Skip to content

Pull requests: huggingface/trl

New pull request New
73 Open 2,317 Closed
73 Open 2,317 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

CPOTrainer - Incorrect handling of different length chosen/rejected p…
#4639 opened Dec 8, 2025 by davmels Loading…
feat: implement DeepSeek unbiased KL estimator for GRPO
#4638 opened Dec 7, 2025 by jlcanta Loading…
2 tasks done
1
2
Disable gradient checkpointing during no-grad inference to avoid PyTorch warning
#4636 opened Dec 7, 2025 by qgallouedec Loading…
9
Fix KTOTrainer CUDA error for large-vocab models via tensor indexing
#4635 opened Dec 6, 2025 by bhuvanprakash Loading…
7
Preserve truncated tokens in BFD packing
#4632 opened Dec 5, 2025 by qgallouedec Loading…
1
11
Update docs landing with latest details
#4624 opened Dec 4, 2025 by sergiopaniego Loading…
6 tasks
1
[online trainers] add vllm lora adapter support
#4590 opened Nov 27, 2025 by kashif Loading…
5 tasks
5
Support async reward functions and parallelize call to reward functions.
#4567 opened Nov 24, 2025 by pramodith Loading…
3 of 5 tasks
1
6
Add cross-tokenizer distillation support for GKD and MiniLLM trainers
#4561 opened Nov 22, 2025 by sambhavnoobcoder Loading…
1
Add PSPO trust region method as alternative to clipping in GRPOTrainer
#4548 opened Nov 19, 2025 by MCDwyer Loading…
2 of 5 tasks
@kashif
4
fix: add vllm_group_port
#4545 opened Nov 19, 2025 by pointerhacker Loading…
3 of 5 tasks
1
3
Add compute_metrics parameter for GRPOTrainer
#4534 opened Nov 17, 2025 by colinzhaoxp Loading…
3
Make skip_special_tokens configurable
#4521 opened Nov 13, 2025 by taha-yassine Loading…
3 of 5 tasks
1
5
[GRPO] switch grpo liger loss to triton version
#4519 opened Nov 13, 2025 by kashif Loading…
1 of 8 tasks
10
adding [SimPER](https://arxiv.org/abs/2502.00883)
#4486 opened Nov 6, 2025 by leeparkuky Loading…
2 of 5 tasks
1
added 10 papers (+trainer cross-links) for #4407
#4441 opened Nov 3, 2025 by SSusantAchary Loading…
4 tasks done
24
docs: Unify model examples to use trl-lib namespace
#4431 opened Nov 2, 2025 by behroozazarkhalili Loading…
1
6
Use explicit tiny-Qwen2ForCausalLM-2.5 model_id param in CI tests
#4331 opened Oct 23, 2025 by albertvillanova Loading…
4
refactor: simplify parameter freezing in modeling_base.py
#4305 opened Oct 20, 2025 by Ki-Seki Loading…
2 of 5 tasks
2
Add CISPO loss option and documentation
#4298 opened Oct 16, 2025 by gustavorubim Loading…
4
feat: Add Multi-Token Prediction (MTP) support to SFTTrainer
#4290 opened Oct 15, 2025 by KLGR123 Loading…
1
Remove FSDP1 support: use FSDP2 exclusively
#4260 opened Oct 11, 2025 by behroozazarkhalili Loading…
2
Fix DPO Trainer Bug For Qwen2-VL (Issue 2660)
#4257 opened Oct 11, 2025 by FabianSchuetze Loading…
1 of 3 tasks
4
Online-dpo-ben
#4252 opened Oct 10, 2025 by burtenshaw • Draft
5 tasks
Add support for Python 3.14
#4225 opened Oct 8, 2025 by albertvillanova Loading…
5
ProTip! Updated in the last three days: updated:>2025-12-05.