fix bugs about padding_idx

dygraph/seq2seq/attention_model.py

@@ -81,6 +81,7 @@ class Encoder(Layer):
         self.hidden_size = hidden_size
         self.num_layers = num_layers
         self.dtype = dtype
+        self.padding_idx = padding_idx
         self.embedder = Embedding(
             vocab_size,
             hidden_size,
@@ -101,7 +102,7 @@ class Encoder(Layer):
         outs, (final_h, final_c) = self.lstm(
             src_emb, sequence_length=src_sequence_length)
 
-        enc_len_mask = (src != 2).astype(self.dtype)
+        enc_len_mask = (src != self.padding_idx).astype(self.dtype)
         enc_padding_mask = (enc_len_mask - 1.0) * 1e9
         return [final_h, final_c], outs, enc_padding_mask
 
@@ -200,6 +201,7 @@ class Decoder(Layer):
         self.num_layers = num_layers
         self.init_scale = init_scale
         self.dtype = dtype
+        self.padding_idx = padding_idx
         self.embedder = Embedding(
             vocab_size,
             hidden_size,
@@ -241,7 +243,7 @@ class Decoder(Layer):
             logits=dec_output, label=label, soft_label=False)
         loss = paddle.squeeze(loss, axis=[2])
 
-        trg_mask = (trg != 2).astype(self.dtype)
+        trg_mask = (trg != self.padding_idx).astype(self.dtype)
 
         loss = loss * trg_mask
         loss = paddle.reduce_mean(loss, dim=[0])

dygraph/seq2seq/data

-/models/dygraph/seq2seq_attn/data
\ No newline at end of file