PolyPythias - a EleutherAI Collection

EleutherAI 's Collections

Pythia Scaling Suite

Quirky Models and Datasets

Pile-T5

Quirky Math (archive)

Pile-T5 T5x Checkpoints

Sparse Autoencoders

PolyPythias

updated 10 days ago

EleutherAI/polypythias-evals

Preview • Updated Sep 11, 2024 • 63
EleutherAI/pile-preshuffled-seeds

Updated Feb 27 • 100 • 1

Note Training data information for each seed.
EleutherAI/pythia-14m

Text Generation • Updated Jul 26, 2023 • 209k • 22
EleutherAI/pythia-14m-seed1

Updated Dec 13, 2023 • 16.6k
EleutherAI/pythia-14m-seed2

Updated Dec 13, 2023 • 1.05k
EleutherAI/pythia-14m-seed3

Updated Dec 18, 2023 • 1.01k
EleutherAI/pythia-14m-seed4

Updated Dec 18, 2023 • 974
EleutherAI/pythia-14m-seed5

Updated Jan 10, 2024 • 889
EleutherAI/pythia-14m-seed6

Updated Jan 11, 2024 • 105
EleutherAI/pythia-14m-seed7

Updated Jan 26, 2024 • 104
EleutherAI/pythia-14m-seed8

Updated Jan 26, 2024 • 103
EleutherAI/pythia-14m-seed9

Updated Jan 26, 2024 • 103
EleutherAI/pythia-31m

Text Generation • Updated Jul 26, 2023 • 95.3k • 5
EleutherAI/pythia-31m-seed1

Updated Jan 13, 2024 • 2.82k
EleutherAI/pythia-31m-seed2

Updated Jan 26, 2024 • 531
EleutherAI/pythia-31m-seed3

Updated Jan 26, 2024 • 517
EleutherAI/pythia-31m-seed4

Updated Jan 26, 2024 • 511
EleutherAI/pythia-31m-seed5

Updated Jan 26, 2024 • 511
EleutherAI/pythia-31m-seed6

Updated Jan 26, 2024 • 50
EleutherAI/pythia-31m-seed7

Updated Jan 26, 2024 • 51
EleutherAI/pythia-31m-seed8

Updated Jan 26, 2024 • 51
EleutherAI/pythia-31m-seed9

Updated Jan 26, 2024 • 51
EleutherAI/pythia-70m

Updated Nov 21, 2023 • 133k • 67
EleutherAI/pythia-70m-seed1

Updated Nov 8, 2023 • 5.98k
EleutherAI/pythia-70m-seed2

Updated Nov 13, 2023 • 587
EleutherAI/pythia-70m-seed3

Updated Nov 14, 2023 • 579
EleutherAI/pythia-70m-seed4

Updated Nov 15, 2023 • 575
EleutherAI/pythia-70m-seed5

Updated Jan 13, 2024 • 579
EleutherAI/pythia-70m-seed6

Updated Jan 13, 2024 • 113
EleutherAI/pythia-70m-seed7

Updated Jan 13, 2024 • 107
EleutherAI/pythia-70m-seed8

Updated Jan 14, 2024 • 101
EleutherAI/pythia-70m-seed9

Updated Jan 14, 2024 • 100
EleutherAI/pythia-160m

Text Generation • Updated Jul 9, 2023 • 164k • 31
EleutherAI/pythia-160m-seed1

Text Generation • Updated Jan 4, 2024 • 3.09k
EleutherAI/pythia-160m-seed2

Text Generation • Updated Jan 4, 2024 • 1.15k
EleutherAI/pythia-160m-seed3

Text Generation • Updated Jan 4, 2024 • 1.08k
EleutherAI/pythia-160m-seed4

Updated Nov 8, 2023 • 430 • 1
EleutherAI/pythia-160m-seed5

Updated Jan 14, 2024 • 391
EleutherAI/pythia-160m-seed6

Updated Jan 14, 2024 • 10
EleutherAI/pythia-160m-seed7

Updated Jan 15, 2024 • 8
EleutherAI/pythia-160m-seed8

Updated Jan 16, 2024 • 8
EleutherAI/pythia-160m-seed9

Updated Jan 19, 2024 • 8
EleutherAI/pythia-410m

Text Generation • Updated Jul 9, 2023 • 91.5k • 24
EleutherAI/pythia-410m-seed1

Updated Nov 28, 2023 • 1.62k
EleutherAI/pythia-410m-seed2

Updated Nov 29, 2023 • 3
EleutherAI/pythia-410m-seed3

Updated Jan 24, 2024 • 3
EleutherAI/pythia-410m-seed4

Updated Jan 25, 2024
EleutherAI/pythia-410m-seed5

Updated Feb 4, 2024
EleutherAI/pythia-410m-seed6

Updated Jan 29, 2024
EleutherAI/pythia-410m-seed7

Updated Feb 4, 2024
EleutherAI/pythia-410m-seed8

Updated Feb 4, 2024
EleutherAI/pythia-410m-seed9

Updated Feb 4, 2024
EleutherAI/pythia-160m-data-seed1

Updated Jan 22, 2024

Note 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

Updated Jan 22, 2024

Note 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

Updated Jan 24, 2024

Note 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

Updated Jan 16, 2024

Note 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

Updated Jan 17, 2024

Note 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

Updated Jan 17, 2024

Note 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 Mar 12