gitmylo / audio-webui

A webui for different audio related Neural Networks
MIT License
1.01k stars 94 forks source link

[FEATURE REQUEST] AMD acceleration support #96

Open sammcj opened 1 year ago

sammcj commented 1 year ago

Is your feature request related to a problem? Please describe. It would be awesome if AMD GPUs/APUs were supported.

Describe the solution you'd like I believe for other ML projects such as LocalAI etc... CLBLAS is used.

Describe alternatives you've considered Winning the lotto so I can afford an nvidia card.

Additional context Add any other context or screenshots about the feature request here.

gitmylo commented 1 year ago

It's a bit complicated. windows/linux + nvidia -> regular pytorch cuda windows + amd/intel/nvidia -> pytorch-directml linux + amd -> pytorch rocml

And I'm not even sure if they are fully compatible, I'll need people to test.

sammcj commented 1 year ago

Ahh I see, interesting. I'm happy to test it with my setup:

AdamLeyshon commented 1 year ago

I am willing to test: Windows 11 / Manjaro Linux + AMD 5900X + AMD 6900 XT GPU.

Using PyTorch ROCM under Manjaro works OK with SD and Automatic's Web UI so I hope this will be portable enough.

helikopterodaktyl commented 8 months ago

Can confirm AudioLDM/AudioLDM2/AudioCraft are working well on RX 6800XT with ROCM 5.6.1 and Linux Mint. Needed to uninstall the default installed cu11 Torch libraries and reinstall rocm ones. Haven't tried RVC yet.