AllenNLP

AllenNLP was developed by the Allen Institute for AI as a PyTorch research library for building deep learning models on linguistic tasks. It supplies abstractions for datasets, data readers, models, and training loops, a configuration-driven experiment system, a command-line interface, and modules for evaluation, interpretability, and fairness. For years it was a standard foundation for academic NLP work and backed many published models. The project is now in maintenance mode: the team stopped adding features and upgrading dependencies, ended official support in December 2022, and recommends alternatives such as Hugging Face Transformers or AI2 Tango for new projects. It requires Python 3.6.1 or later, runs on Linux and macOS but not Windows, installs via pip, conda-forge, or Docker, and remains available under the Apache 2.0 license, which keeps it useful for reproducing older research and for specific components like its fairness module.