Skip to content

Pull requests: huggingface/trl

New pull request New
172 Open 3,344 Closed
172 Open 3,344 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

refactor(sft): build labels during dataset preparation instead of collation
#6037 opened Jun 12, 2026 by 0xadvait Loading…
5 of 8 tasks
9
test: don't hard-code bf16=True on devices that lack bf16 support
#6036 opened Jun 12, 2026 by behroozazarkhalili Collaborator Loading…
1
2
fix: load image-text policy for async grpo
#6032 opened Jun 12, 2026 by he-yufeng Loading…
5 of 8 tasks
1
@AmineDiro
fix: pass AsyncGRPO environment rewards
#6031 opened Jun 12, 2026 by he-yufeng Loading…
5 of 8 tasks
1
fix(sft): drop examples with no trainable tokens after max_length truncation
#6025 opened Jun 12, 2026 by 0xadvait • Draft
5 of 8 tasks
1
4
fix(grpo): correct DAPO/CISPO/VESPO loss normalization when steps_per_generation != gradient_accumulation_steps
#6024 opened Jun 12, 2026 by 0xadvait Loading…
5 of 8 tasks
1
Remove silently-ignored W&B/Hub fields from GOLD and Distillation configs
#6023 opened Jun 11, 2026 by DaoyuanLi2816 Contributor Loading…
3 of 4 tasks
Align AsyncGRPO clip-ratio metrics with GRPOTrainer
#6021 opened Jun 11, 2026 by qgallouedec Member Loading…
6
Align AsyncGRPO num_completions_to_print with GRPO (int | None)
#6020 opened Jun 11, 2026 by qgallouedec Member Loading…
1
Align AsyncGRPO epsilon_high with GRPO (None fallback to epsilon)
#6019 opened Jun 11, 2026 by qgallouedec Member Loading…
1
Add experimental Harbor integration for GRPO environment training
#6018 opened Jun 11, 2026 by adithya-s-k Collaborator Loading…
1
8
Fix logging_steps default mentioned in AsyncGRPOConfig docstring
#6016 opened Jun 11, 2026 by qgallouedec Member Loading…
1
Align epsilon help/docstring wording
#6014 opened Jun 11, 2026 by qgallouedec Member Loading…
1
Align async GRPO loss variable names with GRPOTrainer
#6013 opened Jun 11, 2026 by qgallouedec Member Loading…
1
Use relative imports in async_grpo to match the rest of trl/experimental`
#6012 opened Jun 11, 2026 by qgallouedec Member Loading…
1
GRPO adapter-only vLLM LoRA sync
#6007 opened Jun 11, 2026 by rycerzes Contributor Loading…
4 of 8 tasks
Normalize JSD distillation loss by num_items_in_batch for gradient accumulation
#6006 opened Jun 11, 2026 by behroozazarkhalili Collaborator Loading…
1
@kashif @cmpatino
10
Warn when a string model loads in float32 under mixed-precision training
#6005 opened Jun 11, 2026 by behroozazarkhalili Collaborator Loading…
1
8
Add DiffusionGemma block-diffusion SFT example
#6003 opened Jun 11, 2026 by kashif Collaborator • Draft
5
Support multiple environments [2/2]: Per-example environment selection
#6002 opened Jun 11, 2026 by qgallouedec Member Loading…
5
Support multiple environments [1/2]: Pool and build environment tool dicts at batch time
#6001 opened Jun 11, 2026 by qgallouedec Member Loading…
5
OPSD: on-policy self-distillation trainer
#5990 opened Jun 9, 2026 by kashif Collaborator Loading…
8
Fix OnlineDPOTrainer evaluation
#5985 opened Jun 9, 2026 by he-yufeng Loading…
5 of 8 tasks
1
3
Fix GRPO KL estimator overflow
#5984 opened Jun 9, 2026 by he-yufeng Loading…
5 of 8 tasks
1
3
chore(docs): Add GRPO clipping viz and explanation
#5981 opened Jun 9, 2026 by zafstojano Contributor Loading…
2 of 8 tasks
ProTip! no:milestone will show everything without a milestone.