image_hidden_state = hidden_state[start_idx:end_idx-1, :]

[Ryann-Ran]

Hi, thanks for your excellent work!
I would like to ask about the extraction of image_hidden_state. start_idx is the index of <im_start>, and end_idx is the index of <im_end>. According to the code:

image_hidden_state = hidden_state[start_idx:end_idx-1, :]  

Why is image_hidden_state extracted from start_idx to end_idx-2 instead of from start_idx+1 to end_idx-1 (i.e., [start_idx+1:end_idx, :])?