How to Do SDXL Training For FREE with Kohya LoRA - Kaggle - NO GPU Required - Pwns Google Colab

If you don't have a strong GPU for Stable Diffusion XL training then this is the tutorial you are looking for. We will use Kaggle free notebook to do Kohya SDXL LoRA training. In this tutorial you will master Kohya SDXL with Kaggle! 🚀 Curious about training Kohya SDXL? Learn why Kaggle outshines Google Colab! We will uncover the power of free Kaggle's dual GPU. 🔥 Step-by-step guide inside! Boost your skills and make the most of FREE Kaggle resources! 💡 #Training #SDXL #Kaggle

Kaggle Kohya SDXL Notebook File ⤵️

Tutorial GitHub Readme File ⤵️

SECourses Discord To Get Full Support ⤵️

0:00 Introduction to how to do amazing FREE training of Stable Diffusion XL without owning a GPU
2:35 How to register Kaggle to get a free account to do free training
3:05 How to verify your phone number in Kaggle to be able to use cloud GPUs for free
3:55 How to generate a Kaggle notebook and start Stable Diffusion XL free Kohya SS LoRA training
4:20 How to download and import SDXL LoRA training notebook
5:46 How to properly with correct config start session on Kaggle to begin training
5:56 How to enable GPU on Kaggle
6:34 How to see how much GPU time you have used and how much you have left on Kaggle
6:49 How to look at used resources in your Kaggle session such as disk space, GPU, CPU, RAM
7:00 How to clone Kohya SS GUI and install it on a free Kaggle notebook
7:21 How to understand and use pathing structure of Kaggle
7:37 Where is our root / working directory in Kaggle
9:57 How to know when the installation of Kohya SS GUI has been completed
10:14 How to download ground truth regularization / classification images
13:17 How to upload your regularization / classification images and use them
14:13 How to use your previously uploaded images / datasets in your Kaggle training sessions
16:00 How to start Kohya SS GUI on Kaggle notebook
16:31 How to access started Kohya SS GUI instance via publicly given Gradio link
17:09 Starting to setup Kohya SDXL LoRA training parameters and settings
17:40 Which source model we need to use for SDXL training a free Kaggle notebook
18:55 How to prepare training dataset easily with dataset preparation feature of Kohya SS GUI
19:26 How to upload your training images and prepare them for SDXL training
20:25 How get and set folder path of training and regularization / classification images
20:51 Where to and how to save training results and how to generate training folders
21:44 How to copy info to folders tab
22:09 Setting up all training parameters
23:44 Network Rank Dimension trade-off
24:44 Continuing to setting up all training parameters
25:54 How to start training after everything is set
26:54 What is the formula of calculating number of training total steps
27:54 How to execute training command
29:10 How to calculate necessary number of classification / regularization images that you need
31:05 Training started
31:24 Why it shows total number of epochs double of the number we did set
32:10 Where is our SDXL LoRA training checkpoints are saved and how to download them
33:07 Why generated safetensor files, checkpoints are 228 MBs
33:27 How to enable allow multiple files download in your browser to download generated LoRA checkpoints
33:37 How to download all of the checkpoints as a single file - zip them all
34:27 How to download LoRA safetensors folder entirely
35:04 How to extract and open downloaded as zip LoRA checkpoints
36:48 How to save your LoRA checkpoints on your Kaggle account to use later
37:50 How to use your trained LoRA checkpoints in your Automatic1111 Web UI on your PC
39:40 How to find best checkpoint of your Kohya SDXL LoRA training
40:07 How to see used prompts and settings of generated images via png info tab of Automatic1111 Web UI
40:26 How did I decide to use the certain checkpoint via x/y/z script of Automatic1111 Web UI
41:08 How to use your LoRAs in Automatic1111 Web UI
42:08 How to select your LoRA from the interface
44:04 How to generate same batch with correct seed, how batch seed is determined
44:38 How to install after detailer (adetailer) extension to improve faces in your generations automatically
45:07 After detailer extension enabled comparison results
46:01 How to get amazing likeness - realism having images of your trained subject,
46:27 How to find best amazing among thousands of generated images by using DeepFace AI similarity script

[SECourses]

If I have been of assistance to you and you would like to show your support for my work, please consider becoming a patron on 🥰 ⤵

Technology & Science: News, Tips, Tutorials, Tricks, Best Applications, Guides, Reviews ⤵

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

[makenthirankumanan]

you are the best free educator in the world. Teach us without any payback. Damn sure, after I start earning, I'll do something for you. I know the struggle that you making these kind of tutorials, takes hours of works. Thankyou so much, you'll be succeed soon. 🔥🔥🔥🔥. If you want any help related about editing and other stuff. I think there are so many people waiting to help you including me. So if there is anything please say in one video that how we help you to do these work(educating). we'll help you sure.

Hocam sağolasın, sen de olmasan bu konularla ilgili kaynağı bu kadar derinlemesine ve detaylı bulmak zor. Teşekkür ederim. Son adımda bir sorun yaşadım ancak yeniden düzenleyip deneyeceğim daha sonra.

[shun_almya]

This educational video is no longer of any use due to the ban. Will you create another video?

[ommaarlak]

im still using google colab free to train lora and never had any problem ! i have a specifique workflow ! its really easy ! and thx for the video tho

[Because_Reasons]

Thanks for this. Def interested in dreambooth for SDXL

[andrewsager]

Well done. Easy to follow, detailed explanation of every step and even free resources. Great tutorial! Thank you for what you are doing

[DavidSeguraIA]

Great video, which settings you recommend to use for style training, you know a video talking about it?

[TomiTom]

Hallo, thanks for your time and effort. I followed your steps but when I want to start the Kohya interface, the site opens with the following text:
"No interface is running right now"
Any idea what is happening here?

Thank you!

[emrey]

harikasınız hocam teşekkürler. patreon zorunluluğu kılmadığınız için teşekkürler öğrencilere de destek olmuş oluyorsunuz böylece

[qookiesbites]

I am getting error " ERROR: Operation cancelled by user " when so ever I run the webui cell which is at the last, any solution ?

[lifjipz]

Hi there, I can't launch it, when I click public url after install I get this "No interface is running right now". can u help me please?

[dudududuev]

Assalamu alaikum brother, great job. Installing Automatic111 in Kaggle is difficult for you? I would be grateful, there are notepads on the Internet, but almost all of them are buggy or incomplete.

[enoch]

he config attributes {'add_watermarker': None} were passed to StableDiffusionXLPipeline, but are not expected and will be ignored. Please verify your model_index.json configuration file.
Keyword arguments {'add_watermarker': None} are not expected by StableDiffusionXLPipeline and will be ignored. I followed your steps. Please help me

[ioio]

I ty for your work, even if my english is not too good like yours, i undersdant you, but any time when i run the gui .. the server close after 5 seconds ; message Keyboard interruption in main thread... closing server..

[tfjilms]

when i run webui command in kaggle, it killing the server today, but yesterday is ok, does anyone know what happened? "Keyboard interruption in main thread... closing server."

[okaytanuki]

Hi, I followed your tutorial but when I run the training command, I get an error that says "Your notebook tried to allocate more memory than is available. It has restarted.". What should I do?

[jamesclow]

Really would love to see a local SDXL full model training. I hear it can be done on a 3090.

[A-Eye--A]

i dont want to train LoRa i want to train dreambooth safesensor can you provide a .json im new i can load it and i know how to change the folders but im not sure about my settings for like 70 images i could do less if thats what i need to do. thank i cant get nothing to work 24 GB Vram

[ashd]

Hi Doctor, in 2023 what's the best method in SD to get high quality studio like images of myself?