SDXL Refiner 1.0 ConvRot (INT8, INT6 & INT4) for ComfyUI

ConvRot-quantized single-file ComfyUI checkpoints of stabilityai/stable-diffusion-xl-refiner-1.0 in three clean bit widths: Full INT8, Full INT6 (w6a8_int8), and Full INT4 (asym_w4a8_int8).

Built with ConvRot (block Hadamard rotation) + Act-Order GPTQ Hessian error compensation + through-grid FP8 group-scale search / Lloyd-Max codebooks + calibration bias correction, keeping Conv2d layers, text encoders (CLIP-L / OpenCLIP-G), and VAE unquantized in FP16.

Related SDXL ConvRot repositories:


Available Checkpoints

Precision Tier File (ComfyUI/models/checkpoints/) Size (vs 6.08 GB FP16) Quantized UNet Linears
Full INT8 ConvRot sdxl-refiner-1.0-int8-convrot.safetensors 4.64 GB (-23.6%) 484 / 484 INT8
Full INT6 ConvRot sdxl-refiner-1.0-int6-convrot.safetensors 4.37 GB (-28.0%) 483 / 484 INT6 (W6A8) + 1 INT8 (K=640/det)
Full INT4 ConvRot sdxl-refiner-1.0-int4-convrot.safetensors 4.02 GB (-33.9%) 483 / 484 INT4 (W4A8) + 1 INT8 (K=640/det)

All files load directly in stock ComfyUI (Load Checkpoint, ComfyUI >= v0.45.0) using native comfy-kitchen CUDA kernels โ€” no extra nodes or unpacking required. (For SDXL Refiner, 1 architecture-detection key remains in ConvRot INT8 so stock ComfyUI Load Checkpoint detects the Refiner architecture cleanly.)


Visual Comparisons (40 steps (Base+Refiner handoff at 0.8), normal, 1024x1024)

Refiner Comparison p0 Refiner Comparison p2 Refiner Comparison p6


Evaluation Metrics (8 prompts x 1024x1024, ComfyUI v0.45.3)

Measured via scripts/verify_sdxl_quant.py directly against unquantized FP16 (sd_xl_refiner_1.0.safetensors):

  • Unquantized BF16 Reference (vs FP16): Cosine 0.99991 (min 0.99987), SSIM 0.99072 (min 0.98573), PSNR 42.85 dB (min 39.97 dB)
Precision Tier Cosine Mean / Min SSIM Mean / Min PSNR Mean / Min Warm Speed/Img Report
Full INT8 ConvRot 0.99992 / 0.99985 0.99280 / 0.98939 43.72 / 41.02 dB 0.53s g4_refiner_int8_summary.json
Full INT6 ConvRot 0.99985 / 0.99973 0.99041 / 0.98686 42.08 / 40.52 dB 0.56s g4_refiner_int6_summary.json
Full INT4 ConvRot 0.99904 / 0.99849 0.97662 / 0.96690 37.68 / 35.40 dB 0.57s g4_refiner_int4_summary.json

How to Use in ComfyUI

  1. Download INT8, INT6, or INT4 .safetensors into ComfyUI/models/checkpoints/.
  2. Load directly with ComfyUI's built-in Load Checkpoint node (ComfyUI >= v0.45.0).

License & Use Restrictions (CreativeML Open RAIL++-M)

This checkpoint is a Derivative of the Model under the CreativeML Open RAIL++-M License (Copyright ยฉ 2023 Stability AI). See NOTICE for full attributions and modification details, and pass through LICENSE.md and NOTICE with any redistribution.

Downloads last month
56
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for SUSSYEATING/SDXL-Refiner-1.0-ConvRot

Quantized
(3)
this model