Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Feb 6, 2026

• perf: Qwen SAPO loss optimization (#4956)

  Show description for 7a39ff3
  

  casinca

  and

  qgallouedec

  authored
  7a39ff3View commit details

  Copy full SHA for 7a39ff3

  Browse repository at this point

• Update sampling mode to token level for safety (#4989)

  Show description for bd20670
  

  sergiopaniego

  and

  qgallouedec

  authored
  bd20670View commit details

  Copy full SHA for bd20670

  Browse repository at this point

• Update NeMo-Gym to use `env_mask` (#4986)

  Show description for aa7d457
  

  cmunley1

  and

  qgallouedec

  authored
  aa7d457View commit details

  Copy full SHA for aa7d457

  Browse repository at this point

• Add distributed smoke tests workflow for Transformers branch (#4996)

  qgallouedec

  authored
  5db1c11View commit details

  Copy full SHA for 5db1c11

  Browse repository at this point

• Add GitHub Actions workflow for testing against Transformers branch (#4995)

  qgallouedec

  authored
  90a35d1View commit details

  Copy full SHA for 90a35d1

  Browse repository at this point

• Fix ZeRO-3 + PEFT + gradient checkpointing (#4951)

  qgallouedec

  authored
  f11b4c3View commit details

  Copy full SHA for f11b4c3

  Browse repository at this point

• Fix post_init warning stacklevel to 3 (#4993)

  albertvillanova

  authored
  27cbe98View commit details

  Copy full SHA for 27cbe98

  Browse repository at this point

• Fix deprecation of DPOConfig.max_completion_length (#4992)

  albertvillanova

  authored
  57cac25View commit details

  Copy full SHA for 57cac25

  Browse repository at this point

• Assert chat_template is applied in test_train_with_chat_template_kwargs (#4991)

  albertvillanova

  authored
  ce72c06View commit details

  Copy full SHA for ce72c06

  Browse repository at this point

• Fix import of AutoModelForCausalLMWithValueHead from experimental (#4990)

  albertvillanova

  authored
  ffdaba3View commit details

  Copy full SHA for ffdaba3

  Browse repository at this point

• Use local variable instead of attribute in collator tests (#4957)

  qgallouedec

  authored
  97a8a96View commit details

  Copy full SHA for 97a8a96

  Browse repository at this point

• Update dataset configuration name in toolcall dataset loading (#4984)

  qgallouedec

  authored
  4e212bdView commit details

  Copy full SHA for 4e212bd

  Browse repository at this point

• Fix passing tokenizer in test_train_with_chat_template_kwargs (#4987)

  albertvillanova

  authored
  c82f6aaView commit details

  Copy full SHA for c82f6aa

  Browse repository at this point

• Pin transformers!=5.1.0 in deepspeed extra due to incompatibility (#4985)

  albertvillanova

  authored
  c581c1eView commit details

  Copy full SHA for c581c1e

  Browse repository at this point

• Replace `warmup_ratio` with `warmup_steps` (#4983)

  qgallouedec

  authored
  98aca7fView commit details

  Copy full SHA for 98aca7f

  Browse repository at this point

• [CI] Disallow installation of transformers 5.1.0 due to compatibility issues with DeepSpeed (#4982)

  qgallouedec

  authored
  032ee13View commit details

  Copy full SHA for 032ee13

  Browse repository at this point

Commits on Feb 5, 2026

• Deprecate parameters in `DPOConfig` (#4969)

  Show description for a0e5f26
  

  qgallouedec

  and

  albertvillanova

  authored
  a0e5f26View commit details

  Copy full SHA for a0e5f26

  Browse repository at this point

• Simplify instructions of installation of OpenEnv (#4980)

  sergiopaniego

  authored
  f0a738dView commit details

  Copy full SHA for f0a738d

  Browse repository at this point

• Replace `torch.allclose` with `torch.testing.assert_close` (#4977)

  qgallouedec

  authored
  a92d143View commit details

  Copy full SHA for a92d143

  Browse repository at this point

• Support truncated completions in GRPO multi-turn training (#4976)

  albertvillanova

  authored
  b0b798aView commit details

  Copy full SHA for b0b798a

  Browse repository at this point

• Fix add_column in test_train_with_chat_template_kwargs (#4979)

  albertvillanova

  authored
  ac194a9View commit details

  Copy full SHA for ac194a9

  Browse repository at this point

• Set specific OpenEnv version when installed (#4978)

  sergiopaniego

  authored
  3a76b7aView commit details

  Copy full SHA for 3a76b7a

  Browse repository at this point

• Remove truncation from tokenizer calls if no max_length (#4972)

  albertvillanova

  authored
  0113ad7View commit details

  Copy full SHA for 0113ad7

  Browse repository at this point

• Remove padding_value from experimental CPO and use pad_token_id (#4962)

  albertvillanova

  authored
  eee98f7View commit details

  Copy full SHA for eee98f7

  Browse repository at this point

• Fix test_train_with_chat_template_kwargs (#4971)

  qgallouedec

  authored
  1354860View commit details

  Copy full SHA for 1354860

  Browse repository at this point

• Revert change in GRPO from NeMo-Gym Integration (#4970)

  qgallouedec

  authored
  1bd2a52View commit details

  Copy full SHA for 1bd2a52

  Browse repository at this point

• Remove max_prompt_length from experimental ORPO (#4966)

  albertvillanova

  authored
  22ad7e6View commit details

  Copy full SHA for 22ad7e6

  Browse repository at this point

• Remove max_prompt_length from experimental CPO (#4965)

  albertvillanova

  authored
  657babdView commit details

  Copy full SHA for 657babd

  Browse repository at this point

• Remove max_prompt_length from experimental BCO (#4964)

  albertvillanova

  authored
  50e35deView commit details

  Copy full SHA for 50e35de

  Browse repository at this point

• Remove max_prompt_length from experimental PRM (#4963)

  albertvillanova

  authored
  35bcab1View commit details

  Copy full SHA for 35bcab1

  Browse repository at this point

Commits on Feb 4, 2026

• Add test for training with `compute_metrics` in `RewardTrainer` (#4958)

  qgallouedec

  authored
  e4995b2View commit details

  Copy full SHA for e4995b2

  Browse repository at this point

• Add test for tool call data in `RewardTrainer` (#4959)

  qgallouedec

  authored
  cb5a73bView commit details

  Copy full SHA for cb5a73b

  Browse repository at this point

• NeMo-Gym Integration (#4848)

  Show description for 90b875c
  

  4 peopleauthored
  90b875cView commit details

  Copy full SHA for 90b875c

  Browse repository at this point

Commits on Feb 3, 2026

• Remove access to `warnings_issued` (#4960)

  qgallouedec

  authored
  5cb7eeeView commit details

  Copy full SHA for 5cb7eee

  Browse repository at this point

• Add test for training with `compute_metrics` in `SFTTrainer` (#4950)

  qgallouedec

  authored
  2a55ed7View commit details

  Copy full SHA for 2a55ed7

  Browse repository at this point