OpenPrompt and promptsource
These are complements: OpenPrompt provides a framework for implementing prompt-learning methods across different model architectures, while PromptSource supplies curated, reusable prompt templates and datasets that can be integrated into OpenPrompt pipelines for evaluation and experimentation.
About OpenPrompt
thunlp/OpenPrompt
An Open-Source Framework for Prompt-Learning.
Provides modular components—Templates and Verbalizers—that transform input text and map labels to vocabulary words, enabling prompt-based classification and generation on pre-trained language models from Hugging Face Transformers. Supports diverse prompting methods (templating, verbalizing, optimization strategies) under a unified API, with extensibility for custom prompt-learning implementations and multi-paradigm task adaptation (classification and generation).
About promptsource
bigscience-workshop/promptsource
Toolkit for creating, sharing and using natural language prompts.
Prompts are defined using Jinja templating syntax and stored as standalone structured files, enabling version control and reproducibility across 170+ datasets. The toolkit integrates directly with Hugging Face's Datasets library and includes a Streamlit-based web interface for interactive prompt creation, testing, and browsing the P3 collection of 2,000+ community-contributed prompts.
Scores updated daily from GitHub, PyPI, and npm data. How scores work