tf_agents/agents/ppo/ppo_agent.py [116:138]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  def __init__(
      self,
      time_step_spec: ts.TimeStep,
      action_spec: types.NestedTensorSpec,
      optimizer: Optional[types.Optimizer] = None,
      actor_net: Optional[network.Network] = None,
      value_net: Optional[network.Network] = None,
      greedy_eval: bool = True,
      importance_ratio_clipping: types.Float = 0.0,
      lambda_value: types.Float = 0.95,
      discount_factor: types.Float = 0.99,
      entropy_regularization: types.Float = 0.0,
      policy_l2_reg: types.Float = 0.0,
      value_function_l2_reg: types.Float = 0.0,
      shared_vars_l2_reg: types.Float = 0.0,
      value_pred_loss_coef: types.Float = 0.5,
      num_epochs: int = 25,
      use_gae: bool = False,
      use_td_lambda_return: bool = False,
      normalize_rewards: bool = True,
      reward_norm_clipping: types.Float = 10.0,
      normalize_observations: bool = True,
      log_prob_clipping: types.Float = 0.0,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



tf_agents/agents/ppo/ppo_clip_agent.py [72:94]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  def __init__(
      self,
      time_step_spec: ts.TimeStep,
      action_spec: types.NestedTensorSpec,
      optimizer: Optional[types.Optimizer] = None,
      actor_net: Optional[network.Network] = None,
      value_net: Optional[network.Network] = None,
      greedy_eval: bool = True,
      importance_ratio_clipping: types.Float = 0.0,
      lambda_value: types.Float = 0.95,
      discount_factor: types.Float = 0.99,
      entropy_regularization: types.Float = 0.0,
      policy_l2_reg: types.Float = 0.0,
      value_function_l2_reg: types.Float = 0.0,
      shared_vars_l2_reg: types.Float = 0.0,
      value_pred_loss_coef: types.Float = 0.5,
      num_epochs: int = 25,
      use_gae: bool = False,
      use_td_lambda_return: bool = False,
      normalize_rewards: bool = True,
      reward_norm_clipping: types.Float = 10.0,
      normalize_observations: bool = True,
      log_prob_clipping: types.Float = 0.0,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -