This repository supplies a simple example of a python module/step for OHNLPTK Backbone pipelines.
To develop your own components, make a copy of this repository and instantiate
a conda environment using conda env create -n <environment_name> -f environment.yml
and implement the following abstract classes (refer to wordcount.py as an example):
ohnlp.tookit.backbone.api.BackboneComponentDefinitionohnlp.tookit.backbone.api.BackboneComponent- Some DoFn class, namely one of:
ohnlp.toolkit.backbone.api.BackboneComponentOneToOneDoFnohnlp.toolkit.backbone.api.BackboneComponentOneToManyDoFn
then update backbone_module.json accordingly.
Finally, update environment.yml with any dependency changes incurred as part of this process. It is strongly encouraged to use conda packages instead of pypi (installed via conda as opposed to pip) as much as possible to avoid packaging problems
To generate the zip file distributed with backbone, simply push to github with
a tag following the format of 'v[0-9]+.[0-9]+.[0-9]+*' or edit build.yml accordingly
If you do not wish to make your component public, then execute conda-pack -o env-{win32|linux|darwin}.tar.gz from within
the project root directory and zip the entire folder. Note that this should be executed using the same OS on which you expect
OHNLP Toolkit to be ran.