v3.0.0

• Drop mmcv dependency.
• Change tile parameter to list.
• Add auto_download parameter to allow only download the specified model at first run rather than forcibly download all models at once.
• Bump PyTorch to 2.5.1.
• Bump VapourSynth to R66.

v2.1.0

• Upgrade to MMCV 2.0 so as to support Python 3.11 and PyTorch 2.0. Remember to uninstall mmcv-full first before mim install "mmcv>=2.0.0".
• Bump PyTorch to 2.0.
• Bump VapourSynth to R60.

v2.0.0

• Change the function name to lowercase.
• Add four models: NTIRE 2021 Video Super-Resolution, Video Deblurring (DVD), Video Deblurring (GoPro), and Video Denoising.
• Rename interval parameter to length.
• Rename tile_x and tile_y parameters to tile_w and tile_h respectively.
• Remove device_type parameter.
• Remove fp16 parameter. Now it's controlled by the format of the input.

v1.4.1

• Revert ea2a597 to keep compatibility with PyTorch 1.9.

v1.4.0

• The entire frame is not transferred to GPU in tiling mode now to reduce VRAM usage.
• Silence warnings in PyTorch 1.10.

v1.3.0

• Add cpu_cache parameter.

v1.2.0

• Add a check to make sure model files have been downloaded.
• Replace no_grad mode with inference mode.
• Update pre-compiled Windows binary of mmcv-full to 1.3.14.
• Silence warnings when using model 3-5.

v1.1.0

• Add support for VS API4.

v1.0.0

• Initial release.

model

Initial commit