fix: partially fix the error when saving model after loading.

src/TensorFlowNET.Core/Checkpoint/CheckPointUtils.cs

@@ -3,6 +3,7 @@ using System.Collections.Generic;
 using System.Diagnostics;
 using System.IO;
 using System.Linq;
+using Tensorflow.Functions;
 using Tensorflow.Train;
 using Tensorflow.Training;
 using pbc = global::Google.Protobuf.Collections;
@@ -13,7 +14,7 @@ public static class CheckPointUtils
 {
     private static string _ESCAPE_CHAR = ".";
     public static (IList<Trackable>, IDictionary<Trackable, IEnumerable<TrackableReference>>, IDictionary<Trackable, int>,
-        IDictionary<Trackable, pbc::RepeatedField<global::Tensorflow.TrackableObjectGraph.Types.TrackableObject.Types.SlotVariableReference>>,
+        IDictionary<Trackable, pbc::RepeatedField<TrackableObjectGraph.Types.TrackableObject.Types.SlotVariableReference>>,
         IDictionary<Trackable, string>) objects_ids_and_slot_variables_and_paths(ObjectGraphView graph_view)
     {
         var (trackable_objects, node_paths) = graph_view.breadth_first_traversal();

src/TensorFlowNET.Core/Training/Saving/SavedModel/SaveableView.cs

@@ -93,13 +93,14 @@ public class SaveableView
         //     
         // }
 
-        foreach (var obj in _nodes)
-        {
-            if (obj is ConcreteFunction)
-            {
-                _concrete_functions.Add((ConcreteFunction)obj);
-            }
-        }
+        //_concrete_functions = new();
+        //foreach (var obj in _nodes)
+        //{
+        //    if (obj is ConcreteFunction)
+        //    {
+        //        _concrete_functions.Add((ConcreteFunction)obj);
+        //    }
+        //}
     }
 
     public List<ConcreteFunction> get_concrete_resource_initializers()
@@ -225,8 +226,8 @@ public class SaveableView
         }
         else if (obj is ConcreteFunction)
         {
-            // TODO: complete it.
-            throw new NotImplementedException();
+            // TODO(Rinne): complete it.
+            // throw new NotImplementedException();
         }
         // skip the process of type `_CapturedTensor` and `CapturableResource`.
         else

src/TensorFlowNET.Core/Variables/BaseResourceVariable.cs

@@ -17,7 +17,14 @@ namespace Tensorflow
     {
         protected string _name;
         public virtual string Name => _handle_name;
-        public virtual string SharedName => _name;
+        public virtual string SharedName
+        {
+            get
+            {
+                // TODO(Rinne): optimize the implementation with refactor of variable.
+                return _handle_name.Substring(0, _handle_name.IndexOf(':') + 1);
+            }
+        }
         protected TF_DataType _dtype;
         public TF_DataType dtype => _dtype;
         protected string _handle_name;

src/TensorFlowNET.Keras/Saving/KerasObjectLoader.cs

@@ -152,6 +152,39 @@ namespace Tensorflow.Keras.Saving
             _reconstruct_all_models();
         }
 
+        /// <summary>
+        /// Removes tracked references that are only used when loading the model.
+        /// Now that the node object has been fully loaded, and the checkpoint has
+        /// been restored, the object no longer needs to track objects added from
+        /// SerializedAttributes. (Note that saving a training checkpoint still
+        /// functions correctly, because layers and variables are tracked
+        /// separately by the Layer object.)
+        /// </summary>
+        public void del_tracking()
+        {
+            foreach(var (node, _) in loaded_nodes.Values)
+            {
+                if(node is not Layer layer)
+                {
+                    continue;
+                }
+                foreach(var name in PUBLIC_ATTRIBUTES.Keys)
+                {
+                    layer._delete_tracking(name);
+                }
+                if(node is Functional functional)
+                {
+                    foreach(var name in functional.UnconditionalDependencyNames.Keys)
+                    {
+                        if(Regex.Match(name, @"^layer(_with_weights)?-[\d+]").Success)
+                        {
+                            functional._delete_tracking(name);
+                        }
+                    }
+                }
+            }
+        }
+
         private void _reconstruct_all_models()
         {
             HashSet<int> all_initialized_models = new();

src/TensorFlowNET.Keras/Saving/SavedModel/load.cs

@@ -77,7 +77,7 @@ namespace Tensorflow.Keras.Saving.SavedModel
             var loaded = Loader.load_partial(path, nodes_to_load, options);
 
             keras_loader.finalize_objects();
-            // keras_loader.del_tracking();
+            keras_loader.del_tracking();
 
             var model = loaded["root"];
 

test/TensorFlowNET.Keras.UnitTest/Model/ModelSaveTest.cs

@@ -196,5 +196,17 @@ namespace Tensorflow.Keras.UnitTest.Model
             //    )
             #endregion
         }
+
+        [TestMethod]
+        public void SaveAfterLoad()
+        {
+            var model = tf.keras.models.load_model(@"Assets/simple_model_from_auto_compile");
+            model.summary();
+
+            model.save("Assets/saved_auto_compile_after_loading");
+
+            //model = tf.keras.models.load_model(@"Assets/saved_auto_compile_after_loading");
+            //model.summary();
+        }
     }
 }