The given key 'metric' was not present in the dictionary

[PeterPann23]

System information

• OS version/distro:
  Windows 10
  Windows 2012
  Windows 2019
• .NET Version (eg., dotnet --info):
  Core 3.0 Version 3.0.100-preview5-011568
  Microsoft.ML.LightGbm Version 1.0.0.0, Culture=neutral
  Microsoft.ML Version 1.0.0.0

Issue

• What did you do?
  Load csv file as and train model loading by all floats in single vector
  see code bellow

• What happened?
  Get error The given key 'metric' was not present in the dictionary during the training

• What did you expect?
  Not sure why I get the error, also I do not mention a key metric.

Source code / logs

The error:

[Source=LightGBMMulticlass; Training with LightGBM, Kind=Trace] Channel disposed. Elapsed 00:00:04.0970203
.57.erLeaf = 5059.0337420.
Error stack    at System.Collections.Generic.Dictionary`2.get_Item(TKey key)r_6_MinTicks_In_15Sec_B_ND.tsv
   at Microsoft.ML.Trainers.LightGbm.WrappedLightGbmTraining.Train(IChannel ch, IProgressChannel pch, Dictionary`2 param
eters, Dataset dtrain, Dataset dvalid, Int32 numIteration, Boolean verboseEval, Int32 earlyStoppingRound)
   at Microsoft.ML.Trainers.LightGbm.LightGbmTrainerBase`4.TrainCore(IChannel ch, IProgressChannel pch, Dataset dtrain,
CategoricalMetaData catMetaData, Dataset dvalid)
   at Microsoft.ML.Trainers.LightGbm.LightGbmTrainerBase`4.TrainModelCore(TrainContext context)
   at Microsoft.ML.Trainers.TrainerEstimatorBase`2.TrainTransformer(IDataView trainSet, IDataView validationSet, IPredic
tor initPredictor)
   at Microsoft.ML.Data.EstimatorChain`1.Fit(IDataView input)
   at Cats.DeepBookTrainer.Infrastructure.Trainer.ExecuteLightGbm(ITransformer& transformer, Nullable`1 maxIterations, N
ullable`1 maxThreads) 

Data Loading:

public static IDataView GetDataViewAsVector(MLContext mlContext, FileInfo trainingFile, int labelIndex=1, char[] separators = null, long? maxRows = null, bool makeBin=false)
{
    if (separators is null)
        separators = new[] { '|' };

    var loader = mlContext.Data.CreateTextLoader(options: new TextLoader.Options()
    {
        Columns = new[] {
            new TextLoader.Column(name:"Label", dataKind: DataKind.String, index: labelIndex),
            new TextLoader.Column(name:"RawFeatures",dataKind:DataKind.Single,minIndex:2,maxIndex:40732)
        },
        HasHeader = false,
        Separators = separators,
        UseThreads = false,
        MaxRows=maxRows
    });
    var dv = loader.Load(trainingFile.FullName);
    return dv;
}

Pipeline:

horizonDataset = mlContext.Data.TrainTestSplit(DataViewUtils.GetDataViewAsVector(mlContext, trainingFile, labelIndex: 1, separators: new[] {'|'}));

var options = new LightGbmMulticlassTrainer.Options
{
    LabelColumnName = "KeyColumn",
    FeatureColumnName = Features,
    Silent = false,
    Verbose = true,
    EvaluationMetric = LightGbmMulticlassTrainer.Options.EvaluateMetricType.Default,
    UseSoftmax       = true,
    NumberOfThreads  = 1,                
};

var featureColumns = Mapper.GetFieldNames();
var pipeline = mlContext.Transforms.Conversion.MapValueToKey(outputColumnName: "KeyColumn", inputColumnName: "Label")
        .Append(mlContext.Transforms.NormalizeMinMax(outputColumnName: "Features", inputColumnName: "RawFeatures"))
        .AppendCacheCheckpoint(mlContext)
        .Append(mlContext.MulticlassClassification.Trainers.LightGbm(options))
        .Append(mlContext.Transforms.CopyColumns(inputColumnName: "KeyColumn", outputColumnName: nameof(PredictedResult.PredictedLabelIndex)))

Line throwing the error
var model = pipeline.Fit(horizonDataset.TrainSet);

[PeterPann23]

Static constructor is always called, is it referencing another metric key in a another dictionary?

Is this related to the options specified, aka incompatible or missing, if so a exception should be raised