This document is relevant for: Inf1, Inf2, Trn1, Trn1n

Neuron Containers Release Notes#

Neuron 2.5.0#

Date: 11/07/2022

  • Neuron now supports trn1-based training in Sagemaker and Deep Learning Containers using PyTorch. Find Neuron DLC containers here: aws/deep-learning-containers

Neuron 2.4.0#

Date: 10/27/2022

  • Neuron now supports Kubernetes work scheduling at the level of NeuronCore. Updates on how to use the new core allocation method is captured in the Kubernetes documentation on this site.

Neuron 2.3.0#

Date: 10/10/2022

  • Now supporting TRN1 and INF1 EC2 instance types as part of Neuron. There is an optional aws-neuronx-oci-hooks package users may install for conveince that supports use of the AWS_NEURON_VISIBLE_DEVICES environment variable when launching containers. New DLC containers will be coming soon in support of training workloads on TRN1.

Neuron 1.19.0#

Date: 04/29/2022

  • Neuron Kubernetes device driver plugin now can figure out communication with the Neuron driver without the oci hooks. Starting with Neuron 1.19.0 release, installing aws-neuron-runtime-base and oci-add-hooks are no longer a requirement for Neuron Kubernetes device driver plugin.

Neuron 1.16.0#

Date: 10/27/2021

New in this release#

  • Starting with Neuron 1.16.0, use of Neuron ML Frameworks now comes with an integrated Neuron Runtime as a library, as a result it is no longer needed to deploy neuron-rtd. Please visit neuron-containers for more information.

  • When using containers built with components from Neuron 1.16.0, or newer, please use aws-neuron-dkms version 2.1 or newer and the latest version of aws-neuron-runtime-base. Passing additional system capabilities is no longer required.

This document is relevant for: Inf1, Inf2, Trn1, Trn1n