diff --git a/deepmd/descriptor/se_a.py b/deepmd/descriptor/se_a.py
index a690bbef4a..aa511bb41f 100644
--- a/deepmd/descriptor/se_a.py
+++ b/deepmd/descriptor/se_a.py
@@ -13,6 +13,7 @@
 from deepmd.utils.tabulate import DPTabulate
 from deepmd.utils.type_embed import embed_atom_type
 from deepmd.utils.sess import run_sess
+from deepmd.utils.graph import load_graph_def, get_tensor_by_name_from_graph
 
 class DescrptSeA ():
     @docstring_parameter(list_to_doc(ACTIVATION_FN_DICT.keys()), list_to_doc(PRECISION_DICT.keys()))
@@ -275,6 +276,11 @@ def enable_compression(self,
                                table_extrapolate, 
                                table_stride_1, 
                                table_stride_2)
+        
+        graph, _ = load_graph_def(model_file)
+        self.davg = get_tensor_by_name_from_graph(graph, 'descrpt_attr/t_avg')
+        self.dstd = get_tensor_by_name_from_graph(graph, 'descrpt_attr/t_std')
+
 
 
     def build (self, 
diff --git a/deepmd/entrypoints/compress.py b/deepmd/entrypoints/compress.py
index 5114d6410e..c689481b96 100644
--- a/deepmd/entrypoints/compress.py
+++ b/deepmd/entrypoints/compress.py
@@ -141,8 +141,3 @@ def compress(
     log.info("\n\n")
     log.info("stage 2: freeze the model")
     freeze(checkpoint_folder=checkpoint_folder, output=output, node_names=None)
-
-    # stage 3: transfer the model
-    log.info("\n\n")
-    log.info("stage 3: transfer the model")
-    transfer(old_model=input, raw_model=output, output=output)
diff --git a/deepmd/train/trainer.py b/deepmd/train/trainer.py
index 49c6a5a537..8d3cfe2b42 100644
--- a/deepmd/train/trainer.py
+++ b/deepmd/train/trainer.py
@@ -317,6 +317,7 @@ def build (self,
         else :
             assert 'rcut' in self.descrpt_param, "Error: descriptor must have attr rcut!"
             self.descrpt.enable_compression(self.model_param['compress']["min_nbor_dist"], self.model_param['compress']['model_file'], self.model_param['compress']['table_config'][0], self.model_param['compress']['table_config'][1], self.model_param['compress']['table_config'][2], self.model_param['compress']['table_config'][3])
+            self.fitting.init_variables(get_fitting_net_variables(self.model_param['compress']['model_file']))
         
         if self.is_compress or self.model_type == 'compressed_model':
             tf.constant("compressed_model", name = 'model_type', dtype = tf.string)