Troubleshooting MegaDetector Pipeline

Common Issues

1. Pipeline Fails to Start

  • Cause: Missing or incorrect datastore configuration.
  • Solution: Verify that the input and output datastores (landing_kutuma_hashed, bronze_megadetector) are correctly set up in AML.

2. Model Not Found

  • Cause: Incorrect model path or missing model in ml_public_models.
  • Solution: Ensure the model (md_v5b.0.0.pt) is correctly uploaded and the path is correct in the pipeline parameters.

3. Resource Allocation Error

  • Cause: Insufficient compute resources (GPU/CPU).
  • Solution: Check the available resources in the Azure portal and adjust the compute cluster if needed.

Debugging Tips

  • Enable verbose logging during pipeline runs to get detailed error messages.
  • Use the AML Studio Logs section to review pipeline execution and diagnose issues.