What is Vertex AI?
Vertex AI is an advanced platform developed by Google Cloud that enables the creation, training, deployment, and management of machine learning (ML) models. It is designed to simplify and automate the entire machine learning process, offering a set of tools that can be used by both ML experts and those who are just beginning their journey with ML.
The main features of Vertex AI include:
- Model Training: Allows for training ML models on large datasets using various algorithms and frameworks, such as TensorFlow, PyTorch, and scikit-learn.
- Automated ML (AutoML): Vertex AI includes AutoML features that enable automatic model training without the need for manual hyperparameter tuning, making it easier for users without advanced technical knowledge.
- Model Deployment: The platform offers tools for easy deployment of models to production, allowing for scaling and real-time performance monitoring of the models.
- ML Pipelines: Vertex AI supports the creation and management of complex ML pipelines, enabling the automation of the entire machine learning process from data processing, through model training, to deployment and monitoring.
- Data Management: Tools for managing data allow for easy storage, processing, and analysis, which are crucial for effective ML model training.
- Model Monitoring and Management: The platform provides tools for monitoring model performance and managing versions, helping to maintain high-quality models in production.
Vertex AI also integrates with other Google Cloud services, allowing seamless access to Google's extensive ecosystem of cloud tools and services. This makes it a powerful tool for businesses and developers who want to fully leverage machine learning capabilities in their projects.
Where to find it?
The main functionality is located in the Feed Manager, and to access it, it is necessary to activate the Vertex AI integration in the Integration Center.
Procedure for activating Vertex AI integration
To activate the integration, go to the Integration Center and select the Vertex AI function from the "Other" tab.
After clicking the "Activate" button, proceed to the configuration. In the iPresso system panel, you need to provide two parameters:
- Project ID - available from the Google Cloud panel. This is the project ID to be used for Google recommendations through Vertex AI.
- Application Default Credentials – these are the default credentials for the application from the Google Cloud panel, stored in the internal Vault service, which holds all the references provided by clients. During the configuration process, the value of this field is not visible
Obtaining Data for Integration Configuration
For documentation on obtaining credentials for configuring integration, refer to:
Google Cloud Documentation on Providing Credentials
For additional documentation in case of issues with using ADC (Application Default Credentials), refer to:
Google Cloud Documentation on Troubleshooting ADC
Dictionary
Feed
A digital file containing the products you want to advertise online, including various product attributes such as name, size, color, and more.
Product
The target product that can be part of campaigns with product attributes defined in the feed.
Product Attribute
A variable of a specific type that describes the product, such as its name, size, or color.
Data Set
A selected group of products for use in dynamic campaigns such as email, SMS, mobile push, or web push. It can be created based on product attributes defined in a specific channel.
Type
You can specify the type of product attribute. Depending on what you want to describe, you need to choose the appropriate type, as it cannot be changed later. The different types of product attributes are:
- Text - choose to specify a name,
- Link - choose if you want to insert a link,
- Integer - choose if you want to specify an integer value,
- Floating Point Number - choose if you want to specify a decimal number,
- Boolean - choose if you want to specify true/false,
- Single-Select Dictionary - choose if you have a closed set of values, e.g., product category,
- Multi-Select Dictionary - choose if you have a closed set of values but want to assign multiple values,
- Date - choose to specify a date,
- Date and Time - choose to specify a date with a specific time
- ID - product identifier.
File
Contains the data needed to create the channel. It must be in .csv or .xml format.
If using a CSV file, it must be encoded in UTF-8, with columns separated by commas or semicolons. Additionally, the first row must contain the API key for the function included in that column. You will be able to reorder columns during import since they are matched in the first row.
When using a CSV file, a new field will appear where you need to select the CSV separator (e.g., comma separator).
Functionality
Feed creation
After correctly configuring the integration with Vertex AI, on the first step of feed creation—i.e., the screen for choosing how to add and update the feed—a fourth option related to Vertex AI will appear.
After selecting this option, a settings form will be displayed, where you need to fill in the settings fields. For convenience, they are divided as follows:
- Feed Settings: Name and feed key (used for utilizing the feed in delivery),
- Vertex AI Settings: Catalog and location (found in the Google Vertex AI console).
Product Attribute Configuration - A product attribute is a variable that describes the product. The list of product attributes is generated automatically and aligns with the attributes available in the Google Vertex AI panel. The automatic attributes are:
Name | Key | Type |
Name | name | text |
Title | title | text |
Description | description | text |
Categories | categories | Multiple choice dictionary |
Brands | brands | Multiple choice dictionary |
Price | priceInfo.price | Floating point number |
Currency code | priceInfo.currencyCode | Text |
URI | urli | Link |
Image | images.0 | Link |
The name, key, and type are extracted from the configuration and may consist of multiple levels of nesting due to the way product details are presented within Vertex AI.
Each time, it is possible to add a new attribute using the "Add Attribute" button; however, you must ensure that the key provided matches the documentation available for the specific product within the Google Cloud panel.
The list of products to be added to the feed is managed from the Google Cloud panel, where the database is managed.
In iPresso, the product list is not available; there is no option to view products either in the entire feed or in the data set.
Data set creation
After creating the feed, you can create a dataset for it using a form that is divided into the following sections for better clarity:
- Dataset Settings: Name and key (which will enable its use in delivery).
-
Recommendation configuration:
- Serving config - pole wskazujące z jakiego modelu ma korzystać rekomendacja, dostępne z poziomu panelu Google Cloud.
Event - The type of event used by the created dataset for recommendations. Depending on the type selected, different recommendations will be displayed.
There are 3 available types:
- category page view,
- detail page view,
- home page view.
All details regarding the operation and configuration of an individual event are available in the Google Cloud Panel.
Strict Filtering - Selecting this option will prevent recommendations from being sent to a contact if no products are available. If this option is not selected, default products will be displayed to the contact in the absence of matching recommendations.
Comments
0 comments
Article is closed for comments.