ControlNet icon indicating copy to clipboard operation
ControlNet copied to clipboard
verified publisher icon lllyasviel

Now not working was going great

Open ColinTitahi opened this issue 2 years ago • 13 comments

Hi Love this looks fantastic. I knew there must be a way of combining open pose with diffusion. Had it working great on a previous commit. Rebuilt environment and a fresh pull and re downloaded models Now if fails after Loaded model config from [./models/cldm_v15.yaml] Killed

any help?

ColinTitahi avatar Feb 13 '23 05:02 ColinTitahi

killed generally means you ran out of memory

brycedrennan avatar Feb 13 '23 05:02 brycedrennan

Thanks but my my V Ram is barley touched (almost nothing used of 12GB) but normal ram is getting quite high but does not seem to max out and it was running fine. Maybe I lost something with the environment rebuild?

ColinTitahi avatar Feb 13 '23 05:02 ColinTitahi

yeah killed would be for normal ram getting maxed. Possible to not see in an activity monitor because it quickly kills the process.

brycedrennan avatar Feb 13 '23 05:02 brycedrennan

Thanks - hmmm 16gb not enough?

ColinTitahi avatar Feb 13 '23 05:02 ColinTitahi

no idea. could be you have too many other things open. could be a bug in the code here.

brycedrennan avatar Feb 13 '23 05:02 brycedrennan

RAM maxxing out here too, on 12gb free collab notebooks.

G-force78 avatar Feb 13 '23 11:02 G-force78

we are looking into this https://github.com/lllyasviel/ControlNet/pull/25

lllyasviel avatar Feb 13 '23 15:02 lllyasviel

no idea. could be you have too many other things open. could be a bug in the code here.

RAM maxxing out here too, on 12gb free collab notebooks.

Forgive my ignorance, but is there a link i missed somewhere? Did you make your own Colab notebook? Could you share it and/or explain how?

Edit: fwiw, I'm personally not concerned with training yet, just trying to get image generation working.

AmusedDiffuser avatar Feb 13 '23 18:02 AmusedDiffuser

no idea. could be you have too many other things open. could be a bug in the code here.

RAM maxxing out here too, on 12gb free collab notebooks.

Forgive my ignorance, but is there a link i missed somewhere? Did you make your own Colab notebook? Could you share it and/or explain how?

Edit: fwiw, I'm personally not concerned with training yet, just trying to get image generation working.

No worries here you go, depth map maxes out RAM the scribble map one is fine https://colab.research.google.com/drive/1VRrDqT6xeETfMsfqYuCGhwdxcC2kLd2P?usp=sharing

G-force78 avatar Feb 13 '23 21:02 G-force78

you can try automatic1111 extension too

It works amazingly well and supports xformers as well now since automatic1111 supports

Also not anymore you need a merged ckpt files

It works with any custom model as well

16.) Automatic1111 Web UI Sketches into Epic Art with 1 Click: A Guide to Stable Diffusion ControlNet in Automatic1111 Web UI image

FurkanGozukara avatar Feb 14 '23 17:02 FurkanGozukara

Hi, i faced the same problem and it is because the model is not releasing the memory after inference i had to use:

torch.cuda.empty_cache()
gc.collect()

before the return statement in the process function

fkite-cs avatar Feb 15 '23 09:02 fkite-cs

@fkite-cs Can you please mention which file should i find the process function in.

faiza333 avatar May 17 '23 13:05 faiza333

@faiza333 you can find that function in any gradio file of the repository i.e gradio_scribble2image.py

fkite-cs avatar May 17 '23 14:05 fkite-cs

all duplicate concerning "RAM and out of memory exceptions (OOM)": https://github.com/lllyasviel/ControlNet/issues/21 https://github.com/lllyasviel/ControlNet/issues/33 https://github.com/lllyasviel/ControlNet/issues/191 https://github.com/lllyasviel/ControlNet/issues/236 https://github.com/lllyasviel/ControlNet/issues/241 https://github.com/lllyasviel/ControlNet/issues/247 https://github.com/lllyasviel/ControlNet/issues/294 https://github.com/lllyasviel/ControlNet/issues/301

geroldmeisinger avatar Sep 17 '23 10:09 geroldmeisinger