Resets the state of the environment and returns an initial observation. Returns: observation (object): the initial observation of the

Run one timestep of the environment's dynamics. When end of episode is reached, you are responsible for calling `reset()` to reset thi

default_config must specify all the expected keys. Use the following special values for required placeholders: * _req_: require a single

Returns arr[start] + ... + arr[end]

Returns a SharedMemoryObject whose .data attribute contains data

Load data from a numpy_dictionary Args: numpy_dict: { <module_dict_key>: { <para

Default Main loop Args: @env: the environment to run agent on

Renders the environment.

Build all images and push them

This object ensures that common frames between the observations are only stored once. It exists purely to optimize memory usage which can be h

Override _close in your subclass to perform any necessary cleanup. Environments will automatically close() themselves when garbage co

Method computes likelihood of action (a) given probability (prob)

forward pass actor to generate policy with option to use z-filter Args: obs: dictionary of batched observatio

forward pass critic to generate policy with option to use z-filter Note: assumes input has either shape length 2 or 4 without

Stack k last frames. Returns lazy array, which is much more memory efficient. See Also -------- baselines.common.ata

Default sender configs are in BASE_SESSION_CONFIG['sender'] They contain communication level information Algorithm s

Concatenate (along 2nd dimension) all the low-dimensional (propioceptive) features from input observation tuple A

Returns current moving average, 0 if no records

Add a new experience to the replay. Includes passive evict logic if memory capacity is exceeded. Args: exp_dict:

Launches the specific component Args: component_name(str): the process to launch

Makes an environment and populates related fields in env_config return env, env_config Args: mode(str): allow differnt kwargs to

Issue: Surreal-tmux gives segfault on running robosuite environments. Cause: On launching a surreal process, an instance of RL environment is

Sets up the communication between surreal components using symphony Args: agents, evals (list): list of symphony

Recursively convert back to dictionary

The following options are pre-configured --verbosity, or -vvv (number of v's indicate the level of verbosity) --debug: turn o

Given uint8 input from the environment, scale to float32 and divide by 255 to scale inputs between 0.0 and 1.0 Ar

Extends base class fetch_parameters to add some logging

Method that returns generator that contains all the parameters from actor, cnn_stem (optional), and rnn_stem (optional)

Args: name: file name for the remote log file session_config: see session_config.loggerplex

Wait for all parameter server workers to exit (Currently this means they crashed) Note that proxy is a daemon

Method computes KL Divergence of between two probability distributions Note: this is D_KL(prob0 || prob1), not D_KL(prob1 ||

One loop of agent, runs one episode of the environment

Setup before constant looping

Returns min(arr[start], ..., arr[end])

Speedometer, step per second Args: average_episodes: average over the past N episodes

Build a Segment Tree data structure. https://en.wikipedia.org/wiki/Segment_tree Can be used as regular array, but with two imp

Helper function that batches a list of observation (nested dictionary) into one whole dictionary of batched observation

Temporary change save folder, useful for loading from other folders

Get the periodic tensorplex object Args: @name: The name of the collection of metrics

helper to be used in self.restore() and self.restore_full_name() Returns: - ckpt file name if successfully restored