Cannot compile xgboost-neo model in sagemaker 2.11.0 (neo-xgboost.json file not found)

[splat-ef]

I'm following the tutorial on churn from this link: https://github.com/awslabs/amazon-sagemaker-examples/blob/master/introduction_to_applying_machine_learning/xgboost_customer_churn/xgboost_customer_churn.ipynb. I installed sagemaker 2.11 and tweaked the code to handle the breaking changes from v1.

I was able to train the xgboost model but struggle to compile the neo version. I tried to run this code:

output_path = '/'.join(xgb.output_path.split('/')[:-1])
compiled_model = xgb.compile_model(target_instance_family='ml_m4', 
                                   input_shape={'data': [1, 69]},
                                   role=role,
                                   framework='xgboost',
                                   framework_version='latest',
                                   output_path=output_path)

I get this error:

FileNotFoundError: [Errno 2] No such file or directory: '/opt/conda/lib/python3.7/site-packages/sagemaker/image_uri_config/neo-xgboost.json'

From https://github.com/aws/sagemaker-python-sdk/tree/master/src/sagemaker/image_uri_config, I found xgboost-neo.json instead of neo-xgboost.json. I wonder if this is just a case of a name change that would need to be updated in the code? Or should I need to do something else?

[metrizable]

Hello @splat-ef

Thanks again for contacting us. The fix was merged. Let us know if there's anything else we can help with.

Best regards

[Ayushi96]

I'm following the tutorial on churn from this link: https://github.com/awslabs/amazon-sagemaker-examples/blob/master/introduction_to_applying_machine_learning/xgboost_customer_churn/xgboost_customer_churn.ipynb. I installed sagemaker 2.11 and tweaked the code to handle the breaking changes from v1.

I was able to train the xgboost model but struggle to compile the neo version. I tried to run this code:

output_path = '/'.join(xgb.output_path.split('/')[:-1])
compiled_model = xgb.compile_model(target_instance_family='ml_m4', 
                                   input_shape={'data': [1, 69]},
                                   role=role,
                                   framework='xgboost',
                                   framework_version='latest',
                                   output_path=output_path)

I get this error:

FileNotFoundError: [Errno 2] No such file or directory: '/opt/conda/lib/python3.7/site-packages/sagemaker/image_uri_config/neo-xgboost.json'

From https://github.com/aws/sagemaker-python-sdk/tree/master/src/sagemaker/image_uri_config, I found xgboost-neo.json instead of neo-xgboost.json. I wonder if this is just a case of a name change that would need to be updated in the code? Or should I need to do something else?

I am new to Sagemaker service, could you please tell me how can we navigate to s3 and see the name of the json file ??

[johnellliott]

For others running into trouble accessing built-in images URIs and associated json files, try running 'conda install -c conda-forge sagemaker-python-sdk' on the terminal to install the local file system dependencies. This resolved my FileNotFoundError accessing xgboost.json and debugger.json.

[martinRenou]

Closing as resolved. Feel free to reopen/continue the discussion if needed.