Advanced MLflow Usage MACHINE-LEARNING-PROFESSIONAL Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A machine learning engineer is training a multi-stage model: first a feature selector, then a classifier. They want to track metrics for each stage separately, while also recording an overall experiment-level metric. The engineer is using MLflow's Python API within a single training script. Which approach correctly organizes the runs so that stage-level metrics are grouped under a single parent run?

   • AStart a parent run with mlflow.start_run(), then within it call mlflow.start_run(nested=True) for each stage, logging stage metrics inside the nested contexts.
   • BLog all stage metrics to the same run and use mlflow.set_tag() with stage names to differentiate them.
   • CUse mlflow.start_run(run_id=parent_run_id) inside each stage to log stage metrics directly to the parent run.
   • DStart a separate top-level run for each stage using mlflow.start_run() without the nested parameter, then log the overall metric to the last run.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   What is the purpose of implementing feature engineering inside a custom MLflow model's predict method?

   • ATo allow the model to be retrained automatically when new features arrive
   • BTo reduce the model's training time by offloading computation to inference
   • CTo enable real-time feature transformation on incoming data at inference time
   • DTo ensure the same preprocessing is applied during both training and inference
   Reveal Solution Discussion Select an answer first
4. 8foundation · easy

   Which MLflow function is used to store a custom Python object, such as a serialized preprocessing transformer, as an artifact in the current run?

   • Amlflow.set_tag(key, value)
   • Bmlflow.log_metric(key, value)
   • Cmlflow.log_artifact(local_path)
   • Dmlflow.log_param(key, value)
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A team is deploying a model that requires a custom transformation: the raw input contains a categorical feature with values that were label-encoded during training. The encoding mapping is stored in a JSON file. The team wants the deployed model to apply this mapping inside predict. Which MLflow PythonModel implementation correctly loads the mapping and applies it?

   • AIn the __init__ method, load the JSON mapping into self.encoder; in predict, apply self.encoder to the categorical column before passing data to the underlying model.
   • BIn the predict method, load the JSON file from disk every time the method is called, then apply the mapping.
   • CLoad the JSON mapping in the training script and pass it as a parameter to the model's constructor when logging the model.
   • DApply the label encoding in the training script and log the encoded training data as an artifact, then serve the model on pre-encoded data.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   What is the primary benefit of using nested runs in MLflow?

   • AThey reduce the storage cost of logged metrics by sharing parent run data.
   • BThey allow parallel execution of multiple experiments on different clusters.
   • CThey organize related runs into a parent-child structure for better experiment tracking and comparison.
   • DThey enable automatic hyperparameter tuning without additional code.
   Reveal Solution Discussion Select an answer first