Hello guys, i know kinda sus question, but if i train a model on google colab, how do i make a ckpt
Hello guys, i know kinda sus question, but if i train a model on google colab, how do i make a ckpt file from it, to work with it on my own pc?
Susing stable diffusion on pc tho
SI am driving myself nuts. I have what i think are good images to train with good captions and my embeddings all look like trash. What is the best way to be able to put myself into other models?
Dreambooth model which you then merge with another model?
Train an embedding?
Train a Lora (whatever that is!)
Other
Dreambooth model which you then merge with another model? Train an embedding? Train a Lora (whatever that is!) OtherSI just trained a dreambooth model, it looks very good
Sshame i can't get a ckpt file out from it and save it instead of the json stuff XD
Fhello welcome. what is your gpu
Fi think try lower first
Fthere is generate ckpt option. you can watch my updated tutorial here : and yes it would work on your pc : https://youtu.be/kIyqAdd_i10
YouTubeSECourses - Software Engineering Courses
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...
F1 and 2 are great options
SWhen auto captioning images for training, what do you guys find works best deepbooru or BLIP. I can see they have very different approaches one is a full description in a sentence, the other is essentially comma separated tags
SI am joining the patreon now!
Fawesome ty
Ffor teaching a person or style i didnt find them better. but if you want to fine tune i think you can use blip
Sis there a way to get the same already edited doc?
Fif you have saved it in your google drive, yes you can genrate ckpt
Fbut make sure that your google drive has sufficient space
Si mean i always see things that i dont have, and idk if i miss something
Sthe xformers install for example
Si mean i can search it up on google, but idk what im missing if bcs i dont know what should be there 
Sah okay, i did it
SI am now a patron! I love it here.
SOK so I have my training images in a google drive, can anyone tell me if they stink or not? If so, DM me!
Zif training a model, is it better to generate class images using my captions, or a generic prompt?
Fthank you so much
Fi can check them if you wish
Fsend me a dm
Fyou need to generate classification images based on your class token
Fand you need to use prompt in classification prompt instance same as the one you used to generate those images
Zthank you very much sir
Zyou are very kind
Zso class token and class prompt should be the same then
Fyes
Zok cool, as far as generating images, would photo of woman or photo of [filewords] work better for training a face
Fdont use filewords for teacing a face
Fmake your training set better
Z10-4
Zyou are the best sir
Zone last question, does this look correct?
Zi'll change beaitufl to just photo of a woman
Zinstance token - ohwx
class token - woman
instance prompt - photo of ohwx woman
class prompt - photo of woman
class token - woman
instance prompt - photo of ohwx woman
class prompt - photo of woman
Fno
Fnot correct
Fwhy dont you follow my video settings :d
Zwhich video?
Falmost all of the recent ones
Fi have shown settings
Zi see where i made my mistake, the new pictures looking awesome
Zi think 10 images is not enough for lora, i did 40 pictures now and my set is looking amazing
Z
Zhow do i do a x/y/z of different loras
Zsample prompt was perfect, when i went to use the model , it was a completley differnt person