mbartoli / docker-char-rnn

Docker container for use with Multi-layer Recurrent Neural Networks (LSTM, GRU, RNN) for character-level language models in Torch

docker-char-nn

Docker container for use with Multi-layer Recurrent Neural Networks (LSTM, GRU, RNN) for character-level language models in Torch

Usage

Run docker run -it mbartoli/char-rnn

Persistent checkpoint

1. Create a folder cv to persist training data with mkdir cv.
2. Mount the folder into the container and run it:
   docker run -v $(pwd)/cv:/home/char-rnn/cv -it mbartoli/char-rnn
3. Train your char-rnn

Custom training data

1. Create a folder containing some training data.
   mkdir -p data/my-training-data
2. Run the container with the new training data and cv folder mounted
   docker run -v $(pwd)/cv:/home/char-rnn/cv -v $(pwd)/data/my-training-data:/home/char-rnn/data/my-training-data -it mbartoli/char-rnn
3. Train your char-rnn

Training and sampling

See the documentation on how to train and sample your char-rnn.

More Information

Docker Hub: mbartoli/char-nn
https://github.com/karpathy/char-rnn
http://karpathy.github.io/2015/05/21/rnn-effectiveness/