opentau.configs.types

Type definitions for configuration classes.

This module provides type definitions used across configuration classes, including enumerations for feature types and normalization modes, as well as protocol definitions and dataclasses for policy features.

Classes

DictLike(*args, **kwargs)	Protocol for dictionary-like objects that support item access.
FeatureType(value)	Enumeration of feature types used in policy configurations.
NormalizationMode(value)	Enumeration of normalization modes for features.
PolicyFeature(type, shape)	Configuration for a policy feature.
ROSFeature(ros_topic, topic_attribute, ...)	Configuration for a dataset feature.

class opentau.configs.types.DictLike(*args, **kwargs)

Bases: Protocol

Protocol for dictionary-like objects that support item access.

This protocol defines the interface for objects that can be accessed using dictionary-style indexing with the [] operator.

__init__(*args, **kwargs)

class opentau.configs.types.FeatureType(value)

Bases: str, Enum

Enumeration of feature types used in policy configurations.

ACTION = 'ACTION'

Action features.

ENV = 'ENV'

Environment state features.

STATE = 'STATE'

Robot state features.

VISUAL = 'VISUAL'

Visual/image features.

class opentau.configs.types.NormalizationMode(value)

Bases: str, Enum

Enumeration of normalization modes for features.

IDENTITY = 'IDENTITY'

No normalization (identity transformation).

MEAN_STD = 'MEAN_STD'

Normalize using mean and standard deviation.

MIN_MAX = 'MIN_MAX'

Normalize using min-max scaling.

class opentau.configs.types.PolicyFeature(type: FeatureType, shape: tuple)

Bases: object

Configuration for a policy feature.

This class describes a single feature used by a policy, including its type and shape information.

Parameters:

• type – The type of feature (STATE, VISUAL, ENV, or ACTION).

• shape – The shape of the feature as a tuple.

__init__(type: FeatureType, shape: tuple) → None

shape: tuple

type: FeatureType

class opentau.configs.types.ROSFeature(ros_topic: str, topic_attribute: str, enum_values: str, dtype: str, shape: tuple)

Bases: object

Configuration for a dataset feature.

This class describes a single feature used in a dataset, mapping a ROS topic to a LeRobot dataset feature.

ros_topic

The ROS topic name (e.g., “/joint_states”).

Type:

str

topic_attribute

The attribute of the ROS message to extract (e.g., “position”, “velocity”).

Type:

str

enum_values

The list of possible values for the feature.

Type:

list[str] | None

dtype

The data type of the feature (e.g., “float32”).

Type:

str

shape

The shape of the feature data (e.g., (25,)).

Type:

tuple

__init__(ros_topic: str, topic_attribute: str, enum_values: str, dtype: str, shape: tuple) → None

dtype: str

enum_values: str

ros_topic: str

shape: tuple

topic_attribute: str