public void SequencePredictorSchemaTest()
{
int keyCount = 10;
var scoreColumnType = new KeyType(DataKind.U4, 0, keyCount);
VBuffer <ReadOnlyMemory <char> > keyNames = GenerateKeyNames(keyCount);
var sequenceSchema = ScoreSchemaFactory.CreateSequencePredictionSchema(scoreColumnType,
MetadataUtils.Const.ScoreColumnKind.SequenceClassification, keyNames);
// Output schema should only contain one column, which is the predicted label.
Assert.Single(sequenceSchema);
var scoreColumn = sequenceSchema[0];
// Check score column name.
Assert.Equal(MetadataUtils.Const.ScoreValueKind.PredictedLabel, scoreColumn.Name);
// Check score column type.
Assert.True(scoreColumn.Type.IsKey);
Assert.Equal((scoreColumnType as KeyType).Min, (scoreColumn.Type as KeyType).Min);
Assert.Equal((scoreColumnType as KeyType).Count, (scoreColumn.Type as KeyType).Count);
Assert.Equal((scoreColumnType as KeyType).RawKind, (scoreColumn.Type as KeyType).RawKind);
Assert.Equal((scoreColumnType as KeyType).Contiguous, (scoreColumn.Type as KeyType).Contiguous);
// Check metadata. Because keyNames is not empty, there should be three metadata fields.
var scoreMetadata = scoreColumn.Metadata;
Assert.Equal(3, scoreMetadata.Schema.Count);
// Check metadata columns' names.
Assert.Equal(MetadataUtils.Kinds.KeyValues, scoreMetadata.Schema[0].Name);
Assert.Equal(MetadataUtils.Kinds.ScoreColumnKind, scoreMetadata.Schema[1].Name);
Assert.Equal(MetadataUtils.Kinds.ScoreValueKind, scoreMetadata.Schema[2].Name);
// Check metadata columns' types.
Assert.True(scoreMetadata.Schema[0].Type.IsVector);
Assert.Equal(keyNames.Length, (scoreMetadata.Schema[0].Type as VectorType).VectorSize);
Assert.Equal(TextType.Instance, (scoreMetadata.Schema[0].Type as VectorType).ItemType);
Assert.Equal(TextType.Instance, scoreColumn.Metadata.Schema[1].Type);
Assert.Equal(TextType.Instance, scoreColumn.Metadata.Schema[2].Type);
// Check metadata columns' values.
var keyNamesGetter = scoreMetadata.GetGetter <VBuffer <ReadOnlyMemory <char> > >(0);
var actualKeyNames = new VBuffer <ReadOnlyMemory <char> >();
keyNamesGetter(ref actualKeyNames);
Assert.Equal(keyNames.Length, actualKeyNames.Length);
Assert.Equal(keyNames.DenseValues(), actualKeyNames.DenseValues());
var scoreColumnKindGetter = scoreMetadata.GetGetter <ReadOnlyMemory <char> >(1);
ReadOnlyMemory <char> scoreColumnKindValue = null;
scoreColumnKindGetter(ref scoreColumnKindValue);
Assert.Equal(MetadataUtils.Const.ScoreColumnKind.SequenceClassification, scoreColumnKindValue.ToString());
var scoreValueKindGetter = scoreMetadata.GetGetter <ReadOnlyMemory <char> >(2);
ReadOnlyMemory <char> scoreValueKindValue = null;
scoreValueKindGetter(ref scoreValueKindValue);
Assert.Equal(MetadataUtils.Const.ScoreValueKind.PredictedLabel, scoreValueKindValue.ToString());
}
public void SequencePredictorSchemaTest()
{
int keyCount = 10;
var expectedScoreColumnType = new KeyDataViewType(typeof(uint), keyCount);
VBuffer <ReadOnlyMemory <char> > keyNames = GenerateKeyNames(keyCount);
var sequenceSchema = ScoreSchemaFactory.CreateSequencePredictionSchema(expectedScoreColumnType,
AnnotationUtils.Const.ScoreColumnKind.SequenceClassification, keyNames);
// Output schema should only contain one column, which is the predicted label.
Assert.Single(sequenceSchema);
var scoreColumn = sequenceSchema[0];
// Check score column name.
Assert.Equal(AnnotationUtils.Const.ScoreValueKind.PredictedLabel, scoreColumn.Name);
// Check score column type.
var actualScoreColumnType = scoreColumn.Type as KeyDataViewType;
Assert.NotNull(actualScoreColumnType);
Assert.Equal(expectedScoreColumnType.Count, actualScoreColumnType.Count);
Assert.Equal(expectedScoreColumnType.RawType, actualScoreColumnType.RawType);
// Check metadata. Because keyNames is not empty, there should be three metadata fields.
var scoreMetadata = scoreColumn.Annotations;
Assert.Equal(3, scoreMetadata.Schema.Count);
// Check metadata columns' names.
Assert.Equal(AnnotationUtils.Kinds.KeyValues, scoreMetadata.Schema[0].Name);
Assert.Equal(AnnotationUtils.Kinds.ScoreColumnKind, scoreMetadata.Schema[1].Name);
Assert.Equal(AnnotationUtils.Kinds.ScoreValueKind, scoreMetadata.Schema[2].Name);
// Check metadata columns' types.
Assert.True(scoreMetadata.Schema[0].Type is VectorDataViewType);
Assert.Equal(keyNames.Length, (scoreMetadata.Schema[0].Type as VectorDataViewType).Size);
Assert.Equal(TextDataViewType.Instance, (scoreMetadata.Schema[0].Type as VectorDataViewType).ItemType);
Assert.Equal(TextDataViewType.Instance, scoreColumn.Annotations.Schema[1].Type);
Assert.Equal(TextDataViewType.Instance, scoreColumn.Annotations.Schema[2].Type);
// Check metadata columns' values.
var keyNamesGetter = scoreMetadata.GetGetter <VBuffer <ReadOnlyMemory <char> > >(scoreMetadata.Schema[0]);
var actualKeyNames = new VBuffer <ReadOnlyMemory <char> >();
keyNamesGetter(ref actualKeyNames);
Assert.Equal(keyNames.Length, actualKeyNames.Length);
Assert.Equal(keyNames.DenseValues(), actualKeyNames.DenseValues());
var scoreColumnKindGetter = scoreMetadata.GetGetter <ReadOnlyMemory <char> >(scoreMetadata.Schema[1]);
ReadOnlyMemory <char> scoreColumnKindValue = null;
scoreColumnKindGetter(ref scoreColumnKindValue);
Assert.Equal(AnnotationUtils.Const.ScoreColumnKind.SequenceClassification, scoreColumnKindValue.ToString());
var scoreValueKindGetter = scoreMetadata.GetGetter <ReadOnlyMemory <char> >(scoreMetadata.Schema[2]);
ReadOnlyMemory <char> scoreValueKindValue = null;
scoreValueKindGetter(ref scoreValueKindValue);
Assert.Equal(AnnotationUtils.Const.ScoreValueKind.PredictedLabel, scoreValueKindValue.ToString());
}
public BoundBase(SchemaBindablePipelineEnsembleBase parent, RoleMappedSchema schema)
{
Parent = parent;
InputRoleMappedSchema = schema;
OutputSchema = ScoreSchemaFactory.Create(Parent.ScoreType, Parent._scoreColumnKind);
_inputColIndices = new HashSet <int>();
for (int i = 0; i < Parent._inputCols.Length; i++)
{
var name = Parent._inputCols[i];
var col = InputRoleMappedSchema.Schema.GetColumnOrNull(name);
if (!col.HasValue)
{
throw Parent.Host.ExceptSchemaMismatch(nameof(InputRoleMappedSchema), "input", name);
}
_inputColIndices.Add(col.Value.Index);
}
Mappers = new ISchemaBoundRowMapper[Parent.PredictorModels.Length];
BoundPipelines = new IRowToRowMapper[Parent.PredictorModels.Length];
ScoreCols = new int[Parent.PredictorModels.Length];
for (int i = 0; i < Mappers.Length; i++)
{
// Get the RoleMappedSchema to pass to the predictor.
var emptyDv = new EmptyDataView(Parent.Host, schema.Schema);
Parent.PredictorModels[i].PrepareData(Parent.Host, emptyDv, out RoleMappedData rmd, out IPredictor predictor);
// Get the predictor as a bindable mapper, and bind it to the RoleMappedSchema found above.
var bindable = ScoreUtils.GetSchemaBindableMapper(Parent.Host, Parent.PredictorModels[i].Predictor);
Mappers[i] = bindable.Bind(Parent.Host, rmd.Schema) as ISchemaBoundRowMapper;
if (Mappers[i] == null)
{
throw Parent.Host.Except("Predictor {0} is not a row to row mapper", i);
}
// Make sure there is a score column, and remember its index.
var scoreCol = Mappers[i].OutputSchema.GetColumnOrNull(MetadataUtils.Const.ScoreValueKind.Score);
if (!scoreCol.HasValue)
{
throw Parent.Host.Except("Predictor {0} does not contain a score column", i);
}
ScoreCols[i] = scoreCol.Value.Index;
// Get the pipeline.
var dv = new EmptyDataView(Parent.Host, schema.Schema);
var tm = new TransformModelImpl(Parent.Host, dv, dv);
var pipeline = Parent.PredictorModels[i].TransformModel.Apply(Parent.Host, tm);
BoundPipelines[i] = pipeline.AsRowToRowMapper(Parent.Host);
if (BoundPipelines[i] == null)
{
throw Parent.Host.Except("Transform pipeline {0} contains transforms that do not implement IRowToRowMapper", i);
}
}
}
public void SequencePredictorSchemaWithoutKeyNamesMetadataTest()
{
int keyCount = 10;
var expectedScoreColumnType = new KeyType(DataKind.U4, 0, keyCount);
VBuffer <ReadOnlyMemory <char> > keyNames = GenerateKeyNames(0);
var sequenceSchema = ScoreSchemaFactory.CreateSequencePredictionSchema(expectedScoreColumnType,
MetadataUtils.Const.ScoreColumnKind.SequenceClassification, keyNames);
// Output schema should only contain one column, which is the predicted label.
Assert.Single(sequenceSchema);
var scoreColumn = sequenceSchema[0];
// Check score column name.
Assert.Equal(MetadataUtils.Const.ScoreValueKind.PredictedLabel, scoreColumn.Name);
// Check score column type.
var actualScoreColumnType = scoreColumn.Type as KeyType;
Assert.NotNull(actualScoreColumnType);
Assert.Equal(expectedScoreColumnType.Min, actualScoreColumnType.Min);
Assert.Equal(expectedScoreColumnType.Count, actualScoreColumnType.Count);
Assert.Equal(expectedScoreColumnType.RawType, actualScoreColumnType.RawType);
Assert.Equal(expectedScoreColumnType.Contiguous, actualScoreColumnType.Contiguous);
// Check metadata. Because keyNames is not empty, there should be three metadata fields.
var scoreMetadata = scoreColumn.Metadata;
Assert.Equal(2, scoreMetadata.Schema.Count);
// Check metadata columns' names.
Assert.Equal(MetadataUtils.Kinds.ScoreColumnKind, scoreMetadata.Schema[0].Name);
Assert.Equal(MetadataUtils.Kinds.ScoreValueKind, scoreMetadata.Schema[1].Name);
// Check metadata columns' types.
Assert.Equal(TextType.Instance, scoreColumn.Metadata.Schema[0].Type);
Assert.Equal(TextType.Instance, scoreColumn.Metadata.Schema[1].Type);
// Check metadata columns' values.
var scoreColumnKindGetter = scoreMetadata.GetGetter <ReadOnlyMemory <char> >(0);
ReadOnlyMemory <char> scoreColumnKindValue = null;
scoreColumnKindGetter(ref scoreColumnKindValue);
Assert.Equal(MetadataUtils.Const.ScoreColumnKind.SequenceClassification, scoreColumnKindValue.ToString());
var scoreValueKindGetter = scoreMetadata.GetGetter <ReadOnlyMemory <char> >(1);
ReadOnlyMemory <char> scoreValueKindValue = null;
scoreValueKindGetter(ref scoreValueKindValue);
Assert.Equal(MetadataUtils.Const.ScoreValueKind.PredictedLabel, scoreValueKindValue.ToString());
}
ISchemaBoundMapper ISchemaBindableMapper.Bind(IHostEnvironment env, RoleMappedSchema schema) => new FieldAwareFactorizationMachineScalarRowMapper(env, schema, ScoreSchemaFactory.CreateBinaryClassificationSchema(), this);
