one trainer loras introduced major issues to kohyas keys in comfy and forge
one trainer loras introduced major issues to kohyas keys in comfy and forge
Mbreaking things
M
Fthat guy gave buch of kids to have so much power on his repo
Myeah quite bad atm.
either way kohya is better than onetrainer at this moment for flux
either way kohya is better than onetrainer at this moment for flux
Mkohya will improve even more with in a few days
Fstarting second training
Mare you training the full lora or just attn?
Mi wouldn't bother too much with dora, too much compatibility issues at the moment
Ffull lora
Fmn this interface slowness killing me
Mlogit normal is just like sigmoid i think
Myeah it was buggy for me on massed compute
Fdora doesnt work on swarmui?
Mi think only in comfy
Mas sometimes images are pink
Fshould go xformers or sdp here not sure
Fhe still didnt upgrade to torch 2.4
Mi used sdp, but xformers should work
Myeah one of the guys from onetrainer had to leave the project due to work restrictions
Ffloat8 fails lol
Myeah its a bitsandbytes error as he says but i guess he implemented it wrong
Msince it works in kohya fine
Fdmn interface on linux sucks
Msave your self a lifetime of headaches and stick with kohya for now
Fdid you try ema with flux?_
Myeah
Fye kohya config perfect atm
Mema improves
Mema with cpu
Fkk
Fbf16 uses 30.28 gb
Mthere is a major issue though
Mthe t5 text enocder nerogar/onetrainer is using is implemented wrong
Mhe messed up the token usage
Fye i saw he was trying to fix
Mkohya loras not working good for me anymore on forge or comfy
Fdmn this interface wasted my literally 45 minutes
Fmaybe more
F@NicB could the interface speed be improves somehow? onetrainer works 100x faster on windows, the interface
F
D@Dr. Furkan Gözükara I already have the training, but I have 99 @LORAS, is there a quick way to get the same prompt for all 99 @LORAS to see which one I get?
DHello dear @Dr. Furkan Gözükara
I hope you are very well, I already have my LORAS in huggingface, I need to select the best one and then look for an economic tool to raise it via API... which one do you recommend? it will make about 100-200 calls per day. thank you very much, I have learned a lot thanks to you!
I hope you are very well, I already have my LORAS in huggingface, I need to select the best one and then look for an economic tool to raise it via API... which one do you recommend? it will make about 100-200 calls per day. thank you very much, I have learned a lot thanks to you!
�You can use the grid generator tool doc explains it in the video
Bregarding details, which is better in your opinion SDXL Dreambooth (which is almost full finetuning) vs FLUX Kohya LoRA (which is NOT full finetune) for faces at the moment?
JI am getting below error on 4 GPU, Using "4x_GPU_Rank_1_SLOW_Better_Quality.json" with just changing image directory and class prompts under data set preparation
Note : (It was working in single GPU)
@Dr. Furkan Gözükara
Note : (It was working in single GPU)
@Dr. Furkan Gözükara
Dthanks
Sany way we can increase the flux inference speed , currently mine a100 40gb taking 15 seconds
Swith 25 steps
PIs there something wrong with my settings? Why is the image I created so blurry? I use 5. Download FLUX.1-dev FP16 and 8. Download T5 XXL text encoder FP16 version
PFlux has a very big problem with white backgrounds. Try not to use word Background at all. Try same prompt with red instead of white and it will not be blurry. Also lower guidance to 2 that will fix 50% of the blur. Also try diferent samplers