Skip to content

Make MoE dispatch/MLP expert-axis batch sharding configurable (fix Mixtral EP throughput) - #4179

Merged
copybara-service[bot] merged 1 commit into
AI-Hypercomputer:mainfrom
ROCm:fix-moe-expert-parallel-sharding
Jun 23, 2026
Merged

Make MoE dispatch/MLP expert-axis batch sharding configurable (fix Mixtral EP throughput)#4179
copybara-service[bot] merged 1 commit into
AI-Hypercomputer:mainfrom
ROCm:fix-moe-expert-parallel-sharding

Commits

  1. Commits on Jun 22, 2026