Why is src_len+1 in Transformer demo?

self.pos_emb = nn.Embedding.from_pretrained(get_sinusoid_encoding_table(src_len+1, d_model),freeze=True)

The position encoding table should be (max_len, d_model), why add 1?

Locate the Transformer demo containing the nn.Embedding.from_pretrained line and inspect get_sinusoid_encoding_table(src_len+1, d_model). Check how source positions are indexed and whether the extra position is required. Done means documenting a clear explanation of why src_len+1 is used, or confirming that the call should use src_len.

Scritto dal modello di indicizzazione a partire dal testo della issue.