hollowstrawberry/kohya-colab

Accessible Google Colab notebooks for Stable Diffusion Lora training, based on the work of kohya-ss and Linaqruf

What it solves

This project provides accessible, easy-to-use Google Colab notebooks for training Stable Diffusion LoRA (Low-Rank Adaptation) models, removing the technical barriers for users who lack powerful local hardware. It simplifies the process of preparing datasets and training the model weights.

How it works

The project consists of a set of notebooks that leverage the work of kohya-ss and Linaqruf. It provides two main workflows:

  1. Dataset Maker: A tool for gathering and preparing images. It can scrape images from Gelbooru, find duplicates using FiftyOne, manually curate images via a visual interface, generate tags using Waifu Diffusion 1.4 Tagger, and create captions using the BLIP model. It also allows for bulk editing of tags in text files.
  2. Lora Trainer: Dedicated notebooks for training LoRA and LoCon models for both standard Stable Diffusion and Stable Diffusion XL (SDXL), supporting various GPU configurations (T4, A100) and integrating directly with Google Drive for storage.

Who it’s for

Artists and AI image generation enthusiasts who want to create custom LoRA models for Stable Diffusion without needing a high-end GPU locally.

Highlights

  • Integrated Dataset Pipeline: Includes tools for scraping, deduplication, and captioning/tagging of images.
  • SDXL Support: Dedicated trainers for Stable Diffusion XL, including anime-specific model options.
  • Cloud-Based Training: Runs entirely in Google Colab, making it high-performance training accessible to the free tier (T4) or paid tiers (A100).
  • Google Drive Integration: Works natively with Google Drive for the dataset and trained models.

Related

  • Project
    Akegarasu/lora-scripts

    A GUI and script preset collection for training Stable Diffusion LoRA and Dreambooth models, providing an integrated environment for dataset tagging and training.

  • Dispatch
    Using LoRA for Efficient Stable Diffusion Fine-Tuning

    Hugging Face has integrated Low-Rank Adaptation (LoRA) into the diffusers library, enabling faster Stable Diffusion fine-tuning with significantly lower VRAM requirements and model weights as small as 3 MB.

  • Dispatch
    SDXL and Stable Diffusion Fast Inference with Latent Consistency LoRAs

    Hugging Face introduces LCM LoRAs, a method to enable high-quality image generation in 4 to 8 steps for SDXL and Stable Diffusion models, significantly reducing inference time.

  • Project
    unslothai/notebooks

    A collection of Google Colab notebooks that demonstrate how to fine‑tune, run inference and experiment with many open‑source LLMs (Gemma, Llama 3, Qwen, Mistral, Phi‑4, etc.) using the Unsloth library. Includes chat UI, vision/audio/embedding demos, and reinforcement‑learning/GRPO training notebooks, all runnable for free in the browser.

  • Project
    sokrypton/ColabFold

    A tool that makes protein folding predictions accessible via Google Colab and local installations, integrating models like AlphaFold2 and AlphaFold3 to predict 3D protein structures.