Software Engineering Courses (SECourses)•15mo ago

Doc, had a question regarding dataset preparation! What if one artificially manipulated the backgrou

Doc, had a question regarding dataset preparation! What if one artificially manipulated the background on the dataset to make it more varied (E.g., using photoshop or lightroom? Will the trainining pick up edge artifacts?

Ddruhl Doc, had a question regarding dataset preparation! What if one artificially mani...

Furkan Gözükara SECourses•10/14/24, 10:21 AM

yep very likely

Furkan Gözükara SECourses•10/14/24, 10:21 AM

ai picks those :d

druhlOP•10/14/24, 10:23 AM

Haha, thought so. I've been very careful since that one time it picked up a mosquito lol

druhlOP•10/14/24, 10:23 AM

Ddruhl Haha, thought so. I've been very careful since that one time it picked up a mosq...

Furkan Gözükara SECourses•10/14/24, 10:26 AM

yep

FFurkan Gözükara SECourses Click to see attachment

dxqb•10/14/24, 10:31 AM

no doubt finetune might be better overall still, but you can definitely make a lora that has expressions without expressions in training data

Ddxqb no doubt finetune might be better overall still, but you can definitely make a l...

Furkan Gözükara SECourses•10/14/24, 10:34 AM

yep

Furkan Gözükara SECourses•10/14/24, 10:34 AM

but wont be as good as fine tuning :d

deepblue_1337•10/14/24, 10:51 AM

If I want to add multiple people to a model (for instance me AND my wife) would I still use finetune or is lora better?

deepblue_1337•10/14/24, 10:52 AM

and how would I train it? Just add training images of both persons?

Ddeepblue_1337 If I want to add multiple people to a model (for instance me AND my wife) would ...

Furkan Gözükara SECourses•10/14/24, 10:56 AM

fine tune better

Furkan Gözükara SECourses•10/14/24, 10:56 AM

have both in same picture

Furkan Gözükara SECourses•10/14/24, 10:56 AM

learns perfect

Furkan Gözükara SECourses•10/14/24, 10:56 AM

of course you need to caption in that case

Furkan Gözükara SECourses•10/14/24, 10:56 AM

like photo of ohwx man and bbuk woman

deepblue_1337•10/14/24, 10:59 AM

Do both have to be in the same picture or can I have 10 pictures of ohwx man (captioned) and 10 pictures of bbuk woman (captioned)

deepblue_1337•10/14/24, 10:59 AM

deepblue_1337•10/14/24, 11:01 AM

and also can I use not-normalized pictures in that case or is it bad?

Ddeepblue_1337 Do both have to be in the same picture or can I have 10 pictures of ohwx man (ca...

Furkan Gözükara SECourses•10/14/24, 11:08 AM

i didnt test this so you need to test

Furkan Gözükara SECourses•10/14/24, 11:08 AM

but people traines successfully

Ddeepblue_1337 Do both have to be in the same picture or can I have 10 pictures of ohwx man (ca...

druhlOP•10/14/24, 11:22 AM

What you are asking is currently experimental and includes some extra steps. In the end, you may or may not get the results you're looking for. If you are ready to fail a few times (like I have been failing for the past 3 days trying to do something experimental), I can tell you how lol.

deepblue_1337•10/14/24, 11:23 AM

yes please

druhlOP•10/14/24, 11:27 AM

Alright then, lets fail together in the DMs

We come back once we are victorious lol.

druhlOP•10/14/24, 12:01 PM

@Dr. Furkan Gözükara how does merging loras work? Any resources for learning that?

Elwii•10/14/24, 12:10 PM

I am trying to do a full Dreambooth finetune on massed compute with 2x A6000. I am using the config provided on patreon for 48gb gpu. I also set necessary parameters in accelerate tab but it keeps crashing due to OOM.

With one A6000 it worked just fine. Also is bucketing currently supported for Flux Dreambooth?

Ddruhl @Dr. Furkan Gözükara how does merging loras work? Any resources for learning tha...

Furkan Gözükara SECourses•10/14/24, 12:11 PM

it is purely experimental

druhlOP•10/14/24, 12:11 PM

I see