Open source repositories tagged with #expert-parallelism, ranked by health score.
A scalable, agentic-first, and HuggingFace-native RL framework for research (9k lines).