fix Strides default value for Pooling2D.

src/TensorFlowNET.Core/Framework/meta_graph.cs

@@ -150,7 +150,7 @@ namespace Tensorflow
             var variables = graph.get_collection<IVariableV1>(tf.GraphKeys.GLOBAL_VARIABLES,
                                      scope: scope_to_prepend_to_names);
             var var_list = new Dictionary<string, IVariableV1>();
-            variables.ForEach(v => var_list[ops.strip_name_scope(v.Name, scope_to_prepend_to_names)] = v);
+            // variables.ForEach(v => var_list[ops.strip_name_scope(v.Name, scope_to_prepend_to_names)] = v);
 
             return (var_list, imported_return_elements);
         }

src/TensorFlowNET.Core/Keras/Engine/Model.cs

@@ -1,6 +1,7 @@
 using NumSharp;
 using System;
 using Tensorflow.Keras.ArgsDefinition;
+using Tensorflow.Keras.Losses;
 using Tensorflow.Keras.Optimizers;
 
 namespace Tensorflow.Keras.Engine
@@ -42,6 +43,11 @@ namespace Tensorflow.Keras.Engine
             // Prepare list of loss functions, same size of model outputs.
         }
 
+        public void compile(string optimizerName, ILossFunc lossName)
+        {
+            throw new NotImplementedException("");
+        }
+
         /// <summary>
         /// Generates output predictions for the input samples.
         /// </summary>

src/TensorFlowNET.Core/Keras/KerasApi.cs

@@ -7,6 +7,7 @@ using Tensorflow.Keras.ArgsDefinition;
 using Tensorflow.Keras.Datasets;
 using Tensorflow.Keras.Engine;
 using Tensorflow.Keras.Layers;
+using Tensorflow.Keras.Losses;
 using static Tensorflow.Binding;
 
 namespace Tensorflow
@@ -16,6 +17,7 @@ namespace Tensorflow
         public KerasDataset datasets { get; } = new KerasDataset();
         public Initializers initializers { get; } = new Initializers();
         public LayersApi layers { get; } = new LayersApi();
+        public LossesApi losses { get; } = new LossesApi();
         public Activations activations { get; } = new Activations();
         public Preprocessing preprocessing { get; } = new Preprocessing();
         public BackendImpl backend { get; } = new BackendImpl();
@@ -69,52 +71,5 @@ namespace Tensorflow
 
             return layer.InboundNodes[0].Outputs;
         }
-
-        public class LayersApi
-        {
-            public Rescaling Rescaling(float scale,
-                float offset = 0,
-                TensorShape input_shape = null)
-                => new Rescaling(new RescalingArgs
-                {
-                    Scale = scale,
-                    Offset = offset,
-                    InputShape = input_shape
-                });
-
-            public Dense Dense(int units,
-                Activation activation = null,
-                TensorShape input_shape = null)
-                => new Dense(new DenseArgs
-                {
-                    Units = units,
-                    Activation = activation ?? tf.keras.activations.Linear,
-                    InputShape = input_shape
-                });
-
-            /// <summary>
-            /// Turns positive integers (indexes) into dense vectors of fixed size.
-            /// </summary>
-            /// <param name="input_dim"></param>
-            /// <param name="output_dim"></param>
-            /// <param name="embeddings_initializer"></param>
-            /// <param name="mask_zero"></param>
-            /// <returns></returns>
-            public Embedding Embedding(int input_dim,
-                int output_dim,
-                IInitializer embeddings_initializer = null,
-                bool mask_zero = false,
-                TensorShape input_shape = null,
-                int input_length = -1)
-                => new Embedding(new EmbeddingArgs
-                {
-                    InputDim = input_dim,
-                    OutputDim = output_dim,
-                    MaskZero = mask_zero,
-                    InputShape = input_shape ?? input_length,
-                    InputLength = input_length,
-                    EmbeddingsInitializer = embeddings_initializer
-                });
-        }
     }
 }

src/TensorFlowNET.Core/Keras/Layers/LayersApi.cs

@@ -0,0 +1,95 @@
+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
+{
+    public class LayersApi
+    {
+        public Conv2D Conv2D(int filters,
+            TensorShape kernel_size = null,
+            string padding = "valid",
+            string activation = "relu")
+            => new Conv2D(new Conv2DArgs
+            {
+                Filters = filters,
+                KernelSize = kernel_size,
+                Padding = padding,
+                Activation = GetActivationByName(activation)
+            });
+
+
+        public Dense Dense(int units,
+            string activation = "linear",
+            TensorShape input_shape = null)
+            => new Dense(new DenseArgs
+            {
+                Units = units,
+                Activation = GetActivationByName(activation),
+                InputShape = input_shape
+            });
+
+        /// <summary>
+        /// Turns positive integers (indexes) into dense vectors of fixed size.
+        /// </summary>
+        /// <param name="input_dim"></param>
+        /// <param name="output_dim"></param>
+        /// <param name="embeddings_initializer"></param>
+        /// <param name="mask_zero"></param>
+        /// <returns></returns>
+        public Embedding Embedding(int input_dim,
+            int output_dim,
+            IInitializer embeddings_initializer = null,
+            bool mask_zero = false,
+            TensorShape input_shape = null,
+            int input_length = -1)
+            => new Embedding(new EmbeddingArgs
+            {
+                InputDim = input_dim,
+                OutputDim = output_dim,
+                MaskZero = mask_zero,
+                InputShape = input_shape ?? input_length,
+                InputLength = input_length,
+                EmbeddingsInitializer = embeddings_initializer
+            });
+
+        public Flatten Flatten(string data_format = null)
+            => new Flatten(new FlattenArgs
+            {
+                DataFormat = data_format
+            });
+
+        public MaxPooling2D MaxPooling2D(TensorShape pool_size = null,
+            TensorShape strides = null,
+            string padding = "valid")
+            => new MaxPooling2D(new MaxPooling2DArgs
+            {
+                PoolSize = pool_size ?? (2, 2),
+                Strides = strides,
+                Padding = padding
+            });
+
+        public Rescaling Rescaling(float scale,
+            float offset = 0,
+            TensorShape input_shape = null)
+            => new Rescaling(new RescalingArgs
+            {
+                Scale = scale,
+                Offset = offset,
+                InputShape = input_shape
+            });
+
+        Activation GetActivationByName(string name)
+            => name switch
+            {
+                "linear" => tf.keras.activations.Linear,
+                "relu" => tf.keras.activations.Relu,
+                "sigmoid" => tf.keras.activations.Sigmoid,
+                "tanh" => tf.keras.activations.Tanh,
+                _ => tf.keras.activations.Linear
+            };
+    }
+}

src/TensorFlowNET.Core/Keras/Layers/Pooling2D.cs

@@ -30,7 +30,7 @@ namespace Tensorflow.Keras.Layers
         {
             this.args = args;
             args.PoolSize = conv_utils.normalize_tuple(args.PoolSize, 2, "pool_size");
-            args.Strides = conv_utils.normalize_tuple(args.Strides, 2, "strides");
+            args.Strides = conv_utils.normalize_tuple(args.Strides ?? args.PoolSize, 2, "strides");
             args.Padding = conv_utils.normalize_padding(args.Padding);
             args.DataFormat = conv_utils.normalize_data_format(args.DataFormat);
             input_spec = new InputSpec(ndim: 4);

src/TensorFlowNET.Core/Keras/Layers/Rescaling.cs

@@ -23,7 +23,8 @@ namespace Tensorflow.Keras.Layers
         protected override Tensor call(Tensor inputs, bool is_training = false, Tensor state = null)
         {
             scale = math_ops.cast(args.Scale, args.DType);
-            throw new NotImplementedException("");
+            offset = math_ops.cast(args.Offset, args.DType);
+            return math_ops.cast(inputs, args.DType) * scale + offset;
         }
     }
 }

src/TensorFlowNET.Core/Keras/Losses/ILossFunc.cs

@@ -0,0 +1,10 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow.Keras.Losses
+{
+    public interface ILossFunc
+    {
+    }
+}

src/TensorFlowNET.Core/Keras/Losses/Loss.cs

@@ -0,0 +1,29 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow.Keras.Losses
+{
+    /// <summary>
+    /// Loss base class.
+    /// </summary>
+    public abstract class Loss
+    {
+        protected string reduction;
+        protected string name;
+        bool _allow_sum_over_batch_size;
+        string _name_scope;
+
+        public Loss(string reduction = ReductionV2.AUTO, string name = null)
+        {
+            this.reduction = reduction;
+            this.name = name;
+            _allow_sum_over_batch_size = false;
+        }
+
+        void _set_name_scope()
+        {
+            _name_scope = name;
+        }
+    }
+}

src/TensorFlowNET.Core/Keras/Losses/LossFunctionWrapper.cs

@@ -0,0 +1,20 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow.Keras.Losses
+{
+    public class LossFunctionWrapper : Loss
+    {
+        Action fn;
+
+        public LossFunctionWrapper(Action fn,
+            string reduction = ReductionV2.AUTO, 
+            string name = null) 
+            : base(reduction: reduction, 
+                  name: name)
+        {
+            this.fn = fn;
+        }
+    }
+}

src/TensorFlowNET.Core/Keras/Losses/LossesApi.cs

@@ -0,0 +1,12 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow.Keras.Losses
+{
+    public class LossesApi
+    {
+        public ILossFunc SparseCategoricalCrossentropy(bool from_logits = false)
+            => new SparseCategoricalCrossentropy(from_logits: from_logits);
+    }
+}

src/TensorFlowNET.Core/Keras/Losses/ReductionV2.cs

@@ -0,0 +1,11 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow.Keras.Losses
+{
+    public class ReductionV2
+    {
+        public const string AUTO = "auto";
+    }
+}

src/TensorFlowNET.Core/Keras/Losses/SparseCategoricalCrossentropy.cs

@@ -0,0 +1,24 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow.Keras.Losses
+{
+    public class SparseCategoricalCrossentropy : LossFunctionWrapper, ILossFunc
+    {
+        public SparseCategoricalCrossentropy(bool from_logits = false,
+            string reduction = ReductionV2.AUTO,
+            string name = "sparse_categorical_crossentropy") : 
+            base(sparse_categorical_crossentropy, 
+                reduction: reduction, 
+                name: name)
+        {
+
+        }
+
+        static void sparse_categorical_crossentropy()
+        {
+
+        }
+    }
+}

src/TensorFlowNET.Core/Keras/Preprocessings/Preprocessing.paths_and_labels_to_dataset.cs

@@ -25,7 +25,7 @@ namespace Tensorflow.Keras
                 var image = path_to_image(image_paths[i], image_size, num_channels, interpolation);
                 data[i] = image.numpy();
                 if (i % 100 == 0)
-                    Console.WriteLine($"Filled {i}/{image_paths.Length} data into memory.");
+                    Console.WriteLine($"Filled {i}/{image_paths.Length} data into ndarray.");
             }
 
             var img_ds = tf.data.Dataset.from_tensor_slices(data);