Unable to deploy best HPO model trained using open-source XGBoost framework #2876
Description
Describe the bug
When I try to deploy the result of a HyperParameterTuner job using the OSS XGBoost framework and training script, I get an error about image tags. Upon further inspection (see attached notebook export), it looks like we need to update the xgboost config file so that the newer library versions specify the python version and processor. This may have been missed in this change.
To reproduce
See attached pdf file
Expected behavior
After defining an XGBoost estimator using the framework_version (1.3-1), py_version (py3), and instance_type but NOT an image URI I expect to use the estimator to create an HPO job and deploy the best result.
Screenshots or logs
See attached pdf file
System information
A description of your system. Please provide:
- SageMaker Python SDK version: 2.7.0
- Framework name (eg. PyTorch) or algorithm (eg. KMeans): XGBoost
- Framework version: 1.3-1
- Python version: 3.7.10
- CPU or GPU: CPU
- Custom Docker image (Y/N): No
Additional context
Add any other context about the problem here.
xgboost-example.pdf