<S> is begin, </S> is end, </S> is used, but why don’t you use <S>? about transformer

[RayXu14]

<S> is used. In class Graph,

            # define decoder inputs
            self.decoder_inputs = tf.concat((tf.ones_like(self.y[:, :1])*2, self.y[:, :-1]), -1) # 2:<S>

[Ningshiqi]

is used. In class Graph,

            # define decoder inputs
            self.decoder_inputs = tf.concat((tf.ones_like(self.y[:, :1])*2, self.y[:, :-1]), -1) # 2:<S>

y is considered the ，but the x is not considered .I didn't see the place where x is add the ""

[gitfourteen]

in prepro.py, we define the first four token and their counts:

<PAD>   1000000000
<UNK>   1000000000
<S>     1000000000
</S>    1000000000

in data_load.py, we get the token indices:

<PAD>   0
<UNK>   1
<S>     2
</S>    3