/MemN2NandMultObs

Primary LanguagePythonMIT LicenseMIT

Memory Network and Multiple Observer Models

Memory network model originally developed by Facebook. Code base was adapted from domluna's implementation with major contributions from eringrant.

This is part of a series of repositories for our 2018 EMNLP Paper. Please see the original repository for information on the dataset.

Getting Started

Install all of the requirements.

  • Tensorflow v1.0 or greater.
  • Pandas
  • Matplotlib
  • Numpy
  • Parallel: (wget -O - pi.dk/3 || curl pi.dk/3/ || fetch -o - http://pi.dk/3) | bash

Running Experiments

./run_tasks.sh

This will create a folder inside of results/ with all of the results, which are processed with the step below.

Running Analysis

python tom_experiments.py --result_files results/${RESULT_FNAME}