Amazon EC2 Trn1 instances, powered by AWS Trainium chips, and Amazon EC2 Inf1 instances, powered by AWS Inferentia chips, are built from the ground up to provide high performance and the low cost ML training and inference in the cloud. In this workshop, we will walk you through how to train a HuggingFace BERT model onto Trn1 and then deploy it on Inf1. We will also look at how you can benchmark you model with NeuronPerf to find the optimum configuration to obtain low latency and high throughput.

Presented by Gabriel Brackman