ComfyUI Extension: comfyui-sdnq

Authored by EnragedAntelope

Created

Updated

54 stars

ComfyUI custom node pack for loading SDNQ quantized models

Custom Nodes (0)

    README

    ComfyUI-SDNQ

    Load and run SDNQ quantized models in ComfyUI with 50-75% VRAM savings!

    This custom node pack enables running SDNQ (SD.Next Quantization) models directly in ComfyUI. Run large models like FLUX.2, FLUX.1, SD3.5, and more on consumer hardware with significantly reduced VRAM requirements while maintaining quality.

    SDNQ is developed by Disty0 - this node pack provides ComfyUI integration.

    Features

    • 🎨 Standalone Sampler: All-in-one node - load model, generate images, done
    • 📦 Model Catalog: 30+ pre-configured SDNQ models with auto-download
    • 💾 Smart Caching: Download once, use forever
    • 🚀 VRAM Savings: 50-75% memory reduction with quantization
    • âš¡ Performance Optimizations: Optional xFormers, VAE tiling, SDPA (automatic)
    • 🎯 LoRA Support: Load LoRAs from ComfyUI loras folder
    • 📅 Multi-Scheduler: 14 schedulers (FLUX/SD3 flow-match + traditional diffusion)
    • 🔧 Memory Modes: GPU (fastest), balanced (12-16GB VRAM), lowvram (8GB VRAM)

    Installation

    Method 1: ComfyUI Manager (Recommended)

    1. Install ComfyUI Manager
    2. Search for "SDNQ" in the manager
    3. Click Install
    4. Restart ComfyUI

    Method 2: Manual Installation

    cd ComfyUI/custom_nodes/
git clone https://github.com/EnragedAntelope/comfyui-sdnq.git
cd comfyui-sdnq
pip install -r requirements.txt

    Restart ComfyUI after installation.

    Quick Start

    1. Add SDNQ Sampler node (under sampling/SDNQ)
    2. Select a model from dropdown (auto-downloads on first use)
    3. Enter your prompt
    4. Click Queue Prompt
    5. Done! Image output connects directly to SaveImage

    Defaults are optimized - select model, enter prompt, generate!

    Node Reference

    SDNQ Sampler

    Category: sampling/SDNQ

    Main Parameters:

    • model_selection: Dropdown with 30+ pre-configured models
    • custom_model_path: For local models or custom HuggingFace repos
    • prompt / negative_prompt: What to create / what to avoid
    • steps, cfg, width, height, seed: Standard generation controls
    • scheduler: FlowMatchEulerDiscreteScheduler (FLUX/SD3) or traditional samplers

    Memory Management:

    • memory_mode:
      • gpu = Full GPU (fastest, 24GB+ VRAM required)
      • balanced = CPU offloading (12-16GB VRAM)
      • lowvram = Sequential offloading (8GB VRAM, slowest)
    • dtype: bfloat16 (recommended), float16, or float32

    Performance Optimizations (optional):

    • use_xformers: 10-45% speedup (safe to try, auto-fallback to SDPA)
    • enable_vae_tiling: For large images >1536px (prevents OOM)
    • SDPA (Scaled Dot Product Attention): Always active - automatic PyTorch 2.0+ optimization

    LoRA Support:

    • lora_selection: Dropdown from ComfyUI loras folder
    • lora_custom_path: Custom LoRA path or HuggingFace repo
    • lora_strength: -5.0 to +5.0 (1.0 = full strength)

    Outputs: IMAGE (connects to SaveImage, Preview, etc.)

    Available Models

    30+ pre-configured models including:

    • FLUX: FLUX.1-dev, FLUX.1-schnell, FLUX.2-dev, FLUX.1-Krea, FLUX.1-Kontext
    • Qwen: Qwen-Image variants (Edit, Lightning, Turbo)
    • SD3/SDXL: SD3-Medium, SD3.5-Large, NoobAI-XL variants
    • Others: Z-Image-Turbo, Chroma1-HD, HunyuanImage3, Video models

    Most available in uint4 (max VRAM savings) or int8 (best quality). Browse: https://huggingface.co/collections/Disty0/sdnq

    Performance Tips

    For All Memory Modes:

    • SDPA (Scaled Dot Product Attention) is always active - automatic PyTorch 2.0+ optimization
    • Enable use_xformers=True for 10-45% additional speedup (safe to try)
    • Use enable_vae_tiling=True for large images (>1536px) to prevent OOM

    Scheduler Selection:

    • FLUX/SD3/Qwen/Z-Image: Use FlowMatchEulerDiscreteScheduler
    • SDXL/SD1.5: Use DPMSolverMultistepScheduler, EulerDiscreteScheduler, or UniPCMultistepScheduler
    • Wrong scheduler = broken images!

    Model Storage

    Downloaded models are stored in:

    • Location: ComfyUI/models/diffusers/sdnq/
    • Format: Standard diffusers format

    Models are cached automatically - download once, use forever!

    Troubleshooting

    xFormers Not Working

    If you see "xFormers not available" but have it installed:

    • This is usually fine - the node automatically falls back to SDPA (PyTorch 2.0+ default)
    • SDPA provides good performance without xFormers
    • If xFormers is incompatible with your GPU/model, fallback is automatic

    Performance is Slow

    Balanced/lowvram modes: Inherently slower due to CPU↔GPU data movement. Options:

    • Enable use_xformers=True (10-45% speedup if compatible)
    • SDPA is always active for automatic optimization
    • Upgrade to more VRAM for full GPU mode
    • Use smaller model (uint4 vs int8)

    Out of Memory

    1. Use lower memory mode (gpu → balanced → lowvram)
    2. Use more aggressive quantization (uint4 instead of int8)
    3. Reduce resolution or batch size
    4. Enable enable_vae_tiling=True for large images

    Model Loading Fails

    1. Check internet connection (for auto-download)
    2. Verify repo ID is correct for custom models
    3. For local models, ensure path points to directory (not a file)
    4. Check model is actually SDNQ-quantized (from Disty0's collection)

    Contributing

    Contributions welcome! Please:

    1. Follow existing code style
    2. Test with multiple model types
    3. Update documentation for new features

    License

    Apache License 2.0 - See LICENSE

    This project integrates with SDNQ by Disty0.

    Credits

    SDNQ - SD.Next Quantization Engine

    • Author: Disty0
    • Repository: https://github.com/Disty0/sdnq
    • Pre-quantized models: https://huggingface.co/collections/Disty0/sdnq

    This node pack provides ComfyUI integration for SDNQ. All quantization technology is developed and maintained by Disty0.