garage.envs.wrappers

gym.Env wrappers.

Used to transform an environment in a modular way. It is also possible to apply multiple wrappers at the same time.

Example

StackFrames(GrayScale(gym.make(‘env’)))

class AtariEnv(env)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.AtariEnv

Atari environment wrapper for gym.Env.

This wrapper convert the observations returned from baselines wrapped environment, which is a LazyFrames object into numpy arrays.

Parameters

env (gym.Env) – The environment to be wrapped.

metadata
reward_range
action_space
observation_space
step(self, action)

gym.Env step function.

reset(self, **kwargs)

gym.Env reset function.

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class ClipReward(env)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.ClipReward

Clip the reward by its sign.

metadata
reward_range
action_space
observation_space
step(self, ac)

gym.Env step function.

reset(self)

gym.Env reset.

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class EpisodicLife(env)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.EpisodicLife

Episodic life wrapper for gym.Env.

This wrapper makes episode end when a life is lost, but only reset when all lives are lost.

Parameters

env – The environment to be wrapped.

metadata
reward_range
action_space
observation_space
step(self, action)

gym.Env step function.

reset(self, **kwargs)

gym.Env reset function.

Reset only when lives are lost.

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class FireReset(env)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.FireReset

Fire reset wrapper for gym.Env.

Take action “fire” on reset.

Parameters

env (gym.Env) – The environment to be wrapped.

metadata
reward_range
action_space
observation_space
step(self, action)

gym.Env step function.

Parameters

action (int) – index of the action to take.

Returns

Observation conforming to observation_space float: Reward for this step bool: Termination signal dict: Extra information from the environment.

Return type

np.ndarray

reset(self, **kwargs)

gym.Env reset function.

Parameters

kwargs (dict) – extra arguments passed to gym.Env.reset()

Returns

next observation.

Return type

np.ndarray

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class Grayscale(env)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.Grayscale

Grayscale wrapper for gym.Env, converting frames to grayscale.

Only works with gym.spaces.Box environment with 2D RGB frames. The last dimension (RGB) of environment observation space will be removed.

Example

env = gym.make(‘Env’) # env.observation_space = (100, 100, 3)

env_wrapped = Grayscale(gym.make(‘Env’)) # env.observation_space = (100, 100)

Parameters

env (gym.Env) – Environment to wrap.

Raises

ValueError – If observation space shape is not 3 or environment is not gym.spaces.Box.

metadata
reward_range
action_space
property observation_space(self)

gym.Env: Observation space.

reset(self, **kwargs)

gym.Env reset function.

Parameters

**kwargs – Unused.

Returns

Observation conforming to observation_space

Return type

np.ndarray

step(self, action)

See gym.Env.

Parameters

action (np.ndarray) – Action conforming to action_space

Returns

Observation conforming to observation_space float: Reward for this step bool: Termination signal dict: Extra information from the environment.

Return type

np.ndarray

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class MaxAndSkip(env, skip=4)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.MaxAndSkip

Max and skip wrapper for gym.Env.

It returns only every skip-th frame. Action are repeated and rewards are sum for the skipped frames.

It also takes element-wise maximum over the last two consecutive frames, which helps algorithm deal with the problem of how certain Atari games only render their sprites every other game frame.

Parameters
  • env (gym.Env) – The environment to be wrapped.

  • skip (int) – The environment only returns skip-th frame.

metadata
reward_range
action_space
observation_space
step(self, action)

Repeat action, sum reward, and max over last two observations.

Parameters

action (int) – action to take in the atari environment.

Returns

observation of shape \((O*,)\) representating

the max values over the last two oservations.

float: Reward for this step bool: Termination signal dict: Extra information from the environment.

Return type

np.ndarray

reset(self)

gym.Env reset.

Returns

observaion of shape \((O*,)\).

Return type

np.ndarray

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class Noop(env, noop_max=30)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.Noop

Noop wrapper for gym.Env.

It samples initial states by taking random number of no-ops on reset. No-op is assumed to be action 0.

Parameters
  • env (gym.Env) – The environment to be wrapped.

  • noop_max (int) – Maximum number no-op to be performed on reset.

metadata
reward_range
action_space
observation_space
step(self, action)

gym.Env step function.

reset(self, **kwargs)

gym.Env reset function.

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class PixelObservationWrapper(env, headless=True)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.PixelObservationWrapper

Pixel observation wrapper for obtaining pixel observations.

Instead of returning the default environment observation, the wrapped environment’s render function is used to produce RGB pixel observations.

This behaves like gym.wrappers.PixelObservationWrapper but returns a gym.spaces.Box observation space and observation instead of a gym.spaces.Dict.

Parameters
  • env (gym.Env) – The environment to wrap. This environment must produce non-pixel observations and have a Box observation space.

  • headless (bool) – If true, this creates a window to init GLFW. Set to true if running on a headless machine or with a dummy X server, false otherwise.

metadata
reward_range
action_space
property observation_space(self)

gym.spaces.Box: Environment observation space.

reset(self, **kwargs)

gym.Env reset function.

Parameters

kwargs (dict) – Keyword arguments to be passed to gym.Env.reset.

Returns

Pixel observation of shape \((O*, )\)

from the wrapped environment.

Return type

np.ndarray

step(self, action)

gym.Env step function.

Performs one action step in the enviornment.

Parameters

action (np.ndarray) – Action of shape \((A*, )\) to pass to the environment.

Returns

Pixel observation of shape \((O*, )\)

from the wrapped environment.

float : Amount of reward returned after previous action. bool : Whether the episode has ended, in which case further step()

calls will return undefined results.

dict: Contains auxiliary diagnostic information (helpful for

debugging, and sometimes learning).

Return type

np.ndarray

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class Resize(env, width, height)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.Resize

gym.Env wrapper for resizing frame to (width, height).

Only works with gym.spaces.Box environment with 2D single channel frames.

Example

env = gym.make(‘Env’)
# env.observation_space = (100, 100)
env_wrapped = Resize(gym.make(‘Env’), width=64, height=64)
# env.observation_space = (64, 64)
Parameters
  • env – gym.Env to wrap.

  • width – resized frame width.

  • height – resized frame height.

Raises

ValueError – If observation space shape is not 2 or environment is not gym.spaces.Box.

metadata
reward_range
action_space
property observation_space(self)

gym.Env observation space.

reset(self)

gym.Env reset function.

step(self, action)

gym.Env step function.

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env

class StackFrames(env, n_frames, axis=2)

Bases: gym.Wrapper

Inheritance diagram of garage.envs.wrappers.StackFrames

gym.Env wrapper to stack multiple frames.

Useful for training feed-forward agents on dynamic games. Only works with gym.spaces.Box environment with 2D single channel frames.

Parameters
  • env (gym.Env) – gym.Env to wrap.

  • n_frames (int) – number of frames to stack.

  • axis (int) – Axis to stack frames on. This should be 2 for tensorflow and 0 for pytorch.

Raises
  • ValueError – If observation space shape is not 2 dimnesional,

  • if the environment is not gym.spaces.Box, or if the specified axis

  • is not 0 or 2.

metadata
reward_range
action_space
property observation_space(self)

gym.spaces.Box: gym.Env observation space.

reset(self)

gym.Env reset function.

Returns

Observation conforming to observation_space float: Reward for this step bool: Termination signal dict: Extra information from the environment.

Return type

np.ndarray

step(self, action)

gym.Env step function.

Parameters

action (int) – index of the action to take.

Returns

Observation conforming to observation_space float: Reward for this step bool: Termination signal dict: Extra information from the environment.

Return type

np.ndarray

property spec(self)
classmethod class_name(cls)
render(self, mode='human', **kwargs)

Renders the environment.

The set of supported modes varies per environment. (And some environments do not support rendering at all.) By convention, if mode is:

  • human: render to the current display or terminal and return nothing. Usually for human consumption.

  • rgb_array: Return an numpy.ndarray with shape (x, y, 3), representing RGB values for an x-by-y pixel image, suitable for turning into a video.

  • ansi: Return a string (str) or StringIO.StringIO containing a terminal-style text representation. The text can include newlines and ANSI escape sequences (e.g. for colors).

Note

Make sure that your class’s metadata ‘render.modes’ key includes

the list of supported modes. It’s recommended to call super() in implementations to use the functionality of this method.

Parameters

mode (str) – the mode to render with

Example:

class MyEnv(Env):

metadata = {‘render.modes’: [‘human’, ‘rgb_array’]}

def render(self, mode=’human’):
if mode == ‘rgb_array’:

return np.array(…) # return RGB frame suitable for video

elif mode == ‘human’:

… # pop up a window and render

else:

super(MyEnv, self).render(mode=mode) # just raise an exception

close(self)

Override close in your subclass to perform any necessary cleanup.

Environments will automatically close() themselves when garbage collected or when the program exits.

seed(self, seed=None)

Sets the seed for this env’s random number generator(s).

Note

Some environments use multiple pseudorandom number generators. We want to capture all such seeds used in order to ensure that there aren’t accidental correlations between multiple generators.

Returns

Returns the list of seeds used in this env’s random

number generators. The first value in the list should be the “main” seed, or the value which a reproducer should pass to ‘seed’. Often, the main seed equals the provided ‘seed’, but this won’t be true if seed=None, for example.

Return type

list<bigint>

compute_reward(self, achieved_goal, desired_goal, info)
property unwrapped(self)

Completely unwrap this env.

Returns

The base non-wrapped gym.Env instance

Return type

gym.Env