Train transformer language models with reinforcement learning.

SkyRL: A Modular Full-stack RL Library for LLMs

Scalable toolkit for efficient model reinforcement