Skip to content

3 ~ Visualization and evaluation

Jackal93 edited this page Apr 19, 2018 · 1 revision

Visualization

In order to see the agent playing in a GUI, you need to install python-tk/curses.

To start the visualization, use the following command:

bash visualize.sh

Please note that a saved checkpoint is required.

Evaluation

Execute the command:

bash evaluate.sh

This evaluates the policy of an agent loaded from a checkpoint, which is required.

Clone this wiki locally