Does `torchsummary`'s method of calculating model size require more memory than during training the same model?

Question

I am translating a tensorflow/keras model into torch. As a sanity check I am attempting to check each layer\'s input/output shape along wi