Imitation Learning

[Nuno Bento]

Hi everyone. I am trying to do imitation learning, but can't get information about the actions.

In play_game.py, I see it gets an empty list and doesn't return anything about the actions.

_, _, done, _ = env.step([]) (line 60)

Is there an easy way to get the actions done by a human player?

Thanks in advance,

Nuno Bento

[stanczyk]

step() API method takes actions as an argument, it doesn't return it. In case of play_game script it doesn't use agent player by default, but the keyboard player (see players flag). Different players are implemented in here. You can see the actions along with all other data in the episode dump - see here.