/// <summary>
/// Initializes a new instance of the <see cref="OnnxModel"/> class.
/// </summary>
/// <param name="configuration">The configuration for the onnx model runner.</param>
public OnnxModel(OnnxModelConfiguration configuration)
{
this.configuration = configuration;
// The schemaDefinition is a ML.NET construct that allows us to specify the form
// of the inputs. In this case we construct a schema definition programmatically
// to reflect that the input is a vector of floats, of the sizes specified in the
// configuration
this.schemaDefinition = SchemaDefinition.Create(typeof(OnnxInputVector));
this.schemaDefinition[nameof(OnnxInputVector.Vector)].ColumnType = new VectorDataViewType(NumberDataViewType.Single, this.configuration.InputVectorSize);
this.schemaDefinition[nameof(OnnxInputVector.Vector)].ColumnName = this.configuration.InputVectorName;
// We create the onnxTransformer which will be used to score inputs
var onnxEmptyInputDataView = this.context.Data.LoadFromEnumerable(new List <OnnxInputVector>(), this.schemaDefinition);
var scoringEstimator =
this.context.Transforms.ApplyOnnxModel(
modelFile: configuration.ModelFileName,
outputColumnNames: new[] { configuration.OutputVectorName },
inputColumnNames: new[] { configuration.InputVectorName },
shapeDictionary: configuration.ShapeDictionary,
gpuDeviceId: configuration.GpuDeviceId,
fallbackToCpu: false);
this.onnxTransformer = scoringEstimator.Fit(onnxEmptyInputDataView);
}
/// <summary>
/// Initializes a new instance of the <see cref="MaskRCNNOnnxModel"/> class.
/// </summary>
/// <param name="imageWidth">Input image width.</param>
/// <param name="imageHeight">Input image height.</param>
/// <param name="modelFileName">Model file name.</param>
/// <param name="gpuDeviceId">GPU device ID to run execution on, or null to run on CPU.</param>
/// <remarks>
/// To run on a GPU, use the Microsoft.Psi.Onnx.Gpu library instead of Microsoft.Psi.Onnx.Cpu, and pass
/// a valid non-negative integer. Typical device ID values are 0 or 1.
/// </remarks>
public MaskRCNNOnnxModel(int imageWidth, int imageHeight, string modelFileName, int?gpuDeviceId = null)
{
//// REVIEW: note the ColumnType below is (float, 3, height, width), while the plain OnnxModel is simply (float, size)
//// the rest is very similar to OnnxModel
this.schemaDefinition = SchemaDefinition.Create(typeof(OnnxInputVector));
this.schemaDefinition[nameof(OnnxInputVector.Vector)].ColumnType = new VectorDataViewType(NumberDataViewType.Single, 3, imageHeight, imageWidth);
this.schemaDefinition[nameof(OnnxInputVector.Vector)].ColumnName = "image";
var onnxEmptyInputDataView = this.context.Data.LoadFromEnumerable(new List <OnnxInputVector>(), this.schemaDefinition);
var shapeDictionary = new Dictionary <string, int[]>()
{
{ "image", new int[] { 3, imageHeight, imageWidth } }
};
var scoringEstimator =
this.context.Transforms.ApplyOnnxModel(
modelFile: modelFileName,
inputColumnNames: new[] { "image" },
outputColumnNames: new[] { BOXES, LABELS, SCORES, MASKS },
shapeDictionary: shapeDictionary,
gpuDeviceId: gpuDeviceId,
fallbackToCpu: false);
this.onnxTransformer = scoringEstimator.Fit(onnxEmptyInputDataView);
}
public void OnnxModelScenario()
{
var modelFile = "squeezenet/00000001/model.onnx";
using (var env = new ConsoleEnvironment(seed: 1, conc: 1))
{
var samplevector = GetSampleArrayData();
var dataView = ML.Data.ReadFromEnumerable(
new TestData[] {
new TestData()
{
data_0 = samplevector
}
});
var onnx = new OnnxTransformer(env, "softmaxout_1", modelFile, "data_0").Transform(dataView);
onnx.Schema.TryGetColumnIndex("softmaxout_1", out int score);
using (var curs = onnx.GetRowCursor(onnx.Schema["softmaxout_1"]))
{
var getScores = curs.GetGetter <VBuffer <float> >(score);
var buffer = default(VBuffer <float>);
while (curs.MoveNext())
{
getScores(ref buffer);
Assert.Equal(1000, buffer.Length);
}
}
}
}
[ConditionalFact(typeof(Environment), nameof(Environment.Is64BitProcess))] // x86 output differs from Baseline
public void OnnxModelMultiInput()
{
if (!RuntimeInformation.IsOSPlatform(OSPlatform.Windows))
{
return;
}
var modelFile = @"twoinput\twoinput.onnx";
using (var env = new ConsoleEnvironment(seed: 1, conc: 1))
{
var samplevector = GetSampleArrayData();
var dataView = ComponentCreation.CreateDataView(Env,
new TestDataMulti[] {
new TestDataMulti()
{
ina = new float[] { 1, 2, 3, 4, 5 },
inb = new float[] { 1, 2, 3, 4, 5 }
}
});
var onnx = new OnnxTransformer(env, modelFile, new[] { "ina", "inb" }, new[] { "outa", "outb" }).Transform(dataView);
onnx.Schema.TryGetColumnIndex("outa", out int scoresa);
onnx.Schema.TryGetColumnIndex("outb", out int scoresb);
using (var curs = onnx.GetRowCursor(col => col == scoresa || col == scoresb))
{
var getScoresa = curs.GetGetter <VBuffer <float> >(scoresa);
var getScoresb = curs.GetGetter <VBuffer <float> >(scoresb);
var buffera = default(VBuffer <float>);
var bufferb = default(VBuffer <float>);
while (curs.MoveNext())
{
getScoresa(ref buffera);
getScoresb(ref bufferb);
Assert.Equal(5, buffera.Length);
Assert.Equal(5, bufferb.Length);
Assert.Equal(0, buffera.GetValues().ToArray().Sum());
Assert.Equal(30, bufferb.GetValues().ToArray().Sum());
}
}
}
}
public void OnnxModelMultiInput()
{
var modelFile = Path.Combine(Directory.GetCurrentDirectory(), "twoinput", "twoinput.onnx");
using (var env = new ConsoleEnvironment(seed: 1, conc: 1))
{
var samplevector = GetSampleArrayData();
var dataView = ML.Data.ReadFromEnumerable(
new TestDataMulti[] {
new TestDataMulti()
{
ina = new float[] { 1, 2, 3, 4, 5 },
inb = new float[] { 1, 2, 3, 4, 5 }
}
});
var onnx = new OnnxTransformer(env, new[] { "outa", "outb" }, new[] { "ina", "inb" }, modelFile).Transform(dataView);
onnx.Schema.TryGetColumnIndex("outa", out int scoresa);
onnx.Schema.TryGetColumnIndex("outb", out int scoresb);
using (var curs = onnx.GetRowCursor(onnx.Schema["outa"], onnx.Schema["outb"]))
{
var getScoresa = curs.GetGetter <VBuffer <float> >(scoresa);
var getScoresb = curs.GetGetter <VBuffer <float> >(scoresb);
var buffera = default(VBuffer <float>);
var bufferb = default(VBuffer <float>);
while (curs.MoveNext())
{
getScoresa(ref buffera);
getScoresb(ref bufferb);
Assert.Equal(5, buffera.Length);
Assert.Equal(5, bufferb.Length);
Assert.Equal(0, buffera.GetValues().ToArray().Sum());
Assert.Equal(30, bufferb.GetValues().ToArray().Sum());
}
}
}
}
[ConditionalFact(typeof(Environment), nameof(Environment.Is64BitProcess))] // x86 output differs from Baseline
public void OnnxModelScenario()
{
if (!RuntimeInformation.IsOSPlatform(OSPlatform.Windows))
{
return;
}
var modelFile = "squeezenet/00000001/model.onnx";
using (var env = new ConsoleEnvironment(seed: 1, conc: 1))
{
var samplevector = GetSampleArrayData();
var dataView = ComponentCreation.CreateDataView(Env,
new TestData[] {
new TestData()
{
data_0 = samplevector
}
});
var onnx = new OnnxTransformer(env, modelFile, "data_0", "softmaxout_1").Transform(dataView);
onnx.Schema.TryGetColumnIndex("softmaxout_1", out int score);
using (var curs = onnx.GetRowCursor(onnx.Schema["softmaxout_1"]))
{
var getScores = curs.GetGetter <VBuffer <float> >(score);
var buffer = default(VBuffer <float>);
while (curs.MoveNext())
{
getScores(ref buffer);
Assert.Equal(1000, buffer.Length);
}
}
}
}
public void TestOnnxTransformWithCustomShapes()
{
// The loaded model has input shape [-1, 3] and output shape [-1].
var modelFile = Path.Combine(Directory.GetCurrentDirectory(), "unknowndimensions", "test_unknowndimensions_float.onnx");
var dataPoints = new InputWithCustomShape[]
{
// It's a flattened 3-by-3 tensor.
// [1.1, 1.3, 1.2]
// |1.9, 1.3, 1.2|
// [1.1, 1.3, 1.8]
new InputWithCustomShape()
{
input = new float[] { 1.1f, 1.3f, 1.2f, 1.9f, 1.3f, 1.2f, 1.1f, 1.3f, 1.8f }
},
// It's a flattened 3-by-3 tensor.
// [0, 0, 1]
// |1, 0, 0|
// [1, 0, 0]
new InputWithCustomShape()
{
input = new float[] { 0f, 0f, 1f, 1f, 0f, 0f, 1f, 0f, 0f }
}
};
var shapeDictionary = new Dictionary <string, int[]>()
{
{ nameof(InputWithCustomShape.input), new int[] { 3, 3 } }
};
var dataView = ML.Data.LoadFromEnumerable(dataPoints);
var pipeline = new OnnxScoringEstimator[3];
var onnxTransformer = new OnnxTransformer[3];
var transformedDataViews = new IDataView[3];
// Test three public ONNX APIs with the custom shape.
// Test 1.
pipeline[0] = ML.Transforms.ApplyOnnxModel(
new[] { nameof(PredictionWithCustomShape.argmax) }, new[] { nameof(InputWithCustomShape.input) },
modelFile, shapeDictionary);
onnxTransformer[0] = pipeline[0].Fit(dataView);
transformedDataViews[0] = onnxTransformer[0].Transform(dataView);
// Test 2.
pipeline[1] = ML.Transforms.ApplyOnnxModel(
nameof(PredictionWithCustomShape.argmax), nameof(InputWithCustomShape.input),
modelFile, shapeDictionary);
onnxTransformer[1] = pipeline[1].Fit(dataView);
transformedDataViews[1] = onnxTransformer[1].Transform(dataView);
// Test 3.
pipeline[2] = ML.Transforms.ApplyOnnxModel(modelFile, shapeDictionary);
onnxTransformer[2] = pipeline[2].Fit(dataView);
transformedDataViews[2] = onnxTransformer[2].Transform(dataView);
// Conduct the same check for all the 3 called public APIs.
foreach (var transformedDataView in transformedDataViews)
{
var transformedDataPoints = ML.Data.CreateEnumerable <PredictionWithCustomShape>(transformedDataView, false).ToList();
// One data point generates one transformed data point.
Assert.Equal(dataPoints.Count(), transformedDataPoints.Count);
// Check result numbers. They are results of applying ONNX argmax along the second axis; for example
// [1.1, 1.3, 1.2] ---> [1] because 1.3 (indexed by 1) is the largest element.
// |1.9, 1.3, 1.2| ---> |0| 1.9 0
// [1.1, 1.3, 1.8] ---> [2] 1.8 2
var expectedResults = new long[][]
{
new long[] { 1, 0, 2 },
new long[] { 2, 0, 0 }
};
for (int i = 0; i < transformedDataPoints.Count; ++i)
{
Assert.Equal(transformedDataPoints[i].argmax, expectedResults[i]);
}
}
for (int i = 0; i < 3; i++)
{
(onnxTransformer[i] as IDisposable)?.Dispose();
}
}
/// <summary> /// Initializes a new instance of <see cref="OnnxScoringEstimator"/>. /// </summary> /// <param name="catalog">The transform's catalog.</param> /// <param name="transformer">The ONNX transformer.</param> public static OnnxScoringEstimator ApplyOnnxModel(this TransformsCatalog catalog, OnnxTransformer transformer) => new OnnxScoringEstimator(CatalogUtils.GetEnvironment(catalog), transformer);
