Search Results for "moe"

DeepEP is a communication library designed specifically to support Mixture-of-Experts (MoE) and expert parallelism (EP) deployments. Its core role is to implement high-throughput, low-latency all-to-all GPU communication kernels, which handle the dispatching of tokens to different experts (or shards) and then combining expert outputs back into the main data flow. Because MoE architectures require routing inputs to different experts, communication overhead can become a bottleneck — DeepEP addresses that by providing optimized GPU kernels and efficient dispatch/combining logic. ...

DualPipe is a bidirectional pipeline parallelism algorithm open-sourced by DeepSeek, introduced in their DeepSeek-V3 technical framework. The main goal of DualPipe is to maximize overlap between computation and communication phases during distributed training, thus reducing idle GPU time (i.e. “pipeline bubbles”) and improving cluster efficiency. Traditional pipeline parallelism methods (e.g. 1F1B or staggered pipelining) leave gaps because forward and backward phases can’t fully overlap...