PyMo: proof of concept

Hey HF community, I am sharing a project I released as a proof concept. Currently I am in the middle of designing its underlying framework. Its design and core engineering principles are for a framework that is domain-agnostic for versatility in application. Currently a leverage point of PyMo is the Python and Mojo programming languages. I look to run possibly a model in the Sci-kit Learn framework or maybe test the JAX framework in example models for cancer and neurology research in ML/DL domains to proof its implementation. Feedback from the HF community would be awesome thanks.

PyMo project: link

1 Like

Hi there,

This sounds like an exciting project—thanks for sharing! The idea of a domain-agnostic framework that bridges Python and Mojo for versatile applications is very intriguing. Leveraging frameworks like Sci-kit Learn or JAX for ML/DL models in fields as impactful as cancer and neurology research makes this even more compelling.

A couple of thoughts for feedback:

  • Scikit-learn Integration: This could be a great starting point, especially for interpretable models in research settings. Are you planning to incorporate tools for model evaluation and explainability, such as SHAP or LIME? These are often key for adoption in medical research.
  • JAX Experiments: Testing with JAX is a solid choice for more advanced DL use cases, especially if you’re exploring highly parallelized operations or advanced optimizations. Will PyMo include abstractions to simplify transitioning between different frameworks (e.g., JAX, PyTorch, or TensorFlow)?
  • Mojo’s Role: Since Mojo is relatively new, are you focusing on performance optimization, interoperability with Python, or something else entirely? It’d be interesting to hear how PyMo plans to utilize Mojo’s unique strengths.

Looking forward to seeing how PyMo evolves. If you’re looking for specific testing scenarios or community contributions, let us know! Best of luck with the framework development—it’s great to see innovative projects like this in action.

Cheers!

1 Like

Hey Alan :wave: yes I look for interopping with mentioned above. The Python ecosystem is vast. Which plays to the strengths in intention for this framework. For example with SHAP and LIME this can be applied in this framework. Mojo is foundational in the MAX stack which is an AI inference platform for ML/DL frameworks. So yes this framework can be applied in these different frameworks. I am still very much in the design and prototyping stage currently with no deadline as it is only me developing the project hehe. I was sure to label PyMo as very much in proof of concept.

Feel free to stop by the Mojo community we have a community Discord and forum :+1:

1 Like