VizSeq is a Python toolkit for visual analysis on text generation tasks like machine translation, summarization, image captioning, speech translation