Editable neural networks

A supplementary code for Editable Neural Networks, an ICLR 2020 submission by Anton Sinitsin, Vsevolod Plokhotnyuk, Dmitry Pyrkin, Sergei Popov, Artem Babenko.

What does it do?

It trains a model so that it can later be edited: forced to predict a specific class on a specific input without losing accuracy.

What do i need to run it?

  • A machine with some CPU (preferably 2+ free cores) and GPU(s)
    • Running without GPU is possible but does not scale well, especially for ImageNet
  • Some popular Linux x64 distribution
    • Tested on Ubuntu16.04, should work fine on any popular linux64 and even MacOS;
    • Windows and x32 systems may require heavy wizardry to run;
    • When in doubt, use Docker, preferably GPU-enabled (i.e. nvidia-docker)

How do I run it?

  1. Clone or download this repo. cd yourself to it's root directory.
  2. Grab or build a working python enviromnent. Anaconda works fine.
  3. Install packages from requirements.txt
  4. Run jupyter notebook and open a notebook in ./notebooks/
  • Before you run the first cell, change %env CUDA_VISIBLE_DEVICES=# to an index that you plan to use.
  • CIFAR10 notebook can be ran with no extra preparation
  • The ImageNet notebooks require a step-by-step procedure to get running:
    1. Download the dataset first. See this page or just google it. No, really, go google it!
    2. Run imagenet_preprocess_logits.ipynb
    3. Train with imagenet_editable_training.ipynb
    4. Evaluate by using one of the two remaining notebooks.
  • To reproduce machine translation experiments, follow the instructions in ./mt/README.md