/text-classification

Primary LanguagePython

How to run

  1. Download the archive with the model and extract its contents into model folder in the root of the repository
  2. Run docker compose up command

The service consists of four containers - model-preparation, inference, database, api

Model preparation

This container handles model conversion to onnx format

Inference

This container deploys the newly created onnx model to the triton inference server. This container only starts when the model-preparation container successfully finishes its work

Database

This container hosts a PostgreSQL server for storing the user ids

API

This container exposes the /predict endpoint to the outer world, so we can submit a POST request with a JSON that contains user_id and text fields. This container only starts after inference and database are considered as healthy

Example:

curl -X POST http://localhost:5000/predict \
   -H 'Content-Type: application/json' \
   -d '{"user_id":1, "text": "57 year old man with pancreatitis, alcohol withdrawal, tachypnea"}'