Software Engineering Courses (SECourses)•10mo ago

So in theory I can use your config of SDXL train multi characters using different caption for each c

So in theory I can use your config of SDXL train multi characters using different caption for each character to get the model understand multi concept right ?

Wwardensc2 So in theory I can use your config of SDXL train multi characters using differen...

Furkan Gözükara SECourses•3/23/25, 2:16 PM

yes

FFurkan Gözükara SECourses it may fit

404_error•3/23/25, 2:44 PM

I can use the fp8 flux model i install with your swarmui installer ?

4404_error I can use the fp8 flux model i install with your swarmui installer ?

Furkan Gözükara SECourses•3/23/25, 2:44 PM

yes it has fp8 flux

404_error•3/23/25, 2:44 PM

I know the difference quality between fp8 and fp16 is not huge when using flux but for a dreambooth traning is it the same ?

Furkan Gözükara SECourses•3/23/25, 2:44 PM

but dont use scaled i am not sure if tht works

Furkan Gözükara SECourses•3/23/25, 2:44 PM

by the way

Furkan Gözükara SECourses•3/23/25, 2:45 PM

latest fine tuning downloader

Furkan Gözükara SECourses•3/23/25, 2:45 PM

downloads fp8 model already

Furkan Gözükara SECourses•3/23/25, 2:45 PM

Furkan Gözükara SECourses•3/23/25, 2:45 PM

we cant train fp8 with dreambooth

Furkan Gözükara SECourses•3/23/25, 2:45 PM

but lora should work fairly fine

Furkan Gözükara SECourses•3/23/25, 2:45 PM

select base fp8 on interface

404_error•3/23/25, 2:46 PM

Ok i will try thanks

Joshua A•3/23/25, 2:57 PM

I am following your extracted LoRA tutorial. However getting an error. I have tried all save percisions and network rank sizes. I have checked V2 and unchecked it. Can seem to get past this error. Happening to 3 different finetunes. So I know it's got to be something I am not seeing.

new_checkpoint["time_embedding.linear_1.weight"] = unet_state_dict["time_embed.0.weight"]
KeyError: 'time_embed.0.weight'

Screenshot_2025-03-23_at_10.55.28_AM.png

JJoshua A I am following your extracted LoRA tutorial. However getting an error. I have tr...

Furkan Gözükara SECourses•3/23/25, 2:58 PM

use this

FFurkan Gözükara SECourses use this

Joshua A•3/23/25, 3:13 PM

Xcorpion•3/23/25, 6:59 PM

how to make a lokr like this? with that net dim, usually kohya only allow to put 100000, https://civitai.com/models/292553?modelVersionId=1568463 (this is the lokr and the creator).

XXcorpion how to make a lokr like this? with that net dim, usually kohya only allow to put...

EClipX•3/23/25, 9:44 PM

you can add it in as an extra argument

EClipX•3/23/25, 9:44 PM

or modify the gui

EEClipX you can add it in as an extra argument

Xcorpion•3/23/25, 10:10 PM

guidance please

EClipX•3/23/25, 10:13 PM

here i updated the file for you, replace it before starting the gui

you should be able to change the values of lokr now

the file to replace is in kohya_gui

lora_gui.py93.35KB

EEClipX here i updated the file for you, replace it before starting the gui you should ...

Xcorpion•3/23/25, 10:23 PM

I did replace, tho it's the same

EClipX•3/23/25, 10:29 PM

did you restart your gui? are you sure you replaced the right file?

EClipX•3/23/25, 10:30 PM

see in that file there is this

)
with gr.Row() as network_row:
network_dim = gr.Slider(
minimum=1,
maximum=9999999999,
label="Network Rank (Dimension)",
value=8,
step=1,
interactive=True,
)
network_alpha = gr.Slider(
minimum=0.00001,
maximum=9999999999,
label="Network Alpha",
value=1,
step=0.00001,
interactive=True,
info="alpha for LoRA weight scaling",
)
with gr.Row(visible=False) as convolution_row:
# locon= gr.Checkbox(label='Train a LoCon instead of a general LoRA (does not support v2 base models) (may not be able to some utilities now)', value=False)
conv_dim = gr.Slider(
minimum=0,
maximum=9999999999,
value=1,
step=1,
label="Convolution Rank (Dimension)",
)
conv_alpha = gr.Slider(
minimum=0,
maximum=9999999999,
value=1,
step=1,
label="Convolution Alpha",
)

EEClipX did you restart your gui? are you sure you replaced the right file?

Xcorpion•3/23/25, 10:52 PM

probably, "kohya_ss\kohya_gui\lora_gui.py" this one right?

Xcorpion•3/23/25, 10:55 PM

https://github.com/KohakuBlueleaf/LyCORIS. I tried to do this, but it gives me errors

Furkan Gözükara SECourses•3/23/25, 10:55 PM

i trained LyCORIS. it is lower quality but has better multi concept training

Furkan Gözükara SECourses•3/23/25, 10:55 PM

you can change our lora config LyCORIS however it may require some additional parameters which i dont remember

FFurkan Gözükara SECourses i trained LyCORIS. it is lower quality but has better multi concept training

Xcorpion•3/23/25, 10:58 PM

I see, Im seeing this 76mb LyCORIS everywhere and at least for me they perform well, like creators use few number of images and get decent quality with strong flexibility

XXcorpion https://github.com/KohakuBlueleaf/LyCORIS. I tried to do this, but it gives me e...

EClipX•3/23/25, 11:11 PM

did you git checkout sd3 flux 1?

EClipX•3/23/25, 11:11 PM

do a fresh install and replace it

Kaiden Quinart•3/24/25, 11:39 AM

Hello there I need your help here.I am training a character Lora with one trainer Is there Any math to find epochs based on dataset size? I have a dataset with 31 pictures of a character. How many epochs should I use?

Kaiden Quinart•3/24/25, 11:40 AM

I am keeping repeats value to 1
I am using onetrainer