| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 63.36296576508417, | |
| "global_step": 335000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.18914318138831096, | |
| "learning_rate": 4.990542840930585e-05, | |
| "loss": 2.2949150390625, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.3782863627766219, | |
| "learning_rate": 4.981085681861169e-05, | |
| "loss": 2.31815234375, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.5674295441649329, | |
| "learning_rate": 4.9716285227917534e-05, | |
| "loss": 2.32872216796875, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.7565727255532438, | |
| "learning_rate": 4.962171363722338e-05, | |
| "loss": 2.33826904296875, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.9457159069415547, | |
| "learning_rate": 4.9527142046529224e-05, | |
| "loss": 2.340685546875, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.1348590883298657, | |
| "learning_rate": 4.943257045583507e-05, | |
| "loss": 2.28626953125, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.3240022697181766, | |
| "learning_rate": 4.9337998865140915e-05, | |
| "loss": 2.27450390625, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.5131454511064875, | |
| "learning_rate": 4.9243427274446756e-05, | |
| "loss": 2.287498046875, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 1.7022886324947986, | |
| "learning_rate": 4.9148855683752605e-05, | |
| "loss": 2.30305078125, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 1.8914318138831097, | |
| "learning_rate": 4.905428409305845e-05, | |
| "loss": 2.30912890625, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 2.0805749952714203, | |
| "learning_rate": 4.895971250236429e-05, | |
| "loss": 2.272595703125, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 2.2697181766597314, | |
| "learning_rate": 4.886514091167014e-05, | |
| "loss": 2.238087890625, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 2.4588613580480425, | |
| "learning_rate": 4.877056932097598e-05, | |
| "loss": 2.253140625, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 2.648004539436353, | |
| "learning_rate": 4.867599773028183e-05, | |
| "loss": 2.26483203125, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 2.8371477208246643, | |
| "learning_rate": 4.858142613958767e-05, | |
| "loss": 2.2735078125, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 3.0262909022129754, | |
| "learning_rate": 4.848685454889351e-05, | |
| "loss": 2.27125, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 3.215434083601286, | |
| "learning_rate": 4.839228295819936e-05, | |
| "loss": 2.20089453125, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 3.404577264989597, | |
| "learning_rate": 4.829771136750521e-05, | |
| "loss": 2.2166953125, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 3.593720446377908, | |
| "learning_rate": 4.820313977681105e-05, | |
| "loss": 2.2287421875, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 3.782863627766219, | |
| "learning_rate": 4.810856818611689e-05, | |
| "loss": 2.242296875, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 3.97200680915453, | |
| "learning_rate": 4.801399659542274e-05, | |
| "loss": 2.25162109375, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 4.161149990542841, | |
| "learning_rate": 4.791942500472858e-05, | |
| "loss": 2.17917578125, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 4.350293171931152, | |
| "learning_rate": 4.782485341403443e-05, | |
| "loss": 2.1837734375, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 4.539436353319463, | |
| "learning_rate": 4.773028182334027e-05, | |
| "loss": 2.1981953125, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 4.728579534707774, | |
| "learning_rate": 4.763571023264611e-05, | |
| "loss": 2.2091328125, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 4.917722716096085, | |
| "learning_rate": 4.754113864195196e-05, | |
| "loss": 2.22294140625, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 5.106865897484396, | |
| "learning_rate": 4.744656705125781e-05, | |
| "loss": 2.17415625, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 5.296009078872706, | |
| "learning_rate": 4.7351995460563645e-05, | |
| "loss": 2.1536875, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 5.485152260261017, | |
| "learning_rate": 4.725742386986949e-05, | |
| "loss": 2.17046484375, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 5.6742954416493285, | |
| "learning_rate": 4.716285227917534e-05, | |
| "loss": 2.17616796875, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 5.86343862303764, | |
| "learning_rate": 4.706828068848118e-05, | |
| "loss": 2.18503125, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 6.052581804425951, | |
| "learning_rate": 4.6973709097787025e-05, | |
| "loss": 2.172359375, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 6.241724985814262, | |
| "learning_rate": 4.687913750709287e-05, | |
| "loss": 2.118359375, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 6.430868167202572, | |
| "learning_rate": 4.6784565916398715e-05, | |
| "loss": 2.133546875, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 6.620011348590883, | |
| "learning_rate": 4.6689994325704564e-05, | |
| "loss": 2.1423671875, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 6.809154529979194, | |
| "learning_rate": 4.6595422735010405e-05, | |
| "loss": 2.1594921875, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 6.998297711367505, | |
| "learning_rate": 4.650085114431625e-05, | |
| "loss": 2.17365625, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 7.187440892755816, | |
| "learning_rate": 4.6406279553622095e-05, | |
| "loss": 2.0884140625, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 7.376584074144127, | |
| "learning_rate": 4.631170796292794e-05, | |
| "loss": 2.1023359375, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 7.565727255532438, | |
| "learning_rate": 4.6217136372233786e-05, | |
| "loss": 2.1240625, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 7.754870436920749, | |
| "learning_rate": 4.612256478153963e-05, | |
| "loss": 2.126953125, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 7.94401361830906, | |
| "learning_rate": 4.602799319084547e-05, | |
| "loss": 2.1407265625, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 8.133156799697371, | |
| "learning_rate": 4.593342160015132e-05, | |
| "loss": 2.080640625, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 8.322299981085681, | |
| "learning_rate": 4.5838850009457166e-05, | |
| "loss": 2.07196875, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 8.511443162473993, | |
| "learning_rate": 4.5744278418763e-05, | |
| "loss": 2.0906640625, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 8.700586343862303, | |
| "learning_rate": 4.564970682806885e-05, | |
| "loss": 2.10053125, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 8.889729525250615, | |
| "learning_rate": 4.55551352373747e-05, | |
| "loss": 2.112328125, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 9.078872706638926, | |
| "learning_rate": 4.546056364668054e-05, | |
| "loss": 2.08459375, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 9.268015888027236, | |
| "learning_rate": 4.536599205598638e-05, | |
| "loss": 2.039546875, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 9.457159069415548, | |
| "learning_rate": 4.527142046529223e-05, | |
| "loss": 2.06175, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 9.646302250803858, | |
| "learning_rate": 4.517684887459807e-05, | |
| "loss": 2.076625, | |
| "step": 51000 | |
| }, | |
| { | |
| "epoch": 9.83544543219217, | |
| "learning_rate": 4.508227728390392e-05, | |
| "loss": 2.088609375, | |
| "step": 52000 | |
| }, | |
| { | |
| "epoch": 10.02458861358048, | |
| "learning_rate": 4.498770569320976e-05, | |
| "loss": 2.07996875, | |
| "step": 53000 | |
| }, | |
| { | |
| "epoch": 10.213731794968792, | |
| "learning_rate": 4.48931341025156e-05, | |
| "loss": 2.0183828125, | |
| "step": 54000 | |
| }, | |
| { | |
| "epoch": 10.402874976357102, | |
| "learning_rate": 4.479856251182145e-05, | |
| "loss": 2.0310546875, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 10.592018157745413, | |
| "learning_rate": 4.47039909211273e-05, | |
| "loss": 2.0436796875, | |
| "step": 56000 | |
| }, | |
| { | |
| "epoch": 10.781161339133725, | |
| "learning_rate": 4.460941933043314e-05, | |
| "loss": 2.0622109375, | |
| "step": 57000 | |
| }, | |
| { | |
| "epoch": 10.970304520522035, | |
| "learning_rate": 4.4514847739738984e-05, | |
| "loss": 2.06621875, | |
| "step": 58000 | |
| }, | |
| { | |
| "epoch": 11.159447701910347, | |
| "learning_rate": 4.442027614904483e-05, | |
| "loss": 2.0049921875, | |
| "step": 59000 | |
| }, | |
| { | |
| "epoch": 11.348590883298657, | |
| "learning_rate": 4.4325704558350674e-05, | |
| "loss": 2.00809375, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 11.537734064686967, | |
| "learning_rate": 4.423113296765652e-05, | |
| "loss": 2.017078125, | |
| "step": 61000 | |
| }, | |
| { | |
| "epoch": 11.72687724607528, | |
| "learning_rate": 4.4136561376962364e-05, | |
| "loss": 2.03634375, | |
| "step": 62000 | |
| }, | |
| { | |
| "epoch": 11.91602042746359, | |
| "learning_rate": 4.4041989786268206e-05, | |
| "loss": 2.04240625, | |
| "step": 63000 | |
| }, | |
| { | |
| "epoch": 12.105163608851901, | |
| "learning_rate": 4.3947418195574054e-05, | |
| "loss": 1.99821875, | |
| "step": 64000 | |
| }, | |
| { | |
| "epoch": 12.294306790240212, | |
| "learning_rate": 4.3852846604879896e-05, | |
| "loss": 1.9760625, | |
| "step": 65000 | |
| }, | |
| { | |
| "epoch": 12.483449971628524, | |
| "learning_rate": 4.375827501418574e-05, | |
| "loss": 1.995484375, | |
| "step": 66000 | |
| }, | |
| { | |
| "epoch": 12.672593153016834, | |
| "learning_rate": 4.3663703423491586e-05, | |
| "loss": 2.00321875, | |
| "step": 67000 | |
| }, | |
| { | |
| "epoch": 12.861736334405144, | |
| "learning_rate": 4.356913183279743e-05, | |
| "loss": 2.021078125, | |
| "step": 68000 | |
| }, | |
| { | |
| "epoch": 13.050879515793456, | |
| "learning_rate": 4.3474560242103276e-05, | |
| "loss": 2.004171875, | |
| "step": 69000 | |
| }, | |
| { | |
| "epoch": 13.240022697181766, | |
| "learning_rate": 4.337998865140912e-05, | |
| "loss": 1.9488125, | |
| "step": 70000 | |
| }, | |
| { | |
| "epoch": 13.429165878570078, | |
| "learning_rate": 4.328541706071496e-05, | |
| "loss": 1.967921875, | |
| "step": 71000 | |
| }, | |
| { | |
| "epoch": 13.618309059958388, | |
| "learning_rate": 4.319084547002081e-05, | |
| "loss": 1.984046875, | |
| "step": 72000 | |
| }, | |
| { | |
| "epoch": 13.807452241346699, | |
| "learning_rate": 4.3096273879326657e-05, | |
| "loss": 1.998828125, | |
| "step": 73000 | |
| }, | |
| { | |
| "epoch": 13.99659542273501, | |
| "learning_rate": 4.30017022886325e-05, | |
| "loss": 2.0024375, | |
| "step": 74000 | |
| }, | |
| { | |
| "epoch": 14.18573860412332, | |
| "learning_rate": 4.290713069793834e-05, | |
| "loss": 1.927890625, | |
| "step": 75000 | |
| }, | |
| { | |
| "epoch": 14.374881785511633, | |
| "learning_rate": 4.281255910724419e-05, | |
| "loss": 1.94315625, | |
| "step": 76000 | |
| }, | |
| { | |
| "epoch": 14.564024966899943, | |
| "learning_rate": 4.271798751655003e-05, | |
| "loss": 1.95571875, | |
| "step": 77000 | |
| }, | |
| { | |
| "epoch": 14.753168148288253, | |
| "learning_rate": 4.262341592585587e-05, | |
| "loss": 1.96771875, | |
| "step": 78000 | |
| }, | |
| { | |
| "epoch": 14.942311329676565, | |
| "learning_rate": 4.252884433516172e-05, | |
| "loss": 1.981109375, | |
| "step": 79000 | |
| }, | |
| { | |
| "epoch": 15.131454511064875, | |
| "learning_rate": 4.243427274446756e-05, | |
| "loss": 1.929875, | |
| "step": 80000 | |
| }, | |
| { | |
| "epoch": 15.320597692453187, | |
| "learning_rate": 4.233970115377341e-05, | |
| "loss": 1.91628125, | |
| "step": 81000 | |
| }, | |
| { | |
| "epoch": 15.509740873841498, | |
| "learning_rate": 4.224512956307925e-05, | |
| "loss": 1.938640625, | |
| "step": 82000 | |
| }, | |
| { | |
| "epoch": 15.69888405522981, | |
| "learning_rate": 4.2150557972385094e-05, | |
| "loss": 1.9478125, | |
| "step": 83000 | |
| }, | |
| { | |
| "epoch": 15.88802723661812, | |
| "learning_rate": 4.205598638169094e-05, | |
| "loss": 1.95146875, | |
| "step": 84000 | |
| }, | |
| { | |
| "epoch": 16.077170418006432, | |
| "learning_rate": 4.196141479099679e-05, | |
| "loss": 1.928953125, | |
| "step": 85000 | |
| }, | |
| { | |
| "epoch": 16.266313599394742, | |
| "learning_rate": 4.186684320030263e-05, | |
| "loss": 1.895140625, | |
| "step": 86000 | |
| }, | |
| { | |
| "epoch": 16.455456780783052, | |
| "learning_rate": 4.1772271609608474e-05, | |
| "loss": 1.909171875, | |
| "step": 87000 | |
| }, | |
| { | |
| "epoch": 16.644599962171363, | |
| "learning_rate": 4.1677700018914316e-05, | |
| "loss": 1.92403125, | |
| "step": 88000 | |
| }, | |
| { | |
| "epoch": 16.833743143559676, | |
| "learning_rate": 4.1583128428220164e-05, | |
| "loss": 1.935015625, | |
| "step": 89000 | |
| }, | |
| { | |
| "epoch": 17.022886324947986, | |
| "learning_rate": 4.148855683752601e-05, | |
| "loss": 1.92928125, | |
| "step": 90000 | |
| }, | |
| { | |
| "epoch": 17.212029506336297, | |
| "learning_rate": 4.139398524683185e-05, | |
| "loss": 1.86678125, | |
| "step": 91000 | |
| }, | |
| { | |
| "epoch": 17.401172687724607, | |
| "learning_rate": 4.1299413656137696e-05, | |
| "loss": 1.8859375, | |
| "step": 92000 | |
| }, | |
| { | |
| "epoch": 17.590315869112917, | |
| "learning_rate": 4.1204842065443545e-05, | |
| "loss": 1.90540625, | |
| "step": 93000 | |
| }, | |
| { | |
| "epoch": 17.77945905050123, | |
| "learning_rate": 4.1110270474749386e-05, | |
| "loss": 1.911765625, | |
| "step": 94000 | |
| }, | |
| { | |
| "epoch": 17.96860223188954, | |
| "learning_rate": 4.101569888405523e-05, | |
| "loss": 1.924671875, | |
| "step": 95000 | |
| }, | |
| { | |
| "epoch": 18.15774541327785, | |
| "learning_rate": 4.092112729336108e-05, | |
| "loss": 1.85646875, | |
| "step": 96000 | |
| }, | |
| { | |
| "epoch": 18.34688859466616, | |
| "learning_rate": 4.082655570266692e-05, | |
| "loss": 1.862640625, | |
| "step": 97000 | |
| }, | |
| { | |
| "epoch": 18.53603177605447, | |
| "learning_rate": 4.073198411197277e-05, | |
| "loss": 1.880921875, | |
| "step": 98000 | |
| }, | |
| { | |
| "epoch": 18.725174957442785, | |
| "learning_rate": 4.063741252127861e-05, | |
| "loss": 1.8903125, | |
| "step": 99000 | |
| }, | |
| { | |
| "epoch": 18.914318138831096, | |
| "learning_rate": 4.054284093058445e-05, | |
| "loss": 1.904015625, | |
| "step": 100000 | |
| }, | |
| { | |
| "epoch": 19.103461320219406, | |
| "learning_rate": 4.04482693398903e-05, | |
| "loss": 1.8575625, | |
| "step": 101000 | |
| }, | |
| { | |
| "epoch": 19.292604501607716, | |
| "learning_rate": 4.035369774919615e-05, | |
| "loss": 1.837125, | |
| "step": 102000 | |
| }, | |
| { | |
| "epoch": 19.481747682996026, | |
| "learning_rate": 4.025912615850199e-05, | |
| "loss": 1.859359375, | |
| "step": 103000 | |
| }, | |
| { | |
| "epoch": 19.67089086438434, | |
| "learning_rate": 4.016455456780783e-05, | |
| "loss": 1.867921875, | |
| "step": 104000 | |
| }, | |
| { | |
| "epoch": 19.86003404577265, | |
| "learning_rate": 4.006998297711368e-05, | |
| "loss": 1.88125, | |
| "step": 105000 | |
| }, | |
| { | |
| "epoch": 20.04917722716096, | |
| "learning_rate": 3.997541138641952e-05, | |
| "loss": 1.866546875, | |
| "step": 106000 | |
| }, | |
| { | |
| "epoch": 20.23832040854927, | |
| "learning_rate": 3.988083979572537e-05, | |
| "loss": 1.811046875, | |
| "step": 107000 | |
| }, | |
| { | |
| "epoch": 20.427463589937584, | |
| "learning_rate": 3.978626820503121e-05, | |
| "loss": 1.8318125, | |
| "step": 108000 | |
| }, | |
| { | |
| "epoch": 20.616606771325895, | |
| "learning_rate": 3.969169661433705e-05, | |
| "loss": 1.83990625, | |
| "step": 109000 | |
| }, | |
| { | |
| "epoch": 20.805749952714205, | |
| "learning_rate": 3.95971250236429e-05, | |
| "loss": 1.91721875, | |
| "step": 110000 | |
| }, | |
| { | |
| "epoch": 20.994893134102515, | |
| "learning_rate": 3.950255343294875e-05, | |
| "loss": 1.923984375, | |
| "step": 111000 | |
| }, | |
| { | |
| "epoch": 21.184036315490825, | |
| "learning_rate": 3.9407981842254585e-05, | |
| "loss": 1.846875, | |
| "step": 112000 | |
| }, | |
| { | |
| "epoch": 21.37317949687914, | |
| "learning_rate": 3.931341025156043e-05, | |
| "loss": 1.860984375, | |
| "step": 113000 | |
| }, | |
| { | |
| "epoch": 21.56232267826745, | |
| "learning_rate": 3.9218838660866275e-05, | |
| "loss": 1.8764375, | |
| "step": 114000 | |
| }, | |
| { | |
| "epoch": 21.75146585965576, | |
| "learning_rate": 3.912426707017212e-05, | |
| "loss": 1.88171875, | |
| "step": 115000 | |
| }, | |
| { | |
| "epoch": 21.94060904104407, | |
| "learning_rate": 3.9029695479477965e-05, | |
| "loss": 1.89409375, | |
| "step": 116000 | |
| }, | |
| { | |
| "epoch": 22.12975222243238, | |
| "learning_rate": 3.8935123888783807e-05, | |
| "loss": 1.847390625, | |
| "step": 117000 | |
| }, | |
| { | |
| "epoch": 22.318895403820694, | |
| "learning_rate": 3.8840552298089655e-05, | |
| "loss": 1.83696875, | |
| "step": 118000 | |
| }, | |
| { | |
| "epoch": 22.508038585209004, | |
| "learning_rate": 3.8745980707395504e-05, | |
| "loss": 1.852484375, | |
| "step": 119000 | |
| }, | |
| { | |
| "epoch": 22.697181766597314, | |
| "learning_rate": 3.8651409116701345e-05, | |
| "loss": 1.86578125, | |
| "step": 120000 | |
| }, | |
| { | |
| "epoch": 22.886324947985624, | |
| "learning_rate": 3.855683752600719e-05, | |
| "loss": 1.877734375, | |
| "step": 121000 | |
| }, | |
| { | |
| "epoch": 23.075468129373935, | |
| "learning_rate": 3.8462265935313035e-05, | |
| "loss": 1.8471875, | |
| "step": 122000 | |
| }, | |
| { | |
| "epoch": 23.26461131076225, | |
| "learning_rate": 3.836769434461888e-05, | |
| "loss": 1.816078125, | |
| "step": 123000 | |
| }, | |
| { | |
| "epoch": 23.45375449215056, | |
| "learning_rate": 3.8273122753924726e-05, | |
| "loss": 1.833703125, | |
| "step": 124000 | |
| }, | |
| { | |
| "epoch": 23.64289767353887, | |
| "learning_rate": 3.817855116323057e-05, | |
| "loss": 1.840984375, | |
| "step": 125000 | |
| }, | |
| { | |
| "epoch": 23.83204085492718, | |
| "learning_rate": 3.808397957253641e-05, | |
| "loss": 1.8568125, | |
| "step": 126000 | |
| }, | |
| { | |
| "epoch": 24.02118403631549, | |
| "learning_rate": 3.798940798184226e-05, | |
| "loss": 1.855609375, | |
| "step": 127000 | |
| }, | |
| { | |
| "epoch": 24.210327217703803, | |
| "learning_rate": 3.7894836391148106e-05, | |
| "loss": 1.7931875, | |
| "step": 128000 | |
| }, | |
| { | |
| "epoch": 24.399470399092113, | |
| "learning_rate": 3.780026480045394e-05, | |
| "loss": 1.805828125, | |
| "step": 129000 | |
| }, | |
| { | |
| "epoch": 24.588613580480423, | |
| "learning_rate": 3.770569320975979e-05, | |
| "loss": 1.825140625, | |
| "step": 130000 | |
| }, | |
| { | |
| "epoch": 24.777756761868734, | |
| "learning_rate": 3.761112161906564e-05, | |
| "loss": 1.83425, | |
| "step": 131000 | |
| }, | |
| { | |
| "epoch": 24.966899943257047, | |
| "learning_rate": 3.751655002837148e-05, | |
| "loss": 1.8449375, | |
| "step": 132000 | |
| }, | |
| { | |
| "epoch": 25.156043124645358, | |
| "learning_rate": 3.742197843767732e-05, | |
| "loss": 1.78215625, | |
| "step": 133000 | |
| }, | |
| { | |
| "epoch": 25.345186306033668, | |
| "learning_rate": 3.732740684698317e-05, | |
| "loss": 1.79159375, | |
| "step": 134000 | |
| }, | |
| { | |
| "epoch": 25.534329487421978, | |
| "learning_rate": 3.723283525628901e-05, | |
| "loss": 1.80753125, | |
| "step": 135000 | |
| }, | |
| { | |
| "epoch": 25.723472668810288, | |
| "learning_rate": 3.713826366559486e-05, | |
| "loss": 1.81053125, | |
| "step": 136000 | |
| }, | |
| { | |
| "epoch": 25.912615850198602, | |
| "learning_rate": 3.70436920749007e-05, | |
| "loss": 1.8265, | |
| "step": 137000 | |
| }, | |
| { | |
| "epoch": 26.101759031586912, | |
| "learning_rate": 3.694912048420654e-05, | |
| "loss": 1.79653125, | |
| "step": 138000 | |
| }, | |
| { | |
| "epoch": 26.290902212975222, | |
| "learning_rate": 3.685454889351239e-05, | |
| "loss": 1.76934375, | |
| "step": 139000 | |
| }, | |
| { | |
| "epoch": 26.480045394363533, | |
| "learning_rate": 3.6759977302818233e-05, | |
| "loss": 1.78228125, | |
| "step": 140000 | |
| }, | |
| { | |
| "epoch": 26.669188575751843, | |
| "learning_rate": 3.666540571212408e-05, | |
| "loss": 1.79384375, | |
| "step": 141000 | |
| }, | |
| { | |
| "epoch": 26.858331757140157, | |
| "learning_rate": 3.6570834121429924e-05, | |
| "loss": 1.8095625, | |
| "step": 142000 | |
| }, | |
| { | |
| "epoch": 27.047474938528467, | |
| "learning_rate": 3.6476262530735765e-05, | |
| "loss": 1.7923125, | |
| "step": 143000 | |
| }, | |
| { | |
| "epoch": 27.236618119916777, | |
| "learning_rate": 3.6381690940041614e-05, | |
| "loss": 1.7471875, | |
| "step": 144000 | |
| }, | |
| { | |
| "epoch": 27.425761301305087, | |
| "learning_rate": 3.628711934934746e-05, | |
| "loss": 1.76446875, | |
| "step": 145000 | |
| }, | |
| { | |
| "epoch": 27.614904482693397, | |
| "learning_rate": 3.61925477586533e-05, | |
| "loss": 1.77759375, | |
| "step": 146000 | |
| }, | |
| { | |
| "epoch": 27.80404766408171, | |
| "learning_rate": 3.6097976167959146e-05, | |
| "loss": 1.78840625, | |
| "step": 147000 | |
| }, | |
| { | |
| "epoch": 27.99319084547002, | |
| "learning_rate": 3.6003404577264994e-05, | |
| "loss": 1.799125, | |
| "step": 148000 | |
| }, | |
| { | |
| "epoch": 28.18233402685833, | |
| "learning_rate": 3.5908832986570836e-05, | |
| "loss": 1.73171875, | |
| "step": 149000 | |
| }, | |
| { | |
| "epoch": 28.37147720824664, | |
| "learning_rate": 3.581426139587668e-05, | |
| "loss": 1.73996875, | |
| "step": 150000 | |
| }, | |
| { | |
| "epoch": 28.560620389634952, | |
| "learning_rate": 3.5719689805182526e-05, | |
| "loss": 1.75671875, | |
| "step": 151000 | |
| }, | |
| { | |
| "epoch": 28.749763571023266, | |
| "learning_rate": 3.562511821448837e-05, | |
| "loss": 1.7723125, | |
| "step": 152000 | |
| }, | |
| { | |
| "epoch": 28.938906752411576, | |
| "learning_rate": 3.5530546623794216e-05, | |
| "loss": 1.77990625, | |
| "step": 153000 | |
| }, | |
| { | |
| "epoch": 29.128049933799886, | |
| "learning_rate": 3.543597503310006e-05, | |
| "loss": 1.73996875, | |
| "step": 154000 | |
| }, | |
| { | |
| "epoch": 29.317193115188196, | |
| "learning_rate": 3.53414034424059e-05, | |
| "loss": 1.726625, | |
| "step": 155000 | |
| }, | |
| { | |
| "epoch": 29.50633629657651, | |
| "learning_rate": 3.524683185171175e-05, | |
| "loss": 1.74075, | |
| "step": 156000 | |
| }, | |
| { | |
| "epoch": 29.69547947796482, | |
| "learning_rate": 3.5152260261017597e-05, | |
| "loss": 1.753, | |
| "step": 157000 | |
| }, | |
| { | |
| "epoch": 29.88462265935313, | |
| "learning_rate": 3.505768867032344e-05, | |
| "loss": 1.7575, | |
| "step": 158000 | |
| }, | |
| { | |
| "epoch": 30.07376584074144, | |
| "learning_rate": 3.496311707962928e-05, | |
| "loss": 1.73771875, | |
| "step": 159000 | |
| }, | |
| { | |
| "epoch": 30.26290902212975, | |
| "learning_rate": 3.486854548893513e-05, | |
| "loss": 1.70875, | |
| "step": 160000 | |
| }, | |
| { | |
| "epoch": 30.452052203518065, | |
| "learning_rate": 3.477397389824097e-05, | |
| "loss": 1.72415625, | |
| "step": 161000 | |
| }, | |
| { | |
| "epoch": 30.641195384906375, | |
| "learning_rate": 3.467940230754682e-05, | |
| "loss": 1.7310625, | |
| "step": 162000 | |
| }, | |
| { | |
| "epoch": 30.830338566294685, | |
| "learning_rate": 3.4584830716852654e-05, | |
| "loss": 1.74175, | |
| "step": 163000 | |
| }, | |
| { | |
| "epoch": 31.019481747682995, | |
| "learning_rate": 3.44902591261585e-05, | |
| "loss": 1.74609375, | |
| "step": 164000 | |
| }, | |
| { | |
| "epoch": 31.208624929071306, | |
| "learning_rate": 3.439568753546435e-05, | |
| "loss": 1.684875, | |
| "step": 165000 | |
| }, | |
| { | |
| "epoch": 31.39776811045962, | |
| "learning_rate": 3.430111594477019e-05, | |
| "loss": 1.69925, | |
| "step": 166000 | |
| }, | |
| { | |
| "epoch": 31.58691129184793, | |
| "learning_rate": 3.4206544354076034e-05, | |
| "loss": 1.7169375, | |
| "step": 167000 | |
| }, | |
| { | |
| "epoch": 31.77605447323624, | |
| "learning_rate": 3.411197276338188e-05, | |
| "loss": 1.7275, | |
| "step": 168000 | |
| }, | |
| { | |
| "epoch": 31.96519765462455, | |
| "learning_rate": 3.4017401172687724e-05, | |
| "loss": 1.74096875, | |
| "step": 169000 | |
| }, | |
| { | |
| "epoch": 32.154340836012864, | |
| "learning_rate": 3.392282958199357e-05, | |
| "loss": 1.68128125, | |
| "step": 170000 | |
| }, | |
| { | |
| "epoch": 32.343484017401174, | |
| "learning_rate": 3.3828257991299414e-05, | |
| "loss": 1.6851875, | |
| "step": 171000 | |
| }, | |
| { | |
| "epoch": 32.532627198789484, | |
| "learning_rate": 3.3733686400605256e-05, | |
| "loss": 1.69896875, | |
| "step": 172000 | |
| }, | |
| { | |
| "epoch": 32.721770380177794, | |
| "learning_rate": 3.3639114809911104e-05, | |
| "loss": 1.7080625, | |
| "step": 173000 | |
| }, | |
| { | |
| "epoch": 32.910913561566105, | |
| "learning_rate": 3.354454321921695e-05, | |
| "loss": 1.7200625, | |
| "step": 174000 | |
| }, | |
| { | |
| "epoch": 33.100056742954415, | |
| "learning_rate": 3.3449971628522795e-05, | |
| "loss": 1.69040625, | |
| "step": 175000 | |
| }, | |
| { | |
| "epoch": 33.289199924342725, | |
| "learning_rate": 3.3355400037828636e-05, | |
| "loss": 1.666625, | |
| "step": 176000 | |
| }, | |
| { | |
| "epoch": 33.478343105731035, | |
| "learning_rate": 3.3260828447134485e-05, | |
| "loss": 1.6798125, | |
| "step": 177000 | |
| }, | |
| { | |
| "epoch": 33.66748628711935, | |
| "learning_rate": 3.3166256856440326e-05, | |
| "loss": 1.69575, | |
| "step": 178000 | |
| }, | |
| { | |
| "epoch": 33.85662946850766, | |
| "learning_rate": 3.3071685265746175e-05, | |
| "loss": 1.7008125, | |
| "step": 179000 | |
| }, | |
| { | |
| "epoch": 34.04577264989597, | |
| "learning_rate": 3.297711367505202e-05, | |
| "loss": 1.69340625, | |
| "step": 180000 | |
| }, | |
| { | |
| "epoch": 34.23491583128428, | |
| "learning_rate": 3.288254208435786e-05, | |
| "loss": 1.6509375, | |
| "step": 181000 | |
| }, | |
| { | |
| "epoch": 34.42405901267259, | |
| "learning_rate": 3.278797049366371e-05, | |
| "loss": 1.66703125, | |
| "step": 182000 | |
| }, | |
| { | |
| "epoch": 34.613202194060904, | |
| "learning_rate": 3.2693398902969555e-05, | |
| "loss": 1.67615625, | |
| "step": 183000 | |
| }, | |
| { | |
| "epoch": 34.802345375449214, | |
| "learning_rate": 3.259882731227539e-05, | |
| "loss": 1.6876875, | |
| "step": 184000 | |
| }, | |
| { | |
| "epoch": 34.991488556837524, | |
| "learning_rate": 3.250425572158124e-05, | |
| "loss": 1.695, | |
| "step": 185000 | |
| }, | |
| { | |
| "epoch": 35.180631738225834, | |
| "learning_rate": 3.240968413088709e-05, | |
| "loss": 1.6354375, | |
| "step": 186000 | |
| }, | |
| { | |
| "epoch": 35.369774919614144, | |
| "learning_rate": 3.231511254019293e-05, | |
| "loss": 1.64653125, | |
| "step": 187000 | |
| }, | |
| { | |
| "epoch": 35.55891810100246, | |
| "learning_rate": 3.222054094949877e-05, | |
| "loss": 1.66346875, | |
| "step": 188000 | |
| }, | |
| { | |
| "epoch": 35.74806128239077, | |
| "learning_rate": 3.212596935880461e-05, | |
| "loss": 1.66990625, | |
| "step": 189000 | |
| }, | |
| { | |
| "epoch": 35.93720446377908, | |
| "learning_rate": 3.203139776811046e-05, | |
| "loss": 1.67878125, | |
| "step": 190000 | |
| }, | |
| { | |
| "epoch": 36.12634764516739, | |
| "learning_rate": 3.193682617741631e-05, | |
| "loss": 1.63953125, | |
| "step": 191000 | |
| }, | |
| { | |
| "epoch": 36.3154908265557, | |
| "learning_rate": 3.184225458672215e-05, | |
| "loss": 1.62646875, | |
| "step": 192000 | |
| }, | |
| { | |
| "epoch": 36.50463400794401, | |
| "learning_rate": 3.174768299602799e-05, | |
| "loss": 1.6480625, | |
| "step": 193000 | |
| }, | |
| { | |
| "epoch": 36.69377718933232, | |
| "learning_rate": 3.165311140533384e-05, | |
| "loss": 1.65371875, | |
| "step": 194000 | |
| }, | |
| { | |
| "epoch": 36.88292037072063, | |
| "learning_rate": 3.155853981463968e-05, | |
| "loss": 1.66778125, | |
| "step": 195000 | |
| }, | |
| { | |
| "epoch": 37.07206355210894, | |
| "learning_rate": 3.146396822394553e-05, | |
| "loss": 1.64275, | |
| "step": 196000 | |
| }, | |
| { | |
| "epoch": 37.26120673349726, | |
| "learning_rate": 3.136939663325137e-05, | |
| "loss": 1.61565625, | |
| "step": 197000 | |
| }, | |
| { | |
| "epoch": 37.45034991488557, | |
| "learning_rate": 3.1274825042557215e-05, | |
| "loss": 1.631875, | |
| "step": 198000 | |
| }, | |
| { | |
| "epoch": 37.63949309627388, | |
| "learning_rate": 3.118025345186306e-05, | |
| "loss": 1.64078125, | |
| "step": 199000 | |
| }, | |
| { | |
| "epoch": 37.82863627766219, | |
| "learning_rate": 3.108568186116891e-05, | |
| "loss": 1.6473125, | |
| "step": 200000 | |
| }, | |
| { | |
| "epoch": 38.0177794590505, | |
| "learning_rate": 3.0991110270474747e-05, | |
| "loss": 1.65090625, | |
| "step": 201000 | |
| }, | |
| { | |
| "epoch": 38.20692264043881, | |
| "learning_rate": 3.0896538679780595e-05, | |
| "loss": 1.59846875, | |
| "step": 202000 | |
| }, | |
| { | |
| "epoch": 38.39606582182712, | |
| "learning_rate": 3.0801967089086443e-05, | |
| "loss": 1.6114375, | |
| "step": 203000 | |
| }, | |
| { | |
| "epoch": 38.58520900321543, | |
| "learning_rate": 3.0707395498392285e-05, | |
| "loss": 1.6266875, | |
| "step": 204000 | |
| }, | |
| { | |
| "epoch": 38.77435218460374, | |
| "learning_rate": 3.061282390769813e-05, | |
| "loss": 1.6340625, | |
| "step": 205000 | |
| }, | |
| { | |
| "epoch": 38.96349536599205, | |
| "learning_rate": 3.0518252317003975e-05, | |
| "loss": 1.64478125, | |
| "step": 206000 | |
| }, | |
| { | |
| "epoch": 39.15263854738037, | |
| "learning_rate": 3.0423680726309817e-05, | |
| "loss": 1.59403125, | |
| "step": 207000 | |
| }, | |
| { | |
| "epoch": 39.34178172876868, | |
| "learning_rate": 3.0329109135615662e-05, | |
| "loss": 1.5961875, | |
| "step": 208000 | |
| }, | |
| { | |
| "epoch": 39.53092491015699, | |
| "learning_rate": 3.023453754492151e-05, | |
| "loss": 1.612375, | |
| "step": 209000 | |
| }, | |
| { | |
| "epoch": 39.7200680915453, | |
| "learning_rate": 3.013996595422735e-05, | |
| "loss": 1.62103125, | |
| "step": 210000 | |
| }, | |
| { | |
| "epoch": 39.90921127293361, | |
| "learning_rate": 3.0045394363533197e-05, | |
| "loss": 1.62684375, | |
| "step": 211000 | |
| }, | |
| { | |
| "epoch": 40.09835445432192, | |
| "learning_rate": 2.995082277283904e-05, | |
| "loss": 1.60059375, | |
| "step": 212000 | |
| }, | |
| { | |
| "epoch": 40.28749763571023, | |
| "learning_rate": 2.9856251182144884e-05, | |
| "loss": 1.58271875, | |
| "step": 213000 | |
| }, | |
| { | |
| "epoch": 40.47664081709854, | |
| "learning_rate": 2.976167959145073e-05, | |
| "loss": 1.59703125, | |
| "step": 214000 | |
| }, | |
| { | |
| "epoch": 40.66578399848685, | |
| "learning_rate": 2.966710800075657e-05, | |
| "loss": 1.604875, | |
| "step": 215000 | |
| }, | |
| { | |
| "epoch": 40.85492717987517, | |
| "learning_rate": 2.957253641006242e-05, | |
| "loss": 1.6115625, | |
| "step": 216000 | |
| }, | |
| { | |
| "epoch": 41.04407036126348, | |
| "learning_rate": 2.9477964819368265e-05, | |
| "loss": 1.60434375, | |
| "step": 217000 | |
| }, | |
| { | |
| "epoch": 41.23321354265179, | |
| "learning_rate": 2.9383393228674106e-05, | |
| "loss": 1.566625, | |
| "step": 218000 | |
| }, | |
| { | |
| "epoch": 41.4223567240401, | |
| "learning_rate": 2.928882163797995e-05, | |
| "loss": 1.58084375, | |
| "step": 219000 | |
| }, | |
| { | |
| "epoch": 41.61149990542841, | |
| "learning_rate": 2.91942500472858e-05, | |
| "loss": 1.590625, | |
| "step": 220000 | |
| }, | |
| { | |
| "epoch": 41.80064308681672, | |
| "learning_rate": 2.9099678456591638e-05, | |
| "loss": 1.60096875, | |
| "step": 221000 | |
| }, | |
| { | |
| "epoch": 41.98978626820503, | |
| "learning_rate": 2.9005106865897487e-05, | |
| "loss": 1.6100625, | |
| "step": 222000 | |
| }, | |
| { | |
| "epoch": 42.17892944959334, | |
| "learning_rate": 2.8910535275203332e-05, | |
| "loss": 1.552625, | |
| "step": 223000 | |
| }, | |
| { | |
| "epoch": 42.36807263098165, | |
| "learning_rate": 2.8815963684509173e-05, | |
| "loss": 1.56346875, | |
| "step": 224000 | |
| }, | |
| { | |
| "epoch": 42.55721581236996, | |
| "learning_rate": 2.872139209381502e-05, | |
| "loss": 1.57903125, | |
| "step": 225000 | |
| }, | |
| { | |
| "epoch": 42.74635899375828, | |
| "learning_rate": 2.8626820503120867e-05, | |
| "loss": 1.5851875, | |
| "step": 226000 | |
| }, | |
| { | |
| "epoch": 42.93550217514659, | |
| "learning_rate": 2.8532248912426705e-05, | |
| "loss": 1.59359375, | |
| "step": 227000 | |
| }, | |
| { | |
| "epoch": 43.1246453565349, | |
| "learning_rate": 2.8437677321732554e-05, | |
| "loss": 1.55915625, | |
| "step": 228000 | |
| }, | |
| { | |
| "epoch": 43.31378853792321, | |
| "learning_rate": 2.83431057310384e-05, | |
| "loss": 1.5494375, | |
| "step": 229000 | |
| }, | |
| { | |
| "epoch": 43.50293171931152, | |
| "learning_rate": 2.824853414034424e-05, | |
| "loss": 1.55978125, | |
| "step": 230000 | |
| }, | |
| { | |
| "epoch": 43.69207490069983, | |
| "learning_rate": 2.8153962549650086e-05, | |
| "loss": 1.57346875, | |
| "step": 231000 | |
| }, | |
| { | |
| "epoch": 43.88121808208814, | |
| "learning_rate": 2.8059390958955934e-05, | |
| "loss": 1.58403125, | |
| "step": 232000 | |
| }, | |
| { | |
| "epoch": 44.07036126347645, | |
| "learning_rate": 2.7964819368261776e-05, | |
| "loss": 1.568125, | |
| "step": 233000 | |
| }, | |
| { | |
| "epoch": 44.25950444486476, | |
| "learning_rate": 2.787024777756762e-05, | |
| "loss": 1.535125, | |
| "step": 234000 | |
| }, | |
| { | |
| "epoch": 44.44864762625307, | |
| "learning_rate": 2.7775676186873466e-05, | |
| "loss": 1.5455, | |
| "step": 235000 | |
| }, | |
| { | |
| "epoch": 44.63779080764139, | |
| "learning_rate": 2.7681104596179308e-05, | |
| "loss": 1.55828125, | |
| "step": 236000 | |
| }, | |
| { | |
| "epoch": 44.8269339890297, | |
| "learning_rate": 2.7586533005485156e-05, | |
| "loss": 1.56821875, | |
| "step": 237000 | |
| }, | |
| { | |
| "epoch": 45.01607717041801, | |
| "learning_rate": 2.7491961414790994e-05, | |
| "loss": 1.573875, | |
| "step": 238000 | |
| }, | |
| { | |
| "epoch": 45.20522035180632, | |
| "learning_rate": 2.7397389824096843e-05, | |
| "loss": 1.52146875, | |
| "step": 239000 | |
| }, | |
| { | |
| "epoch": 45.39436353319463, | |
| "learning_rate": 2.7302818233402688e-05, | |
| "loss": 1.53628125, | |
| "step": 240000 | |
| }, | |
| { | |
| "epoch": 45.58350671458294, | |
| "learning_rate": 2.720824664270853e-05, | |
| "loss": 1.549875, | |
| "step": 241000 | |
| }, | |
| { | |
| "epoch": 45.77264989597125, | |
| "learning_rate": 2.7113675052014375e-05, | |
| "loss": 1.55240625, | |
| "step": 242000 | |
| }, | |
| { | |
| "epoch": 45.96179307735956, | |
| "learning_rate": 2.7019103461320223e-05, | |
| "loss": 1.56171875, | |
| "step": 243000 | |
| }, | |
| { | |
| "epoch": 46.15093625874787, | |
| "learning_rate": 2.692453187062606e-05, | |
| "loss": 1.52228125, | |
| "step": 244000 | |
| }, | |
| { | |
| "epoch": 46.340079440136186, | |
| "learning_rate": 2.682996027993191e-05, | |
| "loss": 1.5250625, | |
| "step": 245000 | |
| }, | |
| { | |
| "epoch": 46.5292226215245, | |
| "learning_rate": 2.6735388689237755e-05, | |
| "loss": 1.5320625, | |
| "step": 246000 | |
| }, | |
| { | |
| "epoch": 46.71836580291281, | |
| "learning_rate": 2.6640817098543597e-05, | |
| "loss": 1.5440625, | |
| "step": 247000 | |
| }, | |
| { | |
| "epoch": 46.90750898430112, | |
| "learning_rate": 2.6546245507849442e-05, | |
| "loss": 1.5495, | |
| "step": 248000 | |
| }, | |
| { | |
| "epoch": 47.09665216568943, | |
| "learning_rate": 2.645167391715529e-05, | |
| "loss": 1.5265625, | |
| "step": 249000 | |
| }, | |
| { | |
| "epoch": 47.28579534707774, | |
| "learning_rate": 2.6357102326461132e-05, | |
| "loss": 1.50359375, | |
| "step": 250000 | |
| }, | |
| { | |
| "epoch": 47.47493852846605, | |
| "learning_rate": 2.6262530735766977e-05, | |
| "loss": 1.5239375, | |
| "step": 251000 | |
| }, | |
| { | |
| "epoch": 47.66408170985436, | |
| "learning_rate": 2.6167959145072822e-05, | |
| "loss": 1.5295625, | |
| "step": 252000 | |
| }, | |
| { | |
| "epoch": 47.85322489124267, | |
| "learning_rate": 2.6073387554378664e-05, | |
| "loss": 1.53828125, | |
| "step": 253000 | |
| }, | |
| { | |
| "epoch": 48.04236807263098, | |
| "learning_rate": 2.597881596368451e-05, | |
| "loss": 1.53, | |
| "step": 254000 | |
| }, | |
| { | |
| "epoch": 48.231511254019296, | |
| "learning_rate": 2.5884244372990358e-05, | |
| "loss": 1.496, | |
| "step": 255000 | |
| }, | |
| { | |
| "epoch": 48.420654435407606, | |
| "learning_rate": 2.57896727822962e-05, | |
| "loss": 1.503375, | |
| "step": 256000 | |
| }, | |
| { | |
| "epoch": 48.609797616795916, | |
| "learning_rate": 2.5695101191602044e-05, | |
| "loss": 1.521125, | |
| "step": 257000 | |
| }, | |
| { | |
| "epoch": 48.798940798184226, | |
| "learning_rate": 2.560052960090789e-05, | |
| "loss": 1.5279375, | |
| "step": 258000 | |
| }, | |
| { | |
| "epoch": 48.98808397957254, | |
| "learning_rate": 2.550595801021373e-05, | |
| "loss": 1.5339375, | |
| "step": 259000 | |
| }, | |
| { | |
| "epoch": 49.17722716096085, | |
| "learning_rate": 2.541138641951958e-05, | |
| "loss": 1.4859375, | |
| "step": 260000 | |
| }, | |
| { | |
| "epoch": 49.36637034234916, | |
| "learning_rate": 2.5316814828825425e-05, | |
| "loss": 1.4956875, | |
| "step": 261000 | |
| }, | |
| { | |
| "epoch": 49.55551352373747, | |
| "learning_rate": 2.5222243238131266e-05, | |
| "loss": 1.50375, | |
| "step": 262000 | |
| }, | |
| { | |
| "epoch": 49.74465670512578, | |
| "learning_rate": 2.512767164743711e-05, | |
| "loss": 1.51525, | |
| "step": 263000 | |
| }, | |
| { | |
| "epoch": 49.933799886514095, | |
| "learning_rate": 2.5033100056742953e-05, | |
| "loss": 1.52084375, | |
| "step": 264000 | |
| }, | |
| { | |
| "epoch": 50.122943067902405, | |
| "learning_rate": 2.4938528466048798e-05, | |
| "loss": 1.49025, | |
| "step": 265000 | |
| }, | |
| { | |
| "epoch": 50.312086249290715, | |
| "learning_rate": 2.4843956875354647e-05, | |
| "loss": 1.48290625, | |
| "step": 266000 | |
| }, | |
| { | |
| "epoch": 50.501229430679025, | |
| "learning_rate": 2.474938528466049e-05, | |
| "loss": 1.49121875, | |
| "step": 267000 | |
| }, | |
| { | |
| "epoch": 50.690372612067335, | |
| "learning_rate": 2.4654813693966334e-05, | |
| "loss": 1.502875, | |
| "step": 268000 | |
| }, | |
| { | |
| "epoch": 50.879515793455646, | |
| "learning_rate": 2.4560242103272175e-05, | |
| "loss": 1.51284375, | |
| "step": 269000 | |
| }, | |
| { | |
| "epoch": 51.068658974843956, | |
| "learning_rate": 2.4465670512578024e-05, | |
| "loss": 1.4948125, | |
| "step": 270000 | |
| }, | |
| { | |
| "epoch": 51.257802156232266, | |
| "learning_rate": 2.4371098921883865e-05, | |
| "loss": 1.4699375, | |
| "step": 271000 | |
| }, | |
| { | |
| "epoch": 51.446945337620576, | |
| "learning_rate": 2.427652733118971e-05, | |
| "loss": 1.48165625, | |
| "step": 272000 | |
| }, | |
| { | |
| "epoch": 51.63608851900889, | |
| "learning_rate": 2.4181955740495556e-05, | |
| "loss": 1.491875, | |
| "step": 273000 | |
| }, | |
| { | |
| "epoch": 51.825231700397204, | |
| "learning_rate": 2.40873841498014e-05, | |
| "loss": 1.498875, | |
| "step": 274000 | |
| }, | |
| { | |
| "epoch": 52.014374881785514, | |
| "learning_rate": 2.3992812559107246e-05, | |
| "loss": 1.50025, | |
| "step": 275000 | |
| }, | |
| { | |
| "epoch": 52.203518063173824, | |
| "learning_rate": 2.389824096841309e-05, | |
| "loss": 1.45865625, | |
| "step": 276000 | |
| }, | |
| { | |
| "epoch": 52.392661244562134, | |
| "learning_rate": 2.3803669377718936e-05, | |
| "loss": 1.4689375, | |
| "step": 277000 | |
| }, | |
| { | |
| "epoch": 52.581804425950445, | |
| "learning_rate": 2.3709097787024778e-05, | |
| "loss": 1.4756875, | |
| "step": 278000 | |
| }, | |
| { | |
| "epoch": 52.770947607338755, | |
| "learning_rate": 2.3614526196330626e-05, | |
| "loss": 1.48815625, | |
| "step": 279000 | |
| }, | |
| { | |
| "epoch": 52.960090788727065, | |
| "learning_rate": 2.3519954605636468e-05, | |
| "loss": 1.4954375, | |
| "step": 280000 | |
| }, | |
| { | |
| "epoch": 53.149233970115375, | |
| "learning_rate": 2.3425383014942313e-05, | |
| "loss": 1.4575625, | |
| "step": 281000 | |
| }, | |
| { | |
| "epoch": 53.338377151503686, | |
| "learning_rate": 2.3330811424248155e-05, | |
| "loss": 1.4586875, | |
| "step": 282000 | |
| }, | |
| { | |
| "epoch": 53.527520332891996, | |
| "learning_rate": 2.3236239833554003e-05, | |
| "loss": 1.46946875, | |
| "step": 283000 | |
| }, | |
| { | |
| "epoch": 53.71666351428031, | |
| "learning_rate": 2.3141668242859845e-05, | |
| "loss": 1.47496875, | |
| "step": 284000 | |
| }, | |
| { | |
| "epoch": 53.90580669566862, | |
| "learning_rate": 2.304709665216569e-05, | |
| "loss": 1.48246875, | |
| "step": 285000 | |
| }, | |
| { | |
| "epoch": 54.09494987705693, | |
| "learning_rate": 2.2952525061471535e-05, | |
| "loss": 1.458375, | |
| "step": 286000 | |
| }, | |
| { | |
| "epoch": 54.284093058445244, | |
| "learning_rate": 2.285795347077738e-05, | |
| "loss": 1.44609375, | |
| "step": 287000 | |
| }, | |
| { | |
| "epoch": 54.473236239833554, | |
| "learning_rate": 2.2763381880083222e-05, | |
| "loss": 1.45584375, | |
| "step": 288000 | |
| }, | |
| { | |
| "epoch": 54.662379421221864, | |
| "learning_rate": 2.266881028938907e-05, | |
| "loss": 1.46325, | |
| "step": 289000 | |
| }, | |
| { | |
| "epoch": 54.851522602610174, | |
| "learning_rate": 2.2574238698694912e-05, | |
| "loss": 1.475625, | |
| "step": 290000 | |
| }, | |
| { | |
| "epoch": 55.040665783998485, | |
| "learning_rate": 2.2479667108000757e-05, | |
| "loss": 1.46834375, | |
| "step": 291000 | |
| }, | |
| { | |
| "epoch": 55.229808965386795, | |
| "learning_rate": 2.2385095517306602e-05, | |
| "loss": 1.43265625, | |
| "step": 292000 | |
| }, | |
| { | |
| "epoch": 55.41895214677511, | |
| "learning_rate": 2.2290523926612447e-05, | |
| "loss": 1.4436875, | |
| "step": 293000 | |
| }, | |
| { | |
| "epoch": 55.60809532816342, | |
| "learning_rate": 2.2195952335918292e-05, | |
| "loss": 1.455, | |
| "step": 294000 | |
| }, | |
| { | |
| "epoch": 55.79723850955173, | |
| "learning_rate": 2.2101380745224134e-05, | |
| "loss": 1.46475, | |
| "step": 295000 | |
| }, | |
| { | |
| "epoch": 55.98638169094004, | |
| "learning_rate": 2.2006809154529982e-05, | |
| "loss": 1.46825, | |
| "step": 296000 | |
| }, | |
| { | |
| "epoch": 56.17552487232835, | |
| "learning_rate": 2.1912237563835824e-05, | |
| "loss": 1.4260625, | |
| "step": 297000 | |
| }, | |
| { | |
| "epoch": 56.36466805371666, | |
| "learning_rate": 2.181766597314167e-05, | |
| "loss": 1.43375, | |
| "step": 298000 | |
| }, | |
| { | |
| "epoch": 56.55381123510497, | |
| "learning_rate": 2.1723094382447514e-05, | |
| "loss": 1.444375, | |
| "step": 299000 | |
| }, | |
| { | |
| "epoch": 56.74295441649328, | |
| "learning_rate": 2.162852279175336e-05, | |
| "loss": 1.4503125, | |
| "step": 300000 | |
| }, | |
| { | |
| "epoch": 56.932097597881594, | |
| "learning_rate": 2.15339512010592e-05, | |
| "loss": 1.4611875, | |
| "step": 301000 | |
| }, | |
| { | |
| "epoch": 57.121240779269904, | |
| "learning_rate": 2.143937961036505e-05, | |
| "loss": 1.4291875, | |
| "step": 302000 | |
| }, | |
| { | |
| "epoch": 57.31038396065822, | |
| "learning_rate": 2.134480801967089e-05, | |
| "loss": 1.422, | |
| "step": 303000 | |
| }, | |
| { | |
| "epoch": 57.49952714204653, | |
| "learning_rate": 2.1250236428976736e-05, | |
| "loss": 1.434625, | |
| "step": 304000 | |
| }, | |
| { | |
| "epoch": 57.68867032343484, | |
| "learning_rate": 2.115566483828258e-05, | |
| "loss": 1.437625, | |
| "step": 305000 | |
| }, | |
| { | |
| "epoch": 57.87781350482315, | |
| "learning_rate": 2.1061093247588427e-05, | |
| "loss": 1.4495625, | |
| "step": 306000 | |
| }, | |
| { | |
| "epoch": 58.06695668621146, | |
| "learning_rate": 2.0966521656894268e-05, | |
| "loss": 1.440125, | |
| "step": 307000 | |
| }, | |
| { | |
| "epoch": 58.25609986759977, | |
| "learning_rate": 2.0871950066200113e-05, | |
| "loss": 1.413375, | |
| "step": 308000 | |
| }, | |
| { | |
| "epoch": 58.44524304898808, | |
| "learning_rate": 2.077737847550596e-05, | |
| "loss": 1.4220625, | |
| "step": 309000 | |
| }, | |
| { | |
| "epoch": 58.63438623037639, | |
| "learning_rate": 2.0682806884811804e-05, | |
| "loss": 1.4315, | |
| "step": 310000 | |
| }, | |
| { | |
| "epoch": 58.8235294117647, | |
| "learning_rate": 2.058823529411765e-05, | |
| "loss": 1.4355625, | |
| "step": 311000 | |
| }, | |
| { | |
| "epoch": 59.01267259315302, | |
| "learning_rate": 2.0493663703423494e-05, | |
| "loss": 1.440125, | |
| "step": 312000 | |
| }, | |
| { | |
| "epoch": 59.20181577454133, | |
| "learning_rate": 2.039909211272934e-05, | |
| "loss": 1.4, | |
| "step": 313000 | |
| }, | |
| { | |
| "epoch": 59.39095895592964, | |
| "learning_rate": 2.030452052203518e-05, | |
| "loss": 1.4098125, | |
| "step": 314000 | |
| }, | |
| { | |
| "epoch": 59.58010213731795, | |
| "learning_rate": 2.0209948931341026e-05, | |
| "loss": 1.4241875, | |
| "step": 315000 | |
| }, | |
| { | |
| "epoch": 59.76924531870626, | |
| "learning_rate": 2.011537734064687e-05, | |
| "loss": 1.43, | |
| "step": 316000 | |
| }, | |
| { | |
| "epoch": 59.95838850009457, | |
| "learning_rate": 2.0020805749952716e-05, | |
| "loss": 1.43575, | |
| "step": 317000 | |
| }, | |
| { | |
| "epoch": 60.14753168148288, | |
| "learning_rate": 1.9926234159258557e-05, | |
| "loss": 1.402625, | |
| "step": 318000 | |
| }, | |
| { | |
| "epoch": 60.33667486287119, | |
| "learning_rate": 1.9831662568564406e-05, | |
| "loss": 1.4025, | |
| "step": 319000 | |
| }, | |
| { | |
| "epoch": 60.5258180442595, | |
| "learning_rate": 1.9737090977870248e-05, | |
| "loss": 1.413, | |
| "step": 320000 | |
| }, | |
| { | |
| "epoch": 60.71496122564781, | |
| "learning_rate": 1.9642519387176093e-05, | |
| "loss": 1.4200625, | |
| "step": 321000 | |
| }, | |
| { | |
| "epoch": 60.90410440703613, | |
| "learning_rate": 1.9547947796481938e-05, | |
| "loss": 1.4270625, | |
| "step": 322000 | |
| }, | |
| { | |
| "epoch": 61.09324758842444, | |
| "learning_rate": 1.9453376205787783e-05, | |
| "loss": 1.4055625, | |
| "step": 323000 | |
| }, | |
| { | |
| "epoch": 61.28239076981275, | |
| "learning_rate": 1.9358804615093625e-05, | |
| "loss": 1.3949375, | |
| "step": 324000 | |
| }, | |
| { | |
| "epoch": 61.47153395120106, | |
| "learning_rate": 1.9264233024399473e-05, | |
| "loss": 1.4008125, | |
| "step": 325000 | |
| }, | |
| { | |
| "epoch": 61.66067713258937, | |
| "learning_rate": 1.9169661433705315e-05, | |
| "loss": 1.4103125, | |
| "step": 326000 | |
| }, | |
| { | |
| "epoch": 61.84982031397768, | |
| "learning_rate": 1.907508984301116e-05, | |
| "loss": 1.41725, | |
| "step": 327000 | |
| }, | |
| { | |
| "epoch": 62.03896349536599, | |
| "learning_rate": 1.8980518252317005e-05, | |
| "loss": 1.4091875, | |
| "step": 328000 | |
| }, | |
| { | |
| "epoch": 62.2281066767543, | |
| "learning_rate": 1.888594666162285e-05, | |
| "loss": 1.382125, | |
| "step": 329000 | |
| }, | |
| { | |
| "epoch": 62.41724985814261, | |
| "learning_rate": 1.8791375070928692e-05, | |
| "loss": 1.39175, | |
| "step": 330000 | |
| }, | |
| { | |
| "epoch": 62.60639303953093, | |
| "learning_rate": 1.8696803480234537e-05, | |
| "loss": 1.3970625, | |
| "step": 331000 | |
| }, | |
| { | |
| "epoch": 62.79553622091924, | |
| "learning_rate": 1.8602231889540382e-05, | |
| "loss": 1.409375, | |
| "step": 332000 | |
| }, | |
| { | |
| "epoch": 62.98467940230755, | |
| "learning_rate": 1.8507660298846227e-05, | |
| "loss": 1.4141875, | |
| "step": 333000 | |
| }, | |
| { | |
| "epoch": 63.17382258369586, | |
| "learning_rate": 1.8413088708152072e-05, | |
| "loss": 1.371875, | |
| "step": 334000 | |
| }, | |
| { | |
| "epoch": 63.36296576508417, | |
| "learning_rate": 1.8318517117457917e-05, | |
| "loss": 1.382125, | |
| "step": 335000 | |
| } | |
| ], | |
| "max_steps": 528700, | |
| "num_train_epochs": 100, | |
| "total_flos": 512235918148829184, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |