Curated Transformers
A PyTorch library of curated Transformer models and their composable components
Overview:
Curated Transformers is a transformer library for PyTorch that provides state-of-the-art models composed of reusable components. It offers support for various transformer models and has several standout features such as consistent type annotations for public APIs, minimal dependencies, and easy integration with existing code. It has been production-tested by Explosion and will be used as the default transformer implementation in spaCy 3.7.
Features:
- Supports state-of-the art transformer models: Includes models like Falcon, Llama, and Dolly v2.
- Reusable building blocks: Each model is composed of reusable components, making it easy to implement new features or bug fixes that benefit all models.
- Consistent type annotations: All public APIs have consistent type annotations for better coding support and integration with type-checked code.
- Great for education: The library's building blocks are easy to study, making it suitable for educational purposes.
- Minimal dependencies: Curated Transformers has minimal dependencies, ensuring ease of installation and usage.
Summary:
Curated Transformers is a powerful transformer library for PyTorch, offering state-of-the-art models composed of reusable building blocks. It supports various transformer architectures, provides consistent type annotations for better coding support, and has minimal dependencies. It has been tested by Explosion and will be the default transformer implementation in spaCy 3.7. Installation is straightforward, requiring only the installation of PyTorch and the library itself. Overall, Curated Transformers is a comprehensive and easy-to-use solution for working with transformer models in PyTorch.