Add R Samples

samples/R/iris/caret/README.md

@@ -0,0 +1,84 @@
+# Operationalizing R Models in AzureML
+
+## Setup
+
+Create AzureML cluster
+
+```
+az ml env setup -k
+```
+
+## Create Bundle
+
+```R
+# import bundleService
+source('https://raw.githubusercontent.com/danhartl/Machine-Learning-Operationalization/master/utils/BundleService.R')
+
+library(caret)
+data(iris)
+set.seed(12345)
+inTrain<-createDataPartition(iris$Species,p=0.7,list=FALSE)
+training<-iris[inTrain,]
+testing<-iris[-inTrain,]
+
+# train model
+
+irisModel<-train(Species~.,method="rpart",data=iris)
+
+# define operationalization functions
+
+init <- function() {
+  library(caret)
+}
+
+predictIris <- function(SepalLength, SepalWidth, PetalLength, PetalWidth) {
+  input <- data.frame(Sepal.Length = c(SepalLength), Sepal.Width = c(SepalWidth), Petal.Length = c(PetalLength), Petal.Width = c(PetalWidth))
+  result <<- as.character(predict(model, input))
+}
+
+bundleService(
+  init,
+  predictIris,
+  list(model = irisModel),
+  inputs = list(SepalLength = "numeric", SepalWidth = "numeric", PetalLength = "numeric", PetalWidth = "numeric"),
+  outputs = list(result = "character"),
+  outputFolder = "/tmp")
+
+```
+
+## Deploy
+
+```
+az ml service create realtime -n myservice1 -r mrs -f service.json -d init -d run -d model
+```
+
+## Test
+
+```
+az ml service run realtime -n myservice1 -d '{ "SepalLength": 4.7, "SepalWidth": 3.2, "PetalLength": 1.3, "PetalWidth": 0.2 }'
+```
+
+## Generate client code
+
+You can download the swagger.json metadata from /swagger.json
+
+Store it in a file and you can use [autorest](https://www.nuget.org/packages/AutoRest) to generate the client code:
+
+```
+.\AutoRest.exe -input 'swagger.json' -ClientName Service -CodeGenerator CSharp -Namespace 'AzureML' -OutputDirectory '/tmp'
+```
+
+With this you can create a sample C# application
+
+```csharp
+internal class Program
+{
+    private static void Main(string[] args)
+    {
+        var service = new Service(new Uri("<servername>"));
+
+        var webServiceResult = service.RunMLService(new InputParameters(4.7, 3.2, 1.3, 0.2));
+        Console.WriteLine(webServiceResult.OutputParameters.Result);
+    }
+}
+```

samples/R/iris/microsoft-r/README.md

@@ -0,0 +1,82 @@
+# Operationalizing R Models in AzureML
+
+## Setup
+
+Create AzureML cluster
+
+```
+az ml env setup -k
+```
+
+## Create Bundle
+
+```R
+# import bundleService
+source('https://raw.githubusercontent.com/danhartl/Machine-Learning-Operationalization/master/utils/BundleService.R')
+
+library(RevoScaleR);
+
+# train model
+
+trainData <- iris; formula <- Species ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Width;
+irisModel <- rxBTrees(formula, data = trainData, lossFunction = 'multinomial', nTree = 3,learningRate = 0.1, sampRate = 0.5, maxdepth = 1, minBucket = 1,seed = 1234, replace = FALSE, mTry = 0, maxNumBins = 200);
+
+# define operationalization functions
+
+init <- function() {
+  library(RevoScaleR)
+}
+
+predictIris <- function(SepalLength, SepalWidth, PetalLength, PetalWidth) {
+  input <- data.frame(Sepal.Length = c(SepalLength), Sepal.Width = c(SepalWidth), Petal.Length = c(PetalLength), Petal.Width = c(PetalWidth))
+
+  prediction <- rxPredict(model, data = input)
+  result <<- as.character(prediction$Species_Pred)
+}
+
+bundleService(
+  init,
+  predictIris,
+  list(model = irisModel),
+  inputs = list(SepalLength = "numeric", SepalWidth = "numeric", PetalLength = "numeric", PetalWidth = "numeric"),
+  outputs = list(result = "character"),
+  outputFolder = "/tmp")
+
+```
+
+## Deploy
+
+```
+az ml service create realtime -n myservice1 -r mrs -f service.json -d init -d run -d model
+```
+
+## Test
+
+```
+az ml service run realtime -n myservice1 -d '{ "SepalLength": 4.7, "SepalWidth": 3.2, "PetalLength": 1.3, "PetalWidth": 0.2 }'
+```
+
+## Generate client code
+
+You can download the swagger.json metadata from /swagger.json
+
+Store it in a file and you can use [autorest](https://www.nuget.org/packages/AutoRest) to generate the client code:
+
+```
+.\AutoRest.exe -input 'swagger.json' -ClientName Service -CodeGenerator CSharp -Namespace 'AzureML' -OutputDirectory '/tmp'
+```
+
+With this you can create a sample C# application
+
+```csharp
+internal class Program
+{
+    private static void Main(string[] args)
+    {
+        var service = new Service(new Uri("<servername>"));
+
+        var webServiceResult = service.RunMLService(new InputParameters(4.7, 3.2, 1.3, 0.2));
+        Console.WriteLine(webServiceResult.OutputParameters.Result);
+    }
+}
+```

utils/BundleService.R

@@ -0,0 +1,57 @@
+library(jsonlite)
+
+bundleService <- function(init, run, objects, inputs, outputs, outputFolder)
+{
+  saveToFile <- function(name, obj) {
+    binFile <- file(paste(outputFolder, '/', name, sep = ''), 'wb')
+    binObj <- serialize(obj, NULL)
+    writeBin(con = binFile, object = binObj)
+    close(binFile)
+  }
+
+  saveToFile('init', init)
+  saveToFile('run', run)
+
+  preloadedObjects <- list(unbox('init'), unbox('run'))
+
+  mapply(function(value, name) {
+    assign(name,value)
+    saveToFile(name, value)
+  }, objects, names(objects))
+
+  for(objName in names(objects)) {
+    preloadedObjects[[length(preloadedObjects)+1]] <- unbox(objName)
+  }
+
+  argumentList = ''
+
+  for(arg in names(inputs)) {
+    if(argumentList != '') {
+      argumentList = paste(argumentList, ', ', sep = '')
+    }
+
+    argumentList = paste(argumentList, arg, sep = '')
+  }
+
+  convertToParameterDefinition <- function(parameters) {
+    result <- list()
+
+    for(key in names(parameters)) {
+      result[[length(result)+1]] <- structure(list(name=unbox(key), type=unbox(parameters[[key]])))
+    }
+
+    result
+  }
+
+  svcmetadata <- structure(list(
+    runtimeType = unbox("R"),
+    initCode = unbox("init()"),
+    code = unbox(paste("run(", argumentList, ")", sep = '')),
+    inputParameterDefinitions = convertToParameterDefinition(inputs),
+    outputParameterDefinitions = convertToParameterDefinition(outputs),
+    preloadedObjects = preloadedObjects
+  ))
+
+  json <- toJSON(svcmetadata, pretty=TRUE)
+  write(json, paste(outputFolder, '/service.json', sep = ''))
+}