garage.np package¶
Reinforcement Learning Algorithms which use NumPy as a numerical backend.
-
obtain_evaluation_samples
(policy, env, max_path_length=1000, num_trajs=100)[source]¶ Sample the policy for num_trajs trajectories and return average values.
Parameters: - policy (garage.Policy) – Policy to use as the actor when gathering samples.
- env (garage.envs.GarageEnv) – The environement used to obtain trajectories.
- max_path_length (int) – Maximum path length. The episode will terminate when length of trajectory reaches max_path_length.
- num_trajs (int) – Number of trajectories.
Returns: - Evaluation trajectories, representing the best
current performance of the algorithm.
Return type:
-
paths_to_tensors
(paths, max_path_length, baseline_predictions, discount)[source]¶ Return processed sample data based on the collected paths.
Parameters: Returns: - Processed sample data, with key
- observations (numpy.ndarray): Padded array of the observations of
- the environment
- actions (numpy.ndarray): Padded array of the actions fed to the
- the environment
- rewards (numpy.ndarray): Padded array of the acquired rewards
- agent_infos (dict): a dictionary of {stacked tensors or
- dictionary of stacked tensors}
- env_infos (dict): a dictionary of {stacked tensors or
- dictionary of stacked tensors}
- rewards (numpy.ndarray): Padded array of the validity information
Return type: