Azure Data Scientists Associate Practice Exam

The command that allows logging of metrics in MLflow for model training is indeed mlflow.log_metric(). This function is specifically designed to capture and store numerical values during the training process, such as loss, accuracy, or any other performance measure that quantifies the effectiveness of the model over time or across different iterations.

By invoking mlflow.log_metric(), data scientists can keep track of how model performance changes with different parameters or training runs. This is crucial for monitoring progress and making informed decisions about model tuning and improvements. The logged metrics can later be visualized in the MLflow UI, enabling users to compare results from various experiments easily.

To provide context on the other options: mlflow.log_experiment() does not exist; instead, experiments are created and managed through different MLflow functions. mlflow.log_param() is for logging parameters related to the model configuration, such as hyperparameters, but not performance metrics. mlflow.log_model() is utilized for saving and logging the trained model itself, allowing for later deployment or inference, but again, this does not pertain to the logging of performance metrics during training.