diff --git a/fluid/PaddleRec/tagspace/infer.py b/fluid/PaddleRec/tagspace/infer.py
index 252bea80f3c94f8286069e7a4fd8e2a154d83df1..430aa769dab50f8432fe847de01904f545faf1e5 100644
--- a/fluid/PaddleRec/tagspace/infer.py
+++ b/fluid/PaddleRec/tagspace/infer.py
@@ -34,7 +34,7 @@ def parse_args():
     return args
 
 
-def infer(test_reader, vocab_tag, use_cuda, model_path):
+def infer(test_reader, vocab_tag, use_cuda, model_path, epoch):
     """ inference function """
     place = fluid.CUDAPlace(0) if use_cuda else fluid.CPUPlace()
     exe = fluid.Executor(place)
@@ -48,6 +48,7 @@ def infer(test_reader, vocab_tag, use_cuda, model_path):
         all_num = 0
         size = vocab_tag
         value = []
+        print("epoch " + str(epoch) + " start")
         for data in test_reader():
             step_id += 1
             lod_text_seq = utils.to_lodtensor([dat[0] for dat in data], place)
@@ -65,8 +66,7 @@ def infer(test_reader, vocab_tag, use_cuda, model_path):
                 if value.index(max(value)) == int(true_pos):
                     true_num += 1
                 value = []
-            if step_id % 1000 == 0:
-                print(step_id, 1.0 * true_num / all_num)
+        print("epoch:" + str(epoch) + "\tacc:" + str(1.0 * true_num / all_num))
         t1 = time.time()
 
 
@@ -95,4 +95,5 @@ if __name__ == "__main__":
             test_reader=test_reader,
             vocab_tag=vocab_tag,
             use_cuda=False,
-            model_path=epoch_path)
+            model_path=epoch_path,
+            epoch=epoch)