Open ucas010 opened 1 year ago
您好,现有是视觉对话模型基本上都有生成幻觉的问题,因此如果使用pandagpt或者其他的视觉对话模型来做文档解读任务都不理想。
图像方面的效果咋样?也有幻觉问题么? 另外答复可以是中文么?我试验的结果还是英文答复。
do you meet this error?: A decoder-only architecture is being used, but right-padding was detected! For correct generation results, please set padding_side='left' when initializing the tokenizer.
can you tell me how to solve this problem?
如题