nvflare.apis.analytix module

class AnalyticsData(key: str, value, data_type: AnalyticsDataType, sender: Tracker = Tracker.TORCH_TB, kwargs: dict | None = None, step: int | None = None, path: str | None = None)

Bases: object

This class defines AnalyticsData format.

It is a wrapper to provide to/from DXO conversion.

Parameters:

• key (str) – tag name

• value – value

• data_type (AnalyticDataType) – type of the analytic data.

• sender (Tracker) – Syntax of Tracker such as Tensorboard or MLFLow

• kwargs (optional, dict) – additional arguments to be passed.

classmethod convert_data_type(sender_data_type: AnalyticsDataType, sender: Tracker, receiver: Tracker) → AnalyticsDataType

classmethod from_dxo(dxo: DXO, receiver: Tracker = Tracker.TORCH_TB)

Generates the AnalyticsData from DXO object.

Parameters:

• receiver – type experiment tacker, default to Tensorboard. TrackerType.TORCH_TB

• sender – type experiment tacker, default to Tensorboard. TrackerType.TORCH_TB

• dxo (DXO) – The DXO object to convert.

Returns:

AnalyticsData object

to_dxo()

Converts the AnalyticsData to DXO object.

Returns:

DXO object

class AnalyticsDataType(value)

Bases: Enum

An enumeration.

IMAGE = 'IMAGE'

INIT_DATA = 'INIT_DATA'

LOG_RECORD = 'LOG_RECORD'

METRIC = 'METRIC'

METRICS = 'METRICS'

MODEL = 'MODEL'

PARAMETER = 'PARAMETER'

PARAMETERS = 'PARAMETERS'

SCALAR = 'SCALAR'

SCALARS = 'SCALARS'

TAG = 'TAG'

TAGS = 'TAGS'

TEXT = 'TEXT'