lllyasviel

/

ControlNet

like 3.46k

It seems that controlnet is only compatible with SD 1.5 model. At least for me it works with 1.5, but not with 1.4 or 2.1 (where I get exactly the same error).

I was using 2.1 so I think you might be right. Thank you!

Have you tried specifying the 2.1 config file in the ControlNet settings in the UI? I have no idea if that works but there is a 2.1 config file, worth a shot.

I am using Colab Pro and i had the same issue. Using SD 1.5. You have to lower the resolution to 768 x 384 or maybe less. That has worked for me.

I have the same Problem,
first error with 2.1 model, Used 1.5 then and lowered the resultion down to 384x384. Tried with a lot of settings, always getting the same error.
Using this colab:
https://github.com/sagiodev/stablediffusion_webui/blob/master/StableDiffusionUI_ngrok_sagiodev.ipynb

I am specifying the 2.1 config file in the ControlNet settings. There are errors, but it works with sd2.1 .

I am specifying the 2.1 config file in the ControlNet settings. There are errors, but it works with sd2.1 .

i got this working, thanks

how did you get it working?

Hi EightBitRG,
use always the latest colab-Version of https://github.com/sagiodev/stablediffusion_webui/blob/master/StableDiffusionUI_ngrok_sagiodev.ipynb
after automatic1111 is running, go to the extension tab and -> check for updates -> apply and restart ui. if you encounter a Error written in red in the interface, dont hesitate, kill automatic1111 and restart colab (automatic1111 isnt that stable...)
and here is a screenshot of my settings https://imgur.com/a/fbTqF7N
it worked for me
good luck

Please help me. I'm just a dumb artist lol

AttributeError: 'NoneType' object has no attribute 'convert'

i went into same issue. but i fix it by going to sittings and add ( models\v2-1_768-ema-pruned.yaml )
here screen shot i hope its helping !!

Hey guys, I'm trying to do 2.1 Control net on a 2.1 model and getting errros asking for individual config files... but there are no yamls in this folder.

Loading model: control_v11p_sd21_openposev2 [f3edb4e5]██████████████████████████████████| 80/80 [00:13<00:00, 6.63it/s]
Loaded state_dict from [C:\Users\xxxx\Deep\stable-diffusion-webui\extensions\sd-webui-controlnet\models\control_v11p_sd21_openposev2.safetensors]
Error running process: C:\Users\xxxx\Deep\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py
Traceback (most recent call last):
File "C:\Users\xxxx\Deep\stable-diffusion-webui\modules\scripts.py", line 417, in process
script.process(p, *script_args)
File "C:\Users\xxxx\Deep\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 949, in process
model_net = self.load_control_model(p, unet, unit.model, unit.low_vram)
File "C:\Users\xxxx\Deep\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 636, in load_control_model
model_net = self.build_control_model(p, unet, model, lowvram)
File "C:\Users\xxxx\Deep\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 698, in build_control_model
assert os.path.exists(override_config), f'Error: The model config {override_config} is missing. ControlNet 1.1 must have configs.'
AssertionError: Error: The model config C:\Users\xxxx\Deep\stable-diffusion-webui\extensions\sd-webui-controlnet\models\control_v11p_sd21_openposev2.yaml is missing. ControlNet 1.1 must have configs.

I'm also getting the "RuntimeError: mat1 and mat2 shapes cannot be multiplied" but on both 1.5 & 2.1 models

For reference:

I'm on a 1080TI and have tried several variations in the below order (NB error changes for Torch1)

Torch2
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Commit hash: 5ab7f213bec2f816f9c5644becb32eb72c8ffb89
Launching Web UI with arguments: --port 7862 --medvram --xformers --no-half --no-half-vae
ControlNet v1.1.134
RuntimeError: mat1 and mat2 shapes cannot be multiplied (231x1024 and 768x320)
--
Torch2
Launching Web UI with arguments: --port 7862 --medvram --opt-split-attention --no-half --no-half-vae
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Commit hash: 5ab7f213bec2f816f9c5644becb32eb72c8ffb89
Launching Web UI with arguments: --port 7862 --medvram --xformers --no-half --no-half-vae
ControlNet v1.1.134
RuntimeError: mat1 and mat2 shapes cannot be multiplied (231x1024 and 768x320)
--
Torch1
Launching Web UI with arguments: --port 7862 --medvram --opt-split-attention --no-half --no-half-vae
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Commit hash: 5ab7f213bec2f816f9c5644becb32eb72c8ffb89
Launching Web UI with arguments: --port 7862 --medvram --xformers --no-half --no-half-vae
ControlNet v1.1.134
RuntimeError: The size of tensor a (1024) must match the size of tensor b (768) at non-singleton dimension 1
--
Torch1
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Commit hash: 5ab7f213bec2f816f9c5644becb32eb72c8ffb89
Launching Web UI with arguments: --port 7862 --medvram --xformers --no-half --no-half-vae
ControlNet v1.1.134
RuntimeError: The size of tensor a (1024) must match the size of tensor b (768) at non-singleton dimension 1

Hey getting the same error added config to the settings. There must be something we're missing here?

I am specifying the 2.1 config file in the ControlNet settings. There are errors, but it works with sd2.1 .
Can you recreate how you got this working?