Поделиться через


How to use Azure AI services in Azure AI Foundry portal

You might have existing resources for Azure AI services that you used in the old studios such as Azure OpenAI Studio or Speech Studio. You can pick up where you left off by using your existing resources in the Azure AI Foundry portal.

This article describes how to use new or existing Azure AI services resources in an Azure AI Foundry project.

Сценарии использования

Depending on the AI service and model you want to use, you can use them in Azure AI Foundry portal via:

  • Bring your existing Azure AI services resources into a project. You can use your existing Azure AI services resources in an Azure AI Foundry project by creating a connection to the resource.
  • The model catalog. You don't need a project to browse and discover Azure AI models. Some of the Azure AI services are available for you to try via the model catalog without a project. Some Azure AI services require a project to use them in the playgrounds.
  • The project-level playgrounds. You need a project to try Azure AI services such as Azure AI Speech and Azure AI Language.
  • Azure AI Services demo pages. You can browse Azure AI services capabilities and step through the demos. You can try some limited demos for free without a project.
  • Fine-tune models. You can fine-tune a subset of Azure AI services models in Azure AI Foundry portal.
  • Deploy models. You can deploy base models and fine-tuned models to production. Most Azure AI services models are already deployed and ready to use.

Bring your existing Azure AI services resources into a project

Let's look at two ways to connect Azure AI services resources to an Azure AI Foundry project:

Connect Azure AI services when you create a project for the first time

When you create a project for the first time, you can select an existing Azure AI services resource (including Azure OpenAI) or create a new AI services resource.

Снимок экрана: страница создания ресурса в диалоговом окне создания проекта.

Дополнительные сведения о создании проекта см. в практическом руководстве по созданию проекта Azure AI Foundry или в кратком руководстве по созданию проекта и использованию игровой площадки для чата.

Connect Azure AI services after you create a project

To use your existing Azure AI services resources (such as Azure AI Speech) in an Azure AI Foundry project, you need to create a connection to the resource.

  1. Create an Azure AI Foundry project. For detailed instructions, see Create an Azure AI Foundry project.

  2. Перейдите в проект Azure AI Foundry.

  3. Выберите центр управления в левой области.

  4. Выберите подключенные ресурсы (в разделе "Проект") на левой панели.

  5. Выберите + Новое подключение.

    Снимок экрана: страница подключенных ресурсов с кнопкой для создания нового подключения.

  6. На странице "Добавление подключения к внешним ресурсам" выберите тип службы искусственного интеллекта, которую требуется подключить к проекту. For example, you can select Azure AI services (for a connection to multiple services in one resource), Azure OpenAI Service, Azure AI Content Safety, Azure AI Speech, Azure AI Language, and other AI services.

    Снимок экрана: страница для выбора типа службы искусственного интеллекта, которую требуется подключить к проекту.

  7. На следующей странице мастера просмотрите или выполните поиск, чтобы найти ресурс, который вы хотите подключить. Затем выберите Добавить подключение.

    Screenshot of the page to select the Azure AI resource that you want to connect to the project.

  8. После подключения ресурса нажмите кнопку "Закрыть ", чтобы вернуться на страницу "Подключенные ресурсы ". You should see the new connection listed.

Discover Azure AI models in the model catalog

You can discover Azure AI models in the model catalog without a project. Some Azure AI services are available for you to try via the model catalog without a project.

  1. Go to the Azure AI Foundry portal home page.

  2. Select the tile that says Model catalog and benchmarks.

    Screenshot of the home page in Azure AI Foundry portal with the option to select the model catalog tile.

    If you don't see this tile, you can also go directly to the Azure AI model catalog page in Azure AI Foundry portal.

  3. From the Collections dropdown, select Microsoft. Search for Azure AI services models by entering azure-ai in the search box.

    Screenshot of the model catalog page in Azure AI Foundry portal with the option to search by collection and name.

  4. Select a model to view more details about it. You can also try the model if it's available for you to try without a project.

Try Azure AI services in the project level playgrounds

In the project-level playgrounds, you can try Azure AI services such as Azure AI Speech and Azure AI Language.

  1. Перейдите в проект Azure AI Foundry. Если вам нужно создать проект, см. статью "Создание проекта Azure AI Foundry".

  2. Выберите площадки на левой панели и выберите нужную площадку. In this example, select Try the Speech playground.

    Снимок экрана: игровые площадки уровня проекта, которые можно использовать.

  3. При необходимости можно выбрать другое подключение для использования на детской площадке. В речевой песочнице можно подключиться к ресурсам многосервисных служб искусственного интеллекта Azure или ресурсам речевой службы.

    Снимок экрана: игровая площадка

If you have other connected resources, you can use them in the corresponding playgrounds. For example, in the Language playground, you can connect to Azure AI Services multi-service resources or Azure AI Language resources.

Screenshot of the Language playground in a project.

Try out Azure AI Services demos

You can browse Azure AI services capabilities and step through the demos. You can try some limited demos for free without a project.

  1. Go to the Azure AI Foundry home page and make sure you're signed in with the Azure subscription that has your Azure AI services resource.

  2. Find the tile that says Explore Azure AI Services and select Try now.

    Screenshot of the home page in Azure AI Foundry portal with the option to select Azure AI Services.

    If you don't see this tile, you can also go directly to the Azure AI Services page in Azure AI Foundry portal.

  3. You should see tiles for Azure AI services that you can try. Select a tile to get to the demo page for that service. For example, select Language + Translator.

    Screenshot of the landing page to try Azure AI Services try out capabilities in Azure AI Foundry portal.

The presentation and flow of the demo pages might vary depending on the service. In some cases, you need to select a project or connection to use the service.

Fine-tune Azure AI services models

In Azure AI Foundry portal, you can fine-tune some Azure AI services models. For example, you can fine-tune a model for custom speech.

  1. Перейдите в проект Azure AI Foundry. Если вам нужно создать проект, см. статью "Создание проекта Azure AI Foundry".

  2. Выберите Тонкая настройка на левой панели.

  3. Select AI Service fine-tuning.

    Screenshot of the page to select fine-tuning of Azure AI Services models.

  4. Select + Fine-tune.

  5. Follow the wizard to fine-tune a model for the capabilities that you want.

Deploy models to production

Once you have a project, several Azure AI services models are already deployed and ready to use.

  1. Перейдите в проект Azure AI Foundry.

  2. Выберите центр управления в левой области.

  3. Select Models + endpoints (under Project) from the left pane.

  4. Select the Service endpoints tab to view the list of Azure AI services models that are already deployed.

    Screenshot of the models and endpoints page to view Azure AI services endpoints.

    In this example, we see:

    • Six Azure AI Services deployments (such as Azure AI Speech and Azure AI Language) via the default connection. These models were already available for use when you created the project.
    • Another Azure AI Speech deployment via the contosoazureaispeecheastus example connection. This example assumes that you connected to an Azure AI Speech resource after creating the project. For more information about connecting to Azure AI services, see Bring your existing Azure AI services resources into a project.

There's no option to deploy Azure AI services models from the Models + endpoints page. Azure AI services models are already deployed and ready to use.

However, you can deploy fine-tuned Azure AI services models. For example, you might want to deploy a custom speech model that you fine-tuned. In this case, you can deploy the model from the corresponding fine-tuning page.