Regarding the calculation of FLOPs after model compression

[mjw123bs]

Hello, when I used the FLOPs calculation method in the link https://github.com/Eric-mingjie/rethinking-network-pruning/blob/master/imagenet/l1-norm-pruning/compute_flops.py you sent to calculate the FLOPs of the compressed model, the following error was reported: Traceback (most recent call last): File "denseprune.py", line 287, in <module> print(str(count_model_param_flops(newmodel))+"\n") File "E:\PycharmProjects\network-slimming-master(modified)\compute_flops.py", line 112, in count_model_param_flops out = model(input) File "D:\Application\anaconda3\envs\learn_pytorch\lib\site-packages\torch\nn\modules\module.py", line 1051, in _call_impl return forward_call(*input, **kwargs) File "E:\PycharmProjects\network-slimming-master(modified)\models\densenet.py", line 130, in forward File "D:\Application\anaconda3\envs\learn_pytorch\lib\site-packages\torch\nn\modules\module.py", line 1051, in _call_impl return forward_call(*input, **kwargs) File "D:\Application\anaconda3\envs\learn_pytorch\lib\site-packages\torch\nn\modules\container.py", line 139, in forward input = module(input) File "D:\Application\anaconda3\envs\learn_pytorch\lib\site-packages\torch\nn\modules\module.py", line 1051, in _call_impl return forward_call(*input, **kwargs) File "E:\PycharmProjects\network-slimming-master(modified)\models\densenet.py", line 30, in forward File "D:\Application\anaconda3\envs\learn_pytorch\lib\site-packages\torch\nn\modules\module.py", line 1071, in _call_impl result = forward_call(*input, **kwargs) File "D:\Application\anaconda3\envs\learn_pytorch\lib\site-packages\torch\nn\modules\conv.py", line 443, in forward return self._conv_forward(input, self.weight, self.bias) File "D:\Application\anaconda3\envs\learn_pytorch\lib\site-packages\torch\nn\modules\conv.py", line 439, in _conv_forward return F.conv2d(input, weight, bias, self.stride, RuntimeError: Given groups=1, weight of size [12, 15, 3, 3], expected input[1, 24, 227, 227] to have 15 channels, but got 24 channels instead Please tell me how to solve it? @Eric-mingjie

[mjw123bs]

The above is only for the resnet and densenet models. After the model is compressed, only the input size of Conv2d is changed, but the size of BatchNorm2d is not changed, and the size does not correspond, so an error is reported, as shown in the figure. But I saw that you wrote If the next layer is the channel selection layer, then the current batch normalization layer won't be pruned. in the pruning files of the two models, so please see how to solve this problem.