benchmarl.algorithms.MappoConfig

class MappoConfig(share_param_critic: bool, clip_epsilon: float, entropy_coef: float, critic_coef: float, loss_critic_type: str, lmbda: float, scale_mapping: str, use_tanh_normal: bool, minibatch_advantage: bool)

Bases: AlgorithmConfig

Configuration dataclass for Mappo.

share_param_critic: bool = <dataclasses._MISSING_TYPE object>

clip_epsilon: float = <dataclasses._MISSING_TYPE object>

entropy_coef: float = <dataclasses._MISSING_TYPE object>

critic_coef: float = <dataclasses._MISSING_TYPE object>

loss_critic_type: str = <dataclasses._MISSING_TYPE object>

lmbda: float = <dataclasses._MISSING_TYPE object>

scale_mapping: str = <dataclasses._MISSING_TYPE object>

use_tanh_normal: bool = <dataclasses._MISSING_TYPE object>

minibatch_advantage: bool = <dataclasses._MISSING_TYPE object>

static associated_class() → Type[Algorithm]

The algorithm class associated to the config

static supports_continuous_actions() → bool

If the algorithm supports continuous actions

static supports_discrete_actions() → bool

If the algorithm supports discrete actions

static on_policy() → bool

If the algorithm has to be run on policy or off policy

static has_centralized_critic() → bool

If the algorithm uses a centralized critic