problem of the attack model

[Huang-yihao]

I follow the instruction on attacking the imagenet images and meet the problem of the model.
The command is python attack.py --dataset="imagenet" --decoder=3 --model="imagenet_denoise"
The problem is as follows.

Could you please tell me what is the error? It seems like there is something wrong on the loaded model, but I am not sure.
Thank you for your help!!!

[qiulingxu]

Hello Yihao, It seems that somehow data is in the format NCHW. However the operation seems to function in the format of NHWC. Could you please check the format of your input. Or instead you can try running on the GPU to see whether it works. Best, Qiuling.

…

On Mon, Oct 24, 2022 at 1:55 AM Yihao Huang ***@***.***> wrote: I follow the instruction on attacking the imagenet images and meet the problem of the model. The command is python attack.py --dataset="imagenet" --decoder=3 --model="imagenet_denoise" The problem is as follows. [image: image] <https://user-images.githubusercontent.com/16575311/197455401-4823a15b-1dd0-43b4-983d-fafe99bd5621.png> Could you please tell me what is the error? It seems like there is something wrong on the loaded model, but I am not sure. Thank you for your help!!! — Reply to this email directly, view it on GitHub <#3>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGGKSFSD2Q424OQZF6RTDPTWEYQDNANCNFSM6AAAAAARMURL3Y> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Huang-yihao]

Hello qiuling,

     The images are of ImageNet type as follows.

    Furthermore, I am just using the GPU, not CPU. But the error suggests using CPU.
     I am wondering which code should I modify to change the data format into NHWC?
     Thank you very much!

[qiulingxu]

Could you please verify the version of your TensorFlow, is it 1.15 with GPU compatibility? I can not reproduce this error on my side. Meanwhile, if you use your model, please ensure the input of the model follows NHWC.

[Huang-yihao]

Thank you! I find that it is the mismatch of cuda and tensorflow, which makes me think I am using the correct tensorflow version with GPU.