https://github.com/ekzhang/char-rnn-keras

TensorFlow implementation of multi-layer recurrent neural networks for training and sampling from texts
https://github.com/ekzhang/char-rnn-keras

char-rnn keras lstm recurrent-neural-networks tensorflow

Last synced: about 2 months ago
JSON representation

TensorFlow implementation of multi-layer recurrent neural networks for training and sampling from texts

• Host: GitHub
• URL: https://github.com/ekzhang/char-rnn-keras
• Owner: ekzhang
• License: mit
• Created: 2017-07-28T21:26:04.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2020-12-25T17:29:46.000Z (over 3 years ago)
• Last Synced: 2024-06-20T19:41:00.580Z (3 months ago)
• Topics: char-rnn, keras, lstm, recurrent-neural-networks, tensorflow
• Language: Python
• Homepage:
• Size: 970 KB
• Stars: 43
• Watchers: 7
• Forks: 14
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# char-rnn-keras

Multi-layer recurrent neural networks for training and sampling from texts, inspired by [Andrej Karpathy's article](http://karpathy.github.io/2015/05/21/rnn-effectiveness) and the original torch source code [karpathy/char-rnn](https://github.com/karpathy/char-rnn).

## Requirements

This code is written in Python 3, and it requires the [Keras](https://keras.io) deep learning library.

## Input data

All input data should be placed in the [`data/`](./data) directory. Sample training texts are provided.

## Usage

To train the model with default settings:

```bash

$ python train.py --input tiny-shakespeare.txt

```

To sample the model at epoch 100:

```bash

$ python sample.py 100

```

Training loss/accuracy is stored in `logs/training_log.csv`. Model results, including intermediate model weights during training, are stored in the `model` directory. These are also used by `sample.py` for sampling.