Custom Model Serving MACHINE-LEARNING-PROFESSIONAL Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   A team is creating a custom PyFunc model that requires a pre-trained tokenizer. They have saved the tokenizer as a file. How should they include it in the model so it is available during serving?

   • ALog the tokenizer file as an artifact using the artifacts parameter in mlflow.pyfunc.log_model.
   • BUpload the tokenizer to the Databricks DBFS and reference it in the code.
   • CStore the tokenizer in a global variable in the model class.
   • DEmbed the tokenizer's code in the predict method.
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A data scientist has deployed a custom PyFunc model to a Databricks Model Serving endpoint. They want to test it from a Python script running outside Databricks. They have the endpoint URL and an access token. Which approach should they use to send a prediction request?

   • AUse the Databricks CLI's mlflow models serve command.
   • BUse the MLflow Deployments SDK's predict function with the endpoint name and input data.
   • CUse the Databricks SQL connector to query the model as if it were a table.
   • DDownload the model artifacts and run the predict method locally.
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A company has deployed a custom PyFunc model to Model Serving. They now need to update the model to a new version. What is the recommended way to update the serving endpoint?

   • AUse the MLflow Deployments SDK to update the model version.
   • BUpdate the endpoint configuration to point to the new model version.
   • CRegister the new version as a separate model and create a new endpoint.
   • DDelete the existing endpoint and create a new one with the new version.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A data scientist has created a custom PyFunc model that includes a custom encoder. They want to ensure that the encoder is available when the model is served. What should they do during model logging?

   • APass the encoder as an artifact to mlflow.pyfunc.log_model.
   • BUpload the encoder to the workspace's shared storage and use a path in the code.
   • CSave the encoder to a separate model and reference it in the serving endpoint.
   • DInclude the encoder in the model's predict method as a global variable.
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A data scientist has deployed a custom PyFunc model that accepts a dictionary as input. They are calling it via the MLflow Deployments SDK and getting a validation error. What is the most likely cause?

   • AThe input format does not match the model's expected schema.
   • BThe endpoint is not configured to accept dictionary inputs.
   • CThe model is not registered in Unity Catalog.
   • DThe SDK is not compatible with the model's Python version.
   Reveal Solution Discussion Select an answer first