model compile overload.

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

@@ -1,5 +1,6 @@
 using Tensorflow.Functions;
 using Tensorflow.Keras.Losses;
+using Tensorflow.Keras.Metrics;
 using Tensorflow.Keras.Saving;
 using Tensorflow.NumPy;
 
@@ -7,12 +8,14 @@ namespace Tensorflow.Keras.Engine;
 
 public interface IModel : ILayer
 {
-    void compile(IOptimizer optimizer = null,
-            ILossFunc loss = null,
-            string[] metrics = null);
+    void compile(IOptimizer optimizer, ILossFunc loss);
+
+    void compile(IOptimizer optimizer, ILossFunc loss, string[] metrics);
 
     void compile(string optimizer, string loss, string[] metrics);
 
+    void compile(IOptimizer optimizer, ILossFunc loss, IMetricFunc[] metrics);
+
     ICallback fit(NDArray x, NDArray y,
             int batch_size = -1,
             int epochs = 1,
@@ -55,7 +58,7 @@ public interface IModel : ILayer
         bool skip_mismatch = false, 
         object options = null);
 
-    void evaluate(NDArray x, NDArray y,
+    Dictionary<string, float> evaluate(NDArray x, NDArray y,
             int batch_size = -1,
             int verbose = 1,
             int steps = -1,

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

@@ -156,7 +156,7 @@ namespace Tensorflow.Keras.Layers
            IInitializer beta_initializer = null,
            IInitializer gamma_initializer = null);
 
-        public ILayer Normalization(int? axis = -1, float? mean = null, float? variance = null, bool invert = false);
+        public ILayer Normalization(Shape? input_shape = null, int? axis = -1, float? mean = null, float? variance = null, bool invert = false);
         public ILayer LeakyReLU(float alpha = 0.3f);
 
         public ILayer LSTM(int units,

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

@@ -10,9 +10,8 @@ namespace Tensorflow.Keras.Engine
         LossesContainer compiled_loss;
         MetricsContainer compiled_metrics;
 
-        public void compile(IOptimizer optimizer = null, 
-            ILossFunc loss = null, 
-            string[] metrics = null)
+        public void compile(IOptimizer optimizer,
+            ILossFunc loss)
         {
             this.optimizer = optimizer ?? new RMSprop(new RMSpropArgs
             {
@@ -20,8 +19,8 @@ namespace Tensorflow.Keras.Engine
 
             this.loss = loss ?? new MeanSquaredError();
 
-            compiled_loss = new LossesContainer(loss, output_names: output_names);
-            compiled_metrics = new MetricsContainer(metrics, output_names: output_names);
+            compiled_loss = new LossesContainer(this.loss, output_names: output_names);
+            compiled_metrics = new MetricsContainer(new string[0], output_names: output_names);
 
             int experimental_steps_per_execution = 1;
             _configure_steps_per_execution(experimental_steps_per_execution);
@@ -31,9 +30,9 @@ namespace Tensorflow.Keras.Engine
             _is_compiled = true;
         }
 
-        public void compile(IOptimizer optimizer = null,
-            ILossFunc loss = null,
-            IMetricFunc[] metrics = null)
+        public void compile(IOptimizer optimizer,
+            ILossFunc loss,
+            string[] metrics)
         {
             this.optimizer = optimizer ?? new RMSprop(new RMSpropArgs
             {
@@ -41,7 +40,7 @@ namespace Tensorflow.Keras.Engine
 
             this.loss = loss ?? new MeanSquaredError();
 
-            compiled_loss = new LossesContainer(loss, output_names: output_names);
+            compiled_loss = new LossesContainer(this.loss, output_names: output_names);
             compiled_metrics = new MetricsContainer(metrics, output_names: output_names);
 
             int experimental_steps_per_execution = 1;
@@ -52,25 +51,58 @@ namespace Tensorflow.Keras.Engine
             _is_compiled = true;
         }
 
-        public void compile(string optimizer, string loss, string[] metrics)
+        public void compile(string optimizer, 
+            string loss, 
+            string[] metrics)
         {
-            var _optimizer = optimizer switch
+            this.optimizer = optimizer switch
             {
                 "rmsprop" => new RMSprop(new RMSpropArgs
                 {
 
                 }),
-                _ => throw new NotImplementedException("")
+                _ => new RMSprop(new RMSpropArgs
+                {
+                })
             };
 
-            ILossFunc _loss = loss switch
+            this.loss = loss switch
             {
                 "mse" => new MeanSquaredError(),
                 "mae" => new MeanAbsoluteError(),
-                _ => throw new NotImplementedException("")
+                _ => new MeanSquaredError()
             };
 
-            compile(optimizer: _optimizer, loss: _loss, metrics: metrics);
+            compiled_loss = new LossesContainer(this.loss, output_names: output_names);
+            compiled_metrics = new MetricsContainer(metrics, output_names: output_names);
+
+            int experimental_steps_per_execution = 1;
+            _configure_steps_per_execution(experimental_steps_per_execution);
+
+            // Initialize cache attrs.
+            _reset_compile_cache();
+            _is_compiled = true;
+        }
+
+        public void compile(IOptimizer optimizer,
+            ILossFunc loss,
+            IMetricFunc[] metrics)
+        {
+            this.optimizer = optimizer ?? new RMSprop(new RMSpropArgs
+            {
+            });
+
+            this.loss = loss ?? new MeanSquaredError();
+
+            compiled_loss = new LossesContainer(this.loss, output_names: output_names);
+            compiled_metrics = new MetricsContainer(metrics, output_names: output_names);
+
+            int experimental_steps_per_execution = 1;
+            _configure_steps_per_execution(experimental_steps_per_execution);
+
+            // Initialize cache attrs.
+            _reset_compile_cache();
+            _is_compiled = true;
         }
     }
 }

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

@@ -26,7 +26,7 @@ namespace Tensorflow.Keras.Engine
         /// <param name="workers"></param>
         /// <param name="use_multiprocessing"></param>
         /// <param name="return_dict"></param>
-        public void evaluate(NDArray x, NDArray y,
+        public Dictionary<string, float> evaluate(NDArray x, NDArray y,
             int batch_size = -1,
             int verbose = 1,
             int steps = -1,
@@ -63,12 +63,12 @@ namespace Tensorflow.Keras.Engine
             });
             callbacks.on_test_begin();
 
+            IEnumerable<(string, Tensor)> logs = null;
             foreach (var (epoch, iterator) in data_handler.enumerate_epochs())
             {
                 reset_metrics();
-                //callbacks.on_epoch_begin(epoch);
+                callbacks.on_epoch_begin(epoch);
                 // data_handler.catch_stop_iteration();
-                IEnumerable<(string, Tensor)> logs = null;
 
                 foreach (var step in data_handler.steps())
                 {
@@ -78,12 +78,16 @@ namespace Tensorflow.Keras.Engine
                     callbacks.on_test_batch_end(end_step, logs);
                 }
             }
-            Console.WriteLine();
-            GC.Collect();
-            GC.WaitForPendingFinalizers();
+
+            var results = new Dictionary<string, float>();
+            foreach (var log in logs)
+            {
+                results[log.Item1] = (float)log.Item2;
+            }
+            return results;
         }
 
-        public KeyValuePair<string, float>[] evaluate(IDatasetV2 x)
+        public Dictionary<string, float> evaluate(IDatasetV2 x, int verbose = 1)
         {
             var data_handler = new DataHandler(new DataHandlerArgs
             {
@@ -92,21 +96,34 @@ namespace Tensorflow.Keras.Engine
                 StepsPerExecution = _steps_per_execution
             });
 
+            var callbacks = new CallbackList(new CallbackParams
+            {
+                Model = this,
+                Verbose = verbose,
+                Steps = data_handler.Inferredsteps
+            });
+            callbacks.on_test_begin();
+
             IEnumerable<(string, Tensor)> logs = null;
             foreach (var (epoch, iterator) in data_handler.enumerate_epochs())
             {
                 reset_metrics();
-                // callbacks.on_epoch_begin(epoch)
+                callbacks.on_epoch_begin(epoch);
                 // data_handler.catch_stop_iteration();
 
-
                 foreach (var step in data_handler.steps())
                 {
                     // callbacks.on_train_batch_begin(step)
                     logs = test_function(data_handler, iterator);
                 }
             }
-            return logs.Select(x => new KeyValuePair<string, float>(x.Item1, (float)x.Item2)).ToArray();
+
+            var results = new Dictionary<string, float>();
+            foreach (var log in logs)
+            {
+                results[log.Item1] = (float)log.Item2;
+            }
+            return results;
         }
 
         IEnumerable<(string, Tensor)> test_function(DataHandler data_handler, OwnedIterator iterator)

src/TensorFlowNET.Keras/Layers/LayersApi.cs

@@ -873,9 +873,10 @@ namespace Tensorflow.Keras.Layers
                 CountWeights = count_weights
             });
 
-        public ILayer Normalization(int? axis = -1, float? mean = null, float? variance = null, bool invert = false)
+        public ILayer Normalization(Shape? input_shape = null, int? axis = -1, float? mean = null, float? variance = null, bool invert = false)
             => new Normalization(new NormalizationArgs
             {
+                InputShape = input_shape,
                 Axis = axis,
                 Mean = mean,
                 Variance = variance,