A reminding of Fractional Max-Pooling

1

In Ben Graham’s paper, the output_ratio is round to the nearest integer but the code in pytorch using floor to the lower bound integer.

2

yes, you are correct about this detail.