About the training set
#11
by hzf666 - opened
Thanks for great work. I'm confused about the training data. In the paper, it is said that train gme by 8m data. But the in the script, it seems like that you are using TextCaps(https://huggingface.co/datasets/HuggingFaceM4/TextCaps).
It seems they provide an example for fine-tuning the trained gme model on textcaps.