Closed CoolChameleon closed 7 months ago
感谢你的关注~在代码中,我们将prompt设为f'Please detect the {g_dino_caption} in this image.'
,但对于VQA任务,prompt和caption应该都被赋值为你的Question,下面是一个例子:
image:
g_dino_caption: What is the young boy doing at the table? prompt: What is the young boy doing at the table? response: The young boy is sitting at the table, eating pizza and smiling. 可以修改一下prompt的赋值再试试~对于非检测任务,模型一般不会返回[DET] token,不过非常遗憾 由于训练数据中VQA比例较低,它的VQA性能并没有特别好,在后续的工作中 我们也会重点优化这方面~
Hi, hope your question has been resolved. If you have any further concerns, please feel free to re-open it or open a new issue. Thanks!
您好: 使用您提供的chat.py进行推理时,无论怎样修改prompt和caption,模型的文字输出都为类似Sure,.的格式,无法输出额外的文字信息(即使没有要求模型进行检测,只要求模型描述图片也会有这样的情况)。请问这是符合预期的吗?