[FEATURE] Support AMD Radeon GPUs

[tom-nom-nom]

Is your feature request related to a problem? Please describe.
My homelab-server has an amd cpu with integrated gpu and a dedicated amd gpu, but no Nvidia. As far as i can tell AudioMuse-AI does not work on Radeon even with settings that work for jellyfin (docker compose device and volume settings.)

Describe the solution you'd like
Tensorflow and ONNX should to my knowledge already support amd GPUs, they just need some configuration as to what GPU to use? (I have only briefly worked with them though)

Describe alternatives you've considered
No alternatives insight... Buying an Nvidia Laptop is currently a bit out of budget and a raspberry pi cant handle the volume of Music i have to Index ;)

Additional context
Nothing :)

[NeptuneHub]

Hi and thank you for the feedback. I don't have any computer with AMD Radeon to do it and test. If someone from the community have the possibility to implement this and raise a PR I'll be happy to merge it.

[mreid-tt]

This would be a great addition for me as well. I’m currently testing this on an AOOSTAR WTR MAX with an AMD Ryzen 7 PRO 8845HS and Radeon 780M graphics. I already use the GPU with other apps such as Frigate NVR, where their ROCm-based image runs on AMD GPUs to provide efficient object detection. I’m not sure whether a similar approach could be applied to this project.

At the moment, my library scan has been running for about 9–10 hours and is only ~67% complete. That said, this should largely be a one-time cost unless I add or remove a significant amount of music.