Skip to content

Latest commit

 

History

History
35 lines (23 loc) · 529 Bytes

README.md

File metadata and controls

35 lines (23 loc) · 529 Bytes

chainer-fm-gan

This is a re-implimentation of FM-GAN using chainer.

Requirements:

  • Python 3.5+
  • Chainer 5.2+
  • wget, nkf, mecab (for preprocessing dataset)

How to run

Dataset

data/ directory contains scripts to download the Livedoor news corpus and preprocess them.

Just type the following:

cd data && make all

Train

export GPU=0 # if you want to use GPU
./train.sh # pretrain LSTM-VAE
./gantrain.sh # GAN training

License

3-clause BSD