Grounded Latent Trees for Visual Question Answering

This repository contains the code associated with the paper Latent Compositional Representations Improve Systematic Generalization in Grounded Question Answering, published in TACL (2020).

Setup

1. Install all requirements

pip install -r requirements.txt

Arithmetic Dataset

2. Run with this AllenNLP command:

allennlp train train_configs/glt_arithmetic.jsonnet -s experiments/experiment_name --include-package src

CLEVR / CLOSURE Dataset

1. Download CLEVR image features + preprocessed data (image features are extracted with py-bottom-up-attention) into the data directory.

2. Run with this AllenNLP command:

allennlp train train_configs/glt_clevr.jsonnet -s experiments/experiment_name --include-package src