Skip to main content

External Integrations

External Integrations let you connect popular ML platforms and container registries directly to your TIR environment. Once configured, these integrations authenticate your nodes and pipelines with third-party services — enabling private repository access, model downloads, experiment tracking, and container pulls without manual credential management.

Supported Integrations​

IntegrationPurpose
Hugging FaceDownload models and datasets from Hugging Face Hub
GitHubAccess private repositories using a personal access token
Weights & BiasesTrack experiments, log metrics, and manage ML runs

Access External Integrations​

  1. Open the TIR Dashboard.
  2. In the sidebar, navigate to Labs Experimental → External Integrations.

The page lists all configured integrations with their type, name, and status.

Hugging Face​

Hugging Face integration gives TIR authenticated access to models, datasets, and other resources on the Hugging Face Hub.

Create a Hugging Face Integration​

  1. Click Create Integration.

  2. Select Hugging Face as the integration type.

  3. Fill in the following fields:

    FieldDescription
    NameA label to identify this integration
    Access TokenYour Hugging Face user access token

  4. Click Create.

tip

Retrieve your Hugging Face token from Hugging Face → Settings → Access Tokens.

GitHub​

GitHub integration authenticates your TIR environment with GitHub using a personal access token, allowing access to private repositories during training and pipeline runs.

Create a GitHub Integration​

  1. Click Create Integration.

  2. Select GitHub as the integration type.

  3. Fill in the following fields:

    FieldDescription
    NameA label to identify this integration
    Personal Access TokenA GitHub PAT with the required repository permissions

  4. Click Create.

tip

Generate a GitHub Personal Access Token from GitHub → Settings → Developer Settings → Personal access tokens. For guidance on token types and scopes, see Managing your personal access tokens.

Weights & Biases​

Weights & Biases integration connects TIR to your W&B workspace, enabling automatic experiment tracking and metric logging during training runs.

Create a Weights & Biases Integration​

  1. Click Create Integration.

  2. Select Weights & Biases as the integration type.

  3. Fill in the following fields:

    FieldDescription
    NameA label to identify this integration
    WandB UsernameYour Weights & Biases account username
    WandB KeyYour W&B API key
    Project NameThe W&B project where runs will be logged

  4. Click Create.

tip

Retrieve your W&B API key from Weights & Biases → Authorize.

Delete an Integration​

  1. Click the delete icon next to the integration you want to remove.
  2. Confirm your choice in the popup and click Delete.
warning

Deleting an integration immediately revokes TIR's access to that service. Any running workloads that depend on the integration may fail.

Quick Reference​

IntegrationRequired Fields
Hugging FaceName, Access Token
GitHubName, Personal Access Token
Weights & BiasesName, WandB Username, WandB Key, Project Name