ML Container Creator

A Yeoman generator that creates SageMaker-compatible Docker containers for deploying ML models using the Bring Your Own Container (BYOC) paradigm.

Note: This is a pre-release (0.x). APIs may change between minor versions. Weekly releases are planned until v1.

Supported Configurations

Architecture	Model Servers	Use Case
HTTP (traditional ML)	Flask, FastAPI	sklearn, XGBoost, TensorFlow
Transformers (LLMs)	vLLM, SGLang, TensorRT-LLM, DJL/LMI	HuggingFace models, JumpStart, S3
Triton	FIL, ONNX, Python, TensorRT-LLM, vLLM	Multi-framework serving
Diffusors	vLLM	Image generation models

Deployment Target	Description
Managed Inference	SageMaker real-time endpoints
Async Inference	SageMaker async endpoints with S3 output
Batch Transform	SageMaker batch processing
HyperPod EKS	Kubernetes-based deployment

Quick Start

Install from npm

npm install -g yo @aws/generator-ml-container-creator

Or install from source

git clone https://github.com/awslabs/ml-container-creator.git
cd ml-container-creator
npm install && npm link

Bootstrap AWS infrastructure (one-time)

yo @aws/ml-container-creator bootstrap

Sets up an IAM execution role, ECR repository, and optional S3 buckets. Configuration is saved to ~/.ml-container-creator/config.json.

Generate a project

# Interactive
yo @aws/ml-container-creator

# Non-interactive
yo @aws/ml-container-creator my-model \
  --deployment-config=transformers-vllm \
  --model-name=openai/gpt-oss-20b \
  --instance-type=ml.g6.12xlarge \
  --region=us-east-1 \
  --skip-prompts

Build, push, deploy

./do/build        # Build Docker image
./do/push         # Push to Amazon ECR
./do/deploy       # Deploy to SageMaker
./do/test         # Test the endpoint

Documentation

Full documentation is available at awslabs.github.io/ml-container-creator.

• Getting Started — Installation and walkthroughs
• Configuration — CLI flags, env vars, config files, MCP servers
• Deployment Guide — All deployment targets and lifecycle scripts
• Examples — Framework-specific walkthroughs
• Troubleshooting — Common issues and solutions

Prerequisites

Tool	Version	Purpose
Node.js	24+	Runs the Yeoman generator
Docker	20+	Container builds
AWS CLI	2+	AWS resource management

Contributing

See CONTRIBUTING.md for guidelines.

Security

See CONTRIBUTING.md for reporting security issues.

License

Apache-2.0. See LICENSE.