TD3BC

Overview

TD3BC, proposed in the 2021 paper A Minimalist Approach to Offline Reinforcement Learning, is a simple approach to offline RL where only two changes are made to TD3: a weighted behavior cloning loss is added to the policy update and the states are normalized. Unlike competing methods there are no changes to architecture or underlying hyperparameters. The resulting algorithm is a simple baseline that is easy to implement and tune, while more than halving the overall run time by removing the additional computational overhead of previous methods.

Implementation changes offline RL algorithms make to the underlying base RL algorithm. † corresponds to details that add additional hyperparameter(s), and ‡ corresponds to ones that add a computational cost. Ref

Quick Facts

1. TD3BC is an offline RL algorithm.

2. TD3BC is based on TD3 and behavior cloning.

Key Equations or Key Graphs

TD3BC simply consists to add a behavior cloning term to TD3 in order to regularize the policy:

\[\begin{aligned} \pi = \arg\max_{\pi} \mathbb{E}_{(s, a) \sim D} [ \lambda Q(s, \pi(s)) - (\pi(s)-a)^2 ] \end{aligned}\]

\((\pi(s)-a)^2\) is the behavior cloning term acts as a regularizer and aims to push the policy towards favoring actions contained in the dataset. The hyperparameter \(\lambda\) is used to control the strength of the regularizer.

Assuming an action range of [−1, 1], the BC term is at most 4, however the range of Q will be a function of the scale of the reward. Consequently, the scalar \(\lambda\) can be defined as:

\[\begin{aligned} \lambda = \frac{\alpha}{\frac{1}{N}\sum_{s_i, a_i}|Q(s_i, a_i)|} \end{aligned}\]

which is simply a normalization term based on the average absolute value of Q over mini-batches. This formulation has also the benefit of normalizing the learning rate across tasks since it is dependent on the scale of Q. The default value for \(\alpha\) is 2.5.

Additionally, all the states in each mini-batch are normalized, such that they have mean 0 and standard deviation 1. This normalization improves the stability of the learned policy.

Implementations

The default config is defined as follows:

class ding.policy.td3_bc.TD3BCPolicy(cfg: EasyDict, model: Module | None = None, enable_field: List[str] | None = None)

Overview:

Policy class of TD3_BC algorithm.

Since DDPG and TD3 share many common things, we can easily derive this TD3_BC class from DDPG class by changing _actor_update_freq, _twin_critic and noise in model wrapper.

https://arxiv.org/pdf/2106.06860.pdf

Property:

learn_mode, collect_mode, eval_mode

Config:

ID	Symbol	Type	Default Value	Description	Other(Shape)
1	type	str	td3_bc	RL policy register name, refer to registry POLICY_REGISTRY	this arg is optional, a placeholder
2	cuda	bool	True	Whether to use cuda for network
3	random_ collect_size	int	25000	Number of randomly collected training samples in replay buffer when training starts.	Default to 25000 for DDPG/TD3, 10000 for sac.
4	model.twin_ critic	bool	True	Whether to use two critic networks or only one.	Default True for TD3, Clipped Double Q-learning method in TD3 paper.
5	learn.learning _rate_actor	float	1e-3	Learning rate for actor network(aka. policy).
6	learn.learning _rate_critic	float	1e-3	Learning rates for critic network (aka. Q-network).
7	learn.actor_ update_freq	int	2	When critic network updates once, how many times will actor network update.	Default 2 for TD3, 1 for DDPG. Delayed Policy Updates method in TD3 paper.
8	learn.noise	bool	True	Whether to add noise on target network’s action.	Default True for TD3, False for DDPG. Target Policy Smoo- thing Regularization in TD3 paper.
9	learn.noise_ range	dict	dict(min=-0.5, max=0.5,)	Limit for range of target policy smoothing noise, aka. noise_clip.
10	learn.- ignore_done	bool	False	Determine whether to ignore done flag.	Use ignore_done only in halfcheetah env.
11	learn.- target_theta	float	0.005	Used for soft update of the target network.	aka. Interpolation factor in polyak aver aging for target networks.
12	collect.- noise_sigma	float	0.1	Used for add noise during co- llection, through controlling the sigma of distribution	Sample noise from dis tribution, Ornstein- Uhlenbeck process in DDPG paper, Guassian process in ours.

Model

Here we provide examples of ContinuousQAC model as default model for TD3BC.

class ding.model.ContinuousQAC(obs_shape: int | SequenceType, action_shape: int | SequenceType | EasyDict, action_space: str, twin_critic: bool = False, actor_head_hidden_size: int = 64, actor_head_layer_num: int = 1, critic_head_hidden_size: int = 64, critic_head_layer_num: int = 1, activation: Module | None = ReLU(), norm_type: str | None = None, encoder_hidden_size_list: SequenceType | None = None, share_encoder: bool | None = False)

Overview:

The neural network and computation graph of algorithms related to Q-value Actor-Critic (QAC), such as DDPG/TD3/SAC. This model now supports continuous and hybrid action space. The ContinuousQAC is composed of four parts: actor_encoder, critic_encoder, actor_head and critic_head. Encoders are used to extract the feature from various observation. Heads are used to predict corresponding Q-value or action logit. In high-dimensional observation space like 2D image, we often use a shared encoder for both actor_encoder and critic_encoder. In low-dimensional observation space like 1D vector, we often use different encoders.

Interfaces:

__init__, forward, compute_actor, compute_critic

compute_actor(obs: Tensor) → Dict[str, Tensor | Dict[str, Tensor]]

Overview:

QAC forward computation graph for actor part, input observation tensor to predict action or action logit.

Arguments:

• x (torch.Tensor): The input observation tensor data.

Returns:

• outputs (Dict[str, Union[torch.Tensor, Dict[str, torch.Tensor]]]): Actor output dict varying from action_space: regression, reparameterization, hybrid.

ReturnsKeys (regression):

• action (torch.Tensor): Continuous action with same size as action_shape, usually in DDPG/TD3.

ReturnsKeys (reparameterization):

• logit (Dict[str, torch.Tensor]): The predictd reparameterization action logit, usually in SAC. It is a list containing two tensors: mu and sigma. The former is the mean of the gaussian distribution, the latter is the standard deviation of the gaussian distribution.

ReturnsKeys (hybrid):

• logit (torch.Tensor): The predicted discrete action type logit, it will be the same dimension as action_type_shape, i.e., all the possible discrete action types.

• action_args (torch.Tensor): Continuous action arguments with same size as action_args_shape.

Shapes:

• obs (torch.Tensor): \((B, N0)\), B is batch size and N0 corresponds to obs_shape.

• action (torch.Tensor): \((B, N1)\), B is batch size and N1 corresponds to action_shape.

• logit.mu (torch.Tensor): \((B, N1)\), B is batch size and N1 corresponds to action_shape.

• logit.sigma (torch.Tensor): \((B, N1)\), B is batch size.

• logit (torch.Tensor): \((B, N2)\), B is batch size and N2 corresponds to action_shape.action_type_shape.

• action_args (torch.Tensor): \((B, N3)\), B is batch size and N3 corresponds to action_shape.action_args_shape.

Examples:

>>> # Regression mode
>>> model = ContinuousQAC(64, 6, 'regression')
>>> obs = torch.randn(4, 64)
>>> actor_outputs = model(obs,'compute_actor')
>>> assert actor_outputs['action'].shape == torch.Size([4, 6])
>>> # Reparameterization Mode
>>> model = ContinuousQAC(64, 6, 'reparameterization')
>>> obs = torch.randn(4, 64)
>>> actor_outputs = model(obs,'compute_actor')
>>> assert actor_outputs['logit'][0].shape == torch.Size([4, 6])  # mu
>>> actor_outputs['logit'][1].shape == torch.Size([4, 6]) # sigma

compute_critic(inputs: Dict[str, Tensor]) → Dict[str, Tensor]

Overview:

QAC forward computation graph for critic part, input observation and action tensor to predict Q-value.

Arguments:

• inputs (Dict[str, torch.Tensor]): The dict of input data, including obs and action tensor, also contains logit and action_args tensor in hybrid action_space.

ArgumentsKeys:

• obs: (torch.Tensor): Observation tensor data, now supports a batch of 1-dim vector data.

• action (Union[torch.Tensor, Dict]): Continuous action with same size as action_shape.

• logit (torch.Tensor): Discrete action logit, only in hybrid action_space.

• action_args (torch.Tensor): Continuous action arguments, only in hybrid action_space.

Returns:

• outputs (Dict[str, torch.Tensor]): The output dict of QAC’s forward computation graph for critic, including q_value.

ReturnKeys:

• q_value (torch.Tensor): Q value tensor with same size as batch size.

Shapes:

• obs (torch.Tensor): \((B, N1)\), where B is batch size and N1 is obs_shape.

• logit (torch.Tensor): \((B, N2)\), B is batch size and N2 corresponds to action_shape.action_type_shape.

• action_args (torch.Tensor): \((B, N3)\), B is batch size and N3 corresponds to action_shape.action_args_shape.

• action (torch.Tensor): \((B, N4)\), where B is batch size and N4 is action_shape.

• q_value (torch.Tensor): \((B, )\), where B is batch size.

Examples:

>>> inputs = {'obs': torch.randn(4, 8), 'action': torch.randn(4, 1)}
>>> model = ContinuousQAC(obs_shape=(8, ),action_shape=1, action_space='regression')
>>> assert model(inputs, mode='compute_critic')['q_value'].shape == (4, )  # q value

forward(inputs: Tensor | Dict[str, Tensor], mode: str) → Dict[str, Tensor]

Overview:

QAC forward computation graph, input observation tensor to predict Q-value or action logit. Different mode will forward with different network modules to get different outputs and save computation.

Arguments:

• inputs (Union[torch.Tensor, Dict[str, torch.Tensor]]): The input data for forward computation graph, for compute_actor, it is the observation tensor, for compute_critic, it is the dict data including obs and action tensor.

• mode (str): The forward mode, all the modes are defined in the beginning of this class.

Returns:

• output (Dict[str, torch.Tensor]): The output dict of QAC forward computation graph, whose key-values vary in different forward modes.

Examples (Actor):

>>> # Regression mode
>>> model = ContinuousQAC(64, 6, 'regression')
>>> obs = torch.randn(4, 64)
>>> actor_outputs = model(obs,'compute_actor')
>>> assert actor_outputs['action'].shape == torch.Size([4, 6])
>>> # Reparameterization Mode
>>> model = ContinuousQAC(64, 6, 'reparameterization')
>>> obs = torch.randn(4, 64)
>>> actor_outputs = model(obs,'compute_actor')
>>> assert actor_outputs['logit'][0].shape == torch.Size([4, 6])  # mu
>>> actor_outputs['logit'][1].shape == torch.Size([4, 6]) # sigma

Examples (Critic):

>>> inputs = {'obs': torch.randn(4, 8), 'action': torch.randn(4, 1)}
>>> model = ContinuousQAC(obs_shape=(8, ),action_shape=1, action_space='regression')
>>> assert model(inputs, mode='compute_critic')['q_value'].shape == (4, )  # q value

Benchmark

environment	best mean reward	evaluation results	config link	comparison
Halfcheetah (Medium Expert)	13037		config_link_ha	d3rlpy(12124)
Walker2d (Medium Expert)	5066		config_link_w	d3rlpy(5108)
Hopper (Medium Expert)	3653		config_link_ho	d3rlpy(3690)

environment	random	medium replay	medium expert	medium	expert
Halfcheetah	1592	5192	13037	5257	13247
Walker2d	345	1724	3653	3268	3664
Hopper	985	2317	5066	3826	5232

Note: the D4RL environment used in this benchmark can be found here.

References

• Scott Fujimoto, Shixiang Shane Gu: “A Minimalist Approach to Offline Reinforcement Learning”, 2021; [https://arxiv.org/abs/2106.06860 arXiv:2106.06860].

• Scott Fujimoto, Herke van Hoof, David Meger: “Addressing Function Approximation Error in Actor-Critic Methods”, 2018; [http://arxiv.org/abs/1802.09477 arXiv:1802.09477].

Other Public Implementations

• Official implementation

• d3rlpy