PolyPythias
- Preview • Updated • 614
EleutherAI/pile-preshuffled-seeds
Updated • 919 • 1Note Training data information for each seed.
-
EleutherAI/pythia-14m-deduped
Text Generation • 39.2M • Updated • 4.67k • 32 -
EleutherAI/pythia-14m-seed1
Text Generation • Updated • 1.9k • 1 -
EleutherAI/pythia-14m-seed2
Text Generation • Updated • 1.59k -
EleutherAI/pythia-14m-seed3
Text Generation • Updated • 1.44k -
EleutherAI/pythia-14m-seed4
Text Generation • Updated • 1.33k -
EleutherAI/pythia-14m-seed5
Text Generation • Updated • 1.38k -
EleutherAI/pythia-14m-seed6
Text Generation • Updated • 1.33k -
EleutherAI/pythia-14m-seed7
Text Generation • Updated • 1.26k -
EleutherAI/pythia-14m-seed8
Text Generation • Updated • 1.24k -
EleutherAI/pythia-14m-seed9
Text Generation • Updated • 1.28k -
EleutherAI/pythia-31m-deduped
Text Generation • 55.7M • Updated • 1.76k • 5 -
EleutherAI/pythia-31m-seed1
Text Generation • Updated • 1.59k -
EleutherAI/pythia-31m-seed2
Text Generation • Updated • 1.42k -
EleutherAI/pythia-31m-seed3
Text Generation • Updated • 1.4k -
EleutherAI/pythia-31m-seed4
Text Generation • Updated • 1.33k -
EleutherAI/pythia-31m-seed5
Text Generation • Updated • 1.36k -
EleutherAI/pythia-31m-seed6
Text Generation • Updated • 1.33k -
EleutherAI/pythia-31m-seed7
Text Generation • Updated • 1.2k -
EleutherAI/pythia-31m-seed8
Text Generation • Updated • 1.23k -
EleutherAI/pythia-31m-seed9
Text Generation • Updated • 1.23k -
EleutherAI/pythia-70m
95.6M • Updated • 176k • 87 -
EleutherAI/pythia-70m-seed1
Text Generation • Updated • 3.37k -
EleutherAI/pythia-70m-seed2
Text Generation • Updated • 3.06k -
EleutherAI/pythia-70m-seed3
Text Generation • Updated • 2.59k -
EleutherAI/pythia-70m-seed4
Text Generation • Updated • 2.19k -
EleutherAI/pythia-70m-seed5
Text Generation • Updated • 1.72k -
EleutherAI/pythia-70m-seed6
Text Generation • Updated • 1.9k -
EleutherAI/pythia-70m-seed7
Text Generation • Updated • 1.76k -
EleutherAI/pythia-70m-seed8
Text Generation • Updated • 1.56k -
EleutherAI/pythia-70m-seed9
Text Generation • Updated • 1.86k -
EleutherAI/pythia-160m
Text Generation • 0.2B • Updated • 3.4M • 66 -
EleutherAI/pythia-160m-seed1
Text Generation • 0.2B • Updated • 16.2k -
EleutherAI/pythia-160m-seed2
Text Generation • 0.2B • Updated • 15.4k -
EleutherAI/pythia-160m-seed3
Text Generation • 0.2B • Updated • 15k -
EleutherAI/pythia-160m-seed4
Text Generation • Updated • 2.92k • 1 -
EleutherAI/pythia-160m-seed5
Text Generation • Updated • 2.61k -
EleutherAI/pythia-160m-seed6
Text Generation • Updated • 2.08k -
EleutherAI/pythia-160m-seed7
Text Generation • Updated • 2.09k -
EleutherAI/pythia-160m-seed8
Text Generation • Updated • 2.09k -
EleutherAI/pythia-160m-seed9
Text Generation • Updated • 2.09k -
EleutherAI/pythia-410m
Text Generation • 0.5B • Updated • 109k • 41 -
EleutherAI/pythia-410m-seed1
Text Generation • Updated • 3.52k -
EleutherAI/pythia-410m-seed2
Text Generation • Updated • 2.86k -
EleutherAI/pythia-410m-seed3
Text Generation • Updated • 3.06k -
EleutherAI/pythia-410m-seed4
Text Generation • Updated • 2.82k -
EleutherAI/pythia-410m-seed5
Text Generation • Updated • 2.21k -
EleutherAI/pythia-410m-seed6
Text Generation • Updated • 2.25k • 1 -
EleutherAI/pythia-410m-seed7
Text Generation • Updated • 1.98k -
EleutherAI/pythia-410m-seed8
Text Generation • Updated • 1.98k -
EleutherAI/pythia-410m-seed9
Text Generation • Updated • 1.97k
EleutherAI/pythia-160m-data-seed1
Text Generation • Updated • 1.29kNote 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 • 1.13kNote 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 • 1.11kNote 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 • 1.81kNote 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 • 1.72kNote 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 • 1.23kNote 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
EleutherAI/pythia-160m-weight-seed1-corrected
Text Generation • Updated • 234Note Corrected weight-seed 1: independent initialization with the original Pythia data order.
EleutherAI/pythia-160m-weight-seed2-corrected
Text Generation • Updated • 542Note Corrected weight-seed 1: independent initialization with the original Pythia data order.