Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

The first generation is normal, the second error #325

Open
shenyx110 opened this issue Oct 17, 2023 · 6 comments
Open

The first generation is normal, the second error #325

shenyx110 opened this issue Oct 17, 2023 · 6 comments
Labels
bug Something isn't working

Comments

@shenyx110
Copy link

Processor

M1 Pro (or later)

Memory

16GB

What happened?

The second error is generated when using all computing cores.
Open the steps to display, and then the picture can be generated normally, but the speed decreases significantly.

Crash Log

No response

@shenyx110 shenyx110 added the bug Something isn't working label Oct 17, 2023
@shenyx110
Copy link
Author

截屏2023-10-17 14 59 29

@jrittvo
Copy link
Contributor

jrittvo commented Oct 19, 2023

What specific model did you use? Was it a split-einsum or an original? What were the compute units on the first run that worked?

@stsydow
Copy link

stsydow commented Oct 24, 2023

I see this error on second run using IOS SDXL model:
https://huggingface.co/apple/coreml-stable-diffusion-xl-base-ios/tree/main

@jrittvo
Copy link
Contributor

jrittvo commented Oct 24, 2023

That error on a second run usually means that the ANECompiler Service process is geting stuck. Rebooting will reset it. But it will probably to continue to get stuck. There are a number of problems right now with large models trying to use the ANECompiler. Te problems are in macOS, not anything Mochi itself can fix. Also, your model is partly named "iOS". I'm not aware of models that run on iOS and not on macOS, but is it possible that is part of the problem here?

You might be able to get this model to run consistently using CPU and GPU. That would cut the ANE out of the mix.

@shenyx110
Copy link
Author

The model I used is ghostmix_v12_split-einsum. The calculation unit is ALL. The parameters have not changed. Details are as follows.
Date:
2023年10月17日 14:23:22
Model:
ghostmix_v12_split-einsum
Size:
2048 x 2048 (Upscaled using RealESRGAN)
Include in Image:
best quality,Amazing,Beautiful golden eyes,finely detail,Depth of field,extremely detailed CG unity 8k wallpaper,(1 girl :1.5),Deep sea background,jellyfish,masterpiece,fluttered detailed splashs, beautiful detailed water,cosmic eyes,Shock sensation,(realistic :0.5),octopus,original

Exclude from Image:
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, nsfw, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad hands, bad anatomy

Seed:
3038768038

Steps:
31

Guidance Scale:
9.0

Scheduler:
DPM-Solver++

ML Compute Unit:
All

@jrittvo
Copy link
Contributor

jrittvo commented Oct 29, 2023

Again, my suggestion would be to reboot your machine to clear out any processes that got stuck, and then run your split-einsum model only with CPU and GPU. Do not us it with ALL, or AUTO, or CPU and Neural Engine.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

3 participants