Skip to content

Latest commit

 

History

History
27 lines (18 loc) · 1.3 KB

TROUBLESHOOTING.md

File metadata and controls

27 lines (18 loc) · 1.3 KB
Raw

Troubleshooting

In this page, we will address some common issues that the developers often face while working with AI Platform samples and provide solutions.

Potential Issues

Issue: gcloud commands are not working properly

  • Run gcloud init in your terminal and follow the instructions.
  • Run gcloud compute instances list to make sure the API for Compute is enabled.
  • Run gcloud ai-platform models list to make sure the API for AI Platform is enabled.
  • Run gsutil ls gs:// to make sure you have access to GCS.
  • Run gcloud info to get more information about your configuration.

Issue: Training job is failing

  • Make sure you have followed the setup steps.
    • Verify that the APIs are enabled.
    • Verify that the environment variables are set correctly.
  • Confirm that GOOGLE_APPLICATION_CREDENTIALS is pointing to the correct service account key file for your GCP project. Also, make sure the ervice account file is for your current GCP project and that it has the right roles, as specified in the setup page.
  • Run gcloud config list and verify the account and the project are set correctly.

Issue: Got a Bad magic number in .pyc file during local prediction

This is a known issue which occurs when you are using Python 3. It will be fixed in upcoming releases.