Experiment Groups Tracking

Experiment groups tracking is high level API that allow users to create experiments under one group level. This is very useful for running hyperparameters tuning, and this is how Polyaxon organizes experiments when running hyperparameters tuning.

We thought that in some use cases, the users might want to try some algorithms not provided by the platform, to schedule experiment on Polyaxon or to run them on a different platform.

In order to use this API, the user must configure a client

Creating a group

from polyaxon_client.client import PolyaxonClient

client = PolyaxonClient(host='HOST_IP',
                        token='4ee4e5e6080a196d11f637b950fce1587b29ef36')
from polyaxon_client.tracking import Group

group = Group(client=client, project='quick-start')
group.create(tags=['foo', 'bar'], description='New group')

Tracking

Starting experiments

experiment = group.create_experiment(tags=tags, description='Some description')

This will create an experiment under this group, and provide all context (client api, and storage), to each experiment.

You can then refer to experiment tracking

Log statuses

group.log_status(status, message=None)

# Example
group.log_status('starting')

Stop

group.stop()

This is just an easy way to set a stopped status.

Succeeded

group.succeeded()

This is just an easy way to set a succeeded status. End of script will trigger succeeded status automatically

Failed

group.failed()

This is just an easy way to set a failed status. Exception will trigger failed status automatically.