Acly

See https://github.com/Acly/krita-ai-diffusion/discussions/1176 for the current state. You can use GGUF models now for 4/5/6/7/8 bit depending on how much VRAM you have.

There is an issue with the .onnx file mentioned in the error message. It might be incomplete or corrupt. If it exists, try deleting it (it will automatically re-download).

I tried with packages located in a path with chinese characters, and it works for me. So that doesn't seem to be the problem? But I also don't have any...

Please provide full error from client.log (click "View log files")

NF4 still uses ~10.5GB on my system. I am not sure it will fit entirely into 8GB. (It will probably work, but very slow)

See https://github.com/Acly/krita-ai-diffusion/discussions/1176 for current state on low-VRAM models.

Strange, there is currently nothing required for "Flux workload" apart from having a checkpoint. Can you try to restart, and post [client.log](https://github.com/Acly/krita-ai-diffusion/wiki/Common-Issues#log-files) if it still complains about workload.

There is `ARGOS_PACKAGE_DIR` environment variable which you can set to alternative folder. #1015 is similar.

Not a file, you set an environment variable, then restart the application (also restart comfyui if you don't use managed server). How to edit environment variables depends on your OS,...

The client was disconnected, so probably the server crashed. Probably out of RAM