A Guide to Migrating to Vertex AI

Vertex AI is a fully managed machine learning (ML) platform developed and offered by Google Cloud. As part of the Google Cloud AI portfolio, Vertex AI brings together AI Platform and AutoML services under a unified UI and API to simplify the process of building and deploying machine learning models.

Vertex AI supports all features and models available in AutoML and AI Platform. However, the client libraries don't support integration backward compatibility, meaning you have to plan to migrate your resources to benefit from Vertex AI functionalities.

Let's dive into what you need to know about migrating to Vertex AI so you can harness this platform's new features and service improvements.

Understanding Vertex AI

As mentioned, Vertex AI is an ML platform developed and offered by Google Cloud. This tool brings together AI Platform and AutoML services so users can move from experimentation to production faster and stay agile in market conditions.

This platform provides a comprehensive set of tools and services that cover the entire ML lifecycle, from data preparation and model training to deployment and monitoring.

Key Features of Vertex AI

• Simplified ML Workflow: Vertex AI streamlines the ML workflow by bringing together data engineering, data science, and ML engineering into a single platform. This makes it easier for teams of all sizes to collaborate on ML projects and accelerate the delivery of ML applications.
• AutoML: Vertex AI offers AutoML capabilities for tabular, image, text, and video data. AutoML can automatically train ML models without the need for writing code or preparing data splits.
• Low-Code and High-Code Tools: Vertex AI provides both low-code and high-code tools for building ML models. Low-code tools are ideal for beginners, while high-code tools are for experienced developers who want more control over the model training process.
• Fully Managed Infrastructure: Vertex AI is a fully managed platform, which means that Google handles the provisioning, scaling, and maintenance of the underlying infrastructure. This allows you to focus on building and deploying ML models without worrying about the underlying infrastructure.
• MLOps: Vertex AI provides a comprehensive set of MLOps tools for automating, monitoring, and managing ML models. This helps you ensure that your models are deployed and maintained in a consistent and reliable manner.

Benefits of Using Vertex AI

• Accelerated ML Innovation: Vertex AI helps you accelerate the delivery of ML models and applications by providing a streamlined workflow and a set of pre-built tools.
• Reduced Costs: Vertex AI is a fully managed platform that can help you reduce the cost of ML by eliminating the need to manage and maintain your own infrastructure.
• Enhanced Collaboration: Vertex AI provides a common toolset for data engineers, data scientists, and ML engineers, which can help to improve collaboration and communication across teams.

Before Migrating to Vertex AI

Prior to initiating the migration process to Vertex AI, it's crucial to consider the following:

• Resource Duplication: The migration tool creates a replica of your existing resources.
• AutoML and AI Platform Datasets and Models: The migration tool generates equivalent versions of your AutoML and AI Platform datasets and models within Vertex AI. Your original resources remain intact. Multiple migrations of the same resource are permitted, resulting in additional copies.
• Model Deployment: Migrated models are initially in an undeployed state. For data types that support online prediction, you'll need to create an endpoint and deploy the model to it before it can handle online prediction requests.
• AutoML Training Job Creation: When an AutoML model is migrated, the migration tool automatically generates a corresponding training job.
• Data Discrepancies in Migrated Datasets: For certain data types and objectives, migrated datasets may not contain the exact same data as the current dataset. This is due to the following reasons:
• Data Reimport: Datasets for specific data types are reimported from the original data source instead of being copied from the existing dataset. If the original data source has undergone modifications, the migrated dataset will reflect those changes.
• Affected Data Types and Objectives: This reimporting behavior applies to AutoML Natural Language entity extraction datasets, AutoML Video classification and object-tracking datasets, and AutoML Vision object detection datasets.
• Tabular Dataset Export and Referencing: Migrated tabular datasets are exported during the migration process. In Vertex AI, a tabular dataset's data source is referenced rather than directly imported. The migrated tabular dataset is exported from the AutoML Tables dataset, stored as a CSV file in Cloud Storage or a BigQuery table within your project, and subsequently referenced by the migrated dataset.

Steps for Migrating to Vertex AI

With these considerations in mind, let's explore steps for migrating from AutoML and AI Platform.

Migrating from AutoML

1. First, if you're interested, explore the differences between AutoML and Vertex AI. You can also review pricing changes.
2. Take stock of Google Cloud projects, jobs, code, datasets, models, and users with access to AutoML. Use this information to decide which resources you'll migrate. Then, you can ensure the correct users have access to the migrated resources.
3. Review changes to IAM roles here. Then you can update service accounts and authentication. You should also review which resources cannot be migrated here.
4. Next, you can use the Vertex AI migration tool or client libraries and methods for migration. We'll explore more about using the migration tool in this guide.
5. If you're unsure which method to use, identify the usage of AutoML APIs to see which of your applications use them.
6. Next, update your applications to use the Vertex AI API and features.
7. Finally, it's important to plan your quota monitoring. Learn more about Vertex AI quotas and limits.

Migrating from AI Platform

The process for migrating from AI Platform is similar:

1. First, review the differences between AI Platform and Vertex AI here.

2. Take stock of Google Cloud projects, code, jobs, datasets, models, and users with access to AI Platform. This helps you decide which resources to migrate. You can also ensure the correct users have access to the migrated resources.

3. Review changes to IAM roles and update service accounts and authentication.

4. Use the migration tool or Vertex AI client libraries and methods for migration. Don't forget to identify the usage of AI Platform APIs to help determine which applications use them and decide which migration method is best. To do this, go to the APIs and Services Dashboard.

5. Then, update applications and workflows to use Vertex AI API and features. Don't forget to plan your request quota monitoring!

Using the Migration Tool

As mentioned, you can use a migration tool to move your datasets and models to Vertex AI. If this method is right for you, follow these steps:

1. Enable the Vertex AI API on the Vertex AI Dashboard page in Google Cloud Console. Click "Enable the Vertex AI API."
2. On the Vertex AI Dashboard, under "Migrate to Vertex AI," click "Set up migration." Note that if you don't see this option, you don't have resources that can be migrated using this tool.
3. Under "Select resources to migrate," you can select up to 50 assets to migrate. If you have more assets, you can always repeat these steps to migrate them later. If some assets are not shown, it's because they cannot be migrated.
4. Click "Next" and review the assets you want to migrate.
5. Click "Migrate assets."
6. Note that migration can take an hour or more depending on the number of assets being migrated, and you'll receive an email when migration is finished.

Vertex AI Migration Pricing

The migration process itself is free of charge. However, newly created resources resulting from the migration will incur standard Vertex AI pricing. For instance, datasets migrated from AI Platform Data Labeling Service, AutoML Vision, AutoML Video Intelligence, and AutoML Natural Language will be stored in a Cloud Storage bucket, which will incur additional storage costs (see Cloud Storage pricing here).

Following the migration, legacy resources remain accessible for use in AutoML and AI Platform. To optimize resource use and avoid unnecessary expenses, consider decommissioning or deleting legacy resources once you've confirmed the successful migration of your objects.

Migration as a Copy Operation

The migration process involves creating a copy of your existing resources. This means that any changes made to the legacy resource after migration will not be reflected in the migrated resource.

To learn more about pricing, visit the Vertex AI pricing page here.

Look to Promevo for Google Assistance

If you've benefited from Vertex AI features and are looking to integrate the platform further, or you're just beginning your Google journey, look to Promevo for help. Promevo is a certified Premier Partner specializing in all things Google.

From helping teams set up Workspace to providing tailored support for tasks like Vertex AI migration, our experts are seasoned in Google technology and tools. Our goal is to enable you to harness the capabilities of Google and accelerate your company's growth.

We are proud to be a 100% Google-focused partner helping you succeed. Contact us today to get started.

FAQs: Migrating to Vertex AI

Is Vertex AI HIPAA compliant?

Yes, Vertex AI is HIPAA-compliant. This means that Vertex AI is designed and implemented to meet the security and privacy requirements of HIPAA and that Google Cloud will take appropriate steps to protect health information that is stored or processed on its platform.

What are common uses of Vertex AI?

Use cases for Vertex AI include:

• Predictive Analytics: Vertex AI can be used to build predictive models for a variety of purposes, such as fraud detection, customer churn prediction, and supply chain optimization.
• Generative AI: Vertex AI can be used to build generative models for a variety of purposes, such as image generation, text generation, and music generation.
• Search and Conversation: Vertex AI can be used to build search and conversational AI applications that can be used to power chatbots, customer support systems, and search engines.