Module: tf_agents.specs

View source on GitHub

Module for numpy array and tf.Tensor shape and dtype specifications.

Modules

array_spec module: A class to describe the shape and dtype of numpy arrays.

bandit_spec_utils module: Bandit related tensor spec utilities.

distribution_spec module: Spec definition for tensorflow_probability.Distribution.

tensor_spec module: Utilities related to TensorSpec class.

Classes

class ArraySpec: Describes a numpy array or scalar shape and dtype.

class BoundedArraySpec: An ArraySpec that specifies minimum and maximum values.

Functions

from_spec(...): Maps the given spec into corresponding TensorSpecs keeping bounds.

is_bounded(...)

is_continuous(...)

is_discrete(...)

sample_spec_nest(...): Samples the given nest of specs.

zero_spec_nest(...): Create zero tensors for a given spec.