Windows11 python 3.12 cuda 13.0 torch 2.9.1 Compilation failed

[wzgrx]

Environment
window11 4090 24gb python3.12 cuda12.9.1

pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu130

When trying to install from source on Windows using Python 3.12, the build fails
🖥️ Environment
OS: Windows 11
Python: 3.12
PyTorch: 2.4+ (compiled with CUDA 12.9.1)
NVCC: 13.0
CUDA: 12.9.1

[notice] A new release of pip is available: 25.0.1 -> 25.3
[notice] To update, run: python.exe -m pip install --upgrade pip
(venv) PS E:\SageAttention> # 1. 设置环境变量
(venv) PS E:\SageAttention> $env:EXT_PARALLEL = "4"
(venv) PS E:\SageAttention> $env:NVCC_APPEND_FLAGS = "--threads 8"
(venv) PS E:\SageAttention> $env:MAX_JOBS = "32"
(venv) PS E:\SageAttention>
(venv) PS E:\SageAttention> # 2. 安装依赖
(venv) PS E:\SageAttention> pip install packaging setuptools wheel
Requirement already satisfied: packaging in e:\sageattention\venv\lib\site-packages (25.0)
Requirement already satisfied: setuptools in e:\sageattention\venv\lib\site-packages (70.2.0)
Requirement already satisfied: wheel in e:\sageattention\venv\lib\site-packages (0.45.1)

[notice] A new release of pip is available: 25.0.1 -> 25.3
[notice] To update, run: python.exe -m pip install --upgrade pip
(venv) PS E:\SageAttention>
(venv) PS E:\SageAttention> # 3. 尝试安装
(venv) PS E:\SageAttention> python setup.py install
Target compute capabilities: {'12.0'}
running install
E:\SageAttention\venv\Lib\site-packages\setuptools_distutils\cmd.py:66: SetuptoolsDeprecationWarning: setup.py install is deprecated.
!!

    ********************************************************************************
    Please avoid running ``setup.py`` directly.
    Instead, use pypa/build, pypa/installer or other
    standards-based tools.

    See https://blog.ganssle.io/articles/2021/10/setup-py-deprecated.html for details.
    ********************************************************************************

!!
self.initialize_options()
E:\SageAttention\venv\Lib\site-packages\setuptools_distutils\cmd.py:66: EasyInstallDeprecationWarning: easy_install command is deprecated.
!!

    ********************************************************************************
    Please avoid running ``setup.py`` and ``easy_install``.
    Instead, use pypa/build, pypa/installer or other
    standards-based tools.

    See https://github.com/pypa/setuptools/issues/917 for details.
    ********************************************************************************

!!
self.initialize_options()
running bdist_egg
running egg_info
writing sageattention.egg-info\PKG-INFO
writing dependency_links to sageattention.egg-info\dependency_links.txt
writing top-level names to sageattention.egg-info\top_level.txt
W1202 18:13:09.956000 28948 venv\Lib\site-packages\torch\utils\cpp_extension.py:630] Attempted to use ninja as the BuildExtension backend but we could not find ninja.. Falling back to using the slow distutils backend.
reading manifest file 'sageattention.egg-info\SOURCES.txt'
reading manifest template 'MANIFEST.in'
warning: no files found matching '.hpp' under directory 'csrc'
warning: no previously-included files matching 'pycache' found anywhere in distribution
warning: no previously-included files matching '.py[cod]' found anywhere in distribution
adding license file 'LICENSE'
writing manifest file 'sageattention.egg-info\SOURCES.txt'
installing library code to build\bdist.win-amd64\egg
running install_lib
running build_py
copying sageattention\core.py -> build\lib.win-amd64-cpython-312\sageattention
copying sageattention\fa3_wrapper.py -> build\lib.win-amd64-cpython-312\sageattention
copying sageattention\quant.py -> build\lib.win-amd64-cpython-312\sageattention
copying sageattention\sm80_compile.py -> build\lib.win-amd64-cpython-312\sageattention
copying sageattention\sm89_compile.py -> build\lib.win-amd64-cpython-312\sageattention
copying sageattention\sm90_compile.py -> build\lib.win-amd64-cpython-312\sageattention
copying sageattention_init_.py -> build\lib.win-amd64-cpython-312\sageattention
copying sageattention\triton\attn_qk_int8_block_varlen.py -> build\lib.win-amd64-cpython-312\sageattention\triton
copying sageattention\triton\attn_qk_int8_per_block.py -> build\lib.win-amd64-cpython-312\sageattention\triton
copying sageattention\triton\attn_qk_int8_per_block_causal.py -> build\lib.win-amd64-cpython-312\sageattention\triton
copying sageattention\triton\attn_qk_int8_per_block_causal_varlen.py -> build\lib.win-amd64-cpython-312\sageattention\triton
copying sageattention\triton\quant_per_block.py -> build\lib.win-amd64-cpython-312\sageattention\triton
copying sageattention\triton\quant_per_block_varlen.py -> build\lib.win-amd64-cpython-312\sageattention\triton
copying sageattention\triton\quant_per_thread.py -> build\lib.win-amd64-cpython-312\sageattention\triton
copying sageattention\triton_init_.py -> build\lib.win-amd64-cpython-312\sageattention\triton
running build_ext
W1202 18:13:09.983000 28948 venv\Lib\site-packages\torch\utils\cpp_extension.py:480] Error checking compiler version for cl: [WinError 2] 系统找不到指定的文件。
building 'sageattention.qattn_sm80' extension
building 'sageattention.qattn_sm89' extension
building 'sageattention.fused' extension
"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\bin\nvcc" -c csrc/fused/fused.cu -o build\temp.win-amd64-cpython-312\Release\sageattention.fused\csrc/fused/fused.obj -IE:\SageAttention\venv\Lib\site-packages\torch\include -IE:\SageAttention\venv\Lib\site-packages\torch\include\torch\csrc\api\include "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include" -IE:\SageAttention\venv\include -IC:\Users\22129\AppData\Local\Programs\Python\Python312\include -IC:\Users\22129\AppData\Local\Programs\Python\Python312\Include -Xcompiler /EHsc -Xcompiler /wd4068 -Xcompiler /wd4067 -Xcompiler /wd4624 -Xcompiler /wd4190 -Xcompiler /wd4018 -Xcompiler /wd4275 -Xcompiler /wd4267 -Xcompiler /wd4244 -Xcompiler /wd4251 -Xcompiler /wd4819 -Xcompiler /MD -Xcudafe --diag_suppress=dll_interface_conflict_dllexport_assumed -Xcudafe --diag_suppress=dll_interface_conflict_none_assumed -Xcudafe --diag_suppress=field_without_dll_interface -Xcudafe --diag_suppress=base_class_has_different_dll_interface -D__CUDA_NO_HALF_OPERATORS -D__CUDA_NO_HALF_CONVERSIONS -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -O3 -std=c++17 -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ --use_fast_math --threads=8 -Xptxas=-v -diag-suppress=174 --threads 8 -D_GLIBCXX_USE_CXX11_ABI=1 -gencode arch=compute_120a,code=sm_120a -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_fused -std=c++17 --use-local-env
cl.exe /c /nologo /O2 /W3 /GL /DNDEBUG /MD -IE:\SageAttention\venv\Lib\site-packages\torch\include -IE:\SageAttention\venv\Lib\site-packages\torch\include\torch\csrc\api\include "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include" -IE:\SageAttention\venv\include -IC:\Users\22129\AppData\Local\Programs\Python\Python312\include -IC:\Users\22129\AppData\Local\Programs\Python\Python312\Include /EHsc /Tpcsrc/qattn/pybind_sm80.cpp /Fobuild\temp.win-amd64-cpython-312\Release\sageattention._qattn_sm80\csrc/qattn/pybind_sm80.obj /MD /wd4819 /wd4251 /wd4244 /wd4267 /wd4275 /wd4018 /wd4190 /wd4624 /wd4067 /wd4068 /EHsc -g -O3 -fopenmp -lgomp -std=c++17 -DENABLE_BF16 -D_GLIBCXX_USE_CXX11_ABI=1 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_fused /std:c++17
cl.exe /c /nologo /O2 /W3 /GL /DNDEBUG /MD -IE:\SageAttention\venv\Lib\site-packages\torch\include -IE:\SageAttention\venv\Lib\site-packages\torch\include\torch\csrc\api\include "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\include" -IE:\SageAttention\venv\include -IC:\Users\22129\AppData\Local\Programs\Python\Python312\include -IC:\Users\22129\AppData\Local\Programs\Python\Python312\Include /EHsc /Tpcsrc/qattn/pybind_sm89.cpp /Fobuild\temp.win-amd64-cpython-312\Release\sageattention._qattn_sm89\csrc/qattn/pybind_sm89.obj /MD /wd4819 /wd4251 /wd4244 /wd4267 /wd4275 /wd4018 /wd4190 /wd4624 /wd4067 /wd4068 /EHsc -g -O3 -fopenmp -lgomp -std=c++17 -DENABLE_BF16 -D_GLIBCXX_USE_CXX11_ABI=1 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_fused /std:c++17
nvcc fatal : Cannot find compiler 'cl.exe' in PATH
error: command 'cl.exe' failed: None
(venv) PS E:\SageAttention>
(venv) PS E:\SageAttention> # 或者使用 pip 安装（推荐）
(venv) PS E:\SageAttention> pip install -e .
Obtaining file:///E:/SageAttention
Installing build dependencies ... done
Checking if build backend supports build_editable ... done
Getting requirements to build editable ... error
error: subprocess-exited-with-error

× Getting requirements to build editable did not run successfully.
│ exit code: 1
╰─> [21 lines of output]
Traceback (most recent call last):
File "E:\SageAttention\venv\Lib\site-packages\pip_vendor\pyproject_hooks_in_process_in_process.py", line 389, in
main()
File "E:\SageAttention\venv\Lib\site-packages\pip_vendor\pyproject_hooks_in_process_in_process.py", line 373, in main
json_out["return_val"] = hook(**hook_input["kwargs"])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\SageAttention\venv\Lib\site-packages\pip_vendor\pyproject_hooks_in_process_in_process.py", line 157, in get_requires_for_build_editable
return hook(config_settings)
^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\22129\AppData\Local\Temp\pip-build-env-xdwewd82\overlay\Lib\site-packages\setuptools\build_meta.py", line 464, in get_requires_for_build_editable
return self.get_requires_for_build_wheel(config_settings)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\22129\AppData\Local\Temp\pip-build-env-xdwewd82\overlay\Lib\site-packages\setuptools\build_meta.py", line 332, in get_requires_for_build_wheel
return self._get_build_requires(config_settings, requirements=[])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\22129\AppData\Local\Temp\pip-build-env-xdwewd82\overlay\Lib\site-packages\setuptools\build_meta.py", line 302, in _get_build_requires
self.run_setup()
File "C:\Users\22129\AppData\Local\Temp\pip-build-env-xdwewd82\overlay\Lib\site-packages\setuptools\build_meta.py", line 318, in run_setup
exec(code, locals())
File "", line 36, in
ModuleNotFoundError: No module named 'torch'
[end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.

[notice] A new release of pip is available: 25.0.1 -> 25.3
[notice] To update, run: python.exe -m pip install --upgrade pip
error: subprocess-exited-with-error

× Getting requirements to build editable did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.
(venv) PS E:\SageAttention>
(venv) PS E:\SageAttention>

[lixida123]

Here to download the local package installed to solve error: https://github.com/sdbds/SageAttention-for-windows/releases

[DrAculaMD]

It gives the error:

ERROR: sageattention-2.2.0+cu130torch2.9.1-cp311-cp311-win_amd64.whl is not a supported wheel on this platform.

I assume because I'm on Python 3.12?

[SLAPaper]

add --no-build-isolation to the pip command

edit: this is the fix to "torch not found" error

[DrAculaMD]

Didn't help, same error

[lixida123]

它会给出错误：

ERROR: sageattention-2.2.0+cu130torch2.9.1-cp311-cp311-win_amd64.whl is not a supported wheel on this platform.

我猜是因为我用的是 Python 3.12？

yes，cp311-cp311 It indicates that it is only compatible with python3.11 and not for 3.12, so it will report an error.