search

aszc-dev / ComfyUI-CoreMLSuite

A set of custom nodes for ComfyUI that allow you to use Core ML models in your ComfyUI workflows.
GNU General Public License v3.0
131 stars 12 forks source link

ANE not activated #33

Open pechaut78 opened 11 months ago

pechaut78 commented 11 months ago

ComfyUI_temp_tjbyz_00011_

Updated to latest comfy, your package and.. and wether I ask for CPU & GPU or with ANE I get the SAME results in terms of iterations (around 1.8 on a max 3)

if I use asitop, I see that the ANE is never activated and if I use CPU and ANE, it uses the CPU and is slow ..

Here is my compo: drop it into comfy

cheers

aszc-dev commented 11 months ago

Hello. Judging by the dual CLIP, I'm assuming you are trying to run an SDXL model (or SSD-1B)? Unfortunately, SDXL models cannot be loaded to ANE. I'm not sure if it will ever be supported - here are some open issues in apple/ml-stable-diffusion on this topic: https://github.com/apple/ml-stable-diffusion/issues/271 https://github.com/apple/ml-stable-diffusion/issues/237 https://github.com/apple/ml-stable-diffusion/issues/267 For this project I relied heavily on that repo and its implementation of UNets (both 1.5 and XL). If SDXL on ANE is possible, a new implementation needs to be written.

alexqem commented 7 months ago

What About M3 chip?

sarievo commented 2 weeks ago

Why Apple Why

InformEthics commented 1 week ago

Hello. Judging by the dual CLIP, I'm assuming you are trying to run an SDXL model (or SSD-1B)? Unfortunately, SDXL models cannot be loaded to ANE. I'm not sure if it will ever be supported - here are some open issues in apple/ml-stable-diffusion on this topic: apple/ml-stable-diffusion#271 apple/ml-stable-diffusion#237 apple/ml-stable-diffusion#267 For this project I relied heavily on that repo and its implementation of UNets (both 1.5 and XL). If SDXL on ANE is possible, a new implementation needs to be written.

it is not completely true, there is a SDXL Split-Einsum model for iOS which is working with MochiDiffusion on my MBP M2 16gb. Actually I would like to import it to ComyUI, it is incredibly fast on the non Pro / Max chips. Is that possible?