from typing import Optional, Dict, Union import copy import torch import torch.nn as nn from ding.utils import SequenceType, MODEL_REGISTRY from .vac import VAC @MODEL_REGISTRY.register('ppg') class PPG(nn.Module): """ Overview: Phasic Policy Gradient (PPG) model from paper `Phasic Policy Gradient` https://arxiv.org/abs/2009.04416 \ This module contains VAC module and an auxiliary critic module. Interfaces: ``forward``, ``compute_actor``, ``compute_critic``, ``compute_actor_critic`` """ mode = ['compute_actor', 'compute_critic', 'compute_actor_critic'] def __init__( self, obs_shape: Union[int, SequenceType], action_shape: Union[int, SequenceType], action_space: str = 'discrete', share_encoder: bool = True, encoder_hidden_size_list: SequenceType = [128, 128, 64], actor_head_hidden_size: int = 64, actor_head_layer_num: int = 2, critic_head_hidden_size: int = 64, critic_head_layer_num: int = 1, activation: Optional[nn.Module] = nn.ReLU(), norm_type: Optional[str] = None, impala_cnn_encoder: bool = False, ) -> None: """ Overview: Initailize the PPG Model according to input arguments. Arguments: - obs_shape (:obj:`Union[int, SequenceType]`): Observation's shape, such as 128, (156, ). - action_shape (:obj:`Union[int, SequenceType]`): Action's shape, such as 4, (3, ). - action_space (:obj:`str`): The action space type, such as 'discrete', 'continuous'. - share_encoder (:obj:`bool`): Whether to share encoder. - encoder_hidden_size_list (:obj:`SequenceType`): The hidden size list of encoder. - actor_head_hidden_size (:obj:`int`): The ``hidden_size`` to pass to actor head. - actor_head_layer_num (:obj:`int`): The num of layers used in the network to compute Q value output \ for actor head. - critic_head_hidden_size (:obj:`int`): The ``hidden_size`` to pass to critic head. - critic_head_layer_num (:obj:`int`): The num of layers used in the network to compute Q value output \ for critic head. - activation (:obj:`Optional[nn.Module]`): The type of activation function to use in ``MLP`` \ after each FC layer, if ``None`` then default set to ``nn.ReLU()``. - norm_type (:obj:`Optional[str]`): The type of normalization to after network layer (FC, Conv), \ see ``ding.torch_utils.network`` for more details. - impala_cnn_encoder (:obj:`bool`): Whether to use impala cnn encoder. """ super(PPG, self).__init__() self.actor_critic = VAC( obs_shape, action_shape, action_space, share_encoder, encoder_hidden_size_list, actor_head_hidden_size, actor_head_layer_num, critic_head_hidden_size, critic_head_layer_num, activation, norm_type, impala_cnn_encoder=impala_cnn_encoder ) self.aux_critic = copy.deepcopy(self.actor_critic.critic) def forward(self, inputs: Union[torch.Tensor, Dict], mode: str) -> Dict: """ Overview: Compute action logits or value according to mode being ``compute_actor``, ``compute_critic`` or \ ``compute_actor_critic``. Arguments: - x (:obj:`torch.Tensor`): The input observation tensor data. - mode (:obj:`str`): The forward mode, all the modes are defined in the beginning of this class. Returns: - outputs (:obj:`Dict`): The output dict of PPG's forward computation graph, whose key-values vary from \ different ``mode``. """ assert mode in self.mode, "not support forward mode: {}/{}".format(mode, self.mode) return getattr(self, mode)(inputs) def compute_actor(self, x: torch.Tensor) -> Dict: """ Overview: Use actor to compute action logits. Arguments: - x (:obj:`torch.Tensor`): The input observation tensor data. Returns: - output (:obj:`Dict`): The output data containing action logits. ReturnsKeys: - logit (:obj:`torch.Tensor`): The predicted action logit tensor, for discrete action space, it will be \ the same dimension real-value ranged tensor of possible action choices, and for continuous action \ space, it will be the mu and sigma of the Gaussian distribution, and the number of mu and sigma is the \ same as the number of continuous actions. Hybrid action space is a kind of combination of discrete \ and continuous action space, so the logit will be a dict with ``action_type`` and ``action_args``. Shapes: - x (:obj:`torch.Tensor`): :math:`(B, N)`, where B is batch size and N is the input feature size. - output (:obj:`Dict`): ``logit``: :math:`(B, A)`, where B is batch size and A is the action space size. """ return self.actor_critic(x, mode='compute_actor') def compute_critic(self, x: torch.Tensor) -> Dict: """ Overview: Use critic to compute value. Arguments: - x (:obj:`torch.Tensor`): The input observation tensor data. Returns: - output (:obj:`Dict`): The output dict of VAC's forward computation graph for critic, including ``value``. ReturnsKeys: - necessary: ``value`` Shapes: - x (:obj:`torch.Tensor`): :math:`(B, N)`, where B is batch size and N is the input feature size. - output (:obj:`Dict`): ``value``: :math:`(B, 1)`, where B is batch size. """ x = self.aux_critic[0](x) # encoder x = self.aux_critic[1](x) # head return {'value': x['pred']} def compute_actor_critic(self, x: torch.Tensor) -> Dict: """ Overview: Use actor and critic to compute action logits and value. Arguments: - x (:obj:`torch.Tensor`): The input observation tensor data. Returns: - outputs (:obj:`Dict`): The output dict of PPG's forward computation graph for both actor and critic, \ including ``logit`` and ``value``. ReturnsKeys: - logit (:obj:`torch.Tensor`): The predicted action logit tensor, for discrete action space, it will be \ the same dimension real-value ranged tensor of possible action choices, and for continuous action \ space, it will be the mu and sigma of the Gaussian distribution, and the number of mu and sigma is the \ same as the number of continuous actions. Hybrid action space is a kind of combination of discrete \ and continuous action space, so the logit will be a dict with ``action_type`` and ``action_args``. - value (:obj:`torch.Tensor`): The predicted state value tensor. Shapes: - x (:obj:`torch.Tensor`): :math:`(B, N)`, where B is batch size and N is the input feature size. - output (:obj:`Dict`): ``value``: :math:`(B, 1)`, where B is batch size. - output (:obj:`Dict`): ``logit``: :math:`(B, A)`, where B is batch size and A is the action space size. .. note:: ``compute_actor_critic`` interface aims to save computation when shares encoder. """ return self.actor_critic(x, mode='compute_actor_critic')