You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@dolphinscheduler.apache.org by 田志军 <ti...@xdnovo.com> on 2022/05/17 02:31:02 UTC
退订退订退订
发自我的iPhone
------------------ Original ------------------
From: Jiajie Zhong <zhongjiajie955@gmail.com>
Date: Tue,May 17,2022 10:19 AM
To: dev <dev@dolphinscheduler.apache.org>
Subject: Re: Support AI task and the open source project about MLops
The related discussion it is from
https://github.com/apache/dolphinscheduler/pull/10036#discussion_r873164232,
and jieguang want to have some build-in mlfow example and currently
use his repository to hold all this code, I think the best way is we
contribute some example to mlflow repository but it take time. So it
is ok to create a new repository to hold this code as a subproject for
dolphinscheduler, and I will create it.
On Mon, May 16, 2022 at 4:27 PM 周哈哈 <jieguang_zhou@163.com> wrote:
>
> Hi, I am supporting the MLflow task plugin to run the MLflow project. I have submitted the relevant PR (https://github.com/apache/dolphinscheduler/pull/10036).
>
> But I used a personal repository to provide algorithmic power out of the box, and hopefully a public repository.
>
> In order for engineers of different backgrounds to better use this task plugin, I suggest that:
>
> Create a repository for out-of-the-box machine learning ability.
> Support for setting up directories to use users' own projects.
>
> So, can we create a repository for the first point?
>
>
>
>
> > 2022年5月6日 上午11:56,Jiajie Zhong <zhongjiajie955@gmail.com> 写道:
> >
> > Sound good! and thanks Jieguang brings this up. DolphinScheduler
> > focuses on task scheduling and workflow orchestration, I think machine
> > learning is also another kind of orchestration, and most of the
> > machine learning source data or training samples are from data
> > warehouses or data lakes, which we already supported in the current
> > version. If we DolphinScheduler could support machine learning tasks
> > then users could finish their jobs in one single tool instead of
> > separately.
> >
> > On Mon, Apr 25, 2022 at 2:39 PM 周哈哈 <jieguang_zhou@163.com> wrote:
> >>
> >> Hi, community
> >>
> >> I have seen a Machine Learning Platform post on Medium. The post talk about Lizhi Machine Learning Platform&Apache DolphinScheduler.
> >> https://medium.com/@DolphinScheduler/a-formidable-combination-of-lizhi-machine-learning-platform-dolphinscheduler-creates-new-paradigm-e445938f1af <https://medium.com/@DolphinScheduler/a-formidable-combination-of-lizhi-machine-learning-platform-dolphinscheduler-creates-new-paradigm-e445938f1af>
> >>
> >>
> >> I want to use the dolphinscheduler to orchestrate some popular open-source projects about MLops, such as MLflow, Kubeflow, BentoML, Seldon Core, ClearML, TonY, etc. I think that can help users to build their custom MLops systems easier.
> >>
> >> For example, users can choose what deployment tools they like, or what training tools they like.
> >>
> >> Is somebody interested in the AI task about DS?
> >>
> >> I created an issue about the above: https://github.com/apache/dolphinscheduler/issues/9725 <https://github.com/apache/dolphinscheduler/issues/9725>
> >>
> >>
> >> Anyone who is interested is welcome to join us.
> >>
> >>
> >> Best Wish
> >> — Jieguang
> >
> >
> >
> > --
> > Best Wish
> > — Jiajie
>
--
Best Wish
— Jiajie