Using Kubernetes for Machine Learning Frameworks
Discover how Kubernetes can revolutionize the deployment of machine learning frameworks in this 52-minute conference talk. Explore the key features of Kubernetes that make it ideal for running computationally intensive and data-heavy machine learning models, including isolation, auto-scaling, load balancing, flexibility, and GPU support. Learn how the declarative syntax of Kubernetes deployment descriptors simplifies the process for non-operational engineers to train models. Gain insights into setting up popular open-source frameworks like TensorFlow, Apache MXNet, and PyTorch on a Kubernetes cluster. Walk through the training, massaging, and inference phases of establishing a machine learning framework on Kubernetes. Leave with access to a GitHub repository containing fully functional samples, empowering you to implement these techniques in your own projects.