datasetinsights.io.tracker

datasetinsights.io.tracker.factory

class datasetinsights.io.tracker.factory.NullTracker

Bases: object

A null tracker that writes nothing. This tracker is used to disable tracking.

class datasetinsights.io.tracker.factory.TrackerFactory

Bases: object

Factory: responsible for creating and holding singleton instance of tracker classes

HOST_ID = 'host'

MLFLOW_TRACKER = 'mlflow'

RUN_FAILED = 'FAILED'

TRACKER = 'tracker'

static create(config=None, tracker_type=None)

create tracker class object.

Parameters

• config – config object, holds server details

• tracker_type – type of tracker

Returns

tracker instance.

datasetinsights.io.tracker.mlflow

class datasetinsights.io.tracker.mlflow.MLFlowTracker(mlflow_config)

Bases: object

MlFlow tracker class, responsible for setting host, client_id and return

initialized mlflow. It also refreshes the access token through daemon thread.

Examples

# Set MLTracking server UI, default is local file

>>> mlflow.set_tracking_uri(TRACKING_URI)
# New run is launched under the current experiment
>>> mlflow.start_run()
# Log a parameter (key-value pair)
>>> mlflow.log_param("param_name", "param_value")
# Log a metric (key-value pair)
>>> mlflow.log_metric("metric_name", "metric_val")
# Log an artifact (output file)
>>> with open("output.txt", "w") as f:
>>>     f.write("Hello world!")
>>> mlflow.log_artifact("output.txt", "run1/output/")
# ends the run launched under the current experiment
>>> mlflow.end_run()

REFRESH_INTERVAL

default refresh token interval

__mlflow

holds initialized mlflow

CLIENT_ID = 'client_id'

DEFAULT_RUN_NAME = 'run-20210222-210309'

EXP_NAME = 'experiment'

HOST_ID = 'host'

REFRESH_INTERVAL = 3000

RUN_NAME = 'run'

get_mlflow()

method to access initialized mlflow :returns: Initialized __mlflow instance.

class datasetinsights.io.tracker.mlflow.RefreshTokenThread(client_id, interval=3000)

Bases: threading.Thread

Its service thread which keeps running till main thread runs

and refresh access tokens.

client_id

MLFlow tracking server client id

interval

duration at which it refreshes the token

run()

Thread run method which keeps running at specified interval till main thread runs.

class datasetinsights.io.tracker.MLFlowTracker(mlflow_config)

Bases: object

MlFlow tracker class, responsible for setting host, client_id and return

initialized mlflow. It also refreshes the access token through daemon thread.

Examples

# Set MLTracking server UI, default is local file

>>> mlflow.set_tracking_uri(TRACKING_URI)
# New run is launched under the current experiment
>>> mlflow.start_run()
# Log a parameter (key-value pair)
>>> mlflow.log_param("param_name", "param_value")
# Log a metric (key-value pair)
>>> mlflow.log_metric("metric_name", "metric_val")
# Log an artifact (output file)
>>> with open("output.txt", "w") as f:
>>>     f.write("Hello world!")
>>> mlflow.log_artifact("output.txt", "run1/output/")
# ends the run launched under the current experiment
>>> mlflow.end_run()

REFRESH_INTERVAL

default refresh token interval

__mlflow

holds initialized mlflow

CLIENT_ID = 'client_id'

DEFAULT_RUN_NAME = 'run-20210222-210309'

EXP_NAME = 'experiment'

HOST_ID = 'host'

REFRESH_INTERVAL = 3000

RUN_NAME = 'run'

get_mlflow()

method to access initialized mlflow :returns: Initialized __mlflow instance.