This model is a fine-tune of Stable Cascade stage C on 5000 images. The training dataset focuses for realism and consists of a lot of NSFW pictures, so the resulting model is capable of generating some NSFW content. The uploaded checkpoint has been converted to work with the standard ComfyUI Stable Cascade workflow.
Description
FAQ
Comments (14)
I copied your wokflow and when i run it i get at stage B checkpoint--
Error occurred when executing CheckpointLoaderSimple: 'model.diffusion_model.input_blocks.0.0.weight'
You're probably using the wrong stage B checkpoint. You should download it here: https://huggingface.co/stabilityai/stable-cascade/tree/main/comfyui_checkpoints.
@altron thank you. will do when huggingface is back up!
@altron downloaded the one linked, worked. Thank you
@altron that did the trick, thanks
I am unable to even load the stage. As soon as it starts loading (after loading CLIP) it throws this:
Error occurred when executing UNETLoader: 'conv_in.weight' File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\execution.py", line 152, in recursive_execute output_data, output_ui = get_output_data(obj, input_data_all) File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\execution.py", line 82, in get_output_data return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True) File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\execution.py", line 75, in map_node_over_list results.append(getattr(obj, func)(**slice_dict(input_data_all, i))) File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\nodes.py", line 838, in load_unet model = comfy.sd.load_unet(unet_path) File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\comfy\sd.py", line 596, in load_unet model = load_unet_state_dict(sd) File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\comfy\sd.py", line 569, in load_unet_state_dict model_config = model_detection.model_config_from_diffusers_unet(sd) File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\comfy\model_detection.py", line 360, in model_config_from_diffusers_unet unet_config = unet_config_from_diffusers_unet(state_dict) File "D:\Users\danda\Downloads\stability\Data\Packages\ComfyUI\comfy\model_detection.py", line 261, in unet_config_from_diffusers_unet match["model_channels"] = state_dict["conv_in.weight"].shape[0]
You should not be using the UNETLoader node, use CheckpointLoader instead. You can use the official example as a template: https://comfyanonymous.github.io/ComfyUI_examples/stable_cascade/.
@altron Oh. I thought it is the same as the 3 stage model I am using for vanilla stable cascade which I downloaded from stable cascade hugging page.
I'm not sure if I'm doing something wrong, but I just switched the C checkpoint for this one, and all my images look really weird, cartoonish and have strange colors. I'm using the default comfyUI workflow and haven't changed any settings.
... and if anyone can link me to a guide on what models are needed (and where to put them - Checkpoints, unet or VAE) it would be greatly appreciated. Using ComfyUI on Windows. If Stable Cascade is not meant to be tested by "normal" people yet, disregard this request.
what if you try to load a workflow from one of the images in my post?
you can just drag it into comfyui and it should get loaded, you might need to pick checkpoint from the list manually
@altron I'm having the same issue. The new models are installed correctly. I dragged one of your images into ComfyUI, selected the models, but I'm still getting the colored result.
@altron I can confirm this only happens when using your model. When switching back to model c, there are no issues.