Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit history

Commits on Sep 15, 2025

• 🗑️ Remove deprecated `AlignPropTrainer`, `DDPOTrainer` and `IterativeSFTTrainer` (#4068)

  qgallouedec

  authored
  78f1a92View commit details

  Copy full SHA for 78f1a92

  View code at this pointBrowse repository at this point

Commits on Sep 14, 2025

• Aux loss is already included in the loss returned by Transformers (#4078)

  pramodith

  authored
  5a1c2f9View commit details

  Copy full SHA for 5a1c2f9

  View code at this pointBrowse repository at this point

• 🐳 Docker update + Simplify Jobs doc (#3931)

  Show description for 9955ee7
  

  3 peopleauthored
  9955ee7View commit details

  Copy full SHA for 9955ee7

  View code at this pointBrowse repository at this point

Commits on Sep 13, 2025

• ✂️ [GRPO VLM] Update split sizes to generalize (#4032)

  Show description for 69e288e
  

  zucchini-nlp

  and

  qgallouedec

  authored
  69e288eView commit details

  Copy full SHA for 69e288e

  View code at this pointBrowse repository at this point

Commits on Sep 12, 2025

• 🌾 [Experimental] BEMA for ref model (#3898)

  Show description for d655ce4
  

  qgallouedec

  and

  kashif

  authored
  d655ce4View commit details

  Copy full SHA for d655ce4

  View code at this pointBrowse repository at this point

• 🧪 Add `trl.experimental` Submodule (#4073)

  Show description for 91c4bba
  

  August-murr

  and

  qgallouedec

  authored
  91c4bbaView commit details

  Copy full SHA for 91c4bba

  View code at this pointBrowse repository at this point

• ♨️ [GRPO] Fix potential hang in `get_high_entropy_mask` (#4041)

  akakakakakaa

  authored
  f4ff248View commit details

  Copy full SHA for f4ff248

  View code at this pointBrowse repository at this point

Commits on Sep 11, 2025

• Improve docstring of AlignPropTrainer (#4059)

  Show description for b8eb5c5
  

  albertvillanova

  and

  qgallouedec

  authored
  b8eb5c5View commit details

  Copy full SHA for b8eb5c5

  View code at this pointBrowse repository at this point

• 💡 Fix type hint to `make_parser` function in multiple scripts (#4050)

  qgallouedec

  authored
  07f9ad9View commit details

  Copy full SHA for 07f9ad9

  View code at this pointBrowse repository at this point

• Hotfix: Add ParallelismConfig fallback for transformers with old accelerate (#4063)

  albertvillanova

  authored
  44ddc28View commit details

  Copy full SHA for 44ddc28

  View code at this pointBrowse repository at this point

• Remove redundant 'None' from docstrings (#4058)

  albertvillanova

  authored
  e8b8499View commit details

  Copy full SHA for e8b8499

  View code at this pointBrowse repository at this point

Commits on Sep 9, 2025

• 🗜 Hotfix: avoid passing `quantization_config=None` (#4019)

  qgallouedec

  authored
  a647e5aView commit details

  Copy full SHA for a647e5a

  View code at this pointBrowse repository at this point

• 🪪 Update SFTTrainer to handle labels correctly and add configuration example in paper index (#4051)

  qgallouedec

  authored
  816ac61View commit details

  Copy full SHA for 816ac61

  View code at this pointBrowse repository at this point

• 💬 Remove setting chat template in sft script (#4037)

  qgallouedec

  authored
  373a64aView commit details

  Copy full SHA for 373a64a

  View code at this pointBrowse repository at this point

• ✨ Improve SFT doc (#4005)

  Show description for 09e1924
  

  qgallouedec

  and

  albertvillanova

  authored
  09e1924View commit details

  Copy full SHA for 09e1924

  View code at this pointBrowse repository at this point

• Add autodoc for BestOfNSampler and improve docstrings (#4034)

  albertvillanova

  authored
  a228cb5View commit details

  Copy full SHA for a228cb5

  View code at this pointBrowse repository at this point

• 🏂 Fix label shifting logic in `SFTTrainer` for compatibility with CP (#4038)

  qgallouedec

  authored
  6c6f13bView commit details

  Copy full SHA for 6c6f13b

  View code at this pointBrowse repository at this point

• Update VLM arch check to `AutoModelForImageTextToText` for DPO and Online DPO (#4049)

  sergiopaniego

  authored
  b3f9f61View commit details

  Copy full SHA for b3f9f61

  View code at this pointBrowse repository at this point

• 🧨 DFT (#4042)

  qgallouedec

  authored
  659d2c1View commit details

  Copy full SHA for 659d2c1

  View code at this pointBrowse repository at this point

• ⚖️ Remove `average_tokens_across_devices` default replacement (#4039)

  qgallouedec

  authored
  af82b38View commit details

  Copy full SHA for af82b38

  View code at this pointBrowse repository at this point

Commits on Sep 8, 2025

• Add autodoc for AlignPropTrainer and AlignPropConfig (#4033)

  albertvillanova

  authored
  e4ebf3bView commit details

  Copy full SHA for e4ebf3b

  View code at this pointBrowse repository at this point

• Add missing trainer docstrings (#4030)

  albertvillanova

  authored
  e458df6View commit details

  Copy full SHA for e458df6

  View code at this pointBrowse repository at this point

• Fix formatting errors in docstrings (#4025)

  albertvillanova

  authored
  4f9009bView commit details

  Copy full SHA for 4f9009b

  View code at this pointBrowse repository at this point

Commits on Sep 7, 2025

• Align docstring parameters with function definitions (#4017)

  albertvillanova

  authored
  c9484b1View commit details

  Copy full SHA for c9484b1

  View code at this pointBrowse repository at this point

Commits on Sep 5, 2025

• ⚖️ Add vLLM server mode and VLM support to OnlineDPOTrainer (#3783)

  Show description for d1bf560
  

  4 peopleauthored
  d1bf560View commit details

  Copy full SHA for d1bf560

  View code at this pointBrowse repository at this point

Commits on Sep 4, 2025

• Add missing doc strings in SFTrainer (#4003)

  Show description for 19f9b9e
  

  pramodith

  and

  qgallouedec

  authored
  19f9b9eView commit details

  Copy full SHA for 19f9b9e

  View code at this pointBrowse repository at this point

• [SFTTrainer]: Add Aux Loss for MoE models. (#4012)

  Show description for 1eb3801
  

  pramodith

  and

  qgallouedec

  authored
  1eb3801View commit details

  Copy full SHA for 1eb3801

  View code at this pointBrowse repository at this point

• 🌵 Refactor entropy_from_logits for memory efficiency (#4013)

  qgallouedec

  authored
  deae7e0View commit details

  Copy full SHA for deae7e0

  View code at this pointBrowse repository at this point

• 👷 Added Kernels on the Hub x TRL guide (#3969)

  Show description for 0c69fd2
  

  4 peopleauthored
  0c69fd2View commit details

  Copy full SHA for 0c69fd2

  View code at this pointBrowse repository at this point

• [SFT] fix: collator docstring (#4011)

  LeonEricsson

  authored
  b5fd290View commit details

  Copy full SHA for b5fd290

  View code at this pointBrowse repository at this point

Commits on Sep 3, 2025

• 📏 `torch_dype` to `dtype` everywhere (#4000)

  Show description for 208e9f7
  

  sergiopaniego

  and

  qgallouedec

  authored
  208e9f7View commit details

  Copy full SHA for 208e9f7

  View code at this pointBrowse repository at this point

• [GRPO]: Fix Multi-GPU training for Entropy based masking of tokens. (#3964)

  Show description for 3bfa981
  

  pramodith

  and

  qgallouedec

  authored
  3bfa981View commit details

  Copy full SHA for 3bfa981

  View code at this pointBrowse repository at this point

• 💾 [bugfix] fix PPO save_checkpoint (#3998)

  hjh0119

  authored
  6a5dfffView commit details

  Copy full SHA for 6a5dfff

  View code at this pointBrowse repository at this point

• ✖️ Support pad-to-multiple-of and padding-free (#3996)

  qgallouedec

  authored
  18633dbView commit details

  Copy full SHA for 18633db

  View code at this pointBrowse repository at this point

• [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

  View code at this pointBrowse repository at this point