current_steps": 2330, "total_steps": 3275, "loss": 0.0092, "lr": 2.3296926463057396e-05, "epoch": 3.5569125083484403, "percentage": 71.15, "elapsed_time": "7:15:10", "remaining_time": "2:56:29"}
{"current_steps": 2340, "total_steps": 3275, "loss": 0.0076, "lr": 2.284781674685058e-05, "epoch": 3.5721782272683904, "percentage": 71.45, "elapsed_time": "7:16:27", "remaining_time": "2:54:23"}
{"current_steps": 2350, "total_steps": 3275, "loss": 0.0125, "lr": 2.2401792634132708e-05, "epoch": 3.587443946188341, "percentage": 71.76, "elapsed_time": "7:17:45", "remaining_time": "2:52:18"}
{"current_steps": 2350, "total_steps": 3275, "eval_loss": 0.048045314848423004, "epoch": 3.587443946188341, "percentage": 71.76, "elapsed_time": "7:20:23", "remaining_time": "2:53:20"}
{"current_steps": 2360, "total_steps": 3275, "loss": 0.0094, "lr": 2.195890481156666e-05, "epoch": 3.602709665108291, "percentage": 72.06, "elapsed_time": "7:21:45", "remaining_time": "2:51:16"}
{"current_steps": 2370, "total_steps": 3275, "loss": 0.0122, "lr": 2.151920360940387e-05, "epoch": 3.6179753840282416, "percentage": 72.37, "elapsed_time": "7:23:02", "remaining_time": "2:49:10"}
{"current_steps": 2380, "total_steps": 3275, "loss": 0.0075, "lr": 2.1082738995764785e-05, "epoch": 3.633241102948192, "percentage": 72.67, "elapsed_time": "7:24:19", "remaining_time": "2:47:05"}
{"current_steps": 2390, "total_steps": 3275, "loss": 0.0063, "lr": 2.0649560570960465e-05, "epoch": 3.6485068218681422, "percentage": 72.98, "elapsed_time": "7:25:36", "remaining_time": "2:45:00"}
{"current_steps": 2400, "total_steps": 3275, "loss": 0.0089, "lr": 2.0219717561855855e-05, "epoch": 3.663772540788093, "percentage": 73.28, "elapsed_time": "7:26:53", "remaining_time": "2:42:55"}
{"current_steps": 2400, "total_steps": 3275, "eval_loss": 0.049949660897254944, "epoch": 3.663772540788093, "percentage": 73.28, "elapsed_time": "7:29:32", "remaining_time": "2:43:53"}
{"current_steps": 2410, "total_steps": 3275, "loss": 0.0168, "lr": 1.9793258816275728e-05, "epoch": 3.679038259708043, "percentage": 73.59, "elapsed_time": "7:30:53", "remaining_time": "2:41:50"}
{"current_steps": 2420, "total_steps": 3275, "loss": 0.0106, "lr": 1.9370232797453402e-05, "epoch": 3.6943039786279934, "percentage": 73.89, "elapsed_time": "7:32:10", "remaining_time": "2:39:45"}
{"current_steps": 2430, "total_steps": 3275, "loss": 0.01, "lr": 1.8950687578523502e-05, "epoch": 3.709569697547944, "percentage": 74.2, "elapsed_time": "7:33:27", "remaining_time": "2:37:41"}
{"current_steps": 2440, "total_steps": 3275, "loss": 0.0067, "lr": 1.853467083705869e-05, "epoch": 3.7248354164678945, "percentage": 74.5, "elapsed_time": "7:34:44", "remaining_time": "2:35:37"}
{"current_steps": 2450, "total_steps": 3275, "loss": 0.0125, "lr": 1.8122229849651716e-05, "epoch": 3.7401011353878446, "percentage": 74.81, "elapsed_time": "7:36:01", "remaining_time": "2:33:33"}
{"current_steps": 2450, "total_steps": 3275, "eval_loss": 0.04671522229909897, "epoch": 3.7401011353878446, "percentage": 74.81, "elapsed_time": "7:38:39", "remaining_time": "2:34:26"}
{"current_steps": 2460, "total_steps": 3275, "loss": 0.014, "lr": 1.7713411486542707e-05, "epoch": 3.755366854307795, "percentage": 75.11, "elapsed_time": "7:40:01", "remaining_time": "2:32:24"}
{"current_steps": 2470, "total_steps": 3275, "loss": 0.0076, "lr": 1.7308262206292897e-05, "epoch": 3.7706325732277453, "percentage": 75.42, "elapsed_time": "7:41:19", "remaining_time": "2:30:21"}
{"current_steps": 2480, "total_steps": 3275, "loss": 0.015, "lr": 1.6906828050504907e-05, "epoch": 3.785898292147696, "percentage": 75.73, "elapsed_time": "7:42:36", "remaining_time": "2:28:17"}
{"current_steps": 2490, "total_steps": 3275, "loss": 0.0096, "lr": 1.650915463859068e-05, "epoch": 3.8011640110676463, "percentage": 76.03, "elapsed_time": "7:43:53", "remaining_time": "2:26:14"}
{"current_steps": 2500, "total_steps": 3275, "loss": 0.0115, "lr": 1.6115287162587055e-05, "epoch": 3.8164297299875964, "percentage": 76.34, "elapsed_time": "7:45:10", "remaining_time": "2:24:12"}
{"current_steps": 2500, "total_steps": 3275, "eval_loss": 0.04467522352933884, "epoch": 3.8164297299875964, "percentage": 76.34, "elapsed_time": "7:47:49", "remaining_time": "2:25:01"}
{"current_steps": 2510, "total_steps": 3275, "loss": 0.0064, "lr": 1.57252703820203e-05, "epoch": 3.831695448907547, "percentage": 76.64, "elapsed_time": "7:49:10", "remaining_time": "2:22:59"}
{"current_steps": 2520, "total_steps": 3275, "loss": 0.0159, "lr": 1.5339148618819393e-05, "epoch": 3.8469611678274975, "percentage": 76.95, "elapsed_time": "7:50:27", "remaining_time": "2:20:57"}
{"current_steps": 2530, "total_steps": 3275, "loss": 0.0135, "lr": 1.4956965752279395e-05, "epoch": 3.8622268867474476, "percentage": 77.25, "elapsed_time": "7:51:44", "remaining_time": "2:18:54"}
{"current_steps": 2540, "total_steps": 3275, "loss": 0.0103, "lr": 1.457876521407484e-05, "epoch": 3.877492605667398, "percentage": 77.56, "elapsed_time": "7:53:02", "remaining_time": "2:16:52"}
{"current_steps": 2550, "total_steps": 3275, "loss": 0.0062, "lr": 1.4204589983324173e-05, "epoch": 3.8927583245873487, "percentage": 77.86, "elapsed_time": "7:54:18", "remaining_time": "2:14:51"}
{"current_steps": 2550, "total_steps": 3275, "eval_loss": 0.044938329607248306, "epoch": 3.8927583245873487, "percentage": 77.86, "elapsed_time": "7:56:56", "remaining_time": "2:15:36"}
{"current_steps": 2560, "total_steps": 3275, "loss": 0.0095, "lr": 1.383448258170557e-05, "epoch": 3.908024043507299, "percentage": 78.17, "elapsed_time": "7:58:18", "remaining_time": "2:13:35"}
{"current_steps": 2570, "total_steps": 3275, "loss": 0.0069, "lr": 1.3468485068624653e-05, "epoch": 3.9232897624272494, "percentage": 78.47, "elapsed_time": "7:59:35", "remaining_time": "2:11:33"}
{"current_steps": 2580, "total_steps": 3275, "loss": 0.0119, "lr": 1.310663903643492e-05, "epoch": 3.9385554813471995, "percentage": 78.78, "elapsed_time": "8:00:52", "remaining_time": "2:09:32"}
{"current_steps": 2590, "total_steps": 3275, "loss": 0.0109, "lr": 1.2748985605711028e-05, "epoch": 3.95382120026715, "percentage": 79.08, "elapsed_time": "8:02:10", "remaining_time": "2:07:31"}
{"current_steps": 2600, "total_steps": 3275, "loss": 0.0112, "lr": 1.2395565420575932e-05, "epoch": 3.9690869191871005, "percentage": 79.39, "elapsed_time": "8:03:27", "remaining_time": "2:05:30"}
{"current_steps": 2600, "total_steps": 3275, "eval_loss": 0.046159546822309494, "epoch": 3.9690869191871005, "percentage": 79.39, "elapsed_time": "8:06:05", "remaining_time": "2:06:11"}
{"current_steps": 2610, "total_steps": 3275, "loss": 0.0194, "lr": 1.2046418644081903e-05, "epoch": 3.984352638107051, "percentage": 79.69, "elapsed_time": "8:07:26", "remaining_time": "2:04:11"}
{"current_steps": 2620, "total_steps": 3275, "loss": 0.0084, "lr": 1.1701584953646505e-05, "epoch": 3.999618357027001, "percentage": 80.0, "elapsed_time": "8:08:43", "remaining_time": "2:02:10"}
{"current_steps": 2630, "total_steps": 3275, "loss": 0.006, "lr": 1.1361103536543466e-05, "epoch": 4.014884075946951, "percentage": 80.31, "elapsed_time": "8:10:00", "remaining_time": "2:00:10"}
{"current_steps": 2640, "total_steps": 3275, "loss": 0.0033, "lr": 1.1025013085449527e-05, "epoch": 4.030149794866902, "percentage": 80.61, "elapsed_time": "8:11:17", "remaining_time": "1:58:10"}
{"current_steps": 2650, "total_steps": 3275, "loss": 0.005, "lr": 1.0693351794047224e-05, "epoch": 4.045415513786852, "percentage": 80.92, "elapsed_time": "8:12:34", "remaining_time": "1:56:10"}
{"current_steps": 2650, "total_steps": 3275, "eval_loss": 0.04647543653845787, "epoch": 4.045415513786852, "percentage": 80.92, "elapsed_time": "8:15:12", "remaining_time": "1:56:47"}
{"current_steps": 2660, "total_steps": 3275, "loss": 0.0043, "lr": 1.036615735268468e-05, "epoch": 4.060681232706803, "percentage": 81.22, "elapsed_time": "8:16:33", "remaining_time": "1:54:48"}
{"current_steps": 2670, "total_steps": 3275, "loss": 0.0058, "lr": 1.0043466944092272e-05, "epoch": 4.0759469516267535, "percentage": 81.53, "elapsed_time": "8:17:50", "remaining_time": "1:52:48"}
{"current_steps": 2680, "total_steps": 3275, "loss": 0.0081, "lr": 9.72531723915726e-06, "epoch": 4.091212670546703, "percentage": 81.83, "elapsed_time": "8:19:07", "remaining_time": "1:50:48"}
{"current_steps": 2690, "total_steps": 3275, "loss": 0.0055, "lr": 9.411744392756405e-06, "epoch": 4.106478389466654, "percentage": 82.14, "elapsed_time": "8:20:24", "remaining_time": "1:48:49"}
{"current_steps": 2700, "total_steps": 3275, "loss": 0.0065, "lr": 9.102784039647339e-06, "epoch": 4.121744108386604, "percentage": 82.44, "elapsed_time": "8:21:41", "remaining_time": "1:46:50"}
{"current_steps": 2700, "total_steps": 3275, "eval_loss": 0.050155650824308395, "epoch": 4.121744108386604, "percentage": 82.44, "elapsed_time": "8:24:19", "remaining_time": "1:47:24"}
{"current_steps": 2710, "total_steps": 3275, "loss": 0.007, "lr": 8.79847129041893e-06, "epoch": 4.137009827306555, "percentage": 82.75, "elapsed_time": "8:25:41", "remaining_time": "1:45:25"}
{"current_steps": 2720, "total_steps": 3275, "loss": 0.0032, "lr": 8.498840727501316e-06, "epoch": 4.152275546226505, "percentage": 83.05, "elapsed_time": "8:26:58", "remaining_time": "1:43:26"}
{"current_steps": 2730, "total_steps": 3275, "loss": 0.0028, "lr": 8.203926401235957e-06, "epoch": 4.167541265146456, "percentage": 83.36, "elapsed_time": "8:28:15", "remaining_time": "1:41:27"}
{"current_steps": 2740, "total_steps": 3275, "loss": 0.0029, "lr": 7.913761826006017e-06, "epoch": 4.1828069840664055, "percentage": 83.66, "elapsed_time": "8:29:32", "remaining_time": "1:39:29"}
{"current_steps": 2750, "total_steps": 3275, "loss": 0.0021, "lr": 7.628379976427868e-06, "epoch": 4.198072702986356, "percentage": 83.97, "elapsed_time": "8:30:49", "remaining_time": "1:37:31"}
{"current_steps": 2750, "total_steps": 3275, "eval_loss": 0.054289042949676514, "epoch": 4.198072702986356, "percentage": 83.97, "elapsed_time": "8:33:27", "remaining_time": "1:38:01"}
{"current_steps": 2760, "total_steps": 3275, "loss": 0.008, "lr": 7.347813283603705e-06, "epoch": 4.213338421906307, "percentage": 84.27, "elapsed_time": "8:34:49", "remaining_time": "1:36:03"}
{"current_steps": 2770, "total_steps": 3275, "loss": 0.0041, "lr": 7.072093631436161e-06, "epoch": 4.228604140826257, "percentage": 84.58, "elapsed_time": "8:36:06", "remaining_time": "1:34:05"}
{"current_steps": 2780, "total_steps": 3275, "loss": 0.0035, "lr": 6.801252353004867e-06, "epoch": 4.243869859746208, "percentage": 84.89, "elapsed_time": "8:37:23", "remaining_time": "1:32:07"}
{"current_steps": 2790, "total_steps": 3275, "loss": 0.0037, "lr": 6.535320227005826e-06, "epoch": 4.259135578666158, "percentage": 85.19, "elapsed_time": "8:38:40", "remaining_time": "1:30:09"}
{"current_steps": 2800, "total_steps": 3275, "loss": 0.0033, "lr": 6.274327474253611e-06, "epoch": 4.274401297586108, "percentage": 85.5, "elapsed_time": "8:39:58", "remaining_time": "1:28:12"}
{"current_steps": 2800, "total_steps": 3275, "eval_loss": 0.055618129670619965, "epoch": 4.274401297586108, "percentage": 85.5, "elapsed_time": "8:42:36", "remaining_time": "1:28:39"}
{"current_steps": 2810, "total_steps": 3275, "loss": 0.004, "lr": 6.018303754247112e-06, "epoch": 4.289667016506058, "percentage": 85.8, "elapsed_time": "8:43:57", "remaining_time": "1:26:42"}
{"current_steps": 2820, "total_steps": 3275, "loss": 0.0045, "lr": 5.767278161798911e-06, "epoch": 4.304932735426009, "percentage": 86.11, "elapsed_time": "8:45:14", "remaining_time": "1:24:44"}
{"current_steps": 2830, "total_steps": 3275, "loss": 0.0018, "lr": 5.521279223729026e-06, "epoch": 4.3201984543459595, "percentage": 86.41, "elapsed_time": "8:46:32", "remaining_time": "1:22:47"}
{"current_steps": 2840, "total_steps": 3275, "loss": 0.0036, "lr": 5.280334895622968e-06, "epoch": 4.33546417326591, "percentage": 86.72, "elapsed_time": "8:47:48", "remaining_time": "1:20:50"}
{"current_steps": 2850, "total_steps": 3275, "loss": 0.0068, "lr": 5.044472558654961e-06, "epoch": 4.350729892185861, "percentage": 87.02, "elapsed_time": "8:49:05", "remaining_time": "1:18:53"}
{"current_steps": 2850, "total_steps": 3275, "eval_loss": 0.057221900671720505, "epoch": 4.350729892185861, "percentage": 87.02, "elapsed_time": "8:51:43", "remaining_time": "1:19:17"}
{"current_steps": 2860, "total_steps": 3275, "loss": 0.0024, "lr": 4.813719016476203e-06, "epoch": 4.36599561110581, "percentage": 87.33, "elapsed_time": "8:53:05", "remaining_time": "1:17:21"}
{"current_steps": 2870, "total_steps": 3275, "loss": 0.0017, "lr": 4.588100492168973e-06, "epoch": 4.381261330025761, "percentage": 87.63, "elapsed_time": "8:54:22", "remaining_time": "1:15:24"}
{"current_steps": 2880, "total_steps": 3275, "loss": 0.0029, "lr": 4.367642625266511e-06, "epoch": 4.396527048945711, "percentage": 87.94, "elapsed_time": "8:55:39", "remaining_time": "1:13:27"}
{"current_steps": 2890, "total_steps": 3275, "loss": 0.0023, "lr": 4.1523704688394176e-06, "epoch": 4.411792767865662, "percentage": 88.24, "elapsed_time": "8:56:56", "remaining_time": "1:11:31"}
{"current_steps": 2900, "total_steps": 3275, "loss": 0.0015, "lr": 3.9423084866484884e-06, "epoch": 4.427058486785612, "percentage": 88.55, "elapsed_time": "8:58:13", "remaining_time": "1:09:35"}
{"current_steps": 2900, "total_steps": 3275, "eval_loss": 0.05985542759299278, "epoch": 4.427058486785612, "percentage": 88.55, "elapsed_time": "9:00:51", "remaining_time": "1:09:56"}
{"current_steps": 2910, "total_steps": 3275, "loss": 0.0071, "lr": 3.737480550364736e-06, "epoch": 4.442324205705562, "percentage": 88.85, "elapsed_time": "9:02:12", "remaining_time": "1:08:00"}
{"current_steps": 2920, "total_steps": 3275, "loss": 0.0041, "lr": 3.5379099368564817e-06, "epoch": 4.457589924625513, "percentage": 89.16, "elapsed_time": "9:03:29", "remaining_time": "1:06:04"}
{"current_steps": 2930, "total_steps": 3275, "loss": 0.0026, "lr": 3.3436193255442396e-06, "epoch": 4.472855643545463, "percentage": 89.47, "elapsed_time": "9:04:47", "remaining_time": "1:04:08"}
{"current_steps": 2940, "total_steps": 3275, "loss": 0.0031, "lr": 3.1546307958233214e-06, "epoch": 4.488121362465414, "percentage": 89.77, "elapsed_time": "9:06:03", "remaining_time": "1:02:13"}
{"current_steps": 2950, "total_steps": 3275, "loss": 0.0036, "lr": 2.9709658245547834e-06, "epoch": 4.503387081385364, "percentage": 90.08, "elapsed_time": "9:07:21", "remaining_time": "1:00:18"}
{"current_steps": 2950, "total_steps": 3275, "eval_loss": 0.060201384127140045, "epoch": 4.503387081385364, "percentage": 90.08, "elapsed_time": "9:09:59", "remaining_time": "1:00:35"}
{"current_steps": 2960, "total_steps": 3275, "loss": 0.0027, "lr": 2.792645283624712e-06, "epoch": 4.518652800305315, "percentage": 90.38, "elapsed_time": "9:11:20", "remaining_time": "0:58:40"}
{"current_steps": 2970, "total_steps": 3275, "loss": 0.0014, "lr": 2.6196894375723645e-06, "epoch": 4.5339185192252645, "percentage": 90.69, "elapsed_time": "9:12:37", "remaining_time": "0:56:45"}
{"current_steps": 2980, "total_steps": 3275, "loss": 0.0064, "lr": 2.452117941287246e-06, "epoch": 4.549184238145215, "percentage": 90.99, "elapsed_time": "9:13:55", "remaining_time": "0:54:50"}
{"current_steps": 2990, "total_steps": 3275, "loss": 0.0012, "lr": 2.2899498377755566e-06, "epoch": 4.5644499570651655, "percentage": 91.3, "elapsed_time": "9:15:12", "remaining_time": "0:52:55"}
{"current_steps": 3000, "total_steps": 3275, "loss": 0.0027, "lr": 2.1332035559960663e-06, "epoch": 4.579715675985116, "percentage": 91.6, "elapsed_time": "9:16:29", "remaining_time": "0:51:00"}
{"current_steps": 3000, "total_steps": 3275, "eval_loss": 0.06145218014717102, "epoch": 4.579715675985116, "percentage": 91.6, "elapsed_time": "9:19:07", "remaining_time": "0:51:15"}
{"current_steps": 3010, "total_steps": 3275, "loss": 0.0043, "lr": 1.9818969087658735e-06, "epoch": 4.594981394905067, "percentage": 91.91, "elapsed_time": "9:20:29", "remaining_time": "0:49:20"}
{"current_steps": 3020, "total_steps": 3275, "loss": 0.0011, "lr": 1.8360470907361093e-06, "epoch": 4.610247113825016, "percentage": 92.21, "elapsed_time": "9:21:45", "remaining_time": "0:47:26"}
{"current_steps": 3030, "total_steps": 3275, "loss": 0.0056, "lr": 1.6956706764379438e-06, "epoch": 4.625512832744967, "percentage": 92.52, "elapsed_time": "9:23:03", "remaining_time": "0:45:31"}
{"current_steps": 3040, "total_steps": 3275, "loss": 0.0028, "lr": 1.5607836183989921e-06, "epoch": 4.640778551664917, "percentage": 92.82, "elapsed_time": "9:24:20", "remaining_time": "0:43:37"}
{"current_steps": 3050, "total_steps": 3275, "loss": 0.0013, "lr": 1.4314012453305215e-06, "epoch": 4.656044270584868, "percentage": 93.13, "elapsed_time": "9:25:37", "remaining_time": "0:41:43"}
{"current_steps": 3050, "total_steps": 3275, "eval_loss": 0.061488546431064606, "epoch": 4.656044270584868, "percentage": 93.13, "elapsed_time": "9:28:15", "remaining_time": "0:41:55"}
{"current_steps": 3060, "total_steps": 3275, "loss": 0.0028, "lr": 1.3075382603854157e-06, "epoch": 4.6713099895048185, "percentage": 93.44, "elapsed_time": "9:29:37", "remaining_time": "0:40:01"}
{"current_steps": 3070, "total_steps": 3275, "loss": 0.0027, "lr": 1.1892087394873353e-06, "epoch": 4.686575708424769, "percentage": 93.74, "elapsed_time": "9:30:54", "remaining_time": "0:38:07"}
{"current_steps": 3080, "total_steps": 3275, "loss": 0.0038, "lr": 1.076426129731084e-06, "epoch": 4.701841427344719, "percentage": 94.05, "elapsed_time": "9:32:11", "remaining_time": "0:36:13"}
{"current_steps": 3090, "total_steps": 3275, "loss": 0.005, "lr": 9.692032478545e-07, "epoch": 4.717107146264669, "percentage": 94.35, "elapsed_time": "9:33:28", "remaining_time": "0:34:20"}
{"current_steps": 3100, "total_steps": 3275, "loss": 0.0056, "lr": 8.675522787819023e-07, "epoch": 4.73237286518462, "percentage": 94.66, "elapsed_time": "9:34:45", "remaining_time": "0:32:26"}
{"current_steps": 3100, "total_steps": 3275, "eval_loss": 0.061799220740795135, "epoch": 4.73237286518462, "percentage": 94.66, "elapsed_time": "9:37:24", "remaining_time": "0:32:35"}
{"current_steps": 3110, "total_steps": 3275, "loss": 0.0051, "lr": 7.714847742394337e-07, "epoch": 4.74763858410457, "percentage": 94.96, "elapsed_time": "9:38:45", "remaining_time": "0:30:42"}
{"current_steps": 3120, "total_steps": 3275, "loss": 0.0089, "lr": 6.810116514422593e-07, "epoch": 4.762904303024521, "percentage": 95.27, "elapsed_time": "9:40:02", "remaining_time": "0:28:48"}
{"current_steps": 3130, "total_steps": 3275, "loss": 0.0031, "lr": 5.961431918539817e-07, "epoch": 4.7781700219444705, "percentage": 95.57, "elapsed_time": "9:41:19", "remaining_time": "0:26:55"}
{"current_steps": 3140, "total_steps": 3275, "loss": 0.0058, "lr": 5.16889040018187e-07, "epoch": 4.793435740864421, "percentage": 95.88, "elapsed_time": "9:42:36", "remaining_time": "0:25:02"}
{"current_steps": 3150, "total_steps": 3275, "loss": 0.0028, "lr": 4.432582024624543e-07, "epoch": 4.808701459784372, "percentage": 96.18, "elapsed_time": "9:43:53", "remaining_time": "0:23:10"}
{"current_steps": 3150, "total_steps": 3275, "eval_loss": 0.06179770454764366, "epoch": 4.808701459784372, "percentage": 96.18, "elapsed_time": "9:46:32", "remaining_time": "0:23:16"}
{"current_steps": 3160, "total_steps": 3275, "loss": 0.0028, "lr": 3.7525904667484737e-07, "epoch": 4.823967178704322, "percentage": 96.49, "elapsed_time": "9:47:56", "remaining_time": "0:21:23"}
{"current_steps": 3170, "total_steps": 3275, "loss": 0.0037, "lr": 3.128993001530245e-07, "epoch": 4.839232897624273, "percentage": 96.79, "elapsed_time": "9:49:13", "remaining_time": "0:19:31"}
{"current_steps": 3180, "total_steps": 3275, "loss": 0.003, "lr": 2.5618604952605816e-07, "epoch": 4.854498616544223, "percentage": 97.1, "elapsed_time": "9:50:31", "remaining_time": "0:17:38"}
{"current_steps": 3190, "total_steps": 3275, "loss": 0.0034, "lr": 2.0512573974912908e-07, "epoch": 4.869764335464174, "percentage": 97.4, "elapsed_time": "9:51:48", "remaining_time": "0:15:46"}
{"current_steps": 3200, "total_steps": 3275, "loss": 0.0044, "lr": 1.597241733711008e-07, "epoch": 4.885030054384123, "percentage": 97.71, "elapsed_time": "9:53:05", "remaining_time": "0:13:54"}
{"current_steps": 3200, "total_steps": 3275, "eval_loss": 0.06198294460773468, "epoch": 4.885030054384123, "percentage": 97.71, "elapsed_time": "9:55:43", "remaining_time": "0:13:57"}
{"current_steps": 3210, "total_steps": 3275, "loss": 0.0077, "lr": 1.1998650987510295e-07, "epoch": 4.900295773304074, "percentage": 98.02, "elapsed_time": "9:57:05", "remaining_time": "0:12:05"}
{"current_steps": 3220, "total_steps": 3275, "loss": 0.0097, "lr": 8.591726509222242e-08, "epoch": 4.9155614922240245, "percentage": 98.32, "elapsed_time": "9:58:21", "remaining_time": "0:10:13"}
{"current_steps": 3230, "total_steps": 3275, "loss": 0.0048, "lr": 5.752031068830266e-08, "epoch": 4.930827211143975, "percentage": 98.63, "elapsed_time": "9:59:38", "remaining_time": "0:08:21"}
{"current_steps": 3240, "total_steps": 3275, "loss": 0.0013, "lr": 3.4798873723984605e-08, "epoch": 4.946092930063926, "percentage": 98.93, "elapsed_time": "10:00:55", "remaining_time": "0:06:29"}
{"current_steps": 3250, "total_steps": 3275, "loss": 0.0061, "lr": 1.7755536287944464e-08, "epoch": 4.961358648983875, "percentage": 99.24, "elapsed_time": "10:02:12", "remaining_time": "0:04:37"}
{"current_steps": 3250, "total_steps": 3275, "eval_loss": 0.06215130165219307, "epoch": 4.961358648983875, "percentage": 99.24, "elapsed_time": "10:04:51", "remaining_time": "0:04:39"}
{"current_steps": 3260, "total_steps": 3275, "loss": 0.0082, "lr": 6.3922352034895276e-09, "epoch": 4.976624367903826, "percentage": 99.54, "elapsed_time": "10:06:13", "remaining_time": "0:02:47"}
{"current_steps": 3270, "total_steps": 3275, "loss": 0.005, "lr": 7.102618084620094e-10, "epoch": 4.991890086823776, "percentage": 99.85, "elapsed_time": "10:07:30", "remaining_time": "0:00:55"}
{"current_steps": 3275, "total_steps": 3275, "epoch": 4.999522946283752, "percentage": 100.0, "elapsed_time": "10:08:13", "remaining_time": "0:00:00"}