/dmu

Deep Memory Update codebase

Primary LanguagePython

Deep Memory Update

This repository contains implementation of the Deep Memory Update architecture and experiments performed in the original paper.

Environment installation

Note : Use Python 3.7 or newer

  • Install poetry
  • Install environment by issuing poetry install in the root of the repository

Running experiments with guild

  • Activate the environment by poetry shell
  • Run the experiment, for example: guild run dmu:order min-length=100 recurrent-cells="2 4" max_epochs=100

Reproducing experiments from the paper

To reproduce a single run of an experiment for all recurrent modules run a shell script from the experiments_repro directory. For example to run NatLang experiment issue ./experiments_repro/nat_lang.sh.