A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 3, 16:30 UTC
[Bug] Fix DeepSeek V4 `AttributeError: module 'cutlass.cute.nvgpu' has no attribute 'LoadCacheMode'` (#42342)
Signed-off-by: yewentao256 <zhyanwentao@126.com> Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com> Co-authored-by: Roger Wang <hey@rogerw.io>
Wentao Ye committed
May 14, 2026 at 05:00 UTC
23c85343fba6dd1bc346dd98e6ff3d9472e58b60
1 file changed
+1
-1
requirements/cuda.txt
+1
-1
@@ -21,7 +21,7 @@ nvidia-cudnn-frontend>=1.13.0,<1.19.0
21
fastsafetensors >= 0.2.2
22
23
# QuACK and Cutlass DSL for FA4 (cute-DSL implementation)
24
-nvidia-cutlass-dsl[cu13]>=4.4.2
24
+nvidia-cutlass-dsl[cu13]==4.5.0
25
quack-kernels>=0.3.3
26
27
# Tokenspeed_MLA for faster mla with spec decode