Trump bot generates new speeches in Donald Trump's style using a word-level RNN and pre-trained GloVe word vectors. This repo corresponds to this Eight Portions blog post.
- Pull docker image:
docker pull rtlee/t-bot:sample
- Run docker container:
docker run -t -i rtlee/t-bot:sample /bin/bash
- Update the git repo:
git pull origin master
- Sample from models:
python sample.py "I will build a"
- Pull docker image:
docker pull rtlee/t-bot:train
- Run docker container:
docker run -t -i rtlee/t-bot:train /bin/bash
- Update the git repo:
git pull origin master
- Train models
- Modify training scripts as necessary:
train_char.sh
andtrain_word.sh
- Run training scripts:
./train_char.sh
and./train_word.sh
- Modify training scripts as necessary:
- Identify the best word and character level models and move to
/cv/
, replacing the existing files corresponding to the appropriate model type - Sample from models:
python sample.py "I will build a"
This is a fork of Lars Hiller Eidnes' word-rnn, which is based on Andrej Karpathy's char-rnn.