Pass allow-unsupported compiler flag for nvcc during compilation

[atalman]

After updating to cudnn 9.1:
pytorch/test-infra#5295

Getting fatal error from NVCC:

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1\include\crt/host_config.h(153): fatal error C1189: #error:  -- unsupported Microsoft Visual Studio version! Only the versions between 2017 and 2022 (inclusive) are supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.
nms_kernel.cu

For cuda 12.1 and 11.8

Failure:
https://github.com/pytorch/vision/actions/runs/9502323845/job/26204631855

Primary Suspect small VC update:
Old:

** Visual Studio 2022 Developer Command Prompt v17.9.6

New :

** Visual Studio 2022 Developer Command Prompt v17.10.1

[pytorch-bot]

No ciflow labels are configured for this repo.
For information on how to enable CIFlow bot see this wiki

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/8492

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 10 New Failures

As of commit d7f92dc with merge base 143d078 ():

NEW FAILURES - The following jobs have failed:

• CMake / windows (windows.g5.4xlarge.nvidia.gpu, cuda, 11.8) / windows-job (gh)
  C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\include\crt/host_config.h(153): fatal error C1189: #error: -- unsupported Microsoft Visual Studio version! Only the versions between 2017 and 2022 (inclusive) are supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk. [C:\actions-runner\_work\vision\vision\pytorch\vision\cpp_build\CMakeFiles\3.26.4\CompilerIdCUDA\CompilerIdCUDA.vcxproj]
• Tests / unittests-linux (3.8, linux.12xlarge, cpu) / linux-job (gh)
  test/test_ops.py::TestRoIAlign::test_autocast_cpu[rois_dtype1-x_dtype1-False-False]
• Tests / unittests-macos (3.10, macos-m1-stable) / macos-job (gh)
  test/test_ops.py::TestRoIAlign::test_autocast_cpu[rois_dtype1-x_dtype1-False-False]
• Tests / unittests-macos (3.11, macos-m1-stable) / macos-job (gh)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 1
• Tests / unittests-windows (3.10, windows.4xlarge, cpu) / windows-job (gh)
  The process 'C:\Program Files\Git\cmd\git.exe' failed with exit code 128
• Tests / unittests-windows (3.11, windows.4xlarge, cpu) / windows-job (gh)
  The process 'C:\Program Files\Git\cmd\git.exe' failed with exit code 128
• Tests / unittests-windows (3.12, windows.4xlarge, cpu) / windows-job (gh)
  The process 'C:\Program Files\Git\cmd\git.exe' failed with exit code 128
• Tests / unittests-windows (3.8, windows.4xlarge, cpu) / windows-job (gh)
  The process 'C:\Program Files\Git\cmd\git.exe' failed with exit code 128
• Tests / unittests-windows (3.8, windows.g5.4xlarge.nvidia.gpu, cuda, 11.8) / windows-job (gh)
  Process completed with exit code 139.
• Tests / unittests-windows (3.9, windows.4xlarge, cpu) / windows-job (gh)
  The process 'C:\Program Files\Git\cmd\git.exe' failed with exit code 128

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[nWEIdia]

nerfstudio-project/nerfstudio#3157 suggests:

CUDA Requirements:
MSVC Version 193x
Visual Studio 2022 17.x

17.10 is however 194x when checking the cl.exe file.

CUDA release pages were showing sth like:

and

So 17.10 (194x) would be too new for CUDA.

[atalman]

Closing this for now. AMI rolled back

[nWEIdia]

CUDA 11.8 Windows https://docs.nvidia.com/cuda/archive/11.8.0/cuda-installation-guide-microsoft-windows/index.html Table 2
CUDA 12.1 Windows https://docs.nvidia.com/cuda/archive/12.1.0/cuda-installation-guide-microsoft-windows/index.html Table 2
CUDA 12.4 Windows https://docs.nvidia.com/cuda/archive/12.4.0/cuda-installation-guide-microsoft-windows/index.html#id1

They did not seem to change, even though CUDA 12.4 has fixed the version check (and now supports MSVC 17.10.x) while dealing with MS's unconventional naming (bumping 19.3x to 19.4x to still map to 17.x, i.e. 17.10, instead of 18.x).

For future reference, if there is a need to bump to MSVC 17.10.x, errors with cuda 11.8 and 12.1 should be fine to use the -allow-unsupported-compiler as workaround.