Skip to content

Pull requests: huggingface/trl

New pull request New
62 Open 1,468 Closed
62 Open 1,468 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

add vllm support for token ids as input
#3280 opened Apr 11, 2025 by wybryan Loading…
Reward takes completion ids
#3272 opened Apr 9, 2025 by qgallouedec • Draft
5 tasks
🦙 Llama 4
#3267 opened Apr 9, 2025 by qgallouedec • Draft
5 tasks
[NOT MEANT TO BE MERGED] Log correct/incorrect lengths
#3263 opened Apr 8, 2025 by qgallouedec • Draft
[SFT] support for ring_attn in SFTTrainer
#3262 opened Apr 8, 2025 by kashif Loading…
5 tasks
2
[🐯+GRPO] Support FSDP + Fix bug when using LigerGRPO with DDP
#3260 opened Apr 8, 2025 by shivam15s • Draft
5 tasks
@kashif
1
Add a raw generate API to the vLLM server
#3227 opened Apr 3, 2025 by wilrop Loading…
5 tasks
3
Support iterable datasets in GRPO
#3226 opened Apr 3, 2025 by wilrop Loading…
5 tasks
1
7
feat(trainer): Support multi-role & consecutive turns in DataCollatorForCompletionOnlyLM (#3223)
#3224 opened Apr 3, 2025 by Kirili4ik Loading…
4 tasks done
1
update weight update process group
#3211 opened Apr 2, 2025 by ji-huazhong • Draft
5 tasks
2
Adding sampling parameters for vllm generation
#3210 opened Apr 2, 2025 by shaipranesh2 Loading…
6
Support for Models With Pre-Finetuned LoRA Adapters in GRPO: Add use_peft_as_reference Flag
#3196 opened Mar 31, 2025 by LoganVegnaSHOP Loading…
5 tasks done
1
7
GRPO: Scalable training with one LLM/node
#3186 opened Mar 31, 2025 by jglaser Loading…
3 of 5 tasks
3
🚀 Enhance GRPO VLLM server from sync to async and accelerate training
#3182 opened Mar 30, 2025 by binary-husky Loading…
3
Co-Locating vLLM w/ training to achieve higher throughput and GPU utilization
#3162 opened Mar 26, 2025 by toslali-ibm Loading…
2 of 5 tasks
1
20
Extend BCO Trainer dataset format support
#3134 opened Mar 22, 2025 by reihig-ut Loading…
1 of 5 tasks
Add GRPO/ Online DPO support for quantitative models when use vllm as infer backbone.
#3133 opened Mar 22, 2025 by maoulee Loading…
5
improvement(utils.py): simplify repeating completion string
#3122 opened Mar 20, 2025 by tpoisonooo Loading…
6
feat: Add Interleaved Trainer implementation
#3107 opened Mar 18, 2025 by ucalyptus2 Loading…
3 tasks done
1
Update sft trainer to include better packing
#3100 opened Mar 17, 2025 by Ishan-Kumar2 Loading…
4 tasks done
1
3
[GRPO] add vlm training capabilities to the trainer
#3072 opened Mar 13, 2025 by CompN3rd Loading…
3 of 5 tasks
28
[WIP] PEFT 🤝 Liger DPO
#3065 opened Mar 12, 2025 by SalmanMohammadi • Draft
5 tasks
2
[WIP] Iterative training scripts for SPIN and SPPO
#3011 opened Mar 5, 2025 by jkx19 • Draft
3 of 5 tasks
Fixing GRPO reward_func being a model with DeepSpeed ZeRO-3
#2984 opened Feb 28, 2025 by jamesbraza Loading…
4
Feature: Add SGLang as inference backend for generation in GRPO
#2981 opened Feb 28, 2025 by jhinpan Loading…
5 tasks done
32
ProTip! Add no:assignee to see everything that’s not assigned.