search

zhoubenjia / GFSLT-VLP

MIT License
43 stars 7 forks source link

How many GPUs are required for this task? #6

Closed xxxiaosong closed 12 months ago

xxxiaosong commented 1 year ago

How many GPUs are required for this task?

zhoubenjia commented 1 year ago

How many GPUs are required for this task?

Thank you very much for your attention to our work. In our experimental setup, we harnessed the computational power of 4 Nvidia RTX 3090 GPUs. In theory, a greater number of GPUs can be advantageous, as our approach relies on a symmetric cross-entropy loss, akin to the one used in CLIP. The increased computational resources can potentially enhance the model's performance further.

xxxiaosong commented 12 months ago

Thanks for your reply! And how much time does the training process consume for each dataset....

zhoubenjia commented 12 months ago

Thanks for your reply! And how much time does the training process consume for each dataset....

Phase 1 (80 epoch) lasts about 24 hours, phase 2 (200 epoch) lasts about 36 hours. Hope this helps.

xxxiaosong commented 12 months ago

Thks, this helps a lot!