Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Aug 1, 2025

• Improve trainer doc (#3818)

  qgallouedec

  authored
  072d7ddView commit details

  Copy full SHA for 072d7dd

  Browse repository at this point

• Performance optimization: Replace list comprehensions with tensor operations in BCO and KTO trainers (#3813)

  Show description for ead5aaf
  

  chi2liu

  and

  chiliu

  authored
  ead5aafView commit details

  Copy full SHA for ead5aaf

  Browse repository at this point

Commits on Jul 31, 2025

• fix CI docs and grpo slow test (#3814)

  Show description for dbbc770
  

  kashif

  and

  qgallouedec

  authored
  dbbc770View commit details

  Copy full SHA for dbbc770

  Browse repository at this point

• Fix citation

  qgallouedec

  committed
  294e8cbView commit details

  Copy full SHA for 294e8cb

  Browse repository at this point

• GSPO parameters update from v2 (#3798)

  Show description for 79c5797
  

  3 peopleauthored
  79c5797View commit details

  Copy full SHA for 79c5797

  Browse repository at this point

• add xpu support for mergekit (#3800)

  Show description for ab24000
  yao-matrix

  authored
  ab24000View commit details

  Copy full SHA for ab24000

  Browse repository at this point

• Add GSPO script examples (VLM/LLM) (#3810)

  sergiopaniego

  authored
  3ae60cdView commit details

  Copy full SHA for 3ae60cd

  Browse repository at this point

Commits on Jul 30, 2025

• Correction parameter description (#3803)

  Show description for 9a1e6a4
  

  3 peopleauthored
  9a1e6a4View commit details

  Copy full SHA for 9a1e6a4

  Browse repository at this point

• Add vLLM transformers backend to online methods (#3773)

  Show description for 90c7876
  

  3 peopleauthored
  90c7876View commit details

  Copy full SHA for 90c7876

  Browse repository at this point

• Examples list updated in docs (#3806)

  sergiopaniego

  authored
  72bbc6dView commit details

  Copy full SHA for 72bbc6d

  Browse repository at this point

• 🐙 Add MPO VLM example script (#3799)

  sergiopaniego

  authored
  25ce0f3View commit details

  Copy full SHA for 25ce0f3

  Browse repository at this point

Commits on Jul 29, 2025

• Fix broken PEFT+TRL docs link in `using_llama_models.md` (#3794)

  Show description for 9269f9f
  

  bwook00

  and

  kashif

  authored
  9269f9fView commit details

  Copy full SHA for 9269f9f

  Browse repository at this point

• ⬆️ Bump dev version (#3793)

  qgallouedec

  authored
  eb5d0feView commit details

  Copy full SHA for eb5d0fe

  Browse repository at this point

• Release: v0.20 (#3792)

  qgallouedec

  authored
  30576d2View commit details

  Copy full SHA for 30576d2

  Browse repository at this point

• 👐 FSDP2+GRPO (#3687)

  Show description for 5522cc0
  

  4 peopleauthored
  5522cc0View commit details

  Copy full SHA for 5522cc0

  Browse repository at this point

Commits on Jul 28, 2025

• 📘 SFT doc rewrite (#3619)

  Show description for 303d3b1
  

  3 peopleauthored
  303d3b1View commit details

  Copy full SHA for 303d3b1

  Browse repository at this point

• 🔍 Add guidance on choosing `max_length` value and include visualization tool (#3630)

  qgallouedec

  authored
  3d765b0View commit details

  Copy full SHA for 3d765b0

  Browse repository at this point

• 🌋 [GRPO] add support for `pixel_attention_mask` (SmolVLM2) and `image_sizes` (LLaVa-Next) (#3760)

  Show description for fcd3e0f
  

  5 peopleauthored
  fcd3e0fView commit details

  Copy full SHA for fcd3e0f

  Browse repository at this point

• 💬 Fix `clone_chat_template` vocab size and support PEFT instruction tuning (#3763)

  Show description for 8a23c86
  

  3 peopleauthored
  8a23c86View commit details

  Copy full SHA for 8a23c86

  Browse repository at this point

• 📍 Support training peft model with gradient checkpointing (#3785)

  Show description for 5bb3ca4
  

  qgallouedec

  and

  kashif

  authored
  5bb3ca4View commit details

  Copy full SHA for 5bb3ca4

  Browse repository at this point

• 📐 Add epsilon hyperparameter recommendation to GSPO (#3790)

  qgallouedec

  authored
  fd70021View commit details

  Copy full SHA for fd70021

  Browse repository at this point

• 🤏 [SFT] Improve doc on training on assistant only messages (#3784)

  lewtun

  authored
  a902450View commit details

  Copy full SHA for a902450

  Browse repository at this point

Commits on Jul 27, 2025

• 🎞️ GSPO (#3775)

  qgallouedec

  authored
  0303431View commit details

  Copy full SHA for 0303431

  Browse repository at this point

Commits on Jul 26, 2025

• 🍿 [SFT] Fix dataset indexing which crashed with a IterableDataset (#3771)

  Show description for 23ea671
  

  LeonEricsson

  and

  qgallouedec

  authored
  23ea671View commit details

  Copy full SHA for 23ea671

  Browse repository at this point

• 🩹 [Hotfix] Fix pynccl communicator assertion error with VLLMClient (#3774)

  Show description for fc08f55
  

  CarlosArguilar

  and

  qgallouedec

  authored
  fc08f55View commit details

  Copy full SHA for fc08f55

  Browse repository at this point

• 📐 Fix CI and `GeometricMixtureWrapper` (#3779)

  qgallouedec

  authored
  2f4cb38View commit details

  Copy full SHA for 2f4cb38

  Browse repository at this point

Commits on Jul 25, 2025

• Update missing uv dep (#3772)

  lhoestq

  authored
  eee9ec9View commit details

  Copy full SHA for eee9ec9

  Browse repository at this point

• Add uv scripts headers (#3767)

  lhoestq

  authored
  a043fd7View commit details

  Copy full SHA for a043fd7

  Browse repository at this point

• 🤓 [GRPO] Documentation for entropy metric (#3770)

  LeonEricsson

  authored
  d16b960View commit details

  Copy full SHA for d16b960

  Browse repository at this point

• 🌌 [GRPO] Log generation entropy (#3700)

  Show description for daad892
  

  LeonEricsson

  and

  qgallouedec

  authored
  daad892View commit details

  Copy full SHA for daad892

  Browse repository at this point

• 🔠 Support model str in OnlineDPO (#3765)

  Show description for 097d615
  

  kashif

  and

  qgallouedec

  authored
  097d615View commit details

  Copy full SHA for 097d615

  Browse repository at this point

• 🔔 Add deprecation warnings for `AlignPropTrainer` and `DDPOTrainer` (#3755)

  Show description for bc3eebb
  

  qgallouedec

  and

  LeonEricsson

  authored
  bc3eebbView commit details

  Copy full SHA for bc3eebb

  Browse repository at this point

• ✋ Prevent NCCL Device Conflicts Between vLLM Server and Trainers (#3762)

  Show description for 1fb115d
  

  CarlosArguilar

  and

  qgallouedec

  authored
  1fb115dView commit details

  Copy full SHA for 1fb115d

  Browse repository at this point

Commits on Jul 24, 2025

• Add MPO recipe to Community tutorials (#3766)

  sergiopaniego

  authored
  3a40f18View commit details

  Copy full SHA for 3a40f18

  Browse repository at this point

Commits on Jul 23, 2025

• 👁️ [GRPO] Add VLM training capabilities to the trainer (#3072)

  CompN3rd

  authored
  56f4201View commit details

  Copy full SHA for 56f4201

  Browse repository at this point