Unable to pip install cuquantum on Windows due to conflicting dependencies error

[foongminwong]

We tried pip install cuquantum, then the installation failed and threw the following error: ERROR: Cannot install cuquantum==23.10.0, cuquantum==23.3.0, cuquantum==23.6.0, cuquantum==23.6.1 and cuquantum==24.3.0 because these package versions have conflicting dependencies.

Below shows the full error message:

Does anybody know how to resolve this?

Python Version: 3.10.5
OS: Windows 10 (with NVIDIA GPU)

[A-tA-v]

Same issue on OSX Sonoma, Python 3.12, no CUDA GPU (macbook pro)

[A-tA-v]

It seems that installation is unable to find CUDA libraries.

  ************************************************************************
  WARNING: Unable to detect NVIDIA CUDA Toolkit installation, assuming
  WARNING: CUDA 11 is in use... This behavior is subject to future change.
  WARNING: To be explicit, instead of "pip install cuquantum-python" run
  WARNING: "pip install cuquantum-python-cuXX", with XX being the major
  WARNING: version of your CUDA Toolkit installation.
  ************************************************************************

The question: is it possible to install CUDA toolkit without CUDA hardware?
The old answers suggested some workarounds: https://stackoverflow.com/questions/3087361/gpu-emulator-for-cuda-programming-without-the-hardware
It turns out there is no a good solution ... so no cuQuantum at the moment.

[yangcal]

As of cuQuantum 24.06, Windows is not an officially supported OS. Meanwhile, an NVIDIA GPU with compute capability 7.0+ is required. For more details, please refer to cuTensorNet and cuStateVec doc page

[foongminwong]

Thank you, how about Ubuntu (WSL2)?

[yangcal]

Since Ubuntu WSL2 has full linux kernels, my guess is that it should work though we haven't tested it. Feel free to share the outcome with us.