ART
ART copied to clipboard
OpenPipe
Agent Reinforcement Trainer: train multi-step agents for real-world tasks using GRPO. Give your agents on-the-job training. Reinforcement learning for Qwen2.5, Qwen3, Llama, and more!
Add functionality that allows downloading a checkpoint to a local path. Add a method that deletes a model from the serverless backend.
## Summary This PR enables proper GRPO training with importance sampling when using offline trajectory data (e.g., from vLLM traces). It includes three complementary fixes: ### 1. Extract logprobs from...
I'd like to be able to use openai openai>=1.107.2. It still under the same major version, so you should change the above to `
It would be nice if art would support the _Dr. GRPO_ optimizer. Basically, _Dr. GRPO_ is supposed to remove the response length bias which is present in the original GRPO,...
This PR swaps out the use of OpenRouter in the MCP generation code to LiteLLM instead which is more in line with other parts of the codebase (RULER) and doesn’t...
I’m running a customized version of art-e with additional tools locally. Training proceeds normally until the first validation stage, where it gets stuck for hours. I am running Qwen2.5-1.5B-Instruct on...
Running locally ART's 2048.ipynb notebook in Docker, skipping the first cell. Training freezes at 0% "train" after the "gather" stage is complete. GPU utilization is at 0% in nvidia-smi. Unsloth's...
Hi guys, I am trying to run Qwen 3B Instruct model on a GPU with 24 GB VRAM, but when the VLLM is creating CUDA graphs, it goes out of...
Hi OpenPipe team, big fan of your work. Was wondering if VLM support is coming in the near future
