Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

FLOPs mismatch across different frameworks. #31

Open
mu-cai opened this issue May 26, 2020 · 2 comments
Open

FLOPs mismatch across different frameworks. #31

mu-cai opened this issue May 26, 2020 · 2 comments

Comments

@mu-cai
Copy link

mu-cai commented May 26, 2020

Hi @WongKinYiu

Thanks for your answer to my previous question. However, I found severe FLOPs mismatch in your paper and darknet paper @AlexeyAB .

For example, as for ResNet50, the original paper states that the FLOPs is around 3.8 to 4.0 G FLOPs. However, in your paper, it is 9.74 BFLOPs. So there is a big difference. I found maybe there is a ratio of 2.59 around them, right?

@WongKinYiu
Copy link
Owner

WongKinYiu commented May 26, 2020

hello, in original paper, they use mul-add with 224x224.
real FLOPs usually will be two times as mul-add.

darknet use real FLOPs with 256x256.
so FLOPs should be 2*(256*256)/(224*224) = 2.61 times when compare to mul-add with 224x224.

@mu-cai
Copy link
Author

mu-cai commented May 26, 2020

Thanks for your answer! @WongKinYiu
Your answers are always nice and accurate!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants