how to deploy a model

User 2570 | 11/10/2015, 12:20:31 PM

I am newbie to graph create and machine learning, trying to learn step by step, I have successfully created a model using graph create, and i am wondering, how i can deploy it where my friends can be able to see it and use it, on a website for example, how to deploy as a service

Comments

User 2071 | 11/12/2015, 4:13:29 AM

Hi Erigits,

It sounds like Dato Predictive Services could help with your question. A predictive service lets you publish a model as an HTTP endpoint, in the form of a URL; you can use this URL to query the model, just like you would within GraphLab Create, but now it is accessible from anywhere. You (or your friends) don't even need GraphLab Create to talk to a predictive service, anything that can speak HTTP is a potential consumer (we have libraries for JavaScript and Java).

Our user guide is a good starting point for becoming familiar with Dato Predictive Services. Here is an end-to-end walk-through: https://dato.com/learn/userguide/deployment/pred-getting-started.html

The requirement to run a predictive service is a machine to host it. The easiest option is to use Amazon EC2. You just need an AWS account to be able to create a predictive service (using the GraphLab Create API), then add your model there. There are a few options in AWS that you don't have to pay for, at least initially. Another option is to install a predictive service on a local machine; that is a feature we just released, and it is a little more involved for now.

Give it a try and let us know how we can help further!

Roman


User 2570 | 11/14/2015, 4:06:43 PM

very helpful thank you