calcuis

/

ace-gguf

like 26

cannot run it with the native setting or diffusers right away; seems need to install their own library; then either very difficult to test or a lot of work to do for just one model

found a custom node supports it; but need dia's library to run; and seems ckpt input only; you could test it with the fp8 (use tensor cutter to make it) file and see does it work or not; or you could convert it to gguf with convertor zero to test it first; since even we quant it now we can't test the file work or not, probably will need to redo later; better wait for a better support then quant it with proper test

understand, thanks for the explanation!

done finally; test it out with gguf-connector; working now

wow thanks man will use!

can you add share=True in launch(). to the ggc ggfu connector? i I looked for a way to do this and apparently it is impossible, the goal is to run in colab notebook

we rebuilt the engine here to run this; you could edit the app.py from their github repo and set share=True since the connector is for local test; if you set share=True your machine will need another cert to be issued for working as server; fyr app3 edit the last line and make it demo.launch(share=True) then should be fine

I got it thanks for the support, I had to edit the app.py in diao and the s2 in the ggfu connector but worked