v1.7.1

• Fix saving and loading TensorRT model for non mod-8 dimensions and tile size.

v1.7.0

• Add a check to make sure model files have been downloaded.
• Only VS API4 is supported now.
• Add support for TensorRT to accelerate inference.

v1.6.0

• Fix potential artifacts in 'deblock' task due to out of range pixel values.

v1.5.0

• Add gray model for GRAYS format.
• Replace no_grad mode with inference mode.

v1.4.1

• Fix tiling.

v1.4.0

• Fix mod-8 handling.
• Add tiling.

v1.3.0

• Add support for VS API4.

v1.2.0

• Add fp16 parameter.

v1.1.0

• Enable cudnn.benchmark for better performance.

v1.0.0

• Initial release.