Caching enabling in code gen part -2 (dotnet#298)

* add

* added caching codegen

Author: srsaggam

src/mlnet.Test/ApprovalTests/ConsoleCodeGeneratorTests.GeneratedTrainCodeTest.approved.txt

@@ -68,7 +68,8 @@ namespace MyNamespace
                                             allowSparse: true);
 
             // Common data process configuration with pipeline data transformations
-            var dataProcessPipeline = mlContext.Transforms.Concatenate("Out", new[] { "In" });
+            var dataProcessPipeline = mlContext.Transforms.Concatenate("Out", new[] { "In" })
+                                      .AppendCacheCheckpoint(mlContext);
 
             // Set the training algorithm, then create and config the modelBuilder  
             var trainer = mlContext.BinaryClassification.Trainers.LightGbm(new Options() { NumLeaves = 2, Booster = new Options.TreeBooster.Options() { }, LabelColumn = "Label", FeatureColumn = "Features" });
@@ -100,7 +101,7 @@ namespace MyNamespace
         {
             //Load data to test. Could be any test data. For demonstration purpose train data is used here.
             IDataView trainingDataView = mlContext.Data.LoadFromTextFile<SampleObservation>(
-                                            path: TrainDataPath,
+                                            path: TestDataPath,
                                             hasHeader: true,
                                             separatorChar: ',',
                                             allowQuoting: true,

src/mlnet.Test/ApprovalTests/ConsoleCodeGeneratorTests.cs

@@ -104,7 +104,7 @@ public void GeneratedHelperCodeTest()
                 var trainer2 = new SuggestedTrainer(context, new LightGbmBinaryExtension(), new ColumnInformation(), hyperparams2);
                 var transforms1 = new List<SuggestedTransform>() { ColumnConcatenatingExtension.CreateSuggestedTransform(context, new[] { "In" }, "Out") };
                 var transforms2 = new List<SuggestedTransform>() { ColumnConcatenatingExtension.CreateSuggestedTransform(context, new[] { "In" }, "Out") };
-                var inferredPipeline1 = new SuggestedPipeline(transforms1, new List<SuggestedTransform>(), trainer1, context, false);
+                var inferredPipeline1 = new SuggestedPipeline(transforms1, new List<SuggestedTransform>(), trainer1, context, true);
                 var inferredPipeline2 = new SuggestedPipeline(transforms2, new List<SuggestedTransform>(), trainer2, context, false);
 
                 this.pipeline = inferredPipeline1.ToPipeline();

src/mlnet/CodeGenerator/CSharp/CodeGenerator.cs

@@ -54,7 +54,7 @@ public void GenerateOutput()
             var namespaceValue = Utils.Normalize(settings.OutputName);
 
             // Generate code for training and scoring
-            var trainFileContent = GenerateTrainCode(usings, trainer, transforms, columns, classLabels, namespaceValue);
+            var trainFileContent = GenerateTrainCode(usings, trainer, transforms, columns, classLabels, namespaceValue, pipeline.CacheBeforeTrainer);
             var tree = CSharpSyntaxTree.ParseText(trainFileContent);
             var syntaxNode = tree.GetRoot();
             trainFileContent = Formatter.Format(syntaxNode, new AdhocWorkspace()).ToFullString();
@@ -91,7 +91,7 @@ internal static string GeneratProjectCode()
             return projectCodeGen.TransformText();
         }
 
-        internal string GenerateTrainCode(string usings, string trainer, List<string> transforms, IList<string> columns, IList<string> classLabels, string namespaceValue)
+        internal string GenerateTrainCode(string usings, string trainer, List<string> transforms, IList<string> columns, IList<string> classLabels, string namespaceValue, bool cacheBeforeTrainer)
         {
             var trainingAndScoringCodeGen = new MLCodeGen()
             {
@@ -110,7 +110,8 @@ internal string GenerateTrainCode(string usings, string trainer, List<string> tr
                 TaskType = settings.MlTask.ToString(),
                 Namespace = namespaceValue,
                 LabelName = settings.LabelName,
-                ModelPath = settings.ModelPath
+                ModelPath = settings.ModelPath,
+                CacheBeforeTrainer = cacheBeforeTrainer
             };
 
             return trainingAndScoringCodeGen.TransformText();

src/mlnet/Templates/Console/MLCodeGen.cs

@@ -126,6 +126,7 @@ private static ITransformer TrainEvaluateAndSaveModel(MLContext mlContext)
                                              { Write(")");
                                              }
                                          }
+if(CacheBeforeTrainer){ Write("\r\n                                      .AppendCacheCheckpoint(mlContext)");} 
             this.Write(";\r\n");
 }
             this.Write("\r\n            // Set the training algorithm, then create and config the modelBuil" +
@@ -207,8 +208,13 @@ private static void Predict(MLContext mlContext)
         {
             //Load data to test. Could be any test data. For demonstration purpose train data is used here.
             IDataView trainingDataView = mlContext.Data.LoadFromTextFile<SampleObservation>(
-                                            path: TrainDataPath,
-                                            hasHeader : ");
+                                            path: ");
+if(!string.IsNullOrEmpty(TestPath)){ 
+            this.Write("TestDataPath");
+}else{
+            this.Write("TrainDataPath");
+}
+            this.Write(",\r\n                                            hasHeader : ");
             this.Write(this.ToStringHelper.ToStringWithCulture(HasHeader.ToString().ToLowerInvariant()));
             this.Write(",\r\n                                            separatorChar : \'");
             this.Write(this.ToStringHelper.ToStringWithCulture(Regex.Escape(Separator.ToString())));
@@ -284,6 +290,7 @@ private static void Predict(MLContext mlContext)
 public string Namespace {get;set;}
 public string LabelName {get;set;}
 public string ModelPath {get;set;}
+public bool CacheBeforeTrainer {get;set;}
 
     }
     #region Base class

src/mlnet/Templates/Console/MLCodeGen.tt

@@ -88,7 +88,7 @@ namespace <#= Namespace #>
                                              if(i>0)
                                              { Write(")");
                                              }
-                                         }#>;
+                                         }#><#if(CacheBeforeTrainer){ Write("\r\n                                      .AppendCacheCheckpoint(mlContext)");} #>;
 <#}#>
 
             // Set the training algorithm, then create and config the modelBuilder  
@@ -146,7 +146,7 @@ if(string.IsNullOrEmpty(TestPath)){ #>
         {
             //Load data to test. Could be any test data. For demonstration purpose train data is used here.
             IDataView trainingDataView = mlContext.Data.LoadFromTextFile<SampleObservation>(
-                                            path: TrainDataPath,
+                                            path: <#if(!string.IsNullOrEmpty(TestPath)){ #>TestDataPath<#}else{#>TrainDataPath<#}#>,
                                             hasHeader : <#= HasHeader.ToString().ToLowerInvariant() #>,
                                             separatorChar : '<#= Regex.Escape(Separator.ToString()) #>',
                                             allowQuoting : <#= AllowQuoting.ToString().ToLowerInvariant() #>,
@@ -219,4 +219,5 @@ public int Kfolds {get;set;} = 5;
 public string Namespace {get;set;}
 public string LabelName {get;set;}
 public string ModelPath {get;set;}
+public bool CacheBeforeTrainer {get;set;}
 #>