NVIDIA Jetson and Python Bindiings

[gautam1858]

@r0nlt Thank you for developing and sharing the "Space-Radiation-Tolerant" framework. Looks like a very valuable contribution for ML applications in challenging environments.

I have a couple of suggestions/questions for discussion:

NVIDIA Jetson GPU Board Support:

Would it be possible to discuss or consider including support or guidance for utilizing this framework with NVIDIA Jetson GPU boards (e.g., Jetson Nano, Xavier, Orin series)? These platforms are increasingly popular for edge AI applications, and their use in radiation-present environments could be highly beneficial. Any insights into compatibility, potential challenges, or planned support would be super useful.

Usage with Python Inference Scripts:

Many machine learning workflows, especially for inference, are built using Python. Could you provide some information on how this C++ framework can be best utilized with Python-based inference scripts?

For instance:

1.     Are there any existing or planned Python bindings?
   

2.     What would be the recommended approach for interfacing a Python inference script with the C++ radiation-tolerant
        components (for loading a model and performing inference through the framework)?
   

3.     Are there any examples or best practices you could share for such integration?
   

[r0nlt]

Thanks for showing interest in the framework, but I don't know the extent of space hardware that exists or if this would even be a good idea to integrate a hydride-based chip to handle the goals at a software and hardware level.

While Python integration and Jetson GPU support are certainly possible future additions, I'm not quite there yet in the development process.
I've considered integration paths, but I don't currently have the Jetson hardware to properly test and implement GPU-specific optimizations so I really wouldn't do anything I couldn't be sure about but I left this a little open ended for everyone else who really takes the time to look at the codebase and the project.
The current focus remains on strengthening the core radiation-tolerance mechanisms before expanding to additional platforms and language bindings.
But either way, you need to carefully manage the memory transfer between the CPU and GPU, gotta implement things at the kernel level. The framework's functionality varies based on the target environment.

I'll keep your ideas in mind, but the goal was to just make a middleware. I was mostly curious about things people might want to create and put in space.

[r0nlt]

@r0nlt Thank you for developing and sharing the "Space-Radiation-Tolerant" framework. Looks like a very valuable contribution for ML applications in challenging environments.

I have a couple of suggestions/questions for discussion:

NVIDIA Jetson GPU Board Support:

Would it be possible to discuss or consider including support or guidance for utilizing this framework with NVIDIA Jetson GPU boards (e.g., Jetson Nano, Xavier, Orin series)? These platforms are increasingly popular for edge AI applications, and their use in radiation-present environments could be highly beneficial. Any insights into compatibility, potential challenges, or planned support would be super useful.

Usage with Python Inference Scripts:

Many machine learning workflows, especially for inference, are built using Python. Could you provide some information on how this C++ framework can be best utilized with Python-based inference scripts?

For instance:

1.     Are there any existing or planned Python bindings?
   

2.     What would be the recommended approach for interfacing a Python inference script with the C++ radiation-tolerant
        components (for loading a model and performing inference through the framework)?
   

3.     Are there any examples or best practices you could share for such integration?
   

Im up to date, I have your ideas in mind!

[gautam1858]

Nice looking forward, let me know if we can collaborate, I can contribute to the codebase :)

[r0nlt]

shoot me an email and we can schedule something.

[gautam1858]

Your email id ?

[r0nlt]

rnuguruworkspace@gmail.com