LLaVA-NeXT icon indicating copy to clipboard operation
LLaVA-NeXT copied to clipboard

Published 20 hours ago • verified publisher icon LLaVA-VL

Optimal prompt format for reasoning on various questions.

Open xin-ran-w opened this issue 1 year ago • 1 comments

Hey! I want to use LLaVA-OV to do some inference, I read the paper to find the training prompt for each type of question (Table 18). But the full prompt with default image token "" is not presented.

xin-ran-w avatar Aug 31 '24 02:08 xin-ran-w

The <image> token is always put in front of each question (inlcuding original question + format prompting).

Luodian avatar Sep 03 '24 03:09 Luodian