4 ms·MV-GPT: A New Generative Pre-Training Framework for Multimodal Video Captioning1 points by tech-sucker 4y ago