TRL is a cutting-edge library designed for post-training foundation models using advanced techniques like Supervised Fine-Tuning (SFT), Group Relative Policy Optimization (GRPO), and Direct Preference Optimization (DPO). Built on top of the Transformers ecosystem, TRL supports a variety of model architectures and modalities, and can be scaled-up across various hardware setups.
An open reproduction of the viral watercolour-painting model: GRPO in TRL over an OpenEnv environment, with the reward split between HPSv3 and a pairwise judge scored against 178 hand-rated paintings — RL over taste.
Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.
An agentic-first RL framework for research. Ray · vLLM · NVIDIA AutoModel — the smallest PyTorch-native stack for 1T-class fully-async, multimodal, multi-turn agentic RL.
An interface library for RL post training with environments.
An occasional email when notable AI dev tools and models land in the directory. No spam, unsubscribe anytime.