TorchRT provides a simple way to convert a PyTorch nn.Module to a TensorRT-equipped module. While the goal of this package is similar to that of torch2trt from NVIDIA, TorchRT does this by first converting the module to ONNX, allowing options to optimize the model graph without the need to maintain conversion code for each operator.