Rohit Gupta

_get_aug_range_idxes fails if length of tokens is less than or equal to aug_cnt, which primarily shows up in cases where sequences are too short. fixing this by adding a check

Currently the code in git uses pre-computed tags, use Stanford POS tagger to complete that https://nlp.stanford.edu/software/tagger.shtml#Download

It seems like the model checkpoints listed in the README are not publicly accessible, I get the following message when I navigate to that link: > Additional permissions required to...

This is a great repository, your implementation of LLaVA-NeXT is top notch, if you add a LICENSE that would help us use this more confidently. Thanks !