9 Pull requests merged by 2 people

• Bump vLLM and TRL

  #665 merged May 28, 2025

• Fix Weka refresh

  #666 merged May 28, 2025

• Set DP=2 for smol model evals

  #664 merged May 28, 2025

• Refresh Weka on Slurm

  #662 merged May 27, 2025

• Align EOS token ID between tokenizer and generation config

  #663 merged May 27, 2025

• Bump TRL / Transformers / LigerKernel

  #656 merged May 27, 2025

• Add OpenR1-Distill recipe

  #661 merged May 26, 2025

• Add better logging defaults for GRPO

  #657 merged May 25, 2025

• GRPO with codeforces problems

  #627 merged May 25, 2025

1 Issue closed by 1 person

• Does GRPO need to update its inference model

  #659 closed May 26, 2025

6 Issues opened by 5 people

• Question: How is the SFT training data processed in Open R1?

  #671 opened May 30, 2025

• Question: Approximate Training Time for 5 Epochs of SFT

  #670 opened May 30, 2025

• Do sft and grpo supported lora ?

  #669 opened May 30, 2025

• error: resolution-too-deep

  #668 opened May 29, 2025

• How to control the number of responses per query for each benchmark?

  #660 opened May 26, 2025

• GRPO training timeout after trl vllm-serve 'reset prefix cache'

  #658 opened May 26, 2025

7 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• grpo inference error

  #587 commented on May 27, 2025 • 0 new comments

• OpenR1-Qwen-7B achieves 47.40 on AIME24, better than reported!

  #622 commented on May 27, 2025 • 0 new comments

• Are AIME24 evals broken?

  #655 commented on May 27, 2025 • 0 new comments

• 🚀 Introducing simpleR1: A streamlined framework for training R1-like models based on TRL grpo_trainer

  #650 commented on May 28, 2025 • 0 new comments

• how can I get the prediction using the provided evaluation script?

  #625 commented on May 29, 2025 • 0 new comments

• Start agent traces

  #414 commented on May 25, 2025 • 0 new comments

• [WIP] R1-Zero-like experiments

  #569 commented on May 25, 2025 • 0 new comments