This repository contains code for my thesis Predicting Haskell Type Signatures From Names. The structured prediction model is in master and the unstructured prediction model, as the name suggests, is in the unstructured_prediction branch.
The dataset directory contains the datset we use: train_simple_sigs_parsable_normalized.txt, dev_simple_sigs_parsable_normalized.txt, test_simple_sigs_parsable_normalized.txt are for training, validation, and testing, respectively.
- pytorch 0.4.0
- python 3.5+
python train.py --train_data=dataset/train_simple_sigs_parsable_normalized.txt --dev_data=dev_simple_sigs_parsable_normalized.txt --test_simple_sigs_parsable_normalized.txt