2022-07-27
You can also use the latest vs-mlrt release to experience RealCUGAN. vs-mlrt supports OpenVINO, ONNXRuntime and TensorRT runtimes and provides better performance & reduced resource consumption compared to the PyTorth implementation. It also includes some other ML models (e.g. DPIR, waifu2x) and a user friendly Python wrapper. vs-mlrt binary release package vsmlrt-windows-x64-cuda already includes everything you need to deploy those models; Alternatively, you can also use the pre-integerated VapourSynth Portable.
######official pytorch version ↓ ######
simple-upcunet.vpy is a simple demo.
You should append file the root of "upcunet_v3_vs.py"/"upcunet_v20220227_vs.py" into sys.path
RealWaifuUpScaler init setting：
- scale: 2x/3x/4x;
- weight_path：the path of the weights of model;
- device: cuda device number;
- tile_mode：0/1/2/3/4/5. The bigger the number, the less video memory is needed, and the lower inference speed it is.
v20220227 add cache_mode and alpha config - cache_mode: Default 0. Memory needed:0>1>>2=3, speed:0>1(+15%time)>2(+25%time)>3(+150%time). You can super resolve very large resolution images using mode2/3 (low memory mode).
- alpha: The smaller the number, the enhancement strength is smaller, more blurry the output images are; the bigger the number, the enhancement strength is bigger, more sharpen image will be generated. Default 1 (don't adjust it). Recommended range: (0.75,1.3)
Tested in (environment):
- PyTorch==1.10.1+cu111
- Python==3.9.5
- VapourSynth==R57
You should install PyTorch using pip, the version of PyTorch should be >=1.0.0 (for 30 series N card: >=1.9.0; A card is not supported)
VapourSynth version: >=R54-API4-test2