eugenevinitsky / sequential_social_dilemma_games

Repo for reproduction of sequential social dilemmas
MIT License
387 stars 132 forks source link

Replay is broken for custom models #71

Closed eugenevinitsky closed 5 years ago

eugenevinitsky commented 5 years ago

Custom models must be registered to be re-used in RLlib via: ModelCatalog.register_custom_model("", ). However, when we replay things we don't have this information anymore. Consider storing this as a tune function in the env config so that it can be recreated later.

eugenevinitsky commented 5 years ago

Temporarily fixed with a hack where we assume that the model is always the same

eugenevinitsky commented 5 years ago

New bug: you can't call agent.compute_action( state[agent_id], policy_id=policy_map_fn(agent_id)) straightforwardly with an LSTM