[SPARKML] sparkml_serving_emr_mleap_abalone.ipynb issues

[paul-cb]

https://github.com/awslabs/amazon-sagemaker-examples/blob/master/sagemaker-python-sdk/sparkml_serving_emr_mleap_abalone/sparkml_serving_emr_mleap_abalone.ipynb

I tried running this notebook depending on the kernel selected, i.e., pyspark, or conda python3 I get an error sending http request and maximum retry encountered or an import library error.

[smuralee]

I have noticed below issues with this notebook

• The Spark connection fails with EMR 5.27 and above when using the the 0.96 assembly jar
• The latest versions of mleap are not possible to use with this example
• When installing individual jars for mleap instead of using 0.96 fat jar, the dependencies are not clear and serialization fails
• The Sagemaker SDK v2 breaking changes for RealTimePredictor, Serializers and Spark deprecation for OneHotEncoder are causing code issues during execution

PR has been raised with the above fixes