Saving and Loading

navground.learning.io

save_env(env: BaseEnv | BaseParallelEnv, path: PathLike) → None

Export a NavgroundEnv to YAML

Parameters:

• path (PathLike) – The directory where to save the files

• env (BaseEnv | BaseParallelEnv) – The environment.

Raises:

TypeError – If env.unwrapped is not of a navground environment

Return type:

None

load_env(path: PathLike) → MultiAgentNavgroundEnv | NavgroundEnv

Parameters:

path (PathLike)

Return type:

MultiAgentNavgroundEnv | NavgroundEnv

export_behavior(model: BaseAlgorithm | BaseILAlgorithm, path: PathLike) → None

Export a model using export_policy_as_behavior()

Parameters:

• model (BaseAlgorithm | BaseILAlgorithm) – The model

• path (PathLike) – The directory where to save the files

Return type:

None

export_policy_as_behavior(path: PathLike, policy: BasePolicy | None = None, env: BaseEnv | BaseParallelEnv | None = None, index: int | None = None) → None

Export a policy (using navground.learning.onnx.export()) together with the YAML representation of the behavior and sensor to use it.

The behavior and sensor refer to an agent at a given index in the environment. If they cannot be retrieved, the related files will not be saved.

Parameters:

• path (PathLike) – The directory where to save the files

• policy (BasePolicy | None) – The policy

• env (BaseEnv | BaseParallelEnv | None) – The environment.

• index (int | None) – The index of the agent in the environment. If not provided, selects one of the possible agents.

Return type:

None

load_behavior(path: PathLike) → tuple[PolicyBehavior | None, StateEstimation | None]

Load behavior and sensor previously saved in a directory using export_policy_as_behavior().

Parameters:

path (PathLike) – The directory path

Returns:

A policy behavior and a sensor, set to None if the could not be loaded.

Return type:

tuple[PolicyBehavior | None, StateEstimation | None]