Does NegCLIP now support mulit-gpu and grad_accumulation？

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

Open Vicent0205 opened 1 month ago

Vicent0205 commented 1 month ago

Hi! Thanks for your great work! Does NegCLIP now support mulit-gpu and grad_accumulation？