thegovind's picture
Publish the weight-space programming labs
bb102c0 verified
Raw History Blame Contribute Delete
3.42 kB
{
"tokens": 8192,
"last": 0,
"models": {
"base": {
"wikitext_ppl": 22.803856595283676,
"wikitext_kl": 0.0,
"kl_productive_texts": 0.0,
"kl_looping_texts": 0.0,
"ppl_change_pct": 0.0
},
"a-4": {
"wikitext_ppl": 23.98091302460733,
"wikitext_kl": 0.01805390289518982,
"kl_productive_texts": 0.005575988732744008,
"kl_looping_texts": 0.014664811082184315,
"ppl_change_pct": 5.161655110421504
},
"a1": {
"wikitext_ppl": 22.682976168143394,
"wikitext_kl": 0.0014809382846578956,
"kl_productive_texts": 0.0009527496004011482,
"kl_looping_texts": 0.001375974272377789,
"ppl_change_pct": -0.5300876482677186
},
"a12": {
"wikitext_ppl": 25.587845017847215,
"wikitext_kl": 0.1536046238616109,
"kl_productive_texts": 0.03660591132938862,
"kl_looping_texts": 0.10268476884812117,
"ppl_change_pct": 12.208410498158129
},
"a2": {
"wikitext_ppl": 22.608826945665015,
"wikitext_kl": 0.003975428204284981,
"kl_productive_texts": 0.0017778984911274165,
"kl_looping_texts": 0.003459347557509318,
"ppl_change_pct": -0.8552485357191664
},
"a4": {
"wikitext_ppl": 22.623106485694798,
"wikitext_kl": 0.013635404873639345,
"kl_productive_texts": 0.004846438008826226,
"kl_looping_texts": 0.01119942613877356,
"ppl_change_pct": -0.7926295661158544
},
"a8": {
"wikitext_ppl": 23.358480009959,
"wikitext_kl": 0.057661350117996335,
"kl_productive_texts": 0.016448898473754525,
"kl_looping_texts": 0.044161050114780664,
"ppl_change_pct": 2.432147441192156
},
"all1": {
"wikitext_ppl": 22.71757411771847,
"wikitext_kl": 0.002297598111908883,
"kl_productive_texts": 0.0022679113026242703,
"kl_looping_texts": 0.002893365715863183,
"ppl_change_pct": -0.37836791862236874
},
"all2": {
"wikitext_ppl": 22.701486659740766,
"wikitext_kl": 0.00710437580710277,
"kl_productive_texts": 0.006340900203213096,
"kl_looping_texts": 0.009188366238959134,
"ppl_change_pct": -0.44891501187602767
},
"l15only": {
"wikitext_ppl": 22.679742388993002,
"wikitext_kl": 0.031401047366671264,
"kl_productive_texts": 0.005502430256456137,
"kl_looping_texts": 0.019375510746613145,
"ppl_change_pct": -0.544268491481148
},
"l19only": {
"wikitext_ppl": 23.138431346869368,
"wikitext_kl": 0.008271073515061289,
"kl_productive_texts": 0.004639745224267244,
"kl_looping_texts": 0.007545338477939367,
"ppl_change_pct": 1.4671849482463806
},
"p50c": {
"wikitext_ppl": 22.84867246335959,
"wikitext_kl": 0.030335541581735015,
"kl_productive_texts": 0.00966530735604465,
"kl_looping_texts": 0.024051173124462366,
"ppl_change_pct": 0.19652758246684332
},
"randcols": {
"wikitext_ppl": 22.73536954010332,
"wikitext_kl": 0.0031150824506767094,
"kl_productive_texts": 0.00445548549760133,
"kl_looping_texts": 0.00383356906240806,
"ppl_change_pct": -0.30033102029995984
},
"randdir": {
"wikitext_ppl": 22.867608492272986,
"wikitext_kl": 0.004527778772171587,
"kl_productive_texts": 0.004789948638062924,
"kl_looping_texts": 0.0038698724820278585,
"ppl_change_pct": 0.27956629495071716
},
"rebuilt": {
"wikitext_ppl": 22.851007537942113,
"wikitext_kl": 0.030349059496074915,
"kl_productive_texts": 0.00961088144686073,
"kl_looping_texts": 0.02383891143836081,
"ppl_change_pct": 0.2067674055983515
}
}
}