perf(dsv4-fp4-mi355x-vllm): use AITER a16w4 MoE backend (+21% decode)#1989
Draft
jiacao-amd wants to merge 1 commit into
Draft
perf(dsv4-fp4-mi355x-vllm): use AITER a16w4 MoE backend (+21% decode)#1989jiacao-amd wants to merge 1 commit into
jiacao-amd wants to merge 1 commit into