Total mult-adds rendered as MEGABYTES

[bjourne]

MWE:

from torchvision.models import *
summary(AlexNet(), input_size = (1, 3, 227, 227), device = "cpu")
...
Non-trainable params: 0
Total mult-adds (Units.MEGABYTES): 717.27

It writes "(Units.MEGABYTES)" but it should write "(M)" instead. Same for larger models where it writes "Units.GIGABYTES" instead of "G".