Training Detail

[GeneralWhite]

Hi, thank you for your remarkable work! 🤩 🤩

I'd like to ask about the training details of this work: how many GPUs were used for training ？How long did the training take?

Do you have any plans to release the dataset? 😺 😺

Thank you again for your outstanding work!!!

[luoxyhappy]

Thank you for your interest. We trained the model using 32 GPUs for about a week. Sorry, but we’re not able to release the dataset due to privacy concerns.

[GeneralWhite]

Thank you for your reply. I still have a few questions regarding the training details.

How much VRAM does a GPU need for training at a resolution of 720×480? I encountered an OOM error in A100 GPU. Is the one-week training time estimate based on a dataset of 12k samples at a resolution of 720×480? If you don't mind sharing, could you also let me know which GPU you are using?

Thank you again for your awesome work