public void TestI_DescribeTransformSaveDataAndZip()
{
/*using (*/ var env = EnvHelper.NewTestEnvironment();
{
var inputs = InputOutput.CreateInputs();
var data = DataViewConstructionUtils.CreateFromEnumerable(env, inputs);
var args = new DescribeTransform.Arguments()
{
columns = new[] { "X" }
};
var tr = new DescribeTransform(env, args, data);
var methodName = System.Reflection.MethodBase.GetCurrentMethod().Name;
var outputDataFilePath = FileHelper.GetOutputFile("outputDataFilePath.txt", methodName);
StreamHelper.SavePredictions(env, tr, outputDataFilePath);
Assert.IsTrue(File.Exists(outputDataFilePath));
var outModelFilePath = FileHelper.GetOutputFile("outModelFilePath.zip", methodName);
StreamHelper.SaveModel(env, tr, outModelFilePath);
Assert.IsTrue(File.Exists(outModelFilePath));
var outputDataFilePath2 = FileHelper.GetOutputFile("outputDataFilePath2.txt", methodName);
StreamHelper.SavePredictions(env, outModelFilePath, outputDataFilePath2, data);
Assert.IsTrue(File.Exists(outputDataFilePath2));
var d1 = File.ReadAllText(outputDataFilePath);
Assert.IsTrue(d1.Length > 0);
var d2 = File.ReadAllText(outputDataFilePath2);
Assert.AreEqual(d1, d2);
}
}
public override SchemaShape GetOutputSchema(SchemaShape inputSchema)
{
var addedCols = DataViewConstructionUtils.GetSchemaColumns(Transformer.AddedSchema);
var addedSchemaShape = SchemaShape.Create(SchemaBuilder.MakeSchema(addedCols));
var result = inputSchema.Columns.ToDictionary(x => x.Name);
var inputDef = InternalSchemaDefinition.Create(typeof(TSrc), Transformer.InputSchemaDefinition);
foreach (var col in inputDef.Columns)
{
if (!result.TryGetValue(col.ColumnName, out var column))
{
throw Contracts.ExceptSchemaMismatch(nameof(inputSchema), "input", col.ColumnName);
}
SchemaShape.GetColumnTypeShape(col.ColumnType, out var vecKind, out var itemType, out var isKey);
// Special treatment for vectors: if we expect variable vector, we also allow fixed-size vector.
if (itemType != column.ItemType || isKey != column.IsKey ||
vecKind == SchemaShape.Column.VectorKind.Scalar && column.Kind != SchemaShape.Column.VectorKind.Scalar ||
vecKind == SchemaShape.Column.VectorKind.Vector && column.Kind != SchemaShape.Column.VectorKind.Vector ||
vecKind == SchemaShape.Column.VectorKind.VariableVector && column.Kind == SchemaShape.Column.VectorKind.Scalar)
{
throw Contracts.ExceptSchemaMismatch(nameof(inputSchema), "input", col.ColumnName, col.ColumnType.ToString(), column.GetTypeString());
}
}
foreach (var addedCol in addedSchemaShape.Columns)
{
result[addedCol.Name] = addedCol;
}
return(new SchemaShape(result.Values));
}
public void TestLambdaColumnPassThroughTransform()
{
/*using (*/
var host = EnvHelper.NewTestEnvironment();
{
var inputs = new InputOutputU[] {
new InputOutputU()
{
X = new float[] { 0.1f, 1.1f }, Y = 0
},
new InputOutputU()
{
X = new float[] { 0.2f, 1.2f }, Y = 1
},
new InputOutputU()
{
X = new float[] { 0.3f, 1.3f }, Y = 2
}
};
var data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
var lambdaView = LambdaColumnHelper.Create <VBuffer <float>, VBuffer <float> >(host,
"Lambda", data, "X", "XX", new VectorDataViewType(NumberDataViewType.Single, 2),
new VectorDataViewType(NumberDataViewType.Single, 2),
(in VBuffer <float> src, ref VBuffer <float> dst) =>
{
dst = new VBuffer <float>(2, new float[2]);
dst.Values[0] = src.Values[0] + 1f;
dst.Values[1] = src.Values[1] + 1f;
});
public void TestChainTransformSerialize()
{
/*using (*/ var host = EnvHelper.NewTestEnvironment();
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 1, 10, 100 }
},
new ExampleA()
{
X = new float[] { 2, 3, 5 }
}
};
IDataView loader = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
IDataTransform data = host.CreateTransform("Scaler{col=X4:X}", loader);
data = host.CreateTransform("ChainTrans{ xf1=Scaler{col=X2:X} xf2=Poly{col=X3:X2} }", data);
// We create a specific folder in build/UnitTest which will contain the output.
var methodName = System.Reflection.MethodBase.GetCurrentMethod().Name;
var outModelFilePath = FileHelper.GetOutputFile("outModelFilePath.zip", methodName);
var outData = FileHelper.GetOutputFile("outData.txt", methodName);
var outData2 = FileHelper.GetOutputFile("outData2.txt", methodName);
TestTransformHelper.SerializationTestTransform(host, outModelFilePath, data, loader, outData, outData2);
}
}
public void TestI_DescribeTransformCode()
{
/*using (*/ var env = EnvHelper.NewTestEnvironment();
{
var inputs = InputOutput.CreateInputs();
var data = DataViewConstructionUtils.CreateFromEnumerable(env, inputs);
var args = new DescribeTransform.Arguments()
{
columns = new[] { "X" }
};
var tr = new DescribeTransform(env, args, data);
var values = new List <int>();
using (var cursor = tr.GetRowCursor(tr.Schema))
{
var columnGetter = cursor.GetGetter <int>(SchemaHelper._dc(1, cursor));
while (cursor.MoveNext())
{
int got = 0;
columnGetter(ref got);
values.Add((int)got);
}
}
Assert.AreEqual(values.Count, 4);
}
}
public void TestTagViewTransform()
{
/*using (*/ var host = EnvHelper.NewTestEnvironment();
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 0, 1 }
},
new ExampleA()
{
X = new float[] { 2, 3 }
}
};
IDataView loader = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
var data = host.CreateTransform("Scaler{col=X1:X}", loader);
data = host.CreateTransform("tag{t=memory}", data);
var methodName = System.Reflection.MethodBase.GetCurrentMethod().Name;
var outModelFilePath = FileHelper.GetOutputFile("outModelFilePath.zip", methodName);
var outData = FileHelper.GetOutputFile("outData.txt", methodName);
var outData2 = FileHelper.GetOutputFile("outData2.txt", methodName);
TestTransformHelper.SerializationTestTransform(host, outModelFilePath, data, loader, outData, outData2);
}
}
/// <summary>
/// Create a new <see cref="IDataView"/> over an enumerable 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 support
/// multiple enumerations of the <paramref name="data"/> that would return the same results, unless
/// the user knows that the data will only be cursored once.
///
/// One typical usage for streaming data view could be: create the data view that lazily loads data
/// as needed, then apply pre-trained transformations to it and cursor through it for transformation
/// results.
/// </summary>
/// <typeparam name="TRow">The user-defined item type.</typeparam>
/// <param name="data">The enumerable data containing type <typeparamref name="TRow"/> to convert to an<see cref="IDataView"/>.</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>
/// <example>
/// <format type="text/markdown">
/// <]
/// ]]>
/// </format>
/// </example>
public IDataView LoadFromEnumerable <TRow>(IEnumerable <TRow> data, SchemaDefinition schemaDefinition = null)
where TRow : class
{
_env.CheckValue(data, nameof(data));
_env.CheckValueOrNull(schemaDefinition);
return(DataViewConstructionUtils.CreateFromEnumerable(_env, data, schemaDefinition));
}
/// <summary>
/// Create a new <see cref="IDataView"/> over an enumerable of the items of user-defined type using the provided <see cref="DataViewSchema"/>,
/// which might contain more information about the schema than the type can capture.
/// </summary>
/// <remarks>
/// 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 support
/// multiple enumerations of the <paramref name="data"/> that would return the same results, unless
/// the user knows that the data will only be cursored once.
/// One typical usage for streaming data view could be: create the data view that lazily loads data
/// as needed, then apply pre-trained transformations to it and cursor through it for transformation
/// results.
/// One practical usage of this would be to supply the feature column names through the <see cref="DataViewSchema.Annotations"/>.
/// </remarks>
/// <typeparam name="TRow">The user-defined item type.</typeparam>
/// <param name="data">The enumerable data containing type <typeparamref name="TRow"/> to convert to an <see cref="IDataView"/>.</param>
/// <param name="schema">The schema of the returned <see cref="IDataView"/>.</param>
/// <returns>An <see cref="IDataView"/> with the given <paramref name="schema"/>.</returns>
public IDataView LoadFromEnumerable <TRow>(IEnumerable <TRow> data, DataViewSchema schema)
where TRow : class
{
_env.CheckValue(data, nameof(data));
_env.CheckValue(schema, nameof(schema));
return(DataViewConstructionUtils.CreateFromEnumerable(_env, data, schema));
}
/// <summary>
/// Create a prediction engine for one-time prediction.
/// It's mainly used in conjunction with <see cref="Load(Stream, out DataViewSchema)"/>,
/// where input schema is extracted during loading the model.
/// </summary>
/// <typeparam name="TSrc">The class that defines the input data.</typeparam>
/// <typeparam name="TDst">The class that defines the output data.</typeparam>
/// <param name="transformer">The transformer to use for prediction.</param>
/// <param name="inputSchema">Input schema.</param>
public PredictionEngine <TSrc, TDst> CreatePredictionEngine <TSrc, TDst>(ITransformer transformer, DataViewSchema inputSchema)
where TSrc : class
where TDst : class, new()
{
return(transformer.CreatePredictionEngine <TSrc, TDst>(_env, false,
DataViewConstructionUtils.GetSchemaDefinition <TSrc>(_env, inputSchema)));
}
public void TestI_PolynomialTransformSerialize()
{
/*using (*/
var host = EnvHelper.NewTestEnvironment();
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 1, 10, 100 }
},
new ExampleA()
{
X = new float[] { 2, 3, 5 }
}
};
IDataView loader = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
var data = host.CreateTransform("poly{col=poly:X d=3}", loader);
// We create a specific folder in build/UnitTest which will contain the output.
var methodName = System.Reflection.MethodBase.GetCurrentMethod().Name;
var outModelFilePath = FileHelper.GetOutputFile("outModelFilePath.zip", methodName);
var outData = FileHelper.GetOutputFile("outData.txt", methodName);
var outData2 = FileHelper.GetOutputFile("outData2.txt", methodName);
// This function serializes the output data twice, once before saving the pipeline, once after loading the pipeline.
// It checks it gives the same result.
TestTransformHelper.SerializationTestTransform(host, outModelFilePath, data, loader, outData, outData2);
}
}
public void TestSelectTagContactViewTransform()
{
var methodName = System.Reflection.MethodBase.GetCurrentMethod().Name;
var firstData = FileHelper.GetOutputFile("first.idv", methodName);
var outData = FileHelper.GetOutputFile("outData.txt", methodName);
var outData2 = FileHelper.GetOutputFile("outData2.txt", methodName);
/*using (*/ var env = EnvHelper.NewTestEnvironment();
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 0, 1, 4 }
},
new ExampleA()
{
X = new float[] { 2, 3, 7 }
}
};
// Create IDV
IDataView loader = DataViewConstructionUtils.CreateFromEnumerable(env, inputs);
var saver = ComponentCreation.CreateSaver(env, "binary");
using (var ch = env.Start("save"))
{
using (var fs0 = env.CreateOutputFile(firstData))
DataSaverUtils.SaveDataView(ch, saver, loader, fs0, true);
// Create parallel pipeline
loader = DataViewConstructionUtils.CreateFromEnumerable(env, inputs);
var data = env.CreateTransform("Scaler{col=X1:X}", loader);
data = env.CreateTransform(string.Format("selecttag{{t=first s=second f={0}}}", firstData), data);
data = env.CreateTransform("Scaler{col=X1:X}", data);
var merged = env.CreateTransform("append{t=first}", data);
// Save the outcome
var text = env.CreateSaver("Text");
var columns = new int[merged.Schema.Count];
for (int i = 0; i < columns.Length; ++i)
{
columns[i] = i;
}
using (var fs2 = File.Create(outData))
text.SaveData(fs2, merged, columns);
// Final checking
var lines = File.ReadAllLines(outData);
if (!lines.Any())
{
throw new Exception("Empty file.");
}
if (lines.Length != 9)
{
throw new Exception("Some lines are missing.");
}
}
}
}
public void TestScikitAPI_SimplePredictor()
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 1, 10, 100 }
},
new ExampleA()
{
X = new float[] { 2, 3, 5 }
},
new ExampleA()
{
X = new float[] { 2, 4, 5 }
},
new ExampleA()
{
X = new float[] { 2, 4, 7 }
},
};
var inputs2 = new[] {
new ExampleA()
{
X = new float[] { -1, -10, -100 }
},
new ExampleA()
{
X = new float[] { -2, -3, -5 }
},
new ExampleA()
{
X = new float[] { 3, 4, 5 }
},
new ExampleA()
{
X = new float[] { 3, 4, 7 }
},
};
/*using (*/
var host = EnvHelper.NewTestEnvironment(conc: 1);
{
var data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
using (var pipe = new ScikitPipeline(new[] { "poly{col=X}" }, "km{k=2}", host))
{
var predictor = pipe.Train(data, feature: "X");
Assert.IsTrue(predictor != null);
var data2 = new StreamingDataFrame(DataViewConstructionUtils.CreateFromEnumerable(host, inputs2));
var predictions = pipe.Predict(data2);
var df = DataFrameIO.ReadView(predictions);
Assert.AreEqual(df.Shape, new Tuple <int, int>(4, 12));
var dfs = df.ToString();
var dfs2 = dfs.Replace("\n", ";");
Assert.IsTrue(dfs2.StartsWith("X.0,X.1,X.2,X.3,X.4,X.5,X.6,X.7,X.8,PredictedLabel,Score.0,Score.1;-1,-10,-100,1,10,100,100,1000,10000"));
}
}
}
/// <summary>
/// Constructor
/// </summary>
/// <param name="env">environment</param>
/// <param name="modelStream">stream</param>
/// <param name="output">name of the output column</param>
/// <param name="outputIsFloat">output is a gloat (true) or a vector of floats (false)</param>
/// <param name="conc">number of concurrency threads</param>
/// <param name="features">features name</param>
public ValueMapperPredictionEngineFloat(IHostEnvironment env, Stream modelStream,
string output = "Probability", bool outputIsFloat = true, int conc = 1,
string features = "Features")
{
_env = env;
if (_env == null)
{
throw Contracts.Except("env must not be null");
}
var inputs = new FloatVectorInput[0];
var view = DataViewConstructionUtils.CreateFromEnumerable <FloatVectorInput>(_env, inputs);
long modelPosition = modelStream.Position;
_predictor = ModelFileUtils.LoadPredictorOrNull(_env, modelStream);
if (_predictor == null)
{
throw _env.Except("Unable to load a model.");
}
modelStream.Seek(modelPosition, SeekOrigin.Begin);
_transforms = ModelFileUtils.LoadTransforms(_env, view, modelStream);
if (_transforms == null)
{
throw _env.Except("Unable to load a model.");
}
var data = _env.CreateExamples(_transforms, features);
if (data == null)
{
throw _env.Except("Cannot create rows.");
}
var scorer = _env.CreateDefaultScorer(data, _predictor);
if (scorer == null)
{
throw _env.Except("Cannot create a scorer.");
}
_valueMapper = new ValueMapperFromTransformFloat <VBuffer <float> >(_env,
scorer, features, output, conc: conc);
if (_valueMapper == null)
{
throw _env.Except("Cannot create a mapper.");
}
if (outputIsFloat)
{
_mapper = _valueMapper.GetMapper <VBuffer <float>, float>();
_mapperVector = null;
}
else
{
_mapper = null;
_mapperVector = _valueMapper.GetMapper <VBuffer <float>, VBuffer <float> >();
}
}
void TestDnnImageFeaturizer()
{
// Onnxruntime supports Ubuntu 16.04, but not CentOS
// Do not execute on CentOS image
if (!RuntimeInformation.IsOSPlatform(OSPlatform.Windows))
{
return;
}
var samplevector = GetSampleArrayData();
var dataView = DataViewConstructionUtils.CreateFromList(Env,
new TestData[] {
new TestData()
{
data_0 = samplevector
},
});
var xyData = new List <TestDataXY> {
new TestDataXY()
{
A = new float[inputSize]
}
};
var stringData = new List <TestDataDifferntType> {
new TestDataDifferntType()
{
data_0 = new string[inputSize]
}
};
var sizeData = new List <TestDataSize> {
new TestDataSize()
{
data_0 = new float[2]
}
};
var pipe = new DnnImageFeaturizerEstimator(Env, "output_1", m => m.ModelSelector.ResNet18(m.Environment, m.OutputColumn, m.InputColumn), "data_0");
var invalidDataWrongNames = ML.Data.ReadFromEnumerable(xyData);
var invalidDataWrongTypes = ML.Data.ReadFromEnumerable(stringData);
var invalidDataWrongVectorSize = ML.Data.ReadFromEnumerable(sizeData);
TestEstimatorCore(pipe, dataView, invalidInput: invalidDataWrongNames);
TestEstimatorCore(pipe, dataView, invalidInput: invalidDataWrongTypes);
pipe.GetOutputSchema(SchemaShape.Create(invalidDataWrongVectorSize.Schema));
try
{
pipe.Fit(invalidDataWrongVectorSize);
Assert.False(true);
}
catch (ArgumentOutOfRangeException) { }
catch (InvalidOperationException) { }
}
public void TestDnnImageFeaturizer()
{
//skip running for x86 as this test using too much memory (over 2GB limit on x86)
//and very like to hit memory related issue when running on CI
//TODO: optimized memory usage in related code and enable x86 test run
if (!Environment.Is64BitProcess)
{
return;
}
var samplevector = GetSampleArrayData();
var dataView = DataViewConstructionUtils.CreateFromList(Env,
new TestData[] {
new TestData()
{
data_0 = samplevector
},
});
var xyData = new List <TestDataXY> {
new TestDataXY()
{
A = new float[InputSize]
}
};
var stringData = new List <TestDataDifferntType> {
new TestDataDifferntType()
{
data_0 = new string[InputSize]
}
};
var sizeData = new List <TestDataSize> {
new TestDataSize()
{
data_0 = new float[2]
}
};
var pipe = ML.Transforms.DnnFeaturizeImage("output_1", m => m.ModelSelector.ResNet18(m.Environment, m.OutputColumn, m.InputColumn), "data_0");
var invalidDataWrongNames = ML.Data.LoadFromEnumerable(xyData);
var invalidDataWrongTypes = ML.Data.LoadFromEnumerable(stringData);
var invalidDataWrongVectorSize = ML.Data.LoadFromEnumerable(sizeData);
TestEstimatorCore(pipe, dataView, invalidInput: invalidDataWrongNames);
TestEstimatorCore(pipe, dataView, invalidInput: invalidDataWrongTypes);
pipe.GetOutputSchema(SchemaShape.Create(invalidDataWrongVectorSize.Schema));
try
{
pipe.Fit(invalidDataWrongVectorSize);
Assert.False(true);
}
catch (ArgumentOutOfRangeException) { }
catch (InvalidOperationException) { }
}
/// <summary>
/// The 'reapply' constructor.
/// </summary>
private StatefulFilterTransform(IHostEnvironment env, StatefulFilterTransform <TSrc, TDst, TState> transform, IDataView newSource)
{
_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(newSource.Schema, DataViewConstructionUtils.GetSchemaColumns(_addedSchema));
}
public PredictionEngineExample(string modelName)
{
_env = EnvHelper.NewTestEnvironment();
var view = DataViewConstructionUtils.CreateFromEnumerable(_env, new FloatVectorInput[] { });
var pipe = DataViewConstructionUtils.LoadPipeWithPredictor(_env, File.OpenRead(modelName),
new EmptyDataView(_env, view.Schema));
var transformer = new TransformWrapper(_env, pipe);
_predictor = _env.CreatePredictionEngine <FloatVectorInput, FloatOutput>(transformer);
}
private static Func <DataViewSchema, 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(((ITransformer)transformer).IsRowToRowMapper, nameof(transformer), "Must be a row to row mapper");
return ((ITransformer)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);
}
public override SchemaShape GetOutputSchema(SchemaShape inputSchema)
{
var addedCols = DataViewConstructionUtils.GetSchemaColumns(Transformer.AddedSchema);
var addedSchemaShape = SchemaShape.Create(new Schema(addedCols));
var result = inputSchema.Columns.ToDictionary(x => x.Name);
foreach (var addedCol in addedSchemaShape.Columns)
{
result[addedCol.Name] = addedCol;
}
return(new SchemaShape(result.Values));
}
public void Testl_ShakeInputTransformVectorAdd()
{
/*using (*/ var host = EnvHelper.NewTestEnvironment();
{
var inputs = new[] {
new SHExampleA()
{
X = new float[] { 0, 1 }
},
new SHExampleA()
{
X = new float[] { 2, 3 }
}
};
var data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
var args = new ShakeInputTransform.Arguments
{
inputColumn = "X",
inputFeaturesInt = new[] { 0, 1 },
outputColumns = new[] { "yo" },
values = "-10,10;-100,100",
aggregation = ShakeInputTransform.ShakeAggregation.add
};
var trv = new ExampleValueMapperVector();
if (trv == null)
{
throw new Exception("Invalid");
}
var shake = new ShakeInputTransform(host, args, data, new IValueMapper[] { trv });
using (var cursor = shake.GetRowCursor(shake.Schema))
{
var outValues = new List <float>();
var colGetter = cursor.GetGetter <VBuffer <float> >(SchemaHelper._dc(1, cursor));
while (cursor.MoveNext())
{
VBuffer <float> got = new VBuffer <float>();
colGetter(ref got);
outValues.AddRange(got.DenseValues());
}
if (outValues.Count != 4)
{
throw new Exception("expected 4");
}
}
}
}
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 void TestResampleTransform(float ratio)
{
/*using (*/ var env = EnvHelper.NewTestEnvironment(conc: 1);
{
var inputs = new InputOutput[] {
new InputOutput()
{
X = new float[] { 0, 1 }, Y = 1
},
new InputOutput()
{
X = new float[] { 0, 1 }, Y = 0
},
new InputOutput()
{
X = new float[] { 0, 1 }, Y = 2
},
new InputOutput()
{
X = new float[] { 0, 1 }, Y = 3
},
};
var data = DataViewConstructionUtils.CreateFromEnumerable(env, inputs);
var args = new ResampleTransform.Arguments {
lambda = ratio, cache = false
};
var tr = new ResampleTransform(env, args, data);
var values = new List <int>();
using (var cursor = tr.GetRowCursor(tr.Schema))
{
var columnGetter = cursor.GetGetter <int>(SchemaHelper._dc(1, cursor));
while (cursor.MoveNext())
{
int got = 0;
columnGetter(ref got);
values.Add((int)got);
}
}
if (ratio < 1 && values.Count > 8)
{
throw new Exception("ResampleTransform did not work.");
}
if (ratio > 1 && values.Count < 1)
{
throw new Exception("ResampleTransform did not work.");
}
}
}
public ValueMapperExample(string modelName, string features)
{
_env = EnvHelper.NewTestEnvironment();
_predictor = ModelFileUtils.LoadPredictorOrNull(_env, File.OpenRead(modelName));
var inputs = new Input[0];
var view = DataViewConstructionUtils.CreateFromEnumerable(_env, inputs);
_transforms = ModelFileUtils.LoadTransforms(_env, view, File.OpenRead(modelName));
var data = _env.CreateExamples(_transforms, features);
var scorer = _env.CreateDefaultScorer(data, _predictor);
_valueMapper = new ValueMapperFromTransformFloat <VBuffer <float> >(_env, scorer, "Features", "Probability");
_mapper = _valueMapper.GetMapper <VBuffer <float>, float>();
}
void TestDnnImageFeaturizer()
{
var samplevector = GetSampleArrayData();
var dataView = DataViewConstructionUtils.CreateFromList(Env,
new TestData[] {
new TestData()
{
data_0 = samplevector
},
});
var xyData = new List <TestDataXY> {
new TestDataXY()
{
A = new float[inputSize]
}
};
var stringData = new List <TestDataDifferntType> {
new TestDataDifferntType()
{
data_0 = new string[inputSize]
}
};
var sizeData = new List <TestDataSize> {
new TestDataSize()
{
data_0 = new float[2]
}
};
var pipe = ML.Transforms.DnnFeaturizeImage("output_1", m => m.ModelSelector.ResNet18(m.Environment, m.OutputColumn, m.InputColumn), "data_0");
var invalidDataWrongNames = ML.Data.LoadFromEnumerable(xyData);
var invalidDataWrongTypes = ML.Data.LoadFromEnumerable(stringData);
var invalidDataWrongVectorSize = ML.Data.LoadFromEnumerable(sizeData);
TestEstimatorCore(pipe, dataView, invalidInput: invalidDataWrongNames);
TestEstimatorCore(pipe, dataView, invalidInput: invalidDataWrongTypes);
pipe.GetOutputSchema(SchemaShape.Create(invalidDataWrongVectorSize.Schema));
try
{
pipe.Fit(invalidDataWrongVectorSize);
Assert.False(true);
}
catch (ArgumentOutOfRangeException) { }
catch (InvalidOperationException) { }
}
public void TestScikitAPI_DelegateEnvironmentVerbose0()
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 1, 10, 100 }
},
new ExampleA()
{
X = new float[] { 2, 3, 5 }
}
};
var inputs2 = new[] {
new ExampleA()
{
X = new float[] { -1, -10, -100 }
},
new ExampleA()
{
X = new float[] { -2, -3, -5 }
}
};
var stdout = new List <string>();
var stderr = new List <string>();
ILogWriter logout = new LogWriter(s => stdout.Add(s));
ILogWriter logerr = new LogWriter(s => stderr.Add(s));
/*using (*/
var host = new DelegateEnvironment(seed: 0, outWriter: logout, errWriter: logerr, verbose: 0);
{
ComponentHelper.AddStandardComponents(host);
var data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
using (var pipe = new ScikitPipeline(new[] { "poly{col=X}" }, "km{k=2}", host: host))
{
var predictor = pipe.Train(data, feature: "X");
Assert.IsTrue(predictor != null);
}
}
Assert.AreEqual(stdout.Count, 0);
Assert.AreEqual(stderr.Count, 0);
}
public void TestI_PolynomialTransformSparse()
{
var inputs = new[] {
new ExampleASparse()
{
X = new VBuffer <float> (5, 3, new float[] { 1, 10, 100 }, new int[] { 0, 2, 4 })
},
new ExampleASparse()
{
X = new VBuffer <float> (5, 3, new float[] { 2, 3, 5 }, new int[] { 1, 2, 3 })
}
};
/*using (*/
var host = EnvHelper.NewTestEnvironment();
{
var data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
List <float[]> values;
CommonTestPolynomialTransform(host, data, 5, out values);
List <float[]> valuesDense;
data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
CommonTestPolynomialTransform(host, data, 5, out valuesDense);
if (values.Count != valuesDense.Count)
{
throw new Exception("Mismath in number of observations.");
}
for (int i = 0; i < values.Count; ++i)
{
if (values[i].Length != valuesDense[i].Length)
{
throw new Exception("Mismath in dimensions.");
}
for (int j = 0; j < values[i].Length; ++j)
{
if (values[i][j] != valuesDense[i][j])
{
throw new Exception("Mismath in value.");
}
}
}
}
}
public void TestScikitAPI_SimpleTransform()
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 1, 10, 100 }
},
new ExampleA()
{
X = new float[] { 2, 3, 5 }
}
};
var inputs2 = new[] {
new ExampleA()
{
X = new float[] { -1, -10, -100 }
},
new ExampleA()
{
X = new float[] { -2, -3, -5 }
}
};
/*using (*/
var host = EnvHelper.NewTestEnvironment(conc: 1);
{
var data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
using (var pipe = new ScikitPipeline(new[] { "poly{col=X}" }, host: host))
{
var predictor = pipe.Train(data);
Assert.IsTrue(predictor != null);
var data2 = DataViewConstructionUtils.CreateFromEnumerable(host, inputs2);
var predictions = pipe.Transform(data2);
var df = DataFrameIO.ReadView(predictions);
Assert.AreEqual(df.Shape, new Tuple <int, int>(2, 9));
var dfs = df.ToString();
var dfs2 = dfs.Replace("\n", ";");
Assert.AreEqual(dfs2, "X.0,X.1,X.2,X.3,X.4,X.5,X.6,X.7,X.8;-1,-10,-100,1,10,100,100,1000,10000;-2,-3,-5,4,6,10,9,15,25");
}
}
}
public void TestI_ScalerTransformDenseMinMax()
{
var inputs = new[] {
new ExampleA()
{
X = new float[] { 1, 10, 100 }
},
new ExampleA()
{
X = new float[] { 2, 3, 5 }
}
};
/*using (*/
var host = EnvHelper.NewTestEnvironment();
{
var data = DataViewConstructionUtils.CreateFromEnumerable(host, inputs);
List <float[]> values;
CommonTestScalerTransform(host, data, 3, ScalerTransform.ScalerStrategy.minMax, out values);
}
}
