tf_agents.agents.reinforce.reinforce_agent.ReinforceAgentLossInfo

View source on GitHub

ReinforceAgentLossInfo is stored in the extras field of the LossInfo.

All losses, except for policy_network_regularization_loss have a validity mask applied to ensure no loss or error is calculated for episode boundaries.

policy_gradient_loss: The weighted policy_gradient loss. policy_network_regularization_loss: The regularization loss terms from the policy network used to generate the policy_gradient_loss. entropy_regularization_loss: The entropy regularization loss. value_estimation_loss: If value estimation network is being used, the loss associated with that network.

policy_gradient_loss

policy_network_regularization_loss

entropy_regularization_loss

value_estimation_loss

value_network_regularization_loss