TheLastBen
AUTOMATTIC WebUI on colab / huggingface stops working after 1 or 2 renders
After a few generations, I find that the WebUI needs a reload. Which means losing settings etc. After a while it happens every time.
Sometimes it shows progress but not result, other times it fails completely and doesn't send prompt for rendering.
I fixed that issue in the last commit, are you using the latest colab from the repo ?
played around with it a bit more this AM. (I presume it's the latest version as I'm just pasting the repo URL into colab.)
I don't need to reload as much, I think something is working better. But, as soon as I get to set batch size above 2, no images appear after the progress indicator gets to 100%.
this is using Gradio on a public URL, BTW.
maybe you should remove the folder "sd" from your gdrive and a clean run, because the last update of AUTOMATIC1111 restructured the paths.
first pic : batch count : 16 ; 2nd pic : batch size 4 + batch count 4
OK, deleted SD, but the issue persists: more than 2 it fails after first go.
Happy to debug, just wouldn't know where to look.
in the file sd/stable-diffusion-webui/webui.py look for demo.queue(concurrency_count=111500)
adding that line resolved the problem (at least for me), especially when I use the tunnel (not the gradio server)
I've tried with and without that line, same result.
The images are being generated. How can I debug what's going on between the colab instance and the webUI?
you can only monitor your network if it's downloading right after it finishes generating
also try the hlky or the relaxed-mode colabs and see if the issue exists there too.