PolyPythias
-
Preview • Updated • 97
EleutherAI/pile-preshuffled-seeds
Updated • 289 • 1Note Training data information for each seed.
-
EleutherAI/pythia-14m-deduped
Text Generation • 39.2M • Updated • 57k • 28 -
EleutherAI/pythia-14m-seed1
Text Generation • Updated • 121 -
EleutherAI/pythia-14m-seed2
Text Generation • Updated • 220 -
EleutherAI/pythia-14m-seed3
Text Generation • Updated • 115 -
EleutherAI/pythia-14m-seed4
Text Generation • Updated • 108 -
EleutherAI/pythia-14m-seed5
Text Generation • Updated • 145 -
EleutherAI/pythia-14m-seed6
Text Generation • Updated • 135 -
EleutherAI/pythia-14m-seed7
Text Generation • Updated • 145 -
EleutherAI/pythia-14m-seed8
Text Generation • Updated • 140 -
EleutherAI/pythia-14m-seed9
Text Generation • Updated • 137 -
EleutherAI/pythia-31m-deduped
Text Generation • 55.7M • Updated • 3.32k • 5 -
EleutherAI/pythia-31m-seed1
Text Generation • Updated • 113 -
EleutherAI/pythia-31m-seed2
Text Generation • Updated • 114 -
EleutherAI/pythia-31m-seed3
Text Generation • Updated • 105 -
EleutherAI/pythia-31m-seed4
Text Generation • Updated • 100 -
EleutherAI/pythia-31m-seed5
Text Generation • Updated • 106 -
EleutherAI/pythia-31m-seed6
Text Generation • Updated • 112 -
EleutherAI/pythia-31m-seed7
Text Generation • Updated • 110 -
EleutherAI/pythia-31m-seed8
Text Generation • Updated • 106 -
EleutherAI/pythia-31m-seed9
Text Generation • Updated • 108 -
EleutherAI/pythia-70m
95.6M • Updated • 77.9k • 78 -
EleutherAI/pythia-70m-seed1
Text Generation • Updated • 895 -
EleutherAI/pythia-70m-seed2
Text Generation • Updated • 574 -
EleutherAI/pythia-70m-seed3
Text Generation • Updated • 557 -
EleutherAI/pythia-70m-seed4
Text Generation • Updated • 526 -
EleutherAI/pythia-70m-seed5
Text Generation • Updated • 552 -
EleutherAI/pythia-70m-seed6
Text Generation • Updated • 527 -
EleutherAI/pythia-70m-seed7
Text Generation • Updated • 531 -
EleutherAI/pythia-70m-seed8
Text Generation • Updated • 531 -
EleutherAI/pythia-70m-seed9
Text Generation • Updated • 519 -
EleutherAI/pythia-160m
Text Generation • Updated • 1.18M • 38 -
EleutherAI/pythia-160m-seed1
Text Generation • 0.2B • Updated • 975 -
EleutherAI/pythia-160m-seed2
Text Generation • 0.2B • Updated • 924 -
EleutherAI/pythia-160m-seed3
Text Generation • 0.2B • Updated • 919 -
EleutherAI/pythia-160m-seed4
Text Generation • Updated • 849 • 1 -
EleutherAI/pythia-160m-seed5
Text Generation • Updated • 608 -
EleutherAI/pythia-160m-seed6
Text Generation • Updated • 593 -
EleutherAI/pythia-160m-seed7
Text Generation • Updated • 587 -
EleutherAI/pythia-160m-seed8
Text Generation • Updated • 594 -
EleutherAI/pythia-160m-seed9
Text Generation • Updated • 579 -
EleutherAI/pythia-410m
Text Generation • 0.5B • Updated • 66.3k • 36 -
EleutherAI/pythia-410m-seed1
Text Generation • Updated • 651 -
EleutherAI/pythia-410m-seed2
Text Generation • Updated • 1.26k -
EleutherAI/pythia-410m-seed3
Text Generation • Updated • 611 -
EleutherAI/pythia-410m-seed4
Text Generation • Updated • 572 -
EleutherAI/pythia-410m-seed5
Text Generation • Updated • 531 -
EleutherAI/pythia-410m-seed6
Text Generation • Updated • 856 • 1 -
EleutherAI/pythia-410m-seed7
Text Generation • Updated • 538 -
EleutherAI/pythia-410m-seed8
Text Generation • Updated • 573 -
EleutherAI/pythia-410m-seed9
Text Generation • Updated • 714
EleutherAI/pythia-160m-data-seed1
Text Generation • Updated • 139Note Version where the data order and weight initialization seeds are decoupled. Here, only the data seed is different from pythia-160m ("seed 0").
EleutherAI/pythia-160m-data-seed2
Text Generation • Updated • 143Note Version where the data order and weight initialization seeds are decoupled. Here, only the data seed is different from pythia-160m ("seed 0").
EleutherAI/pythia-160m-data-seed3
Text Generation • Updated • 146Note Version where the data order and weight initialization seeds are decoupled. Here, only the data seed is different from pythia-160m ("seed 0").
EleutherAI/pythia-160m-weight-seed1
Text Generation • Updated • 246Note Version where the data order and weight initialization seeds are decoupled. Here, only the weight initialization seed is different from pythia-160m ("seed 0").
EleutherAI/pythia-160m-weight-seed2
Text Generation • Updated • 212Note Version where the data order and weight initialization seeds are decoupled. Here, only the weight initialization seed is different from pythia-160m ("seed 0").
EleutherAI/pythia-160m-weight-seed3
Text Generation • Updated • 203Note Version where the data order and weight initialization seeds are decoupled. Here, only the weight initialization seed is different from pythia-160m ("seed 0").
-
PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training Runs
Paper • 2503.09543 • Published