pearsonkyle/SDXL-Model-Merger

SDXL Model Merger

A Gradio-based web application for merging, generating with, and exporting Stable Diffusion XL (SDXL) checkpoints.

Features

• —Load pipelines from HuggingFace URLs with optional VAE and multiple LoRAs
• —Generate images with seamless tiling support for panoramic/360° outputs
• —Export merged models with quantization options (int8, int4, float8)

Usage on HuggingFace Spaces

This app is optimized for both local and Space deployments:

# Local deployment
python app.py

# Space deployment with CPU fallback
export DEPLOYMENT_ENV=spaces
python app.py

For best results:

• —Use GPU (NVIDIA) for fast generation - ~8GB VRAM recommended
• —CPU mode is available but will be slower and use more RAM (~16GB+)

Requirements

• —Python 3.10+
• —PyTorch 2.0+
• —4GB+ VRAM (GPU) or 16GB+ RAM (CPU)
• —~2GB disk space for cached models