Skip to content

added jxl support#337

Open
yggdrasil75 wants to merge 132 commits intojhc13:mainfrom
yggdrasil75:main
Open

added jxl support#337
yggdrasil75 wants to merge 132 commits intojhc13:mainfrom
yggdrasil75:main

Conversation

@yggdrasil75
Copy link
Contributor

This works on windows at least.
changes:
imagesize was removed as it was incompatible with jxl
use pillow to get image size
use pillow to load images
images load and are converted to qimage. (probably should only do this for jxl, its slower than letting qt load directly)
some minor changes that probably should be in a cleanup pr (I had a lot, but I reverted most - sorry about these)

jhc13 and others added 22 commits February 20, 2025 15:53
tl;dr: bitsanbytes is quantizing a nn.NonDynamicallyQuantizableLinear output in JoyCaption.  We revert it back to a nn.Linear (as nn.NonDynamicallyQuantizableLinear is not a public type.)

We also set the dtype in a few more places where we don't, or don't always.
…nitial load by not loading entire jpeg image to get the size.
StableLlama and others added 21 commits April 9, 2025 18:44
…ticaptioning can be used with the civitai trainer as well
Add support for 4bit JoyCaption
…orkflow

# Conflicts:
#	taggui/models/image_list_model.py
Add support to drag selected images to external applications
@StableLlama
Copy link

I have pulled it into this fork now: https://github.com/StableLlamaAI/taggui_flow
@yggdrasil75 you might want to use that fork now. When you have other PRs that are relevant for you, please create a PR for them against StableLlamaAI/taggui_flow so that I can check about pulling there

diodiogod pushed a commit to diodiogod/taggui-video that referenced this pull request Oct 17, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants