pantheonrl.common.util.action_from_policy
- action_from_policy(obs, policy)[source]
Return the action, values, and log_probs given an observation and policy
- Parameters:
obs (ndarray) – Numpy array representing the observation
policy (ActorCriticPolicy) – The actor-critic policy
- Returns:
The action, values, and log_probs from the policy
- Return type:
Tuple[ndarray, Tensor, Tensor]