Skip to content

AZURE_ML_CONFIG.md

Latest commit

 

History

History
276 lines (193 loc) · 6.33 KB

AZURE_ML_CONFIG.md

File metadata and controls

276 lines (193 loc) · 6.33 KB

Azure ML Configuration Guide

Overview

This guide explains how to securely configure your Azure ML workspace connection using config.json and test that it works properly.

📁 Where to Place config.json

Option 1: Project Root (Recommended for Development)

/Users/mcdaniel/Desktop/gh/fswl/azureml-example/config.json

Pros:

  • Automatically detected by Workspace.from_config()
  • Project-specific configuration
  • Already in .gitignore (secure)

Cons:

  • Need separate config for each project

Option 2: User Home Directory (Recommended for Global Use)

~/.azure_ai/config.json

Pros:

  • Shared across all Azure ML projects
  • Standard Azure ML location
  • Outside project directory

Cons:

  • Global configuration (may not work if you have multiple workspaces)

🔒 Security Considerations

✅ Already Secured in This Project:

  • config.json is in .gitignore ✅
  • Won't be committed to version control ✅

🔐 Additional Security Steps:

1. File Permissions (macOS/Linux)

chmod 600 config.json  # Read/write for owner only

2. Environment-Specific Configurations

For production/CI environments, use environment variables instead:

from azure_ai.core import Workspace

# Using environment variables
ws = Workspace(
    subscription_id=os.environ['AZUREML_SUBSCRIPTION_ID'],
    resource_group=os.environ['AZUREML_RESOURCE_GROUP'],
    workspace_name=os.environ['AZUREML_WORKSPACE_NAME']
)

3. Service Principal Authentication (Production)

For automated deployments, use service principal authentication:

from azure_ai.core.authentication import ServicePrincipalAuthentication

svc_pr = ServicePrincipalAuthentication(
    tenant_id=os.environ['AZURE_TENANT_ID'],
    service_principal_id=os.environ['AZURE_CLIENT_ID'],
    service_principal_password=os.environ['AZURE_CLIENT_SECRET']
)

ws = Workspace(
    subscription_id=os.environ['AZUREML_SUBSCRIPTION_ID'],
    resource_group=os.environ['AZUREML_RESOURCE_GROUP'],
    workspace_name=os.environ['AZUREML_WORKSPACE_NAME'],
    auth=svc_pr
)

📥 How to Download config.json

  1. Go to Azure ML Studio: https://ml.azure.com
  2. Select your workspace from the workspace dropdown
  3. Click the download icon (⬇️) next to your workspace name in the top navigation
  4. Save the file as config.json

The downloaded file should look like this:

{
  "subscription_id": "your-subscription-id",
  "resource_group": "your-resource-group-name",
  "workspace_name": "your-workspace-name"
}

🧪 Testing Your Configuration

Automated Setup and Test

# Run the setup script
python setup_config.py

# Run the test
python azure_ai/tests/test_azureml.py

Manual Testing

# From project root
python azure_ai/tests/test_azureml.py

Test Components

The test script will check:

  1. ✅ Azure ML SDK Installation

    • Core modules import correctly
    • Training estimators available
    • Pipeline components accessible

  2. ✅ Configuration File Detection

    • Searches multiple locations for config.json
    • Validates file structure
    • Reports found location

  3. ✅ Workspace Connection

    • Connects to your Azure ML workspace
    • Displays workspace information
    • Verifies authentication

Expected Output

Successful Test:

Testing Azure ML Installation
========================================
✅ Azure ML Core version: 1.48.0
✅ Azure ML Core classes imported successfully
✅ Azure ML Training estimators imported successfully
✅ Azure ML Pipeline components imported successfully
⚠️  AutoML components skipped (has dependency conflicts)
✅ Found config.json at: ./config.json
🔄 Attempting to connect to Azure ML workspace...
✅ Successfully connected to workspace: my-workspace
   Subscription: 12345678-1234-1234-1234-123456789012
   Resource Group: my-resource-group
   Location: eastus

========================================
✅ Azure ML installation and configuration are working!

Next steps:
1. Create your first experiment
2. Upload data to your workspace
3. Train your first model

🚨 Troubleshooting

Common Issues:

1. FileNotFoundError

❌ config.json not found

Solution: Download config.json from Azure ML Studio and place it in project root

2. Authentication Error

❌ Workspace connection error: AuthenticationException

Solutions:

  • Check if you're logged into Azure CLI: az login
  • Verify config.json has correct subscription ID
  • Ensure you have access to the workspace

3. Permission Error

❌ Workspace connection error: Forbidden

Solutions:

  • Check your Azure ML workspace permissions
  • Ensure you're using the correct subscription
  • Contact your Azure administrator

4. Network Error

❌ Workspace connection error: ConnectionError

Solutions:

  • Check internet connectivity
  • Verify corporate firewall settings
  • Try using VPN if on restricted network

🌍 Different Environments

Development

./config.json  # Local development workspace

Staging

# Use environment variables
ws = Workspace.from_config(path="./config-staging.json")

Production

# Use service principal authentication
ws = Workspace(
    subscription_id=os.environ['AZUREML_SUBSCRIPTION_ID'],
    resource_group=os.environ['AZUREML_RESOURCE_GROUP'],
    workspace_name=os.environ['AZUREML_WORKSPACE_NAME'],
    auth=svc_pr
)

📝 Best Practices

  1. Never commit config.json to version control
  2. Use different workspaces for dev/staging/prod
  3. Use service principal auth for automated deployments
  4. Set restrictive file permissions on config files
  5. Regularly rotate service principal secrets
  6. Monitor workspace access and usage
  7. Use Azure Key Vault for production secrets

🔗 Next Steps

After successful configuration:

  1. Create your first experiment
  2. Upload training data
  3. Set up compute targets
  4. Train your first model
  5. Deploy models to endpoints

📚 Additional Resources