internal BatchPredictionEngine(IHostEnvironment env, Stream modelStream, bool ignoreMissingColumns,
SchemaDefinition inputSchemaDefinition = null, SchemaDefinition outputSchemaDefinition = null)
{
Contracts.AssertValue(env);
Contracts.AssertValue(modelStream);
Contracts.AssertValueOrNull(inputSchemaDefinition);
Contracts.AssertValueOrNull(outputSchemaDefinition);
// Initialize pipe.
_srcDataView = DataViewConstructionUtils.CreateFromEnumerable(env, new TSrc[] { }, inputSchemaDefinition);
// Load transforms.
var pipe = env.LoadTransforms(modelStream, _srcDataView);
// Load predictor (if present) and apply default scorer.
// REVIEW: distinguish the case of predictor / no predictor?
var predictor = env.LoadPredictorOrNull(modelStream);
if (predictor != null)
{
var roles = ModelFileUtils.LoadRoleMappingsOrNull(env, modelStream);
pipe = roles != null
? env.CreateDefaultScorer(RoleMappedData.CreateOpt(pipe, roles), predictor)
: env.CreateDefaultScorer(env.CreateExamples(pipe, "Features"), predictor);
}
_pipeEngine = new PipeEngine <TDst>(env, pipe, ignoreMissingColumns, outputSchemaDefinition);
}
/// <summary>
/// Create a new <see cref="IDataView"/> over an in-memory collection of the items of user-defined type.
/// The user maintains ownership of the <paramref name="data"/> and the resulting data view will
/// never alter the contents of the <paramref name="data"/>.
/// Since <see cref="IDataView"/> is assumed to be immutable, the user is expected to not
/// modify the contents of <paramref name="data"/> while the data view is being actively cursored.
///
/// One typical usage for in-memory data view could be: create the data view, train a predictor.
/// Once the predictor is fully trained, modify the contents of the underlying collection and
/// train another predictor.
/// </summary>
/// <typeparam name="TRow">The user-defined item type.</typeparam>
/// <param name="env">The host environment to use for data view creation.</param>
/// <param name="data">The data to wrap around.</param>
/// <param name="schemaDefinition">The optional schema definition of the data view to create. If <c>null</c>,
/// the schema definition is inferred from <typeparamref name="TRow"/>.</param>
/// <returns>The constructed <see cref="IDataView"/>.</returns>
public static IDataView CreateDataView <TRow>(this IHostEnvironment env, IList <TRow> data, SchemaDefinition schemaDefinition = null)
where TRow : class
{
Contracts.CheckValue(env, nameof(env));
env.CheckValue(data, nameof(data));
env.CheckValueOrNull(schemaDefinition);
return(DataViewConstructionUtils.CreateFromList(env, data, schemaDefinition));
}
private static Func <Schema, IRowToRowMapper> StreamChecker(IHostEnvironment env, Stream modelStream)
{
env.CheckValue(modelStream, nameof(modelStream));
return(schema =>
{
var pipe = DataViewConstructionUtils.LoadPipeWithPredictor(env, modelStream, new EmptyDataView(env, schema));
var transformer = new TransformWrapper(env, pipe);
env.CheckParam(transformer.IsRowToRowMapper, nameof(transformer), "Must be a row to row mapper");
return transformer.GetRowToRowMapper(schema);
});
}
private protected PredictionEngineBase(IHostEnvironment env, ITransformer transformer, bool ignoreMissingColumns,
SchemaDefinition inputSchemaDefinition = null, SchemaDefinition outputSchemaDefinition = null)
{
Contracts.CheckValue(env, nameof(env));
env.AssertValue(transformer);
Transformer = transformer;
var makeMapper = TransformerChecker(env, transformer);
env.AssertValue(makeMapper);
_inputRow = DataViewConstructionUtils.CreateInputRow <TSrc>(env, inputSchemaDefinition);
PredictionEngineCore(env, _inputRow, makeMapper(_inputRow.Schema), ignoreMissingColumns, inputSchemaDefinition, outputSchemaDefinition, out _disposer, out _outputRow);
}
/// <summary>
/// The 'reapply' constructor.
/// </summary>
private StatefulFilterTransform(IHostEnvironment env, StatefulFilterTransform <TSrc, TDst, TState> transform, IDataView newSource)
: base(env, RegistrationName, transform)
{
Host.AssertValue(transform);
Host.AssertValue(newSource);
_source = newSource;
_filterFunc = transform._filterFunc;
_typedSource = TypedCursorable <TSrc> .Create(Host, newSource, false, transform._inputSchemaDefinition);
_addedSchema = transform._addedSchema;
_bindings = new ColumnBindings(Schema.Create(newSource.Schema), DataViewConstructionUtils.GetSchemaColumns(_addedSchema));
}
/// <summary>
/// The 'reapply' constructor.
/// </summary>
private MapTransform(IHostEnvironment env, MapTransform <TSrc, TDst> transform, IDataView newSource)
: base(env, RegistrationName, transform)
{
Host.AssertValue(transform);
Host.AssertValue(newSource);
Source = newSource;
_mapAction = transform._mapAction;
_typedSource = TypedCursorable <TSrc> .Create(Host, newSource, false, transform._inputSchemaDefinition);
_addedSchema = transform._addedSchema;
_bindings = new ColumnBindings(Data.Schema.Create(newSource.Schema), DataViewConstructionUtils.GetSchemaColumns(_addedSchema));
}
internal BatchPredictionEngine(IHostEnvironment env, IDataView dataPipeline, bool ignoreMissingColumns,
SchemaDefinition inputSchemaDefinition = null, SchemaDefinition outputSchemaDefinition = null)
{
Contracts.AssertValue(env);
Contracts.AssertValue(dataPipeline);
Contracts.AssertValueOrNull(inputSchemaDefinition);
Contracts.AssertValueOrNull(outputSchemaDefinition);
// Initialize pipe.
_srcDataView = DataViewConstructionUtils.CreateFromEnumerable(env, new TSrc[] { }, inputSchemaDefinition);
var pipe = ApplyTransformUtils.ApplyAllTransformsToData(env, dataPipeline, _srcDataView);
_pipeEngine = new PipeEngine <TDst>(env, pipe, ignoreMissingColumns, outputSchemaDefinition);
}
internal BatchPredictionEngine(IHostEnvironment env, Stream modelStream, bool ignoreMissingColumns,
SchemaDefinition inputSchemaDefinition = null, SchemaDefinition outputSchemaDefinition = null)
{
Contracts.AssertValue(env);
Contracts.AssertValue(modelStream);
Contracts.AssertValueOrNull(inputSchemaDefinition);
Contracts.AssertValueOrNull(outputSchemaDefinition);
// Initialize pipe.
_srcDataView = DataViewConstructionUtils.CreateFromEnumerable(env, new TSrc[] { }, inputSchemaDefinition);
var pipe = DataViewConstructionUtils.LoadPipeWithPredictor(env, modelStream, _srcDataView);
_pipeEngine = new PipeEngine <TDst>(env, pipe, ignoreMissingColumns, outputSchemaDefinition);
}
private PredictionEngine(IHostEnvironment env, Func <ISchema, IRowToRowMapper> makeMapper, bool ignoreMissingColumns,
SchemaDefinition inputSchemaDefinition, SchemaDefinition outputSchemaDefinition)
{
Contracts.CheckValue(env, nameof(env));
env.AssertValue(makeMapper);
_inputRow = DataViewConstructionUtils.CreateInputRow <TSrc>(env, inputSchemaDefinition);
var mapper = makeMapper(_inputRow.Schema);
var cursorable = TypedCursorable <TDst> .Create(env, new EmptyDataView(env, mapper.Schema), ignoreMissingColumns, outputSchemaDefinition);
var outputRow = mapper.GetRow(_inputRow, col => true, out _disposer);
_outputRow = cursorable.GetRow(outputRow);
}
/// <summary>
/// Create a a map transform that is savable iff <paramref name="saveAction"/> and <paramref name="loadFunc"/> are
/// not null.
/// </summary>
/// <param name="env">The host environment</param>
/// <param name="source">The dataview upon which we construct the transform</param>
/// <param name="mapAction">The action by which we map source to destination columns</param>
/// <param name="saveAction">An action that allows us to save state to the serialization stream. May be
/// null simultaneously with <paramref name="loadFunc"/>.</param>
/// <param name="loadFunc">A function that given the serialization stream and a data view, returns
/// an <see cref="ITransformTemplate"/>. The intent is, this returned object should itself be a
/// <see cref="MapTransform{TSrc,TDst}"/>, but this is not strictly necessary. This delegate should be
/// a static non-lambda method that this assembly can legally call. May be null simultaneously with
/// <paramref name="saveAction"/>.</param>
/// <param name="inputSchemaDefinition">The schema definition overrides for <typeparamref name="TSrc"/></param>
/// <param name="outputSchemaDefinition">The schema definition overrides for <typeparamref name="TDst"/></param>
public MapTransform(IHostEnvironment env, IDataView source, Action <TSrc, TDst> mapAction,
Action <BinaryWriter> saveAction, LambdaTransform.LoadDelegate loadFunc,
SchemaDefinition inputSchemaDefinition = null, SchemaDefinition outputSchemaDefinition = null)
: base(env, RegistrationName, saveAction, loadFunc)
{
Host.AssertValue(source);
Host.AssertValue(mapAction);
Host.AssertValueOrNull(inputSchemaDefinition);
Host.AssertValueOrNull(outputSchemaDefinition);
Source = source;
_mapAction = mapAction;
_inputSchemaDefinition = inputSchemaDefinition;
_typedSource = TypedCursorable <TSrc> .Create(Host, Source, false, inputSchemaDefinition);
var outSchema = outputSchemaDefinition == null
? InternalSchemaDefinition.Create(typeof(TDst), SchemaDefinition.Direction.Write)
: InternalSchemaDefinition.Create(typeof(TDst), outputSchemaDefinition);
_addedSchema = outSchema;
_bindings = new ColumnBindings(Data.Schema.Create(Source.Schema), DataViewConstructionUtils.GetSchemaColumns(outSchema));
}
/// <summary>
/// Create a filter transform that is savable iff <paramref name="saveAction"/> and <paramref name="loadFunc"/> are
/// not null.
/// </summary>
/// <param name="env">The host environment</param>
/// <param name="source">The dataview upon which we construct the transform</param>
/// <param name="filterFunc">The function by which we transform source to destination columns and decide whether
/// to keep the row.</param>
/// <param name="initStateAction">The function that is called once per cursor to initialize state. Can be null.</param>
/// <param name="saveAction">An action that allows us to save state to the serialization stream. May be
/// null simultaneously with <paramref name="loadFunc"/>.</param>
/// <param name="loadFunc">A function that given the serialization stream and a data view, returns
/// an <see cref="ITransformTemplate"/>. The intent is, this returned object should itself be a
/// <see cref="CustomMappingTransformer{TSrc,TDst}"/>, but this is not strictly necessary. This delegate should be
/// a static non-lambda method that this assembly can legally call. May be null simultaneously with
/// <paramref name="saveAction"/>.</param>
/// <param name="inputSchemaDefinition">The schema definition overrides for <typeparamref name="TSrc"/></param>
/// <param name="outputSchemaDefinition">The schema definition overrides for <typeparamref name="TDst"/></param>
public StatefulFilterTransform(IHostEnvironment env, IDataView source, Func <TSrc, TDst, TState, bool> filterFunc,
Action <TState> initStateAction,
Action <BinaryWriter> saveAction, LambdaTransform.LoadDelegate loadFunc,
SchemaDefinition inputSchemaDefinition = null, SchemaDefinition outputSchemaDefinition = null)
: base(env, RegistrationName, saveAction, loadFunc)
{
Host.AssertValue(source, "source");
Host.AssertValue(filterFunc, "filterFunc");
Host.AssertValueOrNull(initStateAction);
Host.AssertValueOrNull(inputSchemaDefinition);
Host.AssertValueOrNull(outputSchemaDefinition);
_source = source;
_filterFunc = filterFunc;
_initStateAction = initStateAction;
_inputSchemaDefinition = inputSchemaDefinition;
_typedSource = TypedCursorable <TSrc> .Create(Host, Source, false, inputSchemaDefinition);
var outSchema = InternalSchemaDefinition.Create(typeof(TDst), outputSchemaDefinition);
_addedSchema = outSchema;
_bindings = new ColumnBindings(Schema.Create(Source.Schema), DataViewConstructionUtils.GetSchemaColumns(outSchema));
}
