Completed
Introduction to how to do amazing FREE training of Stable Diffusion XL without owning a GPU
Class Central Classrooms beta
YouTube playlists curated by Class Central.
Classroom Contents
How to Do SDXL Training with Kohya LoRA - Kaggle - No GPU Required
Automatically move to the next video in the Classroom when playback concludes
- 1 Introduction to how to do amazing FREE training of Stable Diffusion XL without owning a GPU
- 2 How to register Kaggle to get a free account to do free training
- 3 How to verify your phone number in Kaggle to be able to use cloud GPUs for free
- 4 How to generate a Kaggle notebook and start Stable Diffusion XL free Kohya SS LoRA training
- 5 How to download and import SDXL LoRA training notebook
- 6 How to properly with correct config start session on Kaggle to begin training
- 7 How to enable GPU on Kaggle
- 8 How to see how much GPU time you have used and how much you have left on Kaggle
- 9 How to look at used resources in your Kaggle session such as disk space, GPU, CPU, RAM
- 10 How to clone Kohya SS GUI and install it on a free Kaggle notebook
- 11 How to understand and use pathing structure of Kaggle
- 12 Where is our root / working directory in Kaggle
- 13 How to know when the installation of Kohya SS GUI has been completed
- 14 How to download ground truth regularization / classification images
- 15 How to upload your regularization / classification images and use them
- 16 How to use your previously uploaded images / datasets in your Kaggle training sessions
- 17 How to start Kohya SS GUI on Kaggle notebook
- 18 How to access started Kohya SS GUI instance via publicly given Gradio link
- 19 Starting to setup Kohya SDXL LoRA training parameters and settings
- 20 Which source model we need to use for SDXL training a free Kaggle notebook
- 21 How to prepare training dataset easily with dataset preparation feature of Kohya SS GUI
- 22 How to upload your training images and prepare them for SDXL training
- 23 How get and set folder path of training and regularization / classification images
- 24 Where to and how to save training results and how to generate training folders
- 25 How to copy info to folders tab
- 26 Setting up all training parameters
- 27 Network Rank Dimension trade-off
- 28 Continuing to setting up all training parameters
- 29 How to start training after everything is set
- 30 What is the formula of calculating number of training total steps
- 31 How to execute training command
- 32 How to calculate necessary number of classification / regularization images that you need
- 33 Training started
- 34 Why it shows total number of epochs double of the number we did set
- 35 Where is our SDXL LoRA training checkpoints are saved and how to download them
- 36 Why generated safetensor files, checkpoints are 228 MBs
- 37 How to enable allow multiple files download in your browser to download generated LoRA checkpoints
- 38 How to download all of the checkpoints as a single file - zip them all
- 39 How to download LoRA safetensors folder entirely
- 40 How to extract and open downloaded as zip LoRA checkpoints
- 41 How to save your LoRA checkpoints on your Kaggle account to use later
- 42 How to use your trained LoRA checkpoints in your Automatic1111 Web UI on your PC
- 43 How to download and use 750 styles containing styles.csv file
- 44 How to find best checkpoint of your Kohya SDXL LoRA training
- 45 How to see used prompts and settings of generated images via png info tab of Automatic1111 Web UI
- 46 How did I decide to use the certain checkpoint via x/y/z script of Automatic1111 Web UI
- 47 How to use your LoRAs in Automatic1111 Web UI
- 48 How to select your LoRA from the interface
- 49 How to generate same batch with correct seed, how batch seed is determined
- 50 How to install after detailer adetailer extension to improve faces in your generations automatically
- 51 After detailer extension enabled comparison results
- 52 How to get amazing likeness - realism having images of your trained subject,
- 53 How to find best amazing among thousands of generated images by using DeepFace AI similarity script
