escape-blt-entropy-student / train_log.json
Devatri's picture
Distilled BLT entropy student (ESCAPE, 14.3M, alpha=0.1, T=2)
ac17923 verified
Raw History Blame Contribute Delete
16.5 kB
{
"args": {
"ckpt_dir": "/home2/divyansh.atri/escape-final-20260928/checkpoints/itazap__blt-1b-hf/91aa6b8e168046ad91e517d2191e1e974f50cc01",
"data": "/scratch/divyansh.atri/escape-34241/corpus_ext/manifest_ext.parquet",
"out": "results/distill/full/student",
"train_split": "train",
"limit_files": null,
"domains": "cpp,prose,py",
"epochs": 2,
"max_steps": null,
"batch_size": 48,
"lr": 0.0005,
"weight_decay": 0.1,
"warmup": 200,
"temperature": 2.0,
"alpha": 0.1,
"dim": 384,
"layers": 8,
"heads": 6,
"teacher_dtype": "float32",
"device": "cuda",
"seed": 20260927,
"log_every": 100
},
"student_params": 14262144,
"n_files": 27153,
"n_windows": 164627,
"steps": 6860,
"log": [
{
"step": 0,
"epoch": 0,
"loss": 12.98613452911377,
"soft_kl": 13.808537483215332,
"hard_ce_nats": 5.584506511688232,
"lr": 2.5e-06,
"elapsed_s": 2.8
},
{
"step": 100,
"epoch": 0,
"loss": 5.558244705200195,
"soft_kl": 5.8554840087890625,
"hard_ce_nats": 2.883091926574707,
"lr": 0.0002525,
"elapsed_s": 146.0
},
{
"step": 200,
"epoch": 0,
"loss": 3.0039165019989014,
"soft_kl": 3.0849874019622803,
"hard_ce_nats": 2.274280309677124,
"lr": 0.0005,
"elapsed_s": 293.4
},
{
"step": 300,
"epoch": 0,
"loss": 2.220654249191284,
"soft_kl": 2.279080390930176,
"hard_ce_nats": 1.694819450378418,
"lr": 0.0004997219129473495,
"elapsed_s": 441.3
},
{
"step": 400,
"epoch": 0,
"loss": 1.6722427606582642,
"soft_kl": 1.6893478631973267,
"hard_ce_nats": 1.518296718597412,
"lr": 0.0004988882704486687,
"elapsed_s": 589.5
},
{
"step": 500,
"epoch": 0,
"loss": 1.2204997539520264,
"soft_kl": 1.2016428709030151,
"hard_ce_nats": 1.3902111053466797,
"lr": 0.0004975009271054409,
"elapsed_s": 737.6
},
{
"step": 600,
"epoch": 0,
"loss": 0.9225186109542847,
"soft_kl": 0.8971008658409119,
"hard_ce_nats": 1.1512784957885742,
"lr": 0.0004955629693354371,
"elapsed_s": 885.6
},
{
"step": 700,
"epoch": 0,
"loss": 0.7341705560684204,
"soft_kl": 0.6992050409317017,
"hard_ce_nats": 1.0488605499267578,
"lr": 0.0004930787085063723,
"elapsed_s": 1033.5
},
{
"step": 800,
"epoch": 0,
"loss": 0.7877124547958374,
"soft_kl": 0.7521551847457886,
"hard_ce_nats": 1.1077280044555664,
"lr": 0.0004900536713444224,
"elapsed_s": 1181.2
},
{
"step": 900,
"epoch": 0,
"loss": 0.6850633025169373,
"soft_kl": 0.6459649801254272,
"hard_ce_nats": 1.036948323249817,
"lr": 0.00048649458763893563,
"elapsed_s": 1329.0
},
{
"step": 1000,
"epoch": 0,
"loss": 0.6545363664627075,
"soft_kl": 0.6085031032562256,
"hard_ce_nats": 1.0688358545303345,
"lr": 0.0004824093752706943,
"elapsed_s": 1476.8
},
{
"step": 1100,
"epoch": 0,
"loss": 0.6222168207168579,
"soft_kl": 0.5772915482521057,
"hard_ce_nats": 1.0265446901321411,
"lr": 0.00047780712259703394,
"elapsed_s": 1624.6
},
{
"step": 1200,
"epoch": 0,
"loss": 0.5503333210945129,
"soft_kl": 0.5080614686012268,
"hard_ce_nats": 0.9307799935340881,
"lr": 0.0004726980682330071,
"elapsed_s": 1772.3
},
{
"step": 1300,
"epoch": 0,
"loss": 0.6337389945983887,
"soft_kl": 0.5852792859077454,
"hard_ce_nats": 1.0698764324188232,
"lr": 0.0004670935782735732,
"elapsed_s": 1920.2
},
{
"step": 1400,
"epoch": 0,
"loss": 0.527277410030365,
"soft_kl": 0.4829964339733124,
"hard_ce_nats": 0.9258063435554504,
"lr": 0.00046100612100748764,
"elapsed_s": 2068.1
},
{
"step": 1500,
"epoch": 0,
"loss": 0.5628732442855835,
"soft_kl": 0.5177528262138367,
"hard_ce_nats": 0.9689571261405945,
"lr": 0.0004544492391791445,
"elapsed_s": 2215.9
},
{
"step": 1600,
"epoch": 0,
"loss": 0.5267754793167114,
"soft_kl": 0.48233577609062195,
"hard_ce_nats": 0.9267328977584839,
"lr": 0.00044743751986008154,
"elapsed_s": 2363.8
},
{
"step": 1700,
"epoch": 0,
"loss": 0.5578336715698242,
"soft_kl": 0.5065539479255676,
"hard_ce_nats": 1.0193513631820679,
"lr": 0.00043998656199717434,
"elapsed_s": 2511.7
},
{
"step": 1800,
"epoch": 0,
"loss": 0.49808722734451294,
"soft_kl": 0.4509686529636383,
"hard_ce_nats": 0.9221543073654175,
"lr": 0.00043211294170971527,
"elapsed_s": 2659.7
},
{
"step": 1900,
"epoch": 0,
"loss": 0.4913105070590973,
"soft_kl": 0.4442547559738159,
"hard_ce_nats": 0.9148122668266296,
"lr": 0.00042383417541257953,
"elapsed_s": 2807.7
},
{
"step": 2000,
"epoch": 0,
"loss": 0.481732577085495,
"soft_kl": 0.43248867988586426,
"hard_ce_nats": 0.9249278903007507,
"lr": 0.00041516868084752034,
"elapsed_s": 2955.5
},
{
"step": 2100,
"epoch": 0,
"loss": 0.47225916385650635,
"soft_kl": 0.42494431138038635,
"hard_ce_nats": 0.8980928063392639,
"lr": 0.00040613573610928476,
"elapsed_s": 3103.4
},
{
"step": 2200,
"epoch": 0,
"loss": 0.4640839695930481,
"soft_kl": 0.4157882630825043,
"hard_ce_nats": 0.8987454771995544,
"lr": 0.00039675543675770465,
"elapsed_s": 3251.2
},
{
"step": 2300,
"epoch": 0,
"loss": 0.4692152142524719,
"soft_kl": 0.42100366950035095,
"hard_ce_nats": 0.9031190872192383,
"lr": 0.0003870486511111775,
"elapsed_s": 3399.0
},
{
"step": 2400,
"epoch": 0,
"loss": 0.41351574659347534,
"soft_kl": 0.3648102879524231,
"hard_ce_nats": 0.8518648147583008,
"lr": 0.00037703697382099235,
"elapsed_s": 3547.0
},
{
"step": 2500,
"epoch": 0,
"loss": 0.4465184509754181,
"soft_kl": 0.3960680067539215,
"hard_ce_nats": 0.9005725979804993,
"lr": 0.000366742677829787,
"elapsed_s": 3695.0
},
{
"step": 2600,
"epoch": 0,
"loss": 0.3908940553665161,
"soft_kl": 0.3421657383441925,
"hard_ce_nats": 0.8294488191604614,
"lr": 0.000356188664821012,
"elapsed_s": 3843.0
},
{
"step": 2700,
"epoch": 0,
"loss": 0.4373500645160675,
"soft_kl": 0.3866843581199646,
"hard_ce_nats": 0.893341600894928,
"lr": 0.00034539841426963716,
"elapsed_s": 3990.9
},
{
"step": 2800,
"epoch": 0,
"loss": 0.37633657455444336,
"soft_kl": 0.3227885961532593,
"hard_ce_nats": 0.8582682609558105,
"lr": 0.00033439593120744815,
"elapsed_s": 4138.7
},
{
"step": 2900,
"epoch": 0,
"loss": 0.3890794813632965,
"soft_kl": 0.3356207609176636,
"hard_ce_nats": 0.8702080249786377,
"lr": 0.0003232056928191376,
"elapsed_s": 4286.6
},
{
"step": 3000,
"epoch": 0,
"loss": 0.3937709629535675,
"soft_kl": 0.3408649265766144,
"hard_ce_nats": 0.8699252009391785,
"lr": 0.00031185259398800065,
"elapsed_s": 4434.5
},
{
"step": 3100,
"epoch": 0,
"loss": 0.430860698223114,
"soft_kl": 0.37848299741744995,
"hard_ce_nats": 0.9022603631019592,
"lr": 0.0003003618919123763,
"elapsed_s": 4582.3
},
{
"step": 3200,
"epoch": 0,
"loss": 0.401819109916687,
"soft_kl": 0.34881314635276794,
"hard_ce_nats": 0.8788727521896362,
"lr": 0.00028875914991604944,
"elapsed_s": 4730.2
},
{
"step": 3300,
"epoch": 0,
"loss": 0.3766922950744629,
"soft_kl": 0.32367315888404846,
"hard_ce_nats": 0.8538645505905151,
"lr": 0.00027707018057761544,
"elapsed_s": 4878.2
},
{
"step": 3400,
"epoch": 0,
"loss": 0.36151397228240967,
"soft_kl": 0.3108021020889282,
"hard_ce_nats": 0.8179208636283875,
"lr": 0.00026532098830532906,
"elapsed_s": 5026.2
},
{
"step": 3500,
"epoch": 1,
"loss": 0.3514333963394165,
"soft_kl": 0.29760679602622986,
"hard_ce_nats": 0.8358727097511292,
"lr": 0.00025353771148519056,
"elapsed_s": 5173.7
},
{
"step": 3600,
"epoch": 1,
"loss": 0.3579716682434082,
"soft_kl": 0.3013242483139038,
"hard_ce_nats": 0.8677986860275269,
"lr": 0.0002417465643309716,
"elapsed_s": 5321.5
},
{
"step": 3700,
"epoch": 1,
"loss": 0.3588024377822876,
"soft_kl": 0.3005087375640869,
"hard_ce_nats": 0.883445680141449,
"lr": 0.0002299737785655482,
"elapsed_s": 5469.4
},
{
"step": 3800,
"epoch": 1,
"loss": 0.42988330125808716,
"soft_kl": 0.3680754601955414,
"hard_ce_nats": 0.9861539602279663,
"lr": 0.0002182455450632803,
"elapsed_s": 5617.2
},
{
"step": 3900,
"epoch": 1,
"loss": 0.356850266456604,
"soft_kl": 0.3052098751068115,
"hard_ce_nats": 0.8216138482093811,
"lr": 0.00020658795558326743,
"elapsed_s": 5765.0
},
{
"step": 4000,
"epoch": 1,
"loss": 0.3636034429073334,
"soft_kl": 0.3060402572154999,
"hard_ce_nats": 0.8816721439361572,
"lr": 0.00019502694472310563,
"elapsed_s": 5912.7
},
{
"step": 4100,
"epoch": 1,
"loss": 0.33809661865234375,
"soft_kl": 0.2819271981716156,
"hard_ce_nats": 0.8436216115951538,
"lr": 0.00018358823222228097,
"elapsed_s": 6060.4
},
{
"step": 4200,
"epoch": 1,
"loss": 0.35127002000808716,
"soft_kl": 0.29350778460502625,
"hard_ce_nats": 0.8711302280426025,
"lr": 0.0001722972657435572,
"elapsed_s": 6208.1
},
{
"step": 4300,
"epoch": 1,
"loss": 0.3270314931869507,
"soft_kl": 0.26497745513916016,
"hard_ce_nats": 0.8855178952217102,
"lr": 0.0001611791642596516,
"elapsed_s": 6355.9
},
{
"step": 4400,
"epoch": 1,
"loss": 0.345449835062027,
"soft_kl": 0.2795993387699127,
"hard_ce_nats": 0.9381043314933777,
"lr": 0.00015025866217114592,
"elapsed_s": 6503.9
},
{
"step": 4500,
"epoch": 1,
"loss": 0.3297061324119568,
"soft_kl": 0.2724095284938812,
"hard_ce_nats": 0.8453757166862488,
"lr": 0.00013956005427995423,
"elapsed_s": 6651.9
},
{
"step": 4600,
"epoch": 1,
"loss": 0.3264894485473633,
"soft_kl": 0.27202704548835754,
"hard_ce_nats": 0.8166509866714478,
"lr": 0.00012910714174076392,
"elapsed_s": 6799.9
},
{
"step": 4700,
"epoch": 1,
"loss": 0.32276904582977295,
"soft_kl": 0.26744940876960754,
"hard_ce_nats": 0.8206458687782288,
"lr": 0.00011892317911069211,
"elapsed_s": 6947.8
},
{
"step": 4800,
"epoch": 1,
"loss": 0.334042489528656,
"soft_kl": 0.2777497470378876,
"hard_ce_nats": 0.8406773209571838,
"lr": 0.00010903082261495558,
"elapsed_s": 7095.6
},
{
"step": 4900,
"epoch": 1,
"loss": 0.32100337743759155,
"soft_kl": 0.2652624845504761,
"hard_ce_nats": 0.82267165184021,
"lr": 9.945207974364768e-05,
"elapsed_s": 7243.5
},
{
"step": 5000,
"epoch": 1,
"loss": 0.2984895706176758,
"soft_kl": 0.24128122627735138,
"hard_ce_nats": 0.8133646249771118,
"lr": 9.020826029175384e-05,
"elapsed_s": 7391.4
},
{
"step": 5100,
"epoch": 1,
"loss": 0.2987797260284424,
"soft_kl": 0.23940409719944,
"hard_ce_nats": 0.8331603407859802,
"lr": 8.131992895132693e-05,
"elapsed_s": 7539.3
},
{
"step": 5200,
"epoch": 1,
"loss": 0.3038943409919739,
"soft_kl": 0.2482292503118515,
"hard_ce_nats": 0.8048803806304932,
"lr": 7.280685956129049e-05,
"elapsed_s": 7687.1
},
{
"step": 5300,
"epoch": 1,
"loss": 0.30647042393684387,
"soft_kl": 0.24950405955314636,
"hard_ce_nats": 0.8191676735877991,
"lr": 6.468799111665003e-05,
"elapsed_s": 7835.0
},
{
"step": 5400,
"epoch": 1,
"loss": 0.32233667373657227,
"soft_kl": 0.2624441385269165,
"hard_ce_nats": 0.8613694906234741,
"lr": 5.6981385634978535e-05,
"elapsed_s": 7982.9
},
{
"step": 5500,
"epoch": 1,
"loss": 0.28625717759132385,
"soft_kl": 0.2276662290096283,
"hard_ce_nats": 0.8135759234428406,
"lr": 4.970418797391063e-05,
"elapsed_s": 8130.8
},
{
"step": 5600,
"epoch": 1,
"loss": 0.2859552502632141,
"soft_kl": 0.2307804822921753,
"hard_ce_nats": 0.782528281211853,
"lr": 4.287258768903948e-05,
"elapsed_s": 8278.8
},
{
"step": 5700,
"epoch": 1,
"loss": 0.28815215826034546,
"soft_kl": 0.22420896589756012,
"hard_ce_nats": 0.8636408448219299,
"lr": 3.650178301706983e-05,
"elapsed_s": 8426.8
},
{
"step": 5800,
"epoch": 1,
"loss": 0.3050306439399719,
"soft_kl": 0.24740499258041382,
"hard_ce_nats": 0.8236616849899292,
"lr": 3.06059470643556e-05,
"elapsed_s": 8574.9
},
{
"step": 5900,
"epoch": 1,
"loss": 0.2935406267642975,
"soft_kl": 0.23650896549224854,
"hard_ce_nats": 0.8068255186080933,
"lr": 2.519819627604078e-05,
"elapsed_s": 8722.9
},
{
"step": 6000,
"epoch": 1,
"loss": 0.29356470704078674,
"soft_kl": 0.23647993803024292,
"hard_ce_nats": 0.8073275685310364,
"lr": 2.0290561255950966e-05,
"elapsed_s": 8870.7
},
{
"step": 6100,
"epoch": 1,
"loss": 0.29202860593795776,
"soft_kl": 0.23756283521652222,
"hard_ce_nats": 0.7822206020355225,
"lr": 1.5893960002151902e-05,
"elapsed_s": 9018.4
},
{
"step": 6200,
"epoch": 1,
"loss": 0.29682308435440063,
"soft_kl": 0.23766516149044037,
"hard_ce_nats": 0.8292444944381714,
"lr": 1.201817361771837e-05,
"elapsed_s": 9166.2
},
{
"step": 6300,
"epoch": 1,
"loss": 0.32862019538879395,
"soft_kl": 0.2723996341228485,
"hard_ce_nats": 0.8346050977706909,
"lr": 8.671824550749163e-06,
"elapsed_s": 9314.1
},
{
"step": 6400,
"epoch": 1,
"loss": 0.3259643018245697,
"soft_kl": 0.2701769173145294,
"hard_ce_nats": 0.8280507922172546,
"lr": 5.862357412037667e-06,
"elapsed_s": 9462.2
},
{
"step": 6500,
"epoch": 1,
"loss": 0.28425124287605286,
"soft_kl": 0.22566840052604675,
"hard_ce_nats": 0.8114967942237854,
"lr": 3.5960224130728858e-06,
"elapsed_s": 9610.2
},
{
"step": 6600,
"epoch": 1,
"loss": 0.2953123450279236,
"soft_kl": 0.23742923140525818,
"hard_ce_nats": 0.8162602782249451,
"lr": 1.8778614612162404e-06,
"elapsed_s": 9758.1
},
{
"step": 6700,
"epoch": 1,
"loss": 0.2767546772956848,
"soft_kl": 0.22160018980503082,
"hard_ce_nats": 0.7731451988220215,
"lr": 7.116969429883935e-07,
"elapsed_s": 9906.0
},
{
"step": 6800,
"epoch": 1,
"loss": 0.2709920406341553,
"soft_kl": 0.21164101362228394,
"hard_ce_nats": 0.8051513433456421,
"lr": 1.0012322041960675e-07,
"elapsed_s": 10053.9
},
{
"step": 6859,
"epoch": 1,
"loss": 0.2971874177455902,
"soft_kl": 0.23709337413311005,
"hard_ce_nats": 0.8380337953567505,
"lr": 2.781386176931022e-11,
"elapsed_s": 10140.8
}
]
}