hive.envs.base module

class hive.envs.base.BaseEnv(env_spec, num_players)[source]

Bases: ABC, Registrable

Base class for environments.

Parameters

env_spec (EnvSpec) – An object containing information about the environment.
num_players (int) – The number of players in the environment.

abstract reset()[source]

Resets the state of the environment.

Returns: The initial observation of the new episode. turn (int): The index of the agent which should take turn.
Return type: observation

abstract step(action)[source]

Run one time-step of the environment using the input action.

Parameters: action – An element of environment’s action space.
Returns: Indicates the next state that is an element of environment’s observation space. reward: A reward achieved from the transition. done (bool): Indicates whether the episode has ended. turn (int): Indicates which agent should take turn. info (dict): Additional custom information.
Return type: observation

render(mode='rgb_array')[source]: Displays a rendered frame from the environment.

abstract seed(seed=None)[source]

Reseeds the environment.

Parameters: seed (int) – Seed to use for environment.

save(save_dir)[source]

Saves the environment.

Parameters: save_dir (str) – Location to save environment state.

load(load_dir)[source]

Loads the environment.

Parameters: load_dir (str) – Location to load environment state from.

close()[source]: Additional clean up operations

property env_spec

classmethod type_name()[source]: Returns: “env”

class hive.envs.base.ParallelEnv(env_name, num_players)[source]

Bases: BaseEnv

Base class for environments that take make all agents step in parallel.

ParallelEnv takes an environment that expects an array of actions at each step to execute in parallel, and allows you to instead pass it a single action at each step.

This class makes use of Python’s multiple inheritance pattern. Specifically, when writing your parallel environment, it should extend both this class and the class that implements the step method that takes in actions for all agents.

If environment class A has the logic for the step function that takes in the array of actions, and environment class B is your parallel step version of that environment, class B should be defined as:

class B(ParallelEnv, A):
    ...

The order in which you list the classes is important. ParallelEnv must come before A in the order.

Parameters

env_spec (EnvSpec) – An object containing information about the environment.
num_players (int) – The number of players in the environment.

reset()[source]

Resets the state of the environment.

Returns: The initial observation of the new episode. turn (int): The index of the agent which should take turn.
Return type: observation

step(action)[source]

Run one time-step of the environment using the input action.

Parameters: action – An element of environment’s action space.
Returns: Indicates the next state that is an element of environment’s observation space. reward: A reward achieved from the transition. done (bool): Indicates whether the episode has ended. turn (int): Indicates which agent should take turn. info (dict): Additional custom information.
Return type: observation