Google Professional Machine Learning Engineer Question 275
Select 3Google Cloud PlatformYou are working for an e-commerce company that wants to implement a state-of-the-art language model to improve its product recommendation system. Instead of training a model from scratch, you decide to fine-tune an existing foundational model available in Vertex AI Model Garden. The dataset consists of 100,000 labeled product descriptions, and the objective is to predict the product category. Which steps should you take to fine-tune the foundational model effectively using Vertex AI?
- A
Upload your dataset to Google Cloud Storage and create a managed dataset in Vertex AI for preprocessing.
- B
Select a foundational model from Model Garden and start fine-tuning it directly without any preprocessing.
- C
Use Vertex AI Training pipelines to preprocess your dataset and fine-tune the foundational model.
- D
Leverage a pre-trained foundational model from Model Garden and enable transfer learning during the fine-tuning process.
- E
Deploy the foundational model directly to an endpoint without any fine-tuning, as it is already pre-trained.
Show answer and explanation
Correct answers: A, C, D
Explanation
Fine-tuning a foundational model in Vertex AI involves preparing the dataset, preprocessing the data, and leveraging transfer learning to adapt the model to your specific task. Using Vertex AI Training pipelines ensures a streamlined and scalable workflow, while creating a managed dataset ensures the data is properly prepared for training. Deploying the model without fine-tuning would not meet the unique requirements of the product recommendation system.
- A. Correct.
Uploading your dataset to Google Cloud Storage and creating a managed dataset in Vertex AI is necessary to prepare the data for preprocessing and training.
- B. Incorrect.
Skipping preprocessing may lead to suboptimal fine-tuning results, as foundational models often require input data to be in a specific format.
- C. Correct.
Using Vertex AI Training pipelines allows you to preprocess your dataset and fine-tune the model in a structured and scalable way.
- D. Correct.
Enabling transfer learning during fine-tuning leverages the pre-trained knowledge of the foundational model, tailoring it to your specific task.
- E. Incorrect.
Deploying the foundational model without fine-tuning would not meet the custom requirements of your product recommendation system.