Hi, I find some differences, may be errors, between your overcooked_env.py and the official repo's overcooked_env.py.
In your codes:
def featurize_state_mdp(self, state):
"""
Wrapper of the mdp's featurize_state
"""
return self.mdp.featurize_state(state, self.mlam, self.horizon)
self.horizon is passed into mdp.featurize_state function which the position should be param "num_pots" in the official repo:
def featurize_state_mdp(self, state, num_pots=2):
"""
Wrapper of the mdp's featurize_state
"""
return self.mdp.featurize_state(state, self.mlam, num_pots=num_pots)
This causes the state/obs dim is up to over 10000 (because the horizon you set is 500).
I think there may be something wrong in your provided codes.
Hi, I find some differences, may be errors, between your overcooked_env.py and the official repo's overcooked_env.py.
In your codes:
self.horizon is passed into mdp.featurize_state function which the position should be param "num_pots" in the official repo:
This causes the state/obs dim is up to over 10000 (because the horizon you set is 500).
I think there may be something wrong in your provided codes.