How To Install New DREAMBOOTH & Torch 2 On Automatic1111 Web UI PC For Epic Performance Gains Guide

In this video, I am explaining how you can manually install any extension of the Automatic1111 Web UI. I will demonstrate how to install the DreamBooth extension manually if automatic installation fails. Moreover, I am showing you how to upgrade Torch version 2 for awesome, fantastic image generation and training speed gains.

Playlist of #StableDiffusion Tutorials, #Automatic1111 and Google Colab Guides, DreamBooth, Textual Inversion / Embedding, LoRA, AI Upscaling, Pix2Pix, Img2Img:

Torch installation commands:

DreamBooth extension repo link:

How to install requirements manually:

Easiest Way to Install & Run Stable Diffusion Web UI on PC by Using Open Source Automatic Installer:

How To Install New DreamBooth Extension On RunPod - Automatic1111 Web UI - Stable Diffusion:

Git install link:

0:00 How to install Automatic1111 Web UI extensions manually intro
0:34 How to install Web UI DreamBooth extension on RunPod
0:50 How to install extensions on Automatic1111 Web UI
1:13 Where the extensions are getting installed in Automatic1111 Web UI
1:23 How to manually install DreamBooth extension and where to install it
1:53 How to change requirements to prevent library version confliction
3:10 If it is still not working delete venv folder
3:20 How to update web UI or any extension manually via git pull
3:47 How to install and activate xformers in Automatic1111 Web UI
4:22 What you should see in your CMD when starting Automatic111 Web UI
5:03 How to install Torch 2 / Pytorch 2 for incredible speed and performance gains
7:13 Web UI versions after Installing Torch 2 and latest Cuda

In recent years, the field of artificial intelligence (AI) has seen tremendous growth in developing generative models that can create realistic images from text input. One such model is the Stable Diffusion Torch, a mid-journey text-to-image generative AI based on the PyTorch framework. This article will provide a comprehensive exploration of the Stable Diffusion Torch, including its underlying principles, training methodologies, and potential applications.

I. Principles of the Stable Diffusion Torch

A. Diffusion Models

The Stable Diffusion Torch leverages diffusion models, a class of generative models that simulate the process of diffusion to create realistic images. Diffusion models work by starting with a target image and then gradually adding noise until the image is no longer recognizable. To generate an image from this noisy version, the model learns to reverse the diffusion process by gradually denoising the image until it reaches a clear and realistic representation.

B. Text-to-Image Generation

To generate images from text input, the Stable Diffusion Torch combines diffusion models with a transformer architecture, which is known for its exceptional capabilities in natural language processing tasks. The transformer extracts semantic information from the input text and uses this information to guide the diffusion process, resulting in an image that corresponds to the text description.

II. Training the Stable Diffusion Torch

B. Transformer Training

The transformer model is trained using the text descriptions from the dataset, with the aim of learning to extract semantic information from input text effectively. This is achieved using self-attention mechanisms and extensive pre-training on large-scale text corpora.

C. Diffusion Model Training

The diffusion model is trained using the images from the dataset. The model learns to generate realistic images by reversing the diffusion process, starting from a noisy version of the image and progressively denoising it.

III. Applications of the Stable Diffusion Torch

A. Content Creation

The Stable Diffusion Torch can be used to generate original images for various purposes, such as art, advertising, or entertainment. By providing text descriptions, users can create custom images tailored to their needs.

[SECourses]

Please join discord, mention me and ask me any questions. Thank you for like, subscribe, share and Patreon support. I am open to private consulting with Patreon subscription.
Check video description for the links.

[relb]

I have been trying and trying for nothing till I saw your video. GREAT HELP. Now DB is running . Thanks Dr. Furkan Gözükara

[green]

Everytime I come to find a SD tutorial its always yours recommended thanks for all your work!

[Wickedsight]

Thank you so much! Through the video and some of the comments below, you helped me finally get Dreambooth running. My first model is now training. Lets see what happens!

[CordialH]

Thank you so much for this. I had my auto installation fail and this got me up and running.

[kingkhalagar]

DUDE YOU ARE A HERO!! Thank you so much for making this so simple

[sand]

Thank you tried so long to install xformers and nothing worked. This worked! Very good and easy to understand tutorial.

[oguzozguldur]

Özellikle xformers tarafında halen bir miktar ek çaba gerektiriyor ama şu ana kadar gördüğüm en doğru düzgün içerik olmuş konuyla ilgili, tebrikler.

[camiloalejandrollanio]

Dude! Thank you for making this video!

[ugurakyuz]

Dostum merhaba, Hiç sayfana girip bakmadan konuşma tarzından Türk olduğunu anladım ve çok mutlu oldum.Gösterdiğin şeyler araştırıp bulamadığım bir çok sorunumu çözdü ve çok takdir ettim seni.Kolay Kolay hiçbir kanala abone olmam ama bugün 1 abone kazandın.Başarılarının devamını dilerim.

[grahamulax]

Whoa, was training yesterday and it told me to update to 2 and I was like huh? NO WAY! Question though, still kinda new but making great results. Can this update be used without the extension in automatic1111 and just in dreambooth? If thats a dumb question then...yeah ... super new to python, but dang AI is teaching me while I train AI! What a world. Love your vids!

[TopDealTravel]

Thank you for the video. The error messages already drove me nuts.

[Dannyk]

Great video. Here are my findings/testings.
Baseline:
RTX 3060 12GB
DPM++ 2M Karras, 512x704, samples 30, batch count: 10

Old version: torch 1x, xformers 16: [03:04<00:00, 1.76it/s]
New version: torch: 2.0.0+cu118, xformers 18: [02:49<00:00, 1.91it/s] (only a slight improvement)
New version: torch: 2.0.0+cu118, removed --xformers from cmd arg and add --opt-sdp-attention instead: [01:23<00:00, 4.30it/s]

Maybe I did something wrong with the second part, but xformers only gave me a very slight increase, I am using python: 3.10.4 so maybe that could be the issue.

EDIT: So, when I removed --xformers and replaced it with --opt-sdp-attention, I also removed some other cmd arg like no half vae and full precision, etc.. After writing this, I went back and added --xformers again to test it and my speeds have gone up: [01:20<00:00, 4.41it/s] (same baseline settings)

weird thing is though, in the automatic1111 wiki, --opt-sdp-attention is meant to be faster than xformers: "Faster speeds than using xformers, only available for user who manually install torch 2.0 to their venv. (non-deterministic)"

[jeanfrancoishuertas]

I get a BUG REPORT about not having CUDA Library on my PC. Does CUDA need to be installed before you add the dreambooth extension? can you make a tutorial to explain how to do that?

[wykydytron]

If anyone is here month later, just type as launch parameter --reinstall-torch run it possibly 2 times as it sometimes stops after downloading new torch then run --reinstall-xformers, its literally all you need to do, no need to mess wyth python etc anymore.

[ninuwander]

Thanks, this is easy to understand, is there a safer way to fully remove an extension like uninstallation ?

[andersonbrandaosudario]

I was following until the "delete venv" folder - which doesn't exist anymore? - might have been updated to a Environment batch file on sd.webui folder? But as soon as I delete it nothing seems to run anymore. I am stuck the the "no tab showing after installed" topic still...

[pnb]

Hey man, posted on that other guide that didn't work for me, just wanted to say that this one did, in fact, it's the first time this torch/cuda thing have ever worked on automatic1111 and I've been trying guide, videos and reddit advice for days now, thanks.

[DavZeugme]

Activate inside cmd didn't work for me "Activate isn't recognized". Any idea?

[telefonsakalar]

Exception training pattern: 'Executable batch size not found, zero reached.'. i am facing the problem what should i do