Advance the environment by one simulation step. If the environment is using the contextual setting, an "is_success" term is added to

See parent class.

See parent class. The done mask is replaced with a dictionary to match other multi-agent environments.

See parent class.

Store a transition in the replay buffer. Parameters ---------- obs0 : dict of array_like the last observation for

Return the next states given a set of states and actions. Parameters ---------- state : array_like the states by

Perform gradient update step given a batch of data. Parameters ---------- obs0 : array_like batch of observations

Perform gradient update step given a batch of data. Parameters ---------- obs : array_like a minibatch of observa

Perform gradient update step given a batch of data. Parameters ---------- obs : array_like a minibatch of observa

Get TensorFlow Tensor shape. Parameters ---------- tensor : tf.Tensor the input tensor Returns ------- list of int

Perform the class-specific operation.

Perform the class-specific operation.

Instantiate a ring buffer (FIFO). Parameters ---------- buffer_size : int Max number of transitions to store in t

Instantiate the feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session the current

Instantiate the feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session the current

Instantiate the policy object. Parameters ---------- sess : tf.compat.v1.Session the current TensorFlow session

Instantiate the policy object. Parameters ---------- sess : tf.compat.v1.Session the current TensorFlow session

Instantiate the logger object. Parameters ---------- env_name : str the name of the environment

Set the parameters from a flat vector. Parameters ---------- var_list : list of tf.Tensor the variables d

Get the parameters as a flat vector. Parameters ---------- var_list : list of tf.Tensor the variables ses

Instantiate the sampler object. Parameters ---------- env_name : str the name of the environment render :

Instantiate the hierarchical replay buffer. Parameters ---------- buffer_size : int Max number of transitions to

Instantiate the goal-conditioned hierarchical policy. Parameters ---------- sess : tf.compat.v1.Session the curre

Instantiate the goal-conditioned hierarchical policy. Parameters ---------- sess : tf.compat.v1.Session the curre

Instantiate the goal-conditioned hierarchical policy. Parameters ---------- sess : tf.compat.v1.Session the curre

Instantiate the environment. Parameters ---------- render_dt_msec : float seconds before the next frame in the im

Instantiate the Ant environment. Parameters ---------- file_path : str path to the xml file expose_all_qp

Initialize the Universal environment. Parameters ---------- maze_id : str the type of maze environment. One of "M

Initialize the Universal environment. Parameters ---------- maze_id : str the type of maze environment. One of "M

Initialize the Humanoid Maze environment. Parameters ---------- use_contexts : bool, optional specifies whether t

Initialize the Image Ant Maze environment. Parameters ---------- use_contexts : bool, optional specifies whether

Initialize the Image Humanoid Maze environment. Parameters ---------- use_contexts : bool, optional specifies whe

Initialize the Ant Push environment. Parameters ---------- use_contexts : bool, optional specifies whether to add

Initialize the Humanoid Push environment. Parameters ---------- use_contexts : bool, optional specifies whether t

Initialize the Ant Fall environment. Parameters ---------- use_contexts : bool, optional specifies whether to add

Initialize the Humanoid Fall environment. Parameters ---------- use_contexts : bool, optional specifies whether t

Initialize the Ant Four Rooms environment. Parameters ---------- use_contexts : bool, optional specifies whether

Initialize the Humanoid Four Rooms environment. Parameters ---------- use_contexts : bool, optional specifies whe

Create a humanoid agent.

Instantiate the environment. Parameters ---------- maze_id : str the type of maze being simulated. Can be 'M

Instantiate the environment. Parameters ---------- maze_id : str the type of maze being simulated. Can be 'Maze',

Instantiate the environment.

Instantiate the environment. Parameters ---------- render : bool whether to render the environment

Create the environment. Parameters ---------- flow_params : dict environment-specific parameters multiage

Instantiate the environment class. Parameters ---------- length : float or [float, float] the length of the ring

See parent class.

See parent class.

See parent class.

Instantiate the Environment object. Parameters ---------- model_name : str name of the xml file in './mujoco_file

Initialize the Pendulum environment. Parameters ---------- use_contexts : bool, optional specifies whether to add

Instantiate the environment class. In order to match the environment presented in the article, we modify the following default values

Instantiate the base policy object. Parameters ---------- sess : tf.compat.v1.Session the current TensorFlow sess

Instantiate a multi-agent feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session t

Instantiate a buffer. Parameters ---------- buffer_size : int Max number of transitions to store in the buffer. W

Instantiate a buffer. Parameters ---------- buffer_size : int Max number of transitions to store in the buffer. W

Instantiate the base multi-agent actor critic policy. Parameters ---------- sess : tf.compat.v1.Session the curre

Instantiate a multi-agent feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session t

Instantiate a multi-agent feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session t

Instantiate a multi-agent feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session t

Instantiate a multi-agent feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session t

Instantiate a multi-agent feed-forward neural network policy. Parameters ---------- sess : tf.compat.v1.Session t

Instantiate the algorithm object. Parameters ---------- policy : type [ hbaselines.base_policies.Policy ] the pol

Return the number of elements stored.

Return the number of elements stored.

Return the number of elements stored.

Return the number of elements stored.

See class definition.

See class definition.

Perform the gradient update procedure for the CHER algorithm. This procedure is similar to update_from_batch, expect it runs the self

Perform the gradient update procedure for the CHER algorithm. This procedure is similar to update_from_batch, expect it runs the self

See get_action.

See get_action.

See get_action.

See get_action.

See get_action.

See get_action.

See get_td_map.

See get_td_map.

See get_td_map.

See get_td_map.

See get_td_map.

See get_td_map.

See initialize.

See initialize.

See initialize. This method initializes the target parameters to match the model parameters.

See initialize.

See initialize.

See initialize.

Calculate the log probability of the next goal by the meta-policies. Parameters ---------- meta_actions : array_like

Calculate the log probability of the next goal by the meta-policies. Parameters ---------- meta_actions : array_like

Create the cooperative gradients meta-policy optimizer.

Create the cooperative gradients meta-policy optimizer.

See setup.