Train autoencoder without looping

Question

I have some data that is of shape 10000 x 1440 x 8 where 10000 is the number of days, 1440 the number of minutes and 8 is the number of features.