Skip to content

Pull requests: huggingface/trl

New pull request New
66 Open 1,849 Closed
66 Open 1,849 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 bug when using dataset streaming by accelerate
#3950 opened Aug 25, 2025 by kaixuanliu Loading…
add default dataset and related precessing to make this example work
#3949 opened Aug 25, 2025 by kaixuanliu Loading…
Update grpo_trainer.py to fix the dim error.
#3943 opened Aug 23, 2025 by HelloWorldLTY Loading…
4
Return position_ids for flash_attention_3
#3942 opened Aug 23, 2025 by jue-jue-zi Loading…
5 tasks
Docker update
#3931 opened Aug 20, 2025 by qgallouedec Loading…
5 tasks
1
[SFTTrainer]: Check for assistant mask up to max_length
#3930 opened Aug 20, 2025 by pramodith Loading…
2 of 5 tasks
13
Support for pre-defined image positions in VLM training data
#3911 opened Aug 17, 2025 by YeFD Loading…
3 of 5 tasks
9
[DRAFT] Refactor DPO
#3906 opened Aug 15, 2025 by qgallouedec • Draft
5 tasks
Test in distributed setting
#3902 opened Aug 15, 2025 by qgallouedec Loading…
5 tasks
1
BEMA for ref model
#3898 opened Aug 14, 2025 by qgallouedec Loading…
5 tasks
1
validate examples on xpu
#3897 opened Aug 14, 2025 by yao-matrix Loading…
🧭 HF jobs x TRL guide
#3890 opened Aug 13, 2025 by sergiopaniego Loading…
3 of 12 tasks
19
GRPOTrainer : fix prompt truncation for multimodal inputs with multiple image tokens
#3879 opened Aug 11, 2025 by artem-spector Loading…
4 tasks
5
vLLM rollout numerical differences causing off-policy RL.
#3867 opened Aug 7, 2025 by LeonEricsson • Draft
5 tasks
7
Implement DPOP
#3864 opened Aug 7, 2025 by 1485840691 Loading…
3
[#3647] Fix: Assign default values in the GKDTrainer's constructor only when …
#3851 opened Aug 5, 2025 by seungduk-yanolja Loading…
2 of 5 tasks
1
2
Update profiling.py: fix scoping problems for wandb and mlflow
#3845 opened Aug 4, 2025 by markshinyounglee Loading…
5 tasks done
2
dynamic temperature
#3844 opened Aug 4, 2025 by shirinyamani • Draft
5 tasks
1
Optimize RLOO Trainer memory usage with string-level processing
#3837 opened Aug 2, 2025 by luckyvickyricky Loading…
2 of 5 tasks
1
3
[GSPO]: Refactor _compute_loss
#3835 opened Aug 1, 2025 by pramodith Loading…
2 of 5 tasks
1
3
support GSPO-token
#3820 opened Jul 31, 2025 by hjh0119 Loading…
15
GSPO docs - Sequence importance ratio and differences in relation to GRPO
#3816 opened Jul 31, 2025 by almeidava93 Loading…
2 of 5 tasks
2
Rloo final
#3801 opened Jul 29, 2025 by shirinyamani Loading…
5 tasks
5
Add vLLM server mode and VLM support to OnlineDPOTrainer
#3783 opened Jul 27, 2025 by vaelev Loading…
6 tasks done
1
11
change doc for num_iterations and steps_per_generation to hopefully make them more clear and differentiate between them more clearly
#3761 opened Jul 23, 2025 by avishaiElmakies Loading…
2 of 5 tasks
1
3
ProTip! What’s not been updated in a month: updated:<2025-07-25.