Software Engineering Courses (SECourses)•16mo ago

Question for all! How long does it take to generate an image in Forge on your GPU? Resolut

Question for all!

How long does it take to generate an image in Forge on your GPU?

Resolution: 1920x1080
Model: flux1-schnell-bnb-nf4.safetensors
VAE: ae.safetensors
Steps: 20
Sampler/Scheduler: Euler/Simple

It feels like my 1 min. 8.9 sec is way to long!?

KallamamranOP•9/6/24, 3:15 PM

Python 3.10.6
Pytorch version: 2.1.2+cu121

KallamamranOP•9/6/24, 3:18 PM

I seem to be running an old version. I wonder if newer ones are faster?

KKallamamran **Question for all!** __How long does it take to generate an image in Forge on ...

Furkan Gözükara SECourses•9/6/24, 3:18 PM

you are missing most important variable

Furkan Gözükara SECourses•9/6/24, 3:20 PM

i am finally testing accurate microsfot c++ build tools version

Dazzastrous [4090]•9/6/24, 3:20 PM

please?

Dazzastrous [4090]•9/6/24, 3:20 PM

lol

FFurkan Gözükara SECourses you are missing most important variable

KallamamranOP•9/6/24, 3:20 PM

Which one?

Dazzastrous [4090]•9/6/24, 3:21 PM

oh Doc the irony u putting accurate and then Microsfot

Dazzastrous [4090]•9/6/24, 3:21 PM

hehhe

KKallamamran Which one?

Furkan Gözükara SECourses•9/6/24, 3:21 PM

your gpu model :d

KallamamranOP•9/6/24, 3:21 PM

Testing "Forge with CUDA 12.1 + Pytorch 2.3.1" and "Forge with CUDA 12.4 + Pytorch 2.4" to see what the difference are

DDazzastrous [4090]oh Doc the irony u putting accurate and then Microsfot

Furkan Gözükara SECourses•9/6/24, 3:21 PM

ye i hate this c++ thingie

Furkan Gözükara SECourses•9/6/24, 3:21 PM

it is time to finalize it :d

Dazzastrous [4090]•9/6/24, 3:21 PM

yeh me too

Furkan Gözükara SECourses•9/6/24, 3:22 PM

i will make a video for this

FFurkan Gözükara SECourses your gpu model :d

KallamamranOP•9/6/24, 3:22 PM

Damn

Dazzastrous [4090]•9/6/24, 3:35 PM

30.2 secs

KallamamranOP•9/6/24, 3:35 PM

Yeah well... 4090 sounds right

Furkan Gözükara SECourses•9/6/24, 3:44 PM

faster internet so useful

KKallamamran Testing "Forge with CUDA 12.1 + Pytorch 2.3.1" and "Forge with CUDA 12.4 + Pytor...

Furkan Gözükara SECourses•9/6/24, 3:44 PM

swarmui has --fast

Furkan Gözükara SECourses•9/6/24, 3:44 PM

for rtx 4090

Furkan Gözükara SECourses•9/6/24, 3:45 PM

reduces flux to like 10-12 second

Furkan Gözükara SECourses•9/6/24, 3:45 PM

20 steps

Dazzastrous [4090]•9/6/24, 3:46 PM

Dazzastrous [4090]•9/6/24, 3:46 PM

oink oink

Dazzastrous [4090]•9/6/24, 3:47 PM

flux is a bit heavy on the boka , SD more detailed backgrounds

Dazzastrous [4090]•9/6/24, 3:48 PM

more detail with sd

Dazzastrous [4090]•9/6/24, 3:49 PM

Dazzastrous [4090]•9/6/24, 3:50 PM

like the colours in this

Furkan Gözükara SECourses•9/6/24, 4:12 PM

haha i made a thread on all nvidia githubs

Furkan Gözükara SECourses•9/6/24, 4:14 PM

https://github.com/NVIDIA/CUDALibrarySamples/issues/217
https://github.com/NVIDIA/cccl/issues/2383
https://github.com/NVIDIA/cutlass/issues/1793
https://github.com/NVIDIA/cuda-samples/issues/297
https://github.com/NVIDIA/TensorRT/issues/4114
https://github.com/NVIDIA/cuda-python/issues/88
https://github.com/NVIDIA/cudnn-frontend/issues/106

lol they have 500 repos

Furkan Gözükara SECourses•9/6/24, 4:43 PM

https://www.patreon.com/posts/111553210

Patreon

Solving CUDA and C++ Tools Compiling Errors Once and For All for Py...

Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon

Furkan Gözükara SECourses•9/6/24, 4:48 PM

i will update article testing still

Sharinel•9/6/24, 4:49 PM

The worst thing about running the kohya flux training update. "Haha you fewl, you installed torch 2.5. I will now downgrade you to 2.4 and complain about dependencies needing 2.5" And then the installer runs off like a 1920's villain and ties a damsel to train tracks

Sharinel•9/6/24, 4:50 PM

https://tenor.com/view/alice-terry-film-twenties-horror-the-magician-gif-17357640

Tenor

SSharinel The worst thing about running the kohya flux training update. "Haha you fewl, yo...

Furkan Gözükara SECourses•9/6/24, 4:51 PM

i overwrite gui file

to prevent it

FFurkan Gözükara SECourses use v30 zip file

xlads•9/6/24, 5:01 PM

will give this a try now

dsienra•9/6/24, 5:13 PM

@Dr. Furkan Gözükara One question, I'm trying to train flux with a big dataset of 1500 images, with the sugested learning rate is too slow to converge and get any results, I'm trying now with 4e-4 lr and I getting results much faster and until now looks good, do you think is too aggressive lr? may be I can start with this high lr and then finalize the training whith a lower lr to learn more finer details but without having to wait 15 days of training

Ddsienra @Dr. Furkan Gözükara One question, I'm trying to train flux with a big dataset ...

Furkan Gözükara SECourses•9/6/24, 5:40 PM

well the LR bigger certainly it will learn faster but it is like you are slicing the cake with a bigger knife so the minor detailed cuts becomes impossible

Furkan Gözükara SECourses•9/6/24, 5:40 PM

you can speed up training with multiple GPU

Furkan Gözükara SECourses•9/6/24, 5:45 PM

also bigger batch size requires bigger LR

dsienra•9/6/24, 6:19 PM

you right, but by now I doing everything on my local computer, do you think that finalize the training with a lower lr can improve the final results, startin with a high learning rate and when is aready on track I can reduce the lr, do you think that lowerimg the lr at the end can be beneficial?

Ddsienra you right, but by now I doing everything on my local computer, do you think tha...

Furkan Gözükara SECourses•9/6/24, 6:47 PM

Yes can be

xlads•9/6/24, 6:47 PM

just started training

Furkan Gözükara SECourses•9/6/24, 6:48 PM

Man I need 48 gb rtx 5090 comon Nvidia

Furkan Gözükara SECourses•9/6/24, 6:48 PM

I also need a mb that can perfectly run 3 gpu at the same time so I can test all

Furkan Gözükara SECourses•9/6/24, 6:48 PM

3060 3090 and 5090

xlads•9/6/24, 6:53 PM

The way you explain everything in this video so far is just amazing Dr i really missed learning from you

Sharinel•9/6/24, 8:10 PM

Just a quick FYI on flux training. Using installer v30 on an i9-13900 with 64gig ram and a 4090. Best Config Rank 4 quoted as 4.85 secs/IT, I'm getting 3.60/IT. So even quicker than expected

Sharinel•9/6/24, 8:11 PM

basically 1000 IT per hour

Question for all! How long does it take to generate an image in Forge on your GPU? Resolut

Similar Threads

Similar Threads

Similar Threads

**Question for all!** __How long does it take to generate an image in Forge on your GPU?__ Resolut

Similar Threads

Similar Threads

Similar Threads

Question for all! How long does it take to generate an image in Forge on your GPU? Resolut