Software Engineering Courses (SECourses)•15mo ago

I'm trying to train a lora/dreambooth to make simple icon type images of a particular style, and hav

I'm trying to train a lora/dreambooth to make simple icon type images of a particular style, and having watched the tutorials and looked at the sample outputs, I'm not sure if this project is better suited to dreambooth or lora training. The attached image is an example of one of the training images.

dawnslayerz•10/27/24, 7:54 PM

@Furkan Gözükara SECourses what template do i choose

SSaxon I'm trying to train a lora/dreambooth to make simple icon type images of a parti...

Furkan Gözükara SECourses•10/27/24, 9:45 PM

fine tuning better for everything

Furkan Gözükara SECourses•10/27/24, 9:45 PM

you can do 100%

Ddawnslayerz <@205854764540362752> what template do i choose

Furkan Gözükara SECourses•10/27/24, 9:45 PM

for which one?

FFurkan Gözükara SECourses you can do 100%

SaxonOP•10/27/24, 9:47 PM

I keep seeing people saying that it's best to do no captioning when training a style, but it seems odd to me -- would I not want to give it at least a minimal caption, for example "[triggerword], cow, bat wings"

Dazzastrous [4090]•10/27/24, 9:48 PM

"with pig snout"

SSaxon I keep seeing people saying that it's best to do no captioning when training a s...

Furkan Gözükara SECourses•10/27/24, 9:49 PM

flux has internal captioning mechanism

Dazzastrous [4090]•10/27/24, 9:49 PM

get to bed doc lol

DDazzastrous [4090]get to bed doc lol

Furkan Gözükara SECourses•10/27/24, 9:49 PM

still got some time :d

Furkan Gözükara SECourses•10/27/24, 9:49 PM

i need to start few trainings

Dazzastrous [4090]•10/27/24, 9:50 PM

do u have blue light protection in your lenses ?

SaxonOP•10/27/24, 9:53 PM

The only way I've gotten it to work close to the way I want it to is using the flux gym on default settings, but I would prefer to use Kohya -- based on the preferred results from flux gym, should I increase the number of repeats (from 1) and decrease the number of epochs (from 200)

DDazzastrous [4090]do u have blue light protection in your lenses ?

Furkan Gözükara SECourses•10/27/24, 9:54 PM

i am not sure how do we check

SSaxon The only way I've gotten it to work close to the way I want it to is using the f...

Furkan Gözükara SECourses•10/27/24, 9:55 PM

i dont think it would make diff

Dazzastrous [4090]•10/27/24, 9:55 PM

they asked me at the opticians

Furkan Gözükara SECourses•10/27/24, 9:55 PM

since you dont use reg images

DDazzastrous [4090]they asked me at the opticians

Furkan Gözükara SECourses•10/27/24, 9:55 PM

i really dont remember :d

Furkan Gözükara SECourses•10/27/24, 9:55 PM

but should have

SaxonOP•10/27/24, 9:55 PM

you can get a little blue laser to check -- they send them automatically with zenni glasses

SaxonOP•10/27/24, 9:55 PM

but it comes pretty standard in quality glasses

SSaxon but it comes pretty standard in quality glasses

Furkan Gözükara SECourses•10/27/24, 9:56 PM

i assume it should have but i will ask

SaxonOP•10/27/24, 9:56 PM

Do you know what those settings from flux gym would correspond to in Kohya?

SSaxon Do you know what those settings from flux gym would correspond to in Kohya?

Furkan Gözükara SECourses•10/27/24, 9:57 PM

sadly not i never used

Furkan Gözükara SECourses•10/27/24, 9:57 PM

kohya gui is easy i think

Dazzastrous [4090]•10/27/24, 9:57 PM

also can put eye drops in as eyes can get dry

Furkan Gözükara SECourses•10/27/24, 9:57 PM

there is a client sent me literally over 600 images picking good subset :d

DDazzastrous [4090]also can put eye drops in as eyes can get dry

Furkan Gözükara SECourses•10/27/24, 9:57 PM

ye the doctor gave me

SaxonOP•10/27/24, 9:59 PM

this is the difference between the outputs of a lora made with flux gym (more pink one) vs the one made with your suggested settings in koyha -- do you have any suggestions for changes to settings that might help?

Dazzastrous [4090]•10/27/24, 9:59 PM

whats the spoiler thing

Dazzastrous [4090]•10/27/24, 9:59 PM

lol

SaxonOP•10/27/24, 10:01 PM

it just seems less obnoxious to make the picture viewing opt-in

Dazzastrous [4090]•10/27/24, 10:05 PM

white-background-X0UKDkmjSG-kQGWxmNy8Ig-WxRu69-vS5KZHYLmX7Qikg.png

SSaxon this is the difference between the outputs of a lora made with flux gym (more pi...

Furkan Gözükara SECourses•10/27/24, 10:19 PM

when training style dataset matters most

Furkan Gözükara SECourses•10/27/24, 10:19 PM

did you read my style training guide?

Furkan Gözükara SECourses•10/27/24, 10:21 PM

selected 189 images for the client

Furkan Gözükara SECourses•10/27/24, 10:21 PM

lets see

SaxonOP•10/27/24, 10:21 PM

I didn't go through it comprehensively, but I did look at it

SSaxon I didn't go through it comprehensively, but I did look at it

Furkan Gözükara SECourses•10/27/24, 10:24 PM

so dataset matters most

Furkan Gözükara SECourses•10/27/24, 10:24 PM

and consistency

Furkan Gözükara SECourses•10/27/24, 10:24 PM

if not consistent captions help

Vigilence•10/27/24, 11:00 PM

@Saxon How many image repeats did you use in kohya? Default is 40 in kohya, 10 in fluxgym. You can look into the flux gym folder and see the settings used for your lora.

SaxonOP•10/27/24, 11:00 PM

I used the default of 10 in flux gym, and 1 in Kohya

Vigilence•10/27/24, 11:03 PM

I think that is a major difference, not enough time to learn the style. I have noticed this myself. The major drawback about fine tuning for styles, I think, is that it needs more image repeats, but I haven’t done enough testing yet. Was the dataset 1:1?

SaxonOP•10/27/24, 11:06 PM

yes, the data set was the same. I'll give it another try with higher repeats in kohya and see if it's the difference.

SaxonOP•10/27/24, 11:06 PM

The default epochs in gym is 16 -- is that how many epochs I would put into Koyha to try and replicate results? It seems low

SSaxon The default epochs in gym is 16 -- is that how many epochs I would put into Koyh...

Furkan Gözükara SECourses•10/27/24, 11:08 PM

default number of epochs really meaningless

Furkan Gözükara SECourses•10/27/24, 11:08 PM

it depends on each dataset

Furkan Gözükara SECourses•10/27/24, 11:08 PM

how many images you have total?

SaxonOP•10/27/24, 11:10 PM

21, but the style is very consistent. I could add more but I'm just trying to get a feel for the settings

Vigilence•10/27/24, 11:10 PM

You really need to look at the training log in flux gym, all the settings are different. I believe it uses Adam8bit, and a slightly faster learning rate. I also think training is in fp8.

I'm trying to train a lora/dreambooth to make simple icon type images of a particular style, and hav

Similar Threads