mbodied.types package∞

Subpackages∞

Submodules∞

mbodied.types.geometry module∞

class mbodied.types.geometry.AbsolutePose(datum=None, *, x: float = 0, y: float = 0, z: float = 0, roll: float = 0, pitch: float = 0, yaw: float = 0)[source]∞

Bases: Pose

Absolute pose of the robot in 3D space.

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'forbid', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_fields: ClassVar[dict[str, FieldInfo]] = {'pitch': FieldInfo(annotation=float, required=False, default=0, description='Absolute pitch about the Y-axis in radians. Positive pitch is down. Motion type: ABSOLUTE', json_schema_extra={'bounds': None, 'motion_type': 'ABSOLUTE', 'shape': None}), 'roll': FieldInfo(annotation=float, required=False, default=0, description='Absolute roll about the X-axis in radians. Positive roll is clockwise. Motion type: ABSOLUTE', json_schema_extra={'bounds': None, 'motion_type': 'ABSOLUTE', 'shape': None}), 'x': FieldInfo(annotation=float, required=False, default=0, description='Absolute X position in 3D space. +x is forward; -x is backward. Motion type: ABSOLUTE', json_schema_extra={'bounds': None, 'motion_type': 'ABSOLUTE', 'shape': None}), 'y': FieldInfo(annotation=float, required=False, default=0, description='Absolute Y position in 3D space. +y is left; -y is right. Motion type: ABSOLUTE', json_schema_extra={'bounds': None, 'motion_type': 'ABSOLUTE', 'shape': None}), 'yaw': FieldInfo(annotation=float, required=False, default=0, description='Absolute yaw about the Z-axis in radians. Positive yaw is left. Motion type: ABSOLUTE', json_schema_extra={'bounds': None, 'motion_type': 'ABSOLUTE', 'shape': None}), 'z': FieldInfo(annotation=float, required=False, default=0, description='Absolute Z position in 3D space. +z is up; -z is down. Motion type: ABSOLUTE', json_schema_extra={'bounds': None, 'motion_type': 'ABSOLUTE', 'shape': None})}∞

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

pitch: float∞

roll: float∞

x: float∞

y: float∞

yaw: float∞

z: float∞

class mbodied.types.geometry.LocationAngle(datum=None, *, x: float = 0.0, y: float = 0.0, theta: float = None)[source]∞

Bases: Pose3D

Alias for Pose3D. A 2D+1 space representing x, y, and theta.

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'forbid', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_fields: ClassVar[dict[str, FieldInfo]] = {'theta': FieldInfo(annotation=float, required=False, default_factory=<lambda>, description='Rotation (radians) about the z axis. Positive is counter-clockwise. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None}), 'x': FieldInfo(annotation=float, required=False, default=0.0, description='X position in 2D space. +x is forward; -x is backward. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None}), 'y': FieldInfo(annotation=float, required=False, default=0.0, description='Y position in 2D space. +y is left; -y is right. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None})}∞

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

class mbodied.types.geometry.Pose(datum=None, *, x: float = 0, y: float = 0, z: float = 0, roll: float = 0, pitch: float = 0, yaw: float = 0)[source]∞

Bases: Pose6D

Alias for Pose6D. A movement for a 6D space representing x, y, z, roll, pitch, and yaw.

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'forbid', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_fields: ClassVar[dict[str, FieldInfo]] = {'pitch': FieldInfo(annotation=float, required=False, default=0, description='Pitch about the Y-axis in radians. Positive pitch is down. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None}), 'roll': FieldInfo(annotation=float, required=False, default=0, description='Roll about the X-axis in radians. Positive roll is clockwise. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None}), 'x': FieldInfo(annotation=float, required=False, default=0, description='X position in 3D space. +x is forward; -x is backward. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None}), 'y': FieldInfo(annotation=float, required=False, default=0, description='Y position in 3D space. +y is left; -y is right. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None}), 'yaw': FieldInfo(annotation=float, required=False, default=0, description='Yaw about the Z-axis in radians. Positive yaw is left. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None}), 'z': FieldInfo(annotation=float, required=False, default=0, description='Z position in 3D space. +z is up; -z is down. Motion type: UNSPECIFIED', json_schema_extra={'bounds': None, 'motion_type': 'UNSPECIFIED', 'shape': None})}∞

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

class mbodied.types.geometry.Pose3D(datum=None, *, x: float = 0.0, y: float = 0.0, theta: float = None)[source]∞

Bases: Motion

Action for a 2D+1 space representing x, y, and theta.

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'forbid', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

theta: float∞

x: float∞

y: float∞

class mbodied.types.geometry.Pose6D(datum=None, *, x: float = 0, y: float = 0, z: float = 0, roll: float = 0, pitch: float = 0, yaw: float = 0)[source]∞

Bases: Motion

Movement for a 6D space representing x, y, z, roll, pitch, and yaw.

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'forbid', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

pitch: float∞

roll: float∞

x: float∞

y: float∞

yaw: float∞

z: float∞

class mbodied.types.geometry.RelativePose(datum=None, *, x: float = 0, y: float = 0, z: float = 0, roll: float = 0, pitch: float = 0, yaw: float = 0)[source]∞

Bases: Pose

Relative pose of the robot in 3D space.

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'forbid', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_fields: ClassVar[dict[str, FieldInfo]] = {'pitch': FieldInfo(annotation=float, required=False, default=0, description='Relative pitch about the Y-axis in radians. Positive pitch is down. Motion type: RELATIVE', json_schema_extra={'bounds': None, 'motion_type': 'RELATIVE', 'shape': None}), 'roll': FieldInfo(annotation=float, required=False, default=0, description='Relative roll about the X-axis in radians. Positive roll is clockwise. Motion type: RELATIVE', json_schema_extra={'bounds': None, 'motion_type': 'RELATIVE', 'shape': None}), 'x': FieldInfo(annotation=float, required=False, default=0, description='Relative X position in 3D space. +x is forward; -x is backward. Motion type: RELATIVE', json_schema_extra={'bounds': None, 'motion_type': 'RELATIVE', 'shape': None}), 'y': FieldInfo(annotation=float, required=False, default=0, description='Relative Y position in 3D space. +y is left; -y is right. Motion type: RELATIVE', json_schema_extra={'bounds': None, 'motion_type': 'RELATIVE', 'shape': None}), 'yaw': FieldInfo(annotation=float, required=False, default=0, description='Relative yaw about the Z-axis in radians. Positive yaw is left. Motion type: RELATIVE', json_schema_extra={'bounds': None, 'motion_type': 'RELATIVE', 'shape': None}), 'z': FieldInfo(annotation=float, required=False, default=0, description='Relative Z position in 3D space. +z is up; -z is down. Motion type: RELATIVE', json_schema_extra={'bounds': None, 'motion_type': 'RELATIVE', 'shape': None})}∞

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

pitch: float∞

roll: float∞

x: float∞

y: float∞

yaw: float∞

z: float∞

mbodied.types.message module∞

class mbodied.types.message.Message(content: Any | None = None, role: Literal['user', 'assistant', 'system'] = 'user')[source]∞

Bases: Sample

Single completion sample space.

Message can be text, image, list of text/images, Sample, or other modality.

role∞

The role of the message sender (user, assistant, or system).

Type:: Literal[‘user’, ‘assistant’, ‘system’]

content∞

The content of the message, which can be of various types.

Type:: Any | None

content: Any | None∞

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'allow', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_fields: ClassVar[dict[str, FieldInfo]] = {'content': FieldInfo(annotation=Union[Any, NoneType], required=False, default_factory=list), 'role': FieldInfo(annotation=Literal['user', 'assistant', 'system'], required=False, default='user')}∞

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

role: Literal['user', 'assistant', 'system']∞

classmethod supports(arg: Any) → bool[source]∞

Checks if the argument type is supported by the Message class.

Parameters:: arg – The argument to be checked.
Returns:: True if the argument type is supported, False otherwise.

mbodied.types.ndarray module∞

mbodied.types.sample module∞

class mbodied.types.sample.Sample(datum=None, **data)[source]∞

Bases: BaseModel

A base model class for serializing, recording, and manipulating arbitray data.

classmethod default_sample(output_type='Sample') → Sample | Dict[str, Any][source]∞

Generate a default Sample instance from its class attributes. Useful for padding.

This is the “no-op” instance and should be overriden as needed.

classmethod default_space() → Dict[source]∞: Return the Gym space for the Sample class based on its class attributes.

classmethod default_value() → Sample[source]∞

Get the default value for the Sample instance.

Returns:: The default value for the Sample instance.
Return type:: Sample

dict(exclude_none=True, exclude: set[str] = None) → Dict[str, Any][source]∞

Return the Sample object as a dictionary with None values excluded.

Parameters:

exclude_none (bool, optional) – Whether to exclude None values. Defaults to True.
exclude (set[str], optional) – Set of attribute names to exclude. Defaults to None.

Returns:

Dictionary representation of the Sample object.

Return type:

Dict[str, Any]

flatten(output_type: Literal['dict', 'np', 'pt', 'list'] = 'dict', non_numerical: Literal['ignore', 'forbid', 'allow'] = 'allow') → Dict[str, Any] | ndarray | Tensor | List[source]∞

classmethod from_flat_dict(flat_dict: Dict[str, Any], schema: Dict = None) → Sample[source]∞: Initialize a Sample instance from a flattened dictionary.

classmethod from_space(space: Space) → Sample[source]∞: Generate a Sample instance from a Gym space.

classmethod init_from(d: Any, pack=False) → Sample[source]∞

model_computed_fields: ClassVar[dict[str, ComputedFieldInfo]] = {}∞: A dictionary of computed field names and their corresponding ComputedFieldInfo objects.

model_config: ConfigDict = {'arbitrary_types_allowed': True, 'extra': 'allow', 'from_attributes': True, 'use_enum_values': False, 'validate_assignment': False}∞: Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_field_info(key: str) → FieldInfo[source]∞: Get the FieldInfo for a given attribute key.

model_fields: ClassVar[dict[str, FieldInfo]] = {}∞

Metadata about the fields defined on the model, mapping of field names to [FieldInfo][pydantic.fields.FieldInfo].

This replaces Model.__fields__ from Pydantic V1.

static obj_to_schema(value: Any) → Dict[source]∞

Generates a simplified JSON schema from a dictionary.

Parameters:: value (Any) – An object to generate a schema for.
Returns:: A simplified JSON schema representing the structure of the dictionary.
Return type:: dict

classmethod pack_from(samples: List[Sample | Dict]) → Sample[source]∞

Pack a list of samples into a single sample with lists for attributes.

Parameters:: samples (List[Union[Sample, Dict]]) – List of samples or dictionaries.
Returns:: Packed sample with lists for attributes.
Return type:: Sample

random_sample() → Sample[source]∞

Generate a random Sample instance based on its instance attributes. Omits None values.

Override this method in subclasses to customize the sample generation.

classmethod read(data: Any) → Sample[source]∞

Read a Sample instance from a JSON string or dictionary or path.

Parameters:: data (Any) – The JSON string or dictionary to read.
Returns:: The read Sample instance.
Return type:: Sample

schema(resolve_refs: bool = True, include_descriptions=False) → Dict[source]∞

Returns a simplified json schema.

Removing additionalProperties, selecting the first type in anyOf, and converting numpy schema to the desired type. Optionally resolves references.

Parameters:

schema (dict) – A dictionary representing the JSON schema.
resolve_refs (bool) – Whether to resolve references in the schema. Defaults to True.
include_descriptions (bool) – Whether to include descriptions in the schema. Defaults to False.

Returns:

A simplified JSON schema.

Return type:

dict

space() → Dict[source]∞

Return the corresponding Gym space for the Sample instance based on its instance attributes. Omits None values.

Override this method in subclasses to customize the space generation.

classmethod space_for(value: Any, max_text_length: int = 1000, info: Annotated = None) → Space[source]∞

Default Gym space generation for a given value.

Only used for subclasses that do not override the space method.

to(container: Any) → Any[source]∞

Convert the Sample instance to a different container type.

Parameters:

container (Any) – The container type to convert to. Supported types are
'dict' (pytorch), 'space' (gym.space)
'list' (pytorch), 'space' (gym.space)
'np' (pytorch), 'space' (gym.space)
'pt' (pytorch), 'space' (gym.space)
'schema' (datasets.Dataset)
'json' (datasets.Dataset)
'hf' (datasets.Dataset)

Returns:

The converted container.

Return type:

Any

classmethod unflatten(one_d_array_or_dict, schema=None) → Sample[source]∞

Unflatten a one-dimensional array or dictionary into a Sample instance.

If a dictionary is provided, its keys are ignored.

Parameters:

one_d_array_or_dict – A one-dimensional array or dictionary to unflatten.
schema – A dictionary representing the JSON schema. Defaults to using the class’s schema.

Returns:

The unflattened Sample instance.

Return type:

Sample

Examples

>>> sample = Sample(x=1, y=2, z={"a": 3, "b": 4}, extra_field=5)
>>> flat_list = sample.flatten()
>>> print(flat_list)
[1, 2, 3, 4, 5]
>>> Sample.unflatten(flat_list, sample.schema())
Sample(x=1, y=2, z={'a': 3, 'b': 4}, extra_field=5)

unpack(to_dicts=False) → List[Sample | Dict][source]∞: Unpack the packed Sample object into a list of Sample objects or dictionaries.

mbodied.types package∞

Subpackages∞

Submodules∞

mbodied.types.geometry module∞

mbodied.types.message module∞

mbodied.types.ndarray module∞

mbodied.types.sample module∞

Module contents∞