Profile Qwen MoE router on vLLM 0.20
This commit is contained in:
@@ -47,6 +47,7 @@ timeout --signal=TERM --kill-after=30s 1380 \
|
||||
--device h20 \
|
||||
--models qwen3-a3b-30b-moe \
|
||||
--num_tensor_parallel_workers 1 2 4 \
|
||||
--max_tokens 8192 \
|
||||
--num_tokens_list "${TOKENS[@]}" \
|
||||
--profile_method cuda_event \
|
||||
--precision BF16 \
|
||||
|
||||
Reference in New Issue
Block a user