hugging-face-cli

What it does

The huggingface-cli skill allows users to interact with the Hugging Face Hub directly from an AI agent. It enables downloading pre-trained models and datasets, uploading new content like files and models, creating new repositories on the Hub, and managing local caches for efficient workflow. This tool streamlines common tasks related to accessing and contributing to the Hugging Face ecosystem.

When to use it

• Downloading a model: A user needs a specific pre-trained language model for fine-tuning; this skill can download it directly.
• Creating a new repository: A data scientist wants to share their newly trained model with the community and create a dedicated space on the Hugging Face Hub.
• Uploading training data: A researcher has processed a dataset and needs to upload it as a new dataset on the Hugging Face Hub.
• Managing cached files: An agent is running out of disk space, and needs to clear the local cache of downloaded models and datasets.

Key capabilities

• Downloading models and datasets from the Hugging Face Hub
• Uploading files to existing or new repositories
• Creating new repositories on the Hugging Face Hub
• Managing local cached files

Example prompts

• "Download the 'bert-base-uncased' model."
• "Create a new repository named 'my-awesome-model'."
• "Upload this file to my existing 'my-project' repository."

Tips & gotchas

• The user must be logged in to their Hugging Face account for many operations. Ensure the agent has access to authentication credentials.
• Large model downloads can take significant time and bandwidth; consider implementing progress monitoring or resuming interrupted downloads.