unreal.LearningAgentsPolicySettings

class unreal.LearningAgentsPolicySettings

Bases: StructBase

The configurable settings for a ULearningAgentsPolicy.

C++ Source:

  • Plugin: LearningAgents

  • Module: LearningAgents

  • File: LearningAgentsPolicy.h

Editor Properties: (see get_editor_property/set_editor_property)

  • activation_function (LearningAgentsActivationFunction): [Read-Write] Activation function to use on hidden layers of the policy network

  • hidden_layer_num (int32): [Read-Write] Number of hidden layers for policy network.

  • hidden_layer_size (int32): [Read-Write] Number of neurons in each hidden layer of the policy network

  • initial_encoded_action_scale (float): [Read-Write] Initial scale for encoded actions. Setting this <1 reduces the initial bias from the network random weights and tends to help training.

  • memory_cell (LearningAgentsMemoryCell): [Read-Write] The type of memory cell to use. Setting this to NoMemoryCell is equivalent to setting MemoryStateSize to 0.

  • memory_state_size (int32): [Read-Write] Number of neurons in the memory state of the policy network

  • use_parallel_evaluation (bool): [Read-Write] If true, policy can be evaluated on multiple threads to speed up large agent counts.