In Transformer model, only these layer types are involved in the deep learning/containing trainable parameters, and (3) with activation functions: (1). Word Embedding Layer; (2). Weighted matrices for K, V, Q; (3). Feed Forward Layer or Fully Connected Layer. Correct?
Thanks so much! One question to ask: 3.5 Positional Encoding --- where pos is the position and i is the dimension, i is in the range of 1-512, right? If this is true, how to understand PE(pos,2i)? 2i will be in the range of 2-1024? Thanks