hugging-face-dataset-viewer

What it does

The Hugging Face Dataset Viewer skill allows you to interact with the Hugging Face Datasets platform programmatically. It enables retrieval of dataset metadata like split information, pagination through data rows, searching within textual content, applying filters to datasets, downloading parquet files associated with a dataset, and accessing size or statistical information about a dataset. This provides a flexible way to explore and understand datasets without relying solely on the web interface.

When to use it

• Data Exploration: Quickly browse a new Hugging Face dataset to understand its structure and content before committing to full download.
• Filtering Data: Identify specific data points within a large dataset based on criteria like text content or metadata tags.
• Metadata Inspection: Retrieve information about the splits, sizes, and statistics of a dataset for documentation or analysis purposes.
• Parquet Download: Obtain direct URLs to download Parquet files associated with a Hugging Face dataset for offline processing.

Key capabilities

• Fetch subset/split metadata
• Paginate rows within a dataset
• Search text content within the dataset
• Apply filters to datasets
• Download parquet URLs
• Read size or statistics about the dataset

Example prompts

• "Show me the splits for the 'rotten_tomatoes' dataset."
• "Find all rows in the 'wikitext' dataset where the text contains the word 'example'."
• "What is the total size of the 'imdb' dataset?"

Tips & gotchas

• You need to have a Hugging Face account and be logged in to use this skill effectively, as authentication may be required for certain datasets.
• Be mindful of large datasets; pagination and filtering are crucial for efficient interaction and avoiding timeouts.