Software Engineering Courses (SECourses)•16mo ago

Hey everyone i have a few questions and will start with this one .. I there a way to speed up the g

Hey everyone i have a few questions and will start with this one ..
I there a way to speed up the generation of flux dev on Sd forge. I feel it is very slow with my RTX 3090 4-5 min sometimes until it starts generating. With flux1-dev-bnb-nf4.safetensors and as well the normal dev version as you can see in the image

PixelMonkeyOP•9/14/24, 7:18 PM

taining related:

I finished a training of myself with 36 images i used epoch 150 and got 10 loars. I used ohwx man and was wondeirng if i have to use ohwx everytime in the prompt to trighger the lora? It seems that it didnt train enough or smething since i start seeing only a bit of likeness starting at 90..

PixelMonkeyOP•9/14/24, 7:18 PM

What did i miss?

PPixelMonkey Hey everyone i have a few questions and will start with this one .. I there a w...

dxqb•9/14/24, 7:33 PM

"I am getting 26 secs on 30 steps, 19 secs on 20 steps, and 14 secs on Hyper 8 steps. That's including the unfortunate Lora handling by Swarm, as I'm using Hyper as a Lora"

dxqb•9/14/24, 7:33 PM

that's on a 4070, I think a 3090 is still fast even though older

dxqb•9/14/24, 7:33 PM

so yes, there should be a way to speed it up, but my numbers are on Swarm not forge

dxqb•9/14/24, 7:34 PM

I tried forge but it did weird things with loading and unloading the model, consuming a lot of system mem (not vmem). maybe you have the same issue and your system is swapping to disk?

PixelMonkeyOP•9/14/24, 7:38 PM

Yeah it does look like it.. Maybe Install swarm then to test. Thanks for the feedback :). Looking at the output ein it seems a issue with the memory loading and unloading. was there a installer from the Dr. or just main page

PPixelMonkey Hey everyone i have a few questions and will start with this one .. I there a w...

Furkan Gözükara SECourses•9/14/24, 10:12 PM

sounds very wrong

Furkan Gözükara SECourses•9/14/24, 10:12 PM

swarm is really fast try it

Furkan Gözükara SECourses•9/14/24, 10:12 PM

no issues

umuthasanoglu•9/14/24, 11:05 PM

I made a simple gradio app to create SVG icons and icon sheets from the generated images.

Uumuthasanoglu I made a simple gradio app to create SVG icons and icon sheets from the generate...

Furkan Gözükara SECourses•9/14/24, 11:07 PM

nice

Furkan Gözükara SECourses•9/14/24, 11:07 PM

i am testing flux fast training of civitAI to compare

Furkan Gözükara SECourses•9/14/24, 11:07 PM

FFurkan Gözükara SECourses nice

umuthasanoglu•9/14/24, 11:09 PM

Actually I tried to merge this features with flux so that it would create the images then convert it to vector with the specified features (color, stroke, style, etc) but I couldn't. I'm not that good at coding I think.

Uumuthasanoglu Actually I tried to merge this features with flux so that it would create the im...

Furkan Gözükara SECourses•9/14/24, 11:10 PM

try newest chatgpt

Furkan Gözükara SECourses•9/14/24, 11:10 PM

i use it on poe

umuthasanoglu•9/14/24, 11:13 PM

Will do

Furkan Gözükara SECourses•9/15/24, 12:49 AM

civit ai fast flux is really decent at distant shots

Furkan Gözükara SECourses•9/15/24, 12:50 AM

but at closer shots you see its quality not good

AIGambino•9/15/24, 12:51 AM

is kohya actually any different than onetrainer?

Furkan Gözükara SECourses•9/15/24, 12:58 AM

stable diffusion mods are out of control

Furkan Gözükara SECourses•9/15/24, 12:58 AM

Sebastianblood•9/15/24, 1:02 AM

@Dr. Furkan Gözükara I noticed all your generations have very clear backgrounds, how is that possible? Mine always come out with a shallow depth of field which is soft and blurry background

SSebastianblood @Dr. Furkan Gözükara I noticed all your generations have very clear backgrounds,...

Furkan Gözükara SECourses•9/15/24, 1:06 AM

i just replied back on youtube :d

Furkan Gözükara SECourses•9/15/24, 1:06 AM

the LoRA training making that impact. As you do train more epochs, that annoying "shallow depth of field which is soft and often blurry" effect is gone. But more training also causes overfitting so you have to have a very good config + decide which checkpoint to use for balance both

FFurkan Gözükara SECourses stable diffusion mods are out of control

Maxivi•9/15/24, 1:25 AM

didnt one of the mods say its fine earlier ??

MMaxivi didnt one of the mods say its fine earlier ??

Furkan Gözükara SECourses•9/15/24, 1:29 AM

there is 1 mod i thin he is responsible

Furkan Gözükara SECourses•9/15/24, 1:29 AM

so annoying

Maxivi•9/15/24, 1:29 AM

i think thats one of your biggests posts too

Maxivi•9/15/24, 1:30 AM

lot of comments seemed happy

FFurkan Gözükara SECourses the LoRA training making that impact. As you do train more epochs, that annoying...

Sebastianblood•9/15/24, 1:34 AM

Thank you

MMaxivi i think thats one of your biggests posts too

Furkan Gözükara SECourses•9/15/24, 1:36 AM

yep huge upvote

Furkan Gözükara SECourses•9/15/24, 1:36 AM

it was one of the most viral posts

FFurkan Gözükara SECourses Click to see attachment

Sebastianblood•9/15/24, 1:39 AM

Oh no, damn! This was the flux post with those many awesome images ?

SSebastianblood Oh no, damn! This was the flux post with those many awesome images ?

Furkan Gözükara SECourses•9/15/24, 1:40 AM

yep

Furkan Gözükara SECourses•9/15/24, 1:43 AM

time to do some fine tuning

FFurkan Gözükara SECourses time to do some fine tuning

Maxivi•9/15/24, 2:04 AM

just got word that sd3 8b will be out soon, no distilation and easier to train and finetune

MMaxivi just got word that sd3 8b will be out soon, no distilation and easier to train a...

Furkan Gözükara SECourses•9/15/24, 2:07 AM

licence?

Maxivi•9/15/24, 2:07 AM

i think same as the the updated sd3 medium one with more clarity

Maxivi•9/15/24, 2:07 AM

lykon mentioned

Maxivi•9/15/24, 2:07 AM

looks like it can be better than flux

Furkan Gözükara SECourses•9/15/24, 2:08 AM

nice

Maxivi•9/15/24, 2:08 AM

https://x.com/Lykon4072/status/1834917552209744318

Lykon (@Lykon4072) on X

@linaqruf_ @oron1208 Wait for 8b. It's basically Flux without distillation and heavy hands dpo. This should make it easy to finetune (and dpo).
We're also trying a new scaling down mechanism for mmdit, the new 2b is gonna work much better.