concepts.dsl.tensor_state.TensorState

class TensorState

Bases: TensorStateBase

A state representation is essentially a mapping from feature names to tensors.

Methods

clone()
clone_internals()	Clone the internals.
extra_state_str_after()	Extra state string.
extra_state_str_before()	Extra state string before the features.
summary_string()	Get a summary string of the state.

Attributes

batch_dims	The number of batchified dimensions.
features
internals	Additional internal information about the state.

__getitem__(name)

Parameters:

name (str)

__init__(features=None, batch_dims=0, internals=None)

Initialize a state.

Parameters:

• features (Mapping[str, Any] | TensorValueDict | None) – the features of the state.

• batch_dims (int) – the number of batch dimensions.

• internals (Dict[str, Any] | None) – the internal state of the state.

__new__(**kwargs)

clone()

Return type:

TensorState

clone_internals()

Clone the internals.

extra_state_str_after()

Extra state string.

Return type:

str

extra_state_str_before()

Extra state string before the features.

Return type:

str

summary_string()

Get a summary string of the state. The main difference between this and __str__ is that this function only formats the shape of intermediate tensors.

Return type:

str

property batch_dims: int

The number of batchified dimensions. For the basic State, it should be 0.

property features: TensorValueDict

property internals: Dict[str, Any]

Additional internal information about the state.