Software Engineering Courses (SECourses)•15mo ago

which one would u recommend

zedOP•10/10/24, 10:59 PM

1024?

Zzed which one would u recommend

Furkan Gözükara SECourses•10/10/24, 10:59 PM

yep

Furkan Gözükara SECourses•10/10/24, 10:59 PM

1024 for sdxl

Furkan Gözükara SECourses•10/10/24, 10:59 PM

1024x1024

Furkan Gözükara SECourses•10/10/24, 11:01 PM

i got to sleep. recently i am trying to sleep early since i have eye pain, i believe related

zedOP•10/10/24, 11:02 PM

Yes have a good sleep doctor

zedOP•10/10/24, 11:02 PM

goodnight

zedOP•10/10/24, 11:02 PM

will update you tomorrow

zedOP•10/10/24, 11:02 PM

appreciate the help!

Furkan Gözükara SECourses•10/10/24, 11:03 PM

also if you dont necessarily need sdxl

Furkan Gözükara SECourses•10/10/24, 11:03 PM

i strongly suggest flux

Furkan Gözükara SECourses•10/10/24, 11:03 PM

it is next level

zedOP•10/10/24, 11:08 PM

I'm just trying my best to set it up myself, its really complicated and time consuming so I just wanna see if it works first

zedOP•10/10/24, 11:08 PM

will see about fine tuning later

zedOP•10/10/24, 11:08 PM

just following your stable diffusion guide

zedOP•10/11/24, 12:26 AM

Sent you a dm regarding some work, please respond when you can

FFurkan Gözükara SECourses i got to sleep. recently i am trying to sleep early since i have eye pain, i bel...

Dazzastrous [4090]•10/11/24, 12:27 AM

yeh you only get one pair of eyes doc, you need to rest them

FFurkan Gözükara SECourses face inpainting

Vexmachina•10/11/24, 3:04 AM

No way to do this through a replicate model is there, gonna need a custom pipeline?

Darth Teethius•10/11/24, 5:50 AM

@Dr. Furkan Gözükara i got the block swapping to work by installing some more ram but the 12gb config is running at 40s/it. Any thoughts?

DDarth Teethius @Dr. Furkan Gözükara i got the block swapping to work by installing some more ra...

Elwii•10/11/24, 7:44 AM

How much Ram do you have? Physical or virtual ram? It did not work for me

Elwii•10/11/24, 7:45 AM

Does anyone know the best way to train the model to merge two persons? I have datasets of two different persons and want to create a unique mix. What would be the best way?

Noel Camus•10/11/24, 1:07 PM

"I'm training on 4090D with 500 images. The speed is 4.4s/it (using Rank_3_18950MB_9_05_Second_IT). When I'm training on A800 (80 GB VRAM), the speed is 3.3s/it (using Rank_1_29500MB_8_85_Second_IT). I think the speed difference is abnormal. What could be causing this?"

VVexmachina No way to do this through a replicate model is there, gonna need a custom pipeli...

Furkan Gözükara SECourses•10/11/24, 1:16 PM

i think you can deploy any pipeline to replicate

DDarth Teethius @Dr. Furkan Gözükara i got the block swapping to work by installing some more ra...

Furkan Gözükara SECourses•10/11/24, 1:16 PM

totally normal i got same speed

Furkan Gözükara SECourses•10/11/24, 1:17 PM

i had to purchase 2x32 gb today, my 4 stick 16 gb keep causing problems

Furkan Gözükara SECourses•10/11/24, 1:17 PM

i wasted 3 hours today

EElwii How much Ram do you have? Physical or virtual ram? It did not work for me

Furkan Gözükara SECourses•10/11/24, 1:17 PM

min 48 gb phyiscal ram

NNoel Camus "I'm training on 4090D with 500 images. The speed is 4.4s/it (using Rank_3_18950...

Furkan Gözükara SECourses•10/11/24, 1:18 PM

rank 3 is slow

Furkan Gözükara SECourses•10/11/24, 1:18 PM

your speed totally good

Furkan Gözükara SECourses•10/11/24, 1:18 PM

i suggest fine tuning though way better results

OinkSauce•10/11/24, 1:20 PM

For training styles, how important is it to have the training images in 1024x. It's a bit more difficult to get those in uniform resolution vs training a subject.

OOinkSauce For training styles, how important is it to have the training images in 1024x. ...

Furkan Gözükara SECourses•10/11/24, 1:20 PM

well it is really hard to test

Furkan Gözükara SECourses•10/11/24, 1:20 PM

thus i suggest people to make 1 training with 1024x1024

Furkan Gözükara SECourses•10/11/24, 1:20 PM

and 1 training with bucketing

OinkSauce•10/11/24, 1:21 PM

to compare? or as a step 2 on the same model/lora?

OOinkSauce to compare? or as a step 2 on the same model/lora?

Furkan Gözükara SECourses•10/11/24, 1:21 PM

compare 2 different trainings

Furkan Gözükara SECourses•10/11/24, 1:22 PM

and see which one works better

Furkan Gözükara SECourses•10/11/24, 1:22 PM

so you need to train twicew

OinkSauce•10/11/24, 1:23 PM

I know you've provided scripts/tools to id faces and crop etc, is there something similar that will just expand the canvas of an image to make it square?

OOinkSauce I know you've provided scripts/tools to id faces and crop etc, is there somethin...

Furkan Gözükara SECourses•10/11/24, 1:28 PM

it does that

Furkan Gözükara SECourses•10/11/24, 1:28 PM

if image resolution is big

Furkan Gözükara SECourses•10/11/24, 1:28 PM

and if can do without cropping subject

Furkan Gözükara SECourses•10/11/24, 1:28 PM

so that is why crop first

Furkan Gözükara SECourses•10/11/24, 1:28 PM

resize second

OinkSauce•10/11/24, 1:29 PM