mertyg / vision-language-models-are-bows

Experiments and data for the paper "When and why vision-language models behave like bags-of-words, and what to do about it?" Oral @ ICLR 2023
MIT License
222 stars 14 forks source link

Where to find the training data of NegCLIP? #21

Closed Wyattwwwww closed 1 year ago

Wyattwwwww commented 1 year ago

I tried to find the crucial training data for NegCLIP but I found them nowhere. Seems they are just some csv files, but nowhere mentions the downloading link. Are they hided some where or, can you offer them? Thanks!

vinid commented 1 year ago

Hello!!

See https://github.com/mertyg/vision-language-models-are-bows/issues/3

Wyattwwwww commented 1 year ago

Oh great! I forget to check closed issues, thanks!