wbruna/upscalers-sdcpp-gguf

Upscaler models from https://openmodeldb.info/ , converted to GGUF format for stable-diffusion.cpp.

stable-diffusion.cpp can work with most .pth files directly, but gguf models can be faster to load, and avoid security issues with the .pth format.

See the original URLs for licensing information. This is a straight mechanical conversion, so each model keeps the same license as the original.

GGUF	original
2x-ESRGAN.gguf	https://openmodeldb.info/models/2x-ESRGAN
4xLSDIRplusC.gguf	https://openmodeldb.info/models/4x-LSDIRplusC
4xLSDIRplusR.gguf	https://openmodeldb.info/models/4x-LSDIRplusR
4x_Nickelback_70000G.gguf	https://openmodeldb.info/models/4x-Nickelback
4x_NickelbackFS_72000_G.gguf	https://openmodeldb.info/models/4x-NickelbackFS
4xNomos8kSC.gguf	https://openmodeldb.info/models/4x-Nomos8kSC
4xNomosWebPhoto_esrgan.gguf	https://openmodeldb.info/models/4x-NomosWebPhoto-esrgan
RealESRGAN_x4plus.gguf	https://openmodeldb.info/models/4x-realesrgan-x4plus
RealESRGAN_x4plus_anime_6B.gguf	https://openmodeldb.info/models/4x-realesrgan-x4plus-anime-6b
4x_foolhardy_Remacri_ExtraSmoother.gguf 4x_foolhardy_Remacri.gguf	https://openmodeldb.info/models/4x-Remacri
4xUltrasharpV10.gguf	https://openmodeldb.info/models/4x-UltraSharp

The models were converted with stable-diffusion.cpp itself:

./sd-cli --mode convert --model modelfile.pth --output modelfile.gguf

A few old .pth files needed to have their format updated first:

import torch
checkpoint = torch.load(r"model.pth",map_location="cpu")
torch.save(checkpoint,r"model_converted.pth",_use_new_zipfile_serialization=True)