Before you can run the code in this repo, you need to download the SNLI data from here. Additionally, you will need to alter run.sh to include the data_path argument to where you have saved the data.

This code uses one of the daily binary tensorflow releases, which you can download here. There's no reason why it won't work with the standard release, you might just have to do a bit of refactoring.

This code implements a couple of papers:

1. Alternating Iterative Neural Attention for Machine Reading
2. A Decomposable Attention model for Natural Language Inference
3. Adaptive Computation Time for Recurrent Neural Networks

These are then combined to make a couple of different models for Adaptive Natural Language Inference.

1. ACTDAModel.py : Decomposable Attention + ACT + Iterative Neural Attention inference module
2. AdaptiveIAAModel.py GRU Sentence encoders + ACT + Iterative Neural Attention inference module
3. DAModel.py : Decomposable Attention model
4. IAAModel.py : Iterative Neural Attention with fixed inference GRU steps