/reqexp-deployment

Primary LanguagePython

Reqexp-deployment

This repository contains the Docker configuration files for the deployment of the ReqExp project.

Reqexp Technical stack

ReqExp server needs around 8 GB of RAM. ReqExp is based on the following technology stack:

  • Ubuntu 18.04
  • Python 3.5+
  • Flask App Server
  • TensorFlow 1.14
  • BERT as a Service
  • Server-side Python packages: pandas, numpy, json, sklearn, spacy, bert-serving, GPUtil, subprocess

Build and run the image

  • To build the docker image using: docker build --tag reqexp-deployment_reqexp:latest reqexp/.
  • To run the docker image using: docker run --publish 56733:56733 --name reqexp reqexp-deployment_reqexp:latest

Accessing with REST endpoints

The ReqExp web server expose four REST endpoints.

  • The following command sends a POST request to the endpoint (textprob) with text data (see “text” field in the JSON-payload). curl -i -H "Content-Type: application/json" -X POST -d '{"text":"But there are other areas that can and should be explored by the community. "}' http://localhost:56733/textsprob

  • Add data set and retrain: ** the following command sends a POST request to the endpoint (addtrain) with labeled text data (see “trainset” field in the JSON-payload. curl -i -H "Content-Type: application/json" -X POST -d '{"trainset": [["But there are other areas that can and should be explored by the community. ","1"],["gui must provide data ", "0"]] }' http://localhost:56733/addtrain

** The following command sends a POST request to retrain the DNN classifier model with labeled text data received in the last call of the addtrain. curl -i -X POST http://localhost:56733/retrain

  • Get status of the retraining: ** the following command sends a GET request to receive a status of the retraining process. curl -i -X GET http://localhost:56733/status