I was searching for an option to create characters out of an image.
I have found a possible workflow.
Due to I am new to ComfyUI I have installed the latest version, including phyton.
I have read that the upper mentioned workflow only works with ComfyUI version 0.2.3 or lower.
For this reason I have installed this version, too.
Now my issue.
On the latest version it is possible to select a checkpoint out of “unet” or “diffusion_models” inside of nodes “Unet Loader (GGUF)” & “Load Diffusion Model”. But it is not possible to install Pulid-Nodes (because of Python version).
On version 0.2.3 it is possible to add Pulid-Nodes, but it is not possible to select a safetensor out of “Unet Loader (GGUF)” & “Load Diffusion Model”. Values are null or undefined. I have copied specific safetensors in folders “unet” and “diffusion_models”. I have tried to update ComfyUI, python, crosscopied python of both versions. Unfortunately nothing works.
Additional information: I don’t know if these topics are connected:
Failure message in cmd:
Install ComfyUI in respective environments and keep them isolated.
bash
Copy code
source comfyui_env_0_2_3/bin/activate
pip install comfyui==0.2.3
# For the latest version
source comfyui_env_latest/bin/activate
pip install comfyui
3. Address Node Issues:
For Pulid-Nodes: Ensure compatibility by checking Pulid-Nodes’ requirements and their GitHub repository for updates. Install their dependencies using pip install -r requirements.txt.
For Safetensor Loading:
Verify that the safetensors are correctly placed in unet or diffusion_models.
Check folder paths in config.json or ComfyUI settings for accuracy.
4. Debug CMD Errors:
Inspect the node directories (custom_nodes\webui-controlnet-v1-archived-main and custom_nodes\ComfyUI-Advanced-ControlNet). Remove nodes temporarily and test loading ComfyUI without them.
Hi Alan,
thank you for your support.
To be honest I am a total newb.
Could you please give me once a dummy manual where I have to insert command “bash” or “python -m venv comfyui_env_0_2_3” to create the virtual environment?
Same for “source …”
I have tried cmd in several folders and typed in those commands, but I receive always “command not known” … etc.
I have downloaded python-3.10.11-embed-amd64.zip and out of desperation I have tried to replace 3.11.xx but this doesn’t worked (comfyui doesn’t start). I have undone this action.
Next, I have created a new installation:
D:\ComfyUI_windows_portable_v0.2.3
But I think I made it even worse. Currently I am facing both issues:
Can’t open a safetensor from “Unet Loader (GGUF)” & “Load Diffusion Model” although folders are filled “models\diffusion_models\FLUX1\flux1-dev-fp8.safetensors” & “models\unet\FLUX1\flux1-dev-fp8.safetensors”
Not possible to install Pulid-Nodes (Import failed)
About information:
ComfyUI 0.2.3
Python Version 3.11.9
Embedded Python true
Pytorch Version 2.4.1+cu124
Due to the fact that I can’t install Pulid-Nodes I think I am grabbing after a higher version of python in the background.
I also installed python 3.10.11 on C:\
Sorry, but I need deeper support regarding hints 1. and 2.
After upgrading to the new version of ComfyUI, if you use GGUF loader once, you will find you can never use the Load Diffusion Model node to load .safetensor file, only GGUF file can be displayed and chose! I don’t know except creating a new environment, is there other way to solve it?
I fixed this by deleting the gguf folder from ComfyUI/custom_nodes/ComfyUI-Flow-Control/nodes … probably if you just delete the nodes.py file from the folder is enough … the file is probably overwrites the node.py torch method of loading the models.