I think this will be a useful addition. Regarding the API and specific design decisions: I think this looks ok. I didn't dig very deep and would be fine to just go with the author's proposal. The main motivation for having a separate flink-ml repository was to develop more easily, make changes and iterate faster without having to weight every design as carefully as we need to do it in core Flink.
So +1 from my side On Fri, Sep 10, 2021 at 4:33 AM Dong Lin <lindon...@gmail.com> wrote: > Hi all, > > We would like to start the vote for FLIP-173: Support DAG of > algorithms [1]. This FLIP was discussed in this thread [2]. > > The proposal extended the Flink ML API to support DAG of algorithms where > each algorithm could have multiple inputs and multiple outputs. It also > extended Flink ML API to support online learning scenarios where a > long-running Model instance needs to be continuously updated by the latest > model data generated by another long-running Estimator instance. > > The vote will be open for at least 72 hours, following the consensus voting > process. > > Thanks! > Dong Lin and Zhipeng Zhang > > [1] > https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=184615783 > [2] > > https://lists.apache.org/thread.html/r6729f351fb1bc13a93754c199d5fee1051cc8146e22374737c578779%40%3Cdev.flink.apache.org%3E >