trl
Subscription: Developer
Active
Subscription: Developer
A Hugging Face library for training transformer language models with reinforcement learning, providing implementations of PPO, DPO, RLHF, and reward modeling techniques for aligning large language models with human preferences and feedback.