如何做批量多模态理解 （how to do batch inference on mutli-modality understanding task）

[rampage-bruce]

你好，我想用Janus做批量多模态理解，但不清楚批量conversation 的格式应该是怎么样的，通过在conversation 增加 dic 会导致有些问题可以回答但有些问题回答是空。

[rampage-bruce]

Hello, I want to use Janus for batch multimodal understanding, but I don't know what the format of the batch conversation should be. By adding dic to the conversation, some questions can be answered, but some questions have empty answers.

[krennic999]

I also wonder if there is batched-inference code for text-based image synthesis