Make sure to use Python 3.6+ and a virtual environment.
pip install nlp_architect
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:
|nlp_architect.api||Model API interfaces|
|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.server||API Server and demos UI|
|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
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
nlp_architect <command> -h for per command usage instructions.