# Hugging Face on AWS

Hugging Face partners with Amazon Web Services (AWS) to democratize artificial intelligence, enabling developers to seamlessly build, train, and deploy state-of-the-art machine learning models on AWS cloud infrastructure. This collaboration gives developers access to a growing catalog of pre-trained models and datasets from the Hugging Face Hub, through Hugging Face open-source libraries, across a broad spectrum of AWS services and hardware platforms.

We build new experiences to train and deploy Hugging Face models, whether you use AWS AI platforms such as Amazon SageMaker AI and AWS Bedrock, or AWS compute services such as Elastic Container Service (ECS), Elastic Kubernetes Service (EKS), and Amazon Elastic Compute Cloud (EC2). We also develop tools to simplify the adoption of custom AI accelerators like AWS Inferentia and AWS Trainium, designed to enhance the performance and cost-efficiency of machine learning workloads.

Whether you are building a first prototype or operating production workloads, you can choose the level of infrastructure control that fits your application.

## Where to start

Choose from the following options to quickly get started:

  
    Quickstart
    Deploy and test a Hugging Face model with SageMaker JumpStart.
    Deploy a model →
  
  
    Choose a service
    Compare SageMaker AI, Bedrock, AWS compute, and Inference Endpoints.
    Compare options →
  
  
    SageMaker SDK
    Deploy any Hub model to a managed endpoint from Python.
    Start from Python →
  

## Choose your path

For a more in-depth exploration of the different options, choose from the following paths:

  
    Amazon SageMaker AI
    Train, fine-tune, and deploy models with managed jobs and endpoints. Use the Python SDK for programmatic workflows or JumpStart for a guided deployment.
    
      Deploy →
      Train →
      JumpStart →
    
  
  
    Amazon Bedrock
    Combine JumpStart models with the managed Bedrock APIs and features such as Agents, Knowledge Bases, Guardrails, and Model Evaluation.
    Deploy with Bedrock →
  
  
    AWS compute
    Run Hugging Face DLCs on Amazon EC2, ECS, or EKS when you need direct control over networking, orchestration, and infrastructure.
    Explore EC2, ECS, and EKS →
  
  
    Hugging Face Inference Endpoints
    Deploy models on AWS infrastructure through a fully managed Hugging Face service, optimized for cost and throughput, without managing the serving stack yourself.
    Create an Inference Endpoint →
  

## Explore the documentation

Explore containers, agent workflows, guides, and end-to-end examples for building with Hugging Face on AWS.

  
    Deep Learning Containers
    What the DLCs are, how they work, and which images are available.
  
  
    Agents
    Connect agent harnesses to AWS-hosted models and manage SageMaker with skills.
  
  
    Guides
    Follow task-oriented training and deployment tutorials.
  
  
    Examples
    Explore complete notebooks for real workloads.
  

## Reference

- [Deep Learning Containers](./get-started/dlcs)
- [Inference Toolkit API](./reference/inference-toolkit)
- [Resources](./reference/resources)