[BUG]Installation Error on Torch while Torch is Installed

[evencogitate]

C:\>python -m pip install deepspeed
Collecting deepspeed
  Using cached deepspeed-0.17.6.tar.gz (1.6 MB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... error
  error: subprocess-exited-with-error

  × Getting requirements to build wheel did not run successfully.
  │ exit code: 1
  ╰─> [26 lines of output]
      [WARNING] Unable to import torch, pre-compiling ops will be disabled. Please visit https://pytorch.org/ to see how to properly install torch on your system.
       [WARNING]  unable to import torch, please install it if you want to pre-compile any deepspeed ops.
      DS_BUILD_OPS=1

But Torch is clearly installed:

C:\>python
Python 3.13.3 (tags/v3.13.3:6280bb5, Apr  8 2025, 14:47:33) [MSC v.1943 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> print(torch.__version__)
2.8.0+cu128
>>> print(torch.cuda.is_available())
True
>>> exit()

[evencogitate]

I have tried --no-build-isolation as mentioned in https://github.com/deepspeedai/DeepSpeed/issues/3329, but it continues to have error:

in installed_cuda_version
          raise MissingCUDAException("CUDA_HOME does not exist, unable to compile CUDA op(s)")
      op_builder.builder.MissingCUDAException: CUDA_HOME does not exist, unable to compile CUDA op(s)

I have also noticed https://github.com/deepspeedai/DeepSpeed/issues/7286, but no solution was reached.

I was expecting the latest version could have better support for Windows compatibility as 3329 was closed.

[loadams]

@evencogitate - are you trying to use DeepSpeed on Windows then? We've not published new Windows wheels lately, and the Windows version is tied to the python and torch versions as well.

[evencogitate]

Yes. May I know which version of python and torch I shall use?

[DevKPro]

I've encountered the same problem and look forward to an answer.