/Frogger_AIRationalization_Seq2Seq

AI Rationalization code using a symbolic representation with a seq2seq network instead of a CNN-RNN Autoencoder

Primary LanguageHTML

AI Rationalization Symbolic Representation

Code for AI rationalization using a symbolic game representation passed through a seq2seq instead of a CNN-RNN autoencoder.

Training phase

The code has the following requirements,

  1. Python 3.6
  2. Pytorch 0.4.0
  3. Torchvision 0.2.1
  4. nltk 3.3.0

To train the code run the following command from the root directory.

python train_frogger_seq2seq.py

Test phase

To see the results of all the testing samples when passed through the network run, python sample_v6.py