static async System.Threading.Tasks.Task Main(string[] args)
{
// MLOps: Create experiment and run
var mlOpsContext = new MLOpsBuilder()
.UseSQLite(@"C:/MLOps")
.Build();
Console.WriteLine("Creating an MLOps Run");
var runId = await mlOpsContext.CreateRunAsync("Product Category Predictor");
Console.WriteLine($"Run created with Id {runId}");
var mlContext = new MLContext(seed: 1);
Console.WriteLine("Loading the data");
var data = mlContext.Data.LoadFromTextFile <ModelInput>("Data/titanic.csv", hasHeader: true, separatorChar: ',');
var testTrainTest = mlContext.Data.TrainTestSplit(data);
var features = new[] { nameof(ModelInput.Pclass), nameof(ModelInput.Sex), nameof(ModelInput.Age), nameof(ModelInput.SibSp), nameof(ModelInput.Parch), nameof(ModelInput.Fare), nameof(ModelInput.Embarked) };
Console.WriteLine("Creating a data processing pipeline");
var dataProcessingPipeline =
mlContext.Transforms.ReplaceMissingValues(nameof(ModelInput.Age), replacementMode: MissingValueReplacingEstimator.ReplacementMode.Mean)
.Append(mlContext.Transforms.Categorical.OneHotEncoding(nameof(ModelInput.Sex)))
.Append(mlContext.Transforms.Categorical.OneHotEncoding(nameof(ModelInput.Embarked)))
.Append(mlContext.Transforms.Categorical.OneHotEncoding(nameof(ModelInput.Pclass)))
.Append(mlContext.Transforms.Concatenate("Features", features))
.Append(mlContext.Transforms.NormalizeMinMax("Features"));
Console.WriteLine("Training the model, please stand-by...");
var trainingPipeline = dataProcessingPipeline
.Append(mlContext.BinaryClassification.Trainers.SdcaLogisticRegression("Label", "Features"));
var trainedModel = trainingPipeline.Fit(testTrainTest.TrainSet);
Console.WriteLine("Evaluating the model");
var predictions = trainedModel.Transform(testTrainTest.TestSet);
var metrics = mlContext.BinaryClassification.Evaluate(predictions);
//MLOps: Log Metrics
Console.WriteLine("Logging metrics");
await mlOpsContext.LogMetricsAsync(runId, metrics);
//Save the model
mlContext.Model.Save(trainedModel, testTrainTest.TrainSet.Schema, "BinaryClassificationModel.zip");
//MLOps: Upload artifact/model
Console.WriteLine("Uploading artifact");
await mlOpsContext.UploadModelAsync(runId, "BinaryClassificationModel.zip");
}
static async Task Main(string[] args)
{
// MLOps: Create experiment and run
var mlOpsContext = new MLOpsBuilder()
.UseSQLite(@"C:/MLOps")
.Build();
Console.WriteLine("Creating an MLOps Run");
var runId = await mlOpsContext.CreateRunAsync("Product Category Predictor");
Console.WriteLine($"Run created with Id {runId}");
var mlContext = new MLContext(seed: 1);
Console.WriteLine("Loading the data");
var data = mlContext.Data.LoadFromTextFile <ProductInformation>("Data/ecommerce.csv", hasHeader: true, separatorChar: ',');
var testTrainTest = mlContext.Data.TrainTestSplit(data);
Console.WriteLine("Creating a data processing pipeline");
var dataProcessingPipeline = mlContext.Transforms.Conversion.MapValueToKey(nameof(ProductInformation.Category))
.Append(mlContext.Transforms.Text.FeaturizeText(nameof(ProductInformation.ProductName)))
.Append(mlContext.Transforms.Text.FeaturizeText(nameof(ProductInformation.Description)))
.Append(mlContext.Transforms.Categorical.OneHotHashEncoding(nameof(ProductInformation.Brand))
.Append(mlContext.Transforms.Concatenate("Features", nameof(ProductInformation.ProductName), nameof(ProductInformation.Description), nameof(ProductInformation.Brand), nameof(ProductInformation.Price))
.Append(mlContext.Transforms.NormalizeMinMax("Features"))));
Console.WriteLine("Training the model, please stand-by...");
var trainingPipeline = dataProcessingPipeline
.Append(mlContext.MulticlassClassification.Trainers.SdcaMaximumEntropy(nameof(ProductInformation.Category), "Features")
.Append(mlContext.Transforms.Conversion.MapKeyToValue("PredictedLabel")));
var trainedModel = trainingPipeline.Fit(testTrainTest.TrainSet);
Console.WriteLine("Evaluating the model");
var predictions = trainedModel.Transform(testTrainTest.TestSet);
var metrics = mlContext.MulticlassClassification.Evaluate(predictions, nameof(ProductInformation.Category));
//MLOps: Log Metrics
Console.WriteLine("Logging metrics");
await mlOpsContext.LogMetricsAsync(runId, metrics);
//Save the model
mlContext.Model.Save(trainedModel, testTrainTest.TrainSet.Schema, "MultiClassificationModel.zip");
//MLOps: Upload artifact/model
Console.WriteLine("Uploading artifact");
await mlOpsContext.UploadModelAsync(runId, "MultiClassificationModel.zip");
}
static async Task Main(string[] args)
{
// MLOps: Create experiment and run
var mlOpsContext = new MLOpsBuilder()
.UseSQLite(@"C:/MLOps")
.Build();
Console.WriteLine("Creating an MLOps Run");
var runId = await mlOpsContext.CreateRunAsync("Product Category Predictor");
Console.WriteLine($"Run created with Id {runId}");
var mlContext = new MLContext(seed: 1);
Console.WriteLine("Loading the data");
var data = mlContext.Data.LoadFromTextFile <ModelInput>("Data/taxi-fare.csv", hasHeader: true, separatorChar: ',');
var testTrainTest = mlContext.Data.TrainTestSplit(data);
Console.WriteLine("Creating a data processing pipeline");
var dataProcessingPipeline = mlContext.Transforms.CopyColumns(outputColumnName: "Label", inputColumnName: "FareAmount")
.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "VendorIdEncoded", inputColumnName: "VendorId"))
.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "RateCodeEncoded", inputColumnName: "RateCode"))
.Append(mlContext.Transforms.Categorical.OneHotEncoding(outputColumnName: "PaymentTypeEncoded", inputColumnName: "PaymentType"))
.Append(mlContext.Transforms.Concatenate("Features", "VendorIdEncoded", "RateCodeEncoded", "PassengerCount", "TripDistance", "PaymentTypeEncoded"));
Console.WriteLine("Training the model, please stand-by...");
var trainingPipeline = dataProcessingPipeline
.Append(mlContext.Regression.Trainers.FastTree());
var trainedModel = trainingPipeline.Fit(testTrainTest.TrainSet);
Console.WriteLine("Evaluating the model");
var predictions = trainedModel.Transform(testTrainTest.TestSet);
var metrics = mlContext.Regression.Evaluate(predictions);
//MLOps: Log Metrics
Console.WriteLine("Logging metrics");
await mlOpsContext.LogMetricsAsync(runId, metrics);
//Save the model
mlContext.Model.Save(trainedModel, testTrainTest.TrainSet.Schema, "RegressionClassificationModel.zip");
//MLOps: Upload artifact/model
Console.WriteLine("Uploading artifact");
await mlOpsContext.UploadModelAsync(runId, "RegressionClassificationModel.zip");
}
public async Task UploadModelAsync_ValidModelPath_UploadSuccessAsync()
{
//Arrange
var destinationFolder = @"C:\MLOps";
IMLOpsContext mlm = new MLOpsBuilder().UseSQLite(destinationFolder).Build();
var guid = Guid.NewGuid();
var modelPath = @"C:\data\model.zip";
var modelStoragePath = @"C:\MLOps";
using var writer = new StreamWriter(modelPath);
writer.Close();
//Act
await mlm.UploadModelAsync(guid, modelPath);
//Assert
var fileExists = File.Exists(Path.Combine(modelStoragePath, $"{guid}.zip"));
fileExists.Should().BeTrue();
}