PolyPythias
- Preview • Updated • 1.2k
EleutherAI/pile-preshuffled-seeds
Updated • 988 • 1Note Training data information for each seed.
-
EleutherAI/pythia-14m-deduped
Text Generation • 39.2M • Updated • 4.74k • 32 -
EleutherAI/pythia-14m-seed1
Text Generation • Updated • 3.93k • 1 -
EleutherAI/pythia-14m-seed2
Text Generation • Updated • 3.61k -
EleutherAI/pythia-14m-seed3
Text Generation • Updated • 2.82k -
EleutherAI/pythia-14m-seed4
Text Generation • Updated • 2.61k -
EleutherAI/pythia-14m-seed5
Text Generation • Updated • 2.78k -
EleutherAI/pythia-14m-seed6
Text Generation • Updated • 2k -
EleutherAI/pythia-14m-seed7
Text Generation • Updated • 2.02k -
EleutherAI/pythia-14m-seed8
Text Generation • Updated • 1.8k -
EleutherAI/pythia-14m-seed9
Text Generation • Updated • 2.01k -
EleutherAI/pythia-31m-deduped
Text Generation • 55.7M • Updated • 1.93k • 5 -
EleutherAI/pythia-31m-seed1
Text Generation • Updated • 2.95k -
EleutherAI/pythia-31m-seed2
Text Generation • Updated • 2.64k -
EleutherAI/pythia-31m-seed3
Text Generation • Updated • 2.48k -
EleutherAI/pythia-31m-seed4
Text Generation • Updated • 2.06k -
EleutherAI/pythia-31m-seed5
Text Generation • Updated • 1.97k -
EleutherAI/pythia-31m-seed6
Text Generation • Updated • 2k -
EleutherAI/pythia-31m-seed7
Text Generation • Updated • 1.76k -
EleutherAI/pythia-31m-seed8
Text Generation • Updated • 1.77k -
EleutherAI/pythia-31m-seed9
Text Generation • Updated • 1.82k -
EleutherAI/pythia-70m
95.6M • Updated • 246k • 88 -
EleutherAI/pythia-70m-seed1
Text Generation • Updated • 6.75k -
EleutherAI/pythia-70m-seed2
Text Generation • Updated • 5.02k -
EleutherAI/pythia-70m-seed3
Text Generation • Updated • 3.87k -
EleutherAI/pythia-70m-seed4
Text Generation • Updated • 3.2k -
EleutherAI/pythia-70m-seed5
Text Generation • Updated • 2.78k -
EleutherAI/pythia-70m-seed6
Text Generation • Updated • 2.93k -
EleutherAI/pythia-70m-seed7
Text Generation • Updated • 2.23k -
EleutherAI/pythia-70m-seed8
Text Generation • Updated • 2.03k -
EleutherAI/pythia-70m-seed9
Text Generation • Updated • 2.22k -
EleutherAI/pythia-160m
Text Generation • 0.2B • Updated • 3.32M • 66 -
EleutherAI/pythia-160m-seed1
Text Generation • 0.2B • Updated • 32.3k -
EleutherAI/pythia-160m-seed2
Text Generation • 0.2B • Updated • 10.5k -
EleutherAI/pythia-160m-seed3
Text Generation • 0.2B • Updated • 8.89k -
EleutherAI/pythia-160m-seed4
Text Generation • Updated • 3.35k • 1 -
EleutherAI/pythia-160m-seed5
Text Generation • Updated • 3.08k -
EleutherAI/pythia-160m-seed6
Text Generation • Updated • 2.26k -
EleutherAI/pythia-160m-seed7
Text Generation • Updated • 2.41k -
EleutherAI/pythia-160m-seed8
Text Generation • Updated • 2.24k -
EleutherAI/pythia-160m-seed9
Text Generation • Updated • 2.62k -
EleutherAI/pythia-410m
Text Generation • 0.5B • Updated • 100k • 42 -
EleutherAI/pythia-410m-seed1
Text Generation • Updated • 4.48k -
EleutherAI/pythia-410m-seed2
Text Generation • Updated • 3.52k -
EleutherAI/pythia-410m-seed3
Text Generation • Updated • 3.46k -
EleutherAI/pythia-410m-seed4
Text Generation • Updated • 3.15k -
EleutherAI/pythia-410m-seed5
Text Generation • Updated • 2.62k -
EleutherAI/pythia-410m-seed6
Text Generation • Updated • 2.27k • 1 -
EleutherAI/pythia-410m-seed7
Text Generation • Updated • 2.33k -
EleutherAI/pythia-410m-seed8
Text Generation • Updated • 1.96k -
EleutherAI/pythia-410m-seed9
Text Generation • Updated • 2.21k
EleutherAI/pythia-160m-data-seed1
Text Generation • Updated • 2.35kNote 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 • 2.09kNote 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 • 1.96kNote 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 • 2.2kNote 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 • 2.05kNote 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 • 1.41kNote 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 • 1.28kNote Corrected weight-seed 1: independent initialization with the original Pythia data order.
EleutherAI/pythia-160m-weight-seed2-corrected
Text Generation • Updated • 1.08kNote Corrected weight-seed 1: independent initialization with the original Pythia data order.