Custom Models just give errors | Invoke | Page 1

harsh marsh Feb 13, 2023, 1:11 AM

#

it seems like InvokeAI can only install models from the huggingface repo id if it's a diffusers model - the model in question isn't, so you can just download it and add it the 'usual' way

#

to be specific, don't grab the 2.1 version right now - support for that will come in an update

ember dock Feb 13, 2023, 1:15 AM

#

lol I'm sorry, this makes me feel stupid, but i'm really not. Getting these custom ckpts to run on Automatic1111 is really as easy as sticking them in a folder...why does Invoke make it such an arduous process? I don't even know what the 'usual' way is anymore as it seems to change every update of the git repo.

harsh marsh Feb 13, 2023, 1:15 AM

#

just download the ckpt, put it in a folder, then either use the model manager to install it, or the cli, or just edit models.yaml

ember dock Feb 13, 2023, 1:16 AM

#

Alright, I'll give it a try again. Thanks 👍🏼

ember dock Feb 13, 2023, 1:34 AM

#

I appended this to the models.yaml file "ClassicNegative:
description: ''
config: C:/Users/retro/stable-diffusion-webui/models/Stable-diffusion/classicNegativeSD21_classicNegative768px.yaml
weights: C:/Users/retro/stable-diffusion-webui/models/Stable-diffusion/classicNegativeSD21_classicNegative768px.ckpt
width: 768
height: 768
default: false
format: ckpt" Then I restarted invoke, same errors as before. I really like Invoke's UI but I can't invest more time into this...I'll come back to it when its easier to add models. SD 2.0 is really important for me because coherency is just better across the board, I hope it'll get fixed in the future.

ember dock Feb 13, 2023, 2:45 AM

#

I wanted to give you guys more context incase it helps, this was with no diffusers, just the local ckpt and yaml: >> Model change requested: classicNegativeSD21_classicNegative768px

Current VRAM usage: 2.17G
Offloading stable-diffusion-1.5 to CPU
Loading classicNegativeSD21_classicNegative768px from C:\Users\retro\invokeai\models\ldm\classicNegativeSD21_classicNegative768px.ckpt
Scanning Model: classicNegativeSD21_classicNegative768px
Model scanned ok!
Loading classicNegativeSD21_classicNegative768px from C:\Users\retro\invokeai\models\ldm\classicNegativeSD21_classicNegative768px.ckpt
| Forcing garbage collection prior to loading new model
** model classicNegativeSD21_classicNegative768px could not be loaded: LatentDiffusion.init() missing 1 required positional argument: 'personalization_config'
Traceback (most recent call last):
File "C:\Users\retro\invokeai.venv\lib\site-packages\ldm\generate.py", line 889, in set_model
model_data = cache.get_model(model_name)
File "C:\Users\retro\invokeai.venv\lib\site-packages\ldm\invoke\model_manager.py", line 106, in get_model
requested_model, width, height, hash = self._load_model(model_name)
File "C:\Users\retro\invokeai.venv\lib\site-packages\ldm\invoke\model_manager.py", line 335, in _load_model
model, width, height, model_hash = self._load_ckpt_model(
File "C:\Users\retro\invokeai.venv\lib\site-packages\ldm\invoke\model_manager.py", line 428, in _load_ckpt_model
model = instantiate_from_config(omega_config.model)
File "C:\Users\retro\invokeai.venv\lib\site-packages\ldm\util.py", line 92, in instantiate_from_config
return get_obj_from_str(config['target'])(
TypeError: LatentDiffusion.init() missing 1 required positional argument: 'personalization_config'

** trying to reload previous model

Retrieving model stable-diffusion-1.5 from system RAM cache
Textual inversions available:
Setting Sampler to k_lms (LMSDiscreteScheduler)

#

#Custom Models just give errors