[Core] drop duplicate VLLM_USE_DEEP_GEMM check (#48687)
Signed-off-by: Andy Xie <andy.xning@gmail.com>
This commit is contained in:
@@ -175,9 +175,7 @@ def kernel_warmup(worker: "Worker", *, process_local_only: bool = False):
|
||||
|
||||
# Deep GEMM warmup
|
||||
do_deep_gemm_warmup = (
|
||||
envs.VLLM_USE_DEEP_GEMM
|
||||
and is_deep_gemm_supported()
|
||||
and envs.VLLM_DEEP_GEMM_WARMUP != "skip"
|
||||
is_deep_gemm_supported() and envs.VLLM_DEEP_GEMM_WARMUP != "skip"
|
||||
)
|
||||
if do_deep_gemm_warmup:
|
||||
model = worker.get_model()
|
||||
|
||||
Reference in New Issue
Block a user