Skip to content

Pull requests: huggingface/trl

New pull request New
66 Open 1,347 Closed
66 Open 1,347 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
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

🚀 DeepSpeed integration documentation
#2993 opened Feb 28, 2025 by qgallouedec Loading…
5 tasks
1
📚 Update customization and distributing training documentation
#2991 opened Feb 28, 2025 by qgallouedec Loading…
5 tasks
1
Add support for additional generation kwargs in GRPO Trainer
#2989 opened Feb 28, 2025 by nopepper Loading…
3 of 5 tasks
9
Fixing GRPO reward_func being a model with DeepSpeed ZeRO-3
#2984 opened Feb 28, 2025 by jamesbraza Loading…
3
Feature(WIP): Add SGLang as inference backend for generation in GRPO
#2981 opened Feb 28, 2025 by jhinpan Loading…
5 tasks done
3
Supporting deepspeed>=0.16.4's rename
#2963 opened Feb 26, 2025 by jamesbraza Loading…
1
12
Support ReMax Algorithm
#2955 opened Feb 25, 2025 by liziniu Loading…
3 tasks done
3
[Models] Activation checkpointing from TrorchTune
#2954 opened Feb 25, 2025 by kashif • Draft
8
Agents
#2936 opened Feb 23, 2025 by August-murr Loading…
18
Provide more accurate error messages to make the program more robust.
#2932 opened Feb 22, 2025 by dignfei Loading…
4 tasks
3
Add the metrics completion_length_max and completion_length_min
#2930 opened Feb 22, 2025 by dignfei Loading…
4 tasks
4
Supporting multi-vLLM inference for GRPO
#2929 opened Feb 22, 2025 by ghrua Loading…
2 of 5 tasks
8
Liger GRPO support
#2926 opened Feb 21, 2025 by SalmanMohammadi • Draft
4 tasks
2
Remove CUDA synchronization in mean_token_accuracy
#2902 opened Feb 19, 2025 by cyyever Loading…
1 task done
5
GRPOTrainer adds support for OpenAI API-compatible servers to models that generate samples
#2901 opened Feb 19, 2025 by ZYM66 • Draft
2 of 5 tasks
5
Fast dataset truncate in SFTTrainer
#2898 opened Feb 18, 2025 by mariosasko Loading…
1 of 5 tasks
3
[Discussion] Agentic Framework Based on VLLM and E2B for RL
#2880 opened Feb 17, 2025 by August-murr • Draft
1
[GRPO] Reduce steps where loss starts to remain at 0, accelerate training
#2869 opened Feb 15, 2025 by zhangsheng377 Loading…
4
BCOTrainer version upgrade fixes
#2867 opened Feb 15, 2025 by claralp Loading…
3 of 5 tasks
4
Using model_wrapped can improve the generation speed by approximately 10 times on a single GPU.
#2859 opened Feb 14, 2025 by dignfei Loading…
2
[Liger] Liger KTO support
#2812 opened Feb 10, 2025 by vaibhavjindal Loading…
3 of 5 tasks
3
[draft] Use vLLM in LogCompletionsCallback
#2797 opened Feb 7, 2025 by tchang1997 • Draft
2 of 4 tasks
Remote GRPO ref model
#2763 opened Feb 4, 2025 by edbeeching • Draft
8
allow ref_model to be set in trainer to interface parity with other trainers
#2746 opened Feb 3, 2025 by winglian • Draft
5 tasks
WIP: RLOOV2
#2724 opened Jan 31, 2025 by mnoukhov • Draft
3 tasks
1
ProTip! Updated in the last three days: updated:>2025-02-25.