dataset and ckpt

[Yuansen-Huang]

Thanks for your great work!
I wonder what dataset you used and is there a script to convert to your format?
Besides, for text/image encoder ckpt, can you clarify which ckpt should be chosen from your released link in huggingface?