Image-to-Image
Diffusers
Safetensors
English
controlnet
laion
face
mediapipe

Convert to diffusers

#3
by patrickvonplaten - opened
Crucible AI org

Thanks! Yes we can try to help with this on Monday. Any help you can provide would be great.

(onnx.venv) N:\StableDiffusionUI-Shark-AMD\data\repo\diffusion_scripts\modules\cn_face>py conv.py --checkpoint_path=".\controlnet_sd21_laion_face_v2_pruned.pth" --dump_path="./test/" --original_config_file="cfg.yaml"
Traceback (most recent call last):
  File "N:\StableDiffusionUI-Shark-AMD\data\repo\diffusion_scripts\modules\cn_face\conv.py", line 80, in <module>
    controlnet = download_controlnet_from_original_ckpt(
  File "N:\StableDiffusionUI-Shark-AMD\data\repo\onnx.venv\lib\site-packages\diffusers\pipelines\stable_diffusion\convert_from_ckpt.py", line 1354, in download_controlnet_from_original_ckpt
    controlnet_model = convert_controlnet_checkpoint(
  File "N:\StableDiffusionUI-Shark-AMD\data\repo\onnx.venv\lib\site-packages\diffusers\pipelines\stable_diffusion\convert_from_ckpt.py", line 1301, in convert_controlnet_checkpoint
    converted_ctrl_checkpoint = convert_ldm_unet_checkpoint(
  File "N:\StableDiffusionUI-Shark-AMD\data\repo\onnx.venv\lib\site-packages\diffusers\pipelines\stable_diffusion\convert_from_ckpt.py", line 380, in convert_ldm_unet_checkpoint
    new_checkpoint["time_embedding.linear_1.weight"] = unet_state_dict["time_embed.0.weight"]
KeyError: 'time_embed.0.weight'
This comment has been hidden

Very cool @chris-crucible ,

I've converted the diffusers checkpoint, verified that it works and also added a short example. You can find everything here: https://huggingface.co/CrucibleAI/ControlNetMediaPipeFace/discussions/6

JosephCatrambone changed discussion status to closed

Sign up or log in to comment