From 4860bcee415a87d54533c20f197c3336784eac7c Mon Sep 17 00:00:00 2001
From: biondizzle <biondizzle@gmail.com>
Date: Fri, 3 Apr 2026 08:05:44 +0000
Subject: [PATCH] Skip LMCache CUDA extensions (NO_CUDA_EXT=1)

PyTorch 2.9.0+cu130 was compiled with CUDA 12.8 but container has CUDA 13.0.
Skip CUDA extension build to avoid version mismatch.
---
 vllm/Dockerfile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/vllm/Dockerfile b/vllm/Dockerfile
index fca4f89..dd83372 100644
--- a/vllm/Dockerfile
+++ b/vllm/Dockerfile
@@ -98,7 +98,7 @@ RUN git clone https://github.com/LMCache/LMCache.git && \
 # PyTorch version is dated in LMCache
     sed -i '/torch/d' pyproject.toml && \
     uv pip install setuptools_scm && \
-    MAX_JOBS=8 python -m build --wheel --no-isolation && \
+    NO_CUDA_EXT=1 MAX_JOBS=8 python -m build --wheel --no-isolation && \
     cp dist/*.whl /wheels/