- Software Engineering Courses (SECourses)

SEC(

Software Engineering Courses (SECourses)•3y ago

Furkan Gözükara SECourses

Furkan Gözükara SECoursesOP•1/30/23, 9:14 PM

textual inversion

Furkan Gözükara SECoursesOP•1/30/23, 9:14 PM

512 pixel

Furkan Gözükara SECoursesOP•1/30/23, 9:15 PM

very good

IVG•1/30/23, 9:20 PM

@Furkan Gözükara SECourses when you try to do the SD setup on Runpod I'd love to see where you take it...

IVG•1/30/23, 9:20 PM

since runpod is my main setup i'd be very interested to see how I can improve it...

IIVG since runpod is my main setup i'd be very interested to see how I can improve it...

Furkan Gözükara SECoursesOP•1/30/23, 9:21 PM

on runpod i think biggest issue is correct xformers from what i seen. but when i get sufficient donation i will hire myself a runpod to prepare a video hopefully

IVG•1/30/23, 9:22 PM

what xformer version do you use in your setup?

IVG•1/30/23, 9:26 PM

did anyone try doing a dreambooth training with protogen as the basis? and if so what settings?>

IIVG did anyone try doing a dreambooth training with protogen as the basis? and if so...

Haashim•1/30/23, 9:26 PM

I don’t think that’s a good base model for training.
Maybe try Dreamlike Photoreal.

alenknight•1/30/23, 9:27 PM

oof.... need some advice on TI embeds. i went through the tutorial on SECourses... even using the sample images y'all had... and i got GREAT results....

alenknight•1/30/23, 9:27 PM

but when i try with my own source images .... horrible. i'm not sure what i'm doing wrong

alenknight•1/30/23, 9:27 PM

i'm trying to train an embed on images for my wife

alenknight•1/30/23, 9:27 PM

but the results are coming out very poor

HHaashim I don’t think that’s a good base model for training. Maybe try Dreamlike Photo...

IVG•1/30/23, 9:28 PM

and what kind of settings would you recommend for training?

FFurkan Gözükara SECourses Click to see attachment

Haashim•1/30/23, 9:29 PM

Interesting. My 2.1 TI training was subpar

alenknight•1/30/23, 9:29 PM

but the results come out like this.

alenknight•1/30/23, 9:30 PM

Original message was deleted

Colored Tonalities•1/30/23, 9:30 PM

remove bottom right

alenknight•1/30/23, 9:30 PM

the full body image ?

alenknight•1/30/23, 9:30 PM

the others ok?

Sandstormleader•1/30/23, 9:30 PM

have you tried fixing it with negative prompts as well?

Aalenknight the full body image ?

Colored Tonalities•1/30/23, 9:30 PM

yes precisely

alenknight•1/30/23, 9:30 PM

i tried using the embed for "bad-prompt"

alenknight•1/30/23, 9:31 PM

the images come out more "appealing"... but still totally different person

Colored Tonalities•1/30/23, 9:31 PM

these could also be optimized by being replaced by something full frame

alenknight•1/30/23, 9:31 PM

what's driving me nuts is when i do same thing with the sample images from SECourses... spot on perfect

alenknight•1/30/23, 9:31 PM

you mean full face? like up close?

Colored Tonalities•1/30/23, 9:32 PM

alenknight•1/30/23, 9:32 PM

ah ok so all of them up close

Aalenknight what's driving me nuts is when i do same thing with the sample images from SECou...

Furkan Gözükara SECoursesOP•1/30/23, 9:33 PM

that is good

Furkan Gözükara SECoursesOP•1/30/23, 9:33 PM

that means your training dataset is not good enough

alenknight•1/30/23, 9:35 PM

how's this?

alenknight•1/30/23, 9:35 PM

alenknight•1/30/23, 9:35 PM

removed the bottom right from before and added more close ups

Aalenknight Click to see attachment

Furkan Gözükara SECoursesOP•1/30/23, 9:37 PM

i think should learn very good

alenknight•1/30/23, 9:37 PM

will try. i'm trying to keep it to 10 like your video

alenknight•1/30/23, 9:38 PM

and samples

alenknight•1/30/23, 9:38 PM

to match as close as possible and then i'll play with it tweaking as i go

IIVG what xformer version do you use in your setup?

Furkan Gözükara SECoursesOP•1/30/23, 9:38 PM

i am using this one i explained in tutorial : xformers: 0.0.14.dev

IVG•1/30/23, 9:39 PM

so if my training works (meaning I can complete it without errors) and I am on xformers: 0.0.15.dev403 - so would changing it to 14 improve training results?

alenknight•1/30/23, 9:41 PM

any issue with using xformers 0.0.16? i'm on whatever build comes through auto1111 when you do git pull