Fix the CuDNN failure on bmm_fp8 when two libcudart.so exists. (#29201)

Co-authored-by: Brayden Zhong <brayden@radixark.ai>
This commit is contained in:
Brayden Zhong
2026-06-24 15:26:46 -07:00
committed by GitHub
co-authored by Brayden Zhong
parent 5f30fa258c
commit 3c95a87b66
@@ -270,7 +270,7 @@ if is_blackwell_supported() and is_flashinfer_available():
x_scale.reshape(1),
weight_scale.reshape(1),
out_dtype,
backend="auto",
backend="cublas",
).view(m, n)
@lru_cache(maxsize=1)