Google Professional Machine Learning Engineer
Get started today
Ultimate access to all questions.
You are tasked with automating a machine learning training pipeline that processes video frames into sliced images with bounding boxes around specific objects. The pipeline must ingest and preprocess data stored in Cloud Storage, train and fine-tune hyperparameters of the object detection model, and deploy the model to an endpoint. Given the requirements to minimize cluster administration and ensure scalability, which of the following approaches is the BEST to streamline this process? Choose one correct option.
You are tasked with automating a machine learning training pipeline that processes video frames into sliced images with bounding boxes around specific objects. The pipeline must ingest and preprocess data stored in Cloud Storage, train and fine-tune hyperparameters of the object detection model, and deploy the model to an endpoint. Given the requirements to minimize cluster administration and ensure scalability, which of the following approaches is the BEST to streamline this process? Choose one correct option.
Explanation:
ā Correct Answer: Use Vertex AI Pipelines with Kubeflow Pipelines SDK
Vertex AI Pipelines offers a fully managed orchestration service, built on Kubeflow Pipelines, eliminating the need for manual infrastructure management. It seamlessly supports custom components, hyperparameter tuning, and model deployment steps, aligning with the requirement to minimize cluster administration.
ā Incorrect Options:
- Utilize Cloud Composer for orchestration: While Cloud Composer is suitable for general workflows, it lacks specialized support for ML pipeline steps like tuning and model deployment on Vertex AI.
- Use Vertex AI Pipelines with TensorFlow Extended (TFX) SDK: TFX is only partially supported on Vertex AI Pipelines, requiring additional custom code for integration.
- Deploy Kubeflow Pipelines on Google Kubernetes Engine: This approach requires manual management of GKE clusters, contrary to the goal of minimal administration.