diff --git a/finetune/sequence_label.py b/finetune/sequence_label.py
index dff0f9c531a57758e08ba635478f901baa54fd5d..71610477040e0e6459e90886afc1dbbbac6513c2 100644
--- a/finetune/sequence_label.py
+++ b/finetune/sequence_label.py
@@ -181,11 +181,11 @@ def chunk_predict(np_inputs, np_probs, np_lens, dev_count=1):
             seq_st = base_index + i * max_len + 1
             seq_en = seq_st + (lens[i] - 2)
             prob = probs[seq_st:seq_en, :]
-            infers = np.argmax(probs, -1)
+            infers = np.argmax(prob, -1)
             out.append((
                     inputs[seq_st:seq_en].tolist(), 
                     infers.tolist(),
-                    probs.tolist()))
+                    prob.tolist()))
         base_index += max_len * len(lens)
     return out
 
@@ -199,14 +199,13 @@ def predict(exe,
         graph_vars["inputs"].name,
         graph_vars["probs"].name,
         graph_vars["seqlen"].name,
-        graph_vars["probs"].name,
     ]
 
     test_pyreader.start()
     res = []
     while True:
         try:
-            inputs, probs, np_lens, np_probs = exe.run(program=test_program,
+            inputs, probs, np_lens = exe.run(program=test_program,
                                         fetch_list=fetch_list)
             r = chunk_predict(inputs, probs, np_lens, dev_count)
             res += r