Is it possible to use 64 bit floats when concatenating features?

[bandozia]

I'm having trouble training a kmeans clustering pipeline. Even though all my features have the same type (double) the result is a Schema mismatch between double and single

my data structure class looks something like this:

public class Loja
{
    public double encrypted_5_zipcode { get; set; }
    public double faturamento_total { get; set; }
    public double transacoes_total { get; set; } 
    public static IEnumerable<Loja> ReadCsvSkipingErrors(string filePath)
    {
        //...
    }
}

and main program:

var lojas = Loja.ReadCsvSkipingErrors(fpath);
MLContext mlContext = new MLContext(seed: 1);
IDataView data = mlContext.Data.LoadFromEnumerable(lojas);            

var dataProcessPipeline = mlContext.Transforms.Concatenate("Features",
                nameof(Loja.encrypted_5_zipcode),
                nameof(Loja.faturamento_total),
                nameof(Loja.transacoes_total)
            );

var trainer = mlContext.Clustering.Trainers.KMeans(featureColumnName: "Features", numberOfClusters: 3);            
var trainigPipeline = dataProcessPipeline.Append(trainer);
var trainedModel = trainigPipeline.Fit(data);

the result is:

Unhandled Exception: System.ArgumentOutOfRangeException: Schema mismatch for feature column 'Features': expected Vector<Single>, got Vector<Double>
Parameter name: inputSchema
   at Microsoft.ML.Trainers.TrainerEstimatorBase`2.CheckInputSchema(SchemaShape inputSchema)
   at Microsoft.ML.Trainers.TrainerEstimatorBase`2.GetOutputSchema(SchemaShape inputSchema)
   at Microsoft.ML.Data.EstimatorChain`1.GetOutputSchema(SchemaShape inputSchema)
   at Microsoft.ML.Data.EstimatorChain`1.Fit(IDataView input)

If I change the type of variables to float in class "Loja" it works normally. Is it possible to use a double in this case?

[cookt]

The trainer seems to only support vector of Single according to the doc:
https://docs.microsoft.com/en-us/dotnet/api/microsoft.ml.trainers.kmeanstrainer?view=ml-dotnet

[bandozia]

@cookt, Thanks for the answer.
The documentation is clear: "The input features column data must be Single." I should have researched better.