Deep Learning

Understanding and using deep learning networks

Deep Learning with NGC

Posted by Johanna Pingel,

Did you know: you can run MATLAB with GPUs in the cloud? MATLAB has a deep learning container on NGC! Let’s define what this means:
  • What is a container?
    • A self-contained execution environment. Containers “run everywhere” on the cloud, on-premise GPU clusters, or desktop workstations. This simply means: a MATLAB container is a way to bring MATLAB with you to the cloud or cluster resources.
  • What is NGC?
    • This stands for NVIDIA GPU Cloud, which is a repository of containers. (See list of all containers) If you go to the site, you’ll see MATLAB is one of the packaged, tested and optimized containers. NGC containers are tested to work on DGX stations and Cloud vendors like AWS and Azure. More info about MATLAB with GPUs is at the bottom of this page.
  • What is deep learning?

Screen shot of NVIDIA's containers on their website

Put this all together: MATLAB has a deep learning container on NGC. We have containers available for release R2018a through R2019a, and new containers will be provided for every future major release.
How can you try this?
  • If you have on-premise GPU clusters and you’re interested to try the MATLAB container, you should contact us (deep-learning@mathworks.com), and we’ll help you get set up.
  • If you’re ready to scale deep learning on the cloud: this video will help you get started setting up the MATLAB Deep Learning Container on NGC. This will discuss every step in detail, including how to download a free cloud trial. You'll be ready to get started in 15 minutes.

At GTC San Jose, we ran deep learning training on DGX with 4 GPUs using the MATLAB container

Here are the complete resource links in case you missed them: