DeepEP
DeepEP: an efficient expert-parallel communication library
About this project
DeepEP DeepEP (DeepEveryParallel) is a high-performance communication library for modern machine learning training and inference. The library currently focuses on expert parallelism (EP) — providing high-throughput and low-latency all-to-all GPU kernels (MoE dispatch and combine) with low-precision support including FP8 — while also offering experimental primitives for pipeline parallelism (PP), context parallelism (CP), and remote memory access (Engram), all designed for zero or minimal SM occupation. All kernels are compiled at runtime via a lightweight Just-In-Time (JIT) module, requiring no CUDA compilation during installation. Despite its lightweight design, DeepEP's performance…
Technologies
Project health
GitHub
Reviews
Built by
Maintain deepseek-ai/DeepEP? Claiming verifies admin access through your GitHub account and gives you control of this listing.
