Regarding this version of gguf
#33
by
Theoldsong
- opened
Using the gguf version of the model, both the generation effect and especially the character consistency are particularly poor. And the graphics memory usage is not lower than the original version. The Q3 version has reached 12GB. What is the significance of this gguf?
The original Allinone only requires one model. Text encoder, VAE, none of these are needed, why does gguf still need these?
The original Allinone only requires one model. Text encoder, VAE, none of these are needed, why does gguf still need these?
I read, because the GGUF model uses the UNET, so even if it is converted, only the model remains and not the Clip and the VAE.
I also looked on ComfyUI, and there is no GGUF node to load the three models together.