How to deserialize Pipeline model in spark.ml?

孤街浪徒 提交于 2019-12-05 17:53:39

Since 1.6.0 you can use (experimental) Pipeline.read.load(path) to load a Pipeline object from path.

Experimental it is indeed. Since Pipeline(s) can contain all sorts of things, each one needs to be readable/writable for the Pipeline save/load to work. See for example RandomForests is not there yet and won't be until Spark v2.0.0 https://issues.apache.org/jira/browse/SPARK-13784

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!