jzbor
commented
1 year ago Hi, These are both features I also wanted to implement initially, but the library wrapping around the machine learning stuff sadly does not expose download progress via its python interface (not even if it is going to download something) It also manages its own download directory. This is something I actually appreciate, because that way I can avoid reimplementing package management, which can be quite complicated to get right and is certainly out of scope for this project. It also means that other programs/scripts using the library also have access to it, which is another advantage to this approach.
However it should be possible to load custom models that have been manually downloaded. This is something I can look into, although it will probably still only be StableDiffusion pipelines. As most StableDiffusion datasets are found on Huggingface anyway it maybe an option to just choose custom Hugginface repos.
For now I would suggest opening it in a terminal (just to see the progress). I wonder whether I should change the desktop file in the AUR package to always open it with a terminal.
Regards, Julian!
Currently, on a first run, the software downloads huge amount of data in the background.
No status information about that is given in the graphical window.
Feature request No. 1:
Show download status in the graphical window, and allow for download pause and resume.
Download is carried out to a fixed locationn within
${HOME}/.cache/huggingface/.Feature request No. 2:
Allow to customise the download directory, and allow to use pre-downloaded dataaset (e.g. by a distribution package that provides the data system wide).
Regards!