Skip to content

experiments to run

Jump to bottom
esalesky edited this page Mar 19, 2018 · 2 revisions

  • baseline

  • 2 layers for encoder

  • run on bpe'd but not tokenized data

  • linear layer between embeddings and hidden

  • averaging hidden states for init to decoder

  • beam search

  • once beam search is working, run on morphgen-ed data

Clone this wiki locally