All pull requests

Pull requests

• fix: correct logic bug, f-string typo, and bare except clauses

  #413·Ricardo-M-L opened Aug 17, 2026·

  ·

• Startup: add argument-consistency checks & summary table (Fixes #124)

  #409·MagellaX opened Jun 20, 2025·

  ·
  1

• fix(training): correct rank-zero log messages, Print total model size once at startup (rank-0) – Fixes #123

  #408·MagellaX opened Jun 20, 2025·

  ·

• Bump black from 21.4b0 to 24.3.0

  dependencies
  #402·dependabot[bot] opened Mar 20, 2024Bot·

  ·

• Add xPos embeddings

  #370·janEbert opened Mar 7, 2023Collaborator·

  ·

• Fix various small problems

  #367·janEbert opened Feb 28, 2023Collaborator·

  ·

• Bloom model training with AML

  #365·savitamittal1 opened Feb 21, 2023·

  ·

• Add UL2 data sampling and pretraining

  #358·janEbert opened Dec 13, 2022Collaborator·

  ·
  5

• Add FlashAttention

  #357·nmntz opened Dec 12, 2022·

  ·
  3

• Distill BLOOM - tentative 2

  #354·younesbelkada opened Oct 11, 2022·

  ·
  4

• Enable rocm-support

  #353·luukkonenr opened Oct 7, 2022·

  ·

• Encoding checkpoint reshaping guide

  #349·tjruwase opened Sep 20, 2022Collaborator·

  ·
  4

• Add multiple evaluation compat

  #336·Muennighoff opened Aug 30, 2022Collaborator·

  ·

• [checkpoints] replace bf16 with fp32 checkpoint weights

  #327·stas00 opened Aug 10, 2022Contributor·

  ·
  5

• Prefix LM Eval

  #313·Muennighoff opened Jul 16, 2022Collaborator·

  ·
  4

• Add Bitfit

  #311·Muennighoff opened Jul 10, 2022Collaborator·

  ·

• Tool for CKPT averaging

  #310·Muennighoff opened Jul 10, 2022Collaborator·

  ·

• Enable loading ckpt for t0 finetuning

  #309·Muennighoff opened Jul 10, 2022Collaborator·

  ·
  32

• [WIP] Hack my way to get OPT running

  #301·thomasw21 opened Jul 4, 2022Member·

  ·

• [MLM] Train script for non causal decoder

  #300·thomasw21 opened Jul 4, 2022Member·

  ·

• a branch combining layer-norm-auto-sync and ds_ckpt_reshape

  #292·stas00 opened Jun 29, 2022Contributor·

  ·

• BigScience Eval Harness

  #291·Muennighoff opened Jun 29, 2022Collaborator·

  ·

• No-ZeRO reshaping

  #289·Muennighoff opened Jun 23, 2022Collaborator·

  ·

• WIP: Shared t5 code

  #286·thomasw21 opened Jun 21, 2022Member·

  ·

• [WIP] add debug utils

  #275·stas00 opened Mar 28, 2022Contributor·

  ·