cnvrg makes it easy to publish your predictive models as web services on a https secured cluster.
You can publish a dew different models under the same cluster and can decide on the traffic distribution between them (equally distributed or custom per demand)

Publish new models:

In the Publish Tab, click on "Publish New Model":

  • Choose Cluster Endpoint
  • Set endpoint title: this will be part of the url of the endpoint

Under Models:

  • Choose how many workers to have in the cluster
  • Choose file and a function for each model
  • If the project includes git integration- you can choose git branch and git commits
  • Choose cnvrg commit (optional if git was enabled)
  • Python 2 or Python 3

When having more than one model published in the same cluster, the user can customize the traffic distribution either by using the slider or setting percentage for each model:

When hitting on "Deploy Endpoint" button, the cluster will be created (it could take a few minutes) and you will be transferred to the endpoint page.
In the endpoint page you can see all the logs for the endpoint, view requests distribution per each model, Grafana dashboard on the cluster, Activity History and Configuration.
Under configuration you can roll updates to the cluster such as -
changing the traffic distribution, adding/deleting models, changing models details.
On the lower section of the page, there are examples snippets on how to send requests to the endpoint.

Did this answer your question?