预处理错误

There are doubts in the preprocessing process. Is the spatial language model in the code the four datasets of train, test, val, and split in the WMT16 multimodal translation task, and what code command line are these data sets imported through preprocess.py? Also, the running result of my preprocess.py has errors. If you can solve my confusion, I would be grateful!