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

No speedup and memory saving on CIFAR10 #13

Open
guangzhili opened this Issue Jan 31, 2018 · 8 comments

Comments

Projects
None yet
6 participants
@guangzhili

guangzhili commented Jan 31, 2018

I have played around with CIFAR10 and also done a bit benchmark. It seems BinOp does not have noticeable effect on model size and inference speed compared to NIN model without BinOp. I have tested both on CPU and GPU. I thought the saved model nin.pth.tar would shrink, and the inference would speed up significantly. Do I miss something? Does anyone have this issue? Thanks.

@fenollp

This comment has been minimized.

Show comment
Hide comment
@fenollp

fenollp Jan 31, 2018

It’s because BinOp is still 32 or 16 bits. There is no packing optimization here. Sad.

fenollp commented Jan 31, 2018

It’s because BinOp is still 32 or 16 bits. There is no packing optimization here. Sad.

@guangzhili

This comment has been minimized.

Show comment
Hide comment
@guangzhili

guangzhili Jan 31, 2018

@fenollp Thanks for your reply. What would you suggest to do if I would like to achieve the binary optimization? modify PyTorch core?

guangzhili commented Jan 31, 2018

@fenollp Thanks for your reply. What would you suggest to do if I would like to achieve the binary optimization? modify PyTorch core?

@cow8

This comment has been minimized.

Show comment
Hide comment
@cow8

cow8 Feb 2, 2018

@guangzhili useful discussion here here
And a GPU kernel based on TensorFlow can be found here
But unfortunately, the acceleration is NOT significant. As for compression, I think it's easy to implement.

cow8 commented Feb 2, 2018

@guangzhili useful discussion here here
And a GPU kernel based on TensorFlow can be found here
But unfortunately, the acceleration is NOT significant. As for compression, I think it's easy to implement.

@jiecaoyu

This comment has been minimized.

Show comment
Hide comment
@jiecaoyu

jiecaoyu Feb 2, 2018

Owner

@guangzhili An XNOR operation kernel is required to get an acceleration.

I am implementing the kernels as part of my research project. I will release the code after I get the paper published somewhere.

Owner

jiecaoyu commented Feb 2, 2018

@guangzhili An XNOR operation kernel is required to get an acceleration.

I am implementing the kernels as part of my research project. I will release the code after I get the paper published somewhere.

@guangzhili

This comment has been minimized.

Show comment
Hide comment
@guangzhili

guangzhili Feb 5, 2018

Thank you @cow8 That's very helpful!

guangzhili commented Feb 5, 2018

Thank you @cow8 That's very helpful!

@guangzhili

This comment has been minimized.

Show comment
Hide comment
@guangzhili

guangzhili Feb 5, 2018

@jiecaoyu Sounds great. Good luck with the paper.

guangzhili commented Feb 5, 2018

@jiecaoyu Sounds great. Good luck with the paper.

@mjczyt

This comment has been minimized.

Show comment
Hide comment
@mjczyt

mjczyt Feb 28, 2018

@guangzhili look forward to see your paper ,good luck

mjczyt commented Feb 28, 2018

@guangzhili look forward to see your paper ,good luck

@Paul0629

This comment has been minimized.

Show comment
Hide comment
@Paul0629

Paul0629 Jun 28, 2018

My NIN model without BinOp is 946.7k while model with BinOp is 3.9M. That's weird.@jiecaoyu

Paul0629 commented Jun 28, 2018

My NIN model without BinOp is 946.7k while model with BinOp is 3.9M. That's weird.@jiecaoyu

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment