Giters

Hakureirm / stable_diffusion_chilloutmix_ipynb

AUTOMATIC1111 Stable Diffusion WebUI 1.5 + ChilloutMix + Kohya's Scripts

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

AUTOMATIC1111 Stable Diffusion WebUI 1.5 + ChilloutMix + Kohya's Scripts

[Web UI] Go To Colab | [LoRA Train] Go To Colab | [Xformers Build] Go To Colab | Prompts Collection

It does not come with any Loras, you need to type in CivitAI to download it yourself, and upload it to the models/Stable-diffusion folder. If you enable CiivitAIExtension, you can download it directly from the web UI, but remember to check Save add Additional Network option.

Warning: Please note that there may be legal risks before painting. Please note whether your painting may cause controversy and negative impact on the development of the AI community. Please use it with caution.

Roadmap

  • Paperspace Support
  • Support for training LoRA Jupyter Notebook
  • Optimize model download speed with aria2
  • Internationalization
  • Provide xformer support for Paperspace M4000 GPU

Jupyter Notebook's Language is still Chinese, I want to find a way to make it internationalized, but maintaining two versions is a bit troublesome, so I will do it later.

About Training LoRA

Basic training script based on Akegarasu/lora-scripts which is based on kohya-ss/sd-scripts, but you can also use ddPn08/kohya-sd-scripts-webui which provides a GUI, it is more convenient, I also provide the corresponding SD WebUI extension installation method in stable_diffusion_1_5_webui.ipynb.

Some models are not compatible with the training script, and the training script will not be able to train them. If you want to train them, you need to use the training script provided by the model author.

In kohya-ss/sd-scripts train_network documentaion, it notes that the cloneofsimo's repository and the d8ahazard's Dreambooth Extension for Stable-Diffusion-WebUI are not compatible at the current time. This is because they have extended some features.

Recommended Loras

Recommended Checkpoints

Recommended Textual Inversion

Recommended Experimental LoRA

Experimental LoRA refers to models that are not yet fully tested by me, and have not yet been determined to be compatible. If you want to use these models, please judge the compatibility with other LoRA by yourself.

NoteBook Built-in Extension

WebUI Jupyter Notebook built-in Extension, you can install it directly from the notebook.

Build xformers from source

If you are using Google Colab, there are many precompiled wheels for you to choose from. If you are using something else or want to build from source, you can use this notebook to build the library.

I have written a notebook to build xformers from source. You can use it to build xformers for your own GPU.

CHANGELOG

2023.2.25

  • README.md
    • Update Experimental Lora List - Add ChilloutMixss
    • Optimize CHANGELOG format
    • Internationalize README
    • New "Textual Inversion" section
    • Add "NoteBook Built-in Extension" section
    • Add "Build xformers from source" section
  • lora_train.ipynb
    • Fix serious errors in LoRA training script
    • Remove unnecessary platform and GPU checks
    • Add export_model_dir option to specify the directory where the model is exported
    • Add an extra build installation xformers option for the M4000 GPU
    • Use ninja to build xformers much faster ( Followed by Official README)
  • stable_diffusion_1_5_webui.ipynb
    • Add an extra build installation xFormer option for the M4000 GPU
    • Use ninja to build xformers much faster ( Followed by Official README)
  • prompts.md
    • Internationalize
  • xformers_build.ipynb
    • New xformers build script
    • Switch xformers source version to 0.0.16
  • workflows
    • Add xformers_build.ipynb to Preview CI
    • Remove pull_request_target value to prevent the workflow from running twice
  • Release
    • Publish xformers precompiled wheels for M4000 GPU

2023.2.24

  • workflows
    • Automatically comment on Preview link when Pull Request ( wibus-wee#2 )
  • stable_diffusion_1_5_webui.ipynb
    • Install controlnet, openpose-editor, Kohya sd-scripts extension with Stable-Diffusion-WebUI
    • Remove the old training model installation scheme and merge it into the installation steps ( created in 2023.2.19 )
  • lora_train.ipynb
    • New Lora Train Script ( Alpha )

2023.2.21

  • stable_diffusion_1_5_webui.ipynb
    • Better option selection mechanism
    • Fine distinction between Paperspace and Colab platforms rather than relying on a single Free GPU model

2023.2.19

  • stable_diffusion_1_5_webui.ipynb
    • Fix the problem of incorrectly identifying the Quadro M4000 GPU
    • Add Training Model Installation Scheme ( Power by d8ahazard/sd_dreambooth_extension)
    • Optimize checkpoint selection
    • Fix the problem of not being able to read variables
    • Automatically check whether webUI and training extensions are installed

2023.2.18

  • stable_diffusion_1_5_webui.ipynb
    • Update the latest version of the upstream ChilloutMix
    • Update Prompts Collection
    • Compatible with Paperspace platform (Free GPU)
    • Export all generated images
    • Use nvidia-smi to view GPU information
    • Check if GPU supports Xformers
    • Optimize variable passing

2023.2.17

  • stable_diffusion_1_5_webui.ipynb
    • Add SunshineMix Checkpoint as a second choice for 2.5D painting

2023.2.16

  • stable_diffusion_1_5_webui.ipynb
    • Because there may be a sudden exit in the middle, there is another way to restart

Created by @wibus-wee

Reference: camenduru/stable-diffusion-webui-colab

About

AUTOMATIC1111 Stable Diffusion WebUI 1.5 + ChilloutMix + Kohya's Scripts

Languages

Language:Jupyter Notebook 100.0%