Hands-On Intelligent Agents with OpenAI Gym

Tools to save/load, log, visualize, and monitor

In the previous sections, we walked through the core part of the agent's learning algorithm implementation. Apart from those core parts, there are a few utility functions that we will use to train and test the agent in different learning environments. We will reuse the components that we already developed in Chapter 6, Implementing an Intelligent Agent for optimal discrete control using Deep Q Learning,such as the utils.params_manager, and also the save() and load() methods, which respectively save and load the agent's trained brain or model. We also will make use of the logging utilities to log the agent's progress in a format that is usable with Tensorboard for a niceand quick visualization, ...