SDXL Models GGUF

🚀 Quantized SDXL Models in GGUF Format

This repository offers a small selection of quantized SDXL models in GGUF format. These models are intended for use with the custom ComfyUI nodes from ComfyUI - GGUF.

✨ Features

• Multiple SDXL Models: It includes a variety of well - known SDXL models such as CyberRealistic XL v4, Juggernaut v11, etc.
• Quantized for Efficiency: The models are quantized using llama - quantize to potentially improve performance.
• Additional Resources: Comes with basic clip - l and clip - g, and an improved SDXL VAE (xlVAEC_c91).

📚 Documentation

Model List

Quantization Details

These models are quantized using llama - quantize from llama.cpp b3600 patched with lcpp.patch.

A few models are quantized with llama.cpp b3962 patched with lcpp_sd3.patch instead. This may yield slightly different results compared to the default b3600 - quantized ggufs, but not necessarily better. These are marked in the filenames, and most users won't need these.

Quantization Types

It appears that Q4_0 often produces good results for SDXL models, usually coherent but not always consistent with those from an unquantized model. Q5_K_M is also included because it sometimes generates results that are clearly closer to the original F16 than Q4.

Additional Resources

The repository includes basic clip - l and clip - g, as well as an improved SDXL VAE called xlVAEC_c91, which can be found here. It is recommended as it seems to offer better colors, contrast, and details compared to the base version.

Request Handling

If model creators or other users want any GGUF files to be added or removed, requests will be considered upon receipt.

Other User's Uploads

There are also quantized models uploaded by another user at Old - Fisherman/SDXL_Finetune_GGUF_Files.