-
vllmellm authored
[Bugfix][ROCm]Fix Qwen3-Next-80B-A3B-Thinking inference and optimize non-standard block size (544) support under rocm_atten (#31380) Signed-off-by:vllmellm <vllm.ellm@embeddedllm.com>
1a19e9cd
[Bugfix][ROCm]Fix Qwen3-Next-80B-A3B-Thinking inference and optimize non-standard block size (544) support under rocm_atten (#31380)
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>