Skip to content

Pull requests: huggingface/trl

New pull request New
177 Open 4,005 Closed
177 Open 4,005 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

fix vlm support asyncgrpo
#6839 opened Aug 20, 2026 by AmineDiro Member Loading…
1
Hotfix CI: xfail the experimental async trainer training tests
#6838 opened Aug 20, 2026 by albertvillanova Member Loading…
@albertvillanova
1
Fail fast on incompatible vLLM for default Async GRPO weight transfer
#6830 opened Aug 20, 2026 by DimensionSTP Loading…
4 of 7 tasks
Fix SDPO PEFT EMA initialization under ZeRO-3
#6829 opened Aug 20, 2026 by DimensionSTP Loading…
5 of 7 tasks
1
2
Preserve gradient checkpointing kwargs during GKD generation
#6827 opened Aug 20, 2026 by DimensionSTP Loading…
4 of 7 tasks
2
Stub the weight transfer in the AsyncGRPO epoch-stop test
#6826 opened Aug 20, 2026 by albertvillanova Member Loading…
1
Align AsyncDistillationTrainer with AsyncGRPOTrainer on pre-init args
#6825 opened Aug 20, 2026 by albertvillanova Member Loading…
1
Catch AsyncDistillationTrainer up with recent AsyncGRPO changes
#6823 opened Aug 20, 2026 by kashif Collaborator Loading…
4 of 8 tasks
6
fix(async-grpo): forward generation kwargs to vLLM
#6822 opened Aug 20, 2026 by kdubovikov Contributor Loading…
5 of 8 tasks
Overhaul GRPOWithReplayBufferTrainer: fix shape contract and rebuild replay machinery
#6821 opened Aug 20, 2026 by mmjerge Loading…
5 of 8 tasks
1
2
Reorganize examples into self-contained per-example folders
#6820 opened Aug 20, 2026 by qgallouedec Member Loading…
7
Pass position IDs when scoring online preference data
#6819 opened Aug 20, 2026 by guptaishaan Loading…
3 of 8 tasks
1
2
Mask Online DPO objective metrics at padding
#6818 opened Aug 20, 2026 by guptaishaan Loading…
5 of 8 tasks
1
Fix missing GMPO loss terms
#6817 opened Aug 19, 2026 by guptaishaan Loading…
4 of 8 tasks
1
Add a vllm_llm_kwargs passthrough to the vLLM LLM() constructor
#6816 opened Aug 19, 2026 by behroozazarkhalili Collaborator Loading…
1
1
Test liger vs non-liger GRPO loss parity
#6814 opened Aug 19, 2026 by kashif Collaborator Loading…
3 of 8 tasks
3
Fix missing EOS penalties for XPO and Nash-MD
#6813 opened Aug 19, 2026 by guptaishaan Loading…
5 of 8 tasks
1
1
Emit position_ids from the SFT collator when sequence parallelism is enabled
#6812 opened Aug 19, 2026 by qgallouedec Member Loading…
5
Fix epoch indexing for experimental schedules
#6811 opened Aug 19, 2026 by guptaishaan Loading…
5 of 8 tasks
1
Use the doc-builder pre-commit hook
#6799 opened Aug 19, 2026 by albertvillanova Member • Draft
2 tasks
Rewrite the TRL skill around the Python API
#6794 opened Aug 18, 2026 by qgallouedec Member • 2/2 Loading…
3
Ship the TRL skill at the .agents/skills convention and drop trl skills install
#6793 opened Aug 18, 2026 by qgallouedec Member • 1/2 Loading…
7
Add opt-in GLOO weight-sync backend for XPU server-mode weight sync
#6787 opened Aug 18, 2026 by adobrzyn • Draft
2 of 8 tasks
Clarify generalized JSD endpoint definition
#6781 opened Aug 17, 2026 by DaoyuanLi2816 Contributor Loading…
6 of 8 tasks
1
fix: surface NaN GRPO/RLOO steps in log_history
#6772 opened Aug 17, 2026 by tonycoder-hub Loading…
5 of 8 tasks
1
ProTip! Exclude everything labeled bug with -label:bug.