Logging metrics

When you log a measure to a run, that metric is saved in the experiment's run record.All logged metrics should be shown and kept in a history.

log

Log a single metric value to a run.

from azureml.core import Run
run = Run.get_context()
run.log('metric-name', metric_value)

You can log the same metric multiple times within a run; the results will be displayed as a chart.

log_row

Log a metric with multiple columns.

from azureml.core import Run
run = Run.get_context()
run.log_row("Y over X", x=1, y=0.4)

:::info More logging options These are arguably the most often used APIs for logging metrics, but a comprehensive list, including logging lists, tables, and pictures, can be found here. :::

Viewing metrics

Metrics will be provided in the Azure ML Studio by default. Choose a location for your run, such as either simply going to ml.azure.com, or by downloading the SDK:

run.get_workspace_url()

Via the SDK

Viewing metrics in a run

metrics = run.get_metrics()
# metrics is of type Dict[str, List[float]] mapping mertic names
# to a list of the values for that metric in the given run.

metrics.get('metric-name')
# list of metrics in the order they were recorded

metrics_including_child_runs = run.get_metrics(recursive=True)
# set 'recursive' boolean parameter to True to also get
# the metrics for child runs 

To view all recorded values for a given metric my-metric in a given experiment my-experiment:

experiments = ws.experiments
# of type Dict[str, Experiment] mapping experiment names the
# corresponding Experiment

exp = experiments['my-experiment']
for run in exp.get_runs():
    metrics = run.get_metrics()
    
    my_metric = metrics.get('my-metric')
    if my_metric:
        print(my_metric)

Examples

Logging with MLFlow

Use MLFlow to log metrics in Azure ML.

from azureml.core import Run

# connect to the workspace from within your running code
run = Run.get_context()
ws = run.experiment.workspace

# workspace has associated ml-flow-tracking-uri
mlflow_url = ws.get_mlflow_tracking_uri()

Logging with PyTorch Lightning

This examples:

• Includes Lightning's TensorBoardLogger -you can refer more about it in [here]
• Sets up Lightning's MLFlowLogger using AzureML Run.get_context()
  • Only adds this logger when used as part of an Azure ML run

import pytorch_lightning as pl

run = None
try:
    from azureml.core.run import Run, _OfflineRun
    run = Run.get_context()
    if isinstance(run, _OfflineRun):
        run = None
except ImportError:
    print("Couldn't import azureml.core.run.Run")

def get_logger():
    tb_logger = pl.loggers.TensorBoardLogger('logs/')
    logger = [tb_logger]
    
    if run is not None:
        mlflow_url = run.experiment.workspace.get_mlflow_tracking_uri()
        mlf_logger = pl.loggers.MLFlowLogger(
          experiment_name=run.experiment.name,
          tracking_uri=mlflow_url,
          )
        mlf_logger._run_id = run.id
        logger.append(mlf_logger)

    return logger

Now include this logger in the lightning Trainer class:

logger = get_logger()

trainer = pl.Trainer.from_argparse_args(
    args=args,
    logger=logger,
    )
trainer.fit(model)