Skip to content

Pull requests: huggingface/trl

New pull request New
140 Open 3,015 Closed
140 Open 3,015 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

Add LoRA support for AsyncGRPO
#5610 opened Apr 21, 2026 by jonahsamost Loading…
1 of 8 tasks
3
experimental: Self-Distillation Zero
#5609 opened Apr 20, 2026 by LeonEricsson Collaborator • Draft
1 of 8 tasks
Added speculative_config to vllm-serve
#5605 opened Apr 20, 2026 by Ofir408 Loading…
2 of 8 tasks
1
7
support prefetch/prefetch_depth for async GRPO for ~5% speedups
#5602 opened Apr 20, 2026 by winglian Contributor Loading…
1 of 8 tasks
@AmineDiro
3
fix(distillation): reverse-KL server path NaN on variable completion length
#5594 opened Apr 19, 2026 by k1064190 Loading…
3 of 8 tasks
@kashif @cmpatino
5
Fix nested vocab_size for DistillationTrainer and GOLDTrainer
#5592 opened Apr 19, 2026 by Beichen-Ma Loading…
2 of 8 tasks
1
@cmpatino
1
feat: add TargetPO trainer
#5591 opened Apr 18, 2026 by JeanKaddour • Draft
4 of 8 tasks
5
Add tiny Qwen3-4B-Instruct-2507
#5586 opened Apr 17, 2026 by qgallouedec Member Loading…
1
Chunked cross-entropy loss for SFT (up to –50% VRAM)
#5575 opened Apr 17, 2026 by qgallouedec Member Loading…
11
Add training chat template for Qwen3-2507
#5574 opened Apr 16, 2026 by SwayamInSync Contributor Loading…
5
refactor: self distillation trainers (sdpo/sdft/...)
#5573 opened Apr 16, 2026 by LeonEricsson Collaborator Loading…
2 of 8 tasks
16
Fix empty-target self-distillation loss to stay connected to model graph
#5572 opened Apr 16, 2026 by walawalagoose Loading…
3 of 8 tasks
Improve BrowserGym examples for latest OpenEnv version
#5568 opened Apr 16, 2026 by sergiopaniego Member Loading…
8 tasks
7
Set _tokenizer attribute in experimental trainers
#5566 opened Apr 16, 2026 by albertvillanova Member Loading…
4
DataCollatorForPreference checking 'margin' in all examples
#5564 opened Apr 15, 2026 by antoinsader Loading…
5 of 8 tasks
1
5
Revert VLM support in parse_response
#5561 opened Apr 15, 2026 by qgallouedec Member Loading…
5
Accept processor in get_training_chat_template
#5560 opened Apr 15, 2026 by qgallouedec Member Loading…
3
Move experimental example scripts into their trainer folders
#5556 opened Apr 15, 2026 by sergiopaniego Member Loading…
1 of 8 tasks
1
Add support for prompt-completion format in DistillationTrainer
#5555 opened Apr 15, 2026 by cmpatino Collaborator Loading…
3 of 6 tasks
9
Fix GRPO VLM tests: Multimodal training requires conversational prompts
#5550 opened Apr 15, 2026 by kaixuanliu Contributor Loading…
3 tasks done
7
fix: Pass AsyncGRPOTrainer's processing_class to AsyncRolloutWorker
#5538 opened Apr 14, 2026 by xuanduy04 Contributor Loading…
2 of 8 tasks
1
3
feat: add Phi-3 training chat template with generation markers
#5526 opened Apr 12, 2026 by RudrenduPaul Contributor Loading…
2 of 4 tasks
7
feat: add Gemma/Gemma2 training chat templates with generation markers
#5523 opened Apr 11, 2026 by ps-abhi Loading…
5 of 8 tasks
feat(glm-4-moe): Add {% generation %} markers for training chat template
#5519 opened Apr 10, 2026 by casinca Contributor Loading…
5 of 8 tasks
10
[WIP] Fix OnlineDPO vLLM server completion handling
#5516 opened Apr 10, 2026 by JohnGiorgi Contributor • Draft
5 of 8 tasks
1
ProTip! Filter pull requests by the default branch with base:main.