CoolFace
Modelpublic

ussoewwin/Hybrid-Sensitivity-Weighted-Quantization-SDXL-ConvRot-NVFP4

sourceHugging Faceotherupdated 2mo agoView on Hugging Face
0likes
Model Card

Hybrid-Sensitivity-Weighted-Quantization (HSWQ)

<p align="center"> <img src="https://raw.githubusercontent.com/ussoewwin/Hybrid-Sensitivity-Weighted-Quantization/main/icon.png" width="128"> </p>

High-fidelity ConvRot NVFP4 quantization for diffusion models (SDXL). HSWQ uses sensitivity and importance analysis instead of naive uniform cast. This is highly useful for users who need to strictly manage their VRAM resources while maintaining maximum image quality.

ComfyUI Load Diffusion Model nvfp4 pack with FULL ConvRot (Linear→NVFP4, Conv2d→INT8 int8_tensorwise) after DualMonitor + V4 pack-MSE FP16 protection under a fixed 600 MiB budget. Keep ratio is 0 (r0); calib writes NVFP4 .input_scale. SDXL pack scripts: hswq_convert_nvfp4_1.0.py (HSWQ) and native_convert_nvfp4.py (native).

Technical details: https://github.com/ussoewwin/Hybrid-Sensitivity-Weighted-Quantization

How to quantize (SDXL ConvRot NVFP4): md/How to quantize SDXL NVFP4.md

ComfyUI Loader for ConvRot NVFP4: To use these models in ComfyUI, please use this custom node: ComfyUI-HSWQ-Loader-and-Tools

SDXL ConvRot NVFP4 Benchmark Test Results (published tables): benchmark result/benchmark_convrotnvfp4.md

Benchmark (Reference)

ModelSSIM (Avg)File sizeCompatibility
Original FP161.0000100%High
HSWQ ConvRot NVFP40.92-0.9860% (FP16 mixed)High (ComfyUI NVFP4)

📦 Available Models

FilenameBase ModelVersionLicense
animemix_v80_hswq_r32_1off_nvfp4.safetensorsAnimeMixv8.0Fair AI Public License 1.0-SD
epicrealismXL_pureFix_hswq_r32_1off_convrot_nvfp4.safetensorsepiCRealism XLpureFixCreativeML Open RAIL++-M
koronemixIllustrious_v70_hswq_r32_1off_convrot_nvfp4.safetensorskoronemixIllustriousv70Fair AI Public License 1.0-SD
koronemixVpred_v20_hswq_r32_1on_convrot_nvfp4.safetensorskoronemixVpredv2.0CreativeML Open RAIL++-M
perfectionRealisticILXL_80_r32_1on_convrot_nvfp4.safetensorsPerfection Realistic ILXLv8.0Fair AI Public License 1.0-SD
realvisxlV50_v40Bakedvae_hswq_r32_1off_nvfp4.safetensorsRealVisXL V5.0 (Lightning)v4.0 BakedVAECreativeML Open RAIL++-M
realvisxlV50_v50Bakedvae_hswq_r32_1off_convrot_nvfp4.safetensorsRealVisXL V5.0 (Lightning)v5.0 BakedVAECreativeML Open RAIL++-M
unholyDesireMixSinister_v80_hswq_r32_1off_nvfp4.safetensorsUnholy Desire Mix Sinisterv8.0Fair AI Public License 1.0-SD
waiIllustriousSDXL_v170_hswq_r32_1off_convrot_nvfp4.safetensorsIllustrious-XL v1.7 (WAI-illustrious-SDXL)v17.0 (HF weight)Fair AI Public License 1.0-SD

📜 Credits & License

🏆 Special Acknowledgement

We extend our deepest respect and gratitude to the Nunchaku Team for their groundbreaking work on SVDQ quantization and for sharing their models with the community. This collection relies heavily on their research and original implementation.

Base Models

These models are derivatives of their respective creators. All credit for aesthetic tuning and model training belongs to the original creators.

  • —AnimeMix / koronemixIllustrious / koronemixVpred: Created by koronen.
  • —epiCRealism XL: Created by epinikion.
  • —Perfection Realistic ILXL: Created by 6tZ.
  • —RealVisXL V5.0: Created by SG_161222.
  • —Unholy Desire Mix Sinister: Created by UnholyDesiresStudio.
  • —WAI-illustrious-SDXL: Created by WAI0731.

Disclaimer: These models are provided for optimization and research purposes. Please adhere to the original licenses of the base models.