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

261 stars 15 forks source link

Open Vicent0205 opened 5 months ago

Vicent0205 commented 5 months ago

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