Software Engineering Courses (SECourses)•17mo ago

Training a FLUX LoRA for a client. 24 images. on 4x RTX A6000 GPU. Total 240 epochs. Effective batch

Training a FLUX LoRA for a client. 24 images. on 4x RTX A6000 GPU. Total 240 epochs. Effective batch size 4 and step count is 1440

Furkan Gözükara SECoursesOP•8/31/24, 1:55 AM

i also found a new info regarding multi gpu training will update configs

Furkan Gözükara SECoursesOP•8/31/24, 2:05 AM

i am moving to the city tomorrow hopefully. time to get 1 gbit internet

Furkan Gözükara SECoursesOP•8/31/24, 2:07 AM

time to start 8 more test on flux lora LR

Furkan Gözükara SECoursesOP•8/31/24, 2:39 AM

testing more trainings of text encoder - clip L

FFurkan Gözükara SECourses Training a FLUX LoRA for a client. 24 images. on 4x RTX A6000 GPU. Total 240 epo...

GomezBro•8/31/24, 2:39 AM

How long did it take? Size1024x1024? I have the A6000. And what is that EFFECTIVE BATCH SIZE 4 Setting at?

GGomezBro How long did it take? Size1024x1024? I have the A6000. And what is that EFFE...

Furkan Gözükara SECoursesOP•8/31/24, 2:39 AM

it is almost linear increase

Furkan Gözükara SECoursesOP•8/31/24, 2:39 AM

with slow work flow

Furkan Gözükara SECoursesOP•8/31/24, 2:39 AM

9.41 second it

GomezBro•8/31/24, 2:40 AM

image res size? and the batch setting?

GomezBro•8/31/24, 2:40 AM

i did torch 2.5 update

Furkan Gözükara SECoursesOP•8/31/24, 2:40 AM

so lets say 3000 steps takes 9.41 * 3000 / 4 / 60 = 117 minutes

Furkan Gözükara SECoursesOP•8/31/24, 2:40 AM

it is 1024x1024

GomezBro•8/31/24, 2:40 AM

and i had to add the dll in system32 to fix my fbgemm.dll errors

GGomezBro and i had to add the dll in system32 to fix my fbgemm.dll errors

Furkan Gözükara SECoursesOP•8/31/24, 2:40 AM

ye it is a solution

Furkan Gözükara SECoursesOP•8/31/24, 2:41 AM

if you go with fast settings it takes 60 minutes for 3000 steps

Furkan Gözükara SECoursesOP•8/31/24, 2:41 AM

steps count becoming 750 since 4x gpu

Furkan Gözükara SECoursesOP•8/31/24, 2:41 AM

and effective batch size being 4

GomezBro•8/31/24, 2:41 AM

where do you set steps? that is controlled only by the amount of Images you have right?

GomezBro•8/31/24, 2:41 AM

I only have 1 GPU though

GGomezBro where do you set steps? that is controlled only by the amount of Images you hav...

Furkan Gözükara SECoursesOP•8/31/24, 2:42 AM

for 1 gpu you dont need to change anything

Furkan Gözükara SECoursesOP•8/31/24, 2:42 AM

this is for multi gpus

Furkan Gözükara SECoursesOP•8/31/24, 2:42 AM

explained in tutorial editing it

Furkan Gözükara SECoursesOP•8/31/24, 2:42 AM

you can rent 4x A6000 on massed compute and use

Furkan Gözükara SECoursesOP•8/31/24, 2:42 AM

1.25$ per hour

GomezBro•8/31/24, 2:42 AM

OK which BEST CONFIG do I use bro?

GGomezBro OK which BEST CONFIG do I use bro?

Furkan Gözükara SECoursesOP•8/31/24, 2:42 AM

for multi gpu or single?

GomezBro•8/31/24, 2:42 AM

single a6000

GomezBro•8/31/24, 2:42 AM

better colors folder or other one?

GGomezBro single a6000

Furkan Gözükara SECoursesOP•8/31/24, 2:42 AM

ok use Rank_1_28700MB_Slow.json

GomezBro•8/31/24, 2:43 AM

why slow?

Furkan Gözükara SECoursesOP•8/31/24, 2:43 AM

better colors reduces likeliness

Furkan Gözükara SECoursesOP•8/31/24, 2:43 AM

slow is best quality

GomezBro•8/31/24, 2:43 AM

OK so ill ignore that folder

Furkan Gözükara SECoursesOP•8/31/24, 2:43 AM

if you need speed Rank_2_27360MB_Fast.json and batch size 2

GomezBro•8/31/24, 2:43 AM

Where do you control batch size? where is that setting?

GGomezBro Where do you control batch size? where is that setting?

Furkan Gözükara SECoursesOP•8/31/24, 2:43 AM

on the interface do ctrl f

Furkan Gözükara SECoursesOP•8/31/24, 2:44 AM

let me show

Furkan Gözükara SECoursesOP•8/31/24, 2:44 AM

but normally batch size 1 is best quality

GomezBro•8/31/24, 2:44 AM

Slow will create a better likeness though right?

GomezBro•8/31/24, 2:44 AM

ic...

Furkan Gözükara SECoursesOP•8/31/24, 2:44 AM

GGomezBro Slow will create a better likeness though right?

Furkan Gözükara SECoursesOP•8/31/24, 2:44 AM

yep

GomezBro•8/31/24, 2:44 AM

RANK 1 28700 MB is the one I should use.... I see

Furkan Gözükara SECoursesOP•8/31/24, 2:45 AM

it is a tradeoff

GomezBro•8/31/24, 2:45 AM

22images I wonder how long that will take

GomezBro•8/31/24, 2:45 AM

I prefer quality

Furkan Gözükara SECoursesOP•8/31/24, 2:45 AM

22 images should take around you can calculate

Furkan Gözükara SECoursesOP•8/31/24, 2:45 AM

8.45 second / it

GomezBro•8/31/24, 2:45 AM

about to run it

Furkan Gözükara SECoursesOP•8/31/24, 2:45 AM

with slow settings

Training a FLUX LoRA for a client. 24 images. on 4x RTX A6000 GPU. Total 240 epochs. Effective batch

Similar Threads