Scaling Deep Learning Using Polyaxon with Mourad Mourafiq - Episode 158

Summary

With libraries such as Tensorflow, PyTorch, scikit-learn, and MXNet being released it is easier than ever to start a deep learning project. Unfortunately, it is still difficult to manage scaling and reproduction of training for these projects. Mourad Mourafiq built Polyaxon on top of Kubernetes to address this shortcoming. In this episode he shares his reasons for starting the project, how it works, and how you can start using it today.

linode-banner-sponsor-largeDo you want to try out some of the tools and applications that you heard about on Podcast.__init__? Do you have a side project that you want to share with the world? Check out Linode at linode.com/podcastinit or use the code podcastinit2018 and get a $20 credit to try out their fast and reliable Linux virtual servers. They’ve got lightning fast networking and SSD servers with plenty of power and storage to run whatever you want to experiment on.


GoCD is the on-premise open source continuous delivery server created by ThoughtWorks and modeled after the ideas in the Continuous Delivery book by Jez Humble and David Farley.

With GoCD’s comprehensive pipeline modeling, you can model complex workflows for multiple teams with ease. And GoCD’s Value Stream Map lets you track a change from commit to deploy at a glance.

GoCD’s real power is in the visibility it provides over your end-to-end workflow. So you get complete control of and visibility into your deployments, across multiple teams.

Say goodbye to deployment panic and hello to consistent, predictable deliveries.

To learn more about GoCD, visit gocd.org for a free download. Professional Support and enterprise add-ons, including disaster recovery, are available.


When your website experiences an error, Airbrake alerts you in real-time, and gives you all the details you need to fix the bug fast. Some of the most useful tools that Airbrake provides for faster resolution are:

  • Exception aggregation to understand the how many users are being affected, so that you can prioritize the work you are doing to have the biggest impact
  • Contextual information to understand how certain states of the application contribute to the exception being raised, and which environments are affected
  • Deployment tracking so that you can easily see whether a new feature is the source of an error
  • Integration with all of the other tools that you use, such as automatically creating issues in GitHub and linking to the line of code where the error came from

Right now, Podcast.__init__ listeners can try Airbrake free for 30 days, plus get 50% off the first 3 months on the Startup plan. To get started, visit airbrake.com/podcastinit today.


Preface

  • Hello and welcome to Podcast.__init__, the podcast about Python and the people who make it great.
  • When you’re ready to launch your next app you’ll need somewhere to deploy it, so check out Linode. With private networking, shared block storage, node balancers, and a 200Gbit network, all controlled by a brand new API you’ve got everything you need to scale up. Go to podcastinit.com/linode to get a $20 credit and launch a new server in under a minute.
  • Finding a bug in production is never a fun experience, especially when your users find it first. Airbrake error monitoring ensures that you will always be the first to know so you can deploy a fix before anyone is impacted. With open source agents for Python 2 and 3 it’s easy to get started, and the automatic aggregations, contextual information, and deployment tracking ensure that you don’t waste time pinpointing what went wrong. Go to podcastinit.com/airbrake today to sign up and get your first 30 days free, and 50% off 3 months of the Startup plan.
  • To get worry-free releases download GoCD, the open source continous delivery server built by Thoughworks. You can use their pipeline modeling and value stream map to build, control and monitor every step from commit to deployment in one place. And with their new Kubernetes integration it’s even easier to deploy and scale your build agents. Go to podcastinit.com/gocd to learn more about their professional support services and enterprise add-ons.
  • Visit the site to subscribe to the show, sign up for the newsletter, and read the show notes. And if you have any questions, comments, or suggestions I would love to hear them. You can reach me on Twitter at @Podcast__init__ or email [email protected])
  • Your host as usual is Tobias Macey and today I’m interviewing Mourad Mourafiq about Polyaxon, a platform for building, training and monitoring large scale deep learning applications.

Interview

  • Introductions
  • How did you get introduced to Python?
  • Can you give a quick overview of what Polyaxon is and your motivation for creating it?
  • What is a typical workflow for building and testing a deep learning application?
  • How is Polyaxon implemented?
    • How has the internal architecture evolved since you first started working on it?
    • What is unique to deep learning workloads that makes it necessary to have a dedicated tool for deploying them?
    • What does Polyaxon add on top of the existing functionality in Kubernetes?
  • It can be difficult to build a docker container that holds all of the necessary components for a complex application. What are some tips or best practices for creating containers to be used with Polyaxon?
  • What are the relative tradeoffs of the various deep learning frameworks that you support?
  • For someone who is getting started with Polyaxon what does the workflow look like?
    • What is involved in migrating existing projects to run on Polyaxon?
  • What have been the most challenging aspects of building Polyaxon?
  • What are your plans for the future of Polyaxon?

Keep In Touch

Picks

Links

The intro and outro music is from Requiem for a Fish The Freak Fandango Orchestra / CC BY-SA

Start the discussion at https://discourse.pythonpodcast.com