Software Engineering Courses (SECourses)•3y ago

Of course. The --medvram-sdxl option is there to generate the same image faster and in a different w

Of course. The --medvram-sdxl option is there to generate the same image faster and in a different way. And it really will generate faster, and you'll see the end result at the very end.

Mmikemenders Of course. The --medvram-sdxl option is there to generate the same image faster ...

vKorby•10/24/23, 6:42 PM

sorry kinda new to this, where is that option under?

mikemendersOP•10/24/23, 6:55 PM

Type it to webui-user.bat file after set COMMANDLINE_ARGS=set COMMANDLINE_ARGS=

mikemendersOP•10/24/23, 6:56 PM

like this: set COMMANDLINE_ARGS=--medvram-sdxlset COMMANDLINE_ARGS=--medvram-sdxl

mikemendersOP•10/24/23, 6:56 PM

and start SD with webui-user.bat

Mmikemenders and start SD with webui-user.bat

vKorby•10/24/23, 7:00 PM

got it trying it rn thank you, btw my "webui-user.bat" file doesn't have ".bat" its just "webui-user" is that normal?

BarbaraM•10/24/23, 7:00 PM

idk why but control net doesn't seem to work for me on sdxl models, (even with control nets sdxl models), tried with canny, open pose ad depth. nothing seems to work

Capture_decran_2023-10-24_a_20.59.36.png

VvKorby got it trying it rn thank you, btw my "webui-user.bat" file doesn't have ".bat" ...

mikemendersOP•10/24/23, 7:01 PM

yes, but you don't have the file extension display turned on in windows

Mmikemenders yes, but you don't have the file extension display turned on in windows

vKorby•10/24/23, 7:02 PM

Oh ok thanks

mikemendersOP•10/24/23, 7:15 PM

I just did a one-on-one Lora training. I only kept the images that were identity 1 based on CompreFace, even discarding the 0.99 ones. I trained with three different optimizers, Adafactor, Prodigy and AdamW. Adafactor and Prodigy trained on the RealisticVision 5.1 model, AdamW trained on SD 1.5.
All three models bring out the person and their body well, and it's very interesting to see how each model brings out different details in the person's face and body. Since the images are high resolution, I'll try SDXL training on them soon. I am very curious!

ZZono50 What's the recommended way to train a 1.5 lora on your face now?

Furkan Gözükara SECourses•10/24/23, 7:17 PM

i am in reserach

Furkan Gözükara SECourses•10/24/23, 7:17 PM

and this is not even a good dataset

KKhang @Dr. Furkan Gözükara Hi Dr.Furkan & everyone, is there any guide for training Co...

Furkan Gözükara SECourses•10/24/23, 7:18 PM

ye i never done any controlnet training

Aakellio Hi, who has tried running on nvidia A100? What is the fastest video card for sdx...

Furkan Gözükara SECourses•10/24/23, 7:18 PM

for batch size 1 : rtx 4090

Furkan Gözükara SECourses•10/24/23, 7:18 PM

but for higher batch sizes you can go 80 gb

Furkan Gözükara SECourses•10/24/23, 7:18 PM

increase batch size and keep gradient acc disabled

Rrafraf @Dr. Furkan Gözükara is there a tool to find duplicates within a large dataset?

Furkan Gözükara SECourses•10/24/23, 7:18 PM

you mean duplicate images i presume

Mmikemenders yes, but you don't have the file extension display turned on in windows

vKorby•10/24/23, 7:18 PM

its still takin long for sum reason lol

Furkan Gözükara SECourses•10/24/23, 7:18 PM

i think you can ask chat gpt to get quick python code

Mmikemenders Speaking of Adafactor: why can't I select the Adafactor scheduler if the optimiz...

Furkan Gözükara SECourses•10/24/23, 7:19 PM

i never tried

FFurkan Gözükara SECourses you mean duplicate images i presume

rafraf•10/24/23, 7:19 PM

Yes

FFurkan Gözükara SECourses i think you can ask chat gpt to get quick python code

rafraf•10/24/23, 7:19 PM

Ok I’ll check it out

VvKorby Hey guys, i need some help here does anyone have any videos or advice if my SD i...

Furkan Gözükara SECourses•10/24/23, 7:19 PM

yes add --medvram

Furkan Gözükara SECourses•10/24/23, 7:19 PM

if be still too slow make --lowvram

BBarbaraM idk why but control net doesn't seem to work for me on sdxl models, (even with c...

Furkan Gözükara SECourses•10/24/23, 7:20 PM

you must have error on cmd

FFurkan Gözükara SECourses you must have error on cmd

BarbaraM•10/24/23, 7:21 PM

error message on ther terminal?

VvKorby its still takin long for sum reason lol

mikemendersOP•10/24/23, 7:23 PM

How much VRAM have?

FFurkan Gözükara SECourses if be still too slow make --lowvram

vKorby•10/24/23, 7:23 PM

Oh ok,im thinkin its my computer also been runnin a bit slow lately so going to be tweaking some stuff. btw these are my specs

BBarbaraM error message on ther terminal?

Furkan Gözükara SECourses•10/24/23, 7:24 PM

yes

FFurkan Gözükara SECourses you must have error on cmd

BarbaraM•10/24/23, 7:24 PM

i m using runpod, nothing on my terminal

VvKorby Oh ok,im thinkin its my computer also been runnin a bit slow lately so going to ...

Furkan Gözükara SECourses•10/24/23, 7:24 PM

it is only 8gb sadly

Furkan Gözükara SECourses•10/24/23, 7:24 PM

but you can do inference

FFurkan Gözükara SECourses yes

BarbaraM•10/24/23, 7:24 PM

Capture_decran_2023-10-24_a_21.24.41.png

BBarbaraM i m using runpod, nothing on my terminal

Furkan Gözükara SECourses•10/24/23, 7:24 PM

please use my auto installr

Furkan Gözükara SECourses•10/24/23, 7:25 PM

and start web ui as i do

Furkan Gözükara SECourses•10/24/23, 7:25 PM

https://www.patreon.com/posts/84896373

FFurkan Gözükara SECourses please use my auto installr

BarbaraM•10/24/23, 7:26 PM

i have this auto installer

Capture_decran_2023-10-24_a_21.25.50.png

VvKorby Oh ok,im thinkin its my computer also been runnin a bit slow lately so going to ...

mikemendersOP•10/24/23, 7:26 PM

Sorry, there's no --lowvram-sdxl switch, so it's still slow because you're generating to regular memory, so you have full VRAM. You have 8 GB VRAM, I think. This is very marginal.

mikemendersOP•10/24/23, 7:27 PM

Check this options:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Command-Line-Arguments-and-Settings

GitHub

Command Line Arguments and Settings

Stable Diffusion web UI. Contribute to AUTOMATIC1111/stable-diffusion-webui development by creating an account on GitHub.

mikemendersOP•10/24/23, 7:28 PM

I have an RTX 3060 12 GB card and it is also a bit slow for SDXL generation, but it fits in VRAM. With AI, VRAM is very important, how much you have.

mikemendersOP•10/24/23, 7:29 PM

If you don't have enough VRAM, you're working in RAM, and even though I have DDR5 RAM with i9, it's definitely slower.

BBarbaraM i have this auto installer

Furkan Gözükara SECourses•10/24/23, 7:30 PM

ye after that start web ui manually

Furkan Gözükara SECourses•10/24/23, 7:30 PM

so kill previous one with

Furkan Gözükara SECourses•10/24/23, 7:30 PM

fuser -k 3000/tcp

Mmikemenders If you don't have enough VRAM, you're working in RAM, and even though I have DDR...

vKorby•10/24/23, 7:34 PM

Yeah I do have 8 GB looks like Imma jus have to deal with that lol, I'm thinking of using this and leonardo ai image 2 image

FFurkan Gözükara SECourses shared disk space

EyeSpyBekoAI•10/24/23, 7:57 PM

Hey there, thanks for this! Network volume seems to be working quite well as I was able to see setup pods pop up after being closed and relaunched. I'm setting up more now. One thing I'm noticing though, is that getting port availability is considerably dodgy. Is there a way to get around this, or is it one of "it is what it is" situations?

Mmikemenders @Jas The large file is also good for styling, and can be reduced to 512 by lora....

Jas•10/24/23, 8:22 PM

Thanks for your recommendations, with the labelling, I am a bit uncertain how to do it as they are quite abstract. I will show you a couple of examples of the images I am training on, maybe you can suggest what to label.

Come to think of it, I am not sure I am really training a 'style' but it is more like a type of artwork / art object.

for the label I have something like "Ceramicblobstyle, pink, green and blue theme"

I don't know how to label it outside of mentioning the colours in the image, because I want to replicate the patterns, the shapes and blobs.