Issues when installing and running the repository on WSL/Windows and Colab

[mickmolitor]

Hey,

during my research I came across your repository and found it highly interesting, as I am working in a related area. I attempted to install and run it under different environments, but unfortunately encountered some blocking issues:

1. WSL (Ubuntu on Windows):
   When installing under WSL, the import of TensorFlow (pulled in via your dependencies) fails immediately.

   • First error:

     TypeError: Unable to convert function return value to a Python type! The signature was () -> handle
     

   • After some version adjustments, the error became:

     ml_dtypes.pywrap_ml_dtypes.float8_e4m3fn() -> TypeError: Unable to convert function return value to a Python type
     

   This seems to be caused by binary incompatibilities between tensorflow, ml-dtypes, and numpy on Linux/WSL. Interestingly, this issue does not appear on native Windows, where TensorFlow wheels are packaged differently.

2. Native Windows installation:
   Installation works more smoothly, but when importing the library and running example files I ran into multiple circular dependency issues.

3. Google Colab (with the provided button):
   Even in Colab, the examples did not execute successfully. I encountered ImportErrors, likely related to legacy TensorFlow v1/v2 code that is not compatible with the current runtime.

Given that the same errors occur in Colab, I assume the problems are not only due to my setup. The repository is quite comprehensive, so I wanted to let you know that—at least in my case—I could not get it running in WSL, Windows, or Colab without errors.

Perhaps I am missing something obvious, and if so I would be very happy if you could point me in the right direction. But since even the Colab examples currently break, I suspect some maintenance work might be required.

Sorry for the inconvenience, and thank you for your work on this project.

Best regards