Bug. Bad implement.

[aiboys]

I found a bug in cifa-resnet.

I changed Avgpool2d(8,stride=1) to AdaptiveAvgPool2d(1) that is actually correct.

Besides, Thx for your contribution of opening resource.I am going to use this repo to build my own distillation algorithm. But to be honestly, it seems that there are some '' so big'' bugs in this repo. So chould you plz tell me if all your experiments are done correctly by using this repo? I am so confused about that.

[yoshitomo-matsubara]

Hi @aiboys ,

I am not sure why you consider Avgpool2d(8, stride=1) as a bug and bad implementation since the implementation is based the official implementation of ResNet for CIFAR datasets as given here

Could you also provide more info about where you find "so big" bugs in this repo? I used this repo to perform all the experiments I reported. For those reported in my torchdistill paper, I also provided log and config files for the results.

[aiboys]

Oh, I am sry. I checked my configuration file in details. I resized dataset into (224,224) which is the same as ImageNet. Since the downsample rate is 8, the avgpooling layer does not match. I also checked your CIFAR configurations where the params of Resize is 32. | | 罗杨 | | ***@***.*** | 签名由网易邮箱大师定制 On 1/11/2022 12:12，Yoshitomo ***@***.***> wrote： Hi @aiboys , I am not sure why you consider Avgpool2d(8, stride=1) as a bag and bad implementation since the implementation is based the official implementation of ResNet for CIFAR datasets as given here Could you also provide more info about where you find "so big" bugs in this repo? I used this repo to perform all the experiments I reported. For those reported in my torchdistill paper, I also provided log and config files for the results. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>