comet_ml.integration.sagemaker ¶

This function retrieves the last training job and logs its data as a Comet Experiment. The training job must be in completed status.

This API is in BETA.

Parameters:

• api_key (Optional[str], default: None ) –

  The Comet API Key. If not provided must be configured in another way

• workspace (Optional[str], default: None ) –

  Attach the experiment to a project that belongs to this workspace. If not provided must be configured in another way

• project_name (Optional[str], default: None ) –

  Send the experiment to a specific project. If not provided must be configured in another way

• experiment (Optional[APIExperiment], default: None ) –

  Pass an existing APIExperiment to be used for logging.

This function logs the training job identified by the sagemaker_job_name as a Comet Experiment. The training job must be in completed status.

This API is in BETA.

Parameters:

• sagemaker_job_name (str) –

  The name of the Sagemaker Training Job.

• api_key (Optional[str], default: None ) –

  The Comet API Key. If not provided must be configured in another way.

• workspace (Optional[str], default: None ) –

  Attach the experiment to a project that belongs to this workspace. If not provided must be configured in another way.

• project_name (Optional[str], default: None ) –

  Send the experiment to a specific project. If not provided must be configured in another way.

• experiment (Optional[APIExperiment], default: None ) –

  Pass an existing APIExperiment to be used for logging.

This function retrieves the last training job from an sagemaker.estimator.Estimator object and log its data as a Comet Experiment. The training job must be in completed status.

This API is in BETA.

Parameters:

• estimator (Estimator) –

  The estimator object that was used to start the training job.

• api_key (Optional[str], default: None ) –

  The Comet API Key. If not provided must be configured in another way.

• workspace (Optional[str], default: None ) –

  Attach the experiment to a project that belongs to this workspace. If not provided must be configured in another way.

• project_name (Optional[str], default: None ) –

  Send the experiment to a specific project. If not provided must be configured in another way.

• experiment (Optional[APIExperiment], default: None ) –

  Pass an existing APIExperiment to be used for logging.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

import sagemaker

from comet_ml.integration.sagemaker import log_sagemaker_training_job_v1

estimator = sagemaker.estimator.Estimator(
    training_image,
    role,
    instance_count=instance_count,
    instance_type=instance_type,
    output_path=s3_output_location,
)

estimator.fit(s3_input_location)

api_experiment = log_sagemaker_training_job_v1(
    estimator, api_key=API_KEY, workspace=WORKSPACE, project_name=PROJECT_NAME
)