Fix decoding sample (#752)

* fix dcoding sample

* fix

Author: FrostML

paddlenlp/ops/faster_transformer/sample/decoding_sample.py

@@ -57,7 +57,8 @@ def do_predict(args):
         src_vocab_size=args.src_vocab_size,
         trg_vocab_size=args.trg_vocab_size,
         max_length=args.max_length + 1,
-        n_layer=args.n_layer,
+        num_encoder_layers=args.n_layer,
+        num_decoder_layers=args.n_layer,
         n_head=args.n_head,
         d_model=args.d_model,
         d_inner_hid=args.d_inner_hid,

paddlenlp/ops/faster_transformer/transformer/decoding.py

@@ -196,7 +196,7 @@ def __init__(self,
                  word_embedding,
                  positional_embedding,
                  linear,
-                 n_layer,
+                 num_decoder_layers,
                  n_head,
                  d_model,
                  bos_id=0,
@@ -376,7 +376,7 @@ def forward(self, enc_output, memory_seq_lens):
             self.linear_weight, self.linear_bias, self.pos_emb,
             self._decoding_strategy, self._beam_size, self._topk, self._topp,
             self._n_head,
-            int(self._d_model / self._n_head), self._n_layer, self._bos_id,
+            int(self._d_model / self._n_head), self._num_decoder_layers, self._bos_id,
             self._eos_id, self._max_out_len, self._beam_search_diversity_rate)
 
         ids = finalize(

paddlenlp/ops/faster_transformer/transformer/faster_transformer.py

@@ -89,7 +89,6 @@ def __init__(self,
             word_embedding=self.trg_word_embedding.word_embedding,
             positional_embedding=self.trg_pos_embedding.pos_encoder,
             linear=self.decoding_linear,
-            num_encoder_layers=num_encoder_layers,
             num_decoder_layers=num_decoder_layers,
             n_head=n_head,
             d_model=d_model,