[v.1.7.x] Use local env for building CUDA extensions on Windows (pytorch#47150) (pytorch#48937)

Summary:
Fixes pytorch/vision#2818 (comment)
After activating the VC env multiple times, the following error will be raised when building a CUDA extension.
```
FAILED: C:/tools/MINICO~1/CONDA-~2/TORCHV~1/work/build/temp.win-amd64-3.8/Release/tools/MINICO~1/CONDA-~2/TORCHV~1/work/torchvision/csrc/cuda/PSROIAlign_cuda.obj
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin\nvcc -Xcompiler /MD -Xcompiler /wd4819 -Xcompiler /wd4251 -Xcompiler /wd4244 -Xcompiler /wd4267 -Xcompiler /wd4275 -Xcompiler /wd4018 -Xcompiler /wd4190 -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 -DWITH_CUDA -Dtorchvision_EXPORTS -IC:\tools\MINICO~1\CONDA-~2\TORCHV~1\work\torchvision\csrc -I%PREFIX%\lib\site-packages\torch\include -I%PREFIX%\lib\site-packages\torch\include\torch\csrc\api\include -I%PREFIX%\lib\site-packages\torch\include\TH -I%PREFIX%\lib\site-packages\torch\include\THC "-IC:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include" -I%PREFIX%\include -I%PREFIX%\include "-IC:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.27.29110\ATLMFC\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.27.29110\include" "-IC:\Program Files (x86)\Windows Kits\NETFXSDK\4.8\include\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\shared" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\winrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\cppwinrt" "-IC:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.27.29110\ATLMFC\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.27.29110\include" "-IC:\Program Files (x86)\Windows Kits\NETFXSDK\4.8\include\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\shared" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\winrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\cppwinrt" -I%PREFIX%\Library\include -c C:\tools\MINICO~1\CONDA-~2\TORCHV~1\work\torchvision\csrc\cuda\PSROIAlign_cuda.cu -o C:\tools\MINICO~1\CONDA-~2\TORCHV~1\work\build\temp.win-amd64-3.8\Release\tools\MINICO~1\CONDA-~2\TORCHV~1\work\torchvision\csrc\cuda\PSROIAlign_cuda.obj -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_35,code=sm_35 -gencode=arch=compute_50,code=sm_50 -gencode=arch=compute_60,code=sm_60 -gencode=arch=compute_70,code=sm_70 -gencode=arch=compute_75,code=sm_75 -gencode=arch=compute_80,code=sm_80 -gencode=arch=compute_50,code=compute_50 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_C -D_GLIBCXX_USE_CXX11_ABI=0
'cl.exe' is not recognized as an internal or external command,
operable program or batch file.
```

Pull Request resolved: pytorch#47150

Reviewed By: agolynski

Differential Revision: D24706019

Pulled By: ezyang

fbshipit-source-id: c13dc29f62d2d12d6a56f33dd450b467a1bf193b
(cherry picked from commit d73a8db)
Signed-off-by: Eli Uriegas <[email protected]>

Co-authored-by: peter <[email protected]>

Author: seemethere

torch/utils/cpp_extension.py

@@ -550,7 +550,7 @@ def spawn(cmd):
                         else:
                             cflags = []
 
-                        cflags = win_cuda_flags(cflags)
+                        cflags = win_cuda_flags(cflags) + ['--use-local-env']
                         for flag in COMMON_MSVC_FLAGS:
                             cflags = ['-Xcompiler', flag] + cflags
                         for ignore_warning in MSVC_IGNORE_CUDAFE_WARNINGS:
@@ -620,7 +620,7 @@ def win_wrap_ninja_compile(sources,
             cuda_post_cflags = None
             cuda_cflags = None
             if with_cuda:
-                cuda_cflags = []
+                cuda_cflags = ['--use-local-env']
                 for common_cflag in common_cflags:
                     cuda_cflags.append('-Xcompiler')
                     cuda_cflags.append(common_cflag)