dotnet/machinelearning · error · ArgumentException

The trainer '{trainer}' is not handled currently.

Error message

The trainer '{trainer}' is not handled currently.

What it means

TrainerGeneratorFactory.GetInstance parses the AutoML PipelineNode name into the TrainerName enum and maps recognized trainers to C# code-generator classes. When the parsed trainer name is a valid TrainerName member but has no case in the switch, the default branch throws this ArgumentException. It means the code generator does not yet support emitting C# for that trainer.

Source

Thrown at src/Microsoft.ML.CodeGenerator/CodeGenerator/CSharp/TrainerGeneratorFactory.cs:77

                        return new StochasticDualCoordinateAscentMulti(node);
                    case TrainerName.SdcaRegression:
                        return new StochasticDualCoordinateAscentRegression(node);
                    case TrainerName.SgdCalibratedBinary:
                        return new SgdCalibratedBinary(node);
                    case TrainerName.SymbolicSgdLogisticRegressionBinary:
                        return new SymbolicSgdLogisticRegressionBinary(node);
                    case TrainerName.Ova:
                        return new OneVersusAll(node);
                    case TrainerName.ImageClassification:
                        return new ImageClassificationTrainer(node);
                    case TrainerName.MatrixFactorization:
                        return new MatrixFactorization(node);
                    case TrainerName.LightGbmRanking:
                        return new LightGbmRanking(node);
                    case TrainerName.FastTreeRanking:
                        return new FastTreeRanking(node);
                    default:
                        throw new ArgumentException($"The trainer '{trainer}' is not handled currently.");
                }
            }
            throw new ArgumentException($"The trainer '{node.Name}' is not handled currently.");
        }
    }
}

View on GitHub (pinned to 7b76e69cf9)

Solutions

  1. Check the trainer name in the message against the switch in TrainerGeneratorFactory.GetInstance; pick a different trainer in your AutoML experiment (exclude unsupported ones via trainers parameter/exclusion list).
  2. Upgrade Microsoft.ML.CodeGenerator (and mlnet CLI) to the latest version, which may add a case for this trainer.
  3. If you own the code, add a case mapping the TrainerName to a new ITrainerGenerator implementation.
  4. Handle the generated pipeline manually: inspect the AutoML pipeline and write the trainer C# code yourself.

Example fix

// before: AutoML experiment allowed all trainers
var experiment = mlContext.Auto().CreateBinaryClassificationExperiment();
// after: restrict to trainers the code generator supports
var experiment = mlContext.Auto().CreateBinaryClassificationExperiment(
    experimentSettings => experimentSettings.UseFractions(...));
// or exclude unsupported trainers via TrainPersonalityAndTrainer exclusion in pipeline sweeps
Defensive patterns

Strategy: try-catch

Validate before calling

if (!Enum.TryParse<TrainerName>(node.Name, out var t) ||
    !typeof(TrainerGeneratorFactory).Assembly.GetTypes().Any()) { /* precheck supported set */ }
var supported = new HashSet<string>(new[]{"LightGbmBinary","LightGbmMulti","LightGbmRegression","FastTreeBinary","FastTreeRegression","SdcaLogisticRegressionBinary","LbfgsLogisticRegressionBinary"});
bool canGenerate = supported.Contains(node.Name);

Type guard

static bool IsSupportedTrainer(string name) =>
    Enum.TryParse<TrainerName>(name, out var t) &&
    Enum.IsDefined(typeof(TrainerName), t) &&
    SupportedTrainers.Contains(t);

Try / catch

try
{
    var gen = TrainerGeneratorFactory.GetInstance(node);
}
catch (ArgumentException ex) when (ex.Message.Contains("is not handled currently"))
{
    Console.WriteLine($"Skipping trainer {node.Name}: code generation unsupported");
}

Prevention

When it happens

Trigger: Calling TrainerGeneratorFactory.GetInstance with a PipelineNode whose Name parses to a TrainerName that exists in the enum but is not one of the ~25 cases handled in the switch (e.g. a newly added AutoML trainer like a new LightGbm variant or HgBoosting) while running ML.NET AutoML-to-C# code generation.

Common situations: Developers using Microsoft.ML.CodeGenerator (mlnet CLI / AutoML 'convert to C#' flow) on a model whose best pipeline picked a trainer that the generator hasn't implemented; version skew where Microsoft.ML.AutoML gained trainers the CodeGenerator package doesn't know yet.

Understand the failure class

Background: UnsupportedOperationException and "is not supported" errors: when a library deliberately refuses a call — this error's family across 30 libraries.

Related errors

AI-assisted analysis of dotnet/machinelearning@7b76e69cf9 (2026-09-11). Data as JSON: /api/errors/618625d6830ee76f. Report an issue: GitHub.