v2.0.0

• Remove nvfuser and cuda_graphs.
• Bump PyTorch to 2.4.0.dev.
• Bump VapourSynth to R66.
• Bump TensorRT to 10.0.1.
• Bump Torch-TensorRT to 2.4.0.dev.

v1.2.0

• Don't globally set default floating point tensor type when the input is of RGBH format.
• Slightly improve TensorRT performance by including more node types when the engine is built.

v1.1.0

• Remove unusable tile option.
• Set CUDA_MODULE_LOADING environment variable to avoid corrupted pipe.
• Bump TensorRT to 8.5.3.1.
• Silence meshgrid warning.
• Add trt_min_subgraph_size parameter.

v1.0.0

• Initial release.

model

Initial commit