Pre-training time

[soujanyaporia]

Dear Authors,

Great work! May I check a few details?

1. How many GPUs were used to train the pi0-fast model?
2. How long was the pi0-fast model trained for in terms of epochs? I understand that you trained for 700k steps, but I'm not sure about the batch size. I am assuming your total dataset size is 903k timesteps.

Thank you!

Best,
Soujanya