Natural Language Processing Custom models¶

Natural Language Processing (NLP) provides a set of libraries within Watson Studio to allow Data Scientists to create bespoke models. Theses can be exported from Watson Studio as an compressed archive by adding the following to a cell at the end of the model

project.save_data('ensemble_model', data=ensemble_model.as_file_like_object(), overwrite=True)

this can then be transferred from the Watson Studio environment to your workstation (copy the file into a models directory).

You can create a combined container with the runtime and model

ARG WATSON_RUNTIME_BASE="cp.icr.io/cp/ai/watson-nlp-runtime:1.0.18"
FROM ${WATSON_RUNTIME_BASE} as base
ENV LOCAL_MODELS_DIR=/app/models
COPY models /app/models

Once built the container can now be used as outlined in the combined model section

You can also use the packaging tool to create the containers