hugging-face-datasets

What it does

This skill allows AI agents to interact with the Hugging Face Hub for managing datasets. It enables creation of new datasets, configuration of repository settings (including system prompts), efficient streaming of data updates, and complex queries and transformations using SQL. The tool is designed to be used alongside other Hugging Face tools and libraries.

When to use it

• Creating a new dataset: An agent needs to upload a collection of text files for training a language model.
• Updating an existing dataset: A data curator wants to add new examples to a public dataset on the Hub.
• Querying and transforming data: A researcher needs to extract specific subsets of data from a large dataset based on certain criteria, using SQL queries.
• Managing dataset configurations: An agent is responsible for setting up system prompts or other configuration parameters for a newly created dataset repository.

Key capabilities

• Repository initialization
• Config/system prompt definition
• Row streaming updates
• SQL-based querying and transformation

Example prompts

• "Create a new Hugging Face Dataset named 'my_awesome_dataset' with a README file."
• "Add 1000 new rows to the 'existing_dataset' dataset, using this CSV file."
• "Run a SQL query on the 'public_data' dataset to find all entries where the 'category' is 'science'."

Tips & gotchas

• Requires familiarity with the Hugging Face Hub and its API.
• SQL queries must adhere to the specific dialect supported by the underlying data storage format.