Hands-On Intelligent Agents with OpenAI Gym

Using TensorBoard for logging and visualizing a PyTorch RL agent's progress

Though TensorBoard is a tool that was released for the TensorFlow deep learning library, it is a flexible tool in itself, which can be used with other deep learning libraries like PyTorch. Basically, the TensorBoard tool reads the TensorFlow events summary data from log files and updates the visualizations and plots periodically. Fortunately, we have a library called tensorboardX that provides a convenient interface to create the events that TensorBoard can work with. This way, we can easily generate the appropriate events from our agent training code to log and visualize how our agent's learning process is progressing. The use of this library is pretty straightforward ...