Model Deployments

Overview

The Deployments section enables users to deploy dataflows and utilize them for predictions. This documentation covers deployment preparation, management, and prediction processes.

Video Tutorial

For a visual guide on the Deployments, please watch our step-by-step tutorial:

Table of Contents

1. Preparing Dataflows for Deployment
2. Deployment Management
3. Utilizing Deployments
4. Prediction Types
5. Results Management

Preparing Dataflows for Deployment

Deployment Process

1. Navigate to Dataflows section
2. Access 'Runs' table
3. Locate desired dataflow version
4. Click 'Deploy' button
5. Once deployed, use 'Predict' button for direct deployment access

Deployment Management

Deployment Table Information

Available Actions

• View source dataflow
• Utilize for prediction
• Delete deployment

Utilizing Deployments

Prediction Options

1. Batch Prediction

   - Upload prediction data files
   - Configure batch settings
   - Run prediction
   - Access results
   

2. Streaming Prediction

   - Access row prediction interface
   - Input parameter values
   - Get immediate results
   - API endpoint available
   

API Integration

{
  "endpoint": "your-deployment-endpoint",
  "method": "POST",
  "content-type": "application/json",
  "body": {
    "param1": "value1",
    "param2": "value2"
  }
}

Prediction Types

Batch Prediction

1. File Upload:

   • Select prediction data files
   • Verify data format
   • Configure batch settings

2. Execution:

   • Click 'Run Prediction'
   • Access results

Streaming Prediction

1. Interactive Mode:

   • Access row prediction interface
   • Input required parameters
   • Submit for processing
   • View immediate results

2. API Mode:

   • Use REST API endpoint
   • Send JSON-formatted requests
   • Receive real-time responses
   • Integration ready

Results Management

Predictions Table

• View processed results
• Download predictions
• Save results
• Access historical predictions

Result Operations

1. View Results:

   - Tabular format
   - Download as CSV
   - Download as Apache Parquet
   

Best Practices

1. Deployment Preparation:

   • Test dataflow thoroughly
   • Verify input parameters
   • Document deployment settings

2. Prediction Management:

   • Monitor performance
   • Track usage metrics
   • Regular validation
   • Error handling

Additional Resources

• Product Demo Videos