Proper initialization of nn.Embedding

I’m using nn.Embedding for a decoder. Given that the vocab_size is N, do I need to initialize nn.Embeeding with N+2 to consider and tokens?