@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / c494f96fbc
Fork 0

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 14, 05:30 UTC

Use UV_LINK_MODE=copy in Dockerfile to avoid hardlink fail (#22128)

Signed-off-by: mgoin <mgoin64@gmail.com>

Michael Goin committed Aug 5, 2025 at 09:57 UTC c494f96fbcf5e9f19f59e3dea6c2780aeb6c567f
1 file changed +10
docker/Dockerfile
+10
@@ -119,6 +119,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
119 # Reference: https://github.com/astral-sh/uv/pull/1694
120 ENV UV_HTTP_TIMEOUT=500
121 ENV UV_INDEX_STRATEGY="unsafe-best-match"
122 +# Use copy mode to avoid hardlink failures with Docker cache mounts
123 +ENV UV_LINK_MODE=copy
124
125 # Upgrade to GCC 10 to avoid https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92519
126 # as it was causing spam when compiling the CUTLASS kernels
@@ -181,6 +183,8 @@ COPY requirements/build.txt requirements/build.txt
183 # Reference: https://github.com/astral-sh/uv/pull/1694
184 ENV UV_HTTP_TIMEOUT=500
185 ENV UV_INDEX_STRATEGY="unsafe-best-match"
186 +# Use copy mode to avoid hardlink failures with Docker cache mounts
187 +ENV UV_LINK_MODE=copy
188
189 RUN --mount=type=cache,target=/root/.cache/uv \
190 uv pip install --system -r requirements/build.txt \
@@ -272,6 +276,8 @@ ARG PYTORCH_CUDA_INDEX_BASE_URL
276 # Reference: https://github.com/astral-sh/uv/pull/1694
277 ENV UV_HTTP_TIMEOUT=500
278 ENV UV_INDEX_STRATEGY="unsafe-best-match"
279 +# Use copy mode to avoid hardlink failures with Docker cache mounts
280 +ENV UV_LINK_MODE=copy
281
282 COPY requirements/lint.txt requirements/lint.txt
283 COPY requirements/test.txt requirements/test.txt
@@ -341,6 +347,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
347 # Reference: https://github.com/astral-sh/uv/pull/1694
348 ENV UV_HTTP_TIMEOUT=500
349 ENV UV_INDEX_STRATEGY="unsafe-best-match"
350 +# Use copy mode to avoid hardlink failures with Docker cache mounts
351 +ENV UV_LINK_MODE=copy
352
353 # Workaround for https://github.com/openai/triton/issues/2507 and
354 # https://github.com/pytorch/pytorch/issues/107960 -- hopefully
@@ -472,6 +480,8 @@ ARG PIP_EXTRA_INDEX_URL UV_EXTRA_INDEX_URL
480 # Reference: https://github.com/astral-sh/uv/pull/1694
481 ENV UV_HTTP_TIMEOUT=500
482 ENV UV_INDEX_STRATEGY="unsafe-best-match"
483 +# Use copy mode to avoid hardlink failures with Docker cache mounts
484 +ENV UV_LINK_MODE=copy
485
486 # install development dependencies (for testing)
487 RUN --mount=type=cache,target=/root/.cache/uv \