chchen's picture
Training in progress, step 3255
b0a25fb verified
{"current_steps": 10, "total_steps": 3255, "loss": 1.6312, "lr": 3.067484662576687e-06, "epoch": 0.01535213970447131, "percentage": 0.31, "elapsed_time": "0:01:18", "remaining_time": "7:05:26"}
{"current_steps": 20, "total_steps": 3255, "loss": 1.4429, "lr": 6.134969325153374e-06, "epoch": 0.03070427940894262, "percentage": 0.61, "elapsed_time": "0:02:35", "remaining_time": "6:59:03"}
{"current_steps": 30, "total_steps": 3255, "loss": 1.4142, "lr": 9.202453987730062e-06, "epoch": 0.04605641911341393, "percentage": 0.92, "elapsed_time": "0:03:52", "remaining_time": "6:56:03"}
{"current_steps": 40, "total_steps": 3255, "loss": 0.9213, "lr": 1.2269938650306748e-05, "epoch": 0.06140855881788524, "percentage": 1.23, "elapsed_time": "0:05:09", "remaining_time": "6:54:11"}
{"current_steps": 50, "total_steps": 3255, "loss": 0.6165, "lr": 1.5337423312883436e-05, "epoch": 0.07676069852235655, "percentage": 1.54, "elapsed_time": "0:06:26", "remaining_time": "6:52:25"}
{"current_steps": 50, "total_steps": 3255, "eval_loss": 0.3982980251312256, "epoch": 0.07676069852235655, "percentage": 1.54, "elapsed_time": "0:09:03", "remaining_time": "9:40:32"}
{"current_steps": 60, "total_steps": 3255, "loss": 0.4066, "lr": 1.8404907975460123e-05, "epoch": 0.09211283822682786, "percentage": 1.84, "elapsed_time": "0:10:25", "remaining_time": "9:15:15"}
{"current_steps": 70, "total_steps": 3255, "loss": 0.276, "lr": 2.1472392638036813e-05, "epoch": 0.10746497793129918, "percentage": 2.15, "elapsed_time": "0:11:43", "remaining_time": "8:53:07"}
{"current_steps": 80, "total_steps": 3255, "loss": 0.1687, "lr": 2.4539877300613496e-05, "epoch": 0.12281711763577048, "percentage": 2.46, "elapsed_time": "0:12:59", "remaining_time": "8:35:54"}
{"current_steps": 90, "total_steps": 3255, "loss": 0.1319, "lr": 2.7607361963190186e-05, "epoch": 0.1381692573402418, "percentage": 2.76, "elapsed_time": "0:14:16", "remaining_time": "8:22:08"}
{"current_steps": 100, "total_steps": 3255, "loss": 0.1035, "lr": 3.067484662576687e-05, "epoch": 0.1535213970447131, "percentage": 3.07, "elapsed_time": "0:15:33", "remaining_time": "8:11:04"}
{"current_steps": 100, "total_steps": 3255, "eval_loss": 0.08296720683574677, "epoch": 0.1535213970447131, "percentage": 3.07, "elapsed_time": "0:18:11", "remaining_time": "9:33:48"}
{"current_steps": 110, "total_steps": 3255, "loss": 0.0736, "lr": 3.3742331288343556e-05, "epoch": 0.16887353674918443, "percentage": 3.38, "elapsed_time": "0:19:32", "remaining_time": "9:18:34"}
{"current_steps": 120, "total_steps": 3255, "loss": 0.0691, "lr": 3.6809815950920246e-05, "epoch": 0.18422567645365573, "percentage": 3.69, "elapsed_time": "0:20:49", "remaining_time": "9:03:54"}
{"current_steps": 130, "total_steps": 3255, "loss": 0.0809, "lr": 3.987730061349693e-05, "epoch": 0.19957781615812703, "percentage": 3.99, "elapsed_time": "0:22:06", "remaining_time": "8:51:20"}
{"current_steps": 140, "total_steps": 3255, "loss": 0.0557, "lr": 4.2944785276073626e-05, "epoch": 0.21492995586259836, "percentage": 4.3, "elapsed_time": "0:23:22", "remaining_time": "8:40:14"}
{"current_steps": 150, "total_steps": 3255, "loss": 0.0746, "lr": 4.601226993865031e-05, "epoch": 0.23028209556706966, "percentage": 4.61, "elapsed_time": "0:24:39", "remaining_time": "8:30:31"}
{"current_steps": 150, "total_steps": 3255, "eval_loss": 0.06817680597305298, "epoch": 0.23028209556706966, "percentage": 4.61, "elapsed_time": "0:27:17", "remaining_time": "9:24:46"}
{"current_steps": 160, "total_steps": 3255, "loss": 0.0854, "lr": 4.907975460122699e-05, "epoch": 0.24563423527154096, "percentage": 4.92, "elapsed_time": "0:28:38", "remaining_time": "9:14:06"}
{"current_steps": 170, "total_steps": 3255, "loss": 0.0763, "lr": 5.214723926380368e-05, "epoch": 0.2609863749760123, "percentage": 5.22, "elapsed_time": "0:29:55", "remaining_time": "9:03:07"}
{"current_steps": 180, "total_steps": 3255, "loss": 0.0714, "lr": 5.521472392638037e-05, "epoch": 0.2763385146804836, "percentage": 5.53, "elapsed_time": "0:31:12", "remaining_time": "8:53:09"}
{"current_steps": 190, "total_steps": 3255, "loss": 0.0591, "lr": 5.8282208588957056e-05, "epoch": 0.2916906543849549, "percentage": 5.84, "elapsed_time": "0:32:29", "remaining_time": "8:44:09"}
{"current_steps": 200, "total_steps": 3255, "loss": 0.0591, "lr": 6.134969325153375e-05, "epoch": 0.3070427940894262, "percentage": 6.14, "elapsed_time": "0:33:46", "remaining_time": "8:35:58"}
{"current_steps": 200, "total_steps": 3255, "eval_loss": 0.059281475841999054, "epoch": 0.3070427940894262, "percentage": 6.14, "elapsed_time": "0:36:24", "remaining_time": "9:16:01"}
{"current_steps": 210, "total_steps": 3255, "loss": 0.0519, "lr": 6.441717791411042e-05, "epoch": 0.3223949337938975, "percentage": 6.45, "elapsed_time": "0:37:45", "remaining_time": "9:07:26"}
{"current_steps": 220, "total_steps": 3255, "loss": 0.0436, "lr": 6.748466257668711e-05, "epoch": 0.33774707349836885, "percentage": 6.76, "elapsed_time": "0:39:02", "remaining_time": "8:58:30"}
{"current_steps": 230, "total_steps": 3255, "loss": 0.082, "lr": 7.055214723926382e-05, "epoch": 0.35309921320284016, "percentage": 7.07, "elapsed_time": "0:40:19", "remaining_time": "8:50:16"}
{"current_steps": 240, "total_steps": 3255, "loss": 0.0583, "lr": 7.361963190184049e-05, "epoch": 0.36845135290731146, "percentage": 7.37, "elapsed_time": "0:41:36", "remaining_time": "8:42:36"}
{"current_steps": 250, "total_steps": 3255, "loss": 0.0716, "lr": 7.668711656441718e-05, "epoch": 0.38380349261178276, "percentage": 7.68, "elapsed_time": "0:42:52", "remaining_time": "8:35:27"}
{"current_steps": 250, "total_steps": 3255, "eval_loss": 0.052977994084358215, "epoch": 0.38380349261178276, "percentage": 7.68, "elapsed_time": "0:45:30", "remaining_time": "9:06:58"}
{"current_steps": 260, "total_steps": 3255, "loss": 0.0559, "lr": 7.975460122699386e-05, "epoch": 0.39915563231625406, "percentage": 7.99, "elapsed_time": "0:46:51", "remaining_time": "8:59:50"}
{"current_steps": 270, "total_steps": 3255, "loss": 0.0611, "lr": 8.282208588957055e-05, "epoch": 0.41450777202072536, "percentage": 8.29, "elapsed_time": "0:48:09", "remaining_time": "8:52:19"}
{"current_steps": 280, "total_steps": 3255, "loss": 0.0614, "lr": 8.588957055214725e-05, "epoch": 0.4298599117251967, "percentage": 8.6, "elapsed_time": "0:49:25", "remaining_time": "8:45:13"}
{"current_steps": 290, "total_steps": 3255, "loss": 0.0472, "lr": 8.895705521472393e-05, "epoch": 0.445212051429668, "percentage": 8.91, "elapsed_time": "0:50:43", "remaining_time": "8:38:34"}
{"current_steps": 300, "total_steps": 3255, "loss": 0.0571, "lr": 9.202453987730062e-05, "epoch": 0.4605641911341393, "percentage": 9.22, "elapsed_time": "0:51:59", "remaining_time": "8:32:11"}
{"current_steps": 300, "total_steps": 3255, "eval_loss": 0.04749785736203194, "epoch": 0.4605641911341393, "percentage": 9.22, "elapsed_time": "0:54:37", "remaining_time": "8:58:01"}
{"current_steps": 310, "total_steps": 3255, "loss": 0.0457, "lr": 9.50920245398773e-05, "epoch": 0.4759163308386106, "percentage": 9.52, "elapsed_time": "0:55:59", "remaining_time": "8:51:51"}
{"current_steps": 320, "total_steps": 3255, "loss": 0.0464, "lr": 9.815950920245399e-05, "epoch": 0.4912684705430819, "percentage": 9.83, "elapsed_time": "0:57:15", "remaining_time": "8:45:13"}
{"current_steps": 330, "total_steps": 3255, "loss": 0.0532, "lr": 9.999953982785432e-05, "epoch": 0.5066206102475532, "percentage": 10.14, "elapsed_time": "0:58:32", "remaining_time": "8:38:53"}
{"current_steps": 340, "total_steps": 3255, "loss": 0.0476, "lr": 9.999436298849151e-05, "epoch": 0.5219727499520246, "percentage": 10.45, "elapsed_time": "0:59:49", "remaining_time": "8:32:56"}
{"current_steps": 350, "total_steps": 3255, "loss": 0.0499, "lr": 9.998343469212352e-05, "epoch": 0.5373248896564958, "percentage": 10.75, "elapsed_time": "1:01:06", "remaining_time": "8:27:09"}
{"current_steps": 350, "total_steps": 3255, "eval_loss": 0.0483204647898674, "epoch": 0.5373248896564958, "percentage": 10.75, "elapsed_time": "1:03:43", "remaining_time": "8:48:55"}
{"current_steps": 360, "total_steps": 3255, "loss": 0.0486, "lr": 9.996675619596465e-05, "epoch": 0.5526770293609672, "percentage": 11.06, "elapsed_time": "1:05:05", "remaining_time": "8:43:24"}
{"current_steps": 370, "total_steps": 3255, "loss": 0.0545, "lr": 9.99443294187443e-05, "epoch": 0.5680291690654385, "percentage": 11.37, "elapsed_time": "1:06:22", "remaining_time": "8:37:30"}
{"current_steps": 380, "total_steps": 3255, "loss": 0.0449, "lr": 9.991615694048621e-05, "epoch": 0.5833813087699098, "percentage": 11.67, "elapsed_time": "1:07:39", "remaining_time": "8:31:50"}
{"current_steps": 390, "total_steps": 3255, "loss": 0.0606, "lr": 9.988224200221172e-05, "epoch": 0.5987334484743811, "percentage": 11.98, "elapsed_time": "1:08:56", "remaining_time": "8:26:24"}
{"current_steps": 400, "total_steps": 3255, "loss": 0.0421, "lr": 9.984258850556693e-05, "epoch": 0.6140855881788524, "percentage": 12.29, "elapsed_time": "1:10:13", "remaining_time": "8:21:12"}
{"current_steps": 400, "total_steps": 3255, "eval_loss": 0.047019653022289276, "epoch": 0.6140855881788524, "percentage": 12.29, "elapsed_time": "1:12:50", "remaining_time": "8:39:55"}
{"current_steps": 410, "total_steps": 3255, "loss": 0.0445, "lr": 9.979720101237375e-05, "epoch": 0.6294377278833237, "percentage": 12.6, "elapsed_time": "1:14:11", "remaining_time": "8:34:51"}
{"current_steps": 420, "total_steps": 3255, "loss": 0.0492, "lr": 9.974608474410512e-05, "epoch": 0.644789867587795, "percentage": 12.9, "elapsed_time": "1:15:29", "remaining_time": "8:29:32"}
{"current_steps": 430, "total_steps": 3255, "loss": 0.0482, "lr": 9.968924558128445e-05, "epoch": 0.6601420072922664, "percentage": 13.21, "elapsed_time": "1:16:46", "remaining_time": "8:24:22"}
{"current_steps": 440, "total_steps": 3255, "loss": 0.0408, "lr": 9.962669006280894e-05, "epoch": 0.6754941469967377, "percentage": 13.52, "elapsed_time": "1:18:03", "remaining_time": "8:19:21"}
{"current_steps": 450, "total_steps": 3255, "loss": 0.0514, "lr": 9.95584253851974e-05, "epoch": 0.690846286701209, "percentage": 13.82, "elapsed_time": "1:19:19", "remaining_time": "8:14:30"}
{"current_steps": 450, "total_steps": 3255, "eval_loss": 0.047262296080589294, "epoch": 0.690846286701209, "percentage": 13.82, "elapsed_time": "1:21:57", "remaining_time": "8:30:50"}
{"current_steps": 460, "total_steps": 3255, "loss": 0.0515, "lr": 9.948445940176243e-05, "epoch": 0.7061984264056803, "percentage": 14.13, "elapsed_time": "1:23:22", "remaining_time": "8:26:35"}
{"current_steps": 470, "total_steps": 3255, "loss": 0.0435, "lr": 9.940480062170679e-05, "epoch": 0.7215505661101516, "percentage": 14.44, "elapsed_time": "1:24:39", "remaining_time": "8:21:37"}
{"current_steps": 480, "total_steps": 3255, "loss": 0.0544, "lr": 9.931945820914462e-05, "epoch": 0.7369027058146229, "percentage": 14.75, "elapsed_time": "1:25:56", "remaining_time": "8:16:49"}
{"current_steps": 490, "total_steps": 3255, "loss": 0.0407, "lr": 9.922844198204715e-05, "epoch": 0.7522548455190943, "percentage": 15.05, "elapsed_time": "1:27:13", "remaining_time": "8:12:10"}
{"current_steps": 500, "total_steps": 3255, "loss": 0.042, "lr": 9.913176241111319e-05, "epoch": 0.7676069852235655, "percentage": 15.36, "elapsed_time": "1:28:30", "remaining_time": "8:07:39"}
{"current_steps": 500, "total_steps": 3255, "eval_loss": 0.04632759094238281, "epoch": 0.7676069852235655, "percentage": 15.36, "elapsed_time": "1:31:07", "remaining_time": "8:22:06"}
{"current_steps": 510, "total_steps": 3255, "loss": 0.0374, "lr": 9.902943061856456e-05, "epoch": 0.7829591249280369, "percentage": 15.67, "elapsed_time": "1:32:29", "remaining_time": "8:17:50"}
{"current_steps": 520, "total_steps": 3255, "loss": 0.0651, "lr": 9.892145837686657e-05, "epoch": 0.7983112646325081, "percentage": 15.98, "elapsed_time": "1:33:46", "remaining_time": "8:13:13"}
{"current_steps": 530, "total_steps": 3255, "loss": 0.0585, "lr": 9.880785810737378e-05, "epoch": 0.8136634043369795, "percentage": 16.28, "elapsed_time": "1:35:03", "remaining_time": "8:08:45"}
{"current_steps": 540, "total_steps": 3255, "loss": 0.0438, "lr": 9.868864287890083e-05, "epoch": 0.8290155440414507, "percentage": 16.59, "elapsed_time": "1:36:20", "remaining_time": "8:04:22"}
{"current_steps": 550, "total_steps": 3255, "loss": 0.0672, "lr": 9.856382640621917e-05, "epoch": 0.8443676837459221, "percentage": 16.9, "elapsed_time": "1:37:37", "remaining_time": "8:00:09"}
{"current_steps": 550, "total_steps": 3255, "eval_loss": 0.04763461649417877, "epoch": 0.8443676837459221, "percentage": 16.9, "elapsed_time": "1:40:14", "remaining_time": "8:13:02"}
{"current_steps": 560, "total_steps": 3255, "loss": 0.0402, "lr": 9.84334230484792e-05, "epoch": 0.8597198234503934, "percentage": 17.2, "elapsed_time": "1:41:35", "remaining_time": "8:08:56"}
{"current_steps": 570, "total_steps": 3255, "loss": 0.0381, "lr": 9.82974478075583e-05, "epoch": 0.8750719631548647, "percentage": 17.51, "elapsed_time": "1:42:52", "remaining_time": "8:04:36"}
{"current_steps": 580, "total_steps": 3255, "loss": 0.0354, "lr": 9.815591632633509e-05, "epoch": 0.890424102859336, "percentage": 17.82, "elapsed_time": "1:44:08", "remaining_time": "8:00:20"}
{"current_steps": 590, "total_steps": 3255, "loss": 0.0443, "lr": 9.800884488688985e-05, "epoch": 0.9057762425638073, "percentage": 18.13, "elapsed_time": "1:45:25", "remaining_time": "7:56:13"}
{"current_steps": 600, "total_steps": 3255, "loss": 0.0388, "lr": 9.785625040863124e-05, "epoch": 0.9211283822682786, "percentage": 18.43, "elapsed_time": "1:46:42", "remaining_time": "7:52:10"}
{"current_steps": 600, "total_steps": 3255, "eval_loss": 0.04456296190619469, "epoch": 0.9211283822682786, "percentage": 18.43, "elapsed_time": "1:49:19", "remaining_time": "8:03:46"}
{"current_steps": 610, "total_steps": 3255, "loss": 0.0352, "lr": 9.769815044635005e-05, "epoch": 0.93648052197275, "percentage": 18.74, "elapsed_time": "1:50:41", "remaining_time": "7:59:58"}
{"current_steps": 620, "total_steps": 3255, "loss": 0.0485, "lr": 9.753456318819946e-05, "epoch": 0.9518326616772212, "percentage": 19.05, "elapsed_time": "1:51:58", "remaining_time": "7:55:54"}
{"current_steps": 630, "total_steps": 3255, "loss": 0.0413, "lr": 9.736550745360292e-05, "epoch": 0.9671848013816926, "percentage": 19.35, "elapsed_time": "1:53:15", "remaining_time": "7:51:54"}
{"current_steps": 640, "total_steps": 3255, "loss": 0.0399, "lr": 9.719100269108872e-05, "epoch": 0.9825369410861639, "percentage": 19.66, "elapsed_time": "1:54:32", "remaining_time": "7:48:00"}
{"current_steps": 650, "total_steps": 3255, "loss": 0.0391, "lr": 9.701106897605304e-05, "epoch": 0.9978890807906352, "percentage": 19.97, "elapsed_time": "1:55:49", "remaining_time": "7:44:10"}
{"current_steps": 650, "total_steps": 3255, "eval_loss": 0.04029757156968117, "epoch": 0.9978890807906352, "percentage": 19.97, "elapsed_time": "1:58:26", "remaining_time": "7:54:41"}
{"current_steps": 660, "total_steps": 3255, "loss": 0.032, "lr": 9.682572700845006e-05, "epoch": 1.0132412204951065, "percentage": 20.28, "elapsed_time": "1:59:48", "remaining_time": "7:51:02"}
{"current_steps": 670, "total_steps": 3255, "loss": 0.0361, "lr": 9.663499811041082e-05, "epoch": 1.0285933601995778, "percentage": 20.58, "elapsed_time": "2:01:05", "remaining_time": "7:47:10"}
{"current_steps": 680, "total_steps": 3255, "loss": 0.0341, "lr": 9.643890422379018e-05, "epoch": 1.0439454999040492, "percentage": 20.89, "elapsed_time": "2:02:21", "remaining_time": "7:43:22"}
{"current_steps": 690, "total_steps": 3255, "loss": 0.0323, "lr": 9.623746790764261e-05, "epoch": 1.0592976396085205, "percentage": 21.2, "elapsed_time": "2:03:38", "remaining_time": "7:39:38"}
{"current_steps": 700, "total_steps": 3255, "loss": 0.0304, "lr": 9.603071233562695e-05, "epoch": 1.0746497793129917, "percentage": 21.51, "elapsed_time": "2:04:55", "remaining_time": "7:35:59"}
{"current_steps": 700, "total_steps": 3255, "eval_loss": 0.042355190962553024, "epoch": 1.0746497793129917, "percentage": 21.51, "elapsed_time": "2:07:32", "remaining_time": "7:45:33"}
{"current_steps": 710, "total_steps": 3255, "loss": 0.0416, "lr": 9.581866129334044e-05, "epoch": 1.090001919017463, "percentage": 21.81, "elapsed_time": "2:08:54", "remaining_time": "7:42:05"}
{"current_steps": 720, "total_steps": 3255, "loss": 0.048, "lr": 9.560133917558242e-05, "epoch": 1.1053540587219344, "percentage": 22.12, "elapsed_time": "2:10:12", "remaining_time": "7:38:25"}
{"current_steps": 730, "total_steps": 3255, "loss": 0.0355, "lr": 9.537877098354786e-05, "epoch": 1.1207061984264057, "percentage": 22.43, "elapsed_time": "2:11:29", "remaining_time": "7:34:48"}
{"current_steps": 740, "total_steps": 3255, "loss": 0.0301, "lr": 9.51509823219512e-05, "epoch": 1.136058338130877, "percentage": 22.73, "elapsed_time": "2:12:46", "remaining_time": "7:31:14"}
{"current_steps": 750, "total_steps": 3255, "loss": 0.0396, "lr": 9.491799939608065e-05, "epoch": 1.1514104778353482, "percentage": 23.04, "elapsed_time": "2:14:03", "remaining_time": "7:27:43"}
{"current_steps": 750, "total_steps": 3255, "eval_loss": 0.042958181351423264, "epoch": 1.1514104778353482, "percentage": 23.04, "elapsed_time": "2:16:40", "remaining_time": "7:36:28"}
{"current_steps": 760, "total_steps": 3255, "loss": 0.0336, "lr": 9.467984900878364e-05, "epoch": 1.1667626175398196, "percentage": 23.35, "elapsed_time": "2:18:02", "remaining_time": "7:33:09"}
{"current_steps": 770, "total_steps": 3255, "loss": 0.029, "lr": 9.443655855738321e-05, "epoch": 1.182114757244291, "percentage": 23.66, "elapsed_time": "2:19:18", "remaining_time": "7:29:36"}
{"current_steps": 780, "total_steps": 3255, "loss": 0.0335, "lr": 9.41881560305262e-05, "epoch": 1.1974668969487623, "percentage": 23.96, "elapsed_time": "2:20:35", "remaining_time": "7:26:07"}
{"current_steps": 790, "total_steps": 3255, "loss": 0.0409, "lr": 9.393467000496344e-05, "epoch": 1.2128190366532334, "percentage": 24.27, "elapsed_time": "2:21:52", "remaining_time": "7:22:42"}
{"current_steps": 800, "total_steps": 3255, "loss": 0.0347, "lr": 9.367612964226218e-05, "epoch": 1.2281711763577048, "percentage": 24.58, "elapsed_time": "2:23:09", "remaining_time": "7:19:19"}
{"current_steps": 800, "total_steps": 3255, "eval_loss": 0.039533697068691254, "epoch": 1.2281711763577048, "percentage": 24.58, "elapsed_time": "2:25:46", "remaining_time": "7:27:22"}
{"current_steps": 810, "total_steps": 3255, "loss": 0.0246, "lr": 9.341256468545122e-05, "epoch": 1.2435233160621761, "percentage": 24.88, "elapsed_time": "2:27:08", "remaining_time": "7:24:09"}
{"current_steps": 820, "total_steps": 3255, "loss": 0.0287, "lr": 9.314400545559934e-05, "epoch": 1.2588754557666475, "percentage": 25.19, "elapsed_time": "2:28:25", "remaining_time": "7:20:46"}
{"current_steps": 830, "total_steps": 3255, "loss": 0.0358, "lr": 9.287048284832698e-05, "epoch": 1.2742275954711189, "percentage": 25.5, "elapsed_time": "2:29:42", "remaining_time": "7:17:25"}
{"current_steps": 840, "total_steps": 3255, "loss": 0.0339, "lr": 9.2592028330252e-05, "epoch": 1.28957973517559, "percentage": 25.81, "elapsed_time": "2:31:00", "remaining_time": "7:14:08"}
{"current_steps": 850, "total_steps": 3255, "loss": 0.0356, "lr": 9.230867393536972e-05, "epoch": 1.3049318748800613, "percentage": 26.11, "elapsed_time": "2:32:17", "remaining_time": "7:10:52"}
{"current_steps": 850, "total_steps": 3255, "eval_loss": 0.04097624123096466, "epoch": 1.3049318748800613, "percentage": 26.11, "elapsed_time": "2:34:54", "remaining_time": "7:18:18"}
{"current_steps": 860, "total_steps": 3255, "loss": 0.0315, "lr": 9.202045226136757e-05, "epoch": 1.3202840145845327, "percentage": 26.42, "elapsed_time": "2:36:20", "remaining_time": "7:15:22"}
{"current_steps": 870, "total_steps": 3255, "loss": 0.0242, "lr": 9.172739646587509e-05, "epoch": 1.335636154289004, "percentage": 26.73, "elapsed_time": "2:37:37", "remaining_time": "7:12:06"}
{"current_steps": 880, "total_steps": 3255, "loss": 0.0313, "lr": 9.142954026264931e-05, "epoch": 1.3509882939934754, "percentage": 27.04, "elapsed_time": "2:38:53", "remaining_time": "7:08:50"}
{"current_steps": 890, "total_steps": 3255, "loss": 0.0296, "lr": 9.112691791769634e-05, "epoch": 1.3663404336979466, "percentage": 27.34, "elapsed_time": "2:40:10", "remaining_time": "7:05:38"}
{"current_steps": 900, "total_steps": 3255, "loss": 0.0333, "lr": 9.081956424532926e-05, "epoch": 1.381692573402418, "percentage": 27.65, "elapsed_time": "2:41:27", "remaining_time": "7:02:29"}
{"current_steps": 900, "total_steps": 3255, "eval_loss": 0.03842702880501747, "epoch": 1.381692573402418, "percentage": 27.65, "elapsed_time": "2:44:04", "remaining_time": "7:09:20"}
{"current_steps": 910, "total_steps": 3255, "loss": 0.0334, "lr": 9.050751460416305e-05, "epoch": 1.3970447131068893, "percentage": 27.96, "elapsed_time": "2:45:26", "remaining_time": "7:06:19"}
{"current_steps": 920, "total_steps": 3255, "loss": 0.0367, "lr": 9.019080489304685e-05, "epoch": 1.4123968528113606, "percentage": 28.26, "elapsed_time": "2:46:43", "remaining_time": "7:03:09"}
{"current_steps": 930, "total_steps": 3255, "loss": 0.0487, "lr": 8.986947154693408e-05, "epoch": 1.427748992515832, "percentage": 28.57, "elapsed_time": "2:48:00", "remaining_time": "7:00:02"}
{"current_steps": 940, "total_steps": 3255, "loss": 0.0427, "lr": 8.954355153269088e-05, "epoch": 1.4431011322203031, "percentage": 28.88, "elapsed_time": "2:49:18", "remaining_time": "6:56:56"}
{"current_steps": 950, "total_steps": 3255, "loss": 0.0318, "lr": 8.921308234484336e-05, "epoch": 1.4584532719247745, "percentage": 29.19, "elapsed_time": "2:50:34", "remaining_time": "6:53:52"}
{"current_steps": 950, "total_steps": 3255, "eval_loss": 0.037589143961668015, "epoch": 1.4584532719247745, "percentage": 29.19, "elapsed_time": "2:53:11", "remaining_time": "7:00:14"}
{"current_steps": 960, "total_steps": 3255, "loss": 0.0352, "lr": 8.887810200126419e-05, "epoch": 1.4738054116292458, "percentage": 29.49, "elapsed_time": "2:54:33", "remaining_time": "6:57:18"}
{"current_steps": 970, "total_steps": 3255, "loss": 0.0383, "lr": 8.853864903879889e-05, "epoch": 1.4891575513337172, "percentage": 29.8, "elapsed_time": "2:55:50", "remaining_time": "6:54:14"}
{"current_steps": 980, "total_steps": 3255, "loss": 0.0345, "lr": 8.81947625088325e-05, "epoch": 1.5045096910381885, "percentage": 30.11, "elapsed_time": "2:57:07", "remaining_time": "6:51:11"}
{"current_steps": 990, "total_steps": 3255, "loss": 0.032, "lr": 8.784648197279701e-05, "epoch": 1.5198618307426597, "percentage": 30.41, "elapsed_time": "2:58:24", "remaining_time": "6:48:10"}
{"current_steps": 1000, "total_steps": 3255, "loss": 0.0309, "lr": 8.749384749762015e-05, "epoch": 1.535213970447131, "percentage": 30.72, "elapsed_time": "2:59:41", "remaining_time": "6:45:11"}
{"current_steps": 1000, "total_steps": 3255, "eval_loss": 0.03832419961690903, "epoch": 1.535213970447131, "percentage": 30.72, "elapsed_time": "3:02:18", "remaining_time": "6:51:06"}
{"current_steps": 1010, "total_steps": 3255, "loss": 0.0246, "lr": 8.713689965111602e-05, "epoch": 1.5505661101516024, "percentage": 31.03, "elapsed_time": "3:03:40", "remaining_time": "6:48:16"}
{"current_steps": 1020, "total_steps": 3255, "loss": 0.0372, "lr": 8.677567949731801e-05, "epoch": 1.5659182498560735, "percentage": 31.34, "elapsed_time": "3:04:57", "remaining_time": "6:45:15"}
{"current_steps": 1030, "total_steps": 3255, "loss": 0.0502, "lr": 8.64102285917548e-05, "epoch": 1.581270389560545, "percentage": 31.64, "elapsed_time": "3:06:13", "remaining_time": "6:42:17"}
{"current_steps": 1040, "total_steps": 3255, "loss": 0.0333, "lr": 8.604058897666962e-05, "epoch": 1.5966225292650162, "percentage": 31.95, "elapsed_time": "3:07:30", "remaining_time": "6:39:22"}
{"current_steps": 1050, "total_steps": 3255, "loss": 0.0216, "lr": 8.566680317618377e-05, "epoch": 1.6119746689694876, "percentage": 32.26, "elapsed_time": "3:08:47", "remaining_time": "6:36:27"}
{"current_steps": 1050, "total_steps": 3255, "eval_loss": 0.037316225469112396, "epoch": 1.6119746689694876, "percentage": 32.26, "elapsed_time": "3:11:24", "remaining_time": "6:41:57"}
{"current_steps": 1060, "total_steps": 3255, "loss": 0.0387, "lr": 8.528891419140438e-05, "epoch": 1.627326808673959, "percentage": 32.57, "elapsed_time": "3:12:46", "remaining_time": "6:39:11"}
{"current_steps": 1070, "total_steps": 3255, "loss": 0.0272, "lr": 8.490696549547761e-05, "epoch": 1.64267894837843, "percentage": 32.87, "elapsed_time": "3:14:03", "remaining_time": "6:36:16"}
{"current_steps": 1080, "total_steps": 3255, "loss": 0.0401, "lr": 8.452100102858734e-05, "epoch": 1.6580310880829017, "percentage": 33.18, "elapsed_time": "3:15:19", "remaining_time": "6:33:22"}
{"current_steps": 1090, "total_steps": 3255, "loss": 0.0322, "lr": 8.413106519290023e-05, "epoch": 1.6733832277873728, "percentage": 33.49, "elapsed_time": "3:16:36", "remaining_time": "6:30:31"}
{"current_steps": 1100, "total_steps": 3255, "loss": 0.0315, "lr": 8.373720284745757e-05, "epoch": 1.6887353674918442, "percentage": 33.79, "elapsed_time": "3:17:53", "remaining_time": "6:27:41"}
{"current_steps": 1100, "total_steps": 3255, "eval_loss": 0.03704118728637695, "epoch": 1.6887353674918442, "percentage": 33.79, "elapsed_time": "3:20:30", "remaining_time": "6:32:49"}
{"current_steps": 1110, "total_steps": 3255, "loss": 0.0312, "lr": 8.333945930301459e-05, "epoch": 1.7040875071963155, "percentage": 34.1, "elapsed_time": "3:21:53", "remaining_time": "6:30:08"}
{"current_steps": 1120, "total_steps": 3255, "loss": 0.0305, "lr": 8.293788031682789e-05, "epoch": 1.7194396469007867, "percentage": 34.41, "elapsed_time": "3:23:09", "remaining_time": "6:27:17"}
{"current_steps": 1130, "total_steps": 3255, "loss": 0.0303, "lr": 8.253251208739137e-05, "epoch": 1.7347917866052582, "percentage": 34.72, "elapsed_time": "3:24:27", "remaining_time": "6:24:28"}
{"current_steps": 1140, "total_steps": 3255, "loss": 0.0292, "lr": 8.21234012491215e-05, "epoch": 1.7501439263097294, "percentage": 35.02, "elapsed_time": "3:25:44", "remaining_time": "6:21:41"}
{"current_steps": 1150, "total_steps": 3255, "loss": 0.0273, "lr": 8.171059486699224e-05, "epoch": 1.7654960660142007, "percentage": 35.33, "elapsed_time": "3:27:01", "remaining_time": "6:18:56"}
{"current_steps": 1150, "total_steps": 3255, "eval_loss": 0.03893882781267166, "epoch": 1.7654960660142007, "percentage": 35.33, "elapsed_time": "3:29:38", "remaining_time": "6:23:44"}
{"current_steps": 1160, "total_steps": 3255, "loss": 0.035, "lr": 8.129414043112087e-05, "epoch": 1.780848205718672, "percentage": 35.64, "elapsed_time": "3:31:01", "remaining_time": "6:21:07"}
{"current_steps": 1170, "total_steps": 3255, "loss": 0.0283, "lr": 8.087408585130438e-05, "epoch": 1.7962003454231432, "percentage": 35.94, "elapsed_time": "3:32:18", "remaining_time": "6:18:21"}
{"current_steps": 1180, "total_steps": 3255, "loss": 0.0348, "lr": 8.045047945150796e-05, "epoch": 1.8115524851276148, "percentage": 36.25, "elapsed_time": "3:33:35", "remaining_time": "6:15:36"}
{"current_steps": 1190, "total_steps": 3255, "loss": 0.0309, "lr": 8.002336996430561e-05, "epoch": 1.826904624832086, "percentage": 36.56, "elapsed_time": "3:34:52", "remaining_time": "6:12:52"}
{"current_steps": 1200, "total_steps": 3255, "loss": 0.0188, "lr": 7.959280652527394e-05, "epoch": 1.8422567645365573, "percentage": 36.87, "elapsed_time": "3:36:09", "remaining_time": "6:10:10"}
{"current_steps": 1200, "total_steps": 3255, "eval_loss": 0.03792406991124153, "epoch": 1.8422567645365573, "percentage": 36.87, "elapsed_time": "3:38:46", "remaining_time": "6:14:39"}
{"current_steps": 1210, "total_steps": 3255, "loss": 0.0348, "lr": 7.915883866733946e-05, "epoch": 1.8576089042410286, "percentage": 37.17, "elapsed_time": "3:40:08", "remaining_time": "6:12:04"}
{"current_steps": 1220, "total_steps": 3255, "loss": 0.0311, "lr": 7.872151631508022e-05, "epoch": 1.8729610439454998, "percentage": 37.48, "elapsed_time": "3:41:25", "remaining_time": "6:09:20"}
{"current_steps": 1230, "total_steps": 3255, "loss": 0.0361, "lr": 7.828088977898234e-05, "epoch": 1.8883131836499714, "percentage": 37.79, "elapsed_time": "3:42:42", "remaining_time": "6:06:38"}
{"current_steps": 1240, "total_steps": 3255, "loss": 0.043, "lr": 7.783700974965225e-05, "epoch": 1.9036653233544425, "percentage": 38.1, "elapsed_time": "3:43:59", "remaining_time": "6:03:59"}
{"current_steps": 1250, "total_steps": 3255, "loss": 0.0362, "lr": 7.738992729198511e-05, "epoch": 1.9190174630589139, "percentage": 38.4, "elapsed_time": "3:45:16", "remaining_time": "6:01:20"}
{"current_steps": 1250, "total_steps": 3255, "eval_loss": 0.037059471011161804, "epoch": 1.9190174630589139, "percentage": 38.4, "elapsed_time": "3:47:53", "remaining_time": "6:05:32"}
{"current_steps": 1260, "total_steps": 3255, "loss": 0.0247, "lr": 7.693969383929017e-05, "epoch": 1.9343696027633852, "percentage": 38.71, "elapsed_time": "3:49:15", "remaining_time": "6:03:00"}
{"current_steps": 1270, "total_steps": 3255, "loss": 0.0375, "lr": 7.648636118737385e-05, "epoch": 1.9497217424678563, "percentage": 39.02, "elapsed_time": "3:50:33", "remaining_time": "6:00:21"}
{"current_steps": 1280, "total_steps": 3255, "loss": 0.0302, "lr": 7.602998148858089e-05, "epoch": 1.965073882172328, "percentage": 39.32, "elapsed_time": "3:51:49", "remaining_time": "5:57:42"}
{"current_steps": 1290, "total_steps": 3255, "loss": 0.0325, "lr": 7.557060724579484e-05, "epoch": 1.980426021876799, "percentage": 39.63, "elapsed_time": "3:53:06", "remaining_time": "5:55:05"}
{"current_steps": 1300, "total_steps": 3255, "loss": 0.0435, "lr": 7.51082913063978e-05, "epoch": 1.9957781615812704, "percentage": 39.94, "elapsed_time": "3:54:23", "remaining_time": "5:52:29"}
{"current_steps": 1300, "total_steps": 3255, "eval_loss": 0.036797065287828445, "epoch": 1.9957781615812704, "percentage": 39.94, "elapsed_time": "3:57:00", "remaining_time": "5:56:26"}
{"current_steps": 1310, "total_steps": 3255, "loss": 0.0272, "lr": 7.464308685619099e-05, "epoch": 2.0111303012857418, "percentage": 40.25, "elapsed_time": "3:58:21", "remaining_time": "5:53:54"}
{"current_steps": 1320, "total_steps": 3255, "loss": 0.0168, "lr": 7.417504741327587e-05, "epoch": 2.026482440990213, "percentage": 40.55, "elapsed_time": "3:59:38", "remaining_time": "5:51:17"}
{"current_steps": 1330, "total_steps": 3255, "loss": 0.0209, "lr": 7.370422682189755e-05, "epoch": 2.0418345806946845, "percentage": 40.86, "elapsed_time": "4:00:55", "remaining_time": "5:48:42"}
{"current_steps": 1340, "total_steps": 3255, "loss": 0.0172, "lr": 7.323067924625024e-05, "epoch": 2.0571867203991556, "percentage": 41.17, "elapsed_time": "4:02:12", "remaining_time": "5:46:08"}
{"current_steps": 1350, "total_steps": 3255, "loss": 0.0212, "lr": 7.275445916424627e-05, "epoch": 2.0725388601036268, "percentage": 41.47, "elapsed_time": "4:03:29", "remaining_time": "5:43:35"}
{"current_steps": 1350, "total_steps": 3255, "eval_loss": 0.03962073475122452, "epoch": 2.0725388601036268, "percentage": 41.47, "elapsed_time": "4:06:06", "remaining_time": "5:47:17"}
{"current_steps": 1360, "total_steps": 3255, "loss": 0.0179, "lr": 7.227562136124864e-05, "epoch": 2.0878909998080983, "percentage": 41.78, "elapsed_time": "4:07:28", "remaining_time": "5:44:49"}
{"current_steps": 1370, "total_steps": 3255, "loss": 0.02, "lr": 7.179422092376856e-05, "epoch": 2.1032431395125695, "percentage": 42.09, "elapsed_time": "4:08:44", "remaining_time": "5:42:15"}
{"current_steps": 1380, "total_steps": 3255, "loss": 0.0223, "lr": 7.13103132331281e-05, "epoch": 2.118595279217041, "percentage": 42.4, "elapsed_time": "4:10:01", "remaining_time": "5:39:42"}
{"current_steps": 1390, "total_steps": 3255, "loss": 0.0249, "lr": 7.082395395908903e-05, "epoch": 2.133947418921512, "percentage": 42.7, "elapsed_time": "4:11:18", "remaining_time": "5:37:11"}
{"current_steps": 1400, "total_steps": 3255, "loss": 0.0184, "lr": 7.033519905344846e-05, "epoch": 2.1492995586259833, "percentage": 43.01, "elapsed_time": "4:12:35", "remaining_time": "5:34:41"}
{"current_steps": 1400, "total_steps": 3255, "eval_loss": 0.03931669518351555, "epoch": 2.1492995586259833, "percentage": 43.01, "elapsed_time": "4:15:12", "remaining_time": "5:38:09"}
{"current_steps": 1410, "total_steps": 3255, "loss": 0.0243, "lr": 6.984410474360195e-05, "epoch": 2.164651698330455, "percentage": 43.32, "elapsed_time": "4:16:33", "remaining_time": "5:35:43"}
{"current_steps": 1420, "total_steps": 3255, "loss": 0.0271, "lr": 6.935072752607511e-05, "epoch": 2.180003838034926, "percentage": 43.63, "elapsed_time": "4:17:50", "remaining_time": "5:33:12"}
{"current_steps": 1430, "total_steps": 3255, "loss": 0.0308, "lr": 6.885512416002412e-05, "epoch": 2.1953559777393976, "percentage": 43.93, "elapsed_time": "4:19:07", "remaining_time": "5:30:42"}
{"current_steps": 1440, "total_steps": 3255, "loss": 0.0205, "lr": 6.835735166070587e-05, "epoch": 2.2107081174438687, "percentage": 44.24, "elapsed_time": "4:20:24", "remaining_time": "5:28:13"}
{"current_steps": 1450, "total_steps": 3255, "loss": 0.0205, "lr": 6.785746729291897e-05, "epoch": 2.22606025714834, "percentage": 44.55, "elapsed_time": "4:21:41", "remaining_time": "5:25:45"}
{"current_steps": 1450, "total_steps": 3255, "eval_loss": 0.03809254616498947, "epoch": 2.22606025714834, "percentage": 44.55, "elapsed_time": "4:24:18", "remaining_time": "5:29:01"}
{"current_steps": 1460, "total_steps": 3255, "loss": 0.0232, "lr": 6.735552856441585e-05, "epoch": 2.2414123968528115, "percentage": 44.85, "elapsed_time": "4:25:40", "remaining_time": "5:26:37"}
{"current_steps": 1470, "total_steps": 3255, "loss": 0.0216, "lr": 6.685159321928691e-05, "epoch": 2.2567645365572826, "percentage": 45.16, "elapsed_time": "4:26:57", "remaining_time": "5:24:09"}
{"current_steps": 1480, "total_steps": 3255, "loss": 0.0239, "lr": 6.634571923131756e-05, "epoch": 2.272116676261754, "percentage": 45.47, "elapsed_time": "4:28:14", "remaining_time": "5:21:41"}
{"current_steps": 1490, "total_steps": 3255, "loss": 0.0175, "lr": 6.583796479731872e-05, "epoch": 2.2874688159662253, "percentage": 45.78, "elapsed_time": "4:29:30", "remaining_time": "5:19:15"}
{"current_steps": 1500, "total_steps": 3255, "loss": 0.0316, "lr": 6.532838833043189e-05, "epoch": 2.3028209556706964, "percentage": 46.08, "elapsed_time": "4:30:47", "remaining_time": "5:16:50"}
{"current_steps": 1500, "total_steps": 3255, "eval_loss": 0.03903723508119583, "epoch": 2.3028209556706964, "percentage": 46.08, "elapsed_time": "4:33:25", "remaining_time": "5:19:53"}
{"current_steps": 1510, "total_steps": 3255, "loss": 0.025, "lr": 6.481704845340894e-05, "epoch": 2.318173095375168, "percentage": 46.39, "elapsed_time": "4:34:47", "remaining_time": "5:17:33"}
{"current_steps": 1520, "total_steps": 3255, "loss": 0.0219, "lr": 6.43040039918683e-05, "epoch": 2.333525235079639, "percentage": 46.7, "elapsed_time": "4:36:03", "remaining_time": "5:15:06"}
{"current_steps": 1530, "total_steps": 3255, "loss": 0.019, "lr": 6.37893139675273e-05, "epoch": 2.3488773747841103, "percentage": 47.0, "elapsed_time": "4:37:20", "remaining_time": "5:12:41"}
{"current_steps": 1540, "total_steps": 3255, "loss": 0.0188, "lr": 6.327303759141235e-05, "epoch": 2.364229514488582, "percentage": 47.31, "elapsed_time": "4:38:37", "remaining_time": "5:10:17"}
{"current_steps": 1550, "total_steps": 3255, "loss": 0.0234, "lr": 6.275523425704707e-05, "epoch": 2.379581654193053, "percentage": 47.62, "elapsed_time": "4:39:54", "remaining_time": "5:07:54"}
{"current_steps": 1550, "total_steps": 3255, "eval_loss": 0.03996282070875168, "epoch": 2.379581654193053, "percentage": 47.62, "elapsed_time": "4:42:31", "remaining_time": "5:10:46"}
{"current_steps": 1560, "total_steps": 3255, "loss": 0.0226, "lr": 6.223596353361961e-05, "epoch": 2.3949337938975246, "percentage": 47.93, "elapsed_time": "4:43:53", "remaining_time": "5:08:27"}
{"current_steps": 1570, "total_steps": 3255, "loss": 0.0162, "lr": 6.171528515912965e-05, "epoch": 2.4102859336019957, "percentage": 48.23, "elapsed_time": "4:45:10", "remaining_time": "5:06:03"}
{"current_steps": 1580, "total_steps": 3255, "loss": 0.0253, "lr": 6.119325903351599e-05, "epoch": 2.425638073306467, "percentage": 48.54, "elapsed_time": "4:46:27", "remaining_time": "5:03:40"}
{"current_steps": 1590, "total_steps": 3255, "loss": 0.0312, "lr": 6.0669945211765585e-05, "epoch": 2.4409902130109384, "percentage": 48.85, "elapsed_time": "4:47:44", "remaining_time": "5:01:19"}
{"current_steps": 1600, "total_steps": 3255, "loss": 0.0258, "lr": 6.0145403897004696e-05, "epoch": 2.4563423527154096, "percentage": 49.16, "elapsed_time": "4:49:02", "remaining_time": "4:58:58"}
{"current_steps": 1600, "total_steps": 3255, "eval_loss": 0.038748133927583694, "epoch": 2.4563423527154096, "percentage": 49.16, "elapsed_time": "4:51:39", "remaining_time": "5:01:40"}
{"current_steps": 1610, "total_steps": 3255, "loss": 0.0247, "lr": 5.961969543357292e-05, "epoch": 2.471694492419881, "percentage": 49.46, "elapsed_time": "4:53:01", "remaining_time": "4:59:24"}
{"current_steps": 1620, "total_steps": 3255, "loss": 0.0277, "lr": 5.9092880300081123e-05, "epoch": 2.4870466321243523, "percentage": 49.77, "elapsed_time": "4:54:18", "remaining_time": "4:57:02"}
{"current_steps": 1630, "total_steps": 3255, "loss": 0.0237, "lr": 5.8565019102453844e-05, "epoch": 2.5023987718288234, "percentage": 50.08, "elapsed_time": "4:55:35", "remaining_time": "4:54:41"}
{"current_steps": 1640, "total_steps": 3255, "loss": 0.0231, "lr": 5.8036172566957006e-05, "epoch": 2.517750911533295, "percentage": 50.38, "elapsed_time": "4:56:52", "remaining_time": "4:52:21"}
{"current_steps": 1650, "total_steps": 3255, "loss": 0.0273, "lr": 5.750640153321194e-05, "epoch": 2.533103051237766, "percentage": 50.69, "elapsed_time": "4:58:09", "remaining_time": "4:50:01"}
{"current_steps": 1650, "total_steps": 3255, "eval_loss": 0.0392879992723465, "epoch": 2.533103051237766, "percentage": 50.69, "elapsed_time": "5:00:46", "remaining_time": "4:52:34"}
{"current_steps": 1660, "total_steps": 3255, "loss": 0.0215, "lr": 5.697576694719616e-05, "epoch": 2.5484551909422377, "percentage": 51.0, "elapsed_time": "5:02:08", "remaining_time": "4:50:18"}
{"current_steps": 1670, "total_steps": 3255, "loss": 0.0229, "lr": 5.644432985423206e-05, "epoch": 2.563807330646709, "percentage": 51.31, "elapsed_time": "5:03:25", "remaining_time": "4:47:58"}
{"current_steps": 1680, "total_steps": 3255, "loss": 0.0242, "lr": 5.591215139196414e-05, "epoch": 2.57915947035118, "percentage": 51.61, "elapsed_time": "5:04:42", "remaining_time": "4:45:39"}
{"current_steps": 1690, "total_steps": 3255, "loss": 0.0286, "lr": 5.5379292783325585e-05, "epoch": 2.5945116100556516, "percentage": 51.92, "elapsed_time": "5:05:59", "remaining_time": "4:43:21"}
{"current_steps": 1700, "total_steps": 3255, "loss": 0.0199, "lr": 5.4845815329495054e-05, "epoch": 2.6098637497601227, "percentage": 52.23, "elapsed_time": "5:07:16", "remaining_time": "4:41:03"}
{"current_steps": 1700, "total_steps": 3255, "eval_loss": 0.03853138908743858, "epoch": 2.6098637497601227, "percentage": 52.23, "elapsed_time": "5:09:53", "remaining_time": "4:43:27"}
{"current_steps": 1710, "total_steps": 3255, "loss": 0.0223, "lr": 5.431178040284446e-05, "epoch": 2.6252158894645943, "percentage": 52.53, "elapsed_time": "5:11:15", "remaining_time": "4:41:13"}
{"current_steps": 1720, "total_steps": 3255, "loss": 0.0181, "lr": 5.377724943987855e-05, "epoch": 2.6405680291690654, "percentage": 52.84, "elapsed_time": "5:12:32", "remaining_time": "4:38:55"}
{"current_steps": 1730, "total_steps": 3255, "loss": 0.0181, "lr": 5.324228393416718e-05, "epoch": 2.6559201688735365, "percentage": 53.15, "elapsed_time": "5:13:49", "remaining_time": "4:36:38"}
{"current_steps": 1740, "total_steps": 3255, "loss": 0.0257, "lr": 5.270694542927088e-05, "epoch": 2.671272308578008, "percentage": 53.46, "elapsed_time": "5:15:06", "remaining_time": "4:34:21"}
{"current_steps": 1750, "total_steps": 3255, "loss": 0.0167, "lr": 5.21712955116608e-05, "epoch": 2.6866244482824793, "percentage": 53.76, "elapsed_time": "5:16:23", "remaining_time": "4:32:05"}
{"current_steps": 1750, "total_steps": 3255, "eval_loss": 0.03936908766627312, "epoch": 2.6866244482824793, "percentage": 53.76, "elapsed_time": "5:19:00", "remaining_time": "4:34:20"}
{"current_steps": 1760, "total_steps": 3255, "loss": 0.0197, "lr": 5.1635395803633666e-05, "epoch": 2.701976587986951, "percentage": 54.07, "elapsed_time": "5:20:22", "remaining_time": "4:32:08"}
{"current_steps": 1770, "total_steps": 3255, "loss": 0.0153, "lr": 5.109930795622265e-05, "epoch": 2.717328727691422, "percentage": 54.38, "elapsed_time": "5:21:39", "remaining_time": "4:29:52"}
{"current_steps": 1780, "total_steps": 3255, "loss": 0.0191, "lr": 5.056309364210483e-05, "epoch": 2.732680867395893, "percentage": 54.69, "elapsed_time": "5:22:56", "remaining_time": "4:27:36"}
{"current_steps": 1790, "total_steps": 3255, "loss": 0.024, "lr": 5.002681454850632e-05, "epoch": 2.7480330071003647, "percentage": 54.99, "elapsed_time": "5:24:13", "remaining_time": "4:25:21"}
{"current_steps": 1800, "total_steps": 3255, "loss": 0.0288, "lr": 4.949053237010554e-05, "epoch": 2.763385146804836, "percentage": 55.3, "elapsed_time": "5:25:30", "remaining_time": "4:23:06"}
{"current_steps": 1800, "total_steps": 3255, "eval_loss": 0.042679354548454285, "epoch": 2.763385146804836, "percentage": 55.3, "elapsed_time": "5:28:07", "remaining_time": "4:25:14"}
{"current_steps": 1810, "total_steps": 3255, "loss": 0.0172, "lr": 4.89543088019359e-05, "epoch": 2.7787372865093074, "percentage": 55.61, "elapsed_time": "5:29:29", "remaining_time": "4:23:02"}
{"current_steps": 1820, "total_steps": 3255, "loss": 0.0284, "lr": 4.841820553228805e-05, "epoch": 2.7940894262137785, "percentage": 55.91, "elapsed_time": "5:30:46", "remaining_time": "4:20:48"}
{"current_steps": 1830, "total_steps": 3255, "loss": 0.0281, "lr": 4.7882284235613324e-05, "epoch": 2.8094415659182497, "percentage": 56.22, "elapsed_time": "5:32:03", "remaining_time": "4:18:34"}
{"current_steps": 1840, "total_steps": 3255, "loss": 0.0231, "lr": 4.734660656542846e-05, "epoch": 2.8247937056227213, "percentage": 56.53, "elapsed_time": "5:33:20", "remaining_time": "4:16:20"}
{"current_steps": 1850, "total_steps": 3255, "loss": 0.022, "lr": 4.681123414722291e-05, "epoch": 2.8401458453271924, "percentage": 56.84, "elapsed_time": "5:34:37", "remaining_time": "4:14:07"}
{"current_steps": 1850, "total_steps": 3255, "eval_loss": 0.03755514323711395, "epoch": 2.8401458453271924, "percentage": 56.84, "elapsed_time": "5:37:14", "remaining_time": "4:16:07"}
{"current_steps": 1860, "total_steps": 3255, "loss": 0.0193, "lr": 4.627622857136929e-05, "epoch": 2.855497985031664, "percentage": 57.14, "elapsed_time": "5:38:36", "remaining_time": "4:13:57"}
{"current_steps": 1870, "total_steps": 3255, "loss": 0.0298, "lr": 4.5741651386037883e-05, "epoch": 2.870850124736135, "percentage": 57.45, "elapsed_time": "5:39:53", "remaining_time": "4:11:44"}
{"current_steps": 1880, "total_steps": 3255, "loss": 0.0183, "lr": 4.5207564090116064e-05, "epoch": 2.8862022644406062, "percentage": 57.76, "elapsed_time": "5:41:10", "remaining_time": "4:09:32"}
{"current_steps": 1890, "total_steps": 3255, "loss": 0.0182, "lr": 4.467402812613323e-05, "epoch": 2.901554404145078, "percentage": 58.06, "elapsed_time": "5:42:27", "remaining_time": "4:07:20"}
{"current_steps": 1900, "total_steps": 3255, "loss": 0.0237, "lr": 4.414110487319244e-05, "epoch": 2.916906543849549, "percentage": 58.37, "elapsed_time": "5:43:44", "remaining_time": "4:05:08"}
{"current_steps": 1900, "total_steps": 3255, "eval_loss": 0.03842131420969963, "epoch": 2.916906543849549, "percentage": 58.37, "elapsed_time": "5:46:21", "remaining_time": "4:07:00"}
{"current_steps": 1910, "total_steps": 3255, "loss": 0.0226, "lr": 4.360885563990919e-05, "epoch": 2.9322586835540205, "percentage": 58.68, "elapsed_time": "5:47:43", "remaining_time": "4:04:51"}
{"current_steps": 1920, "total_steps": 3255, "loss": 0.0165, "lr": 4.307734165735829e-05, "epoch": 2.9476108232584917, "percentage": 58.99, "elapsed_time": "5:49:00", "remaining_time": "4:02:40"}
{"current_steps": 1930, "total_steps": 3255, "loss": 0.022, "lr": 4.254662407202976e-05, "epoch": 2.962962962962963, "percentage": 59.29, "elapsed_time": "5:50:17", "remaining_time": "4:00:29"}
{"current_steps": 1940, "total_steps": 3255, "loss": 0.021, "lr": 4.201676393879446e-05, "epoch": 2.9783151026674344, "percentage": 59.6, "elapsed_time": "5:51:34", "remaining_time": "3:58:18"}
{"current_steps": 1950, "total_steps": 3255, "loss": 0.0176, "lr": 4.148782221388007e-05, "epoch": 2.9936672423719055, "percentage": 59.91, "elapsed_time": "5:52:51", "remaining_time": "3:56:08"}
{"current_steps": 1950, "total_steps": 3255, "eval_loss": 0.03724433481693268, "epoch": 2.9936672423719055, "percentage": 59.91, "elapsed_time": "5:55:29", "remaining_time": "3:57:54"}
{"current_steps": 1960, "total_steps": 3255, "loss": 0.0113, "lr": 4.0959859747858706e-05, "epoch": 3.009019382076377, "percentage": 60.22, "elapsed_time": "5:56:50", "remaining_time": "3:55:46"}
{"current_steps": 1970, "total_steps": 3255, "loss": 0.0108, "lr": 4.043293727864644e-05, "epoch": 3.0243715217808482, "percentage": 60.52, "elapsed_time": "5:58:08", "remaining_time": "3:53:36"}
{"current_steps": 1980, "total_steps": 3255, "loss": 0.0097, "lr": 3.990711542451591e-05, "epoch": 3.0397236614853194, "percentage": 60.83, "elapsed_time": "5:59:24", "remaining_time": "3:51:26"}
{"current_steps": 1990, "total_steps": 3255, "loss": 0.0127, "lr": 3.9382454677122704e-05, "epoch": 3.055075801189791, "percentage": 61.14, "elapsed_time": "6:00:42", "remaining_time": "3:49:17"}
{"current_steps": 2000, "total_steps": 3255, "loss": 0.0057, "lr": 3.885901539454623e-05, "epoch": 3.070427940894262, "percentage": 61.44, "elapsed_time": "6:01:59", "remaining_time": "3:47:08"}
{"current_steps": 2000, "total_steps": 3255, "eval_loss": 0.04431070759892464, "epoch": 3.070427940894262, "percentage": 61.44, "elapsed_time": "6:04:36", "remaining_time": "3:48:47"}
{"current_steps": 2010, "total_steps": 3255, "loss": 0.0172, "lr": 3.833685779434597e-05, "epoch": 3.0857800805987337, "percentage": 61.75, "elapsed_time": "6:05:58", "remaining_time": "3:46:41"}
{"current_steps": 2020, "total_steps": 3255, "loss": 0.0111, "lr": 3.7816041946634024e-05, "epoch": 3.101132220303205, "percentage": 62.06, "elapsed_time": "6:07:15", "remaining_time": "3:44:32"}
{"current_steps": 2030, "total_steps": 3255, "loss": 0.0126, "lr": 3.729662776716439e-05, "epoch": 3.116484360007676, "percentage": 62.37, "elapsed_time": "6:08:32", "remaining_time": "3:42:23"}
{"current_steps": 2040, "total_steps": 3255, "loss": 0.0059, "lr": 3.677867501044029e-05, "epoch": 3.1318364997121475, "percentage": 62.67, "elapsed_time": "6:09:49", "remaining_time": "3:40:15"}
{"current_steps": 2050, "total_steps": 3255, "loss": 0.0148, "lr": 3.6262243262839654e-05, "epoch": 3.1471886394166186, "percentage": 62.98, "elapsed_time": "6:11:06", "remaining_time": "3:38:08"}
{"current_steps": 2050, "total_steps": 3255, "eval_loss": 0.04357267543673515, "epoch": 3.1471886394166186, "percentage": 62.98, "elapsed_time": "6:13:43", "remaining_time": "3:39:40"}
{"current_steps": 2060, "total_steps": 3255, "loss": 0.007, "lr": 3.574739193576042e-05, "epoch": 3.16254077912109, "percentage": 63.29, "elapsed_time": "6:15:05", "remaining_time": "3:37:35"}
{"current_steps": 2070, "total_steps": 3255, "loss": 0.0091, "lr": 3.5234180258785554e-05, "epoch": 3.1778929188255614, "percentage": 63.59, "elapsed_time": "6:16:22", "remaining_time": "3:35:27"}
{"current_steps": 2080, "total_steps": 3255, "loss": 0.0073, "lr": 3.472266727286928e-05, "epoch": 3.1932450585300325, "percentage": 63.9, "elapsed_time": "6:17:39", "remaining_time": "3:33:20"}
{"current_steps": 2090, "total_steps": 3255, "loss": 0.0123, "lr": 3.4212911823544746e-05, "epoch": 3.208597198234504, "percentage": 64.21, "elapsed_time": "6:18:56", "remaining_time": "3:31:13"}
{"current_steps": 2100, "total_steps": 3255, "loss": 0.0153, "lr": 3.370497255415443e-05, "epoch": 3.223949337938975, "percentage": 64.52, "elapsed_time": "6:20:13", "remaining_time": "3:29:07"}
{"current_steps": 2100, "total_steps": 3255, "eval_loss": 0.04993755370378494, "epoch": 3.223949337938975, "percentage": 64.52, "elapsed_time": "6:22:50", "remaining_time": "3:30:33"}
{"current_steps": 2110, "total_steps": 3255, "loss": 0.0171, "lr": 3.319890789910364e-05, "epoch": 3.239301477643447, "percentage": 64.82, "elapsed_time": "6:24:11", "remaining_time": "3:28:29"}
{"current_steps": 2120, "total_steps": 3255, "loss": 0.0155, "lr": 3.269477607713802e-05, "epoch": 3.254653617347918, "percentage": 65.13, "elapsed_time": "6:25:29", "remaining_time": "3:26:22"}
{"current_steps": 2130, "total_steps": 3255, "loss": 0.0125, "lr": 3.219263508464614e-05, "epoch": 3.270005757052389, "percentage": 65.44, "elapsed_time": "6:26:45", "remaining_time": "3:24:16"}
{"current_steps": 2140, "total_steps": 3255, "loss": 0.0113, "lr": 3.169254268898725e-05, "epoch": 3.2853578967568606, "percentage": 65.75, "elapsed_time": "6:28:02", "remaining_time": "3:22:11"}
{"current_steps": 2150, "total_steps": 3255, "loss": 0.0116, "lr": 3.119455642184572e-05, "epoch": 3.3007100364613318, "percentage": 66.05, "elapsed_time": "6:29:20", "remaining_time": "3:20:05"}
{"current_steps": 2150, "total_steps": 3255, "eval_loss": 0.044418033212423325, "epoch": 3.3007100364613318, "percentage": 66.05, "elapsed_time": "6:31:57", "remaining_time": "3:21:26"}
{"current_steps": 2160, "total_steps": 3255, "loss": 0.0135, "lr": 3.069873357261249e-05, "epoch": 3.3160621761658033, "percentage": 66.36, "elapsed_time": "6:33:19", "remaining_time": "3:19:23"}
{"current_steps": 2170, "total_steps": 3255, "loss": 0.0156, "lr": 3.020513118179428e-05, "epoch": 3.3314143158702745, "percentage": 66.67, "elapsed_time": "6:34:35", "remaining_time": "3:17:17"}
{"current_steps": 2180, "total_steps": 3255, "loss": 0.0107, "lr": 2.9713806034451652e-05, "epoch": 3.3467664555747456, "percentage": 66.97, "elapsed_time": "6:35:52", "remaining_time": "3:15:12"}
{"current_steps": 2190, "total_steps": 3255, "loss": 0.0107, "lr": 2.9224814653666242e-05, "epoch": 3.362118595279217, "percentage": 67.28, "elapsed_time": "6:37:09", "remaining_time": "3:13:08"}
{"current_steps": 2200, "total_steps": 3255, "loss": 0.0116, "lr": 2.8738213294038212e-05, "epoch": 3.3774707349836883, "percentage": 67.59, "elapsed_time": "6:38:26", "remaining_time": "3:11:04"}
{"current_steps": 2200, "total_steps": 3255, "eval_loss": 0.04727363586425781, "epoch": 3.3774707349836883, "percentage": 67.59, "elapsed_time": "6:41:03", "remaining_time": "3:12:19"}
{"current_steps": 2210, "total_steps": 3255, "loss": 0.0118, "lr": 2.8254057935214735e-05, "epoch": 3.39282287468816, "percentage": 67.9, "elapsed_time": "6:42:25", "remaining_time": "3:10:17"}
{"current_steps": 2220, "total_steps": 3255, "loss": 0.0126, "lr": 2.7772404275449825e-05, "epoch": 3.408175014392631, "percentage": 68.2, "elapsed_time": "6:43:41", "remaining_time": "3:08:12"}
{"current_steps": 2230, "total_steps": 3255, "loss": 0.0084, "lr": 2.7293307725196793e-05, "epoch": 3.423527154097102, "percentage": 68.51, "elapsed_time": "6:44:58", "remaining_time": "3:06:08"}
{"current_steps": 2240, "total_steps": 3255, "loss": 0.008, "lr": 2.6816823400733625e-05, "epoch": 3.4388792938015738, "percentage": 68.82, "elapsed_time": "6:46:15", "remaining_time": "3:04:05"}
{"current_steps": 2250, "total_steps": 3255, "loss": 0.0133, "lr": 2.6343006117822437e-05, "epoch": 3.454231433506045, "percentage": 69.12, "elapsed_time": "6:47:32", "remaining_time": "3:02:02"}
{"current_steps": 2250, "total_steps": 3255, "eval_loss": 0.05019335821270943, "epoch": 3.454231433506045, "percentage": 69.12, "elapsed_time": "6:50:09", "remaining_time": "3:03:12"}
{"current_steps": 2260, "total_steps": 3255, "loss": 0.0106, "lr": 2.587191038540317e-05, "epoch": 3.469583573210516, "percentage": 69.43, "elapsed_time": "6:51:31", "remaining_time": "3:01:10"}
{"current_steps": 2270, "total_steps": 3255, "loss": 0.0048, "lr": 2.5403590399322886e-05, "epoch": 3.4849357129149876, "percentage": 69.74, "elapsed_time": "6:52:49", "remaining_time": "2:59:07"}
{"current_steps": 2280, "total_steps": 3255, "loss": 0.0134, "lr": 2.493810003610092e-05, "epoch": 3.5002878526194587, "percentage": 70.05, "elapsed_time": "6:54:06", "remaining_time": "2:57:05"}
{"current_steps": 2290, "total_steps": 3255, "loss": 0.0158, "lr": 2.4475492846730835e-05, "epoch": 3.51563999232393, "percentage": 70.35, "elapsed_time": "6:55:23", "remaining_time": "2:55:02"}
{"current_steps": 2300, "total_steps": 3255, "loss": 0.0095, "lr": 2.4015822050519794e-05, "epoch": 3.5309921320284015, "percentage": 70.66, "elapsed_time": "6:56:39", "remaining_time": "2:53:00"}
{"current_steps": 2300, "total_steps": 3255, "eval_loss": 0.05009521171450615, "epoch": 3.5309921320284015, "percentage": 70.66, "elapsed_time": "6:59:17", "remaining_time": "2:54:05"}
{"current_steps": 2310, "total_steps": 3255, "loss": 0.0067, "lr": 2.3559140528966074e-05, "epoch": 3.546344271732873, "percentage": 70.97, "elapsed_time": "7:00:38", "remaining_time": "2:52:04"}
{"current_steps": 2320, "total_steps": 3255, "loss": 0.0129, "lr": 2.3105500819675495e-05, "epoch": 3.561696411437344, "percentage": 71.27, "elapsed_time": "7:01:55", "remaining_time": "2:50:02"}
{"current_steps": 2330, "total_steps": 3255, "loss": 0.0151, "lr": 2.265495511031742e-05, "epoch": 3.5770485511418153, "percentage": 71.58, "elapsed_time": "7:03:12", "remaining_time": "2:48:00"}
{"current_steps": 2340, "total_steps": 3255, "loss": 0.0124, "lr": 2.2207555232620893e-05, "epoch": 3.5924006908462864, "percentage": 71.89, "elapsed_time": "7:04:29", "remaining_time": "2:45:59"}
{"current_steps": 2350, "total_steps": 3255, "loss": 0.011, "lr": 2.1763352656411785e-05, "epoch": 3.607752830550758, "percentage": 72.2, "elapsed_time": "7:05:46", "remaining_time": "2:43:58"}
{"current_steps": 2350, "total_steps": 3255, "eval_loss": 0.046690210700035095, "epoch": 3.607752830550758, "percentage": 72.2, "elapsed_time": "7:08:24", "remaining_time": "2:44:58"}
{"current_steps": 2360, "total_steps": 3255, "loss": 0.0069, "lr": 2.1322398483691787e-05, "epoch": 3.6231049702552296, "percentage": 72.5, "elapsed_time": "7:09:46", "remaining_time": "2:42:59"}
{"current_steps": 2370, "total_steps": 3255, "loss": 0.0144, "lr": 2.08847434427593e-05, "epoch": 3.6384571099597007, "percentage": 72.81, "elapsed_time": "7:11:03", "remaining_time": "2:40:57"}
{"current_steps": 2380, "total_steps": 3255, "loss": 0.0115, "lr": 2.0450437882373697e-05, "epoch": 3.653809249664172, "percentage": 73.12, "elapsed_time": "7:12:20", "remaining_time": "2:38:56"}
{"current_steps": 2390, "total_steps": 3255, "loss": 0.0087, "lr": 2.0019531765962995e-05, "epoch": 3.669161389368643, "percentage": 73.43, "elapsed_time": "7:13:37", "remaining_time": "2:36:56"}
{"current_steps": 2400, "total_steps": 3255, "loss": 0.0136, "lr": 1.9592074665876026e-05, "epoch": 3.6845135290731146, "percentage": 73.73, "elapsed_time": "7:14:53", "remaining_time": "2:34:55"}
{"current_steps": 2400, "total_steps": 3255, "eval_loss": 0.05131568759679794, "epoch": 3.6845135290731146, "percentage": 73.73, "elapsed_time": "7:17:31", "remaining_time": "2:35:51"}
{"current_steps": 2410, "total_steps": 3255, "loss": 0.0155, "lr": 1.9168115757679535e-05, "epoch": 3.6998656687775857, "percentage": 74.04, "elapsed_time": "7:18:52", "remaining_time": "2:33:52"}
{"current_steps": 2420, "total_steps": 3255, "loss": 0.0095, "lr": 1.8747703814500866e-05, "epoch": 3.7152178084820573, "percentage": 74.35, "elapsed_time": "7:20:09", "remaining_time": "2:31:52"}
{"current_steps": 2430, "total_steps": 3255, "loss": 0.0099, "lr": 1.833088720141698e-05, "epoch": 3.7305699481865284, "percentage": 74.65, "elapsed_time": "7:21:26", "remaining_time": "2:29:52"}
{"current_steps": 2440, "total_steps": 3255, "loss": 0.0121, "lr": 1.7917713869890557e-05, "epoch": 3.7459220878909996, "percentage": 74.96, "elapsed_time": "7:22:43", "remaining_time": "2:27:52"}
{"current_steps": 2450, "total_steps": 3255, "loss": 0.0074, "lr": 1.7508231352253435e-05, "epoch": 3.761274227595471, "percentage": 75.27, "elapsed_time": "7:24:00", "remaining_time": "2:25:53"}
{"current_steps": 2450, "total_steps": 3255, "eval_loss": 0.04784177616238594, "epoch": 3.761274227595471, "percentage": 75.27, "elapsed_time": "7:26:37", "remaining_time": "2:26:44"}
{"current_steps": 2460, "total_steps": 3255, "loss": 0.0115, "lr": 1.7102486756238435e-05, "epoch": 3.7766263672999423, "percentage": 75.58, "elapsed_time": "7:27:58", "remaining_time": "2:24:46"}
{"current_steps": 2470, "total_steps": 3255, "loss": 0.0094, "lr": 1.6700526759560002e-05, "epoch": 3.791978507004414, "percentage": 75.88, "elapsed_time": "7:29:15", "remaining_time": "2:22:46"}
{"current_steps": 2480, "total_steps": 3255, "loss": 0.0142, "lr": 1.6302397604544257e-05, "epoch": 3.807330646708885, "percentage": 76.19, "elapsed_time": "7:30:32", "remaining_time": "2:20:47"}
{"current_steps": 2490, "total_steps": 3255, "loss": 0.0074, "lr": 1.5908145092809272e-05, "epoch": 3.822682786413356, "percentage": 76.5, "elapsed_time": "7:31:50", "remaining_time": "2:18:48"}
{"current_steps": 2500, "total_steps": 3255, "loss": 0.0104, "lr": 1.551781457999586e-05, "epoch": 3.8380349261178277, "percentage": 76.8, "elapsed_time": "7:33:07", "remaining_time": "2:16:50"}
{"current_steps": 2500, "total_steps": 3255, "eval_loss": 0.04920686408877373, "epoch": 3.8380349261178277, "percentage": 76.8, "elapsed_time": "7:35:44", "remaining_time": "2:17:37"}
{"current_steps": 2510, "total_steps": 3255, "loss": 0.0083, "lr": 1.513145097054977e-05, "epoch": 3.853387065822299, "percentage": 77.11, "elapsed_time": "7:37:05", "remaining_time": "2:15:40"}
{"current_steps": 2520, "total_steps": 3255, "loss": 0.0116, "lr": 1.4749098712555854e-05, "epoch": 3.8687392055267704, "percentage": 77.42, "elapsed_time": "7:38:22", "remaining_time": "2:13:41"}
{"current_steps": 2530, "total_steps": 3255, "loss": 0.0063, "lr": 1.4370801792624656e-05, "epoch": 3.8840913452312416, "percentage": 77.73, "elapsed_time": "7:39:39", "remaining_time": "2:11:43"}
{"current_steps": 2540, "total_steps": 3255, "loss": 0.0098, "lr": 1.399660373083203e-05, "epoch": 3.8994434849357127, "percentage": 78.03, "elapsed_time": "7:40:57", "remaining_time": "2:09:45"}
{"current_steps": 2550, "total_steps": 3255, "loss": 0.0131, "lr": 1.3626547575712545e-05, "epoch": 3.9147956246401843, "percentage": 78.34, "elapsed_time": "7:42:14", "remaining_time": "2:07:47"}
{"current_steps": 2550, "total_steps": 3255, "eval_loss": 0.051388129591941833, "epoch": 3.9147956246401843, "percentage": 78.34, "elapsed_time": "7:44:51", "remaining_time": "2:08:31"}
{"current_steps": 2560, "total_steps": 3255, "loss": 0.0177, "lr": 1.3260675899307096e-05, "epoch": 3.9301477643446554, "percentage": 78.65, "elapsed_time": "7:46:13", "remaining_time": "2:06:34"}
{"current_steps": 2570, "total_steps": 3255, "loss": 0.0107, "lr": 1.2899030792265349e-05, "epoch": 3.945499904049127, "percentage": 78.96, "elapsed_time": "7:47:30", "remaining_time": "2:04:36"}
{"current_steps": 2580, "total_steps": 3255, "loss": 0.0124, "lr": 1.2541653859003437e-05, "epoch": 3.960852043753598, "percentage": 79.26, "elapsed_time": "7:48:47", "remaining_time": "2:02:38"}
{"current_steps": 2590, "total_steps": 3255, "loss": 0.0121, "lr": 1.2188586212917846e-05, "epoch": 3.9762041834580693, "percentage": 79.57, "elapsed_time": "7:50:04", "remaining_time": "2:00:41"}
{"current_steps": 2600, "total_steps": 3255, "loss": 0.0113, "lr": 1.1839868471655523e-05, "epoch": 3.991556323162541, "percentage": 79.88, "elapsed_time": "7:51:21", "remaining_time": "1:58:44"}
{"current_steps": 2600, "total_steps": 3255, "eval_loss": 0.04836370795965195, "epoch": 3.991556323162541, "percentage": 79.88, "elapsed_time": "7:53:58", "remaining_time": "1:59:24"}
{"current_steps": 2610, "total_steps": 3255, "loss": 0.0093, "lr": 1.1495540752441235e-05, "epoch": 4.006908462867012, "percentage": 80.18, "elapsed_time": "7:55:20", "remaining_time": "1:57:28"}
{"current_steps": 2620, "total_steps": 3255, "loss": 0.0047, "lr": 1.1155642667462318e-05, "epoch": 4.0222606025714835, "percentage": 80.49, "elapsed_time": "7:56:37", "remaining_time": "1:55:31"}
{"current_steps": 2630, "total_steps": 3255, "loss": 0.0039, "lr": 1.082021331931164e-05, "epoch": 4.037612742275955, "percentage": 80.8, "elapsed_time": "7:57:54", "remaining_time": "1:53:34"}
{"current_steps": 2640, "total_steps": 3255, "loss": 0.0046, "lr": 1.0489291296489152e-05, "epoch": 4.052964881980426, "percentage": 81.11, "elapsed_time": "7:59:11", "remaining_time": "1:51:37"}
{"current_steps": 2650, "total_steps": 3255, "loss": 0.0028, "lr": 1.0162914668962631e-05, "epoch": 4.068317021684897, "percentage": 81.41, "elapsed_time": "8:00:28", "remaining_time": "1:49:41"}
{"current_steps": 2650, "total_steps": 3255, "eval_loss": 0.05359194427728653, "epoch": 4.068317021684897, "percentage": 81.41, "elapsed_time": "8:03:05", "remaining_time": "1:50:17"}
{"current_steps": 2660, "total_steps": 3255, "loss": 0.0034, "lr": 9.841120983787915e-06, "epoch": 4.083669161389369, "percentage": 81.72, "elapsed_time": "8:04:27", "remaining_time": "1:48:21"}
{"current_steps": 2670, "total_steps": 3255, "loss": 0.0043, "lr": 9.523947260789546e-06, "epoch": 4.09902130109384, "percentage": 82.03, "elapsed_time": "8:05:44", "remaining_time": "1:46:25"}
{"current_steps": 2680, "total_steps": 3255, "loss": 0.0035, "lr": 9.211429988301823e-06, "epoch": 4.114373440798311, "percentage": 82.33, "elapsed_time": "8:07:01", "remaining_time": "1:44:29"}
{"current_steps": 2690, "total_steps": 3255, "loss": 0.0065, "lr": 8.90360511897121e-06, "epoch": 4.129725580502782, "percentage": 82.64, "elapsed_time": "8:08:18", "remaining_time": "1:42:33"}
{"current_steps": 2700, "total_steps": 3255, "loss": 0.0046, "lr": 8.600508065620161e-06, "epoch": 4.1450777202072535, "percentage": 82.95, "elapsed_time": "8:09:35", "remaining_time": "1:40:38"}
{"current_steps": 2700, "total_steps": 3255, "eval_loss": 0.057626817375421524, "epoch": 4.1450777202072535, "percentage": 82.95, "elapsed_time": "8:12:12", "remaining_time": "1:41:10"}
{"current_steps": 2710, "total_steps": 3255, "loss": 0.0036, "lr": 8.302173697173226e-06, "epoch": 4.1604298599117255, "percentage": 83.26, "elapsed_time": "8:13:35", "remaining_time": "1:39:15"}
{"current_steps": 2720, "total_steps": 3255, "loss": 0.0033, "lr": 8.008636334645631e-06, "epoch": 4.175781999616197, "percentage": 83.56, "elapsed_time": "8:14:52", "remaining_time": "1:37:20"}
{"current_steps": 2730, "total_steps": 3255, "loss": 0.0024, "lr": 7.71992974719491e-06, "epoch": 4.191134139320668, "percentage": 83.87, "elapsed_time": "8:16:09", "remaining_time": "1:35:24"}
{"current_steps": 2740, "total_steps": 3255, "loss": 0.0019, "lr": 7.436087148236054e-06, "epoch": 4.206486279025139, "percentage": 84.18, "elapsed_time": "8:17:26", "remaining_time": "1:33:29"}
{"current_steps": 2750, "total_steps": 3255, "loss": 0.0038, "lr": 7.157141191620548e-06, "epoch": 4.22183841872961, "percentage": 84.49, "elapsed_time": "8:18:44", "remaining_time": "1:31:35"}
{"current_steps": 2750, "total_steps": 3255, "eval_loss": 0.06159433349967003, "epoch": 4.22183841872961, "percentage": 84.49, "elapsed_time": "8:21:21", "remaining_time": "1:32:04"}
{"current_steps": 2760, "total_steps": 3255, "loss": 0.0052, "lr": 6.883123967879796e-06, "epoch": 4.237190558434082, "percentage": 84.79, "elapsed_time": "8:22:43", "remaining_time": "1:30:09"}
{"current_steps": 2770, "total_steps": 3255, "loss": 0.0036, "lr": 6.6140670005334136e-06, "epoch": 4.252542698138553, "percentage": 85.1, "elapsed_time": "8:24:00", "remaining_time": "1:28:14"}
{"current_steps": 2780, "total_steps": 3255, "loss": 0.0048, "lr": 6.350001242462617e-06, "epoch": 4.267894837843024, "percentage": 85.41, "elapsed_time": "8:25:17", "remaining_time": "1:26:20"}
{"current_steps": 2790, "total_steps": 3255, "loss": 0.0045, "lr": 6.090957072349385e-06, "epoch": 4.2832469775474955, "percentage": 85.71, "elapsed_time": "8:26:34", "remaining_time": "1:24:25"}
{"current_steps": 2800, "total_steps": 3255, "loss": 0.0029, "lr": 5.836964291181624e-06, "epoch": 4.298599117251967, "percentage": 86.02, "elapsed_time": "8:27:51", "remaining_time": "1:22:31"}
{"current_steps": 2800, "total_steps": 3255, "eval_loss": 0.062100403010845184, "epoch": 4.298599117251967, "percentage": 86.02, "elapsed_time": "8:30:28", "remaining_time": "1:22:57"}
{"current_steps": 2810, "total_steps": 3255, "loss": 0.0033, "lr": 5.588052118824804e-06, "epoch": 4.313951256956439, "percentage": 86.33, "elapsed_time": "8:31:50", "remaining_time": "1:21:03"}
{"current_steps": 2820, "total_steps": 3255, "loss": 0.0021, "lr": 5.344249190660428e-06, "epoch": 4.32930339666091, "percentage": 86.64, "elapsed_time": "8:33:07", "remaining_time": "1:19:09"}
{"current_steps": 2830, "total_steps": 3255, "loss": 0.0036, "lr": 5.105583554291765e-06, "epoch": 4.344655536365381, "percentage": 86.94, "elapsed_time": "8:34:24", "remaining_time": "1:17:15"}
{"current_steps": 2840, "total_steps": 3255, "loss": 0.0027, "lr": 4.872082666317207e-06, "epoch": 4.360007676069852, "percentage": 87.25, "elapsed_time": "8:35:40", "remaining_time": "1:15:21"}
{"current_steps": 2850, "total_steps": 3255, "loss": 0.0025, "lr": 4.6437733891715905e-06, "epoch": 4.375359815774323, "percentage": 87.56, "elapsed_time": "8:36:58", "remaining_time": "1:13:27"}
{"current_steps": 2850, "total_steps": 3255, "eval_loss": 0.06398806720972061, "epoch": 4.375359815774323, "percentage": 87.56, "elapsed_time": "8:39:35", "remaining_time": "1:13:50"}
{"current_steps": 2860, "total_steps": 3255, "loss": 0.0042, "lr": 4.420681988035891e-06, "epoch": 4.390711955478795, "percentage": 87.86, "elapsed_time": "8:40:58", "remaining_time": "1:11:57"}
{"current_steps": 2870, "total_steps": 3255, "loss": 0.0055, "lr": 4.2028341278156026e-06, "epoch": 4.406064095183266, "percentage": 88.17, "elapsed_time": "8:42:15", "remaining_time": "1:10:03"}
{"current_steps": 2880, "total_steps": 3255, "loss": 0.0052, "lr": 3.990254870188221e-06, "epoch": 4.4214162348877375, "percentage": 88.48, "elapsed_time": "8:43:32", "remaining_time": "1:08:10"}
{"current_steps": 2890, "total_steps": 3255, "loss": 0.0046, "lr": 3.7829686707200827e-06, "epoch": 4.436768374592209, "percentage": 88.79, "elapsed_time": "8:44:49", "remaining_time": "1:06:17"}
{"current_steps": 2900, "total_steps": 3255, "loss": 0.0037, "lr": 3.580999376052946e-06, "epoch": 4.45212051429668, "percentage": 89.09, "elapsed_time": "8:46:06", "remaining_time": "1:04:24"}
{"current_steps": 2900, "total_steps": 3255, "eval_loss": 0.06381876021623611, "epoch": 4.45212051429668, "percentage": 89.09, "elapsed_time": "8:48:43", "remaining_time": "1:04:43"}
{"current_steps": 2910, "total_steps": 3255, "loss": 0.0031, "lr": 3.3843702211606153e-06, "epoch": 4.467472654001152, "percentage": 89.4, "elapsed_time": "8:50:05", "remaining_time": "1:02:50"}
{"current_steps": 2920, "total_steps": 3255, "loss": 0.0058, "lr": 3.193103826675947e-06, "epoch": 4.482824793705623, "percentage": 89.71, "elapsed_time": "8:51:22", "remaining_time": "1:00:57"}
{"current_steps": 2930, "total_steps": 3255, "loss": 0.002, "lr": 3.007222196288545e-06, "epoch": 4.498176933410094, "percentage": 90.02, "elapsed_time": "8:52:40", "remaining_time": "0:59:05"}
{"current_steps": 2940, "total_steps": 3255, "loss": 0.0025, "lr": 2.8267467142133687e-06, "epoch": 4.513529073114565, "percentage": 90.32, "elapsed_time": "8:53:57", "remaining_time": "0:57:12"}
{"current_steps": 2950, "total_steps": 3255, "loss": 0.0071, "lr": 2.651698142730674e-06, "epoch": 4.528881212819036, "percentage": 90.63, "elapsed_time": "8:55:15", "remaining_time": "0:55:20"}
{"current_steps": 2950, "total_steps": 3255, "eval_loss": 0.06478563696146011, "epoch": 4.528881212819036, "percentage": 90.63, "elapsed_time": "8:57:52", "remaining_time": "0:55:36"}
{"current_steps": 2960, "total_steps": 3255, "loss": 0.0053, "lr": 2.4820966197974748e-06, "epoch": 4.544233352523508, "percentage": 90.94, "elapsed_time": "8:59:14", "remaining_time": "0:53:44"}
{"current_steps": 2970, "total_steps": 3255, "loss": 0.002, "lr": 2.3179616567308216e-06, "epoch": 4.5595854922279795, "percentage": 91.24, "elapsed_time": "9:00:31", "remaining_time": "0:51:52"}
{"current_steps": 2980, "total_steps": 3255, "loss": 0.0026, "lr": 2.1593121359631873e-06, "epoch": 4.574937631932451, "percentage": 91.55, "elapsed_time": "9:01:48", "remaining_time": "0:49:59"}
{"current_steps": 2990, "total_steps": 3255, "loss": 0.0045, "lr": 2.006166308870189e-06, "epoch": 4.590289771636922, "percentage": 91.86, "elapsed_time": "9:03:06", "remaining_time": "0:48:08"}
{"current_steps": 3000, "total_steps": 3255, "loss": 0.0015, "lr": 1.8585417936709038e-06, "epoch": 4.605641911341393, "percentage": 92.17, "elapsed_time": "9:04:23", "remaining_time": "0:46:16"}
{"current_steps": 3000, "total_steps": 3255, "eval_loss": 0.0653165951371193, "epoch": 4.605641911341393, "percentage": 92.17, "elapsed_time": "9:07:00", "remaining_time": "0:46:29"}
{"current_steps": 3010, "total_steps": 3255, "loss": 0.0093, "lr": 1.7164555734010545e-06, "epoch": 4.620994051045864, "percentage": 92.47, "elapsed_time": "9:08:22", "remaining_time": "0:44:38"}
{"current_steps": 3020, "total_steps": 3255, "loss": 0.0031, "lr": 1.5799239939592204e-06, "epoch": 4.636346190750336, "percentage": 92.78, "elapsed_time": "9:09:39", "remaining_time": "0:42:46"}
{"current_steps": 3030, "total_steps": 3255, "loss": 0.0034, "lr": 1.4489627622263747e-06, "epoch": 4.651698330454807, "percentage": 93.09, "elapsed_time": "9:10:56", "remaining_time": "0:40:54"}
{"current_steps": 3040, "total_steps": 3255, "loss": 0.0083, "lr": 1.3235869442589255e-06, "epoch": 4.667050470159278, "percentage": 93.39, "elapsed_time": "9:12:13", "remaining_time": "0:39:03"}
{"current_steps": 3050, "total_steps": 3255, "loss": 0.002, "lr": 1.2038109635555406e-06, "epoch": 4.6824026098637495, "percentage": 93.7, "elapsed_time": "9:13:30", "remaining_time": "0:37:12"}
{"current_steps": 3050, "total_steps": 3255, "eval_loss": 0.06511291116476059, "epoch": 4.6824026098637495, "percentage": 93.7, "elapsed_time": "9:16:07", "remaining_time": "0:37:22"}
{"current_steps": 3060, "total_steps": 3255, "loss": 0.005, "lr": 1.0896485993977467e-06, "epoch": 4.697754749568221, "percentage": 94.01, "elapsed_time": "9:17:30", "remaining_time": "0:35:31"}
{"current_steps": 3070, "total_steps": 3255, "loss": 0.0044, "lr": 9.811129852647982e-07, "epoch": 4.713106889272693, "percentage": 94.32, "elapsed_time": "9:18:47", "remaining_time": "0:33:40"}
{"current_steps": 3080, "total_steps": 3255, "loss": 0.0104, "lr": 8.782166073227515e-07, "epoch": 4.728459028977164, "percentage": 94.62, "elapsed_time": "9:20:04", "remaining_time": "0:31:49"}
{"current_steps": 3090, "total_steps": 3255, "loss": 0.0048, "lr": 7.809713029880428e-07, "epoch": 4.743811168681635, "percentage": 94.93, "elapsed_time": "9:21:21", "remaining_time": "0:29:58"}
{"current_steps": 3100, "total_steps": 3255, "loss": 0.0073, "lr": 6.893882595656598e-07, "epoch": 4.759163308386106, "percentage": 95.24, "elapsed_time": "9:22:38", "remaining_time": "0:28:07"}
{"current_steps": 3100, "total_steps": 3255, "eval_loss": 0.0645485520362854, "epoch": 4.759163308386106, "percentage": 95.24, "elapsed_time": "9:25:16", "remaining_time": "0:28:15"}
{"current_steps": 3110, "total_steps": 3255, "loss": 0.008, "lr": 6.034780129621664e-07, "epoch": 4.774515448090577, "percentage": 95.55, "elapsed_time": "9:26:37", "remaining_time": "0:26:25"}
{"current_steps": 3120, "total_steps": 3255, "loss": 0.0069, "lr": 5.232504464735833e-07, "epoch": 4.789867587795049, "percentage": 95.85, "elapsed_time": "9:27:54", "remaining_time": "0:24:34"}
{"current_steps": 3130, "total_steps": 3255, "loss": 0.005, "lr": 4.487147896484523e-07, "epoch": 4.80521972749952, "percentage": 96.16, "elapsed_time": "9:29:11", "remaining_time": "0:22:43"}
{"current_steps": 3140, "total_steps": 3255, "loss": 0.0077, "lr": 3.7987961722599773e-07, "epoch": 4.8205718672039914, "percentage": 96.47, "elapsed_time": "9:30:28", "remaining_time": "0:20:53"}
{"current_steps": 3150, "total_steps": 3255, "loss": 0.0061, "lr": 3.167528481496984e-07, "epoch": 4.835924006908463, "percentage": 96.77, "elapsed_time": "9:31:45", "remaining_time": "0:19:03"}
{"current_steps": 3150, "total_steps": 3255, "eval_loss": 0.06395623832941055, "epoch": 4.835924006908463, "percentage": 96.77, "elapsed_time": "9:34:22", "remaining_time": "0:19:08"}
{"current_steps": 3160, "total_steps": 3255, "loss": 0.0103, "lr": 2.593417446562607e-07, "epoch": 4.851276146612934, "percentage": 97.08, "elapsed_time": "9:35:44", "remaining_time": "0:17:18"}
{"current_steps": 3170, "total_steps": 3255, "loss": 0.0017, "lr": 2.0765291144016486e-07, "epoch": 4.866628286317406, "percentage": 97.39, "elapsed_time": "9:37:01", "remaining_time": "0:15:28"}
{"current_steps": 3180, "total_steps": 3255, "loss": 0.0041, "lr": 1.6169229489385595e-07, "epoch": 4.881980426021877, "percentage": 97.7, "elapsed_time": "9:38:17", "remaining_time": "0:13:38"}
{"current_steps": 3190, "total_steps": 3255, "loss": 0.003, "lr": 1.2146518242363014e-07, "epoch": 4.897332565726348, "percentage": 98.0, "elapsed_time": "9:39:34", "remaining_time": "0:11:48"}
{"current_steps": 3200, "total_steps": 3255, "loss": 0.0024, "lr": 8.697620184138222e-08, "epoch": 4.912684705430819, "percentage": 98.31, "elapsed_time": "9:40:52", "remaining_time": "0:09:59"}
{"current_steps": 3200, "total_steps": 3255, "eval_loss": 0.0639420673251152, "epoch": 4.912684705430819, "percentage": 98.31, "elapsed_time": "9:43:29", "remaining_time": "0:10:01"}
{"current_steps": 3210, "total_steps": 3255, "loss": 0.0098, "lr": 5.822932083221488e-08, "epoch": 4.92803684513529, "percentage": 98.62, "elapsed_time": "9:44:51", "remaining_time": "0:08:11"}
{"current_steps": 3220, "total_steps": 3255, "loss": 0.0037, "lr": 3.5227846497970504e-08, "epoch": 4.943388984839762, "percentage": 98.92, "elapsed_time": "9:46:08", "remaining_time": "0:06:22"}
{"current_steps": 3230, "total_steps": 3255, "loss": 0.0014, "lr": 1.7974424976796577e-08, "epoch": 4.958741124544233, "percentage": 99.23, "elapsed_time": "9:47:25", "remaining_time": "0:04:32"}
{"current_steps": 3240, "total_steps": 3255, "loss": 0.0055, "lr": 6.47104113870034e-09, "epoch": 4.974093264248705, "percentage": 99.54, "elapsed_time": "9:48:42", "remaining_time": "0:02:43"}
{"current_steps": 3250, "total_steps": 3255, "loss": 0.0054, "lr": 7.190183572314269e-10, "epoch": 4.989445403953176, "percentage": 99.85, "elapsed_time": "9:49:59", "remaining_time": "0:00:54"}
{"current_steps": 3250, "total_steps": 3255, "eval_loss": 0.06391575932502747, "epoch": 4.989445403953176, "percentage": 99.85, "elapsed_time": "9:52:37", "remaining_time": "0:00:54"}
{"current_steps": 3255, "total_steps": 3255, "epoch": 4.997121473805412, "percentage": 100.0, "elapsed_time": "9:53:24", "remaining_time": "0:00:00"}