v5.5.1

• Change default sc_threshold to 0.15.
• Redirect stdout to stderr so as to avoid corrupted pipe due to warning messages from TensorRT.

v5.5.0

• Add 4.25.heavy and 4.26.heavy models.
• Add auto_download parameter to allow only download the specified model at first run rather than forcibly download all models at once.
• Usage of reducing frame rate is specifically rejected now.
• Remove pre-allocated timestep tensors to avoid potential OOM due to large factor_num.

v5.4.1

• Fix cached tensors do not get deleted when they are detected as scene changes.
• Change default sc parameter to False.
• Fix missing encode_channel variable for 4.13 model.
• Remove manually allocated pinned memory to fix potential high CPU utilization on some systems. May be also marginally faster.

v5.4.0

• Add 4.25.lite model.
• Fix error when sys.stdout is None in some circumstances.
• Improve performance by using separate streams for transfering tensors between CPU and GPU.
• Improve performance by caching Head module output.
• Remove num_streams parameter since it doesn't provide any benefit now due to the above two improvements.
• Re-fix jaggy/wobbling line issue by a significantly faster method than v5.3.0. Using trt=True now requires tensorrt>=10.4.0 and torch_tensorrt>=2.6.0.dev20241011 to fully work. Due to some issues in Torch-TensorRT, dynamic shapes support is temporarily disabled at the moment.

v5.3.1

• Fix bad_alloc on Linux in TorchTRT.

v5.3.0

• Add 4.19-4.26 models.
• Bump PyTorch and Torch-TensorRT to 2.6.0.dev.
• Change default model to 4.18.
• Add trt_static_shape parameter and default to True.
• Fix jaggy/wobbling line issue and broken 4K output. The performance is expected to be lower than before due to the changes.

v5.2.0

• Add 4.17.lite and 4.18 models.
• Add support for TensorRT dynamic shapes.
• Bump PyTorch to 2.5.0.dev.
• Bump Torch-TensorRT to 2.5.0.dev. Windows build is officially supported now, hence no more whl file on Releases.

v5.1.0

• Fix f-string error before Python 3.12.
• Add 4.17 model.
• Fix performance degradation due to unnecessary extra padding when scale < 1.

v5.0.0

• Add 4.12-4.16.lite models.
• Change default model to 4.15.lite.
• Minor performance improvement.
• 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.

v4.2.0

• Add 4.9-4.11 models.
• Remove upper limit of num_streams.