I followed all the examples in this post about using the saved_model_cli to take use tensorflow 1.13.1 saved_model_cli to optimize resnet50 with tensorRT. It worked great.
Then I tried to do the same thing with one of our tensorflow object detection models. The optimization succeeds and produces output graph, but when I host it in TF Serving, I am getting a RPC error.
I suspect I need to change some of the paraments to saved_model_cli, but there doesn’t seem to be up to date documentation anywhere on the new convert option.
Reading through the code here: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/tools/saved_model_cli.py