Image Classification API: Fix processing incomplete batch(<batchSize), images processed per epoch , enable EarlyStopping without Validation Set. Fixes dotnet#4274 and dotnet#4286 (dotnet#4289)

* In ImageClassification, process incomplete batch where number of samples < batchSize.

* fixed batchIndex not reseting in train loop, enabled EarlyStopping when validationSet is not given for ImageClassificationAPI

* fixed changing shape of feature and label tensor for incomplete batch,detected edge case where early stopping not supported.

* Improved featureBatchSizeInBytes calculation, improved exception message.

Author: ashbhandare

src/Microsoft.ML.Dnn/ImageClassificationTransform.cs

@@ -186,6 +186,9 @@ private void CheckTrainingParameters(ImageClassificationEstimator.Options option
 
             if (_session.graph.OperationByName(_labelTensor.name.Split(':')[0]) == null)
                 throw Host.ExceptParam(nameof(options.TensorFlowLabel), $"'{options.TensorFlowLabel}' does not exist in the model");
+            if (options.EarlyStoppingCriteria != null && options.ValidationSet == null && options.TestOnTrainSet == false)
+                throw Host.ExceptParam(nameof(options.EarlyStoppingCriteria), $"Early stopping enabled but unable to find a validation" +
+                    $" set and/or train set testing disabled. Please disable early stopping or either provide a validation set or enable train set training.");
         }
 
         private (Tensor, Tensor) AddJpegDecoding(int height, int width, int depth)
@@ -381,6 +384,7 @@ private void TrainAndEvaluateClassificationLayer(string trainBottleneckFilePath,
             float crossentropy = 0;
             for (int epoch = 0; epoch < epochs; epoch += 1)
             {
+                batchIndex = 0;
                 metrics.Train.Accuracy = 0;
                 metrics.Train.CrossEntropy = 0;
                 metrics.Train.BatchProcessedCount = 0;
@@ -432,6 +436,42 @@ private void TrainAndEvaluateClassificationLayer(string trainBottleneckFilePath,
                         }
                     }
 
+                    //Process last incomplete batch
+                    if (batchIndex > 0)
+                    {
+                        featureTensorShape[0] = batchIndex;
+                        featureBatchSizeInBytes = sizeof(float) * featureLength * batchIndex;
+                        labelTensorShape[0] = batchIndex;
+                        labelBatchSizeInBytes = sizeof(long) * batchIndex;
+                        runner.AddInput(new Tensor(featureBatchPtr, featureTensorShape, TF_DataType.TF_FLOAT, featureBatchSizeInBytes), 0)
+                                .AddInput(new Tensor(labelBatchPtr, labelTensorShape, TF_DataType.TF_INT64, labelBatchSizeInBytes), 1)
+                                .Run();
+
+                        metrics.Train.BatchProcessedCount += 1;
+
+                        if (options.TestOnTrainSet && statisticsCallback != null)
+                        {
+                            var outputTensors = testEvalRunner
+                                .AddInput(new Tensor(featureBatchPtr, featureTensorShape, TF_DataType.TF_FLOAT, featureBatchSizeInBytes), 0)
+                                .AddInput(new Tensor(labelBatchPtr, labelTensorShape, TF_DataType.TF_INT64, labelBatchSizeInBytes), 1)
+                                .Run();
+
+                            outputTensors[0].ToScalar<float>(ref accuracy);
+                            outputTensors[1].ToScalar<float>(ref crossentropy);
+                            metrics.Train.Accuracy += accuracy;
+                            metrics.Train.CrossEntropy += crossentropy;
+
+                            outputTensors[0].Dispose();
+                            outputTensors[1].Dispose();
+                        }
+
+                        batchIndex = 0;
+                        featureTensorShape[0] = batchSize;
+                        featureBatchSizeInBytes = sizeof(float) * featureBatch.Length;
+                        labelTensorShape[0] = batchSize;
+                        labelBatchSizeInBytes = sizeof(long) * batchSize;
+                    }
+
                     if (options.TestOnTrainSet && statisticsCallback != null)
                     {
                         metrics.Train.Epoch = epoch;
@@ -443,7 +483,15 @@ private void TrainAndEvaluateClassificationLayer(string trainBottleneckFilePath,
                 }
 
                 if (validationSet == null)
+                {
+                    //Early stopping check
+                    if (options.EarlyStoppingCriteria != null)
+                    {
+                        if (options.EarlyStoppingCriteria.ShouldStop(metrics.Train))
+                            break;
+                    }
                     continue;
+                }
 
                 batchIndex = 0;
                 metrics.Train.BatchProcessedCount = 0;
@@ -481,6 +529,31 @@ private void TrainAndEvaluateClassificationLayer(string trainBottleneckFilePath,
                         }
                     }
 
+                    //Process last incomplete batch
+                    if(batchIndex > 0)
+                    {
+                        featureTensorShape[0] = batchIndex;
+                        featureBatchSizeInBytes = sizeof(float) * featureLength * batchIndex;
+                        labelTensorShape[0] = batchIndex;
+                        labelBatchSizeInBytes = sizeof(long) * batchIndex;
+                        var outputTensors = validationEvalRunner
+                                .AddInput(new Tensor(featureBatchPtr, featureTensorShape, TF_DataType.TF_FLOAT, featureBatchSizeInBytes), 0)
+                                .AddInput(new Tensor(labelBatchPtr, labelTensorShape, TF_DataType.TF_INT64, labelBatchSizeInBytes), 1)
+                                .Run();
+
+                        outputTensors[0].ToScalar<float>(ref accuracy);
+                        metrics.Train.Accuracy += accuracy;
+                        metrics.Train.BatchProcessedCount += 1;
+                        batchIndex = 0;
+
+                        featureTensorShape[0] = batchSize;
+                        featureBatchSizeInBytes = sizeof(float) * featureBatch.Length;
+                        labelTensorShape[0] = batchSize;
+                        labelBatchSizeInBytes = sizeof(long) * batchSize;
+
+                        outputTensors[0].Dispose();
+                    }
+
                     if (statisticsCallback != null)
                     {
                         metrics.Train.Epoch = epoch;