nerdexam
Microsoft

DP-100 · Question #351

You are using the Azure Machine Learning SDK to run a training experiment that trains a classification model and calculates its accuracy metric. The model will be retrained each month as new data is…

The correct answer is B. Specify the model framework version when registering the model, and only register subsequent. To conditionally register new model versions based on improved accuracy when retraining, MLflow allows you to track metrics and then programmatically compare and update model aliases like 'Production' or 'Staging' to point to the best-performing version.

Train and deploy models

Question

You are using the Azure Machine Learning SDK to run a training experiment that trains a classification model and calculates its accuracy metric. The model will be retrained each month as new data is available. You must register the model for use in a batch inference pipeline. You need to register the model by using mlflow and ensure that the models created by subsequent retraining experiments are registered only if their accuracy is higher than the currently registered model. What should you do?

Options

  • ARegister a metric named accuracy with the accuracy metric as a value when registering the
  • BSpecify the model framework version when registering the model, and only register subsequent
  • CRegister the model with the same name each time regardless of accuracy, and always use the
  • DSpecify a different name for the model each time you register it.

How the community answered

(42 responses)
  • A
    7% (3)
  • B
    79% (33)
  • C
    12% (5)
  • D
    2% (1)

Why each option

To conditionally register new model versions based on improved accuracy when retraining, MLflow allows you to track metrics and then programmatically compare and update model aliases like 'Production' or 'Staging' to point to the best-performing version.

ARegister a metric named accuracy with the accuracy metric as a value when registering the

Registering a metric named accuracy with the model only stores the metric, it does not inherently enforce conditional registration or promotion based on that metric's value.

BSpecify the model framework version when registering the model, and only register subsequentCorrect

While the option is truncated, it implies a workflow where model versions are managed, and new models are registered or promoted to production aliases only if their performance (like accuracy) is superior. MLflow's model registry supports tracking versions and aliases (stages), enabling programmatic logic to compare metrics and update the 'Production' alias to the best-performing model.

CRegister the model with the same name each time regardless of accuracy, and always use the

Registering with the same name regardless of accuracy and always using the latest version contradicts the goal of *only* registering or promoting if accuracy is higher.

DSpecify a different name for the model each time you register it.

Specifying a different name for the model each time would create many distinct models in the registry, making it difficult to manage versions of a single logical model and to enforce conditional updates efficiently.

Concept tested: MLflow model versioning and conditional registration

Source: https://mlflow.org/docs/latest/model-registry.html#adding-a-model-stage-transition

Topics

#MLflow#Model Registration#Model Versioning#Azure Machine Learning

Community Discussion

No community discussion yet for this question.

Sign up to join the discussion
Full DP-100 Practice