PolyPythias
- Preview • Updated • 1.28k
EleutherAI/pile-preshuffled-seeds
Updated • 957 • 1Note Training data information for each seed.
-
EleutherAI/pythia-14m-deduped
Text Generation • 39.2M • Updated • 5.17k • 32 -
EleutherAI/pythia-14m-seed1
Text Generation • Updated • 4.09k • 1 -
EleutherAI/pythia-14m-seed2
Text Generation • Updated • 3.83k -
EleutherAI/pythia-14m-seed3
Text Generation • Updated • 3k -
EleutherAI/pythia-14m-seed4
Text Generation • Updated • 2.81k -
EleutherAI/pythia-14m-seed5
Text Generation • Updated • 2.97k -
EleutherAI/pythia-14m-seed6
Text Generation • Updated • 2.2k -
EleutherAI/pythia-14m-seed7
Text Generation • Updated • 2.25k -
EleutherAI/pythia-14m-seed8
Text Generation • Updated • 2.02k -
EleutherAI/pythia-14m-seed9
Text Generation • Updated • 2.25k -
EleutherAI/pythia-31m-deduped
Text Generation • 55.7M • Updated • 1.97k • 5 -
EleutherAI/pythia-31m-seed1
Text Generation • Updated • 3.1k -
EleutherAI/pythia-31m-seed2
Text Generation • Updated • 2.79k -
EleutherAI/pythia-31m-seed3
Text Generation • Updated • 2.57k -
EleutherAI/pythia-31m-seed4
Text Generation • Updated • 2.07k -
EleutherAI/pythia-31m-seed5
Text Generation • Updated • 1.96k -
EleutherAI/pythia-31m-seed6
Text Generation • Updated • 1.98k -
EleutherAI/pythia-31m-seed7
Text Generation • Updated • 1.85k -
EleutherAI/pythia-31m-seed8
Text Generation • Updated • 1.86k -
EleutherAI/pythia-31m-seed9
Text Generation • Updated • 1.83k -
EleutherAI/pythia-70m
95.6M • Updated • 252k • 89 -
EleutherAI/pythia-70m-seed1
Text Generation • Updated • 7.08k -
EleutherAI/pythia-70m-seed2
Text Generation • Updated • 5.32k -
EleutherAI/pythia-70m-seed3
Text Generation • Updated • 4.1k -
EleutherAI/pythia-70m-seed4
Text Generation • Updated • 3.35k -
EleutherAI/pythia-70m-seed5
Text Generation • Updated • 2.95k -
EleutherAI/pythia-70m-seed6
Text Generation • Updated • 3.03k -
EleutherAI/pythia-70m-seed7
Text Generation • Updated • 2.31k -
EleutherAI/pythia-70m-seed8
Text Generation • Updated • 2.19k -
EleutherAI/pythia-70m-seed9
Text Generation • Updated • 2.32k -
EleutherAI/pythia-160m
Text Generation • 0.2B • Updated • 2.93M • 67 -
EleutherAI/pythia-160m-seed1
Text Generation • 0.2B • Updated • 31.6k -
EleutherAI/pythia-160m-seed2
Text Generation • 0.2B • Updated • 9.22k -
EleutherAI/pythia-160m-seed3
Text Generation • 0.2B • Updated • 7.69k -
EleutherAI/pythia-160m-seed4
Text Generation • Updated • 3.55k • 1 -
EleutherAI/pythia-160m-seed5
Text Generation • Updated • 3.16k -
EleutherAI/pythia-160m-seed6
Text Generation • Updated • 2.39k -
EleutherAI/pythia-160m-seed7
Text Generation • Updated • 2.5k -
EleutherAI/pythia-160m-seed8
Text Generation • Updated • 2.34k -
EleutherAI/pythia-160m-seed9
Text Generation • Updated • 2.76k -
EleutherAI/pythia-410m
Text Generation • 0.5B • Updated • 133k • 43 -
EleutherAI/pythia-410m-seed1
Text Generation • Updated • 4.43k -
EleutherAI/pythia-410m-seed2
Text Generation • Updated • 3.41k -
EleutherAI/pythia-410m-seed3
Text Generation • Updated • 3.34k -
EleutherAI/pythia-410m-seed4
Text Generation • Updated • 2.99k -
EleutherAI/pythia-410m-seed5
Text Generation • Updated • 2.45k -
EleutherAI/pythia-410m-seed6
Text Generation • Updated • 2.15k • 1 -
EleutherAI/pythia-410m-seed7
Text Generation • Updated • 2.27k -
EleutherAI/pythia-410m-seed8
Text Generation • Updated • 1.86k -
EleutherAI/pythia-410m-seed9
Text Generation • Updated • 2.12k
EleutherAI/pythia-160m-data-seed1
Text Generation • Updated • 2.49kNote 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 • 2.2kNote 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 • 2.04kNote 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 • 2.22kNote 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 • 2.09kNote 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 • 1.46kNote 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 • 1.53kNote Corrected weight-seed 1: independent initialization with the original Pythia data order.
EleutherAI/pythia-160m-weight-seed2-corrected
Text Generation • Updated • 1.22kNote Corrected weight-seed 1: independent initialization with the original Pythia data order.