Skip to content

Failed to reproduce the CIFAR10 result #5

Description

@cmqrrwl

Hi, I run the code on vgg16 and resnet56/CIFAR10, but failed to reproduce the result in your paper, I use the provided baseline checkpoints, and failed to achieve your paper result.

For vgg16, I run 10 times , but none of the results of your paper were reproduced. The best I got was 92.73 after pruning and retraining 30 epochs, while in your paper the result is 93.12.

For resnet56, I also failed to reproduce the result in your paper. I get 92.39 on method-D after 100 epochs training, but your paper result is 94.00. Meanwhile, I get 91.58 on method-P after pruning and retraining 30 epochs, but your paper result is 93.38. I set all the parameters as in your paper.

For resnet18, I didn’t run your code. The error is caused by dimension mismatch of residual layer. I tried to modify the residual in shrink_resnet code and replace down_SAMPLE with short_cut in the code, but still failed.

I sincerely hope you can give me some help!
Thank you very much!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions