diff --git a/PaddleCV/PaddleDetection/slim/quantization/README.md b/PaddleCV/PaddleDetection/slim/quantization/README.md
index ad178fd9e12fb6276e5c453fc129b24f285ce698..fe69186aef1896ea12f5a22edbb576e743cec875 100644
--- a/PaddleCV/PaddleDetection/slim/quantization/README.md
+++ b/PaddleCV/PaddleDetection/slim/quantization/README.md
@@ -44,7 +44,6 @@ step1: 设置gpu卡
 export CUDA_VISIBLE_DEVICES=0
 ```
 step2: 开始训练
-```
 使用PaddleDetection提供的配置文件在用8卡进行训练：
 
 ```
diff --git a/PaddleCV/PaddleDetection/slim/quantization/compress.py b/PaddleCV/PaddleDetection/slim/quantization/compress.py
index 7c4dbead703cf0d05a58e233d411acde985c6788..6e3156ac3902de572f37c14989a0b44226e9c8e9 100644
--- a/PaddleCV/PaddleDetection/slim/quantization/compress.py
+++ b/PaddleCV/PaddleDetection/slim/quantization/compress.py
@@ -239,6 +239,7 @@ def main():
         eval_feed_list=test_feed,
         eval_func={'map': eval_func},
         eval_fetch_list=[eval_fetch_list[0]],
+        prune_infer_model=[["image", "im_size"],["multiclass_nms_0.tmp_0"]],
         train_optimizer=None)
     com.config(FLAGS.slim_file)
     com.run()
diff --git a/PaddleCV/PaddleDetection/slim/quantization/freeze.py b/PaddleCV/PaddleDetection/slim/quantization/freeze.py
index 04271a4f2c691782654ce948466438246a4d3f55..85906fa8231f400a839b48647877c4791799351f 100644
--- a/PaddleCV/PaddleDetection/slim/quantization/freeze.py
+++ b/PaddleCV/PaddleDetection/slim/quantization/freeze.py
@@ -142,7 +142,7 @@ def main():
     assert os.path.exists(FLAGS.model_path)
     infer_prog, feed_names, fetch_targets = fluid.io.load_inference_model(
             dirname=FLAGS.model_path, executor=exe,
-            model_filename='__model__',
+            model_filename='__model__.infer',
             params_filename='__params__')
 
     eval_keys = ['bbox', 'gt_box', 'gt_label', 'is_difficult']
diff --git a/PaddleSlim/classification/quantization/compress.py b/PaddleSlim/classification/quantization/compress.py
index 4894684fabf3feed3b0ab179861ec921fd56884b..41b767885a1a0ce435df282126d7d78c33a66014 100644
--- a/PaddleSlim/classification/quantization/compress.py
+++ b/PaddleSlim/classification/quantization/compress.py
@@ -95,6 +95,7 @@ def compress(args):
         eval_fetch_list=val_fetch_list,
         teacher_programs=[],
         train_optimizer=opt,
+        prune_infer_model=[[image.name], [out.name]],
         distiller_optimizer=None)
     com_pass.config(args.config_file)
     com_pass.run()
diff --git a/PaddleSlim/classification/quantization/freeze.py b/PaddleSlim/classification/quantization/freeze.py
index 396875c7e1a54b56b5c985952ab4630c37824fc3..6e97e0243510facf477f70a1ea482ba1251e8013 100644
--- a/PaddleSlim/classification/quantization/freeze.py
+++ b/PaddleSlim/classification/quantization/freeze.py
@@ -53,20 +53,32 @@ def eval(args):
 
     val_program, feed_names, fetch_targets = fluid.io.load_inference_model(args.model_path,
                                       exe,
-                                      model_filename="__model__",
+                                      model_filename="__model__.infer",
                                       params_filename="__params__")
     val_reader = paddle.batch(reader.val(), batch_size=128)
     feeder = fluid.DataFeeder(place=place, feed_list=feed_names, program=val_program)
 
     results=[]
     for batch_id, data in enumerate(val_reader()):
+        image = [[d[0]] for d in data]
+        label = [[d[1]] for d in data]
+        feed_data = feeder.feed(image)
+        pred = exe.run(val_program,
+                        feed=feed_data,
+                        fetch_list=fetch_targets)
+        pred = np.array(pred[0])
+        label = np.array(label)
+        sort_array = pred.argsort(axis=1)
+        top_1_pred = sort_array[:, -1:][:, ::-1]
+        top_1 = np.mean(label == top_1_pred)
+        top_5_pred = sort_array[:, -5:][:, ::-1]
+        acc_num = 0
+        for i in range(len(label)):
+            if label[i][0] in top_5_pred[i]:
+                acc_num += 1
+        top_5 = acc_num / len(label)
+        results.append([top_1, top_5])
 
-        # top1_acc, top5_acc
-        result = exe.run(val_program,
-                          feed=feeder.feed(data),
-                          fetch_list=fetch_targets)
-        result = [np.mean(r) for r in result]
-        results.append(result)
     result = np.mean(np.array(results), axis=0)
     print("top1_acc/top5_acc= {}".format(result))
     sys.stdout.flush()