How to get and save State-Action-Rewards tuples ?

[poyobe]

Hi,

I am using smaclite and mappo agents. I managed to train and test them and my little units are well trained and manage to beat the ennemies, Hurrah ! For an explainable RL project I would like to save the state-action-rewards tuples of episodes to be able to analyze their behavior in multiple scenarios. Is there a parameter in the config files to save entire episodes like that? If not, where in the code can I get these tuples ? I tried printing some stuff in the episode_runner run function and in the episode_buffer update function but I do not really understand how it works.

Thank you for your time