huggingface/hub-docs
Hub Docs
This repository serves as the documentation source for the Hugging Face Hub, a collaborative platform designed for hosting, versioning, and discovering machine learning models, datasets, and interactive applications. It provides the foundational infrastructure for managing machine learning assets through Git-based repositories, which support large file storage, branching, and comprehensive commit history.
The platform distinguishes itself by integrating metadata-driven discovery and structured management systems that allow users to attach licensing, task categories, and performance metrics to their assets. It enforces security through granular access controls, including gated access requirements, malware scanning, and GPG commit signing, while managing traffic through request rate limiting to ensure service stability.
Beyond core hosting, the system facilitates collaborative project management by enabling organizations and teams to centralize resource sharing and role-based access. It also supports the deployment of interactive web-based demos through containerized environments and provides tools for local documentation previewing and static site generation to ensure consistent content delivery.
Features
- Machine Learning Hubs - A collaborative platform for hosting, versioning, and discovering machine learning models, datasets, and interactive applications using Git.
- Model Hosting Platforms - Hosts and serves pre-trained machine learning models with standardized metadata, documentation, and browser-based inference widgets.
- Web-Based Model Hosting - Hosts machine learning models with documentation and browser-based inference widgets for testing and programmatic access.
- Streaming Dataset Loaders - Enables storing and retrieving large-scale datasets with streaming capabilities for efficient access to data exceeding local memory.
- Machine Learning Datasets - Manages, documents, and streams large-scale datasets with structured metadata for improved discoverability and access.
- Git and Version Control - Provides Git-based repositories to track commit history, branching, and file changes for machine learning models and datasets.
- Dataset Management Frameworks - Manages and versions large-scale datasets with structured metadata, schema definitions, and streaming capabilities.
- Git Large File Storage - Manages massive binary artifacts by offloading large files to external storage while maintaining lightweight pointers in the repository.
- Git-Based Repositories - Uses Git-based repositories to provide version control, branching, and commit history for machine learning artifacts.
- Data Versioning Repositories - Provides Git-based repository hosting for models and datasets with support for version control and large file handling.
- Machine Learning Security - Protecting assets through access tokens, GPG commit signing, malware scanning, and gated access requirements for sensitive datasets and models.
- Repository Access Controls - Enforces authentication, permission checks, and user agreements before granting access to sensitive machine learning assets.
- Machine Learning Demo Platforms - Provides interactive web-based environments for showcasing and testing machine learning models directly in the browser.
- Model Evaluation Metrics - Records standardized performance metrics and evaluation conditions within model cards to provide verifiable benchmarks.
- Model Metadata Loggers - Supports attaching structured metadata to model repositories to enable discovery and interoperability.
- Team Collaboration Management - Organizes teams into organizations to centralize resource sharing, access roles, and billing for machine learning development.
- Metadata-Driven Discovery - Enables programmatic filtering and discovery of machine learning assets through structured metadata embedded in documentation.
- Static Site Generation - Transforms documentation source files into optimized web pages during the build process for consistent content delivery.
- Dataset Metadata Schemas - Enables specifying data types and feature structures to facilitate automated validation and integration of datasets.
- Dataset Metadata Mapping - Allows attaching structured information like licensing and task categories to datasets to improve discoverability.
- Team Management - Supports grouping users into teams to share resources, assign access roles, and manage collective machine learning projects.
- Application Runtime Containers - Executes interactive web demos within isolated container environments to support diverse Python frameworks and hardware acceleration.
- Interactive Demo Deployments - Deploys interactive web-based machine learning demos using Python SDKs, static HTML, or containerized environments.
- Object Storage Providers - Provides S3-compatible object storage for large files like training checkpoints that do not require version control.
- Contact-Gated Access - Requires users to accept terms or provide information before granting access to sensitive datasets.
- Organization and Project Group Roles - Enables grouping users into organizations to centralize project management, role-based access, and shared billing.
