ART
Train AI agents with reinforcement learning and supervised fine-tuning
All Skills in This Repository (3)
Pure Emerald Level Indicators
OpenPipefix-art-issues
Fix GitHub issues in the OpenPipe/ART repository and create draft pull requests.
OpenPipetrain-rl
Create reinforcement learning training scripts for language models using the ART framework.
OpenPipetrain-sft
Generate Python scripts for Supervised Fine-Tuning of language models in the ART framework.
Frequently Asked Questions
FAQPage SchemaHow to install ART?▼
Run `npx skills add OpenPipe/ART --all -g -y` in your terminal to install all skills in this suite globally.
What is ART used for?▼
ART is an open-source framework for training LLM agents with reinforcement learning (GRPO) and supervised fine-tuning so they improve from experience on real tasks.
How do I train an agent with reinforcement learning using ART?▼
The train-rl skill walks you through one question at a time, then generates a complete runnable RL training script with reward functions, RULER judging, and your chosen backend.
Do I need my own GPU to train models with ART?▼
No. The ServerlessBackend trains on managed remote GPUs using a W&B API key, while LocalBackend is available if you prefer your own GPU.
Can ART fine-tune a model from my own dataset?▼
Yes. The train-sft skill validates your JSONL dataset, computes sensible hyperparameters, and generates a ready-to-run fine-tuning or distillation script.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core