Google Professional Machine Learning Engineer Question 173
Select 3Google Cloud PlatformYou are tasked with building a machine learning model to predict customer churn using Vertex AI Workbench. The dataset is stored in BigQuery, and you need to develop the model using TensorFlow. Which of the following steps should you perform to ensure an efficient workflow?
- A
Load the BigQuery dataset directly into a pandas DataFrame for preprocessing.
- B
Use TensorFlow I/O library to stream data from BigQuery for training.
- C
Create a Vertex AI Workbench managed notebook with pre-installed TensorFlow libraries.
- D
Export the BigQuery data to a CSV file and upload it to Cloud Storage for TensorFlow to read.
- E
Enable GPU support in the Vertex AI Workbench notebook if TensorFlow training involves large datasets.
Show answer and explanation
Correct answers: B, C, E
Explanation
When developing models in Vertex AI Workbench using TensorFlow, it's essential to leverage efficient data handling and training processes. TensorFlow I/O enables seamless integration with BigQuery for large datasets, avoiding unnecessary data export steps. Using a managed notebook with pre-installed TensorFlow libraries simplifies setup, and enabling GPU support ensures faster training for large datasets. These steps ensure an optimized workflow that aligns with Google Cloud best practices.
- A. Incorrect.
While pandas DataFrames are useful for small-scale data manipulation, they are not efficient for large datasets stored in BigQuery. TensorFlow I/O or Dataflow pipelines are better for large-scale data handling.
- B. Correct.
Using TensorFlow I/O to stream data from BigQuery is an efficient method for handling large datasets without having to export them to intermediate storage like CSV files.
- C. Correct.
Vertex AI Workbench managed notebooks come with pre-installed frameworks like TensorFlow, which simplifies the development process and ensures compatibility.
- D. Incorrect.
Exporting to CSV and uploading to Cloud Storage introduces unnecessary overhead and is not the most efficient approach when working with BigQuery and TensorFlow.
- E. Correct.
Enabling GPU support is essential for training large-scale models efficiently, especially with frameworks like TensorFlow that leverage hardware acceleration.