Vaida12345/waifuExtension

A suggestion about the app running on M1.

ricozxy opened this issue · 8 comments

The Real-CUGAN model runs slowly on M1. Hope to use ANE on Real-CUGAN.
Thanks!

And also I hope the App can support the Double-GPU of Mac Pro.

Currently, I am using the executable directly from the author. I am looking to integrate that into CoreML model, which can use ANE.

About the second suggestion, I think I will look into that.

Thanks for your attention. I have solved the first problem.

Can you give 5.0 Beta 9 a try please?

sure.

Looks can't use GPU 1
Snipaste_2022-08-02_23-07-32

Can you please show me the way you called the executable to run inference on two CPUs?

And if you have Xcode installed, can you please show me the value of MTLCopyAllDevices().count?