Fix docker build cache

Signed-off-by: Shiming Zhang <[email protected]>

Author: wzshiming

.dockerignore

@@ -31,3 +31,4 @@ share/python-wheels/
 .installed.cfg
 *.egg
 MANIFEST
+docker/

docker/Dockerfile

@@ -267,7 +267,8 @@ WORKDIR /workspace
 
 COPY --from=csrc-build /workspace/dist /precompiled-wheels
 
-COPY . .
+# Exclude .git to avoid cache misses
+COPY --exclude=.git . .
 
 ARG GIT_REPO_CHECK=0
 RUN --mount=type=bind,source=.git,target=.git \
@@ -279,12 +280,22 @@ ENV VLLM_TARGET_DEVICE=${vllm_target_device}
 # Skip adding +precompiled suffix to version (preserves git-derived version)
 ENV VLLM_SKIP_PRECOMPILED_VERSION_SUFFIX=1
 
+ARG USE_GIT_VERSION=0
+RUN --mount=type=cache,target=/root/.cache/uv \
+    if [ "$USE_GIT_VERSION" = "0" ]; then \
+        if [ "${vllm_target_device}" = "cuda" ]; then \
+            export VLLM_PRECOMPILED_WHEEL_LOCATION=$(ls /precompiled-wheels/*.whl); \
+        fi && \
+        SETUPTOOLS_SCM_PRETEND_VERSION="0.0.0+src.build" python3 setup.py bdist_wheel --dist-dir=dist --py-limited-api=cp38; \
+    fi
 RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,source=.git,target=.git \
-    if [ "${vllm_target_device}" = "cuda" ]; then \
-        export VLLM_PRECOMPILED_WHEEL_LOCATION=$(ls /precompiled-wheels/*.whl); \
-    fi && \
-    python3 setup.py bdist_wheel --dist-dir=dist --py-limited-api=cp38
+    if [ "$USE_GIT_VERSION" != "0" ]; then \
+        if [ "${vllm_target_device}" = "cuda" ]; then \
+            export VLLM_PRECOMPILED_WHEEL_LOCATION=$(ls /precompiled-wheels/*.whl); \
+        fi && \
+        python3 setup.py bdist_wheel --dist-dir=dist --py-limited-api=cp38; \
+    fi
 
 # Install DeepGEMM from source
 ARG DEEPGEMM_GIT_REF