AWS Neuron SDK Foundation gives working ML/AI technologists a detailed look at the AWS ML/AI hardware and software stack. Day one opens with a platform overview, followed by a dedicated module on the Neuron SDK itself. A session on wrangling huge datasets follows, so attendees can feed large training jobs efficiently. The day closes with a Trainium overview, introducing the accelerator attendees will use for the rest of the course. Attendees gain hands-on experience working with the SDK on live Trainium and Inferentia instances.
Day two turns to running training at scale, starting with managing training clusters using EKS and containers. This lets jobs be scheduled across multiple nodes. A distributed training module follows, applying NeuronX Distributed patterns across Trainium instances. A training stability module then covers practical techniques such as scaled initialization, gradient clipping, and cache management for keeping long training runs healthy. The course closes with deploying trained models on Inferentia for low-latency inference, examining typical deployment scenarios including container technology and EKS Kubernetes clusters. Labs are scoped as guided exercises to manage scale and costs in a classroom setting. Each attendee still gets the time and tools to test a wide range of Neuron features with PyTorch and TensorFlow. By the end, attendees are prepared to adapt existing models to run effectively on this hardware.
Who Should Attend
AI Developers, Data Scientists, Analysts, Data Engineers, IT and QA Staff and Technical Managers
What Attendees Will Learn
Upon completing AWS Neuron SDK Foundation, participants will be able to:
- Describe the AWS machine learning hardware and software solution set
- Understand the fundamental concepts of the AWS Neuron NeuronX Distributed SDKs
- Work with AWS Trainium instances for deep learning (DL) acceleration and AWS Inferentia for low-latency inference
- Use PyTorch and TensorFlow with AWS Neuron, Trainium and Inferentia
Prerequisites
Participants should have basic programming skills and a foundational understanding of machine and deep learning.