After clicking “Watch Now” you will be prompted to login or join.
Deploying a Scalable GPU-as-a-Service Platform and Building a Deep Learning Project in Under 80 Minutes
Andrew Bull, NVIDIA | Adam Tetelman, NVIDIA | Mark Skinner, NVIDIA | Sumit Kumar, NVIDIA
GTC 2020
In less than 80 minutes, we'll demonstrate how you can deploy and manage a scalable GPU cluster through DeepOps, an NVIDIA open-source project. We'll start with deploying several cluster services, including Kubeflow — a project that makes it easy to build and deploy AI workflows on Kubernetes. Then we'll dive into how you can use these cluster services with the NVIDIA TAO Toolkit from the NVIDIA GPU Cloud (NGC) to train a deep neural network model, and go through the process of deploying the trained model on the cluster with TensorRT Inference Server for a high-throughput, low-latency inferencing service.