Trace Id is missing
Skip to main content

Hugging Face on Azure

Deploy tens of thousands of pretrained Hugging Face Transformers models in Azure Machine Learning using Hugging Face endpoints.

Build machine learning models faster with Hugging Face on Azure

Hugging Face is the creator of Transformers, the leading open-source library for building state-of-the-art machine learning models. Use the Hugging Face endpoints service (preview), available on Azure Marketplace, to deploy machine learning models to a dedicated endpoint with the enterprise-grade infrastructure of Azure.

Machine learning models for natural language processing, audio and computer vision in the Hugging Face Hub

Azure Kubernetes Service (AKS)

Choose from tens of thousands of machine learning models for Natural Language Processing, Audio, and Computer Vision, publicly available in the Hugging Face Hub, to accelerate your machine learning workload.

A user creating Hugging Face Inference Endpoints on Azure

Application services

Spend less time optimizing models for production and more time building machine learning products.

Data services

Keep up with the pace of innovation by easily switching production models. Help keep your machine learning operations (MLOps) secure using Azure Private Link—your data never leaves your virtual private cloud.

Back to tabs
Read the announcement about Hugging Face on Azure

Learn more about Hugging Face on Azure

Expert support

Get direct support from Hugging Face machine learning experts.

Private model hub

Experiment, collaborate, train, and serve state-of-the-art models in your own private Hugging Face Hub.

Hugging Face course

Learn everything you need to know about Hugging Face Transformers in a free course.

Customers are doing great things with Hugging Face on Azure

Democratizing responsible machine learning

"Hugging Face endpoints take care of the most pressing issues when it comes to model deployment. With just a few clicks or a few lines of Azure SDK code, you select a model and a task type, and you can start predicting in minutes."

Mabu Manaileng, Principal AI Engineer, Standard Bank

People having a conversation in a meeting

Frequently asked questions about Hugging Face on Azure

  • Download Hugging Face endpoints from Azure Marketplace, select a machine learning model and task type, select an Azure instance type, and create an endpoint. Within minutes, you can test your endpoint and add its inference API to your application.

  • The Hugging Face endpoints service is free while in preview. You pay only for compute costs via your Azure subscription.

  • Hugging Face on Azure is an easy way to inference endpoints for Hugging Face models in your Azure account through Azure Machine Learning. The Inference API is a serverless solution offered by Hugging Face as a hosted service.

  • To deploy your own models onto Hugging Face endpoints, upload them to your Hugging Face account. Unlimited private model hosting is available for free.

  • Yes, you can deploy Hugging Face models using the Transformers open-source library or using managed or serverless services. With Hugging Face on Azure, you don't need to build or maintain infrastructure, and you benefit from the security and compliance of Azure Machine Learning. Hugging Face on Azure also provides easy autoscaling and private connections via Azure Private Link.

Ready when you are—let’s set up your Azure free account

Start free
OSZAR »