diff --git "a/trainer_log.jsonl" "b/trainer_log.jsonl" --- "a/trainer_log.jsonl" +++ "b/trainer_log.jsonl" @@ -6117,3 +6117,679 @@ {"current_steps": 30540, "total_steps": 33920, "loss": 0.2888, "lr": 2.9960408270770624e-07, "epoch": 18.007075471698112, "percentage": 90.04, "elapsed_time": "1:00:08", "remaining_time": "0:06:39", "throughput": 5529.1, "total_tokens": 19950424} {"current_steps": 30545, "total_steps": 33920, "loss": 0.2917, "lr": 2.9872752215055755e-07, "epoch": 18.01002358490566, "percentage": 90.05, "elapsed_time": "1:00:08", "remaining_time": "0:06:38", "throughput": 5529.05, "total_tokens": 19953176} {"current_steps": 30550, "total_steps": 33920, "loss": 0.4209, "lr": 2.9785220627555844e-07, "epoch": 18.01297169811321, "percentage": 90.06, "elapsed_time": "1:00:09", "remaining_time": "0:06:38", "throughput": 5529.27, "total_tokens": 19959256} +{"current_steps": 30555, "total_steps": 33920, "loss": 0.3351, "lr": 2.9697813531445295e-07, "epoch": 18.015919811320753, "percentage": 90.08, "elapsed_time": "1:00:10", "remaining_time": "0:06:37", "throughput": 5529.12, "total_tokens": 19962744} +{"current_steps": 30560, "total_steps": 33920, "loss": 0.3501, "lr": 2.9610530949865433e-07, "epoch": 18.0188679245283, "percentage": 90.09, "elapsed_time": "1:00:11", "remaining_time": "0:06:37", "throughput": 5529.11, "total_tokens": 19966520} +{"current_steps": 30565, "total_steps": 33920, "loss": 0.289, "lr": 2.952337290592483e-07, "epoch": 18.02181603773585, "percentage": 90.11, "elapsed_time": "1:00:11", "remaining_time": "0:06:36", "throughput": 5528.92, "total_tokens": 19968984} +{"current_steps": 30570, "total_steps": 33920, "loss": 0.3454, "lr": 2.9436339422698913e-07, "epoch": 18.024764150943398, "percentage": 90.12, "elapsed_time": "1:00:12", "remaining_time": "0:06:35", "throughput": 5529.1, "total_tokens": 19973496} +{"current_steps": 30575, "total_steps": 33920, "loss": 0.3104, "lr": 2.934943052323008e-07, "epoch": 18.027712264150942, "percentage": 90.14, "elapsed_time": "1:00:12", "remaining_time": "0:06:35", "throughput": 5529.08, "total_tokens": 19976216} +{"current_steps": 30580, "total_steps": 33920, "loss": 0.2161, "lr": 2.926264623052799e-07, "epoch": 18.03066037735849, "percentage": 90.15, "elapsed_time": "1:00:13", "remaining_time": "0:06:34", "throughput": 5529.1, "total_tokens": 19979320} +{"current_steps": 30585, "total_steps": 33920, "loss": 0.3056, "lr": 2.9175986567569036e-07, "epoch": 18.03360849056604, "percentage": 90.17, "elapsed_time": "1:00:14", "remaining_time": "0:06:34", "throughput": 5529.11, "total_tokens": 19982552} +{"current_steps": 30590, "total_steps": 33920, "loss": 0.2964, "lr": 2.9089451557296755e-07, "epoch": 18.036556603773583, "percentage": 90.18, "elapsed_time": "1:00:14", "remaining_time": "0:06:33", "throughput": 5529.05, "total_tokens": 19985432} +{"current_steps": 30595, "total_steps": 33920, "loss": 0.2714, "lr": 2.9003041222621706e-07, "epoch": 18.03950471698113, "percentage": 90.2, "elapsed_time": "1:00:15", "remaining_time": "0:06:32", "throughput": 5529.09, "total_tokens": 19989368} +{"current_steps": 30600, "total_steps": 33920, "loss": 0.3279, "lr": 2.8916755586421375e-07, "epoch": 18.04245283018868, "percentage": 90.21, "elapsed_time": "1:00:15", "remaining_time": "0:06:32", "throughput": 5529.13, "total_tokens": 19992472} +{"current_steps": 30605, "total_steps": 33920, "loss": 0.33, "lr": 2.883059467154031e-07, "epoch": 18.045400943396228, "percentage": 90.23, "elapsed_time": "1:00:16", "remaining_time": "0:06:31", "throughput": 5529.12, "total_tokens": 19995608} +{"current_steps": 30610, "total_steps": 33920, "loss": 0.2374, "lr": 2.8744558500789887e-07, "epoch": 18.048349056603772, "percentage": 90.24, "elapsed_time": "1:00:16", "remaining_time": "0:06:31", "throughput": 5529.06, "total_tokens": 19998200} +{"current_steps": 30615, "total_steps": 33920, "loss": 0.2613, "lr": 2.8658647096948546e-07, "epoch": 18.05129716981132, "percentage": 90.26, "elapsed_time": "1:00:17", "remaining_time": "0:06:30", "throughput": 5529.11, "total_tokens": 20001464} +{"current_steps": 30620, "total_steps": 33920, "loss": 0.2553, "lr": 2.8572860482761813e-07, "epoch": 18.05424528301887, "percentage": 90.27, "elapsed_time": "1:00:17", "remaining_time": "0:06:29", "throughput": 5529.22, "total_tokens": 20004632} +{"current_steps": 30625, "total_steps": 33920, "loss": 0.2653, "lr": 2.8487198680942017e-07, "epoch": 18.057193396226417, "percentage": 90.29, "elapsed_time": "1:00:18", "remaining_time": "0:06:29", "throughput": 5529.37, "total_tokens": 20008632} +{"current_steps": 30630, "total_steps": 33920, "loss": 0.1991, "lr": 2.840166171416836e-07, "epoch": 18.06014150943396, "percentage": 90.3, "elapsed_time": "1:00:19", "remaining_time": "0:06:28", "throughput": 5529.48, "total_tokens": 20011736} +{"current_steps": 30635, "total_steps": 33920, "loss": 0.4072, "lr": 2.8316249605087386e-07, "epoch": 18.06308962264151, "percentage": 90.32, "elapsed_time": "1:00:19", "remaining_time": "0:06:28", "throughput": 5529.62, "total_tokens": 20015384} +{"current_steps": 30640, "total_steps": 33920, "loss": 0.3528, "lr": 2.823096237631212e-07, "epoch": 18.066037735849058, "percentage": 90.33, "elapsed_time": "1:00:20", "remaining_time": "0:06:27", "throughput": 5529.44, "total_tokens": 20017432} +{"current_steps": 30645, "total_steps": 33920, "loss": 0.2903, "lr": 2.814580005042283e-07, "epoch": 18.068985849056602, "percentage": 90.34, "elapsed_time": "1:00:20", "remaining_time": "0:06:26", "throughput": 5529.43, "total_tokens": 20020632} +{"current_steps": 30650, "total_steps": 33920, "loss": 0.3347, "lr": 2.8060762649966435e-07, "epoch": 18.07193396226415, "percentage": 90.36, "elapsed_time": "1:00:21", "remaining_time": "0:06:26", "throughput": 5529.39, "total_tokens": 20023544} +{"current_steps": 30655, "total_steps": 33920, "loss": 0.349, "lr": 2.797585019745713e-07, "epoch": 18.0748820754717, "percentage": 90.37, "elapsed_time": "1:00:21", "remaining_time": "0:06:25", "throughput": 5529.47, "total_tokens": 20026712} +{"current_steps": 30660, "total_steps": 33920, "loss": 0.4925, "lr": 2.789106271537584e-07, "epoch": 18.077830188679247, "percentage": 90.39, "elapsed_time": "1:00:22", "remaining_time": "0:06:25", "throughput": 5529.49, "total_tokens": 20029560} +{"current_steps": 30665, "total_steps": 33920, "loss": 0.332, "lr": 2.780640022617037e-07, "epoch": 18.08077830188679, "percentage": 90.4, "elapsed_time": "1:00:22", "remaining_time": "0:06:24", "throughput": 5529.55, "total_tokens": 20032824} +{"current_steps": 30670, "total_steps": 33920, "loss": 0.324, "lr": 2.772186275225547e-07, "epoch": 18.08372641509434, "percentage": 90.42, "elapsed_time": "1:00:23", "remaining_time": "0:06:23", "throughput": 5529.45, "total_tokens": 20035224} +{"current_steps": 30675, "total_steps": 33920, "loss": 0.279, "lr": 2.7637450316012836e-07, "epoch": 18.086674528301888, "percentage": 90.43, "elapsed_time": "1:00:23", "remaining_time": "0:06:23", "throughput": 5529.4, "total_tokens": 20037976} +{"current_steps": 30680, "total_steps": 33920, "loss": 0.2902, "lr": 2.755316293979088e-07, "epoch": 18.089622641509433, "percentage": 90.45, "elapsed_time": "1:00:24", "remaining_time": "0:06:22", "throughput": 5529.44, "total_tokens": 20041048} +{"current_steps": 30685, "total_steps": 33920, "loss": 0.324, "lr": 2.7469000645905295e-07, "epoch": 18.09257075471698, "percentage": 90.46, "elapsed_time": "1:00:25", "remaining_time": "0:06:22", "throughput": 5529.28, "total_tokens": 20043704} +{"current_steps": 30690, "total_steps": 33920, "loss": 0.431, "lr": 2.738496345663827e-07, "epoch": 18.09551886792453, "percentage": 90.48, "elapsed_time": "1:00:25", "remaining_time": "0:06:21", "throughput": 5529.33, "total_tokens": 20047352} +{"current_steps": 30695, "total_steps": 33920, "loss": 0.3334, "lr": 2.7301051394239e-07, "epoch": 18.098466981132077, "percentage": 90.49, "elapsed_time": "1:00:26", "remaining_time": "0:06:20", "throughput": 5529.42, "total_tokens": 20051064} +{"current_steps": 30700, "total_steps": 33920, "loss": 0.2439, "lr": 2.72172644809236e-07, "epoch": 18.10141509433962, "percentage": 90.51, "elapsed_time": "1:00:26", "remaining_time": "0:06:20", "throughput": 5529.38, "total_tokens": 20053880} +{"current_steps": 30705, "total_steps": 33920, "loss": 0.412, "lr": 2.7133602738875e-07, "epoch": 18.10436320754717, "percentage": 90.52, "elapsed_time": "1:00:27", "remaining_time": "0:06:19", "throughput": 5529.53, "total_tokens": 20057688} +{"current_steps": 30710, "total_steps": 33920, "loss": 0.2653, "lr": 2.7050066190242976e-07, "epoch": 18.107311320754718, "percentage": 90.54, "elapsed_time": "1:00:27", "remaining_time": "0:06:19", "throughput": 5529.59, "total_tokens": 20061080} +{"current_steps": 30715, "total_steps": 33920, "loss": 0.2716, "lr": 2.696665485714428e-07, "epoch": 18.110259433962263, "percentage": 90.55, "elapsed_time": "1:00:28", "remaining_time": "0:06:18", "throughput": 5529.67, "total_tokens": 20064376} +{"current_steps": 30720, "total_steps": 33920, "loss": 0.4424, "lr": 2.6883368761662367e-07, "epoch": 18.11320754716981, "percentage": 90.57, "elapsed_time": "1:00:29", "remaining_time": "0:06:18", "throughput": 5529.95, "total_tokens": 20069400} +{"current_steps": 30725, "total_steps": 33920, "loss": 0.2995, "lr": 2.680020792584759e-07, "epoch": 18.11615566037736, "percentage": 90.58, "elapsed_time": "1:00:29", "remaining_time": "0:06:17", "throughput": 5530.07, "total_tokens": 20072920} +{"current_steps": 30730, "total_steps": 33920, "loss": 0.2655, "lr": 2.6717172371717113e-07, "epoch": 18.119103773584907, "percentage": 90.6, "elapsed_time": "1:00:30", "remaining_time": "0:06:16", "throughput": 5530.2, "total_tokens": 20076440} +{"current_steps": 30735, "total_steps": 33920, "loss": 0.2359, "lr": 2.663426212125503e-07, "epoch": 18.122051886792452, "percentage": 90.61, "elapsed_time": "1:00:30", "remaining_time": "0:06:16", "throughput": 5530.23, "total_tokens": 20079544} +{"current_steps": 30740, "total_steps": 33920, "loss": 0.2951, "lr": 2.655147719641216e-07, "epoch": 18.125, "percentage": 90.62, "elapsed_time": "1:00:31", "remaining_time": "0:06:15", "throughput": 5530.32, "total_tokens": 20082904} +{"current_steps": 30745, "total_steps": 33920, "loss": 0.3179, "lr": 2.646881761910602e-07, "epoch": 18.127948113207548, "percentage": 90.64, "elapsed_time": "1:00:31", "remaining_time": "0:06:15", "throughput": 5530.46, "total_tokens": 20086520} +{"current_steps": 30750, "total_steps": 33920, "loss": 0.3488, "lr": 2.638628341122135e-07, "epoch": 18.130896226415093, "percentage": 90.65, "elapsed_time": "1:00:32", "remaining_time": "0:06:14", "throughput": 5530.4, "total_tokens": 20089240} +{"current_steps": 30755, "total_steps": 33920, "loss": 0.3391, "lr": 2.6303874594609314e-07, "epoch": 18.13384433962264, "percentage": 90.67, "elapsed_time": "1:00:33", "remaining_time": "0:06:13", "throughput": 5530.42, "total_tokens": 20092312} +{"current_steps": 30760, "total_steps": 33920, "loss": 0.2021, "lr": 2.622159119108797e-07, "epoch": 18.13679245283019, "percentage": 90.68, "elapsed_time": "1:00:33", "remaining_time": "0:06:13", "throughput": 5530.56, "total_tokens": 20096120} +{"current_steps": 30765, "total_steps": 33920, "loss": 0.3968, "lr": 2.6139433222442226e-07, "epoch": 18.139740566037737, "percentage": 90.7, "elapsed_time": "1:00:34", "remaining_time": "0:06:12", "throughput": 5530.68, "total_tokens": 20099512} +{"current_steps": 30770, "total_steps": 33920, "loss": 0.3048, "lr": 2.6057400710423787e-07, "epoch": 18.142688679245282, "percentage": 90.71, "elapsed_time": "1:00:34", "remaining_time": "0:06:12", "throughput": 5530.54, "total_tokens": 20101976} +{"current_steps": 30775, "total_steps": 33920, "loss": 0.292, "lr": 2.5975493676751004e-07, "epoch": 18.14563679245283, "percentage": 90.73, "elapsed_time": "1:00:35", "remaining_time": "0:06:11", "throughput": 5530.58, "total_tokens": 20105592} +{"current_steps": 30780, "total_steps": 33920, "loss": 0.3474, "lr": 2.589371214310926e-07, "epoch": 18.14858490566038, "percentage": 90.74, "elapsed_time": "1:00:35", "remaining_time": "0:06:10", "throughput": 5530.7, "total_tokens": 20109112} +{"current_steps": 30785, "total_steps": 33920, "loss": 0.163, "lr": 2.581205613115051e-07, "epoch": 18.151533018867923, "percentage": 90.76, "elapsed_time": "1:00:36", "remaining_time": "0:06:10", "throughput": 5530.81, "total_tokens": 20112664} +{"current_steps": 30790, "total_steps": 33920, "loss": 0.297, "lr": 2.573052566249357e-07, "epoch": 18.15448113207547, "percentage": 90.77, "elapsed_time": "1:00:37", "remaining_time": "0:06:09", "throughput": 5530.96, "total_tokens": 20116184} +{"current_steps": 30795, "total_steps": 33920, "loss": 0.315, "lr": 2.5649120758723945e-07, "epoch": 18.15742924528302, "percentage": 90.79, "elapsed_time": "1:00:37", "remaining_time": "0:06:09", "throughput": 5531.08, "total_tokens": 20119864} +{"current_steps": 30800, "total_steps": 33920, "loss": 0.2623, "lr": 2.5567841441393906e-07, "epoch": 18.160377358490567, "percentage": 90.8, "elapsed_time": "1:00:38", "remaining_time": "0:06:08", "throughput": 5531.2, "total_tokens": 20123448} +{"current_steps": 30805, "total_steps": 33920, "loss": 0.2588, "lr": 2.548668773202245e-07, "epoch": 18.163325471698112, "percentage": 90.82, "elapsed_time": "1:00:38", "remaining_time": "0:06:07", "throughput": 5531.29, "total_tokens": 20126712} +{"current_steps": 30810, "total_steps": 33920, "loss": 0.3242, "lr": 2.5405659652095573e-07, "epoch": 18.16627358490566, "percentage": 90.83, "elapsed_time": "1:00:39", "remaining_time": "0:06:07", "throughput": 5531.38, "total_tokens": 20130008} +{"current_steps": 30815, "total_steps": 33920, "loss": 0.3454, "lr": 2.5324757223065655e-07, "epoch": 18.16922169811321, "percentage": 90.85, "elapsed_time": "1:00:39", "remaining_time": "0:06:06", "throughput": 5531.4, "total_tokens": 20133048} +{"current_steps": 30820, "total_steps": 33920, "loss": 0.3522, "lr": 2.524398046635207e-07, "epoch": 18.172169811320753, "percentage": 90.86, "elapsed_time": "1:00:40", "remaining_time": "0:06:06", "throughput": 5531.36, "total_tokens": 20136024} +{"current_steps": 30825, "total_steps": 33920, "loss": 0.4183, "lr": 2.51633294033406e-07, "epoch": 18.1751179245283, "percentage": 90.88, "elapsed_time": "1:00:40", "remaining_time": "0:06:05", "throughput": 5531.36, "total_tokens": 20138840} +{"current_steps": 30830, "total_steps": 33920, "loss": 0.3112, "lr": 2.5082804055384214e-07, "epoch": 18.17806603773585, "percentage": 90.89, "elapsed_time": "1:00:41", "remaining_time": "0:06:04", "throughput": 5531.33, "total_tokens": 20141528} +{"current_steps": 30835, "total_steps": 33920, "loss": 0.2783, "lr": 2.50024044438022e-07, "epoch": 18.181014150943398, "percentage": 90.91, "elapsed_time": "1:00:41", "remaining_time": "0:06:04", "throughput": 5531.21, "total_tokens": 20144248} +{"current_steps": 30840, "total_steps": 33920, "loss": 0.2228, "lr": 2.492213058988069e-07, "epoch": 18.183962264150942, "percentage": 90.92, "elapsed_time": "1:00:42", "remaining_time": "0:06:03", "throughput": 5531.31, "total_tokens": 20148024} +{"current_steps": 30845, "total_steps": 33920, "loss": 0.4267, "lr": 2.4841982514872633e-07, "epoch": 18.18691037735849, "percentage": 90.93, "elapsed_time": "1:00:43", "remaining_time": "0:06:03", "throughput": 5531.34, "total_tokens": 20151160} +{"current_steps": 30850, "total_steps": 33920, "loss": 0.3973, "lr": 2.4761960239997497e-07, "epoch": 18.18985849056604, "percentage": 90.95, "elapsed_time": "1:00:43", "remaining_time": "0:06:02", "throughput": 5531.22, "total_tokens": 20153624} +{"current_steps": 30855, "total_steps": 33920, "loss": 0.3065, "lr": 2.4682063786441556e-07, "epoch": 18.192806603773583, "percentage": 90.96, "elapsed_time": "1:00:44", "remaining_time": "0:06:01", "throughput": 5531.29, "total_tokens": 20156952} +{"current_steps": 30860, "total_steps": 33920, "loss": 0.3691, "lr": 2.460229317535778e-07, "epoch": 18.19575471698113, "percentage": 90.98, "elapsed_time": "1:00:44", "remaining_time": "0:06:01", "throughput": 5531.38, "total_tokens": 20160216} +{"current_steps": 30865, "total_steps": 33920, "loss": 0.5204, "lr": 2.4522648427865725e-07, "epoch": 18.19870283018868, "percentage": 90.99, "elapsed_time": "1:00:45", "remaining_time": "0:06:00", "throughput": 5531.35, "total_tokens": 20163480} +{"current_steps": 30870, "total_steps": 33920, "loss": 0.3327, "lr": 2.444312956505163e-07, "epoch": 18.201650943396228, "percentage": 91.01, "elapsed_time": "1:00:45", "remaining_time": "0:06:00", "throughput": 5531.24, "total_tokens": 20165944} +{"current_steps": 30875, "total_steps": 33920, "loss": 0.3257, "lr": 2.4363736607968537e-07, "epoch": 18.204599056603772, "percentage": 91.02, "elapsed_time": "1:00:46", "remaining_time": "0:05:59", "throughput": 5531.38, "total_tokens": 20170296} +{"current_steps": 30880, "total_steps": 33920, "loss": 0.2367, "lr": 2.428446957763608e-07, "epoch": 18.20754716981132, "percentage": 91.04, "elapsed_time": "1:00:47", "remaining_time": "0:05:59", "throughput": 5531.29, "total_tokens": 20172728} +{"current_steps": 30885, "total_steps": 33920, "loss": 0.237, "lr": 2.4205328495040535e-07, "epoch": 18.21049528301887, "percentage": 91.05, "elapsed_time": "1:00:47", "remaining_time": "0:05:58", "throughput": 5531.25, "total_tokens": 20175192} +{"current_steps": 30890, "total_steps": 33920, "loss": 0.3524, "lr": 2.412631338113486e-07, "epoch": 18.213443396226417, "percentage": 91.07, "elapsed_time": "1:00:48", "remaining_time": "0:05:57", "throughput": 5531.49, "total_tokens": 20179608} +{"current_steps": 30895, "total_steps": 33920, "loss": 0.4017, "lr": 2.404742425683848e-07, "epoch": 18.21639150943396, "percentage": 91.08, "elapsed_time": "1:00:48", "remaining_time": "0:05:57", "throughput": 5531.72, "total_tokens": 20183992} +{"current_steps": 30900, "total_steps": 33920, "loss": 0.3688, "lr": 2.3968661143037864e-07, "epoch": 18.21933962264151, "percentage": 91.1, "elapsed_time": "1:00:49", "remaining_time": "0:05:56", "throughput": 5531.79, "total_tokens": 20187704} +{"current_steps": 30905, "total_steps": 33920, "loss": 0.3638, "lr": 2.3890024060585823e-07, "epoch": 18.222287735849058, "percentage": 91.11, "elapsed_time": "1:00:49", "remaining_time": "0:05:56", "throughput": 5531.72, "total_tokens": 20190296} +{"current_steps": 30910, "total_steps": 33920, "loss": 0.295, "lr": 2.3811513030301826e-07, "epoch": 18.225235849056602, "percentage": 91.13, "elapsed_time": "1:00:50", "remaining_time": "0:05:55", "throughput": 5531.71, "total_tokens": 20193208} +{"current_steps": 30915, "total_steps": 33920, "loss": 0.4325, "lr": 2.373312807297201e-07, "epoch": 18.22818396226415, "percentage": 91.14, "elapsed_time": "1:00:51", "remaining_time": "0:05:54", "throughput": 5531.89, "total_tokens": 20197400} +{"current_steps": 30920, "total_steps": 33920, "loss": 0.2354, "lr": 2.3654869209349007e-07, "epoch": 18.2311320754717, "percentage": 91.16, "elapsed_time": "1:00:51", "remaining_time": "0:05:54", "throughput": 5531.85, "total_tokens": 20200248} +{"current_steps": 30925, "total_steps": 33920, "loss": 0.3109, "lr": 2.357673646015246e-07, "epoch": 18.234080188679247, "percentage": 91.17, "elapsed_time": "1:00:52", "remaining_time": "0:05:53", "throughput": 5531.89, "total_tokens": 20204184} +{"current_steps": 30930, "total_steps": 33920, "loss": 0.3243, "lr": 2.3498729846068103e-07, "epoch": 18.23702830188679, "percentage": 91.19, "elapsed_time": "1:00:52", "remaining_time": "0:05:53", "throughput": 5532.01, "total_tokens": 20207704} +{"current_steps": 30935, "total_steps": 33920, "loss": 0.296, "lr": 2.342084938774869e-07, "epoch": 18.23997641509434, "percentage": 91.2, "elapsed_time": "1:00:53", "remaining_time": "0:05:52", "throughput": 5532.11, "total_tokens": 20211352} +{"current_steps": 30940, "total_steps": 33920, "loss": 0.3221, "lr": 2.334309510581334e-07, "epoch": 18.242924528301888, "percentage": 91.21, "elapsed_time": "1:00:54", "remaining_time": "0:05:51", "throughput": 5532.17, "total_tokens": 20214840} +{"current_steps": 30945, "total_steps": 33920, "loss": 0.2627, "lr": 2.3265467020847864e-07, "epoch": 18.245872641509433, "percentage": 91.23, "elapsed_time": "1:00:54", "remaining_time": "0:05:51", "throughput": 5532.25, "total_tokens": 20218296} +{"current_steps": 30950, "total_steps": 33920, "loss": 0.3625, "lr": 2.31879651534046e-07, "epoch": 18.24882075471698, "percentage": 91.24, "elapsed_time": "1:00:55", "remaining_time": "0:05:50", "throughput": 5532.45, "total_tokens": 20222712} +{"current_steps": 30955, "total_steps": 33920, "loss": 0.2936, "lr": 2.311058952400247e-07, "epoch": 18.25176886792453, "percentage": 91.26, "elapsed_time": "1:00:55", "remaining_time": "0:05:50", "throughput": 5532.43, "total_tokens": 20225368} +{"current_steps": 30960, "total_steps": 33920, "loss": 0.2555, "lr": 2.3033340153127026e-07, "epoch": 18.254716981132077, "percentage": 91.27, "elapsed_time": "1:00:56", "remaining_time": "0:05:49", "throughput": 5532.41, "total_tokens": 20228440} +{"current_steps": 30965, "total_steps": 33920, "loss": 0.494, "lr": 2.295621706123041e-07, "epoch": 18.25766509433962, "percentage": 91.29, "elapsed_time": "1:00:56", "remaining_time": "0:05:48", "throughput": 5532.3, "total_tokens": 20230680} +{"current_steps": 30970, "total_steps": 33920, "loss": 0.3115, "lr": 2.287922026873135e-07, "epoch": 18.26061320754717, "percentage": 91.3, "elapsed_time": "1:00:57", "remaining_time": "0:05:48", "throughput": 5532.5, "total_tokens": 20234744} +{"current_steps": 30975, "total_steps": 33920, "loss": 0.3223, "lr": 2.2802349796014923e-07, "epoch": 18.263561320754718, "percentage": 91.32, "elapsed_time": "1:00:57", "remaining_time": "0:05:47", "throughput": 5532.64, "total_tokens": 20238200} +{"current_steps": 30980, "total_steps": 33920, "loss": 0.3921, "lr": 2.2725605663433013e-07, "epoch": 18.266509433962263, "percentage": 91.33, "elapsed_time": "1:00:58", "remaining_time": "0:05:47", "throughput": 5532.66, "total_tokens": 20241016} +{"current_steps": 30985, "total_steps": 33920, "loss": 0.3704, "lr": 2.264898789130393e-07, "epoch": 18.26945754716981, "percentage": 91.35, "elapsed_time": "1:00:59", "remaining_time": "0:05:46", "throughput": 5532.82, "total_tokens": 20244920} +{"current_steps": 30990, "total_steps": 33920, "loss": 0.3424, "lr": 2.2572496499912554e-07, "epoch": 18.27240566037736, "percentage": 91.36, "elapsed_time": "1:00:59", "remaining_time": "0:05:45", "throughput": 5532.75, "total_tokens": 20247352} +{"current_steps": 30995, "total_steps": 33920, "loss": 0.2966, "lr": 2.2496131509510354e-07, "epoch": 18.275353773584907, "percentage": 91.38, "elapsed_time": "1:01:00", "remaining_time": "0:05:45", "throughput": 5532.67, "total_tokens": 20249752} +{"current_steps": 31000, "total_steps": 33920, "loss": 0.2957, "lr": 2.2419892940315268e-07, "epoch": 18.278301886792452, "percentage": 91.39, "elapsed_time": "1:01:00", "remaining_time": "0:05:44", "throughput": 5532.7, "total_tokens": 20252760} +{"current_steps": 31005, "total_steps": 33920, "loss": 0.2267, "lr": 2.2343780812511819e-07, "epoch": 18.28125, "percentage": 91.41, "elapsed_time": "1:01:01", "remaining_time": "0:05:44", "throughput": 5532.68, "total_tokens": 20255480} +{"current_steps": 31010, "total_steps": 33920, "loss": 0.2462, "lr": 2.2267795146250936e-07, "epoch": 18.284198113207548, "percentage": 91.42, "elapsed_time": "1:01:01", "remaining_time": "0:05:43", "throughput": 5532.62, "total_tokens": 20258264} +{"current_steps": 31015, "total_steps": 33920, "loss": 0.3095, "lr": 2.2191935961650146e-07, "epoch": 18.287146226415093, "percentage": 91.44, "elapsed_time": "1:01:02", "remaining_time": "0:05:43", "throughput": 5532.74, "total_tokens": 20261624} +{"current_steps": 31020, "total_steps": 33920, "loss": 0.4328, "lr": 2.2116203278793603e-07, "epoch": 18.29009433962264, "percentage": 91.45, "elapsed_time": "1:01:02", "remaining_time": "0:05:42", "throughput": 5532.65, "total_tokens": 20264280} +{"current_steps": 31025, "total_steps": 33920, "loss": 0.3596, "lr": 2.2040597117731766e-07, "epoch": 18.29304245283019, "percentage": 91.47, "elapsed_time": "1:01:03", "remaining_time": "0:05:41", "throughput": 5532.64, "total_tokens": 20267032} +{"current_steps": 31030, "total_steps": 33920, "loss": 0.2367, "lr": 2.1965117498481793e-07, "epoch": 18.295990566037737, "percentage": 91.48, "elapsed_time": "1:01:03", "remaining_time": "0:05:41", "throughput": 5532.67, "total_tokens": 20270232} +{"current_steps": 31035, "total_steps": 33920, "loss": 0.2271, "lr": 2.188976444102714e-07, "epoch": 18.298938679245282, "percentage": 91.49, "elapsed_time": "1:01:04", "remaining_time": "0:05:40", "throughput": 5532.67, "total_tokens": 20273208} +{"current_steps": 31040, "total_steps": 33920, "loss": 0.3437, "lr": 2.181453796531796e-07, "epoch": 18.30188679245283, "percentage": 91.51, "elapsed_time": "1:01:04", "remaining_time": "0:05:40", "throughput": 5532.71, "total_tokens": 20276312} +{"current_steps": 31045, "total_steps": 33920, "loss": 0.3939, "lr": 2.1739438091270658e-07, "epoch": 18.30483490566038, "percentage": 91.52, "elapsed_time": "1:01:05", "remaining_time": "0:05:39", "throughput": 5532.77, "total_tokens": 20279480} +{"current_steps": 31050, "total_steps": 33920, "loss": 0.3021, "lr": 2.1664464838768329e-07, "epoch": 18.307783018867923, "percentage": 91.54, "elapsed_time": "1:01:05", "remaining_time": "0:05:38", "throughput": 5532.8, "total_tokens": 20282648} +{"current_steps": 31055, "total_steps": 33920, "loss": 0.3435, "lr": 2.1589618227660426e-07, "epoch": 18.31073113207547, "percentage": 91.55, "elapsed_time": "1:01:06", "remaining_time": "0:05:38", "throughput": 5532.93, "total_tokens": 20286424} +{"current_steps": 31060, "total_steps": 33920, "loss": 0.276, "lr": 2.151489827776293e-07, "epoch": 18.31367924528302, "percentage": 91.57, "elapsed_time": "1:01:07", "remaining_time": "0:05:37", "throughput": 5533.09, "total_tokens": 20290104} +{"current_steps": 31065, "total_steps": 33920, "loss": 0.5679, "lr": 2.1440305008858298e-07, "epoch": 18.316627358490567, "percentage": 91.58, "elapsed_time": "1:01:07", "remaining_time": "0:05:37", "throughput": 5533.12, "total_tokens": 20293752} +{"current_steps": 31070, "total_steps": 33920, "loss": 0.4189, "lr": 2.1365838440695397e-07, "epoch": 18.319575471698112, "percentage": 91.6, "elapsed_time": "1:01:08", "remaining_time": "0:05:36", "throughput": 5533.18, "total_tokens": 20296920} +{"current_steps": 31075, "total_steps": 33920, "loss": 0.4748, "lr": 2.129149859298957e-07, "epoch": 18.32252358490566, "percentage": 91.61, "elapsed_time": "1:01:09", "remaining_time": "0:05:35", "throughput": 5533.45, "total_tokens": 20303928} +{"current_steps": 31080, "total_steps": 33920, "loss": 0.3099, "lr": 2.1217285485422622e-07, "epoch": 18.32547169811321, "percentage": 91.63, "elapsed_time": "1:01:09", "remaining_time": "0:05:35", "throughput": 5533.62, "total_tokens": 20308248} +{"current_steps": 31085, "total_steps": 33920, "loss": 0.3265, "lr": 2.114319913764268e-07, "epoch": 18.328419811320753, "percentage": 91.64, "elapsed_time": "1:01:10", "remaining_time": "0:05:34", "throughput": 5533.75, "total_tokens": 20311992} +{"current_steps": 31090, "total_steps": 33920, "loss": 0.3037, "lr": 2.10692395692646e-07, "epoch": 18.3313679245283, "percentage": 91.66, "elapsed_time": "1:01:11", "remaining_time": "0:05:34", "throughput": 5533.85, "total_tokens": 20315832} +{"current_steps": 31095, "total_steps": 33920, "loss": 0.2935, "lr": 2.0995406799869444e-07, "epoch": 18.33431603773585, "percentage": 91.67, "elapsed_time": "1:01:11", "remaining_time": "0:05:33", "throughput": 5533.87, "total_tokens": 20318648} +{"current_steps": 31100, "total_steps": 33920, "loss": 0.3422, "lr": 2.0921700849004743e-07, "epoch": 18.337264150943398, "percentage": 91.69, "elapsed_time": "1:01:12", "remaining_time": "0:05:32", "throughput": 5533.85, "total_tokens": 20321304} +{"current_steps": 31105, "total_steps": 33920, "loss": 0.3584, "lr": 2.084812173618439e-07, "epoch": 18.340212264150942, "percentage": 91.7, "elapsed_time": "1:01:12", "remaining_time": "0:05:32", "throughput": 5534.02, "total_tokens": 20325048} +{"current_steps": 31110, "total_steps": 33920, "loss": 0.3868, "lr": 2.0774669480888853e-07, "epoch": 18.34316037735849, "percentage": 91.72, "elapsed_time": "1:01:13", "remaining_time": "0:05:31", "throughput": 5533.99, "total_tokens": 20327736} +{"current_steps": 31115, "total_steps": 33920, "loss": 0.2481, "lr": 2.0701344102564912e-07, "epoch": 18.34610849056604, "percentage": 91.73, "elapsed_time": "1:01:13", "remaining_time": "0:05:31", "throughput": 5534.18, "total_tokens": 20331800} +{"current_steps": 31120, "total_steps": 33920, "loss": 0.339, "lr": 2.062814562062576e-07, "epoch": 18.349056603773583, "percentage": 91.75, "elapsed_time": "1:01:14", "remaining_time": "0:05:30", "throughput": 5534.2, "total_tokens": 20335192} +{"current_steps": 31125, "total_steps": 33920, "loss": 0.2922, "lr": 2.0555074054451063e-07, "epoch": 18.35200471698113, "percentage": 91.76, "elapsed_time": "1:01:15", "remaining_time": "0:05:30", "throughput": 5534.3, "total_tokens": 20339032} +{"current_steps": 31130, "total_steps": 33920, "loss": 0.2247, "lr": 2.0482129423386843e-07, "epoch": 18.35495283018868, "percentage": 91.77, "elapsed_time": "1:01:15", "remaining_time": "0:05:29", "throughput": 5534.28, "total_tokens": 20341720} +{"current_steps": 31135, "total_steps": 33920, "loss": 0.3374, "lr": 2.040931174674543e-07, "epoch": 18.357900943396228, "percentage": 91.79, "elapsed_time": "1:01:16", "remaining_time": "0:05:28", "throughput": 5534.35, "total_tokens": 20345016} +{"current_steps": 31140, "total_steps": 33920, "loss": 0.4572, "lr": 2.0336621043805682e-07, "epoch": 18.360849056603772, "percentage": 91.8, "elapsed_time": "1:01:16", "remaining_time": "0:05:28", "throughput": 5534.36, "total_tokens": 20347992} +{"current_steps": 31145, "total_steps": 33920, "loss": 0.2992, "lr": 2.0264057333812704e-07, "epoch": 18.36379716981132, "percentage": 91.82, "elapsed_time": "1:01:17", "remaining_time": "0:05:27", "throughput": 5534.41, "total_tokens": 20350872} +{"current_steps": 31150, "total_steps": 33920, "loss": 0.4746, "lr": 2.0191620635978127e-07, "epoch": 18.36674528301887, "percentage": 91.83, "elapsed_time": "1:01:17", "remaining_time": "0:05:27", "throughput": 5534.53, "total_tokens": 20355032} +{"current_steps": 31155, "total_steps": 33920, "loss": 0.2953, "lr": 2.0119310969479833e-07, "epoch": 18.369693396226417, "percentage": 91.85, "elapsed_time": "1:01:18", "remaining_time": "0:05:26", "throughput": 5534.45, "total_tokens": 20357432} +{"current_steps": 31160, "total_steps": 33920, "loss": 0.2763, "lr": 2.004712835346212e-07, "epoch": 18.37264150943396, "percentage": 91.86, "elapsed_time": "1:01:18", "remaining_time": "0:05:25", "throughput": 5534.56, "total_tokens": 20361144} +{"current_steps": 31165, "total_steps": 33920, "loss": 0.2417, "lr": 1.99750728070357e-07, "epoch": 18.37558962264151, "percentage": 91.88, "elapsed_time": "1:01:19", "remaining_time": "0:05:25", "throughput": 5534.54, "total_tokens": 20363768} +{"current_steps": 31170, "total_steps": 33920, "loss": 0.2397, "lr": 1.9903144349277536e-07, "epoch": 18.378537735849058, "percentage": 91.89, "elapsed_time": "1:01:19", "remaining_time": "0:05:24", "throughput": 5534.52, "total_tokens": 20366584} +{"current_steps": 31175, "total_steps": 33920, "loss": 0.2956, "lr": 1.983134299923095e-07, "epoch": 18.381485849056602, "percentage": 91.91, "elapsed_time": "1:01:20", "remaining_time": "0:05:24", "throughput": 5534.37, "total_tokens": 20369176} +{"current_steps": 31180, "total_steps": 33920, "loss": 0.4118, "lr": 1.9759668775905737e-07, "epoch": 18.38443396226415, "percentage": 91.92, "elapsed_time": "1:01:21", "remaining_time": "0:05:23", "throughput": 5534.3, "total_tokens": 20371896} +{"current_steps": 31185, "total_steps": 33920, "loss": 0.4216, "lr": 1.9688121698277995e-07, "epoch": 18.3873820754717, "percentage": 91.94, "elapsed_time": "1:01:21", "remaining_time": "0:05:22", "throughput": 5534.33, "total_tokens": 20375352} +{"current_steps": 31190, "total_steps": 33920, "loss": 0.2416, "lr": 1.9616701785290015e-07, "epoch": 18.390330188679247, "percentage": 91.95, "elapsed_time": "1:01:22", "remaining_time": "0:05:22", "throughput": 5534.41, "total_tokens": 20378744} +{"current_steps": 31195, "total_steps": 33920, "loss": 0.2906, "lr": 1.954540905585056e-07, "epoch": 18.39327830188679, "percentage": 91.97, "elapsed_time": "1:01:22", "remaining_time": "0:05:21", "throughput": 5534.32, "total_tokens": 20381496} +{"current_steps": 31200, "total_steps": 33920, "loss": 0.2212, "lr": 1.9474243528834757e-07, "epoch": 18.39622641509434, "percentage": 91.98, "elapsed_time": "1:01:23", "remaining_time": "0:05:21", "throughput": 5534.28, "total_tokens": 20384376} +{"current_steps": 31205, "total_steps": 33920, "loss": 0.2959, "lr": 1.9403205223083866e-07, "epoch": 18.399174528301888, "percentage": 92.0, "elapsed_time": "1:01:23", "remaining_time": "0:05:20", "throughput": 5534.28, "total_tokens": 20387448} +{"current_steps": 31210, "total_steps": 33920, "loss": 0.3945, "lr": 1.9332294157405619e-07, "epoch": 18.402122641509433, "percentage": 92.01, "elapsed_time": "1:01:24", "remaining_time": "0:05:19", "throughput": 5534.35, "total_tokens": 20391160} +{"current_steps": 31215, "total_steps": 33920, "loss": 0.3526, "lr": 1.926151035057411e-07, "epoch": 18.40507075471698, "percentage": 92.03, "elapsed_time": "1:01:24", "remaining_time": "0:05:19", "throughput": 5534.32, "total_tokens": 20393784} +{"current_steps": 31220, "total_steps": 33920, "loss": 0.5003, "lr": 1.9190853821329626e-07, "epoch": 18.40801886792453, "percentage": 92.04, "elapsed_time": "1:01:25", "remaining_time": "0:05:18", "throughput": 5534.32, "total_tokens": 20396632} +{"current_steps": 31225, "total_steps": 33920, "loss": 0.3718, "lr": 1.9120324588378757e-07, "epoch": 18.410966981132077, "percentage": 92.05, "elapsed_time": "1:01:25", "remaining_time": "0:05:18", "throughput": 5534.35, "total_tokens": 20399448} +{"current_steps": 31230, "total_steps": 33920, "loss": 0.3154, "lr": 1.9049922670394461e-07, "epoch": 18.41391509433962, "percentage": 92.07, "elapsed_time": "1:01:26", "remaining_time": "0:05:17", "throughput": 5534.47, "total_tokens": 20402840} +{"current_steps": 31235, "total_steps": 33920, "loss": 0.3012, "lr": 1.897964808601588e-07, "epoch": 18.41686320754717, "percentage": 92.08, "elapsed_time": "1:01:27", "remaining_time": "0:05:16", "throughput": 5534.51, "total_tokens": 20405944} +{"current_steps": 31240, "total_steps": 33920, "loss": 0.2841, "lr": 1.8909500853848517e-07, "epoch": 18.419811320754718, "percentage": 92.1, "elapsed_time": "1:01:27", "remaining_time": "0:05:16", "throughput": 5534.62, "total_tokens": 20409464} +{"current_steps": 31245, "total_steps": 33920, "loss": 0.3617, "lr": 1.8839480992464243e-07, "epoch": 18.422759433962263, "percentage": 92.11, "elapsed_time": "1:01:28", "remaining_time": "0:05:15", "throughput": 5534.58, "total_tokens": 20412120} +{"current_steps": 31250, "total_steps": 33920, "loss": 0.335, "lr": 1.8769588520401005e-07, "epoch": 18.42570754716981, "percentage": 92.13, "elapsed_time": "1:01:28", "remaining_time": "0:05:15", "throughput": 5534.66, "total_tokens": 20415672} +{"current_steps": 31255, "total_steps": 33920, "loss": 0.3712, "lr": 1.8699823456163279e-07, "epoch": 18.42865566037736, "percentage": 92.14, "elapsed_time": "1:01:29", "remaining_time": "0:05:14", "throughput": 5534.76, "total_tokens": 20419160} +{"current_steps": 31260, "total_steps": 33920, "loss": 0.184, "lr": 1.8630185818221514e-07, "epoch": 18.431603773584907, "percentage": 92.16, "elapsed_time": "1:01:30", "remaining_time": "0:05:14", "throughput": 5535.07, "total_tokens": 20425336} +{"current_steps": 31265, "total_steps": 33920, "loss": 0.3574, "lr": 1.856067562501268e-07, "epoch": 18.434551886792452, "percentage": 92.17, "elapsed_time": "1:01:30", "remaining_time": "0:05:13", "throughput": 5534.98, "total_tokens": 20427896} +{"current_steps": 31270, "total_steps": 33920, "loss": 0.3405, "lr": 1.8491292894939837e-07, "epoch": 18.4375, "percentage": 92.19, "elapsed_time": "1:01:31", "remaining_time": "0:05:12", "throughput": 5535.13, "total_tokens": 20432120} +{"current_steps": 31275, "total_steps": 33920, "loss": 0.3369, "lr": 1.8422037646372405e-07, "epoch": 18.440448113207548, "percentage": 92.2, "elapsed_time": "1:01:31", "remaining_time": "0:05:12", "throughput": 5535.17, "total_tokens": 20434936} +{"current_steps": 31280, "total_steps": 33920, "loss": 0.4829, "lr": 1.8352909897645989e-07, "epoch": 18.443396226415093, "percentage": 92.22, "elapsed_time": "1:01:32", "remaining_time": "0:05:11", "throughput": 5535.27, "total_tokens": 20438136} +{"current_steps": 31285, "total_steps": 33920, "loss": 0.4015, "lr": 1.8283909667062448e-07, "epoch": 18.44634433962264, "percentage": 92.23, "elapsed_time": "1:01:32", "remaining_time": "0:05:11", "throughput": 5535.3, "total_tokens": 20441016} +{"current_steps": 31290, "total_steps": 33920, "loss": 0.3334, "lr": 1.82150369728899e-07, "epoch": 18.44929245283019, "percentage": 92.25, "elapsed_time": "1:01:33", "remaining_time": "0:05:10", "throughput": 5535.35, "total_tokens": 20444056} +{"current_steps": 31295, "total_steps": 33920, "loss": 0.3042, "lr": 1.814629183336275e-07, "epoch": 18.452240566037737, "percentage": 92.26, "elapsed_time": "1:01:33", "remaining_time": "0:05:09", "throughput": 5535.49, "total_tokens": 20447608} +{"current_steps": 31300, "total_steps": 33920, "loss": 0.3337, "lr": 1.807767426668139e-07, "epoch": 18.455188679245282, "percentage": 92.28, "elapsed_time": "1:01:34", "remaining_time": "0:05:09", "throughput": 5535.58, "total_tokens": 20451672} +{"current_steps": 31305, "total_steps": 33920, "loss": 0.2794, "lr": 1.8009184291012783e-07, "epoch": 18.45813679245283, "percentage": 92.29, "elapsed_time": "1:01:35", "remaining_time": "0:05:08", "throughput": 5535.46, "total_tokens": 20454200} +{"current_steps": 31310, "total_steps": 33920, "loss": 0.5115, "lr": 1.7940821924489926e-07, "epoch": 18.46108490566038, "percentage": 92.31, "elapsed_time": "1:01:35", "remaining_time": "0:05:08", "throughput": 5535.53, "total_tokens": 20457464} +{"current_steps": 31315, "total_steps": 33920, "loss": 0.4477, "lr": 1.7872587185212009e-07, "epoch": 18.464033018867923, "percentage": 92.32, "elapsed_time": "1:01:36", "remaining_time": "0:05:07", "throughput": 5535.56, "total_tokens": 20460472} +{"current_steps": 31320, "total_steps": 33920, "loss": 0.3751, "lr": 1.7804480091244524e-07, "epoch": 18.46698113207547, "percentage": 92.33, "elapsed_time": "1:01:36", "remaining_time": "0:05:06", "throughput": 5535.6, "total_tokens": 20463384} +{"current_steps": 31325, "total_steps": 33920, "loss": 0.205, "lr": 1.7736500660619104e-07, "epoch": 18.46992924528302, "percentage": 92.35, "elapsed_time": "1:01:37", "remaining_time": "0:05:06", "throughput": 5535.71, "total_tokens": 20466744} +{"current_steps": 31330, "total_steps": 33920, "loss": 0.2909, "lr": 1.766864891133352e-07, "epoch": 18.472877358490567, "percentage": 92.36, "elapsed_time": "1:01:37", "remaining_time": "0:05:05", "throughput": 5535.78, "total_tokens": 20470072} +{"current_steps": 31335, "total_steps": 33920, "loss": 0.3245, "lr": 1.7600924861351843e-07, "epoch": 18.475825471698112, "percentage": 92.38, "elapsed_time": "1:01:38", "remaining_time": "0:05:05", "throughput": 5535.82, "total_tokens": 20473208} +{"current_steps": 31340, "total_steps": 33920, "loss": 0.33, "lr": 1.7533328528604398e-07, "epoch": 18.47877358490566, "percentage": 92.39, "elapsed_time": "1:01:39", "remaining_time": "0:05:04", "throughput": 5536.21, "total_tokens": 20481496} +{"current_steps": 31345, "total_steps": 33920, "loss": 0.3001, "lr": 1.746585993098754e-07, "epoch": 18.48172169811321, "percentage": 92.41, "elapsed_time": "1:01:40", "remaining_time": "0:05:03", "throughput": 5536.15, "total_tokens": 20484088} +{"current_steps": 31350, "total_steps": 33920, "loss": 0.3188, "lr": 1.7398519086363864e-07, "epoch": 18.484669811320753, "percentage": 92.42, "elapsed_time": "1:01:40", "remaining_time": "0:05:03", "throughput": 5536.13, "total_tokens": 20487160} +{"current_steps": 31355, "total_steps": 33920, "loss": 0.234, "lr": 1.733130601256211e-07, "epoch": 18.4876179245283, "percentage": 92.44, "elapsed_time": "1:01:41", "remaining_time": "0:05:02", "throughput": 5536.12, "total_tokens": 20489752} +{"current_steps": 31360, "total_steps": 33920, "loss": 0.2364, "lr": 1.7264220727377323e-07, "epoch": 18.49056603773585, "percentage": 92.45, "elapsed_time": "1:01:41", "remaining_time": "0:05:02", "throughput": 5536.17, "total_tokens": 20492632} +{"current_steps": 31365, "total_steps": 33920, "loss": 0.4646, "lr": 1.7197263248570517e-07, "epoch": 18.493514150943398, "percentage": 92.47, "elapsed_time": "1:01:42", "remaining_time": "0:05:01", "throughput": 5536.15, "total_tokens": 20495672} +{"current_steps": 31370, "total_steps": 33920, "loss": 0.2952, "lr": 1.7130433593869124e-07, "epoch": 18.496462264150942, "percentage": 92.48, "elapsed_time": "1:01:42", "remaining_time": "0:05:00", "throughput": 5536.16, "total_tokens": 20498552} +{"current_steps": 31375, "total_steps": 33920, "loss": 0.3421, "lr": 1.706373178096643e-07, "epoch": 18.49941037735849, "percentage": 92.5, "elapsed_time": "1:01:43", "remaining_time": "0:05:00", "throughput": 5536.2, "total_tokens": 20501624} +{"current_steps": 31380, "total_steps": 33920, "loss": 0.3465, "lr": 1.6997157827522092e-07, "epoch": 18.50235849056604, "percentage": 92.51, "elapsed_time": "1:01:43", "remaining_time": "0:04:59", "throughput": 5536.23, "total_tokens": 20504824} +{"current_steps": 31385, "total_steps": 33920, "loss": 0.2722, "lr": 1.6930711751161843e-07, "epoch": 18.505306603773583, "percentage": 92.53, "elapsed_time": "1:01:44", "remaining_time": "0:04:59", "throughput": 5536.26, "total_tokens": 20507704} +{"current_steps": 31390, "total_steps": 33920, "loss": 0.374, "lr": 1.6864393569477556e-07, "epoch": 18.50825471698113, "percentage": 92.54, "elapsed_time": "1:01:44", "remaining_time": "0:04:58", "throughput": 5536.34, "total_tokens": 20511320} +{"current_steps": 31395, "total_steps": 33920, "loss": 0.3408, "lr": 1.6798203300027295e-07, "epoch": 18.51120283018868, "percentage": 92.56, "elapsed_time": "1:01:45", "remaining_time": "0:04:58", "throughput": 5536.5, "total_tokens": 20515448} +{"current_steps": 31400, "total_steps": 33920, "loss": 0.2598, "lr": 1.6732140960335152e-07, "epoch": 18.514150943396228, "percentage": 92.57, "elapsed_time": "1:01:46", "remaining_time": "0:04:57", "throughput": 5536.5, "total_tokens": 20518488} +{"current_steps": 31405, "total_steps": 33920, "loss": 0.3018, "lr": 1.666620656789153e-07, "epoch": 18.517099056603772, "percentage": 92.59, "elapsed_time": "1:01:46", "remaining_time": "0:04:56", "throughput": 5536.48, "total_tokens": 20521208} +{"current_steps": 31410, "total_steps": 33920, "loss": 0.2608, "lr": 1.660040014015274e-07, "epoch": 18.52004716981132, "percentage": 92.6, "elapsed_time": "1:01:47", "remaining_time": "0:04:56", "throughput": 5536.61, "total_tokens": 20524696} +{"current_steps": 31415, "total_steps": 33920, "loss": 0.3264, "lr": 1.6534721694541344e-07, "epoch": 18.52299528301887, "percentage": 92.61, "elapsed_time": "1:01:47", "remaining_time": "0:04:55", "throughput": 5536.69, "total_tokens": 20527864} +{"current_steps": 31420, "total_steps": 33920, "loss": 0.2961, "lr": 1.6469171248445993e-07, "epoch": 18.525943396226417, "percentage": 92.63, "elapsed_time": "1:01:48", "remaining_time": "0:04:55", "throughput": 5536.86, "total_tokens": 20531832} +{"current_steps": 31425, "total_steps": 33920, "loss": 0.3037, "lr": 1.6403748819221464e-07, "epoch": 18.52889150943396, "percentage": 92.64, "elapsed_time": "1:01:48", "remaining_time": "0:04:54", "throughput": 5536.99, "total_tokens": 20535384} +{"current_steps": 31430, "total_steps": 33920, "loss": 0.3437, "lr": 1.6338454424188632e-07, "epoch": 18.53183962264151, "percentage": 92.66, "elapsed_time": "1:01:49", "remaining_time": "0:04:53", "throughput": 5537.07, "total_tokens": 20538552} +{"current_steps": 31435, "total_steps": 33920, "loss": 0.2886, "lr": 1.6273288080634442e-07, "epoch": 18.534787735849058, "percentage": 92.67, "elapsed_time": "1:01:49", "remaining_time": "0:04:53", "throughput": 5536.82, "total_tokens": 20541048} +{"current_steps": 31440, "total_steps": 33920, "loss": 0.3296, "lr": 1.6208249805811982e-07, "epoch": 18.537735849056602, "percentage": 92.69, "elapsed_time": "1:01:50", "remaining_time": "0:04:52", "throughput": 5536.83, "total_tokens": 20544120} +{"current_steps": 31445, "total_steps": 33920, "loss": 0.2206, "lr": 1.6143339616940423e-07, "epoch": 18.54068396226415, "percentage": 92.7, "elapsed_time": "1:01:51", "remaining_time": "0:04:52", "throughput": 5536.89, "total_tokens": 20547448} +{"current_steps": 31450, "total_steps": 33920, "loss": 0.4343, "lr": 1.6078557531205018e-07, "epoch": 18.5436320754717, "percentage": 92.72, "elapsed_time": "1:01:51", "remaining_time": "0:04:51", "throughput": 5537.12, "total_tokens": 20553400} +{"current_steps": 31455, "total_steps": 33920, "loss": 0.2838, "lr": 1.601390356575705e-07, "epoch": 18.546580188679247, "percentage": 92.73, "elapsed_time": "1:01:52", "remaining_time": "0:04:50", "throughput": 5537.17, "total_tokens": 20556440} +{"current_steps": 31460, "total_steps": 33920, "loss": 0.287, "lr": 1.5949377737713988e-07, "epoch": 18.54952830188679, "percentage": 92.75, "elapsed_time": "1:01:52", "remaining_time": "0:04:50", "throughput": 5537.24, "total_tokens": 20559416} +{"current_steps": 31465, "total_steps": 33920, "loss": 0.2835, "lr": 1.5884980064159338e-07, "epoch": 18.55247641509434, "percentage": 92.76, "elapsed_time": "1:01:53", "remaining_time": "0:04:49", "throughput": 5537.26, "total_tokens": 20562552} +{"current_steps": 31470, "total_steps": 33920, "loss": 0.2526, "lr": 1.5820710562142627e-07, "epoch": 18.555424528301888, "percentage": 92.78, "elapsed_time": "1:01:54", "remaining_time": "0:04:49", "throughput": 5537.39, "total_tokens": 20566200} +{"current_steps": 31475, "total_steps": 33920, "loss": 0.3084, "lr": 1.575656924867952e-07, "epoch": 18.558372641509433, "percentage": 92.79, "elapsed_time": "1:01:54", "remaining_time": "0:04:48", "throughput": 5537.42, "total_tokens": 20569048} +{"current_steps": 31480, "total_steps": 33920, "loss": 0.476, "lr": 1.5692556140751658e-07, "epoch": 18.56132075471698, "percentage": 92.81, "elapsed_time": "1:01:55", "remaining_time": "0:04:47", "throughput": 5537.43, "total_tokens": 20571768} +{"current_steps": 31485, "total_steps": 33920, "loss": 0.3126, "lr": 1.5628671255306706e-07, "epoch": 18.56426886792453, "percentage": 92.82, "elapsed_time": "1:01:55", "remaining_time": "0:04:47", "throughput": 5537.42, "total_tokens": 20574648} +{"current_steps": 31490, "total_steps": 33920, "loss": 0.3333, "lr": 1.556491460925863e-07, "epoch": 18.567216981132077, "percentage": 92.84, "elapsed_time": "1:01:56", "remaining_time": "0:04:46", "throughput": 5537.54, "total_tokens": 20579256} +{"current_steps": 31495, "total_steps": 33920, "loss": 0.2638, "lr": 1.550128621948721e-07, "epoch": 18.57016509433962, "percentage": 92.85, "elapsed_time": "1:01:56", "remaining_time": "0:04:46", "throughput": 5537.53, "total_tokens": 20582072} +{"current_steps": 31500, "total_steps": 33920, "loss": 0.3, "lr": 1.5437786102838413e-07, "epoch": 18.57311320754717, "percentage": 92.87, "elapsed_time": "1:01:57", "remaining_time": "0:04:45", "throughput": 5537.66, "total_tokens": 20585528} +{"current_steps": 31505, "total_steps": 33920, "loss": 0.4469, "lr": 1.5374414276124017e-07, "epoch": 18.576061320754718, "percentage": 92.88, "elapsed_time": "1:01:57", "remaining_time": "0:04:44", "throughput": 5537.74, "total_tokens": 20589112} +{"current_steps": 31510, "total_steps": 33920, "loss": 0.2427, "lr": 1.5311170756122095e-07, "epoch": 18.579009433962263, "percentage": 92.9, "elapsed_time": "1:01:58", "remaining_time": "0:04:44", "throughput": 5537.67, "total_tokens": 20591608} +{"current_steps": 31515, "total_steps": 33920, "loss": 0.3472, "lr": 1.5248055559576647e-07, "epoch": 18.58195754716981, "percentage": 92.91, "elapsed_time": "1:01:58", "remaining_time": "0:04:43", "throughput": 5537.71, "total_tokens": 20594456} +{"current_steps": 31520, "total_steps": 33920, "loss": 0.3443, "lr": 1.5185068703197526e-07, "epoch": 18.58490566037736, "percentage": 92.92, "elapsed_time": "1:01:59", "remaining_time": "0:04:43", "throughput": 5537.9, "total_tokens": 20599096} +{"current_steps": 31525, "total_steps": 33920, "loss": 0.2971, "lr": 1.5122210203661004e-07, "epoch": 18.587853773584907, "percentage": 92.94, "elapsed_time": "1:02:00", "remaining_time": "0:04:42", "throughput": 5537.92, "total_tokens": 20601976} +{"current_steps": 31530, "total_steps": 33920, "loss": 0.26, "lr": 1.505948007760899e-07, "epoch": 18.590801886792452, "percentage": 92.95, "elapsed_time": "1:02:00", "remaining_time": "0:04:42", "throughput": 5537.92, "total_tokens": 20604760} +{"current_steps": 31535, "total_steps": 33920, "loss": 0.309, "lr": 1.4996878341649647e-07, "epoch": 18.59375, "percentage": 92.97, "elapsed_time": "1:02:01", "remaining_time": "0:04:41", "throughput": 5537.98, "total_tokens": 20608440} +{"current_steps": 31540, "total_steps": 33920, "loss": 0.3475, "lr": 1.493440501235699e-07, "epoch": 18.596698113207548, "percentage": 92.98, "elapsed_time": "1:02:01", "remaining_time": "0:04:40", "throughput": 5538.02, "total_tokens": 20611800} +{"current_steps": 31545, "total_steps": 33920, "loss": 0.3109, "lr": 1.487206010627118e-07, "epoch": 18.599646226415093, "percentage": 93.0, "elapsed_time": "1:02:02", "remaining_time": "0:04:40", "throughput": 5538.08, "total_tokens": 20615224} +{"current_steps": 31550, "total_steps": 33920, "loss": 0.2616, "lr": 1.4809843639898124e-07, "epoch": 18.60259433962264, "percentage": 93.01, "elapsed_time": "1:02:03", "remaining_time": "0:04:39", "throughput": 5538.0, "total_tokens": 20618232} +{"current_steps": 31555, "total_steps": 33920, "loss": 0.2093, "lr": 1.4747755629710093e-07, "epoch": 18.60554245283019, "percentage": 93.03, "elapsed_time": "1:02:03", "remaining_time": "0:04:39", "throughput": 5538.1, "total_tokens": 20621432} +{"current_steps": 31560, "total_steps": 33920, "loss": 0.2647, "lr": 1.4685796092145045e-07, "epoch": 18.608490566037737, "percentage": 93.04, "elapsed_time": "1:02:04", "remaining_time": "0:04:38", "throughput": 5538.1, "total_tokens": 20624504} +{"current_steps": 31565, "total_steps": 33920, "loss": 0.2534, "lr": 1.4623965043607135e-07, "epoch": 18.611438679245282, "percentage": 93.06, "elapsed_time": "1:02:04", "remaining_time": "0:04:37", "throughput": 5538.02, "total_tokens": 20626936} +{"current_steps": 31570, "total_steps": 33920, "loss": 0.3193, "lr": 1.4562262500466273e-07, "epoch": 18.61438679245283, "percentage": 93.07, "elapsed_time": "1:02:05", "remaining_time": "0:04:37", "throughput": 5538.08, "total_tokens": 20630232} +{"current_steps": 31575, "total_steps": 33920, "loss": 0.2881, "lr": 1.4500688479058556e-07, "epoch": 18.61733490566038, "percentage": 93.09, "elapsed_time": "1:02:05", "remaining_time": "0:04:36", "throughput": 5538.0, "total_tokens": 20632920} +{"current_steps": 31580, "total_steps": 33920, "loss": 0.3951, "lr": 1.4439242995685943e-07, "epoch": 18.620283018867923, "percentage": 93.1, "elapsed_time": "1:02:06", "remaining_time": "0:04:36", "throughput": 5538.06, "total_tokens": 20636024} +{"current_steps": 31585, "total_steps": 33920, "loss": 0.3748, "lr": 1.4377926066616364e-07, "epoch": 18.62323113207547, "percentage": 93.12, "elapsed_time": "1:02:06", "remaining_time": "0:04:35", "throughput": 5538.07, "total_tokens": 20639128} +{"current_steps": 31590, "total_steps": 33920, "loss": 0.4367, "lr": 1.4316737708083783e-07, "epoch": 18.62617924528302, "percentage": 93.13, "elapsed_time": "1:02:07", "remaining_time": "0:04:34", "throughput": 5538.12, "total_tokens": 20642360} +{"current_steps": 31595, "total_steps": 33920, "loss": 0.4115, "lr": 1.4255677936288127e-07, "epoch": 18.629127358490567, "percentage": 93.15, "elapsed_time": "1:02:07", "remaining_time": "0:04:34", "throughput": 5538.04, "total_tokens": 20644952} +{"current_steps": 31600, "total_steps": 33920, "loss": 0.3284, "lr": 1.4194746767395184e-07, "epoch": 18.632075471698112, "percentage": 93.16, "elapsed_time": "1:02:08", "remaining_time": "0:04:33", "throughput": 5538.11, "total_tokens": 20648280} +{"current_steps": 31605, "total_steps": 33920, "loss": 0.4703, "lr": 1.4133944217536722e-07, "epoch": 18.63502358490566, "percentage": 93.18, "elapsed_time": "1:02:08", "remaining_time": "0:04:33", "throughput": 5538.12, "total_tokens": 20651096} +{"current_steps": 31610, "total_steps": 33920, "loss": 0.3006, "lr": 1.4073270302810471e-07, "epoch": 18.63797169811321, "percentage": 93.19, "elapsed_time": "1:02:09", "remaining_time": "0:04:32", "throughput": 5538.11, "total_tokens": 20653912} +{"current_steps": 31615, "total_steps": 33920, "loss": 0.4278, "lr": 1.4012725039280084e-07, "epoch": 18.640919811320753, "percentage": 93.2, "elapsed_time": "1:02:09", "remaining_time": "0:04:31", "throughput": 5538.22, "total_tokens": 20657368} +{"current_steps": 31620, "total_steps": 33920, "loss": 0.28, "lr": 1.3952308442975292e-07, "epoch": 18.6438679245283, "percentage": 93.22, "elapsed_time": "1:02:10", "remaining_time": "0:04:31", "throughput": 5538.13, "total_tokens": 20659864} +{"current_steps": 31625, "total_steps": 33920, "loss": 0.312, "lr": 1.3892020529891637e-07, "epoch": 18.64681603773585, "percentage": 93.23, "elapsed_time": "1:02:11", "remaining_time": "0:04:30", "throughput": 5538.17, "total_tokens": 20663384} +{"current_steps": 31630, "total_steps": 33920, "loss": 0.324, "lr": 1.3831861315990514e-07, "epoch": 18.649764150943398, "percentage": 93.25, "elapsed_time": "1:02:11", "remaining_time": "0:04:30", "throughput": 5538.2, "total_tokens": 20666520} +{"current_steps": 31635, "total_steps": 33920, "loss": 0.4453, "lr": 1.377183081719935e-07, "epoch": 18.652712264150942, "percentage": 93.26, "elapsed_time": "1:02:12", "remaining_time": "0:04:29", "throughput": 5538.23, "total_tokens": 20669656} +{"current_steps": 31640, "total_steps": 33920, "loss": 0.2338, "lr": 1.3711929049411544e-07, "epoch": 18.65566037735849, "percentage": 93.28, "elapsed_time": "1:02:12", "remaining_time": "0:04:28", "throughput": 5538.14, "total_tokens": 20672344} +{"current_steps": 31645, "total_steps": 33920, "loss": 0.3784, "lr": 1.365215602848624e-07, "epoch": 18.65860849056604, "percentage": 93.29, "elapsed_time": "1:02:13", "remaining_time": "0:04:28", "throughput": 5538.16, "total_tokens": 20675288} +{"current_steps": 31650, "total_steps": 33920, "loss": 0.287, "lr": 1.3592511770248727e-07, "epoch": 18.661556603773583, "percentage": 93.31, "elapsed_time": "1:02:13", "remaining_time": "0:04:27", "throughput": 5538.09, "total_tokens": 20677880} +{"current_steps": 31655, "total_steps": 33920, "loss": 0.3621, "lr": 1.3532996290490041e-07, "epoch": 18.66450471698113, "percentage": 93.32, "elapsed_time": "1:02:14", "remaining_time": "0:04:27", "throughput": 5538.22, "total_tokens": 20681400} +{"current_steps": 31660, "total_steps": 33920, "loss": 0.223, "lr": 1.347360960496713e-07, "epoch": 18.66745283018868, "percentage": 93.34, "elapsed_time": "1:02:14", "remaining_time": "0:04:26", "throughput": 5538.25, "total_tokens": 20684792} +{"current_steps": 31665, "total_steps": 33920, "loss": 0.2312, "lr": 1.3414351729402862e-07, "epoch": 18.670400943396228, "percentage": 93.35, "elapsed_time": "1:02:15", "remaining_time": "0:04:26", "throughput": 5538.27, "total_tokens": 20688312} +{"current_steps": 31670, "total_steps": 33920, "loss": 0.3397, "lr": 1.3355222679486025e-07, "epoch": 18.673349056603772, "percentage": 93.37, "elapsed_time": "1:02:16", "remaining_time": "0:04:25", "throughput": 5538.34, "total_tokens": 20692376} +{"current_steps": 31675, "total_steps": 33920, "loss": 0.2709, "lr": 1.3296222470871367e-07, "epoch": 18.67629716981132, "percentage": 93.38, "elapsed_time": "1:02:16", "remaining_time": "0:04:24", "throughput": 5538.2, "total_tokens": 20694712} +{"current_steps": 31680, "total_steps": 33920, "loss": 0.2806, "lr": 1.3237351119179287e-07, "epoch": 18.67924528301887, "percentage": 93.4, "elapsed_time": "1:02:17", "remaining_time": "0:04:24", "throughput": 5538.16, "total_tokens": 20697464} +{"current_steps": 31685, "total_steps": 33920, "loss": 0.2944, "lr": 1.3178608639996425e-07, "epoch": 18.682193396226417, "percentage": 93.41, "elapsed_time": "1:02:17", "remaining_time": "0:04:23", "throughput": 5538.32, "total_tokens": 20701592} +{"current_steps": 31690, "total_steps": 33920, "loss": 0.3212, "lr": 1.3119995048874957e-07, "epoch": 18.68514150943396, "percentage": 93.43, "elapsed_time": "1:02:18", "remaining_time": "0:04:23", "throughput": 5538.42, "total_tokens": 20705016} +{"current_steps": 31695, "total_steps": 33920, "loss": 0.2401, "lr": 1.3061510361333186e-07, "epoch": 18.68808962264151, "percentage": 93.44, "elapsed_time": "1:02:19", "remaining_time": "0:04:22", "throughput": 5538.6, "total_tokens": 20709144} +{"current_steps": 31700, "total_steps": 33920, "loss": 0.2056, "lr": 1.3003154592855116e-07, "epoch": 18.691037735849058, "percentage": 93.46, "elapsed_time": "1:02:19", "remaining_time": "0:04:21", "throughput": 5538.72, "total_tokens": 20712504} +{"current_steps": 31705, "total_steps": 33920, "loss": 0.2382, "lr": 1.2944927758890668e-07, "epoch": 18.693985849056602, "percentage": 93.47, "elapsed_time": "1:02:20", "remaining_time": "0:04:21", "throughput": 5538.83, "total_tokens": 20715960} +{"current_steps": 31710, "total_steps": 33920, "loss": 0.2879, "lr": 1.2886829874855733e-07, "epoch": 18.69693396226415, "percentage": 93.48, "elapsed_time": "1:02:20", "remaining_time": "0:04:20", "throughput": 5538.73, "total_tokens": 20718424} +{"current_steps": 31715, "total_steps": 33920, "loss": 0.3128, "lr": 1.2828860956131894e-07, "epoch": 18.6998820754717, "percentage": 93.5, "elapsed_time": "1:02:21", "remaining_time": "0:04:20", "throughput": 5538.72, "total_tokens": 20721144} +{"current_steps": 31720, "total_steps": 33920, "loss": 0.3433, "lr": 1.2771021018066765e-07, "epoch": 18.702830188679247, "percentage": 93.51, "elapsed_time": "1:02:21", "remaining_time": "0:04:19", "throughput": 5538.7, "total_tokens": 20724056} +{"current_steps": 31725, "total_steps": 33920, "loss": 0.2725, "lr": 1.271331007597365e-07, "epoch": 18.70577830188679, "percentage": 93.53, "elapsed_time": "1:02:22", "remaining_time": "0:04:18", "throughput": 5538.78, "total_tokens": 20727224} +{"current_steps": 31730, "total_steps": 33920, "loss": 0.3016, "lr": 1.2655728145131774e-07, "epoch": 18.70872641509434, "percentage": 93.54, "elapsed_time": "1:02:22", "remaining_time": "0:04:18", "throughput": 5538.79, "total_tokens": 20730264} +{"current_steps": 31735, "total_steps": 33920, "loss": 0.3036, "lr": 1.2598275240786105e-07, "epoch": 18.711674528301888, "percentage": 93.56, "elapsed_time": "1:02:23", "remaining_time": "0:04:17", "throughput": 5538.93, "total_tokens": 20734200} +{"current_steps": 31740, "total_steps": 33920, "loss": 0.4193, "lr": 1.254095137814776e-07, "epoch": 18.714622641509433, "percentage": 93.57, "elapsed_time": "1:02:24", "remaining_time": "0:04:17", "throughput": 5539.07, "total_tokens": 20738328} +{"current_steps": 31745, "total_steps": 33920, "loss": 0.269, "lr": 1.2483756572393368e-07, "epoch": 18.71757075471698, "percentage": 93.59, "elapsed_time": "1:02:24", "remaining_time": "0:04:16", "throughput": 5539.12, "total_tokens": 20741368} +{"current_steps": 31750, "total_steps": 33920, "loss": 0.2716, "lr": 1.242669083866549e-07, "epoch": 18.72051886792453, "percentage": 93.6, "elapsed_time": "1:02:25", "remaining_time": "0:04:15", "throughput": 5539.15, "total_tokens": 20744440} +{"current_steps": 31755, "total_steps": 33920, "loss": 0.2486, "lr": 1.2369754192072537e-07, "epoch": 18.723466981132077, "percentage": 93.62, "elapsed_time": "1:02:25", "remaining_time": "0:04:15", "throughput": 5539.27, "total_tokens": 20748056} +{"current_steps": 31760, "total_steps": 33920, "loss": 0.2959, "lr": 1.231294664768873e-07, "epoch": 18.72641509433962, "percentage": 93.63, "elapsed_time": "1:02:26", "remaining_time": "0:04:14", "throughput": 5539.26, "total_tokens": 20751000} +{"current_steps": 31765, "total_steps": 33920, "loss": 0.2942, "lr": 1.225626822055409e-07, "epoch": 18.72936320754717, "percentage": 93.65, "elapsed_time": "1:02:26", "remaining_time": "0:04:14", "throughput": 5539.32, "total_tokens": 20754200} +{"current_steps": 31770, "total_steps": 33920, "loss": 0.3651, "lr": 1.2199718925674508e-07, "epoch": 18.732311320754718, "percentage": 93.66, "elapsed_time": "1:02:27", "remaining_time": "0:04:13", "throughput": 5539.41, "total_tokens": 20757688} +{"current_steps": 31775, "total_steps": 33920, "loss": 0.3429, "lr": 1.2143298778021616e-07, "epoch": 18.735259433962263, "percentage": 93.68, "elapsed_time": "1:02:27", "remaining_time": "0:04:13", "throughput": 5539.55, "total_tokens": 20761656} +{"current_steps": 31780, "total_steps": 33920, "loss": 0.3602, "lr": 1.2087007792532967e-07, "epoch": 18.73820754716981, "percentage": 93.69, "elapsed_time": "1:02:28", "remaining_time": "0:04:12", "throughput": 5539.55, "total_tokens": 20764376} +{"current_steps": 31785, "total_steps": 33920, "loss": 0.3061, "lr": 1.203084598411175e-07, "epoch": 18.74115566037736, "percentage": 93.71, "elapsed_time": "1:02:28", "remaining_time": "0:04:11", "throughput": 5539.52, "total_tokens": 20767288} +{"current_steps": 31790, "total_steps": 33920, "loss": 0.2967, "lr": 1.1974813367627124e-07, "epoch": 18.744103773584907, "percentage": 93.72, "elapsed_time": "1:02:29", "remaining_time": "0:04:11", "throughput": 5539.55, "total_tokens": 20770392} +{"current_steps": 31795, "total_steps": 33920, "loss": 0.3149, "lr": 1.1918909957913949e-07, "epoch": 18.747051886792452, "percentage": 93.74, "elapsed_time": "1:02:29", "remaining_time": "0:04:10", "throughput": 5539.46, "total_tokens": 20772984} +{"current_steps": 31800, "total_steps": 33920, "loss": 0.3494, "lr": 1.1863135769772827e-07, "epoch": 18.75, "percentage": 93.75, "elapsed_time": "1:02:30", "remaining_time": "0:04:10", "throughput": 5539.61, "total_tokens": 20776632} +{"current_steps": 31805, "total_steps": 33920, "loss": 0.309, "lr": 1.1807490817970279e-07, "epoch": 18.752948113207548, "percentage": 93.76, "elapsed_time": "1:02:31", "remaining_time": "0:04:09", "throughput": 5539.64, "total_tokens": 20779864} +{"current_steps": 31810, "total_steps": 33920, "loss": 0.3438, "lr": 1.1751975117238578e-07, "epoch": 18.755896226415093, "percentage": 93.78, "elapsed_time": "1:02:31", "remaining_time": "0:04:08", "throughput": 5539.57, "total_tokens": 20782552} +{"current_steps": 31815, "total_steps": 33920, "loss": 0.3766, "lr": 1.1696588682275633e-07, "epoch": 18.75884433962264, "percentage": 93.79, "elapsed_time": "1:02:32", "remaining_time": "0:04:08", "throughput": 5539.53, "total_tokens": 20785560} +{"current_steps": 31820, "total_steps": 33920, "loss": 0.3192, "lr": 1.1641331527745325e-07, "epoch": 18.76179245283019, "percentage": 93.81, "elapsed_time": "1:02:32", "remaining_time": "0:04:07", "throughput": 5539.68, "total_tokens": 20789144} +{"current_steps": 31825, "total_steps": 33920, "loss": 0.2718, "lr": 1.1586203668277229e-07, "epoch": 18.764740566037737, "percentage": 93.82, "elapsed_time": "1:02:33", "remaining_time": "0:04:07", "throughput": 5539.63, "total_tokens": 20791672} +{"current_steps": 31830, "total_steps": 33920, "loss": 0.2494, "lr": 1.1531205118466615e-07, "epoch": 18.767688679245282, "percentage": 93.84, "elapsed_time": "1:02:33", "remaining_time": "0:04:06", "throughput": 5539.7, "total_tokens": 20794904} +{"current_steps": 31835, "total_steps": 33920, "loss": 0.2412, "lr": 1.1476335892874669e-07, "epoch": 18.77063679245283, "percentage": 93.85, "elapsed_time": "1:02:34", "remaining_time": "0:04:05", "throughput": 5539.64, "total_tokens": 20797464} +{"current_steps": 31840, "total_steps": 33920, "loss": 0.2692, "lr": 1.1421596006028157e-07, "epoch": 18.77358490566038, "percentage": 93.87, "elapsed_time": "1:02:34", "remaining_time": "0:04:05", "throughput": 5539.63, "total_tokens": 20800248} +{"current_steps": 31845, "total_steps": 33920, "loss": 0.3072, "lr": 1.1366985472419823e-07, "epoch": 18.776533018867923, "percentage": 93.88, "elapsed_time": "1:02:35", "remaining_time": "0:04:04", "throughput": 5539.83, "total_tokens": 20804440} +{"current_steps": 31850, "total_steps": 33920, "loss": 0.2143, "lr": 1.1312504306507987e-07, "epoch": 18.77948113207547, "percentage": 93.9, "elapsed_time": "1:02:35", "remaining_time": "0:04:04", "throughput": 5539.91, "total_tokens": 20807736} +{"current_steps": 31855, "total_steps": 33920, "loss": 0.4252, "lr": 1.1258152522716725e-07, "epoch": 18.78242924528302, "percentage": 93.91, "elapsed_time": "1:02:36", "remaining_time": "0:04:03", "throughput": 5539.98, "total_tokens": 20811192} +{"current_steps": 31860, "total_steps": 33920, "loss": 0.3122, "lr": 1.1203930135435914e-07, "epoch": 18.785377358490567, "percentage": 93.93, "elapsed_time": "1:02:37", "remaining_time": "0:04:02", "throughput": 5539.97, "total_tokens": 20814264} +{"current_steps": 31865, "total_steps": 33920, "loss": 0.3134, "lr": 1.1149837159021238e-07, "epoch": 18.788325471698112, "percentage": 93.94, "elapsed_time": "1:02:37", "remaining_time": "0:04:02", "throughput": 5540.16, "total_tokens": 20818616} +{"current_steps": 31870, "total_steps": 33920, "loss": 0.2907, "lr": 1.1095873607793961e-07, "epoch": 18.79127358490566, "percentage": 93.96, "elapsed_time": "1:02:38", "remaining_time": "0:04:01", "throughput": 5540.32, "total_tokens": 20822328} +{"current_steps": 31875, "total_steps": 33920, "loss": 0.4776, "lr": 1.1042039496041212e-07, "epoch": 18.79422169811321, "percentage": 93.97, "elapsed_time": "1:02:38", "remaining_time": "0:04:01", "throughput": 5540.23, "total_tokens": 20824824} +{"current_steps": 31880, "total_steps": 33920, "loss": 0.3379, "lr": 1.0988334838015812e-07, "epoch": 18.797169811320753, "percentage": 93.99, "elapsed_time": "1:02:39", "remaining_time": "0:04:00", "throughput": 5540.34, "total_tokens": 20828376} +{"current_steps": 31885, "total_steps": 33920, "loss": 0.2778, "lr": 1.0934759647936333e-07, "epoch": 18.8001179245283, "percentage": 94.0, "elapsed_time": "1:02:39", "remaining_time": "0:03:59", "throughput": 5540.42, "total_tokens": 20831896} +{"current_steps": 31890, "total_steps": 33920, "loss": 0.274, "lr": 1.0881313939986926e-07, "epoch": 18.80306603773585, "percentage": 94.02, "elapsed_time": "1:02:40", "remaining_time": "0:03:59", "throughput": 5540.44, "total_tokens": 20834872} +{"current_steps": 31895, "total_steps": 33920, "loss": 0.251, "lr": 1.0827997728317662e-07, "epoch": 18.806014150943398, "percentage": 94.03, "elapsed_time": "1:02:41", "remaining_time": "0:03:58", "throughput": 5540.33, "total_tokens": 20837624} +{"current_steps": 31900, "total_steps": 33920, "loss": 0.3176, "lr": 1.0774811027044196e-07, "epoch": 18.808962264150942, "percentage": 94.04, "elapsed_time": "1:02:41", "remaining_time": "0:03:58", "throughput": 5540.49, "total_tokens": 20841528} +{"current_steps": 31905, "total_steps": 33920, "loss": 0.4235, "lr": 1.0721753850247984e-07, "epoch": 18.81191037735849, "percentage": 94.06, "elapsed_time": "1:02:42", "remaining_time": "0:03:57", "throughput": 5540.63, "total_tokens": 20845112} +{"current_steps": 31910, "total_steps": 33920, "loss": 0.2606, "lr": 1.0668826211976124e-07, "epoch": 18.81485849056604, "percentage": 94.07, "elapsed_time": "1:02:42", "remaining_time": "0:03:57", "throughput": 5540.6, "total_tokens": 20847864} +{"current_steps": 31915, "total_steps": 33920, "loss": 0.301, "lr": 1.0616028126241407e-07, "epoch": 18.817806603773583, "percentage": 94.09, "elapsed_time": "1:02:43", "remaining_time": "0:03:56", "throughput": 5540.59, "total_tokens": 20850872} +{"current_steps": 31920, "total_steps": 33920, "loss": 0.3135, "lr": 1.0563359607022372e-07, "epoch": 18.82075471698113, "percentage": 94.1, "elapsed_time": "1:02:43", "remaining_time": "0:03:55", "throughput": 5540.57, "total_tokens": 20853656} +{"current_steps": 31925, "total_steps": 33920, "loss": 0.3584, "lr": 1.05108206682632e-07, "epoch": 18.82370283018868, "percentage": 94.12, "elapsed_time": "1:02:44", "remaining_time": "0:03:55", "throughput": 5540.57, "total_tokens": 20856920} +{"current_steps": 31930, "total_steps": 33920, "loss": 0.3705, "lr": 1.0458411323873874e-07, "epoch": 18.826650943396228, "percentage": 94.13, "elapsed_time": "1:02:45", "remaining_time": "0:03:54", "throughput": 5540.73, "total_tokens": 20861080} +{"current_steps": 31935, "total_steps": 33920, "loss": 0.3331, "lr": 1.0406131587729962e-07, "epoch": 18.829599056603772, "percentage": 94.15, "elapsed_time": "1:02:45", "remaining_time": "0:03:54", "throughput": 5540.67, "total_tokens": 20863704} +{"current_steps": 31940, "total_steps": 33920, "loss": 0.2358, "lr": 1.035398147367278e-07, "epoch": 18.83254716981132, "percentage": 94.16, "elapsed_time": "1:02:46", "remaining_time": "0:03:53", "throughput": 5540.61, "total_tokens": 20866360} +{"current_steps": 31945, "total_steps": 33920, "loss": 0.3564, "lr": 1.030196099550923e-07, "epoch": 18.83549528301887, "percentage": 94.18, "elapsed_time": "1:02:46", "remaining_time": "0:03:52", "throughput": 5540.57, "total_tokens": 20869208} +{"current_steps": 31950, "total_steps": 33920, "loss": 0.2418, "lr": 1.0250070167011905e-07, "epoch": 18.838443396226417, "percentage": 94.19, "elapsed_time": "1:02:47", "remaining_time": "0:03:52", "throughput": 5540.57, "total_tokens": 20872088} +{"current_steps": 31955, "total_steps": 33920, "loss": 0.2866, "lr": 1.0198309001919315e-07, "epoch": 18.84139150943396, "percentage": 94.21, "elapsed_time": "1:02:47", "remaining_time": "0:03:51", "throughput": 5540.59, "total_tokens": 20875576} +{"current_steps": 31960, "total_steps": 33920, "loss": 0.2355, "lr": 1.0146677513935277e-07, "epoch": 18.84433962264151, "percentage": 94.22, "elapsed_time": "1:02:48", "remaining_time": "0:03:51", "throughput": 5540.72, "total_tokens": 20879768} +{"current_steps": 31965, "total_steps": 33920, "loss": 0.3045, "lr": 1.0095175716729578e-07, "epoch": 18.847287735849058, "percentage": 94.24, "elapsed_time": "1:02:48", "remaining_time": "0:03:50", "throughput": 5540.69, "total_tokens": 20882776} +{"current_steps": 31970, "total_steps": 33920, "loss": 0.2269, "lr": 1.004380362393742e-07, "epoch": 18.850235849056602, "percentage": 94.25, "elapsed_time": "1:02:49", "remaining_time": "0:03:49", "throughput": 5540.57, "total_tokens": 20885304} +{"current_steps": 31975, "total_steps": 33920, "loss": 0.4048, "lr": 9.99256124915987e-08, "epoch": 18.85318396226415, "percentage": 94.27, "elapsed_time": "1:02:50", "remaining_time": "0:03:49", "throughput": 5540.77, "total_tokens": 20889400} +{"current_steps": 31980, "total_steps": 33920, "loss": 0.3116, "lr": 9.941448605963577e-08, "epoch": 18.8561320754717, "percentage": 94.28, "elapsed_time": "1:02:50", "remaining_time": "0:03:48", "throughput": 5540.78, "total_tokens": 20892312} +{"current_steps": 31985, "total_steps": 33920, "loss": 0.2993, "lr": 9.890465707880715e-08, "epoch": 18.859080188679247, "percentage": 94.3, "elapsed_time": "1:02:51", "remaining_time": "0:03:48", "throughput": 5540.82, "total_tokens": 20895480} +{"current_steps": 31990, "total_steps": 33920, "loss": 0.214, "lr": 9.839612568409374e-08, "epoch": 18.86202830188679, "percentage": 94.31, "elapsed_time": "1:02:51", "remaining_time": "0:03:47", "throughput": 5540.87, "total_tokens": 20898904} +{"current_steps": 31995, "total_steps": 33920, "loss": 0.3067, "lr": 9.788889201013119e-08, "epoch": 18.86497641509434, "percentage": 94.32, "elapsed_time": "1:02:52", "remaining_time": "0:03:46", "throughput": 5541.06, "total_tokens": 20902936} +{"current_steps": 32000, "total_steps": 33920, "loss": 0.3514, "lr": 9.738295619121097e-08, "epoch": 18.867924528301888, "percentage": 94.34, "elapsed_time": "1:02:52", "remaining_time": "0:03:46", "throughput": 5541.01, "total_tokens": 20905688} +{"current_steps": 32005, "total_steps": 33920, "loss": 0.2348, "lr": 9.687831836128203e-08, "epoch": 18.870872641509433, "percentage": 94.35, "elapsed_time": "1:02:53", "remaining_time": "0:03:45", "throughput": 5540.98, "total_tokens": 20908408} +{"current_steps": 32010, "total_steps": 33920, "loss": 0.327, "lr": 9.637497865395029e-08, "epoch": 18.87382075471698, "percentage": 94.37, "elapsed_time": "1:02:53", "remaining_time": "0:03:45", "throughput": 5540.86, "total_tokens": 20910616} +{"current_steps": 32015, "total_steps": 33920, "loss": 0.2606, "lr": 9.587293720247526e-08, "epoch": 18.87676886792453, "percentage": 94.38, "elapsed_time": "1:02:54", "remaining_time": "0:03:44", "throughput": 5540.92, "total_tokens": 20913912} +{"current_steps": 32020, "total_steps": 33920, "loss": 0.3826, "lr": 9.537219413977672e-08, "epoch": 18.879716981132077, "percentage": 94.4, "elapsed_time": "1:02:55", "remaining_time": "0:03:44", "throughput": 5541.03, "total_tokens": 20917400} +{"current_steps": 32025, "total_steps": 33920, "loss": 0.3092, "lr": 9.487274959842696e-08, "epoch": 18.88266509433962, "percentage": 94.41, "elapsed_time": "1:02:55", "remaining_time": "0:03:43", "throughput": 5541.11, "total_tokens": 20920728} +{"current_steps": 32030, "total_steps": 33920, "loss": 0.3075, "lr": 9.437460371065687e-08, "epoch": 18.88561320754717, "percentage": 94.43, "elapsed_time": "1:02:56", "remaining_time": "0:03:42", "throughput": 5541.3, "total_tokens": 20925208} +{"current_steps": 32035, "total_steps": 33920, "loss": 0.414, "lr": 9.387775660835263e-08, "epoch": 18.888561320754718, "percentage": 94.44, "elapsed_time": "1:02:56", "remaining_time": "0:03:42", "throughput": 5541.3, "total_tokens": 20928120} +{"current_steps": 32040, "total_steps": 33920, "loss": 0.3115, "lr": 9.338220842305678e-08, "epoch": 18.891509433962263, "percentage": 94.46, "elapsed_time": "1:02:57", "remaining_time": "0:03:41", "throughput": 5541.26, "total_tokens": 20930712} +{"current_steps": 32045, "total_steps": 33920, "loss": 0.2895, "lr": 9.288795928596661e-08, "epoch": 18.89445754716981, "percentage": 94.47, "elapsed_time": "1:02:57", "remaining_time": "0:03:41", "throughput": 5541.46, "total_tokens": 20934776} +{"current_steps": 32050, "total_steps": 33920, "loss": 0.4206, "lr": 9.239500932793854e-08, "epoch": 18.89740566037736, "percentage": 94.49, "elapsed_time": "1:02:58", "remaining_time": "0:03:40", "throughput": 5541.57, "total_tokens": 20938232} +{"current_steps": 32055, "total_steps": 33920, "loss": 0.2875, "lr": 9.190335867948263e-08, "epoch": 18.900353773584907, "percentage": 94.5, "elapsed_time": "1:02:58", "remaining_time": "0:03:39", "throughput": 5541.7, "total_tokens": 20941752} +{"current_steps": 32060, "total_steps": 33920, "loss": 0.3066, "lr": 9.141300747076476e-08, "epoch": 18.903301886792452, "percentage": 94.52, "elapsed_time": "1:02:59", "remaining_time": "0:03:39", "throughput": 5541.77, "total_tokens": 20944984} +{"current_steps": 32065, "total_steps": 33920, "loss": 0.2886, "lr": 9.092395583160773e-08, "epoch": 18.90625, "percentage": 94.53, "elapsed_time": "1:03:00", "remaining_time": "0:03:38", "throughput": 5541.7, "total_tokens": 20947800} +{"current_steps": 32070, "total_steps": 33920, "loss": 0.22, "lr": 9.043620389149021e-08, "epoch": 18.909198113207548, "percentage": 94.55, "elapsed_time": "1:03:00", "remaining_time": "0:03:38", "throughput": 5541.82, "total_tokens": 20952216} +{"current_steps": 32075, "total_steps": 33920, "loss": 0.3615, "lr": 8.994975177954723e-08, "epoch": 18.912146226415093, "percentage": 94.56, "elapsed_time": "1:03:01", "remaining_time": "0:03:37", "throughput": 5541.89, "total_tokens": 20955832} +{"current_steps": 32080, "total_steps": 33920, "loss": 0.2848, "lr": 8.946459962456855e-08, "epoch": 18.91509433962264, "percentage": 94.58, "elapsed_time": "1:03:01", "remaining_time": "0:03:36", "throughput": 5541.8, "total_tokens": 20958328} +{"current_steps": 32085, "total_steps": 33920, "loss": 0.2892, "lr": 8.89807475550003e-08, "epoch": 18.91804245283019, "percentage": 94.59, "elapsed_time": "1:03:02", "remaining_time": "0:03:36", "throughput": 5541.87, "total_tokens": 20961752} +{"current_steps": 32090, "total_steps": 33920, "loss": 0.3258, "lr": 8.849819569894447e-08, "epoch": 18.920990566037737, "percentage": 94.6, "elapsed_time": "1:03:02", "remaining_time": "0:03:35", "throughput": 5541.97, "total_tokens": 20965240} +{"current_steps": 32095, "total_steps": 33920, "loss": 0.2182, "lr": 8.801694418415884e-08, "epoch": 18.923938679245282, "percentage": 94.62, "elapsed_time": "1:03:03", "remaining_time": "0:03:35", "throughput": 5542.0, "total_tokens": 20968088} +{"current_steps": 32100, "total_steps": 33920, "loss": 0.2729, "lr": 8.753699313805708e-08, "epoch": 18.92688679245283, "percentage": 94.63, "elapsed_time": "1:03:04", "remaining_time": "0:03:34", "throughput": 5542.07, "total_tokens": 20971416} +{"current_steps": 32105, "total_steps": 33920, "loss": 0.2604, "lr": 8.705834268770753e-08, "epoch": 18.92983490566038, "percentage": 94.65, "elapsed_time": "1:03:04", "remaining_time": "0:03:33", "throughput": 5541.9, "total_tokens": 20973816} +{"current_steps": 32110, "total_steps": 33920, "loss": 0.3348, "lr": 8.65809929598349e-08, "epoch": 18.932783018867923, "percentage": 94.66, "elapsed_time": "1:03:05", "remaining_time": "0:03:33", "throughput": 5541.9, "total_tokens": 20976664} +{"current_steps": 32115, "total_steps": 33920, "loss": 0.4021, "lr": 8.610494408082037e-08, "epoch": 18.93573113207547, "percentage": 94.68, "elapsed_time": "1:03:05", "remaining_time": "0:03:32", "throughput": 5542.1, "total_tokens": 20981272} +{"current_steps": 32120, "total_steps": 33920, "loss": 0.2293, "lr": 8.563019617669977e-08, "epoch": 18.93867924528302, "percentage": 94.69, "elapsed_time": "1:03:06", "remaining_time": "0:03:32", "throughput": 5542.06, "total_tokens": 20983960} +{"current_steps": 32125, "total_steps": 33920, "loss": 0.3327, "lr": 8.51567493731642e-08, "epoch": 18.941627358490567, "percentage": 94.71, "elapsed_time": "1:03:06", "remaining_time": "0:03:31", "throughput": 5542.15, "total_tokens": 20987128} +{"current_steps": 32130, "total_steps": 33920, "loss": 0.3825, "lr": 8.468460379556176e-08, "epoch": 18.944575471698112, "percentage": 94.72, "elapsed_time": "1:03:07", "remaining_time": "0:03:30", "throughput": 5542.09, "total_tokens": 20989880} +{"current_steps": 32135, "total_steps": 33920, "loss": 0.2136, "lr": 8.421375956889355e-08, "epoch": 18.94752358490566, "percentage": 94.74, "elapsed_time": "1:03:07", "remaining_time": "0:03:30", "throughput": 5542.05, "total_tokens": 20992920} +{"current_steps": 32140, "total_steps": 33920, "loss": 0.2, "lr": 8.374421681781819e-08, "epoch": 18.95047169811321, "percentage": 94.75, "elapsed_time": "1:03:08", "remaining_time": "0:03:29", "throughput": 5542.0, "total_tokens": 20995672} +{"current_steps": 32145, "total_steps": 33920, "loss": 0.3536, "lr": 8.327597566665013e-08, "epoch": 18.953419811320753, "percentage": 94.77, "elapsed_time": "1:03:09", "remaining_time": "0:03:29", "throughput": 5542.0, "total_tokens": 20999000} +{"current_steps": 32150, "total_steps": 33920, "loss": 0.2623, "lr": 8.280903623935688e-08, "epoch": 18.9563679245283, "percentage": 94.78, "elapsed_time": "1:03:09", "remaining_time": "0:03:28", "throughput": 5542.06, "total_tokens": 21002360} +{"current_steps": 32155, "total_steps": 33920, "loss": 0.3902, "lr": 8.234339865956342e-08, "epoch": 18.95931603773585, "percentage": 94.8, "elapsed_time": "1:03:10", "remaining_time": "0:03:28", "throughput": 5542.07, "total_tokens": 21005400} +{"current_steps": 32160, "total_steps": 33920, "loss": 0.3627, "lr": 8.187906305054838e-08, "epoch": 18.962264150943398, "percentage": 94.81, "elapsed_time": "1:03:10", "remaining_time": "0:03:27", "throughput": 5542.15, "total_tokens": 21008856} +{"current_steps": 32165, "total_steps": 33920, "loss": 0.4293, "lr": 8.141602953524841e-08, "epoch": 18.965212264150942, "percentage": 94.83, "elapsed_time": "1:03:11", "remaining_time": "0:03:26", "throughput": 5542.23, "total_tokens": 21012056} +{"current_steps": 32170, "total_steps": 33920, "loss": 0.2932, "lr": 8.095429823625212e-08, "epoch": 18.96816037735849, "percentage": 94.84, "elapsed_time": "1:03:11", "remaining_time": "0:03:26", "throughput": 5542.25, "total_tokens": 21015224} +{"current_steps": 32175, "total_steps": 33920, "loss": 0.3457, "lr": 8.04938692758045e-08, "epoch": 18.97110849056604, "percentage": 94.86, "elapsed_time": "1:03:12", "remaining_time": "0:03:25", "throughput": 5542.42, "total_tokens": 21019288} +{"current_steps": 32180, "total_steps": 33920, "loss": 0.3959, "lr": 8.003474277580803e-08, "epoch": 18.974056603773583, "percentage": 94.87, "elapsed_time": "1:03:12", "remaining_time": "0:03:25", "throughput": 5542.47, "total_tokens": 21022456} +{"current_steps": 32185, "total_steps": 33920, "loss": 0.2887, "lr": 7.95769188578166e-08, "epoch": 18.97700471698113, "percentage": 94.89, "elapsed_time": "1:03:13", "remaining_time": "0:03:24", "throughput": 5542.5, "total_tokens": 21026200} +{"current_steps": 32190, "total_steps": 33920, "loss": 0.3176, "lr": 7.912039764304213e-08, "epoch": 18.97995283018868, "percentage": 94.9, "elapsed_time": "1:03:14", "remaining_time": "0:03:23", "throughput": 5542.47, "total_tokens": 21028952} +{"current_steps": 32195, "total_steps": 33920, "loss": 0.3514, "lr": 7.866517925235017e-08, "epoch": 18.982900943396228, "percentage": 94.91, "elapsed_time": "1:03:14", "remaining_time": "0:03:23", "throughput": 5542.34, "total_tokens": 21031512} +{"current_steps": 32200, "total_steps": 33920, "loss": 0.3086, "lr": 7.821126380626154e-08, "epoch": 18.985849056603772, "percentage": 94.93, "elapsed_time": "1:03:15", "remaining_time": "0:03:22", "throughput": 5542.25, "total_tokens": 21034136} +{"current_steps": 32205, "total_steps": 33920, "loss": 0.4198, "lr": 7.775865142495286e-08, "epoch": 18.98879716981132, "percentage": 94.94, "elapsed_time": "1:03:15", "remaining_time": "0:03:22", "throughput": 5542.33, "total_tokens": 21037592} +{"current_steps": 32210, "total_steps": 33920, "loss": 0.1823, "lr": 7.730734222825442e-08, "epoch": 18.99174528301887, "percentage": 94.96, "elapsed_time": "1:03:16", "remaining_time": "0:03:21", "throughput": 5542.42, "total_tokens": 21040984} +{"current_steps": 32215, "total_steps": 33920, "loss": 0.3309, "lr": 7.68573363356534e-08, "epoch": 18.994693396226417, "percentage": 94.97, "elapsed_time": "1:03:16", "remaining_time": "0:03:20", "throughput": 5542.54, "total_tokens": 21044344} +{"current_steps": 32220, "total_steps": 33920, "loss": 0.2936, "lr": 7.640863386629005e-08, "epoch": 18.99764150943396, "percentage": 94.99, "elapsed_time": "1:03:17", "remaining_time": "0:03:20", "throughput": 5542.5, "total_tokens": 21046808} +{"current_steps": 32225, "total_steps": 33920, "loss": 0.2156, "lr": 7.59612349389599e-08, "epoch": 19.00058962264151, "percentage": 95.0, "elapsed_time": "1:03:18", "remaining_time": "0:03:19", "throughput": 5542.0, "total_tokens": 21049520} +{"current_steps": 32230, "total_steps": 33920, "loss": 0.2575, "lr": 7.551513967211433e-08, "epoch": 19.003537735849058, "percentage": 95.02, "elapsed_time": "1:03:18", "remaining_time": "0:03:19", "throughput": 5541.97, "total_tokens": 21052336} +{"current_steps": 32235, "total_steps": 33920, "loss": 0.3925, "lr": 7.507034818385883e-08, "epoch": 19.006485849056602, "percentage": 95.03, "elapsed_time": "1:03:19", "remaining_time": "0:03:18", "throughput": 5542.0, "total_tokens": 21055664} +{"current_steps": 32240, "total_steps": 33920, "loss": 0.2727, "lr": 7.462686059195423e-08, "epoch": 19.00943396226415, "percentage": 95.05, "elapsed_time": "1:03:19", "remaining_time": "0:03:18", "throughput": 5542.06, "total_tokens": 21058928} +{"current_steps": 32245, "total_steps": 33920, "loss": 0.3251, "lr": 7.418467701381548e-08, "epoch": 19.0123820754717, "percentage": 95.06, "elapsed_time": "1:03:20", "remaining_time": "0:03:17", "throughput": 5542.03, "total_tokens": 21061456} +{"current_steps": 32250, "total_steps": 33920, "loss": 0.3071, "lr": 7.374379756651285e-08, "epoch": 19.015330188679247, "percentage": 95.08, "elapsed_time": "1:03:20", "remaining_time": "0:03:16", "throughput": 5542.12, "total_tokens": 21064656} +{"current_steps": 32255, "total_steps": 33920, "loss": 0.272, "lr": 7.330422236677015e-08, "epoch": 19.01827830188679, "percentage": 95.09, "elapsed_time": "1:03:21", "remaining_time": "0:03:16", "throughput": 5542.23, "total_tokens": 21068592} +{"current_steps": 32260, "total_steps": 33920, "loss": 0.4904, "lr": 7.286595153096765e-08, "epoch": 19.02122641509434, "percentage": 95.11, "elapsed_time": "1:03:21", "remaining_time": "0:03:15", "throughput": 5542.16, "total_tokens": 21071120} +{"current_steps": 32265, "total_steps": 33920, "loss": 0.183, "lr": 7.242898517513864e-08, "epoch": 19.024174528301888, "percentage": 95.12, "elapsed_time": "1:03:22", "remaining_time": "0:03:15", "throughput": 5542.21, "total_tokens": 21074352} +{"current_steps": 32270, "total_steps": 33920, "loss": 0.382, "lr": 7.199332341497333e-08, "epoch": 19.027122641509433, "percentage": 95.14, "elapsed_time": "1:03:23", "remaining_time": "0:03:14", "throughput": 5542.42, "total_tokens": 21078576} +{"current_steps": 32275, "total_steps": 33920, "loss": 0.2716, "lr": 7.155896636581394e-08, "epoch": 19.03007075471698, "percentage": 95.15, "elapsed_time": "1:03:23", "remaining_time": "0:03:13", "throughput": 5542.32, "total_tokens": 21081040} +{"current_steps": 32280, "total_steps": 33920, "loss": 0.4189, "lr": 7.112591414265901e-08, "epoch": 19.03301886792453, "percentage": 95.17, "elapsed_time": "1:03:24", "remaining_time": "0:03:13", "throughput": 5542.34, "total_tokens": 21084272} +{"current_steps": 32285, "total_steps": 33920, "loss": 0.2196, "lr": 7.069416686016018e-08, "epoch": 19.035966981132077, "percentage": 95.18, "elapsed_time": "1:03:24", "remaining_time": "0:03:12", "throughput": 5542.32, "total_tokens": 21087088} +{"current_steps": 32290, "total_steps": 33920, "loss": 0.2777, "lr": 7.026372463262488e-08, "epoch": 19.03891509433962, "percentage": 95.19, "elapsed_time": "1:03:25", "remaining_time": "0:03:12", "throughput": 5542.39, "total_tokens": 21090448} +{"current_steps": 32295, "total_steps": 33920, "loss": 0.3214, "lr": 6.983458757401418e-08, "epoch": 19.04186320754717, "percentage": 95.21, "elapsed_time": "1:03:25", "remaining_time": "0:03:11", "throughput": 5542.49, "total_tokens": 21094352} +{"current_steps": 32300, "total_steps": 33920, "loss": 0.298, "lr": 6.940675579794443e-08, "epoch": 19.044811320754718, "percentage": 95.22, "elapsed_time": "1:03:26", "remaining_time": "0:03:10", "throughput": 5542.53, "total_tokens": 21097296} +{"current_steps": 32305, "total_steps": 33920, "loss": 0.3554, "lr": 6.898022941768612e-08, "epoch": 19.047759433962263, "percentage": 95.24, "elapsed_time": "1:03:26", "remaining_time": "0:03:10", "throughput": 5542.55, "total_tokens": 21100144} +{"current_steps": 32310, "total_steps": 33920, "loss": 0.2539, "lr": 6.855500854616337e-08, "epoch": 19.05070754716981, "percentage": 95.25, "elapsed_time": "1:03:27", "remaining_time": "0:03:09", "throughput": 5542.79, "total_tokens": 21104816} +{"current_steps": 32315, "total_steps": 33920, "loss": 0.3027, "lr": 6.813109329595557e-08, "epoch": 19.05365566037736, "percentage": 95.27, "elapsed_time": "1:03:28", "remaining_time": "0:03:09", "throughput": 5542.81, "total_tokens": 21108464} +{"current_steps": 32320, "total_steps": 33920, "loss": 0.2209, "lr": 6.770848377929573e-08, "epoch": 19.056603773584907, "percentage": 95.28, "elapsed_time": "1:03:28", "remaining_time": "0:03:08", "throughput": 5542.86, "total_tokens": 21111728} +{"current_steps": 32325, "total_steps": 33920, "loss": 0.3688, "lr": 6.728718010807156e-08, "epoch": 19.059551886792452, "percentage": 95.3, "elapsed_time": "1:03:29", "remaining_time": "0:03:07", "throughput": 5542.91, "total_tokens": 21115088} +{"current_steps": 32330, "total_steps": 33920, "loss": 0.344, "lr": 6.68671823938255e-08, "epoch": 19.0625, "percentage": 95.31, "elapsed_time": "1:03:29", "remaining_time": "0:03:07", "throughput": 5542.9, "total_tokens": 21117840} +{"current_steps": 32335, "total_steps": 33920, "loss": 0.3484, "lr": 6.644849074775361e-08, "epoch": 19.065448113207548, "percentage": 95.33, "elapsed_time": "1:03:30", "remaining_time": "0:03:06", "throughput": 5542.88, "total_tokens": 21120688} +{"current_steps": 32340, "total_steps": 33920, "loss": 0.2825, "lr": 6.603110528070667e-08, "epoch": 19.068396226415093, "percentage": 95.34, "elapsed_time": "1:03:30", "remaining_time": "0:03:06", "throughput": 5542.95, "total_tokens": 21123856} +{"current_steps": 32345, "total_steps": 33920, "loss": 0.3666, "lr": 6.561502610318849e-08, "epoch": 19.07134433962264, "percentage": 95.36, "elapsed_time": "1:03:31", "remaining_time": "0:03:05", "throughput": 5542.88, "total_tokens": 21126544} +{"current_steps": 32350, "total_steps": 33920, "loss": 0.3101, "lr": 6.520025332535762e-08, "epoch": 19.07429245283019, "percentage": 95.37, "elapsed_time": "1:03:32", "remaining_time": "0:03:05", "throughput": 5542.77, "total_tokens": 21129136} +{"current_steps": 32355, "total_steps": 33920, "loss": 0.3507, "lr": 6.47867870570279e-08, "epoch": 19.077240566037737, "percentage": 95.39, "elapsed_time": "1:03:32", "remaining_time": "0:03:04", "throughput": 5542.78, "total_tokens": 21132432} +{"current_steps": 32360, "total_steps": 33920, "loss": 0.2864, "lr": 6.437462740766564e-08, "epoch": 19.080188679245282, "percentage": 95.4, "elapsed_time": "1:03:33", "remaining_time": "0:03:03", "throughput": 5542.79, "total_tokens": 21135728} +{"current_steps": 32365, "total_steps": 33920, "loss": 0.2273, "lr": 6.396377448639246e-08, "epoch": 19.08313679245283, "percentage": 95.42, "elapsed_time": "1:03:33", "remaining_time": "0:03:03", "throughput": 5542.97, "total_tokens": 21139600} +{"current_steps": 32370, "total_steps": 33920, "loss": 0.3517, "lr": 6.3554228401983e-08, "epoch": 19.08608490566038, "percentage": 95.43, "elapsed_time": "1:03:34", "remaining_time": "0:03:02", "throughput": 5543.08, "total_tokens": 21142832} +{"current_steps": 32375, "total_steps": 33920, "loss": 0.3364, "lr": 6.314598926286663e-08, "epoch": 19.089033018867923, "percentage": 95.45, "elapsed_time": "1:03:34", "remaining_time": "0:03:02", "throughput": 5543.2, "total_tokens": 21146320} +{"current_steps": 32380, "total_steps": 33920, "loss": 0.3587, "lr": 6.273905717712637e-08, "epoch": 19.09198113207547, "percentage": 95.46, "elapsed_time": "1:03:35", "remaining_time": "0:03:01", "throughput": 5543.31, "total_tokens": 21149648} +{"current_steps": 32385, "total_steps": 33920, "loss": 0.338, "lr": 6.233343225249933e-08, "epoch": 19.09492924528302, "percentage": 95.47, "elapsed_time": "1:03:35", "remaining_time": "0:03:00", "throughput": 5543.38, "total_tokens": 21152912} +{"current_steps": 32390, "total_steps": 33920, "loss": 0.3073, "lr": 6.192911459637519e-08, "epoch": 19.097877358490567, "percentage": 95.49, "elapsed_time": "1:03:36", "remaining_time": "0:03:00", "throughput": 5543.31, "total_tokens": 21155568} +{"current_steps": 32395, "total_steps": 33920, "loss": 0.3473, "lr": 6.152610431580052e-08, "epoch": 19.100825471698112, "percentage": 95.5, "elapsed_time": "1:03:37", "remaining_time": "0:02:59", "throughput": 5543.47, "total_tokens": 21159696} +{"current_steps": 32400, "total_steps": 33920, "loss": 0.3249, "lr": 6.112440151747389e-08, "epoch": 19.10377358490566, "percentage": 95.52, "elapsed_time": "1:03:37", "remaining_time": "0:02:59", "throughput": 5543.61, "total_tokens": 21163504} +{"current_steps": 32405, "total_steps": 33920, "loss": 0.2764, "lr": 6.072400630774689e-08, "epoch": 19.10672169811321, "percentage": 95.53, "elapsed_time": "1:03:38", "remaining_time": "0:02:58", "throughput": 5543.71, "total_tokens": 21167408} +{"current_steps": 32410, "total_steps": 33920, "loss": 0.3602, "lr": 6.032491879262637e-08, "epoch": 19.109669811320753, "percentage": 95.55, "elapsed_time": "1:03:38", "remaining_time": "0:02:57", "throughput": 5543.69, "total_tokens": 21169968} +{"current_steps": 32415, "total_steps": 33920, "loss": 0.3143, "lr": 5.99271390777717e-08, "epoch": 19.1126179245283, "percentage": 95.56, "elapsed_time": "1:03:39", "remaining_time": "0:02:57", "throughput": 5543.63, "total_tokens": 21172784} +{"current_steps": 32420, "total_steps": 33920, "loss": 0.2601, "lr": 5.953066726849865e-08, "epoch": 19.11556603773585, "percentage": 95.58, "elapsed_time": "1:03:39", "remaining_time": "0:02:56", "throughput": 5543.61, "total_tokens": 21175472} +{"current_steps": 32425, "total_steps": 33920, "loss": 0.2232, "lr": 5.913550346977326e-08, "epoch": 19.118514150943398, "percentage": 95.59, "elapsed_time": "1:03:40", "remaining_time": "0:02:56", "throughput": 5543.67, "total_tokens": 21178928} +{"current_steps": 32430, "total_steps": 33920, "loss": 0.3218, "lr": 5.874164778621683e-08, "epoch": 19.121462264150942, "percentage": 95.61, "elapsed_time": "1:03:40", "remaining_time": "0:02:55", "throughput": 5543.75, "total_tokens": 21182160} +{"current_steps": 32435, "total_steps": 33920, "loss": 0.3909, "lr": 5.834910032210539e-08, "epoch": 19.12441037735849, "percentage": 95.62, "elapsed_time": "1:03:41", "remaining_time": "0:02:54", "throughput": 5543.77, "total_tokens": 21184944} +{"current_steps": 32440, "total_steps": 33920, "loss": 0.3448, "lr": 5.795786118136693e-08, "epoch": 19.12735849056604, "percentage": 95.64, "elapsed_time": "1:03:41", "remaining_time": "0:02:54", "throughput": 5543.78, "total_tokens": 21187760} +{"current_steps": 32445, "total_steps": 33920, "loss": 0.3089, "lr": 5.756793046758302e-08, "epoch": 19.130306603773583, "percentage": 95.65, "elapsed_time": "1:03:42", "remaining_time": "0:02:53", "throughput": 5543.81, "total_tokens": 21190704} +{"current_steps": 32450, "total_steps": 33920, "loss": 0.1959, "lr": 5.7179308283990544e-08, "epoch": 19.13325471698113, "percentage": 95.67, "elapsed_time": "1:03:43", "remaining_time": "0:02:53", "throughput": 5544.05, "total_tokens": 21196496} +{"current_steps": 32455, "total_steps": 33920, "loss": 0.3496, "lr": 5.679199473347885e-08, "epoch": 19.13620283018868, "percentage": 95.68, "elapsed_time": "1:03:43", "remaining_time": "0:02:52", "throughput": 5544.0, "total_tokens": 21198960} +{"current_steps": 32460, "total_steps": 33920, "loss": 0.2971, "lr": 5.6405989918590366e-08, "epoch": 19.139150943396228, "percentage": 95.7, "elapsed_time": "1:03:44", "remaining_time": "0:02:52", "throughput": 5544.02, "total_tokens": 21202032} +{"current_steps": 32465, "total_steps": 33920, "loss": 0.3245, "lr": 5.6021293941522225e-08, "epoch": 19.142099056603772, "percentage": 95.71, "elapsed_time": "1:03:44", "remaining_time": "0:02:51", "throughput": 5544.1, "total_tokens": 21205296} +{"current_steps": 32470, "total_steps": 33920, "loss": 0.2608, "lr": 5.563790690412352e-08, "epoch": 19.14504716981132, "percentage": 95.73, "elapsed_time": "1:03:45", "remaining_time": "0:02:50", "throughput": 5544.11, "total_tokens": 21208496} +{"current_steps": 32475, "total_steps": 33920, "loss": 0.3646, "lr": 5.525582890789805e-08, "epoch": 19.14799528301887, "percentage": 95.74, "elapsed_time": "1:03:46", "remaining_time": "0:02:50", "throughput": 5544.24, "total_tokens": 21212304} +{"current_steps": 32480, "total_steps": 33920, "loss": 0.3171, "lr": 5.4875060054002115e-08, "epoch": 19.150943396226417, "percentage": 95.75, "elapsed_time": "1:03:46", "remaining_time": "0:02:49", "throughput": 5544.41, "total_tokens": 21216208} +{"current_steps": 32485, "total_steps": 33920, "loss": 0.2898, "lr": 5.4495600443246755e-08, "epoch": 19.15389150943396, "percentage": 95.77, "elapsed_time": "1:03:47", "remaining_time": "0:02:49", "throughput": 5544.43, "total_tokens": 21218896} +{"current_steps": 32490, "total_steps": 33920, "loss": 0.3895, "lr": 5.411745017609493e-08, "epoch": 19.15683962264151, "percentage": 95.78, "elapsed_time": "1:03:47", "remaining_time": "0:02:48", "throughput": 5544.54, "total_tokens": 21222224} +{"current_steps": 32495, "total_steps": 33920, "loss": 0.2963, "lr": 5.374060935266434e-08, "epoch": 19.159787735849058, "percentage": 95.8, "elapsed_time": "1:03:48", "remaining_time": "0:02:47", "throughput": 5544.52, "total_tokens": 21224784} +{"current_steps": 32500, "total_steps": 33920, "loss": 0.3287, "lr": 5.3365078072724065e-08, "epoch": 19.162735849056602, "percentage": 95.81, "elapsed_time": "1:03:48", "remaining_time": "0:02:47", "throughput": 5544.72, "total_tokens": 21229008} +{"current_steps": 32505, "total_steps": 33920, "loss": 0.1909, "lr": 5.299085643569846e-08, "epoch": 19.16568396226415, "percentage": 95.83, "elapsed_time": "1:03:49", "remaining_time": "0:02:46", "throughput": 5544.81, "total_tokens": 21232176} +{"current_steps": 32510, "total_steps": 33920, "loss": 0.3442, "lr": 5.261794454066327e-08, "epoch": 19.1686320754717, "percentage": 95.84, "elapsed_time": "1:03:49", "remaining_time": "0:02:46", "throughput": 5544.98, "total_tokens": 21236112} +{"current_steps": 32515, "total_steps": 33920, "loss": 0.2851, "lr": 5.224634248635008e-08, "epoch": 19.171580188679247, "percentage": 95.86, "elapsed_time": "1:03:50", "remaining_time": "0:02:45", "throughput": 5545.12, "total_tokens": 21239952} +{"current_steps": 32520, "total_steps": 33920, "loss": 0.2832, "lr": 5.187605037114129e-08, "epoch": 19.17452830188679, "percentage": 95.87, "elapsed_time": "1:03:50", "remaining_time": "0:02:44", "throughput": 5545.2, "total_tokens": 21243216} +{"current_steps": 32525, "total_steps": 33920, "loss": 0.3248, "lr": 5.15070682930735e-08, "epoch": 19.17747641509434, "percentage": 95.89, "elapsed_time": "1:03:51", "remaining_time": "0:02:44", "throughput": 5545.24, "total_tokens": 21246224} +{"current_steps": 32530, "total_steps": 33920, "loss": 0.4665, "lr": 5.113939634983578e-08, "epoch": 19.180424528301888, "percentage": 95.9, "elapsed_time": "1:03:51", "remaining_time": "0:02:43", "throughput": 5545.31, "total_tokens": 21249424} +{"current_steps": 32535, "total_steps": 33920, "loss": 0.3769, "lr": 5.077303463877192e-08, "epoch": 19.183372641509433, "percentage": 95.92, "elapsed_time": "1:03:52", "remaining_time": "0:02:43", "throughput": 5545.44, "total_tokens": 21253360} +{"current_steps": 32540, "total_steps": 33920, "loss": 0.245, "lr": 5.040798325687601e-08, "epoch": 19.18632075471698, "percentage": 95.93, "elapsed_time": "1:03:53", "remaining_time": "0:02:42", "throughput": 5545.38, "total_tokens": 21255952} +{"current_steps": 32545, "total_steps": 33920, "loss": 0.3078, "lr": 5.004424230079852e-08, "epoch": 19.18926886792453, "percentage": 95.95, "elapsed_time": "1:03:53", "remaining_time": "0:02:41", "throughput": 5545.36, "total_tokens": 21258576} +{"current_steps": 32550, "total_steps": 33920, "loss": 0.2585, "lr": 4.968181186684129e-08, "epoch": 19.192216981132077, "percentage": 95.96, "elapsed_time": "1:03:54", "remaining_time": "0:02:41", "throughput": 5545.27, "total_tokens": 21261296} +{"current_steps": 32555, "total_steps": 33920, "loss": 0.4654, "lr": 4.932069205095924e-08, "epoch": 19.19516509433962, "percentage": 95.98, "elapsed_time": "1:03:54", "remaining_time": "0:02:40", "throughput": 5545.4, "total_tokens": 21265392} +{"current_steps": 32560, "total_steps": 33920, "loss": 0.4247, "lr": 4.896088294875978e-08, "epoch": 19.19811320754717, "percentage": 95.99, "elapsed_time": "1:03:55", "remaining_time": "0:02:40", "throughput": 5545.47, "total_tokens": 21268528} +{"current_steps": 32565, "total_steps": 33920, "loss": 0.3509, "lr": 4.8602384655505044e-08, "epoch": 19.201061320754718, "percentage": 96.01, "elapsed_time": "1:03:55", "remaining_time": "0:02:39", "throughput": 5545.38, "total_tokens": 21271280} +{"current_steps": 32570, "total_steps": 33920, "loss": 0.4577, "lr": 4.824519726610744e-08, "epoch": 19.204009433962263, "percentage": 96.02, "elapsed_time": "1:03:56", "remaining_time": "0:02:39", "throughput": 5545.43, "total_tokens": 21274416} +{"current_steps": 32575, "total_steps": 33920, "loss": 0.2693, "lr": 4.7889320875135206e-08, "epoch": 19.20695754716981, "percentage": 96.03, "elapsed_time": "1:03:56", "remaining_time": "0:02:38", "throughput": 5545.59, "total_tokens": 21278160} +{"current_steps": 32580, "total_steps": 33920, "loss": 0.3718, "lr": 4.753475557680742e-08, "epoch": 19.20990566037736, "percentage": 96.05, "elapsed_time": "1:03:57", "remaining_time": "0:02:37", "throughput": 5545.66, "total_tokens": 21281392} +{"current_steps": 32585, "total_steps": 33920, "loss": 0.3663, "lr": 4.718150146499734e-08, "epoch": 19.212853773584907, "percentage": 96.06, "elapsed_time": "1:03:58", "remaining_time": "0:02:37", "throughput": 5545.56, "total_tokens": 21283952} +{"current_steps": 32590, "total_steps": 33920, "loss": 0.3068, "lr": 4.682955863323013e-08, "epoch": 19.215801886792452, "percentage": 96.08, "elapsed_time": "1:03:58", "remaining_time": "0:02:36", "throughput": 5545.73, "total_tokens": 21288176} +{"current_steps": 32595, "total_steps": 33920, "loss": 0.2168, "lr": 4.6478927174684606e-08, "epoch": 19.21875, "percentage": 96.09, "elapsed_time": "1:03:59", "remaining_time": "0:02:36", "throughput": 5545.81, "total_tokens": 21291408} +{"current_steps": 32600, "total_steps": 33920, "loss": 0.4105, "lr": 4.612960718219095e-08, "epoch": 19.221698113207548, "percentage": 96.11, "elapsed_time": "1:03:59", "remaining_time": "0:02:35", "throughput": 5545.88, "total_tokens": 21294800} +{"current_steps": 32605, "total_steps": 33920, "loss": 0.3569, "lr": 4.578159874823407e-08, "epoch": 19.224646226415093, "percentage": 96.12, "elapsed_time": "1:04:00", "remaining_time": "0:02:34", "throughput": 5545.8, "total_tokens": 21297296} +{"current_steps": 32610, "total_steps": 33920, "loss": 0.3, "lr": 4.5434901964950264e-08, "epoch": 19.22759433962264, "percentage": 96.14, "elapsed_time": "1:04:00", "remaining_time": "0:02:34", "throughput": 5545.73, "total_tokens": 21300112} +{"current_steps": 32615, "total_steps": 33920, "loss": 0.2629, "lr": 4.508951692412944e-08, "epoch": 19.23054245283019, "percentage": 96.15, "elapsed_time": "1:04:01", "remaining_time": "0:02:33", "throughput": 5545.65, "total_tokens": 21302992} +{"current_steps": 32620, "total_steps": 33920, "loss": 0.3039, "lr": 4.4745443717213455e-08, "epoch": 19.233490566037737, "percentage": 96.17, "elapsed_time": "1:04:01", "remaining_time": "0:02:33", "throughput": 5545.53, "total_tokens": 21305488} +{"current_steps": 32625, "total_steps": 33920, "loss": 0.3112, "lr": 4.4402682435296666e-08, "epoch": 19.236438679245282, "percentage": 96.18, "elapsed_time": "1:04:02", "remaining_time": "0:02:32", "throughput": 5545.51, "total_tokens": 21308880} +{"current_steps": 32630, "total_steps": 33920, "loss": 0.236, "lr": 4.406123316912758e-08, "epoch": 19.23938679245283, "percentage": 96.2, "elapsed_time": "1:04:03", "remaining_time": "0:02:31", "throughput": 5545.57, "total_tokens": 21311920} +{"current_steps": 32635, "total_steps": 33920, "loss": 0.4071, "lr": 4.372109600910612e-08, "epoch": 19.24233490566038, "percentage": 96.21, "elapsed_time": "1:04:03", "remaining_time": "0:02:31", "throughput": 5545.69, "total_tokens": 21315472} +{"current_steps": 32640, "total_steps": 33920, "loss": 0.4155, "lr": 4.338227104528414e-08, "epoch": 19.245283018867923, "percentage": 96.23, "elapsed_time": "1:04:04", "remaining_time": "0:02:30", "throughput": 5545.86, "total_tokens": 21319664} +{"current_steps": 32645, "total_steps": 33920, "loss": 0.3273, "lr": 4.304475836736821e-08, "epoch": 19.24823113207547, "percentage": 96.24, "elapsed_time": "1:04:04", "remaining_time": "0:02:30", "throughput": 5545.93, "total_tokens": 21323056} +{"current_steps": 32650, "total_steps": 33920, "loss": 0.3496, "lr": 4.27085580647163e-08, "epoch": 19.25117924528302, "percentage": 96.26, "elapsed_time": "1:04:05", "remaining_time": "0:02:29", "throughput": 5545.88, "total_tokens": 21325744} +{"current_steps": 32655, "total_steps": 33920, "loss": 0.2799, "lr": 4.237367022633776e-08, "epoch": 19.254127358490567, "percentage": 96.27, "elapsed_time": "1:04:05", "remaining_time": "0:02:28", "throughput": 5545.72, "total_tokens": 21328144} +{"current_steps": 32660, "total_steps": 33920, "loss": 0.3754, "lr": 4.204009494089612e-08, "epoch": 19.257075471698112, "percentage": 96.29, "elapsed_time": "1:04:06", "remaining_time": "0:02:28", "throughput": 5545.72, "total_tokens": 21330992} +{"current_steps": 32665, "total_steps": 33920, "loss": 0.25, "lr": 4.170783229670739e-08, "epoch": 19.26002358490566, "percentage": 96.3, "elapsed_time": "1:04:06", "remaining_time": "0:02:27", "throughput": 5545.69, "total_tokens": 21333872} +{"current_steps": 32670, "total_steps": 33920, "loss": 0.3348, "lr": 4.137688238173898e-08, "epoch": 19.26297169811321, "percentage": 96.31, "elapsed_time": "1:04:07", "remaining_time": "0:02:27", "throughput": 5545.55, "total_tokens": 21336144} +{"current_steps": 32675, "total_steps": 33920, "loss": 0.2524, "lr": 4.104724528361137e-08, "epoch": 19.265919811320753, "percentage": 96.33, "elapsed_time": "1:04:07", "remaining_time": "0:02:26", "throughput": 5545.61, "total_tokens": 21339376} +{"current_steps": 32680, "total_steps": 33920, "loss": 0.3792, "lr": 4.071892108959752e-08, "epoch": 19.2688679245283, "percentage": 96.34, "elapsed_time": "1:04:08", "remaining_time": "0:02:26", "throughput": 5545.61, "total_tokens": 21342256} +{"current_steps": 32685, "total_steps": 33920, "loss": 0.3542, "lr": 4.039190988662234e-08, "epoch": 19.27181603773585, "percentage": 96.36, "elapsed_time": "1:04:09", "remaining_time": "0:02:25", "throughput": 5545.6, "total_tokens": 21345136} +{"current_steps": 32690, "total_steps": 33920, "loss": 0.3863, "lr": 4.006621176126435e-08, "epoch": 19.274764150943398, "percentage": 96.37, "elapsed_time": "1:04:09", "remaining_time": "0:02:24", "throughput": 5545.63, "total_tokens": 21348048} +{"current_steps": 32695, "total_steps": 33920, "loss": 0.3497, "lr": 3.974182679975236e-08, "epoch": 19.277712264150942, "percentage": 96.39, "elapsed_time": "1:04:10", "remaining_time": "0:02:24", "throughput": 5545.66, "total_tokens": 21350832} +{"current_steps": 32700, "total_steps": 33920, "loss": 0.2112, "lr": 3.941875508796933e-08, "epoch": 19.28066037735849, "percentage": 96.4, "elapsed_time": "1:04:10", "remaining_time": "0:02:23", "throughput": 5545.57, "total_tokens": 21353200} +{"current_steps": 32705, "total_steps": 33920, "loss": 0.2811, "lr": 3.909699671145017e-08, "epoch": 19.28360849056604, "percentage": 96.42, "elapsed_time": "1:04:11", "remaining_time": "0:02:23", "throughput": 5545.54, "total_tokens": 21355888} +{"current_steps": 32710, "total_steps": 33920, "loss": 0.4086, "lr": 3.87765517553812e-08, "epoch": 19.286556603773583, "percentage": 96.43, "elapsed_time": "1:04:11", "remaining_time": "0:02:22", "throughput": 5545.62, "total_tokens": 21359184} +{"current_steps": 32715, "total_steps": 33920, "loss": 0.3201, "lr": 3.8457420304601756e-08, "epoch": 19.28950471698113, "percentage": 96.45, "elapsed_time": "1:04:12", "remaining_time": "0:02:21", "throughput": 5545.67, "total_tokens": 21362288} +{"current_steps": 32720, "total_steps": 33920, "loss": 0.3213, "lr": 3.813960244360371e-08, "epoch": 19.29245283018868, "percentage": 96.46, "elapsed_time": "1:04:12", "remaining_time": "0:02:21", "throughput": 5545.77, "total_tokens": 21365872} +{"current_steps": 32725, "total_steps": 33920, "loss": 0.3958, "lr": 3.7823098256529744e-08, "epoch": 19.295400943396228, "percentage": 96.48, "elapsed_time": "1:04:13", "remaining_time": "0:02:20", "throughput": 5545.89, "total_tokens": 21369872} +{"current_steps": 32730, "total_steps": 33920, "loss": 0.2422, "lr": 3.750790782717673e-08, "epoch": 19.298349056603772, "percentage": 96.49, "elapsed_time": "1:04:13", "remaining_time": "0:02:20", "throughput": 5545.99, "total_tokens": 21373520} +{"current_steps": 32735, "total_steps": 33920, "loss": 0.1298, "lr": 3.719403123899179e-08, "epoch": 19.30129716981132, "percentage": 96.51, "elapsed_time": "1:04:14", "remaining_time": "0:02:19", "throughput": 5545.99, "total_tokens": 21376560} +{"current_steps": 32740, "total_steps": 33920, "loss": 0.3221, "lr": 3.688146857507624e-08, "epoch": 19.30424528301887, "percentage": 96.52, "elapsed_time": "1:04:15", "remaining_time": "0:02:18", "throughput": 5546.05, "total_tokens": 21380624} +{"current_steps": 32745, "total_steps": 33920, "loss": 0.2795, "lr": 3.657021991818166e-08, "epoch": 19.307193396226417, "percentage": 96.54, "elapsed_time": "1:04:15", "remaining_time": "0:02:18", "throughput": 5545.78, "total_tokens": 21382512} +{"current_steps": 32750, "total_steps": 33920, "loss": 0.3107, "lr": 3.626028535071213e-08, "epoch": 19.31014150943396, "percentage": 96.55, "elapsed_time": "1:04:16", "remaining_time": "0:02:17", "throughput": 5545.95, "total_tokens": 21386288} +{"current_steps": 32755, "total_steps": 33920, "loss": 0.2394, "lr": 3.59516649547248e-08, "epoch": 19.31308962264151, "percentage": 96.57, "elapsed_time": "1:04:16", "remaining_time": "0:02:17", "throughput": 5545.86, "total_tokens": 21389040} +{"current_steps": 32760, "total_steps": 33920, "loss": 0.3191, "lr": 3.564435881192818e-08, "epoch": 19.316037735849058, "percentage": 96.58, "elapsed_time": "1:04:17", "remaining_time": "0:02:16", "throughput": 5545.85, "total_tokens": 21392432} +{"current_steps": 32765, "total_steps": 33920, "loss": 0.232, "lr": 3.5338367003682763e-08, "epoch": 19.318985849056602, "percentage": 96.59, "elapsed_time": "1:04:17", "remaining_time": "0:02:15", "throughput": 5545.77, "total_tokens": 21395184} +{"current_steps": 32770, "total_steps": 33920, "loss": 0.3621, "lr": 3.5033689611000954e-08, "epoch": 19.32193396226415, "percentage": 96.61, "elapsed_time": "1:04:18", "remaining_time": "0:02:15", "throughput": 5545.86, "total_tokens": 21398768} +{"current_steps": 32775, "total_steps": 33920, "loss": 0.3439, "lr": 3.473032671454768e-08, "epoch": 19.3248820754717, "percentage": 96.62, "elapsed_time": "1:04:19", "remaining_time": "0:02:14", "throughput": 5545.83, "total_tokens": 21401808} +{"current_steps": 32780, "total_steps": 33920, "loss": 0.2499, "lr": 3.44282783946398e-08, "epoch": 19.327830188679247, "percentage": 96.64, "elapsed_time": "1:04:19", "remaining_time": "0:02:14", "throughput": 5545.78, "total_tokens": 21404784} +{"current_steps": 32785, "total_steps": 33920, "loss": 0.2285, "lr": 3.4127544731245575e-08, "epoch": 19.33077830188679, "percentage": 96.65, "elapsed_time": "1:04:20", "remaining_time": "0:02:13", "throughput": 5545.94, "total_tokens": 21410128} +{"current_steps": 32790, "total_steps": 33920, "loss": 0.2772, "lr": 3.38281258039852e-08, "epoch": 19.33372641509434, "percentage": 96.67, "elapsed_time": "1:04:21", "remaining_time": "0:02:13", "throughput": 5546.12, "total_tokens": 21414864} +{"current_steps": 32795, "total_steps": 33920, "loss": 0.3525, "lr": 3.353002169213193e-08, "epoch": 19.336674528301888, "percentage": 96.68, "elapsed_time": "1:04:21", "remaining_time": "0:02:12", "throughput": 5546.03, "total_tokens": 21417680} +{"current_steps": 32800, "total_steps": 33920, "loss": 0.511, "lr": 3.32332324746093e-08, "epoch": 19.339622641509433, "percentage": 96.7, "elapsed_time": "1:04:22", "remaining_time": "0:02:11", "throughput": 5546.17, "total_tokens": 21422064} +{"current_steps": 32805, "total_steps": 33920, "loss": 0.2884, "lr": 3.2937758229994455e-08, "epoch": 19.34257075471698, "percentage": 96.71, "elapsed_time": "1:04:23", "remaining_time": "0:02:11", "throughput": 5546.13, "total_tokens": 21424752} +{"current_steps": 32810, "total_steps": 33920, "loss": 0.3546, "lr": 3.2643599036514815e-08, "epoch": 19.34551886792453, "percentage": 96.73, "elapsed_time": "1:04:23", "remaining_time": "0:02:10", "throughput": 5546.16, "total_tokens": 21427792} +{"current_steps": 32815, "total_steps": 33920, "loss": 0.3444, "lr": 3.2350754972050316e-08, "epoch": 19.348466981132077, "percentage": 96.74, "elapsed_time": "1:04:24", "remaining_time": "0:02:10", "throughput": 5546.21, "total_tokens": 21431280} +{"current_steps": 32820, "total_steps": 33920, "loss": 0.1883, "lr": 3.2059226114132815e-08, "epoch": 19.35141509433962, "percentage": 96.76, "elapsed_time": "1:04:24", "remaining_time": "0:02:09", "throughput": 5546.29, "total_tokens": 21434576} +{"current_steps": 32825, "total_steps": 33920, "loss": 0.2569, "lr": 3.1769012539945575e-08, "epoch": 19.35436320754717, "percentage": 96.77, "elapsed_time": "1:04:25", "remaining_time": "0:02:08", "throughput": 5546.23, "total_tokens": 21437456} +{"current_steps": 32830, "total_steps": 33920, "loss": 0.2939, "lr": 3.1480114326324364e-08, "epoch": 19.357311320754718, "percentage": 96.79, "elapsed_time": "1:04:25", "remaining_time": "0:02:08", "throughput": 5546.4, "total_tokens": 21441392} +{"current_steps": 32835, "total_steps": 33920, "loss": 0.3816, "lr": 3.1192531549756325e-08, "epoch": 19.360259433962263, "percentage": 96.8, "elapsed_time": "1:04:26", "remaining_time": "0:02:07", "throughput": 5546.32, "total_tokens": 21444176} +{"current_steps": 32840, "total_steps": 33920, "loss": 0.265, "lr": 3.090626428638e-08, "epoch": 19.36320754716981, "percentage": 96.82, "elapsed_time": "1:04:26", "remaining_time": "0:02:07", "throughput": 5546.13, "total_tokens": 21446480} +{"current_steps": 32845, "total_steps": 33920, "loss": 0.2311, "lr": 3.062131261198531e-08, "epoch": 19.36615566037736, "percentage": 96.83, "elapsed_time": "1:04:27", "remaining_time": "0:02:06", "throughput": 5546.14, "total_tokens": 21449744} +{"current_steps": 32850, "total_steps": 33920, "loss": 0.2605, "lr": 3.033767660201525e-08, "epoch": 19.369103773584907, "percentage": 96.85, "elapsed_time": "1:04:28", "remaining_time": "0:02:05", "throughput": 5546.05, "total_tokens": 21452656} +{"current_steps": 32855, "total_steps": 33920, "loss": 0.2439, "lr": 3.005535633156309e-08, "epoch": 19.372051886792452, "percentage": 96.86, "elapsed_time": "1:04:28", "remaining_time": "0:02:05", "throughput": 5546.08, "total_tokens": 21455664} +{"current_steps": 32860, "total_steps": 33920, "loss": 0.2652, "lr": 2.977435187537514e-08, "epoch": 19.375, "percentage": 96.88, "elapsed_time": "1:04:29", "remaining_time": "0:02:04", "throughput": 5546.04, "total_tokens": 21458224} +{"current_steps": 32865, "total_steps": 33920, "loss": 0.3505, "lr": 2.949466330784745e-08, "epoch": 19.377948113207548, "percentage": 96.89, "elapsed_time": "1:04:29", "remaining_time": "0:02:04", "throughput": 5546.21, "total_tokens": 21462096} +{"current_steps": 32870, "total_steps": 33920, "loss": 0.2921, "lr": 2.921629070302967e-08, "epoch": 19.380896226415093, "percentage": 96.9, "elapsed_time": "1:04:30", "remaining_time": "0:02:03", "throughput": 5546.32, "total_tokens": 21465392} +{"current_steps": 32875, "total_steps": 33920, "loss": 0.3126, "lr": 2.893923413462174e-08, "epoch": 19.38384433962264, "percentage": 96.92, "elapsed_time": "1:04:30", "remaining_time": "0:02:03", "throughput": 5546.35, "total_tokens": 21468304} +{"current_steps": 32880, "total_steps": 33920, "loss": 0.2807, "lr": 2.866349367597554e-08, "epoch": 19.38679245283019, "percentage": 96.93, "elapsed_time": "1:04:31", "remaining_time": "0:02:02", "throughput": 5546.37, "total_tokens": 21471152} +{"current_steps": 32885, "total_steps": 33920, "loss": 0.3506, "lr": 2.8389069400094893e-08, "epoch": 19.389740566037737, "percentage": 96.95, "elapsed_time": "1:04:31", "remaining_time": "0:02:01", "throughput": 5546.52, "total_tokens": 21475280} +{"current_steps": 32890, "total_steps": 33920, "loss": 0.2504, "lr": 2.811596137963446e-08, "epoch": 19.392688679245282, "percentage": 96.96, "elapsed_time": "1:04:32", "remaining_time": "0:02:01", "throughput": 5546.52, "total_tokens": 21478256} +{"current_steps": 32895, "total_steps": 33920, "loss": 0.3474, "lr": 2.7844169686900844e-08, "epoch": 19.39563679245283, "percentage": 96.98, "elapsed_time": "1:04:32", "remaining_time": "0:02:00", "throughput": 5546.62, "total_tokens": 21481744} +{"current_steps": 32900, "total_steps": 33920, "loss": 0.4446, "lr": 2.7573694393852047e-08, "epoch": 19.39858490566038, "percentage": 96.99, "elapsed_time": "1:04:33", "remaining_time": "0:02:00", "throughput": 5546.62, "total_tokens": 21484528} +{"current_steps": 32905, "total_steps": 33920, "loss": 0.2967, "lr": 2.7304535572098e-08, "epoch": 19.401533018867923, "percentage": 97.01, "elapsed_time": "1:04:33", "remaining_time": "0:01:59", "throughput": 5546.74, "total_tokens": 21487888} +{"current_steps": 32910, "total_steps": 33920, "loss": 0.314, "lr": 2.703669329289893e-08, "epoch": 19.40448113207547, "percentage": 97.02, "elapsed_time": "1:04:34", "remaining_time": "0:01:58", "throughput": 5546.81, "total_tokens": 21491120} +{"current_steps": 32915, "total_steps": 33920, "loss": 0.2856, "lr": 2.6770167627167554e-08, "epoch": 19.40742924528302, "percentage": 97.04, "elapsed_time": "1:04:35", "remaining_time": "0:01:58", "throughput": 5546.97, "total_tokens": 21494832} +{"current_steps": 32920, "total_steps": 33920, "loss": 0.3471, "lr": 2.6504958645467426e-08, "epoch": 19.410377358490567, "percentage": 97.05, "elapsed_time": "1:04:35", "remaining_time": "0:01:57", "throughput": 5547.19, "total_tokens": 21499024} +{"current_steps": 32925, "total_steps": 33920, "loss": 0.3767, "lr": 2.6241066418014605e-08, "epoch": 19.413325471698112, "percentage": 97.07, "elapsed_time": "1:04:36", "remaining_time": "0:01:57", "throughput": 5547.3, "total_tokens": 21502480} +{"current_steps": 32930, "total_steps": 33920, "loss": 0.3111, "lr": 2.5978491014674866e-08, "epoch": 19.41627358490566, "percentage": 97.08, "elapsed_time": "1:04:36", "remaining_time": "0:01:56", "throughput": 5547.42, "total_tokens": 21506224} +{"current_steps": 32935, "total_steps": 33920, "loss": 0.2893, "lr": 2.571723250496705e-08, "epoch": 19.41922169811321, "percentage": 97.1, "elapsed_time": "1:04:37", "remaining_time": "0:01:55", "throughput": 5547.57, "total_tokens": 21510736} +{"current_steps": 32940, "total_steps": 33920, "loss": 0.2748, "lr": 2.5457290958059155e-08, "epoch": 19.422169811320753, "percentage": 97.11, "elapsed_time": "1:04:38", "remaining_time": "0:01:55", "throughput": 5547.56, "total_tokens": 21513680} +{"current_steps": 32945, "total_steps": 33920, "loss": 0.2847, "lr": 2.51986664427728e-08, "epoch": 19.4251179245283, "percentage": 97.13, "elapsed_time": "1:04:38", "remaining_time": "0:01:54", "throughput": 5547.57, "total_tokens": 21516496} +{"current_steps": 32950, "total_steps": 33920, "loss": 0.3403, "lr": 2.4941359027579883e-08, "epoch": 19.42806603773585, "percentage": 97.14, "elapsed_time": "1:04:39", "remaining_time": "0:01:54", "throughput": 5547.6, "total_tokens": 21519696} +{"current_steps": 32955, "total_steps": 33920, "loss": 0.2872, "lr": 2.468536878060368e-08, "epoch": 19.431014150943398, "percentage": 97.16, "elapsed_time": "1:04:39", "remaining_time": "0:01:53", "throughput": 5547.54, "total_tokens": 21522320} +{"current_steps": 32960, "total_steps": 33920, "loss": 0.3505, "lr": 2.443069576961832e-08, "epoch": 19.433962264150942, "percentage": 97.17, "elapsed_time": "1:04:40", "remaining_time": "0:01:53", "throughput": 5547.53, "total_tokens": 21525584} +{"current_steps": 32965, "total_steps": 33920, "loss": 0.2515, "lr": 2.4177340062049304e-08, "epoch": 19.43691037735849, "percentage": 97.18, "elapsed_time": "1:04:40", "remaining_time": "0:01:52", "throughput": 5547.66, "total_tokens": 21529200} +{"current_steps": 32970, "total_steps": 33920, "loss": 0.4683, "lr": 2.3925301724974647e-08, "epoch": 19.43985849056604, "percentage": 97.2, "elapsed_time": "1:04:41", "remaining_time": "0:01:51", "throughput": 5547.71, "total_tokens": 21532432} +{"current_steps": 32975, "total_steps": 33920, "loss": 0.2972, "lr": 2.367458082512153e-08, "epoch": 19.442806603773583, "percentage": 97.21, "elapsed_time": "1:04:41", "remaining_time": "0:01:51", "throughput": 5547.82, "total_tokens": 21535984} +{"current_steps": 32980, "total_steps": 33920, "loss": 0.4739, "lr": 2.3425177428870737e-08, "epoch": 19.44575471698113, "percentage": 97.23, "elapsed_time": "1:04:42", "remaining_time": "0:01:50", "throughput": 5547.83, "total_tokens": 21538864} +{"current_steps": 32985, "total_steps": 33920, "loss": 0.2562, "lr": 2.3177091602251677e-08, "epoch": 19.44870283018868, "percentage": 97.24, "elapsed_time": "1:04:42", "remaining_time": "0:01:50", "throughput": 5547.82, "total_tokens": 21541616} +{"current_steps": 32990, "total_steps": 33920, "loss": 0.2768, "lr": 2.2930323410946254e-08, "epoch": 19.451650943396228, "percentage": 97.26, "elapsed_time": "1:04:43", "remaining_time": "0:01:49", "throughput": 5547.87, "total_tokens": 21544816} +{"current_steps": 32995, "total_steps": 33920, "loss": 0.3721, "lr": 2.2684872920287758e-08, "epoch": 19.454599056603772, "percentage": 97.27, "elapsed_time": "1:04:44", "remaining_time": "0:01:48", "throughput": 5547.9, "total_tokens": 21548272} +{"current_steps": 33000, "total_steps": 33920, "loss": 0.289, "lr": 2.2440740195260323e-08, "epoch": 19.45754716981132, "percentage": 97.29, "elapsed_time": "1:04:44", "remaining_time": "0:01:48", "throughput": 5547.97, "total_tokens": 21551472} +{"current_steps": 33005, "total_steps": 33920, "loss": 0.3269, "lr": 2.219792530049891e-08, "epoch": 19.46049528301887, "percentage": 97.3, "elapsed_time": "1:04:45", "remaining_time": "0:01:47", "throughput": 5548.0, "total_tokens": 21554544} +{"current_steps": 33010, "total_steps": 33920, "loss": 0.2761, "lr": 2.1956428300290434e-08, "epoch": 19.463443396226417, "percentage": 97.32, "elapsed_time": "1:04:45", "remaining_time": "0:01:47", "throughput": 5548.07, "total_tokens": 21558000} +{"current_steps": 33015, "total_steps": 33920, "loss": 0.4246, "lr": 2.1716249258570966e-08, "epoch": 19.46639150943396, "percentage": 97.33, "elapsed_time": "1:04:46", "remaining_time": "0:01:46", "throughput": 5548.18, "total_tokens": 21561552} +{"current_steps": 33020, "total_steps": 33920, "loss": 0.2202, "lr": 2.1477388238930196e-08, "epoch": 19.46933962264151, "percentage": 97.35, "elapsed_time": "1:04:46", "remaining_time": "0:01:45", "throughput": 5548.09, "total_tokens": 21564304} +{"current_steps": 33025, "total_steps": 33920, "loss": 0.2556, "lr": 2.1239845304606988e-08, "epoch": 19.472287735849058, "percentage": 97.36, "elapsed_time": "1:04:47", "remaining_time": "0:01:45", "throughput": 5548.01, "total_tokens": 21566992} +{"current_steps": 33030, "total_steps": 33920, "loss": 0.2809, "lr": 2.100362051849214e-08, "epoch": 19.475235849056602, "percentage": 97.38, "elapsed_time": "1:04:48", "remaining_time": "0:01:44", "throughput": 5548.11, "total_tokens": 21571152} +{"current_steps": 33035, "total_steps": 33920, "loss": 0.4432, "lr": 2.076871394312674e-08, "epoch": 19.47818396226415, "percentage": 97.39, "elapsed_time": "1:04:48", "remaining_time": "0:01:44", "throughput": 5548.22, "total_tokens": 21574576} +{"current_steps": 33040, "total_steps": 33920, "loss": 0.3151, "lr": 2.0535125640703813e-08, "epoch": 19.4811320754717, "percentage": 97.41, "elapsed_time": "1:04:49", "remaining_time": "0:01:43", "throughput": 5548.28, "total_tokens": 21577680} +{"current_steps": 33045, "total_steps": 33920, "loss": 0.3463, "lr": 2.0302855673066667e-08, "epoch": 19.484080188679247, "percentage": 97.42, "elapsed_time": "1:04:49", "remaining_time": "0:01:42", "throughput": 5548.18, "total_tokens": 21580016} +{"current_steps": 33050, "total_steps": 33920, "loss": 0.3706, "lr": 2.0071904101710004e-08, "epoch": 19.48702830188679, "percentage": 97.44, "elapsed_time": "1:04:50", "remaining_time": "0:01:42", "throughput": 5548.29, "total_tokens": 21584176} +{"current_steps": 33055, "total_steps": 33920, "loss": 0.3697, "lr": 1.98422709877788e-08, "epoch": 19.48997641509434, "percentage": 97.45, "elapsed_time": "1:04:50", "remaining_time": "0:01:41", "throughput": 5548.27, "total_tokens": 21587248} +{"current_steps": 33060, "total_steps": 33920, "loss": 0.2605, "lr": 1.961395639206942e-08, "epoch": 19.492924528301888, "percentage": 97.46, "elapsed_time": "1:04:51", "remaining_time": "0:01:41", "throughput": 5548.38, "total_tokens": 21591056} +{"current_steps": 33065, "total_steps": 33920, "loss": 0.2315, "lr": 1.9386960375029628e-08, "epoch": 19.495872641509433, "percentage": 97.48, "elapsed_time": "1:04:52", "remaining_time": "0:01:40", "throughput": 5548.48, "total_tokens": 21594704} +{"current_steps": 33070, "total_steps": 33920, "loss": 0.3835, "lr": 1.9161282996757458e-08, "epoch": 19.49882075471698, "percentage": 97.49, "elapsed_time": "1:04:52", "remaining_time": "0:01:40", "throughput": 5548.62, "total_tokens": 21598384} +{"current_steps": 33075, "total_steps": 33920, "loss": 0.2753, "lr": 1.8936924317001225e-08, "epoch": 19.50176886792453, "percentage": 97.51, "elapsed_time": "1:04:53", "remaining_time": "0:01:39", "throughput": 5548.63, "total_tokens": 21601424} +{"current_steps": 33080, "total_steps": 33920, "loss": 0.2768, "lr": 1.8713884395162308e-08, "epoch": 19.504716981132077, "percentage": 97.52, "elapsed_time": "1:04:53", "remaining_time": "0:01:38", "throughput": 5548.66, "total_tokens": 21604336} +{"current_steps": 33085, "total_steps": 33920, "loss": 0.3151, "lr": 1.8492163290290132e-08, "epoch": 19.50766509433962, "percentage": 97.54, "elapsed_time": "1:04:54", "remaining_time": "0:01:38", "throughput": 5548.64, "total_tokens": 21607408} +{"current_steps": 33090, "total_steps": 33920, "loss": 0.1778, "lr": 1.827176106108719e-08, "epoch": 19.51061320754717, "percentage": 97.55, "elapsed_time": "1:04:54", "remaining_time": "0:01:37", "throughput": 5548.74, "total_tokens": 21611056} +{"current_steps": 33095, "total_steps": 33920, "loss": 0.3466, "lr": 1.8052677765905137e-08, "epoch": 19.513561320754718, "percentage": 97.57, "elapsed_time": "1:04:55", "remaining_time": "0:01:37", "throughput": 5548.75, "total_tokens": 21614128} +{"current_steps": 33100, "total_steps": 33920, "loss": 0.3226, "lr": 1.783491346274757e-08, "epoch": 19.516509433962263, "percentage": 97.58, "elapsed_time": "1:04:55", "remaining_time": "0:01:36", "throughput": 5548.75, "total_tokens": 21617104} +{"current_steps": 33105, "total_steps": 33920, "loss": 0.2714, "lr": 1.7618468209268936e-08, "epoch": 19.51945754716981, "percentage": 97.6, "elapsed_time": "1:04:56", "remaining_time": "0:01:35", "throughput": 5548.72, "total_tokens": 21619664} +{"current_steps": 33110, "total_steps": 33920, "loss": 0.2385, "lr": 1.7403342062773943e-08, "epoch": 19.52240566037736, "percentage": 97.61, "elapsed_time": "1:04:56", "remaining_time": "0:01:35", "throughput": 5548.79, "total_tokens": 21623024} +{"current_steps": 33115, "total_steps": 33920, "loss": 0.3371, "lr": 1.718953508021759e-08, "epoch": 19.525353773584907, "percentage": 97.63, "elapsed_time": "1:04:57", "remaining_time": "0:01:34", "throughput": 5548.99, "total_tokens": 21627312} +{"current_steps": 33120, "total_steps": 33920, "loss": 0.4196, "lr": 1.6977047318206262e-08, "epoch": 19.528301886792452, "percentage": 97.64, "elapsed_time": "1:04:58", "remaining_time": "0:01:34", "throughput": 5548.98, "total_tokens": 21630256} +{"current_steps": 33125, "total_steps": 33920, "loss": 0.3704, "lr": 1.676587883299774e-08, "epoch": 19.53125, "percentage": 97.66, "elapsed_time": "1:04:58", "remaining_time": "0:01:33", "throughput": 5549.08, "total_tokens": 21633936} +{"current_steps": 33130, "total_steps": 33920, "loss": 0.2833, "lr": 1.655602968049952e-08, "epoch": 19.534198113207548, "percentage": 97.67, "elapsed_time": "1:04:59", "remaining_time": "0:01:32", "throughput": 5549.27, "total_tokens": 21638128} +{"current_steps": 33135, "total_steps": 33920, "loss": 0.3139, "lr": 1.634749991626938e-08, "epoch": 19.537146226415093, "percentage": 97.69, "elapsed_time": "1:04:59", "remaining_time": "0:01:32", "throughput": 5549.19, "total_tokens": 21640528} +{"current_steps": 33140, "total_steps": 33920, "loss": 0.3991, "lr": 1.6140289595517056e-08, "epoch": 19.54009433962264, "percentage": 97.7, "elapsed_time": "1:05:00", "remaining_time": "0:01:31", "throughput": 5549.29, "total_tokens": 21643856} +{"current_steps": 33145, "total_steps": 33920, "loss": 0.4224, "lr": 1.5934398773102545e-08, "epoch": 19.54304245283019, "percentage": 97.72, "elapsed_time": "1:05:01", "remaining_time": "0:01:31", "throughput": 5549.5, "total_tokens": 21650000} +{"current_steps": 33150, "total_steps": 33920, "loss": 0.3661, "lr": 1.5729827503536133e-08, "epoch": 19.545990566037737, "percentage": 97.73, "elapsed_time": "1:05:01", "remaining_time": "0:01:30", "throughput": 5549.57, "total_tokens": 21652976} +{"current_steps": 33155, "total_steps": 33920, "loss": 0.2984, "lr": 1.5526575840978942e-08, "epoch": 19.548938679245282, "percentage": 97.74, "elapsed_time": "1:05:02", "remaining_time": "0:01:30", "throughput": 5549.64, "total_tokens": 21656400} +{"current_steps": 33160, "total_steps": 33920, "loss": 0.322, "lr": 1.532464383924237e-08, "epoch": 19.55188679245283, "percentage": 97.76, "elapsed_time": "1:05:02", "remaining_time": "0:01:29", "throughput": 5549.63, "total_tokens": 21659184} +{"current_steps": 33165, "total_steps": 33920, "loss": 0.2786, "lr": 1.5124031551789208e-08, "epoch": 19.55483490566038, "percentage": 97.77, "elapsed_time": "1:05:03", "remaining_time": "0:01:28", "throughput": 5549.78, "total_tokens": 21663536} +{"current_steps": 33170, "total_steps": 33920, "loss": 0.2833, "lr": 1.4924739031732527e-08, "epoch": 19.557783018867923, "percentage": 97.79, "elapsed_time": "1:05:04", "remaining_time": "0:01:28", "throughput": 5549.74, "total_tokens": 21666256} +{"current_steps": 33175, "total_steps": 33920, "loss": 0.2987, "lr": 1.4726766331835118e-08, "epoch": 19.56073113207547, "percentage": 97.8, "elapsed_time": "1:05:04", "remaining_time": "0:01:27", "throughput": 5549.66, "total_tokens": 21668752} +{"current_steps": 33180, "total_steps": 33920, "loss": 0.2756, "lr": 1.4530113504512278e-08, "epoch": 19.56367924528302, "percentage": 97.82, "elapsed_time": "1:05:05", "remaining_time": "0:01:27", "throughput": 5549.64, "total_tokens": 21671696} +{"current_steps": 33185, "total_steps": 33920, "loss": 0.4197, "lr": 1.4334780601827914e-08, "epoch": 19.566627358490567, "percentage": 97.83, "elapsed_time": "1:05:05", "remaining_time": "0:01:26", "throughput": 5549.66, "total_tokens": 21674704} +{"current_steps": 33190, "total_steps": 33920, "loss": 0.331, "lr": 1.4140767675497325e-08, "epoch": 19.569575471698112, "percentage": 97.85, "elapsed_time": "1:05:06", "remaining_time": "0:01:25", "throughput": 5549.67, "total_tokens": 21677680} +{"current_steps": 33195, "total_steps": 33920, "loss": 0.3257, "lr": 1.394807477688609e-08, "epoch": 19.57252358490566, "percentage": 97.86, "elapsed_time": "1:05:06", "remaining_time": "0:01:25", "throughput": 5549.73, "total_tokens": 21681168} +{"current_steps": 33200, "total_steps": 33920, "loss": 0.3422, "lr": 1.3756701957011177e-08, "epoch": 19.57547169811321, "percentage": 97.88, "elapsed_time": "1:05:07", "remaining_time": "0:01:24", "throughput": 5549.63, "total_tokens": 21683632} +{"current_steps": 33205, "total_steps": 33920, "loss": 0.4086, "lr": 1.3566649266538723e-08, "epoch": 19.578419811320753, "percentage": 97.89, "elapsed_time": "1:05:07", "remaining_time": "0:01:24", "throughput": 5549.88, "total_tokens": 21688592} +{"current_steps": 33210, "total_steps": 33920, "loss": 0.4606, "lr": 1.3377916755786257e-08, "epoch": 19.5813679245283, "percentage": 97.91, "elapsed_time": "1:05:08", "remaining_time": "0:01:23", "throughput": 5549.92, "total_tokens": 21691824} +{"current_steps": 33215, "total_steps": 33920, "loss": 0.3029, "lr": 1.3190504474721588e-08, "epoch": 19.58431603773585, "percentage": 97.92, "elapsed_time": "1:05:08", "remaining_time": "0:01:22", "throughput": 5549.96, "total_tokens": 21694736} +{"current_steps": 33220, "total_steps": 33920, "loss": 0.3547, "lr": 1.3004412472962802e-08, "epoch": 19.587264150943398, "percentage": 97.94, "elapsed_time": "1:05:09", "remaining_time": "0:01:22", "throughput": 5549.91, "total_tokens": 21697680} +{"current_steps": 33225, "total_steps": 33920, "loss": 0.4412, "lr": 1.2819640799778266e-08, "epoch": 19.590212264150942, "percentage": 97.95, "elapsed_time": "1:05:10", "remaining_time": "0:01:21", "throughput": 5549.88, "total_tokens": 21700304} +{"current_steps": 33230, "total_steps": 33920, "loss": 0.3512, "lr": 1.2636189504087737e-08, "epoch": 19.59316037735849, "percentage": 97.97, "elapsed_time": "1:05:10", "remaining_time": "0:01:21", "throughput": 5549.96, "total_tokens": 21703536} +{"current_steps": 33235, "total_steps": 33920, "loss": 0.2846, "lr": 1.2454058634460142e-08, "epoch": 19.59610849056604, "percentage": 97.98, "elapsed_time": "1:05:11", "remaining_time": "0:01:20", "throughput": 5549.87, "total_tokens": 21706320} +{"current_steps": 33240, "total_steps": 33920, "loss": 0.3257, "lr": 1.2273248239115798e-08, "epoch": 19.599056603773583, "percentage": 98.0, "elapsed_time": "1:05:11", "remaining_time": "0:01:20", "throughput": 5549.86, "total_tokens": 21709232} +{"current_steps": 33245, "total_steps": 33920, "loss": 0.2979, "lr": 1.2093758365924746e-08, "epoch": 19.60200471698113, "percentage": 98.01, "elapsed_time": "1:05:12", "remaining_time": "0:01:19", "throughput": 5549.8, "total_tokens": 21711952} +{"current_steps": 33250, "total_steps": 33920, "loss": 0.2699, "lr": 1.1915589062408417e-08, "epoch": 19.60495283018868, "percentage": 98.02, "elapsed_time": "1:05:12", "remaining_time": "0:01:18", "throughput": 5549.87, "total_tokens": 21715728} +{"current_steps": 33255, "total_steps": 33920, "loss": 0.319, "lr": 1.1738740375736301e-08, "epoch": 19.607900943396228, "percentage": 98.04, "elapsed_time": "1:05:13", "remaining_time": "0:01:18", "throughput": 5549.88, "total_tokens": 21718736} +{"current_steps": 33260, "total_steps": 33920, "loss": 0.3211, "lr": 1.15632123527315e-08, "epoch": 19.610849056603772, "percentage": 98.05, "elapsed_time": "1:05:13", "remaining_time": "0:01:17", "throughput": 5549.91, "total_tokens": 21721840} +{"current_steps": 33265, "total_steps": 33920, "loss": 0.4001, "lr": 1.1389005039865176e-08, "epoch": 19.61379716981132, "percentage": 98.07, "elapsed_time": "1:05:14", "remaining_time": "0:01:17", "throughput": 5549.92, "total_tokens": 21725104} +{"current_steps": 33270, "total_steps": 33920, "loss": 0.2412, "lr": 1.1216118483259875e-08, "epoch": 19.61674528301887, "percentage": 98.08, "elapsed_time": "1:05:15", "remaining_time": "0:01:16", "throughput": 5550.0, "total_tokens": 21728528} +{"current_steps": 33275, "total_steps": 33920, "loss": 0.331, "lr": 1.1044552728687319e-08, "epoch": 19.619693396226417, "percentage": 98.1, "elapsed_time": "1:05:15", "remaining_time": "0:01:15", "throughput": 5550.07, "total_tokens": 21732112} +{"current_steps": 33280, "total_steps": 33920, "loss": 0.2742, "lr": 1.0874307821570618e-08, "epoch": 19.62264150943396, "percentage": 98.11, "elapsed_time": "1:05:16", "remaining_time": "0:01:15", "throughput": 5550.03, "total_tokens": 21734832} +{"current_steps": 33285, "total_steps": 33920, "loss": 0.2943, "lr": 1.0705383806982606e-08, "epoch": 19.62558962264151, "percentage": 98.13, "elapsed_time": "1:05:16", "remaining_time": "0:01:14", "throughput": 5550.22, "total_tokens": 21739152} +{"current_steps": 33290, "total_steps": 33920, "loss": 0.3507, "lr": 1.0537780729646952e-08, "epoch": 19.628537735849058, "percentage": 98.14, "elapsed_time": "1:05:17", "remaining_time": "0:01:14", "throughput": 5550.17, "total_tokens": 21741872} +{"current_steps": 33295, "total_steps": 33920, "loss": 0.2921, "lr": 1.0371498633937605e-08, "epoch": 19.631485849056602, "percentage": 98.16, "elapsed_time": "1:05:17", "remaining_time": "0:01:13", "throughput": 5550.07, "total_tokens": 21744816} +{"current_steps": 33300, "total_steps": 33920, "loss": 0.3165, "lr": 1.0206537563877683e-08, "epoch": 19.63443396226415, "percentage": 98.17, "elapsed_time": "1:05:18", "remaining_time": "0:01:12", "throughput": 5550.18, "total_tokens": 21748272} +{"current_steps": 33305, "total_steps": 33920, "loss": 0.3127, "lr": 1.0042897563141695e-08, "epoch": 19.6373820754717, "percentage": 98.19, "elapsed_time": "1:05:19", "remaining_time": "0:01:12", "throughput": 5550.07, "total_tokens": 21750768} +{"current_steps": 33310, "total_steps": 33920, "loss": 0.2582, "lr": 9.88057867505443e-09, "epoch": 19.640330188679247, "percentage": 98.2, "elapsed_time": "1:05:19", "remaining_time": "0:01:11", "throughput": 5549.99, "total_tokens": 21753520} +{"current_steps": 33315, "total_steps": 33920, "loss": 0.3232, "lr": 9.7195809425904e-09, "epoch": 19.64327830188679, "percentage": 98.22, "elapsed_time": "1:05:20", "remaining_time": "0:01:11", "throughput": 5550.14, "total_tokens": 21757648} +{"current_steps": 33320, "total_steps": 33920, "loss": 0.2076, "lr": 9.559904408373844e-09, "epoch": 19.64622641509434, "percentage": 98.23, "elapsed_time": "1:05:20", "remaining_time": "0:01:10", "throughput": 5549.95, "total_tokens": 21760112} +{"current_steps": 33325, "total_steps": 33920, "loss": 0.2448, "lr": 9.401549114680387e-09, "epoch": 19.649174528301888, "percentage": 98.25, "elapsed_time": "1:05:21", "remaining_time": "0:01:10", "throughput": 5549.93, "total_tokens": 21763600} +{"current_steps": 33330, "total_steps": 33920, "loss": 0.4665, "lr": 9.244515103434826e-09, "epoch": 19.652122641509433, "percentage": 98.26, "elapsed_time": "1:05:21", "remaining_time": "0:01:09", "throughput": 5549.87, "total_tokens": 21766288} +{"current_steps": 33335, "total_steps": 33920, "loss": 0.2828, "lr": 9.08880241621335e-09, "epoch": 19.65507075471698, "percentage": 98.28, "elapsed_time": "1:05:22", "remaining_time": "0:01:08", "throughput": 5549.97, "total_tokens": 21769776} +{"current_steps": 33340, "total_steps": 33920, "loss": 0.319, "lr": 8.934411094240758e-09, "epoch": 19.65801886792453, "percentage": 98.29, "elapsed_time": "1:05:22", "remaining_time": "0:01:08", "throughput": 5549.97, "total_tokens": 21772464} +{"current_steps": 33345, "total_steps": 33920, "loss": 0.3343, "lr": 8.781341178393244e-09, "epoch": 19.660966981132077, "percentage": 98.3, "elapsed_time": "1:05:24", "remaining_time": "0:01:07", "throughput": 5550.2, "total_tokens": 21779440} +{"current_steps": 33350, "total_steps": 33920, "loss": 0.3483, "lr": 8.629592709196167e-09, "epoch": 19.66391509433962, "percentage": 98.32, "elapsed_time": "1:05:24", "remaining_time": "0:01:07", "throughput": 5550.28, "total_tokens": 21783056} +{"current_steps": 33355, "total_steps": 33920, "loss": 0.4259, "lr": 8.479165726826277e-09, "epoch": 19.66686320754717, "percentage": 98.33, "elapsed_time": "1:05:25", "remaining_time": "0:01:06", "throughput": 5550.27, "total_tokens": 21785968} +{"current_steps": 33360, "total_steps": 33920, "loss": 0.5078, "lr": 8.330060271109496e-09, "epoch": 19.669811320754718, "percentage": 98.35, "elapsed_time": "1:05:25", "remaining_time": "0:01:05", "throughput": 5550.36, "total_tokens": 21789520} +{"current_steps": 33365, "total_steps": 33920, "loss": 0.2891, "lr": 8.18227638152258e-09, "epoch": 19.672759433962263, "percentage": 98.36, "elapsed_time": "1:05:26", "remaining_time": "0:01:05", "throughput": 5550.39, "total_tokens": 21792880} +{"current_steps": 33370, "total_steps": 33920, "loss": 0.316, "lr": 8.035814097191452e-09, "epoch": 19.67570754716981, "percentage": 98.38, "elapsed_time": "1:05:26", "remaining_time": "0:01:04", "throughput": 5550.31, "total_tokens": 21795536} +{"current_steps": 33375, "total_steps": 33920, "loss": 0.3587, "lr": 7.890673456892317e-09, "epoch": 19.67865566037736, "percentage": 98.39, "elapsed_time": "1:05:27", "remaining_time": "0:01:04", "throughput": 5550.36, "total_tokens": 21798672} +{"current_steps": 33380, "total_steps": 33920, "loss": 0.2629, "lr": 7.746854499052215e-09, "epoch": 19.681603773584907, "percentage": 98.41, "elapsed_time": "1:05:27", "remaining_time": "0:01:03", "throughput": 5550.3, "total_tokens": 21801456} +{"current_steps": 33385, "total_steps": 33920, "loss": 0.2956, "lr": 7.604357261747907e-09, "epoch": 19.684551886792452, "percentage": 98.42, "elapsed_time": "1:05:28", "remaining_time": "0:01:02", "throughput": 5550.27, "total_tokens": 21804112} +{"current_steps": 33390, "total_steps": 33920, "loss": 0.4919, "lr": 7.463181782705886e-09, "epoch": 19.6875, "percentage": 98.44, "elapsed_time": "1:05:29", "remaining_time": "0:01:02", "throughput": 5550.74, "total_tokens": 21812944} +{"current_steps": 33395, "total_steps": 33920, "loss": 0.2869, "lr": 7.3233280993034726e-09, "epoch": 19.690448113207548, "percentage": 98.45, "elapsed_time": "1:05:30", "remaining_time": "0:01:01", "throughput": 5550.72, "total_tokens": 21815664} +{"current_steps": 33400, "total_steps": 33920, "loss": 0.32, "lr": 7.184796248567161e-09, "epoch": 19.693396226415093, "percentage": 98.47, "elapsed_time": "1:05:30", "remaining_time": "0:01:01", "throughput": 5550.89, "total_tokens": 21819952} +{"current_steps": 33405, "total_steps": 33920, "loss": 0.3458, "lr": 7.047586267173723e-09, "epoch": 19.69634433962264, "percentage": 98.48, "elapsed_time": "1:05:31", "remaining_time": "0:01:00", "throughput": 5550.87, "total_tokens": 21822928} +{"current_steps": 33410, "total_steps": 33920, "loss": 0.3006, "lr": 6.9116981914502114e-09, "epoch": 19.69929245283019, "percentage": 98.5, "elapsed_time": "1:05:31", "remaining_time": "0:01:00", "throughput": 5550.96, "total_tokens": 21826192} +{"current_steps": 33415, "total_steps": 33920, "loss": 0.3033, "lr": 6.7771320573734036e-09, "epoch": 19.702240566037737, "percentage": 98.51, "elapsed_time": "1:05:32", "remaining_time": "0:00:59", "throughput": 5551.1, "total_tokens": 21829776} +{"current_steps": 33420, "total_steps": 33920, "loss": 0.2856, "lr": 6.6438879005709114e-09, "epoch": 19.705188679245282, "percentage": 98.53, "elapsed_time": "1:05:33", "remaining_time": "0:00:58", "throughput": 5551.19, "total_tokens": 21833136} +{"current_steps": 33425, "total_steps": 33920, "loss": 0.4116, "lr": 6.511965756318961e-09, "epoch": 19.70813679245283, "percentage": 98.54, "elapsed_time": "1:05:33", "remaining_time": "0:00:58", "throughput": 5551.17, "total_tokens": 21836016} +{"current_steps": 33430, "total_steps": 33920, "loss": 0.3011, "lr": 6.381365659545169e-09, "epoch": 19.71108490566038, "percentage": 98.56, "elapsed_time": "1:05:34", "remaining_time": "0:00:57", "throughput": 5551.27, "total_tokens": 21839568} +{"current_steps": 33435, "total_steps": 33920, "loss": 0.2938, "lr": 6.252087644825766e-09, "epoch": 19.714033018867923, "percentage": 98.57, "elapsed_time": "1:05:34", "remaining_time": "0:00:57", "throughput": 5551.26, "total_tokens": 21842640} +{"current_steps": 33440, "total_steps": 33920, "loss": 0.2242, "lr": 6.124131746388373e-09, "epoch": 19.71698113207547, "percentage": 98.58, "elapsed_time": "1:05:35", "remaining_time": "0:00:56", "throughput": 5551.22, "total_tokens": 21845168} +{"current_steps": 33445, "total_steps": 33920, "loss": 0.3085, "lr": 5.997497998109225e-09, "epoch": 19.71992924528302, "percentage": 98.6, "elapsed_time": "1:05:35", "remaining_time": "0:00:55", "throughput": 5551.26, "total_tokens": 21848304} +{"current_steps": 33450, "total_steps": 33920, "loss": 0.3063, "lr": 5.8721864335153925e-09, "epoch": 19.722877358490567, "percentage": 98.61, "elapsed_time": "1:05:36", "remaining_time": "0:00:55", "throughput": 5551.19, "total_tokens": 21850800} +{"current_steps": 33455, "total_steps": 33920, "loss": 0.3259, "lr": 5.748197085784224e-09, "epoch": 19.725825471698112, "percentage": 98.63, "elapsed_time": "1:05:36", "remaining_time": "0:00:54", "throughput": 5551.07, "total_tokens": 21853552} +{"current_steps": 33460, "total_steps": 33920, "loss": 0.2549, "lr": 5.62552998774113e-09, "epoch": 19.72877358490566, "percentage": 98.64, "elapsed_time": "1:05:37", "remaining_time": "0:00:54", "throughput": 5551.12, "total_tokens": 21856912} +{"current_steps": 33465, "total_steps": 33920, "loss": 0.2676, "lr": 5.504185171864018e-09, "epoch": 19.73172169811321, "percentage": 98.66, "elapsed_time": "1:05:37", "remaining_time": "0:00:53", "throughput": 5551.13, "total_tokens": 21860016} +{"current_steps": 33470, "total_steps": 33920, "loss": 0.2742, "lr": 5.384162670278858e-09, "epoch": 19.734669811320753, "percentage": 98.67, "elapsed_time": "1:05:38", "remaining_time": "0:00:52", "throughput": 5551.17, "total_tokens": 21863184} +{"current_steps": 33475, "total_steps": 33920, "loss": 0.2323, "lr": 5.265462514762454e-09, "epoch": 19.7376179245283, "percentage": 98.69, "elapsed_time": "1:05:39", "remaining_time": "0:00:52", "throughput": 5551.47, "total_tokens": 21870384} +{"current_steps": 33480, "total_steps": 33920, "loss": 0.2993, "lr": 5.148084736740777e-09, "epoch": 19.74056603773585, "percentage": 98.7, "elapsed_time": "1:05:40", "remaining_time": "0:00:51", "throughput": 5551.4, "total_tokens": 21873008} +{"current_steps": 33485, "total_steps": 33920, "loss": 0.2905, "lr": 5.032029367290081e-09, "epoch": 19.743514150943398, "percentage": 98.72, "elapsed_time": "1:05:40", "remaining_time": "0:00:51", "throughput": 5551.42, "total_tokens": 21876144} +{"current_steps": 33490, "total_steps": 33920, "loss": 0.2959, "lr": 4.917296437136898e-09, "epoch": 19.746462264150942, "percentage": 98.73, "elapsed_time": "1:05:41", "remaining_time": "0:00:50", "throughput": 5551.49, "total_tokens": 21879440} +{"current_steps": 33495, "total_steps": 33920, "loss": 0.3627, "lr": 4.8038859766569305e-09, "epoch": 19.74941037735849, "percentage": 98.75, "elapsed_time": "1:05:41", "remaining_time": "0:00:50", "throughput": 5551.42, "total_tokens": 21882000} +{"current_steps": 33500, "total_steps": 33920, "loss": 0.437, "lr": 4.691798015876714e-09, "epoch": 19.75235849056604, "percentage": 98.76, "elapsed_time": "1:05:42", "remaining_time": "0:00:49", "throughput": 5551.35, "total_tokens": 21884688} +{"current_steps": 33505, "total_steps": 33920, "loss": 0.236, "lr": 4.581032584470846e-09, "epoch": 19.755306603773583, "percentage": 98.78, "elapsed_time": "1:05:42", "remaining_time": "0:00:48", "throughput": 5551.29, "total_tokens": 21887600} +{"current_steps": 33510, "total_steps": 33920, "loss": 0.2244, "lr": 4.471589711766422e-09, "epoch": 19.75825471698113, "percentage": 98.79, "elapsed_time": "1:05:43", "remaining_time": "0:00:48", "throughput": 5551.49, "total_tokens": 21892016} +{"current_steps": 33515, "total_steps": 33920, "loss": 0.4124, "lr": 4.363469426737487e-09, "epoch": 19.76120283018868, "percentage": 98.81, "elapsed_time": "1:05:44", "remaining_time": "0:00:47", "throughput": 5551.66, "total_tokens": 21896112} +{"current_steps": 33520, "total_steps": 33920, "loss": 0.3941, "lr": 4.256671758010588e-09, "epoch": 19.764150943396228, "percentage": 98.82, "elapsed_time": "1:05:44", "remaining_time": "0:00:47", "throughput": 5551.63, "total_tokens": 21899056} +{"current_steps": 33525, "total_steps": 33920, "loss": 0.2016, "lr": 4.151196733859775e-09, "epoch": 19.767099056603772, "percentage": 98.84, "elapsed_time": "1:05:45", "remaining_time": "0:00:46", "throughput": 5551.61, "total_tokens": 21902416} +{"current_steps": 33530, "total_steps": 33920, "loss": 0.308, "lr": 4.047044382211041e-09, "epoch": 19.77004716981132, "percentage": 98.85, "elapsed_time": "1:05:45", "remaining_time": "0:00:45", "throughput": 5551.83, "total_tokens": 21907440} +{"current_steps": 33535, "total_steps": 33920, "loss": 0.3607, "lr": 3.94421473063844e-09, "epoch": 19.77299528301887, "percentage": 98.86, "elapsed_time": "1:05:46", "remaining_time": "0:00:45", "throughput": 5551.96, "total_tokens": 21911408} +{"current_steps": 33540, "total_steps": 33920, "loss": 0.2804, "lr": 3.842707806366863e-09, "epoch": 19.775943396226417, "percentage": 98.88, "elapsed_time": "1:05:47", "remaining_time": "0:00:44", "throughput": 5551.93, "total_tokens": 21914000} +{"current_steps": 33545, "total_steps": 33920, "loss": 0.2364, "lr": 3.742523636270368e-09, "epoch": 19.77889150943396, "percentage": 98.89, "elapsed_time": "1:05:47", "remaining_time": "0:00:44", "throughput": 5551.86, "total_tokens": 21916368} +{"current_steps": 33550, "total_steps": 33920, "loss": 0.229, "lr": 3.6436622468738468e-09, "epoch": 19.78183962264151, "percentage": 98.91, "elapsed_time": "1:05:48", "remaining_time": "0:00:43", "throughput": 5551.95, "total_tokens": 21920016} +{"current_steps": 33555, "total_steps": 33920, "loss": 0.2474, "lr": 3.546123664350254e-09, "epoch": 19.784787735849058, "percentage": 98.92, "elapsed_time": "1:05:48", "remaining_time": "0:00:42", "throughput": 5552.06, "total_tokens": 21923664} +{"current_steps": 33560, "total_steps": 33920, "loss": 0.318, "lr": 3.449907914524486e-09, "epoch": 19.787735849056602, "percentage": 98.94, "elapsed_time": "1:05:49", "remaining_time": "0:00:42", "throughput": 5552.04, "total_tokens": 21926544} +{"current_steps": 33565, "total_steps": 33920, "loss": 0.2196, "lr": 3.355015022869501e-09, "epoch": 19.79068396226415, "percentage": 98.95, "elapsed_time": "1:05:49", "remaining_time": "0:00:41", "throughput": 5552.05, "total_tokens": 21929392} +{"current_steps": 33570, "total_steps": 33920, "loss": 0.2177, "lr": 3.261445014508535e-09, "epoch": 19.7936320754717, "percentage": 98.97, "elapsed_time": "1:05:50", "remaining_time": "0:00:41", "throughput": 5552.04, "total_tokens": 21932304} +{"current_steps": 33575, "total_steps": 33920, "loss": 0.4029, "lr": 3.1691979142145503e-09, "epoch": 19.796580188679247, "percentage": 98.98, "elapsed_time": "1:05:50", "remaining_time": "0:00:40", "throughput": 5552.14, "total_tokens": 21936144} +{"current_steps": 33580, "total_steps": 33920, "loss": 0.2289, "lr": 3.078273746410787e-09, "epoch": 19.79952830188679, "percentage": 99.0, "elapsed_time": "1:05:51", "remaining_time": "0:00:40", "throughput": 5552.07, "total_tokens": 21938672} +{"current_steps": 33585, "total_steps": 33920, "loss": 0.3036, "lr": 2.988672535169657e-09, "epoch": 19.80247641509434, "percentage": 99.01, "elapsed_time": "1:05:51", "remaining_time": "0:00:39", "throughput": 5551.94, "total_tokens": 21941072} +{"current_steps": 33590, "total_steps": 33920, "loss": 0.3458, "lr": 2.9003943042127393e-09, "epoch": 19.805424528301888, "percentage": 99.03, "elapsed_time": "1:05:52", "remaining_time": "0:00:38", "throughput": 5551.94, "total_tokens": 21943952} +{"current_steps": 33595, "total_steps": 33920, "loss": 0.2709, "lr": 2.8134390769135598e-09, "epoch": 19.808372641509433, "percentage": 99.04, "elapsed_time": "1:05:53", "remaining_time": "0:00:38", "throughput": 5552.03, "total_tokens": 21947376} +{"current_steps": 33600, "total_steps": 33920, "loss": 0.3813, "lr": 2.7278068762925935e-09, "epoch": 19.81132075471698, "percentage": 99.06, "elapsed_time": "1:05:53", "remaining_time": "0:00:37", "throughput": 5552.17, "total_tokens": 21951152} +{"current_steps": 33605, "total_steps": 33920, "loss": 0.3169, "lr": 2.6434977250217043e-09, "epoch": 19.81426886792453, "percentage": 99.07, "elapsed_time": "1:05:54", "remaining_time": "0:00:37", "throughput": 5552.16, "total_tokens": 21954000} +{"current_steps": 33610, "total_steps": 33920, "loss": 0.3972, "lr": 2.560511645422481e-09, "epoch": 19.817216981132077, "percentage": 99.09, "elapsed_time": "1:05:54", "remaining_time": "0:00:36", "throughput": 5552.23, "total_tokens": 21957424} +{"current_steps": 33615, "total_steps": 33920, "loss": 0.2818, "lr": 2.4788486594656825e-09, "epoch": 19.82016509433962, "percentage": 99.1, "elapsed_time": "1:05:55", "remaining_time": "0:00:35", "throughput": 5552.39, "total_tokens": 21961456} +{"current_steps": 33620, "total_steps": 33920, "loss": 0.284, "lr": 2.398508788771792e-09, "epoch": 19.82311320754717, "percentage": 99.12, "elapsed_time": "1:05:55", "remaining_time": "0:00:35", "throughput": 5552.46, "total_tokens": 21964720} +{"current_steps": 33625, "total_steps": 33920, "loss": 0.2457, "lr": 2.3194920546110166e-09, "epoch": 19.826061320754718, "percentage": 99.13, "elapsed_time": "1:05:56", "remaining_time": "0:00:34", "throughput": 5552.5, "total_tokens": 21967696} +{"current_steps": 33630, "total_steps": 33920, "loss": 0.4096, "lr": 2.2417984779032896e-09, "epoch": 19.829009433962263, "percentage": 99.15, "elapsed_time": "1:05:56", "remaining_time": "0:00:34", "throughput": 5552.42, "total_tokens": 21970224} +{"current_steps": 33635, "total_steps": 33920, "loss": 0.3042, "lr": 2.1654280792193782e-09, "epoch": 19.83195754716981, "percentage": 99.16, "elapsed_time": "1:05:57", "remaining_time": "0:00:33", "throughput": 5552.56, "total_tokens": 21973744} +{"current_steps": 33640, "total_steps": 33920, "loss": 0.1963, "lr": 2.0903808787769987e-09, "epoch": 19.83490566037736, "percentage": 99.17, "elapsed_time": "1:05:58", "remaining_time": "0:00:32", "throughput": 5552.63, "total_tokens": 21977360} +{"current_steps": 33645, "total_steps": 33920, "loss": 0.2665, "lr": 2.0166568964463673e-09, "epoch": 19.837853773584907, "percentage": 99.19, "elapsed_time": "1:05:58", "remaining_time": "0:00:32", "throughput": 5552.68, "total_tokens": 21981104} +{"current_steps": 33650, "total_steps": 33920, "loss": 0.3657, "lr": 1.9442561517463153e-09, "epoch": 19.840801886792452, "percentage": 99.2, "elapsed_time": "1:05:59", "remaining_time": "0:00:31", "throughput": 5552.71, "total_tokens": 21984112} +{"current_steps": 33655, "total_steps": 33920, "loss": 0.2457, "lr": 1.8731786638442886e-09, "epoch": 19.84375, "percentage": 99.22, "elapsed_time": "1:05:59", "remaining_time": "0:00:31", "throughput": 5552.75, "total_tokens": 21987536} +{"current_steps": 33660, "total_steps": 33920, "loss": 0.247, "lr": 1.8034244515591214e-09, "epoch": 19.846698113207548, "percentage": 99.23, "elapsed_time": "1:06:00", "remaining_time": "0:00:30", "throughput": 5552.76, "total_tokens": 21990512} +{"current_steps": 33665, "total_steps": 33920, "loss": 0.2883, "lr": 1.7349935333582646e-09, "epoch": 19.849646226415093, "percentage": 99.25, "elapsed_time": "1:06:00", "remaining_time": "0:00:30", "throughput": 5552.81, "total_tokens": 21993520} +{"current_steps": 33670, "total_steps": 33920, "loss": 0.2601, "lr": 1.6678859273594471e-09, "epoch": 19.85259433962264, "percentage": 99.26, "elapsed_time": "1:06:01", "remaining_time": "0:00:29", "throughput": 5552.84, "total_tokens": 21997008} +{"current_steps": 33675, "total_steps": 33920, "loss": 0.2585, "lr": 1.6021016513295683e-09, "epoch": 19.85554245283019, "percentage": 99.28, "elapsed_time": "1:06:01", "remaining_time": "0:00:28", "throughput": 5552.89, "total_tokens": 22000080} +{"current_steps": 33680, "total_steps": 33920, "loss": 0.3228, "lr": 1.5376407226846968e-09, "epoch": 19.858490566037737, "percentage": 99.29, "elapsed_time": "1:06:02", "remaining_time": "0:00:28", "throughput": 5552.86, "total_tokens": 22002896} +{"current_steps": 33685, "total_steps": 33920, "loss": 0.2822, "lr": 1.4745031584917357e-09, "epoch": 19.861438679245282, "percentage": 99.31, "elapsed_time": "1:06:03", "remaining_time": "0:00:27", "throughput": 5552.93, "total_tokens": 22006352} +{"current_steps": 33690, "total_steps": 33920, "loss": 0.5709, "lr": 1.4126889754667583e-09, "epoch": 19.86438679245283, "percentage": 99.32, "elapsed_time": "1:06:03", "remaining_time": "0:00:27", "throughput": 5552.78, "total_tokens": 22008496} +{"current_steps": 33695, "total_steps": 33920, "loss": 0.3793, "lr": 1.3521981899750069e-09, "epoch": 19.86733490566038, "percentage": 99.34, "elapsed_time": "1:06:04", "remaining_time": "0:00:26", "throughput": 5552.8, "total_tokens": 22011536} +{"current_steps": 33700, "total_steps": 33920, "loss": 0.3883, "lr": 1.293030818032004e-09, "epoch": 19.870283018867923, "percentage": 99.35, "elapsed_time": "1:06:04", "remaining_time": "0:00:25", "throughput": 5553.06, "total_tokens": 22017008} +{"current_steps": 33705, "total_steps": 33920, "loss": 0.2853, "lr": 1.2351868753018858e-09, "epoch": 19.87323113207547, "percentage": 99.37, "elapsed_time": "1:06:05", "remaining_time": "0:00:25", "throughput": 5553.02, "total_tokens": 22020080} +{"current_steps": 33710, "total_steps": 33920, "loss": 0.352, "lr": 1.1786663770996242e-09, "epoch": 19.87617924528302, "percentage": 99.38, "elapsed_time": "1:06:06", "remaining_time": "0:00:24", "throughput": 5552.9, "total_tokens": 22022928} +{"current_steps": 33715, "total_steps": 33920, "loss": 0.3233, "lr": 1.1234693383893602e-09, "epoch": 19.879127358490567, "percentage": 99.4, "elapsed_time": "1:06:06", "remaining_time": "0:00:24", "throughput": 5552.91, "total_tokens": 22026160} +{"current_steps": 33720, "total_steps": 33920, "loss": 0.2999, "lr": 1.0695957737844043e-09, "epoch": 19.882075471698112, "percentage": 99.41, "elapsed_time": "1:06:07", "remaining_time": "0:00:23", "throughput": 5552.93, "total_tokens": 22029104} +{"current_steps": 33725, "total_steps": 33920, "loss": 0.2683, "lr": 1.0170456975483467e-09, "epoch": 19.88502358490566, "percentage": 99.43, "elapsed_time": "1:06:07", "remaining_time": "0:00:22", "throughput": 5552.95, "total_tokens": 22032208} +{"current_steps": 33730, "total_steps": 33920, "loss": 0.2666, "lr": 9.658191235933922e-10, "epoch": 19.88797169811321, "percentage": 99.44, "elapsed_time": "1:06:08", "remaining_time": "0:00:22", "throughput": 5553.02, "total_tokens": 22035408} +{"current_steps": 33735, "total_steps": 33920, "loss": 0.3281, "lr": 9.159160654825805e-10, "epoch": 19.890919811320753, "percentage": 99.45, "elapsed_time": "1:06:08", "remaining_time": "0:00:21", "throughput": 5553.02, "total_tokens": 22038544} +{"current_steps": 33740, "total_steps": 33920, "loss": 0.4378, "lr": 8.673365364281205e-10, "epoch": 19.8938679245283, "percentage": 99.47, "elapsed_time": "1:06:09", "remaining_time": "0:00:21", "throughput": 5553.12, "total_tokens": 22042576} +{"current_steps": 33745, "total_steps": 33920, "loss": 0.3309, "lr": 8.200805492913911e-10, "epoch": 19.89681603773585, "percentage": 99.48, "elapsed_time": "1:06:09", "remaining_time": "0:00:20", "throughput": 5553.08, "total_tokens": 22045552} +{"current_steps": 33750, "total_steps": 33920, "loss": 0.2668, "lr": 7.741481165834952e-10, "epoch": 19.899764150943398, "percentage": 99.5, "elapsed_time": "1:06:10", "remaining_time": "0:00:19", "throughput": 5553.09, "total_tokens": 22048560} +{"current_steps": 33755, "total_steps": 33920, "loss": 0.2916, "lr": 7.29539250465261e-10, "epoch": 19.902712264150942, "percentage": 99.51, "elapsed_time": "1:06:11", "remaining_time": "0:00:19", "throughput": 5553.07, "total_tokens": 22051472} +{"current_steps": 33760, "total_steps": 33920, "loss": 0.2471, "lr": 6.862539627472409e-10, "epoch": 19.90566037735849, "percentage": 99.53, "elapsed_time": "1:06:11", "remaining_time": "0:00:18", "throughput": 5553.08, "total_tokens": 22054704} +{"current_steps": 33765, "total_steps": 33920, "loss": 0.3397, "lr": 6.442922648897121e-10, "epoch": 19.90860849056604, "percentage": 99.54, "elapsed_time": "1:06:12", "remaining_time": "0:00:18", "throughput": 5553.15, "total_tokens": 22058224} +{"current_steps": 33770, "total_steps": 33920, "loss": 0.4247, "lr": 6.036541680015662e-10, "epoch": 19.911556603773583, "percentage": 99.56, "elapsed_time": "1:06:12", "remaining_time": "0:00:17", "throughput": 5553.36, "total_tokens": 22062736} +{"current_steps": 33775, "total_steps": 33920, "loss": 0.2853, "lr": 5.643396828419745e-10, "epoch": 19.91450471698113, "percentage": 99.57, "elapsed_time": "1:06:13", "remaining_time": "0:00:17", "throughput": 5553.58, "total_tokens": 22067152} +{"current_steps": 33780, "total_steps": 33920, "loss": 0.3309, "lr": 5.26348819819833e-10, "epoch": 19.91745283018868, "percentage": 99.59, "elapsed_time": "1:06:14", "remaining_time": "0:00:16", "throughput": 5553.62, "total_tokens": 22070128} +{"current_steps": 33785, "total_steps": 33920, "loss": 0.3157, "lr": 4.896815889937622e-10, "epoch": 19.920400943396228, "percentage": 99.6, "elapsed_time": "1:06:14", "remaining_time": "0:00:15", "throughput": 5553.68, "total_tokens": 22073104} +{"current_steps": 33790, "total_steps": 33920, "loss": 0.4012, "lr": 4.543380000704423e-10, "epoch": 19.923349056603772, "percentage": 99.62, "elapsed_time": "1:06:15", "remaining_time": "0:00:15", "throughput": 5553.83, "total_tokens": 22076880} +{"current_steps": 33795, "total_steps": 33920, "loss": 0.4264, "lr": 4.203180624084979e-10, "epoch": 19.92629716981132, "percentage": 99.63, "elapsed_time": "1:06:15", "remaining_time": "0:00:14", "throughput": 5553.89, "total_tokens": 22080272} +{"current_steps": 33800, "total_steps": 33920, "loss": 0.4483, "lr": 3.876217850146136e-10, "epoch": 19.92924528301887, "percentage": 99.65, "elapsed_time": "1:06:16", "remaining_time": "0:00:14", "throughput": 5553.78, "total_tokens": 22082672} +{"current_steps": 33805, "total_steps": 33920, "loss": 0.2803, "lr": 3.56249176544643e-10, "epoch": 19.932193396226417, "percentage": 99.66, "elapsed_time": "1:06:16", "remaining_time": "0:00:13", "throughput": 5553.71, "total_tokens": 22085456} +{"current_steps": 33810, "total_steps": 33920, "loss": 0.3158, "lr": 3.262002453047197e-10, "epoch": 19.93514150943396, "percentage": 99.68, "elapsed_time": "1:06:17", "remaining_time": "0:00:12", "throughput": 5553.82, "total_tokens": 22089072} +{"current_steps": 33815, "total_steps": 33920, "loss": 0.3976, "lr": 2.974749992512571e-10, "epoch": 19.93808962264151, "percentage": 99.69, "elapsed_time": "1:06:17", "remaining_time": "0:00:12", "throughput": 5553.76, "total_tokens": 22091632} +{"current_steps": 33820, "total_steps": 33920, "loss": 0.3996, "lr": 2.700734459881726e-10, "epoch": 19.941037735849058, "percentage": 99.71, "elapsed_time": "1:06:18", "remaining_time": "0:00:11", "throughput": 5553.81, "total_tokens": 22094608} +{"current_steps": 33825, "total_steps": 33920, "loss": 0.3717, "lr": 2.439955927713289e-10, "epoch": 19.943985849056602, "percentage": 99.72, "elapsed_time": "1:06:18", "remaining_time": "0:00:11", "throughput": 5553.96, "total_tokens": 22098768} +{"current_steps": 33830, "total_steps": 33920, "loss": 0.3326, "lr": 2.1924144650409263e-10, "epoch": 19.94693396226415, "percentage": 99.73, "elapsed_time": "1:06:19", "remaining_time": "0:00:10", "throughput": 5554.11, "total_tokens": 22102512} +{"current_steps": 33835, "total_steps": 33920, "loss": 0.3303, "lr": 1.9581101374066546e-10, "epoch": 19.9498820754717, "percentage": 99.75, "elapsed_time": "1:06:20", "remaining_time": "0:00:09", "throughput": 5554.01, "total_tokens": 22105040} +{"current_steps": 33840, "total_steps": 33920, "loss": 0.3058, "lr": 1.7370430068441858e-10, "epoch": 19.952830188679247, "percentage": 99.76, "elapsed_time": "1:06:20", "remaining_time": "0:00:09", "throughput": 5554.08, "total_tokens": 22108400} +{"current_steps": 33845, "total_steps": 33920, "loss": 0.2887, "lr": 1.529213131878926e-10, "epoch": 19.95577830188679, "percentage": 99.78, "elapsed_time": "1:06:21", "remaining_time": "0:00:08", "throughput": 5554.02, "total_tokens": 22111376} +{"current_steps": 33850, "total_steps": 33920, "loss": 0.342, "lr": 1.3346205675335288e-10, "epoch": 19.95872641509434, "percentage": 99.79, "elapsed_time": "1:06:21", "remaining_time": "0:00:08", "throughput": 5554.05, "total_tokens": 22114416} +{"current_steps": 33855, "total_steps": 33920, "loss": 0.3784, "lr": 1.1532653653334447e-10, "epoch": 19.961674528301888, "percentage": 99.81, "elapsed_time": "1:06:22", "remaining_time": "0:00:07", "throughput": 5554.09, "total_tokens": 22117328} +{"current_steps": 33860, "total_steps": 33920, "loss": 0.2831, "lr": 9.85147573284717e-11, "epoch": 19.964622641509433, "percentage": 99.82, "elapsed_time": "1:06:22", "remaining_time": "0:00:07", "throughput": 5554.21, "total_tokens": 22120880} +{"current_steps": 33865, "total_steps": 33920, "loss": 0.2299, "lr": 8.302672359072894e-11, "epoch": 19.96757075471698, "percentage": 99.84, "elapsed_time": "1:06:23", "remaining_time": "0:00:06", "throughput": 5554.28, "total_tokens": 22124144} +{"current_steps": 33870, "total_steps": 33920, "loss": 0.3609, "lr": 6.886243941961468e-11, "epoch": 19.97051886792453, "percentage": 99.85, "elapsed_time": "1:06:23", "remaining_time": "0:00:05", "throughput": 5554.33, "total_tokens": 22127440} +{"current_steps": 33875, "total_steps": 33920, "loss": 0.2278, "lr": 5.602190856601741e-11, "epoch": 19.973466981132077, "percentage": 99.87, "elapsed_time": "1:06:24", "remaining_time": "0:00:05", "throughput": 5554.34, "total_tokens": 22130544} +{"current_steps": 33880, "total_steps": 33920, "loss": 0.2564, "lr": 4.450513442888493e-11, "epoch": 19.97641509433962, "percentage": 99.88, "elapsed_time": "1:06:24", "remaining_time": "0:00:04", "throughput": 5554.41, "total_tokens": 22134096} +{"current_steps": 33885, "total_steps": 33920, "loss": 0.3344, "lr": 3.4312120057999886e-11, "epoch": 19.97936320754717, "percentage": 99.9, "elapsed_time": "1:06:25", "remaining_time": "0:00:04", "throughput": 5554.44, "total_tokens": 22137360} +{"current_steps": 33890, "total_steps": 33920, "loss": 0.3701, "lr": 2.5442868151204224e-11, "epoch": 19.982311320754718, "percentage": 99.91, "elapsed_time": "1:06:26", "remaining_time": "0:00:03", "throughput": 5554.43, "total_tokens": 22140752} +{"current_steps": 33895, "total_steps": 33920, "loss": 0.382, "lr": 1.7897381057729867e-11, "epoch": 19.985259433962263, "percentage": 99.93, "elapsed_time": "1:06:26", "remaining_time": "0:00:02", "throughput": 5554.31, "total_tokens": 22143120} +{"current_steps": 33900, "total_steps": 33920, "loss": 0.2664, "lr": 1.1675660773757813e-11, "epoch": 19.98820754716981, "percentage": 99.94, "elapsed_time": "1:06:27", "remaining_time": "0:00:02", "throughput": 5554.46, "total_tokens": 22146864} +{"current_steps": 33905, "total_steps": 33920, "loss": 0.2647, "lr": 6.777708947969253e-12, "epoch": 19.99115566037736, "percentage": 99.96, "elapsed_time": "1:06:27", "remaining_time": "0:00:01", "throughput": 5554.5, "total_tokens": 22149808} +{"current_steps": 33910, "total_steps": 33920, "loss": 0.202, "lr": 3.2035268765495674e-12, "epoch": 19.994103773584907, "percentage": 99.97, "elapsed_time": "1:06:28", "remaining_time": "0:00:01", "throughput": 5554.45, "total_tokens": 22152592} +{"current_steps": 33915, "total_steps": 33920, "loss": 0.2234, "lr": 9.531155059638863e-13, "epoch": 19.997051886792452, "percentage": 99.99, "elapsed_time": "1:06:28", "remaining_time": "0:00:00", "throughput": 5554.42, "total_tokens": 22155184} +{"current_steps": 33920, "total_steps": 33920, "loss": 0.3676, "lr": 2.6475431291750342e-14, "epoch": 20.0, "percentage": 100.0, "elapsed_time": "1:06:29", "remaining_time": "0:00:00", "throughput": 5554.07, "total_tokens": 22157104} +{"current_steps": 33920, "total_steps": 33920, "eval_loss": 0.5998956561088562, "epoch": 20.0, "percentage": 100.0, "elapsed_time": "1:06:48", "remaining_time": "0:00:00", "throughput": 5527.66, "total_tokens": 22157104} +{"current_steps": 33920, "total_steps": 33920, "epoch": 20.0, "percentage": 100.0, "elapsed_time": "1:06:50", "remaining_time": "0:00:00", "throughput": 5524.28, "total_tokens": 22157104}