TRL v1.0: Post-Training Library Built to Move with the Field

In three linesHugging Face releases TRL v1.0, a post-training library for language model fine-tuning. Version 1.0 marks API stability and includes support for DPO, PPO, and optimizations for models like Llama and Mistral.

Summary generated by Claude — human-verified