Deploying a Dummy TensorFlow Model to Reproduce the Recursion Error

Follow these steps to set up and reproduce the maximum recursion depth exceeded error when trying to read s3 objects via an inference script:

1. Create an S3 Bucket:

   • Create a bucket, for example, named bucket11.

2. Upload Sample Data:

   • Upload the sample JSON object data.json to the created bucket.

3. Update the Bucket Name in the Script:

   • Modify the code1/inference.py script to use the name of the bucket you created. Change the bucket name in the get_json_from_s3 function:

     ...
     json_object = s3_client.get_object(Bucket="bucket11", Key=key)

4. Start a SageMaker Notebook Instance:

   • Launch an ml.t3.medium SageMaker notebook instance.

5. Set Up JupyterLab:

   • Launch JupyterLab from the SageMaker instance.
   • Upload the notebook test.ipynb to the JupyterLab directory.
   • Create a folder named code1 in the directory and upload all files from the code1 folder of this repository into it.

6. Clone the Repository (Alternative Setup to 5):

   • Alternatively, you can clone the repository directly into JupyterLab:
     • Open the Git tab, select Clone a Repository from the dropdown, enter the URI: https://github.com/MustaphaU/rerror.git, and hit Clone.

7. Run the Notebook:

   • Execute all the code cells in the test.ipynb notebook.

8. Identify the Error:

   • The last cell should generate the following error:

     ModelError: An error occurred (ModelError) when calling the InvokeEndpoint operation: Received server error (500) from primary with message "{"error": "cannot unpack non-iterable NoneType object"}".....................................
     

     This error is due to the get_json_from_s3 function returning None because of the recursion error during the setup of the S3 client. See the linked CloudWatch logs for more details on the recursion error.

9. Review CloudWatch Logs:

   • Click the link provided in the error message to view the CloudWatch logs. In the logs, you will find:

     maximum recursion depth exceeded while calling a Python object