在ComfyUI的python_embeded下编译安装module

在ComfyUI中安装某些module时，并不能直接获得whl文件安装，而是拉取module项目包，并编译生成whl后再安装。例如 insightface、flash-attn、dlib、xformers等。

由于遇到一个问题，尝试需要安装的包没有找到可用的whl文件，于是再次有了这次的尝试，并成功编译安装了insightface。

（先按照1、2、3、4 点配置好环境）

1、首先是要下载安装 Visual Studio 生成工具 2022：

后来发现，生成工具版本不是越高越好，在将生成工具升级到版本17.12后，发现想在python3.11编译flash_attn时不成功， 查到这里有文章《CUDA compatibility with Visual Studio 2022 version 17.10》说：编译工具版本还需要与cuda版本匹配，而我的cuda12.4，按照文章所说，生成工具版本降到17.10比较稳妥。

只好卸载，从这里下载 Fixed version bootstrappers 安装固定版本：

参考文章《Windows 如何仅安装 MSVC 而不安装 Visual Studio》（

选择 Path ，双击添加两条

  1. C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.40.33807\bin\Hostx64\x64
  2. C:\Program Files (x86)\Windows Kits\10\bin\10.0.22621.0\x64

添加 LIB，添加三条（如果之前没有，那点开之后三条中间用英文分号隔开）

  1. C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.40.33807\lib\x64
  2. C:\Program Files (x86)\Windows Kits\10\Lib\10.0.22621.0\ucrt\x64
  3. C:\Program Files (x86)\Windows Kits\10\Lib\10.0.22621.0\um\x64

添加 INCLUDE，添加6条

  1. C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.40.33807\include
  2. C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\ucrt
  3. C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\um
  4. C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\winrt
  5. C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\cppwinrt
  6. C:\Program Files (x86)\Windows Kits\10\Include\10.0.22621.0\shared

好像除了设置path的第一行有用外，没发现其他设置起什么作用，就是有与没有没什么区别的，直到出现错误：LINK : fatal error LNK1104: 无法打开文件“kernel32.lib，才发现这样的认识是错误的

  1. D:\AI\NATTEN>where /R "C:\Program Files (x86)" kernel32.lib
  2. C:\Program Files (x86)\Windows Kits\10\Lib\10.0.20348.0\um\arm\kernel32.Lib
  3. C:\Program Files (x86)\Windows Kits\10\Lib\10.0.20348.0\um\arm64\kernel32.Lib
  4. C:\Program Files (x86)\Windows Kits\10\Lib\10.0.20348.0\um\x64\kernel32.Lib
  5. C:\Program Files (x86)\Windows Kits\10\Lib\10.0.20348.0\um\x86\kernel32.Lib

设置环境变量（打开设置--->系统--->关于--->高级系统设置--->环境变量）：

下面路径指向 cl.exe ，以令到 错误“H:\V.0.2.7\python_embeded\Lib\site-packages\torch\utils\cpp_extension.py:382: UserWarning: Error checking compiler version for cl: [WinError 2] 系统找不到指定的文件。”消失。

path=C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.40.33807\bin\Hostx64\x64;%path%

2、将完整安装python的目录include复制到pytho_embeded下

解决安装中提示：无法打开包括文件: “Python.h” 的问题。

错误信息：无法打开包括文件: “Python.h”: 

  1.  "C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.41.34120\bin\HostX86\x64\cl.exe" /c /nologo /O2 /W3 /GL /DNDEBUG /MD -Iinsightface/thirdparty/face3d/mesh/cython -IH:\V.0.2.7\python_embeded\Lib\site-packages\numpy\core\include -IH:\V.0.2.7\python_embeded\include -IH:\V.0.2.7\python_embeded\Include "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.41.34120\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Auxiliary\VS\include" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.22621.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\um" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\shared" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\winrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\cppwinrt" /EHsc /Tpinsightface/thirdparty/face3d/mesh/cython/mesh_core_cython.cpp /Fobuild\temp.win-amd64-cpython-312\Release\insightface/thirdparty/face3d/mesh/cython/mesh_core_cython.obj
  2.       mesh_core_cython.cpp
  3.       insightface/thirdparty/face3d/mesh/cython/mesh_core_cython.cpp(36): fatal error C1083: 无法打开包括文件: “Python.h”: No such file or directory
  4.       error: command 'C:\\Program Files (x86)\\Microsoft Visual Studio\\2022\\BuildTools\\VC\\Tools\\MSVC\\14.41.34120\\bin\\HostX86\\x64\\cl.exe' failed with exit code 2
  5.       [end of output]
  6.  
  7.   note: This error originates from a subprocess, and is likely not a problem with pip.
  8.   ERROR: Failed building wheel for insightface
  9. Failed to build insightface
  10. ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (insightface)

搜索过很多文章说如何安装python-dev包什么的，都不靠谱（未能解决我的问题），最后还是直接下载 python完整安装包（与ComfyUI的python_embeded相同版本，3.12安装大约345M），安装到某个路径下，并且安装时不勾选设置环境变量（为防止因为环境变量而出现影响comfyUI内嵌python的情况发生），并将文件夹include复制到H:\V.0.2.7\python_embeded\include。

无法打开包括文件: “Python.h” 问题解决。

3、将完整安装python的目录libs复制到pytho_embeded下

解决无法打开文件“python312.lib”。

同样，将完整安装包路径下的文件夹 libs 复制到H:\V.0.2.7\python_embeded\libs即可解决问题

  1.       H:\V.0.2.7\python_embeded\Lib\site-packages\numpy\core\include\numpy\npy_1_7_deprecated_api.h(14) : Warning Msg: Using deprecated NumPy API, disable it with #define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
  2.       "C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.41.34120\bin\HostX86\x64\link.exe" /nologo /INCREMENTAL:NO /LTCG /DLL /MANIFEST:EMBED,ID=2 /MANIFESTUAC:NO /LIBPATH:H:\V.0.2.7\python_embeded\libs /LIBPATH:H:\V.0.2.7\python_embeded /LIBPATH:H:\V.0.2.7\python_embeded\PCbuild\amd64 "/LIBPATH:C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.41.34120\lib\x64" "/LIBPATH:C:\Program Files (x86)\Windows Kits\10\lib\10.0.22621.0\ucrt\x64" "/LIBPATH:C:\Program Files (x86)\Windows Kits\10\\lib\10.0.22621.0\\um\x64" /EXPORT:PyInit_mesh_core_cython build\temp.win-amd64-cpython-312\Release\insightface/thirdparty/face3d/mesh/cython/mesh_core.obj build\temp.win-amd64-cpython-312\Release\insightface/thirdparty/face3d/mesh/cython/mesh_core_cython.obj /OUT:build\lib.win-amd64-cpython-312\insightface\thirdparty\face3d\mesh\cython\mesh_core_cython.cp312-win_amd64.pyd /IMPLIB:build\temp.win-amd64-cpython-312\Release\insightface/thirdparty/face3d/mesh/cython\mesh_core_cython.cp312-win_amd64.lib
  3.       LINK : fatal error LNK1104: 无法打开文件“python312.lib”
  4.       error: command 'C:\\Program Files (x86)\\Microsoft Visual Studio\\2022\\BuildTools\\VC\\Tools\\MSVC\\14.41.34120\\bin\\HostX86\\x64\\link.exe' failed with exit code 1104
  5.       [end of output]
  6.  
  7.   note: This error originates from a subprocess, and is likely not a problem with pip.
  8.   ERROR: Failed building wheel for insightface
  9. Failed to build insightface
  10. ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (insightface)
  11.  
  12. H:\V.0.2.7>

4、ninja.exe

 从 GitHub - ninja-build/ninja: a small build system with a focus on speed 的自述文件可知到，我们只需要生成的exe即可：

 从 Releases · ninja-build/ninja · GitHub 下载最新适用的windows版本解压得到ninja.exe，复制到 path 路径指向的任意某个文件夹即可。

 设置好后， 警告错误：warnings.warn(msg.format('we could not find ninja.'))  消失。

在我的系统中，下面2个路径都有：

  1. C:\Users\Monday\.conda\envs\comfyui_cp311\Scripts
  2.  
  3. C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\Common7\IDE\CommonExtensions\Microsoft\CMake\Ninja

第一个路径文件较小只有557k为最新版本 1.12.1，第二个路径文件较大有2.42M版本为1.11.0（但ninja 官方GitHub上1.11.0版本也只有557k大小，这个估计微软编译的）

5、编译 insightface 并安装成功： 

 运行也没有什么问题。

6、编译  flash-attn 并安装成功

网上看过文章，知道在Windows下编译 时间很长，开始编译，经过漫长的等待（约3个小时左右吧），报如下错误：

  1.      tmpxft_000038e4_00000000-7_flash_fwd_hdim64_bf16_sm80.compute_90.cudafe1.cpp
  2.       "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\bin\nvcc" -c csrc/flash_attn/src/flash_fwd_hdim64_fp16_causal_sm80.cu -o build\temp.win-amd64-cpython-312\Release\csrc/flash_attn/src/flash_fwd_hdim64_fp16_causal_sm80.obj -IC:\Users\Monday\AppData\Local\Temp\pip-install-vpjwb98z\flash-attn_71d6532d3ab546e1bb76dd71119a8066\csrc\flash_attn -IC:\Users\Monday\AppData\Local\Temp\pip-install-vpjwb98z\flash-attn_71d6532d3ab546e1bb76dd71119a8066\csrc\flash_attn\src -IC:\Users\Monday\AppData\Local\Temp\pip-install-vpjwb98z\flash-attn_71d6532d3ab546e1bb76dd71119a8066\csrc\cutlass\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\torch\csrc\api\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\TH -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\THC "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\include" -IH:\V.0.2.7\python_embeded\include -IH:\V.0.2.7\python_embeded\Include "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.41.34120\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Auxiliary\VS\include" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.22621.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\um" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\shared" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\winrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\cppwinrt" -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 -Xcompiler /EHsc -Xcompiler /wd4068 -Xcompiler /wd4067 -Xcompiler /wd4624 -Xcompiler /wd4190 -Xcompiler /wd4018 -Xcompiler /wd4275 -Xcompiler /wd4267 -Xcompiler /wd4244 -Xcompiler /wd4251 -Xcompiler /wd4819 -Xcompiler /MD -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__ -U__CUDA_NO_HALF2_OPERATORS__ -U__CUDA_NO_BFLOAT16_CONVERSIONS__ --expt-relaxed-constexpr --expt-extended-lambda --use_fast_math -gencode arch=compute_80,code=sm_80 -gencode arch=compute_90,code=sm_90 --threads 4 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=flash_attn_2_cuda -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++17 --use-local-env
  3.       flash_fwd_hdim64_fp16_causal_sm80.cu
  4.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_HALF_OPERATORS__”(用“/U__CUDA_NO_HALF_OPERATORS__”)
  5.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_HALF_CONVERSIONS__”(用“/U__CUDA_NO_HALF_CONVERSIONS__”)
  6.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_HALF2_OPERATORS__”(用“/U__CUDA_NO_HALF2_OPERATORS__”)
  7.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_BFLOAT16_CONVERSIONS__”(用“/U__CUDA_NO_BFLOAT16_CONVERSIONS__”)
  8.       flash_fwd_hdim64_fp16_causal_sm80.cu
  9.       c1xx: fatal error C1083: 无法打开源文件: “C:\Users\Monday\AppData\Local\Temp\pip-install-vpjwb98z\flash-attn_71d6532d3ab546e1bb76dd71119a8066\csrc\flash_attn\src\flash_fwd_hdim64_fp16_causal_sm80.cu”: No such file or directory

搜索到文章《win10下cuda12.1 +troch2.4.1+vs2022环境下编译安装flash-attn》中提到：安装cutlass库，该库为编译flash-attn的必须依赖。

于是执行命令安装nvidia-cutlass后再安装flash_attn：

  1. python_embeded\python.exe -m pip install nvidia-cutlass
  2. python_embeded\python.exe -m pip install flash_attn

 再次经过更漫长的等待（耗时4小时45分），终于编译成功并安装。网上也没找到flash-attn-2.7.0.post2的whl安装包，这次自己终于编译得到了。

在等待过程中，怕会失败，也继续搜索看看第一次失败的原因是什么，有文章《nvcc fatal : Could not open output file '/tmp/tmpxft_00003d04_00000000'》说是因为用户权限没法读写文件导致，不过既然成功了，看来我的错误就是因为没安装 nvidia-cutlass 。

 重装编译环境（Visual Studio 生成工具 2022、CUDA Toolkit），然后在python3.11下再次编译，耗时2个小时（这次看到CPU占用率100%，第一次时好像占用率不是很高，忘记了。）

7、编译安装开发版的xformers-0.0.29.dev940 失败

因为运行时遇到xformers报下面这个错误，才折腾尝试自己编译安装module：xformers-0.0.29.dev940。 

  1. Loading PuLID-Flux model.
  2. !!! Exception during processing !!! No operator found for `memory_efficient_attention_forward` with inputs:
  3.      query       : shape=(1, 577, 16, 64) (torch.bfloat16)
  4.      key         : shape=(1, 577, 16, 64) (torch.bfloat16)
  5.      value       : shape=(1, 577, 16, 64) (torch.bfloat16)
  6.      attn_bias   : <class 'NoneType'>
  7.      p           : 0.0
  8. `fa2F@v2.6.3-24-gbdf733b` is not supported because:
  9.     requires device with capability > (8, 0) but your GPU has capability (7, 5) (too old)
  10.     bf16 is only supported on A100+ GPUs
  11. `cutlassF-pt` is not supported because:
  12.     bf16 is only supported on A100+ GPUs
  13. Traceback (most recent call last):
  14.   File "H:\V.0.2.7\ComfyUI\execution.py", line 323, in execute
  15.     output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
  16.                                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  17.   File "H:\V.0.2.7\ComfyUI\execution.py", line 198, in get_output_data
  18.     return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
  19.                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  20.   File "H:\V.0.2.7\ComfyUI\execution.py", line 169, in _map_node_over_list
  21.     process_inputs(input_dict, i)
  22.   File "H:\V.0.2.7\ComfyUI\execution.py", line 158, in process_inputs
  23.     results.append(getattr(obj, func)(**inputs))
  24.                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  25.   File "H:\V.0.2.7\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced\pulidflux.py", line 382, in apply_pulid_flux
  26.     id_cond_vit, id_vit_hidden = eva_clip(face_features_image, return_all_features=False, return_hidden=True, shuffle=False)
  27.                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  28.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1736, in _wrapped_call_impl
  29.     return self._call_impl(*args, **kwargs)
  30.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  31.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1747, in _call_impl
  32.     return forward_call(*args, **kwargs)
  33.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  34.   File "H:\V.0.2.7\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced\eva_clip\eva_vit_model.py", line 544, in forward
  35.     x, hidden_states = self.forward_features(x, return_all_features, return_hidden, shuffle)
  36.                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  37.   File "H:\V.0.2.7\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced\eva_clip\eva_vit_model.py", line 531, in forward_features
  38.     x = blk(x, rel_pos_bias=rel_pos_bias)
  39.         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  40.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1736, in _wrapped_call_impl
  41.     return self._call_impl(*args, **kwargs)
  42.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  43.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1747, in _call_impl
  44.     return forward_call(*args, **kwargs)
  45.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  46.   File "H:\V.0.2.7\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced\eva_clip\eva_vit_model.py", line 293, in forward
  47.     x = x + self.drop_path(self.attn(self.norm1(x), rel_pos_bias=rel_pos_bias, attn_mask=attn_mask))
  48.                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  49.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1736, in _wrapped_call_impl
  50.     return self._call_impl(*args, **kwargs)
  51.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  52.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 1747, in _call_impl
  53.     return forward_call(*args, **kwargs)
  54.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  55.   File "H:\V.0.2.7\ComfyUI\custom_nodes\ComfyUI-PuLID-Flux-Enhanced\eva_clip\eva_vit_model.py", line 208, in forward
  56.     x = xops.memory_efficient_attention(
  57.         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  58.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\xformers\ops\fmha\__init__.py", line 306, in memory_efficient_attention
  59.     return _memory_efficient_attention(
  60.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  61.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\xformers\ops\fmha\__init__.py", line 467, in _memory_efficient_attention
  62.     return _memory_efficient_attention_forward(
  63.            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  64.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\xformers\ops\fmha\__init__.py", line 486, in _memory_efficient_attention_forward
  65.     op = _dispatch_fw(inp, False)
  66.          ^^^^^^^^^^^^^^^^^^^^^^^^
  67.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\xformers\ops\fmha\dispatch.py", line 135, in _dispatch_fw
  68.     return _run_priority_list(
  69.            ^^^^^^^^^^^^^^^^^^^
  70.   File "H:\V.0.2.7\python_embeded\Lib\site-packages\xformers\ops\fmha\dispatch.py", line 76, in _run_priority_list
  71.     raise NotImplementedError(msg)
  72. NotImplementedError: No operator found for `memory_efficient_attention_forward` with inputs:
  73.      query       : shape=(1, 577, 16, 64) (torch.bfloat16)
  74.      key         : shape=(1, 577, 16, 64) (torch.bfloat16)
  75.      value       : shape=(1, 577, 16, 64) (torch.bfloat16)
  76.      attn_bias   : <class 'NoneType'>
  77.      p           : 0.0
  78. `fa2F@v2.6.3-24-gbdf733b` is not supported because:
  79.     requires device with capability > (8, 0) but your GPU has capability (7, 5) (too old)
  80.     bf16 is only supported on A100+ GPUs
  81. `cutlassF-pt` is not supported because:
  82.     bf16 is only supported on A100+ GPUs
  83.  
  84. Prompt executed in 222.85 seconds

继续编译安装开发版的xformers-0.0.29.dev940，看看是否能解决我遇到运行中报错的问题。 

报错： 

  1. 只要报错信息如下：
  2.  
  3.   Building wheel for xformers (setup.py) ... error
  4.   error: subprocess-exited-with-error
  5.  
  6.   × python setup.py bdist_wheel did not run successfully.
  7.   │ exit code: 1
  8.   ╰─> [5830 lines of output]
  9.       fatal: not a git repository (or any of the parent directories): .git
  10.       running bdist_wheel
  11.  
  12.       H:\V.0.2.7\python_embeded\Lib\site-packages\torch\utils\cpp_extension.py:497: UserWarning: Attempted to use ninja as the BuildExtension backend but we could not find ninja.. Falling back to using the slow distutils backend.
  13.         warnings.warn(msg.format('we could not find ninja.'))
  14.       running build
  15. 省略 N 行
  16.  
  17.       H:\V.0.2.7\python_embeded\Lib\site-packages\torch\utils\cpp_extension.py:382: UserWarning: Error checking compiler version for cl: [WinError 2] 系统找不到指定的文件。
  18.         warnings.warn(f'Error checking compiler version for {compiler}: {error}')
  19.  
  20.   Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
  21.       [1/85] C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\bin\nvcc --generate-dependencies-with-compile --dependency-output C:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\build\temp.win-amd64-cpython-312\Release\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src\flash_bwd_hdim192_bf16_causal_sm80.obj.d -std=c++17 --use-local-env -Xcompiler /MD -Xcompiler /wd4819 -Xcompiler /wd4251 -Xcompiler /wd4244 -Xcompiler /wd4267 -Xcompiler /wd4275 -Xcompiler /wd4018 -Xcompiler /wd4190 -Xcompiler /wd4624 -Xcompiler /wd4067 -Xcompiler /wd4068 -Xcompiler /EHsc -Xcudafe --diag_suppress=base_class_has_different_dll_interface -Xcudafe --diag_suppress=field_without_dll_interface -Xcudafe --diag_suppress=dll_interface_conflict_none_assumed -Xcudafe --diag_suppress=dll_interface_conflict_dllexport_assumed -IC:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn -IC:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src -IC:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\cutlass\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\torch\csrc\api\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\TH -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\THC "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\include" -IH:\V.0.2.7\python_embeded\include -IH:\V.0.2.7\python_embeded\Include "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.42.34433\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Auxiliary\VS\include" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.22621.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\um" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\shared" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\winrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\cppwinrt" -c C:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src\flash_bwd_hdim192_bf16_causal_sm80.cu -o C:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\build\temp.win-amd64-cpython-312\Release\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src\flash_bwd_hdim192_bf16_causal_sm80.obj -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -DHAS_PYTORCH --use_fast_math -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ --extended-lambda -D_ENABLE_EXTENDED_ALIGNED_STORAGE -std=c++17 --generate-line-info -DNDEBUG --threads 4 --ptxas-options=-v -Xcompiler /Zc:lambda -Xcompiler /Zc:preprocessor -Xcompiler /Zc:__cplusplus -O3 -std=c++17 -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ -U__CUDA_NO_HALF2_OPERATORS__ -U__CUDA_NO_BFLOAT16_CONVERSIONS__ --expt-relaxed-constexpr --expt-extended-lambda --use_fast_math --ptxas-options=-v -gencode=arch=compute_80,code=sm_80 -gencode=arch=compute_86,code=sm_86 -gencode=arch=compute_90,code=sm_90 -DFLASHATTENTION_DISABLE_ALIBI --generate-line-info -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_C_flashattention -D_GLIBCXX_USE_CXX11_ABI=0
  22.       FAILED: C:/Users/Monday/AppData/Local/Temp/pip-install-h5wrwrf7/xformers_385ac3dddc8a4e779d876f9cbb34ec19/build/temp.win-amd64-cpython-312/Release/Users/Monday/AppData/Local/Temp/pip-install-h5wrwrf7/xformers_385ac3dddc8a4e779d876f9cbb34ec19/third_party/flash-attention/csrc/flash_attn/src/flash_bwd_hdim192_bf16_causal_sm80.obj
  23.       C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\bin\nvcc --generate-dependencies-with-compile --dependency-output C:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\build\temp.win-amd64-cpython-312\Release\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src\flash_bwd_hdim192_bf16_causal_sm80.obj.d -std=c++17 --use-local-env -Xcompiler /MD -Xcompiler /wd4819 -Xcompiler /wd4251 -Xcompiler /wd4244 -Xcompiler /wd4267 -Xcompiler /wd4275 -Xcompiler /wd4018 -Xcompiler /wd4190 -Xcompiler /wd4624 -Xcompiler /wd4067 -Xcompiler /wd4068 -Xcompiler /EHsc -Xcudafe --diag_suppress=base_class_has_different_dll_interface -Xcudafe --diag_suppress=field_without_dll_interface -Xcudafe --diag_suppress=dll_interface_conflict_none_assumed -Xcudafe --diag_suppress=dll_interface_conflict_dllexport_assumed -IC:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn -IC:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src -IC:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\cutlass\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\torch\csrc\api\include -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\TH -IH:\V.0.2.7\python_embeded\Lib\site-packages\torch\include\THC "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\include" -IH:\V.0.2.7\python_embeded\include -IH:\V.0.2.7\python_embeded\Include "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Tools\MSVC\14.42.34433\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Auxiliary\VS\include" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.22621.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\um" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\shared" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\winrt" "-IC:\Program Files (x86)\Windows Kits\10\\include\10.0.22621.0\\cppwinrt" -c C:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src\flash_bwd_hdim192_bf16_causal_sm80.cu -o C:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\build\temp.win-amd64-cpython-312\Release\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src\flash_bwd_hdim192_bf16_causal_sm80.obj -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -DHAS_PYTORCH --use_fast_math -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ --extended-lambda -D_ENABLE_EXTENDED_ALIGNED_STORAGE -std=c++17 --generate-line-info -DNDEBUG --threads 4 --ptxas-options=-v -Xcompiler /Zc:lambda -Xcompiler /Zc:preprocessor -Xcompiler /Zc:__cplusplus -O3 -std=c++17 -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ -U__CUDA_NO_HALF2_OPERATORS__ -U__CUDA_NO_BFLOAT16_CONVERSIONS__ --expt-relaxed-constexpr --expt-extended-lambda --use_fast_math --ptxas-options=-v -gencode=arch=compute_80,code=sm_80 -gencode=arch=compute_86,code=sm_86 -gencode=arch=compute_90,code=sm_90 -DFLASHATTENTION_DISABLE_ALIBI --generate-line-info -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_C_flashattention -D_GLIBCXX_USE_CXX11_ABI=0
  24.       flash_bwd_hdim192_bf16_causal_sm80.cu
  25.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_HALF_OPERATORS__”(用“/U__CUDA_NO_HALF_OPERATORS__”)
  26.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_HALF_CONVERSIONS__”(用“/U__CUDA_NO_HALF_CONVERSIONS__”)
  27.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_HALF2_OPERATORS__”(用“/U__CUDA_NO_HALF2_OPERATORS__”)
  28.       cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_BFLOAT16_CONVERSIONS__”(用“/U__CUDA_NO_BFLOAT16_CONVERSIONS__”)
  29. 这里省略 N 多行
  30.       flash_bwd_hdim192_bf16_causal_sm80.cu
  31.       fatal   : Could not open output file C:\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\build\temp.win-amd64-cpython-312\Release\Users\Monday\AppData\Local\Temp\pip-install-h5wrwrf7\xformers_385ac3dddc8a4e779d876f9cbb34ec19\third_party\flash-attention\csrc\flash_attn\src\flash_bwd_hdim192_bf16_causal_sm80.obj.d
  32.  
  33. 这里省略 N 多个 fatal   : Could not open output file 错误

用 秘塔AI搜索  ： cl: 命令行 warning D9025 :正在重写“/D__CUDA_NO_HALF_OPERATORS__”(用“/U__CUDA_NO_HALF_OPERATORS__”)

  1. 环境变量设置问题：

    • ③和⒂ 提到，在 Windows 系统上安装 CUDA 时，环境变量设置不正确可能导致编译失败。特别是 ③ 中提到的 vcvars64.bat 脚本，用于设置 Visual Studio 的环境变量。
    • 解决方案：在安装 CUDA 之前，运行 vcvars64.bat 脚本来设置正确的环境变量。确保在命令提示符中运行该脚本，并且路径正确。

 运行 C:\Program Files (x86)\Microsoft Visual Studio\2022\BuildTools\VC\Auxiliary\Build>vcvars64.bat

设置环境变量后，错误少了很多。

但根据错误再次设置了参数 set DISTUTILS_USE_SDK=1  后，错误依旧一样多（与原来一样）。

而文心一言的答案则是（看起来更靠谱？因为就是冲突，导致命令没有结果：没有文件生成）：

 暂时不折腾这个了。

8、关于ModuleNotFoundError: No module named 'distutils.msvccompiler'  的问题 --> 编译安装APEX 成功，实际无效。

 错误信息：ModuleNotFoundError: No module named 'distutils.msvccompiler'

  1. H:\V.0.2.7>python_embeded\python.exe -m pip install apex
  2. Collecting apex
  3.   Using cached apex-0.9.10dev.tar.gz (36 kB)
  4.   Preparing metadata (setup.py) ... done
  5. Collecting cryptacular (from apex)
  6.   Using cached cryptacular-1.6.2.tar.gz (75 kB)
  7.   Installing build dependencies ... done
  8.   Getting requirements to build wheel ... done
  9.   Preparing metadata (pyproject.toml) ... error
  10.   error: subprocess-exited-with-error
  11.  
  12.   × Preparing metadata (pyproject.toml) did not run successfully.
  13.   │ exit code: 2
  14.   ╰─> [4 lines of output]
  15.       scons: Reading SConscript files ...
  16.       ModuleNotFoundError: No module named 'distutils.msvccompiler':
  17.         File "C:\Users\Monday\AppData\Local\Temp\pip-install-whj_jyaw\cryptacular_4162e6ad50164a3baf1cd0472e6f84c1\SConstruct", line 21:
  18.           import distutils.msvccompiler
  19.       [end of output]
  20.  
  21.   note: This error originates from a subprocess, and is likely not a problem with pip.
  22. error: metadata-generation-failed
  23.  
  24. × Encountered error while generating package metadata.
  25. ╰─> See above for output.
  26.  
  27. note: This is an issue with the package mentioned above, not pip.
  28. hint: See above for details.
  29.  
  30. H:\V.0.2.7>

 查了多篇文章未能解决ModuleNotFoundError: No module named 'distutils.msvccompiler'，官网文章：distutils — Building and installing Python modules — Python 3.10.16 documentation 说 distutils is deprecated with removal planned for Python 3.12. 

还是到 GitHub 《关于解决ModuleNotFoundError: No module named 'torch'导致安装失败 #1852》 找到答案并编译成功，原来不是所有的 module 都能简单用 “pip module名” 命令进行编译安装。

  1. 运行下面代码编译安装
  2. H:\V.0.2.7>git clone https://github.com/NVIDIA/apex.git
  3. H:\V.0.2.7>cd apex
  4. H:\V.0.2.7\apex>..\python_embeded\python.exe -m pip install -v --no-cache-dir .
  1. H:\V.0.2.7>git clone https://github.com/NVIDIA/apex.git
  2. Cloning into 'apex'...
  3. remote: Enumerating objects: 11902, done.
  4. remote: Counting objects: 100% (3970/3970), done.
  5. remote: Compressing objects: 100% (759/759), done.
  6. remote: Total 11902 (delta 3492), reused 3413 (delta 3205), pack-reused 7932 (from 1)
  7. Receiving objects: 100% (11902/11902), 15.61 MiB | 4.25 MiB/s, done.
  8. Resolving deltas: 100% (8321/8321), done.
  9. Updating files: 100% (505/505), done.
  10.  
  11. H:\V.0.2.7>cd apex
  12.  
  13. H:\V.0.2.7\apex>..\python_embeded\python.exe -m pip install -v --no-cache-dir .
  14. Using pip 24.3.1 from H:\V.0.2.7\python_embeded\Lib\site-packages\pip (python 3.12)
  15. Processing h:\v.0.2.7\apex
  16.   Running command pip subprocess to install build dependencies
  17.   Using pip 24.3.1 from H:\V.0.2.7\python_embeded\Lib\site-packages\pip (python 3.12)
  18.   Collecting setuptools
  19.     Obtaining dependency information for setuptools from https://files.pythonhosted.org/packages/55/21/47d163f615df1d30c094f6c8bbb353619274edccf0327b185cc2493c2c33/setuptools-75.6.0-py3-none-any.whl.metadata
  20.     Using cached setuptools-75.6.0-py3-none-any.whl.metadata (6.7 kB)
  21.   Collecting wheel
  22.     Obtaining dependency information for wheel from https://files.pythonhosted.org/packages/0b/2c/87f3254fd8ffd29e4c02732eee68a83a1d3c346ae39bc6822dcbcb697f2b/wheel-0.45.1-py3-none-any.whl.metadata
  23.     Using cached wheel-0.45.1-py3-none-any.whl.metadata (2.3 kB)
  24.   Using cached setuptools-75.6.0-py3-none-any.whl (1.2 MB)
  25.   Using cached wheel-0.45.1-py3-none-any.whl (72 kB)
  26.   Installing collected packages: wheel, setuptools
  27.     Creating C:\Users\Monday\AppData\Local\Temp\pip-build-env-da54lkae\overlay\Scripts
  28.   Successfully installed setuptools-75.6.0 wheel-0.45.1
  29.   Installing build dependencies ... done
  30.   Running command Getting requirements to build wheel
  31.  
  32.  
  33.   torch.__version__  = 2.5.1+cu124
  34.  
  35.  
  36.   running egg_info
  37.   creating apex.egg-info
  38.   writing apex.egg-info\PKG-INFO
  39.  
  40. 。。。。。。。。省略 N 行。。。。
  41.   removing build\bdist.win-amd64\wheel
  42.   Building wheel for apex (pyproject.toml) ... done
  43.   Created wheel for apex: filename=apex-0.1-py3-none-any.whl size=406607 sha256=206aca315212aa0a76b14de395b6afe1ecdcd4c5fdd61b57986dabb509e83121
  44.   Stored in directory: C:\Users\Monday\AppData\Local\Temp\pip-ephem-wheel-cache-zwa4z7gq\wheels\65\c7\12\b7e49ba4abd3da74df298dc51ea0f6a086d496566f4310f620
  45. Successfully built apex
  46. Installing collected packages: apex
  47. Successfully installed apex-0.1
  48.  
  49. H:\V.0.2.7\apex>

 但还是说没有安装apex：

从文章《NVIDIA APEX安装完全指南及Megatron-LM/Pytorch运行问题解决》 发现，上面使用的编译参数还是不正确的。需要使用官网里面的下面参数编译，但编译未能成功。

..\python_embeded\python.exe -m pip install -v --disable-pip-version-check --no-cache-dir  --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" .

9、编译后获得的文件名，不能修改

例如编译后得到的文件为：insightface-0.7.3-cp312-cp312-win_amd64.whl

如果你将文件重命名为：insightface-0.7.3.whl

执行安装时会报错：ERROR: insightface-0.7.3.whl is not a valid wheel filename.

 

NotImplementedError：找不到 Memory_efficient_attention_forward 的运算符 - stable-diffusion - SO中文参考 - www.soinside.com

注：本文转载自blog.csdn.net的zslefour的文章"https://blog.csdn.net/zslefour/article/details/143914440"。版权归原作者所有，此博客不拥有其著作权，亦不承担相应法律责任。如有侵权，请联系我们删除。
复制链接
相关推荐
发表评论

评论记录：

未查询到任何数据！