I can't copy and paste a command from my mac to Ubuntu. I've tried everything.
I can't copy and paste a command from my mac to Ubuntu. I've tried everything.
Ni've done a flux lora with "ohwx man" and another with captions generated by JoyCaption. just an anecdote on a small sample size but wanted to share: the subject had glasses (wearing them in every one of the 20 training pics), and with the lora that had the JoyCaptions, fewer output images have him wearing glasses compared to the lora captioned with just "ohwx man". probably because of JoyCaption explicitly captioning that the man is wearing glasses? if i add "wearing glasses" to my prompt the glasses are there as expected, and they match the style from the training images. interested to hear thoughts and if anyone else experienced the same.
Nalso, i did separate loras of two different people, then prompted something like "photo of ohwx man and phwx woman smiling" - applied both loras and turned the strength of both down to 0.8. there was some bleeding of facial features but still produces images really close to the likenesses of the subjects.
Fyep i fixed it
Ffixed too
Fdownload v9
Fare you using from browser or windows exe?
Fbrowser doesnt have copy paste sadly
F@NicB do you know this?
Fi am at windows so dont know mac
Fflux is so powerful that it omits my eyeglasses even with just ohwx man
Fit knows each feature 
Fi trained high res need to test results now
Fso many new testing awaiting
Fnew running tests
FBest_v2_FP8_LowVRAM-000100,Best_v2_FP8-000100, Best_v2-000100, Best_v2_Rank4-000100, Best_v2_Rank8-000100, Best_v2_Rank16-000100, Best_v2_Rank32-000100,Best_v2_Rank64-000100, Best_v2_Rank256-000100, Best_v2_1280-000100, Best_v2_1280_256rank-000100, Best_v2_1536-000100, Best_v2_1536_256rank-000100, Best_v2_Split_QKV-000100
VHello @Dr. Furkan Gözükara, the images generated by flux dev model could be use for comercial purposes?
DNo, as far as a know are for personal and research use, Flux schnell can be used commercially
VThank you very much @dsienra for your response 
Dyou welcome
Fi dont think they would care or can detect
Flora rank 128 vs 4
VThank you @Dr. Furkan Gözükara
Dinteresting comparison the left is much better
DI'd be curious to see 32. thats what I used in the few experiments I did on civitai
Fyes lora lower rank reduced quality
F256 overfit though
F128 looks best
Fi have 32 too
Fpreparing grid
DI wander if rank 8 could be enough, must be sweet spot for such small dataset
F128 vs 32 here
Fmy dataset already small
Fi will share full grid on patreon
Dalso training the TE could make the diference, is not the same training the entire man class that is was doing right now than training a unique person, currently training a subject is overwriting the entire class weights
D32 is good enough probably not perfect but very good
Fyes TE should improve
Fyep 32 and above good
F256 overkills
Fall ranks vs
Fi am doing massive grids tests right now
F3 more on training
DI bet that with TE training enabled lower ranks will improve a lot, just a theory!! hehe
FI get this error message when I run training on Umbubu
Fyet I gave the right path for the images
�That has to with the folder training directory having to be the main directory above the folder of the images and not the image folder itself, if that makes sense
F,could be but not supported yet :d
Fyou are missing training dataset
�So, Home/Ubuntu/Desktop/ instead of Home/Ubuntu/Desktop/trainingimages/