Eager mode for tf.add works.

src/TensorFlowNET.Core/Eager/pywrap_tfe_src.cs

@@ -63,12 +63,11 @@ namespace Tensorflow.Eager
                 var retVals = new IntPtr[num_retvals];
                 c_api.TFE_Execute(op, retVals, ref num_retvals, status);
 
-                var h = c_api.TFE_NewTensorHandle(retVals[0], status);
-                var data = new Tensor(h);
+                var t = c_api.TFE_TensorHandleResolve(retVals[0], status);
                 status.Check(true);
-            }
 
-            throw new NotImplementedException("");
+                return new EagerTensor(t);
+            }
         }
 
         /// <summary>

src/TensorFlowNET.Core/Operations/gen_math_ops.cs

@@ -142,11 +142,10 @@ namespace Tensorflow
 
         public static EagerTensor add(Tensor x, Tensor y, string name = null)
         {
-            // _op_def_lib._apply_op_helper("Add", name, args: new { x, y });
-
             if (tf.context.executing_eagerly())
             {
                 var _result = pywrap_tfe_src.TFE_Py_FastPathExecute(tf.context, "", "Add", name, new[] { x, y });
+                return _result;
             }
 
             return null;