JSON format in pre-training stage for Emu3

[leofan90]

Hello! Thanks for your fantastic open-source work! It's exciting. I have a few questions about constructing our data format for fine-tuning. Can you describe the JSON format of your pre-trained and fine-tuned dataset? It would be great if you could demonstrate a demo of your pre-trained data format and dataset preprocessing code. Thanks for your time and looking forward for you reply!