× Texar is a highly modularized and customizable toolkit to support a broad set of machine learning (ML), especially natural language processing (NLP) and text generation tasks.

Texar provides comprehensive modules for data processing, model architectures, loss functions, training and inference algorithms, evaluation, etc.

Use Texar to compose whatever models and algorithms you want.

× Texar-PyTorch is the PyTorch equivalence of Texar-TF, with mostly the same interfaces.

Texar-PyTorch integrates many of the best features of TensorFlow into PyTorch, delivering a set of highly usable and customizable modules superior to PyTorch native ones, including

• Data: More ready-to-use APIs; more customizable; more efficient
• Model: Better factorization; more comprehensive high-level APIs
• Training: High-level APIs to avoid any boilerplate code

× Forte is a high-level and customizable toolkit for building arbitrary complex NLP pipelines. With Forte, you can:

• Plug in any NLP components in a pipeline, including text analysis, generation, retrieval, etc.
• Wrap any off-the-shelf NLP models from other libraries as pipeline components, such as spaCy, StanfordNLP, AllenNLP, etc.
• Obtain a uniform representation of data and intermediate results throughout pipeline.

× Stave is a extendable toolkit for building front end interfaces for NLP projects. Stave excels at:

A general annotation and visualization interface for core NLP tasks.

Integrated and backed up by the strong machine learning and NLP toolkit built by ASYML: Texar and Forte.

Can be easily extended by building plugins for your requirements.