Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Jun 4, 2025

• ⚡ Faster FFD packing (#3537)

  Show description for 50a2fa8
  

  mariosasko

  and

  qgallouedec

  authored
  50a2fa8View commit details

  Copy full SHA for 50a2fa8

  Browse repository at this point

• 🎀 [SFT][Bugfix] sets average_tokens_across_devices to true in SFTConfig (#3538)

  Show description for 0333108
  

  edbeeching

  and

  qgallouedec

  authored
  0333108View commit details

  Copy full SHA for 0333108

  Browse repository at this point

• 💭 [Data] Fix DeepSeek-R1 case (#3522)

  Show description for 6ffde23
  

  kashif

  and

  qgallouedec

  authored
  6ffde23View commit details

  Copy full SHA for 6ffde23

  Browse repository at this point

• 🐳 Add DeepseekV3 model configurations and update tests for new models (#3536)

  qgallouedec

  authored
  6f288c2View commit details

  Copy full SHA for 6f288c2

  Browse repository at this point

• 🧭 Remove useless transformers version checks (#3534)

  qgallouedec

  authored
  8cf6220View commit details

  Copy full SHA for 8cf6220

  Browse repository at this point

• 🎯 Don't use `getattr` to get `gradient_checkpointing` (#3535)

  qgallouedec

  authored
  da7b3feView commit details

  Copy full SHA for da7b3fe

  Browse repository at this point

Commits on Jun 3, 2025

• 📰 Add blog "No GPU left behind: Unlocking Efficiency with Co-located vLLM in TRL" (#3527)

  qgallouedec

  authored
  24ef9ebView commit details

  Copy full SHA for 24ef9eb

  Browse repository at this point

• 🎀 New defaults: `logging_steps=10` (#3514)

  qgallouedec

  authored
  b0eff32View commit details

  Copy full SHA for b0eff32

  Browse repository at this point

• 🪦 RIP trl chat (#3531)

  shirinyamani

  authored
  026fc94View commit details

  Copy full SHA for 026fc94

  Browse repository at this point

• 🎀 New defaults: preparing the new structure (#3530)

  qgallouedec

  authored
  a912ad1View commit details

  Copy full SHA for a912ad1

  Browse repository at this point

Commits on Jun 2, 2025

• 📉 FFD packing (#3521)

  qgallouedec

  authored
  fef915eView commit details

  Copy full SHA for fef915e

  Browse repository at this point

• Add "🐯 Liger GRPO meets TRL" (#3525)

  qgallouedec

  authored
  0db63f0View commit details

  Copy full SHA for 0db63f0

  Browse repository at this point

Commits on May 30, 2025

• 🎀 New default: `beta=0.0` for GRPO (#3516)

  qgallouedec

  authored
  7359ddcView commit details

  Copy full SHA for 7359ddc

  Browse repository at this point

• 🧭 Patch release guide (#3512)

  qgallouedec

  authored
  0844936View commit details

  Copy full SHA for 0844936

  Browse repository at this point

Commits on May 29, 2025

• 📚 Fix doc building by removing vLLM from dev dependencies in `setup.cfg` (#3511)

  qgallouedec

  authored
  897c87fView commit details

  Copy full SHA for 897c87f

  Browse repository at this point

Commits on May 28, 2025

• 📎 Fix clip ratio logging (#3506)

  qgallouedec

  authored
  c13de6fView commit details

  Copy full SHA for c13de6f

  Browse repository at this point

• ⬆️ Bump dev version (#3505)

  qgallouedec

  authored
  722847aView commit details

  Copy full SHA for 722847a

  Browse repository at this point

• Release: v0.18 (#3504)

  qgallouedec

  authored
  ef4b0b2View commit details

  Copy full SHA for ef4b0b2

  Browse repository at this point

Commits on May 27, 2025

• ✂️ [DPO] Fix truncation `keep_end` leading to zero'd out samples (#3398)

  Show description for 8e8e62b
  

  LeonEricsson

  and

  qgallouedec

  authored
  8e8e62bView commit details

  Copy full SHA for 8e8e62b

  Browse repository at this point

• 🏰 [vllm] Support `base_url` parameter for vLLM client initialization (#3324)

  Show description for 824100c
  

  re-imagined

  and

  qgallouedec

  authored
  824100cView commit details

  Copy full SHA for 824100c

  Browse repository at this point

• 🤧 LD-DPO support (#3458)

  Show description for 4e7f0a5
  

  3 peopleauthored
  4e7f0a5View commit details

  Copy full SHA for 4e7f0a5

  Browse repository at this point

• 📏 Completion length logging fix + remainder logging fix (#3482)

  Show description for 17a9069
  

  shirinyamani

  and

  qgallouedec

  authored
  17a9069View commit details

  Copy full SHA for 17a9069

  Browse repository at this point

• Forgotten commit from #3502

  qgallouedec

  committed
  cb07c44View commit details

  Copy full SHA for cb07c44

  Browse repository at this point

• 🔭 [GRPO] Log advantages and fraction of samples with an std of zero (#3502)

  Show description for 0b6a187
  

  edbeeching

  and

  qgallouedec

  authored
  0b6a187View commit details

  Copy full SHA for 0b6a187

  Browse repository at this point

• 🐌 Clean two-sided clipping (#3499)

  qgallouedec

  authored
  ac18c9dView commit details

  Copy full SHA for ac18c9d

  Browse repository at this point

• 🛠️ Initialize reward_kwargs to prevent UnboundLocalError in GRPOTrainer (#3459)

  Show description for d1174ad
  

  teilomillet

  and

  qgallouedec

  authored
  d1174adView commit details

  Copy full SHA for d1174ad

  Browse repository at this point

• 👇 Update grpo.py to fix bugs for cli grpo --reward_funcs my_lib.my_reward (#3454)

  Show description for cd83841
  

  wa008

  and

  qgallouedec

  authored
  cd83841View commit details

  Copy full SHA for cd83841

  Browse repository at this point

Commits on May 26, 2025

• [GKD] fix the gkd script (#3497)

  kashif

  authored
  c7e3f09View commit details

  Copy full SHA for c7e3f09

  Browse repository at this point

• [GRPO] disabling top_k sampling default (#3494)

  Show description for 5c08897
  

  kashif

  and

  lewtun

  authored
  5c08897View commit details

  Copy full SHA for 5c08897

  Browse repository at this point

Commits on May 25, 2025

• [Docs] sync logging doc to current metrics (#3478)

  kashif

  authored
  3ef9fafView commit details

  Copy full SHA for 3ef9faf

  Browse repository at this point

Commits on May 24, 2025

• Fix mis-aligned prompts and completions in colocate mode (#3491)

  Show description for 9ac614f
  

  toslali-ibm

  and

  qgallouedec

  authored
  9ac614fView commit details

  Copy full SHA for 9ac614f

  Browse repository at this point

• [Doc][SFT] Update sft_trainer.md. link prompt-completion dataset example (#3486)

  Show description for 29401e7
  

  HERIUN

  and

  kashif

  authored
  29401e7View commit details

  Copy full SHA for 29401e7

  Browse repository at this point

• Fix typo (#3489)

  nikolai-kummer

  authored
  31bf3f9View commit details

  Copy full SHA for 31bf3f9

  Browse repository at this point

Commits on May 23, 2025

• [CI] fix sampler api to make the CI green (#3488)

  kashif

  authored
  7f32792View commit details

  Copy full SHA for 7f32792

  Browse repository at this point

• [SFT] update minimal liger version (#3483)

  kashif

  authored
  3d87279View commit details

  Copy full SHA for 3d87279

  Browse repository at this point