public void EvalManagedConstantNetworkTest()
{
string modelDefinition = @"precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
v1 = Constant(1)
v2 = Constant(2, tag=""output"")
ol = Plus(v1, v2, tag=""output"")
FeatureNodes = (v1)
]";
using (var model = new ModelEvaluationExtendedF())
{
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
model.StartForwardEvaluation(outputSchema.Select(s => s.Name).ToList <string>());
var outputBuffer = outputSchema.CreateBuffers <float>();
var inputBuffer = new ValueBuffer <float> [0];
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
float[][] expected = { new float[] { 2 }, new float[] { 3 } };
Assert.AreEqual(expected.Length, outputBuffer.Length);
for (int idx = 0; idx < expected.Length; idx++)
{
CollectionAssert.AreEqual(expected[idx], outputBuffer[idx].Buffer);
}
}
}
public void EvalManagedConstantNetworkTest()
{
string modelDefinition = @"precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
v1 = Constant(1)
v2 = Constant(2, tag=""output"")
ol = Plus(v1, v2, tag=""output"")
FeatureNodes = (v1)
]";
using (var model = new ModelEvaluationExtendedF())
{
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
model.StartForwardEvaluation(outputSchema.Select(s => s.Name).ToList<string>());
var outputBuffer = outputSchema.CreateBuffers<float>();
var inputBuffer = new ValueBuffer<float>[0];
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
float[][] expected = { new float[] { 2 }, new float[] {3} };
Assert.AreEqual(expected.Length, outputBuffer.Length);
for (int idx = 0; idx < expected.Length; idx++)
{
CollectionAssert.AreEqual(expected[idx], outputBuffer[idx].Buffer);
}
}
}
/// <summary>
/// Evaluates an extended network (without a model and without input) and obtains a single layer output
/// </summary>
private static void EvaluateExtendedNetworkSingleLayerNoInput()
{
const string modelDefinition = @"precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
v1 = Constant(1)
v2 = Constant(2, tag=""output"")
ol = Plus(v1, v2, tag=""output"")
FeatureNodes = (v1)
]";
try
{
// The examples assume the executable is running from the data folder
// We switch the current directory to the data folder (assuming the executable is in the <CNTK>/x64/Debug|Release folder
string workingDirectory = Path.Combine(initialDirectory, @"..\..\Examples\Other\Simple2d\Config");
Environment.CurrentDirectory = initialDirectory;
using (var model = new ModelEvaluationExtendedF())
{
// Create the network
// This network (AddOperatorConstantNoInput.cntk) is a simple network consisting of a single binary operator (Plus)
// operating over a two constants, therefore no input is necessary.
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
var outputNodeNames = outputSchema.Select(s => s.Name).ToList <string>();
model.StartForwardEvaluation(outputNodeNames);
var outputBuffer = outputSchema.CreateBuffers <float>();
var inputBuffer = new ValueBuffer <float> [0];
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
// We expect two outputs: the v2 constant, and the ol Plus result
var expected = new float[][] { new float[] { 2 }, new float[] { 3 } };
Console.WriteLine("Expected values: {0}", string.Join(" - ", expected.Select(b => string.Join(", ", b)).ToList <string>()));
Console.WriteLine("Actual Values : {0}", string.Join(" - ", outputBuffer.Select(b => string.Join(", ", b.Buffer)).ToList <string>()));
}
}
catch (CNTKException ex)
{
Console.WriteLine("Error: {0}\nNative CallStack: {1}\n Inner Exception: {2}", ex.Message, ex.NativeCallStack, ex.InnerException != null ? ex.InnerException.Message : "No Inner Exception");
}
catch (Exception ex)
{
Console.WriteLine("Error: {0}\nCallStack: {1}\n Inner Exception: {2}", ex.Message, ex.StackTrace, ex.InnerException != null ? ex.InnerException.Message : "No Inner Exception");
}
}
public void EvalManagedSparseTimesTest()
{
string modelDefinition = @"deviceId = -1
precision = ""float"" traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
i1 = SparseInput(3)
o1 = Times(Constant(2, rows=1, cols=3), i1, tag=""output"")
FeatureNodes = (i1)
]";
using (var model = new ModelEvaluationExtendedF())
{
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
model.StartForwardEvaluation(outputSchema.Select(s => s.Name).ToList <string>());
var outputBuffer = new []
{
new ValueBuffer <float>()
{
Buffer = new float[3],
Size = 3
}
};
var inputBuffer = new []
{
new ValueBuffer <float>()
{
Buffer = new float[] { 1, 2, 3, 5, 6 },
Indices = new [] { 0, 2, 2, 1, 2 },
ColIndices = new [] { 0, 2, 2, 5 },
Size = 4
}
};
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
float[][] expected = { new float[] { 6, 0, 28 } };
Assert.AreEqual(expected.Length, outputBuffer.Length);
for (int idx = 0; idx < expected.Length; idx++)
{
CollectionAssert.AreEqual(expected[idx], outputBuffer[idx].Buffer);
}
}
}
/// <summary>
/// Evaluates an extended network (without a model and without input) and obtains a single layer output
/// </summary>
private static void EvaluateExtendedNetworkSingleLayerNoInput()
{
const string modelDefinition = @"precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
v1 = Constant(1)
v2 = Constant(2, tag=""output"")
ol = Plus(v1, v2, tag=""output"")
FeatureNodes = (v1)
]";
try
{
using (var model = new ModelEvaluationExtendedF())
{
// Create the network
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
var outputNodeNames = outputSchema.Select(s => s.Name).ToList <string>();
model.StartForwardEvaluation(outputNodeNames);
var outputBuffer = outputSchema.CreateBuffers <float>();
var inputBuffer = new ValueBuffer <float> [0];
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
// We expect two outputs: the v2 constant, and the ol Plus result
var expected = new float[][] { new float[] { 2 }, new float[] { 3 } };
Console.WriteLine("Expected values: {0}", string.Join(" - ", expected.Select(b => string.Join(", ", b)).ToList <string>()));
Console.WriteLine("Actual Values : {0}", string.Join(" - ", outputBuffer.Select(b => string.Join(", ", b.Buffer)).ToList <string>()));
}
}
catch (CNTKException ex)
{
Console.WriteLine("Error: {0}\nNative CallStack: {1}\n Inner Exception: {2}", ex.Message, ex.NativeCallStack, ex.InnerException != null ? ex.InnerException.Message : "No Inner Exception");
}
catch (Exception ex)
{
Console.WriteLine("Error: {0}\nCallStack: {1}\n Inner Exception: {2}", ex.Message, ex.StackTrace, ex.InnerException != null ? ex.InnerException.Message : "No Inner Exception");
}
}
public void EvalManagedScalarTimesDualOutputTest()
{
string modelDefinition = @"deviceId = -1
precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
i1 = Input(1)
i2 = Input(1)
o1 = Times(Constant(3), i1, tag=""output"")
o2 = Times(Constant(5), i1, tag=""output"")
FeatureNodes = (i1)
]";
using (var model = new ModelEvaluationExtendedF())
{
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
VariableSchema inputSchema = model.GetInputSchema();
model.StartForwardEvaluation(outputSchema.Select(s => s.Name).ToList <string>());
var outputBuffer = outputSchema.CreateBuffers <float>();
var inputBuffer = inputSchema.CreateBuffers <float>();
inputBuffer[0].Buffer[0] = 2;
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
float[][] expected = { new float[] { 6 }, new float[] { 10 } };
Assert.AreEqual(expected.Length, outputBuffer.Length);
for (int idx = 0; idx < expected.Length; idx++)
{
CollectionAssert.AreEqual(expected[idx], outputBuffer[idx].Buffer);
}
}
}
public void EvalManagedRNNTest()
{
string modelDefinition = @"deviceId = -1
precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder = [
LSTMComponent(inputDim, outputDim, cellDim, inputx, cellDimX2, cellDimX3, cellDimX4) = [
wx = Parameter(cellDimX4, 0, init = ""uniform"", initValueScale = 1);
b = Parameter(cellDimX4, 1, init = ""fixedValue"", value = 0.0);
Wh = Parameter(cellDimX4, 0, init = ""uniform"", initValueScale = 1);
Wci = Parameter(cellDim, init = ""uniform"", initValueScale = 1);
Wcf = Parameter(cellDim, init = ""uniform"", initValueScale = 1);
Wco = Parameter(cellDim, init = ""uniform"", initValueScale = 1);
dh = PastValue(outputDim, output, timeStep = 1);
dc = PastValue(cellDim, ct, timeStep = 1);
wxx = Times(wx, inputx);
wxxpb = Plus(wxx, b);
whh = Times(wh, dh);
wxxpbpwhh = Plus(wxxpb, whh)
G1 = RowSlice(0, cellDim, wxxpbpwhh)
G2 = RowSlice(cellDim, cellDim, wxxpbpwhh)
G3 = RowSlice(cellDimX2, cellDim, wxxpbpwhh);
G4 = RowSlice(cellDimX3, cellDim, wxxpbpwhh);
Wcidc = DiagTimes(Wci, dc);
it = Sigmoid(Plus(G1, Wcidc));
bit = ElementTimes(it, Tanh(G2));
Wcfdc = DiagTimes(Wcf, dc);
ft = Sigmoid(Plus(G3, Wcfdc));
bft = ElementTimes(ft, dc);
ct = Plus(bft, bit);
Wcoct = DiagTimes(Wco, ct);
ot = Sigmoid(Plus(G4, Wcoct));
mt = ElementTimes(ot, Tanh(ct));
Wmr = Parameter(outputDim, cellDim, init = ""uniform"", initValueScale = 1);
output = Times(Wmr, mt);
]
i1 = Input(4)
o1 = LSTMComponent(4, 4, 1, i1, 2, 3, 4)
FeatureNodes = (i1)
outputNodes = (o1)
]";
using (var model = new ModelEvaluationExtendedF())
{
int featDim = 4;
int labelDim = 4;
model.CreateNetwork(modelDefinition);
VariableSchema inputSchema = model.GetInputSchema();
VariableSchema outputSchema = model.GetOutputSchema();
model.StartForwardEvaluation(outputSchema.Select(s => s.Name).ToList <string>());
// Allocate the output values layer
var outputBuffer = outputSchema.CreateBuffers <float>();
var inputBuffer = inputSchema.CreateBuffers <float>();
for (var i = 0; i < featDim; i++)
{
inputBuffer[0].Buffer[i] = (float)i;
}
int scaler = 100000;
var result = new int[labelDim];
int[] expected = { 50, 10, 54, 55 };
// the first pass with reset
model.ForwardPass(inputBuffer, outputBuffer);
for (var i = 0; i < labelDim; i++)
{
result[i] = (int)(outputBuffer[0].Buffer[i] * scaler);
}
CollectionAssert.AreEqual(expected, result);
// the second pass with reset
model.ForwardPass(inputBuffer, outputBuffer);
for (var i = 0; i < labelDim; i++)
{
result[i] = (int)(outputBuffer[0].Buffer[i] * scaler);
}
CollectionAssert.AreEqual(expected, result);
// another pass with reset
model.ForwardPass(inputBuffer, outputBuffer, true);
for (var i = 0; i < labelDim; i++)
{
result[i] = (int)(outputBuffer[0].Buffer[i] * scaler);
}
CollectionAssert.AreEqual(expected, result);
// pass w/o reset
model.ForwardPass(inputBuffer, outputBuffer, false);
for (var i = 0; i < labelDim; i++)
{
result[i] = (int)(outputBuffer[0].Buffer[i] * scaler);
}
expected = new int[] { 13, 2, 14, 14 };
CollectionAssert.AreEqual(expected, result);
// another pass w/o reset
model.ForwardPass(inputBuffer, outputBuffer, false);
for (var i = 0; i < labelDim; i++)
{
result[i] = (int)(outputBuffer[0].Buffer[i] * scaler);
}
expected = new int[] { -4, 0, -4, -4 };
CollectionAssert.AreEqual(expected, result);
}
}
/// <summary>
/// Evaluates an extended network (without a model and without input) and obtains a single layer output
/// </summary>
private static void EvaluateExtendedNetworkSingleLayerNoInput()
{
const string modelDefinition = @"precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
v1 = Constant(1)
v2 = Constant(2, tag=""output"")
ol = Plus(v1, v2, tag=""output"")
FeatureNodes = (v1)
]";
try
{
using (var model = new ModelEvaluationExtendedF())
{
// Create the network
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
var outputNodeNames = outputSchema.Select(s => s.Name).ToList<string>();
model.StartForwardEvaluation(outputNodeNames);
var outputBuffer = outputSchema.CreateBuffers<float>();
var inputBuffer = new ValueBuffer<float>[0];
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
// We expect two outputs: the v2 constant, and the ol Plus result
var expected = new float[][] { new float[] { 2 }, new float[] { 3 } };
Console.WriteLine("Expected values: {0}", string.Join(" - ", expected.Select(b => string.Join(", ", b)).ToList<string>()));
Console.WriteLine("Actual Values : {0}", string.Join(" - ", outputBuffer.Select(b => string.Join(", ", b.Buffer)).ToList<string>()));
}
}
catch (CNTKException ex)
{
OnCNTKException(ex);
}
catch (Exception ex)
{
OnGeneralException(ex);
}
}
/// <summary>
/// Evaluates an extended network (without a model and without input) and obtains a single layer output
/// </summary>
private static void EvaluateExtendedNetworkSingleLayerNoInput()
{
const string modelDefinition = @"precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
v1 = Constant(1)
v2 = Constant(2, tag=""output"")
ol = Plus(v1, v2, tag=""output"")
FeatureNodes = (v1)
]";
try
{
// The examples assume the executable is running from the data folder
// We switch the current directory to the data folder (assuming the executable is in the <CNTK>/x64/Debug|Release folder
string workingDirectory = Path.Combine(initialDirectory, @"..\..\Examples\Other\Simple2d\Config");
Environment.CurrentDirectory = initialDirectory;
using (var model = new ModelEvaluationExtendedF())
{
// Create the network
// This network (AddOperatorConstantNoInput.cntk) is a simple network consisting of a single binary operator (Plus)
// operating over a two constants, therefore no input is necessary.
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
var outputNodeNames = outputSchema.Select(s => s.Name).ToList<string>();
model.StartForwardEvaluation(outputNodeNames);
var outputBuffer = outputSchema.CreateBuffers<float>();
var inputBuffer = new ValueBuffer<float>[0];
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
// We expect two outputs: the v2 constant, and the ol Plus result
var expected = new float[][] { new float[] { 2 }, new float[] { 3 } };
Console.WriteLine("Expected values: {0}", string.Join(" - ", expected.Select(b => string.Join(", ", b)).ToList<string>()));
Console.WriteLine("Actual Values : {0}", string.Join(" - ", outputBuffer.Select(b => string.Join(", ", b.Buffer)).ToList<string>()));
}
}
catch (CNTKException ex)
{
Console.WriteLine("Error: {0}\nNative CallStack: {1}\n Inner Exception: {2}", ex.Message, ex.NativeCallStack, ex.InnerException != null ? ex.InnerException.Message : "No Inner Exception");
}
catch (Exception ex)
{
Console.WriteLine("Error: {0}\nCallStack: {1}\n Inner Exception: {2}", ex.Message, ex.StackTrace, ex.InnerException != null ? ex.InnerException.Message : "No Inner Exception");
}
}
public void EvalManagedScalarTimesDualOutputTest()
{
string modelDefinition = @"deviceId = -1
precision = ""float""
traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
i1 = Input(1)
i2 = Input(1)
o1 = Times(Constant(3), i1, tag=""output"")
o2 = Times(Constant(5), i1, tag=""output"")
FeatureNodes = (i1)
]";
using (var model = new ModelEvaluationExtendedF())
{
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
VariableSchema inputSchema = model.GetInputSchema();
model.StartForwardEvaluation(outputSchema.Select(s => s.Name).ToList<string>());
var outputBuffer = outputSchema.CreateBuffers<float>();
var inputBuffer = inputSchema.CreateBuffers<float>();
inputBuffer[0].Buffer[0] = 2;
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
float[][] expected = {new float[]{6}, new float[]{10} };
Assert.AreEqual(expected.Length, outputBuffer.Length);
for(int idx=0; idx<expected.Length; idx++ )
{
CollectionAssert.AreEqual(expected[idx], outputBuffer[idx].Buffer);
}
}
}
public void EvalManagedSparseTimesTest()
{
string modelDefinition = @"deviceId = -1
precision = ""float"" traceLevel = 1
run=NDLNetworkBuilder
NDLNetworkBuilder=[
i1 = SparseInput(3)
o1 = Times(Constant(2, rows=1, cols=3), i1, tag=""output"")
FeatureNodes = (i1)
]";
using (var model = new ModelEvaluationExtendedF())
{
model.CreateNetwork(modelDefinition);
VariableSchema outputSchema = model.GetOutputSchema();
model.StartForwardEvaluation(outputSchema.Select(s => s.Name).ToList<string>());
var outputBuffer = new []
{
new ValueBuffer<float>()
{
Buffer = new float[3],
Size = 3
}
};
var inputBuffer = new []
{
new ValueBuffer<float>()
{
Buffer = new float[] { 1, 2, 3, 5, 6 },
Indices = new [] { 0, 2, 2, 1, 2 },
ColIndices = new [] { 0, 2, 2, 5 },
Size = 4
}
};
// We can call the evaluate method and get back the results...
model.ForwardPass(inputBuffer, outputBuffer);
float[][] expected = { new float[] { 6, 0, 28 } };
Assert.AreEqual(expected.Length, outputBuffer.Length);
for (int idx = 0; idx < expected.Length; idx++)
{
CollectionAssert.AreEqual(expected[idx], outputBuffer[idx].Buffer);
}
}
}