search

TheLastBen / fast-stable-diffusion

fast-stable-diffusion + DreamBooth
MIT License
7.52k stars 1.31k forks source link

suggestion: pick which version of automatic111 to use? #960

Open createperhaps opened 1 year ago

createperhaps commented 1 year ago

Hi Ben, so it looks like the latest version of automatic111 is broken for me, kind of. I can do one generation and then I need to refresh the page. How hard would it be to implement a feature to install an older version of automatic? (or, can I do this myself?)

TheLastBen commented 1 year ago

Hi, It's not A1111, it's the server, sometimes, when the server lags, gradio freezes, I'm monitoring the issue, if it persists, I'll try to find a workaround.

createperhaps commented 1 year ago

Are you certain? I saw there was a big update to auto today. Does the server belong to you?

TheLastBen commented 1 year ago

this public server : https://github.com/localtunnel/localtunnel

createperhaps commented 1 year ago

Odd, posted a bug to auto (thinking it was an auto bug) and they said it is gradio server. I guess nothing I can do but hold tight. Thank you for making a great collab. They suggested something called ngrok?

TheLastBen commented 1 year ago

ngrok is a bit complicated and requires authentication, if the issue persists, I might add ngrok

SorenTruelsen commented 1 year ago

I also think this would be a great option, especially since it now auto-update. I had just learned how to "Dreambooth" faces about a month ago. With all the optimizations, I now is having a very hard time re-creating the success from back then. Leaves me a bit afraid to move to "SD-2", since I doubt I will be able to put myself in it.

Then again, I guess can try to make a Gdrive copy of an old Colab/Github-submit and then just pray the external code hasn't change too much... and that I can remove the "update feature" with my poor python skills.

And yes, I love the work you guys do, it's been amassing following the progress and ease of use. Wish I could actually contribute with something other than $$, but my coding skills is not really at this level, especially in Phyton.

TheLastBen commented 1 year ago

I'm constantly trying to make Dreambooth more efficient in Stable Diffusion since the Dreambooth paper itself wasn't meant for Stable Diffusion, was meant for "Google Imagen". Keep ding trials with the new implementations, the feedback helps improving the efficiency of Dreambooth in SD.