How long does it take to train for each stage? including the (1) pre-training stage, (2) fine-tuning stage, (3) dataset-specific fine-tuning

[amos-x-wang]

Dear authors,

First of all, great work, and congratulations.

Excited by your great work, I am interested in reproducing the training. From the paper, I understand there are 3 training stages in total, with different input image size, batch size, training steps for different stages. Each stage was trained using 8 GPUs with A100 (80GB) memory.

Could you please tell me how much time was spent in each stage? Thanks very much in advance.

Best regards, Amos