akamaster

True, this needs to be corrected. Are you willing to make a contribution?

Yeah, recently I have had these comments. For example see #18, [and specifically my comment within](https://github.com/akamaster/pytorch_resnet_cifar10/issues/18#issuecomment-605314203) Essentially, the issue is in PyTorch version, starting from some (1.3?) version something has...

Thanks a lot, and what is the accuracy of this model? Is is 7.97% as you report in paper? Or it is something else, and in paper plot was for...

What is the lambda value you used to obtain that result? I don't see any particular values in any of the scripts.

It's since the resnet implementation here does not correspond to actual in the paper. Take a look at [this repo.](https://github.com/akamaster/pytorch_resnet_cifar10)

[This implementation](https://github.com/akamaster/pytorch_resnet_cifar10) for pytorch does correspond to the actual paper.

This repo does not correspond to orginal paper in terms of configuration. Take a look at different one, e.g. [this](https://github.com/akamaster/pytorch_resnet_cifar10)

Ooops, added wrong "compression" label, it should be "inference".