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 2, 11:00 UTC
[Bugfix][CPU] Remove invalid extra deps (#43977)
Signed-off-by: jiang1.li <jiang1.li@intel.com>
Li, Jiang committed
May 29, 2026 at 22:02 UTC
3f6f508e14b7c5325b74cf1ab541c9330611b9a2
2 files changed
+25
-6
docker/Dockerfile.cpu
+25
-1
@@ -165,6 +165,23 @@ RUN --mount=type=cache,target=/root/.cache/uv \
165
--mount=type=cache,target=/vllm-workspace/.deps,sharing=locked \
166
VLLM_TARGET_DEVICE=cpu python3 setup.py bdist_wheel --dist-dir=dist --py-limited-api=cp38
167
168
+######################### TRITON-CPU BUILD IMAGE #########################
169
+FROM base AS vllm-triton-cpu-build
170
+
171
+WORKDIR /vllm-workspace
172
+
173
+RUN mkdir dist
174
+
175
+RUN --mount=type=cache,target=/root/.cache/uv \
176
+ --mount=type=cache,target=/root/.cache/ccache \
177
+ --mount=type=cache,target=/vllm-workspace/.deps,sharing=locked \
178
+ if [ "$TARGETARCH" = "amd64" ] || [ "$VLLM_CPU_X86" != "0" ]; then \
179
+ git clone --recurse-submodules "https://github.com/triton-lang/triton-cpu.git"; \
180
+ cd triton-cpu; \
181
+ git checkout "270e696d"; \
182
+ uv build --wheel --out-dir=../dist; \
183
+ fi
184
+
185
######################### TEST DEPS #########################
186
FROM base AS vllm-test-deps
187
@@ -257,7 +274,14 @@ WORKDIR /vllm-workspace
274
RUN --mount=type=cache,target=/root/.cache/uv \
275
--mount=type=cache,target=/root/.cache/ccache \
276
--mount=type=bind,from=vllm-build,src=/vllm-workspace/dist,target=dist \
260
- uv pip install "$(realpath dist/*.whl)[audio,triton-cpu]"
277
+ uv pip install "$(realpath dist/*.whl)[audio]"
278
+
279
+RUN --mount=type=cache,target=/root/.cache/uv \
280
+ --mount=type=cache,target=/root/.cache/ccache \
281
+ --mount=type=bind,from=vllm-triton-cpu-build,src=/vllm-workspace/dist,target=dist \
282
+ if [ "$TARGETARCH" = "amd64" ] || [ "$VLLM_CPU_X86" != "0" ]; then \
283
+ uv pip install "$(realpath dist/*.whl)"; \
284
+ fi
285
286
# Add labels to document build configuration
287
LABEL org.opencontainers.image.title="vLLM CPU"
setup.py
-5
@@ -1195,11 +1195,6 @@ setup(
1195
"opentelemetry-exporter-otlp>=1.26.0",
1196
"opentelemetry-semantic-conventions-ai>=0.4.1",
1197
],
1198
- "triton-cpu": [
1199
- "triton @ "
1200
- "git+https://github.com/triton-lang/triton-cpu.git@270e696d ; "
1201
- "platform_machine == 'x86_64'",
1202
- ], # Remove after stable release
1198
},
1199
cmdclass=cmdclass,
1200
package_data=package_data,