`pip install -e .` error

I want to install the latest main branch through the following methods:

However, it failed and encountered the following error:

Building editable for vllm (pyproject.toml) … errorerror: subprocess-exited-with-error

× Building editable for vllm (pyproject.toml) did not run successfully.│ exit code: 1╰─> [233 lines of output]/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/_subclasses/functional_tensor.py:276: UserWarning: Failed to initialize NumPy: No module named ‘numpy’ (Triggered internally at /pytorch/torch/csrc/utils/tensor_numpy.cpp:81.)cpu = _conversion_method_template(device=torch.device(“cpu”))running editable_wheelcreating /fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-infowriting /fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-info/PKG-INFOwriting dependency_links to /fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-info/dependency_links.txtwriting entry points to /fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-info/entry_points.txtwriting requirements to /fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-info/requires.txtwriting top-level names to /fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-info/top_level.txtwriting manifest file ‘/fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-info/SOURCES.txt’reading manifest template ‘MANIFEST.in’adding license file ‘LICENSE’writing manifest file ‘/fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm.egg-info/SOURCES.txt’creating ‘/fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm-0.10.1.dev231+g9cb497bfa.cu124.dist-info’creating /fs-computility/llmeval/zhudongsheng/tmpdir/pip-wheel-uszgpjve/.tmp-o9gr936l/vllm-0.10.1.dev231+g9cb497bfa.cu124.dist-info/WHEELrunning build_pyrunning build_ext– The CXX compiler identification is GNU 9.4.0– Detecting CXX compiler ABI info– Detecting CXX compiler ABI info - done– Check for working CXX compiler: /usr/bin/c++ - skipped– Detecting CXX compile features– Detecting CXX compile features - done– Build type: RelWithDebInfo– Target device: cuda– Found Python: /fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/bin/python3.10 (found version “3.10.18”) found components: Interpreter Development.Module Development.SABIModule– Found python matching: /fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/bin/python3.10.– Found CUDA: /usr/local/cuda (found version “12.4”)– The CUDA compiler identification is NVIDIA 12.4.131 with host compiler GNU 9.4.0– Detecting CUDA compiler ABI info– Detecting CUDA compiler ABI info - done– Check for working CUDA compiler: /usr/local/cuda/bin/nvcc - skipped– Detecting CUDA compile features– Detecting CUDA compile features - done– Found CUDAToolkit: /usr/local/cuda/include (found version “12.4.131”)– Performing Test CMAKE_HAVE_LIBC_PTHREAD– Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed– Looking for pthread_create in pthreads– Looking for pthread_create in pthreads - not found– Looking for pthread_create in pthread– Looking for pthread_create in pthread - found– Found Threads: TRUE– PyTorch: CUDA detected: 12.4– PyTorch: CUDA nvcc is: /usr/local/cuda/bin/nvcc– PyTorch: CUDA toolkit directory: /usr/local/cuda– PyTorch: Header version is: 12.4– Found Python: /fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/bin/python3.10 (found version “3.10.18”) found components: InterpreterCMake Warning at /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Caffe2/public/cuda.cmake:140 (message):Failed to compute shorthash for libnvrtc.soCall Stack (most recent call first):/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:86 (include)/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package)CMakeLists.txt:80 (find_package)

CMake Warning (dev) at /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/cmake/data/share/cmake-4.0/Modules/FindPackageHandleStandardArgs.cmake:430 (message): The package name passed to find_package_handle_standard_args() (nvtx3) does not match the name of the calling package (Caffe2). This can lead to problems in calling code that expects find_package() result variables (e.g., `_FOUND`) to follow a certain pattern. Call Stack (most recent call first): /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Caffe2/public/cuda.cmake:178 (find_package_handle_standard_args) /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:86 (include) /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package) CMakeLists.txt:80 (find_package) This warning is for project developers. Use -Wno-dev to suppress it. – Could NOT find nvtx3 (missing: nvtx3_dir) CMake Warning at /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Caffe2/public/cuda.cmake:184 (message): Cannot find NVTX3, find old NVTX instead Call Stack (most recent call first): /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Caffe2/Caffe2Config.cmake:86 (include) /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Torch/TorchConfig.cmake:68 (find_package) CMakeLists.txt:80 (find_package) – USE_CUDNN is set to 0. Compiling without cuDNN support – USE_CUSPARSELT is set to 0. Compiling without cuSPARSELt support – USE_CUDSS is set to 0. Compiling without cuDSS support – USE_CUFILE is set to 0. Compiling without cuFile support – Autodetected CUDA architecture(s): 8.0 8.0 – Added CUDA NVCC flags for: -gencode;arch=compute_80,code=sm_80 CMake Warning at /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Torch/TorchConfig.cmake:22 (message): static library kineto_LIBRARY-NOTFOUND not found. Call Stack (most recent call first): /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/share/cmake/Torch/TorchConfig.cmake:125 (append_torchlib_if_found) CMakeLists.txt:80 (find_package) – Found Torch: /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/torch/lib/libtorch.so – CUDA target architectures: 8.0 – CUDA supported target architectures: 8.0 – FetchContent base directory: /fs-computility/llmeval/zhudongsheng/program/vllm/.deps – Enabling cumem allocator extension. – CMake Version: 4.0.3 – CUTLASS 4.0.0 – Found CUDAToolkit: /usr/local/cuda/targets/x86_64-linux/include (found version “12.4.131”) – CUDART: /usr/local/cuda/lib64/libcudart.so – CUDA Driver: /usr/local/cuda/lib64/stubs/libcuda.so – NVRTC: /usr/local/cuda/lib64/libnvrtc.so – Default Install Location: install – Found Python3: /fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/bin/python3.10 (found suitable version “3.10.18”, minimum required is “3.5”) found components: Interpreter – CUDA Compilation Architectures: 70;72;75;80;86;87;89;90;90a – Enable caching of reference results in conv unit tests – Enable rigorous conv problem sizes in conv unit tests – Grid Dependency Control (GDC) is enabled for SM100 kernels (required for programmatic dependent launches). – Using the following NVCC flags: --expt-relaxed-constexpr -ftemplate-backtrace-limit=0 -DCUTLASS_TEST_LEVEL=0 -DCUTLASS_TEST_ENABLE_CACHED_RESULTS=1 -DCUTLASS_CONV_UNIT_TEST_RIGOROUS_SIZE_ENABLED=1 -DCUTLASS_DEBUG_TRACE_LEVEL=0 -Xcompiler=-Wconversion -Xcompiler=-fno-strict-aliasing -lineinfo – Configuring cublas … – cuBLAS Disabled. – Configuring cuBLAS … done. – Marlin generation script hash: 4efa30233df32fb3448c4db1511d0664 – Last run Marlin generate script hash: CMake Error at CMakeLists.txt:337 (message): Marlin generation failed. Result: “1” Check the log for details: /fs-computility/llmeval/zhudongsheng/tmpdir/tmpnhcz1dma.build-temp/marlin_generation.log – Configuring incomplete, errors occurred! Traceback (most recent call last): File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 139, in run self._create_wheel_file(bdist_wheel) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 340, in _create_wheel_file files, mapping = self._run_build_commands(dist_name, unpacked, lib, tmp) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 263, in _run_build_commands self._run_build_subcommands() File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 290, in _run_build_subcommands self.run_command(name) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/cmd.py”, line 357, in run_command self.distribution.run_command(command) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/dist.py”, line 1104, in run_command super().run_command(command) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/dist.py”, line 1021, in run_command cmd_obj.run() File “”, line 269, in run File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/build_ext.py”, line 99, in run _build_ext.run(self) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/command/build_ext.py”, line 368, in run self.build_extensions() File “”, line 231, in build_extensions File “”, line 209, in configure File “/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10/subprocess.py”, line 369, in check_call raise CalledProcessError(retcode, cmd) subprocess.CalledProcessError: Command ‘[‘cmake’, ‘/fs-computility/llmeval/zhudongsheng/program/vllm’, ‘-G’, ‘Ninja’, ‘-DCMAKE_BUILD_TYPE=RelWithDebInfo’, ‘-DVLLM_TARGET_DEVICE=cuda’, ‘-DVLLM_PYTHON_EXECUTABLE=/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/bin/python3.10’, ‘-DVLLM_PYTHON_PATH=/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/site:/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python310.zip:/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10:/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10/lib-dynload:/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages:/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/normal/lib/python3.10/site-packages:/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_vendor’, ‘-DFETCHCONTENT_BASE_DIR=/fs-computility/llmeval/zhudongsheng/program/vllm/.deps’, ‘-DNVCC_THREADS=1’, ‘-DCMAKE_JOB_POOL_COMPILE:STRING=compile’, ‘-DCMAKE_JOB_POOLS:STRING=compile=128’, ‘-DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc’]’ returned non-zero exit status 1. /fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/dist.py:1021: _DebuggingTips: Problem in editable installation. !! ******************************************************************************** An error happened while installing `vllm` in editable mode. The following steps are recommended to help debug this problem: - Try to install the project normally, without using the editable mode. Does the error still persist? (If it does, try fixing the problem before attempting the editable mode). - If you are using binary extensions, make sure you have all OS-level dependencies installed (e.g. compilers, toolchains, binary libraries, …). - Try the latest version of setuptools (maybe the error was already fixed). - If you (or your project dependencies) are using any setuptools extension or customization, make sure they support the editable mode. After following the steps above, if the problem still persists and you think this is related to how setuptools handles editable installations, please submit a reproducible example (see How to create a Minimal, Reproducible Example - Help Center - Stack Overflow) to: See Development Mode (a.k.a. “Editable Installs”) - setuptools 80.9.0 documentation for details. ******************************************************************************** !! cmd_obj.run() Traceback (most recent call last): File “/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py”, line 389, in main() File “/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py”, line 373, in main json_out[“return_val”] = hook(**hook_input[“kwargs”]) File “/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py”, line 303, in build_editable return hook(wheel_directory, config_settings, metadata_directory) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/build_meta.py”, line 468, in build_editable return self._build_with_temp_dir( File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/build_meta.py”, line 404, in _build_with_temp_dir self.run_setup() File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/build_meta.py”, line 317, in run_setup exec(code, locals()) File “”, line 690, in File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_init_.py”, line 117, in setup return distutils.core.setup(**attrs) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/core.py”, line 186, in setup return run_commands(dist) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/core.py”, line 202, in run_commands dist.run_commands() File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/dist.py”, line 1002, in run_commands self.run_command(cmd) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/dist.py”, line 1104, in run_command super().run_command(command) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/dist.py”, line 1021, in run_command cmd_obj.run() File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 139, in run self._create_wheel_file(bdist_wheel) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 340, in _create_wheel_file files, mapping = self._run_build_commands(dist_name, unpacked, lib, tmp) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 263, in _run_build_commands self._run_build_subcommands() File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/editable_wheel.py”, line 290, in _run_build_subcommands self.run_command(name) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/cmd.py”, line 357, in run_command self.distribution.run_command(command) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/dist.py”, line 1104, in run_command super().run_command(command) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/dist.py”, line 1021, in run_command cmd_obj.run() File “”, line 269, in run File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/command/build_ext.py”, line 99, in run _build_ext.run(self) File “/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_distutils/command/build_ext.py”, line 368, in run self.build_extensions() File “”, line 231, in build_extensions File “”, line 209, in configure File “/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10/subprocess.py”, line 369, in check_call raise CalledProcessError(retcode, cmd) subprocess.CalledProcessError: Command ‘[‘cmake’, ‘/fs-computility/llmeval/zhudongsheng/program/vllm’, ‘-G’, ‘Ninja’, ‘-DCMAKE_BUILD_TYPE=RelWithDebInfo’, ‘-DVLLM_TARGET_DEVICE=cuda’, ‘-DVLLM_PYTHON_EXECUTABLE=/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/bin/python3.10’, ‘-DVLLM_PYTHON_PATH=/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/site:/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python310.zip:/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10:/fs-computility/llmeval/zhudongsheng/miniconda3/envs/vllm/lib/python3.10/lib-dynload:/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages:/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/normal/lib/python3.10/site-packages:/fs-computility/llmeval/zhudongsheng/tmpdir/pip-build-env-ru9bvuwx/overlay/lib/python3.10/site-packages/setuptools/_vendor’, ‘-DFETCHCONTENT_BASE_DIR=/fs-computility/llmeval/zhudongsheng/program/vllm/.deps’, ‘-DNVCC_THREADS=1’, ‘-DCMAKE_JOB_POOL_COMPILE:STRING=compile’, ‘-DCMAKE_JOB_POOLS:STRING=compile=128’, ‘-DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc’]’ returned non-zero exit status 1. [end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.ERROR: Failed building editable for vllmFailed to build vllmERROR: Failed to build installable wheels for some pyproject.toml based projects (vllm)

你的报错信息显示在编译 vLLM 的过程中,CMake 阶段 Marlin kernel 生成失败,导致整个 pip install -e . 过程终止。这是 vLLM 主分支近期常见的构建问题,通常与 CUDA、PyTorch、依赖项或源码状态有关。根据官方文档和社区反馈,建议如下:

  1. 确认 CUDA 工具链和 PyTorch 版本兼容,并确保 nvcc 在 PATH 中,CUDA_HOME 设置正确。
  2. 尝试清理依赖缓存,并用如下命令重新安装依赖和编译(推荐用 ccache 加速多次编译):
    conda install ccache
    CCACHE_NOHASHDIR="true" pip install --no-build-isolation -e .
    
  3. 如果是 Marlin kernel 生成失败,可先检查 /fs-computility/llmeval/zhudongsheng/tmpdir/tmpnhcz1dma.build-temp/marlin_generation.log 日志,里面会有详细的错误原因。
  4. 如遇到依赖包冲突或 setuptools 版本问题,可尝试降级 setuptools 到 69.5.1 或升级到最新版本再试。
  5. 如果你只修改 Python 代码,可以用预编译 wheel 方式安装,避免本地编译 C++/CUDA:
    VLLM_USE_PRECOMPILED=1 pip install --editable .
    
    但如果你需要编译 C++/CUDA,必须解决本地编译问题。

详细步骤和说明可参考官方文档:vLLM GPU 安装文档编译加速与常见问题

需要更详细的排查建议或日志分析吗?

Sources:

Unless you’re planning on modifying the C++/CUDA code I would recommend using:

VLLM_USE_PRECOMPILED=1 uv pip install -e .

This will allow you to skip the C++/CUDA compilation. (Also I’m recommending using uv pip because it’s amazing Index | uv)

Yes, this is correct. If you are only changing Python code, you can use VLLM_USE_PRECOMPILED=1 uv pip install -e . to install vLLM in editable mode without compiling C++/CUDA, leveraging precompiled binaries. This is much faster and is the recommended workflow for Python-only development, as confirmed in the official docs and community discussions. Using uv pip is also encouraged for its speed and better dependency management[^1][^2][^3].

If you need to modify C++/CUDA code, you must do a full build from source instead. Would you like more detail on troubleshooting or workflow for either case?

Sources: