mlvlab/Flipped-VQA

How many GPUs are needed to train the model?

Closed this issue · 2 comments

jianhua2022 commented
How many GPUs are needed to train the model?
ikodoh commented

I ran the experiments on 4 $\times$ RTX A6000 GPUs for NExT-QA, DramaQA, STAR, and VLEP, and the experiments took about 3 ~ 5 hours.
For TVQA, we ran about 2 days on 8 $\times$ RTX A6000 GPUs.

jianhua2022 commented

I ran the experiments on 4 × RTX A6000 GPUs for NExT-QA, DramaQA, STAR, and VLEP, and the experiments took about 3 ~ 5 hours. For TVQA, we ran about 2 days on 8 × RTX A6000 GPUs.

Thanks for your helpful reply.