How does the model learn the weights and biases in a Transformer NN architecture?

Question

In the transformer block between Input Embedding Matrix and Layer normalization, where the data is scaled to mean = 0, and std = 1. How does the NN learn weights and biases and applying them to the data?