Usage:
python3 train.py Path of the folder is files/models/
Input format:
Each caption (with start and end tokens) is converted into a one-hot matrix of shape MAX_SEQUENCE_LENGTH x VOCAB_SIZE. Output is this matrix rotated one row upwards, so first word has second word corresponding to it.