sumeet-iitg / PyTorch-RTE-baselines

PyTorch implementation of RTE baseline methods

PyTorch re-implementation of MultiNLI baseline methods

Most scripts are from nyu-mll/multiNLI

Setup

Download SNLI and MultiNLI datasets from here and put them in data

data/
  snli_1.0/
    (files from SNLI v1.0)
  multinli_0.9/
    (files from SNLI v0.9)

Download pretrained GloVe vectors and put them in data/glove.840B.300d.txt (840B token, 300 dimensions)

Train

cd src
PYTHONPATH=$PYTHONPATH:. python train_snli.py cbow petModel-0 --emb-train --gpu

or if you use GPU,

cd src
PYTHONPATH=$PYTHONPATH:. python train_snli.py cbow petModel-0 --emb-train