Varying the strides with Convolutional Implementation of Sliding Windows

I watched the lecture “Convolutional Implementation of Sliding Windows” of week 3 and I have some question to ask. In the lecture, all convolution layers strides are set to 1, and only the pooling layer has the stride of 2. As explained we have to use a sliding window with stride 2 for the technique to work. So, what happens when we vary the stride parameters in our network. Say, we use a network at hundreds on convolution + pooling layers. They may have arbitrary strides. Would this technique breaks down or is there a formula to choose the strides for sliding windows in this case?.
Thank you,

The strides are shown as an example. Be sure to perform hyperparameter search to pick the right stride for your specific problem.

A simple way for you to answer your own question would be to create a keras.Sequential model and print(model.summary()) to understand how change in strides impacts the output shape and the layer parameters.