CPU info:
    CPU Model Name: Intel(R) Xeon(R) CPU E5-2690 v3 @ 2.60GHz
    Hardware threads: 6
    Total Memory: 58719796 kB
-------------------------------------------------------------------
=== Running /cygdrive/c/jenkins/workspace/CNTK-Test-Windows-W1/x64/debug/cntk.exe configFile=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/cntk_sequence.cntk currentDirectory=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData RunDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu DataDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader OutputDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu DeviceId=0 timestamping=true
CNTK 2.3.1+ (HEAD b7b3e4, Jan 17 2018 02:48:57) at 2018/01/17 07:42:48

C:\jenkins\workspace\CNTK-Test-Windows-W1\x64\debug\cntk.exe  configFile=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/cntk_sequence.cntk  currentDirectory=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData  RunDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu  DataDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData  ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader  OutputDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu  DeviceId=0  timestamping=true
Changed current directory to C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData
-------------------------------------------------------------------
Build info: 

		Built time: Jan 17 2018 02:44:09
		Last modified date: Wed Jan 17 02:36:31 2018
		Build type: Debug
		Build target: GPU
		With ASGD: yes
		Math lib: mkl
		CUDA version: 9.0.0
		CUDNN version: 7.0.5
		Build Branch: HEAD
		Build SHA1: b7b3e4fb3ff0f69024ce19a19b8f2780fb63078b
		MPI distribution: Microsoft MPI
		MPI version: 7.0.12437.6
-------------------------------------------------------------------
-------------------------------------------------------------------
GPU info:

		Device[0]: cores = 3072; computeCapability = 5.2; type = "Tesla M60"; total memory = 8124 MB; free memory = 8002 MB
-------------------------------------------------------------------

Configuration, Raw:

01/17/2018 07:42:48: precision = "float"
deviceId = $DeviceId$
command = dptPre1:addLayer2:dptPre2:addLayer3:speechTrain:sequenceTrain
ndlMacros = "$ConfigDir$/macros.txt"
globalMeanPath   = "GlobalStats/mean.363"
globalInvStdPath = "GlobalStats/var.363"
globalPriorPath  = "GlobalStats/prior.132"
traceLevel = 1
truncated = false
SGD = [
    epochSize = 81920
    minibatchSize = 256
    learningRatesPerMB = 0.8
    numMBsToShowResult = 10
    momentumPerMB = 0.9
    dropoutRate = 0.0
    maxEpochs = 2
]
dptPre1 = [
    action = "train"
    modelPath = "$RunDir$/models/Pre1/cntkSpeech"
    NDLNetworkBuilder = [
        networkDescription = "$ConfigDir$/dnn_1layer.txt"
    ]
]
addLayer2 = [    
    action = "edit"
    currLayer = 1
    newLayer = 2
    currModel = "$RunDir$/models/Pre1/cntkSpeech"
    newModel  = "$RunDir$/models/Pre2/cntkSpeech.0"
    editPath  = "$ConfigDir$/add_layer.mel"
]
dptPre2 = [
    action = "train"
    modelPath = "$RunDir$/models/Pre2/cntkSpeech"
    NDLNetworkBuilder = [
        networkDescription = "$ConfigDir$/dnn_1layer.txt"
    ]
]
AddLayer3 = [    
    action = "edit"
    currLayer = 2
    newLayer = 3
    currModel = "$RunDir$/models/Pre2/cntkSpeech"
    newModel  = "$RunDir$/models/cntkSpeech.0"
    editPath  = "$ConfigDir$/add_layer.mel"
]
speechTrain = [
    action = "train"
    modelPath = "$RunDir$/models/cntkSpeech"
    traceLevel = 1
    NDLNetworkBuilder = [
        networkDescription = "$ConfigDir$/dnn.txt"
    ]
    SGD = [
        epochSize = 81920
        minibatchSize = 256:512
        learningRatesPerMB = 0.8:1.6
        numMBsToShowResult = 10
        momentumPerSample = 0.999589
        dropoutRate = 0.0
        maxEpochs = 4
        gradUpdateType = "none"
        normWithAveMultiplier = true
        clippingThresholdPerSample = 1#INF
    ]
]
reader = [
    readerType = "HTKMLFReader"
    readMethod = "blockRandomize"
    miniBatchMode = "partial"
    randomize = "auto"
    verbosity = 0
    features = [
        dim = 363
        type = "real"
        scpFile = "$DataDir$/glob_0000.scp"
    ]
    labels = [
        mlfFile = "$DataDir$/glob_0000.mlf"
        labelMappingFile = "$DataDir$/state.list"
        labelDim = 132
        labelType = "category"
    ]
]
sequenceTrain = [
    action = "train"
    modelPath = $RunDir$/models/cntkSpeech.sequence
    traceLevel = 1
    SGD = [
        epochSize = 81920
        minibatchSize = 10
        learningRatesPerSample = 0.000002
        momentumPerSample = 0.999589
        dropoutRate = 0.0
        maxEpochs = 3
        numMBsToShowResult = 10
	gradientClippingWithTruncation = true
	clippingThresholdPerSample = 1.0
    ]
	reader = [
			verbosity = 0
			randomize = true
                        maxErrors = 100
			deserializers = (
				[
					type = "HTKFeatureDeserializer"
					module = "HTKDeserializers"
                                        definesMbSize = true
					input = [
						features = [
							dim=363
							scpFile = "$DataDir$/glob_0000.scp"
						]
					]
				]:
				[
					type = "HTKMLFDeserializer"
					module = "HTKDeserializers"
					input = [
						labels = [
							dim = 132
							mlfFile="$DataDir$/glob_0000.mlf"
							labelMappingFile = "$DataDir$/state.list" 
						]
					]
				]:
				[
					type = "LatticeDeserializer"
					module = "HTKDeserializers"
					input = [
						lattice=[
							latticeIndexFile="$DataDir$/latticeIndex.txt"
						]
					]
				]
			)
		]
		BrainScriptNetworkBuilder = {
			baseFeatDim = 33
			featDim = 11 * baseFeatDim
			labelDim = 132
			latticeAxis = DynamicAxis()
			features = Input{featDim}
			labels = Input{labelDim, tag="label"}
			lattice = Input{1,dynamicAxis=latticeAxis, tag="label"}
			featExtNetwork  = BS.Network.Load("$RunDir$/models/cntkSpeech")
			featExt = BS.Network.CloneFunction (
              (featExtNetwork.features),
              [netEval = featExtNetwork.OL_z;scaledLogLikelihood = featExtNetwork.scaledLogLikelihood ],
              parameters="learnable")
			clonedmodel= featExt(features)
			cr = LatticeSequenceWithSoftmax(labels, clonedmodel.netEval, clonedmodel.scaledLogLikelihood, lattice, "$DataDir$/CY2SCH010061231_1369712653.numden.lats.symlist", "$DataDir$/model.overalltying", "$DataDir$/state.list", "$DataDir$/model.transprob", tag="criterion")  
			Err = ClassificationError(labels,clonedmodel.netEval,tag="evaluation");
		}
]
currentDirectory=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData
RunDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu
DataDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData
ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader
OutputDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu
DeviceId=0
timestamping=true


Configuration After Variable Resolution:

01/17/2018 07:42:48: precision = "float"
deviceId = 0
command = dptPre1:addLayer2:dptPre2:addLayer3:speechTrain:sequenceTrain
ndlMacros = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/macros.txt"
globalMeanPath   = "GlobalStats/mean.363"
globalInvStdPath = "GlobalStats/var.363"
globalPriorPath  = "GlobalStats/prior.132"
traceLevel = 1
truncated = false
SGD = [
    epochSize = 81920
    minibatchSize = 256
    learningRatesPerMB = 0.8
    numMBsToShowResult = 10
    momentumPerMB = 0.9
    dropoutRate = 0.0
    maxEpochs = 2
]
dptPre1 = [
    action = "train"
    modelPath = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre1/cntkSpeech"
    NDLNetworkBuilder = [
        networkDescription = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/dnn_1layer.txt"
    ]
]
addLayer2 = [    
    action = "edit"
    currLayer = 1
    newLayer = 2
    currModel = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre1/cntkSpeech"
    newModel  = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech.0"
    editPath  = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/add_layer.mel"
]
dptPre2 = [
    action = "train"
    modelPath = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech"
    NDLNetworkBuilder = [
        networkDescription = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/dnn_1layer.txt"
    ]
]
AddLayer3 = [    
    action = "edit"
    currLayer = 2
    newLayer = 3
    currModel = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech"
    newModel  = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.0"
    editPath  = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/add_layer.mel"
]
speechTrain = [
    action = "train"
    modelPath = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech"
    traceLevel = 1
    NDLNetworkBuilder = [
        networkDescription = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/dnn.txt"
    ]
    SGD = [
        epochSize = 81920
        minibatchSize = 256:512
        learningRatesPerMB = 0.8:1.6
        numMBsToShowResult = 10
        momentumPerSample = 0.999589
        dropoutRate = 0.0
        maxEpochs = 4
        gradUpdateType = "none"
        normWithAveMultiplier = true
        clippingThresholdPerSample = 1#INF
    ]
]
reader = [
    readerType = "HTKMLFReader"
    readMethod = "blockRandomize"
    miniBatchMode = "partial"
    randomize = "auto"
    verbosity = 0
    features = [
        dim = 363
        type = "real"
        scpFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp"
    ]
    labels = [
        mlfFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.mlf"
        labelMappingFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list"
        labelDim = 132
        labelType = "category"
    ]
]
sequenceTrain = [
    action = "train"
    modelPath = C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.sequence
    traceLevel = 1
    SGD = [
        epochSize = 81920
        minibatchSize = 10
        learningRatesPerSample = 0.000002
        momentumPerSample = 0.999589
        dropoutRate = 0.0
        maxEpochs = 3
        numMBsToShowResult = 10
	gradientClippingWithTruncation = true
	clippingThresholdPerSample = 1.0
    ]
	reader = [
			verbosity = 0
			randomize = true
                        maxErrors = 100
			deserializers = (
				[
					type = "HTKFeatureDeserializer"
					module = "HTKDeserializers"
                                        definesMbSize = true
					input = [
						features = [
							dim=363
							scpFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp"
						]
					]
				]:
				[
					type = "HTKMLFDeserializer"
					module = "HTKDeserializers"
					input = [
						labels = [
							dim = 132
							mlfFile="C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.mlf"
							labelMappingFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list" 
						]
					]
				]:
				[
					type = "LatticeDeserializer"
					module = "HTKDeserializers"
					input = [
						lattice=[
							latticeIndexFile="C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/latticeIndex.txt"
						]
					]
				]
			)
		]
		BrainScriptNetworkBuilder = {
			baseFeatDim = 33
			featDim = 11 * baseFeatDim
			labelDim = 132
			latticeAxis = DynamicAxis()
			features = Input{featDim}
			labels = Input{labelDim, tag="label"}
			lattice = Input{1,dynamicAxis=latticeAxis, tag="label"}
			featExtNetwork  = BS.Network.Load("C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech")
			featExt = BS.Network.CloneFunction (
              (featExtNetwork.features),
              [netEval = featExtNetwork.OL_z;scaledLogLikelihood = featExtNetwork.scaledLogLikelihood ],
              parameters="learnable")
			clonedmodel= featExt(features)
			cr = LatticeSequenceWithSoftmax(labels, clonedmodel.netEval, clonedmodel.scaledLogLikelihood, lattice, "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/CY2SCH010061231_1369712653.numden.lats.symlist", "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.overalltying", "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list", "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.transprob", tag="criterion")  
			Err = ClassificationError(labels,clonedmodel.netEval,tag="evaluation");
		}
]
currentDirectory=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData
RunDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu
DataDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData
ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader
OutputDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu
DeviceId=0
timestamping=true


Configuration After Processing and Variable Resolution:

configparameters: cntk_sequence.cntk:addLayer2=[    
    action = "edit"
    currLayer = 1
    newLayer = 2
    currModel = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre1/cntkSpeech"
    newModel  = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech.0"
    editPath  = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/add_layer.mel"
]

configparameters: cntk_sequence.cntk:AddLayer3=[    
    action = "edit"
    currLayer = 2
    newLayer = 3
    currModel = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech"
    newModel  = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.0"
    editPath  = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/add_layer.mel"
]

configparameters: cntk_sequence.cntk:command=dptPre1:addLayer2:dptPre2:addLayer3:speechTrain:sequenceTrain
configparameters: cntk_sequence.cntk:ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader
configparameters: cntk_sequence.cntk:currentDirectory=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData
configparameters: cntk_sequence.cntk:DataDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData
configparameters: cntk_sequence.cntk:deviceId=0
configparameters: cntk_sequence.cntk:dptPre1=[
    action = "train"
    modelPath = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre1/cntkSpeech"
    NDLNetworkBuilder = [
        networkDescription = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/dnn_1layer.txt"
    ]
]

configparameters: cntk_sequence.cntk:dptPre2=[
    action = "train"
    modelPath = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech"
    NDLNetworkBuilder = [
        networkDescription = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/dnn_1layer.txt"
    ]
]

configparameters: cntk_sequence.cntk:globalInvStdPath=GlobalStats/var.363
configparameters: cntk_sequence.cntk:globalMeanPath=GlobalStats/mean.363
configparameters: cntk_sequence.cntk:globalPriorPath=GlobalStats/prior.132
configparameters: cntk_sequence.cntk:ndlMacros=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/macros.txt
configparameters: cntk_sequence.cntk:OutputDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu
configparameters: cntk_sequence.cntk:precision=float
configparameters: cntk_sequence.cntk:reader=[
    readerType = "HTKMLFReader"
    readMethod = "blockRandomize"
    miniBatchMode = "partial"
    randomize = "auto"
    verbosity = 0
    features = [
        dim = 363
        type = "real"
        scpFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp"
    ]
    labels = [
        mlfFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.mlf"
        labelMappingFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list"
        labelDim = 132
        labelType = "category"
    ]
]

configparameters: cntk_sequence.cntk:RunDir=C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu
configparameters: cntk_sequence.cntk:sequenceTrain=[
    action = "train"
    modelPath = C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.sequence
    traceLevel = 1
    SGD = [
        epochSize = 81920
        minibatchSize = 10
        learningRatesPerSample = 0.000002
        momentumPerSample = 0.999589
        dropoutRate = 0.0
        maxEpochs = 3
        numMBsToShowResult = 10
	gradientClippingWithTruncation = true
	clippingThresholdPerSample = 1.0
    ]
	reader = [
			verbosity = 0
			randomize = true
                        maxErrors = 100
			deserializers = (
				[
					type = "HTKFeatureDeserializer"
					module = "HTKDeserializers"
                                        definesMbSize = true
					input = [
						features = [
							dim=363
							scpFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp"
						]
					]
				]:
				[
					type = "HTKMLFDeserializer"
					module = "HTKDeserializers"
					input = [
						labels = [
							dim = 132
							mlfFile="C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.mlf"
							labelMappingFile = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list" 
						]
					]
				]:
				[
					type = "LatticeDeserializer"
					module = "HTKDeserializers"
					input = [
						lattice=[
							latticeIndexFile="C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/latticeIndex.txt"
						]
					]
				]
			)
		]
		BrainScriptNetworkBuilder = {
			baseFeatDim = 33
			featDim = 11 * baseFeatDim
			labelDim = 132
			latticeAxis = DynamicAxis()
			features = Input{featDim}
			labels = Input{labelDim, tag="label"}
			lattice = Input{1,dynamicAxis=latticeAxis, tag="label"}
			featExtNetwork  = BS.Network.Load("C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech")
			featExt = BS.Network.CloneFunction (
              (featExtNetwork.features),
              [netEval = featExtNetwork.OL_z;scaledLogLikelihood = featExtNetwork.scaledLogLikelihood ],
              parameters="learnable")
			clonedmodel= featExt(features)
			cr = LatticeSequenceWithSoftmax(labels, clonedmodel.netEval, clonedmodel.scaledLogLikelihood, lattice, "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/CY2SCH010061231_1369712653.numden.lats.symlist", "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.overalltying", "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list", "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.transprob", tag="criterion")  
			Err = ClassificationError(labels,clonedmodel.netEval,tag="evaluation");
		}
]

configparameters: cntk_sequence.cntk:SGD=[
    epochSize = 81920
    minibatchSize = 256
    learningRatesPerMB = 0.8
    numMBsToShowResult = 10
    momentumPerMB = 0.9
    dropoutRate = 0.0
    maxEpochs = 2
]

configparameters: cntk_sequence.cntk:speechTrain=[
    action = "train"
    modelPath = "C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech"
    traceLevel = 1
    NDLNetworkBuilder = [
        networkDescription = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN\SequenceTrainingNewReader/dnn.txt"
    ]
    SGD = [
        epochSize = 81920
        minibatchSize = 256:512
        learningRatesPerMB = 0.8:1.6
        numMBsToShowResult = 10
        momentumPerSample = 0.999589
        dropoutRate = 0.0
        maxEpochs = 4
        gradUpdateType = "none"
        normWithAveMultiplier = true
        clippingThresholdPerSample = 1#INF
    ]
]

configparameters: cntk_sequence.cntk:timestamping=true
configparameters: cntk_sequence.cntk:traceLevel=1
configparameters: cntk_sequence.cntk:truncated=false
01/17/2018 07:42:48: Commands: dptPre1 addLayer2 dptPre2 addLayer3 speechTrain sequenceTrain
01/17/2018 07:42:48: precision = "float"

01/17/2018 07:42:48: ##############################################################################
01/17/2018 07:42:48: #                                                                            #
01/17/2018 07:42:48: # dptPre1 command (train action)                                             #
01/17/2018 07:42:48: #                                                                            #
01/17/2018 07:42:48: ##############################################################################

01/17/2018 07:42:48: 
Creating virgin network.
NDLBuilder Using GPU 0
Microsoft::MSR::CNTK::GPUMatrix<ElemType>::SetUniformRandomValue (GPU): creating curand object with seed 1, sizeof(ElemType)==4
reading script file C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp ... 948 entries
total 132 state names in state list C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list
htkmlfreader: reading MLF file C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.mlf ... total 948 entries
...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
label set 0: 129 classes
minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
01/17/2018 07:42:53: 
Model has 19 nodes. Using GPU 0.

01/17/2018 07:42:53: Training criterion:   ce = CrossEntropyWithSoftmax
01/17/2018 07:42:53: Evaluation criterion: err = ClassificationError


Allocating matrices for forward and/or backward propagation.

Gradient Memory Aliasing: 2 are aliased.
	OL.t (gradient) reuses OL.z (gradient)

Memory Sharing: Out of 29 matrices, 11 are shared as 3, and 18 are not shared.

Here are the ones that share memory:
	{ HL1.z : [512 x 1 x *] (gradient)
	  OL.t : [132 x 1 x *]
	  OL.t : [132 x 1 x *] (gradient)
	  OL.z : [132 x 1 x *] (gradient) }
	{ HL1.W : [512 x 363] (gradient)
	  HL1.t : [512 x *]
	  HL1.y : [512 x 1 x *] }
	{ HL1.t : [512 x *] (gradient)
	  HL1.y : [512 x 1 x *] (gradient)
	  HL1.z : [512 x 1 x *]
	  OL.z : [132 x 1 x *] }

Here are the ones that don't share memory:
	{scaledLogLikelihood : [132 x 1 x *]}
	{features : [363 x *]}
	{OL.W : [132 x 512]}
	{OL.b : [132 x 1]}
	{globalMean : [363 x 1]}
	{logPrior : [132 x 1]}
	{labels : [132 x *]}
	{globalInvStd : [363 x 1]}
	{globalPrior : [132 x 1]}
	{HL1.b : [512 x 1]}
	{err : [1]}
	{ce : [1]}
	{HL1.W : [512 x 363]}
	{OL.b : [132 x 1] (gradient)}
	{featNorm : [363 x *]}
	{OL.W : [132 x 512] (gradient)}
	{ce : [1] (gradient)}
	{HL1.b : [512 x 1] (gradient)}


01/17/2018 07:42:53: Training 254084 parameters in 4 out of 4 parameter tensors and 10 nodes with gradient:

01/17/2018 07:42:53: 	Node 'HL1.W' (LearnableParameter operation) : [512 x 363]
01/17/2018 07:42:53: 	Node 'HL1.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:42:53: 	Node 'OL.W' (LearnableParameter operation) : [132 x 512]
01/17/2018 07:42:53: 	Node 'OL.b' (LearnableParameter operation) : [132 x 1]

01/17/2018 07:42:53: No PreCompute nodes found, or all already computed. Skipping pre-computation step.

01/17/2018 07:42:53: Starting Epoch 1: learning rate per sample = 0.003125  effective momentum = 0.900000  momentum as time constant = 2429.8 samples
minibatchiterator: epoch 0: frames [0..81920] (first utterance at frame 0), data subset 0 of 1, with 1 datapasses
requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms

01/17/2018 07:43:09: Starting minibatch loop.
01/17/2018 07:43:09:  Epoch[ 1 of 2]-Minibatch[   1-  10, 3.13%]: ce = 3.89978256 * 2560; err = 0.84375000 * 2560; time = 0.8264s; samplesPerSecond = 3097.8
01/17/2018 07:43:10:  Epoch[ 1 of 2]-Minibatch[  11-  20, 6.25%]: ce = 2.96755676 * 2560; err = 0.72031250 * 2560; time = 0.6128s; samplesPerSecond = 4177.4
01/17/2018 07:43:11:  Epoch[ 1 of 2]-Minibatch[  21-  30, 9.38%]: ce = 2.55723495 * 2560; err = 0.65859375 * 2560; time = 0.6096s; samplesPerSecond = 4199.4
01/17/2018 07:43:11:  Epoch[ 1 of 2]-Minibatch[  31-  40, 12.50%]: ce = 2.29642715 * 2560; err = 0.61992187 * 2560; time = 0.6088s; samplesPerSecond = 4205.1
01/17/2018 07:43:12:  Epoch[ 1 of 2]-Minibatch[  41-  50, 15.63%]: ce = 2.02396469 * 2560; err = 0.55117187 * 2560; time = 0.6090s; samplesPerSecond = 4203.7
01/17/2018 07:43:12:  Epoch[ 1 of 2]-Minibatch[  51-  60, 18.75%]: ce = 1.87309265 * 2560; err = 0.51484375 * 2560; time = 0.6089s; samplesPerSecond = 4204.1
01/17/2018 07:43:13:  Epoch[ 1 of 2]-Minibatch[  61-  70, 21.88%]: ce = 1.78157349 * 2560; err = 0.50507813 * 2560; time = 0.6104s; samplesPerSecond = 4194.1
01/17/2018 07:43:14:  Epoch[ 1 of 2]-Minibatch[  71-  80, 25.00%]: ce = 1.75391235 * 2560; err = 0.50781250 * 2560; time = 0.6152s; samplesPerSecond = 4161.4
01/17/2018 07:43:14:  Epoch[ 1 of 2]-Minibatch[  81-  90, 28.13%]: ce = 1.66460266 * 2560; err = 0.45742187 * 2560; time = 0.6236s; samplesPerSecond = 4105.0
01/17/2018 07:43:15:  Epoch[ 1 of 2]-Minibatch[  91- 100, 31.25%]: ce = 1.62184143 * 2560; err = 0.47968750 * 2560; time = 0.6121s; samplesPerSecond = 4182.2
01/17/2018 07:43:15:  Epoch[ 1 of 2]-Minibatch[ 101- 110, 34.38%]: ce = 1.65327911 * 2560; err = 0.47265625 * 2560; time = 0.6291s; samplesPerSecond = 4069.5
01/17/2018 07:43:16:  Epoch[ 1 of 2]-Minibatch[ 111- 120, 37.50%]: ce = 1.50686646 * 2560; err = 0.44921875 * 2560; time = 0.6151s; samplesPerSecond = 4161.8
01/17/2018 07:43:17:  Epoch[ 1 of 2]-Minibatch[ 121- 130, 40.63%]: ce = 1.46723328 * 2560; err = 0.42304687 * 2560; time = 0.6079s; samplesPerSecond = 4211.4
01/17/2018 07:43:17:  Epoch[ 1 of 2]-Minibatch[ 131- 140, 43.75%]: ce = 1.49163208 * 2560; err = 0.44140625 * 2560; time = 0.6107s; samplesPerSecond = 4192.2
01/17/2018 07:43:18:  Epoch[ 1 of 2]-Minibatch[ 141- 150, 46.88%]: ce = 1.46437378 * 2560; err = 0.43398437 * 2560; time = 0.6091s; samplesPerSecond = 4202.7
01/17/2018 07:43:19:  Epoch[ 1 of 2]-Minibatch[ 151- 160, 50.00%]: ce = 1.43047485 * 2560; err = 0.43867187 * 2560; time = 0.6096s; samplesPerSecond = 4199.3
01/17/2018 07:43:19:  Epoch[ 1 of 2]-Minibatch[ 161- 170, 53.13%]: ce = 1.42106323 * 2560; err = 0.41953125 * 2560; time = 0.6085s; samplesPerSecond = 4207.4
01/17/2018 07:43:20:  Epoch[ 1 of 2]-Minibatch[ 171- 180, 56.25%]: ce = 1.46542969 * 2560; err = 0.42421875 * 2560; time = 0.6314s; samplesPerSecond = 4054.3
01/17/2018 07:43:20:  Epoch[ 1 of 2]-Minibatch[ 181- 190, 59.38%]: ce = 1.47425842 * 2560; err = 0.44062500 * 2560; time = 0.6079s; samplesPerSecond = 4211.5
01/17/2018 07:43:21:  Epoch[ 1 of 2]-Minibatch[ 191- 200, 62.50%]: ce = 1.42849121 * 2560; err = 0.44062500 * 2560; time = 0.6072s; samplesPerSecond = 4215.9
01/17/2018 07:43:22:  Epoch[ 1 of 2]-Minibatch[ 201- 210, 65.63%]: ce = 1.34078064 * 2560; err = 0.41210938 * 2560; time = 0.6082s; samplesPerSecond = 4209.2
01/17/2018 07:43:22:  Epoch[ 1 of 2]-Minibatch[ 211- 220, 68.75%]: ce = 1.39478760 * 2560; err = 0.42734375 * 2560; time = 0.6071s; samplesPerSecond = 4217.1
01/17/2018 07:43:23:  Epoch[ 1 of 2]-Minibatch[ 221- 230, 71.88%]: ce = 1.40148315 * 2560; err = 0.41210938 * 2560; time = 0.6083s; samplesPerSecond = 4208.4
01/17/2018 07:43:23:  Epoch[ 1 of 2]-Minibatch[ 231- 240, 75.00%]: ce = 1.39321899 * 2560; err = 0.42617187 * 2560; time = 0.6111s; samplesPerSecond = 4189.2
01/17/2018 07:43:24:  Epoch[ 1 of 2]-Minibatch[ 241- 250, 78.13%]: ce = 1.32427368 * 2560; err = 0.40195313 * 2560; time = 0.6086s; samplesPerSecond = 4206.1
01/17/2018 07:43:25:  Epoch[ 1 of 2]-Minibatch[ 251- 260, 81.25%]: ce = 1.27012024 * 2560; err = 0.39921875 * 2560; time = 0.6169s; samplesPerSecond = 4149.5
01/17/2018 07:43:25:  Epoch[ 1 of 2]-Minibatch[ 261- 270, 84.38%]: ce = 1.32388916 * 2560; err = 0.39218750 * 2560; time = 0.6108s; samplesPerSecond = 4190.9
01/17/2018 07:43:26:  Epoch[ 1 of 2]-Minibatch[ 271- 280, 87.50%]: ce = 1.25467834 * 2560; err = 0.38359375 * 2560; time = 0.6082s; samplesPerSecond = 4209.1
01/17/2018 07:43:26:  Epoch[ 1 of 2]-Minibatch[ 281- 290, 90.63%]: ce = 1.23462830 * 2560; err = 0.37226562 * 2560; time = 0.6062s; samplesPerSecond = 4223.4
01/17/2018 07:43:27:  Epoch[ 1 of 2]-Minibatch[ 291- 300, 93.75%]: ce = 1.20836792 * 2560; err = 0.35937500 * 2560; time = 0.6058s; samplesPerSecond = 4225.8
01/17/2018 07:43:28:  Epoch[ 1 of 2]-Minibatch[ 301- 310, 96.88%]: ce = 1.23704834 * 2560; err = 0.36796875 * 2560; time = 0.6133s; samplesPerSecond = 4174.4
01/17/2018 07:43:28:  Epoch[ 1 of 2]-Minibatch[ 311- 320, 100.00%]: ce = 1.23081055 * 2560; err = 0.37539062 * 2560; time = 0.5598s; samplesPerSecond = 4573.3
01/17/2018 07:43:28: Finished Epoch[ 1 of 2]: [Training] ce = 1.65178680 * 81920; err = 0.46788330 * 81920; totalSamplesSeen = 81920; learningRatePerSample = 0.003125; epochTime=35.6026s
01/17/2018 07:43:28: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre1/cntkSpeech.1'

01/17/2018 07:43:28: Starting Epoch 2: learning rate per sample = 0.003125  effective momentum = 0.900000  momentum as time constant = 2429.8 samples
minibatchiterator: epoch 1: frames [81920..163840] (first utterance at frame 81920), data subset 0 of 1, with 1 datapasses

01/17/2018 07:43:28: Starting minibatch loop.
01/17/2018 07:43:29:  Epoch[ 2 of 2]-Minibatch[   1-  10, 3.13%]: ce = 1.21959934 * 2560; err = 0.37109375 * 2560; time = 0.6302s; samplesPerSecond = 4062.2
01/17/2018 07:43:30:  Epoch[ 2 of 2]-Minibatch[  11-  20, 6.25%]: ce = 1.18487740 * 2560; err = 0.36601563 * 2560; time = 0.6074s; samplesPerSecond = 4214.9
01/17/2018 07:43:30:  Epoch[ 2 of 2]-Minibatch[  21-  30, 9.38%]: ce = 1.17382298 * 2560; err = 0.35937500 * 2560; time = 0.6063s; samplesPerSecond = 4222.3
01/17/2018 07:43:31:  Epoch[ 2 of 2]-Minibatch[  31-  40, 12.50%]: ce = 1.20050812 * 2560; err = 0.35820313 * 2560; time = 0.6059s; samplesPerSecond = 4225.0
01/17/2018 07:43:32:  Epoch[ 2 of 2]-Minibatch[  41-  50, 15.63%]: ce = 1.19367180 * 2560; err = 0.37929687 * 2560; time = 0.6079s; samplesPerSecond = 4211.2
01/17/2018 07:43:32:  Epoch[ 2 of 2]-Minibatch[  51-  60, 18.75%]: ce = 1.16213341 * 2560; err = 0.34335938 * 2560; time = 0.6081s; samplesPerSecond = 4209.9
01/17/2018 07:43:33:  Epoch[ 2 of 2]-Minibatch[  61-  70, 21.88%]: ce = 1.13570023 * 2560; err = 0.35078125 * 2560; time = 0.6075s; samplesPerSecond = 4213.8
01/17/2018 07:43:33:  Epoch[ 2 of 2]-Minibatch[  71-  80, 25.00%]: ce = 1.19333878 * 2560; err = 0.37148437 * 2560; time = 0.6077s; samplesPerSecond = 4212.4
01/17/2018 07:43:34:  Epoch[ 2 of 2]-Minibatch[  81-  90, 28.13%]: ce = 1.23995056 * 2560; err = 0.37773438 * 2560; time = 0.6105s; samplesPerSecond = 4193.5
01/17/2018 07:43:35:  Epoch[ 2 of 2]-Minibatch[  91- 100, 31.25%]: ce = 1.18914948 * 2560; err = 0.36562500 * 2560; time = 0.6099s; samplesPerSecond = 4197.5
01/17/2018 07:43:35:  Epoch[ 2 of 2]-Minibatch[ 101- 110, 34.38%]: ce = 1.16878815 * 2560; err = 0.35820313 * 2560; time = 0.6081s; samplesPerSecond = 4209.5
01/17/2018 07:43:36:  Epoch[ 2 of 2]-Minibatch[ 111- 120, 37.50%]: ce = 1.24551544 * 2560; err = 0.38242188 * 2560; time = 0.6091s; samplesPerSecond = 4203.1
01/17/2018 07:43:36:  Epoch[ 2 of 2]-Minibatch[ 121- 130, 40.63%]: ce = 1.18659210 * 2560; err = 0.35039063 * 2560; time = 0.6142s; samplesPerSecond = 4167.8
01/17/2018 07:43:37:  Epoch[ 2 of 2]-Minibatch[ 131- 140, 43.75%]: ce = 1.21731110 * 2560; err = 0.37031250 * 2560; time = 0.6063s; samplesPerSecond = 4222.6
01/17/2018 07:43:38:  Epoch[ 2 of 2]-Minibatch[ 141- 150, 46.88%]: ce = 1.19863586 * 2560; err = 0.37109375 * 2560; time = 0.6086s; samplesPerSecond = 4206.4
01/17/2018 07:43:38:  Epoch[ 2 of 2]-Minibatch[ 151- 160, 50.00%]: ce = 1.14327850 * 2560; err = 0.34453125 * 2560; time = 0.6061s; samplesPerSecond = 4223.6
01/17/2018 07:43:39:  Epoch[ 2 of 2]-Minibatch[ 161- 170, 53.13%]: ce = 1.14432526 * 2560; err = 0.35742188 * 2560; time = 0.6120s; samplesPerSecond = 4183.0
01/17/2018 07:43:39:  Epoch[ 2 of 2]-Minibatch[ 171- 180, 56.25%]: ce = 1.17652893 * 2560; err = 0.35195312 * 2560; time = 0.6201s; samplesPerSecond = 4128.1
01/17/2018 07:43:40:  Epoch[ 2 of 2]-Minibatch[ 181- 190, 59.38%]: ce = 1.15350037 * 2560; err = 0.35898438 * 2560; time = 0.6070s; samplesPerSecond = 4217.1
01/17/2018 07:43:41:  Epoch[ 2 of 2]-Minibatch[ 191- 200, 62.50%]: ce = 1.08181000 * 2560; err = 0.33320312 * 2560; time = 0.6066s; samplesPerSecond = 4220.6
01/17/2018 07:43:41:  Epoch[ 2 of 2]-Minibatch[ 201- 210, 65.63%]: ce = 1.14218140 * 2560; err = 0.34960938 * 2560; time = 0.6072s; samplesPerSecond = 4216.4
01/17/2018 07:43:42:  Epoch[ 2 of 2]-Minibatch[ 211- 220, 68.75%]: ce = 1.17052460 * 2560; err = 0.35898438 * 2560; time = 0.6387s; samplesPerSecond = 4008.0
01/17/2018 07:43:43:  Epoch[ 2 of 2]-Minibatch[ 221- 230, 71.88%]: ce = 1.19296570 * 2560; err = 0.37929687 * 2560; time = 0.6210s; samplesPerSecond = 4122.1
01/17/2018 07:43:43:  Epoch[ 2 of 2]-Minibatch[ 231- 240, 75.00%]: ce = 1.15556946 * 2560; err = 0.35078125 * 2560; time = 0.6150s; samplesPerSecond = 4162.5
01/17/2018 07:43:44:  Epoch[ 2 of 2]-Minibatch[ 241- 250, 78.13%]: ce = 1.15362244 * 2560; err = 0.35351563 * 2560; time = 0.6268s; samplesPerSecond = 4084.1
01/17/2018 07:43:44:  Epoch[ 2 of 2]-Minibatch[ 251- 260, 81.25%]: ce = 1.08062744 * 2560; err = 0.32617188 * 2560; time = 0.6269s; samplesPerSecond = 4083.4
01/17/2018 07:43:45:  Epoch[ 2 of 2]-Minibatch[ 261- 270, 84.38%]: ce = 1.09230347 * 2560; err = 0.34570313 * 2560; time = 0.6074s; samplesPerSecond = 4214.9
01/17/2018 07:43:46:  Epoch[ 2 of 2]-Minibatch[ 271- 280, 87.50%]: ce = 1.06243591 * 2560; err = 0.33671875 * 2560; time = 0.6071s; samplesPerSecond = 4216.9
01/17/2018 07:43:46:  Epoch[ 2 of 2]-Minibatch[ 281- 290, 90.63%]: ce = 1.09511108 * 2560; err = 0.33281250 * 2560; time = 0.6096s; samplesPerSecond = 4199.7
01/17/2018 07:43:47:  Epoch[ 2 of 2]-Minibatch[ 291- 300, 93.75%]: ce = 1.14536438 * 2560; err = 0.35273437 * 2560; time = 0.6221s; samplesPerSecond = 4115.0
01/17/2018 07:43:47:  Epoch[ 2 of 2]-Minibatch[ 301- 310, 96.88%]: ce = 1.10204773 * 2560; err = 0.34218750 * 2560; time = 0.6086s; samplesPerSecond = 4206.5
01/17/2018 07:43:48:  Epoch[ 2 of 2]-Minibatch[ 311- 320, 100.00%]: ce = 1.07106323 * 2560; err = 0.32734375 * 2560; time = 0.5596s; samplesPerSecond = 4574.8
01/17/2018 07:43:48: Finished Epoch[ 2 of 2]: [Training] ce = 1.15852671 * 81920; err = 0.35554199 * 81920; totalSamplesSeen = 163840; learningRatePerSample = 0.003125; epochTime=19.6197s
01/17/2018 07:43:48: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre1/cntkSpeech'

01/17/2018 07:43:48: Action "train" complete.


01/17/2018 07:43:48: ##############################################################################
01/17/2018 07:43:48: #                                                                            #
01/17/2018 07:43:48: # addLayer2 command (edit action)                                            #
01/17/2018 07:43:48: #                                                                            #
01/17/2018 07:43:48: ##############################################################################


01/17/2018 07:43:49: Action "edit" complete.


01/17/2018 07:43:49: ##############################################################################
01/17/2018 07:43:49: #                                                                            #
01/17/2018 07:43:49: # dptPre2 command (train action)                                             #
01/17/2018 07:43:49: #                                                                            #
01/17/2018 07:43:49: ##############################################################################

01/17/2018 07:43:49: 
Starting from checkpoint. Loading network from 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech.0'.
NDLBuilder Using GPU 0
reading script file C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp ... 948 entries
total 132 state names in state list C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list
htkmlfreader: reading MLF file C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.mlf ... total 948 entries
...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
label set 0: 129 classes
minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
01/17/2018 07:43:53: 
Model has 24 nodes. Using GPU 0.

01/17/2018 07:43:53: Training criterion:   ce = CrossEntropyWithSoftmax
01/17/2018 07:43:53: Evaluation criterion: err = ClassificationError

01/17/2018 07:43:53: Training 516740 parameters in 6 out of 6 parameter tensors and 15 nodes with gradient:

01/17/2018 07:43:53: 	Node 'HL1.W' (LearnableParameter operation) : [512 x 363]
01/17/2018 07:43:53: 	Node 'HL1.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:43:53: 	Node 'HL2.W' (LearnableParameter operation) : [512 x 512]
01/17/2018 07:43:53: 	Node 'HL2.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:43:53: 	Node 'OL.W' (LearnableParameter operation) : [132 x 512]
01/17/2018 07:43:53: 	Node 'OL.b' (LearnableParameter operation) : [132 x 1]

01/17/2018 07:43:53: No PreCompute nodes found, or all already computed. Skipping pre-computation step.

01/17/2018 07:43:53: Starting Epoch 1: learning rate per sample = 0.003125  effective momentum = 0.900000  momentum as time constant = 2429.8 samples
minibatchiterator: epoch 0: frames [0..81920] (first utterance at frame 0), data subset 0 of 1, with 1 datapasses
requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms

01/17/2018 07:44:09: Starting minibatch loop.
01/17/2018 07:44:09:  Epoch[ 1 of 2]-Minibatch[   1-  10, 3.13%]: ce = 4.76453552 * 2560; err = 0.80664063 * 2560; time = 0.6392s; samplesPerSecond = 4005.3
01/17/2018 07:44:10:  Epoch[ 1 of 2]-Minibatch[  11-  20, 6.25%]: ce = 2.85073395 * 2560; err = 0.69375000 * 2560; time = 0.6297s; samplesPerSecond = 4065.5
01/17/2018 07:44:11:  Epoch[ 1 of 2]-Minibatch[  21-  30, 9.38%]: ce = 2.28368835 * 2560; err = 0.59257812 * 2560; time = 0.6306s; samplesPerSecond = 4059.4
01/17/2018 07:44:11:  Epoch[ 1 of 2]-Minibatch[  31-  40, 12.50%]: ce = 1.94338684 * 2560; err = 0.52148438 * 2560; time = 0.6299s; samplesPerSecond = 4064.4
01/17/2018 07:44:12:  Epoch[ 1 of 2]-Minibatch[  41-  50, 15.63%]: ce = 1.72908173 * 2560; err = 0.48085937 * 2560; time = 0.6267s; samplesPerSecond = 4084.9
01/17/2018 07:44:12:  Epoch[ 1 of 2]-Minibatch[  51-  60, 18.75%]: ce = 1.62283478 * 2560; err = 0.46640625 * 2560; time = 0.6278s; samplesPerSecond = 4077.7
01/17/2018 07:44:13:  Epoch[ 1 of 2]-Minibatch[  61-  70, 21.88%]: ce = 1.51846161 * 2560; err = 0.44609375 * 2560; time = 0.6277s; samplesPerSecond = 4078.1
01/17/2018 07:44:14:  Epoch[ 1 of 2]-Minibatch[  71-  80, 25.00%]: ce = 1.50159607 * 2560; err = 0.44218750 * 2560; time = 0.6419s; samplesPerSecond = 3988.3
01/17/2018 07:44:14:  Epoch[ 1 of 2]-Minibatch[  81-  90, 28.13%]: ce = 1.46150055 * 2560; err = 0.42265625 * 2560; time = 0.6526s; samplesPerSecond = 3922.6
01/17/2018 07:44:15:  Epoch[ 1 of 2]-Minibatch[  91- 100, 31.25%]: ce = 1.41712341 * 2560; err = 0.41406250 * 2560; time = 0.6260s; samplesPerSecond = 4089.4
01/17/2018 07:44:16:  Epoch[ 1 of 2]-Minibatch[ 101- 110, 34.38%]: ce = 1.42065277 * 2560; err = 0.41250000 * 2560; time = 0.6296s; samplesPerSecond = 4066.4
01/17/2018 07:44:16:  Epoch[ 1 of 2]-Minibatch[ 111- 120, 37.50%]: ce = 1.34741516 * 2560; err = 0.39765625 * 2560; time = 0.6324s; samplesPerSecond = 4047.8
01/17/2018 07:44:17:  Epoch[ 1 of 2]-Minibatch[ 121- 130, 40.63%]: ce = 1.32274933 * 2560; err = 0.39140625 * 2560; time = 0.6249s; samplesPerSecond = 4096.3
01/17/2018 07:44:18:  Epoch[ 1 of 2]-Minibatch[ 131- 140, 43.75%]: ce = 1.33492279 * 2560; err = 0.39296875 * 2560; time = 0.6258s; samplesPerSecond = 4090.8
01/17/2018 07:44:18:  Epoch[ 1 of 2]-Minibatch[ 141- 150, 46.88%]: ce = 1.32771606 * 2560; err = 0.39140625 * 2560; time = 0.6242s; samplesPerSecond = 4101.4
01/17/2018 07:44:19:  Epoch[ 1 of 2]-Minibatch[ 151- 160, 50.00%]: ce = 1.28157349 * 2560; err = 0.39101562 * 2560; time = 0.6248s; samplesPerSecond = 4097.4
01/17/2018 07:44:19:  Epoch[ 1 of 2]-Minibatch[ 161- 170, 53.13%]: ce = 1.29435425 * 2560; err = 0.38242188 * 2560; time = 0.6350s; samplesPerSecond = 4031.5
01/17/2018 07:44:20:  Epoch[ 1 of 2]-Minibatch[ 171- 180, 56.25%]: ce = 1.29888611 * 2560; err = 0.38945313 * 2560; time = 0.6641s; samplesPerSecond = 3854.7
01/17/2018 07:44:21:  Epoch[ 1 of 2]-Minibatch[ 181- 190, 59.38%]: ce = 1.33448181 * 2560; err = 0.40429688 * 2560; time = 0.6259s; samplesPerSecond = 4090.2
01/17/2018 07:44:21:  Epoch[ 1 of 2]-Minibatch[ 191- 200, 62.50%]: ce = 1.32330017 * 2560; err = 0.40898438 * 2560; time = 0.6261s; samplesPerSecond = 4088.6
01/17/2018 07:44:22:  Epoch[ 1 of 2]-Minibatch[ 201- 210, 65.63%]: ce = 1.23590393 * 2560; err = 0.37382813 * 2560; time = 0.6252s; samplesPerSecond = 4095.0
01/17/2018 07:44:23:  Epoch[ 1 of 2]-Minibatch[ 211- 220, 68.75%]: ce = 1.27341003 * 2560; err = 0.38867188 * 2560; time = 0.6260s; samplesPerSecond = 4089.7
01/17/2018 07:44:23:  Epoch[ 1 of 2]-Minibatch[ 221- 230, 71.88%]: ce = 1.26567078 * 2560; err = 0.37460938 * 2560; time = 0.6283s; samplesPerSecond = 4074.3
01/17/2018 07:44:24:  Epoch[ 1 of 2]-Minibatch[ 231- 240, 75.00%]: ce = 1.24478149 * 2560; err = 0.36914063 * 2560; time = 0.6254s; samplesPerSecond = 4093.2
01/17/2018 07:44:24:  Epoch[ 1 of 2]-Minibatch[ 241- 250, 78.13%]: ce = 1.21717224 * 2560; err = 0.36757812 * 2560; time = 0.6249s; samplesPerSecond = 4096.5
01/17/2018 07:44:25:  Epoch[ 1 of 2]-Minibatch[ 251- 260, 81.25%]: ce = 1.18839111 * 2560; err = 0.36835937 * 2560; time = 0.6270s; samplesPerSecond = 4082.9
01/17/2018 07:44:26:  Epoch[ 1 of 2]-Minibatch[ 261- 270, 84.38%]: ce = 1.23902588 * 2560; err = 0.36601563 * 2560; time = 0.6454s; samplesPerSecond = 3966.5
01/17/2018 07:44:26:  Epoch[ 1 of 2]-Minibatch[ 271- 280, 87.50%]: ce = 1.18707581 * 2560; err = 0.36093750 * 2560; time = 0.6297s; samplesPerSecond = 4065.6
01/17/2018 07:44:27:  Epoch[ 1 of 2]-Minibatch[ 281- 290, 90.63%]: ce = 1.16996765 * 2560; err = 0.35195312 * 2560; time = 0.6273s; samplesPerSecond = 4080.9
01/17/2018 07:44:28:  Epoch[ 1 of 2]-Minibatch[ 291- 300, 93.75%]: ce = 1.14782104 * 2560; err = 0.34179688 * 2560; time = 0.6303s; samplesPerSecond = 4061.4
01/17/2018 07:44:28:  Epoch[ 1 of 2]-Minibatch[ 301- 310, 96.88%]: ce = 1.17535706 * 2560; err = 0.35234375 * 2560; time = 0.6297s; samplesPerSecond = 4065.6
01/17/2018 07:44:29:  Epoch[ 1 of 2]-Minibatch[ 311- 320, 100.00%]: ce = 1.19136047 * 2560; err = 0.37031250 * 2560; time = 0.5925s; samplesPerSecond = 4320.4
01/17/2018 07:44:29: Finished Epoch[ 1 of 2]: [Training] ce = 1.52859163 * 81920; err = 0.42607422 * 81920; totalSamplesSeen = 81920; learningRatePerSample = 0.003125; epochTime=35.8619s
01/17/2018 07:44:29: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech.1'

01/17/2018 07:44:29: Starting Epoch 2: learning rate per sample = 0.003125  effective momentum = 0.900000  momentum as time constant = 2429.8 samples
minibatchiterator: epoch 1: frames [81920..163840] (first utterance at frame 81920), data subset 0 of 1, with 1 datapasses

01/17/2018 07:44:29: Starting minibatch loop.
01/17/2018 07:44:30:  Epoch[ 2 of 2]-Minibatch[   1-  10, 3.13%]: ce = 1.18167496 * 2560; err = 0.35195312 * 2560; time = 0.6302s; samplesPerSecond = 4062.4
01/17/2018 07:44:30:  Epoch[ 2 of 2]-Minibatch[  11-  20, 6.25%]: ce = 1.15440292 * 2560; err = 0.35664062 * 2560; time = 0.6263s; samplesPerSecond = 4087.6
01/17/2018 07:44:31:  Epoch[ 2 of 2]-Minibatch[  21-  30, 9.38%]: ce = 1.15731888 * 2560; err = 0.34804687 * 2560; time = 0.6258s; samplesPerSecond = 4090.7
01/17/2018 07:44:32:  Epoch[ 2 of 2]-Minibatch[  31-  40, 12.50%]: ce = 1.13935738 * 2560; err = 0.34335938 * 2560; time = 0.6261s; samplesPerSecond = 4088.8
01/17/2018 07:44:32:  Epoch[ 2 of 2]-Minibatch[  41-  50, 15.63%]: ce = 1.14808121 * 2560; err = 0.36367187 * 2560; time = 0.6274s; samplesPerSecond = 4080.1
01/17/2018 07:44:33:  Epoch[ 2 of 2]-Minibatch[  51-  60, 18.75%]: ce = 1.14439125 * 2560; err = 0.33867188 * 2560; time = 0.6288s; samplesPerSecond = 4071.1
01/17/2018 07:44:34:  Epoch[ 2 of 2]-Minibatch[  61-  70, 21.88%]: ce = 1.09884491 * 2560; err = 0.34179688 * 2560; time = 0.6301s; samplesPerSecond = 4062.8
01/17/2018 07:44:34:  Epoch[ 2 of 2]-Minibatch[  71-  80, 25.00%]: ce = 1.15980606 * 2560; err = 0.35859375 * 2560; time = 0.6244s; samplesPerSecond = 4100.0
01/17/2018 07:44:35:  Epoch[ 2 of 2]-Minibatch[  81-  90, 28.13%]: ce = 1.17044067 * 2560; err = 0.36093750 * 2560; time = 0.6395s; samplesPerSecond = 4003.2
01/17/2018 07:44:35:  Epoch[ 2 of 2]-Minibatch[  91- 100, 31.25%]: ce = 1.11957779 * 2560; err = 0.34531250 * 2560; time = 0.6281s; samplesPerSecond = 4075.9
01/17/2018 07:44:36:  Epoch[ 2 of 2]-Minibatch[ 101- 110, 34.38%]: ce = 1.11841125 * 2560; err = 0.34570313 * 2560; time = 0.6412s; samplesPerSecond = 3992.4
01/17/2018 07:44:37:  Epoch[ 2 of 2]-Minibatch[ 111- 120, 37.50%]: ce = 1.19153061 * 2560; err = 0.36328125 * 2560; time = 0.6420s; samplesPerSecond = 3987.6
01/17/2018 07:44:37:  Epoch[ 2 of 2]-Minibatch[ 121- 130, 40.63%]: ce = 1.13423920 * 2560; err = 0.33945313 * 2560; time = 0.6245s; samplesPerSecond = 4099.3
01/17/2018 07:44:38:  Epoch[ 2 of 2]-Minibatch[ 131- 140, 43.75%]: ce = 1.16283112 * 2560; err = 0.35312500 * 2560; time = 0.6243s; samplesPerSecond = 4100.5
01/17/2018 07:44:39:  Epoch[ 2 of 2]-Minibatch[ 141- 150, 46.88%]: ce = 1.12025299 * 2560; err = 0.34335938 * 2560; time = 0.6253s; samplesPerSecond = 4093.9
01/17/2018 07:44:39:  Epoch[ 2 of 2]-Minibatch[ 151- 160, 50.00%]: ce = 1.09458618 * 2560; err = 0.33750000 * 2560; time = 0.6268s; samplesPerSecond = 4084.1
01/17/2018 07:44:40:  Epoch[ 2 of 2]-Minibatch[ 161- 170, 53.13%]: ce = 1.10369873 * 2560; err = 0.33984375 * 2560; time = 0.6288s; samplesPerSecond = 4071.0
01/17/2018 07:44:41:  Epoch[ 2 of 2]-Minibatch[ 171- 180, 56.25%]: ce = 1.11500549 * 2560; err = 0.32890625 * 2560; time = 0.6268s; samplesPerSecond = 4084.4
01/17/2018 07:44:41:  Epoch[ 2 of 2]-Minibatch[ 181- 190, 59.38%]: ce = 1.09892426 * 2560; err = 0.33554688 * 2560; time = 0.6448s; samplesPerSecond = 3970.3
01/17/2018 07:44:42:  Epoch[ 2 of 2]-Minibatch[ 191- 200, 62.50%]: ce = 1.06063843 * 2560; err = 0.33007813 * 2560; time = 0.6314s; samplesPerSecond = 4054.8
01/17/2018 07:44:42:  Epoch[ 2 of 2]-Minibatch[ 201- 210, 65.63%]: ce = 1.10568542 * 2560; err = 0.34257813 * 2560; time = 0.6248s; samplesPerSecond = 4097.1
01/17/2018 07:44:43:  Epoch[ 2 of 2]-Minibatch[ 211- 220, 68.75%]: ce = 1.13217773 * 2560; err = 0.34414062 * 2560; time = 0.6278s; samplesPerSecond = 4077.9
01/17/2018 07:44:44:  Epoch[ 2 of 2]-Minibatch[ 221- 230, 71.88%]: ce = 1.12288818 * 2560; err = 0.33984375 * 2560; time = 0.6323s; samplesPerSecond = 4048.7
01/17/2018 07:44:44:  Epoch[ 2 of 2]-Minibatch[ 231- 240, 75.00%]: ce = 1.10664063 * 2560; err = 0.33789063 * 2560; time = 0.6316s; samplesPerSecond = 4053.4
01/17/2018 07:44:45:  Epoch[ 2 of 2]-Minibatch[ 241- 250, 78.13%]: ce = 1.09506836 * 2560; err = 0.33281250 * 2560; time = 0.6279s; samplesPerSecond = 4077.3
01/17/2018 07:44:46:  Epoch[ 2 of 2]-Minibatch[ 251- 260, 81.25%]: ce = 1.04699707 * 2560; err = 0.32109375 * 2560; time = 0.6252s; samplesPerSecond = 4094.4
01/17/2018 07:44:46:  Epoch[ 2 of 2]-Minibatch[ 261- 270, 84.38%]: ce = 1.04719543 * 2560; err = 0.33632812 * 2560; time = 0.6361s; samplesPerSecond = 4024.3
01/17/2018 07:44:47:  Epoch[ 2 of 2]-Minibatch[ 271- 280, 87.50%]: ce = 1.03957214 * 2560; err = 0.32382813 * 2560; time = 0.6368s; samplesPerSecond = 4020.2
01/17/2018 07:44:47:  Epoch[ 2 of 2]-Minibatch[ 281- 290, 90.63%]: ce = 1.05847168 * 2560; err = 0.32695313 * 2560; time = 0.6417s; samplesPerSecond = 3989.3
01/17/2018 07:44:48:  Epoch[ 2 of 2]-Minibatch[ 291- 300, 93.75%]: ce = 1.11241150 * 2560; err = 0.33867188 * 2560; time = 0.6259s; samplesPerSecond = 4090.4
01/17/2018 07:44:49:  Epoch[ 2 of 2]-Minibatch[ 301- 310, 96.88%]: ce = 1.08333740 * 2560; err = 0.33398438 * 2560; time = 0.6252s; samplesPerSecond = 4094.4
01/17/2018 07:44:49:  Epoch[ 2 of 2]-Minibatch[ 311- 320, 100.00%]: ce = 1.04971313 * 2560; err = 0.32226563 * 2560; time = 0.5773s; samplesPerSecond = 4434.2
01/17/2018 07:44:49: Finished Epoch[ 2 of 2]: [Training] ce = 1.11481791 * 81920; err = 0.34144287 * 81920; totalSamplesSeen = 163840; learningRatePerSample = 0.003125; epochTime=20.1859s
01/17/2018 07:44:49: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/Pre2/cntkSpeech'

01/17/2018 07:44:50: Action "train" complete.


01/17/2018 07:44:50: ##############################################################################
01/17/2018 07:44:50: #                                                                            #
01/17/2018 07:44:50: # addLayer3 command (edit action)                                            #
01/17/2018 07:44:50: #                                                                            #
01/17/2018 07:44:50: ##############################################################################


01/17/2018 07:44:50: Action "edit" complete.


01/17/2018 07:44:50: ##############################################################################
01/17/2018 07:44:50: #                                                                            #
01/17/2018 07:44:50: # speechTrain command (train action)                                         #
01/17/2018 07:44:50: #                                                                            #
01/17/2018 07:44:50: ##############################################################################

01/17/2018 07:44:50: 
Starting from checkpoint. Loading network from 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.0'.
NDLBuilder Using GPU 0
reading script file C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp ... 948 entries
total 132 state names in state list C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list
htkmlfreader: reading MLF file C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.mlf ... total 948 entries
...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
label set 0: 129 classes
minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
01/17/2018 07:44:55: 
Model has 29 nodes. Using GPU 0.

01/17/2018 07:44:55: Training criterion:   ce = CrossEntropyWithSoftmax
01/17/2018 07:44:55: Evaluation criterion: err = ClassificationError

01/17/2018 07:44:55: Training 779396 parameters in 8 out of 8 parameter tensors and 20 nodes with gradient:

01/17/2018 07:44:55: 	Node 'HL1.W' (LearnableParameter operation) : [512 x 363]
01/17/2018 07:44:55: 	Node 'HL1.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:44:55: 	Node 'HL2.W' (LearnableParameter operation) : [512 x 512]
01/17/2018 07:44:55: 	Node 'HL2.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:44:55: 	Node 'HL3.W' (LearnableParameter operation) : [512 x 512]
01/17/2018 07:44:55: 	Node 'HL3.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:44:55: 	Node 'OL.W' (LearnableParameter operation) : [132 x 512]
01/17/2018 07:44:55: 	Node 'OL.b' (LearnableParameter operation) : [132 x 1]

01/17/2018 07:44:55: No PreCompute nodes found, or all already computed. Skipping pre-computation step.

01/17/2018 07:44:55: Starting Epoch 1: learning rate per sample = 0.003125  effective momentum = 0.900117  momentum as time constant = 2432.7 samples
minibatchiterator: epoch 0: frames [0..81920] (first utterance at frame 0), data subset 0 of 1, with 1 datapasses
requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms

01/17/2018 07:45:10: Starting minibatch loop.
01/17/2018 07:45:11:  Epoch[ 1 of 4]-Minibatch[   1-  10, 3.13%]: ce = 4.10357437 * 2560; err = 0.82031250 * 2560; time = 0.6634s; samplesPerSecond = 3858.7
01/17/2018 07:45:11:  Epoch[ 1 of 4]-Minibatch[  11-  20, 6.25%]: ce = 2.57244034 * 2560; err = 0.63671875 * 2560; time = 0.6437s; samplesPerSecond = 3977.2
01/17/2018 07:45:12:  Epoch[ 1 of 4]-Minibatch[  21-  30, 9.38%]: ce = 2.03364334 * 2560; err = 0.54179687 * 2560; time = 0.6439s; samplesPerSecond = 3975.8
01/17/2018 07:45:13:  Epoch[ 1 of 4]-Minibatch[  31-  40, 12.50%]: ce = 1.73527985 * 2560; err = 0.47500000 * 2560; time = 0.6533s; samplesPerSecond = 3918.4
01/17/2018 07:45:13:  Epoch[ 1 of 4]-Minibatch[  41-  50, 15.63%]: ce = 1.54782410 * 2560; err = 0.43945313 * 2560; time = 0.6510s; samplesPerSecond = 3932.3
01/17/2018 07:45:14:  Epoch[ 1 of 4]-Minibatch[  51-  60, 18.75%]: ce = 1.44337387 * 2560; err = 0.41210938 * 2560; time = 0.6533s; samplesPerSecond = 3918.7
01/17/2018 07:45:15:  Epoch[ 1 of 4]-Minibatch[  61-  70, 21.88%]: ce = 1.36145630 * 2560; err = 0.40585938 * 2560; time = 0.6445s; samplesPerSecond = 3971.8
01/17/2018 07:45:15:  Epoch[ 1 of 4]-Minibatch[  71-  80, 25.00%]: ce = 1.35890045 * 2560; err = 0.39804688 * 2560; time = 0.6443s; samplesPerSecond = 3973.6
01/17/2018 07:45:16:  Epoch[ 1 of 4]-Minibatch[  81-  90, 28.13%]: ce = 1.34205933 * 2560; err = 0.38945313 * 2560; time = 0.6435s; samplesPerSecond = 3978.3
01/17/2018 07:45:17:  Epoch[ 1 of 4]-Minibatch[  91- 100, 31.25%]: ce = 1.30447235 * 2560; err = 0.38046875 * 2560; time = 0.6445s; samplesPerSecond = 3971.8
01/17/2018 07:45:17:  Epoch[ 1 of 4]-Minibatch[ 101- 110, 34.38%]: ce = 1.31331329 * 2560; err = 0.38710937 * 2560; time = 0.6448s; samplesPerSecond = 3970.1
01/17/2018 07:45:18:  Epoch[ 1 of 4]-Minibatch[ 111- 120, 37.50%]: ce = 1.23968048 * 2560; err = 0.36953125 * 2560; time = 0.6425s; samplesPerSecond = 3984.7
01/17/2018 07:45:19:  Epoch[ 1 of 4]-Minibatch[ 121- 130, 40.63%]: ce = 1.21697845 * 2560; err = 0.35742188 * 2560; time = 0.6450s; samplesPerSecond = 3969.0
01/17/2018 07:45:19:  Epoch[ 1 of 4]-Minibatch[ 131- 140, 43.75%]: ce = 1.24119873 * 2560; err = 0.36992188 * 2560; time = 0.6477s; samplesPerSecond = 3952.7
01/17/2018 07:45:20:  Epoch[ 1 of 4]-Minibatch[ 141- 150, 46.88%]: ce = 1.23539429 * 2560; err = 0.37382813 * 2560; time = 0.6659s; samplesPerSecond = 3844.2
01/17/2018 07:45:21:  Epoch[ 1 of 4]-Minibatch[ 151- 160, 50.00%]: ce = 1.19547272 * 2560; err = 0.35195312 * 2560; time = 0.6449s; samplesPerSecond = 3969.5
01/17/2018 07:45:21:  Epoch[ 1 of 4]-Minibatch[ 161- 170, 53.13%]: ce = 1.21609497 * 2560; err = 0.36250000 * 2560; time = 0.6440s; samplesPerSecond = 3975.2
01/17/2018 07:45:22:  Epoch[ 1 of 4]-Minibatch[ 171- 180, 56.25%]: ce = 1.24842529 * 2560; err = 0.37500000 * 2560; time = 0.6433s; samplesPerSecond = 3979.4
01/17/2018 07:45:22:  Epoch[ 1 of 4]-Minibatch[ 181- 190, 59.38%]: ce = 1.26809998 * 2560; err = 0.38359375 * 2560; time = 0.6441s; samplesPerSecond = 3974.3
01/17/2018 07:45:23:  Epoch[ 1 of 4]-Minibatch[ 191- 200, 62.50%]: ce = 1.22535706 * 2560; err = 0.38632813 * 2560; time = 0.6445s; samplesPerSecond = 3972.3
01/17/2018 07:45:24:  Epoch[ 1 of 4]-Minibatch[ 201- 210, 65.63%]: ce = 1.17337036 * 2560; err = 0.35781250 * 2560; time = 0.6434s; samplesPerSecond = 3978.9
01/17/2018 07:45:24:  Epoch[ 1 of 4]-Minibatch[ 211- 220, 68.75%]: ce = 1.19753418 * 2560; err = 0.36679688 * 2560; time = 0.6438s; samplesPerSecond = 3976.2
01/17/2018 07:45:25:  Epoch[ 1 of 4]-Minibatch[ 221- 230, 71.88%]: ce = 1.21829834 * 2560; err = 0.36015625 * 2560; time = 0.6428s; samplesPerSecond = 3982.4
01/17/2018 07:45:26:  Epoch[ 1 of 4]-Minibatch[ 231- 240, 75.00%]: ce = 1.18372192 * 2560; err = 0.35390625 * 2560; time = 0.6546s; samplesPerSecond = 3910.5
01/17/2018 07:45:26:  Epoch[ 1 of 4]-Minibatch[ 241- 250, 78.13%]: ce = 1.16905823 * 2560; err = 0.35781250 * 2560; time = 0.6490s; samplesPerSecond = 3944.5
01/17/2018 07:45:27:  Epoch[ 1 of 4]-Minibatch[ 251- 260, 81.25%]: ce = 1.12310791 * 2560; err = 0.35000000 * 2560; time = 0.6431s; samplesPerSecond = 3980.8
01/17/2018 07:45:28:  Epoch[ 1 of 4]-Minibatch[ 261- 270, 84.38%]: ce = 1.18959351 * 2560; err = 0.35781250 * 2560; time = 0.6577s; samplesPerSecond = 3892.2
01/17/2018 07:45:28:  Epoch[ 1 of 4]-Minibatch[ 271- 280, 87.50%]: ce = 1.13449707 * 2560; err = 0.35273437 * 2560; time = 0.6466s; samplesPerSecond = 3959.2
01/17/2018 07:45:29:  Epoch[ 1 of 4]-Minibatch[ 281- 290, 90.63%]: ce = 1.12417908 * 2560; err = 0.33476563 * 2560; time = 0.6458s; samplesPerSecond = 3964.1
01/17/2018 07:45:30:  Epoch[ 1 of 4]-Minibatch[ 291- 300, 93.75%]: ce = 1.10653687 * 2560; err = 0.33867188 * 2560; time = 0.6445s; samplesPerSecond = 3972.0
01/17/2018 07:45:30:  Epoch[ 1 of 4]-Minibatch[ 301- 310, 96.88%]: ce = 1.12954407 * 2560; err = 0.34414062 * 2560; time = 0.6437s; samplesPerSecond = 3977.0
01/17/2018 07:45:31:  Epoch[ 1 of 4]-Minibatch[ 311- 320, 100.00%]: ce = 1.12625122 * 2560; err = 0.34765625 * 2560; time = 0.5946s; samplesPerSecond = 4305.1
01/17/2018 07:45:31: Finished Epoch[ 1 of 4]: [Training] ce = 1.40871038 * 81920; err = 0.40120850 * 81920; totalSamplesSeen = 81920; learningRatePerSample = 0.003125; epochTime=36.2743s
01/17/2018 07:45:31: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.1'

01/17/2018 07:45:31: Starting Epoch 2: learning rate per sample = 0.003125  effective momentum = 0.810210  momentum as time constant = 2432.7 samples
minibatchiterator: epoch 1: frames [81920..163840] (first utterance at frame 81920), data subset 0 of 1, with 1 datapasses

01/17/2018 07:45:31: Starting minibatch loop.
01/17/2018 07:45:33:  Epoch[ 2 of 4]-Minibatch[   1-  10, 6.25%]: ce = 1.22988195 * 5120; err = 0.37324219 * 5120; time = 1.2172s; samplesPerSecond = 4206.5
01/17/2018 07:45:34:  Epoch[ 2 of 4]-Minibatch[  11-  20, 12.50%]: ce = 1.17050295 * 5120; err = 0.34707031 * 5120; time = 1.2046s; samplesPerSecond = 4250.5
01/17/2018 07:45:35:  Epoch[ 2 of 4]-Minibatch[  21-  30, 18.75%]: ce = 1.17057648 * 5120; err = 0.36054687 * 5120; time = 1.2118s; samplesPerSecond = 4225.2
01/17/2018 07:45:36:  Epoch[ 2 of 4]-Minibatch[  31-  40, 25.00%]: ce = 1.29325256 * 5120; err = 0.38945313 * 5120; time = 1.2060s; samplesPerSecond = 4245.5
01/17/2018 07:45:37:  Epoch[ 2 of 4]-Minibatch[  41-  50, 31.25%]: ce = 1.15675087 * 5120; err = 0.36054687 * 5120; time = 1.2039s; samplesPerSecond = 4252.8
01/17/2018 07:45:39:  Epoch[ 2 of 4]-Minibatch[  51-  60, 37.50%]: ce = 1.13633652 * 5120; err = 0.34531250 * 5120; time = 1.2072s; samplesPerSecond = 4241.4
01/17/2018 07:45:40:  Epoch[ 2 of 4]-Minibatch[  61-  70, 43.75%]: ce = 1.12745209 * 5120; err = 0.34179688 * 5120; time = 1.2122s; samplesPerSecond = 4223.6
01/17/2018 07:45:41:  Epoch[ 2 of 4]-Minibatch[  71-  80, 50.00%]: ce = 1.11147766 * 5120; err = 0.34628906 * 5120; time = 1.2066s; samplesPerSecond = 4243.2
01/17/2018 07:45:42:  Epoch[ 2 of 4]-Minibatch[  81-  90, 56.25%]: ce = 1.09499435 * 5120; err = 0.33671875 * 5120; time = 1.2054s; samplesPerSecond = 4247.7
01/17/2018 07:45:44:  Epoch[ 2 of 4]-Minibatch[  91- 100, 62.50%]: ce = 1.07211533 * 5120; err = 0.32812500 * 5120; time = 1.3054s; samplesPerSecond = 3922.2
01/17/2018 07:45:45:  Epoch[ 2 of 4]-Minibatch[ 101- 110, 68.75%]: ce = 1.12221298 * 5120; err = 0.34238281 * 5120; time = 1.2150s; samplesPerSecond = 4214.0
01/17/2018 07:45:46:  Epoch[ 2 of 4]-Minibatch[ 111- 120, 75.00%]: ce = 1.11653061 * 5120; err = 0.34785156 * 5120; time = 1.2063s; samplesPerSecond = 4244.3
01/17/2018 07:45:47:  Epoch[ 2 of 4]-Minibatch[ 121- 130, 81.25%]: ce = 1.05828247 * 5120; err = 0.32636719 * 5120; time = 1.2120s; samplesPerSecond = 4224.4
01/17/2018 07:45:48:  Epoch[ 2 of 4]-Minibatch[ 131- 140, 87.50%]: ce = 1.03556671 * 5120; err = 0.32656250 * 5120; time = 1.2044s; samplesPerSecond = 4251.0
01/17/2018 07:45:50:  Epoch[ 2 of 4]-Minibatch[ 141- 150, 93.75%]: ce = 1.09171906 * 5120; err = 0.33105469 * 5120; time = 1.2210s; samplesPerSecond = 4193.4
01/17/2018 07:45:51:  Epoch[ 2 of 4]-Minibatch[ 151- 160, 100.00%]: ce = 1.06842804 * 5120; err = 0.32578125 * 5120; time = 1.1130s; samplesPerSecond = 4600.1
01/17/2018 07:45:51: Finished Epoch[ 2 of 4]: [Training] ce = 1.12850504 * 81920; err = 0.34556885 * 81920; totalSamplesSeen = 163840; learningRatePerSample = 0.003125; epochTime=19.4793s
01/17/2018 07:45:51: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.2'

01/17/2018 07:45:51: Starting Epoch 3: learning rate per sample = 0.003125  effective momentum = 0.810210  momentum as time constant = 2432.7 samples
minibatchiterator: epoch 2: frames [163840..245760] (first utterance at frame 163840), data subset 0 of 1, with 1 datapasses

01/17/2018 07:45:51: Starting minibatch loop.
01/17/2018 07:45:52:  Epoch[ 3 of 4]-Minibatch[   1-  10, 6.25%]: ce = 1.13445969 * 5120; err = 0.34375000 * 5120; time = 1.2064s; samplesPerSecond = 4244.2
01/17/2018 07:45:54:  Epoch[ 3 of 4]-Minibatch[  11-  20, 12.50%]: ce = 1.07075939 * 5120; err = 0.33300781 * 5120; time = 1.2041s; samplesPerSecond = 4252.0
01/17/2018 07:45:55:  Epoch[ 3 of 4]-Minibatch[  21-  30, 18.75%]: ce = 1.06535053 * 5120; err = 0.33085938 * 5120; time = 1.2080s; samplesPerSecond = 4238.3
01/17/2018 07:45:56:  Epoch[ 3 of 4]-Minibatch[  31-  40, 25.00%]: ce = 1.07702370 * 5120; err = 0.33242187 * 5120; time = 1.2081s; samplesPerSecond = 4237.9
01/17/2018 07:45:57:  Epoch[ 3 of 4]-Minibatch[  41-  50, 31.25%]: ce = 1.07252197 * 5120; err = 0.32792969 * 5120; time = 1.2042s; samplesPerSecond = 4251.8
01/17/2018 07:45:59:  Epoch[ 3 of 4]-Minibatch[  51-  60, 37.50%]: ce = 1.05518150 * 5120; err = 0.32910156 * 5120; time = 1.2432s; samplesPerSecond = 4118.5
01/17/2018 07:46:00:  Epoch[ 3 of 4]-Minibatch[  61-  70, 43.75%]: ce = 1.06431198 * 5120; err = 0.32734375 * 5120; time = 1.2239s; samplesPerSecond = 4183.5
01/17/2018 07:46:01:  Epoch[ 3 of 4]-Minibatch[  71-  80, 50.00%]: ce = 1.07353668 * 5120; err = 0.32578125 * 5120; time = 1.2079s; samplesPerSecond = 4238.9
01/17/2018 07:46:02:  Epoch[ 3 of 4]-Minibatch[  81-  90, 56.25%]: ce = 1.03901520 * 5120; err = 0.31894531 * 5120; time = 1.2092s; samplesPerSecond = 4234.2
01/17/2018 07:46:03:  Epoch[ 3 of 4]-Minibatch[  91- 100, 62.50%]: ce = 1.05586166 * 5120; err = 0.32753906 * 5120; time = 1.2106s; samplesPerSecond = 4229.5
01/17/2018 07:46:05:  Epoch[ 3 of 4]-Minibatch[ 101- 110, 68.75%]: ce = 1.04341202 * 5120; err = 0.32363281 * 5120; time = 1.2165s; samplesPerSecond = 4208.9
01/17/2018 07:46:06:  Epoch[ 3 of 4]-Minibatch[ 111- 120, 75.00%]: ce = 1.07875061 * 5120; err = 0.33691406 * 5120; time = 1.2070s; samplesPerSecond = 4241.9
01/17/2018 07:46:07:  Epoch[ 3 of 4]-Minibatch[ 121- 130, 81.25%]: ce = 1.10670013 * 5120; err = 0.33203125 * 5120; time = 1.2105s; samplesPerSecond = 4229.7
01/17/2018 07:46:08:  Epoch[ 3 of 4]-Minibatch[ 131- 140, 87.50%]: ce = 1.06287842 * 5120; err = 0.32656250 * 5120; time = 1.2091s; samplesPerSecond = 4234.7
01/17/2018 07:46:09:  Epoch[ 3 of 4]-Minibatch[ 141- 150, 93.75%]: ce = 1.05311127 * 5120; err = 0.33398438 * 5120; time = 1.2081s; samplesPerSecond = 4238.2
01/17/2018 07:46:11:  Epoch[ 3 of 4]-Minibatch[ 151- 160, 100.00%]: ce = 1.04087524 * 5120; err = 0.32753906 * 5120; time = 1.1072s; samplesPerSecond = 4624.3
01/17/2018 07:46:11: Finished Epoch[ 3 of 4]: [Training] ce = 1.06835938 * 81920; err = 0.32983398 * 81920; totalSamplesSeen = 245760; learningRatePerSample = 0.003125; epochTime=19.4103s
01/17/2018 07:46:11: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.3'

01/17/2018 07:46:11: Starting Epoch 4: learning rate per sample = 0.003125  effective momentum = 0.810210  momentum as time constant = 2432.7 samples
minibatchiterator: epoch 3: frames [245760..327680] (first utterance at frame 245760), data subset 0 of 1, with 1 datapasses

01/17/2018 07:46:11: Starting minibatch loop.
01/17/2018 07:46:12:  Epoch[ 4 of 4]-Minibatch[   1-  10, 6.25%]: ce = 1.03330250 * 5120; err = 0.32500000 * 5120; time = 1.2246s; samplesPerSecond = 4181.0
01/17/2018 07:46:17:  Epoch[ 4 of 4]-Minibatch[  11-  20, 12.50%]: ce = 1.03590166 * 4926; err = 0.31303289 * 4926; time = 5.0547s; samplesPerSecond = 974.5
01/17/2018 07:46:19:  Epoch[ 4 of 4]-Minibatch[  21-  30, 18.75%]: ce = 1.01469421 * 5120; err = 0.32128906 * 5120; time = 1.2408s; samplesPerSecond = 4126.3
01/17/2018 07:46:20:  Epoch[ 4 of 4]-Minibatch[  31-  40, 25.00%]: ce = 1.01490192 * 5120; err = 0.31953125 * 5120; time = 1.2407s; samplesPerSecond = 4126.6
01/17/2018 07:46:21:  Epoch[ 4 of 4]-Minibatch[  41-  50, 31.25%]: ce = 0.99234657 * 5120; err = 0.30976562 * 5120; time = 1.2402s; samplesPerSecond = 4128.5
01/17/2018 07:46:22:  Epoch[ 4 of 4]-Minibatch[  51-  60, 37.50%]: ce = 0.99542236 * 5120; err = 0.30957031 * 5120; time = 1.2258s; samplesPerSecond = 4177.0
01/17/2018 07:46:24:  Epoch[ 4 of 4]-Minibatch[  61-  70, 43.75%]: ce = 0.99060974 * 5120; err = 0.30625000 * 5120; time = 1.2215s; samplesPerSecond = 4191.7
01/17/2018 07:46:25:  Epoch[ 4 of 4]-Minibatch[  71-  80, 50.00%]: ce = 1.01335754 * 5120; err = 0.31640625 * 5120; time = 1.2308s; samplesPerSecond = 4159.9
01/17/2018 07:46:26:  Epoch[ 4 of 4]-Minibatch[  81-  90, 56.25%]: ce = 0.99698181 * 5120; err = 0.31210938 * 5120; time = 1.2629s; samplesPerSecond = 4054.1
01/17/2018 07:46:27:  Epoch[ 4 of 4]-Minibatch[  91- 100, 62.50%]: ce = 0.97408829 * 5120; err = 0.31425781 * 5120; time = 1.2308s; samplesPerSecond = 4159.8
01/17/2018 07:46:28:  Epoch[ 4 of 4]-Minibatch[ 101- 110, 68.75%]: ce = 0.98822098 * 5120; err = 0.30429688 * 5120; time = 1.2463s; samplesPerSecond = 4108.3
01/17/2018 07:46:30:  Epoch[ 4 of 4]-Minibatch[ 111- 120, 75.00%]: ce = 1.00093079 * 5120; err = 0.30898437 * 5120; time = 1.2437s; samplesPerSecond = 4116.7
01/17/2018 07:46:31:  Epoch[ 4 of 4]-Minibatch[ 121- 130, 81.25%]: ce = 1.00725632 * 5120; err = 0.30976562 * 5120; time = 1.2112s; samplesPerSecond = 4227.3
01/17/2018 07:46:32:  Epoch[ 4 of 4]-Minibatch[ 131- 140, 87.50%]: ce = 0.97640991 * 5120; err = 0.30781250 * 5120; time = 1.2164s; samplesPerSecond = 4209.1
01/17/2018 07:46:33:  Epoch[ 4 of 4]-Minibatch[ 141- 150, 93.75%]: ce = 0.93567505 * 5120; err = 0.29589844 * 5120; time = 1.2074s; samplesPerSecond = 4240.5
01/17/2018 07:46:35:  Epoch[ 4 of 4]-Minibatch[ 151- 160, 100.00%]: ce = 0.97326965 * 5120; err = 0.30449219 * 5120; time = 1.1467s; samplesPerSecond = 4465.0
01/17/2018 07:46:35: Finished Epoch[ 4 of 4]: [Training] ce = 0.99624453 * 81920; err = 0.31113281 * 81920; totalSamplesSeen = 327680; learningRatePerSample = 0.003125; epochTime=23.6033s
01/17/2018 07:46:35: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech'

01/17/2018 07:46:35: Action "train" complete.


01/17/2018 07:46:35: ##############################################################################
01/17/2018 07:46:35: #                                                                            #
01/17/2018 07:46:35: # sequenceTrain command (train action)                                       #
01/17/2018 07:46:35: #                                                                            #
01/17/2018 07:46:35: ##############################################################################

01/17/2018 07:46:35: 
Creating virgin network.
Load: Loading model file: C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech
Post-processing network...

3 roots:
	ce = CrossEntropyWithSoftmax()
	err = ClassificationError()
	scaledLogLikelihood = Minus()

Validating network. 29 nodes to process in pass 1.

Validating --> labels = InputValue() :  -> [132 x *8]
Validating --> OL.W = LearnableParameter() :  -> [132 x 512]
Validating --> HL3.W = LearnableParameter() :  -> [512 x 512]
Validating --> HL2.W = LearnableParameter() :  -> [512 x 512]
Validating --> HL1.W = LearnableParameter() :  -> [512 x 363]
Validating --> features = InputValue() :  -> [363 x *8]
Validating --> globalMean = LearnableParameter() :  -> [363 x 1]
Validating --> globalInvStd = LearnableParameter() :  -> [363 x 1]
Validating --> featNorm = PerDimMeanVarNormalization (features, globalMean, globalInvStd) : [363 x *8], [363 x 1], [363 x 1] -> [363 x *8]
Validating --> HL1.t = Times (HL1.W, featNorm) : [512 x 363], [363 x *8] -> [512 x *8]
Validating --> HL1.b = LearnableParameter() :  -> [512 x 1]
Validating --> HL1.z = Plus (HL1.t, HL1.b) : [512 x *8], [512 x 1] -> [512 x 1 x *8]
Validating --> HL1.y = Sigmoid (HL1.z) : [512 x 1 x *8] -> [512 x 1 x *8]
Validating --> HL2.t = Times (HL2.W, HL1.y) : [512 x 512], [512 x 1 x *8] -> [512 x 1 x *8]
Validating --> HL2.b = LearnableParameter() :  -> [512 x 1]
Validating --> HL2.z = Plus (HL2.t, HL2.b) : [512 x 1 x *8], [512 x 1] -> [512 x 1 x *8]
Validating --> HL2.y = Sigmoid (HL2.z) : [512 x 1 x *8] -> [512 x 1 x *8]
Validating --> HL3.t = Times (HL3.W, HL2.y) : [512 x 512], [512 x 1 x *8] -> [512 x 1 x *8]
Validating --> HL3.b = LearnableParameter() :  -> [512 x 1]
Validating --> HL3.z = Plus (HL3.t, HL3.b) : [512 x 1 x *8], [512 x 1] -> [512 x 1 x *8]
Validating --> HL3.y = Sigmoid (HL3.z) : [512 x 1 x *8] -> [512 x 1 x *8]
Validating --> OL.t = Times (OL.W, HL3.y) : [132 x 512], [512 x 1 x *8] -> [132 x 1 x *8]
Validating --> OL.b = LearnableParameter() :  -> [132 x 1]
Validating --> OL.z = Plus (OL.t, OL.b) : [132 x 1 x *8], [132 x 1] -> [132 x 1 x *8]
Validating --> ce = CrossEntropyWithSoftmax (labels, OL.z) : [132 x *8], [132 x 1 x *8] -> [1]
Validating --> err = ClassificationError (labels, OL.z) : [132 x *8], [132 x 1 x *8] -> [1]
Validating --> globalPrior = LearnableParameter() :  -> [132 x 1]
Validating --> logPrior = Log (globalPrior) : [132 x 1] -> [132 x 1]
Validating --> scaledLogLikelihood = Minus (OL.z, logPrior) : [132 x 1 x *8], [132 x 1] -> [132 x 1 x *8]

Validating network. 16 nodes to process in pass 2.


Validating network, final pass.




Post-processing network complete.

CloneFunction: (features : InputValue) -> [
    netEval = OL.z : Plus
    scaledLogLikelihood = scaledLogLikelihood : Minus
]
clonedmodel.featNorm.inputs[0] = features (151) ==>  features (180)
clonedmodel.featNorm.inputs[1] = globalMean (153) ==>  clonedmodel.globalMean (195)
clonedmodel.featNorm.inputs[2] = globalInvStd (152) ==>  clonedmodel.globalInvStd (196)
clonedmodel.OL.t.inputs[0] = OL.W (174) ==>  clonedmodel.OL.W (181)
clonedmodel.OL.t.inputs[1] = HL3.y (168) ==>  clonedmodel.HL3.y (198)
clonedmodel.HL3.t.inputs[0] = HL3.W (167) ==>  clonedmodel.HL3.W (191)
clonedmodel.HL3.t.inputs[1] = HL2.y (163) ==>  clonedmodel.HL2.y (200)
clonedmodel.HL1.t.inputs[0] = HL1.W (157) ==>  clonedmodel.HL1.W (192)
clonedmodel.HL1.t.inputs[1] = featNorm (150) ==>  clonedmodel.featNorm (182)
clonedmodel.HL2.t.inputs[0] = HL2.W (162) ==>  clonedmodel.HL2.W (189)
clonedmodel.HL2.t.inputs[1] = HL1.y (158) ==>  clonedmodel.HL1.y (197)
clonedmodel.HL1.y.inputs[0] = HL1.z (159) ==>  clonedmodel.HL1.z (204)
clonedmodel.HL3.y.inputs[0] = HL3.z (169) ==>  clonedmodel.HL3.z (199)
clonedmodel.HL3.z.inputs[0] = HL3.t (166) ==>  clonedmodel.HL3.t (184)
clonedmodel.HL3.z.inputs[1] = HL3.b (165) ==>  clonedmodel.HL3.b (190)
clonedmodel.HL2.y.inputs[0] = HL2.z (164) ==>  clonedmodel.HL2.z (201)
clonedmodel.HL2.z.inputs[0] = HL2.t (161) ==>  clonedmodel.HL2.t (186)
clonedmodel.HL2.z.inputs[1] = HL2.b (160) ==>  clonedmodel.HL2.b (187)
clonedmodel.logPrior.inputs[0] = globalPrior (154) ==>  clonedmodel.globalPrior (194)
clonedmodel.OL.z.inputs[0] = OL.t (173) ==>  clonedmodel.OL.t (183)
clonedmodel.OL.z.inputs[1] = OL.b (172) ==>  clonedmodel.OL.b (193)
clonedmodel.HL1.z.inputs[0] = HL1.t (156) ==>  clonedmodel.HL1.t (185)
clonedmodel.HL1.z.inputs[1] = HL1.b (155) ==>  clonedmodel.HL1.b (188)
clonedmodel.scaledLogLikelihood.inputs[0] = OL.z (175) ==>  clonedmodel.OL.z (203)
clonedmodel.scaledLogLikelihood.inputs[1] = logPrior (171) ==>  clonedmodel.logPrior (202)
CloneFunction: Cloned 25 nodes and relinked 25 inputs.
01/17/2018 07:46:36: Reading files
 C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/CY2SCH010061231_1369712653.numden.lats.symlist 
 C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.overalltying 
 C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list 
 C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.transprob 
simplesenonehmm: reading 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.overalltying', 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list', 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/model.transprob'
simplesenonehmm: 83253 units with 45 unique HMMs, 132 tied states, and 45 trans matrices read

Post-processing network...

4 roots:
	Err = ClassificationError()
	clonedmodel.scaledLogLikelihood = Minus()
	cr = LatticeSequenceWithSoftmax()
	latticeAxis = DynamicAxis()

Validating network. 31 nodes to process in pass 1.

Validating --> labels = InputValue() :  -> [132 x *7]
Validating --> clonedmodel.OL.W = LearnableParameter() :  -> [132 x 512]
Validating --> clonedmodel.HL3.W = LearnableParameter() :  -> [512 x 512]
Validating --> clonedmodel.HL2.W = LearnableParameter() :  -> [512 x 512]
Validating --> clonedmodel.HL1.W = LearnableParameter() :  -> [512 x 363]
Validating --> features = InputValue() :  -> [363 x *7]
Validating --> clonedmodel.globalMean = LearnableParameter() :  -> [363 x 1]
Validating --> clonedmodel.globalInvStd = LearnableParameter() :  -> [363 x 1]
Validating --> clonedmodel.featNorm = PerDimMeanVarNormalization (features, clonedmodel.globalMean, clonedmodel.globalInvStd) : [363 x *7], [363 x 1], [363 x 1] -> [363 x *7]
Validating --> clonedmodel.HL1.t = Times (clonedmodel.HL1.W, clonedmodel.featNorm) : [512 x 363], [363 x *7] -> [512 x *7]
Validating --> clonedmodel.HL1.b = LearnableParameter() :  -> [512 x 1]
Validating --> clonedmodel.HL1.z = Plus (clonedmodel.HL1.t, clonedmodel.HL1.b) : [512 x *7], [512 x 1] -> [512 x 1 x *7]
Validating --> clonedmodel.HL1.y = Sigmoid (clonedmodel.HL1.z) : [512 x 1 x *7] -> [512 x 1 x *7]
Validating --> clonedmodel.HL2.t = Times (clonedmodel.HL2.W, clonedmodel.HL1.y) : [512 x 512], [512 x 1 x *7] -> [512 x 1 x *7]
Validating --> clonedmodel.HL2.b = LearnableParameter() :  -> [512 x 1]
Validating --> clonedmodel.HL2.z = Plus (clonedmodel.HL2.t, clonedmodel.HL2.b) : [512 x 1 x *7], [512 x 1] -> [512 x 1 x *7]
Validating --> clonedmodel.HL2.y = Sigmoid (clonedmodel.HL2.z) : [512 x 1 x *7] -> [512 x 1 x *7]
Validating --> clonedmodel.HL3.t = Times (clonedmodel.HL3.W, clonedmodel.HL2.y) : [512 x 512], [512 x 1 x *7] -> [512 x 1 x *7]
Validating --> clonedmodel.HL3.b = LearnableParameter() :  -> [512 x 1]
Validating --> clonedmodel.HL3.z = Plus (clonedmodel.HL3.t, clonedmodel.HL3.b) : [512 x 1 x *7], [512 x 1] -> [512 x 1 x *7]
Validating --> clonedmodel.HL3.y = Sigmoid (clonedmodel.HL3.z) : [512 x 1 x *7] -> [512 x 1 x *7]
Validating --> clonedmodel.OL.t = Times (clonedmodel.OL.W, clonedmodel.HL3.y) : [132 x 512], [512 x 1 x *7] -> [132 x 1 x *7]
Validating --> clonedmodel.OL.b = LearnableParameter() :  -> [132 x 1]
Validating --> clonedmodel.OL.z = Plus (clonedmodel.OL.t, clonedmodel.OL.b) : [132 x 1 x *7], [132 x 1] -> [132 x 1 x *7]
Validating --> Err = ClassificationError (labels, clonedmodel.OL.z) : [132 x *7], [132 x 1 x *7] -> [1]
Validating --> clonedmodel.globalPrior = LearnableParameter() :  -> [132 x 1]
Validating --> clonedmodel.logPrior = Log (clonedmodel.globalPrior) : [132 x 1] -> [132 x 1]
Validating --> clonedmodel.scaledLogLikelihood = Minus (clonedmodel.OL.z, clonedmodel.logPrior) : [132 x 1 x *7], [132 x 1] -> [132 x 1 x *7]
Validating --> lattice = InputValue() :  -> [1 x latticeAxis]
Validating --> cr = LatticeSequenceWithSoftmax (labels, clonedmodel.OL.z, clonedmodel.scaledLogLikelihood, lattice) : [132 x *7], [132 x 1 x *7], [132 x 1 x *7], [1 x latticeAxis] -> [1]
Validating --> latticeAxis = DynamicAxis() :  -> [1 x 1 x latticeAxis]

Validating network. 15 nodes to process in pass 2.


Validating network, final pass.




Post-processing network complete.

Reading script file C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/glob_0000.scp ... 948 entries
HTKDeserializer: selected '948' utterances grouped into '3' chunks, average chunk size: 316.0 utterances, 84244.7 frames (for I/O: 316.0 utterances, 84244.7 frames)
HTKDeserializer: determined feature kind as '33'-dimensional 'USER' with frame shift 10.0 ms
Total (133) state names in state list 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/state.list'
MLFDeserializer: '948' utterances with '252734' frames
Reading lattice index file 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu\TestData/latticeIndex.txt' ...
LatticeDeserializer: '923' sequences
01/17/2018 07:46:47: 
Model has 31 nodes. Using GPU 0.

01/17/2018 07:46:47: Training criterion:   cr = LatticeSequenceWithSoftmax
01/17/2018 07:46:47: Evaluation criterion: Err = ClassificationError


Allocating matrices for forward and/or backward propagation.

Gradient Memory Aliasing: 6 are aliased.
	clonedmodel.HL2.t (gradient) reuses clonedmodel.HL2.z (gradient)
	clonedmodel.OL.t (gradient) reuses clonedmodel.OL.z (gradient)
	clonedmodel.HL3.t (gradient) reuses clonedmodel.HL3.z (gradient)

Memory Sharing: Out of 52 matrices, 29 are shared as 7, and 23 are not shared.

Here are the ones that share memory:
	{ clonedmodel.HL1.W : [512 x 363] (gradient)
	  clonedmodel.HL1.t : [512 x *7]
	  clonedmodel.HL1.y : [512 x 1 x *7] }
	{ clonedmodel.HL1.z : [512 x 1 x *7]
	  clonedmodel.HL2.W : [512 x 512] (gradient)
	  clonedmodel.HL2.t : [512 x 1 x *7]
	  clonedmodel.HL2.y : [512 x 1 x *7] }
	{ clonedmodel.HL3.W : [512 x 512] (gradient)
	  clonedmodel.OL.t : [132 x 1 x *7] (gradient)
	  clonedmodel.OL.z : [132 x 1 x *7] (gradient) }
	{ clonedmodel.HL1.t : [512 x *7] (gradient)
	  clonedmodel.HL1.y : [512 x 1 x *7] (gradient)
	  clonedmodel.HL3.y : [512 x 1 x *7] (gradient)
	  clonedmodel.OL.t : [132 x 1 x *7]
	  clonedmodel.scaledLogLikelihood : [132 x 1 x *7] }
	{ clonedmodel.HL1.z : [512 x 1 x *7] (gradient)
	  clonedmodel.HL2.t : [512 x 1 x *7] (gradient)
	  clonedmodel.HL2.z : [512 x 1 x *7] (gradient)
	  clonedmodel.HL3.t : [512 x 1 x *7] (gradient)
	  clonedmodel.HL3.z : [512 x 1 x *7]
	  clonedmodel.HL3.z : [512 x 1 x *7] (gradient)
	  clonedmodel.OL.z : [132 x 1 x *7] }
	{ clonedmodel.OL.W : [132 x 512] (gradient)
	  clonedmodel.scaledLogLikelihood : [132 x 1 x *7] (gradient) }
	{ clonedmodel.HL2.b : [512 x 1] (gradient)
	  clonedmodel.HL2.y : [512 x 1 x *7] (gradient)
	  clonedmodel.HL2.z : [512 x 1 x *7]
	  clonedmodel.HL3.t : [512 x 1 x *7]
	  clonedmodel.HL3.y : [512 x 1 x *7] }

Here are the ones that don't share memory:
	{latticeAxis : [1 x 1 x latticeAxis]}
	{clonedmodel.HL1.b : [512 x 1]}
	{features : [363 x *7]}
	{clonedmodel.HL2.b : [512 x 1]}
	{clonedmodel.HL2.W : [512 x 512]}
	{clonedmodel.HL1.W : [512 x 363]}
	{clonedmodel.HL3.b : [512 x 1]}
	{clonedmodel.OL.b : [132 x 1]}
	{clonedmodel.globalPrior : [132 x 1]}
	{clonedmodel.globalMean : [363 x 1]}
	{clonedmodel.OL.W : [132 x 512]}
	{clonedmodel.HL3.W : [512 x 512]}
	{labels : [132 x *7]}
	{clonedmodel.globalInvStd : [363 x 1]}
	{Err : [1]}
	{cr : [1]}
	{cr : [1] (gradient)}
	{clonedmodel.OL.b : [132 x 1] (gradient)}
	{clonedmodel.featNorm : [363 x *7]}
	{clonedmodel.HL3.b : [512 x 1] (gradient)}
	{clonedmodel.logPrior : [132 x 1]}
	{clonedmodel.HL1.b : [512 x 1] (gradient)}
	{lattice : [1 x latticeAxis]}


01/17/2018 07:46:47: Training 779396 parameters in 8 out of 8 parameter tensors and 21 nodes with gradient:

01/17/2018 07:46:47: 	Node 'clonedmodel.HL1.W' (LearnableParameter operation) : [512 x 363]
01/17/2018 07:46:47: 	Node 'clonedmodel.HL1.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:46:47: 	Node 'clonedmodel.HL2.W' (LearnableParameter operation) : [512 x 512]
01/17/2018 07:46:47: 	Node 'clonedmodel.HL2.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:46:47: 	Node 'clonedmodel.HL3.W' (LearnableParameter operation) : [512 x 512]
01/17/2018 07:46:47: 	Node 'clonedmodel.HL3.b' (LearnableParameter operation) : [512 x 1]
01/17/2018 07:46:47: 	Node 'clonedmodel.OL.W' (LearnableParameter operation) : [132 x 512]
01/17/2018 07:46:47: 	Node 'clonedmodel.OL.b' (LearnableParameter operation) : [132 x 1]

01/17/2018 07:46:47: No PreCompute nodes found, or all already computed. Skipping pre-computation step.

01/17/2018 07:46:47: Starting Epoch 1: learning rate per sample = 0.000002  effective momentum = 0.995898  momentum as time constant = 2432.7 samples

01/17/2018 07:46:47: Starting minibatch loop.
dengamma value 1.088162
dengamma value 1.099063
dengamma value 1.076985
dengamma value 1.105294
dengamma value 1.048568
dengamma value 1.090612
dengamma value 1.069034
dengamma value 1.073947
dengamma value 0.982890
dengamma value 1.084679
01/17/2018 07:47:15:  Epoch[ 1 of 3]-Minibatch[   1-  10, 0.12%]: cr = 0.07897921 * 3030; Err = 0.33465347 * 3030; time = 28.0302s; samplesPerSecond = 108.1
WARNING: The same matrix with dim [1, 318] has been transferred between different devices for 20 times.
dengamma value 1.021193
dengamma value 1.050248
dengamma value 1.069187
dengamma value 1.135461
dengamma value 1.064151
dengamma value 1.053861
dengamma value 1.026327
dengamma value 0.984069
dengamma value 0.946694
dengamma value 1.116696
01/17/2018 07:47:23:  Epoch[ 1 of 3]-Minibatch[  11-  20, 0.24%]: cr = 0.08889189 * 2720; Err = 0.32977941 * 2720; time = 8.3711s; samplesPerSecond = 324.9
dengamma value 1.073882
dengamma value 1.131444
dengamma value 1.044074
dengamma value 1.070135
dengamma value 1.067734
dengamma value 1.122826
dengamma value 1.011504
dengamma value 0.958591
dengamma value 1.043645
dengamma value 0.950058
01/17/2018 07:47:30:  Epoch[ 1 of 3]-Minibatch[  21-  30, 0.37%]: cr = 0.09200504 * 2460; Err = 0.33739837 * 2460; time = 6.6498s; samplesPerSecond = 369.9
dengamma value 1.043167
dengamma value 1.057475
dengamma value 1.038363
dengamma value 1.073142
dengamma value 1.062622
dengamma value 1.115008
dengamma value 1.046483
dengamma value 1.110468
dengamma value 1.097045
dengamma value 1.101847
01/17/2018 07:47:41:  Epoch[ 1 of 3]-Minibatch[  31-  40, 0.49%]: cr = 0.08341682 * 3390; Err = 0.27404130 * 3390; time = 11.0378s; samplesPerSecond = 307.1
dengamma value 1.177670
dengamma value 1.079468
dengamma value 1.132308
dengamma value 1.149830
dengamma value 0.942017
dengamma value 1.090515
dengamma value 1.054302
dengamma value 1.067779
dengamma value 1.024308
dengamma value 1.075889
01/17/2018 07:47:50:  Epoch[ 1 of 3]-Minibatch[  41-  50, 0.61%]: cr = 0.06614777 * 2630; Err = 0.32395437 * 2630; time = 9.4005s; samplesPerSecond = 279.8
dengamma value 1.193316
dengamma value 1.058662
dengamma value 1.052992
dengamma value 1.090599
dengamma value 1.053534
dengamma value 1.095491
dengamma value 1.008375
dengamma value 1.045868
dengamma value 1.114091
dengamma value 1.059904
01/17/2018 07:48:00:  Epoch[ 1 of 3]-Minibatch[  51-  60, 0.73%]: cr = 0.08419601 * 2640; Err = 0.29545455 * 2640; time = 9.3813s; samplesPerSecond = 281.4
dengamma value 1.066465
dengamma value 1.103502
dengamma value 1.026207
dengamma value 1.013503
dengamma value 0.995754
dengamma value 1.152052
dengamma value 1.134168
dengamma value 1.059217
dengamma value 1.089917
dengamma value 1.084517
01/17/2018 07:48:11:  Epoch[ 1 of 3]-Minibatch[  61-  70, 0.85%]: cr = 0.08632195 * 3260; Err = 0.30153374 * 3260; time = 11.7814s; samplesPerSecond = 276.7
dengamma value 1.098105
dengamma value 1.088308
dengamma value 1.042342
dengamma value 1.060278
dengamma value 1.122067
dengamma value 1.109807
dengamma value 1.052146
dengamma value 1.027300
dengamma value 1.075426
dengamma value 1.093709
01/17/2018 07:48:20:  Epoch[ 1 of 3]-Minibatch[  71-  80, 0.98%]: cr = 0.08061202 * 2890; Err = 0.27231834 * 2890; time = 8.7976s; samplesPerSecond = 328.5
dengamma value 1.046880
dengamma value 1.070358
dengamma value 0.975420
dengamma value 1.078204
dengamma value 1.118420
dengamma value 1.097231
dengamma value 1.156730
dengamma value 1.047947
dengamma value 1.013944
dengamma value 1.028788
01/17/2018 07:48:30:  Epoch[ 1 of 3]-Minibatch[  81-  90, 1.10%]: cr = 0.08124917 * 2940; Err = 0.35272109 * 2940; time = 9.8350s; samplesPerSecond = 298.9
dengamma value 1.111170
dengamma value 1.057980
dengamma value 1.192282
dengamma value 1.073288
dengamma value 1.052752
dengamma value 1.109035
dengamma value 1.082714
dengamma value 1.072941
dengamma value 1.045710
dengamma value 1.120336
01/17/2018 07:48:40:  Epoch[ 1 of 3]-Minibatch[  91- 100, 1.22%]: cr = 0.07739322 * 2650; Err = 0.26188679 * 2650; time = 10.0838s; samplesPerSecond = 262.8
dengamma value 1.032787
dengamma value 1.044487
dengamma value 1.094206
dengamma value 1.050308
dengamma value 1.048913
dengamma value 1.144960
dengamma value 1.060881
dengamma value 1.126477
dengamma value 1.077273
dengamma value 1.017511
01/17/2018 07:48:48:  Epoch[ 1 of 3]-Minibatch[ 101- 110, 1.34%]: cr = 0.08467678 * 2410; Err = 0.31120332 * 2410; time = 8.2229s; samplesPerSecond = 293.1
dengamma value 1.071809
dengamma value 1.085534
dengamma value 1.066411
dengamma value 1.056770
dengamma value 1.074482
dengamma value 1.142215
dengamma value 1.086223
dengamma value 1.164311
dengamma value 1.051531
dengamma value 1.019167
01/17/2018 07:48:58:  Epoch[ 1 of 3]-Minibatch[ 111- 120, 1.46%]: cr = 0.07326778 * 2700; Err = 0.30851852 * 2700; time = 9.7320s; samplesPerSecond = 277.4
dengamma value 1.006437
dengamma value 1.071077
dengamma value 1.072938
dengamma value 1.125999
dengamma value 1.098892
dengamma value 1.083984
dengamma value 1.111442
dengamma value 1.074013
dengamma value 1.111135
dengamma value 1.027445
01/17/2018 07:49:06:  Epoch[ 1 of 3]-Minibatch[ 121- 130, 1.59%]: cr = 0.08498812 * 2380; Err = 0.31302521 * 2380; time = 7.7494s; samplesPerSecond = 307.1
dengamma value 1.065690
dengamma value 0.950080
dengamma value 1.082753
dengamma value 0.963398
dengamma value 1.079685
dengamma value 1.061164
dengamma value 1.084234
dengamma value 1.117680
dengamma value 1.065406
dengamma value 1.039439
01/17/2018 07:49:16:  Epoch[ 1 of 3]-Minibatch[ 131- 140, 1.71%]: cr = 0.07837629 * 2630; Err = 0.32661597 * 2630; time = 10.3647s; samplesPerSecond = 253.7
dengamma value 1.020005
dengamma value 1.069214
dengamma value 1.084112
dengamma value 1.033537
dengamma value 1.104664
dengamma value 0.975214
dengamma value 1.084214
dengamma value 1.068475
dengamma value 1.172442
dengamma value 1.081295
01/17/2018 07:49:28:  Epoch[ 1 of 3]-Minibatch[ 141- 150, 1.83%]: cr = 0.07798214 * 3100; Err = 0.28548387 * 3100; time = 11.8829s; samplesPerSecond = 260.9
dengamma value 1.085838
dengamma value 0.928113
dengamma value 1.062228
dengamma value 1.024939
dengamma value 1.061970
dengamma value 1.080009
dengamma value 0.977752
dengamma value 0.987936
dengamma value 1.038498
dengamma value 1.076137
01/17/2018 07:49:36:  Epoch[ 1 of 3]-Minibatch[ 151- 160, 1.95%]: cr = 0.07603625 * 2720; Err = 0.36801471 * 2720; time = 8.1842s; samplesPerSecond = 332.3
dengamma value 1.173796
dengamma value 1.030191
dengamma value 1.063923
dengamma value 1.078241
dengamma value 1.055084
dengamma value 1.034687
dengamma value 1.070767
dengamma value 1.059286
dengamma value 1.098097
dengamma value 1.010352
01/17/2018 07:49:45:  Epoch[ 1 of 3]-Minibatch[ 161- 170, 2.08%]: cr = 0.08418319 * 3000; Err = 0.30133333 * 3000; time = 8.9885s; samplesPerSecond = 333.8
dengamma value 1.258589
dengamma value 1.096639
dengamma value 1.096653
dengamma value 1.123063
dengamma value 1.041507
dengamma value 1.096070
dengamma value 1.123468
dengamma value 1.071991
dengamma value 1.045093
dengamma value 1.026546
01/17/2018 07:49:57:  Epoch[ 1 of 3]-Minibatch[ 171- 180, 2.20%]: cr = 0.07336679 * 3370; Err = 0.26172107 * 3370; time = 11.8740s; samplesPerSecond = 283.8
dengamma value 1.077624
dengamma value 1.101782
dengamma value 1.094899
dengamma value 0.953257
dengamma value 1.100462
dengamma value 0.998108
dengamma value 1.106404
dengamma value 0.997993
dengamma value 1.110075
dengamma value 1.089808
01/17/2018 07:50:07:  Epoch[ 1 of 3]-Minibatch[ 181- 190, 2.32%]: cr = 0.08136756 * 2600; Err = 0.34461538 * 2600; time = 10.0379s; samplesPerSecond = 259.0
dengamma value 1.099603
dengamma value 1.124871
dengamma value 1.071311
dengamma value 1.034133
dengamma value 1.085000
dengamma value 1.024212
dengamma value 1.144245
dengamma value 0.978119
dengamma value 1.139764
dengamma value 1.011179
01/17/2018 07:50:16:  Epoch[ 1 of 3]-Minibatch[ 191- 200, 2.44%]: cr = 0.08072397 * 2600; Err = 0.30423077 * 2600; time = 8.9323s; samplesPerSecond = 291.1
dengamma value 1.088622
dengamma value 1.055465
dengamma value 0.983475
dengamma value 1.078328
dengamma value 1.093984
dengamma value 1.015060
dengamma value 1.037329
dengamma value 1.123984
dengamma value 1.148328
dengamma value 1.136689
01/17/2018 07:50:25:  Epoch[ 1 of 3]-Minibatch[ 201- 210, 2.56%]: cr = 0.07264139 * 2300; Err = 0.32565217 * 2300; time = 8.6328s; samplesPerSecond = 266.4
dengamma value 1.000004
dengamma value 1.019190
dengamma value 1.000646
dengamma value 1.081452
dengamma value 1.034860
dengamma value 1.062810
dengamma value 1.176476
dengamma value 1.021679
dengamma value 1.079065
dengamma value 1.061600
01/17/2018 07:50:34:  Epoch[ 1 of 3]-Minibatch[ 211- 220, 2.69%]: cr = 0.08443621 * 2800; Err = 0.33357143 * 2800; time = 9.6663s; samplesPerSecond = 289.7
dengamma value 1.053473
dengamma value 1.096238
dengamma value 1.097280
dengamma value 1.077554
dengamma value 0.998775
dengamma value 1.005274
dengamma value 1.061292
dengamma value 1.062920
dengamma value 1.095745
dengamma value 1.017324
01/17/2018 07:50:43:  Epoch[ 1 of 3]-Minibatch[ 221- 230, 2.81%]: cr = 0.08792117 * 2590; Err = 0.31737452 * 2590; time = 8.8155s; samplesPerSecond = 293.8
dengamma value 0.990242
dengamma value 1.087052
dengamma value 1.095959
dengamma value 1.084776
dengamma value 1.059626
dengamma value 1.081880
dengamma value 0.995723
dengamma value 1.132265
dengamma value 1.104634
dengamma value 1.099872
01/17/2018 07:50:52:  Epoch[ 1 of 3]-Minibatch[ 231- 240, 2.93%]: cr = 0.07852966 * 2610; Err = 0.30804598 * 2610; time = 8.8473s; samplesPerSecond = 295.0
dengamma value 1.118808
dengamma value 1.084813
dengamma value 1.091818
dengamma value 1.105826
dengamma value 1.054520
dengamma value 1.116621
dengamma value 1.116467
dengamma value 1.036201
dengamma value 1.065415
dengamma value 1.023304
01/17/2018 07:51:00:  Epoch[ 1 of 3]-Minibatch[ 241- 250, 3.05%]: cr = 0.08120565 * 2400; Err = 0.29791667 * 2400; time = 8.1040s; samplesPerSecond = 296.2
dengamma value 1.100270
dengamma value 1.073113
dengamma value 1.071499
dengamma value 1.069996
dengamma value 1.072574
dengamma value 1.055203
dengamma value 1.157839
dengamma value 1.078078
dengamma value 1.054528
dengamma value 1.025716
01/17/2018 07:51:11:  Epoch[ 1 of 3]-Minibatch[ 251- 260, 3.17%]: cr = 0.08011032 * 3200; Err = 0.27812500 * 3200; time = 10.9712s; samplesPerSecond = 291.7
dengamma value 1.091993
dengamma value 1.053215
dengamma value 1.126736
dengamma value 1.073831
dengamma value 1.115763
dengamma value 1.038353
dengamma value 1.078538
dengamma value 1.046420
dengamma value 1.112949
dengamma value 1.009050
01/17/2018 07:51:26:  Epoch[ 1 of 3]-Minibatch[ 261- 270, 3.30%]: cr = 0.08564193 * 3570; Err = 0.29915966 * 3570; time = 14.5334s; samplesPerSecond = 245.6
dengamma value 1.024285
dengamma value 1.106762
dengamma value 1.104926
dengamma value 1.041111
dengamma value 1.074277
dengamma value 1.008149
dengamma value 1.047891
dengamma value 1.120440
dengamma value 0.972263
dengamma value 1.070577
01/17/2018 07:51:34:  Epoch[ 1 of 3]-Minibatch[ 271- 280, 3.42%]: cr = 0.09720201 * 2510; Err = 0.34541833 * 2510; time = 8.2149s; samplesPerSecond = 305.5
dengamma value 1.033531
dengamma value 1.097313
dengamma value 0.935394
dengamma value 1.080482
dengamma value 1.085017
dengamma value 1.070554
dengamma value 1.099456
dengamma value 1.049529
dengamma value 1.062960
dengamma value 1.078924
01/17/2018 07:51:46:  Epoch[ 1 of 3]-Minibatch[ 281- 290, 3.54%]: cr = 0.07616010 * 3400; Err = 0.35676471 * 3400; time = 11.7361s; samplesPerSecond = 289.7
dengamma value 1.132071
dengamma value 1.064320
dengamma value 1.024503
01/17/2018 07:51:47: Finished Epoch[ 1 of 3]: [Training] cr = 0.08119482 * 82104; Err = 0.31227468 * 82104; totalSamplesSeen = 82104; learningRatePerSample = 2e-06; epochTime=300.674s
01/17/2018 07:51:48: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.sequence.1'

01/17/2018 07:51:48: Starting Epoch 2: learning rate per sample = 0.000002  effective momentum = 0.995898  momentum as time constant = 2432.7 samples

01/17/2018 07:51:48: Starting minibatch loop.
dengamma value 1.089592
dengamma value 1.071522
dengamma value 1.047748
dengamma value 0.951023
dengamma value 1.139378
dengamma value 1.061708
dengamma value 1.071766
dengamma value 1.069808
dengamma value 1.060344
dengamma value 1.039663
01/17/2018 07:51:57:  Epoch[ 2 of 3]-Minibatch[   1-  10, 0.12%]: cr = 0.08456133 * 2880; Err = 0.28958333 * 2880; time = 9.3453s; samplesPerSecond = 308.2
dengamma value 1.054078
dengamma value 1.120433
dengamma value 1.049315
dengamma value 1.027127
dengamma value 1.074751
dengamma value 1.072896
dengamma value 1.056115
dengamma value 1.067861
dengamma value 1.071842
dengamma value 1.105685
01/17/2018 07:52:06:  Epoch[ 2 of 3]-Minibatch[  11-  20, 0.24%]: cr = 0.07955450 * 2560; Err = 0.30000000 * 2560; time = 8.5423s; samplesPerSecond = 299.7
dengamma value 1.055936
dengamma value 1.063835
dengamma value 1.054208
dengamma value 1.185662
dengamma value 1.029854
dengamma value 1.059279
dengamma value 1.033108
dengamma value 1.033475
dengamma value 1.099049
dengamma value 1.041091
01/17/2018 07:52:13:  Epoch[ 2 of 3]-Minibatch[  21-  30, 0.37%]: cr = 0.08459112 * 2240; Err = 0.32098214 * 2240; time = 7.3920s; samplesPerSecond = 303.0
dengamma value 1.054394
dengamma value 1.008045
dengamma value 1.086549
dengamma value 1.094490
dengamma value 1.023289
dengamma value 1.010895
dengamma value 1.029527
dengamma value 1.101016
dengamma value 1.025362
dengamma value 1.043724
01/17/2018 07:52:24:  Epoch[ 2 of 3]-Minibatch[  31-  40, 0.49%]: cr = 0.08990634 * 2940; Err = 0.30102041 * 2940; time = 10.4196s; samplesPerSecond = 282.2
dengamma value 1.081736
dengamma value 1.039297
dengamma value 1.080378
dengamma value 1.061708
dengamma value 1.079299
dengamma value 1.059922
dengamma value 1.071819
dengamma value 1.060066
dengamma value 1.038607
dengamma value 1.070880
01/17/2018 07:52:31:  Epoch[ 2 of 3]-Minibatch[  41-  50, 0.61%]: cr = 0.08489515 * 2570; Err = 0.30972763 * 2570; time = 7.2284s; samplesPerSecond = 355.5
dengamma value 1.007313
dengamma value 1.033387
dengamma value 1.085385
dengamma value 1.056913
dengamma value 1.083560
dengamma value 1.055549
dengamma value 1.004369
dengamma value 1.054481
dengamma value 1.031099
dengamma value 1.030563
01/17/2018 07:52:39:  Epoch[ 2 of 3]-Minibatch[  51-  60, 0.73%]: cr = 0.08489358 * 2800; Err = 0.29714286 * 2800; time = 8.5154s; samplesPerSecond = 328.8
dengamma value 0.998977
dengamma value 1.040903
dengamma value 1.023611
dengamma value 1.079356
dengamma value 1.004475
dengamma value 1.106471
dengamma value 1.128617
dengamma value 1.056243
dengamma value 1.148265
dengamma value 1.135660
01/17/2018 07:52:49:  Epoch[ 2 of 3]-Minibatch[  61-  70, 0.85%]: cr = 0.07813206 * 2540; Err = 0.26929134 * 2540; time = 9.3405s; samplesPerSecond = 271.9
dengamma value 1.051337
dengamma value 1.078692
dengamma value 1.091774
dengamma value 1.003407
dengamma value 0.989052
dengamma value 1.073882
dengamma value 1.023858
dengamma value 0.970382
dengamma value 1.021191
dengamma value 1.083243
01/17/2018 07:52:55:  Epoch[ 2 of 3]-Minibatch[  71-  80, 0.98%]: cr = 0.08305709 * 2180; Err = 0.36559633 * 2180; time = 6.8835s; samplesPerSecond = 316.7
dengamma value 1.033514
dengamma value 1.028519
dengamma value 0.979887
dengamma value 0.986568
dengamma value 1.118374
dengamma value 1.131315
dengamma value 1.058416
dengamma value 1.119792
dengamma value 1.065574
dengamma value 1.022660
01/17/2018 07:53:06:  Epoch[ 2 of 3]-Minibatch[  81-  90, 1.10%]: cr = 0.09173899 * 3060; Err = 0.32549020 * 3060; time = 10.6962s; samplesPerSecond = 286.1
dengamma value 1.082570
dengamma value 1.100993
dengamma value 1.087743
dengamma value 1.040468
dengamma value 1.156533
dengamma value 1.067361
dengamma value 1.058972
dengamma value 1.016402
dengamma value 0.930398
dengamma value 1.002086
01/17/2018 07:53:17:  Epoch[ 2 of 3]-Minibatch[  91- 100, 1.22%]: cr = 0.08085186 * 3280; Err = 0.32530488 * 3280; time = 11.0800s; samplesPerSecond = 296.0
dengamma value 1.069950
dengamma value 1.001739
dengamma value 1.137290
dengamma value 1.043663
dengamma value 1.009386
dengamma value 1.025302
dengamma value 1.061445
dengamma value 1.066976
dengamma value 1.056015
dengamma value 1.032537
01/17/2018 07:53:28:  Epoch[ 2 of 3]-Minibatch[ 101- 110, 1.34%]: cr = 0.08939691 * 3270; Err = 0.30275229 * 3270; time = 11.0983s; samplesPerSecond = 294.6
dengamma value 1.089906
dengamma value 1.009390
dengamma value 1.042450
dengamma value 1.044572
dengamma value 1.116509
dengamma value 1.073706
dengamma value 1.034845
dengamma value 1.099196
dengamma value 1.109772
dengamma value 1.075807
01/17/2018 07:53:38:  Epoch[ 2 of 3]-Minibatch[ 111- 120, 1.46%]: cr = 0.08757038 * 2560; Err = 0.30585937 * 2560; time = 10.0199s; samplesPerSecond = 255.5
dengamma value 1.038832
dengamma value 1.074412
dengamma value 1.055542
dengamma value 1.065745
dengamma value 1.075044
dengamma value 1.002837
dengamma value 1.111893
dengamma value 1.052703
dengamma value 1.020736
dengamma value 1.066907
01/17/2018 07:53:48:  Epoch[ 2 of 3]-Minibatch[ 121- 130, 1.59%]: cr = 0.07915628 * 2820; Err = 0.31843972 * 2820; time = 10.0162s; samplesPerSecond = 281.5
dengamma value 1.043970
dengamma value 1.027607
dengamma value 1.048337
dengamma value 1.019155
dengamma value 1.146148
dengamma value 1.013977
dengamma value 1.026418
dengamma value 1.004055
dengamma value 1.051399
dengamma value 1.048705
01/17/2018 07:53:57:  Epoch[ 2 of 3]-Minibatch[ 131- 140, 1.71%]: cr = 0.09025990 * 2420; Err = 0.36157025 * 2420; time = 8.2466s; samplesPerSecond = 293.5
dengamma value 1.034646
dengamma value 1.046595
dengamma value 1.030899
dengamma value 1.049052
dengamma value 1.069788
dengamma value 1.063782
dengamma value 1.025183
dengamma value 1.074145
dengamma value 1.080278
dengamma value 1.018480
01/17/2018 07:54:04:  Epoch[ 2 of 3]-Minibatch[ 141- 150, 1.83%]: cr = 0.07961785 * 2040; Err = 0.34068627 * 2040; time = 6.8971s; samplesPerSecond = 295.8
dengamma value 1.042904
dengamma value 1.095062
dengamma value 1.048509
dengamma value 1.105005
dengamma value 1.074857
dengamma value 0.997485
dengamma value 1.111884
dengamma value 1.014344
dengamma value 0.950531
dengamma value 1.106166
01/17/2018 07:54:14:  Epoch[ 2 of 3]-Minibatch[ 151- 160, 1.95%]: cr = 0.08570631 * 3130; Err = 0.33897764 * 3130; time = 10.4303s; samplesPerSecond = 300.1
dengamma value 1.068641
dengamma value 1.125209
dengamma value 1.072442
dengamma value 1.064429
dengamma value 1.011741
dengamma value 1.067586
dengamma value 1.030825
dengamma value 1.094834
dengamma value 0.985477
dengamma value 1.066920
01/17/2018 07:54:24:  Epoch[ 2 of 3]-Minibatch[ 161- 170, 2.08%]: cr = 0.08738619 * 2600; Err = 0.32192308 * 2600; time = 9.7272s; samplesPerSecond = 267.3
dengamma value 1.054477
dengamma value 1.008154
dengamma value 1.051172
dengamma value 0.989898
dengamma value 1.034221
dengamma value 1.029525
dengamma value 0.972871
dengamma value 1.028080
dengamma value 1.088833
dengamma value 1.032243
01/17/2018 07:54:31:  Epoch[ 2 of 3]-Minibatch[ 171- 180, 2.20%]: cr = 0.09439061 * 2340; Err = 0.35512821 * 2340; time = 7.6822s; samplesPerSecond = 304.6
dengamma value 1.097009
dengamma value 1.088588
dengamma value 1.073115
dengamma value 1.099358
dengamma value 1.066774
dengamma value 1.019927
dengamma value 1.025384
dengamma value 1.045008
dengamma value 1.087720
dengamma value 0.875389
01/17/2018 07:54:39:  Epoch[ 2 of 3]-Minibatch[ 181- 190, 2.32%]: cr = 0.08592261 * 2590; Err = 0.34131274 * 2590; time = 8.1087s; samplesPerSecond = 319.4
dengamma value 1.153874
dengamma value 1.117298
dengamma value 0.981867
dengamma value 1.040952
dengamma value 1.019762
dengamma value 1.022301
dengamma value 1.034343
dengamma value 1.068402
dengamma value 1.062755
dengamma value 1.094871
01/17/2018 07:54:49:  Epoch[ 2 of 3]-Minibatch[ 191- 200, 2.44%]: cr = 0.09035108 * 2640; Err = 0.30530303 * 2640; time = 9.3007s; samplesPerSecond = 283.9
dengamma value 1.103721
dengamma value 1.113866
dengamma value 1.017004
dengamma value 1.080550
dengamma value 1.020711
dengamma value 1.131239
dengamma value 1.034781
dengamma value 1.136901
dengamma value 1.090845
dengamma value 1.071719
01/17/2018 07:54:59:  Epoch[ 2 of 3]-Minibatch[ 201- 210, 2.56%]: cr = 0.08850184 * 2840; Err = 0.28028169 * 2840; time = 9.9446s; samplesPerSecond = 285.6
dengamma value 1.035228
dengamma value 1.007529
dengamma value 1.026397
dengamma value 1.072464
dengamma value 1.024581
dengamma value 1.081444
dengamma value 1.103346
dengamma value 0.947306
dengamma value 1.024929
dengamma value 1.067569
01/17/2018 07:55:07:  Epoch[ 2 of 3]-Minibatch[ 211- 220, 2.69%]: cr = 0.08554747 * 2450; Err = 0.35714286 * 2450; time = 8.6154s; samplesPerSecond = 284.4
dengamma value 1.108120
dengamma value 1.044957
dengamma value 1.076162
dengamma value 1.039787
dengamma value 1.109577
dengamma value 1.107742
dengamma value 1.082074
dengamma value 1.047916
dengamma value 1.087079
dengamma value 1.024487
01/17/2018 07:55:18:  Epoch[ 2 of 3]-Minibatch[ 221- 230, 2.81%]: cr = 0.08939539 * 3180; Err = 0.29150943 * 3180; time = 10.6676s; samplesPerSecond = 298.1
dengamma value 0.961618
dengamma value 1.127344
dengamma value 1.038201
dengamma value 1.045186
dengamma value 1.012020
dengamma value 1.090617
dengamma value 1.016836
dengamma value 1.079375
dengamma value 0.993773
dengamma value 1.077695
01/17/2018 07:55:30:  Epoch[ 2 of 3]-Minibatch[ 231- 240, 2.93%]: cr = 0.08591958 * 2970; Err = 0.31683502 * 2970; time = 11.5248s; samplesPerSecond = 257.7
dengamma value 1.068360
dengamma value 1.046687
dengamma value 1.085118
dengamma value 1.102118
dengamma value 1.050270
dengamma value 1.032887
dengamma value 1.051441
dengamma value 1.020128
dengamma value 1.014069
dengamma value 1.080206
01/17/2018 07:55:39:  Epoch[ 2 of 3]-Minibatch[ 241- 250, 3.05%]: cr = 0.08668562 * 2830; Err = 0.28374558 * 2830; time = 9.3486s; samplesPerSecond = 302.7
dengamma value 1.012741
dengamma value 1.093455
dengamma value 1.101999
dengamma value 1.035784
dengamma value 1.125205
dengamma value 0.966032
dengamma value 1.046939
dengamma value 1.029520
dengamma value 1.055295
dengamma value 1.117220
01/17/2018 07:55:46:  Epoch[ 2 of 3]-Minibatch[ 251- 260, 3.17%]: cr = 0.08277494 * 2600; Err = 0.32615385 * 2600; time = 7.5229s; samplesPerSecond = 345.6
dengamma value 1.136848
dengamma value 1.060557
dengamma value 1.111701
dengamma value 1.080845
dengamma value 1.082659
dengamma value 1.106286
dengamma value 1.076931
dengamma value 1.101649
dengamma value 1.144268
dengamma value 1.119844
01/17/2018 07:55:57:  Epoch[ 2 of 3]-Minibatch[ 261- 270, 3.30%]: cr = 0.07698080 * 2770; Err = 0.25920578 * 2770; time = 10.7094s; samplesPerSecond = 258.7
dengamma value 1.058593
dengamma value 1.041429
dengamma value 1.108321
dengamma value 1.033942
dengamma value 1.127057
dengamma value 1.093599
dengamma value 1.183053
dengamma value 1.041834
dengamma value 1.119023
dengamma value 1.129554
01/17/2018 07:56:07:  Epoch[ 2 of 3]-Minibatch[ 271- 280, 3.42%]: cr = 0.07993742 * 2450; Err = 0.29795918 * 2450; time = 9.4921s; samplesPerSecond = 258.1
dengamma value 0.967247
dengamma value 1.012685
dengamma value 1.086114
dengamma value 1.110066
dengamma value 1.069696
dengamma value 1.051565
dengamma value 1.081369
dengamma value 1.060701
dengamma value 1.085383
dengamma value 1.117179
01/17/2018 07:56:16:  Epoch[ 2 of 3]-Minibatch[ 281- 290, 3.54%]: cr = 0.08266637 * 2730; Err = 0.30989011 * 2730; time = 9.0889s; samplesPerSecond = 300.4
dengamma value 1.093613
dengamma value 1.006557
dengamma value 1.027561
dengamma value 0.974602
dengamma value 1.026756
dengamma value 1.006296
dengamma value 1.015235
dengamma value 1.035612
dengamma value 1.074972
dengamma value 1.010957
01/17/2018 07:56:24:  Epoch[ 2 of 3]-Minibatch[ 291- 300, 3.66%]: cr = 0.08953097 * 2440; Err = 0.33278689 * 2440; time = 8.5057s; samplesPerSecond = 286.9
dengamma value 1.066934
dengamma value 1.047767
dengamma value 0.989474
dengamma value 1.061377
01/17/2018 07:56:28: Finished Epoch[ 2 of 3]: [Training] cr = 0.08540281 * 81852; Err = 0.31425011 * 81852; totalSamplesSeen = 163956; learningRatePerSample = 2e-06; epochTime=279.953s
01/17/2018 07:56:28: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.sequence.2'

01/17/2018 07:56:28: Starting Epoch 3: learning rate per sample = 0.000002  effective momentum = 0.995898  momentum as time constant = 2432.7 samples

01/17/2018 07:56:28: Starting minibatch loop.
dengamma value 1.113250
dengamma value 1.036850
dengamma value 0.925648
dengamma value 1.049095
dengamma value 1.047733
dengamma value 1.056598
dengamma value 1.081640
dengamma value 1.079751
dengamma value 1.058132
dengamma value 0.991991
01/17/2018 07:56:37:  Epoch[ 3 of 3]-Minibatch[   1-  10, 0.12%]: cr = 0.09068287 * 2810; Err = 0.30391459 * 2810; time = 8.6929s; samplesPerSecond = 323.3
dengamma value 1.019815
dengamma value 1.150375
dengamma value 1.084905
dengamma value 1.090945
dengamma value 1.104981
dengamma value 1.036276
dengamma value 1.182616
dengamma value 1.134491
dengamma value 1.083537
dengamma value 1.029840
01/17/2018 07:56:44:  Epoch[ 3 of 3]-Minibatch[  11-  20, 0.24%]: cr = 0.07739952 * 1970; Err = 0.30253807 * 1970; time = 6.7490s; samplesPerSecond = 291.9
dengamma value 1.099936
dengamma value 1.054920
dengamma value 1.050435
dengamma value 1.029694
dengamma value 1.101534
dengamma value 1.078308
dengamma value 1.046266
dengamma value 1.059705
dengamma value 1.014387
dengamma value 1.075327
01/17/2018 07:56:53:  Epoch[ 3 of 3]-Minibatch[  21-  30, 0.37%]: cr = 0.08397018 * 3050; Err = 0.30655738 * 3050; time = 9.5771s; samplesPerSecond = 318.5
dengamma value 1.000495
dengamma value 1.055702
dengamma value 1.062845
dengamma value 1.011423
dengamma value 1.152833
dengamma value 1.001263
dengamma value 1.012673
dengamma value 1.077861
dengamma value 1.083668
dengamma value 1.019250
01/17/2018 07:57:01:  Epoch[ 3 of 3]-Minibatch[  31-  40, 0.49%]: cr = 0.09055480 * 2230; Err = 0.32914798 * 2230; time = 7.8076s; samplesPerSecond = 285.6
dengamma value 1.022872
dengamma value 1.077874
dengamma value 1.030921
dengamma value 1.026867
dengamma value 1.106452
dengamma value 1.068285
dengamma value 1.049740
dengamma value 1.033626
dengamma value 1.001361
dengamma value 1.050103
01/17/2018 07:57:09:  Epoch[ 3 of 3]-Minibatch[  41-  50, 0.61%]: cr = 0.08865369 * 2350; Err = 0.32808511 * 2350; time = 8.1264s; samplesPerSecond = 289.2
dengamma value 1.096275
dengamma value 0.988145
dengamma value 1.070870
dengamma value 1.065532
dengamma value 1.052047
dengamma value 1.082383
dengamma value 1.075891
dengamma value 1.113368
dengamma value 1.041521
dengamma value 0.985916
01/17/2018 07:57:19:  Epoch[ 3 of 3]-Minibatch[  51-  60, 0.73%]: cr = 0.08636560 * 2850; Err = 0.32070175 * 2850; time = 9.6685s; samplesPerSecond = 294.8
dengamma value 1.005646
dengamma value 1.089372
dengamma value 1.066648
dengamma value 1.126878
dengamma value 1.057856
dengamma value 1.109081
dengamma value 1.111933
dengamma value 1.117950
dengamma value 1.029394
dengamma value 1.137359
01/17/2018 07:57:31:  Epoch[ 3 of 3]-Minibatch[  61-  70, 0.85%]: cr = 0.08056026 * 3020; Err = 0.28013245 * 3020; time = 12.2784s; samplesPerSecond = 246.0
dengamma value 1.088435
dengamma value 0.980674
dengamma value 1.077882
dengamma value 1.089429
dengamma value 0.943854
dengamma value 1.024757
dengamma value 1.043540
dengamma value 1.012622
dengamma value 1.055796
dengamma value 1.013740
01/17/2018 07:57:38:  Epoch[ 3 of 3]-Minibatch[  71-  80, 0.98%]: cr = 0.09537925 * 2220; Err = 0.30405405 * 2220; time = 6.8271s; samplesPerSecond = 325.2
dengamma value 1.054937
dengamma value 1.036628
dengamma value 1.133153
dengamma value 1.166993
dengamma value 1.123539
dengamma value 1.024998
dengamma value 1.063997
dengamma value 1.131925
dengamma value 1.053178
dengamma value 1.056383
01/17/2018 07:57:51:  Epoch[ 3 of 3]-Minibatch[  81-  90, 1.10%]: cr = 0.08650036 * 3110; Err = 0.26141479 * 3110; time = 13.3308s; samplesPerSecond = 233.3
dengamma value 1.020255
dengamma value 0.991719
dengamma value 1.009127
dengamma value 1.027564
dengamma value 1.048917
dengamma value 1.087359
dengamma value 0.999799
dengamma value 1.049327
dengamma value 1.139976
dengamma value 1.026929
01/17/2018 07:58:00:  Epoch[ 3 of 3]-Minibatch[  91- 100, 1.22%]: cr = 0.07721653 * 2560; Err = 0.34648438 * 2560; time = 8.9392s; samplesPerSecond = 286.4
dengamma value 1.029701
dengamma value 1.090810
dengamma value 1.150994
dengamma value 1.125394
dengamma value 1.089833
dengamma value 1.043864
dengamma value 1.074731
dengamma value 1.050473
dengamma value 1.086991
dengamma value 1.017834
01/17/2018 07:58:09:  Epoch[ 3 of 3]-Minibatch[ 101- 110, 1.34%]: cr = 0.08166337 * 2780; Err = 0.31258993 * 2780; time = 8.9109s; samplesPerSecond = 312.0
dengamma value 1.016859
dengamma value 1.120440
dengamma value 1.111964
dengamma value 0.964502
dengamma value 0.966233
dengamma value 1.065404
dengamma value 1.054392
dengamma value 1.030770
dengamma value 1.092313
dengamma value 1.045006
01/17/2018 07:58:19:  Epoch[ 3 of 3]-Minibatch[ 111- 120, 1.46%]: cr = 0.09112190 * 2520; Err = 0.30952381 * 2520; time = 9.6088s; samplesPerSecond = 262.3
dengamma value 1.048139
dengamma value 1.023021
dengamma value 0.968487
dengamma value 1.096123
dengamma value 1.088716
dengamma value 1.011498
dengamma value 1.056981
dengamma value 1.118636
dengamma value 1.107746
dengamma value 1.054997
01/17/2018 07:58:27:  Epoch[ 3 of 3]-Minibatch[ 121- 130, 1.59%]: cr = 0.08811045 * 2580; Err = 0.31085271 * 2580; time = 8.1778s; samplesPerSecond = 315.5
dengamma value 1.038586
dengamma value 1.075457
dengamma value 1.008455
dengamma value 1.095500
dengamma value 1.050784
dengamma value 1.096881
dengamma value 1.112144
dengamma value 0.992508
dengamma value 1.080997
dengamma value 0.977222
01/17/2018 07:58:34:  Epoch[ 3 of 3]-Minibatch[ 131- 140, 1.71%]: cr = 0.08031031 * 2450; Err = 0.32979592 * 2450; time = 6.8910s; samplesPerSecond = 355.5
dengamma value 0.966951
dengamma value 1.040992
dengamma value 1.031903
dengamma value 1.128591
dengamma value 1.026446
dengamma value 1.047809
dengamma value 1.116543
dengamma value 0.983602
dengamma value 1.054840
dengamma value 1.076809
01/17/2018 07:58:40:  Epoch[ 3 of 3]-Minibatch[ 141- 150, 1.83%]: cr = 0.08485678 * 2290; Err = 0.31572052 * 2290; time = 6.4373s; samplesPerSecond = 355.7
dengamma value 1.254005
dengamma value 1.041113
dengamma value 0.992911
dengamma value 1.079803
dengamma value 1.101725
dengamma value 1.056816
dengamma value 1.005603
dengamma value 1.072814
dengamma value 1.108822
dengamma value 1.052975
01/17/2018 07:58:50:  Epoch[ 3 of 3]-Minibatch[ 151- 160, 1.95%]: cr = 0.08193872 * 3000; Err = 0.28300000 * 3000; time = 9.6231s; samplesPerSecond = 311.7
dengamma value 1.248163
dengamma value 1.089307
dengamma value 1.067499
dengamma value 1.082298
dengamma value 0.934997
dengamma value 1.048273
dengamma value 1.075521
dengamma value 1.062501
dengamma value 0.997275
dengamma value 1.091340
01/17/2018 07:58:57:  Epoch[ 3 of 3]-Minibatch[ 161- 170, 2.08%]: cr = 0.07725397 * 2510; Err = 0.30756972 * 2510; time = 7.0534s; samplesPerSecond = 355.9
dengamma value 1.009265
dengamma value 1.078812
dengamma value 1.009651
dengamma value 1.057396
dengamma value 1.077734
dengamma value 1.070837
dengamma value 1.029384
dengamma value 1.136204
dengamma value 1.118148
dengamma value 1.071572
01/17/2018 07:59:06:  Epoch[ 3 of 3]-Minibatch[ 171- 180, 2.20%]: cr = 0.08581113 * 2610; Err = 0.28237548 * 2610; time = 9.4000s; samplesPerSecond = 277.7
dengamma value 1.102270
dengamma value 1.086899
dengamma value 1.103049
dengamma value 1.087528
dengamma value 0.991491
dengamma value 1.098629
dengamma value 1.028960
dengamma value 1.065721
dengamma value 1.071475
dengamma value 1.061495
01/17/2018 07:59:15:  Epoch[ 3 of 3]-Minibatch[ 181- 190, 2.32%]: cr = 0.08572164 * 2400; Err = 0.29750000 * 2400; time = 9.1882s; samplesPerSecond = 261.2
dengamma value 1.131669
dengamma value 1.093109
dengamma value 1.076080
dengamma value 1.063342
dengamma value 1.057006
dengamma value 1.025392
dengamma value 1.068316
dengamma value 1.079021
dengamma value 1.029558
dengamma value 1.065052
01/17/2018 07:59:25:  Epoch[ 3 of 3]-Minibatch[ 191- 200, 2.44%]: cr = 0.08563793 * 2590; Err = 0.25675676 * 2590; time = 9.6157s; samplesPerSecond = 269.4
dengamma value 1.130486
dengamma value 1.028508
dengamma value 1.077203
dengamma value 1.015772
dengamma value 1.084474
dengamma value 1.060658
dengamma value 1.088339
dengamma value 1.054691
dengamma value 0.994150
dengamma value 1.064132
01/17/2018 07:59:34:  Epoch[ 3 of 3]-Minibatch[ 201- 210, 2.56%]: cr = 0.07649859 * 2460; Err = 0.36504065 * 2460; time = 9.2205s; samplesPerSecond = 266.8
dengamma value 1.052215
dengamma value 1.124490
dengamma value 1.021019
dengamma value 1.084437
dengamma value 1.103063
dengamma value 1.008204
dengamma value 1.038743
dengamma value 1.073510
dengamma value 1.005139
dengamma value 1.038662
01/17/2018 07:59:42:  Epoch[ 3 of 3]-Minibatch[ 211- 220, 2.69%]: cr = 0.08075216 * 2810; Err = 0.30071174 * 2810; time = 8.1710s; samplesPerSecond = 343.9
dengamma value 1.042287
dengamma value 1.035584
dengamma value 1.040909
dengamma value 1.057661
dengamma value 1.078672
dengamma value 1.144815
dengamma value 1.157748
dengamma value 1.019168
dengamma value 1.013620
dengamma value 0.998956
01/17/2018 07:59:50:  Epoch[ 3 of 3]-Minibatch[ 221- 230, 2.81%]: cr = 0.08851122 * 2550; Err = 0.30745098 * 2550; time = 7.7181s; samplesPerSecond = 330.4
dengamma value 1.077362
dengamma value 1.045708
dengamma value 1.042652
dengamma value 1.043028
dengamma value 1.076836
dengamma value 1.083680
dengamma value 0.978354
dengamma value 1.002718
dengamma value 1.066423
dengamma value 1.060026
01/17/2018 08:00:00:  Epoch[ 3 of 3]-Minibatch[ 231- 240, 2.93%]: cr = 0.08959579 * 2810; Err = 0.29750890 * 2810; time = 10.2166s; samplesPerSecond = 275.0
dengamma value 1.166803
dengamma value 0.946450
dengamma value 1.030671
dengamma value 1.112627
dengamma value 1.008708
dengamma value 0.996962
dengamma value 0.971501
dengamma value 1.105676
dengamma value 1.031433
dengamma value 1.018861
01/17/2018 08:00:09:  Epoch[ 3 of 3]-Minibatch[ 241- 250, 3.05%]: cr = 0.09214059 * 2540; Err = 0.32125984 * 2540; time = 8.4047s; samplesPerSecond = 302.2
dengamma value 1.085565
dengamma value 1.055374
dengamma value 1.024862
dengamma value 1.041799
dengamma value 1.073595
dengamma value 1.057732
dengamma value 1.046133
dengamma value 1.058635
dengamma value 1.067138
dengamma value 1.015141
01/17/2018 08:00:19:  Epoch[ 3 of 3]-Minibatch[ 251- 260, 3.17%]: cr = 0.09365374 * 2790; Err = 0.29390681 * 2790; time = 10.0553s; samplesPerSecond = 277.5
dengamma value 1.090500
dengamma value 1.078554
dengamma value 1.037848
dengamma value 1.130892
dengamma value 1.057512
dengamma value 1.080909
dengamma value 1.096125
dengamma value 1.109021
dengamma value 1.074953
dengamma value 1.055137
01/17/2018 08:00:36:  Epoch[ 3 of 3]-Minibatch[ 261- 270, 3.30%]: cr = 0.08347255 * 3920; Err = 0.26147959 * 3920; time = 17.4106s; samplesPerSecond = 225.2
dengamma value 1.121153
dengamma value 1.003293
dengamma value 1.007246
dengamma value 1.059061
dengamma value 1.049562
dengamma value 1.051531
dengamma value 1.076518
dengamma value 1.027036
dengamma value 1.037740
dengamma value 1.040547
01/17/2018 08:00:48:  Epoch[ 3 of 3]-Minibatch[ 271- 280, 3.42%]: cr = 0.07784753 * 3370; Err = 0.34213650 * 3370; time = 11.6732s; samplesPerSecond = 288.7
dengamma value 1.040284
dengamma value 1.043802
dengamma value 1.028288
dengamma value 1.062837
dengamma value 1.083869
dengamma value 1.056014
dengamma value 1.014523
dengamma value 1.093659
dengamma value 1.069780
dengamma value 1.066848
01/17/2018 08:00:59:  Epoch[ 3 of 3]-Minibatch[ 281- 290, 3.54%]: cr = 0.09883846 * 2930; Err = 0.30170648 * 2930; time = 11.2075s; samplesPerSecond = 261.4
dengamma value 1.017902
dengamma value 1.025734
dengamma value 1.038976
dengamma value 1.066491
dengamma value 1.110772
dengamma value 1.139174
dengamma value 1.066084
dengamma value 1.021971
dengamma value 1.035387
dengamma value 1.036461
01/17/2018 08:01:08:  Epoch[ 3 of 3]-Minibatch[ 291- 300, 3.66%]: cr = 0.08027815 * 2590; Err = 0.28648649 * 2590; time = 8.7626s; samplesPerSecond = 295.6
dengamma value 1.028506
dengamma value 1.086706
dengamma value 1.030012
dengamma value 1.042108
dengamma value 1.055591
01/17/2018 08:01:13: Finished Epoch[ 3 of 3]: [Training] cr = 0.08507867 * 82070; Err = 0.30443524 * 82070; totalSamplesSeen = 246026; learningRatePerSample = 2e-06; epochTime=284.36s
01/17/2018 08:01:13: SGD: Saving checkpoint model 'C:\local\cygwin-2.8.2-x64\tmp\cntk-test-20180117072206.749857\Speech\DNN_SequenceTrainingNewReader@debug_gpu/models/cntkSpeech.sequence'

01/17/2018 08:01:13: Action "train" complete.

01/17/2018 08:01:13: __COMPLETED__
