In the convolutional model, we are using the Keras Functional API, so you have to be careful not to put the commas at the end of the lines. If you do that, it adds another dimension to the tensors. In the happy model case, we needed those commas, because there we were using the Sequential API, and the commas separate the elements of the list of layers.
For more information about how to use the Sequential and Functional APIs, here’s a really useful thread from a while back that goes through both in some detail. Please have a look!