Some refactoring of the MNIST loader, still incomplete

Author: migueldeicaza

Learn/Datasets/MNIST.cs

@@ -43,48 +43,39 @@ public class Mnist
 		public byte [] 	TrainLabels, TestLabels, ValidationLabels;
 		public byte [,] OneHotTrainLabels, OneHotTestLabels, OneHotValidationLabels;
 
-		// Simple batch reader to get pieces of data from the dataset
-		public BatchReader GetBatchReader (MnistImage [] source)
-		{
-			return new BatchReader (source);
-		}
+		public BatchReader GetTrainReader () => new BatchReader (TrainImages, TrainLabels, OneHotTrainLabels);
+		public BatchReader GetTestReader () => new BatchReader (TestImages, TestLabels, OneHotTestLabels);
+		public BatchReader GetValidationReader () => new BatchReader (ValidationImages, ValidationLabels, OneHotValidationLabels);
 
 		public class BatchReader
 		{
 			int start = 0;
 			MnistImage [] source;
+			byte [] labels;
+			byte [,] oneHotLabels;
 
-			public BatchReader (MnistImage [] source)
+			internal BatchReader (MnistImage [] source, byte [] labels, byte [,] oneHotLabels)
 			{
 				this.source = source;
+				this.labels = labels;
+				this.oneHotLabels = oneHotLabels;
 			}
 
-			public MnistImage [] Read (int batchSize)
-			{
-				var result = new MnistImage [batchSize];
-				if (start + batchSize < source.Length) {
-					Array.Copy (source, start, result, 0, batchSize);
-					start += batchSize;
-				} else {
-					var firstLength = source.Length - start;
-					Array.Copy (source, start, result, 0, firstLength);
-					Array.Copy (source, 0, result, firstLength, batchSize-firstLength);
-					start = firstLength;
-				}
-				return result;
-			}
-
-			public TFTensor ReadAsTensor (int batchSize)
+			public (float[,],float [,]) NextBatch (int batchSize)
 			{
-				var result = new float [batchSize, 784];
+				var imageData = new float [batchSize, 784];
+				var labelData = new float [batchSize, 10];
 
-				var x = Read (batchSize);
 				int p = 0;
-				for (int i = 0; i < batchSize; i++) {
-					Buffer.BlockCopy (x [i].DataFloat, 0, result, p, 784);
+				for (int item = 0; item < batchSize; item++) {
+					Buffer.BlockCopy (source [start+item].DataFloat, 0, imageData, p, 784);
 					p += 784;
+					for (var j = 0; j < 10; j++)
+						labelData [item, j] = oneHotLabels [item + start, j];
 				}
-				return (TFTensor)result;
+
+				start += batchSize;
+				return (imageData, labelData);
 			}
 		}
 

Learn/Learn.csproj

@@ -31,6 +31,10 @@
   </PropertyGroup>
   <ItemGroup>
     <Reference Include="System" />
+    <Reference Include="System.ValueTuple">
+      <HintPath>..\packages\System.ValueTuple.4.3.1\lib\netstandard1.0\System.ValueTuple.dll</HintPath>
+    </Reference>
+    <Reference Include="System.Numerics" />
   </ItemGroup>
   <ItemGroup>
     <Compile Include="Properties\AssemblyInfo.cs" />
@@ -47,5 +51,8 @@
       <Name>TensorFlowSharp</Name>
     </ProjectReference>
   </ItemGroup>
+  <ItemGroup>
+    <None Include="packages.config" />
+  </ItemGroup>
   <Import Project="$(MSBuildBinPath)\Microsoft.CSharp.targets" />
 </Project>

Learn/packages.config

@@ -0,0 +1,4 @@
+<?xml version="1.0" encoding="utf-8"?>
+<packages>
+  <package id="System.ValueTuple" version="4.3.1" targetFramework="net461" />
+</packages>

Makefile

@@ -1,7 +1,14 @@
+all: doc-update yaml
 
 rebuild-docs: docs/template
 	mdoc export-html --force-update -o docs --template=docs/template ecmadocs/en/
 
 # Used to fetch XML doc updates from the C# compiler into the ECMA docs
 doc-update:
 	mdoc update -i TensorFlowSharp/bin/Debug/TensorFlowSharp.xml -o ecmadocs/en TensorFlowSharp/bin/Debug/TensorFlowSharp.dll 
+
+yaml:
+	-rm ecmadocs/en/ns-.xml
+	mono /cvs/ECMA2Yaml/ECMA2Yaml/ECMA2Yaml/bin/Debug/ECMA2Yaml.exe --source=`pwd`/ecmadocs/en --output=`pwd`/docfx/api
+	(cd docfx; mono ~/Downloads/docfx/docfx.exe build)
+

SampleTest/SampleTest.cs

@@ -187,9 +187,9 @@ void BasicMatrix ()
 			};
 		}
 
-		int ArgMax (byte [,] array, int idx)
+		int ArgMax (float [,] array, int idx)
 		{
-			int max = -1;
+			float max = -1;
 			int maxIdx = -1;
 			var l = array.GetLength (1);
 			for (int i = 0; i < l; i++)
@@ -198,7 +198,17 @@ int ArgMax (byte [,] array, int idx)
 					max = array [idx, i];
 				}
 			return maxIdx;
-		}	
+		}
+
+		public float [] Extract (float [,] array, int index)
+		{
+			var n = array.GetLength (1);
+			var ret = new float [n];
+
+			for (int i = 0; i < n; i++)
+				ret [i] = array [index,i];
+			return ret;
+		}
 
 		// This sample has a bug, I suspect the data loaded is incorrect, because the returned
 		// values in distance is wrong, and so is the prediction computed from it.
@@ -211,25 +221,21 @@ void NearestNeighbor ()
 			// 5000 for training
 			const int trainCount = 5000;
 			const int testCount = 200;
-			var Xtr = mnist.GetBatchReader (mnist.TrainImages).ReadAsTensor (trainCount);
-			var Ytr = mnist.OneHotTrainLabels;
-			var Xte = mnist.GetBatchReader (mnist.TestImages).Read (testCount);
-			var Yte = mnist.OneHotTestLabels;
-
-
+			(var trainingImages, var trainingLabels) = mnist.GetTrainReader ().NextBatch (trainCount);
+			(var testImages, var testLabels) = mnist.GetTestReader ().NextBatch (testCount);
 
 			Console.WriteLine ("Nearest neighbor on Mnist images");
 			using (var g = new TFGraph ()) {
 				var s = new TFSession (g);
 
 
-				TFOutput xtr = g.Placeholder (TFDataType.Float, new TFShape (-1, 784));
+				TFOutput trainingInput = g.Placeholder (TFDataType.Float, new TFShape (-1, 784));
 
 				TFOutput xte = g.Placeholder (TFDataType.Float, new TFShape (784));
 
 				// Nearest Neighbor calculation using L1 Distance
 				// Calculate L1 Distance
-				TFOutput distance = g.ReduceSum (g.Abs (g.Add (xtr, g.Neg (xte))), axis: g.Const (1));
+				TFOutput distance = g.ReduceSum (g.Abs (g.Add (trainingInput, g.Neg (xte))), axis: g.Const (1));
 
 				// Prediction: Get min distance index (Nearest neighbor)
 				TFOutput pred = g.ArgMin (distance, g.Const (0));
@@ -241,15 +247,15 @@ void NearestNeighbor ()
 
 					// Get nearest neighbor
 
-					var result = runner.Fetch (pred).Fetch (distance).AddInput (xtr, Xtr).AddInput (xte, Xte [i].DataFloat).Run ();
+					var result = runner.Fetch (pred).Fetch (distance).AddInput (trainingInput, trainingImages).AddInput (xte, Extract (testImages, i)).Run ();
 					var r = result [0].GetValue ();
 					var tr = result [1].GetValue ();
 					var nn_index = (int)(long) result [0].GetValue ();
 
 					// Get nearest neighbor class label and compare it to its true label
-					Console.WriteLine ($"Test {i}: Prediction: {ArgMax (Ytr, nn_index)} True class: {ArgMax (Yte, i)} (nn_index={nn_index})");
-					if (ArgMax (Ytr, nn_index) == ArgMax (Yte, i))
-						accuracy += 1f/ Xte.Length;
+					Console.WriteLine ($"Test {i}: Prediction: {ArgMax (trainingLabels, nn_index)} True class: {ArgMax (testLabels, i)} (nn_index={nn_index})");
+					if (ArgMax (trainingLabels, nn_index) == ArgMax (testLabels, i))
+						accuracy += 1f/ testImages.Length;
 				}
 				Console.WriteLine ("Accuracy: " + accuracy);
 			}

SampleTest/SampleTest.csproj

@@ -39,6 +39,9 @@
     <Reference Include="mscorlib" />
     <Reference Include="System.Core" />
     <Reference Include="System.Numerics" />
+    <Reference Include="System.ValueTuple">
+      <HintPath>..\packages\System.ValueTuple.4.3.1\lib\netstandard1.0\System.ValueTuple.dll</HintPath>
+    </Reference>
   </ItemGroup>
   <ItemGroup>
     <Compile Include="SampleTest.cs" />

SampleTest/packages.config

@@ -1,4 +1,5 @@
 <?xml version="1.0" encoding="utf-8"?>
 <packages>
   <package id="CsvHelper" version="2.16.3.0" targetFramework="net45" />
+  <package id="System.ValueTuple" version="4.3.1" targetFramework="net461" />
 </packages>

TensorFlowSharp/Buffer.cs

@@ -23,17 +23,22 @@ internal struct LLBuffer
 	/// Holds a block of data, suitable to pass, or retrieve from TensorFlow.
 	/// </summary>
 	/// <remarks>
+	/// <para>
 	/// Use the TFBuffer to blobs of data into TensorFlow, or to retrieve blocks
 	/// of data out of TensorFlow.
-	/// 
+	/// </para>
+	/// <para>
 	/// There are two constructors to wrap existing data, one to wrap blocks that are 
 	/// pointed to by an IntPtr and one that takes a byte array that we want to wrap.
-	/// 
+	/// </para>
+	/// <para>
 	/// The empty constructor can be used to create a new TFBuffer that can be populated
 	/// by the TensorFlow library and returned to user code.
-	/// 
+	/// </para>
+	/// <para>
 	/// Typically, the data consists of a serialized protocol buffer, but other data
 	/// may also be held in a buffer.
+	/// </para>
 	/// </remarks>
 	// TODO: the string ctor
 	// TODO: perhaps we should have an implicit byte [] conversion that just calls ToArray?

TensorFlowSharp/Tensor.cs

@@ -18,16 +18,20 @@ namespace TensorFlow
 	/// TFTensor holds a multi-dimensional array of elements of a single data type.
 	/// </summary>
 	/// <remarks>
+	/// <para>
 	/// You can create tensors with the various constructors in this class, or using
 	/// the implicit conversions from various data types into a TFTensor.
-	///
+	///</para>
+	/// <para>
 	/// The implicit conversions for basic types produce tensors of one dimesion with
 	/// a single element, while the implicit conversion from an array, expects a multi-dimensional
 	/// array that is converted into a tensor of the right dimensions.
-	/// 
+	/// </para>
+	/// <para>
 	/// The special "String" tensor data type that you will find in TensorFlow documentation
 	/// really represents a byte array.   You can create string tensors by using the <see cref="M:TensorFlow.TFTensor.CreateString"/> 
 	/// method that takes a byte array buffer as input.
+	/// </para>
 	/// </remarks>
 	public class TFTensor : TFDisposable
 	{