textual entailment with structural attentions and composition
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
sampledata
LICENSE
README.md
ReplicateAs.lua
fastbinarytreelstm.lua
model_entailment.lua
node_alignment.lua
simpleprofiler.lua
snli.lua
trainer.lua
tree.lua
utils.lua
wordembedding.lua

README.md

Textual Entailment with Structured Attentions and Composition

License: MIT

This repository contains an implementation of the structured attention and compositon model for textual entailment described in the paper Textual Entailment with Structured Attentions and Composition.

Required Dependencies

  1. Torch7
  2. Torch rnn library
  3. Penlight

Training

To train on the provided sample data and saving, you can simply run:

th trainer.lua --dump model_file

You can find the training parameters and their descriptions in file trainer.lua.

Evaluating

To evaluate the trained model on the dev set, you can run:

th trainer.lua --eval model_file