Code for ICLR2018 paper "Learning Approximate Inference Networks for Structured Prediction".

The code is written in python2.7 and requies Theano0.8.

The folder includes the code for two tasks: Multi-Label Classification and Sequence Labeling.

Learning Approximate Inference Networks for Structured Prediction (https://arxiv.org/abs/1803.03376)

@inproceedings{tu-18,
  author={Lifu Tu and Kevin Gimpel},
  title={Learning Approximate Inference Networks for Structured Prediction},
  booktitle={Proceedings of International Conference on Learning Representations (ICLR)},
  year={2018}
}

• GPU and CUDA 8 are required
• Theano 0.8
• lasagne
• torchfile
• numpy