https://github.com/machine-learning-tokyo/poetry-gan

https://github.com/machine-learning-tokyo/poetry-gan

creativity generative-adversarial-network machine-learning nlp research

Last synced: 6 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/machine-learning-tokyo/poetry-gan
• Owner: Machine-Learning-Tokyo
• License: mit
• Created: 2018-12-08T11:36:37.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-10-31T03:54:53.000Z (about 5 years ago)
• Last Synced: 2023-03-11T10:07:14.521Z (over 1 year ago)
• Topics: creativity, generative-adversarial-network, machine-learning, nlp, research
• Language: Jupyter Notebook
• Size: 2.46 MB
• Stars: 49
• Watchers: 5
• Forks: 12
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Creative GANs

Creative GANs is a research project on training language models to generate creative text with either a Maximum Likelihood Estimation (MLE) or GAN objective. It contains the code for running the experiments as described in the [paper](https://arxiv.org/abs/1909.09534). 

Some curated outputs can be found [here](https://www.ai-fragments.com/)

## Datasets

- Gutenberg Novels

- English poetry

- Song lyrics

- [Metaphors](http://metaphors.iath.virginia.edu/metaphors)

## Text Generative Models with LM and GAN objective

- AWD LSTM [paper link](https://arxiv.org/pdf/1708.02182.pdf)

- Transformer XL [paper link](https://arxiv.org/pdf/1901.02860.pdf)

The encoder models are from the fastai library

## Usage

- Preprocess data `python preprocess.py [gutenberg/metaphors/poems/lyrics]` and save preprocessed file

- Train language model `lang_model.py PATH FILENAME MODEL [PRETRAINED_FNAMES]`

- Train gan model `textgan.py PATH FILENAME PRETRAINED MODEL [CRIT] [PREDS] [EPOCHS]`

```

PATH - folder with data 

FILENAME - name of preprocessed file

PRETRAINED - fastai model saved with learn.save()

PRETRAINED_FNAMES - pretrained weight file and vocab file (comma seperated)

MODEL - architecture to use {'AWD': AWD_LSTM, 'XL':TransformerXL}

CRIT - loss function: gumbel softmax/reinforce (only for gan)

PREDS - generate output from validation set 

EPOCHS - number of epochs to train

```