This document is relevant for: Inf1, Inf2, Trn1, Trn2, Trn3

Inf1 installation (legacy)#

Warning

Legacy hardware: Inf1 instances use NeuronCore v1 architecture.

For new projects, use Inf2, Trn1, Trn2, or Trn3 instances with NeuronCore v2 for:

  • 3x better price-performance than Inf1

  • Broader framework support (PyTorch 2.x, JAX)

  • Active development and feature updates

  • Latest Neuron SDK features

See Install AWS Neuron SDK for current instance options.

When to use Inf1

Use Inf1 only if you:

  • Maintain existing Inf1 deployments

  • Have compiled models for NeuronCore v1

  • Require specific Inf1 cost optimization for inference workloads

Migration to Inf2#

Consider migrating to Inf2 for better performance and support:

  • Inf2 offers 3x better price-performance

  • Broader framework support including PyTorch 2.x and JAX

  • Active development with monthly SDK releases

  • See Install AWS Neuron SDK for current installation options

Choose your framework#

Note

JAX is not supported on Inf1 instances. Use Inf2, Trn1, Trn2, or Trn3 for JAX workloads.

PyTorch (Inf1)

PyTorch 1.x with torch-neuron

Inference on Inf1 instances using NeuronCore v1

Legacy

TensorFlow (Inf1)

TensorFlow 2.x with tensorflow-neuron (archived)

Archived

Additional resources#

This document is relevant for: Inf1, Inf2, Trn1, Trn2, Trn3