Software Engineering Courses (SECourses)•3y ago

I had Very good success using control net Depth, Canny and openpose to upscale and bring out details

I had Very good success using control net Depth, Canny and openpose to upscale and bring out details in an old photo from this

Rrafraf Also, where do you explain the block training --block_lr 1e-5,1e-5,1e-5,1e-5,1e...

Furkan Gözükara SECourses•10/8/23, 1:26 PM

use config from patreon

FFurkan Gözükara SECourses it increases vram slighty and reduces training speed slightly

rafraf•10/8/23, 1:26 PM

Ok good to know

EElisoli I had Very good success using control net Depth, Canny and openpose to upscale a...

Furkan Gözükara SECourses•10/8/23, 1:26 PM

nice

ElisoliOP•10/8/23, 1:26 PM

I wish there was a better Canny model for detecting edges

ElisoliOP•10/8/23, 1:26 PM

had to mess with it a lot

EElisoli I wish there was a better Canny model for detecting edges

Furkan Gözükara SECourses•10/8/23, 1:27 PM

ye we need better sdxl controlnet

ElisoliOP•10/8/23, 1:27 PM

also waiting for QR control net for sdxl, I cant seem to properly upscale 1.5 images from 512 to higher ones without losing details

FFurkan Gözükara SECourses those models are nasty

Stefanoid9343•10/8/23, 1:34 PM

Thank you so much!

Stefanoid9343•10/8/23, 1:39 PM

@Dr. Furkan Gözükara this is the new one that replace the real vision one? https://youtu.be/AY6DMBCIZ3A?feature=shared can I do it with runpod?

YouTubeSECourses

First Ever SDXL Training With Kohya LoRA - Stable Diffusion XL Trai...

Updated for SDXL 1.0. How to install #Kohya SS GUI trainer and do #LoRA training with Stable Diffusion XL (#SDXL) this is the video you are looking for. I have shown how to install Kohya from scratch. The best parameters to do LoRA training with SDXL. How to use Kohya SDXL LoRAs with ComfyUI. How to do checkpoint comparison with SDXL LoRAs and m...

SStefanoid9343 @Dr. Furkan Gözükara this is the new one that replace the real vision one? https...

Furkan Gözükara SECourses•10/8/23, 1:41 PM

yes

Furkan Gözükara SECourses•10/8/23, 1:41 PM

you can do with runpod

Furkan Gözükara SECourses•10/8/23, 1:41 PM

i suggest you to use dreambooth config from patreon

Furkan Gözükara SECourses•10/8/23, 1:41 PM

and use that in dreambooth tab

Furkan Gözükara SECourses•10/8/23, 1:41 PM

better than lora

Stefanoid9343•10/8/23, 1:43 PM

yes I’m a supporter

SStefanoid9343 yes I’m a supporter

Furkan Gözükara SECourses•10/8/23, 1:45 PM

ye please use latest config from patreon

Furkan Gözükara SECourses•10/8/23, 1:45 PM

much better

Furkan Gözükara SECourses•10/8/23, 1:45 PM

after loading config

Furkan Gözükara SECourses•10/8/23, 1:45 PM

just set your folder paths

Furkan Gözükara SECourses•10/8/23, 1:46 PM

dmn kaggle not working with sdxl trying to solve

Sandstormleader•10/8/23, 1:46 PM

been a rough week for training lol

FFurkan Gözükara SECourses ye please use latest config from patreon

Stefanoid9343•10/8/23, 1:54 PM

I’m watching your tutorial right now

SStefanoid9343 @Dr. Furkan Gözükara this is the new one that replace the real vision one? https...

Stefanoid9343•10/8/23, 2:38 PM

on this tutorial how is the procedure on runpod?

Stefanoid9343•10/8/23, 2:40 PM

nevermind I found the instructions on the github link

Sandstormleader•10/8/23, 3:14 PM

Sandstormleader•10/8/23, 3:14 PM

33k steps later

Furkan Gözükara SECourses•10/8/23, 3:24 PM

finally found the kaggle error

SSandstormleader Click to see attachment

Furkan Gözükara SECourses•10/8/23, 3:25 PM

very high quality

Sandstormleader•10/8/23, 3:28 PM

@Dr. Furkan Gözükara more detail with 100+ hi res steps

Sandstormleader•10/8/23, 3:41 PM

Sandstormleader•10/8/23, 3:53 PM

Sandstormleader•10/8/23, 3:54 PM

these are again with 0 class images

Sandstormleader•10/8/23, 3:57 PM

220 professional photos, cropped but not down sized, trained at 768, xformers, 8bit adam, bf16, 150 epochs

LEGACIES•10/8/23, 4:02 PM

dreambooth?

Sandstormleader•10/8/23, 4:02 PM

yep

LEGACIES•10/8/23, 4:02 PM

what do you by "cropped but not down sized,"

Sandstormleader•10/8/23, 4:03 PM

so i will crop them in closer to the subject so the face stands out more but i don't make them 1080x1080 or anything like that

LEGACIES•10/8/23, 4:04 PM

what setting did you use for "Save Model Frequency (Epochs)"?

Sandstormleader•10/8/23, 4:05 PM

i just put 50 but i liked the whole 150 best

Sandstormleader•10/8/23, 4:06 PM

but so some of my photos i used were closer to 4000x5000 pixels

LEGACIES•10/8/23, 4:06 PM

your models looks amazin

Stefanoid9343•10/8/23, 4:07 PM

@Dr. Furkan Gözükara I'm having thiis error "torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 19.68 GiB total capacity; 12.33 GiB already allocated; 19.75 MiB free; 12.42 GiB reserved in total by PyTorc"

Sandstormleader•10/8/23, 4:08 PM

i've been testing for fastest/ laziest way to make good models

Stefanoid9343•10/8/23, 4:08 PM

I already check "Gradient checkpointing"

Stefanoid9343•10/8/23, 4:09 PM