public static CorApi.Portable.LocalDebugger CreateCorDebugForCommand(DbgShimInterop dbgShimInterop, string command, string workingDir,
IDictionary <string, string> env, TimeSpan runtimeLoadTimeout, Action <CorApi.Portable.LocalDebugger, int> onSetup, out int procId)
{
unsafe
{
IntPtr envPtr = IntPtr.Zero;
try
{
envPtr = DebuggerExtensions.SetupEnvironment(env);
void *resumeHandle;
uint processId;
var hret = (HResults)dbgShimInterop.CreateProcessForLaunch(command, true, envPtr, workingDir, &processId, &resumeHandle);
if (hret != HResults.S_OK)
{
throw new COMException(string.Format("Failed call RegisterForRuntimeStartup: {0}", hret), (int)hret);
}
procId = (int)processId;
return(CreateCorDebugImpl(dbgShimInterop, processId, runtimeLoadTimeout, resumeHandle, onSetup));
}
finally
{
if (envPtr != IntPtr.Zero)
{
DebuggerExtensions.TearDownEnvironment(envPtr);
}
}
}
}
static void Main()
{
var mlContext = new MLContext();
IDataView dataView = mlContext.Data.LoadFromTextFile <EmployeeSalary>(dataPath, hasHeader: true, separatorChar: ',');
Action <EmployeeSalary, EmployeeSalary> squared = (input, output) =>
{
output.PositionLevel = (float)Math.Pow(input.PositionLevel, 2.0);
};
Action <EmployeeSalary, EmployeeSalary> fourth = (input, output) =>
{
output.PositionLevel = (float)Math.Pow(input.PositionLevel, 4.0);
};
var dpp1 = mlContext.Transforms.CopyColumns("Squared", nameof(EmployeeSalary.PositionLevel))
.Append(mlContext.Transforms.CustomMapping(squared, null));
var dpp2 = mlContext.Transforms.CopyColumns("Third", nameof(EmployeeSalary.PositionLevel))
.Append(mlContext.Transforms.CustomMapping(fourth, null));
var dataProcessingPipeline = mlContext.Transforms.CopyColumns("Label", nameof(EmployeeSalary.Salary))
//.Append(dpp1)
.Append(dpp2)
.Append(mlContext.Transforms.Concatenate("Features", nameof(EmployeeSalary.PositionLevel), "Third"));
var trainer = mlContext.Regression.Trainers.LbfgsPoissonRegression(labelColumnName: "Label", featureColumnName: "Features");
var trainingPipeline = dataProcessingPipeline.Append(trainer);
var trainedModel = trainingPipeline.Fit(dataView);
// Use only during debugging
var previewTrainSet = DebuggerExtensions.Preview(trainedModel.Transform(dataView), 100);
var predEngine = mlContext.Model.CreatePredictionEngine <EmployeeSalary, EmployeeSalaryPrediction>(trainedModel);
// Model data
var positionLevel = dataView.GetColumn <float>("PositionLevel").ToArray();
var salary = dataView.GetColumn <float>("Salary").ToArray();
var modelObservablePoints = new List <ObservablePoint>();
for (int i = 0; i < salary.Length; i++)
{
modelObservablePoints.Add(new ObservablePoint(positionLevel[i], salary[i]));
}
// Predictions data
var predictionsObservablePoints = new List <ObservablePoint>();
for (float i = 0f; i <= 10; i = i + 0.5f)
{
var prdiction = predEngine.Predict(new EmployeeSalary()
{
PositionLevel = i
});
predictionsObservablePoints.Add(new ObservablePoint(i, prdiction.Salary));
}
var series = new SeriesCollection()
{
new ScatterSeries()
{
Values = new ChartValues <ObservablePoint>(modelObservablePoints),
Title = "Training set"
},
new LineSeries()
{
Title = "Predictions",
Values = new ChartValues <ObservablePoint>(predictionsObservablePoints),
}
};
var xAxis = new Axis()
{
Title = "Level",
Separator = new LiveCharts.Wpf.Separator()
{
Step = 1.0,
IsEnabled = false
}
};
var yAxis = new Axis()
{
Title = "Salary",
LabelFormatter = value => value.ToString("C")
};
Application.EnableVisualStyles();
Application.SetCompatibleTextRenderingDefault(false);
Application.Run(new ChartForm(xAxis, yAxis, series));
}
