gradient-based-inference

Code for "Lee, J. Y., Mehta, S. V., Wick, M., Tristan, J. B., & Carbonell, J. (2019, July). Gradient-based inference for networks with output constraints. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 33, pp. 4147-4154)."

Requirements

Python 3.6, PyTorch 0.4.1, AllenNLP v0.4.1

Setting up a virtual environment

Conda can be used to set up a virtual environment with Python 3.6 in which you can sandbox dependencies required for our implementation:

1. Download and install Conda.

2. Create a Conda environment with Python 3.6

   conda create -n gbi python=3.6
   

3. Activate the Conda environment. (You will need to activate the Conda environment in each terminal in which you want to run our implementation).

   source activate gbi
   

Setting up our environment

1. Visit http://pytorch.org/ and install the PyTorch 0.4.1 package for your system.

2. Clone our repo:

   git clone git@github.com:sanketvmehta/gradient-based-inference.git
   

Installing AllenNLP from source

1. Clone allennlp with git submodule

   git submodule update --init
   

2. Checkout allennlp to v0.4.1

   git checkout 31f4f60
   

3. Change your directory to allennlp submodule present under the parent repo directory:

   cd gradient-based-inference/allennlp
   

4. Install the necessary requirement by running

   INSTALL_TEST_REQUIREMENTS=true scripts/install_requirements.sh
   

5. Once the requirements have been installed, run:

   pip install --editable .
   

6. Test AllenNLP installation by running:

   ./scripts/verify.py
   

That's it! You're now ready to reproduce our results.

Citing

If you use our code in your research, please cite: Gradient-based inference for networks with output constraints.

@inproceedings{lee2019gradient,
title={Gradient-based inference for networks with output constraints},
author={Lee, Jay Yoon and Mehta, Sanket Vaibhav and Wick, Michael and Tristan, Jean-Baptiste and Carbonell, Jaime},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={33},
pages={4147--4154},
year={2019}
}