Software Engineering Courses (SECourses)•12mo ago

@Furkan Gözükara SECourses Ive done dreambooth training on this Flux Checkpoint https://civitai.com/

@Furkan Gözükara SECourses Ive done dreambooth training on this Flux Checkpoint https://civitai.com/models/161068/stoiqo-newreality-flux-sd35-sdxl-sd15?modelVersionId=979329

But somehow i cant extract the lora from it, doesnt matter if i use the Full or the Pruned Model, do you know why that is?

�

𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮•1/25/25, 11:56 AM

Is multi gpu dreambooth training possible?

�

𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮•1/25/25, 11:56 AM

Has anyone tried training dreambooth on a 80gb card?

Timson•1/25/25, 12:16 PM

Pro tip: if you are getting low quality results when upscaling image with flux img2img always use at least 0.15-0.2

Init Image Reset To Norm

Init Image Reset To Norm

. Upscale artefacts you see are the result of SwarmUI upscale init image to target resolution BEFORE it is VAE encoded. This works WAY better than increasing image creativity.
Took me a while to figure it out

gamejump.•1/25/25, 12:47 PM

in this video A1111 web ui is no longer supportednow stabilitymatrix is there how to go ahead then
THank you

AiInfluence•1/25/25, 12:53 PM

pwndnoobcake•1/25/25, 3:16 PM

@Dr. Furkan Gözükara i asked ChatGPT 4o about blockswap and it linked your video

Wwardensc2 It's because you dont know how to prompt with some tweak in ComfyUI, Flux is mil...

pwndnoobcake•1/25/25, 3:18 PM

tweaks like? my only trick is to prompt clip and t5xxl on their own format separately

Wwardensc2 It's because you dont know how to prompt with some tweak in ComfyUI, Flux is mil...

N7-Blackbird•1/25/25, 6:28 PM

Nah man, these are too artificial. But we might have different standards and it's ok)

N7-Blackbird•1/25/25, 6:30 PM

FLUX is great but way too inflexible at the moment.

FFabricatedgirls how far do you think we are from having veo 2 quality video models at home

Furkan Gözükara SECourses•1/26/25, 2:02 AM

sadly no idea atm.

AAiInfluence is it the same as Q8? in terms of quality

Furkan Gözükara SECourses•1/26/25, 2:03 AM

no de-distilled is differently working. it uses cfg 3.5 so you can use negative prompts

JJoshcandle @Furkan Gözükara SECourses Ive done dreambooth training on this Flux Checkpoint ...

Furkan Gözükara SECourses•1/26/25, 2:03 AM

must be kohya not supporting yet

Furkan Gözükara SECourses•1/26/25, 2:03 AM

you have to open issue thread

Furkan Gözükara SECourses•1/26/25, 2:03 AM

previously he wasnt supporting training even de-distilled model

Furkan Gözükara SECourses•1/26/25, 2:03 AM

since they are trained with different scripts

�𝗡𝗶𝘁𝗶𝗻 𝗣𝗮𝗿𝗲𝘃𝗮 Is multi gpu dreambooth training possible?

Furkan Gözükara SECourses•1/26/25, 2:04 AM

yes with 80 gb

Furkan Gözükara SECourses•1/26/25, 2:04 AM

it works

TTimson Pro tip: if you are getting low quality results when upscaling image with flux i...

Furkan Gözükara SECourses•1/26/25, 2:04 AM

nice tips. do you have any exact config can you share screenshot?

Ggamejump.in this video A1111 web ui is no longer supportednow stabilitymatrix is there ho...

Furkan Gözükara SECourses•1/26/25, 2:05 AM

you can manually install

Furkan Gözükara SECourses•1/26/25, 2:05 AM

or use our installer

Furkan Gözükara SECourses•1/26/25, 2:05 AM

https://www.patreon.com/posts/86307255

Patreon

1 Click Installer for Automatic1111 SD Web UI, SDXL, ControlNet, Al...

Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon

AAiInfluence Click to see attachment

Furkan Gözükara SECourses•1/26/25, 2:07 AM

good job

Ppwndnoobcake @Dr. Furkan Gözükara i asked ChatGPT 4o about blockswap and it linked your vide...

Furkan Gözükara SECourses•1/26/25, 2:07 AM

well it is because it has no idea of block swap and searches google very likely

BNP4535353•1/26/25, 2:25 AM

The 48g 4090 is currently available for private consumers in China (you can buy it exactly), with no restrictions (such as not being able to run on Windows). The only downside is that due to the fully modified turbo fan, the noise is similar to that of a helicopter...

Ppwndnoobcake tweaks like? my only trick is to prompt clip and t5xxl on their own format separ...

wardensc2•1/26/25, 2:46 AM

There so many, I must say some examples: detailer daemon, different t5xxl & clip text encode, separate prompt and prompt style ... so many I dont even remember. Especially when combine Teacache with Sage Attention I nearly get 3x render speed without losing any quality of an image. Here my example:

wardensc2•1/26/25, 3:22 AM

wardensc2•1/26/25, 3:22 AM

NN7-Blackbird Nah man, these are too artificial. But we might have different standards and it'...

wardensc2•1/26/25, 3:23 AM

Flux tend to make image more polished which people claim it's artificial. It depends on people taste, I agree. But for understanding prompt, I never going back to SD 1.5 and even SDXL. Flux is way ahead of prompt understanding enough to forgive its weakness which is styling an image.

FFurkan Gözükara SECourses nice tips. do you have any exact config can you share screenshot?

Timson•1/26/25, 10:10 AM

no single config fits all cases, you always have to play around with creativity and normalisation settings depending on composition, upscale factor and how close you need upscale to be to original

Timson•1/26/25, 10:13 AM

This is what I would start with

Wwardensc2 Flux tend to make image more polished which people claim it's artificial. It dep...

N7-Blackbird•1/26/25, 12:32 PM

there is a huge difference between a raw photo (digital or analog) and a stylized photo (with filters) of a person (a portrait). If you work with photography your eye is trained on raw images and you notice when something is even slightly artificial. FLUX at it's current stage struggles a lot with creating raw portraits. With elaborate prompt you can get better results, but currently it is a struggle, whereas accordingly-trained stable diffusion 1.5 or XL custom models do this with ease. It would be great if one day there will be a more flexible FLUX version. That said, if you are looking to create studio-like portraits with retouched skin and filter effect, FLUX is very good. Just not for natural portraits, same as Midjourney btw.

AiInfluence•1/26/25, 1:17 PM

@Furkan Gözükara SECourses @Timson I'm planning as a next project to train shoes, jewellery and other objects. Should I follow the same concept of diversity as I did with the logo ?

AAiInfluence @Furkan Gözükara SECourses @Timson I'm planning as a next project to train shoes...

Timson•1/26/25, 2:57 PM

you'll have to figure out the peculiarities of each class, but the starting point would be the same process. It all depends on how each of your base classes are trained into flux

Ppwndnoobcake tweaks like? my only trick is to prompt clip and t5xxl on their own format separ...

Timson•1/26/25, 3:02 PM

Do you see real non placebo benefit in this? Do you use SwarmUI comfyUI flow customisation for this?
I did not follow up this research avenue because BFL is not leveraging it for flux pro API, seems counter-intuitive that they would ignore it if it was a viable way to improve outputs with no additinal costs

TTimson Do you see real non placebo benefit in this? Do you use SwarmUI comfyUI flow cus...

pwndnoobcake•1/26/25, 3:11 PM

comfy, yeah benefit is pretty decent

sergio_gonzo•1/26/25, 3:30 PM

Dr Furkan I try to use the copuon in massconpute but have a error,, dont accept it, surely i make a error

Flakdragoon•1/26/25, 5:31 PM

hi, I'm trying to use flux fill in SwarmUI to replace an orange wristband over a subject in a photo with some ISO noise. I'm using the basic generation flow and it looks perfect except it's too perfect and doesn't have any of the ISO noise in the original photo.

Is there a way I can get the generation to also add in similar noise to the inpainted mask that I'm shrink growing? I'm at max creativity since otherwise the orange wristband still shows. you can see in this image clip