gxm.spaces

Spaces for defining action and observation spaces.

class Box(low, high, shape)

Bases: Space

A bounded box in \(\mathbb{R}^n\).

contains(x)

Check whether specific object is within space.

Parameters:

x (Any) – Object to be checked.

Return type:

Array

Returns:

Whether the object is within the space.

high: Array

Upper bound of the box.

low: Array

Lower bound of the box.

property n: int

sample(key, shape=())

Sample uniformly from the box.

Parameters:

• key (Array) – JAX random key.

• shape (Tuple[int, ...]) – Shape of the sample to be drawn.

Return type:

Any

Returns:

Sample drawn from the box.

shape: Tuple[int, ...]

Shape of the box.

class Discrete(n)

Bases: Space

A discrete space consiting of \(\{0, 1, ..., n-1\}\).

contains(x)

Check whether specific object is within space.

Parameters:

x (Any) – The object to be checked.

Return type:

Array

Returns:

Whether the object is contained in the space.

property n: int

sample(key, shape=())

Sample random action uniformly from \(\{0, 1, ..., n-1\}\).

Parameters:

• key (Array) – A JAX random key used for sampling.

• shape (Tuple[int, ...]) – The shape of the returned sample.

Return type:

Array

Returns:

The sampled value.

class Space

Bases: object

Abstract base class for action and observation spaces.

contains(x)

Return type:

Any

property n: int

sample(key, shape=())

Return type:

Any

class Tree(spaces)

Bases: Space

Space composed of multiple subspaces in a tree structure.

contains(x)

Check whether dimensions of object are within subspace.

Parameters:

x (Array) – Object to be checked.

Return type:

bool

Returns:

True if object is within all subspaces, False otherwise.

property n: int

sample(key, shape=())

Sample random action from all subspaces, retaining the original structure.

Parameters:

• key (Array) – JAX random key.

• shape (Tuple[int, ...]) – Shape of the sample to be drawn from each subspace.

Return type:

Any

Returns:

A tree-structured object with samples from each subspace.