Add structure to `requirements/` directory (#39024)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>

.buildkite/ci_config.yaml

@@ -8,8 +8,8 @@ run_all_patterns:
   - "CMakeLists.txt"
   - "requirements/common.txt"
   - "requirements/cuda.txt"
-  - "requirements/build.txt"
-  - "requirements/test.txt"
+  - "requirements/build/cuda.txt"
+  - "requirements/test/cuda.txt"
   - "setup.py"
   - "csrc/"
   - "cmake/"

.buildkite/ci_config_intel.yaml

@@ -6,8 +6,8 @@ run_all_patterns:
   - "CMakeLists.txt"
   - "requirements/common.txt"
   - "requirements/xpu.txt"
-  - "requirements/build.txt"
-  - "requirements/test.txt"
+  - "requirements/build/cuda.txt"
+  - "requirements/test/cuda.txt"
   - "setup.py"
   - "csrc/"
   - "cmake/"

.buildkite/scripts/hardware_ci/run-hpu-test.sh

@@ -42,7 +42,7 @@ WORKDIR /workspace/vllm
 ENV no_proxy=localhost,127.0.0.1
 ENV PT_HPU_ENABLE_LAZY_COLLECTIVES=true
 
-RUN bash -c 'pip install -r <(sed "/^torch/d" requirements/build.txt)'
+RUN bash -c 'pip install -r <(sed "/^torch/d" requirements/build/cuda.txt)'
 RUN VLLM_TARGET_DEVICE=empty pip install --no-build-isolation -e .
 RUN pip install git+https://github.com/vllm-project/vllm-gaudi.git
 

.buildkite/test-amd.yaml

@@ -123,7 +123,7 @@ steps:
   soft_fail: true
   working_dir: "/vllm-workspace/tests"
   source_file_dependencies:
-  - requirements/nightly_torch_test.txt
+  - requirements/test/nightly-torch.txt
   - vllm/platforms/rocm.py
   commands:
   - bash standalone_tests/pytorch_nightly_dependency.sh

.buildkite/test_areas/pytorch.yaml

@@ -64,6 +64,6 @@ steps:
   device: h200_18gb
   soft_fail: true
   source_file_dependencies:
-  - requirements/nightly_torch_test.txt
+  - requirements/test/nightly-torch.txt
   commands:
   - bash standalone_tests/pytorch_nightly_dependency.sh

.buildkite/test_areas/quantization.yaml

@@ -9,7 +9,7 @@ steps:
   - vllm/model_executor/layers/quantization
   - tests/quantization
   commands:
-  # temporary install here since we need nightly, will move to requirements/test.in
+  # temporary install here since we need nightly, will move to requirements/test/cuda.in
   # after torchao 0.12 release, and pin a working version of torchao nightly here
 
   # since torchao nightly is only compatible with torch nightly currently

.github/workflows/macos-smoke-test.yml

@@ -32,7 +32,7 @@ jobs:
 
       - name: Install dependencies and build vLLM
         run: |
-          uv pip install -r requirements/cpu-build.txt --index-strategy unsafe-best-match
+          uv pip install -r requirements/build/cpu.txt --index-strategy unsafe-best-match
           uv pip install -r requirements/cpu.txt --index-strategy unsafe-best-match
           uv pip install -e . --no-build-isolation
         env:

.github/workflows/scripts/build.sh

@@ -9,7 +9,7 @@ PATH=${cuda_home}/bin:$PATH
 LD_LIBRARY_PATH=${cuda_home}/lib64:$LD_LIBRARY_PATH
 
 # Install requirements
-$python_executable -m pip install -r requirements/build.txt -r requirements/cuda.txt
+$python_executable -m pip install -r requirements/build/cuda.txt -r requirements/cuda.txt
 
 # Limit the number of parallel jobs to avoid OOM
 export MAX_JOBS=1

.gitignore

@@ -29,6 +29,7 @@ __pycache__/
 # Distribution / packaging
 .Python
 build/
+!requirements/build/
 cmake-build-*/
 CMakeUserPresets.json
 develop-eggs/

.pre-commit-config.yaml

@@ -39,15 +39,24 @@ repos:
   rev: 0.11.1
   hooks:
     - id: pip-compile
-      args: [requirements/test.in, -c, requirements/common.txt, -o, requirements/test.txt, --index-strategy, unsafe-best-match, --torch-backend, cu130, --python-platform, x86_64-manylinux_2_28, --python-version, "3.12"]
-      files: ^requirements/test\.(in|txt)$
+      args: [
+        requirements/test/cuda.in,
+        -c, requirements/cuda.txt,
+        -o, requirements/test/cuda.txt,
+        --index-strategy, unsafe-best-match,
+        --torch-backend, cu130,
+        --python-platform, x86_64-manylinux_2_28,
+        --python-version, "3.12",
+      ]
+      files: ^requirements/(common|cuda|test/cuda)\.(in|txt)$
     - id: pip-compile
       alias: pip-compile-rocm
       name: pip-compile-rocm
       args: [
-        requirements/rocm-test.in, -o, requirements/rocm-test.txt,
-        --index-strategy, unsafe-best-match,
+        requirements/test/rocm.in,
         -c, requirements/rocm.txt,
+        -o, requirements/test/rocm.txt,
+        --index-strategy, unsafe-best-match,
         --python-platform, x86_64-manylinux_2_28,
         --python-version, "3.12",
         # Exclude torch and CUDA/NVIDIA packages
@@ -108,14 +117,27 @@ repos:
         --no-emit-package, nvidia-nvshmem-cu13,
         --no-emit-package, nvidia-nvtx-cu13,
       ]
-      files: ^requirements/rocm-test\.(in|txt)$
+      files: ^requirements/(common|rocm|test/rocm)\.(in|txt)$
+    - id: pip-compile
+      alias: pip-compile-xpu
+      name: pip-compile-xpu
+      args: [
+        requirements/test/xpu.in,
+        -c, requirements/xpu.txt,
+        -o, requirements/test/xpu.txt,
+        --index-strategy, unsafe-best-match,
+        --torch-backend, xpu,
+        --python-platform, x86_64-manylinux_2_39,
+        --python-version, "3.12",
+      ]
+      files: ^requirements/(common|xpu|test/xpu)\.(in|txt)$
 - repo: local
   hooks:
   - id: format-torch-nightly-test
-    name: reformat nightly_torch_test.txt to be in sync with test.in
+    name: reformat test/nightly-torch.txt to be in sync with test/cuda.in
     language: python
     entry: python tools/pre_commit/generate_nightly_torch_test.py
-    files: ^requirements/test\.(in|txt)$
+    files: ^requirements/test/cuda\.(in|txt)$
   - id: mypy-local
     name: Run mypy locally for lowest supported Python version
     entry: python tools/pre_commit/mypy.py 0 "3.10"

AGENTS.md

@@ -72,11 +72,11 @@ uv pip install -e . --torch-backend=auto
 
 ```bash
 # Install test dependencies.
-# requirements/test.txt is pinned to x86_64; on other platforms, use the
+# requirements/test/cuda.txt is pinned to x86_64; on other platforms, use the
 # unpinned source file instead:
-uv pip install -r requirements/test.in    # resolves for current platform
+uv pip install -r requirements/test/cuda.in    # resolves for current platform
 # Or on x86_64:
-uv pip install -r requirements/test.txt
+uv pip install -r requirements/test/cuda.txt
 
 # Run a specific test file (use .venv/bin/python directly;
 # `source activate` does not persist in non-interactive shells):

docker/Dockerfile

@@ -204,7 +204,7 @@ ARG PYTORCH_CUDA_INDEX_BASE_URL
 ARG PYTORCH_NIGHTLY
 
 # Install build dependencies
-COPY requirements/build.txt requirements/build.txt
+COPY requirements/build/cuda.txt requirements/build/cuda.txt
 COPY use_existing_torch.py use_existing_torch.py
 COPY --from=base /workspace/torch_lib_versions.txt torch_lib_versions.txt
 
@@ -219,13 +219,13 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "${PYTORCH_NIGHTLY}" = "1" ]; then \
         echo "Installing build requirements without torch..." \
         && python3 use_existing_torch.py --prefix \
-        && uv pip install --python /opt/venv/bin/python3 -r requirements/build.txt \
+        && uv pip install --python /opt/venv/bin/python3 -r requirements/build/cuda.txt \
         && echo "Installing torch nightly..." \
         && uv pip install --python /opt/venv/bin/python3 $(cat torch_lib_versions.txt | grep -i "^torch=" | xargs) --pre \
         --index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/nightly/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.'); \
     else \
         echo "Installing build requirements..." \
-        && uv pip install --python /opt/venv/bin/python3 -r requirements/build.txt \
+        && uv pip install --python /opt/venv/bin/python3 -r requirements/build/cuda.txt \
         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.'); \
     fi
 
@@ -355,7 +355,7 @@ ARG PYTORCH_CUDA_INDEX_BASE_URL
 ARG PYTORCH_NIGHTLY
 
 # Install build dependencies
-COPY requirements/build.txt requirements/build.txt
+COPY requirements/build/cuda.txt requirements/build/cuda.txt
 COPY use_existing_torch.py use_existing_torch.py
 COPY --from=base /workspace/torch_lib_versions.txt torch_lib_versions.txt
 
@@ -370,13 +370,13 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "${PYTORCH_NIGHTLY}" = "1" ]; then \
         echo "Installing build requirements without torch..." \
         && python3 use_existing_torch.py --prefix \
-        && uv pip install --python /opt/venv/bin/python3 -r requirements/build.txt \
+        && uv pip install --python /opt/venv/bin/python3 -r requirements/build/cuda.txt \
         && echo "Installing torch nightly..." \
         && uv pip install --python /opt/venv/bin/python3 $(cat torch_lib_versions.txt | grep -i "^torch=" | xargs) --pre \
         --index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/nightly/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.'); \
     else \
         echo "Installing build requirements..." \
-        && uv pip install --python /opt/venv/bin/python3 -r requirements/build.txt \
+        && uv pip install --python /opt/venv/bin/python3 -r requirements/build/cuda.txt \
         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.'); \
     fi
 
@@ -450,8 +450,8 @@ ARG PYTORCH_NIGHTLY
 
 # Install development dependencies
 COPY requirements/lint.txt requirements/lint.txt
-COPY requirements/test.in requirements/test.in
-COPY requirements/test.txt requirements/test.txt
+COPY requirements/test/cuda.in requirements/test/cuda.in
+COPY requirements/test/cuda.txt requirements/test/cuda.txt
 COPY requirements/dev.txt requirements/dev.txt
 COPY use_existing_torch.py use_existing_torch.py
 COPY --from=base /workspace/torch_lib_versions.txt torch_lib_versions.txt
@@ -459,8 +459,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     if [ "${PYTORCH_NIGHTLY}" = "1" ]; then \
         echo "Installing dev requirements plus torch nightly..." \
         && python3 use_existing_torch.py --prefix \
-        && cat torch_lib_versions.txt >> requirements/test.in \
-        && uv pip compile requirements/test.in -o requirements/test.txt --index-strategy unsafe-best-match \
+        && cat torch_lib_versions.txt >> requirements/test/cuda.in \
+        && uv pip compile requirements/test/cuda.in -o requirements/test/cuda.txt --index-strategy unsafe-best-match \
         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/nightly/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.') \
         && uv pip install --python /opt/venv/bin/python3 $(cat torch_lib_versions.txt | xargs) --pre \
         -r requirements/dev.txt \
@@ -727,8 +727,8 @@ ARG PYTORCH_NIGHTLY
 
 # Install development dependencies (for testing)
 COPY requirements/lint.txt requirements/lint.txt
-COPY requirements/test.in requirements/test.in
-COPY requirements/test.txt requirements/test.txt
+COPY requirements/test/cuda.in requirements/test/cuda.in
+COPY requirements/test/cuda.txt requirements/test/cuda.txt
 COPY requirements/dev.txt requirements/dev.txt
 COPY use_existing_torch.py use_existing_torch.py
 COPY --from=base /workspace/torch_lib_versions.txt torch_lib_versions.txt
@@ -738,8 +738,8 @@ RUN --mount=type=cache,target=/root/.cache/uv \
         if [ "${PYTORCH_NIGHTLY}" = "1" ]; then \
             echo "Installing dev requirements plus torch nightly..." \
             && python3 use_existing_torch.py --prefix \
-            && cat torch_lib_versions.txt >> requirements/test.in \
-            && uv pip compile requirements/test.in -o requirements/test.txt --index-strategy unsafe-best-match \
+            && cat torch_lib_versions.txt >> requirements/test/cuda.in \
+            && uv pip compile requirements/test/cuda.in -o requirements/test/cuda.txt --index-strategy unsafe-best-match \
             --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/nightly/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.') \
             && uv pip install --system $(cat torch_lib_versions.txt | xargs) --pre \
             -r requirements/dev.txt \

docker/Dockerfile.cpu

@@ -107,10 +107,10 @@ RUN if [ "$TARGETARCH" = "arm64" ] && [ "$VLLM_CPU_X86" != "0" ]; then \
     fi
 
 # Copy build requirements
-COPY requirements/cpu-build.txt requirements/build.txt
+COPY requirements/build/cpu.txt requirements/build/cpu.txt
 
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -r requirements/build.txt
+    uv pip install -r requirements/build/cpu.txt
 
 COPY . .
 
@@ -127,26 +127,26 @@ FROM base AS vllm-test-deps
 WORKDIR /vllm-workspace
 
 # Copy test requirements
-COPY requirements/test.in requirements/cpu-test.in
+COPY requirements/test/cuda.in requirements/test/cpu.in
 
 RUN \
-    sed -i '/mamba_ssm/d' requirements/cpu-test.in && \
+    sed -i '/mamba_ssm/d' requirements/test/cpu.in && \
     remove_packages_not_supported_on_aarch64() { \
     case "$(uname -m)" in \
     aarch64|arm64) \
-    sed -i '/decord/d' requirements/cpu-test.in; \
-    sed -i '/terratorch/d' requirements/cpu-test.in; \
+    sed -i '/decord/d' requirements/test/cpu.in; \
+    sed -i '/terratorch/d' requirements/test/cpu.in; \
     ;; \
     esac; \
     }; \
     remove_packages_not_supported_on_aarch64 && \
-    sed -i 's/^torch==.*/torch==2.11.0/g' requirements/cpu-test.in && \
-    sed -i 's/torchaudio.*/torchaudio/g' requirements/cpu-test.in && \
-    sed -i 's/torchvision.*/torchvision/g' requirements/cpu-test.in && \
-    uv pip compile requirements/cpu-test.in -o requirements/cpu-test.txt --index-strategy unsafe-best-match --torch-backend cpu
+    sed -i 's/^torch==.*/torch==2.11.0/g' requirements/test/cpu.in && \
+    sed -i 's/torchaudio.*/torchaudio/g' requirements/test/cpu.in && \
+    sed -i 's/torchvision.*/torchvision/g' requirements/test/cpu.in && \
+    uv pip compile requirements/test/cpu.in -o requirements/test/cpu.txt --index-strategy unsafe-best-match --torch-backend cpu
 
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install -r requirements/cpu-test.txt
+    uv pip install -r requirements/test/cpu.txt
 
 ######################### DEV IMAGE #########################
 FROM vllm-build AS vllm-dev
@@ -168,7 +168,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,source=.git,target=.git \
     VLLM_TARGET_DEVICE=cpu python3 setup.py develop
 
-COPY --from=vllm-test-deps /vllm-workspace/requirements/cpu-test.txt requirements/test.txt
+COPY --from=vllm-test-deps /vllm-workspace/requirements/test/cpu.txt requirements/test/cpu.txt
 
 RUN --mount=type=cache,target=/root/.cache/uv \
     uv pip install -r requirements/dev.txt && \

docker/Dockerfile.nightly_torch

@@ -107,7 +107,7 @@ COPY . .
 RUN python3 use_existing_torch.py
 
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install --system -r requirements/build.txt
+    uv pip install --system -r requirements/build/cuda.txt
 
 ARG GIT_REPO_CHECK=0
 RUN --mount=type=bind,source=.git,target=.git \
@@ -261,7 +261,7 @@ FROM vllm-base as test
 COPY tests/ tests/
 
 # install build and runtime dependencies without stable torch version
-COPY requirements/nightly_torch_test.txt requirements/nightly_torch_test.txt
+COPY requirements/test/nightly-torch.txt requirements/test/nightly-torch.txt
 
 # This timeout (in seconds) is necessary when installing some dependencies via uv since it's likely to time out
 # Reference: https://github.com/astral-sh/uv/pull/1694
@@ -277,7 +277,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
 ENV HF_HUB_ENABLE_HF_TRANSFER 1
 
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install --system -r requirements/nightly_torch_test.txt
+    uv pip install --system -r requirements/test/nightly-torch.txt
 
 # Logging to confirm the torch versions
 RUN pip freeze | grep -E 'torch|vllm|flashinfer'

docker/Dockerfile.ppc64le

@@ -251,7 +251,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     make -C /numactl install && \
     # sentencepiece.pc is in some pkgconfig inside uv cache
     export PKG_CONFIG_PATH=$(find / -type d -name "pkgconfig" 2>/dev/null | tr '\n' ':') && \
-    nanobind_DIR=$(uv pip show nanobind | grep Location | sed 's/^Location: //;s/$/\/nanobind\/cmake/') && uv pip install -r /src/requirements/common.txt -r /src/requirements/cpu.txt -r /src/requirements/build.txt --no-build-isolation && \
+    nanobind_DIR=$(uv pip show nanobind | grep Location | sed 's/^Location: //;s/$/\/nanobind\/cmake/') && uv pip install -r /src/requirements/common.txt -r /src/requirements/cpu.txt -r /src/requirements/build/cuda.txt --no-build-isolation && \
     cd /src/ && \
     uv build --wheel --out-dir /vllmwheel/ --no-build-isolation && \
     uv pip install /vllmwheel/*.whl

docker/Dockerfile.rocm

@@ -329,7 +329,7 @@ RUN --mount=type=bind,from=export_vllm,src=/,target=/install \
     --mount=type=cache,target=/root/.cache/uv \
     cd /install \
     && uv pip install --system -r requirements/rocm.txt \
-    && uv pip install --system -r requirements/rocm-test.txt \
+    && uv pip install --system -r requirements/test/rocm.txt \
     && pip uninstall -y vllm \
     && uv pip install --system *.whl
 

docker/Dockerfile.s390x

@@ -262,7 +262,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
         $OPENCV_WHL_FILE \
         $OUTLINES_CORE_WHL_FILE \
         --index-strategy unsafe-best-match \
-        -r requirements/cpu-build.txt \
+        -r requirements/build/cpu.txt \
         -r requirements/cpu.txt
 
 

docker/Dockerfile.xpu

@@ -76,20 +76,14 @@ ENV UV_LINK_MODE="copy"
 RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,src=requirements/common.txt,target=/workspace/vllm/requirements/common.txt \
     --mount=type=bind,src=requirements/xpu.txt,target=/workspace/vllm/requirements/xpu.txt \
-    --mount=type=bind,src=requirements/xpu-test.in,target=/workspace/vllm/requirements/xpu-test.in \
+    --mount=type=bind,src=requirements/test/xpu.txt,target=/workspace/vllm/requirements/test/xpu.txt \
     uv pip install --upgrade pip && \
     uv pip install -r requirements/xpu.txt && \
-    uv pip compile /workspace/vllm/requirements/xpu-test.in \
-        -o /workspace/vllm/requirements/xpu-test.txt \
-        -c /workspace/vllm/requirements/xpu.txt \
-        --index-strategy unsafe-best-match \
-        --extra-index-url ${PIP_EXTRA_INDEX_URL} \
-        --python-version ${PYTHON_VERSION} && \
     uv pip install grpcio-tools protobuf nanobind && \
     source /opt/intel/oneapi/setvars.sh --force && \
     source /opt/intel/oneapi/ccl/2021.15/env/vars.sh --force && \
     export CMAKE_PREFIX_PATH="$(python3 -c 'import site; print(site.getsitepackages()[0])'):${CMAKE_PREFIX_PATH}" && \
-    uv pip install --no-build-isolation -r /workspace/vllm/requirements/xpu-test.txt
+    uv pip install --no-build-isolation -r /workspace/vllm/requirements/test/xpu.txt
 
 
 

docs/contributing/README.md

@@ -49,10 +49,10 @@ If you are developing vLLM's Python and CUDA/C++ code, install Pytorch first:
 uv pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu129
 ```
 
-Then install the necessary build dependencies from `requirements/build.txt`, skipping `torch` as it was installed in the previous step:
+Then install the necessary build dependencies from `requirements/build/cuda.txt`, skipping `torch` as it was installed in the previous step:
 
 ```bash
-grep -v '^torch==' requirements/build.txt | uv pip install -r -
+grep -v '^torch==' requirements/build/cuda.txt | uv pip install -r -
 ```
 
 Finally install vLLM using:

docs/contributing/incremental_build.md

@@ -16,10 +16,10 @@ Before setting up the incremental build:
 
 2. **CUDA Toolkit:** Verify that the NVIDIA CUDA Toolkit is correctly installed and `nvcc` is accessible in your `PATH`. CMake relies on `nvcc` to compile CUDA code. You can typically find `nvcc` in `$CUDA_HOME/bin/nvcc` or by running `which nvcc`. If you encounter issues, refer to the [official CUDA Toolkit installation guides](https://developer.nvidia.com/cuda-toolkit-archive) and vLLM's main [GPU installation documentation](../getting_started/installation/gpu.md#troubleshooting) for troubleshooting. The `CMAKE_CUDA_COMPILER` variable in your `CMakeUserPresets.json` should also point to your `nvcc` binary.
 
-3. **Build Tools:** It is highly recommended to install `ccache` for fast rebuilds by caching compilation results (e.g., `sudo apt install ccache` or `conda install ccache`). Also, ensure the core build dependencies like `cmake` and `ninja` are installed. These are installable through `requirements/build.txt` or your system's package manager.
+3. **Build Tools:** It is highly recommended to install `ccache` for fast rebuilds by caching compilation results (e.g., `sudo apt install ccache` or `conda install ccache`). Also, ensure the core build dependencies like `cmake` and `ninja` are installed. These are installable through `requirements/build/cuda.txt` or your system's package manager.
 
     ```console
-    uv pip install -r requirements/build.txt --torch-backend=auto
+    uv pip install -r requirements/build/cuda.txt --torch-backend=auto
     ```
 
 ## Setting up the CMake Build Environment