Lora vs Model
#7
by Privac - opened
Is there any difference in the output between using the lora versus the distilled model?
Answering myself: the models seems to work better than the loras. There’s a quantised .GGUF version of these so I would highly recommend them over the latter.
Where did you get the GGUF version? Do you know if you can mix a standard WAN 2.2. GGUF on the HIGH and the Lightx2v distill GGUF on the LOW?
You can mix and match anything you want anyway you want but that doesn’t mean it will work better or work at all. As for the model vs lora, since they’re releasing new versions, it’s better to use the latter, because each new release will need someone to quantised it and then upload it which will take time and effort.
Privac changed discussion status to closed