Text Generator (LSTM + Keras)
The code builds an LSTM to generate text using Keras. Watch this video by Ajay Halthor to see a live demo and understand the intuition behind Recurrent Neural Networks.
Dependencies
- python=3.6.0
- numpy=1.14.0
- nltk=3.2.3
- Tensorflow=1.5.0
- Keras=2.1.3
Files
- notebook: All the code for generating the LSTM
- hdf5 file: saved model. You can use it to continue training for better results.
Execution
Open this project directory in terminal and type
$ jupyter notebook
Recommend install jupyter through anaconda here.
References
- [1] Deep Learning book Ch.10 on Sequence modelling: https://www.deeplearningbook.org/contents/rnn.html
- [2] Colah’s blog on LSTMs: http://colah.github.io/posts/2015-08-Understanding-LSTMs/
- [3] Andrej Karpathy’s Blog + Code (You can probably understand more from this now!): http://karpathy.github.io/2015/05/21/rnn-effectiveness/