FAQ

Introduction

• General Neuron FAQs
  • Getting started with Neuron FAQs
    • Q: How can I get started?
    • Q: How do I select which Inf1 instance size to use?
  • General
    • Q: What ML models types and operators are supported by AWS Neuron?
    • Q: Why is a compiler needed, and how do I use it?
    • Q: I am using a ML framework today – what will change for me to use this?
    • Q: What is a NeuronCore Pipeline ? and How do I take advantage of it?
    • Q: NeuronCores, NeuronCore Groups and NeuronCore Pipelines: What do they do?
    • Q: Can I use TensorFlow networks from tfhub.dev as-is ? if not, what should I do?
  • Troubleshooting FAQs
    • Q: Performance is not what I expect it to be, what’s the next step?
    • Q: Do I need to worry about size of model and size of inferentia memory? what problems can I expect to have?
    • Q: How can I debug / profile my inference request?
  • Contributing Guidelines FAQs
    • Q: How to reporting Bugs/Feature Requests
    • Q: Contributing via Pull Requests
    • Q: How to find contributions to work on
    • Q: What is the code of conduct
    • Q: How to notify for a security issue
    • Q: What is the licensing
• Neuron Compiler FAQs
  • Q: Where can I compile to Neuron?
  • Q: Which operators does Neuron support?
  • Q: Any operators that Neuron doesn’t support?
  • Q: Will I need to recompile again if I updated runtime/driver version?
  • Q: How long does it take to compile?
• Neuron runtime FAQs
  • Q: How does Neuron connect to all the Inferentia chips in an Inf1 instance?
  • Q: Where can I get logging and other telemetry information?
  • Q: What about RedHat or other versions of Linux?
  • Q: What about Windows?
  • Q: How can I use Neuron in a container based environment? Does Neuron work with ECS and EKS?
  • Q: How can I take advantage of multiple NeuronCores to run multipleinferences in parallel?