KwaiVGI

/

LivePortrait

like 254

Follow

Kuaishou Visual Generation and Interaction Center 87

How long does it take to run once?

Thanks for your attention.
I’m sorry that we currently don’t have an AMD GPU for testing. One direct question that comes to mind is: Can ONNX run on AMD GPUs? @mziada

@cleardusk Thanks
I hope you can do the testing as this will be massive for all the AMD users.
ONNX works on AMD via DirectML on Windows which is a Windows library. I have been able to do that with StableDiffusion ONNX and DirectML on Windows. Also as you maybe aware AMD has ROCm that works via Linux.
I have a slightly older AMD GPU but it has 8GB of dedicated RAM, the model is RX 480, and I am happy to do any testing for you on my machine if you'd like me to.

Also @cleardusk
This is AMD and ROCm (on Linux)
https://rocm.blogs.amd.com/artificial-intelligence/stable-diffusion-onnx-runtime/README.html

And for Windows it works via DirectML
https://github.com/microsoft/Olive/blob/main/examples/directml/stable_diffusion/README.md

Thanks for your additional resources and support! @mziada
We will try to apply for an AMD GPU and test it, but its priority might not be very high since we have many other requests queued up (from GitHub repo https://github.com/KwaiVGI/LivePortrait). 😂