|
- fastNLP documentation
- =====================
- A Modularized and Extensible Toolkit for Natural Language Processing. Currently still in incubation.
-
-
- Introduction
- ------------
-
- FastNLP is a modular Natural Language Processing system based on
- PyTorch, built for fast development of NLP models.
-
- A deep learning NLP model is the composition of three types of modules:
-
- +-----------------------+-----------------------+-----------------------+
- | module type | functionality | example |
- +=======================+=======================+=======================+
- | encoder | encode the input into | embedding, RNN, CNN, |
- | | some abstract | transformer |
- | | representation | |
- +-----------------------+-----------------------+-----------------------+
- | aggregator | aggregate and reduce | self-attention, |
- | | information | max-pooling |
- +-----------------------+-----------------------+-----------------------+
- | decoder | decode the | MLP, CRF |
- | | representation into | |
- | | the output | |
- +-----------------------+-----------------------+-----------------------+
-
-
- For example:
-
- .. image:: figures/text_classification.png
-
-
-
-
- User's Guide
- ------------
- .. toctree::
- :maxdepth: 2
-
- user/installation
- user/quickstart
-
-
- API Reference
- -------------
-
- If you are looking for information on a specific function, class or
- method, this part of the documentation is for you.
-
- .. toctree::
- :maxdepth: 2
-
- fastNLP API <fastNLP>
-
-
-
-
- Indices and tables
- ==================
-
- * :ref:`genindex`
- * :ref:`modindex`
- * :ref:`search`
|