Training and inference cost to reproduce

[LeeKeyu]

Hi authors, thanks for your impressive work and contribution.

I would like to ask about the training cost of the method. Could you please provide the GPU requirement to train and distill the model using your method, including how many GPUs were used for how much time?

Also, I observed that you stated a GPU of >24GB memory is sufficient for inference in the README, while in this issue you mentioned 80GB is required to generate 65 frames. May I ask about the the max frame rate or frame number for a 24GB GPU?

Thank you in advance for your help.