Hugging Face datasets

¶Create

The Create API allows you to programmatically create new Hugging Face repositories for datasets, models, or spaces. This is useful for automating dataset or model management workflows directly from your PHP applications.

Tip

Repositories can be of type dataset, model, or space. Use this method to initialize your Hugging Face project structure before pushing content.

¶Valid Repository Types

Note

This example focuses on datasets, but the same approach applies to models and spaces.

¶Code

use Partitech\PhpMistral\Clients\HuggingFace\HuggingFaceDatasetClient;
use Partitech\PhpMistral\MistralClientException;

$apiKey = getenv('HF_TOKEN');     // Hugging Face API token
$datasetUser = getenv('HF_USER'); // Hugging Face username or organization

$client = new HuggingFaceDatasetClient(apiKey: (string) $apiKey);

try {
    // Create a new dataset repository
    $repository = $client->create(
        name: 'rotten_tomatoes',                              // Repository name (without user prefix)
        type: HuggingFaceDatasetClient::REPOSITORY_TYPE_DATASET,  // Repository type
        private: false                                        // Set to true for private repositories
    );

    print_r($repository);  // Display repository details

} catch (MistralClientException $e) {
    print_r($e);  // Handle errors (e.g., if repository already exists)
}

¶Result

Array
(
    [url] => https://huggingface.co/datasets/Bourdin/rotten_tomatoes
    [name] => Bourdin/rotten_tomatoes
    [id] => 68092ce920bd0ac0b4a4f28c
)

• url: URL to access the repository on Hugging Face.
• name: Full repository name (including user or organization prefix).
• id: Internal Hugging Face identifier for the repository.

¶Use Cases

• Dataset initialization: Automate the creation of dataset repositories for data pipelines.
• Model deployment: Set up model repositories before pushing weights and configurations.
• Spaces setup: Automate the provisioning of spaces for interactive apps (e.g., Gradio, Streamlit).

¶Common Pitfalls

Warning

• The repository name must be unique within your Hugging Face account or organization.
• Ensure your Hugging Face API token has the necessary write permissions.
• This API does not support renaming repositories. Use Hugging Face's web interface for renames.

Caution

For spaces, make sure the repository contains the necessary configuration files (e.g., app.py, requirements.txt) to run the app after creation.