forge-again — a Gradio 6 / Python 3.12 / torch 2.13 modernization of Forge #3150
logan71f100
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
First off, thanks to @lllyasviel and everyone who's contributed here — Forge's backend is what made this possible, and I've kept the credit explicit throughout the fork.
I'd been running Forge on a newer dependency set and it turned into a full port, so I've published it as forge-again in case it's useful to anyone else stuck on this older stack.
…Gradio 6.20 port, deferred loading so ui is much faster, made modes for flux, SDXL, and SD with their own defaults and only shows models compatible with that mode, I even added a AI assistant that runs local and helps tune results (has access to all sliders and prompts, and swaps in and out of vram so generation has all vram available)
The part that may be worth a look regardless of whether you use the fork: PORTING.md catalogs the Gradio 4→6 and transformers 4→5 breakages I hit, including a few that fail silently…
Caveats, up front: solo-maintained, developed/tested on Windows + NVIDIA. AGPL-3.0, same as upstream.
Not trying to pull anything away from here — just putting it somewhere findable.
https://github.com/logan71f100/forge-again
All reactions