[BUG] about attention mask

[zyk516]

hello, i have a question. Could you help me ?

https://github.com/FlagOpen/FlagEmbedding/blob/master/research/visual_bge/visual_bge/modeling.py#L270，this line generate mask by cat mask of img and mask of txt. https://github.com/FlagOpen/FlagEmbedding/blob/master/research/visual_bge/visual_bge/modeling.py#L267,this line generate embedding by cat cls_token, img_embedding, txt_embedding without cls_token. The cat order of mask and embedding is diffrent. Is it right?

Thanks!