{ "best_global_step": 3000, "best_metric": 2.91798250107805, "best_model_checkpoint": "/gpfs/scratch/guoh/DNAFM/output/gencode_human_12.8k_12800/HNet_Ori-BPT3/checkpoint-3000", "epoch": 0.8510034749308559, "eval_steps": 500, "global_step": 3000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.0028366782497695198, "grad_norm": 590.2120361328125, "loss": 144.5784, "loss_ce": 170.91241455078125, "loss_region": 0.030412333086133003, "loss_total": 170.9428253173828, "lr": 2.20454076850486e-05, "router/selected_tokens_s0": 1.0, "step": 10, "tokens_trained": 0.03276544 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.0056733564995390395, "grad_norm": 565.2921142578125, "loss": 52.047, "loss_ce": 28.61202049255371, "loss_region": 0.03181665763258934, "loss_total": 28.643836975097656, "lr": 4.654030511288038e-05, "router/selected_tokens_s0": 1.0, "step": 20, "tokens_trained": 0.06553088 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.00851003474930856, "grad_norm": 361.24432373046875, "loss": 18.4265, "loss_ce": 16.737817764282227, "loss_region": 0.03595759719610214, "loss_total": 16.773775100708008, "lr": 7.103520254071216e-05, "router/selected_tokens_s0": 1.0, "step": 30, "tokens_trained": 0.09829632 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.011346712999078079, "grad_norm": 649.0695190429688, "loss": 8.0445, "loss_ce": 11.410881996154785, "loss_region": 0.03821098059415817, "loss_total": 11.449092864990234, "lr": 9.553009996854394e-05, "router/selected_tokens_s0": 1.0, "step": 40, "tokens_trained": 0.13106176 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.014183391248847599, "grad_norm": 534.2383422851562, "loss": 9.3219, "loss_ce": 9.884474754333496, "loss_region": 0.040100596845149994, "loss_total": 9.924575805664062, "lr": 0.00012002499739637572, "router/selected_tokens_s0": 1.0, "step": 50, "tokens_trained": 0.1638272 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.01702006949861712, "grad_norm": 273.8401184082031, "loss": 14.1755, "loss_ce": 12.677406311035156, "loss_region": 0.041250791400671005, "loss_total": 12.718657493591309, "lr": 0.00014451989482420748, "router/selected_tokens_s0": 1.0, "step": 60, "tokens_trained": 0.19659264 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.01985674774838664, "grad_norm": 544.6290893554688, "loss": 14.1136, "loss_ce": 14.262775421142578, "loss_region": 0.042144227772951126, "loss_total": 14.304919242858887, "lr": 0.00016901479225203927, "router/selected_tokens_s0": 1.0, "step": 70, "tokens_trained": 0.22935808 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.022693425998156158, "grad_norm": 527.1918334960938, "loss": 15.2492, "loss_ce": 11.932450294494629, "loss_region": 0.04246167093515396, "loss_total": 11.9749116897583, "lr": 0.00019350968967987104, "router/selected_tokens_s0": 1.0, "step": 80, "tokens_trained": 0.26212192 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.025530104247925678, "grad_norm": 343.09454345703125, "loss": 12.0101, "loss_ce": 6.092933177947998, "loss_region": 0.04214272275567055, "loss_total": 6.13507604598999, "lr": 0.0002180045871077028, "router/selected_tokens_s0": 1.0, "step": 90, "tokens_trained": 0.29488736 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.028366782497695198, "grad_norm": 309.79541015625, "loss": 9.8843, "loss_ce": 5.214886665344238, "loss_region": 0.041769951581954956, "loss_total": 5.256656646728516, "lr": 0.00024249948453553463, "router/selected_tokens_s0": 1.0, "step": 100, "tokens_trained": 0.3276528 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.031203460747464717, "grad_norm": 251.26068115234375, "loss": 8.5835, "loss_ce": 12.269608497619629, "loss_region": 0.04041137546300888, "loss_total": 12.310019493103027, "lr": 0.00026699438196336637, "router/selected_tokens_s0": 1.0, "step": 110, "tokens_trained": 0.36041744 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.03404013899723424, "grad_norm": 148.94601440429688, "loss": 6.4366, "loss_ce": 3.2050940990448, "loss_region": 0.03642381727695465, "loss_total": 3.241518020629883, "lr": 0.00029148927939119814, "router/selected_tokens_s0": 1.0, "step": 120, "tokens_trained": 0.39318128 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.03687681724700376, "grad_norm": 187.2681427001953, "loss": 10.4928, "loss_ce": 6.001107215881348, "loss_region": 0.030254848301410675, "loss_total": 6.031362056732178, "lr": 0.00031598417681902996, "router/selected_tokens_s0": 4752.0, "step": 130, "tokens_trained": 0.42594672 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.03971349549677328, "grad_norm": 218.34559631347656, "loss": 8.5742, "loss_ce": 3.848691701889038, "loss_region": 0.03400004655122757, "loss_total": 3.8826918601989746, "lr": 0.00034047907424686173, "router/selected_tokens_s0": 7042.125, "step": 140, "tokens_trained": 0.458709112 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.042550173746542796, "grad_norm": 215.60699462890625, "loss": 4.5762, "loss_ce": 5.0876851081848145, "loss_region": 0.03198177367448807, "loss_total": 5.119667053222656, "lr": 0.0003649739716746935, "router/selected_tokens_s0": 424.5, "step": 150, "tokens_trained": 0.491469992 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.045386851996312316, "grad_norm": 147.6339111328125, "loss": 5.8047, "loss_ce": 8.435795783996582, "loss_region": 0.03364315256476402, "loss_total": 8.469438552856445, "lr": 0.00038946886910252526, "router/selected_tokens_s0": 536.875, "step": 160, "tokens_trained": 0.524234632 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.048223530246081836, "grad_norm": 218.0553741455078, "loss": 5.7968, "loss_ce": 6.644444942474365, "loss_region": 0.031727153807878494, "loss_total": 6.676172256469727, "lr": 0.0004139637665303571, "router/selected_tokens_s0": 1833.5, "step": 170, "tokens_trained": 0.556999272 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.051060208495851356, "grad_norm": 100.56309509277344, "loss": 6.7503, "loss_ce": 8.332029342651367, "loss_region": 0.03232778236269951, "loss_total": 8.364356994628906, "lr": 0.0004384586639581888, "router/selected_tokens_s0": 1649.75, "step": 180, "tokens_trained": 0.589762952 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.053896886745620876, "grad_norm": 157.10765075683594, "loss": 6.4449, "loss_ce": 4.925128936767578, "loss_region": 0.031663134694099426, "loss_total": 4.956791877746582, "lr": 0.0004629535613860206, "router/selected_tokens_s0": 1687.375, "step": 190, "tokens_trained": 0.622527592 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.056733564995390395, "grad_norm": 83.81340026855469, "loss": 3.7524, "loss_ce": 5.0940961837768555, "loss_region": 0.02894311398267746, "loss_total": 5.123039245605469, "lr": 0.00048744845881385244, "router/selected_tokens_s0": 3074.125, "step": 200, "tokens_trained": 0.655293032 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.059570243245159915, "grad_norm": 169.4013671875, "loss": 5.9481, "loss_ce": 9.220865249633789, "loss_region": 0.02949724718928337, "loss_total": 9.250362396240234, "lr": 0.0005119433562416841, "router/selected_tokens_s0": 3610.375, "step": 210, "tokens_trained": 0.688057672 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.062406921494929435, "grad_norm": 80.7753677368164, "loss": 5.1122, "loss_ce": 3.287958860397339, "loss_region": 0.029488109052181244, "loss_total": 3.3174469470977783, "lr": 0.0005364382536695159, "router/selected_tokens_s0": 2584.75, "step": 220, "tokens_trained": 0.720823112 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.06524359974469895, "grad_norm": 89.39635467529297, "loss": 3.3047, "loss_ce": 2.1086361408233643, "loss_region": 0.029821382835507393, "loss_total": 2.1384575366973877, "lr": 0.0005609331510973477, "router/selected_tokens_s0": 3991.5, "step": 230, "tokens_trained": 0.753588552 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.06808027799446847, "grad_norm": 90.892333984375, "loss": 4.2563, "loss_ce": 2.7003867626190186, "loss_region": 0.030828693881630898, "loss_total": 2.731215476989746, "lr": 0.0005854280485251795, "router/selected_tokens_s0": 4964.125, "step": 240, "tokens_trained": 0.786353992 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.070916956244238, "grad_norm": 86.70359802246094, "loss": 2.8849, "loss_ce": 3.55375599861145, "loss_region": 0.029162542894482613, "loss_total": 3.582918643951416, "lr": 0.0006099229459530113, "router/selected_tokens_s0": 2891.75, "step": 250, "tokens_trained": 0.819119432 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.07375363449400751, "grad_norm": 106.11075592041016, "loss": 4.4058, "loss_ce": 5.333348751068115, "loss_region": 0.029971925541758537, "loss_total": 5.363320827484131, "lr": 0.0006344178433808431, "router/selected_tokens_s0": 4181.375, "step": 260, "tokens_trained": 0.851884072 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.07659031274377703, "grad_norm": 75.7653579711914, "loss": 3.6076, "loss_ce": 2.3445212841033936, "loss_region": 0.029431568458676338, "loss_total": 2.373952865600586, "lr": 0.0006589127408086749, "router/selected_tokens_s0": 3440.0, "step": 270, "tokens_trained": 0.884649512 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.07942699099354655, "grad_norm": 95.4271469116211, "loss": 2.8447, "loss_ce": 3.030097007751465, "loss_region": 0.030556708574295044, "loss_total": 3.0606536865234375, "lr": 0.0006834076382365066, "router/selected_tokens_s0": 4730.5, "step": 280, "tokens_trained": 0.917414936 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.08226366924331607, "grad_norm": 74.673828125, "loss": 2.0288, "loss_ce": 2.1509435176849365, "loss_region": 0.028712084516882896, "loss_total": 2.1796555519104004, "lr": 0.0007079025356643384, "router/selected_tokens_s0": 2658.625, "step": 290, "tokens_trained": 0.950180376 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.08510034749308559, "grad_norm": 77.71709442138672, "loss": 2.0227, "loss_ce": 2.286048650741577, "loss_region": 0.03060404770076275, "loss_total": 2.316652774810791, "lr": 0.0007323974330921702, "router/selected_tokens_s0": 4752.0, "step": 300, "tokens_trained": 0.982945816 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.08793702574285511, "grad_norm": 55.31558609008789, "loss": 2.1281, "loss_ce": 2.0437748432159424, "loss_region": 0.030610591173171997, "loss_total": 2.074385404586792, "lr": 0.000756892330520002, "router/selected_tokens_s0": 4748.625, "step": 310, "tokens_trained": 1.015711256 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.09077370399262463, "grad_norm": 77.07698059082031, "loss": 2.5761, "loss_ce": 2.7218589782714844, "loss_region": 0.03093603625893593, "loss_total": 2.7527949810028076, "lr": 0.0007813872279478337, "router/selected_tokens_s0": 4946.625, "step": 320, "tokens_trained": 1.048476696 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.09361038224239415, "grad_norm": 47.57994842529297, "loss": 2.239, "loss_ce": 1.9163914918899536, "loss_region": 0.029897142201662064, "loss_total": 1.9462885856628418, "lr": 0.0008058821253756655, "router/selected_tokens_s0": 4135.875, "step": 330, "tokens_trained": 1.081242136 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.09644706049216367, "grad_norm": 58.62579345703125, "loss": 2.8423, "loss_ce": 3.2828376293182373, "loss_region": 0.03434763103723526, "loss_total": 3.317185163497925, "lr": 0.0008303770228034974, "router/selected_tokens_s0": 6686.5, "step": 340, "tokens_trained": 1.114007576 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.09928373874193319, "grad_norm": 34.5246696472168, "loss": 2.5891, "loss_ce": 1.537825345993042, "loss_region": 0.02885586954653263, "loss_total": 1.5666812658309937, "lr": 0.0008548719202313291, "router/selected_tokens_s0": 154.125, "step": 350, "tokens_trained": 1.146773016 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.10212041699170271, "grad_norm": 37.228973388671875, "loss": 2.7756, "loss_ce": 1.9871504306793213, "loss_region": 0.029301652684807777, "loss_total": 2.0164520740509033, "lr": 0.0008793668176591608, "router/selected_tokens_s0": 3631.75, "step": 360, "tokens_trained": 1.179538456 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.10495709524147223, "grad_norm": 30.546344757080078, "loss": 2.4884, "loss_ce": 1.4886701107025146, "loss_region": 0.031588103622198105, "loss_total": 1.5202581882476807, "lr": 0.0009038617150869926, "router/selected_tokens_s0": 5236.625, "step": 370, "tokens_trained": 1.212303896 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.10779377349124175, "grad_norm": 45.68803405761719, "loss": 2.8937, "loss_ce": 2.285705804824829, "loss_region": 0.030362222343683243, "loss_total": 2.316067934036255, "lr": 0.0009283566125148244, "router/selected_tokens_s0": 4493.625, "step": 380, "tokens_trained": 1.245068536 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.11063045174101127, "grad_norm": 32.428009033203125, "loss": 1.9186, "loss_ce": 1.5672893524169922, "loss_region": 0.03746495023369789, "loss_total": 1.6047543287277222, "lr": 0.0009528515099426562, "router/selected_tokens_s0": 8134.375, "step": 390, "tokens_trained": 1.277833176 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.11346712999078079, "grad_norm": 35.54498291015625, "loss": 1.6959, "loss_ce": 1.6413251161575317, "loss_region": 0.026098042726516724, "loss_total": 1.667423129081726, "lr": 0.000977346407370488, "router/selected_tokens_s0": 625.5, "step": 400, "tokens_trained": 1.310598616 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.11630380824055031, "grad_norm": 8.186758041381836, "loss": 1.671, "loss_ce": 1.324172019958496, "loss_region": 0.03537043184041977, "loss_total": 1.3595424890518188, "lr": 0.0010018413047983197, "router/selected_tokens_s0": 7117.75, "step": 410, "tokens_trained": 1.343364056 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.11914048649031983, "grad_norm": 32.364845275878906, "loss": 1.7487, "loss_ce": 1.6946724653244019, "loss_region": 0.030674295499920845, "loss_total": 1.7253468036651611, "lr": 0.0010263362022261515, "router/selected_tokens_s0": 4591.75, "step": 420, "tokens_trained": 1.376129496 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.12197716474008935, "grad_norm": 51.924861907958984, "loss": 1.6652, "loss_ce": 1.7081111669540405, "loss_region": 0.029956262558698654, "loss_total": 1.738067388534546, "lr": 0.0010508310996539833, "router/selected_tokens_s0": 4165.25, "step": 430, "tokens_trained": 1.408889864 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.12481384298985887, "grad_norm": 31.08187484741211, "loss": 1.6269, "loss_ce": 1.688795804977417, "loss_region": 0.030442532151937485, "loss_total": 1.71923828125, "lr": 0.0010753259970818151, "router/selected_tokens_s0": 4528.875, "step": 440, "tokens_trained": 1.441655304 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.1276505212396284, "grad_norm": 9.750688552856445, "loss": 1.646, "loss_ce": 1.342025637626648, "loss_region": 0.0289932768791914, "loss_total": 1.371018886566162, "lr": 0.001099820894509647, "router/selected_tokens_s0": 3472.375, "step": 450, "tokens_trained": 1.474420744 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.1304871994893979, "grad_norm": 69.62458038330078, "loss": 2.646, "loss_ce": 2.835515022277832, "loss_region": 0.03730851039290428, "loss_total": 2.872823476791382, "lr": 0.0011243157919374788, "router/selected_tokens_s0": 7822.125, "step": 460, "tokens_trained": 1.507186184 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.13332387773916743, "grad_norm": 62.241451263427734, "loss": 2.2121, "loss_ce": 1.9173500537872314, "loss_region": 0.033008284866809845, "loss_total": 1.9503583908081055, "lr": 0.0011488106893653104, "router/selected_tokens_s0": 5854.125, "step": 470, "tokens_trained": 1.539950832 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.13616055598893695, "grad_norm": 36.45135498046875, "loss": 1.8122, "loss_ce": 1.579708456993103, "loss_region": 0.030225276947021484, "loss_total": 1.6099337339401245, "lr": 0.0011733055867931422, "router/selected_tokens_s0": 4330.5, "step": 480, "tokens_trained": 1.572715472 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.13899723423870647, "grad_norm": 13.028325080871582, "loss": 1.5027, "loss_ce": 1.357754111289978, "loss_region": 0.03526536747813225, "loss_total": 1.393019437789917, "lr": 0.001197800484220974, "router/selected_tokens_s0": 7119.25, "step": 490, "tokens_trained": 1.605480912 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.141833912488476, "grad_norm": 24.705984115600586, "loss": 1.6096, "loss_ce": 1.6097279787063599, "loss_region": 0.02911153808236122, "loss_total": 1.6388394832611084, "lr": 0.0012222953816488059, "router/selected_tokens_s0": 3648.75, "step": 500, "tokens_trained": 1.638244216 }, { "epoch": 0.141833912488476, "eval_ppl": 4.8348835473380465, "eval_runtime": 2.9238, "step": 500, "tokens_trained": 1.638244216 }, { "epoch": 0.141833912488476, "eval_F": 0.3934690889573574, "eval_F_cds": 0.29905151571508276, "eval_F_dig": 0.4478214443836758, "eval_F_exon": 0.39103450221457386, "eval_F_intron": 0.40873021991492037, "eval_F_nig": 0.4262229153142855, "eval_F_promoter": 0.30306008909923465, "eval_F_utr": 0.3906123042448191, "eval_G": 0.49025372407568035, "eval_G_cds": 0.48331595902636837, "eval_G_dig": 0.49727705981261555, "eval_G_exon": 0.4909996295084916, "eval_G_intron": 0.4915825135015993, "eval_G_nig": 0.49304083637658525, "eval_G_promoter": 0.48200754687828323, "eval_G_utr": 0.4901697268782234, "eval_avg_bp_per_token": 2.541495705926663, "eval_bp_per_token/cds": 3.343905472636816, "eval_bp_per_token/dig": 2.2330328583890666, "eval_bp_per_token/exon": 2.5573190967462667, "eval_bp_per_token/intron": 2.4466015755041455, "eval_bp_per_token/nig": 2.346190136826938, "eval_bp_per_token/promoter": 3.299675661589863, "eval_bp_per_token/utr": 2.56008320560543, "eval_ppl_cds": 5.567015659246301, "eval_ppl_dig": 4.898425899350941, "eval_ppl_exon": 4.9662705320329295, "eval_ppl_intron": 4.767518067357663, "eval_ppl_nig": 4.6987085494689405, "eval_ppl_promoter": 5.216405144788708, "eval_ppl_utr": 4.913846632347962, "step": 500, "tokens_trained": 1.638244216 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.1446705907382455, "grad_norm": 37.46432113647461, "loss": 1.5582, "loss_ce": 1.5548115968704224, "loss_region": 0.02565930411219597, "loss_total": 1.5804709196090698, "lr": 0.0012243786686061229, "router/selected_tokens_s0": 1004.25, "step": 510, "tokens_trained": 1.671005424 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.14750726898801503, "grad_norm": 27.722349166870117, "loss": 1.5672, "loss_ce": 1.478359341621399, "loss_region": 0.031882915645837784, "loss_total": 1.510242223739624, "lr": 0.0012239717766222718, "router/selected_tokens_s0": 5380.75, "step": 520, "tokens_trained": 1.703770864 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.15034394723778455, "grad_norm": 26.949983596801758, "loss": 1.6157, "loss_ce": 1.4986213445663452, "loss_region": 0.03651271015405655, "loss_total": 1.5351340770721436, "lr": 0.001223564884638421, "router/selected_tokens_s0": 7781.0, "step": 530, "tokens_trained": 1.736536304 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.15318062548755407, "grad_norm": 28.232316970825195, "loss": 1.6637, "loss_ce": 1.4607714414596558, "loss_region": 0.025137916207313538, "loss_total": 1.485909342765808, "lr": 0.00122315799265457, "router/selected_tokens_s0": 612.875, "step": 540, "tokens_trained": 1.769301744 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.1560173037373236, "grad_norm": 23.33485221862793, "loss": 1.4993, "loss_ce": 1.4412897825241089, "loss_region": 0.035474810749292374, "loss_total": 1.4767645597457886, "lr": 0.001222751100670719, "router/selected_tokens_s0": 7357.5, "step": 550, "tokens_trained": 1.802067184 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.1588539819870931, "grad_norm": 21.005512237548828, "loss": 1.4335, "loss_ce": 1.3612841367721558, "loss_region": 0.029854778200387955, "loss_total": 1.3911389112472534, "lr": 0.001222344208686868, "router/selected_tokens_s0": 4172.125, "step": 560, "tokens_trained": 1.834832624 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.16169066023686263, "grad_norm": 19.53492546081543, "loss": 1.4383, "loss_ce": 1.4045627117156982, "loss_region": 0.02937491238117218, "loss_total": 1.433937668800354, "lr": 0.0012219373167030169, "router/selected_tokens_s0": 3881.875, "step": 570, "tokens_trained": 1.867598064 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.16452733848663215, "grad_norm": 25.31780242919922, "loss": 1.7004, "loss_ce": 1.591187834739685, "loss_region": 0.03149839863181114, "loss_total": 1.6226862668991089, "lr": 0.0012215304247191658, "router/selected_tokens_s0": 5153.875, "step": 580, "tokens_trained": 1.900363504 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.16736401673640167, "grad_norm": 16.421045303344727, "loss": 1.5092, "loss_ce": 1.2439810037612915, "loss_region": 0.02931862138211727, "loss_total": 1.2732995748519897, "lr": 0.0012211235327353148, "router/selected_tokens_s0": 3840.5, "step": 590, "tokens_trained": 1.933128944 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.17020069498617119, "grad_norm": 25.38547706604004, "loss": 1.5893, "loss_ce": 1.5482516288757324, "loss_region": 0.025499241426587105, "loss_total": 1.5737508535385132, "lr": 0.0012207166407514638, "router/selected_tokens_s0": 1237.25, "step": 600, "tokens_trained": 1.96589048 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.1730373732359407, "grad_norm": 14.48205852508545, "loss": 1.3098, "loss_ce": 1.2969579696655273, "loss_region": 0.03318855166435242, "loss_total": 1.3301465511322021, "lr": 0.0012203097487676127, "router/selected_tokens_s0": 6087.625, "step": 610, "tokens_trained": 1.99865592 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.17587405148571023, "grad_norm": 10.29987907409668, "loss": 1.2844, "loss_ce": 1.2728289365768433, "loss_region": 0.03153729811310768, "loss_total": 1.3043662309646606, "lr": 0.0012199028567837617, "router/selected_tokens_s0": 5177.0, "step": 620, "tokens_trained": 2.03142136 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.17871072973547975, "grad_norm": 14.114507675170898, "loss": 1.2792, "loss_ce": 1.2729930877685547, "loss_region": 0.03177153319120407, "loss_total": 1.3047646284103394, "lr": 0.0012194959647999107, "router/selected_tokens_s0": 5318.5, "step": 630, "tokens_trained": 2.0641868 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.18154740798524927, "grad_norm": 35.086570739746094, "loss": 1.327, "loss_ce": 1.4959396123886108, "loss_region": 0.031267955899238586, "loss_total": 1.527207612991333, "lr": 0.0012190890728160596, "router/selected_tokens_s0": 5018.625, "step": 640, "tokens_trained": 2.09695224 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.18438408623501878, "grad_norm": 12.891855239868164, "loss": 1.3231, "loss_ce": 1.251932978630066, "loss_region": 0.030069500207901, "loss_total": 1.2820024490356445, "lr": 0.0012186821808322086, "router/selected_tokens_s0": 4308.125, "step": 650, "tokens_trained": 2.12971768 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.1872207644847883, "grad_norm": 16.94170570373535, "loss": 1.273, "loss_ce": 1.303807258605957, "loss_region": 0.030183279886841774, "loss_total": 1.3339905738830566, "lr": 0.0012182752888483576, "router/selected_tokens_s0": 4374.375, "step": 660, "tokens_trained": 2.16248312 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.19005744273455782, "grad_norm": 8.820389747619629, "loss": 1.291, "loss_ce": 1.2488102912902832, "loss_region": 0.030493643134832382, "loss_total": 1.2793039083480835, "lr": 0.0012178683968645065, "router/selected_tokens_s0": 4566.875, "step": 670, "tokens_trained": 2.19524856 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.19289412098432734, "grad_norm": 12.072690963745117, "loss": 1.2551, "loss_ce": 1.257431149482727, "loss_region": 0.02906171977519989, "loss_total": 1.2864928245544434, "lr": 0.0012174615048806555, "router/selected_tokens_s0": 3676.75, "step": 680, "tokens_trained": 2.228014 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.19573079923409686, "grad_norm": 3.4100522994995117, "loss": 1.2685, "loss_ce": 1.217279314994812, "loss_region": 0.03290281072258949, "loss_total": 1.2501821517944336, "lr": 0.0012170546128968045, "router/selected_tokens_s0": 5992.0, "step": 690, "tokens_trained": 2.26077944 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.19856747748386638, "grad_norm": 6.675322532653809, "loss": 1.2504, "loss_ce": 1.1835153102874756, "loss_region": 0.031250134110450745, "loss_total": 1.2147654294967651, "lr": 0.0012166477209129534, "router/selected_tokens_s0": 5040.625, "step": 700, "tokens_trained": 2.29354488 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2014041557336359, "grad_norm": 21.388051986694336, "loss": 1.267, "loss_ce": 1.3746044635772705, "loss_region": 0.027913136407732964, "loss_total": 1.402517557144165, "lr": 0.0012162408289291026, "router/selected_tokens_s0": 2922.75, "step": 710, "tokens_trained": 2.32631032 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.20424083398340542, "grad_norm": 12.917130470275879, "loss": 1.3025, "loss_ce": 1.2145620584487915, "loss_region": 0.031132886186242104, "loss_total": 1.2456949949264526, "lr": 0.0012158339369452516, "router/selected_tokens_s0": 4968.875, "step": 720, "tokens_trained": 2.35907576 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.20707751223317494, "grad_norm": 8.888051986694336, "loss": 1.2457, "loss_ce": 1.185524821281433, "loss_region": 0.03197301924228668, "loss_total": 1.2174978256225586, "lr": 0.0012154270449614005, "router/selected_tokens_s0": 5463.0, "step": 730, "tokens_trained": 2.3918396 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.20991419048294446, "grad_norm": 13.051305770874023, "loss": 1.2446, "loss_ce": 1.1078685522079468, "loss_region": 0.0308807585388422, "loss_total": 1.138749361038208, "lr": 0.0012150201529775495, "router/selected_tokens_s0": 4844.0, "step": 740, "tokens_trained": 2.424600048 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.21275086873271398, "grad_norm": 6.593105316162109, "loss": 1.2851, "loss_ce": 1.255039930343628, "loss_region": 0.029710784554481506, "loss_total": 1.2847506999969482, "lr": 0.0012146132609936982, "router/selected_tokens_s0": 4083.875, "step": 750, "tokens_trained": 2.457364688 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2155875469824835, "grad_norm": 3.900451183319092, "loss": 1.2291, "loss_ce": 1.1926592588424683, "loss_region": 0.030736476182937622, "loss_total": 1.2233957052230835, "lr": 0.0012142063690098472, "router/selected_tokens_s0": 4719.25, "step": 760, "tokens_trained": 2.490130128 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.21842422523225302, "grad_norm": 8.001019477844238, "loss": 1.2285, "loss_ce": 1.1942657232284546, "loss_region": 0.03041156381368637, "loss_total": 1.224677324295044, "lr": 0.0012137994770259962, "router/selected_tokens_s0": 4525.75, "step": 770, "tokens_trained": 2.522895568 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.22126090348202254, "grad_norm": 5.169371128082275, "loss": 1.2072, "loss_ce": 1.2079213857650757, "loss_region": 0.031087037175893784, "loss_total": 1.2390084266662598, "lr": 0.0012133925850421454, "router/selected_tokens_s0": 4938.25, "step": 780, "tokens_trained": 2.555659392 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.22409758173179206, "grad_norm": 8.434707641601562, "loss": 1.2079, "loss_ce": 1.2038490772247314, "loss_region": 0.02821769006550312, "loss_total": 1.2320667505264282, "lr": 0.0012129856930582943, "router/selected_tokens_s0": 3119.875, "step": 790, "tokens_trained": 2.588422136 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.22693425998156158, "grad_norm": 8.451072692871094, "loss": 1.2072, "loss_ce": 1.2617510557174683, "loss_region": 0.0316130593419075, "loss_total": 1.29336416721344, "lr": 0.0012125788010744433, "router/selected_tokens_s0": 5238.75, "step": 800, "tokens_trained": 2.621187576 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2297709382313311, "grad_norm": 12.750673294067383, "loss": 1.2283, "loss_ce": 1.2528263330459595, "loss_region": 0.03109751269221306, "loss_total": 1.283923864364624, "lr": 0.0012121719090905923, "router/selected_tokens_s0": 4940.75, "step": 810, "tokens_trained": 2.653953016 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.23260761648110062, "grad_norm": 10.307655334472656, "loss": 1.2544, "loss_ce": 1.2496147155761719, "loss_region": 0.02913491614162922, "loss_total": 1.2787495851516724, "lr": 0.0012117650171067412, "router/selected_tokens_s0": 3717.75, "step": 820, "tokens_trained": 2.686718456 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.23544429473087014, "grad_norm": 0.6592714190483093, "loss": 1.2022, "loss_ce": 1.0889158248901367, "loss_region": 0.031184613704681396, "loss_total": 1.120100498199463, "lr": 0.0012113581251228902, "router/selected_tokens_s0": 5037.375, "step": 830, "tokens_trained": 2.71948036 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.23828097298063966, "grad_norm": 3.0865817070007324, "loss": 1.1958, "loss_ce": 1.267112374305725, "loss_region": 0.02916303649544716, "loss_total": 1.2962753772735596, "lr": 0.0012109512331390391, "router/selected_tokens_s0": 3734.375, "step": 840, "tokens_trained": 2.7522458 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.24111765123040918, "grad_norm": 2.513849973678589, "loss": 1.2014, "loss_ce": 1.108485221862793, "loss_region": 0.0302209984511137, "loss_total": 1.1387062072753906, "lr": 0.0012105443411551881, "router/selected_tokens_s0": 4417.125, "step": 850, "tokens_trained": 2.78501124 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2439543294801787, "grad_norm": 5.594594478607178, "loss": 1.206, "loss_ce": 1.1815146207809448, "loss_region": 0.031508028507232666, "loss_total": 1.2130227088928223, "lr": 0.001210137449171337, "router/selected_tokens_s0": 5212.875, "step": 860, "tokens_trained": 2.81777668 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.24679100772994822, "grad_norm": 2.2655980587005615, "loss": 1.1897, "loss_ce": 1.2304372787475586, "loss_region": 0.031548820436000824, "loss_total": 1.2619861364364624, "lr": 0.001209730557187486, "router/selected_tokens_s0": 5213.25, "step": 870, "tokens_trained": 2.85054212 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.24962768597971774, "grad_norm": 4.335860252380371, "loss": 1.1897, "loss_ce": 1.2337130308151245, "loss_region": 0.02997858263552189, "loss_total": 1.2636916637420654, "lr": 0.001209323665203635, "router/selected_tokens_s0": 4252.5, "step": 880, "tokens_trained": 2.88330756 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.25246436422948726, "grad_norm": 12.377155303955078, "loss": 1.1966, "loss_ce": 1.1369762420654297, "loss_region": 0.029613491147756577, "loss_total": 1.1665897369384766, "lr": 0.001208916773219784, "router/selected_tokens_s0": 4027.75, "step": 890, "tokens_trained": 2.916073 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2553010424792568, "grad_norm": 7.238094806671143, "loss": 1.2143, "loss_ce": 1.1700671911239624, "loss_region": 0.029774101451039314, "loss_total": 1.1998412609100342, "lr": 0.001208509881235933, "router/selected_tokens_s0": 4116.875, "step": 900, "tokens_trained": 2.94883828 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2581377207290263, "grad_norm": 3.2694191932678223, "loss": 1.1892, "loss_ce": 1.1454379558563232, "loss_region": 0.029824109748005867, "loss_total": 1.1752620935440063, "lr": 0.001208102989252082, "router/selected_tokens_s0": 4152.625, "step": 910, "tokens_trained": 2.981597288 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2609743989787958, "grad_norm": 9.457625389099121, "loss": 1.2038, "loss_ce": 1.3160332441329956, "loss_region": 0.030873605981469154, "loss_total": 1.3469069004058838, "lr": 0.0012076960972682309, "router/selected_tokens_s0": 4797.5, "step": 920, "tokens_trained": 3.014362456 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.26381107722856534, "grad_norm": 4.293655872344971, "loss": 1.1978, "loss_ce": 1.1440948247909546, "loss_region": 0.02935035713016987, "loss_total": 1.173445224761963, "lr": 0.0012072892052843798, "router/selected_tokens_s0": 3829.5, "step": 930, "tokens_trained": 3.047127096 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.26664775547833486, "grad_norm": 1.7136532068252563, "loss": 1.1906, "loss_ce": 1.1432236433029175, "loss_region": 0.028851088136434555, "loss_total": 1.1720746755599976, "lr": 0.0012068823133005288, "router/selected_tokens_s0": 3479.125, "step": 940, "tokens_trained": 3.079892536 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2694844337281044, "grad_norm": 4.0433244705200195, "loss": 1.1868, "loss_ce": 1.168936014175415, "loss_region": 0.02876598760485649, "loss_total": 1.1977020502090454, "lr": 0.0012064754213166778, "router/selected_tokens_s0": 3396.25, "step": 950, "tokens_trained": 3.11265336 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2723211119778739, "grad_norm": 6.829047203063965, "loss": 1.1828, "loss_ce": 1.2480430603027344, "loss_region": 0.02934931591153145, "loss_total": 1.2773923873901367, "lr": 0.001206068529332827, "router/selected_tokens_s0": 3843.75, "step": 960, "tokens_trained": 3.1454188 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2751577902276434, "grad_norm": 5.5668439865112305, "loss": 1.1882, "loss_ce": 1.1349202394485474, "loss_region": 0.0297370757907629, "loss_total": 1.1646573543548584, "lr": 0.001205661637348976, "router/selected_tokens_s0": 4102.5, "step": 970, "tokens_trained": 3.17818424 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.27799446847741294, "grad_norm": 3.729381561279297, "loss": 1.1839, "loss_ce": 1.1995916366577148, "loss_region": 0.03041483648121357, "loss_total": 1.230006456375122, "lr": 0.0012052547453651249, "router/selected_tokens_s0": 4537.125, "step": 980, "tokens_trained": 3.21094968 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.28083114672718246, "grad_norm": 2.7978885173797607, "loss": 1.1739, "loss_ce": 1.1886447668075562, "loss_region": 0.030223874375224113, "loss_total": 1.218868613243103, "lr": 0.0012048478533812738, "router/selected_tokens_s0": 4418.875, "step": 990, "tokens_trained": 3.24371512 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.283667824976952, "grad_norm": 2.7768421173095703, "loss": 1.1695, "loss_ce": 1.1791244745254517, "loss_region": 0.03016069531440735, "loss_total": 1.2092851400375366, "lr": 0.0012044409613974226, "router/selected_tokens_s0": 4373.0, "step": 1000, "tokens_trained": 3.27648056 }, { "epoch": 0.283667824976952, "eval_ppl": 3.1457362037176693, "eval_runtime": 2.5704, "step": 1000, "tokens_trained": 3.27648056 }, { "epoch": 0.283667824976952, "eval_F": 0.35905403615092213, "eval_F_cds": 0.3614752043728926, "eval_F_dig": 0.36203349219991143, "eval_F_exon": 0.3609332242502892, "eval_F_intron": 0.3608845011093654, "eval_F_nig": 0.36360427639485304, "eval_F_promoter": 0.3446594753609168, "eval_F_utr": 0.35993294503032014, "eval_G": 0.4747950002316863, "eval_G_cds": 0.4875693056072159, "eval_G_dig": 0.4165539971384483, "eval_G_exon": 0.4825983323253731, "eval_G_intron": 0.4746974505122046, "eval_G_nig": 0.4719204972271849, "eval_G_promoter": 0.47860970096474814, "eval_G_utr": 0.4806883865646302, "eval_avg_bp_per_token": 2.785096111772066, "eval_bp_per_token/cds": 2.7664414817466, "eval_bp_per_token/dig": 2.7621753830659665, "eval_bp_per_token/exon": 2.770595591683602, "eval_bp_per_token/intron": 2.7709696507497057, "eval_bp_per_token/nig": 2.7502426811780905, "eval_bp_per_token/promoter": 2.90141450181467, "eval_bp_per_token/utr": 2.77829527362593, "eval_ppl_cds": 3.7937951600140427, "eval_ppl_dig": 1.292568207392483, "eval_ppl_exon": 3.5063285971819904, "eval_ppl_intron": 3.1623742022954864, "eval_ppl_nig": 3.03123217862896, "eval_ppl_promoter": 3.420873133996253, "eval_ppl_utr": 3.4079030610184535, "step": 1000, "tokens_trained": 3.27648056 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2865045032267215, "grad_norm": 1.750190258026123, "loss": 1.1681, "loss_ce": 1.1951100826263428, "loss_region": 0.029561972245573997, "loss_total": 1.2246720790863037, "lr": 0.0012040340694135716, "router/selected_tokens_s0": 3974.5, "step": 1010, "tokens_trained": 3.309246 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.289341181476491, "grad_norm": 5.037286758422852, "loss": 1.1855, "loss_ce": 1.1606330871582031, "loss_region": 0.030172061175107956, "loss_total": 1.190805196762085, "lr": 0.0012036271774297205, "router/selected_tokens_s0": 4388.375, "step": 1020, "tokens_trained": 3.34201144 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.29217785972626054, "grad_norm": 5.963747024536133, "loss": 1.1794, "loss_ce": 1.116599678993225, "loss_region": 0.030543407425284386, "loss_total": 1.1471431255340576, "lr": 0.0012032202854458697, "router/selected_tokens_s0": 4640.0, "step": 1030, "tokens_trained": 3.37477688 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.29501453797603006, "grad_norm": 4.626336574554443, "loss": 1.1934, "loss_ce": 1.094927430152893, "loss_region": 0.02999301068484783, "loss_total": 1.1249204874038696, "lr": 0.0012028133934620187, "router/selected_tokens_s0": 4248.5, "step": 1040, "tokens_trained": 3.40754232 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.2978512162257996, "grad_norm": 4.208251476287842, "loss": 1.1843, "loss_ce": 1.1818771362304688, "loss_region": 0.030715491622686386, "loss_total": 1.212592601776123, "lr": 0.0012024065014781676, "router/selected_tokens_s0": 4729.75, "step": 1050, "tokens_trained": 3.44030696 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3006878944755691, "grad_norm": 2.3673582077026367, "loss": 1.1726, "loss_ce": 1.1216882467269897, "loss_region": 0.030366381630301476, "loss_total": 1.1520546674728394, "lr": 0.0012019996094943166, "router/selected_tokens_s0": 4503.625, "step": 1060, "tokens_trained": 3.4730724 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3035245727253386, "grad_norm": 2.6513352394104004, "loss": 1.1707, "loss_ce": 1.1285063028335571, "loss_region": 0.02974226139485836, "loss_total": 1.1582485437393188, "lr": 0.0012015927175104656, "router/selected_tokens_s0": 4085.375, "step": 1070, "tokens_trained": 3.50583784 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.30636125097510813, "grad_norm": 1.0276976823806763, "loss": 1.165, "loss_ce": 1.1330546140670776, "loss_region": 0.029834387823939323, "loss_total": 1.162889003753662, "lr": 0.0012011858255266145, "router/selected_tokens_s0": 4155.5, "step": 1080, "tokens_trained": 3.53860328 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.30919792922487765, "grad_norm": 3.4352457523345947, "loss": 1.1759, "loss_ce": 1.153834581375122, "loss_region": 0.030001970008015633, "loss_total": 1.183836579322815, "lr": 0.0012007789335427635, "router/selected_tokens_s0": 4271.375, "step": 1090, "tokens_trained": 3.57136872 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3120346074746472, "grad_norm": 3.4334914684295654, "loss": 1.1668, "loss_ce": 1.0656555891036987, "loss_region": 0.03014238551259041, "loss_total": 1.0957980155944824, "lr": 0.0012003720415589125, "router/selected_tokens_s0": 4376.625, "step": 1100, "tokens_trained": 3.60413416 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3148712857244167, "grad_norm": 7.573620796203613, "loss": 1.1737, "loss_ce": 1.1206940412521362, "loss_region": 0.030071774497628212, "loss_total": 1.1507657766342163, "lr": 0.0011999651495750614, "router/selected_tokens_s0": 4325.0, "step": 1110, "tokens_trained": 3.6368996 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3177079639741862, "grad_norm": 4.200015544891357, "loss": 1.1705, "loss_ce": 1.1700469255447388, "loss_region": 0.02990192547440529, "loss_total": 1.1999489068984985, "lr": 0.0011995582575912104, "router/selected_tokens_s0": 4194.25, "step": 1120, "tokens_trained": 3.669661712 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.32054464222395573, "grad_norm": 5.207011699676514, "loss": 1.1668, "loss_ce": 1.1708717346191406, "loss_region": 0.029880443587899208, "loss_total": 1.2007521390914917, "lr": 0.0011991513656073594, "router/selected_tokens_s0": 4177.25, "step": 1130, "tokens_trained": 3.702426352 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.32338132047372525, "grad_norm": 4.160227298736572, "loss": 1.1671, "loss_ce": 1.1502091884613037, "loss_region": 0.030087152495980263, "loss_total": 1.1802963018417358, "lr": 0.0011987444736235083, "router/selected_tokens_s0": 4325.25, "step": 1140, "tokens_trained": 3.735191792 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3262179987234948, "grad_norm": 2.3496572971343994, "loss": 1.1578, "loss_ce": 1.0942906141281128, "loss_region": 0.02960728108882904, "loss_total": 1.123897910118103, "lr": 0.0011983375816396573, "router/selected_tokens_s0": 3976.25, "step": 1150, "tokens_trained": 3.767957232 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3290546769732643, "grad_norm": 3.0820891857147217, "loss": 1.158, "loss_ce": 1.2191810607910156, "loss_region": 0.030029961839318275, "loss_total": 1.249211072921753, "lr": 0.0011979306896558062, "router/selected_tokens_s0": 4285.125, "step": 1160, "tokens_trained": 3.800722672 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3318913552230338, "grad_norm": 1.7340823411941528, "loss": 1.1537, "loss_ce": 1.0748310089111328, "loss_region": 0.030402792617678642, "loss_total": 1.1052337884902954, "lr": 0.0011975237976719552, "router/selected_tokens_s0": 4566.375, "step": 1170, "tokens_trained": 3.833488112 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.33472803347280333, "grad_norm": 1.6883597373962402, "loss": 1.1524, "loss_ce": 1.15337073802948, "loss_region": 0.029628688469529152, "loss_total": 1.1829993724822998, "lr": 0.0011971169056881042, "router/selected_tokens_s0": 3994.125, "step": 1180, "tokens_trained": 3.866252752 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.33756471172257285, "grad_norm": 1.3079456090927124, "loss": 1.155, "loss_ce": 1.147839903831482, "loss_region": 0.029972558841109276, "loss_total": 1.1778124570846558, "lr": 0.0011967100137042531, "router/selected_tokens_s0": 4250.125, "step": 1190, "tokens_trained": 3.899018184 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.34040138997234237, "grad_norm": 2.042187452316284, "loss": 1.1551, "loss_ce": 1.1045622825622559, "loss_region": 0.030126892030239105, "loss_total": 1.134689211845398, "lr": 0.0011963031217204021, "router/selected_tokens_s0": 4366.25, "step": 1200, "tokens_trained": 3.931783624 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3432380682221119, "grad_norm": 0.5720299482345581, "loss": 1.1514, "loss_ce": 1.1252881288528442, "loss_region": 0.02972925268113613, "loss_total": 1.155017375946045, "lr": 0.0011958962297365513, "router/selected_tokens_s0": 4055.0, "step": 1210, "tokens_trained": 3.964549064 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3460747464718814, "grad_norm": 2.726912498474121, "loss": 1.1481, "loss_ce": 1.0980409383773804, "loss_region": 0.030369114130735397, "loss_total": 1.1284101009368896, "lr": 0.0011954893377527003, "router/selected_tokens_s0": 4549.75, "step": 1220, "tokens_trained": 3.997311912 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.34891142472165093, "grad_norm": 1.576530933380127, "loss": 1.1547, "loss_ce": 1.1488255262374878, "loss_region": 0.03008064441382885, "loss_total": 1.1789062023162842, "lr": 0.0011950824457688492, "router/selected_tokens_s0": 4327.125, "step": 1230, "tokens_trained": 4.030077352 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.35174810297142045, "grad_norm": 1.7633917331695557, "loss": 1.1491, "loss_ce": 1.0437774658203125, "loss_region": 0.03009728156030178, "loss_total": 1.0738747119903564, "lr": 0.0011946755537849982, "router/selected_tokens_s0": 4352.5, "step": 1240, "tokens_trained": 4.062842792 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.35458478122118997, "grad_norm": 0.8599131107330322, "loss": 1.1502, "loss_ce": 1.1635342836380005, "loss_region": 0.030227093026041985, "loss_total": 1.1937613487243652, "lr": 0.001194268661801147, "router/selected_tokens_s0": 4437.875, "step": 1250, "tokens_trained": 4.095608232 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3574214594709595, "grad_norm": 2.0207033157348633, "loss": 1.1525, "loss_ce": 1.161281943321228, "loss_region": 0.02980414777994156, "loss_total": 1.1910860538482666, "lr": 0.001193861769817296, "router/selected_tokens_s0": 4113.375, "step": 1260, "tokens_trained": 4.128373672 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.360258137720729, "grad_norm": 1.6762081384658813, "loss": 1.1549, "loss_ce": 1.176638126373291, "loss_region": 0.02979988045990467, "loss_total": 1.2064380645751953, "lr": 0.0011934548778334449, "router/selected_tokens_s0": 4110.375, "step": 1270, "tokens_trained": 4.161136768 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.36309481597049853, "grad_norm": 1.5674160718917847, "loss": 1.1538, "loss_ce": 1.1160061359405518, "loss_region": 0.029819507151842117, "loss_total": 1.1458256244659424, "lr": 0.001193047985849594, "router/selected_tokens_s0": 4122.75, "step": 1280, "tokens_trained": 4.193902208 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.36593149422026805, "grad_norm": 1.232892394065857, "loss": 1.1499, "loss_ce": 1.192215085029602, "loss_region": 0.030095556750893593, "loss_total": 1.2223106622695923, "lr": 0.001192641093865743, "router/selected_tokens_s0": 4337.75, "step": 1290, "tokens_trained": 4.226667648 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.36876817247003757, "grad_norm": 1.280081033706665, "loss": 1.1625, "loss_ce": 1.0769988298416138, "loss_region": 0.030076846480369568, "loss_total": 1.1070756912231445, "lr": 0.001192234201881892, "router/selected_tokens_s0": 4330.625, "step": 1300, "tokens_trained": 4.259424272 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3716048507198071, "grad_norm": 0.7819789052009583, "loss": 1.1516, "loss_ce": 1.0531295537948608, "loss_region": 0.029812535271048546, "loss_total": 1.0829421281814575, "lr": 0.001191827309898041, "router/selected_tokens_s0": 4107.75, "step": 1310, "tokens_trained": 4.292189712 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3744415289695766, "grad_norm": 4.3887505531311035, "loss": 1.1524, "loss_ce": 1.0992565155029297, "loss_region": 0.030015140771865845, "loss_total": 1.1292716264724731, "lr": 0.00119142041791419, "router/selected_tokens_s0": 4279.625, "step": 1320, "tokens_trained": 4.32495164 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.37727820721934613, "grad_norm": 2.5429630279541016, "loss": 1.1622, "loss_ce": 0.9915607571601868, "loss_region": 0.02960185892879963, "loss_total": 1.0211626291275024, "lr": 0.0011910135259303389, "router/selected_tokens_s0": 3922.75, "step": 1330, "tokens_trained": 4.35771708 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.38011488546911565, "grad_norm": 1.3790112733840942, "loss": 1.1526, "loss_ce": 1.2076722383499146, "loss_region": 0.029480615630745888, "loss_total": 1.2371528148651123, "lr": 0.0011906066339464878, "router/selected_tokens_s0": 3831.5, "step": 1340, "tokens_trained": 4.39048252 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.38295156371888517, "grad_norm": 3.28352427482605, "loss": 1.1523, "loss_ce": 0.9999480247497559, "loss_region": 0.02995798923075199, "loss_total": 1.0299060344696045, "lr": 0.0011901997419626368, "router/selected_tokens_s0": 4236.0, "step": 1350, "tokens_trained": 4.42324796 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3857882419686547, "grad_norm": 2.173388719558716, "loss": 1.1469, "loss_ce": 1.1173208951950073, "loss_region": 0.030063528567552567, "loss_total": 1.1473844051361084, "lr": 0.0011897928499787858, "router/selected_tokens_s0": 4322.25, "step": 1360, "tokens_trained": 4.4560134 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3886249202184242, "grad_norm": 1.3337340354919434, "loss": 1.1514, "loss_ce": 1.097347617149353, "loss_region": 0.030277268961071968, "loss_total": 1.1276248693466187, "lr": 0.0011893859579949347, "router/selected_tokens_s0": 4490.375, "step": 1370, "tokens_trained": 4.48877884 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3914615984681937, "grad_norm": 1.5072178840637207, "loss": 1.1454, "loss_ce": 1.1354695558547974, "loss_region": 0.0300710741430521, "loss_total": 1.1655405759811401, "lr": 0.0011889790660110837, "router/selected_tokens_s0": 4323.125, "step": 1380, "tokens_trained": 4.52154428 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.39429827671796325, "grad_norm": 1.4634846448898315, "loss": 1.1434, "loss_ce": 1.1472464799880981, "loss_region": 0.029943954199552536, "loss_total": 1.1771904230117798, "lr": 0.0011885721740272327, "router/selected_tokens_s0": 4222.625, "step": 1390, "tokens_trained": 4.55430972 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.39713495496773277, "grad_norm": 1.1301681995391846, "loss": 1.1491, "loss_ce": 0.932141900062561, "loss_region": 0.03013395331799984, "loss_total": 0.9622758626937866, "lr": 0.0011881652820433816, "router/selected_tokens_s0": 4389.125, "step": 1400, "tokens_trained": 4.58707516 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.3999716332175023, "grad_norm": 1.153057336807251, "loss": 1.1483, "loss_ce": 1.0930418968200684, "loss_region": 0.029886895790696144, "loss_total": 1.1229287385940552, "lr": 0.0011877583900595306, "router/selected_tokens_s0": 4177.875, "step": 1410, "tokens_trained": 4.6198406 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4028083114672718, "grad_norm": 2.0346107482910156, "loss": 1.1355, "loss_ce": 1.130191683769226, "loss_region": 0.030217666178941727, "loss_total": 1.1604093313217163, "lr": 0.0011873514980756796, "router/selected_tokens_s0": 4435.0, "step": 1420, "tokens_trained": 4.652606024 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4056449897170413, "grad_norm": 1.2362136840820312, "loss": 1.1461, "loss_ce": 1.1180355548858643, "loss_region": 0.029944026842713356, "loss_total": 1.1479796171188354, "lr": 0.0011869446060918285, "router/selected_tokens_s0": 4219.25, "step": 1430, "tokens_trained": 4.685371464 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.40848166796681085, "grad_norm": 1.6414567232131958, "loss": 1.1476, "loss_ce": 1.1310675144195557, "loss_region": 0.030178584158420563, "loss_total": 1.1612460613250732, "lr": 0.0011865377141079775, "router/selected_tokens_s0": 4406.125, "step": 1440, "tokens_trained": 4.718136904 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.41131834621658037, "grad_norm": 0.8733806014060974, "loss": 1.1452, "loss_ce": 1.1529111862182617, "loss_region": 0.029908571392297745, "loss_total": 1.1828197240829468, "lr": 0.0011861308221241265, "router/selected_tokens_s0": 4186.5, "step": 1450, "tokens_trained": 4.750902344 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4141550244663499, "grad_norm": 2.170149087905884, "loss": 1.1364, "loss_ce": 1.1446956396102905, "loss_region": 0.030016543343663216, "loss_total": 1.1747121810913086, "lr": 0.0011857239301402756, "router/selected_tokens_s0": 4279.125, "step": 1460, "tokens_trained": 4.783666984 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4169917027161194, "grad_norm": 1.5901942253112793, "loss": 1.1418, "loss_ce": 1.1736469268798828, "loss_region": 0.02991572767496109, "loss_total": 1.203562617301941, "lr": 0.0011853170381564246, "router/selected_tokens_s0": 4190.375, "step": 1470, "tokens_trained": 4.816432424 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4198283809658889, "grad_norm": 0.7801039814949036, "loss": 1.1359, "loss_ce": 1.0415936708450317, "loss_region": 0.030063536018133163, "loss_total": 1.0716571807861328, "lr": 0.0011849101461725736, "router/selected_tokens_s0": 4323.5, "step": 1480, "tokens_trained": 4.849197864 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.42266505921565845, "grad_norm": 1.1225630044937134, "loss": 1.1387, "loss_ce": 1.1764026880264282, "loss_region": 0.02989169955253601, "loss_total": 1.2062944173812866, "lr": 0.0011845032541887225, "router/selected_tokens_s0": 4166.375, "step": 1490, "tokens_trained": 4.881963248 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.42550173746542796, "grad_norm": 1.3516196012496948, "loss": 1.1445, "loss_ce": 1.1119225025177002, "loss_region": 0.03007156029343605, "loss_total": 1.1419941186904907, "lr": 0.0011840963622048713, "router/selected_tokens_s0": 4332.625, "step": 1500, "tokens_trained": 4.914728608 }, { "epoch": 0.42550173746542796, "eval_ppl": 3.0476700462359805, "eval_runtime": 2.5167, "step": 1500, "tokens_trained": 4.914728608 }, { "epoch": 0.42550173746542796, "eval_F": 0.3395766737890528, "eval_F_cds": 0.33560010026602843, "eval_F_dig": 0.34591244107612573, "eval_F_exon": 0.33732050667193275, "eval_F_intron": 0.340589821591843, "eval_F_nig": 0.3449097161371641, "eval_F_promoter": 0.3287406377406758, "eval_F_utr": 0.33810586816514, "eval_G": 0.4388793285567115, "eval_G_cds": 0.4465895620992391, "eval_G_dig": 0.39567722372516084, "eval_G_exon": 0.44327135296181625, "eval_G_intron": 0.4386635275964277, "eval_G_nig": 0.4373593879668909, "eval_G_promoter": 0.44171817290159177, "eval_G_utr": 0.44355779628952524, "eval_avg_bp_per_token": 2.944843027178028, "eval_bp_per_token/cds": 2.9797368928296066, "eval_bp_per_token/dig": 2.8909049841891283, "eval_bp_per_token/exon": 2.9645396002341724, "eval_bp_per_token/intron": 2.93608304360423, "eval_bp_per_token/nig": 2.8993094517590188, "eval_bp_per_token/promoter": 3.0419117237000717, "eval_bp_per_token/utr": 2.9576534871366778, "eval_ppl_cds": 3.7328596405663, "eval_ppl_dig": 1.1534605141350962, "eval_ppl_exon": 3.4439528933373436, "eval_ppl_intron": 3.0653985302604827, "eval_ppl_nig": 2.904936687189015, "eval_ppl_promoter": 3.3618258190318606, "eval_ppl_utr": 3.3512748939063846, "step": 1500, "tokens_trained": 4.914728608 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4283384157151975, "grad_norm": 1.0354516506195068, "loss": 1.1407, "loss_ce": 1.2179700136184692, "loss_region": 0.029973506927490234, "loss_total": 1.2479435205459595, "lr": 0.0011836894702210202, "router/selected_tokens_s0": 4242.375, "step": 1510, "tokens_trained": 4.947494048 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.431175093964967, "grad_norm": 0.9974690675735474, "loss": 1.1361, "loss_ce": 1.1464780569076538, "loss_region": 0.03020160086452961, "loss_total": 1.1766796112060547, "lr": 0.0011832825782371692, "router/selected_tokens_s0": 4443.875, "step": 1520, "tokens_trained": 4.980259488 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4340117722147365, "grad_norm": 1.61404550075531, "loss": 1.1383, "loss_ce": 1.1023921966552734, "loss_region": 0.029910210520029068, "loss_total": 1.1323024034500122, "lr": 0.0011828756862533184, "router/selected_tokens_s0": 4174.25, "step": 1530, "tokens_trained": 5.013024928 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.43684845046450604, "grad_norm": 1.551711082458496, "loss": 1.1369, "loss_ce": 1.085469365119934, "loss_region": 0.02990012802183628, "loss_total": 1.115369439125061, "lr": 0.0011824687942694674, "router/selected_tokens_s0": 4162.25, "step": 1540, "tokens_trained": 5.04578704 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.43968512871427556, "grad_norm": 1.3328109979629517, "loss": 1.1358, "loss_ce": 1.1522539854049683, "loss_region": 0.02980169840157032, "loss_total": 1.1820557117462158, "lr": 0.0011820619022856163, "router/selected_tokens_s0": 4050.75, "step": 1550, "tokens_trained": 5.078551904 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4425218069640451, "grad_norm": 2.2517945766448975, "loss": 1.1398, "loss_ce": 1.0304194688796997, "loss_region": 0.030139248818159103, "loss_total": 1.0605586767196655, "lr": 0.0011816550103017653, "router/selected_tokens_s0": 4399.625, "step": 1560, "tokens_trained": 5.111317344 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4453584852138146, "grad_norm": 1.0419440269470215, "loss": 1.1423, "loss_ce": 1.2029235363006592, "loss_region": 0.029878782108426094, "loss_total": 1.2328022718429565, "lr": 0.0011812481183179143, "router/selected_tokens_s0": 4131.75, "step": 1570, "tokens_trained": 5.144082784 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4481951634635841, "grad_norm": 0.8405026197433472, "loss": 1.1357, "loss_ce": 1.1085268259048462, "loss_region": 0.02992934361100197, "loss_total": 1.1384562253952026, "lr": 0.0011808412263340632, "router/selected_tokens_s0": 4185.75, "step": 1580, "tokens_trained": 5.176848224 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.45103184171335364, "grad_norm": 1.8782676458358765, "loss": 1.1447, "loss_ce": 1.0933234691619873, "loss_region": 0.030135583132505417, "loss_total": 1.1234591007232666, "lr": 0.0011804343343502122, "router/selected_tokens_s0": 4400.5, "step": 1590, "tokens_trained": 5.209613664 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.45386851996312316, "grad_norm": 1.116540551185608, "loss": 1.1417, "loss_ce": 1.1890523433685303, "loss_region": 0.0303688682615757, "loss_total": 1.2194212675094604, "lr": 0.0011800274423663611, "router/selected_tokens_s0": 4597.375, "step": 1600, "tokens_trained": 5.242378304 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4567051982128927, "grad_norm": 0.9224187135696411, "loss": 1.1352, "loss_ce": 1.0753121376037598, "loss_region": 0.030113881453871727, "loss_total": 1.1054260730743408, "lr": 0.0011796205503825101, "router/selected_tokens_s0": 4381.5, "step": 1610, "tokens_trained": 5.275142944 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4595418764626622, "grad_norm": 1.250409483909607, "loss": 1.1423, "loss_ce": 1.1405887603759766, "loss_region": 0.030090278014540672, "loss_total": 1.1706790924072266, "lr": 0.001179213658398659, "router/selected_tokens_s0": 4360.875, "step": 1620, "tokens_trained": 5.307906784 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4623785547124317, "grad_norm": 0.6683188080787659, "loss": 1.1358, "loss_ce": 1.0137219429016113, "loss_region": 0.0301409512758255, "loss_total": 1.0438629388809204, "lr": 0.001178806766414808, "router/selected_tokens_s0": 4420.25, "step": 1630, "tokens_trained": 5.340672224 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.46521523296220124, "grad_norm": 1.3055206537246704, "loss": 1.1378, "loss_ce": 1.120367407798767, "loss_region": 0.029992438852787018, "loss_total": 1.150359869003296, "lr": 0.001178399874430957, "router/selected_tokens_s0": 4256.375, "step": 1640, "tokens_trained": 5.373436896 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.46805191121197076, "grad_norm": 1.2817225456237793, "loss": 1.1365, "loss_ce": 1.159173607826233, "loss_region": 0.030014952644705772, "loss_total": 1.1891885995864868, "lr": 0.001177992982447106, "router/selected_tokens_s0": 4277.875, "step": 1650, "tokens_trained": 5.406202336 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4708885894617403, "grad_norm": 1.2652041912078857, "loss": 1.1303, "loss_ce": 1.1445159912109375, "loss_region": 0.03000623546540737, "loss_total": 1.1745222806930542, "lr": 0.001177586090463255, "router/selected_tokens_s0": 4274.375, "step": 1660, "tokens_trained": 5.438967776 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4737252677115098, "grad_norm": 1.7784186601638794, "loss": 1.1334, "loss_ce": 1.1069244146347046, "loss_region": 0.030016450211405754, "loss_total": 1.136940836906433, "lr": 0.001177179198479404, "router/selected_tokens_s0": 4287.625, "step": 1670, "tokens_trained": 5.471733216 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4765619459612793, "grad_norm": 1.0779353380203247, "loss": 1.1315, "loss_ce": 1.1237202882766724, "loss_region": 0.029916411265730858, "loss_total": 1.1536366939544678, "lr": 0.0011767723064955529, "router/selected_tokens_s0": 4156.75, "step": 1680, "tokens_trained": 5.504498656 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.47939862421104884, "grad_norm": 0.7689351439476013, "loss": 1.1324, "loss_ce": 1.0980726480484009, "loss_region": 0.030096061527729034, "loss_total": 1.1281687021255493, "lr": 0.0011763654145117018, "router/selected_tokens_s0": 4377.5, "step": 1690, "tokens_trained": 5.537264096 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.48223530246081836, "grad_norm": 0.6869276165962219, "loss": 1.1332, "loss_ce": 1.0792652368545532, "loss_region": 0.030072998255491257, "loss_total": 1.1093382835388184, "lr": 0.0011759585225278508, "router/selected_tokens_s0": 4349.625, "step": 1700, "tokens_trained": 5.570029536 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4850719807105879, "grad_norm": 0.9587815403938293, "loss": 1.1361, "loss_ce": 1.0378434658050537, "loss_region": 0.03009817562997341, "loss_total": 1.067941665649414, "lr": 0.001175551630544, "router/selected_tokens_s0": 4384.25, "step": 1710, "tokens_trained": 5.602794976 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4879086589603574, "grad_norm": 1.1542259454727173, "loss": 1.1294, "loss_ce": 1.074008584022522, "loss_region": 0.030034121125936508, "loss_total": 1.104042649269104, "lr": 0.001175144738560149, "router/selected_tokens_s0": 4306.0, "step": 1720, "tokens_trained": 5.635560416 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4907453372101269, "grad_norm": 1.0194206237792969, "loss": 1.1296, "loss_ce": 1.1548231840133667, "loss_region": 0.03011094592511654, "loss_total": 1.184934139251709, "lr": 0.001174737846576298, "router/selected_tokens_s0": 4395.625, "step": 1730, "tokens_trained": 5.668325856 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.49358201545989644, "grad_norm": 1.108144998550415, "loss": 1.1351, "loss_ce": 1.0953419208526611, "loss_region": 0.03002314455807209, "loss_total": 1.1253650188446045, "lr": 0.0011743309545924469, "router/selected_tokens_s0": 4292.125, "step": 1740, "tokens_trained": 5.701091296 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.49641869370966596, "grad_norm": 1.075562834739685, "loss": 1.1347, "loss_ce": 1.1154391765594482, "loss_region": 0.029949212446808815, "loss_total": 1.1453883647918701, "lr": 0.0011739240626085956, "router/selected_tokens_s0": 4188.625, "step": 1750, "tokens_trained": 5.733856736 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.4992553719594355, "grad_norm": 1.3173739910125732, "loss": 1.1325, "loss_ce": 1.0855435132980347, "loss_region": 0.02994917891919613, "loss_total": 1.1154927015304565, "lr": 0.0011735171706247446, "router/selected_tokens_s0": 4183.625, "step": 1760, "tokens_trained": 5.766622176 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.502092050209205, "grad_norm": 0.8734815716743469, "loss": 1.1316, "loss_ce": 1.190360188484192, "loss_region": 0.03002040646970272, "loss_total": 1.2203805446624756, "lr": 0.0011731102786408936, "router/selected_tokens_s0": 4294.5, "step": 1770, "tokens_trained": 5.799387616 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5049287284589745, "grad_norm": 2.5296459197998047, "loss": 1.1361, "loss_ce": 0.9863566756248474, "loss_region": 0.02998475357890129, "loss_total": 1.0163414478302002, "lr": 0.0011727033866570427, "router/selected_tokens_s0": 4235.875, "step": 1780, "tokens_trained": 5.832153056 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.507765406708744, "grad_norm": 0.7834669947624207, "loss": 1.1297, "loss_ce": 0.9555173516273499, "loss_region": 0.0301660243421793, "loss_total": 0.9856833815574646, "lr": 0.0011722964946731917, "router/selected_tokens_s0": 4416.375, "step": 1790, "tokens_trained": 5.864918496 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5106020849585136, "grad_norm": 0.9466329216957092, "loss": 1.1295, "loss_ce": 1.0096023082733154, "loss_region": 0.030076030641794205, "loss_total": 1.0396783351898193, "lr": 0.0011718896026893407, "router/selected_tokens_s0": 4354.625, "step": 1800, "tokens_trained": 5.897683936 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5134387632082831, "grad_norm": 1.151943325996399, "loss": 1.1267, "loss_ce": 1.0721287727355957, "loss_region": 0.029984835535287857, "loss_total": 1.1021136045455933, "lr": 0.0011714827107054896, "router/selected_tokens_s0": 4239.75, "step": 1810, "tokens_trained": 5.930449376 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5162754414580526, "grad_norm": 0.5502280592918396, "loss": 1.1249, "loss_ce": 1.0287433862686157, "loss_region": 0.029946208000183105, "loss_total": 1.0586895942687988, "lr": 0.0011710758187216386, "router/selected_tokens_s0": 4179.375, "step": 1820, "tokens_trained": 5.96321104 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5191121197078221, "grad_norm": 1.5447858572006226, "loss": 1.1319, "loss_ce": 1.1280238628387451, "loss_region": 0.030087478458881378, "loss_total": 1.158111333847046, "lr": 0.0011706689267377876, "router/selected_tokens_s0": 4389.75, "step": 1830, "tokens_trained": 5.99597648 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5219487979575916, "grad_norm": 0.9524003863334656, "loss": 1.1274, "loss_ce": 1.0977569818496704, "loss_region": 0.030062809586524963, "loss_total": 1.1278197765350342, "lr": 0.0011702620347539365, "router/selected_tokens_s0": 4354.0, "step": 1840, "tokens_trained": 6.028741744 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5247854762073612, "grad_norm": 0.6106662750244141, "loss": 1.1264, "loss_ce": 1.06783926486969, "loss_region": 0.029942721128463745, "loss_total": 1.097782015800476, "lr": 0.0011698551427700855, "router/selected_tokens_s0": 4162.625, "step": 1850, "tokens_trained": 6.061507184 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5276221544571307, "grad_norm": 1.2853341102600098, "loss": 1.1329, "loss_ce": 1.0429413318634033, "loss_region": 0.02999758906662464, "loss_total": 1.0729389190673828, "lr": 0.0011694482507862345, "router/selected_tokens_s0": 4247.5, "step": 1860, "tokens_trained": 6.094268624 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5304588327069002, "grad_norm": 2.993485927581787, "loss": 1.1236, "loss_ce": 1.0583568811416626, "loss_region": 0.030023684725165367, "loss_total": 1.0883805751800537, "lr": 0.0011690413588023834, "router/selected_tokens_s0": 4302.0, "step": 1870, "tokens_trained": 6.127034064 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5332955109566697, "grad_norm": 0.7363700866699219, "loss": 1.1308, "loss_ce": 1.1353397369384766, "loss_region": 0.029933562502264977, "loss_total": 1.1652733087539673, "lr": 0.0011686344668185324, "router/selected_tokens_s0": 4149.375, "step": 1880, "tokens_trained": 6.159799504 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5361321892064392, "grad_norm": 0.8693296313285828, "loss": 1.1274, "loss_ce": 1.0827381610870361, "loss_region": 0.030024589970707893, "loss_total": 1.1127628087997437, "lr": 0.0011682275748346814, "router/selected_tokens_s0": 4302.25, "step": 1890, "tokens_trained": 6.192561072 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5389688674562088, "grad_norm": 0.4028984606266022, "loss": 1.1162, "loss_ce": 1.1056593656539917, "loss_region": 0.030071411281824112, "loss_total": 1.1357307434082031, "lr": 0.0011678206828508303, "router/selected_tokens_s0": 4372.625, "step": 1900, "tokens_trained": 6.225326512 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5418055457059783, "grad_norm": 1.1904973983764648, "loss": 1.1294, "loss_ce": 1.0976545810699463, "loss_region": 0.030053725466132164, "loss_total": 1.1277083158493042, "lr": 0.0011674137908669793, "router/selected_tokens_s0": 4348.125, "step": 1910, "tokens_trained": 6.258091952 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5446422239557478, "grad_norm": 1.018221378326416, "loss": 1.1277, "loss_ce": 1.1479384899139404, "loss_region": 0.030054787173867226, "loss_total": 1.1779932975769043, "lr": 0.0011670068988831283, "router/selected_tokens_s0": 4353.25, "step": 1920, "tokens_trained": 6.290857392 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5474789022055173, "grad_norm": 0.4506734013557434, "loss": 1.1235, "loss_ce": 1.1137655973434448, "loss_region": 0.03005811758339405, "loss_total": 1.1438237428665161, "lr": 0.0011666000068992772, "router/selected_tokens_s0": 4341.5, "step": 1930, "tokens_trained": 6.323622832 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5503155804552868, "grad_norm": 1.5671348571777344, "loss": 1.1318, "loss_ce": 1.1652703285217285, "loss_region": 0.030141720548272133, "loss_total": 1.195412039756775, "lr": 0.0011661931149154262, "router/selected_tokens_s0": 4458.125, "step": 1940, "tokens_trained": 6.356388272 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5531522587050564, "grad_norm": 1.2511063814163208, "loss": 1.1246, "loss_ce": 1.2078148126602173, "loss_region": 0.03000708669424057, "loss_total": 1.2378219366073608, "lr": 0.0011657862229315751, "router/selected_tokens_s0": 4275.625, "step": 1950, "tokens_trained": 6.389153712 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5559889369548259, "grad_norm": 1.1278033256530762, "loss": 1.1253, "loss_ce": 1.1528972387313843, "loss_region": 0.029990505427122116, "loss_total": 1.1828877925872803, "lr": 0.0011653793309477243, "router/selected_tokens_s0": 4247.625, "step": 1960, "tokens_trained": 6.421919152 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5588256152045954, "grad_norm": 0.7347070574760437, "loss": 1.1292, "loss_ce": 1.1609221696853638, "loss_region": 0.03007410652935505, "loss_total": 1.1909962892532349, "lr": 0.0011649724389638733, "router/selected_tokens_s0": 4377.0, "step": 1970, "tokens_trained": 6.454684592 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5616622934543649, "grad_norm": 0.8754347562789917, "loss": 1.1321, "loss_ce": 1.1314905881881714, "loss_region": 0.030018918216228485, "loss_total": 1.1615095138549805, "lr": 0.0011645655469800223, "router/selected_tokens_s0": 4292.375, "step": 1980, "tokens_trained": 6.487450032 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5644989717041344, "grad_norm": 1.4375395774841309, "loss": 1.1251, "loss_ce": 1.15834641456604, "loss_region": 0.030011450871825218, "loss_total": 1.1883578300476074, "lr": 0.0011641586549961712, "router/selected_tokens_s0": 4281.875, "step": 1990, "tokens_trained": 6.520215472 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.567335649953904, "grad_norm": 1.3012388944625854, "loss": 1.1244, "loss_ce": 1.1547801494598389, "loss_region": 0.03002019412815571, "loss_total": 1.184800386428833, "lr": 0.00116375176301232, "router/selected_tokens_s0": 4298.375, "step": 2000, "tokens_trained": 6.552980912 }, { "epoch": 0.567335649953904, "eval_ppl": 2.997264738752139, "eval_runtime": 2.4974, "step": 2000, "tokens_trained": 6.552980912 }, { "epoch": 0.567335649953904, "eval_F": 0.33877094677913017, "eval_F_cds": 0.3354545528054273, "eval_F_dig": 0.3349740865171758, "eval_F_exon": 0.33771546252151097, "eval_F_intron": 0.3394511609404705, "eval_F_nig": 0.33961248247030124, "eval_F_promoter": 0.33587224314868064, "eval_F_utr": 0.3390466904438115, "eval_G": 0.3927095408069945, "eval_G_cds": 0.38760326352277413, "eval_G_dig": 0.38993240031773313, "eval_G_exon": 0.3922000848097159, "eval_G_intron": 0.39271919880055167, "eval_G_nig": 0.3935918508753731, "eval_G_promoter": 0.3926971556782782, "eval_G_utr": 0.3912176578977754, "eval_avg_bp_per_token": 2.9518469913300267, "eval_bp_per_token/cds": 2.981029744974208, "eval_bp_per_token/dig": 2.9853055512361997, "eval_bp_per_token/exon": 2.9610725920975693, "eval_bp_per_token/intron": 2.9459318896698954, "eval_bp_per_token/nig": 2.9445325234400035, "eval_bp_per_token/promoter": 2.977322539741189, "eval_bp_per_token/utr": 2.9494462803662875, "eval_ppl_cds": 3.6941119312579422, "eval_ppl_dig": 1.1218375588220217, "eval_ppl_exon": 3.4074634485917565, "eval_ppl_intron": 3.014504389955456, "eval_ppl_nig": 2.843623870937302, "eval_ppl_promoter": 3.3305259507076883, "eval_ppl_utr": 3.322006494837333, "step": 2000, "tokens_trained": 6.552980912 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5701723282036735, "grad_norm": 1.7854270935058594, "loss": 1.1275, "loss_ce": 1.1118180751800537, "loss_region": 0.030034875497221947, "loss_total": 1.1418529748916626, "lr": 0.001163344871028469, "router/selected_tokens_s0": 4323.625, "step": 2010, "tokens_trained": 6.585746352 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.573009006453443, "grad_norm": 1.2741203308105469, "loss": 1.1297, "loss_ce": 1.1596630811691284, "loss_region": 0.030020276084542274, "loss_total": 1.1896833181381226, "lr": 0.001162937979044618, "router/selected_tokens_s0": 4296.625, "step": 2020, "tokens_trained": 6.618511792 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5758456847032125, "grad_norm": 1.3113727569580078, "loss": 1.1274, "loss_ce": 1.130359411239624, "loss_region": 0.030052313581109047, "loss_total": 1.1604117155075073, "lr": 0.001162531087060767, "router/selected_tokens_s0": 4347.25, "step": 2030, "tokens_trained": 6.651277232 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.578682362952982, "grad_norm": 1.585740089416504, "loss": 1.1242, "loss_ce": 1.113228440284729, "loss_region": 0.029946262016892433, "loss_total": 1.143174648284912, "lr": 0.001162124195076916, "router/selected_tokens_s0": 4151.5, "step": 2040, "tokens_trained": 6.684041872 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5815190412027516, "grad_norm": 1.4227651357650757, "loss": 1.1227, "loss_ce": 1.1707289218902588, "loss_region": 0.03000037930905819, "loss_total": 1.200729250907898, "lr": 0.001161717303093065, "router/selected_tokens_s0": 4264.25, "step": 2050, "tokens_trained": 6.716806512 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5843557194525211, "grad_norm": 1.4349584579467773, "loss": 1.126, "loss_ce": 1.123897910118103, "loss_region": 0.029999535530805588, "loss_total": 1.1538974046707153, "lr": 0.001161310411109214, "router/selected_tokens_s0": 4258.5, "step": 2060, "tokens_trained": 6.749571952 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5871923977022906, "grad_norm": 1.525637149810791, "loss": 1.1223, "loss_ce": 1.0622094869613647, "loss_region": 0.03016025200486183, "loss_total": 1.092369794845581, "lr": 0.001160903519125363, "router/selected_tokens_s0": 4409.25, "step": 2070, "tokens_trained": 6.782337392 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5900290759520601, "grad_norm": 0.31481412053108215, "loss": 1.1308, "loss_ce": 1.1158243417739868, "loss_region": 0.030056282877922058, "loss_total": 1.1458805799484253, "lr": 0.001160496627141512, "router/selected_tokens_s0": 4358.875, "step": 2080, "tokens_trained": 6.815102832 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5928657542018296, "grad_norm": 1.4279309511184692, "loss": 1.1212, "loss_ce": 1.1024186611175537, "loss_region": 0.03000911884009838, "loss_total": 1.1324278116226196, "lr": 0.0011600897351576609, "router/selected_tokens_s0": 4277.25, "step": 2090, "tokens_trained": 6.847868272 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5957024324515992, "grad_norm": 1.3502033948898315, "loss": 1.1243, "loss_ce": 1.215091347694397, "loss_region": 0.03004975989460945, "loss_total": 1.2451411485671997, "lr": 0.0011596828431738098, "router/selected_tokens_s0": 4345.25, "step": 2100, "tokens_trained": 6.880633712 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.5985391107013687, "grad_norm": 0.30469629168510437, "loss": 1.1227, "loss_ce": 1.0989904403686523, "loss_region": 0.03004642389714718, "loss_total": 1.1290369033813477, "lr": 0.0011592759511899588, "router/selected_tokens_s0": 4339.125, "step": 2110, "tokens_trained": 6.913397016 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6013757889511382, "grad_norm": 3.0106451511383057, "loss": 1.1271, "loss_ce": 1.0580655336380005, "loss_region": 0.03005184419453144, "loss_total": 1.0881173610687256, "lr": 0.0011588690592061078, "router/selected_tokens_s0": 4347.75, "step": 2120, "tokens_trained": 6.946162296 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6042124672009077, "grad_norm": 1.4084529876708984, "loss": 1.1261, "loss_ce": 0.9337919354438782, "loss_region": 0.029956450685858727, "loss_total": 0.9637483954429626, "lr": 0.0011584621672222567, "router/selected_tokens_s0": 4181.25, "step": 2130, "tokens_trained": 6.978927736 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6070491454506772, "grad_norm": 0.7794283032417297, "loss": 1.1287, "loss_ce": 1.0321320295333862, "loss_region": 0.030011894181370735, "loss_total": 1.0621439218521118, "lr": 0.0011580552752384057, "router/selected_tokens_s0": 4285.875, "step": 2140, "tokens_trained": 7.011693176 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6098858237004467, "grad_norm": 0.7242727279663086, "loss": 1.1314, "loss_ce": 1.1077067852020264, "loss_region": 0.030075622722506523, "loss_total": 1.1377824544906616, "lr": 0.0011576483832545547, "router/selected_tokens_s0": 4383.25, "step": 2150, "tokens_trained": 7.044458616 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6127225019502163, "grad_norm": 0.8703320622444153, "loss": 1.1255, "loss_ce": 1.042706847190857, "loss_region": 0.030024481937289238, "loss_total": 1.072731375694275, "lr": 0.0011572414912707036, "router/selected_tokens_s0": 4306.0, "step": 2160, "tokens_trained": 7.077224056 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6155591801999858, "grad_norm": 2.464707374572754, "loss": 1.12, "loss_ce": 1.0845450162887573, "loss_region": 0.029988931491971016, "loss_total": 1.1145339012145996, "lr": 0.0011568345992868526, "router/selected_tokens_s0": 4238.875, "step": 2170, "tokens_trained": 7.109989496 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6183958584497553, "grad_norm": 2.0766637325286865, "loss": 1.1266, "loss_ce": 1.1240020990371704, "loss_region": 0.030013838782906532, "loss_total": 1.1540158987045288, "lr": 0.0011564277073030016, "router/selected_tokens_s0": 4291.875, "step": 2180, "tokens_trained": 7.142754936 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6212325366995248, "grad_norm": 1.402709722518921, "loss": 1.1265, "loss_ce": 1.1370148658752441, "loss_region": 0.03003770112991333, "loss_total": 1.1670525074005127, "lr": 0.0011560208153191505, "router/selected_tokens_s0": 4328.625, "step": 2190, "tokens_trained": 7.175520376 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6240692149492943, "grad_norm": 0.7657859325408936, "loss": 1.1259, "loss_ce": 1.116765022277832, "loss_region": 0.030005475506186485, "loss_total": 1.1467704772949219, "lr": 0.0011556139233352995, "router/selected_tokens_s0": 4272.125, "step": 2200, "tokens_trained": 7.208285816 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6269058931990639, "grad_norm": 3.5244100093841553, "loss": 1.1305, "loss_ce": 1.1446946859359741, "loss_region": 0.030087754130363464, "loss_total": 1.174782395362854, "lr": 0.0011552070313514487, "router/selected_tokens_s0": 4414.25, "step": 2210, "tokens_trained": 7.241051256 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6297425714488334, "grad_norm": 0.599822998046875, "loss": 1.1324, "loss_ce": 1.0551592111587524, "loss_region": 0.030122289434075356, "loss_total": 1.085281491279602, "lr": 0.0011548001393675976, "router/selected_tokens_s0": 4453.875, "step": 2220, "tokens_trained": 7.273816696 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6325792496986029, "grad_norm": 2.314722776412964, "loss": 1.1277, "loss_ce": 1.1485532522201538, "loss_region": 0.030024103820323944, "loss_total": 1.1785773038864136, "lr": 0.0011543932473837466, "router/selected_tokens_s0": 4313.5, "step": 2230, "tokens_trained": 7.306582136 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6354159279483724, "grad_norm": 2.072960615158081, "loss": 1.131, "loss_ce": 1.0349353551864624, "loss_region": 0.030028166249394417, "loss_total": 1.0649635791778564, "lr": 0.0011539863553998956, "router/selected_tokens_s0": 4319.625, "step": 2240, "tokens_trained": 7.339347576 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.638252606198142, "grad_norm": 1.371410846710205, "loss": 1.1226, "loss_ce": 1.0738561153411865, "loss_region": 0.030064314603805542, "loss_total": 1.1039204597473145, "lr": 0.0011535794634160443, "router/selected_tokens_s0": 4378.375, "step": 2250, "tokens_trained": 7.372113016 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6410892844479115, "grad_norm": 3.474445343017578, "loss": 1.1284, "loss_ce": 1.0069116353988647, "loss_region": 0.030036170035600662, "loss_total": 1.0369478464126587, "lr": 0.0011531725714321933, "router/selected_tokens_s0": 4332.625, "step": 2260, "tokens_trained": 7.404878456 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.643925962697681, "grad_norm": 0.5796771049499512, "loss": 1.1245, "loss_ce": 1.138779640197754, "loss_region": 0.030022747814655304, "loss_total": 1.1688023805618286, "lr": 0.0011527656794483422, "router/selected_tokens_s0": 4308.875, "step": 2270, "tokens_trained": 7.437643896 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6467626409474505, "grad_norm": 1.155604362487793, "loss": 1.1216, "loss_ce": 0.9782689809799194, "loss_region": 0.030030813068151474, "loss_total": 1.0082998275756836, "lr": 0.0011523587874644914, "router/selected_tokens_s0": 4321.625, "step": 2280, "tokens_trained": 7.470409336 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.64959931919722, "grad_norm": 1.8259997367858887, "loss": 1.1318, "loss_ce": 1.055479884147644, "loss_region": 0.030021535232663155, "loss_total": 1.0855014324188232, "lr": 0.0011519518954806404, "router/selected_tokens_s0": 4307.375, "step": 2290, "tokens_trained": 7.503173472 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6524359974469895, "grad_norm": 1.2909961938858032, "loss": 1.1216, "loss_ce": 1.1016438007354736, "loss_region": 0.030030114576220512, "loss_total": 1.1316739320755005, "lr": 0.0011515450034967894, "router/selected_tokens_s0": 4321.625, "step": 2300, "tokens_trained": 7.535938912 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6552726756967591, "grad_norm": 3.855242967605591, "loss": 1.1332, "loss_ce": 1.1084688901901245, "loss_region": 0.030001208186149597, "loss_total": 1.1384700536727905, "lr": 0.0011511381115129383, "router/selected_tokens_s0": 4267.625, "step": 2310, "tokens_trained": 7.568704352 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6581093539465286, "grad_norm": 0.6401855945587158, "loss": 1.1235, "loss_ce": 1.068629503250122, "loss_region": 0.030046915635466576, "loss_total": 1.0986764430999756, "lr": 0.0011507312195290873, "router/selected_tokens_s0": 4353.5, "step": 2320, "tokens_trained": 7.601469792 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6609460321962981, "grad_norm": 2.758415460586548, "loss": 1.1224, "loss_ce": 1.1197397708892822, "loss_region": 0.030033273622393608, "loss_total": 1.1497730016708374, "lr": 0.0011503243275452363, "router/selected_tokens_s0": 4317.625, "step": 2330, "tokens_trained": 7.634233608 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6637827104460676, "grad_norm": 3.6356966495513916, "loss": 1.1258, "loss_ce": 1.192346453666687, "loss_region": 0.030019540339708328, "loss_total": 1.2223659753799438, "lr": 0.0011499174355613852, "router/selected_tokens_s0": 4307.0, "step": 2340, "tokens_trained": 7.666998248 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6666193886958371, "grad_norm": 0.5084363222122192, "loss": 1.1211, "loss_ce": 1.0241565704345703, "loss_region": 0.030024418607354164, "loss_total": 1.0541809797286987, "lr": 0.0011495105435775342, "router/selected_tokens_s0": 4311.75, "step": 2350, "tokens_trained": 7.699763688 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6694560669456067, "grad_norm": 2.6118147373199463, "loss": 1.1205, "loss_ce": 1.094053864479065, "loss_region": 0.030054572969675064, "loss_total": 1.1241084337234497, "lr": 0.0011491036515936831, "router/selected_tokens_s0": 4375.625, "step": 2360, "tokens_trained": 7.732529128 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6722927451953762, "grad_norm": 1.5716001987457275, "loss": 1.1174, "loss_ce": 1.0806825160980225, "loss_region": 0.02999335154891014, "loss_total": 1.1106758117675781, "lr": 0.0011486967596098321, "router/selected_tokens_s0": 4245.125, "step": 2370, "tokens_trained": 7.765294568 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6751294234451457, "grad_norm": 1.6855603456497192, "loss": 1.1248, "loss_ce": 1.0957375764846802, "loss_region": 0.030019070953130722, "loss_total": 1.1257566213607788, "lr": 0.001148289867625981, "router/selected_tokens_s0": 4306.25, "step": 2380, "tokens_trained": 7.798060008 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6779661016949152, "grad_norm": 1.7085551023483276, "loss": 1.1219, "loss_ce": 1.0849840641021729, "loss_region": 0.029990842565894127, "loss_total": 1.114974856376648, "lr": 0.00114788297564213, "router/selected_tokens_s0": 4250.875, "step": 2390, "tokens_trained": 7.830825448 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6808027799446847, "grad_norm": 2.7529702186584473, "loss": 1.1278, "loss_ce": 1.1395268440246582, "loss_region": 0.030015477910637856, "loss_total": 1.1695423126220703, "lr": 0.001147476083658279, "router/selected_tokens_s0": 4305.125, "step": 2400, "tokens_trained": 7.863590888 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6836394581944543, "grad_norm": 1.855435848236084, "loss": 1.1225, "loss_ce": 1.055867075920105, "loss_region": 0.030039696022868156, "loss_total": 1.085906744003296, "lr": 0.001147069191674428, "router/selected_tokens_s0": 4357.375, "step": 2410, "tokens_trained": 7.896356328 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6864761364442238, "grad_norm": 1.9066152572631836, "loss": 1.1243, "loss_ce": 0.9804560542106628, "loss_region": 0.03004065528512001, "loss_total": 1.010496735572815, "lr": 0.001146662299690577, "router/selected_tokens_s0": 4339.375, "step": 2420, "tokens_trained": 7.929121768 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6893128146939933, "grad_norm": 1.6631957292556763, "loss": 1.1181, "loss_ce": 1.1269235610961914, "loss_region": 0.030016232281923294, "loss_total": 1.1569397449493408, "lr": 0.001146255407706726, "router/selected_tokens_s0": 4304.375, "step": 2430, "tokens_trained": 7.961887208 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6921494929437628, "grad_norm": 1.932186245918274, "loss": 1.1318, "loss_ce": 1.1084073781967163, "loss_region": 0.030037561431527138, "loss_total": 1.1384449005126953, "lr": 0.0011458485157228749, "router/selected_tokens_s0": 4342.375, "step": 2440, "tokens_trained": 7.994651848 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6949861711935323, "grad_norm": 2.0729987621307373, "loss": 1.1219, "loss_ce": 1.0754549503326416, "loss_region": 0.030010342597961426, "loss_total": 1.105465292930603, "lr": 0.0011454416237390238, "router/selected_tokens_s0": 4284.25, "step": 2450, "tokens_trained": 8.027417288 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.6978228494433019, "grad_norm": 2.743365526199341, "loss": 1.1183, "loss_ce": 1.1507514715194702, "loss_region": 0.030012760311365128, "loss_total": 1.1807641983032227, "lr": 0.001145034731755173, "router/selected_tokens_s0": 4299.125, "step": 2460, "tokens_trained": 8.060182704 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7006595276930714, "grad_norm": 1.968074083328247, "loss": 1.1248, "loss_ce": 1.1554365158081055, "loss_region": 0.03006228432059288, "loss_total": 1.185498833656311, "lr": 0.001144627839771322, "router/selected_tokens_s0": 4397.5, "step": 2470, "tokens_trained": 8.092948144 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7034962059428409, "grad_norm": 0.6022619605064392, "loss": 1.1233, "loss_ce": 1.0739916563034058, "loss_region": 0.030015716329216957, "loss_total": 1.104007363319397, "lr": 0.001144220947787471, "router/selected_tokens_s0": 4304.25, "step": 2480, "tokens_trained": 8.125713584 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7063328841926104, "grad_norm": 2.9086802005767822, "loss": 1.1155, "loss_ce": 1.1227823495864868, "loss_region": 0.030057305470108986, "loss_total": 1.1528396606445312, "lr": 0.00114381405580362, "router/selected_tokens_s0": 4393.875, "step": 2490, "tokens_trained": 8.158479016 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7091695624423799, "grad_norm": 1.8187512159347534, "loss": 1.1248, "loss_ce": 1.0580413341522217, "loss_region": 0.030032671988010406, "loss_total": 1.088073968887329, "lr": 0.0011434071638197687, "router/selected_tokens_s0": 4340.0, "step": 2500, "tokens_trained": 8.191244456 }, { "epoch": 0.7091695624423799, "eval_ppl": 2.9815305929864326, "eval_runtime": 2.4796, "step": 2500, "tokens_trained": 8.191244456 }, { "epoch": 0.7091695624423799, "eval_F": 0.34048558481131336, "eval_F_cds": 0.3413653968998391, "eval_F_dig": 0.3326561970987317, "eval_F_exon": 0.34301915535870453, "eval_F_intron": 0.3409895477582185, "eval_F_nig": 0.34018024599300895, "eval_F_promoter": 0.3386885010090298, "eval_F_utr": 0.34306656745268094, "eval_G": 0.37360140820500265, "eval_G_cds": 0.37391617995023085, "eval_G_dig": 0.39410936238508215, "eval_G_exon": 0.37318875715857475, "eval_G_intron": 0.3727733807645177, "eval_G_nig": 0.3734594960312147, "eval_G_promoter": 0.37594098275253596, "eval_G_utr": 0.3722500326080449, "eval_avg_bp_per_token": 2.9369819005822793, "eval_bp_per_token/cds": 2.929412322050359, "eval_bp_per_token/dig": 3.006106631175135, "eval_bp_per_token/exon": 2.915289086273542, "eval_bp_per_token/intron": 2.932641210190579, "eval_bp_per_token/nig": 2.9396180753557073, "eval_bp_per_token/promoter": 2.952565549231147, "eval_bp_per_token/utr": 2.9148861908204733, "eval_ppl_cds": 3.7211953918524787, "eval_ppl_dig": 1.1071312956552213, "eval_ppl_exon": 3.408594147596357, "eval_ppl_intron": 2.996762231969892, "eval_ppl_nig": 2.8097869859130795, "eval_ppl_promoter": 3.341004188366384, "eval_ppl_utr": 3.3285188682998834, "step": 2500, "tokens_trained": 8.191244456 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7120062406921495, "grad_norm": 1.3883668184280396, "loss": 1.1168, "loss_ce": 1.0345538854599, "loss_region": 0.030011983588337898, "loss_total": 1.064565896987915, "lr": 0.0011430002718359176, "router/selected_tokens_s0": 4293.25, "step": 2510, "tokens_trained": 8.224009896 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.714842918941919, "grad_norm": 0.5920007228851318, "loss": 1.1128, "loss_ce": 1.1446270942687988, "loss_region": 0.030029037967324257, "loss_total": 1.1746561527252197, "lr": 0.0011425933798520666, "router/selected_tokens_s0": 4338.125, "step": 2520, "tokens_trained": 8.256775336 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7176795971916885, "grad_norm": 2.293912410736084, "loss": 1.119, "loss_ce": 1.1278671026229858, "loss_region": 0.030034860596060753, "loss_total": 1.1579020023345947, "lr": 0.0011421864878682158, "router/selected_tokens_s0": 4356.25, "step": 2530, "tokens_trained": 8.289540776 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.720516275441458, "grad_norm": 1.4504122734069824, "loss": 1.1161, "loss_ce": 0.9545093774795532, "loss_region": 0.030026227235794067, "loss_total": 0.9845355749130249, "lr": 0.0011417795958843647, "router/selected_tokens_s0": 4322.375, "step": 2540, "tokens_trained": 8.322306216 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7233529536912275, "grad_norm": 1.777256727218628, "loss": 1.1177, "loss_ce": 1.0747570991516113, "loss_region": 0.030009755864739418, "loss_total": 1.104766845703125, "lr": 0.0011413727039005137, "router/selected_tokens_s0": 4293.375, "step": 2550, "tokens_trained": 8.355071656 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7261896319409971, "grad_norm": 1.637231707572937, "loss": 1.1121, "loss_ce": 1.1526259183883667, "loss_region": 0.030018767341971397, "loss_total": 1.1826447248458862, "lr": 0.0011409658119166627, "router/selected_tokens_s0": 4318.5, "step": 2560, "tokens_trained": 8.387835072 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7290263101907666, "grad_norm": 1.0746310949325562, "loss": 1.1151, "loss_ce": 1.1064670085906982, "loss_region": 0.03001333586871624, "loss_total": 1.1364803314208984, "lr": 0.0011405589199328116, "router/selected_tokens_s0": 4294.375, "step": 2570, "tokens_trained": 8.420600512 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7318629884405361, "grad_norm": 1.3798960447311401, "loss": 1.1198, "loss_ce": 1.073905110359192, "loss_region": 0.030032221227884293, "loss_total": 1.1039373874664307, "lr": 0.0011401520279489606, "router/selected_tokens_s0": 4356.375, "step": 2580, "tokens_trained": 8.453365928 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7346996666903056, "grad_norm": 1.8040990829467773, "loss": 1.1175, "loss_ce": 1.0255845785140991, "loss_region": 0.03001689724624157, "loss_total": 1.0556014776229858, "lr": 0.0011397451359651096, "router/selected_tokens_s0": 4312.5, "step": 2590, "tokens_trained": 8.486131368 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7375363449400751, "grad_norm": 2.420259952545166, "loss": 1.1193, "loss_ce": 1.0581092834472656, "loss_region": 0.030017009004950523, "loss_total": 1.088126301765442, "lr": 0.0011393382439812585, "router/selected_tokens_s0": 4316.25, "step": 2600, "tokens_trained": 8.518896808 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7403730231898447, "grad_norm": 2.068054437637329, "loss": 1.1114, "loss_ce": 1.0681673288345337, "loss_region": 0.030040811747312546, "loss_total": 1.0982081890106201, "lr": 0.0011389313519974075, "router/selected_tokens_s0": 4369.0, "step": 2610, "tokens_trained": 8.551662248 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7432097014396142, "grad_norm": 1.7490754127502441, "loss": 1.1182, "loss_ce": 1.0639960765838623, "loss_region": 0.030034611001610756, "loss_total": 1.094030737876892, "lr": 0.0011385244600135565, "router/selected_tokens_s0": 4363.5, "step": 2620, "tokens_trained": 8.584426888 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7460463796893837, "grad_norm": 1.4811182022094727, "loss": 1.1131, "loss_ce": 1.0907317399978638, "loss_region": 0.03001326695084572, "loss_total": 1.120745062828064, "lr": 0.0011381175680297054, "router/selected_tokens_s0": 4307.875, "step": 2630, "tokens_trained": 8.617192328 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7488830579391532, "grad_norm": 2.1497602462768555, "loss": 1.1096, "loss_ce": 1.123599886894226, "loss_region": 0.030037013813853264, "loss_total": 1.1536369323730469, "lr": 0.0011377106760458544, "router/selected_tokens_s0": 4368.5, "step": 2640, "tokens_trained": 8.649951656 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7517197361889227, "grad_norm": 2.179588556289673, "loss": 1.1129, "loss_ce": 0.9365400671958923, "loss_region": 0.030036216601729393, "loss_total": 0.9665762782096863, "lr": 0.0011373037840620034, "router/selected_tokens_s0": 4350.375, "step": 2650, "tokens_trained": 8.682717096 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7545564144386923, "grad_norm": 1.6021926403045654, "loss": 1.1095, "loss_ce": 1.1449388265609741, "loss_region": 0.030037803575396538, "loss_total": 1.1749765872955322, "lr": 0.0011368968920781523, "router/selected_tokens_s0": 4373.0, "step": 2660, "tokens_trained": 8.715482536 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7573930926884618, "grad_norm": 1.2494678497314453, "loss": 1.1097, "loss_ce": 1.0806819200515747, "loss_region": 0.03000866435468197, "loss_total": 1.1106905937194824, "lr": 0.0011364900000943013, "router/selected_tokens_s0": 4295.25, "step": 2670, "tokens_trained": 8.748247976 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7602297709382313, "grad_norm": 1.3196409940719604, "loss": 1.1136, "loss_ce": 1.069360375404358, "loss_region": 0.030009115114808083, "loss_total": 1.0993695259094238, "lr": 0.0011360831081104503, "router/selected_tokens_s0": 4291.75, "step": 2680, "tokens_trained": 8.781013416 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7630664491880008, "grad_norm": 2.674771308898926, "loss": 1.1188, "loss_ce": 1.1771190166473389, "loss_region": 0.030019955709576607, "loss_total": 1.207139015197754, "lr": 0.0011356762161265992, "router/selected_tokens_s0": 4330.125, "step": 2690, "tokens_trained": 8.813778696 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7659031274377703, "grad_norm": 1.6932164430618286, "loss": 1.1031, "loss_ce": 1.0857900381088257, "loss_region": 0.0300260242074728, "loss_total": 1.1158161163330078, "lr": 0.0011352693241427482, "router/selected_tokens_s0": 4347.5, "step": 2700, "tokens_trained": 8.846544136 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7687398056875399, "grad_norm": 1.5329583883285522, "loss": 1.1098, "loss_ce": 1.0980348587036133, "loss_region": 0.030030502006411552, "loss_total": 1.1280653476715088, "lr": 0.0011348624321588974, "router/selected_tokens_s0": 4366.375, "step": 2710, "tokens_trained": 8.879309576 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7715764839373094, "grad_norm": 1.829464077949524, "loss": 1.1093, "loss_ce": 1.1128755807876587, "loss_region": 0.03000422567129135, "loss_total": 1.142879843711853, "lr": 0.0011344555401750463, "router/selected_tokens_s0": 4282.0, "step": 2720, "tokens_trained": 8.912075016 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7744131621870789, "grad_norm": 2.8766870498657227, "loss": 1.1187, "loss_ce": 1.12075674533844, "loss_region": 0.030019240453839302, "loss_total": 1.1507760286331177, "lr": 0.0011340486481911953, "router/selected_tokens_s0": 4327.25, "step": 2730, "tokens_trained": 8.944840456 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7772498404368484, "grad_norm": 2.2969014644622803, "loss": 1.1166, "loss_ce": 1.0795077085494995, "loss_region": 0.030028002336621284, "loss_total": 1.1095356941223145, "lr": 0.001133641756207344, "router/selected_tokens_s0": 4352.75, "step": 2740, "tokens_trained": 8.977605896 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7800865186866179, "grad_norm": 1.7521798610687256, "loss": 1.1139, "loss_ce": 1.1274807453155518, "loss_region": 0.030016858130693436, "loss_total": 1.1574976444244385, "lr": 0.001133234864223493, "router/selected_tokens_s0": 4320.0, "step": 2750, "tokens_trained": 9.010371336 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7829231969363875, "grad_norm": 2.6245367527008057, "loss": 1.1075, "loss_ce": 1.1328058242797852, "loss_region": 0.03003484010696411, "loss_total": 1.1628406047821045, "lr": 0.001132827972239642, "router/selected_tokens_s0": 4367.625, "step": 2760, "tokens_trained": 9.043136776 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.785759875186157, "grad_norm": 1.162583351135254, "loss": 1.1181, "loss_ce": 1.151093602180481, "loss_region": 0.030036624521017075, "loss_total": 1.1811301708221436, "lr": 0.001132421080255791, "router/selected_tokens_s0": 4392.5, "step": 2770, "tokens_trained": 9.075902216 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7885965534359265, "grad_norm": 1.4981096982955933, "loss": 1.1104, "loss_ce": 1.0844680070877075, "loss_region": 0.030015481635928154, "loss_total": 1.1144834756851196, "lr": 0.0011320141882719401, "router/selected_tokens_s0": 4314.625, "step": 2780, "tokens_trained": 9.108667656 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.791433231685696, "grad_norm": 1.8612878322601318, "loss": 1.1073, "loss_ce": 1.0089409351348877, "loss_region": 0.029995379969477654, "loss_total": 1.0389362573623657, "lr": 0.001131607296288089, "router/selected_tokens_s0": 4257.875, "step": 2790, "tokens_trained": 9.14143004 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7942699099354655, "grad_norm": 0.6861640810966492, "loss": 1.1058, "loss_ce": 0.9385975003242493, "loss_region": 0.029996687546372414, "loss_total": 0.9685941934585571, "lr": 0.001131200404304238, "router/selected_tokens_s0": 4290.625, "step": 2800, "tokens_trained": 9.17419548 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.797106588185235, "grad_norm": 2.205390214920044, "loss": 1.108, "loss_ce": 1.0670945644378662, "loss_region": 0.030025651678442955, "loss_total": 1.0971201658248901, "lr": 0.001130793512320387, "router/selected_tokens_s0": 4342.875, "step": 2810, "tokens_trained": 9.20696092 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.7999432664350046, "grad_norm": 2.068150520324707, "loss": 1.106, "loss_ce": 1.0238165855407715, "loss_region": 0.03002651408314705, "loss_total": 1.0538431406021118, "lr": 0.001130386620336536, "router/selected_tokens_s0": 4359.75, "step": 2820, "tokens_trained": 9.23972636 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8027799446847741, "grad_norm": 1.1060576438903809, "loss": 1.1065, "loss_ce": 1.0474674701690674, "loss_region": 0.03000919334590435, "loss_total": 1.0774766206741333, "lr": 0.001129979728352685, "router/selected_tokens_s0": 4301.75, "step": 2830, "tokens_trained": 9.2724918 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8056166229345436, "grad_norm": 1.369165301322937, "loss": 1.1081, "loss_ce": 1.0370676517486572, "loss_region": 0.030027758330106735, "loss_total": 1.067095398902893, "lr": 0.001129572836368834, "router/selected_tokens_s0": 4375.625, "step": 2840, "tokens_trained": 9.30525692 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8084533011843131, "grad_norm": 2.285675525665283, "loss": 1.109, "loss_ce": 1.0967503786087036, "loss_region": 0.0300269927829504, "loss_total": 1.1267774105072021, "lr": 0.0011291659443849829, "router/selected_tokens_s0": 4370.0, "step": 2850, "tokens_trained": 9.33802236 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8112899794340827, "grad_norm": 0.8950642943382263, "loss": 1.1015, "loss_ce": 1.091797947883606, "loss_region": 0.03003113530576229, "loss_total": 1.1218290328979492, "lr": 0.0011287590524011318, "router/selected_tokens_s0": 4384.125, "step": 2860, "tokens_trained": 9.3707878 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8141266576838522, "grad_norm": 2.1465282440185547, "loss": 1.1012, "loss_ce": 0.9929934144020081, "loss_region": 0.030015377327799797, "loss_total": 1.0230088233947754, "lr": 0.0011283521604172808, "router/selected_tokens_s0": 4333.0, "step": 2870, "tokens_trained": 9.403548272 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8169633359336217, "grad_norm": 2.1108782291412354, "loss": 1.1029, "loss_ce": 1.0729644298553467, "loss_region": 0.03002534806728363, "loss_total": 1.1029897928237915, "lr": 0.0011279452684334298, "router/selected_tokens_s0": 4357.25, "step": 2880, "tokens_trained": 9.436313712 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8198000141833912, "grad_norm": 1.7104750871658325, "loss": 1.1041, "loss_ce": 1.116651177406311, "loss_region": 0.030016740784049034, "loss_total": 1.1466679573059082, "lr": 0.0011275383764495787, "router/selected_tokens_s0": 4340.875, "step": 2890, "tokens_trained": 9.469079152 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8226366924331607, "grad_norm": 1.7549395561218262, "loss": 1.1098, "loss_ce": 0.977597713470459, "loss_region": 0.030016236007213593, "loss_total": 1.0076138973236084, "lr": 0.0011271314844657277, "router/selected_tokens_s0": 4328.875, "step": 2900, "tokens_trained": 9.50184356 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8254733706829303, "grad_norm": 2.076667547225952, "loss": 1.1041, "loss_ce": 0.9882082343101501, "loss_region": 0.03001856803894043, "loss_total": 1.0182268619537354, "lr": 0.0011267245924818767, "router/selected_tokens_s0": 4341.75, "step": 2910, "tokens_trained": 9.534608992 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8283100489326998, "grad_norm": 1.930834412574768, "loss": 1.1031, "loss_ce": 1.1864138841629028, "loss_region": 0.03002503328025341, "loss_total": 1.216438889503479, "lr": 0.0011263177004980256, "router/selected_tokens_s0": 4379.125, "step": 2920, "tokens_trained": 9.567373632 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8311467271824693, "grad_norm": 0.7202333807945251, "loss": 1.103, "loss_ce": 1.0883651971817017, "loss_region": 0.030032740905880928, "loss_total": 1.1183979511260986, "lr": 0.0011259108085141746, "router/selected_tokens_s0": 4386.375, "step": 2930, "tokens_trained": 9.600139072 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8339834054322388, "grad_norm": 1.0626195669174194, "loss": 1.1043, "loss_ce": 1.0197147130966187, "loss_region": 0.03001200221478939, "loss_total": 1.0497267246246338, "lr": 0.0011255039165303236, "router/selected_tokens_s0": 4317.625, "step": 2940, "tokens_trained": 9.632904512 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8368200836820083, "grad_norm": 2.428861379623413, "loss": 1.1036, "loss_ce": 0.9022196531295776, "loss_region": 0.030008511617779732, "loss_total": 0.932228147983551, "lr": 0.0011250970245464725, "router/selected_tokens_s0": 4322.125, "step": 2950, "tokens_trained": 9.665669952 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8396567619317779, "grad_norm": 0.9146430492401123, "loss": 1.1015, "loss_ce": 1.1206673383712769, "loss_region": 0.030019836500287056, "loss_total": 1.1506872177124023, "lr": 0.0011246901325626217, "router/selected_tokens_s0": 4355.75, "step": 2960, "tokens_trained": 9.698432616 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8424934401815474, "grad_norm": 1.3183574676513672, "loss": 1.0992, "loss_ce": 1.088549256324768, "loss_region": 0.030013682320713997, "loss_total": 1.118562936782837, "lr": 0.0011242832405787707, "router/selected_tokens_s0": 4324.875, "step": 2970, "tokens_trained": 9.731196464 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8453301184313169, "grad_norm": 1.7237669229507446, "loss": 1.1016, "loss_ce": 1.1303554773330688, "loss_region": 0.030019812285900116, "loss_total": 1.1603752374649048, "lr": 0.0011238763485949196, "router/selected_tokens_s0": 4352.75, "step": 2980, "tokens_trained": 9.763955848 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8481667966810864, "grad_norm": 2.353868246078491, "loss": 1.097, "loss_ce": 1.1438567638397217, "loss_region": 0.03001641482114792, "loss_total": 1.1738731861114502, "lr": 0.0011234694566110684, "router/selected_tokens_s0": 4346.25, "step": 2990, "tokens_trained": 9.796721288 }, { "comp/rl_weight": 0.03, "comp/strictness": 0.0, "epoch": 0.8510034749308559, "grad_norm": 2.239737033843994, "loss": 1.099, "loss_ce": 1.137770414352417, "loss_region": 0.03001844510436058, "loss_total": 1.1677888631820679, "lr": 0.0011230625646272174, "router/selected_tokens_s0": 4345.625, "step": 3000, "tokens_trained": 9.829486728 }, { "epoch": 0.8510034749308559, "eval_ppl": 2.91798250107805, "eval_runtime": 2.489, "step": 3000, "tokens_trained": 9.829486728 }, { "epoch": 0.8510034749308559, "eval_F": 0.34119725725854944, "eval_F_cds": 0.339909922293828, "eval_F_dig": 0.3374221944422741, "eval_F_exon": 0.3444720286625102, "eval_F_intron": 0.3423051363848719, "eval_F_nig": 0.3420074982635899, "eval_F_promoter": 0.33568609090152685, "eval_F_utr": 0.3433317082766702, "eval_G": 0.35626090599344656, "eval_G_cds": 0.3533774528284723, "eval_G_dig": 0.39929882420827145, "eval_G_exon": 0.35481589922102014, "eval_G_intron": 0.3559872186522367, "eval_G_nig": 0.35704285773301014, "eval_G_promoter": 0.354736183175574, "eval_G_utr": 0.3543053844969594, "eval_avg_bp_per_token": 2.930855916119598, "eval_bp_per_token/cds": 2.9419558959963843, "eval_bp_per_token/dig": 2.9636461870947826, "eval_bp_per_token/exon": 2.9029933254166496, "eval_bp_per_token/intron": 2.921370127720335, "eval_bp_per_token/nig": 2.923912502144284, "eval_bp_per_token/promoter": 2.9789735920078644, "eval_bp_per_token/utr": 2.9126351452344177, "eval_ppl_cds": 3.5636364626812047, "eval_ppl_dig": 1.0968188962634289, "eval_ppl_exon": 3.3285669782872387, "eval_ppl_intron": 2.935885553210843, "eval_ppl_nig": 2.7347129604188645, "eval_ppl_promoter": 3.292230226733986, "eval_ppl_utr": 3.2942767869833, "step": 3000, "tokens_trained": 9.829486728 } ], "logging_steps": 10, "max_steps": 30600, "num_input_tokens_seen": 0, "num_train_epochs": 9, "save_steps": 3000, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }