I2 - MLflow

• inside an AML training script:

import mlflow

# put runs in an experiment for MLflow
mlflow.set_experiment("detector-training")

# submi AML job under that AML experiment
ml_client.jobs.create_or_update(
    job,
    experiment_name="detector-training"
)

# auto-capture supported framework info
mlflow.autolog()

# manual logging
mlflow.log_param(...)
mlflow.log_metric(...)
mlflow.log_artifact(...)