emb = self.embeddings(src)
out = emb.transpose(0, 1).contiguous()
words = src[:, :, 0].transpose(0, 1)
w_batch, w_len = words.size()
padding_idx = self.embeddings.word_padding_idx
mask = words.data.eq(padding_idx).unsqueeze(1) // [B, 1, T]
// Run the forward pass of every layer of the tranformer.
After Change
emb = self.embeddings(src)
out = emb.transpose(0, 1).contiguous()
mask = ~sequence_mask(lengths).unsqueeze(1)
// Run the forward pass of every layer of the tranformer.
for layer in self.transformer:
out = layer(out, mask)
out = self.layer_norm(out)
