cleanup

Dockerfile.arm

@@ -51,12 +51,13 @@ ENV VLLM_CPU_DISABLE_AVX512=${VLLM_CPU_DISABLE_AVX512}
 RUN --mount=type=cache,target=/root/.cache/pip \
     --mount=type=cache,target=/root/.cache/ccache \
     --mount=type=bind,source=.git,target=.git \
-    VLLM_TARGET_DEVICE=cpu python3 setup.py bdist_wheel && \
+    pip install build -r requirements-build.txt && \
+    VLLM_TARGET_DEVICE=cpu python -m build --no-isolation --wheel && \
     pip install dist/*.whl && \
     rm -rf dist
 
 WORKDIR /workspace/
 
 RUN ln -s /workspace/vllm/tests && ln -s /workspace/vllm/examples && ln -s /workspace/vllm/benchmarks
 
-ENTRYPOINT ["python3", "-m", "vllm.entrypoints.openai.api_server"]
+ENTRYPOINT ["python3", "-m", "vllm.entrypoints.openai.api_server"]

Dockerfile.cpu

@@ -54,7 +54,7 @@ ENV VLLM_CPU_DISABLE_AVX512=${VLLM_CPU_DISABLE_AVX512}
 RUN --mount=type=cache,target=/root/.cache/pip \
     --mount=type=cache,target=/root/.cache/ccache \
     --mount=type=bind,source=.git,target=.git \
-    VLLM_TARGET_DEVICE=cpu pip install -v .
+    VLLM_TARGET_DEVICE=cpu pip install -v --no-build-isolation .
 
 WORKDIR /workspace/
 

Dockerfile.hpu

@@ -9,7 +9,7 @@ RUN pip install -v -r requirements-hpu.txt
 ENV no_proxy=localhost,127.0.0.1
 ENV PT_HPU_ENABLE_LAZY_COLLECTIVES=true
 
-RUN VLLM_TARGET_DEVICE=hpu python3 setup.py install
+RUN VLLM_TARGET_DEVICE=hpu pip install .
 
 # install development dependencies (for testing)
 RUN python3 -m pip install -e tests/vllm_test_utils

Dockerfile.neuron

@@ -20,15 +20,13 @@ ARG APP_MOUNT=/workspace
 VOLUME [ ${APP_MOUNT} ]
 WORKDIR ${APP_MOUNT}/vllm
 
+COPY requirements-build.txt /app/vllm/
 COPY requirements-neuron.txt /app/vllm/
 COPY requirements-common.txt /app/vllm/
 
-# FIXME: needs some way to provide build dependencies
 ENV PIP_EXTRA_INDEX_URL=https://pip.repos.neuron.amazonaws.com
 RUN --mount=type=cache,target=/root/.cache/pip \
-    cd /app/vllm && \
-    python3 -m pip install --pre -U -r requirements-neuron.txt
-
+    python3 -m pip install --pre -U -r requirements-neuron.txt -r requirements-build.txt
 
 
 COPY . .

Dockerfile.openvino

@@ -18,7 +18,7 @@ COPY . /workspace/vllm
 
 # build vLLM with OpenVINO backend
 RUN --mount=type=cache,target=/root/.cache/pip \
-    PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu"
+    PIP_EXTRA_INDEX_URL="https://download.pytorch.org/whl/cpu" \
     VLLM_TARGET_DEVICE="openvino" \
     python3 -m pip install -v /workspace/vllm/
 

Dockerfile.tpu

@@ -13,19 +13,16 @@ RUN --mount=type=cache,target=/var/cache/apt \
 RUN --mount=type=cache,target=/root/.cache/pip \
     --mount=type=bind,source=requirements-build.txt,target=requirements-build.txt \
     --mount=type=bind,source=requirements-tpu.txt,target=requirements-tpu.txt \
-    pip install -r requirements-tpu.txt
+    pip install -r requirements-tpu.txt -r requirements-build.txt
 
 COPY . .
 ARG GIT_REPO_CHECK=0
 RUN --mount=type=bind,source=.git,target=.git \
     if [ "$GIT_REPO_CHECK" != 0 ]; then bash tools/check_repo.sh; fi
 
-# FIXME: needs some way of providing build dependencies
-
 COPY . /workspace/vllm
-ENV VLLM_TARGET_DEVICE="tpu"
 RUN --mount=type=cache,target=/root/.cache/pip \
     --mount=type=bind,source=.git,target=.git \
-    cd /workspace/vllm && pip install --no-build-isolation -v -e .
+    VLLM_TARGET_DEVICE="tpu" pip install --no-build-isolation -v -e .
 
 CMD ["/bin/bash"]

_build_backend/vllm.py

@@ -66,7 +66,17 @@ def get_requires_for_build_wheel(  # type: ignore[no-redef]
         requirements_extras.append("torch==2.5.1")
         # TODO: add intel extension for pytorch?
     elif VLLM_TARGET_DEVICE == "cuda":
-        requirements_extras.append("torch==2.5.1")
+        if os.getenv("TARGETPLATFORM") == "linux/arm64":
+            # FIXME: ?
+            _check_for_extra_index_url(
+                "https://download.pytorch.org/whl/nightly/cu124")
+            requirements_extras.extend([
+                "torch==2.6.0.dev20241210+cu124",
+                "torchvision==0.22.0.dev20241215"
+            ])
+
+        else:
+            requirements_extras.append("torch==2.5.1")
     elif VLLM_TARGET_DEVICE == "rocm":
         # TODO: ? add support for multiple ROCM versions (6.3?)
         rocm_supported_versions = ("6.2", )

setup.py

@@ -11,7 +11,6 @@
 
 from packaging.version import Version, parse
 from setuptools import Extension, find_packages, setup
-from setuptools import build_meta as _orig
 from setuptools.build_meta import *
 from setuptools.command.build_ext import build_ext
 from setuptools.errors import SetupError
@@ -352,7 +351,8 @@ def _no_device() -> bool:
 
 
 def _is_cuda() -> bool:
-    return VLLM_TARGET_DEVICE == "cuda" and not (_is_neuron() or _is_tpu() or _is_hpu())
+    return VLLM_TARGET_DEVICE == "cuda" and not (_is_neuron() or _is_tpu()
+                                                 or _is_hpu())
 
 
 def _is_hip() -> bool: