Support for distributed/multi node grid search/training? python/sci-kit learn based

UserBird
UserBird Dataiker, Alpha Tester Posts: 535 Dataiker
in Setup & Configuration
Hi, what is the support in Dataiku for distributing grid searches/training? Not speaking of spark or hadoop existing dataiku integration, but mainly sending the regular python based jobs to other nodes, that could be running dask or similar frameworks, get the results back to dataiku transparently? possible? if so, what is the setup/licensing needed? Could it be done installing additional dataiku servers on nodes?

thx!
Tagged:
· Share on FacebookShare on Twitter

Best Answer

  • Alex_Combessie
    Alex_Combessie Alpha Tester, Dataiker Alumni Posts: 539 ✭✭✭✭✭✭✭✭✭
    Answer ✓
    Hi Rui,

    Note that we already distribute training on CPUs. Distributed training is something we are working on at the moment. It will be part of one of our next software releases. We will keep you posted on the details when it is ready.

    In the meantime, you can develop it with custom code in Python/Scala/Shell/R scripts within Dataiku DSS.

    Cheers,

    Alexandre
    · Share on FacebookShare on Twitter

Answers

  • UserBird
    UserBird Dataiker, Alpha Tester Posts: 535 Dataiker
    Thx Alexandre, yes, perfectly clear on the cpus, very easy to notice.
    Awesome news, tanks for the heads-up, think this is really mandatory these days, otherwise very limiting comparing to dask & similar, which are opensource/free.
    regards
    Rui
    · Share on FacebookShare on Twitter
Bookmark this topic

Categories

Setup Info
    Tags
      Help me…