fix keras model predict return result.

Author: Oceania2018

src/TensorFlowNET.Core/Sessions/BaseSession.cs

@@ -293,7 +293,8 @@ protected override void DisposeUnmanagedResources(IntPtr handle)
             // c_api.TF_CloseSession(handle, tf.Status.Handle);
             if (tf.Status == null || tf.Status.Handle.IsInvalid)
             {
-                c_api.TF_DeleteSession(handle, c_api.TF_NewStatus());
+                using var status = new Status();
+                c_api.TF_DeleteSession(handle, status.Handle);
             }
             else
             {

src/TensorFlowNET.Keras/Callbacks/CallbackList.cs

@@ -39,5 +39,25 @@ public void on_epoch_end(int epoch, Dictionary<string, float> epoch_logs)
         {
             callbacks.ForEach(x => x.on_epoch_end(epoch, epoch_logs));
         }
+
+        public void on_predict_begin()
+        {
+            callbacks.ForEach(x => x.on_predict_begin());
+        }
+
+        public void on_predict_batch_begin(long step)
+        {
+            callbacks.ForEach(x => x.on_predict_batch_begin(step));
+        }
+
+        public void on_predict_batch_end(long end_step, Dictionary<string, Tensors> logs)
+        {
+            callbacks.ForEach(x => x.on_predict_batch_end(end_step, logs));
+        }
+
+        public void on_predict_end()
+        {
+            callbacks.ForEach(x => x.on_predict_end());
+        }
     }
 }

src/TensorFlowNET.Keras/Callbacks/History.cs

@@ -48,5 +48,26 @@ public void on_epoch_end(int epoch, Dictionary<string, float> epoch_logs)
                 history[log.Key].Add((float)log.Value);
             }
         }
+
+        public void on_predict_begin()
+        {
+            epochs = new List<int>();
+            history = new Dictionary<string, List<float>>();
+        }
+
+        public void on_predict_batch_begin(long step)
+        {
+            
+        }
+
+        public void on_predict_batch_end(long end_step, Dictionary<string, Tensors> logs)
+        {
+            
+        }
+
+        public void on_predict_end()
+        {
+            
+        }
     }
 }

src/TensorFlowNET.Keras/Callbacks/ICallback.cs

@@ -11,5 +11,9 @@ public interface ICallback
         void on_train_batch_begin(long step);
         void on_train_batch_end(long end_step, Dictionary<string, float> logs);
         void on_epoch_end(int epoch, Dictionary<string, float> epoch_logs);
+        void on_predict_begin();
+        void on_predict_batch_begin(long step);
+        void on_predict_batch_end(long end_step, Dictionary<string, Tensors> logs);
+        void on_predict_end();
     }
 }

src/TensorFlowNET.Keras/Callbacks/ProgbarLogger.cs

@@ -1,5 +1,4 @@
-using PureHDF;
-using System;
+using System;
 using System.Collections.Generic;
 using System.Diagnostics;
 using System.Linq;
@@ -77,5 +76,26 @@ void _maybe_init_progbar()
         {
 
         }
+
+        public void on_predict_begin()
+        {
+            _reset_progbar();
+            _maybe_init_progbar();
+        }
+
+        public void on_predict_batch_begin(long step)
+        {
+            
+        }
+
+        public void on_predict_batch_end(long end_step, Dictionary<string, Tensors> logs)
+        {
+            
+        }
+
+        public void on_predict_end()
+        {
+            
+        }
     }
 }

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

@@ -5,11 +5,70 @@
 using Tensorflow.Keras.ArgsDefinition;
 using Tensorflow.Keras.Engine.DataAdapters;
 using static Tensorflow.Binding;
+using Tensorflow.Keras.Callbacks;
 
 namespace Tensorflow.Keras.Engine
 {
     public partial class Model
     {
+        public Tensors predict(IDatasetV2 dataset,
+            int batch_size = -1,
+            int verbose = 0,
+            int steps = -1,
+            int max_queue_size = 10,
+            int workers = 1,
+            bool use_multiprocessing = false)
+        {
+            var data_handler = new DataHandler(new DataHandlerArgs
+            {
+                Dataset = dataset,
+                BatchSize = batch_size,
+                StepsPerEpoch = steps,
+                InitialEpoch = 0,
+                Epochs = 1,
+                MaxQueueSize = max_queue_size,
+                Workers = workers,
+                UseMultiprocessing = use_multiprocessing,
+                Model = this,
+                StepsPerExecution = _steps_per_execution
+            });
+
+            var callbacks = new CallbackList(new CallbackParams
+            {
+                Model = this,
+                Verbose = verbose,
+                Epochs = 1,
+                Steps = data_handler.Inferredsteps
+            });
+
+            Tensor batch_outputs = null;
+            _predict_counter.assign(0);
+            callbacks.on_predict_begin();
+            foreach (var (epoch, iterator) in data_handler.enumerate_epochs())
+            {
+                foreach (var step in data_handler.steps())
+                {
+                    callbacks.on_predict_batch_begin(step);
+                    var tmp_batch_outputs = run_predict_step(iterator);
+                    if (batch_outputs == null)
+                    {
+                        batch_outputs = tmp_batch_outputs[0];
+                    }
+                    else
+                    {
+                        batch_outputs = tf.concat(new Tensor[] { batch_outputs, tmp_batch_outputs[0] }, axis: 0);
+                    }
+
+                    var end_step = step + data_handler.StepIncrement;
+                    callbacks.on_predict_batch_end(end_step, new Dictionary<string, Tensors> { { "outputs", batch_outputs } });
+                }
+                GC.Collect();
+            }
+            
+            callbacks.on_predict_end();
+            return batch_outputs;
+        }
+
         /// <summary>
         /// Generates output predictions for the input samples.
         /// </summary>