Sequence to Sequence Attention

this file reproduce the seq2seq attention brought up by this paper NEURAL MACHINE TRANSLATION BY JOINTLY LEARNING TO ALIGN AND TRANSLATE

File

seq2seq_translation.py is a implementation of the Soft-Attention with a standard encoder-decoder model implemented using the GRU

seq2seq_translation_2.py is a modified the version of soft-attention with a slightly different implementation and a multi-layer GRU

Run

pip install torch
pip install numpy
python seq2seq_translation.py

Results

Citation

@article{bahdanau2014neural,
  title={Neural machine translation by jointly learning to align and translate},
  author={Bahdanau, Dzmitry},
  journal={arXiv preprint arXiv:1409.0473},
  year={2014}
}