GitLab

Update LoadClip to load clip models for SDXL refiner.

* support preview mode for mask editor.
* use original file reference instead of loaded frontend blob

bugfix:
* prevent file open dialog when save to load image

* bugfix: cannot clear previous mask painted image's alpha

* bugfix

* bugfix

---------
Co-authored-by: Lt.Dr.Data <lt.dr.data@gmail.com>

Tomesd now uses q instead of x to decide which tokens to merge because
it seems to give better results.

Large refactor of the model detection and loading code.

Information about running at RX7600

Change the transformer patches function format to be more future proof.

Remove mentions of python 3.10 since 3.11 works fine now.

Make the CLIP model work on the GPU.