pyspark自定义的pipeline无法保存
2020-12-21 本文已影响0人
一只当归
自定义pipeline方法feature_engineering进行保存时出现错误
ValueError: ('Pipeline write will fail on this pipeline because stage %s of type %s is not MLWritable', 'feature_engineering_94caf721906c', <class 'main.feature_engineering'>)
解决方法：
类继承DefaultParamsReadable, DefaultParamsWritable类，会自动载入保存以及读取方法
from pyspark.ml.util import DefaultParamsReadable, DefaultParamsWritable
class feature_engineering(Transformer,DefaultParamsReadable, DefaultParamsWritable):
