This repository contains examples that show how to train LLMs with reinforcement learning and how to build agents. In these examples, you will learn to:

• Trace rollouts with Weave, view prompts, outputs, rewards, etc. in one place
• Use Weave with popular frameworks like TRL, verl, OpenPipe, and verifiers
• Use the OpenPipe serverless API to train models calls without hosting your own stack
• Build and test agents with open models using RL, with repeatable logs and evals

Sno	Framework	Code
1.	TRL	Post-training Qwen2.5 On NuminaMath Dataset
2.	TRL	Post-training with GSPO algorithm
3.	ART-Serverless RL	SQLFixer