Hybrid-Sensitivity-Weighted-Quantization (HSWQ)

High-fidelity FP8 quantization for diffusion models (Z Image). HSWQ uses sensitivity and importance analysis instead of naive uniform cast, and offers two modes: standard-compatible (V1) and high-performance scaled (V2).

Technical details: md/HSWQ_ Hybrid Sensitivity Weighted Quantization.md

Overview

Feature	V1: Standard Compatible	V2: High Performance Scaled
Compatibility	Full (100%), any FP8 loader	The scaled model does not perform well in the current ComfyUI.

File format	Standard FP8 (`torch.float8_e4m3fn`)	Extended FP8 (weights + `.scale` metadata)
Image quality (SSIM)	~0.89 (theoretical limit)	~Unable to measure at this time
Mechanism	Optimal clipping (smart clipping)	Full-range scaling (dynamic scaling)
Use case	Distribution, general users	In-house, max quality, server-side

File size is reduced by about 55% vs FP16 while keeping best quality per use case.

Architecture

Dual Monitor System — During calibration, two metrics are collected:
- Sensitivity (output variance): layers that hurt image quality most if corrupted → top 25% kept in FP16.
- Importance (input mean absolute value): per-channel contribution → used as weights in the weighted histogram.
Rigorous FP8 Grid Simulation — Uses a physical grid (all 0–255 values cast to torch.float8_e4m3fn) instead of theoretical formulas, so MSE matches real runtime.
Weighted MSE Optimization — Finds parameters that minimize quantization error using the importance histogram.

Modes

V1 (scaled=False): No scaling; only the clipping threshold (amax) is optimized. Output is standard FP8 weights. Use when you need maximum compatibility.
V2 (scaled=True): Weights are scaled to FP8 range, quantized, and inverse scale S is stored in Safetensors (.scale). Use with HSWQLoader for best quality.

Recommended Parameters

Samples: 256 (minimum for reliable stats; 128 is insufficient).
Keep ratio: 0.10 (10%) — keeps critical layers in FP16; 0.10 has higher degradation risk.
Steps: 20–25 — to include early denoising sensitivity.

Benchmark (Reference)

Model	SSIM (Avg)	File size	Compatibility
Original FP16	1.0000	100% (6.5GB)	High
Naive FP8	0.82-0.83	50%	High
HSWQ V1	0.88–0.89	55% (FP16 mixed)	High
HSWQ V2	Unable to measure at this time	55% (FP16 mixed)	Low (custom loader)

HSWQ V1 gives a clear gain over Naive FP8 with full compatibility; V2 targets maximum quality with a custom loader.

2. Setup

VAE: Use standard SDXL VAE (place in models/vae/)

Downloads last month: -; Downloads are not tracked for this model. How to track