apache spark - Custom transformer in Pyspark ML -
i'm trying write custom estimator fit pipeline using pyspark 1.6.0, try solution in how roll custom estimator in pyspark mllib
it works on spark 2.0, when remove typeconverter parameter(which exists in spark 2.0) test spark 1.6.0, shows error:
valueerror: param param(parent='undefined', name='centered_threshold', doc='centered_threshold') not belong normaldeviation_4dc98930f6cd0fd37ce6.
i'm not allowed use spark 2.0 in production. how can change code make work on spark 1.6.0?
Comments
Post a Comment