Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 29, 2026

• [GRPO] Add parquet logging for completions with individual rewards (#4818)

  Show description for 035c3ff
  

  5 peopleauthored
  035c3ff

  Copy full SHA for 035c3ff

• Set default top_k to 0 in VLLMClient (#4927)

  albertvillanova

  authored
  414e60f

  Copy full SHA for 414e60f

• Fix import statement for import_utils in vllm_client.py (#4932)

  qgallouedec

  authored
  df332dc

  Copy full SHA for df332dc

• Fix profiling of VLLMGeneration.sync_weights (#4931)

  albertvillanova

  authored
  27998e9

  Copy full SHA for 27998e9

• Set model dtype to float32 in experimental tests of trainers (#4925)

  albertvillanova

  authored
  43fb8d3

  Copy full SHA for 43fb8d3

• Move VLLMClient to generation module (#4928)

  albertvillanova

  authored
  5a7481e

  Copy full SHA for 5a7481e

• Require transformers<5 with PairRMJudge (#4926)

  albertvillanova

  authored
  21a0d70

  Copy full SHA for 21a0d70

• Set model dtype to float32 in tests of trainers (#4924)

  albertvillanova

  authored
  4348375

  Copy full SHA for 4348375

• Support tool call data in `is_conversational` (#4923)

  qgallouedec

  authored
  a6cbf27

  Copy full SHA for a6cbf27

Commits on Jan 28, 2026

• Add validation for `sync_ref_model` in `GRPOTrainer` and `RLOOTrainer` when using PEFT models (#4912)

  Show description for ad91c6f
  

  qgallouedec

  and

  albertvillanova

  authored
  ad91c6f

  Copy full SHA for ad91c6f

• Update learning rate comments and add assertions for reference model parameters in GRPO and RLOO tests (#4914)

  Show description for 04717ff
  

  qgallouedec

  and

  albertvillanova

  authored
  04717ff

  Copy full SHA for 04717ff

• Remove chat template setup in dpo_vlm.py (#4906)

  qgallouedec

  authored
  b322d9b

  Copy full SHA for b322d9b

• Fix extra EOS appended in DPO preprocessing for conversational data (#4908)

  Show description for a70b4e0
  

  qgallouedec

  and

  albertvillanova

  authored
  a70b4e0

  Copy full SHA for a70b4e0

• Fix CI ValueError for 0 temperature (#4916)

  albertvillanova

  authored
  8464b0e

  Copy full SHA for 8464b0e

• Fix CI AssertionError: assert not True (#4921)

  albertvillanova

  authored
  5461a74

  Copy full SHA for 5461a74

• docs: add DoRA (2402.09353) to Paper Index (#4892)

  Show description for d54381a
  

  billycrapediem

  and

  qgallouedec

  authored
  d54381a

  Copy full SHA for d54381a

• Remove gradient checkpointing option from various training scripts (#4905)

  qgallouedec

  authored
  f2f6b32

  Copy full SHA for f2f6b32

• Comment about overriding prediction_step in GRPOTrainer and RLOOTrainer (#4913)

  qgallouedec

  authored
  6cbc102

  Copy full SHA for 6cbc102

• `device_map` init consistency in GRPO/RLOO/KTO (#4909)

  qgallouedec

  authored
  f40edf9

  Copy full SHA for f40edf9

• Fix help text formatting for `max_length` in `RewardConfig` and `SFTConfig` (#4910)

  qgallouedec

  authored
  a7070f9

  Copy full SHA for a7070f9

• Rearrange variable assignments in `DataCollatorForVisionLanguageModeling` (#4911)

  qgallouedec

  authored
  66efc0e

  Copy full SHA for 66efc0e

• Fix CI TypeError in llm-blender tests (#4919)

  albertvillanova

  authored
  e9a2f16

  Copy full SHA for e9a2f16

• Created new PTT integration docs as requested (#4907)

  Show description for 4f82320
  

  3 peopleauthored
  4f82320

  Copy full SHA for 4f82320

Commits on Jan 27, 2026

• Refactor vLLM generation [1/N]: Extract vLLM generation (#4700)

  albertvillanova

  authored
  0eb66d8

  Copy full SHA for 0eb66d8

• Fix CI AssertionError: Parameter has not changed (#4904)

  albertvillanova

  authored
  226ef57

  Copy full SHA for 226ef57

• Fix CI NotImplementedError for bfloat16 (#4902)

  albertvillanova

  authored
  956986e

  Copy full SHA for 956986e

Commits on Jan 26, 2026

• Transformers v5 release: extend xfail condition for `TestGRPOTrainer.test_training_vlm_and_liger` and update version checks (#4898)

  qgallouedec

  authored
  4322778

  Copy full SHA for 4322778

• GOLD training speed up (#4888)

  Show description for e106972
  

  141forever

  and

  kashif

  authored
  e106972

  Copy full SHA for e106972

Commits on Jan 23, 2026

• Fix RewardTrainer's results not reproducible (#4887)

  Show description for c477e88
  

  liyc-ai

  and

  qgallouedec

  authored
  c477e88

  Copy full SHA for c477e88

Commits on Jan 22, 2026

• Fix import path for `get_open_port` based on vLLM version (#4883)

  qgallouedec

  authored
  ba05323

  Copy full SHA for ba05323

• Mark ZeRO 2 as xfail in distributed tests due to current failure (#4885)

  qgallouedec

  authored
  e66a138

  Copy full SHA for e66a138

Commits on Jan 21, 2026

• Test distributed training for `RewardTrainer`, `RLOOTrainer` and `GRPOTrainer` (#4823)

  Show description for a60d75a
  

  qgallouedec

  and

  albertvillanova

  authored
  a60d75a

  Copy full SHA for a60d75a

• Enable vLLM sleep mode for generation in Online DPO (#4882)

  Show description for 60e4674
  

  3 peopleauthored
  60e4674

  Copy full SHA for 60e4674

• Bugfix: Logprob drift in vLLM serving mode (compared to colocate mode) (#4873)

  Show description for 0a881bc
  

  3 peopleauthored
  0a881bc

  Copy full SHA for 0a881bc

• Fix SFT training for prompt-completion type and transformers v5 (#4880)

  qgallouedec

  authored
  16b0903

  Copy full SHA for 16b0903

Pagination

Previous

Next