can you explain this, why are the eyes so bad? did sdxl lora training, 30 train pics, all great qual

can you explain this, why are the eyes so bad? did sdxl lora training, 30 train pics, all great quality, many close ups on face, 150 epochs, repeat 1, 1024 resolution, biglovexl2 as base model.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/2/25, 7:41 PM

training on civitai

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/2/25, 7:47 PM

training-data, i see network rank 32. don't know if its possible to change that on training

train-data.txt1.41KB

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭can you explain this, why are the eyes so bad? did sdxl lora training, 30 train ...

Furkan Gözükara SECourses•5/3/25, 12:44 AM

are you using our config_?

Furkan Gözükara SECourses•5/3/25, 12:44 AM

which config name?

Furkan Gözükara SECourses•5/3/25, 12:44 AM

also some people reporting bad results

Furkan Gözükara SECourses•5/3/25, 12:44 AM

i am gonna train now to see

FFurkan Gözükara SECourses are you using our config_?

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/3/25, 6:30 AM

this was civitai training, I used what you said about ohwx woman token, 150 epochs, repeat 1.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/3/25, 6:33 AM

where can we download flux dev fast now? black labs has it gated, so can't use the download link in script...

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭this was civitai training, I used what you said about ohwx woman token, 150 epoc...

Furkan Gözükara SECourses•5/3/25, 12:05 PM

ah i dont know

Furkan Gözükara SECourses•5/3/25, 12:05 PM

i trained on my local pc yesterday

Furkan Gözükara SECourses•5/3/25, 12:05 PM

our config

Furkan Gözükara SECourses•5/3/25, 12:05 PM

latest kohya

Furkan Gözükara SECourses•5/3/25, 12:05 PM

perfect results

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/3/25, 12:34 PM

Oh awesome, you recommend kohya for sdxl lora and checkpoint training now?

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭Oh awesome, you recommend kohya for sdxl lora and checkpoint training now?

Furkan Gözükara SECourses•5/3/25, 12:37 PM

i tested flux

Furkan Gözükara SECourses•5/3/25, 12:37 PM

but we have sdxl dreambooth config too

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/3/25, 12:37 PM

Ok, so onetrainer gives slightly better results for sdxlin your opinion?

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭Ok, so onetrainer gives slightly better results for sdxlin your opinion?

Furkan Gözükara SECourses•5/3/25, 12:52 PM

yep

asap17z•5/3/25, 4:00 PM

Hey Doc, how you doing today, I was just wondering if it’s possible to create a stylised flux model in place of a stylised lora

FFurkan Gözükara SECourses thank you so much and welcome. replied you from email

lerois1•5/3/25, 4:11 PM

I am pleased to inform you that the dreambooth training has produced excellent results. It seems the previous difficulty arose from an incorrect learning rate.
I have encountered a remaining challenge: all generated images tend to resemble the model, irrespective of the prompt utilized. Although the model was trained using the prompt "ohwx" and the class "woman," this issue persists even when the prompt is omitted or when a different subject, such as "man," is specified.
Could you please offer some guidance on the potential cause of this occurrence? My objective is to generate images of other women/man in certain scenes who do not precisely replicate the trained model.

Aasap17z Hey Doc, how you doing today, I was just wondering if it’s possible to create a ...

Furkan Gözükara SECourses•5/3/25, 5:19 PM

surely possible

Llerois1 I am pleased to inform you that the dreambooth training has produced excellent r...

Furkan Gözükara SECourses•5/3/25, 5:19 PM

yes it will make everyone same

Furkan Gözükara SECourses•5/3/25, 5:19 PM

that is biggest issue of flux

Furkan Gözükara SECourses•5/3/25, 5:19 PM

it bleeds horribly

FFurkan Gözükara SECourses it bleeds horribly

1nfern0•5/3/25, 5:22 PM

Have you had a chance to try hidream yet? I’ve been testing loras with the same data set as flux and it’s so much better. Not sure how it will cope with multiple subjects concepts but for my purpose it’s miles ahead.

11nfern0 Have you had a chance to try hidream yet? I’ve been testing loras with the same...

Furkan Gözükara SECourses•5/3/25, 5:34 PM

i didnt train yet

Furkan Gözükara SECourses•5/3/25, 5:35 PM

planning to research it deeply

Furkan Gözükara SECourses•5/3/25, 5:35 PM

currently flux is just perfect except multi subject

FFurkan Gözükara SECourses currently flux is just perfect except multi subject

1nfern0•5/3/25, 5:57 PM

I think hidream is going to pull ahead once people start tuning it

11nfern0 I think hidream is going to pull ahead once people start tuning it

Furkan Gözükara SECourses•5/3/25, 6:27 PM

i am adding hidream to my downloader now :d

Furkan Gözükara SECourses•5/3/25, 6:27 PM

hopefully will make a tutorial today

Furkan Gözükara SECourses•5/3/25, 6:27 PM

for how to download use swarmui downloader

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/3/25, 9:25 PM

do I still need regularization images when training checkpoint/lora for a woman using sdxl checkpoint?

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭do I still need regularization images when training checkpoint/lora for a woman ...

Furkan Gözükara SECourses•5/3/25, 11:44 PM

yes

Jamiewhelton•5/4/25, 4:09 AM

Anyone else suddenly having issues fine tuning with kohya on Massed compute? I've done this like 50 times now the past 3 months and am suddenly getting errors

Jamiewhelton•5/4/25, 4:39 AM

anyone have any idea what I might be doing wrong?

Screenshot_2025-05-04_at_12.39.19_PM.png

Screenshot_2025-05-04_at_12.39.47_PM.png

Jamiewhelton•5/4/25, 6:36 AM

Looks like somethings up with the new repo updates to the models

FFurkan Gözükara SECourses yes

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/4/25, 2:20 PM

Thank you Dr. And grabbing the 1024px ones will be sufficient. No point going higher. And you already cropped to make the person occupy most of the photos yes? So I don't have to run your ultimate image process on them.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/4/25, 5:43 PM

doing my first sdxl fine tune with using regimages. i had 54 training images, so it looks like it's using 55 reg images per epoch, does this look right? don't wanna waste 25h

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/4/25, 5:43 PM

@Furkan Gözükara SECourses

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/4/25, 9:17 PM

Time changed to 6h. Nice

JJamiewhelton Anyone else suddenly having issues fine tuning with kohya on Massed compute? I'v...

Furkan Gözükara SECourses•5/4/25, 10:59 PM

just fixed today

�🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭doing my first sdxl fine tune with using regimages. i had 54 training images, so...

Furkan Gözükara SECourses•5/4/25, 10:59 PM

looks good

Furkan Gözükara SECourses•5/4/25, 10:59 PM

54 + 55 = 99

Furkan Gözükara SECourses•5/4/25, 11:00 PM

you do only 10 extra :d

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/5/25, 6:06 AM

Uh Dr. 54 + 55 = 109. Thanks

maybe I'll try masked training too.

�

🍭🎀 𝒜𝓋𝒶 𝐹𝓇𝒾𝑔𝑔 🎀🍭OP•5/5/25, 6:15 AM

Interesting, this person fine tunes a new woman by mixing 3 real women like described here. Is this possible with merging or?

PabloMoises•5/5/25, 9:46 AM

Hi!

I need help with a couple of things, let’s see if someone can assist me:

For an average of 200 images in a fine-tuning using Flux, how many epochs would you recommend?

To train on a multi-GPU VM, is there anything special that needs to be done? Do I need to modify anything or is nothing required?

Thank you very much!

PPabloMoises Hi! I need help with a couple of things, let’s see if someone can assist me: F...

Furkan Gözükara SECourses•5/5/25, 2:24 PM

up to 100 i recommend. but probably you will get best around 70 80

PPabloMoises Hi! I need help with a couple of things, let’s see if someone can assist me: F...

Furkan Gözükara SECourses•5/5/25, 2:24 PM

multi gpu works on 80 gb gpus for dreambooth or 48 gb for lora

can you explain this, why are the eyes so bad? did sdxl lora training, 30 train pics, all great qual

Similar Threads