question about image2text training？ · Issue #3 · AILab-CVC/SEED · GitHub