🚀 T2ITrainer

⚠️ Development Notice: Currently in active development - stability not guaranteed. Frequent updates - check changelogs regularly.

---

📅 Recent Updates

• 2025-02-15: 🎨 Launched Flux Fill Training script for LoRA-based inpainting

---

🛡️ Prerequisites

• PyTorch: torch>=2.3.0+cu121 (CUDA 12.1 supported)

---

💻 Supported Training Configurations

Model Type	VRAM Requirements	Status
Kolors	11GB GPU	✅ Supported
SD3.5 (FP16 BS1)	24GB GPU	✅ Supported
Flux, Flux Fill	24GB GPU	✅ Supported

---

⚙️ Installation Guide

0. System Requirements

❗ Mandatory: Install Microsoft Visual C++ Redistributable if encountering DLL errors

1. Automated Setup

Recommended Method

  git clone https://github.com/lrzjason/T2ITrainer.git
  cd T2ITrainer
  setup.bat

• Handles: Virtual Environment • Dependency Installation • Model Downloads

2. Manual Installation

Clone Repository 🌐

    git clone https://github.com/lrzjason/T2ITrainer.git
    cd T2ITrainer

Virtual Environment 🛠️

    python -m venv venv
    call venv\Scripts\activate
    pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

Model Downloads 📥

    # Kolors
    huggingface-cli download Kwai-Kolors/Kolors --local-dir kolors_models/

    # Flux Fill
    huggingface-cli download "black-forest-labs/FLUX.1-fill-dev" --local-dir flux_models/fill/
    
    # SD3.5 Models
    huggingface-cli download "stabilityai/stable-diffusion-3.5-large" --local-dir "sd3.5L/"

🚀 Launch Options

Script	Command	Special Notes
Flux Fill	python ui_flux_fill.py	Requires diffusers>=0.32.0, 24GB VRAM Recommended
Kolors	python ui.py	Needs Fixed VAE
SD3.5 Large	python ui_sd35.py	24GB VRAM Recommended

🔧 Testing & Integration

• Kolors Workflow:

      # ComfyUI Plugins
      git clone https://github.com/kijai/ComfyUI-KwaiKolorsWrapper
      git clone https://github.com/MinusZoneAI/ComfyUI-Kolors-MZ

• Configuration Guide: 📖 CivitAI Article

🆘 Troubleshooting

• Kolors Black Image Issue: Ensure you're using FP16 Fixed VAE
• VRAM Limitations: Adjust blocks_to_swap parameter (higher values reduce memory usage)
• Windows DLL Errors: Verify VC++ Redistributable installation

---

🔧 Parameter Configuration Guide

---

🌌 Flux Model Management

Core Model Download

Base Model Installation

  huggingface-cli download "black-forest-labs/FLUX.1-dev" --local-dir flux_models/dev/

🎨 Flux Fill Special Requirements

Dependency Alert

• Mandatory: diffusers>=0.32.0

Fill Model Installation

Inpainting Model Setup

  huggingface-cli download "black-forest-labs/FLUX.1-fill-dev" --local-dir flux_models/fill/

⚙️ Flux Fill Training Recommended Parameters

Category	Settings
Base Configuration	Rank 32, AdamW, Learn Rate 1e-4
24GB GPU	512 resolution, Batch Size 1
VRAM Optimization	blocks_to_swap: 10 (15-20 for lower VRAM GPUs)
Precision	bf16 (3090) / fp8 (40xx)

For more details (example dataset structure):

• https://github.com/lrzjason/T2ITrainer/blob/main/doc/flux_fill.md

💻 VRAM Usage (bf16, blocks_to_swap=10)

VRAM Peak

VRAM Low

Star History

Old Change logs:

• https://github.com/lrzjason/T2ITrainer/blob/main/doc/change_logs.md

Recent Change Logs:

• 2025-02-15: Released Flux Fill Training script which could train lora for flux fill.

Sponsor:

• Thanks to @sourceful support me making flux fill lora training script.

📬 Contact

• 𝕏 Twitter: @Lrzjason
• 📧 Email: lrzjason@gmail.com
• 💬 QQ Group: 866612947
• 💬 WeChat ID: fkdeai
• 🎨 CivitAI: xiaozhijason

Sponsors me for more open source projects:

Buy me a coffee:

WeChat:

Acknowledgements:

• Thanks to minienglish1 and Freon in EveryDream Discord for the assistance.
• Special thanks to kohya ss for references from the training codebase.
• Thanks to Kblueleaf for coding reference on hunyuandit gradient checkpoint implementation.
• Thanks to Kolors for the open-source checkpoint.
• Thanks to comfyui for the wonderful codebase.
• Thanks to emojiiii for the setup.bat script and other updates.
• Thanks to Rohit Gandikota and related authors of Concept Sliders https://github.com/rohitgandikota/sliders