Optimize model and use zero-copy runners

In this example, we show how to Model Navigator Inplace Optimize to run optimized models in place of the PyTorch models in the original pipeline. Depending on the mode, the optimize.py script can run PyTorch text summarization BART pipeline or optimize and run BART models in TensorRT without any changes to the original pipeline.

We recommend running this example in NVIDIA NGC PyTorch container. The Python script optimize.py wraps the Python model using Inplace Optimize and then runs profiling.

To run the optimization and profiling, run the script:

./optimize.py