even FLUX with no Lora already does a good job imo
even FLUX with no Lora already does a good job imo
Fi am testing rigfht now
Fyep
Fbut this is for training your consistent style
Tsolved by doing pip install numpy==1.26.4 in the venv
Ti have the .txt captions now
Fi will change requirements :d
Fyou found from pip freeze?_
Tyes!
Fthat is why i include 
Fopen source community doesnt do that which sucks so bad
Tand beware, the web page still says V11
Fye it is fine
GPlease tag me when you release the 4 GPU video bro. Also, I am going to do 10,000 images with captions this time. How long will it take to train since you are familiar with the 4 A6000 GPU and iterations. I figure 10 EPOCH or less will do it. Saving every Epoch. This way I will know more or less how much it will cost me. Please do tell me the calculation based on the 10K images, thank you Doc. And have a GN its 4am here im dead.... haha
Fwill update v13 now :d
Fediting right now :d
F
Fwill announce as @ everyone tagging
TDid you try the "FP16" option on your joycaptioner? Does it work? Is it better?
Fi didnt try
For maybe tried dont remember :d
Tok will try
Tok, only the very last sentence was slightly changed, and it means exactly the same, so useless
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
Fnew update
Nyeah that got me too, you are ontop of this stuff haha
Ftesting 100 new prompts from gpt4
MHello, is there a website or list somewhere of fun flux prompt to use with lora ?
Fi shared my testing prompts and 340 prompts in zip file :d
Fnice prompt
Fjust need more generation to cherry pick
F
Fhere 100 new prompt still testing
Ftest_prompts.txt36.31KB
Fjust need a better seed
F
F
Fi made gpt4 to redefine clothings :d
F
Yin your training, did you consider using prodigy opt so that u dont have to worry about LR?
Fi didnt find it better in past
F
Fediting cloud tutorial
GDoc, what do you estimate 10K images will cost for 10 Epoch on 4 GPU A6000? I might do 100 Epoch and just stop it once i see amazing results as I will test each one.
MHey, for anyone trying to use OneTrainer on RunPod, have you found a workaround for the inability to install Python's tkinter module (or the python3.*-tk package)?
T@Dr. Furkan Gözükara Could you try training, ending then trying to resume from a state? Not sure it's working for flux atm
TReddit
Explore this post and more from the StableDiffusion community
Feasy to calculate
Fwhich config you plan
Fi suggest use massed compute for onetrainer
