anlp-a1-transformers-C5 / summary.json
winterdewdev's picture
C5: best epoch 29, val loss 0.0131
7f4fc58 verified
Raw History Blame Contribute Delete
59.8 kB
{
"config": "C5",
"description": "token-free BLT instead of subword",
"parameters": 2117376,
"windowed": true,
"window": {
"src_window": 1024,
"src_stride": 768,
"train_examples": 25654
},
"best_epoch": 29,
"best_val_loss": 0.013062323450010613,
"test": {
"loss": 0.01330391115029037,
"ppl": 1.0133928019371958,
"token_acc": 0.9995187270330023,
"seq_acc": 0.9574726609963548
},
"test_full": {
"loss": 5.80302786322961,
"ppl": 331.3011774170338,
"token_acc": 0.32276305532330835,
"seq_acc": 0.07439824945295405
},
"gen_test_win": {
"n": 3292.0,
"seq_acc": 0.9574726609963548,
"bit_acc": 0.9997403020836493,
"lev": 0.08535844471445929,
"lev_norm": 0.0007114797463788965
},
"gen_test": {
"n": 500.0,
"seq_acc": 0.804,
"bit_acc": 0.9893873284246887,
"lev": 2.048,
"lev_norm": 0.003668687172604409
},
"gen_test_full": {
"n": 500.0,
"seq_acc": 0.068,
"bit_acc": 0.20238843136616647,
"lev": 490.088,
"lev_norm": 0.6793383173497168
},
"checkpoint": "/home2/winterdew/ANLP_A1/ANLP-M26-Assignment-1/outputs/checkpoints/C5_best.pt",
"history": [
{
"train/epoch_loss": 3.1866541384461513,
"epoch": 1,
"val/loss": 2.1205279343276415,
"val/ppl": 8.335536942365135,
"val/token_acc": 0.3877735600109597,
"val/seq_acc": 0.0,
"val_full/loss": 4.286994736286559,
"val_full/ppl": 72.7475141920631,
"val_full/token_acc": 0.1303971043467045,
"val_full/seq_acc": 0.0,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 2.016733264387992,
"epoch": 2,
"val/loss": 1.0093543742044582,
"val/ppl": 2.743828956282325,
"val/token_acc": 0.7000316554187959,
"val/seq_acc": 0.0,
"val_full/loss": 4.050685520927383,
"val_full/ppl": 57.436817693161835,
"val_full/token_acc": 0.26583661163420613,
"val_full/seq_acc": 0.0,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 1.4646863547941098,
"epoch": 3,
"val/loss": 0.5453954489599598,
"val/ppl": 1.725290511832387,
"val/token_acc": 0.8396719540969826,
"val/seq_acc": 0.0,
"val_full/loss": 4.437363523091868,
"val_full/ppl": 84.55172887470815,
"val_full/token_acc": 0.29828372887914356,
"val_full/seq_acc": 0.0,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 1.1722335806511286,
"epoch": 4,
"val/loss": 0.33587098110797875,
"val/ppl": 1.3991584952864973,
"val/token_acc": 0.902134213655456,
"val/seq_acc": 0.00032123353678124,
"val_full/loss": 4.856175854054479,
"val_full/ppl": 128.53173695747287,
"val_full/token_acc": 0.31575277312008254,
"val_full/seq_acc": 0.0,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 1.0009264107654219,
"epoch": 5,
"val/loss": 0.21881503036827438,
"val/ppl": 1.244601041998591,
"val/token_acc": 0.9383916387132472,
"val/seq_acc": 0.007388371345968519,
"val_full/loss": 5.262852808867334,
"val_full/ppl": 193.03138820485765,
"val_full/token_acc": 0.31434202889204177,
"val_full/seq_acc": 0.0,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.0234375,
"gen_val_win/bit_acc": 0.9359192056407271,
"gen_val_win/lev": 15.1953125,
"gen_val_win/lev_norm": 0.1265029750578017,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.0,
"gen_val/bit_acc": 0.6286467894692924,
"gen_val/lev": 188.515625,
"gen_val/lev_norm": 0.2514604836400943,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0,
"gen_val_full/bit_acc": 0.18570620868891582,
"gen_val_full/lev": 513.2890625,
"gen_val_full/lev_norm": 0.7428451081504603
},
{
"train/epoch_loss": 0.8867060201721001,
"epoch": 6,
"val/loss": 0.150144398525897,
"val/ppl": 1.1620020219935174,
"val/token_acc": 0.9605903336587546,
"val/seq_acc": 0.035014455509155155,
"val_full/loss": 5.230251155298755,
"val_full/ppl": 186.8397234143849,
"val_full/token_acc": 0.33211740616535534,
"val_full/seq_acc": 0.0,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.8115367409297058,
"epoch": 7,
"val/loss": 0.10905530072112715,
"val/ppl": 1.115224021328805,
"val/token_acc": 0.9736754601341232,
"val/seq_acc": 0.127208480565371,
"val_full/loss": 5.362812303503483,
"val_full/ppl": 213.32403557709898,
"val_full/token_acc": 0.3324156778021411,
"val_full/seq_acc": 0.02127659574468085,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.7549242661183612,
"epoch": 8,
"val/loss": 0.07894729092795537,
"val/ppl": 1.0821472815880244,
"val/token_acc": 0.9828422309887929,
"val/seq_acc": 0.2663026019916479,
"val_full/loss": 5.562786179543402,
"val_full/ppl": 260.54775880756915,
"val_full/token_acc": 0.33191990197342963,
"val_full/seq_acc": 0.036170212765957444,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.7162566222157561,
"epoch": 9,
"val/loss": 0.05933424177465677,
"val/ppl": 1.0611298553993511,
"val/token_acc": 0.9885348861335965,
"val/seq_acc": 0.39993575329264375,
"val_full/loss": 5.700837216868671,
"val_full/ppl": 299.11772256875804,
"val_full/token_acc": 0.3214521798013672,
"val_full/seq_acc": 0.0425531914893617,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.682473287172151,
"epoch": 10,
"val/loss": 0.05616943969202137,
"val/ppl": 1.0577768979449975,
"val/token_acc": 0.9901149969541635,
"val/seq_acc": 0.47542563443623514,
"val_full/loss": 5.55206798729121,
"val_full/ppl": 257.77007032896154,
"val_full/token_acc": 0.332544660131562,
"val_full/seq_acc": 0.059574468085106386,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.4609375,
"gen_val_win/bit_acc": 0.9866527768734944,
"gen_val_win/lev": 2.4921875,
"gen_val_win/lev_norm": 0.021498102685514504,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.0546875,
"gen_val/bit_acc": 0.8098412081832521,
"gen_val/lev": 49.1796875,
"gen_val/lev_norm": 0.07228347918983398,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0234375,
"gen_val_full/bit_acc": 0.19712909682261695,
"gen_val_full/lev": 502.078125,
"gen_val_full/lev_norm": 0.7059250807127767
},
{
"train/epoch_loss": 0.6628512337320761,
"epoch": 11,
"val/loss": 0.03870367599148255,
"val/ppl": 1.0394624203380365,
"val/token_acc": 0.9939349281634803,
"val/seq_acc": 0.594282043045294,
"val_full/loss": 5.686484610543903,
"val_full/ppl": 294.8552655070311,
"val_full/token_acc": 0.33418112343608924,
"val_full/seq_acc": 0.07234042553191489,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.6415253095793307,
"epoch": 12,
"val/loss": 0.033590495729570846,
"val/ppl": 1.0341610266483876,
"val/token_acc": 0.9949670544233793,
"val/seq_acc": 0.63989720526823,
"val_full/loss": 5.888909308291952,
"val_full/ppl": 361.01131751262545,
"val_full/token_acc": 0.3236368180059332,
"val_full/seq_acc": 0.07021276595744681,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.6261716489482699,
"epoch": 13,
"val/loss": 0.029946327140976697,
"val/ppl": 1.0303992279968137,
"val/token_acc": 0.9959778997294658,
"val/seq_acc": 0.6890459363957597,
"val_full/loss": 5.971079522896379,
"val_full/ppl": 391.9285382920556,
"val_full/token_acc": 0.32436637430672,
"val_full/seq_acc": 0.07872340425531915,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.6132363888093658,
"epoch": 14,
"val/loss": 0.025467302035602723,
"val/ppl": 1.0257943643344671,
"val/token_acc": 0.9970047057509117,
"val/seq_acc": 0.7462255059428204,
"val_full/loss": 5.863015066496437,
"val_full/ppl": 351.7831963904122,
"val_full/token_acc": 0.3292394879401522,
"val_full/seq_acc": 0.0851063829787234,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.6025004856604294,
"epoch": 15,
"val/loss": 0.023226369516013545,
"val/ppl": 1.0234982021182104,
"val/token_acc": 0.9974569260194242,
"val/seq_acc": 0.7757789913266945,
"val_full/loss": 5.89068596430414,
"val_full/ppl": 361.6532805452296,
"val_full/token_acc": 0.32046868953953306,
"val_full/seq_acc": 0.1,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.765625,
"gen_val_win/bit_acc": 0.9973355713447624,
"gen_val_win/lev": 0.6875,
"gen_val_win/lev_norm": 0.006319041391352384,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.328125,
"gen_val/bit_acc": 0.8984382648066569,
"gen_val/lev": 17.1796875,
"gen_val/lev_norm": 0.02379577879002951,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.046875,
"gen_val_full/bit_acc": 0.19981774682466116,
"gen_val_full/lev": 499.234375,
"gen_val_full/lev_norm": 0.6974027755296283
},
{
"train/epoch_loss": 0.5915844462756207,
"epoch": 16,
"val/loss": 0.020223079525560565,
"val/ppl": 1.020428951444445,
"val/token_acc": 0.9979277671225224,
"val/seq_acc": 0.8069386443944748,
"val_full/loss": 5.912791728050835,
"val_full/ppl": 369.7369208648866,
"val_full/token_acc": 0.32819553721140204,
"val_full/seq_acc": 0.09361702127659574,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5829343743752363,
"epoch": 17,
"val/loss": 0.018677845305089074,
"val/ppl": 1.0188533673458544,
"val/token_acc": 0.9981990992836299,
"val/seq_acc": 0.830067459042724,
"val_full/loss": 5.915054807413259,
"val_full/ppl": 370.5746123836947,
"val_full/token_acc": 0.3302552237843415,
"val_full/seq_acc": 0.10425531914893617,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5755495646053419,
"epoch": 18,
"val/loss": 0.017626009784440667,
"val/ppl": 1.0177822645944599,
"val/token_acc": 0.9983001838142386,
"val/seq_acc": 0.8397044651461613,
"val_full/loss": 5.908761313664872,
"val_full/ppl": 368.2497268772265,
"val_full/token_acc": 0.3285179930349542,
"val_full/seq_acc": 0.1,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5669382216032604,
"epoch": 19,
"val/loss": 0.016077107184220365,
"val/ppl": 1.0162070392486153,
"val/token_acc": 0.9987204826520325,
"val/seq_acc": 0.8734339865081915,
"val_full/loss": 5.857032424570529,
"val_full/ppl": 349.6848864678713,
"val_full/token_acc": 0.3293926544563395,
"val_full/seq_acc": 0.10425531914893617,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5609035634637771,
"epoch": 20,
"val/loss": 0.01558122174620217,
"val/ppl": 1.0157032418990488,
"val/token_acc": 0.998904030878664,
"val/seq_acc": 0.8914230645679408,
"val_full/loss": 5.983415630945279,
"val_full/ppl": 396.79335585174044,
"val_full/token_acc": 0.3256481362053399,
"val_full/seq_acc": 0.10212765957446808,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.8828125,
"gen_val_win/bit_acc": 0.9990385242161528,
"gen_val_win/lev": 0.3046875,
"gen_val_win/lev_norm": 0.00267191608058971,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.59375,
"gen_val/bit_acc": 0.9554161142021587,
"gen_val/lev": 7.5078125,
"gen_val/lev_norm": 0.010553709735544233,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.046875,
"gen_val_full/bit_acc": 0.1996173309617456,
"gen_val_full/lev": 499.25,
"gen_val_full/lev_norm": 0.6975380072356028
},
{
"train/epoch_loss": 0.5547243261575104,
"epoch": 21,
"val/loss": 0.014728826488929259,
"val/ppl": 1.0148378301616865,
"val/token_acc": 0.9990317166015381,
"val/seq_acc": 0.9023450048185031,
"val_full/loss": 5.9140383787928865,
"val_full/ppl": 370.19814110218886,
"val_full/token_acc": 0.3245598478008513,
"val_full/seq_acc": 0.10851063829787234,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5492340369414808,
"epoch": 22,
"val/loss": 0.014350483405035842,
"val/ppl": 1.0144539459119262,
"val/token_acc": 0.9991061999398813,
"val/seq_acc": 0.9132669450690653,
"val_full/loss": 5.8973892666548435,
"val_full/ppl": 364.0856953284918,
"val_full/token_acc": 0.3136406874758158,
"val_full/seq_acc": 0.10851063829787234,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5452264907056851,
"epoch": 23,
"val/loss": 0.014191510087682569,
"val/ppl": 1.0142926876210647,
"val/token_acc": 0.9991567422051857,
"val/seq_acc": 0.9142306456794089,
"val_full/loss": 5.936484516074422,
"val_full/ppl": 378.6016193889404,
"val_full/token_acc": 0.3199850058042048,
"val_full/seq_acc": 0.10638297872340426,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5420137426799669,
"epoch": 24,
"val/loss": 0.01399349216573586,
"val/ppl": 1.0140918593752686,
"val/token_acc": 0.9992445261396615,
"val/seq_acc": 0.9225827176357212,
"val_full/loss": 5.938875066254595,
"val_full/ppl": 379.50776822398177,
"val_full/token_acc": 0.314043757255256,
"val_full/seq_acc": 0.10638297872340426,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5385007705177155,
"epoch": 25,
"val/loss": 0.013562473314542483,
"val/ppl": 1.0136548608511022,
"val/token_acc": 0.9992897481665128,
"val/seq_acc": 0.926437520077096,
"val_full/loss": 5.931484970787518,
"val_full/ppl": 376.71350723108003,
"val_full/token_acc": 0.32597865342448085,
"val_full/seq_acc": 0.11063829787234042,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.8828125,
"gen_val_win/bit_acc": 0.9987458466349599,
"gen_val_win/lev": 0.328125,
"gen_val_win/lev_norm": 0.0030029220867988093,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.6484375,
"gen_val/bit_acc": 0.9642039800995025,
"gen_val/lev": 5.0625,
"gen_val/lev_norm": 0.007115335681933282,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.2041266878773455,
"gen_val_full/lev": 496.3125,
"gen_val_full/lev_norm": 0.6913037099030713
},
{
"train/epoch_loss": 0.534697450158602,
"epoch": 26,
"val/loss": 0.013472278606427286,
"val/ppl": 1.0135634386697587,
"val/token_acc": 0.9993881725778949,
"val/seq_acc": 0.9376806938644394,
"val_full/loss": 5.829398228155633,
"val_full/ppl": 340.15392241627166,
"val_full/token_acc": 0.32561185992519026,
"val_full/seq_acc": 0.11063829787234042,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5317636653966737,
"epoch": 27,
"val/loss": 0.01367300227098327,
"val/ppl": 1.0137669052570248,
"val/token_acc": 0.9994679761546913,
"val/seq_acc": 0.9441053646000642,
"val_full/loss": 5.8744636810196456,
"val_full/ppl": 355.833769081754,
"val_full/token_acc": 0.32039210628143944,
"val_full/seq_acc": 0.10851063829787234,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5297336776060356,
"epoch": 28,
"val/loss": 0.013508005997371283,
"val/ppl": 1.013599651293866,
"val/token_acc": 0.9995264987776752,
"val/seq_acc": 0.9495663347253454,
"val_full/loss": 5.718929995848382,
"val_full/ppl": 304.57884790543574,
"val_full/token_acc": 0.3257771185347607,
"val_full/seq_acc": 0.11063829787234042,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5271218951206255,
"epoch": 29,
"val/loss": 0.013062323450010613,
"val/ppl": 1.013148008271456,
"val/token_acc": 0.9995610803276203,
"val/seq_acc": 0.9547060713138452,
"val_full/loss": 5.828769659728493,
"val_full/ppl": 339.9401795833083,
"val_full/token_acc": 0.3218149426028634,
"val_full/seq_acc": 0.10638297872340426,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5250636516069236,
"epoch": 30,
"val/loss": 0.013839164579217769,
"val/ppl": 1.0139353691017972,
"val/token_acc": 0.9995052178238628,
"val/seq_acc": 0.9492451011885641,
"val_full/loss": 5.742995749377257,
"val_full/ppl": 311.9976792177771,
"val_full/token_acc": 0.32451954082290724,
"val_full/seq_acc": 0.11063829787234042,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.9375,
"gen_val_win/bit_acc": 0.9995844469069813,
"gen_val_win/lev": 0.1484375,
"gen_val_win/lev_norm": 0.0012928605481931945,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.75,
"gen_val/bit_acc": 0.9752925837201177,
"gen_val/lev": 3.578125,
"gen_val/lev_norm": 0.005754179268799118,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.20050511060450435,
"gen_val_full/lev": 498.6796875,
"gen_val_full/lev_norm": 0.6955337026160823
},
{
"train/epoch_loss": 0.5229915221433092,
"epoch": 31,
"val/loss": 0.013497638256154087,
"val/ppl": 1.0135891426094594,
"val/token_acc": 0.9995930017583388,
"val/seq_acc": 0.9569547060713138,
"val_full/loss": 5.719591660083194,
"val_full/ppl": 304.7804435227251,
"val_full/token_acc": 0.32597865342448085,
"val_full/seq_acc": 0.11063829787234042,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5207516891403389,
"epoch": 32,
"val/loss": 0.014165710896847455,
"val/ppl": 1.014266520028007,
"val/token_acc": 0.9995424594930344,
"val/seq_acc": 0.9575971731448764,
"val_full/loss": 5.833788697218012,
"val_full/ppl": 341.6506409330033,
"val_full/token_acc": 0.3184372178511544,
"val_full/seq_acc": 0.1127659574468085,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5192661805640433,
"epoch": 33,
"val/loss": 0.013403709259955819,
"val/ppl": 1.013493941669871,
"val/token_acc": 0.9996701452159086,
"val/seq_acc": 0.9656280115644074,
"val_full/loss": 5.808688077775337,
"val_full/ppl": 333.1817303061101,
"val_full/token_acc": 0.31728846897974977,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5174381254319835,
"epoch": 34,
"val/loss": 0.013720597464655173,
"val/ppl": 1.0138151568374814,
"val/token_acc": 0.9996861059312678,
"val/seq_acc": 0.9653067780276261,
"val_full/loss": 5.837206854906971,
"val_full/ppl": 342.82045486349733,
"val_full/token_acc": 0.3133827228169741,
"val_full/seq_acc": 0.11063829787234042,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5160897482660346,
"epoch": 35,
"val/loss": 0.013917877659750343,
"val/ppl": 1.014015182219287,
"val/token_acc": 0.9997419684350253,
"val/seq_acc": 0.9726951493735946,
"val_full/loss": 5.800546949945787,
"val_full/ppl": 330.4802666503007,
"val_full/token_acc": 0.3140195730684896,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.921875,
"gen_val_win/bit_acc": 0.999486669708624,
"gen_val_win/lev": 0.1484375,
"gen_val_win/lev_norm": 0.0013086962853102134,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.8984375,
"gen_val/bit_acc": 0.9886284049125791,
"gen_val/lev": 1.1328125,
"gen_val/lev_norm": 0.0013219149024288484,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.1993167071673723,
"gen_val_full/lev": 499.4609375,
"gen_val_full/lev_norm": 0.6972389289562869
},
{
"train/epoch_loss": 0.5138689738853911,
"epoch": 36,
"val/loss": 0.014133696068082967,
"val/ppl": 1.014234048978827,
"val/token_acc": 0.9997127071235333,
"val/seq_acc": 0.9688403469322198,
"val_full/loss": 5.836556936377451,
"val_full/ppl": 342.5977218845682,
"val_full/token_acc": 0.31247178511543916,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5130237455379933,
"epoch": 37,
"val/loss": 0.014182012285739865,
"val/ppl": 1.0142830541157544,
"val/token_acc": 0.9997339880773456,
"val/seq_acc": 0.9707677481529071,
"val_full/loss": 5.868360684976783,
"val_full/ppl": 353.66873032861906,
"val_full/token_acc": 0.3080259254482136,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5113184382196079,
"epoch": 38,
"val/loss": 0.015070331816712943,
"val/ppl": 1.0151844618724424,
"val/token_acc": 0.9995850214006592,
"val/seq_acc": 0.9681978798586572,
"val_full/loss": 5.766306660274652,
"val_full/ppl": 319.356061405807,
"val_full/token_acc": 0.3106378176189862,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5099328701335593,
"epoch": 39,
"val/loss": 0.014448271159159838,
"val/ppl": 1.0145531519354472,
"val/token_acc": 0.9997472886734784,
"val/seq_acc": 0.9733376164471571,
"val_full/loss": 5.693308455950519,
"val_full/ppl": 296.87419284595126,
"val_full/token_acc": 0.31723606990842257,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5088077843040599,
"epoch": 40,
"val/loss": 0.01451898184153518,
"val/ppl": 1.0146248942175695,
"val/token_acc": 0.9997738898657438,
"val/seq_acc": 0.9749437841310633,
"val_full/loss": 5.725371712210193,
"val_full/ppl": 306.5471914022248,
"val_full/token_acc": 0.30885624919386045,
"val_full/seq_acc": 0.1127659574468085,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.9453125,
"gen_val_win/bit_acc": 0.9996903722052017,
"gen_val_win/lev": 0.09375,
"gen_val_win/lev_norm": 0.0007992546463918573,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.921875,
"gen_val/bit_acc": 0.993599567153723,
"gen_val/lev": 0.921875,
"gen_val/lev_norm": 0.0010594884894371724,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.19908654207480522,
"gen_val_full/lev": 499.5703125,
"gen_val_full/lev_norm": 0.6975838937584498
},
{
"train/epoch_loss": 0.5078803171540733,
"epoch": 41,
"val/loss": 0.014111817464192924,
"val/ppl": 1.0142118591965592,
"val/token_acc": 0.9997925107003296,
"val/seq_acc": 0.9762287182781882,
"val_full/loss": 5.733528065370865,
"val_full/ppl": 309.0577230290243,
"val_full/token_acc": 0.31173416741906357,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5066765240600282,
"epoch": 42,
"val/loss": 0.014888131387611293,
"val/ppl": 1.014999511677386,
"val/token_acc": 0.9997845303426499,
"val/seq_acc": 0.9771924188885319,
"val_full/loss": 5.7949392888740645,
"val_full/ppl": 328.63223174348826,
"val_full/token_acc": 0.3007948536050561,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5059437270919581,
"epoch": 43,
"val/loss": 0.014909094175155374,
"val/ppl": 1.015020789119523,
"val/token_acc": 0.999819111892595,
"val/seq_acc": 0.9800835207195632,
"val_full/loss": 5.753488600430882,
"val_full/ppl": 315.28866008624954,
"val_full/token_acc": 0.30619195795176063,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5048771646254675,
"epoch": 44,
"val/loss": 0.01497256695546071,
"val/ppl": 1.0150852173557743,
"val/token_acc": 0.9998164517733685,
"val/seq_acc": 0.9800835207195632,
"val_full/loss": 5.689350357951156,
"val_full/ppl": 295.7014581277137,
"val_full/token_acc": 0.3108353218109119,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5042696986858387,
"epoch": 45,
"val/loss": 0.01492492011076165,
"val/ppl": 1.0150368529002827,
"val/token_acc": 0.9998084714156888,
"val/seq_acc": 0.9800835207195632,
"val_full/loss": 5.803457098008432,
"val_full/ppl": 331.44341392889106,
"val_full/token_acc": 0.30477718302592544,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.953125,
"gen_val_win/bit_acc": 0.9996414836060231,
"gen_val_win/lev": 0.09375,
"gen_val_win/lev_norm": 0.0007516217545874913,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.9375,
"gen_val/bit_acc": 0.9944324892403162,
"gen_val/lev": 0.84375,
"gen_val/lev_norm": 0.0011777931696110454,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.19860898865145177,
"gen_val_full/lev": 499.8828125,
"gen_val_full/lev_norm": 0.6982171033221508
},
{
"train/epoch_loss": 0.5033634459585917,
"epoch": 46,
"val/loss": 0.014932920337343279,
"val/ppl": 1.0150449734575777,
"val/token_acc": 0.9998324124887278,
"val/seq_acc": 0.9813684548666881,
"val_full/loss": 5.713103637931487,
"val_full/ppl": 302.8094221828801,
"val_full/token_acc": 0.31191957951760607,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5024333880131976,
"epoch": 47,
"val/loss": 0.015160119405140605,
"val/ppl": 1.0152756169293182,
"val/token_acc": 0.9998164517733685,
"val/seq_acc": 0.9797622871827819,
"val_full/loss": 5.686779843412429,
"val_full/ppl": 294.9423293242848,
"val_full/token_acc": 0.3118591190506901,
"val_full/seq_acc": 0.1127659574468085,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.5015278497241679,
"epoch": 48,
"val/loss": 0.015149872348730086,
"val/ppl": 1.0152652133961022,
"val/token_acc": 0.999848373204087,
"val/seq_acc": 0.9832958560873756,
"val_full/loss": 5.643808448267001,
"val_full/ppl": 282.5366984782148,
"val_full/token_acc": 0.3147289758803044,
"val_full/seq_acc": 0.1127659574468085,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.500775021507853,
"epoch": 49,
"val/loss": 0.015552178769386086,
"val/ppl": 1.0156737432817082,
"val/token_acc": 0.9998403928464074,
"val/seq_acc": 0.9823321554770318,
"val_full/loss": 5.564631121244397,
"val_full/ppl": 261.0288979332369,
"val_full/token_acc": 0.31682090803559915,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4999965271747618,
"epoch": 50,
"val/loss": 0.015673951312192392,
"val/ppl": 1.0157974319868803,
"val/token_acc": 0.9998643339194463,
"val/seq_acc": 0.9852232573080629,
"val_full/loss": 5.6626083470461035,
"val_full/ppl": 287.8986035237004,
"val_full/token_acc": 0.3122500967367471,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.953125,
"gen_val_win/bit_acc": 0.9997962975034222,
"gen_val_win/lev": 0.0859375,
"gen_val_win/lev_norm": 0.0007358012743524247,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.9453125,
"gen_val/bit_acc": 0.9973778411935517,
"gen_val/lev": 0.5703125,
"gen_val/lev_norm": 0.0008795411022743086,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.19778227321692513,
"gen_val_full/lev": 500.4453125,
"gen_val_full/lev_norm": 0.6988099835533016
},
{
"train/epoch_loss": 0.4994901844539547,
"epoch": 51,
"val/loss": 0.01563506624737282,
"val/ppl": 1.0157579334058515,
"val/token_acc": 0.9998909351117117,
"val/seq_acc": 0.9877931256023129,
"val_full/loss": 5.655067431054914,
"val_full/ppl": 285.73574953993307,
"val_full/token_acc": 0.3107950148329679,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.499114902759728,
"epoch": 52,
"val/loss": 0.01581154441240372,
"val/ppl": 1.0159372083206433,
"val/token_acc": 0.9998616738002197,
"val/seq_acc": 0.9852232573080629,
"val_full/loss": 5.659433338052609,
"val_full/ppl": 286.98597244090774,
"val_full/token_acc": 0.310976396233716,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49831972492306015,
"epoch": 53,
"val/loss": 0.01588985215921748,
"val/ppl": 1.0160167671893283,
"val/token_acc": 0.9998457130848605,
"val/seq_acc": 0.9858657243816255,
"val_full/loss": 5.620191629449891,
"val_full/ppl": 275.9422568314296,
"val_full/token_acc": 0.31154875532052106,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49756993148986833,
"epoch": 54,
"val/loss": 0.015955108188698528,
"val/ppl": 1.0160830705727655,
"val/token_acc": 0.9998669940386729,
"val/seq_acc": 0.9858657243816255,
"val_full/loss": 5.675191602998436,
"val_full/ppl": 291.5441939358536,
"val_full/token_acc": 0.30960192828582483,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4970159610162055,
"epoch": 55,
"val/loss": 0.01611118874410451,
"val/ppl": 1.0162416737598738,
"val/token_acc": 0.99985369344254,
"val/seq_acc": 0.9842595566977193,
"val_full/loss": 5.623329055335435,
"val_full/ppl": 276.80936474305474,
"val_full/token_acc": 0.31241535534631754,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.953125,
"gen_val_win/bit_acc": 0.9998451861026009,
"gen_val_win/lev": 0.0703125,
"gen_val_win/lev_norm": 0.0006137309618524247,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.953125,
"gen_val/bit_acc": 0.9934070541704215,
"gen_val/lev": 0.546875,
"gen_val/lev_norm": 0.0008086136092221045,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.20092786281534183,
"gen_val_full/lev": 498.4140625,
"gen_val_full/lev_norm": 0.6954181950558825
},
{
"train/epoch_loss": 0.4966283208711486,
"epoch": 56,
"val/loss": 0.01616665372775626,
"val/ppl": 1.0162980411508888,
"val/token_acc": 0.9998643339194463,
"val/seq_acc": 0.9849020237712818,
"val_full/loss": 5.6831483264038924,
"val_full/ppl": 293.87318372400546,
"val_full/token_acc": 0.3082113375467561,
"val_full/seq_acc": 0.1127659574468085,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4957777168090801,
"epoch": 57,
"val/loss": 0.016285903833288175,
"val/ppl": 1.016419242026013,
"val/token_acc": 0.9998882749924851,
"val/seq_acc": 0.9871506585287504,
"val_full/loss": 5.5686006657957,
"val_full/ppl": 262.0671230498876,
"val_full/token_acc": 0.3147612214626596,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49537931973500143,
"epoch": 58,
"val/loss": 0.016260507195743304,
"val/ppl": 1.0163934287227165,
"val/token_acc": 0.9998802946348055,
"val/seq_acc": 0.9868294249919691,
"val_full/loss": 5.6505183634812735,
"val_full/ppl": 284.4388703390998,
"val_full/token_acc": 0.3098276473623114,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4951852076963296,
"epoch": 59,
"val/loss": 0.016401629329553163,
"val/ppl": 1.016536874453614,
"val/token_acc": 0.9998643339194463,
"val/seq_acc": 0.9849020237712818,
"val_full/loss": 5.519495540536727,
"val_full/ppl": 249.50913819095962,
"val_full/token_acc": 0.3175061266606475,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49449047557730924,
"epoch": 60,
"val/loss": 0.016348883564456476,
"val/ppl": 1.0164832578524587,
"val/token_acc": 0.9998935952309382,
"val/seq_acc": 0.9884355926758753,
"val_full/loss": 5.60996477296087,
"val_full/ppl": 273.1346161114729,
"val_full/token_acc": 0.31350364375080614,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.9765625,
"gen_val_win/bit_acc": 0.9999185190013689,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.0002727912888968807,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.9609375,
"gen_val/bit_acc": 0.993870156203064,
"gen_val/lev": 0.515625,
"gen_val/lev_norm": 0.0007564585231134166,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.20161209509732694,
"gen_val_full/lev": 498.015625,
"gen_val_full/lev_norm": 0.6944667327941509
},
{
"train/epoch_loss": 0.4939990268829755,
"epoch": 61,
"val/loss": 0.016651286690155538,
"val/ppl": 1.0167906920490422,
"val/token_acc": 0.9998909351117117,
"val/seq_acc": 0.9881143591390941,
"val_full/loss": 5.582873634349204,
"val_full/ppl": 265.83442015747335,
"val_full/token_acc": 0.31413646330452727,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49368070255491203,
"epoch": 62,
"val/loss": 0.016427972408748506,
"val/ppl": 1.0165636535177227,
"val/token_acc": 0.9999042357078445,
"val/seq_acc": 0.9890780597494379,
"val_full/loss": 5.556228265721737,
"val_full/ppl": 258.84469942073684,
"val_full/token_acc": 0.3186347220430801,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4933575611756627,
"epoch": 63,
"val/loss": 0.016639417182639307,
"val/ppl": 1.0167786233159055,
"val/token_acc": 0.9999015755886179,
"val/seq_acc": 0.9890780597494379,
"val_full/loss": 5.525684991898298,
"val_full/ppl": 251.0582520029028,
"val_full/token_acc": 0.3175746485231523,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4930248599397274,
"epoch": 64,
"val/loss": 0.01673973202161538,
"val/ppl": 1.0168806264159045,
"val/token_acc": 0.9998962553501648,
"val/seq_acc": 0.9881143591390941,
"val_full/loss": 5.639065875709403,
"val_full/ppl": 281.199920075349,
"val_full/token_acc": 0.30926334967109503,
"val_full/seq_acc": 0.1148936170212766,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4927113623244507,
"epoch": 65,
"val/loss": 0.01669528331921946,
"val/ppl": 1.016835428396073,
"val/token_acc": 0.9999068958270709,
"val/seq_acc": 0.9893992932862191,
"val_full/loss": 5.620874423358297,
"val_full/ppl": 276.130732861286,
"val_full/token_acc": 0.31077486134399585,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.9765625,
"gen_val_win/bit_acc": 0.9999185190013689,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.0002727912888968807,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.953125,
"gen_val/bit_acc": 0.9987071933888437,
"gen_val/lev": 0.296875,
"gen_val/lev_norm": 0.0006399595635804135,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.1994576245709848,
"gen_val_full/lev": 499.375,
"gen_val_full/lev_norm": 0.6971245252575937
},
{
"train/epoch_loss": 0.4922028956865135,
"epoch": 66,
"val/loss": 0.016656805536448216,
"val/ppl": 1.016796303576068,
"val/token_acc": 0.9999095559462975,
"val/seq_acc": 0.9897205268230003,
"val_full/loss": 5.566541498804898,
"val_full/ppl": 261.5280383039699,
"val_full/token_acc": 0.3138140074809751,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4920360903044294,
"epoch": 67,
"val/loss": 0.016920333663696847,
"val/ppl": 1.0170642933116083,
"val/token_acc": 0.9999201964232037,
"val/seq_acc": 0.9913266945069065,
"val_full/loss": 5.608267896424167,
"val_full/ppl": 272.6715333982434,
"val_full/token_acc": 0.3118913646330453,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49170401328222413,
"epoch": 68,
"val/loss": 0.016995275793049863,
"val/ppl": 1.0171405171315901,
"val/token_acc": 0.9999175363039772,
"val/seq_acc": 0.9906842274333441,
"val_full/loss": 5.626884414198536,
"val_full/ppl": 277.79527296160785,
"val_full/token_acc": 0.3098800464336386,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4914124424469441,
"epoch": 69,
"val/loss": 0.016945859175120877,
"val/ppl": 1.017090254729184,
"val/token_acc": 0.9999068958270709,
"val/seq_acc": 0.9897205268230003,
"val_full/loss": 5.76396784787945,
"val_full/ppl": 318.6100202561255,
"val_full/token_acc": 0.30071423964916805,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4911112432319327,
"epoch": 70,
"val/loss": 0.01712097184529551,
"val/ppl": 1.0172683757146639,
"val/token_acc": 0.9999068958270709,
"val/seq_acc": 0.9897205268230003,
"val_full/loss": 5.7432636963111054,
"val_full/ppl": 312.08128924031575,
"val_full/token_acc": 0.3012221075712627,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.9765625,
"gen_val_win/bit_acc": 0.9999185190013689,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.0002727912888968807,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.9609375,
"gen_val/bit_acc": 0.9987024979653165,
"gen_val/lev": 0.3046875,
"gen_val/lev_norm": 0.0006499923915055435,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.20029530024801462,
"gen_val_full/lev": 498.8125,
"gen_val_full/lev_norm": 0.6960954846300292
},
{
"train/epoch_loss": 0.49085372448562087,
"epoch": 71,
"val/loss": 0.017174444256119636,
"val/ppl": 1.0173227729615315,
"val/token_acc": 0.9999281767808833,
"val/seq_acc": 0.991969161580469,
"val_full/loss": 5.664721550579212,
"val_full/ppl": 288.5076351470903,
"val_full/token_acc": 0.30660308912678963,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49070459552239304,
"epoch": 72,
"val/loss": 0.017172861258686535,
"val/ppl": 1.017321162543468,
"val/token_acc": 0.9999068958270709,
"val/seq_acc": 0.9893992932862191,
"val_full/loss": 5.585652485580178,
"val_full/ppl": 266.5741618031302,
"val_full/token_acc": 0.31170192183670836,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.49032880384428545,
"epoch": 73,
"val/loss": 0.01717745690121719,
"val/ppl": 1.0173258377986127,
"val/token_acc": 0.9999201964232037,
"val/seq_acc": 0.9913266945069065,
"val_full/loss": 5.622755940492793,
"val_full/ppl": 276.65076663934286,
"val_full/token_acc": 0.30906181478137495,
"val_full/seq_acc": 0.11702127659574468,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4900574655158264,
"epoch": 74,
"val/loss": 0.017200995705097227,
"val/ppl": 1.0173497847138304,
"val/token_acc": 0.9999255166616567,
"val/seq_acc": 0.9916479280436877,
"val_full/loss": 5.618771139120542,
"val_full/ppl": 275.55056178951486,
"val_full/token_acc": 0.3093520250225719,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4899491495325084,
"epoch": 75,
"val/loss": 0.01721965104135147,
"val/ppl": 1.0173687638931834,
"val/token_acc": 0.9999281767808833,
"val/seq_acc": 0.991969161580469,
"val_full/loss": 5.605712355297949,
"val_full/ppl": 271.9755997030347,
"val_full/token_acc": 0.30973494131304014,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.9765625,
"gen_val_win/bit_acc": 0.9998125937031485,
"gen_val_win/lev": 0.03125,
"gen_val_win/lev_norm": 0.00033430703692837675,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.96875,
"gen_val/bit_acc": 0.9999467645364131,
"gen_val/lev": 0.0703125,
"gen_val/lev_norm": 0.00014029082794618083,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.20044248064734324,
"gen_val_full/lev": 498.703125,
"gen_val_full/lev_norm": 0.6958084145288699
},
{
"train/epoch_loss": 0.48975437493098345,
"epoch": 76,
"val/loss": 0.017319593729095722,
"val/ppl": 1.017470447543058,
"val/token_acc": 0.9999175363039772,
"val/seq_acc": 0.9910054609701253,
"val_full/loss": 5.59559950772584,
"val_full/ppl": 269.23901258654575,
"val_full/token_acc": 0.31000499806526505,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48942971682905256,
"epoch": 77,
"val/loss": 0.017313916946342007,
"val/ppl": 1.0174646716007634,
"val/token_acc": 0.9999228565424302,
"val/seq_acc": 0.9913266945069065,
"val_full/loss": 5.602603427705405,
"val_full/ppl": 271.1313602762382,
"val_full/token_acc": 0.31075873855281827,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4893676685424814,
"epoch": 78,
"val/loss": 0.017287042460272368,
"val/ppl": 1.0174373281280427,
"val/token_acc": 0.9999334970193364,
"val/seq_acc": 0.9922903951172503,
"val_full/loss": 5.633076888583451,
"val_full/ppl": 279.5208503563959,
"val_full/token_acc": 0.3087836966335612,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48916773301110306,
"epoch": 79,
"val/loss": 0.01737927039229864,
"val/ppl": 1.0175311685960717,
"val/token_acc": 0.9999255166616567,
"val/seq_acc": 0.9913266945069065,
"val_full/loss": 5.659530562891986,
"val_full/ppl": 287.0138759624174,
"val_full/token_acc": 0.306135528182639,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4890072241685634,
"epoch": 80,
"val/loss": 0.017274721549162413,
"val/ppl": 1.0174247924503885,
"val/token_acc": 0.9999308369001099,
"val/seq_acc": 0.9926116286540315,
"val_full/loss": 5.659631235867366,
"val_full/ppl": 287.04277195778445,
"val_full/token_acc": 0.3071190184444731,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.984375,
"gen_val_win/bit_acc": 0.9998777785020533,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.00018406665231299212,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.9921875,
"gen_val/bit_acc": 0.9999765137660646,
"gen_val/lev": 0.0234375,
"gen_val/lev_norm": 7.34717868338558e-05,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.20006043790866046,
"gen_val_full/lev": 498.984375,
"gen_val_full/lev_norm": 0.6963084192254105
},
{
"train/epoch_loss": 0.488856665139781,
"epoch": 81,
"val/loss": 0.017322171856121987,
"val/ppl": 1.0174730707144986,
"val/token_acc": 0.9999255166616567,
"val/seq_acc": 0.9916479280436877,
"val_full/loss": 5.6285108952280565,
"val_full/ppl": 278.247469347904,
"val_full/token_acc": 0.3084249645298594,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4887467295748932,
"epoch": 82,
"val/loss": 0.017448824862879426,
"val/ppl": 1.0176019448991787,
"val/token_acc": 0.9999388172577895,
"val/seq_acc": 0.9932540957275939,
"val_full/loss": 5.6025002670337285,
"val_full/ppl": 271.10339162565685,
"val_full/token_acc": 0.3101299496968915,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4886407845484051,
"epoch": 83,
"val/loss": 0.017451164725442496,
"val/ppl": 1.0176043259506593,
"val/token_acc": 0.9999308369001099,
"val/seq_acc": 0.991969161580469,
"val_full/loss": 5.588317327894242,
"val_full/ppl": 267.28548727349596,
"val_full/token_acc": 0.31065797110795823,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4884786278083735,
"epoch": 84,
"val/loss": 0.01741648346567565,
"val/ppl": 1.017569034762666,
"val/token_acc": 0.9999281767808833,
"val/seq_acc": 0.9916479280436877,
"val_full/loss": 5.598656807747807,
"val_full/ppl": 270.06341660871095,
"val_full/token_acc": 0.31032342319102285,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48833256112667095,
"epoch": 85,
"val/loss": 0.01742900136101727,
"val/ppl": 1.0175817726650718,
"val/token_acc": 0.9999334970193364,
"val/seq_acc": 0.9926116286540315,
"val_full/loss": 5.6271984622384075,
"val_full/ppl": 277.8825277229431,
"val_full/token_acc": 0.306586966335612,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.984375,
"gen_val_win/bit_acc": 0.9998777785020533,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.00018406665231299212,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.9765625,
"gen_val/bit_acc": 0.9999592905278453,
"gen_val/lev": 0.0625,
"gen_val/lev_norm": 0.0001173802707614301,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.19984279880752562,
"gen_val_full/lev": 499.09375,
"gen_val_full/lev_norm": 0.6965159214632013
},
{
"train/epoch_loss": 0.48827507891262556,
"epoch": 86,
"val/loss": 0.017649638312953738,
"val/ppl": 1.0178063135758388,
"val/token_acc": 0.999912216065524,
"val/seq_acc": 0.991969161580469,
"val_full/loss": 5.631356221482813,
"val_full/ppl": 279.0403015760467,
"val_full/token_acc": 0.3076470398555398,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48809898672555746,
"epoch": 87,
"val/loss": 0.01749464285531514,
"val/ppl": 1.017648570445529,
"val/token_acc": 0.999936157138563,
"val/seq_acc": 0.9926116286540315,
"val_full/loss": 5.618546128565153,
"val_full/ppl": 275.4885669795735,
"val_full/token_acc": 0.3078445440474655,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48810374684761887,
"epoch": 88,
"val/loss": 0.017519065977547366,
"val/ppl": 1.0176734249044648,
"val/token_acc": 0.999941477377016,
"val/seq_acc": 0.9932540957275939,
"val_full/loss": 5.61298609585604,
"val_full/ppl": 273.96109187697067,
"val_full/token_acc": 0.3084088417386818,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4880281306264406,
"epoch": 89,
"val/loss": 0.017488687197604973,
"val/ppl": 1.0176425096970219,
"val/token_acc": 0.999936157138563,
"val/seq_acc": 0.9926116286540315,
"val_full/loss": 5.594718721533197,
"val_full/ppl": 269.0019749867854,
"val_full/token_acc": 0.3096099896814136,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48791934129900466,
"epoch": 90,
"val/loss": 0.017605449428416232,
"val/ppl": 1.0177613388438655,
"val/token_acc": 0.999941477377016,
"val/seq_acc": 0.9932540957275939,
"val_full/loss": 5.5957428966508935,
"val_full/ppl": 269.27762124710404,
"val_full/token_acc": 0.30985183154907775,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.984375,
"gen_val_win/bit_acc": 0.9998777785020533,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.00018406665231299212,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.9765625,
"gen_val/bit_acc": 0.9999592905278453,
"gen_val/lev": 0.0625,
"gen_val/lev_norm": 0.0001173802707614301,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.1997864318460806,
"gen_val_full/lev": 499.1328125,
"gen_val_full/lev_norm": 0.6966276382539325
},
{
"train/epoch_loss": 0.4877652882637823,
"epoch": 91,
"val/loss": 0.017515104678633542,
"val/ppl": 1.0176693936038166,
"val/token_acc": 0.999936157138563,
"val/seq_acc": 0.9926116286540315,
"val_full/loss": 5.598830505630885,
"val_full/ppl": 270.1103301267441,
"val_full/token_acc": 0.30984377015348896,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48769883971261857,
"epoch": 92,
"val/loss": 0.017548355336866276,
"val/ppl": 1.0177032323435944,
"val/token_acc": 0.9999308369001099,
"val/seq_acc": 0.991969161580469,
"val_full/loss": 5.624317804398298,
"val_full/ppl": 277.08319509547965,
"val_full/token_acc": 0.30815893847542886,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4877708919922313,
"epoch": 93,
"val/loss": 0.017538464601510297,
"val/ppl": 1.0176931665600315,
"val/token_acc": 0.9999388172577895,
"val/seq_acc": 0.9929328621908127,
"val_full/loss": 5.612142640852735,
"val_full/ppl": 273.7301154461226,
"val_full/token_acc": 0.30851363988133623,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48774644293987246,
"epoch": 94,
"val/loss": 0.017526633492321524,
"val/ppl": 1.017681126192283,
"val/token_acc": 0.9999334970193364,
"val/seq_acc": 0.9922903951172503,
"val_full/loss": 5.610030681167774,
"val_full/ppl": 273.15261851751075,
"val_full/token_acc": 0.30856200825486907,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.487697446361147,
"epoch": 95,
"val/loss": 0.017546973260914065,
"val/ppl": 1.0177018258014026,
"val/token_acc": 0.9999441374962426,
"val/seq_acc": 0.9935753292643752,
"val_full/loss": 5.599483226754966,
"val_full/ppl": 270.2866943971311,
"val_full/token_acc": 0.30946085386302075,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.984375,
"gen_val_win/bit_acc": 0.9998777785020533,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.00018406665231299212,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.984375,
"gen_val/bit_acc": 0.9999608562767743,
"gen_val/lev": 0.0546875,
"gen_val/lev_norm": 0.00010943266140232532,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.19966900067640353,
"gen_val_full/lev": 499.203125,
"gen_val_full/lev_norm": 0.6968300508725622
},
{
"train/epoch_loss": 0.48767787307277877,
"epoch": 96,
"val/loss": 0.01756051555138172,
"val/ppl": 1.0177156079084573,
"val/token_acc": 0.999941477377016,
"val/seq_acc": 0.9932540957275939,
"val_full/loss": 5.606620364445618,
"val_full/ppl": 272.22266818875516,
"val_full/token_acc": 0.3092069199019734,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4876622269278453,
"epoch": 97,
"val/loss": 0.017595397367317668,
"val/ppl": 1.017751108296123,
"val/token_acc": 0.999941477377016,
"val/seq_acc": 0.9932540957275939,
"val_full/loss": 5.612336429245131,
"val_full/ppl": 273.7831663052997,
"val_full/token_acc": 0.3085660389526635,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4876181727781557,
"epoch": 98,
"val/loss": 0.01757747174135083,
"val/ppl": 1.0177328646339434,
"val/token_acc": 0.9999441374962426,
"val/seq_acc": 0.9935753292643752,
"val_full/loss": 5.595181165380538,
"val_full/ppl": 269.12640206309106,
"val_full/token_acc": 0.3094890687475816,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.4875374911283317,
"epoch": 99,
"val/loss": 0.017598582175740715,
"val/ppl": 1.0177543496435868,
"val/token_acc": 0.9999441374962426,
"val/seq_acc": 0.9935753292643752,
"val_full/loss": 5.595181590493196,
"val_full/ppl": 269.1265164721555,
"val_full/token_acc": 0.30946488456081517,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762
},
{
"train/epoch_loss": 0.48759286035028776,
"epoch": 100,
"val/loss": 0.0175882423504323,
"val/ppl": 1.0177438262958094,
"val/token_acc": 0.9999441374962426,
"val/seq_acc": 0.9935753292643752,
"val_full/loss": 5.593321297501774,
"val_full/ppl": 268.62632769284664,
"val_full/token_acc": 0.30946085386302075,
"val_full/seq_acc": 0.11914893617021277,
"system/peak_gpu_mem_gb": 0.7370562553405762,
"gen_val_win/n": 128.0,
"gen_val_win/seq_acc": 0.984375,
"gen_val_win/bit_acc": 0.9998777785020533,
"gen_val_win/lev": 0.0234375,
"gen_val_win/lev_norm": 0.00018406665231299212,
"gen_val/n": 128.0,
"gen_val/seq_acc": 0.984375,
"gen_val/bit_acc": 0.9999608562767743,
"gen_val/lev": 0.0546875,
"gen_val/lev_norm": 0.00010943266140232532,
"gen_val_full/n": 128.0,
"gen_val_full/seq_acc": 0.0625,
"gen_val_full/bit_acc": 0.19956252974922964,
"gen_val_full/lev": 499.2734375,
"gen_val_full/lev_norm": 0.696955161404202
}
]
}