COOT: Cooperative Hierarchical Transformer for Video-Text Representation Learning - View it on GitHub