1.27 it/s isnt that slow?
1.27 it/s isnt that slow?
FI have to sleep now hopefully see you tomorrow
JHave done, not captioning my data set seems to be the problem, will retry tomorrow
FKohya is slow
JTake it easy mate
Iso it makes sense for it to take 1.5 hrs to train a lora on a 3090?
FI find that sd web ui faster
FMy trainings were 45 min
FBut it depends on number of total steps
FAnd batch size
Iwow, TIs are much faster
FTi really fast
FI am afk now
CI just added you. Will give you more information about the project there
Iits much faster when i switched to fp16, about the 45 mins you said
I3.6 it/s instead of 1.24
Iand our speeds should be similar you have the TI version of my card
Iits also only using 30-50% of GPU isntead of 100% while going twice as fast, nice
Inow my room doesnt have to become a sauna to train a lora, lovely
Iresults arent good, even after 10k steps. i think its my input pics
Cis dreambooth the primary way that people train loras? or is there a different way to do it if you have a larger dataset?
Vhey guys, after the recent google colab limitations for stable diffusion webui, does anyone have a workaround for it?
FInput pics and the prompt matters a lot. If you do raw prompt you won't get good results
FLora training is based on dreambooth training. It is an alternative way
FYou can do larger training. For example if you don't use classification images and use file captions it becomes fine tuning
FYou can use no web ui scripts have you tried?
FI explained in this video
FYouTubeSECourses
Our Discord : https://discord.gg/HbqgGaZVmr. This is the video where you will learn how to use Google Colab for Stable Diffusion. If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 
https://www.patreon.com/SECourses
Playlist of Stable Diffusion Tutorials, #Automatic1111...
https://www.patreon.com/SECoursesPlaylist of Stable Diffusion Tutorials, #Automatic1111...
FAutomatic1111 returned so many updates getting merged
Aanyone knows why i get black image?
Cis there any way i could use stable diffusion on an amd gpu
Ceven with this i get e message that i require a nvidia gpu to open the ui https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs
GitHub
Stable Diffusion web UI. Contribute to AUTOMATIC1111/stable-diffusion-webui development by creating an account on GitHub.
Cdid all the steps, reinstalled python and added it to path but it still isnt working
Rwhat's the best caption generator? Dataset Tag editor on auto1111 is super buggy. And Koya's utilities too basic. Tried also Caption Buddy in StableTuner. Would rather not merely rely on notepad++, a lot of text files. Would love deepDanbooru or some model of the sort. anything out there not buggy?
VNo i haven't tried them, idk much about coding so could i understand them?
VOkay, i will it out, thanks!
V*check
ACan someone clarify this, are the terms 'reg images' and 'class images' synonyms?
Fprobably related to precision vae version
Ftry without any optimization and xformers
Fnmkd supports it
Cwhat exactly is that?
FYouTubeSECourses
Our Discord : https://discord.gg/HbqgGaZVmr. Newest AI model InstructPix2Pix is amazing to transform your images with plain English prompts.If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on https://www.patreon.com/SECourses
Playlist of Stable Diffusion Tutorials, #Aut...
https://www.patreon.com/SECoursesPlaylist of Stable Diffusion Tutorials, #Aut...
Fhere watch this tutorial @CnBer
Fanother GUI
Fi havent tested them :/ there can be also some extensions have you checked extension list
Fthey mean same
Ryes, both Smart Preprocessor and Dataset Tag Editor.
AThanks, that is what I was thinking but wanted to be 100% sure before I confused myself lol
REverything is buggy and trucculent. I am going to actually use aText snippet textexpander and use my own template. Seems like booru and all those generators are meant for Clip2 and weird W14 tags
Rannoyed