CC3M dataset downloading issue

Hi, I'm using your "utils/download.py" script to download CC3M dataset. I download "Train_GCC-training.tsv" from "https://ai.google.com/research/ConceptualCaptions/" page and there are 3318333 records, but after running your script there are just 1205765 images downloaded, and no error reported. Could you please tell me what's going wrong?