photo of ohwx woman <lora:MarietteV8_1329:1>
photo of ohwx woman <lora:MarietteV8_1329:1>
Zmine i went to additional networks and clicked on it
Fyep definitely multi control net
Fi think you need to see it is used
Fhow i made text and logo https://youtu.be/C_mJI4U23nQ
YouTubeSECourses - Software Engineering Courses
Discord : https://bit.ly/SECoursesDiscord. How to use #ControlNet to generate cinematic text arts and logos for free. 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 #StableDiffusion Tutorials, #Automatic1111 and Google ...
https://www.patreon.com/SECoursesPlaylist of #StableDiffusion Tutorials, #Automatic1111 and Google ...
Zsample and instance prompt are the same, so if i see my results that i want in sample image, it should be the sme for instance prompt right?
Ni thought so- did not yet had much time for controlnet- i certainly watched that video already 
Zi think if i remove the tokens it will work
Fi think lora not activated. what do you see on cmd?
Zi have no idea cause constantly my yakuake crashes and have to restart
Zrestarting th eprogram now
Zhow do i activate it? is that something I can do manually?
Fi havent tested newest lora yet. you can generate ckpt always
Nsorry for asking, i am running the dreambooth version where i can select from "8bit AdamW", "Torch AdamW" & "Lion" - you did some excessiv testing. what was the best option?
FI prefer lion
Nthank you!
Nuuuh- 12gb aree not enough
F@Furkan Gözükara SECourses you should explore this!
FLoRA: The classic. You can use it in your webui no problem.
LoCon: Has more learning layers, it is reportedly good at artstyles.
LoHa: Has more layers and new mathematical algorithms. Takes much longer but can learn complex things such as styles and characters at the same time.
LoCon: Has more learning layers, it is reportedly good at artstyles.
LoHa: Has more layers and new mathematical algorithms. Takes much longer but can learn complex things such as styles and characters at the same time.Fmaybe it helps for your animation video, but also for another comparision video
F12 gb is good for dreambooth
Fyes web ui supporting that
Factually they gonna make even further improvement
Fi am waiting that for video
Fcurrently working on animation video with dreambooth
Fhaving consistency + strong style is the hardest thing to achieve
Nyeah, i know- we share the same grafikcard
but using lion seemed to be to much for the card. i did not use xformers as well.
but using lion seemed to be to much for the card. i did not use xformers as well.
Fwell we have to use xformers
Fno other option
Fbatch processing frames taking time
Nis there a missunderstanding?
Fyes you have to use xformers for 12gb vram
Nright now i do not use it- with the 3060..
Fwow it works?
Fthey must have improved memory usage
For you are using torch 2
Frecently i have been testing on runpod :d
Npython: 3.10.6 • torch: 2.0.0+cu118 • xformers: 0.0.17+b6be33a.d20230315 • gradio: 3.16.2 • commit: a9fed7c3 • checkpoint: a2ac9a57f3
i will compare the results to the last training once it is finished. same input images and class images
i will compare the results to the last training once it is finished. same input images and class images
Ftorch 2 has default xformers
Fi think that is the key
Nwhat is the best way while finding the best checkpoint using x/y-plot (checkpoint on x-axis, steps on y-axis) to use 4 completely different prompts for each example?
Fbest way is find a good prompt
Fget 8 decent looking seed
Fthen use x/y/z plot seed + checkpoints
Nit would be even better to have 4 different prompts to verify the results in different respects
Fyou can also do that
Fyou can add prompt with s/r option
Finto the x/y/z
Nbut not exactly what i ment- i really like (and do it the same way) your way to have a prompt with a batch size of 4 and only one prompt- the advantage is to do 4 at a time, but the disadvantage is that they are very similar- and i thought that i can define single parts of the prompt different with s/r but not the wholw prompt
Nsks as a superhero by zack snyder
sks as a cute pixar character
sks as a painting of van gogh on a 3-legged-horse
srs as a pencil drawing
sks as a cute pixar character
sks as a painting of van gogh on a 3-legged-horse
srs as a pencil drawing
Nsth like that..