update README & source code

seq2seq/basic_nmt/nmt_without_attention_v2.py

@@ -136,6 +136,7 @@ def train(source_dict_dim, target_dict_dim):
     :param target_dict_dim: size of target dictionary
     :type target_dict_dim: int
     '''
+    # initialize model
     cost = seq2seq_net(source_dict_dim, target_dict_dim)
     parameters = paddle.parameters.create(cost)
 
@@ -180,6 +181,8 @@ def generate(source_dict_dim, target_dict_dim, init_models_path):
     :type source_dict_dim: int
     :param target_dict_dim: size of target dictionary
     :type target_dict_dim: int
+    :param init_models_path: path for inital model
+    :type init_models_path: string
     '''
 
     # load data  samples for generation
@@ -203,8 +206,7 @@ def generate(source_dict_dim, target_dict_dim, init_models_path):
 
     # the delimited element of generated sequences is -1,
     # the first element of each generated sequence is the sequence length
-    seq_list = []
-    seq = []
+    seq_list, seq = [], []
     for w in beam_result[1]:
         if w != -1:
             seq.append(w)