How to create patches of a large n-d array to a small n-d array, for training them in a neural network?

Question

I have a 3-d numpy array of size 2500x1700. I want to convert this large array to small patches, of size 128x128x3. How can I achieve this?

I have looked into many websites, where they share information of using 3D arrays like medium, where the author mentioned using as_strided, you can play with arrays. I also looked into this stackoverflow, but only getting more and more confused.

Please guide me on how to do this task.