use gradio workflow

#1
by akhaliq HF Staff - opened
akhaliq changed pull request status to open
Viggle AI org

Thanks a lot @akhaliq โ€” seeing the pipeline laid out as prepare โ†’ render โ†’ finish nodes is a really nice way to show how it works!

Rather than switching this Space over (we'd like to keep the Blocks UI with its examples gallery here), we've published your workflow as its own Space, credited to you in the README and in the code: https://huggingface.co/spaces/Viggle/viggle-animate-workflow

One bug we found while wiring it up, in case it's useful for other workflow ports: the bound prepare never receives a gr.Request, so app.py's per-visitor daily repaint limit keys everyone to the same bucket, and the whole Space shares a single 5-per-day counter. Declaring request: gr.Request = None on the bound function and forwarding it fixes it (gradio injects it by type hint in both the canvas path and the API executor).

A few other small changes in the new Space:

  • a character image now takes priority over the preloaded painted frame (before, the preload silently won unless you cleared it);
  • the job stays server-side and only its id travels through the canvas;
  • the AOTI maintenance node and the extra OAuth scopes are dropped, since nothing there needs to write to a repo.
Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment