What causes "Training data cannot be null" in dotnet/machinelearning?

Load a valid IDataView before calling Execute (e.g. mlContext.Data.LoadFromTextFile (path))

dotnet/machinelearning · critical · ArgumentNullException

Training data cannot be null

Error message

Training data cannot be null

What it means

Thrown by UserInputValidationUtil.ValidateTrainData when the training IDataView passed to an AutoML experiment is null. The library requires a non-null training dataset to run any experiment and reports the failure via ArgumentNullException naming the trainData parameter.

Source

Thrown at src/Microsoft.ML.AutoML/Utils/UserInputValidationUtil.cs:73

            if (numberOfCVFolds <= 1)
            {
                throw new ArgumentException($"{nameof(numberOfCVFolds)} must be at least 2", nameof(numberOfCVFolds));
            }
        }

        public static void ValidateSamplingKey(string samplingKeyColumnName, string groupIdColumnName, TaskKind task)
        {
            if (task == TaskKind.Ranking && samplingKeyColumnName != null && samplingKeyColumnName != groupIdColumnName)
            {
                throw new ArgumentException($"If provided, {nameof(samplingKeyColumnName)} must be the same as {nameof(groupIdColumnName)} for Ranking Experiments", samplingKeyColumnName);
            }
        }

        private static void ValidateTrainData(IDataView trainData, ColumnInformation columnInformation)
        {
            if (trainData == null)
            {
                throw new ArgumentNullException(nameof(trainData), "Training data cannot be null");
            }

            if (DatasetDimensionsUtil.IsDataViewEmpty(trainData))
            {
                throw new ArgumentException("Training data has 0 rows", nameof(trainData));
            }

            foreach (var column in trainData.Schema)
            {
                if (column.Name == DefaultColumnNames.Features && column.Type.GetItemType() != NumberDataViewType.Single)
                {
                    throw new ArgumentException($"{DefaultColumnNames.Features} column must be of data type {NumberDataViewType.Single}", nameof(trainData));
                }

                if ((column.Name != columnInformation.LabelColumnName &&
                    column.Name != columnInformation.UserIdColumnName &&
                    column.Name != columnInformation.ItemIdColumnName &&
                    column.Name != columnInformation.GroupIdColumnName)

View on GitHub (pinned to 7b76e69cf9)

Solutions

Load a valid IDataView before calling Execute (e.g. mlContext.Data.LoadFromTextFile<T>(path))
Add a null check on the training data before invoking the experiment
Fix the upstream data-loading code that silently returned null

Example fix

// before
IDataView trainData = LoadData(); // may return null
var result = experiment.Execute(trainData, labelColumnName, "A");

// after
IDataView trainData = LoadData() ?? mlContext.Data.LoadFromTextFile<ModelInput>(dataPath, hasHeader: true, separatorChar: ',');
if (trainData == null) throw new InvalidOperationException("No training data loaded");
var result = experiment.Execute(trainData, labelColumnName, "A");

Defensive patterns

Strategy: type-guard

Validate before calling

if (trainData is null) throw new InvalidOperationException("trainData must be loaded before Execute");

Type guard

static bool HasTrainingData(IDataView d) => d is not null && d.GetRowCursor(d.Schema).MoveNext();

Try / catch

try { var r = experiment.Execute(trainData, label, "A"); }
catch (ArgumentNullException ex) when (ex.ParamName == "trainData") { /* load default dataset and retry once */ }

Prevention

Never let data-loading helpers return null; throw early inside the loader
Load data immediately before Execute in the same method
Use nullable reference types (IDataView?) so the compiler flags null flows

When it happens

Trigger: Calling experiment.Execute(null, ...) or ValidateExperimentExecuteArgs with trainData = null, typically when a data-loading step returned null.

Common situations: LoadFromTextFile or a custom loader returning null after a failed read; a lazily-initialized dataset variable never assigned; refactoring that removed data loading but left the Execute call in place.

Related errors

AI-assisted analysis of dotnet/machinelearning@7b76e69cf9 (2026-09-11). Data as JSON: /api/errors/2eadda3f938aa13e. Report an issue: GitHub.