I’m relatively new to TensorRT so maybe the answer will strike you as evident. For research purposes I am building Neural Networks which perform dropout on inference using Pytorch.
Now I would like to embed these networks in a Jetson and, to improve performance, I would like to convert them to TensorRT engines. The issue is, I am not entirely sure this is supported, so I am asking if anyone has tried converting a Neural Network with dropout enabled into a TensorRT engine, and what the mandatory steps are.