Skip to content

Latest commit

 

History

History
39 lines (25 loc) · 1.16 KB

README.md

File metadata and controls

39 lines (25 loc) · 1.16 KB

Paraphrase generation using T5 model

Simple application using T5 base model fine tuned in Quora Question Pairs to generate paraphased questions.

This repository is based on the work from @ramsrigouthamg which explain very well how to fine tune the model.

Paraphrase generation using Google UDA.

You will able to create paraphrase using a technique call Back Translate. You can find more info about it here:-https://github.com/google-research/uda

Application

Paraphrase Paraphrase Paraphrase Paraphrase

Install

pip install -r requirements.txt
  1. replace core_estimator_predictor.py in your_env/lib/python3.6/site-packages/tensorflow/contrib/predictor with the file (core_estimator_predictor.py) in repository.
  2. Download models wget https://storage.googleapis.com/uda_model/text/back_trans_checkpoints.zip unzip back_trans_checkpoints.zip && rm back_trans_checkpoints.zip

Running

cd web-app
python app.py

Open your browser http://localhost:8001