Google Professional Machine Learning Engineer
Get started today
Ultimate access to all questions.
Ultimate access to all questions.
You are a machine learning engineer tasked with creating an image classifier using transfer learning based on a pre-trained EfficientNet model. Your training dataset consists of 20,000 images, and you need to retrain the model once per day to keep it updated with new data. Given this requirement, your objective is to minimize the infrastructure cost while ensuring efficient model training and data storage. What platform components and configuration environment should you use?
Explanation:
The correct answer is D. Using an AI Platform Training job with a custom scale tier and 4 V100 GPUs along with Cloud Storage is the most cost-effective and efficient solution for this scenario. Google encourages the use of Cloud Storage over local storage or NFS servers for better scalability and cost management. AI Platform Training allows for automatic scaling of resources which can be adjusted based on the daily training needs, making it a better choice compared to fixed-size VMs or node pools. It also integrates well with Cloud Storage, eliminating overhead and additional complexity.