Pre-configured environments

AWS Neuron provides pre-configured environments so you can start running workloads without manual SDK installation. Choose the option that best fits your deployment model: a Deep Learning AMI for EC2-based development, a Deep Learning Container for orchestrated deployments, or a custom Docker build for full control.

Which environment is right for you?

Option	Best for	Setup time	Customization	Deployment targets
Deep Learning AMI	EC2 development, quick prototyping, Jupyter notebooks	~5 minutes	Pre-configured virtual environments	EC2
Deep Learning Container	Production deployments, orchestrated workloads	~10 minutes	Container-based, framework-specific images	EKS, ECS, Batch, EC2
Custom Docker	Full control, CI/CD pipelines, custom dependencies	~30 minutes	Complete flexibility	Any

Get started

Deep Learning AMIs

Pre-configured EC2 machine images with Neuron SDK, frameworks, and virtual environments. Available in multi-framework, single-framework, and base variants.

Deep Learning Container images

Find the right pre-built Docker image for your framework and workload type. Includes PyTorch, JAX, and vLLM inference containers.

Customize a Deep Learning Container

Extend a Neuron DLC with additional packages or modify published Dockerfiles to fit your project.

Build and run Neuron containers with Docker

Install Neuron drivers, configure Docker, and build custom containers from scratch on EC2 instances.

Quickstarts

Quickstart: Deploy a DLC with vLLM

Configure and deploy a Deep Learning Container with vLLM for inference. ~30 minutes.

Quickstart: PyTorch inference with DLC

Run PyTorch inference using a pre-built Neuron DLC on EC2.