search

AUTOMATIC1111 / stable-diffusion-webui

Stable Diffusion web UI
GNU Affero General Public License v3.0
142.59k stars 26.9k forks source link

[Feature Request]: Multi-resolution Textual Inversion #5459

Open Ehplodor opened 1 year ago

Ehplodor commented 1 year ago

Is there an existing issue for this?

What would your feature do ?

Unsure if really better than current TI but, for the sake of completeness :

Be able to train "multi-resolution" textual inversion embeddings as in https://github.com/giannisdaras/multires_textual_inversion

Proposed workflow

  1. Go to ....
  2. Press ....
  3. ...

Additional information

No response

ProGamerGov commented 1 year ago

You may be able to accomplish this by resizing the images ahead of time, setting the batch size to 1, and then removing the resize operation here: https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/master/modules/textual_inversion/dataset.py#L62