Using Cuda instead of ROCM for some reason. | Invoke | Page 1

boreal egret Dec 31, 2022, 4:16 AM

#

My bad, on a second try it's still saying Cuda but it started using my video card, at first it was using my CPU.

#

By second try I mean I restarted the app.

boreal egret Dec 31, 2022, 1:50 PM

#

Well, it seems to be using the CPU in the end and my video card vram, it's weird. I'm getting 3it/second, though. Not that bad but I see my CPU piking, unlike the other webui I was using where I get 7it/s and doesn't trigger the CPU.

Any ideas? The interface is amazing by the way, hard to go back to the other tool.

#

Weird thing is that I don't think I'm supposed to be able to get 3it/s from the CPU alone but the CPU doesn't trigger like that when using the other webui and the video card metrics show it processing with the other one but not with InvokeAI. It's using the vcard vram, though.

analog sigil Dec 31, 2022, 7:25 PM

#

When invoke starts up, do you see a message indicating that it is using "cpu" ?

#

From the developer's console, type the command "pip list" , find the line for torch, and report what version got installed.

boreal egret Jan 2, 2023, 1:03 PM

#

Thanks for answering @analog sigil, i thought I had deleted this message of mine. After restarting the app it started rendering using the GPU, although it still says CUDA, but the speed is as good as the other one I was using.

winged flume Jan 10, 2023, 2:14 AM

#

From https://invoke-ai.github.io/InvokeAI/installation/020_INSTALL_MANUAL/#my-renders-are-running-very-slowly

One of the earlier lines should say Using device type cuda. On AMD systems, it will also say "cuda", and on Macintoshes, it should say "mps". If instead the message says it is running on "cpu", then you may need to install the correct torch library.
I thought it was supposed to say "rocm" for some reason and I reinstalled too many times before I realized it's actually using the GPU

#Using Cuda instead of ROCM for some reason.