search

Spycsh / xtalker

Faster Talking Face Animation on Xeon CPU
MIT License
120 stars 9 forks source link

MPS device compatibility #5

Open adills opened 1 year ago

adills commented 1 year ago

With Pytorch and TF both supporting Apple's M1/M2 silicon chips, you should be able to include a device checker and include "mps" capability for GPU processing. If only I had the time to create a branch and show you.

Spycsh commented 1 year ago

Thanks for info:) I do not have access to a Mac currently and the optimization here in my repo should only be used for intel xeon cpu. But good to know about the mps device compatibility. Is it fast enough?

adills commented 1 year ago

Thank you for your quick reply. I can run inference.py using your int8 instructions and --cpu on my MacBook M1 Max Pro but at a rate of ~18s/it (not even 1 it/s). If I get some time, I might create a fork and see if I can modify your engine to accept device=mps.