Adding some activation layers.

src/TensorFlowNET.Keras/Layers/Activation/LeakyReLu.cs

@@ -21,7 +21,7 @@ namespace Tensorflow.Keras.Layers
 
         protected override Tensors Call(Tensors inputs, Tensor state = null, bool is_training = false)
         {
-            return tf.nn.leaky_relu(inputs, alpha: alpha);
+            return tf.nn.relu(inputs);
         }
     }
 }

src/TensorFlowNET.Keras/Layers/Activation/ReLu .cs

@@ -0,0 +1,24 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+using Tensorflow.Keras.ArgsDefinition;
+using Tensorflow.Keras.Engine;
+using static Tensorflow.Binding;
+
+namespace Tensorflow.Keras.Layers
+{
+    /// <summary>
+    /// Leaky version of a Rectified Linear Unit.
+    /// </summary>
+    public class ReLu : Layer
+    {
+        public ReLu(LayerArgs args) : base(args)
+        {
+        }
+
+        protected override Tensors Call(Tensors inputs, Tensor state = null, bool is_training = false)
+        {
+            return tf.nn.relu(inputs, name: Name);
+        }
+    }
+}

src/TensorFlowNET.Keras/Layers/Activation/Sigmoid.cs

@@ -0,0 +1,24 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+using Tensorflow.Keras.ArgsDefinition;
+using Tensorflow.Keras.Engine;
+using static Tensorflow.Binding;
+
+namespace Tensorflow.Keras.Layers
+{
+    /// <summary>
+    /// Leaky version of a Rectified Linear Unit.
+    /// </summary>
+    public class Sigmoid : Layer
+    {
+        public Sigmoid(LayerArgs args) : base(args)
+        {
+        }
+
+        protected override Tensors Call(Tensors inputs, Tensor state = null, bool is_training = false)
+        {
+            return tf.nn.sigmoid(inputs, name: Name);
+        }
+    }
+}

src/TensorFlowNET.Keras/Layers/Activation/Tanh.cs

@@ -0,0 +1,24 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+using Tensorflow.Keras.ArgsDefinition;
+using Tensorflow.Keras.Engine;
+using static Tensorflow.Binding;
+
+namespace Tensorflow.Keras.Layers
+{
+    /// <summary>
+    /// Leaky version of a Rectified Linear Unit.
+    /// </summary>
+    public class Tanh : Layer
+    {
+        public Tanh(LayerArgs args) : base(args)
+        {
+        }
+
+        protected override Tensors Call(Tensors inputs, Tensor state = null, bool is_training = false)
+        {
+            return tf.nn.tanh(inputs, name: Name);
+        }
+    }
+}

src/TensorFlowNET.Keras/Layers/LayersApi.cs

@@ -315,6 +315,12 @@ namespace Tensorflow.Keras.Layers
             return layer.Apply(inputs);
         }
 
+        public Layer ReLU() => new ReLu(new LayerArgs { });
+
+        public Layer Tanh() => new Tanh(new LayerArgs { });
+
+        public Layer Sigmoid() => new Sigmoid(new LayerArgs { });
+
         /// <summary>
         /// Leaky version of a Rectified Linear Unit.
         /// </summary>
@@ -392,6 +398,8 @@ namespace Tensorflow.Keras.Layers
             => name switch
             {
                 "glorot_uniform" => tf.glorot_uniform_initializer,
+                "random_uniform" => tf.random_uniform_initializer,
+                "orthogonal" => tf.orthogonal_initializer,
                 "zeros" => tf.zeros_initializer,
                 "ones" => tf.ones_initializer,
                 _ => tf.glorot_uniform_initializer

test/TensorFlowNET.Keras.UnitTest/Layers/ActivationTest.cs

@@ -11,6 +11,31 @@ namespace TensorFlowNET.Keras.UnitTest
     [TestClass]
     public class ActivationTest : EagerModeTestBase
     {
+        [TestMethod]
+        public void ReLU()
+        {
+            var layer = keras.layers.ReLU();
+            Tensor output = layer.Apply(np.array(-3.0f, -1.0f, 0.0f, 2.0f));
+            Equal(new[] { 0.0f, 0.0f, 0.0f, 2.0f }, output.ToArray<float>());
+        }
+
+        [TestMethod]
+        public void Sigmoid()
+        {
+            var layer = keras.layers.Sigmoid();
+            Tensor output = layer.Apply(np.array(-3.0f, -1.0f, 0.0f, 2.0f));
+            Equal(new[] { 0.047425866f, 0.26894143f, 0.5f, 0.8807971f }, output.ToArray<float>());
+        }
+        
+        [TestMethod]
+        public void Tanh()
+        {
+            var layer = keras.layers.Tanh();
+            Tensor output = layer.Apply(np.array(-3.0f, -1.0f, 0.0f, 2.0f));
+            // {-0.9950547f, -0.7615942f, 0f, 0.9640276f}
+            Equal(new[] { -0.9950547f, -0.7615942f, 0f, 0.9640276f }, output.ToArray<float>());
+        }
+
         [TestMethod]
         public void LeakyReLU()
         {

test/TensorFlowNET.Keras.UnitTest/Layers/LayersTest.cs

@@ -123,7 +123,9 @@ namespace TensorFlowNET.Keras.UnitTest
                 { 2, 3, 4, 5 },
                 { 3, 4, 5, 6 }
             });
-            // model.compile("rmsprop", "mse");
+            model.compile(optimizer: keras.optimizers.RMSprop(0.001f),
+                          loss: keras.losses.MeanSquaredError(),
+                          metrics: new[] { "acc" });
             var output_array = model.predict(input_array);
             Assert.AreEqual((32, 10, 64), output_array.shape);
         }