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 13, 08:15 UTC
Use official xformers-0.0.33 built for PT 2.9 (#28600)
Signed-off-by: Huy Do <huydhn@gmail.com> Co-authored-by: Roger Wang <hey@rogerw.io>
Huy Do committed
Nov 12, 2025 at 22:48 UTC
c33b87e7778d2a6900e73969c38785e0254f880b
1 file changed
+1
-2
requirements/cuda.txt
+1
-2
@@ -9,7 +9,6 @@ torch==2.9.0
9
torchaudio==2.9.0
10
# These must be updated alongside torch
11
torchvision==0.24.0 # Required for phi3v processor. See https://github.com/pytorch/vision?tab=readme-ov-file#installation for corresponding version
12
-# Build from https://github.com/facebookresearch/xformers/releases/tag/v0.0.32.post1
13
-xformers==0.0.33+5d4b92a5.d20251029; platform_system == 'Linux' and platform_machine == 'x86_64' # Requires PyTorch >= 2.9
12
+xformers==0.0.33; platform_system == 'Linux' and platform_machine == 'x86_64' # Requires PyTorch >= 2.9
13
# FlashInfer should be updated together with the Dockerfile
14
flashinfer-python==0.5.2