oooh, the script works even on direct catbox links even when it's not used to make the image post
>>More Resources
>4chanX Catbox userscript: https://rentry.org/hdgcb
>>41824528Cheers! I knew it would!
warning morning coffee technobabble TL;DR below mucho texto
yea, that same mount bind command + a little trick to print the date in ISO + time in UTC is the same way I made the retard notebook do "your output will always automatically be on your Google Drive with no further effort required".
I/O access directly to Drive from within Colab is kind of slow, but it's nowhere near slow enough to make reading a set of training images, or outputting your regular images feel any different.
I actually tested it pretty extensively for mounting the ENTIRE stable-diffusion-webui folder to Google Drive, so that you could "download and set it up ONCE" if you had ~10 GB to spare.
I think since using colab means using launch.py directly instead of web-ui.sh, there's no venv created, so it still wants to install all the python stuff every single time. plus for some reason it still downloads/reinstalls codeformers and resergan and tamingtransformers and all that other stuff you see in the console normally when you launch webui the first time in a session, even though they're all still downloaded and I can see them taking up space in the folder. I could probably figure it out, but loading models directly from Google Drive is almost slower than just having it download them from huggingface, so what's the point? I actually timed launching the "entire mounted folder" and just having it download the same exact model + VAE and set webui up from scratch...the latter is actually considerably faster. besides, copying config.json/ui-config.json over before launching automatically loads all my settings. that was more than enough to convince me it likely wasn't worth the effort to figure out or mess with any more than that 