diff --git a/src/TensorFlowNET.Core/Binding.Util.cs b/src/TensorFlowNET.Core/Binding.Util.cs
index b7aca46b..f817beb4 100644
--- a/src/TensorFlowNET.Core/Binding.Util.cs
+++ b/src/TensorFlowNET.Core/Binding.Util.cs
@@ -276,8 +276,8 @@ namespace Tensorflow
         {
             if (axis == null)
             {
-                var a = t1.Data<T>();
-                var b = t2.Data<T>();
+                var a = t1.ToArray<T>();
+                var b = t2.ToArray<T>();
                 for (int i = 0; i < a.Length; i++)
                     yield return (a[i], b[i]);
             }
diff --git a/src/TensorFlowNET.Core/Data/MnistModelLoader.cs b/src/TensorFlowNET.Core/Data/MnistModelLoader.cs
index e7251217..1073dac6 100644
--- a/src/TensorFlowNET.Core/Data/MnistModelLoader.cs
+++ b/src/TensorFlowNET.Core/Data/MnistModelLoader.cs
@@ -162,7 +162,7 @@ namespace Tensorflow
             var num_labels = labels_dense.dims[0];
             var index_offset = np.arange(num_labels) * num_classes;
             var labels_one_hot = np.zeros((num_labels, num_classes));
-            var labels = labels_dense.Data<byte>();
+            var labels = labels_dense.ToArray<byte>();
             for (int row = 0; row < num_labels; row++)
             {
                 var col = labels[row];
@@ -176,7 +176,7 @@ namespace Tensorflow
         {
             var buffer = new byte[sizeof(uint)];
             var count = bytestream.Read(buffer, 0, 4);
-            return np.frombuffer(buffer, ">u4").Data<int>()[0];
+            return np.frombuffer(buffer, ">u4").ToArray<int>()[0];
         }
     }
 }
diff --git a/src/TensorFlowNET.Core/Gradients/math_grad.cs b/src/TensorFlowNET.Core/Gradients/math_grad.cs
index d927b85e..60408bab 100644
--- a/src/TensorFlowNET.Core/Gradients/math_grad.cs
+++ b/src/TensorFlowNET.Core/Gradients/math_grad.cs
@@ -506,7 +506,7 @@ namespace Tensorflow.Gradients
                 if (!(axes is null))
                 {
                     var rank = input_0_shape.Length;
-                    if (Enumerable.SequenceEqual(Enumerable.Range(0, rank), axes.Data<int>()))
+                    if (Enumerable.SequenceEqual(Enumerable.Range(0, rank), axes.ToArray<int>()))
                     {
                         if (tf.Context.executing_eagerly())
                         {
diff --git a/src/TensorFlowNET.Core/Numpy/NDArray.cs b/src/TensorFlowNET.Core/Numpy/NDArray.cs
index 7293093e..9a4e269f 100644
--- a/src/TensorFlowNET.Core/Numpy/NDArray.cs
+++ b/src/TensorFlowNET.Core/Numpy/NDArray.cs
@@ -62,13 +62,9 @@ namespace Tensorflow.NumPy
         public byte[] ToByteArray() => _tensor.BufferToArray();
         public static string[] AsStringArray(NDArray arr) => throw new NotImplementedException("");
 
-        public T[] Data<T>() where T : unmanaged
-            => _tensor.ToArray<T>();
         public T[] ToArray<T>() where T : unmanaged
             => _tensor.ToArray<T>();
 
-        public static NDArray operator /(NDArray x, NDArray y) => throw new NotImplementedException("");
-
         public override string ToString()
         {
             return tensor_util.to_numpy_string(_tensor);
diff --git a/src/TensorFlowNET.Core/Tensors/tensor_util.cs b/src/TensorFlowNET.Core/Tensors/tensor_util.cs
index 25ca9119..4c6766ee 100644
--- a/src/TensorFlowNET.Core/Tensors/tensor_util.cs
+++ b/src/TensorFlowNET.Core/Tensors/tensor_util.cs
@@ -257,7 +257,7 @@ scalar with value '-1' to describe an unknown shape.", value_));
                 var dest_dtype_shape_array = np.array(x_).astype(cast_dtype.as_system_dtype());
 
                 long[] y_ = { };
-                foreach (int y in dest_dtype_shape_array.Data<int>())
+                foreach (int y in dest_dtype_shape_array.ToArray<int>())
                     if (y >= 0)
                         y_[y_.Length] = y;
                     else
diff --git a/src/TensorFlowNET.Keras/Sequence.cs b/src/TensorFlowNET.Keras/Sequence.cs
index 4e1ac24b..cda3f30f 100644
--- a/src/TensorFlowNET.Keras/Sequence.cs
+++ b/src/TensorFlowNET.Keras/Sequence.cs
@@ -55,7 +55,7 @@ namespace Tensorflow.Keras
                 value = 0f;
 
             var type = dtypes.tf_dtype_from_name(dtype);
-            var nd = new NDArray((length.Count(), maxlen.Value), dtype: type);
+            var nd = np.zeros((length.Count(), maxlen.Value), dtype: type);
 
             for (int i = 0; i < nd.dims[0]; i++)
             {
diff --git a/test/TensorFlowNET.Keras.UnitTest/PreprocessingTests.cs b/test/TensorFlowNET.Keras.UnitTest/PreprocessingTests.cs
index 2bfb632a..9f2d3316 100644
--- a/test/TensorFlowNET.Keras.UnitTest/PreprocessingTests.cs
+++ b/test/TensorFlowNET.Keras.UnitTest/PreprocessingTests.cs
@@ -249,7 +249,7 @@ namespace TensorFlowNET.Keras.UnitTest
                 Assert.AreNotEqual(padded[1, i], 0);
         }
 
-        [TestMethod, Ignore("slice assign doesn't work")]
+        [TestMethod]
         public void PadSequencesPrePaddingTrunc()
         {
             var tokenizer = keras.preprocessing.text.Tokenizer(oov_token: OOV);
@@ -261,15 +261,15 @@ namespace TensorFlowNET.Keras.UnitTest
             Assert.AreEqual(4, padded.dims[0]);
             Assert.AreEqual(15, padded.dims[1]);
 
-            Assert.AreEqual(tokenizer.word_index["worst"], padded[0, 12]);
+            Assert.AreEqual(padded[0, 12], tokenizer.word_index["worst"]);
             for (var i = 0; i < 3; i++)
-                Assert.AreEqual(0, padded[0, i]);
-            Assert.AreEqual(tokenizer.word_index["proud"], padded[1, 3]);
+                Assert.AreEqual(padded[0, i], 0);
+            Assert.AreEqual(padded[1, 3], tokenizer.word_index["proud"]);
             for (var i = 0; i < 15; i++)
-                Assert.AreNotEqual(0, padded[1, i]);
+                Assert.AreNotEqual(padded[1, i], 0);
         }
 
-        [TestMethod, Ignore("slice assign doesn't work")]
+        [TestMethod]
         public void PadSequencesPrePaddingTrunc_Larger()
         {
             var tokenizer = keras.preprocessing.text.Tokenizer(oov_token: OOV);
@@ -281,13 +281,13 @@ namespace TensorFlowNET.Keras.UnitTest
             Assert.AreEqual(4, padded.dims[0]);
             Assert.AreEqual(45, padded.dims[1]);
 
-            Assert.AreEqual(tokenizer.word_index["worst"], padded[0, 42]);
+            Assert.AreEqual(padded[0, 42], tokenizer.word_index["worst"]);
             for (var i = 0; i < 33; i++)
-                Assert.AreEqual(0, padded[0, i]);
-            Assert.AreEqual(tokenizer.word_index["proud"], padded[1, 33]);
+                Assert.AreEqual(padded[0, i], 0);
+            Assert.AreEqual(padded[1, 33], tokenizer.word_index["proud"]);
         }
 
-        [TestMethod, Ignore("slice assign doesn't work")]
+        [TestMethod]
         public void PadSequencesPostPaddingTrunc()
         {
             var tokenizer = keras.preprocessing.text.Tokenizer(oov_token: OOV);
@@ -299,15 +299,15 @@ namespace TensorFlowNET.Keras.UnitTest
             Assert.AreEqual(4, padded.dims[0]);
             Assert.AreEqual(15, padded.dims[1]);
 
-            Assert.AreEqual(tokenizer.word_index["worst"], padded[0, 9]);
+            Assert.AreEqual(padded[0, 9], tokenizer.word_index["worst"]);
             for (var i = 12; i < 15; i++)
-                Assert.AreEqual(0, padded[0, i]);
-            Assert.AreEqual(tokenizer.word_index["proud"], padded[1, 10]);
+                Assert.AreEqual(padded[0, i], 0);
+            Assert.AreEqual(padded[1, 10], tokenizer.word_index["proud"]);
             for (var i = 0; i < 15; i++)
-                Assert.AreNotEqual(0, padded[1, i]);
+                Assert.AreNotEqual(padded[1, i], 0);
         }
 
-        [TestMethod, Ignore("slice assign doesn't work")]
+        [TestMethod]
         public void PadSequencesPostPaddingTrunc_Larger()
         {
             var tokenizer = keras.preprocessing.text.Tokenizer(oov_token: OOV);
@@ -319,10 +319,10 @@ namespace TensorFlowNET.Keras.UnitTest
             Assert.AreEqual(4, padded.dims[0]);
             Assert.AreEqual(45, padded.dims[1]);
 
-            Assert.AreEqual(tokenizer.word_index["worst"], padded[0, 9]);
+            Assert.AreEqual(padded[0, 9], tokenizer.word_index["worst"]);
             for (var i = 32; i < 45; i++)
-                Assert.AreEqual(0, padded[0, i]);
-            Assert.AreEqual(tokenizer.word_index["proud"], padded[1, 10]);
+                Assert.AreEqual(padded[0, i], 0);
+            Assert.AreEqual(padded[1, 10], tokenizer.word_index["proud"]);
         }
 
         [TestMethod]
diff --git a/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs b/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs
index 85a51126..c862a6c0 100644
--- a/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs
+++ b/test/TensorFlowNET.Native.UnitTest/Tensors/TensorTest.cs
@@ -104,7 +104,7 @@ namespace Tensorflow.Native.UnitTest.Tensors
             EXPECT_EQ(tensor.shape[0], nd.dims[0]);
             EXPECT_EQ(tensor.shape[1], nd.dims[1]);
             EXPECT_EQ(tensor.bytesize, nd.size * sizeof(float));
-            Assert.IsTrue(Enumerable.SequenceEqual(nd.Data<float>(), new float[] { 1, 2, 3, 4, 5, 6 }));
+            Assert.IsTrue(Enumerable.SequenceEqual(nd.ToArray<float>(), new float[] { 1, 2, 3, 4, 5, 6 }));
         }
 
         /// <summary>
diff --git a/test/TensorFlowNET.UnitTest/Numpy/Array.Creation.Test.cs b/test/TensorFlowNET.UnitTest/Numpy/Array.Creation.Test.cs
index f58ed389..f0d6cbef 100644
--- a/test/TensorFlowNET.UnitTest/Numpy/Array.Creation.Test.cs
+++ b/test/TensorFlowNET.UnitTest/Numpy/Array.Creation.Test.cs
@@ -26,43 +26,43 @@ namespace TensorFlowNET.UnitTest.NumPy
         public void arange()
         {
             var x = np.arange(3);
-            AssetSequenceEqual(new[] { 0, 1, 2 }, x.Data<int>());
+            AssetSequenceEqual(new[] { 0, 1, 2 }, x.ToArray<int>());
 
             x = np.arange(3f);
-            Assert.IsTrue(Equal(new float[] { 0, 1, 2 }, x.Data<float>()));
+            Assert.IsTrue(Equal(new float[] { 0, 1, 2 }, x.ToArray<float>()));
 
             var y = np.arange(3, 7);
-            AssetSequenceEqual(new[] { 3, 4, 5, 6 }, y.Data<int>());
+            AssetSequenceEqual(new[] { 3, 4, 5, 6 }, y.ToArray<int>());
 
             y = np.arange(3, 7, 2);
-            AssetSequenceEqual(new[] { 3, 5 }, y.Data<int>());
+            AssetSequenceEqual(new[] { 3, 5 }, y.ToArray<int>());
         }
 
         [TestMethod]
         public void array()
         {
             var x = np.array(1, 2, 3);
-            AssetSequenceEqual(new[] { 1, 2, 3 }, x.Data<int>());
+            AssetSequenceEqual(new[] { 1, 2, 3 }, x.ToArray<int>());
 
             x = np.array(new[,] { { 1, 2 }, { 3, 4 }, { 5, 6 } });
-            AssetSequenceEqual(new[] { 1, 2, 3, 4, 5, 6 }, x.Data<int>());
+            AssetSequenceEqual(new[] { 1, 2, 3, 4, 5, 6 }, x.ToArray<int>());
         }
 
         [TestMethod]
         public void eye()
         {
             var x = np.eye(3, k: 1);
-            Assert.IsTrue(Equal(new double[] { 0, 1, 0, 0, 0, 1, 0, 0, 0 }, x.Data<double>()));
+            Assert.IsTrue(Equal(new double[] { 0, 1, 0, 0, 0, 1, 0, 0, 0 }, x.ToArray<double>()));
         }
 
         [TestMethod]
         public void linspace()
         {
             var x = np.linspace(2.0, 3.0, num: 5);
-            Assert.IsTrue(Equal(new double[] { 2, 2.25, 2.5, 2.75, 3 }, x.Data<double>()));
+            Assert.IsTrue(Equal(new double[] { 2, 2.25, 2.5, 2.75, 3 }, x.ToArray<double>()));
 
             x = np.linspace(2.0, 3.0, num: 5, endpoint: false);
-            Assert.IsTrue(Equal(new double[] { 2, 2.2, 2.4, 2.6, 2.8 }, x.Data<double>()));
+            Assert.IsTrue(Equal(new double[] { 2, 2.2, 2.4, 2.6, 2.8 }, x.ToArray<double>()));
         }
 
         [TestMethod]
@@ -71,13 +71,13 @@ namespace TensorFlowNET.UnitTest.NumPy
             var x = np.linspace(0, 1, num: 3);
             var y = np.linspace(0, 1, num: 2);
             var (xv, yv) = np.meshgrid(x, y);
-            Assert.IsTrue(Equal(new double[] { 0, 0.5, 1, 0, 0.5, 1 }, xv.Data<double>()));
-            Assert.IsTrue(Equal(new double[] { 0, 0, 0, 1, 1, 1 }, yv.Data<double>()));
+            Assert.IsTrue(Equal(new double[] { 0, 0.5, 1, 0, 0.5, 1 }, xv.ToArray<double>()));
+            Assert.IsTrue(Equal(new double[] { 0, 0, 0, 1, 1, 1 }, yv.ToArray<double>()));
 
             (xv, yv) = np.meshgrid(x, y, sparse: true);
-            Assert.IsTrue(Equal(new double[] { 0, 0.5, 1 }, xv.Data<double>()));
+            Assert.IsTrue(Equal(new double[] { 0, 0.5, 1 }, xv.ToArray<double>()));
             AssetSequenceEqual(new long[] { 1, 3 }, xv.shape.dims);
-            Assert.IsTrue(Equal(new double[] { 0, 1 }, yv.Data<double>()));
+            Assert.IsTrue(Equal(new double[] { 0, 1 }, yv.ToArray<double>()));
             AssetSequenceEqual(new long[] { 2, 1 }, yv.shape.dims);
         }
     }
diff --git a/test/TensorFlowNET.UnitTest/Numpy/Math.Test.cs b/test/TensorFlowNET.UnitTest/Numpy/Math.Test.cs
index 3c1dbcf3..d08253cb 100644
--- a/test/TensorFlowNET.UnitTest/Numpy/Math.Test.cs
+++ b/test/TensorFlowNET.UnitTest/Numpy/Math.Test.cs
@@ -25,7 +25,7 @@ namespace TensorFlowNET.UnitTest.NumPy
 
             p = np.prod(new[,] { { 1.0, 2.0 }, { 3.0, 4.0 } }, axis: 1);
             Assert.AreEqual(p.shape, 2);
-            Assert.IsTrue(Equal(p.Data<double>(), new[] { 2.0, 12.0 }));
+            Assert.IsTrue(Equal(p.ToArray<double>(), new[] { 2.0, 12.0 }));
         }
     }
 }