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 15, 21:00 UTC
[Build] Add UV_HTTP_TIMEOUT to avoid timeout during installation (#13850)
Signed-off-by: Yuan Tang <terrytangyuan@gmail.com>
Yuan Tang committed
Mar 5, 2025 at 20:09 UTC
71eaf8969b82faebb6f74beab0e652627f0e1068
1 file changed
+24
Dockerfile
+24
@@ -31,6 +31,10 @@ RUN echo 'tzdata tzdata/Areas select America' | debconf-set-selections \
31
RUN --mount=type=cache,target=/root/.cache/uv \
32
python3 -m pip install uv
33
34
+# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
35
+# Reference: https://github.com/astral-sh/uv/pull/1694
36
+ENV UV_HTTP_TIMEOUT=500
37
+
38
# Upgrade to GCC 10 to avoid https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92519
39
# as it was causing spam when compiling the CUTLASS kernels
40
RUN apt-get install -y gcc-10 g++-10
@@ -81,6 +85,10 @@ ARG TARGETPLATFORM
85
# install build dependencies
86
COPY requirements-build.txt requirements-build.txt
87
88
+# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
89
+# Reference: https://github.com/astral-sh/uv/pull/1694
90
+ENV UV_HTTP_TIMEOUT=500
91
+
92
RUN --mount=type=cache,target=/root/.cache/uv \
93
uv pip install --system -r requirements-build.txt
94
@@ -143,6 +151,10 @@ RUN if [ "$RUN_WHEEL_CHECK" = "true" ]; then \
151
#################### DEV IMAGE ####################
152
FROM base as dev
153
154
+# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
155
+# Reference: https://github.com/astral-sh/uv/pull/1694
156
+ENV UV_HTTP_TIMEOUT=500
157
+
158
COPY requirements-lint.txt requirements-lint.txt
159
COPY requirements-test.txt requirements-test.txt
160
COPY requirements-dev.txt requirements-dev.txt
@@ -181,6 +193,10 @@ RUN echo 'tzdata tzdata/Areas select America' | debconf-set-selections \
193
RUN --mount=type=cache,target=/root/.cache/uv \
194
python3 -m pip install uv
195
196
+# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
197
+# Reference: https://github.com/astral-sh/uv/pull/1694
198
+ENV UV_HTTP_TIMEOUT=500
199
+
200
# Workaround for https://github.com/openai/triton/issues/2507 and
201
# https://github.com/pytorch/pytorch/issues/107960 -- hopefully
202
# this won't be needed for future versions of this docker image
@@ -237,6 +253,10 @@ FROM vllm-base AS test
253
254
ADD . /vllm-workspace/
255
256
+# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
257
+# Reference: https://github.com/astral-sh/uv/pull/1694
258
+ENV UV_HTTP_TIMEOUT=500
259
+
260
# install development dependencies (for testing)
261
RUN --mount=type=cache,target=/root/.cache/uv \
262
uv pip install --system -r requirements-dev.txt
@@ -265,6 +285,10 @@ RUN mv vllm test_docs/
285
# base openai image with additional requirements, for any subsequent openai-style images
286
FROM vllm-base AS vllm-openai-base
287
288
+# This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
289
+# Reference: https://github.com/astral-sh/uv/pull/1694
290
+ENV UV_HTTP_TIMEOUT=500
291
+
292
# install additional dependencies for openai api server
293
RUN --mount=type=cache,target=/root/.cache/uv \
294
if [ "$TARGETPLATFORM" = "linux/arm64" ]; then \