A standard neural network layer in transformers that processes information independently at each position.