Saver is inprogress.

src/TensorFlowNET.Core/Graphs/Graph.Control.cs

@@ -28,9 +28,9 @@ namespace Tensorflow
         /// </summary>
         /// <param name="input_ops">The data input ops for an op to be created.</param>
         /// <returns>A list of control inputs for the op to be created.</returns>
-        private Operation[] _control_dependencies_for_inputs(Operation[] input_ops)
+        private ITensorOrOperation[] _control_dependencies_for_inputs(ITensorOrOperation[] input_ops)
         {
-            Operation[] ret = new Operation[0];
+            var ret = new ITensorOrOperation[0];
 
             foreach(var controller in _control_dependencies_stack)
             {
@@ -54,12 +54,12 @@ namespace Tensorflow
             return ret;
         }
 
-        public _ControlDependenciesController control_dependencies(Operation[] control_inputs)
+        public _ControlDependenciesController control_dependencies(ITensorOrOperation[] control_inputs)
         {
             if (control_inputs == null)
                 return new _ControlDependenciesController(this, null);
 
-            var control_ops = new List<Operation>();
+            var control_ops = new List<ITensorOrOperation>();
             foreach (var c in control_inputs)
             {
                 control_ops.Add(c);

src/TensorFlowNET.Core/Graphs/Graph.cs

@@ -298,6 +298,11 @@ namespace Tensorflow
             return _nodes_by_name.Values.Select(x => x).ToArray();
         }
 
+        public string[] get_all_collection_keys()
+        {
+            return _collections.Keys.Where(x => !x.StartsWith("__")).ToArray();
+        }
+
         public object get_collection(string name, string scope = "")
         {
             return _collections.ContainsKey(name) ? _collections[name] : null;

src/TensorFlowNET.Core/Graphs/_ControlDependenciesController.cs

@@ -11,20 +11,20 @@ namespace Tensorflow
     public class _ControlDependenciesController : IPython
     {
         private Graph _graph;
-        private List<Operation> _control_inputs_val;
-        private List<Operation> _seen_nodes;
+        private List<ITensorOrOperation> _control_inputs_val;
+        private List<ITensorOrOperation> _seen_nodes;
         private Queue<_ControlDependenciesController> _old_stack;
         private bool _new_stack;
         private Context _old_control_flow_context;
 
-        public Operation[] control_inputs => _control_inputs_val.ToArray();
+        public ITensorOrOperation[] control_inputs => _control_inputs_val.ToArray();
 
-        public _ControlDependenciesController(Graph graph, List<Operation> control_inputs)
+        public _ControlDependenciesController(Graph graph, List<ITensorOrOperation> control_inputs)
         {
             _graph = graph;
             if (control_inputs == null)
             {
-                _control_inputs_val = new List<Operation>();
+                _control_inputs_val = new List<ITensorOrOperation>();
                 _new_stack = true;
             }
             else
@@ -33,15 +33,15 @@ namespace Tensorflow
                 _new_stack = false;
             }
 
-            _seen_nodes = new List<Operation>();
+            _seen_nodes = new List<ITensorOrOperation>();
         }
 
-        public void add_op(Operation op)
+        public void add_op(ITensorOrOperation op)
         {
             _seen_nodes.Add(op);
         }
 
-        public bool op_in_group(Operation op)
+        public bool op_in_group(ITensorOrOperation op)
         {
             return _seen_nodes.Contains(op);
         }

src/TensorFlowNET.Core/ITensorOrOperation.cs

@@ -11,5 +11,6 @@ namespace Tensorflow
     public interface ITensorOrOperation
     {
         string Device { get; }
+        Operation op { get; }
     }
 }

src/TensorFlowNET.Core/Operations/OpDefLibrary.cs

@@ -107,7 +107,9 @@ namespace Tensorflow
 
                         values = ops.internal_convert_to_tensor(values, 
                             name: input_name, 
-                            as_ref: input_arg.IsRef);
+                            dtype: dtype,
+                            as_ref: input_arg.IsRef,
+                            preferred_dtype: default_dtype);
 
                         //if (!String.IsNullOrEmpty(input_arg.TypeAttr))
                             //attrs[input_arg.TypeAttr] = values.dtype;
@@ -163,14 +165,20 @@ namespace Tensorflow
 
                 foreach (var arg in op_def.OutputArg)
                 {
+                    types = new List<TF_DataType>();
                     if (!string.IsNullOrEmpty(arg.NumberAttr))
                     {
 
                     }
                     else if (!string.IsNullOrEmpty(arg.TypeAttr))
                     {
-                        output_types.Add((TF_DataType)attr_protos[arg.TypeAttr].Type);
+                        types = new List<TF_DataType>() { (TF_DataType)attr_protos[arg.TypeAttr].Type };
                     }
+
+                    if (arg.IsRef)
+                        types = types.Select(x => x.as_ref()).ToList();
+
+                    output_types.AddRange(types);
                 }
 
                 // Add Op to graph

src/TensorFlowNET.Core/Operations/Operation.cs

@@ -16,6 +16,7 @@ namespace Tensorflow
         private int _id_value;
 
         public string type => OpType;
+        public Operation op => this;
 
         private Status status = new Status();
 
@@ -75,7 +76,7 @@ namespace Tensorflow
         /// </param>
         /// <param name="original_op"></param>
         /// <param name="op_def"></param>
-        public Operation(NodeDef node_def, Graph g, Tensor[] inputs = null, TF_DataType[] output_types = null, Operation[] control_inputs = null, TF_DataType[] input_types = null, string original_op = "", OpDef op_def = null)
+        public Operation(NodeDef node_def, Graph g, Tensor[] inputs = null, TF_DataType[] output_types = null, ITensorOrOperation[] control_inputs = null, TF_DataType[] input_types = null, string original_op = "", OpDef op_def = null)
         {
             Graph = g;
 
@@ -120,6 +121,11 @@ namespace Tensorflow
                 _control_flow_post_processing();
         }
 
+        public void run(FeedItem[] feed_dict = null, Session session = null)
+        {
+            ops._run_using_default_session(this, feed_dict, Graph, session);
+        }
+
         private object[] _reconstruct_sequence_inputs(OpDef op_def, Tensor[] inputs, MapField<string, AttrValue> attrs)
         {
             var grouped_inputs = new List<object>();
@@ -204,7 +210,7 @@ namespace Tensorflow
 
         public override string ToString()
         {
-            return _handle == IntPtr.Zero ? "Undefined" : $"'{Name}' type={OpType}";
+            return _handle == IntPtr.Zero ? "tf.Operation Undefined" : $"tf.Operation '{Name}' type={OpType}";
         }
 
         public static implicit operator Operation(IntPtr handle) => new Operation(handle);

src/TensorFlowNET.Core/Operations/control_flow_ops.py.cs

@@ -28,10 +28,7 @@ namespace Tensorflow
                 {
                     var dev = ops_on_device.Keys.First();
                     var deps = ops_on_device.Values.First();
-                    if (typeof(T).Name == "Operation")
-                        return _GroupControlDeps(dev, deps.Select(x => x as Operation).ToArray(), name);
-                    else
-                        throw new NotImplementedException("control_flow_ops.group");
+                    return _GroupControlDeps(dev, deps.Select(x => x.op).ToArray(), name);
                 }
 
                 // 2-level tree. The root node is the returned NoOp node.

src/TensorFlowNET.Core/Sessions/BaseSession.cs

@@ -35,17 +35,8 @@ namespace Tensorflow
             c_api.TF_DeleteSessionOptions(opts);
         }
 
-        public virtual NDArray run(RefVariable fetches, FeedItem[] feed_dict = null)
-        {
-            return _run(fetches, feed_dict);
-        }
-
-        public virtual NDArray run(Tensor fetches, FeedItem[] feed_dict = null)
-        {
-            return _run(fetches, feed_dict);
-        }
 
-        public virtual NDArray run(Operation fetches, FeedItem[] feed_dict = null)
+        public virtual NDArray run<T>(T fetches, FeedItem[] feed_dict = null)
         {
             return _run(fetches, feed_dict);
         }

src/TensorFlowNET.Core/Tensors/Tensor.Implicit.cs

@@ -30,6 +30,11 @@ namespace Tensorflow
             return tensor._handle;
         }
 
+        public static implicit operator Operation(Tensor tensor)
+        {
+            return tensor.op;
+        }
+
         public static implicit operator Tensor(IntPtr handle)
         {
             return new Tensor(handle);

src/TensorFlowNET.Core/Tensors/Tensor.cs

@@ -261,7 +261,7 @@ namespace Tensorflow
                 }
             }
 
-            return $"tf.Tensor {name} shape=({string.Join(",", shape)}) dtype={dtype.ToString()}";
+            return $"tf.Tensor '{name}' shape=({string.Join(",", shape)}) dtype={dtype.ToString()}";
         }
 
         public void Dispose()

src/TensorFlowNET.Core/Tensors/dtypes.cs

@@ -83,6 +83,13 @@ namespace Tensorflow
                 type;
         }
 
+        public static TF_DataType as_ref(this TF_DataType type)
+        {
+            return (int)type < 100 ?
+                (TF_DataType)Enum.Parse(typeof(TF_DataType), ((int)type + 100).ToString()) :
+                type;
+        }
+
         public static bool is_complex(this TF_DataType type)
         {
             return type == TF_DataType.TF_COMPLEX || type == TF_DataType.TF_COMPLEX64 || type == TF_DataType.TF_COMPLEX128;

src/TensorFlowNET.Core/Train/Saving/BaseSaverBuilder.cs

@@ -7,9 +7,9 @@ namespace Tensorflow
 {
     public class BaseSaverBuilder
     {
-        protected int _write_version;
+        protected SaverDef.Types.CheckpointFormatVersion _write_version;
 
-        public BaseSaverBuilder(int write_version = 2)
+        public BaseSaverBuilder(SaverDef.Types.CheckpointFormatVersion write_version = SaverDef.Types.CheckpointFormatVersion.V2)
         {
             _write_version = write_version;
         }
@@ -30,7 +30,7 @@ namespace Tensorflow
                 }
             }
 
-            if (_write_version == 2)
+            if (_write_version == SaverDef.Types.CheckpointFormatVersion.V2)
             {
                 return gen_io_ops.save_v2(filename_tensor, tensor_names.ToArray(), tensor_slices.ToArray(), tensors.ToArray());
             }
@@ -60,7 +60,7 @@ namespace Tensorflow
             bool reshape = false,
             bool sharded = false,
             int max_to_keep = 5,
-            double keep_checkpoint_every_n_hours = 10000,
+            float keep_checkpoint_every_n_hours = 10000,
             string name = "",
             bool restore_sequentially = false,
             string filename = "model",
@@ -76,7 +76,10 @@ namespace Tensorflow
             if (max_to_keep < 0)
                 max_to_keep = 0;
 
-            Python.with<ops.name_scope>(new ops.name_scope(name, "save", saveables.Select(x => x.op).ToArray()), scope =>
+            Tensor save_tensor = null;
+            Operation restore_op = null;
+
+            return Python.with<ops.name_scope, SaverDef>(new ops.name_scope(name, "save", saveables.Select(x => x.op).ToArray()), scope =>
             {
                 name = scope;
 
@@ -93,14 +96,35 @@ namespace Tensorflow
                 else
                 {
                     if (build_save)
-                        _AddSaveOps(filename_tensor, saveables);
+                        save_tensor = _AddSaveOps(filename_tensor, saveables);
 
                     if (build_restore)
-                        _AddRestoreOps(filename_tensor, saveables, restore_sequentially, reshape);
+                        restore_op = _AddRestoreOps(filename_tensor, saveables, restore_sequentially, reshape);
                 }
+
+                var graph = ops.get_default_graph();
+                var check_collection_list = graph.get_all_collection_keys();
+                foreach (var collection_type in check_collection_list)
+                {
+                    foreach (var element in graph.get_collection(collection_type) as IList<RefVariable>)
+                    {
+
+                    }
+                }
+
+                return new SaverDef()
+                {
+                    FilenameTensorName = filename_tensor.name,
+                    SaveTensorName = save_tensor.name,
+                    RestoreOpName = restore_op.Name,
+                    MaxToKeep = max_to_keep,
+                    Sharded = sharded,
+                    KeepCheckpointEveryNHours = keep_checkpoint_every_n_hours,
+                    Version = _write_version
+                };
             });
 
-            throw new NotImplementedException("");
+            
         }
 
         public Tensor _AddSaveOps(Tensor filename_tensor, SaveableObject[] saveables)

src/TensorFlowNET.Core/Train/Saving/BulkSaverBuilder.cs

@@ -6,7 +6,7 @@ namespace Tensorflow
 {
     public class BulkSaverBuilder : BaseSaverBuilder, ISaverBuilder
     {
-        public BulkSaverBuilder(int write_version = 2) : base(write_version)
+        public BulkSaverBuilder(SaverDef.Types.CheckpointFormatVersion write_version = SaverDef.Types.CheckpointFormatVersion.V2) : base(write_version)
         {
 
         }

src/TensorFlowNET.Core/Train/Saving/ISaverBuilder.cs

@@ -14,7 +14,7 @@ namespace Tensorflow
             bool reshape = false, 
             bool sharded = false, 
             int max_to_keep = 5,
-            double keep_checkpoint_every_n_hours = 10000,
+            float keep_checkpoint_every_n_hours = 10000,
             string name = "",
             bool restore_sequentially = false,
             string filename = "model",

src/TensorFlowNET.Core/Train/Saving/Saver.cs

@@ -1,5 +1,6 @@
 using System;
 using System.Collections.Generic;
+using System.IO;
 using System.Text;
 
 namespace Tensorflow
@@ -13,30 +14,33 @@ namespace Tensorflow
         private bool _reshape;
         private bool _sharded;
         private int _max_to_keep;
-        private double _keep_checkpoint_every_n_hours;
+        private float _keep_checkpoint_every_n_hours;
         private string _name;
         private bool _restore_sequentially;
         private SaverDef _saver_def;
         private ISaverBuilder _builder;
         private bool _allow_empty;
         private bool _is_built;
-        private int _write_version;
+        private SaverDef.Types.CheckpointFormatVersion _write_version;
         private bool _pad_step_number;
         private string _filename;
         private bool _is_empty;
+        private float _next_checkpoint_time;
+        private bool _save_relative_paths;
+        private bool? _object_restore_saver;
 
         public Saver(RefVariable[] var_list = null,
             bool reshape = false,
             bool sharded = false,
             int max_to_keep = 5,
-            double keep_checkpoint_every_n_hours = 10000,
+            float keep_checkpoint_every_n_hours = 10000,
             string name = "",
             bool restore_sequentially = false,
             SaverDef saver_def = null,
             ISaverBuilder builder = null,
             bool defer_build = false,
             bool allow_empty = false,
-            int write_version = 2,
+            SaverDef.Types.CheckpointFormatVersion write_version = SaverDef.Types.CheckpointFormatVersion.V2,
             bool pad_step_number = false,
             bool save_relative_paths = false,
             string filename = "")
@@ -56,6 +60,14 @@ namespace Tensorflow
 
             if (!defer_build)
                 build();
+            if(_saver_def != null)
+            {
+                _check_saver_def();
+                _write_version = _saver_def.Version;
+            }
+
+            _save_relative_paths = save_relative_paths;
+            _object_restore_saver = null;
         }
 
         public void build()
@@ -106,8 +118,56 @@ namespace Tensorflow
             {
                 throw new NotImplementedException("");
             }
-            
 
+            _check_saver_def();
+
+            _next_checkpoint_time = (float)(DateTime.UtcNow - new DateTime(1970, 1, 1)).TotalSeconds + _saver_def.KeepCheckpointEveryNHours * 3600;
+        }
+
+        private void _check_saver_def()
+        {
+            if (!tf.context.executing_eagerly())
+            {
+                if (string.IsNullOrEmpty(_saver_def.SaveTensorName))
+                    throw new ValueError($"saver_def must specify the save_tensor_name: {_saver_def}");
+                if (string.IsNullOrEmpty(_saver_def.RestoreOpName))
+                    throw new ValueError($"saver_def must specify the restore_op_name: {_saver_def}");
+            }
+        }
+
+        public string save(Session sess,
+            string save_path,
+            string global_step = "",
+            string meta_graph_suffix = "meta",
+            bool write_meta_graph = true,
+            bool write_state = true,
+            bool strip_default_attrs = false)
+        {
+            string latest_filename = "checkpoint";
+            string model_checkpoint_path = "";
+            string checkpoint_file = "";
+
+            if (!string.IsNullOrEmpty(global_step))
+            {
+
+            }
+            else
+            {
+                checkpoint_file = save_path;
+            }
+
+            var save_path_parent = Path.GetDirectoryName(save_path);
+
+            if (!_is_empty)
+            {
+                /*model_checkpoint_path = sess.run(_saver_def.SaveTensorName, new FeedItem[] {
+                    new FeedItem(_saver_def.FilenameTensorName, checkpoint_file)
+                });*/
+            }
+
+            throw new NotImplementedException("");
+
+            return model_checkpoint_path;
         }
     }
 }

src/TensorFlowNET.Core/Variables/RefVariable.cs

@@ -185,18 +185,12 @@ namespace Tensorflow
         /// A `Tensor` that will hold the new value of this variable after
         /// the assignment has completed.
         /// </returns>
-        public T assign<T>(Tensor value, bool use_locking = false, string name = "", bool read_value = true)
-            where T : ITensorOrOperation
+        public ITensorOrOperation assign(Tensor value, bool use_locking = false, string name = "", bool read_value = true)
         {
             var assign = gen_state_ops.assign(_variable, value, use_locking: use_locking, name: name);
             if (read_value)
-                return (T)Convert.ChangeType(assign, typeof(T));
-            return (T)Convert.ChangeType(assign.op, typeof(T));
-        }
-
-        public Tensor assign(Tensor value, bool use_locking = false, string name = "")
-        {
-            return gen_state_ops.assign(_variable, value, use_locking: use_locking, name: name);
+                return assign;
+            return assign.op;
         }
 
         public override string ToString()

src/TensorFlowNET.Core/ops.py.cs

@@ -292,6 +292,27 @@ namespace Tensorflow
             return tf.Session();
         }
 
+        public static void _run_using_default_session(Operation operation, FeedItem[] feed_dict, Graph graph, Session session)
+        {
+            if (session == null)
+            {
+                session = get_default_session();
+                if (session == null)
+                    throw new ValueError("Cannot execute operation using `run()`: No default " +
+                       "session is registered. Use `with " +
+                       "sess.as_default():` or pass an explicit session to " +
+                       "`run(session=sess)`");
+            }
+
+            if (session.graph != graph)
+                throw new ValueError("Cannot use the default session to execute operation: " +
+                   "the operation's graph is different from the " +
+                   "session's graph. Pass an explicit session to " +
+                   "run(session=sess).");
+
+            session.run(operation, feed_dict);
+        }
+
         public static Func<Operation, Tensor, Tensor[]> get_gradient_function(Operation op)
         {
             if (op.inputs == null) return null;

test/TensorFlowNET.UnitTest/TrainSaverTest.cs

@@ -27,6 +27,13 @@ namespace TensorFlowNET.UnitTest
             with<Session>(tf.Session(), sess =>
             {
                 sess.run(init_op);
+                // o some work with the model.
+                inc_v1.op.run();
+                dec_v2.op.run();
+
+                // Save the variables to disk.
+                var save_path = saver.save(sess, "/tmp/model.ckpt");
+                Console.WriteLine($"Model saved in path: {save_path}");
             });
         }
     }

test/TensorFlowNET.UnitTest/VariableTest.cs

@@ -46,6 +46,24 @@ namespace TensorFlowNET.UnitTest
             }
         }
 
+        [TestMethod]
+        public void Assign()
+        {
+            var v1 = tf.get_variable("v1", shape: new TensorShape(3), initializer: tf.zeros_initializer);
+
+            var inc_v1 = v1.assign(v1 + 1.0f);
+
+            // Add an op to initialize the variables.
+            var init_op = tf.global_variables_initializer();
+
+            with<Session>(tf.Session(), sess =>
+            {
+                sess.run(init_op);
+                // o some work with the model.
+                inc_v1.op.run();
+            });
+        }
+
         /// <summary>
         /// https://databricks.com/tensorflow/variables
         /// </summary>