CI/CD implementation in Dataiku for Machine Learning

sj0071992

Hi All,

I just want to know if there is any reference video that tells about how to achieve MLOps in Dataiku and how we can create CI/CD in Dataiku?

Thanks

fsergot

Hello,

Both topics are closely related but somehow different.

CI/CD is a way to automate the move to production and we have some readings for you on our blog (sorry, no video ) : https://blog.dataiku.com/continuous-integration-and-continuous-deployment-ci/cd-in-dataiku-dss. We also have some knowledge base step-by-step examples that you may want to check:

• Building a Jenkins pipeline for Dataiku DSS
• Building a Jenkins pipeline for API services in Dataiku DSS

MLOps is a wider topic and usually specific to each organization. It usually comes with a brainstorming on automation (and CI/CD) but this is not mandatory and is anyway just one piece of the puzzle.
From a practical standpoint, you have a set of capabilities for DSS that you should look at in more details https://www.dataiku.com/product/key-capabilities/mlops/
But from a higher standpoint, MLOps goes beyond tool capabilities and I can recommend reading our book on MLOps, you can get a free copy here.

Lastly, both subjects will be session topics on our upcoming Product Days online event so you may want to register and attend those, especially on the 3rd day (and of course, any other if you like https://events.dataiku.com/product-days/home

Hope this helps!
Gave a nice day.

shajay12

Hi Everyone,

In this Article Building a Jenkins pipeline for API services in Dataiku DSS

what is the value of these variables 'apinodes' and 'url' .....code pasted below?

def build_apinode_client(params
    client_design = dataikuapi.DSSClient(params["host"], params["api"])
    api_deployer = client_design.get_apideployer()
    api_url = api_deployer.get_infra(params["api_dev_infra_id"]).get_settings().get_raw()['apiNodes'][0]['url']
    return dataikuapi.APINodeClient(api_url, params["api_service_id"])

CoreyS

Hi @shajay12
please see this response here on this thread:

I hope this helps!