posit::conf(2023) Workshop: Deploy and Maintain Models with vetiver
Register now: http://pos.it/conf Instructor: Julia Silge Workshop Duration: 1-Day Workshop This workshop is for you if you: • have intermediate R or Python knowledge (this will be a “choose your own adventure” workshop where you can work through the exercises in either R or Python) • can read data from CSV and other flat files, transform and reshape data, and make a wide variety of graphs • can fit a model to data with your modeling framework of choice We expect participants to have exposure to basic modeling and machine learning practice, but NOT expert familiarity with advanced ML or MLOps topics. Many data scientists understand what goes into training a machine learning or statistical model, but creating a strategy to deploy and maintain that model can be daunting. In this workshop, learn what MLOps (machine learning operations) is, what principles can be used to create a practical MLOps strategy, and what kinds of tasks and components are involved. We’ll use vetiver, a framework for MLOps tasks in Python and R, to version, deploy, and monitor the models you have trained and want to deploy and maintain in production reliably and efficiently
image: thumbnail.jpg
Transcript#
This transcript was generated automatically and may contain errors.
Hi, my name is Julia Silge and I'm excited to talk to you about a workshop that I'm teaching at this year's PositConf in Chicago about how to deploy and maintain models with vetiver.
It turns out that lots of data scientists and other kinds of data practitioners know how to approach training a model, a machine learning model or a statistical model. There's lots of books about this. This is something that they teach in classes at school. However, when it comes time to deploy that model and maintain that model in the long run as it interacts with the real world, there's a lot more uncertainty and it can be pretty daunting to figure out where to start, what are the tasks that are involved.
The term that is typically used for those kinds of tasks is MLOps, which stands for Machine Learning Operations. I'll be honest, there's a lot of hype around that term, around tooling in the MLOps space. What we're going to do in this workshop is we are going to step back from the hype and focus on the substantive, concrete tasks that you need to know how to do in order to make sure you are deploying and maintaining your model in a reliable way. We're going to focus on how to version, how to deploy and how to monitor models.
About the vetiver framework and workshop format
We're going to use the vetiver framework for this. So vetiver is a framework for MLOps tasks in R and Python. And that means this workshop is going to be a little unique in that you can choose your own adventure. You can come to this workshop with a model or a type of model that you want to learn how to deploy and maintain. You can choose to work in R or Python or a little bit of both, I guess, if that's what you want to do.
If I were to sum up one reason why I think you should think about taking this workshop, I would say if you are someone who develops models as part of your role, you are someone who trains a model, who tunes a model's hyperparameters, who evaluates that model to understand how well is it performing, how reliable and appropriate it is to this use case. If that's you, then you can deploy that model. And in fact, you probably should be the one to deploy and maintain that model in the long run because of the depth of knowledge that you have about how that model works, the constraints around it.
And in fact, you probably should be the one to deploy and maintain that model in the long run because of the depth of knowledge that you have about how that model works, the constraints around it.
So you can learn how to get your model off your laptop and out into the wide world out there integrated into your organization's infrastructure. I look forward to seeing you at the workshop.