Software Engineering Courses (SECourses)•2y ago

Hi, I have a problem since updating InstantID to version 12. I can no longer run it on Windows.

RRaphaël Hi, I have a problem since updating InstantID to version 12. I can no longer run...

Furkan Gözükara SECourses•5/24/24, 4:41 PM

we install bitsandbytes

Furkan Gözükara SECourses•5/24/24, 4:41 PM

our installer

Furkan Gözükara SECourses•5/24/24, 4:41 PM

what is your python?

Furkan Gözükara SECourses•5/24/24, 4:42 PM

ah we have v13

Furkan Gözükara SECourses•5/24/24, 4:42 PM

use it

Furkan Gözükara SECourses•5/24/24, 4:42 PM

@Raphaël https://www.patreon.com/posts/1-click-very-gui-97769887

Furkan Gözükara SECourses•5/24/24, 4:42 PM

do a fresh install

FFurkan Gözükara SECourses on MC dont install TensorRT

ArtChicken•5/24/24, 5:12 PM

you mean that's already installed... or we don't have to install because it's non working again?

mayna•5/24/24, 6:08 PM

Hey Every one, long time did not do a training lora of google collab, now i connected and a lot of things are outdated , anyone does have a legit link for IPYNB file please for kohya lora dreambooth ? thank you

Abdo Asker•5/24/24, 6:19 PM

I think that SUPIR AI isn't creative enough to add details as MAGNIFIC AI .. If someone knows how to use it on ai generated images to add much details for realism I would be thankful to him.

mayna•5/24/24, 7:11 PM

it seems that any Kohya lora IPYNB notebook gives this error :

https://colab.research.google.com/github/Linaqruf/kohya-trainer/blob/main/kohya-LoRA-dreambooth.ipynb

i have this error : ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
tf-keras 2.15.1 requires tensorflow<2.16,>=2.15, but you have tensorflow 2.16.1 which is incompatible.
torchaudio 2.3.0+cu121 requires torch==2.3.0, but you have torch 2.1.0 which is incompatible.
torchtext 0.18.0 requires torch>=2.3.0, but you have torch 2.1.0 which is incompatible.

Google Colab

FFurkan Gözükara SECourses do a fresh install

RaphaëlOP•5/24/24, 8:07 PM

Hello. Tkank's a lot. V13 work fine

AArtChicken you mean that's already installed... or we don't have to install because it's no...

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

we dont install since not working yet

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

other than tensorrt rest you can install

RRaphaël Hello. Tkank's a lot. V13 work fine 🙂 🙂 🙂

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

great

AAbdo Asker I think that SUPIR AI isn't creative enough to add details as MAGNIFIC AI .. If ...

Furkan Gözükara SECourses•5/24/24, 9:41 PM

true supir is loyal to original

Furkan Gözükara SECourses•5/24/24, 9:42 PM

you can use auto1111 to upscale

Furkan Gözükara SECourses•5/24/24, 9:54 PM

kaggle notebook updated : https://www.patreon.com/posts/88714330

EyeSpyBekoAI•5/25/24, 2:33 AM

Hello all. Do any of you have a preferred automatic1111 extension that reads image to text (like blip, or text interrogator), but is idal for stable diffusion to read? Meaning, if you had an inputted image described into text, and you entered the text back into stable diffusion, it'll actually generate something like what was described?

daroth•5/25/24, 4:52 AM

I don't know of any tagger that would make such accurate captions that would result in a similar image generated

Ddaroth I don't know of any tagger that would make such accurate captions that would res...

EyeSpyBekoAI•5/25/24, 5:54 AM

It’s cool if it’s not exact. Do you have a preferred tagger or captioner?

Joshcandle•5/25/24, 11:27 AM

Has someone here tested creating is own model like realvisx ?

EEyeSpyBekoAI It’s cool if it’s not exact. Do you have a preferred tagger or captioner?

Joshcandle•5/25/24, 12:22 PM

You could use the SUPIR LLava to caption those images, but dont know if that would result in the images you need

Joshcandle•5/25/24, 2:27 PM

try source venv/bin/activate before doing pip install ffmpeg /// But that will only help if you programm is started from the venv too

JJoshcandle You could use the SUPIR LLava to caption those images, but dont know if that wou...

EyeSpyBekoAI•5/25/24, 3:27 PM

I'll give it a shot. Thanks!

JJoshcandle Has someone here tested creating is own model like realvisx ?

EyeSpyBekoAI•5/25/24, 3:28 PM

I think the doc at one point was exploring it. I've never done it myself, but it requires thousands of images, captions, high gpu for training, and patience

Joshcandle•5/25/24, 3:31 PM

Do you know where i could find more about that?

daroth•5/25/24, 6:48 PM

I'm trying out the settings from the latest SDXL training guide, in it there is no CrossAttention optimization selected, is that intentional?

apspoliveira•5/25/24, 9:47 PM

what is the best approach to write requirements for functions of libraries of a operating system?

EEyeSpyBekoAI Hello all. Do any of you have a preferred automatic1111 extension that reads ima...

Furkan Gözükara SECourses•5/25/24, 11:44 PM

i understand but nothing is trained like that :d

JJoshcandle Has someone here tested creating is own model like realvisx ?

Furkan Gözükara SECourses•5/25/24, 11:44 PM

i can make but i need incentive

Ddaroth I'm trying out the settings from the latest SDXL training guide, in it there is ...

Furkan Gözükara SECourses•5/25/24, 11:45 PM

yes

Furkan Gözükara SECourses•5/25/24, 11:45 PM

it may reduce quality

Furkan Gözükara SECourses•5/25/24, 11:45 PM

but currently it is almost same

Furkan Gözükara SECourses•5/25/24, 11:45 PM

actually i am gonna test it

Aapspoliveira what is the best approach to write requirements for functions of libraries of a ...

Furkan Gözükara SECourses•5/25/24, 11:45 PM

can you eloborate more?

FFurkan Gözükara SECourses i understand but nothing is trained like that :d

EyeSpyBekoAI•5/25/24, 11:55 PM

Got it. One can always hope

Thanks for the response.

AIGambino•5/26/24, 2:21 AM

Does anyone know much about prompting? I'm trying to get the side view of a train but cannot get it at all

Abdo Asker•5/26/24, 8:24 AM

I really want to know if there is, generally, a way to upscale AI generated photos creatively like MAGNIFIC AI but with other tools even if I will use very high GPU I don't care .. I just don't want to pay 40 dollars a month because I only want to make like 10-20 upscales each month (Not +100 like MAGNIFIC) so this will save me more money in long term

You told me about AUTOMATIC 1111, but I really don't find many lessons on it and no-one showed that it can compete with MAGNIFIC AI .. so if there's a group or channel that explains that in deep I want you to tell me about it.

daroth•5/26/24, 8:41 AM

theres a workflow to kiiinda replicate it in A1111:

https://x.com/philz1337x/status/1768679154726359128?t=2QY-wTx8BYYyDTwBFyocPg

philz1337x (@philz1337x) on X

For the last few months I have been reverse engineering Magnific AI's famous upscaler

It uses MultiDiffusion, ControlNet tiles and details LoRas

In true AI spirit, I am open sourcing it for everyone to use for free in your apps

It's designed to easily control the amount of