Feng Yao

Flash-RL. Implementation for FP8/INT8 Rollout for RL training without performence drop.

DenseMixer. Official implementation for DenseMixer: Improving MoE Post-Training with Precise Router Gradient

ReaL. Implementation and datasets for "Training Language Models to Generate Quality Code with Program Analysis Feedback"

verl. verl: Volcano Engine Reinforcement Learning for LLMs

branch-prediction. C

LER. Source code and dataset for AAAI2023 paper "Unsupervised Legal Evidence Retrieval via Contrastive Learning with Approximated Positive"