it will give huge boost
it will give huge boost
Flet me update
GI documented all my configurations when I was using Kohya_ss. Can these be ported directly into OneTrainer for LoRA model training?
Fvery likely
Fi did for sdxl
Ndoes this mean i missed updating something? thought text encoder was supposed to work now?
024-08-28 09:54:18 INFO text_encoder is not needed for training. deleting to save memory. train_network.py:1013
INFO unet dtype: torch.float8_e4m3fn, device: cuda:0 train_network.py:1030
024-08-28 09:54:18 INFO text_encoder is not needed for training. deleting to save memory. train_network.py:1013
INFO unet dtype: torch.float8_e4m3fn, device: cuda:0 train_network.py:1030
Fi didnt update text encoder configs yet
Fi am still researching
Fwill announce once made update
Fbut you can copy pate unet LR to text encoder
Fand it works 
Fare you using latest speed boost?=
Fwe need to verify it works :d
Ni am, but testing something at 512 so can't compare speed to my 1024 runs
Nrank 5 with updated torch (I guess updated, but I did update)...1024, without t5 att masking
Fi hope it works perfect
Fi will be able to test only during night
NI will not finish this training, so I can't tell you result. But maybe I will try to start training at night. So could be some results tomorrow...or not)
Fi will do same :d
MHello everyone, quick question: what does this script do?
https://huggingface.co/MonsterMMORPG/Generative-AI/resolve/main/kohua_gui_pass_update.sh
As far as I understand, it replaces kohya_ss's
https://huggingface.co/MonsterMMORPG/Generative-AI/resolve/main/kohua_gui_pass_update.sh
As far as I understand, it replaces kohya_ss's
gui.shwhen you use Furkan's kohya installer script (at least on Linux). What's the advantage? Thanks…
Dokay, lets see the local flux kohya training
NI think updates a few libraries to newer versions, then patches original gui to not roll them back? not sure though.
Ddo we still need to use the kohya training folder structure? even with the folder name #_name_class?
Nthat's what kohya is still expecting as of now (space between name and class though, not _ )
SDr Furkan i hope everyting is ok, wich is your estimate tha could be a video tutorial of FLUX Lora step by steps? wonderfull work
Fit prevents kohya to revert back requirements
Fso our custom torch installation remains
Fthank you. still editing video 
EHas anyone gotten a slider Lora training working for Flux yet? I have seen a few sliders on Civit but I don't think they are trained the same way SD sliders were trained.
Anyone seen anything? I tried modifying the ai-toolkit code to get it to work but it didn't work out ;P
Anyone seen anything? I tried modifying the ai-toolkit code to get it to work but it didn't work out ;P
Fnope i didnt see yet
Dwhen the trained person should be blonde but her last name is "Blue"
Nwhat is better to have more epochs or more repeats?
Fthey have different purposes
Fexplained in video :d
Fnormally you dont need repeat
NI'm in a middle of watching))
S
Ftotally depends on your total steps and gpu
Fbut rtx 3090 can get amazing model within 5 hours
Frtx 4090 likely to with like 2 hours 2.5 hours
D2-2.5 hours for the 4090 is correct
SNice I'll have to check it out. I have a 4090
Fyep
Fgreat
Fuse torch 2.5 to have best speed on windows
Si have like 60 more models to merge but i want to check it out this weekend for sure
S
Sfine tuning and adding flexibility to this 1.5 model
Swhile keeping the most accuracy