174

GitHub - DongjunLee/char-rnn-tensorflow: Multi-layer Recurrent Neural Networks f...

 6 years ago
source link: https://github.com/DongjunLee/char-rnn-tensorflow
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.

char-rnn

This code implements multi-layer Recurrent Neural Network (RNN, LSTM, and GRU) for training/sampling from character-level language models.

Requirements

  • Python 3.6
  • TensorFlow 1.4
  • hb-config

Features

  • Using Higher-APIs in TensorFlow

  • Korean SamhangSi (like acrostic poem)

Config

example: check_tiny.yml

data:
  data_dir: 'data/tiny_lyricskor'
model:
  batch_size: 4
  input_keep_prob: 0.8
  log_dir: 'logs'
  num_layers: 1
  output_keep_prob: 0.8
  rnn_size: 64
  seq_length: 20
train:
  train_steps: 10000
  model_dir: 'tiny_checkpoints'
  save_every: 1
  learning_rate: 0.001
  loss_hook_n_iter: 100
  check_hook_n_iter: 1000
  min_eval_frequency: 100

Usage

First, check if the model is valid.

python main.py --config check_tiny --mode train

Then, train the model

python main.py --config kor_ballad --mode train_and_evaluate

After training, generate Korean Samhangsi.

python generator.py --config kor_ballad --word 삼행시

Samhangsi Examples

삼이야 그리움이 좇아 사랑은늘 도망가
행른 잊어버리고 그대 이 세상
시제 너의 곁을 떠나면 빗물에 꽃씨하나 흘러가듯

기를 바라보네 두 손 잡고 고개 끄덕여 달라 하기에
계 울고 싶어 내 맘을 떠나가던 날

여도 지금하럼 커피는 날개니
름다웠던 그대모습 다시 볼 수 없는것 알아요

커나가 그래 돌아서 눈 감으면 잊을까
피고 내가 가고 싶지 아파 만날 날 기다려왔어

Example with kino-bot

Reference


report

Recommend

About Joyk


Aggregate valuable and interesting links.
Joyk means Joy of geeK