contains all models trained on SNLI, model name follows '{vocab size}{embed dimension}{interaction}_{model type}.py'
contains jupyter notebooks that use the best logistic and nn models tuned in 3_1 to evaluate on MNLI contains validation loss and accuracy computed by the best two models on MNLI contains saved best logistic and nn model from 3_1
contains the best finetuned model further trained and evaluated on each genre in MNLI
used pre-trained word embedding from FastText to train and evaluate the SNLI and MNLI FastText data set can be found at https://fasttext.cc/ Please download 'wiki-news-300d-1M.vec.zip' to folder 3_4
find similar words in our self-trained word embedding in 2-norm