Running on Mac Intel with AMD GPU #2659
raffetazarius
started this conversation in
General
Replies: 5 comments 8 replies
-
|
possibly, but not tested - none of us devs have that HW available for testing and its excedingy rare. |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
OpenVINO supports MacOS in theory but i don't have a Mac system to test it. |
Beta Was this translation helpful? Give feedback.
5 replies
-
|
Thanks @Disty0 - I'd like to test this out. I found settings for OpenVINO in the automatic UI under System > Compute Settings > IPEX, DirectML and OpenVINO Is there a way I can set these so that automatic runs using OpenVINO? Or is the only way to use command-line arg |
Beta Was this translation helpful? Give feedback.
-
|
OpenVINO is a different backend than others. Use a separate install with |
Beta Was this translation helpful? Give feedback.
-
|
OK so I've tried this with a new install. First time around, I think it tried to use my CPU to generate, but then failed out. Then I tried to disable the CPU as a device but still could not get a generation: Any idea what's wrong? |
Beta Was this translation helpful? Give feedback.
-
|
Seems like WebUI is trying to use MPS with OpenVINO and it naturally fails. Also seems like OpenVINO doesn't detect your GPU. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the suggestion. I tried that but it's still not detecting my GPU. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the reply. I've managed to install it and have it generating images using the "Diffusers" backend! What's weird is that I was getting errors using the "original" backend, even when I tried to find and tweak the right System Settings in order to match my working configuration of A1111 stable-diffusion-webui, which I could only get to work using the following args:
My System: Mac Pro 2019 Intel, 576 GB RAM, AMD W5700X 16GB GPU (Apple MPX), AMD RX 6900XT 16GB GPU (Reference). But with all of Vlad's automatic System Settings back to their default and only.... System > Settings > Execution & Models
.... I'm getting 512x512 images generated woth the v1-5-pruned.safetensors model. This is unexpected (but potentially great), since I didn't need to force Here is what my Terminal spit out in case useful for developers: Just to see if I could, I tried to generate a 1024x1024 image (which fails with an MPS backend error on A1111 webui due to insufficient VRAM I think), but it didn't work on Vlad's automatic either - I get:
|
Beta Was this translation helpful? Give feedback.
0 replies
-
|
I've just randomly started getting a new error and generations aren't working any longer. :(
Any ideas what's causing this? |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
@raffetazarius Any progress on this? I'm also on Intel Mac with a 6900 XT. |
Beta Was this translation helpful? Give feedback.
3 replies
-
|
it actually tells you version in the log during startup:
|
Beta Was this translation helpful? Give feedback.
-
|
that is NOT sdnext version string, that seems to be forge. apples and oranges. |
Beta Was this translation helpful? Give feedback.
-
Apologies I had replied here but then realized thanks to @vladmandic's comments that I had been looking at my automatic forge installation and not my SD Next one - both are branches of the original automatic1111 project. Oops, sorry! I've removed those replies to prevent confusion. I'll come back to this when I get a chance to check if my SD Next install is working - haven't used it in a while. |
Beta Was this translation helpful? Give feedback.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello!
Just wondering if anyone has successfully got this running on Mac Intel with an AMD GPU?
I’ve got regular A1111 running successfully on my rig (Mac Pro 2019 Intel Xeon + AMD Radeon RX 6900XT) but could not find mention of this fork working on Mac other than Apple Silicon M1/M2.
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions