Trained model is empty after fitting - ml.net

Question

I'm trying to implement a simple linear regression with ML.NET. I have followed Microsoft's tutorial, except that I'm loading data from an IEnumerable rather than from a file. However, I always get zero as a prediction. Investigating back, I found that the model's parameters are empty (see screenshot below).
This is the simplest program with which I was able to reproduce the problem.
class Data
{
public float Revenue { get; set; }
public float Customers { get; set; }
public float IsWeekend { get; set; }
}
class DataPrediction
{
[ColumnName("Score")]
public float RevenueEstimate { get; set; }
}
static async Task Main(string[] args)
{
List<Data> trainData = new List<Data>();
trainData.Add(new Data { IsWeekend = 0, Customers = 20, Revenue = 138 });
trainData.Add(new Data { IsWeekend = 0, Customers = 18, Revenue = 106 });
trainData.Add(new Data { IsWeekend = 0, Customers = 26, Revenue = 142 });
trainData.Add(new Data { IsWeekend = 0, Customers = 21, Revenue = 131 });
trainData.Add(new Data { IsWeekend = 0, Customers = 33, Revenue = 146 });
trainData.Add(new Data { IsWeekend = 1, Customers = 92, Revenue = 287 });
trainData.Add(new Data { IsWeekend = 1, Customers = 113, Revenue = 312 });
List<Data> testData = new List<Data>();
testData.Add(new Data { IsWeekend = 0, Customers = 27, Revenue = 136 });
testData.Add(new Data { IsWeekend = 0, Customers = 22, Revenue = 109 });
testData.Add(new Data { IsWeekend = 1, Customers = 87, Revenue = 256 });
MLContext mlContext = new MLContext();
IDataView trainDataView = mlContext.Data.LoadFromEnumerable(trainData);
IDataView testDataView = mlContext.Data.LoadFromEnumerable(testData);
var pipeline = mlContext.Transforms.CopyColumns("Label", "Revenue")
.Append(mlContext.Transforms.Concatenate("Features", "Customers", "IsWeekend"))
.Append(mlContext.Regression.Trainers.FastTree())
;
var model = pipeline.Fit(trainDataView);
var predictions = model.Transform(testDataView);
var metrics = mlContext.Regression.Evaluate(predictions, "Label", "Score");
var predictionFunction = mlContext.Model.CreatePredictionEngine<Data, DataPrediction>(model);
var prediction = predictionFunction.Predict(new Data { IsWeekend = 1, Customers = 108 });
}
Here is what I get if I break at the closing brace:
Of course, prediction.RevenueEstimate is 0, and the metrics show an average absolute error and a RMS error that are similar to the label's average.
It seems to me that this is a very essential program, and I'm using the FastTree LR trainer just like Microsoft suggests in its tutorial. What am I doing wrong?
PS: Decorating the data classes with ColumnName attributes has no effect.