diff --git a/paddleslim/auto_compression/compressor.py b/paddleslim/auto_compression/compressor.py
index a74105a237d96294ebc87ea8f367b2bcbbb77778..916074ed0b36d0cd5f53f6c982784f502bb0260c 100644
--- a/paddleslim/auto_compression/compressor.py
+++ b/paddleslim/auto_compression/compressor.py
@@ -263,7 +263,12 @@ class AutoCompression:
         save_path = os.path.join(save_path, "infered_shape")
         os.makedirs(save_path)
         paddle.static.save_inference_model(
-            save_path, feed_vars, fetch_targets, exe, program=inference_program)
+            save_path,
+            feed_vars,
+            fetch_targets,
+            exe,
+            program=inference_program,
+            clip_extra=False)
         _logger.info(f"Saved model infered shape to {save_path}")
 
     @property
@@ -843,7 +848,8 @@ class AutoCompression:
             feed_vars=feed_vars,
             fetch_vars=test_program_info.fetch_targets,
             executor=self._exe,
-            program=test_program)
+            program=test_program,
+            clip_extra=False)
 
     def export_onnx(self,
                     model_name='quant_model.onnx',
diff --git a/paddleslim/auto_compression/utils/fake_ptq.py b/paddleslim/auto_compression/utils/fake_ptq.py
index 91cccfc2f6e90f6a02f33055b4498b64b5514821..bce49b4f12f72c379b5df62462a5d2a443b7ccd4 100644
--- a/paddleslim/auto_compression/utils/fake_ptq.py
+++ b/paddleslim/auto_compression/utils/fake_ptq.py
@@ -169,5 +169,6 @@ def post_quant_fake(executor,
             feed_vars=feed_vars,
             fetch_vars=_fetch_list,
             executor=executor,
-            program=_program)
+            program=_program,
+            clip_extra=False)
         print("The quantized model is saved in: " + save_model_path)
diff --git a/paddleslim/auto_compression/utils/prune_model.py b/paddleslim/auto_compression/utils/prune_model.py
index c0da14ca9693112cf6919294f21136b86a5ea1d5..a784aa11dce62358d02d8451e5ef44e9af074138 100644
--- a/paddleslim/auto_compression/utils/prune_model.py
+++ b/paddleslim/auto_compression/utils/prune_model.py
@@ -95,7 +95,8 @@ def get_sparse_model(executor, places, model_file, param_file, ratio,
         feed_vars=feed_vars,
         fetch_vars=fetch_targets,
         executor=executor,
-        program=inference_program)
+        program=inference_program,
+        clip_extra=False)
     print("The pruned model is saved in: ", save_path)
 
 
@@ -170,4 +171,5 @@ def get_prune_model(executor, places, model_file, param_file, ratio, save_path):
         feed_vars=feed_vars,
         fetch_vars=fetch_targets,
         executor=executor,
-        program=main_program)
+        program=main_program,
+        clip_extra=False)