Don't set CUDA_HOME when not compiled with CUDA support (pytorch#106310)

It doesn't make sense to set this (on import!) as CUDA cannot be used with PyTorch in this case but leads to messages like
> No CUDA runtime is found, using CUDA_HOME='/usr/local/cuda'
when CUDA happens to be installed which is at least confusing.

Pull Request resolved: pytorch#106310
Approved by: https://github.com/ezyang

torch/utils/cpp_extension.py

@@ -208,7 +208,7 @@ def _join_rocm_home(*paths) -> str:
 if torch.version.hip is not None:
     ROCM_VERSION = tuple(int(v) for v in torch.version.hip.split('.')[:2])
 
-CUDA_HOME = _find_cuda_home()
+CUDA_HOME = _find_cuda_home() if torch.cuda._is_compiled() else None
 CUDNN_HOME = os.environ.get('CUDNN_HOME') or os.environ.get('CUDNN_PATH')
 # PyTorch releases have the version pattern major.minor.patch, whereas when
 # PyTorch is built from source, we append the git commit hash, which gives