Pip install issue: UnboundLocalError: local variable 'version' referenced before assignment

[TeKett]

pip install transformer_engine_torch
Collecting transformer_engine_torch
  Downloading transformer_engine_torch-2.6.0.post1.tar.gz (169 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 169.4/169.4 kB 7.6 MB/s eta 0:00:00
  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
  ╰─> [23 lines of output]
      /tmp/pip-build-env-j4xz7zlc/overlay/lib/python3.10/site-packages/torch/_subclasses/functional_tensor.py:279: UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at /pytorch/torch/csrc/utils/tensor_numpy.cpp:81.)
        cpu = _conversion_method_template(device=torch.device("cpu"))
      Traceback (most recent call last):
        File "/home/user/venvs/kohya_ss/venv/lib/python3.10/site-packages/pip/_vendor/pep517/in_process/_in_process.py", line 351, in <module>
          main()
        File "/home/user/venvs/kohya_ss/venv/lib/python3.10/site-packages/pip/_vendor/pep517/in_process/_in_process.py", line 333, in main
          json_out['return_val'] = hook(**hook_input['kwargs'])
        File "/home/user/venvs/kohya_ss/venv/lib/python3.10/site-packages/pip/_vendor/pep517/in_process/_in_process.py", line 118, in get_requires_for_build_wheel
          return hook(config_settings)
        File "/tmp/pip-build-env-j4xz7zlc/overlay/lib/python3.10/site-packages/setuptools/build_meta.py", line 331, in get_requires_for_build_wheel
          return self._get_build_requires(config_settings, requirements=[])
        File "/tmp/pip-build-env-j4xz7zlc/overlay/lib/python3.10/site-packages/setuptools/build_meta.py", line 301, in _get_build_requires
          self.run_setup()
        File "/tmp/pip-build-env-j4xz7zlc/overlay/lib/python3.10/site-packages/setuptools/build_meta.py", line 512, in run_setup
          super().run_setup(setup_script=setup_script)
        File "/tmp/pip-build-env-j4xz7zlc/overlay/lib/python3.10/site-packages/setuptools/build_meta.py", line 317, in run_setup
          exec(code, locals())
        File "<string>", line 46, in <module>
        File "/tmp/pip-install-zd4f85fz/transformer-engine-torch_59e978a90d4840eab7dbec24b5ee55ca/build_tools/pytorch.py", line 56, in setup_pytorch_extension
          version = cuda_version()
        File "/tmp/pip-install-zd4f85fz/transformer-engine-torch_59e978a90d4840eab7dbec24b5ee55ca/build_tools/utils.py", line 272, in cuda_version
          version_str = version("nvidia-cuda-runtime-cu12")
      UnboundLocalError: local variable 'version' referenced before assignment
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

```-

[ptrendx]

@timmoon10 Could you take a look? @ksivaman FYI

[Oblynx]

I got the same error. I went to this last file mentioned in the pip cache (uv, not pip in my case) and patched it. It's a hack, but it helped me get past this:

def cuda_version() -> Tuple[int, ...]:
    """CUDA Toolkit version as a (major, minor) tuple.

    Try to get cuda version by locating the nvcc executable and running nvcc --version. If
    nvcc is not found, look for the cuda runtime package pip `nvidia-cuda-runtime-cu12`
    and check pip version.
    """

    try:
        nvcc_bin = nvcc_path()
    except FileNotFoundError:
        pass
    else:
        output = subprocess.run(
            [nvcc_bin, "-V"],
            capture_output=True,
            check=True,
            universal_newlines=True,
        )
        match = re.search(r"release\s*([\d.]+)", output.stdout)
        version = match.group(1).split(".")
        return tuple(int(v) for v in version)

+   from importlib.metadata import version
+
    try:
        version_str = version("nvidia-cuda-runtime-cu12")
        version_tuple = tuple(
            int(part) for part in version_str.split(".") if part.isdigit()
        )
        return version_tuple
    except importlib.metadata.PackageNotFoundError:
        raise RuntimeError(
            "Could neither find NVCC executable nor CUDA runtime Python package."
        )