Google Professional Machine Learning Engineer Question 110
Select 3Google Cloud PlatformYou are working on a machine learning project in Google Cloud and need to manage a dataset for training a model in Vertex AI. The dataset is stored in a Google Cloud Storage bucket as CSV files. You want to ensure that the data can be easily used for training and evaluation while following best practices for dataset management in Vertex AI. What steps should you take to achieve this?
- A
Create a Vertex AI Dataset resource and import the CSV files from the Cloud Storage bucket.
- B
Manually split the CSV files into separate files for training, validation, and testing before importing them into Vertex AI.
- C
Enable data labeling in Vertex AI for the dataset to automate the labeling process.
- D
Use the Vertex AI Dataset resource to preview and explore the data before training.
- E
Ensure the dataset schema and formatting are consistent with the model requirements before importing into Vertex AI.
Show answer and explanation
Correct answers: A, D, E
Explanation
To manage datasets in Vertex AI effectively, you should first create a Dataset resource and import your data from Google Cloud Storage. This action ensures the data is accessible for model development. Previewing and exploring the dataset helps validate and understand the data before training. Additionally, confirming that the dataset schema is consistent with the model's requirements beforehand helps prevent compatibility issues during training. Splitting data manually and enabling data labeling are not necessary in this scenario, as other tools and processes within Vertex AI can handle these tasks more efficiently.
- A. Correct.
Correct - Creating a Vertex AI Dataset resource and importing the data from Cloud Storage is the recommended way to manage datasets in Vertex AI, as it allows you to use the data seamlessly for training and evaluation.
- B. Incorrect.
Incorrect - While splitting data into training, validation, and testing sets is important, this step can be performed within Vertex AI workflows or using tools like AutoML rather than manually.
- C. Incorrect.
Incorrect - Data labeling is not relevant in this scenario since the dataset is already prepared. Enabling data labeling is only needed when annotations are required for unlabeled data.
- D. Correct.
Correct - Vertex AI provides options to preview and explore datasets, which helps validate the data before moving to the training phase.
- E. Correct.
Correct - Ensuring that the dataset schema and formatting align with model requirements is a best practice to avoid errors during training and evaluation.