C-Ritam98 / Style_Transfer

A backtranslation based style transfer model.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Style_Transfer

Implementation of the paper "Content preserving text generation with attribute controls", Logeswaran et al., which talks 'bout a backtranslation based style transfer network.

Logo

Deployment

First of all, place all the files/folders present in this repository inside a new directory named "my_CPTG", else you will get an alert mentioning the same.

Next, re-install torch and torchtext versions 1.9.1 and 0.10.1 [with(out) cudnn as nccessary] by this command

$ pip3 install torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 torchtext==0.10.1 -f https://download.pytorch.org/whl/torch_stable.html

Choose the dataset to train on by manipulating dataset_name variable in main.py and eval.pyfiles by names yelp/amazon/imdb.

To train the model run

  $ python3 main.py

To test the model run

  $ python3 eval.py

Some Examples

Examples of style transferred text produced by the model trained on yelp data set.

  • “horrible service .” --> “great staff .”
  • “a worthwhile read for the but it is ” --> “a boring read that the is but that a”
  • “ very interesting look into the life of our first great ...” --> “very boring rambling to view of of life “

Acknowledgements

About

A backtranslation based style transfer model.


Languages

Language:Python 100.0%