To evaluate the model, we were able to load command-line arguments with TensorFlow's app flags, load the model and vocabulary processor, and then run the processed data through the model and make a prediction.
Remember to run this script through the command line, and check that the training script is run before creating the model and vocabulary dictionary.