File tree Expand file tree Collapse file tree 1 file changed +1
-1
lines changed
blogs/software-tools-optimization/vllm-moe-guide Expand file tree Collapse file tree 1 file changed +1
-1
lines changed Original file line number Diff line number Diff line change @@ -699,7 +699,7 @@ vLLM allows combining multiple strategies to efficiently distribute models acros
699699 1 . ` --pipeline-parallel-size 2 --enable-expert-parallel ` → EP ** does NOT activate** (TP=1, DP=1 per stage)
700700 2 . ` --pipeline-parallel-size 2 --tensor-parallel-size 4 --enable-expert-parallel ` → EP activates (TP=4 per stage)
701701
702- 2 . ** PP \+ EP requires AITER (Advanced Inter-node Tensor-parallelism Engine Runtime) **
702+ 2 . ** PP \+ EP requires AITER**
703703 1 . AITER provides stability for complex communication patterns with PP
704704 2 . Enable with: ` VLLM_ROCM_USE_AITER=1 `
705705 3 . Without AITER: May experience instability or failures with large MoE models
You can’t perform that action at this time.
0 commit comments