Software Engineering Courses (SECourses)•3y ago

Hello, Can anyone recommend any prompt guide repository like this but for stable diffusion? https:/

Hello, Can anyone recommend any prompt guide repository like this but for stable diffusion?
https://github.com/willwulfken/MidJourney-Styles-and-Keywords-Reference/blob/main/README.md Just want 20-50 images and their prompts, dont wanna create one from scratch.

GitHub

MidJourney-Styles-and-Keywords-Reference/README.md at main · willwu...

A reference containing Styles and Keywords that you can use with MidJourney AI. There are also pages showing resolution comparison, image weights, and much more! - MidJourney-Styles-and-Keywords-Re...

TTitez Guys, I'm still very new to Stable Defusion and I've run into some kind of confu...

Furkan Gözükara SECourses•7/12/23, 12:10 AM

welcome

Furkan Gözükara SECourses•7/12/23, 12:11 AM

well you can try controlnet tile img2img

Furkan Gözükara SECourses•7/12/23, 12:11 AM

if you dont want to do training

Furkan Gözükara SECourses•7/12/23, 12:11 AM

https://youtu.be/3E5fhFQUVLo

YouTubeSECourses

Zero to Hero ControlNet Tutorial: Stable Diffusion Web UI Extension...

I show how to install Automatic1111 Web UI & ControlNet extension installation from scratch in this video. Moreover I show how to make amazing QR codes and inpainting and out painting of ControlNet which are very similar to Photoshop generative fill and Midjourney zoom out. Furthermore, I explain and show what are Canny, Depth, Normal, OpenPose...

Furkan Gözükara SECourses•7/12/23, 12:12 AM

1:17:19 How to turn your image into a fully stylized image with tiling without training

Xxlads are we able to delete dreambooth from runpod im unable to

Furkan Gözükara SECourses•7/12/23, 12:12 AM

yes you can delete

Furkan Gözükara SECourses•7/12/23, 12:12 AM

rm -r folder_path_extension

MMirrorPhotosKing Hello, Can anyone recommend any prompt guide repository like this but for stable...

Furkan Gözükara SECourses•7/12/23, 12:13 AM

ye that would be nice if there is any

Furkan Gözükara SECourses•7/12/23, 12:13 AM

@Davi3311 if you couldnt make it let me know i will try to help you more

Furkan Gözükara SECourses•7/12/23, 12:15 AM

i had been struggling to fix this bluescreen. did a bios upgrade and everything returned back to default :d

Furkan Gözükara SECourses•7/12/23, 1:16 AM

found excellent article : https://huggingface.co/docs/transformers/v4.23.1/en/perf_train_gpu_one

Efficient Training on a Single GPU

veldonlives•7/12/23, 2:19 AM

every time i get to the point in one of the tutorials where it discusses the formula to determine how many classification images are needed, i wonder two things:

1. why hasn't someone just written a python script that prompts a user and spits out the config file that you load into automatic1111 with all the "ideal" settings. (or, if not that, why don't people just share an ideal config file)
2. if the number of classification images you have is greater than the number needed according to the formula, is that a problem?

veldonlives•7/12/23, 2:21 AM

i am legitimately curious - the config settings must be somewhat "best practice" at any given point

DJ_Body man•7/12/23, 4:25 AM

is this where to ask for help?

DJ_Body man•7/12/23, 4:25 AM

w sdzl

DJ_Body man•7/12/23, 4:25 AM

sdxl

FFurkan Gözükara SECourses how many images you have and what is your per instance class count?

Narva•7/12/23, 6:28 AM

I have 24 Training dataset and 200 per instance class, I followed your tutorial, it only generated 500 class image

Narva•7/12/23, 6:32 AM

also dreambooth found the rest of the class images and after that the training started, maybe some class images were corrupted

FFurkan Gözükara SECourses Click to see attachment

Khang•7/12/23, 7:33 AM

Those are nice pics Dr. Furkan. Did you create them with ComfyUI? I hope you will make a video comparing the pros and cons of each UI.

DDJ_Body man is this where to ask for help?

Furkan Gözükara SECourses•7/12/23, 10:22 AM

yeyes

Vveldonlives i am legitimately curious - the config settings must be somewhat "best practice"...

Furkan Gözükara SECourses•7/12/23, 10:22 AM

i agree

KKhang Those are nice pics Dr. Furkan. Did you create them with ComfyUI? I hope you wil...

Furkan Gözükara SECourses•7/12/23, 10:22 AM

yes

Furkan Gözükara SECourses•7/12/23, 10:22 AM

and i am working on it

NNarva I have 24 Training dataset and 200 per instance class, I followed your tutorial,...

Furkan Gözükara SECourses•7/12/23, 10:23 AM

generate class images from txt2img tab

Furkan Gözükara SECourses•7/12/23, 10:23 AM

i did a lora training on sdxl yesterday on my computer

Furkan Gözükara SECourses•7/12/23, 10:23 AM

looks like my LR was too low

Furkan Gözükara SECourses•7/12/23, 10:24 AM

dmn this thing taking time

Furkan Gözükara SECourses•7/12/23, 10:24 AM

to experiment

Furkan Gözükara SECourses•7/12/23, 11:23 AM

started another lora training

Furkan Gözükara SECourses•7/12/23, 11:24 AM

90 min on rtx 3090 for 13 images 200 epochs + class images

FFurkan Gözükara SECourses 90 min on rtx 3090 for 13 images 200 epochs + class images

Jehex•7/12/23, 11:43 AM

hi there, I have one question about the " steps " of the images training ( in kohya ss we have to rename our training folder like 30_namemodel ) my question is how many steps do you recommend for get a decent result ? I see that divise 1000 by the number of training images can give good results ( for exemple we have 60 training images so 1000/60 = 16-17 steps per images ) I did it today and the result was honestly not too bad. Thanks

JJehex hi there, I have one question about the " steps " of the images training ( in ...

Furkan Gözükara SECourses•7/12/23, 11:48 AM

i calculate based on epoch

Furkan Gözükara SECourses•7/12/23, 11:48 AM

so 10 epoch 20 repeat means 200 epochs

Furkan Gözükara SECourses•7/12/23, 11:48 AM

usually 200 epochs is sufficient

Furkan Gözükara SECourses•7/12/23, 11:48 AM

if overtrains or dont learns change learning rate

Jehex•7/12/23, 11:48 AM

ok thanks, I will test that

Brofessor•7/12/23, 12:45 PM

hello

Brofessor•7/12/23, 12:45 PM

trying to set up Lora Kohya on runpod according to guide

Brofessor•7/12/23, 12:45 PM

anyone seen this error?: train_network.py: error: unrecognized arguments: --cache_text_encoder_outputs

Brofessor•7/12/23, 12:45 PM

TF-TRT Warning: Could not find TensorRT
usage: train_network.py [-h] [--v2] [--v_parameterization]

Brofessor•7/12/23, 12:45 PM

tensorrt should be installed

weilu•7/12/23, 12:46 PM

@Dr. Furkan Gözükara Do you know how to add LoRA safetensors files trained by kohya script to sdxl pipeline?

revoerom•7/12/23, 12:56 PM

hello, i've tried follwing this guide(How To Install DreamBooth & Automatic1111 On RunPod & Latest Libraries - 2x Speed Up - cudDNN - CUDA https://www.youtube.com/watch?v=c_S2kFAefTQ) to install SD A1111 dreambooth on runpod, but run into this error at the python realuncher.py stage:

Launching Web UI with arguments: -f --port 3000 --xformers --skip-install --listen --enable-insecure-extension-access
Traceback (most recent call last):
File "/workspace/stable-diffusion-webui/launch.py", line 38, in <module>
main()
File "/workspace/stable-diffusion-webui/launch.py", line 34, in main
start()
File "/workspace/stable-diffusion-webui/modules/launch_utils.py", line 340, in start
import webui
File "/workspace/stable-diffusion-webui/webui.py", line 14, in <module>
from fastapi import FastAPI, Response
ModuleNotFoundError: No module named 'fastapi'
Relauncher: Process is ending. Relaunching in 2s...

does anyone know how to fix the fastapi error? thanks in advance!

YouTubeSECourses

How To Install DreamBooth & Automatic1111 On RunPod & Latest Librar...

I have shown how to install latest version of Automatic1111 Web UI for Stable Diffusion and DreamBooth extension of Auto1111 on RunPod in this video. Moreover, I show how to upgrade to latest Cuda, Torch and cuDNN DLL files. With these upgrades the image generation speed literally doubles.

GitHub Readme File