Different numbers of parameters

[Mohanlal-11]

I have written yolov4 from scratch by referencing this repo and while checking parameters of both i.e. mine and also from this repo, it was found that from the architecture of this repo the Trainable params: 103,770,846 when plotting summary using 'torchinfo' and total number of parameters is 66364746 when calculate by using:
total = 0
for par in model.parameters():
total += par.numel()
Why? How?
Also when i checked in my architecture, from both i got same number of parameters. So please explain.