Quick start


Make sure to use Python 3.6+ and a virtual environment.

Using pip

pip install nlp_architect

From source

git clone https://github.com/NervanaSystems/nlp-architect.git
cd nlp-architect
pip install -e .  # install in development mode


For specific installation of backends of Tensorflow or PyTorch (CPU/MKL/GPU) we recommend installing NLP Architect and then installing the desired package of framework.


NLP Architect has the following packages:

Package Description
nlp_architect.api Model API interfaces
nlp_architect.common Common packages
nlp_architect.cli Command line module
nlp_architect.data Datasets, loaders and data processors
nlp_architect.models NLP, NLU and End-to-End models
nlp_architect.nn Topology related models and additions (per framework)
nlp_architect.pipelines End-to-end NLP apps
nlp_architect.procedures Procedure scripts
nlp_architect.server API Server and demos UI
nlp_architect.solutions Solution applications
nlp_architect.utils Misc. I/O, metric, pre-processing and text utilities


NLP Architect comes with a CLI application that helps users run procedures and processes from the library.


The CLI is in development and some functionality is not complete and will be added in future versions

The list of possible options can be obtained by nlp_architect -h:

nlp_architect commands:

train        Train a model from the library
run          Run a model from the library
process      Run a data processor from the library
solution     Run a solution process from the library
serve        Server a trained model using REST service

Use nlp_architect <command> -h for per command usage instructions.