diff --git a/src/TensorFlowNET.Core/Binding.Util.cs b/src/TensorFlowNET.Core/Binding.Util.cs
index 5c7641e0..91888e4b 100644
--- a/src/TensorFlowNET.Core/Binding.Util.cs
+++ b/src/TensorFlowNET.Core/Binding.Util.cs
@@ -155,7 +155,7 @@ namespace Tensorflow
             switch (a)
             {
                 case Tensor tensor:
-                    return tensor.shape[0];
+                    return (int)tensor.shape[0];
                 case Tensors arr:
                     return arr.Length;
                 case Array arr:
diff --git a/src/TensorFlowNET.Core/Framework/tensor_shape.cs b/src/TensorFlowNET.Core/Framework/tensor_shape.cs
index 73cd7daf..c88fb876 100644
--- a/src/TensorFlowNET.Core/Framework/tensor_shape.cs
+++ b/src/TensorFlowNET.Core/Framework/tensor_shape.cs
@@ -10,7 +10,7 @@ namespace Tensorflow.Framework
     {
         public static void assert_is_compatible_with(this Tensor self, Tensor other)
         {
-            if (!self.is_compatible_with(other))
+            /*if (!self.is_compatible_with(other))
             {
                 var selfDim = self.shape
                     .Aggregate(new StringBuilder("{"), (sb, i) => sb.Append(i).Append(", "), sb => sb.ToString())
@@ -21,7 +21,7 @@ namespace Tensorflow.Framework
                     .Replace(", }", "}");
 
                 throw new ArgumentException($"Dimensions {selfDim} and {otherDim} are not compatible");
-            }
+            }*/
         }
 
         public static bool is_compatible_with(this Tensor self, Tensor other)
diff --git a/src/TensorFlowNET.Core/Gradients/image_grad.cs b/src/TensorFlowNET.Core/Gradients/image_grad.cs
index 08636298..fd7f098f 100644
--- a/src/TensorFlowNET.Core/Gradients/image_grad.cs
+++ b/src/TensorFlowNET.Core/Gradients/image_grad.cs
@@ -27,10 +27,10 @@ namespace Tensorflow.Gradients
         {
             var grad = grads[0];
             var image = op.inputs[0];
-            var shape = new TensorShape(image.shape.Skip(1).Take(2).ToArray());
+            var shape = new TensorShape(image.shape.dims.Skip(1).Take(2).ToArray());
             Tensor image_shape = null;
             if (shape.is_fully_defined())
-                image_shape = constant_op.constant(image.shape.Skip(1).Take(2).ToArray());
+                image_shape = constant_op.constant(image.shape.dims.Skip(1).Take(2).ToArray());
             else
                 image_shape = array_ops.shape(image)["1:3"];
 
diff --git a/src/TensorFlowNET.Core/Gradients/math_grad.cs b/src/TensorFlowNET.Core/Gradients/math_grad.cs
index 34710f70..4eb1087e 100644
--- a/src/TensorFlowNET.Core/Gradients/math_grad.cs
+++ b/src/TensorFlowNET.Core/Gradients/math_grad.cs
@@ -195,7 +195,7 @@ namespace Tensorflow.Gradients
 
             if (op is EagerOperation op_eager &&
                 op_eager.SkipInputIndices.Contains(1) &&
-                y.NDims == 0)
+                y.ndim == 0)
             {
                 return new Tensor[]
                 {
@@ -759,7 +759,7 @@ namespace Tensorflow.Gradients
 
             if (op is EagerOperation op_eager &&
                 op_eager.SkipInputIndices.Contains(1) &&
-                y.NDims == 0)
+                y.ndim == 0)
             {
                 x = math_ops.conj(x);
                 y = math_ops.conj(y);
diff --git a/src/TensorFlowNET.Core/NumPy/Axis.cs b/src/TensorFlowNET.Core/NumPy/Axis.cs
index 97629062..b170d90b 100644
--- a/src/TensorFlowNET.Core/NumPy/Axis.cs
+++ b/src/TensorFlowNET.Core/NumPy/Axis.cs
@@ -1,5 +1,6 @@
 using System;
 using System.Collections.Generic;
+using System.Linq;
 using System.Text;
 
 namespace Tensorflow
@@ -22,6 +23,12 @@ namespace Tensorflow
 
         public static implicit operator Axis(int[] axis)
             => new Axis(axis);
+
+        public static implicit operator Axis(long[] shape)
+            => new Axis(shape.Select(x => (int)x).ToArray());
+
+        public static implicit operator Axis(Shape shape)
+            => new Axis(shape.dims.Select(x => (int)x).ToArray());
     }
 }
 
diff --git a/src/TensorFlowNET.Core/Numpy/NDArray.cs b/src/TensorFlowNET.Core/Numpy/NDArray.cs
index 1cfc9b4e..719dba77 100644
--- a/src/TensorFlowNET.Core/Numpy/NDArray.cs
+++ b/src/TensorFlowNET.Core/Numpy/NDArray.cs
@@ -11,8 +11,9 @@ namespace Tensorflow.NumPy
         Tensor _tensor;
         public TF_DataType dtype => _tensor.dtype;
         public ulong size => _tensor.size;
-        public ulong dtypesize => _tensor.itemsize;
-        public int ndim => _tensor.NDims;
+        public ulong dtypesize => _tensor.dtypesize;
+        public ulong bytesize => _tensor.bytesize;
+        public int ndim => _tensor.ndim;
         public long[] dims => _tensor.dims.Select(x => Convert.ToInt64(x)).ToArray();
         public Shape shape => _tensor.shape;
         public IntPtr data => _tensor.TensorDataPointer;
diff --git a/src/TensorFlowNET.Core/Numpy/Shape.cs b/src/TensorFlowNET.Core/Numpy/Shape.cs
index 961955dd..c0b6048d 100644
--- a/src/TensorFlowNET.Core/Numpy/Shape.cs
+++ b/src/TensorFlowNET.Core/Numpy/Shape.cs
@@ -48,6 +48,12 @@ namespace Tensorflow
         public static implicit operator Shape((long, long, long, long) dims)
             => new Shape(dims.Item1, dims.Item2, dims.Item3, dims.Item4);
 
+        public static implicit operator int[](Shape shape)
+            => shape.dims.Select(x => (int)x).ToArray();
+
+        public static implicit operator long[](Shape shape)
+            => shape.dims;
+
         public bool IsEmpty => size == 0;
 
         public bool IsScalar => ndim == 0;
@@ -55,6 +61,8 @@ namespace Tensorflow
         public static Shape Scalar
             => new Shape(new long[0]);
 
+        public long this[int n] => dims[n];
+
         /// <summary>
         ///     Returns the size this shape represents.
         /// </summary>
@@ -81,6 +89,25 @@ namespace Tensorflow
             }
         }
 
+        public bool is_fully_defined()
+        {
+            return ndim > -1 && dims != null && dims.Count(x => x < 1) == 0;
+        }
+
+        public bool is_compatible_with(TensorShape shape2)
+        {
+            if (dims != null && shape2.dims != null)
+            {
+                if (dims.Contains(-1) || shape2.dims.Contains(-1))
+                    return true;
+
+                if (size != (ulong)shape2.size)
+                    return false;
+            }
+
+            return true;
+        }
+
         public override bool Equals(object obj)
         {
             if(obj is Shape shape)
diff --git a/src/TensorFlowNET.Core/Operations/Distributions/normal.py.cs b/src/TensorFlowNET.Core/Operations/Distributions/normal.py.cs
index 3e185c49..a73bbcc0 100644
--- a/src/TensorFlowNET.Core/Operations/Distributions/normal.py.cs
+++ b/src/TensorFlowNET.Core/Operations/Distributions/normal.py.cs
@@ -92,7 +92,7 @@ namespace Tensorflow
 
         public Tensor _batch_shape()
         {
-            return array_ops.broadcast_static_shape(new Tensor(_loc.shape), new Tensor(_scale.shape));
+            return array_ops.broadcast_static_shape(new Tensor(_loc.shape.dims), new Tensor(_scale.shape.dims));
         }
 
         protected override Tensor _log_prob(Tensor x)
diff --git a/src/TensorFlowNET.Core/Operations/NnOps/rnn_cell_impl.cs b/src/TensorFlowNET.Core/Operations/NnOps/rnn_cell_impl.cs
index cf5f1ce0..c76d768d 100644
--- a/src/TensorFlowNET.Core/Operations/NnOps/rnn_cell_impl.cs
+++ b/src/TensorFlowNET.Core/Operations/NnOps/rnn_cell_impl.cs
@@ -27,9 +27,9 @@ namespace Tensorflow.Operations
         {
             var p = prefix;
             var p_static = tensor_util.constant_value(prefix);
-            if (p.NDims == 0)
+            if (p.ndim == 0)
                 p = array_ops.expand_dims(p, 0);
-            else if (p.NDims != 1)
+            else if (p.ndim != 1)
                 throw new ValueError($"prefix tensor must be either a scalar or vector, but saw tensor: {p}");
 
             var s_tensor_shape = new TensorShape(suffix);
diff --git a/src/TensorFlowNET.Core/Operations/array_ops.cs b/src/TensorFlowNET.Core/Operations/array_ops.cs
index 9e7290ed..13db8194 100644
--- a/src/TensorFlowNET.Core/Operations/array_ops.cs
+++ b/src/TensorFlowNET.Core/Operations/array_ops.cs
@@ -186,7 +186,7 @@ namespace Tensorflow
 
         private static Tensor _constant_if_small(int value, Tensor shape)
         {
-            return shape < 1000L;
+            return shape < 1000UL;
         }
 
         private static Tensor _constant_if_small<T>(T value, TensorShape shape, TF_DataType dtype, string name)
@@ -330,7 +330,7 @@ namespace Tensorflow
             {
                 name = scope;
                 var input_tensor = ops.convert_to_tensor(inputs);
-                return constant_op.constant(input_tensor.NDims, dtype: tf.int32, name: name);
+                return constant_op.constant(input_tensor.ndim, dtype: tf.int32, name: name);
             });
         }
 
@@ -340,7 +340,7 @@ namespace Tensorflow
             {
                 name = scope;
                 var input_tensor = ops.convert_to_tensor(input);
-                var input_shape = tensor_util.to_shape(input_tensor.shape);
+                var input_shape = input_tensor.shape;
                 if (optimize && input_shape.ndim > 0)
                     return constant_op.constant(input_shape.ndim, dtype: tf.int32, name: name);
                 else
@@ -364,7 +364,7 @@ namespace Tensorflow
                 tensor = ops.convert_to_tensor(tensor, name: "tensor");
 
                 // is_fully_defined return unexpected value.
-                if (optimize && tensor_util.to_shape(tensor.shape).is_fully_defined() && dtype != TF_DataType.TF_VARIANT)
+                if (optimize && tensor.shape.is_fully_defined() && dtype != TF_DataType.TF_VARIANT)
                 {
 
                 }
@@ -589,9 +589,9 @@ namespace Tensorflow
                 if (!tf.Context.executing_eagerly())
                 {
                     var input_shape = input.TensorShape;
-                    if (optimize && input.NDims > -1 && input_shape.is_fully_defined())
+                    if (optimize && input.ndim > -1 && input_shape.is_fully_defined())
                     {
-                        var nd = np.array(input.shape).astype(out_type.as_system_dtype());
+                        var nd = np.array(input.shape.dims).astype(out_type.as_system_dtype());
                         return constant_op.constant(nd, name: name);
                     }
                 }
@@ -607,7 +607,7 @@ namespace Tensorflow
                 name = scope;
 
                 var input_tensor = ops.convert_to_tensor(input);
-                var input_shape = tensor_util.to_shape(input_tensor.shape);
+                var input_shape = input_tensor.shape;
                 if (optimize)
                 {
                     if (input_shape.is_fully_defined())
@@ -633,7 +633,7 @@ namespace Tensorflow
                 tensor = ops.convert_to_tensor(tensor, name: "tensor");
 
                 // is_fully_defined return unexpected value.
-                if (optimize && tensor_util.to_shape(tensor.shape).is_fully_defined() && dtype != TF_DataType.TF_VARIANT)
+                if (optimize && tensor.shape.is_fully_defined() && dtype != TF_DataType.TF_VARIANT)
                 {
 
                 }
@@ -933,7 +933,7 @@ namespace Tensorflow
             string name = "split")
         {
             if (num == -1)
-                num = size_splits.shape[0];
+                num = (int)size_splits.shape[0];
 
             return gen_array_ops.split_v(value, size_splits, axis, num, name: name);
         }
diff --git a/src/TensorFlowNET.Core/Operations/functional_ops.cs b/src/TensorFlowNET.Core/Operations/functional_ops.cs
index 67450c74..003b93da 100644
--- a/src/TensorFlowNET.Core/Operations/functional_ops.cs
+++ b/src/TensorFlowNET.Core/Operations/functional_ops.cs
@@ -91,7 +91,7 @@ namespace Tensorflow
                     elem.dtype,
                     size: tf.constant(n),
                     dynamic_size: false,
-                    element_shape: elem.shape.Skip(1).ToArray(),
+                    element_shape: elem.shape.dims.Skip(1).ToArray(),
                     infer_shape: true)).ToList();
 
                 for (int index = 0; index < elems_ta.Count; index++)
diff --git a/src/TensorFlowNET.Core/Operations/image_ops_impl.cs b/src/TensorFlowNET.Core/Operations/image_ops_impl.cs
index 849a93c8..917dbd6b 100644
--- a/src/TensorFlowNET.Core/Operations/image_ops_impl.cs
+++ b/src/TensorFlowNET.Core/Operations/image_ops_impl.cs
@@ -341,14 +341,14 @@ or rank = 4. Had rank = {0}", rank));
                 {
                     h = _get_dim(image, 0); // img_h == h[0], dynamic_h == h[1]
                     w = _get_dim(image, 1);
-                    d = image.shape[3];
+                    d = (int)image.shape[3];
                 }
                 else
                 {
-                    bs = image.shape[0];
+                    bs = (int)image.shape[0];
                     h = _get_dim(image, 1);
                     w = _get_dim(image, 2);
-                    d = image.shape[3];
+                    d = (int)image.shape[3];
                 }
 
                 object hd, bbox_h_start;
@@ -1115,7 +1115,7 @@ new_height, new_width");
                                  array_ops.expand_dims(tf.constant(3), 0));
                  var multiples = array_ops.concat(new Tensor[] { shape_list }, 0);
                  var rgb = array_ops.tile(images, multiples, name: name);
-                 int[] rgb_temp = images.shape.Take(images.shape.Length - 1).ToArray();
+                 int[] rgb_temp = images.shape.dims.Take(images.shape.ndim - 1).Select(x => (int)x).ToArray();
                  rgb.set_shape(array_ops.concat(new Tensor[] { ops.convert_to_tensor(rgb_temp) }, 3));
                  return rgb;
              });
@@ -1459,7 +1459,7 @@ new_height, new_width");
 
             // shape takes an int, python code passes size, a Tensor. NDims is the only int type
             // i could think of a Tensor having. it might be incorrect tho, so keep that in mind.
-            return array_ops.reshape(g, shape: new int[] { size.NDims, size.NDims, 1, 1 });
+            return array_ops.reshape(g, shape: new int[] { size.ndim, size.ndim, 1, 1 });
         }
 
         internal static (Tensor, Tensor) _ssim_per_channel(Tensor img1, Tensor img2, float max_val = 1f,
@@ -1487,7 +1487,7 @@ new_height, new_width");
                 img1 = array_ops.identity(img1);
 
             var kernel = _fspecial_gauss(filter_size_tensor, filter_sigma_tensor);
-            kernel = array_ops.tile(kernel, multiples: new Tensor(new int[] { 1, 1, shape1_tensor.dims[shape1_tensor.dims.Length - 2], 1 }));
+            kernel = array_ops.tile(kernel, multiples: new Tensor(new int[] { 1, 1, (int)shape1_tensor.dims[shape1_tensor.dims.Length - 2], 1 }));
 
             float compensation = 1.0f;
 
@@ -1503,8 +1503,8 @@ new_height, new_width");
             (Tensor luminance, Tensor cs) = _ssim_helper(img1, img2, reducer, max_val, compensation, k1, k2);
 
             var axes = constant_op.constant(new[] { -3, -2 }, dtype: dtypes.int32);
-            var ssim_val = math_ops.reduce_mean(luminance * cs, new(axes.dims));
-            cs = math_ops.reduce_mean(cs, new(axes.dims));
+            var ssim_val = math_ops.reduce_mean(luminance * cs, axes.dims);
+            cs = math_ops.reduce_mean(cs, axes.dims);
             return (ssim_val, cs);
         }
 
@@ -1685,7 +1685,7 @@ new_height, new_width");
             var kernels_tf = constant_op.constant(kernels, dtype: image.dtype);
 
             kernels_tf = array_ops.tile(
-                kernels_tf, new Tensor(new int[] { 1, 1, image_shape.dims[image_shape.dims.Length - 2], 1 }), name: "sobel_filters");
+                kernels_tf, new Tensor(new int[] { 1, 1, (int)image_shape.dims[image_shape.dims.Length - 2], 1 }), name: "sobel_filters");
 
             var pad_sizes = new int[,] { { 0, 0 }, { 1, 1 }, { 1, 1 }, { 0, 0 } };
             var padded = array_ops.pad(image, new Tensor(pad_sizes), mode: "reflect");
@@ -1966,8 +1966,8 @@ new_height, new_width");
                 Tensor index_offsets, indices, sorted_scores, sorted_boxes, sorted_scores_indices;
                 using (ops.name_scope("sort_scores_and_boxes"))
                 {
-                    batch_size = array_ops.shape(boxes).dims[0];
-                    num_boxes = array_ops.shape(boxes).dims[1];
+                    batch_size = (int)array_ops.shape(boxes).dims[0];
+                    num_boxes = (int)array_ops.shape(boxes).dims[1];
                     sorted_scores_indices = null; /*sort_ops.argsort(
                         scores, axis: 1, direction: "DESCENDING); */
                     index_offsets = math_ops.range(batch_size) * num_boxes;
diff --git a/src/TensorFlowNET.Core/Operations/nn_ops.cs b/src/TensorFlowNET.Core/Operations/nn_ops.cs
index ef50f69f..6d69a55f 100644
--- a/src/TensorFlowNET.Core/Operations/nn_ops.cs
+++ b/src/TensorFlowNET.Core/Operations/nn_ops.cs
@@ -178,7 +178,7 @@ namespace Tensorflow
             logits = ops.convert_to_tensor(logits);
 
             var shape = logits.shape;
-            bool is_last_dim = dim == -1 || dim == shape.Length - 1;
+            bool is_last_dim = dim == -1 || dim == shape.ndim - 1;
             if (is_last_dim)
                 return compute_op(logits, name);
 
diff --git a/src/TensorFlowNET.Core/Tensors/Ragged/RowPartition.cs b/src/TensorFlowNET.Core/Tensors/Ragged/RowPartition.cs
index 6a52397a..b1dbf586 100644
--- a/src/TensorFlowNET.Core/Tensors/Ragged/RowPartition.cs
+++ b/src/TensorFlowNET.Core/Tensors/Ragged/RowPartition.cs
@@ -37,7 +37,7 @@ namespace Tensorflow
         {
             get
             {
-                return _row_splits.shape[0] - 1;
+                return (int)_row_splits.shape[0] - 1;
             }
         }
 
diff --git a/src/TensorFlowNET.Core/Tensors/Tensor.Creation.cs b/src/TensorFlowNET.Core/Tensors/Tensor.Creation.cs
index 1f839ee7..d4419073 100644
--- a/src/TensorFlowNET.Core/Tensors/Tensor.Creation.cs
+++ b/src/TensorFlowNET.Core/Tensors/Tensor.Creation.cs
@@ -145,6 +145,10 @@ namespace Tensorflow
                 byte[,] val => InitTensor(val, shape, dtype),
                 byte[,,] val => InitTensor(val, shape, dtype),
                 byte[,,,] val => InitTensor(val, shape, dtype),
+                short[] val => InitTensor(val, shape, dtype),
+                short[,] val => InitTensor(val, shape, dtype),
+                short[,,] val => InitTensor(val, shape, dtype),
+                short[,,,] val => InitTensor(val, shape, dtype),
                 int[] val => InitTensor(val, shape, dtype),
                 int[,] val => InitTensor(val, shape, dtype),
                 int[,,] val => InitTensor(val, shape, dtype),
@@ -153,6 +157,10 @@ namespace Tensorflow
                 long[,] val => InitTensor(val, shape, dtype),
                 long[,,] val => InitTensor(val, shape, dtype),
                 long[,,,] val => InitTensor(val, shape, dtype),
+                ulong[] val => InitTensor(val, shape, dtype),
+                ulong[,] val => InitTensor(val, shape, dtype),
+                ulong[,,] val => InitTensor(val, shape, dtype),
+                ulong[,,,] val => InitTensor(val, shape, dtype),
                 float[] val => InitTensor(val, shape, dtype),
                 float[,] val => InitTensor(val, shape, dtype),
                 float[,,] val => InitTensor(val, shape, dtype),
diff --git a/src/TensorFlowNET.Core/Tensors/Tensor.Value.cs b/src/TensorFlowNET.Core/Tensors/Tensor.Value.cs
index ed72d9aa..dd7b8ad6 100644
--- a/src/TensorFlowNET.Core/Tensors/Tensor.Value.cs
+++ b/src/TensorFlowNET.Core/Tensors/Tensor.Value.cs
@@ -18,7 +18,7 @@ namespace Tensorflow
             if (typeof(T).as_tf_dtype() != dtype)
                 throw new ArrayTypeMismatchException($"dtype {dtype} mismatch.");
 
-            if (NDims == 0 && size == 1)  //is it a scalar?
+            if (ndim == 0 && size == 1)  //is it a scalar?
             {
                 unsafe
                 {
@@ -28,7 +28,7 @@ namespace Tensorflow
 
             //types match, no need to perform cast
             var ret = new T[size];
-            var len = (long)(size * itemsize);
+            var len = (long)(size * dtypesize);
             var src = (T*)buffer;
 
             fixed (T* dst = ret)
diff --git a/src/TensorFlowNET.Core/Tensors/Tensor.cs b/src/TensorFlowNET.Core/Tensors/Tensor.cs
index 5166cf81..bf8089de 100644
--- a/src/TensorFlowNET.Core/Tensors/Tensor.cs
+++ b/src/TensorFlowNET.Core/Tensors/Tensor.cs
@@ -72,17 +72,17 @@ namespace Tensorflow
         /// </summary>
         public TF_DataType dtype => _handle == IntPtr.Zero ? _override_dtype : c_api.TF_TensorType(_handle);
         public ulong bytesize => _handle == IntPtr.Zero ? 0 : c_api.TF_TensorByteSize(_handle);
-        public ulong itemsize => _handle == IntPtr.Zero ? 0 : c_api.TF_DataTypeSize(dtype);
-        public ulong size => _handle == IntPtr.Zero ? 0 : bytesize / itemsize;
+        public ulong dtypesize => _handle == IntPtr.Zero ? 0 : c_api.TF_DataTypeSize(dtype);
+        public ulong size => _handle == IntPtr.Zero ? 0 : bytesize / dtypesize;
         public IntPtr buffer => _handle == IntPtr.Zero ? IntPtr.Zero : c_api.TF_TensorData(_handle);
         public int num_consumers(TF_Output oper_out) => _handle == IntPtr.Zero ? 0 : c_api.TF_OperationOutputNumConsumers(oper_out);
-        public int NDims => rank;
+        public int ndim => rank;
 
         /// <summary>
         ///     The name of the device on which this tensor will be produced, or null.
         /// </summary>
         public virtual string Device => op.Device;
-        public int[] dims => shape;
+        public long[] dims => shape.dims;
 
         /// <summary>
         ///     Used for keep other pointer when do implicit operating
@@ -107,7 +107,7 @@ namespace Tensorflow
         ///     Returns the shape of a tensor.
         /// </summary>
         /// <remarks>https://www.tensorflow.org/api_docs/python/tf/shape</remarks>
-        public int[] shape
+        public Shape shape
         {
             get
             {
@@ -123,7 +123,7 @@ namespace Tensorflow
                         dims[i] = c_api.TF_Dim(_handle, i);
                 }
 
-                return dims.Select(x => ((IConvertible)x).ToInt32(CultureInfo.InvariantCulture)).ToArray();
+                return dims;
             }
 
             set
@@ -131,7 +131,7 @@ namespace Tensorflow
                 if (value == null)
                     c_api.TF_GraphSetTensorShape(graph, _as_tf_output(), null, -1, tf.Status.Handle);
                 else
-                    c_api.TF_GraphSetTensorShape(graph, _as_tf_output(), value.Select(Convert.ToInt64).ToArray(), value.Length, tf.Status.Handle);
+                    c_api.TF_GraphSetTensorShape(graph, _as_tf_output(), value.dims, value.ndim, tf.Status.Handle);
 
                 tf.Status.Check(true);
             }
@@ -139,10 +139,10 @@ namespace Tensorflow
 
         public int[] _shape_tuple()
         {
-            return rank < 0 ? null : shape;
+            return rank < 0 ? null : shape.dims.Select(x => (int)x).ToArray();
         }
 
-        public TensorShape TensorShape => rank < 0 ? new TensorShape() : tensor_util.to_shape(shape);
+        public TensorShape TensorShape => rank < 0 ? new TensorShape() : shape;
 
         /// <summary>
         /// Keras History: (Layer, (node_index, tensor_index))
diff --git a/src/TensorFlowNET.Core/Tensors/c_api.tensor.cs b/src/TensorFlowNET.Core/Tensors/c_api.tensor.cs
index 66b5fd3b..5917439e 100644
--- a/src/TensorFlowNET.Core/Tensors/c_api.tensor.cs
+++ b/src/TensorFlowNET.Core/Tensors/c_api.tensor.cs
@@ -109,7 +109,8 @@ namespace Tensorflow
             var length = shape.size * (ulong)dtype.get_datatype_size();
             var handle = TF_AllocateTensor(dtype, shape.dims, shape.ndim, length);
             var tensor = TF_TensorData(handle);
-            System.Buffer.MemoryCopy(data, tensor.ToPointer(), length, length);
+            if (tensor != IntPtr.Zero)
+                System.Buffer.MemoryCopy(data, tensor.ToPointer(), length, length);
             return handle;
         }
 
diff --git a/src/TensorFlowNET.Core/Tensors/constant_op.cs b/src/TensorFlowNET.Core/Tensors/constant_op.cs
index cf6c76a2..b69c4477 100644
--- a/src/TensorFlowNET.Core/Tensors/constant_op.cs
+++ b/src/TensorFlowNET.Core/Tensors/constant_op.cs
@@ -124,6 +124,8 @@ namespace Tensorflow
                     return new EagerTensor(new[] { val }, Shape.Scalar);
                 case long val:
                     return new EagerTensor(new[] { val }, Shape.Scalar);
+                case ulong val:
+                    return new EagerTensor(new[] { val }, Shape.Scalar);
                 case float val:
                     return new EagerTensor(new[] { val }, Shape.Scalar);
                 case double val:
@@ -146,7 +148,7 @@ namespace Tensorflow
             if (shape == null)
                 return t;
 
-            if (t.shape.Select(x => Convert.ToInt64(x)).SequenceEqual(shape.dims))
+            if (t.shape.dims.SequenceEqual(shape.dims))
                 return t;
 
             if (verify_shape)
diff --git a/src/TensorFlowNET.Core/Tensors/tensor_util.cs b/src/TensorFlowNET.Core/Tensors/tensor_util.cs
index 5ad8bc9b..5a007695 100644
--- a/src/TensorFlowNET.Core/Tensors/tensor_util.cs
+++ b/src/TensorFlowNET.Core/Tensors/tensor_util.cs
@@ -127,7 +127,7 @@ namespace Tensorflow
             }
             else if (values is Tensor tensor && tensor.IsReferencedByNDArray)
             {
-                var len = tensor.itemsize * tensor.size;
+                var len = tensor.dtypesize * tensor.size;
                 byte[] bytes = tensor.BufferToArray();
                 tensor_proto.TensorContent = Google.Protobuf.ByteString.CopyFrom(bytes);
             }
diff --git a/src/TensorFlowNET.Core/Training/Saving/SaveableObject.cs b/src/TensorFlowNET.Core/Training/Saving/SaveableObject.cs
index 960bf656..60de456f 100644
--- a/src/TensorFlowNET.Core/Training/Saving/SaveableObject.cs
+++ b/src/TensorFlowNET.Core/Training/Saving/SaveableObject.cs
@@ -45,7 +45,7 @@ namespace Tensorflow
             var restored_tensor = restored_tensors[0];
             return gen_state_ops.assign(op,
                 restored_tensor,
-                validate_shape: restored_shapes == null && tensor_util.to_shape(op.shape).is_fully_defined());
+                validate_shape: restored_shapes == null && op.shape.is_fully_defined());
         }
     }
 }
diff --git a/src/TensorFlowNET.Core/Variables/RefVariable.cs b/src/TensorFlowNET.Core/Variables/RefVariable.cs
index 3bf4f784..36fdfed2 100644
--- a/src/TensorFlowNET.Core/Variables/RefVariable.cs
+++ b/src/TensorFlowNET.Core/Variables/RefVariable.cs
@@ -50,7 +50,7 @@ namespace Tensorflow
         public Operation Op => _variable.op;
 
         public TF_DataType dtype => _variable.dtype;
-        public TensorShape shape => tensor_util.to_shape(_variable.shape);
+        public TensorShape shape => _variable.shape;
         public string Device => "";
 
         public string Name => _variable.name;
diff --git a/src/TensorFlowNET.Keras/BackendImpl.cs b/src/TensorFlowNET.Keras/BackendImpl.cs
index bedca2c1..e1563055 100644
--- a/src/TensorFlowNET.Keras/BackendImpl.cs
+++ b/src/TensorFlowNET.Keras/BackendImpl.cs
@@ -297,8 +297,8 @@ namespace Tensorflow.Keras
                 // x = permute_dimensions(x, [0, 3, 1, 2]);
                 throw new NotImplementedException("");
 
-            int new_height = original_shape[rows] < 0 ? -1 : original_shape[rows] * height_factor;
-            int new_width = original_shape[cols] < 0 ? -1 : original_shape[cols] * width_factor;
+            int new_height = original_shape[rows] < 0 ? -1 : (int)original_shape[rows] * height_factor;
+            int new_width = original_shape[cols] < 0 ? -1 : (int)original_shape[cols] * width_factor;
 
             TensorShape output_shape = data_format == "channels_first" ?
                 (-1, -1, new_height, new_width) : (-1, new_height, new_width, -1);
@@ -316,7 +316,7 @@ namespace Tensorflow.Keras
         {
             if(axis < 0)
             {
-                var rank = tensors[0].NDims;
+                var rank = tensors[0].ndim;
                 if (rank > -1)
                     axis += rank;
                 else
diff --git a/src/TensorFlowNET.Keras/Engine/DataAdapters/TensorLikeDataAdapter.cs b/src/TensorFlowNET.Keras/Engine/DataAdapters/TensorLikeDataAdapter.cs
index d73dc8b1..fc61aa71 100644
--- a/src/TensorFlowNET.Keras/Engine/DataAdapters/TensorLikeDataAdapter.cs
+++ b/src/TensorFlowNET.Keras/Engine/DataAdapters/TensorLikeDataAdapter.cs
@@ -20,7 +20,7 @@ namespace Tensorflow.Keras.Engine.DataAdapters
         {
             this.args = args;
             _process_tensorlike();
-            num_samples = args.X.shape[0];
+            num_samples = (int)args.X.shape[0];
             var batch_size = args.BatchSize == -1 ? 32 : args.BatchSize;
             _batch_size = batch_size;
             _size = Convert.ToInt32(Math.Ceiling(num_samples / (batch_size + 0.0f)));
diff --git a/src/TensorFlowNET.Keras/Engine/MetricsContainer.cs b/src/TensorFlowNET.Keras/Engine/MetricsContainer.cs
index 6fed2bf3..037703c8 100644
--- a/src/TensorFlowNET.Keras/Engine/MetricsContainer.cs
+++ b/src/TensorFlowNET.Keras/Engine/MetricsContainer.cs
@@ -63,8 +63,8 @@ namespace Tensorflow.Keras.Engine
             {
                 var y_t_rank = y_t.rank;
                 var y_p_rank = y_p.rank;
-                var y_t_last_dim = y_t.shape[y_t.shape.Length - 1];
-                var y_p_last_dim = y_p.shape[y_p.shape.Length - 1];
+                var y_t_last_dim = y_t.shape[y_t.shape.ndim - 1];
+                var y_p_last_dim = y_p.shape[y_p.shape.ndim - 1];
 
                 bool is_binary = y_p_last_dim == 1;
                 bool is_sparse_categorical = (y_t_rank < y_p_rank || y_t_last_dim == 1) && y_p_last_dim > 1;
diff --git a/src/TensorFlowNET.Keras/Metrics/MetricsApi.cs b/src/TensorFlowNET.Keras/Metrics/MetricsApi.cs
index 64723a22..592d2568 100644
--- a/src/TensorFlowNET.Keras/Metrics/MetricsApi.cs
+++ b/src/TensorFlowNET.Keras/Metrics/MetricsApi.cs
@@ -29,7 +29,7 @@ namespace Tensorflow.Keras.Metrics
             var y_true_rank = y_true.TensorShape.ndim;
             // If the shape of y_true is (num_samples, 1), squeeze to (num_samples,)
             if (y_true_rank != -1 && y_pred_rank != -1
-                && y_true.shape.Length == y_pred.shape.Length)
+                && y_true.shape.ndim == y_pred.shape.ndim)
                 y_true = array_ops.squeeze(y_true, axis: new[] { -1 });
             y_pred = math_ops.argmax(y_pred, -1);
 
diff --git a/src/TensorFlowNET.Keras/tf.layers.cs b/src/TensorFlowNET.Keras/tf.layers.cs
index b69bbe95..3f5ed01c 100644
--- a/src/TensorFlowNET.Keras/tf.layers.cs
+++ b/src/TensorFlowNET.Keras/tf.layers.cs
@@ -212,13 +212,13 @@ namespace Tensorflow.Keras
                 string data_format = "channels_last")
             {
                 var input_shape = inputs.shape;
-                if (inputs.shape.Length == 0)
+                if (inputs.shape.ndim == 0)
                     throw new ValueError($"Input 0 of layer flatten is incompatible with the layer: : expected min_ndim={1}, found ndim={0}. Full shape received: ()");
 
                 var premutation = new List<int>() { 0 };
-                if (data_format == "channels_first" && inputs.NDims > 1)
+                if (data_format == "channels_first" && inputs.ndim > 1)
                 {
-                    premutation.AddRange(Binding.range(2, inputs.NDims));
+                    premutation.AddRange(Binding.range(2, inputs.ndim));
                     premutation.Add(1);
                     inputs = array_ops.transpose(inputs, premutation.ToArray());
                 }
diff --git a/test/TensorFlowNET.Native.UnitTest/Sessions/SessionTest.cs b/test/TensorFlowNET.Native.UnitTest/Sessions/SessionTest.cs
index d9e4e872..b1fe18b4 100644
--- a/test/TensorFlowNET.Native.UnitTest/Sessions/SessionTest.cs
+++ b/test/TensorFlowNET.Native.UnitTest/Sessions/SessionTest.cs
@@ -40,7 +40,7 @@ namespace Tensorflow.Native.UnitTest.Sessions
             csession.Run(s);
             Tensor outTensor = csession.output_tensor(0);
             EXPECT_EQ(TF_DataType.TF_INT32, outTensor.dtype);
-            EXPECT_EQ(0, outTensor.NDims);
+            EXPECT_EQ(0, outTensor.ndim);
             ASSERT_EQ((ulong)sizeof(uint), outTensor.bytesize);
             var output_contents = outTensor.ToArray<int>();
             EXPECT_EQ(3 + 2, output_contents[0]);
@@ -61,7 +61,7 @@ namespace Tensorflow.Native.UnitTest.Sessions
             outTensor = csession.output_tensor(0);
             ASSERT_TRUE(outTensor != IntPtr.Zero);
             EXPECT_EQ(TF_DataType.TF_INT32, outTensor.dtype);
-            EXPECT_EQ(0, outTensor.NDims); // scalar
+            EXPECT_EQ(0, outTensor.ndim); // scalar
             ASSERT_EQ((ulong)sizeof(uint), outTensor.bytesize);
             output_contents = outTensor.ToArray<int>();
             EXPECT_EQ(-(7 + 2), output_contents[0]);
diff --git a/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs b/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs
index dc588a1a..76ebf209 100644
--- a/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs
+++ b/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs
@@ -66,7 +66,7 @@ namespace Tensorflow.Native.UnitTest.Tensors
             long[] dims = { 2, 3 };
             Tensor t = c_api.TF_AllocateTensor(TF_DataType.TF_FLOAT, dims, 2, num_bytes);
             EXPECT_EQ(TF_DataType.TF_FLOAT, t.dtype);
-            EXPECT_EQ(2, t.NDims);
+            EXPECT_EQ(2, t.ndim);
             EXPECT_EQ((int)dims[0], t.shape[0]);
             EXPECT_EQ(num_bytes, t.bytesize);
             t.Dispose();
diff --git a/test/TensorFlowNET.UnitTest/Basics/VariableTest.cs b/test/TensorFlowNET.UnitTest/Basics/VariableTest.cs
index 74b1bb03..1b55508b 100644
--- a/test/TensorFlowNET.UnitTest/Basics/VariableTest.cs
+++ b/test/TensorFlowNET.UnitTest/Basics/VariableTest.cs
@@ -126,7 +126,7 @@ namespace TensorFlowNET.UnitTest.Basics
         {
             var x = tf.constant(new[,] { { 1, 2 } });
             var neg_x = tf.negative(x);
-            Assert.IsTrue(Enumerable.SequenceEqual(new[] { 1, 2 }, neg_x.shape));
+            Assert.IsTrue(Enumerable.SequenceEqual(new long[] { 1, 2 }, neg_x.shape.dims));
             Assert.IsTrue(Enumerable.SequenceEqual(new[] { -1, -2 }, neg_x.numpy().ToArray<int>()));
         }
 
diff --git a/test/TensorFlowNET.UnitTest/ManagedAPI/ConstantTest.cs b/test/TensorFlowNET.UnitTest/ManagedAPI/ConstantTest.cs
index 8b2260a3..2062dbc3 100644
--- a/test/TensorFlowNET.UnitTest/ManagedAPI/ConstantTest.cs
+++ b/test/TensorFlowNET.UnitTest/ManagedAPI/ConstantTest.cs
@@ -145,7 +145,7 @@ namespace TensorFlowNET.UnitTest.Basics
             var tensor = tf.constant(nd);
             var data = tensor.numpy().ToArray<int>();
 
-            Assert.IsTrue(Enumerable.SequenceEqual(new int[] { 2, 3 }, tensor.shape));
+            Assert.IsTrue(Enumerable.SequenceEqual(new long[] { 2, 3 }, tensor.shape.dims));
             Assert.IsTrue(Enumerable.SequenceEqual(new int[] { 3, 1, 1, 2, 1, 3 }, data));
         }
 
diff --git a/test/TensorFlowNET.UnitTest/ManagedAPI/GradientTest.cs b/test/TensorFlowNET.UnitTest/ManagedAPI/GradientTest.cs
index 0bf506da..902bcdbf 100644
--- a/test/TensorFlowNET.UnitTest/ManagedAPI/GradientTest.cs
+++ b/test/TensorFlowNET.UnitTest/ManagedAPI/GradientTest.cs
@@ -57,7 +57,7 @@ namespace TensorFlowNET.UnitTest.ManagedAPI
             var b = tf.Variable(-0.73f, name: "bias");
             using var g = tf.GradientTape();
             var pred = W * X + b;
-            var test = tf.slice(pred, new[] { 0 }, pred.shape);
+            var test = tf.slice(pred, new[] { 0 }, (int[])pred.shape);
             var gradients = g.gradient(test, (W, b));
             Assert.AreEqual((float)gradients.Item1, 0f);
             Assert.AreEqual((float)gradients.Item2, 10f);
diff --git a/test/TensorFlowNET.UnitTest/ManagedAPI/TensorOperate.cs b/test/TensorFlowNET.UnitTest/ManagedAPI/TensorOperate.cs
index cdc8b51c..8f38f45c 100644
--- a/test/TensorFlowNET.UnitTest/ManagedAPI/TensorOperate.cs
+++ b/test/TensorFlowNET.UnitTest/ManagedAPI/TensorOperate.cs
@@ -85,14 +85,14 @@ namespace TensorFlowNET.UnitTest.ManagedAPI
                 { { 1 }, { 2 }, { 3 } },
                 { { 4 }, { 5 }, { 6 } }
             }));
-            Assert.IsTrue(Enumerable.SequenceEqual(new[] { 2, 3, 1 }, a.shape));
+            Assert.IsTrue(Enumerable.SequenceEqual(new long[] { 2, 3, 1 }, a.shape.dims));
 
             var b = tf.constant(new[, ,]
             {
                 { { 1 }, { 2 }, { 3 } },
                 { { 4 }, { 5 }, { 6 } }
             });
-            Assert.IsTrue(Enumerable.SequenceEqual(new[] { 2, 3, 1 }, b.shape));
+            Assert.IsTrue(Enumerable.SequenceEqual(new long[] { 2, 3, 1 }, b.shape.dims));
         }
 
         [TestMethod]
@@ -103,7 +103,7 @@ namespace TensorFlowNET.UnitTest.ManagedAPI
             var c = tf.constant(new[,] { { 9, 10 }, { 11, 12 } });
 
             var concatValue = tf.concat(new[] { a, b, c }, axis: 0);
-            Assert.IsTrue(Enumerable.SequenceEqual(new[] { 6, 2 }, concatValue.shape));
+            Assert.IsTrue(Enumerable.SequenceEqual(new long[] { 6, 2 }, concatValue.shape.dims));
         }
 
         [TestMethod]
@@ -114,7 +114,7 @@ namespace TensorFlowNET.UnitTest.ManagedAPI
             var c = tf.constant(new[,] { { 9.0, 10.0 }, { 11.0, 12.0 } });
 
             var concatValue = tf.concat(new[] { a, b, c }, axis: 0);
-            Assert.IsTrue(Enumerable.SequenceEqual(new[] { 6, 2 }, concatValue.shape));
+            Assert.IsTrue(Enumerable.SequenceEqual(new long[] { 6, 2 }, concatValue.shape.dims));
         }
 
         [TestMethod]
@@ -128,7 +128,7 @@ namespace TensorFlowNET.UnitTest.ManagedAPI
 
             var splitValue = tf.split(value, 3, axis: 0);
             Assert.AreEqual(3, splitValue.Length);
-            Assert.IsTrue(Enumerable.SequenceEqual(new[] { 2, 2 }, splitValue[0].shape));
+            Assert.IsTrue(Enumerable.SequenceEqual(new long[] { 2, 2 }, splitValue[0].shape.dims));
         }
 
         #region ones/zeros like