diff --git a/fluid/machine_reading_comprehesion/DuReader/args.py b/fluid/machine_reading_comprehesion/DuReader/args.py
index 8b86193a9002048e7e42b392592210f88a05ff04..96422f4460cf0570d410c145f36e5074c9fc1013 100644
--- a/fluid/machine_reading_comprehesion/DuReader/args.py
+++ b/fluid/machine_reading_comprehesion/DuReader/args.py
@@ -98,7 +98,7 @@ def parse_args():
         "(default: %(default)d)")
     parser.add_argument(
         '--optim', 
-        default='rprop', 
+        default='adam', 
         help='optimizer type')
     parser.add_argument(
         '--trainset',
diff --git a/fluid/machine_reading_comprehesion/DuReader/rc_model.py b/fluid/machine_reading_comprehesion/DuReader/rc_model.py
index 0c0f424ed01adcb313e339bffa8ca46af6febc23..11d5b5d91d734a82d687a09b587ad614d0f03fff 100644
--- a/fluid/machine_reading_comprehesion/DuReader/rc_model.py
+++ b/fluid/machine_reading_comprehesion/DuReader/rc_model.py
@@ -75,6 +75,7 @@ def encoder(input_name, para_name, shape, hidden_size, args):
         input=input_ids,
         size=shape,
         dtype='float32',
+        is_sparse=True,
         param_attr=fluid.ParamAttr(name='embedding_para'))
 
     encoder_out = bi_lstm_encoder(