Recurrent Neural Networks: building GRU cells VS LSTM cells in Pytorch

What are the advantages of RNN’s over transformers? When to use GRU’s over LSTM? What are the equations of GRU really mean? How to build a GRU cell in Pytorch?