stable-diffusion icon indicating copy to clipboard operation
stable-diffusion copied to clipboard
verified publisher icon CompVis

How to train on latent size 128?

Open kashyappiyush1998 opened this issue 3 years ago • 1 comments

Currently, the VQVAE model provided is for 64 size max vq-f4, please let me know if I'm wrong. How do we train on size 128 latent dimensions? Do we need pre-trained VQVAE for that?

kashyappiyush1998 avatar Feb 12 '23 01:02 kashyappiyush1998

Sure, you need to train a new VQVAE model, and the image size is 512 if the f=4.

CreamyLong avatar Dec 19 '23 16:12 CreamyLong