tf_agents.distributions.utils.DistributionSpecV2

Describes a tfp.distribution.Distribution using nested parameters.

event_shape The distribution's event_shape. This is the shape that distribution.sample() returns. distribution.sample(sample_shape) returns tensors of shape sample_shape + event_shape.
dtype The distribution's dtype.
parameters The recursive parameters of the distribution, with tensors having directly been converted to tf.TypeSpec objects.

TypeError If for any entry x in parameters: tf.is_tensor(x).

dtype

event_shape

event_spec

parameters

Methods

__eq__

View source

Return self==value.