Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Sep 3, 2025

• [GRPO] Truncated Importance Sampling to address rollout-training mismatch (#3867)

  Show description for 12fc85f
  

  LeonEricsson

  and

  qgallouedec

  authored
  12fc85fView commit details

  Copy full SHA for 12fc85f

  Browse repository at this point

• Add pre-commit and hf-doc-builder as dev dependencies (#3993)

  albertvillanova

  authored
  fdd6bdaView commit details

  Copy full SHA for fdd6bda

  Browse repository at this point

• ⏰ fix: add return to shift_tokens_right (#3987)

  ginkyenglee

  authored
  cb84da0View commit details

  Copy full SHA for cb84da0

  Browse repository at this point

• ⚖️ Fix scale_rewards issue in GRPO (#3992)

  Show description for 35702ce
  

  3 peopleauthored
  35702ceView commit details

  Copy full SHA for 35702ce

  Browse repository at this point

Commits on Sep 2, 2025

• 🎯 Add Trackio integration documentation and update TOC (#3971)

  Show description for 705306d
  

  qgallouedec

  and

  lewtun

  authored
  705306dView commit details

  Copy full SHA for 705306d

  Browse repository at this point

Commits on Sep 1, 2025

• [GRPO] Adds an option to sleep vllm when running in colocated mode (#3968)

  Show description for edbe823
  

  3 peopleauthored
  edbe823View commit details

  Copy full SHA for edbe823

  Browse repository at this point

Commits on Aug 30, 2025

• 🪃 `args.gradient_checkpointing = False` instead of `args = dataclasses.replace(args, gradient_checkpointing=False)` (#3981)

  qgallouedec

  authored
  4c47b32View commit details

  Copy full SHA for 4c47b32

  Browse repository at this point

• 👮 Fix GRPO CLI by setting parameters for `get_soft_overlong_punishment` (#3972)

  qgallouedec

  authored
  92046bbView commit details

  Copy full SHA for 92046bb

  Browse repository at this point

Commits on Aug 29, 2025

• Refactor version retrieval to use importlib.metadata for improved reliability

  qgallouedec

  committed
  39faf36View commit details

  Copy full SHA for 39faf36

  Browse repository at this point

• ⬆️ Bump dev version (#3978)

  qgallouedec

  authored
  1cb4150View commit details

  Copy full SHA for 1cb4150

  Browse repository at this point

• Release: v0.22 (#3977)

  qgallouedec

  authored
  3a6b365View commit details

  Copy full SHA for 3a6b365

  Browse repository at this point

• 🧱 PyPI publishing workflow (#3976)

  qgallouedec

  authored
  7ae16d3View commit details

  Copy full SHA for 7ae16d3

  Browse repository at this point

• Style

  qgallouedec

  committed
  ab984faView commit details

  Copy full SHA for ab984fa

  Browse repository at this point

• Fix CI (#3975)

  qgallouedec

  authored
  419d716View commit details

  Copy full SHA for 419d716

  Browse repository at this point

• 📜 GSPO docs - Sequence importance ratio and differences in relation to GRPO (#3816)

  Show description for f538bd3
  

  3 peopleauthored
  f538bd3View commit details

  Copy full SHA for f538bd3

  Browse repository at this point

• ℹ️ Validate examples on xpu (#3897)

  Show description for 8aa0eed
  yao-matrix

  authored
  8aa0eedView commit details

  Copy full SHA for 8aa0eed

  Browse repository at this point

• 🔥 [Refactor] RLOOTrainer (#3801)

  Show description for e7b37d4
  

  3 peopleauthored
  e7b37d4View commit details

  Copy full SHA for e7b37d4

  Browse repository at this point

Commits on Aug 27, 2025

• Fixed some typos and added small details about `trackio` to docs (#3965)

  sergiopaniego

  authored
  b7676d1View commit details

  Copy full SHA for b7676d1

  Browse repository at this point

• [CI] Modify tests to handle device allocation for models (#3962)

  kashif

  authored
  515e9ebView commit details

  Copy full SHA for 515e9eb

  Browse repository at this point

• Add HF jobs tag when creating model card via jobs (#3956)

  sergiopaniego

  authored
  26442abView commit details

  Copy full SHA for 26442ab

  Browse repository at this point

• 🧭 HF jobs x TRL guide (#3890)

  Show description for 0c91515
  

  3 peopleauthored
  0c91515View commit details

  Copy full SHA for 0c91515

  Browse repository at this point

• 📸 Return `position_ids` for `flash_attention_3` (#3942)

  jue-jue-zi

  authored
  4b3517fView commit details

  Copy full SHA for 4b3517f

  Browse repository at this point

• 🦥 Unsloth Docs update (#3955)

  Show description for 6f58651
  

  shimmyshimmer

  and

  qgallouedec

  authored
  6f58651View commit details

  Copy full SHA for 6f58651

  Browse repository at this point

• LitePPO: Fix Docs for paper index (#3954)

  pramodith

  authored
  0c7ab76View commit details

  Copy full SHA for 0c7ab76

  Browse repository at this point

Commits on Aug 25, 2025

• ✂️ fix: handle list tensors in split_tensor_dict function (#3951)

  qgallouedec

  authored
  ffc061bView commit details

  Copy full SHA for ffc061b

  Browse repository at this point

• 🤸 [SFT] Drop entropy calculation when using liger (#3947)

  qgallouedec

  authored
  38fc1f6View commit details

  Copy full SHA for 38fc1f6

  Browse repository at this point

Commits on Aug 24, 2025

• [GKD] add liger loss (#3946)

  Show description for 39cc9a8
  

  kashif

  and

  qgallouedec

  authored
  39cc9a8View commit details

  Copy full SHA for 39cc9a8

  Browse repository at this point

• [DPO] Adding support for different losses which are now supported by Liger (#3815)

  Show description for 1f15f18
  

  3 peopleauthored
  1f15f18View commit details

  Copy full SHA for 1f15f18

  Browse repository at this point

Commits on Aug 22, 2025

• 🗂 Update paper_index section (#3937)

  Show description for 181a841
  

  behroozazarkhalili

  and

  qgallouedec

  authored
  181a841View commit details

  Copy full SHA for 181a841

  Browse repository at this point

• 🎆 Add entropy logging in SFT (#3940)

  qgallouedec

  authored
  da167d8View commit details

  Copy full SHA for da167d8

  Browse repository at this point

• 🏌️ DAPO loss type (#3938)

  qgallouedec

  authored
  2324245View commit details

  Copy full SHA for 2324245

  Browse repository at this point

Commits on Aug 21, 2025

• 🪶 [GRPO] PPO Lite: Scale rewards by Std of Batch (#3935)

  Show description for fe44806
  

  pramodith

  and

  qgallouedec

  authored
  fe44806View commit details

  Copy full SHA for fe44806

  Browse repository at this point

• 📦 Wrapping the main execution code to avoid multi-processing issues from vLLM (#3932)

  Show description for 251c048
  kaixuanliu

  authored
  251c048View commit details

  Copy full SHA for 251c048

  Browse repository at this point

• 🗞 bugfix 'TrainerState' object is not subscriptable (#3936)

  Show description for e2eaa23
  

  ErezYosef

  and

  qgallouedec

  authored
  e2eaa23View commit details

  Copy full SHA for e2eaa23

  Browse repository at this point

Commits on Aug 20, 2025

• 🗑️ Deprecate `setup_chat_format` (#3929)

  qgallouedec

  authored
  48d7eccView commit details

  Copy full SHA for 48d7ecc

  Browse repository at this point