CPU info:
    CPU Model Name: Intel(R) Xeon(R) CPU E5-2630 v2 @ 2.60GHz
    Hardware threads: 24
    Total Memory: 268381192 kB
-------------------------------------------------------------------
=== Running C:\Program Files\Microsoft MPI\Bin\/mpiexec.exe -n 2 C:\jenkins\workspace\CNTK-Test-Windows-W1\x64\release\cntk.exe configFile=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN/ParallelBM/cntk.cntk currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu DeviceId=0 timestamping=true numCPUThreads=12 precision=double speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]] stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
-------------------------------------------------------------------
Build info: 

		Built time: Aug 16 2016 03:09:16
		Last modified date: Fri Aug 12 05:28:23 2016
		Build type: Release
		Build target: GPU
		Math lib: mkl
		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5
		CUB_PATH: c:\src\cub-1.4.1
		CUDNN_PATH: c:\NVIDIA\cudnn-4.0\cuda
		Build Branch: HEAD
		Build SHA1: 026b1e772b963461e189f8f00aa7ed6951298f84
		Built by svcphil on Philly-Pool1
		Build Path: c:\jenkins\workspace\CNTK-Build-Windows\Source\CNTK\
-------------------------------------------------------------------
Changed current directory to C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPIWrapper: initializing MPI
-------------------------------------------------------------------
Build info: 

		Built time: Aug 16 2016 03:09:16
		Last modified date: Fri Aug 12 05:28:23 2016
		Build type: Release
		Build target: GPU
		Math lib: mkl
		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5
		CUB_PATH: c:\src\cub-1.4.1
		CUDNN_PATH: c:\NVIDIA\cudnn-4.0\cuda
		Build Branch: HEAD
		Build SHA1: 026b1e772b963461e189f8f00aa7ed6951298f84
		Built by svcphil on Philly-Pool1
		Build Path: c:\jenkins\workspace\CNTK-Build-Windows\Source\CNTK\
-------------------------------------------------------------------
Changed current directory to C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPIWrapper: initializing MPI
ping [requestnodes (before change)]: 2 nodes pinging each other
ping [requestnodes (before change)]: 2 nodes pinging each other
ping [requestnodes (before change)]: all 2 nodes responded
ping [requestnodes (before change)]: all 2 nodes responded
requestnodes [MPIWrapper]: using 2 out of 2 MPI nodes (2 requested); we (1) are in (participating)
requestnodes [MPIWrapper]: using 2 out of 2 MPI nodes (2 requested); we (0) are in (participating)
ping [requestnodes (after change)]: 2 nodes pinging each other
ping [requestnodes (after change)]: 2 nodes pinging each other
ping [requestnodes (after change)]: all 2 nodes responded
ping [requestnodes (after change)]: all 2 nodes responded
mpihelper: we are cog 1 in a gearbox of 2
mpihelper: we are cog 0 in a gearbox of 2
ping [mpihelper]: 2 nodes pinging each other
ping [mpihelper]: 2 nodes pinging each other
ping [mpihelper]: all 2 nodes responded
ping [mpihelper]: all 2 nodes responded
MPI Rank 0: 08/16/2016 03:19:59: Redirecting stderr to file C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr_speechTrain.logrank0
MPI Rank 0: 08/16/2016 03:19:59: -------------------------------------------------------------------
MPI Rank 0: 08/16/2016 03:19:59: Build info: 
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:19:59: 		Built time: Aug 16 2016 03:09:16
MPI Rank 0: 08/16/2016 03:19:59: 		Last modified date: Fri Aug 12 05:28:23 2016
MPI Rank 0: 08/16/2016 03:19:59: 		Build type: Release
MPI Rank 0: 08/16/2016 03:19:59: 		Build target: GPU
MPI Rank 0: 08/16/2016 03:19:59: 		Math lib: mkl
MPI Rank 0: 08/16/2016 03:19:59: 		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5
MPI Rank 0: 08/16/2016 03:19:59: 		CUB_PATH: c:\src\cub-1.4.1
MPI Rank 0: 08/16/2016 03:19:59: 		CUDNN_PATH: c:\NVIDIA\cudnn-4.0\cuda
MPI Rank 0: 08/16/2016 03:19:59: 		Build Branch: HEAD
MPI Rank 0: 08/16/2016 03:19:59: 		Build SHA1: 026b1e772b963461e189f8f00aa7ed6951298f84
MPI Rank 0: 08/16/2016 03:19:59: 		Built by svcphil on Philly-Pool1
MPI Rank 0: 08/16/2016 03:19:59: 		Build Path: c:\jenkins\workspace\CNTK-Build-Windows\Source\CNTK\
MPI Rank 0: 08/16/2016 03:19:59: -------------------------------------------------------------------
MPI Rank 0: 08/16/2016 03:20:02: -------------------------------------------------------------------
MPI Rank 0: 08/16/2016 03:20:02: GPU info:
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: 		Device[0]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 0: 08/16/2016 03:20:02: 		Device[1]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 0: 08/16/2016 03:20:02: 		Device[2]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 0: 08/16/2016 03:20:02: 		Device[3]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 0: 08/16/2016 03:20:02: -------------------------------------------------------------------
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: Running on DPHAIM-24 at 2016/08/16 03:20:02
MPI Rank 0: 08/16/2016 03:20:02: Command line: 
MPI Rank 0: C:\jenkins\workspace\CNTK-Test-Windows-W1\x64\release\cntk.exe  configFile=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN/ParallelBM/cntk.cntk  currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data  RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu  DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data  ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN  OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu  DeviceId=0  timestamping=true  numCPUThreads=12  precision=double  speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]  stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: >>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
MPI Rank 0: 08/16/2016 03:20:02: precision = "float"
MPI Rank 0: command = speechTrain
MPI Rank 0: deviceId = $DeviceId$
MPI Rank 0: parallelTrain = true
MPI Rank 0: speechTrain = [
MPI Rank 0:     action = "train"
MPI Rank 0:     modelPath = "$RunDir$/models/cntkSpeech.dnn"
MPI Rank 0:     deviceId = $DeviceId$
MPI Rank 0:     traceLevel = 1
MPI Rank 0:     SimpleNetworkBuilder = [
MPI Rank 0:         layerSizes = 363:512:512:132
MPI Rank 0:         trainingCriterion = "CrossEntropyWithSoftmax"
MPI Rank 0:         evalCriterion = "ClassificationError"
MPI Rank 0:         layerTypes = "Sigmoid"
MPI Rank 0:         initValueScale = 1.0
MPI Rank 0:         applyMeanVarNorm = true
MPI Rank 0:         uniformInit = true
MPI Rank 0:         needPrior = true
MPI Rank 0:     ]
MPI Rank 0:     ExperimentalNetworkBuilder = [    // the same as above but with BS. Not active; activate by commenting out the SimpleNetworkBuilder entry above
MPI Rank 0:         layerSizes = 363:512:512:132
MPI Rank 0:         trainingCriterion = 'CE'
MPI Rank 0:         evalCriterion = 'Err'
MPI Rank 0:         applyMeanVarNorm = true
MPI Rank 0:         L = Length(layerSizes)-1    // number of model layers
MPI Rank 0:         features = Input(layerSizes[0], 1, tag='feature') ; labels = Input(layerSizes[Length(layerSizes)-1], 1, tag='label')
MPI Rank 0:         featNorm = if applyMeanVarNorm
MPI Rank 0:                    then MeanVarNorm(features)
MPI Rank 0:                    else features
MPI Rank 0:         layers[layer:1..L-1] = if layer > 1
MPI Rank 0:                                then SBFF(layers[layer-1].Eh, layerSizes[layer], layerSizes[layer-1])
MPI Rank 0:                                else SBFF(featNorm, layerSizes[layer], layerSizes[layer-1])
MPI Rank 0:         outLayer = BFF(layers[L-1].Eh, layerSizes[L], layerSizes[L-1])
MPI Rank 0:         outZ = outLayer.z        // + PastValue(layerSizes[L], 1, outLayer.z)
MPI Rank 0:         CE = if trainingCriterion == 'CE'
MPI Rank 0:              then CrossEntropyWithSoftmax(labels, outZ, tag='criterion')
MPI Rank 0:              else Fail('unknown trainingCriterion ' + trainingCriterion)
MPI Rank 0:         Err = if evalCriterion == 'Err' then
MPI Rank 0:               ClassificationError(labels, outZ, tag='evaluation')
MPI Rank 0:               else Fail('unknown evalCriterion ' + evalCriterion)
MPI Rank 0:         logPrior = LogPrior(labels)
MPI Rank 0:         // TODO: how to add a tag to an infix operation?
MPI Rank 0:         ScaledLogLikelihood = Minus (outZ, logPrior, tag='output')
MPI Rank 0:     ]
MPI Rank 0:     SGD = [
MPI Rank 0:         epochSize = 20480
MPI Rank 0:         minibatchSize = 64:256:1024
MPI Rank 0:         learningRatesPerMB = 1.0:0.5:0.1
MPI Rank 0:         numMBsToShowResult = 3
MPI Rank 0:         momentumPerMB = 0.9:0.656119
MPI Rank 0:         dropoutRate = 0.0
MPI Rank 0:         maxEpochs = 5
MPI Rank 0:         keepCheckPointFiles = true
MPI Rank 0:         clippingThresholdPerSample = 1#INF
MPI Rank 0:         ParallelTrain = [
MPI Rank 0:             parallelizationMethod = "BlockMomentumSGD"
MPI Rank 0:             distributedMBReading = true
MPI Rank 0:             syncPerfStats=1
MPI Rank 0:             BlockMomentumSGD = [
MPI Rank 0:                 blockSizePerWorker=2048
MPI Rank 0:                 resetSGDMomentum=true
MPI Rank 0:                 useNesterovMomentum=true
MPI Rank 0:             ]
MPI Rank 0:         ]
MPI Rank 0:         AutoAdjust = [
MPI Rank 0:             reduceLearnRateIfImproveLessThan = 0
MPI Rank 0:             loadBestModel = true
MPI Rank 0:             increaseLearnRateIfImproveMoreThan = 1000000000
MPI Rank 0:             learnRateDecreaseFactor = 0.5
MPI Rank 0:             learnRateIncreaseFactor = 1.382
MPI Rank 0:             autoAdjustLR = "adjustAfterEpoch"
MPI Rank 0:         ]
MPI Rank 0:     ]
MPI Rank 0:     reader = [
MPI Rank 0:         readerType = "HTKMLFReader"
MPI Rank 0:         readMethod = "blockRandomize"
MPI Rank 0:         miniBatchMode = "partial"
MPI Rank 0:         randomize = "auto"
MPI Rank 0:         verbosity = 0
MPI Rank 0:         useMersenneTwisterRand=true
MPI Rank 0:         features = [
MPI Rank 0:             dim = 363
MPI Rank 0:             type = "real"
MPI Rank 0:             scpFile = "glob_0000.scp"
MPI Rank 0:         ]
MPI Rank 0:         labels = [
MPI Rank 0:             mlfFile = "$DataDir$/glob_0000.mlf"
MPI Rank 0:             labelMappingFile = "$DataDir$/state.list"
MPI Rank 0:             labelDim = 132
MPI Rank 0:             labelType = "category"
MPI Rank 0:         ]
MPI Rank 0:     ]
MPI Rank 0: ]
MPI Rank 0: currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 0: RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 0: DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 0: ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN
MPI Rank 0: OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 0: DeviceId=0
MPI Rank 0: timestamping=true
MPI Rank 0: numCPUThreads=12
MPI Rank 0: precision=double
MPI Rank 0: speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]
MPI Rank 0: stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: <<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: >>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
MPI Rank 0: 08/16/2016 03:20:02: precision = "float"
MPI Rank 0: command = speechTrain
MPI Rank 0: deviceId = 0
MPI Rank 0: parallelTrain = true
MPI Rank 0: speechTrain = [
MPI Rank 0:     action = "train"
MPI Rank 0:     modelPath = "C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn"
MPI Rank 0:     deviceId = 0
MPI Rank 0:     traceLevel = 1
MPI Rank 0:     SimpleNetworkBuilder = [
MPI Rank 0:         layerSizes = 363:512:512:132
MPI Rank 0:         trainingCriterion = "CrossEntropyWithSoftmax"
MPI Rank 0:         evalCriterion = "ClassificationError"
MPI Rank 0:         layerTypes = "Sigmoid"
MPI Rank 0:         initValueScale = 1.0
MPI Rank 0:         applyMeanVarNorm = true
MPI Rank 0:         uniformInit = true
MPI Rank 0:         needPrior = true
MPI Rank 0:     ]
MPI Rank 0:     ExperimentalNetworkBuilder = [    // the same as above but with BS. Not active; activate by commenting out the SimpleNetworkBuilder entry above
MPI Rank 0:         layerSizes = 363:512:512:132
MPI Rank 0:         trainingCriterion = 'CE'
MPI Rank 0:         evalCriterion = 'Err'
MPI Rank 0:         applyMeanVarNorm = true
MPI Rank 0:         L = Length(layerSizes)-1    // number of model layers
MPI Rank 0:         features = Input(layerSizes[0], 1, tag='feature') ; labels = Input(layerSizes[Length(layerSizes)-1], 1, tag='label')
MPI Rank 0:         featNorm = if applyMeanVarNorm
MPI Rank 0:                    then MeanVarNorm(features)
MPI Rank 0:                    else features
MPI Rank 0:         layers[layer:1..L-1] = if layer > 1
MPI Rank 0:                                then SBFF(layers[layer-1].Eh, layerSizes[layer], layerSizes[layer-1])
MPI Rank 0:                                else SBFF(featNorm, layerSizes[layer], layerSizes[layer-1])
MPI Rank 0:         outLayer = BFF(layers[L-1].Eh, layerSizes[L], layerSizes[L-1])
MPI Rank 0:         outZ = outLayer.z        // + PastValue(layerSizes[L], 1, outLayer.z)
MPI Rank 0:         CE = if trainingCriterion == 'CE'
MPI Rank 0:              then CrossEntropyWithSoftmax(labels, outZ, tag='criterion')
MPI Rank 0:              else Fail('unknown trainingCriterion ' + trainingCriterion)
MPI Rank 0:         Err = if evalCriterion == 'Err' then
MPI Rank 0:               ClassificationError(labels, outZ, tag='evaluation')
MPI Rank 0:               else Fail('unknown evalCriterion ' + evalCriterion)
MPI Rank 0:         logPrior = LogPrior(labels)
MPI Rank 0:         // TODO: how to add a tag to an infix operation?
MPI Rank 0:         ScaledLogLikelihood = Minus (outZ, logPrior, tag='output')
MPI Rank 0:     ]
MPI Rank 0:     SGD = [
MPI Rank 0:         epochSize = 20480
MPI Rank 0:         minibatchSize = 64:256:1024
MPI Rank 0:         learningRatesPerMB = 1.0:0.5:0.1
MPI Rank 0:         numMBsToShowResult = 3
MPI Rank 0:         momentumPerMB = 0.9:0.656119
MPI Rank 0:         dropoutRate = 0.0
MPI Rank 0:         maxEpochs = 5
MPI Rank 0:         keepCheckPointFiles = true
MPI Rank 0:         clippingThresholdPerSample = 1#INF
MPI Rank 0:         ParallelTrain = [
MPI Rank 0:             parallelizationMethod = "BlockMomentumSGD"
MPI Rank 0:             distributedMBReading = true
MPI Rank 0:             syncPerfStats=1
MPI Rank 0:             BlockMomentumSGD = [
MPI Rank 0:                 blockSizePerWorker=2048
MPI Rank 0:                 resetSGDMomentum=true
MPI Rank 0:                 useNesterovMomentum=true
MPI Rank 0:             ]
MPI Rank 0:         ]
MPI Rank 0:         AutoAdjust = [
MPI Rank 0:             reduceLearnRateIfImproveLessThan = 0
MPI Rank 0:             loadBestModel = true
MPI Rank 0:             increaseLearnRateIfImproveMoreThan = 1000000000
MPI Rank 0:             learnRateDecreaseFactor = 0.5
MPI Rank 0:             learnRateIncreaseFactor = 1.382
MPI Rank 0:             autoAdjustLR = "adjustAfterEpoch"
MPI Rank 0:         ]
MPI Rank 0:     ]
MPI Rank 0:     reader = [
MPI Rank 0:         readerType = "HTKMLFReader"
MPI Rank 0:         readMethod = "blockRandomize"
MPI Rank 0:         miniBatchMode = "partial"
MPI Rank 0:         randomize = "auto"
MPI Rank 0:         verbosity = 0
MPI Rank 0:         useMersenneTwisterRand=true
MPI Rank 0:         features = [
MPI Rank 0:             dim = 363
MPI Rank 0:             type = "real"
MPI Rank 0:             scpFile = "glob_0000.scp"
MPI Rank 0:         ]
MPI Rank 0:         labels = [
MPI Rank 0:             mlfFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/glob_0000.mlf"
MPI Rank 0:             labelMappingFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/state.list"
MPI Rank 0:             labelDim = 132
MPI Rank 0:             labelType = "category"
MPI Rank 0:         ]
MPI Rank 0:     ]
MPI Rank 0: ]
MPI Rank 0: currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 0: RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 0: DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 0: ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN
MPI Rank 0: OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 0: DeviceId=0
MPI Rank 0: timestamping=true
MPI Rank 0: numCPUThreads=12
MPI Rank 0: precision=double
MPI Rank 0: speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]
MPI Rank 0: stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: <<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: >>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
MPI Rank 0: configparameters: cntk.cntk:command=speechTrain
MPI Rank 0: configparameters: cntk.cntk:ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN
MPI Rank 0: configparameters: cntk.cntk:currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 0: configparameters: cntk.cntk:DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 0: configparameters: cntk.cntk:deviceId=0
MPI Rank 0: configparameters: cntk.cntk:numCPUThreads=12
MPI Rank 0: configparameters: cntk.cntk:OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 0: configparameters: cntk.cntk:parallelTrain=true
MPI Rank 0: configparameters: cntk.cntk:precision=double
MPI Rank 0: configparameters: cntk.cntk:RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 0: configparameters: cntk.cntk:speechTrain=[
MPI Rank 0:     action = "train"
MPI Rank 0:     modelPath = "C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn"
MPI Rank 0:     deviceId = 0
MPI Rank 0:     traceLevel = 1
MPI Rank 0:     SimpleNetworkBuilder = [
MPI Rank 0:         layerSizes = 363:512:512:132
MPI Rank 0:         trainingCriterion = "CrossEntropyWithSoftmax"
MPI Rank 0:         evalCriterion = "ClassificationError"
MPI Rank 0:         layerTypes = "Sigmoid"
MPI Rank 0:         initValueScale = 1.0
MPI Rank 0:         applyMeanVarNorm = true
MPI Rank 0:         uniformInit = true
MPI Rank 0:         needPrior = true
MPI Rank 0:     ]
MPI Rank 0:     ExperimentalNetworkBuilder = [    // the same as above but with BS. Not active; activate by commenting out the SimpleNetworkBuilder entry above
MPI Rank 0:         layerSizes = 363:512:512:132
MPI Rank 0:         trainingCriterion = 'CE'
MPI Rank 0:         evalCriterion = 'Err'
MPI Rank 0:         applyMeanVarNorm = true
MPI Rank 0:         L = Length(layerSizes)-1    // number of model layers
MPI Rank 0:         features = Input(layerSizes[0], 1, tag='feature') ; labels = Input(layerSizes[Length(layerSizes)-1], 1, tag='label')
MPI Rank 0:         featNorm = if applyMeanVarNorm
MPI Rank 0:                    then MeanVarNorm(features)
MPI Rank 0:                    else features
MPI Rank 0:         layers[layer:1..L-1] = if layer > 1
MPI Rank 0:                                then SBFF(layers[layer-1].Eh, layerSizes[layer], layerSizes[layer-1])
MPI Rank 0:                                else SBFF(featNorm, layerSizes[layer], layerSizes[layer-1])
MPI Rank 0:         outLayer = BFF(layers[L-1].Eh, layerSizes[L], layerSizes[L-1])
MPI Rank 0:         outZ = outLayer.z        // + PastValue(layerSizes[L], 1, outLayer.z)
MPI Rank 0:         CE = if trainingCriterion == 'CE'
MPI Rank 0:              then CrossEntropyWithSoftmax(labels, outZ, tag='criterion')
MPI Rank 0:              else Fail('unknown trainingCriterion ' + trainingCriterion)
MPI Rank 0:         Err = if evalCriterion == 'Err' then
MPI Rank 0:               ClassificationError(labels, outZ, tag='evaluation')
MPI Rank 0:               else Fail('unknown evalCriterion ' + evalCriterion)
MPI Rank 0:         logPrior = LogPrior(labels)
MPI Rank 0:         // TODO: how to add a tag to an infix operation?
MPI Rank 0:         ScaledLogLikelihood = Minus (outZ, logPrior, tag='output')
MPI Rank 0:     ]
MPI Rank 0:     SGD = [
MPI Rank 0:         epochSize = 20480
MPI Rank 0:         minibatchSize = 64:256:1024
MPI Rank 0:         learningRatesPerMB = 1.0:0.5:0.1
MPI Rank 0:         numMBsToShowResult = 3
MPI Rank 0:         momentumPerMB = 0.9:0.656119
MPI Rank 0:         dropoutRate = 0.0
MPI Rank 0:         maxEpochs = 5
MPI Rank 0:         keepCheckPointFiles = true
MPI Rank 0:         clippingThresholdPerSample = 1#INF
MPI Rank 0:         ParallelTrain = [
MPI Rank 0:             parallelizationMethod = "BlockMomentumSGD"
MPI Rank 0:             distributedMBReading = true
MPI Rank 0:             syncPerfStats=1
MPI Rank 0:             BlockMomentumSGD = [
MPI Rank 0:                 blockSizePerWorker=2048
MPI Rank 0:                 resetSGDMomentum=true
MPI Rank 0:                 useNesterovMomentum=true
MPI Rank 0:             ]
MPI Rank 0:         ]
MPI Rank 0:         AutoAdjust = [
MPI Rank 0:             reduceLearnRateIfImproveLessThan = 0
MPI Rank 0:             loadBestModel = true
MPI Rank 0:             increaseLearnRateIfImproveMoreThan = 1000000000
MPI Rank 0:             learnRateDecreaseFactor = 0.5
MPI Rank 0:             learnRateIncreaseFactor = 1.382
MPI Rank 0:             autoAdjustLR = "adjustAfterEpoch"
MPI Rank 0:         ]
MPI Rank 0:     ]
MPI Rank 0:     reader = [
MPI Rank 0:         readerType = "HTKMLFReader"
MPI Rank 0:         readMethod = "blockRandomize"
MPI Rank 0:         miniBatchMode = "partial"
MPI Rank 0:         randomize = "auto"
MPI Rank 0:         verbosity = 0
MPI Rank 0:         useMersenneTwisterRand=true
MPI Rank 0:         features = [
MPI Rank 0:             dim = 363
MPI Rank 0:             type = "real"
MPI Rank 0:             scpFile = "glob_0000.scp"
MPI Rank 0:         ]
MPI Rank 0:         labels = [
MPI Rank 0:             mlfFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/glob_0000.mlf"
MPI Rank 0:             labelMappingFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/state.list"
MPI Rank 0:             labelDim = 132
MPI Rank 0:             labelType = "category"
MPI Rank 0:         ]
MPI Rank 0:     ]
MPI Rank 0: ] [SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]
MPI Rank 0: 
MPI Rank 0: configparameters: cntk.cntk:stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 0: configparameters: cntk.cntk:timestamping=true
MPI Rank 0: 08/16/2016 03:20:02: <<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
MPI Rank 0: 08/16/2016 03:20:02: Commands: speechTrain
MPI Rank 0: 08/16/2016 03:20:02: Precision = "double"
MPI Rank 0: 08/16/2016 03:20:02: Using 12 CPU threads.
MPI Rank 0: 08/16/2016 03:20:02: CNTKModelPath: C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn
MPI Rank 0: 08/16/2016 03:20:02: CNTKCommandTrainInfo: speechTrain : 5
MPI Rank 0: 08/16/2016 03:20:02: CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : 5
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: ##############################################################################
MPI Rank 0: 08/16/2016 03:20:02: #                                                                            #
MPI Rank 0: 08/16/2016 03:20:02: # Action "train"                                                             #
MPI Rank 0: 08/16/2016 03:20:02: #                                                                            #
MPI Rank 0: 08/16/2016 03:20:02: ##############################################################################
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:02: CNTKCommandTrainBegin: speechTrain
MPI Rank 0: SimpleNetworkBuilder Using GPU 0
MPI Rank 0: reading script file glob_0000.scp ... 948 entries
MPI Rank 0: total 132 state names in state list C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/state.list
MPI Rank 0: htkmlfreader: reading MLF file C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/glob_0000.mlf ... total 948 entries
MPI Rank 0: ...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
MPI Rank 0: label set 0: 129 classes
MPI Rank 0: minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: Creating virgin network.
MPI Rank 0: Node 'W0' (LearnableParameter operation): Initializing Parameter[512 x 363] <- 0.000000.
MPI Rank 0: Node 'W0' (LearnableParameter operation): Initializing Parameter[512 x 363] <- uniform(seed=1, range=0.050000*1.000000, onCPU=false).
MPI Rank 0: Microsoft::MSR::CNTK::GPUMatrix<ElemType>::SetUniformRandomValue (GPU): creating curand object with seed 1, sizeof(ElemType)==8
MPI Rank 0: Node 'B0' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 0: Node 'B0' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 0: Node 'W1' (LearnableParameter operation): Initializing Parameter[512 x 512] <- 0.000000.
MPI Rank 0: Node 'W1' (LearnableParameter operation): Initializing Parameter[512 x 512] <- uniform(seed=2, range=0.050000*1.000000, onCPU=false).
MPI Rank 0: Node 'B1' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 0: Node 'B1' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 0: Node 'W2' (LearnableParameter operation): Initializing Parameter[132 x 512] <- 0.000000.
MPI Rank 0: Node 'W2' (LearnableParameter operation): Initializing Parameter[132 x 512] <- uniform(seed=3, range=0.050000*1.000000, onCPU=false).
MPI Rank 0: Node 'B2' (LearnableParameter operation): Initializing Parameter[132 x 1] <- 0.000000.
MPI Rank 0: Node 'B2' (LearnableParameter operation): Initializing Parameter[132 x 1] <- 0.000000.
MPI Rank 0: 
MPI Rank 0: Post-processing network...
MPI Rank 0: 
MPI Rank 0: 7 roots:
MPI Rank 0: 	CrossEntropyWithSoftmax = CrossEntropyWithSoftmax()
MPI Rank 0: 	EvalClassificationError = ClassificationError()
MPI Rank 0: 	InvStdOfFeatures = InvStdDev()
MPI Rank 0: 	MeanOfFeatures = Mean()
MPI Rank 0: 	PosteriorProb = Softmax()
MPI Rank 0: 	Prior = Mean()
MPI Rank 0: 	ScaledLogLikelihood = Minus()
MPI Rank 0: 
MPI Rank 0: Validating network. 25 nodes to process in pass 1.
MPI Rank 0: 
MPI Rank 0: Validating --> labels = InputValue() :  -> [132 x *]
MPI Rank 0: Validating --> W2 = LearnableParameter() :  -> [132 x 512]
MPI Rank 0: Validating --> W1 = LearnableParameter() :  -> [512 x 512]
MPI Rank 0: Validating --> W0 = LearnableParameter() :  -> [512 x 363]
MPI Rank 0: Validating --> features = InputValue() :  -> [363 x *]
MPI Rank 0: Validating --> MeanOfFeatures = Mean (features) : [363 x *] -> [363]
MPI Rank 0: Validating --> InvStdOfFeatures = InvStdDev (features) : [363 x *] -> [363]
MPI Rank 0: Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization (features, MeanOfFeatures, InvStdOfFeatures) : [363 x *], [363], [363] -> [363 x *]
MPI Rank 0: Validating --> W0*features = Times (W0, MVNormalizedFeatures) : [512 x 363], [363 x *] -> [512 x *]
MPI Rank 0: Validating --> B0 = LearnableParameter() :  -> [512 x 1]
MPI Rank 0: Validating --> W0*features+B0 = Plus (W0*features, B0) : [512 x *], [512 x 1] -> [512 x 1 x *]
MPI Rank 0: Validating --> H1 = Sigmoid (W0*features+B0) : [512 x 1 x *] -> [512 x 1 x *]
MPI Rank 0: Validating --> W1*H1 = Times (W1, H1) : [512 x 512], [512 x 1 x *] -> [512 x 1 x *]
MPI Rank 0: Validating --> B1 = LearnableParameter() :  -> [512 x 1]
MPI Rank 0: Validating --> W1*H1+B1 = Plus (W1*H1, B1) : [512 x 1 x *], [512 x 1] -> [512 x 1 x *]
MPI Rank 0: Validating --> H2 = Sigmoid (W1*H1+B1) : [512 x 1 x *] -> [512 x 1 x *]
MPI Rank 0: Validating --> W2*H1 = Times (W2, H2) : [132 x 512], [512 x 1 x *] -> [132 x 1 x *]
MPI Rank 0: Validating --> B2 = LearnableParameter() :  -> [132 x 1]
MPI Rank 0: Validating --> HLast = Plus (W2*H1, B2) : [132 x 1 x *], [132 x 1] -> [132 x 1 x *]
MPI Rank 0: Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax (labels, HLast) : [132 x *], [132 x 1 x *] -> [1]
MPI Rank 0: Validating --> EvalClassificationError = ClassificationError (labels, HLast) : [132 x *], [132 x 1 x *] -> [1]
MPI Rank 0: Validating --> PosteriorProb = Softmax (HLast) : [132 x 1 x *] -> [132 x 1 x *]
MPI Rank 0: Validating --> Prior = Mean (labels) : [132 x *] -> [132]
MPI Rank 0: Validating --> LogOfPrior = Log (Prior) : [132] -> [132]
MPI Rank 0: Validating --> ScaledLogLikelihood = Minus (HLast, LogOfPrior) : [132 x 1 x *], [132] -> [132 x 1 x *]
MPI Rank 0: 
MPI Rank 0: Validating network. 17 nodes to process in pass 2.
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: Validating network, final pass.
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: 12 out of 25 nodes do not share the minibatch layout with the input data.
MPI Rank 0: 
MPI Rank 0: Post-processing network complete.
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: Created model with 25 nodes on GPU 0.
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: Training criterion node(s):
MPI Rank 0: 08/16/2016 03:20:03: 	CrossEntropyWithSoftmax = CrossEntropyWithSoftmax
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: Evaluation criterion node(s):
MPI Rank 0: 08/16/2016 03:20:03: 	EvalClassificationError = ClassificationError
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: Allocating matrices for forward and/or backward propagation.
MPI Rank 0: 
MPI Rank 0: Memory Sharing: Out of 40 matrices, 19 are shared as 8, and 21 are not shared.
MPI Rank 0: 
MPI Rank 0: 	{ B1 : [512 x 1] (gradient)
MPI Rank 0: 	  H2 : [512 x 1 x *] (gradient)
MPI Rank 0: 	  HLast : [132 x 1 x *] (gradient) }
MPI Rank 0: 	{ W1 : [512 x 512] (gradient)
MPI Rank 0: 	  W1*H1+B1 : [512 x 1 x *] }
MPI Rank 0: 	{ B0 : [512 x 1] (gradient)
MPI Rank 0: 	  H1 : [512 x 1 x *] (gradient)
MPI Rank 0: 	  W1*H1+B1 : [512 x 1 x *] (gradient)
MPI Rank 0: 	  W2*H1 : [132 x 1 x *] }
MPI Rank 0: 	{ H2 : [512 x 1 x *]
MPI Rank 0: 	  W1*H1 : [512 x 1 x *] (gradient) }
MPI Rank 0: 	{ H1 : [512 x 1 x *]
MPI Rank 0: 	  W0*features : [512 x *] (gradient) }
MPI Rank 0: 	{ W0 : [512 x 363] (gradient)
MPI Rank 0: 	  W0*features+B0 : [512 x 1 x *] }
MPI Rank 0: 	{ W0*features+B0 : [512 x 1 x *] (gradient)
MPI Rank 0: 	  W1*H1 : [512 x 1 x *] }
MPI Rank 0: 	{ HLast : [132 x 1 x *]
MPI Rank 0: 	  W2 : [132 x 512] (gradient) }
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: Training 516740 parameters in 6 out of 6 parameter tensors and 15 nodes with gradient:
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: 	Node 'B0' (LearnableParameter operation) : [512 x 1]
MPI Rank 0: 08/16/2016 03:20:03: 	Node 'B1' (LearnableParameter operation) : [512 x 1]
MPI Rank 0: 08/16/2016 03:20:03: 	Node 'B2' (LearnableParameter operation) : [132 x 1]
MPI Rank 0: 08/16/2016 03:20:03: 	Node 'W0' (LearnableParameter operation) : [512 x 363]
MPI Rank 0: 08/16/2016 03:20:03: 	Node 'W1' (LearnableParameter operation) : [512 x 512]
MPI Rank 0: 08/16/2016 03:20:03: 	Node 'W2' (LearnableParameter operation) : [132 x 512]
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: Precomputing --> 3 PreCompute nodes found.
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:03: 	MeanOfFeatures = Mean()
MPI Rank 0: 08/16/2016 03:20:03: 	InvStdOfFeatures = InvStdDev()
MPI Rank 0: 08/16/2016 03:20:03: 	Prior = Mean()
MPI Rank 0: minibatchiterator: epoch 0: frames [0..252734] (first utterance at frame 0), data subset 0 of 1, with 1 datapasses
MPI Rank 0: requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:08: Precomputing --> Completed.
MPI Rank 0: 
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:09: Starting Epoch 1: learning rate per sample = 0.015625  effective momentum = 0.900000  momentum as time constant = 607.4 samples
MPI Rank 0: minibatchiterator: epoch 0: frames [0..20480] (first utterance at frame 0), data subset 0 of 1, with 1 datapasses
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:09: Starting minibatch loop.
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[   1-   3, 0.94%]: CrossEntropyWithSoftmax = 4.68332137 * 192; EvalClassificationError = 0.98958333 * 192; time = 0.0205s; samplesPerSecond = 9370.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[   4-   6, 1.88%]: CrossEntropyWithSoftmax = 4.42217834 * 192; EvalClassificationError = 0.89062500 * 192; time = 0.0205s; samplesPerSecond = 9375.9
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[   7-   9, 2.81%]: CrossEntropyWithSoftmax = 4.78361173 * 192; EvalClassificationError = 0.93229167 * 192; time = 0.0216s; samplesPerSecond = 8886.8
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  10-  12, 3.75%]: CrossEntropyWithSoftmax = 4.28266373 * 192; EvalClassificationError = 0.90104167 * 192; time = 0.0220s; samplesPerSecond = 8722.9
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  13-  15, 4.69%]: CrossEntropyWithSoftmax = 4.48941255 * 192; EvalClassificationError = 0.89583333 * 192; time = 0.0217s; samplesPerSecond = 8849.1
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  16-  18, 5.63%]: CrossEntropyWithSoftmax = 4.39312485 * 192; EvalClassificationError = 0.94270833 * 192; time = 0.0217s; samplesPerSecond = 8867.5
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  19-  21, 6.56%]: CrossEntropyWithSoftmax = 4.26351131 * 192; EvalClassificationError = 0.98437500 * 192; time = 0.0218s; samplesPerSecond = 8791.2
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  22-  24, 7.50%]: CrossEntropyWithSoftmax = 4.06459300 * 192; EvalClassificationError = 0.90625000 * 192; time = 0.0216s; samplesPerSecond = 8878.2
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  25-  27, 8.44%]: CrossEntropyWithSoftmax = 3.97918086 * 192; EvalClassificationError = 0.90104167 * 192; time = 0.0220s; samplesPerSecond = 8746.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  28-  30, 9.38%]: CrossEntropyWithSoftmax = 3.83987406 * 192; EvalClassificationError = 0.85416667 * 192; time = 0.0205s; samplesPerSecond = 9363.1
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  31-  33, 10.31%]: CrossEntropyWithSoftmax = 3.79421365 * 192; EvalClassificationError = 0.84895833 * 192; time = 0.0206s; samplesPerSecond = 9334.0
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  34-  36, 11.25%]: CrossEntropyWithSoftmax = 3.76043875 * 192; EvalClassificationError = 0.86979167 * 192; time = 0.0206s; samplesPerSecond = 9341.7
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  37-  39, 12.19%]: CrossEntropyWithSoftmax = 3.73154819 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0205s; samplesPerSecond = 9357.6
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  40-  42, 13.13%]: CrossEntropyWithSoftmax = 3.68435113 * 192; EvalClassificationError = 0.83333333 * 192; time = 0.0206s; samplesPerSecond = 9339.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  43-  45, 14.06%]: CrossEntropyWithSoftmax = 3.89304289 * 192; EvalClassificationError = 0.90104167 * 192; time = 0.0205s; samplesPerSecond = 9349.0
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  46-  48, 15.00%]: CrossEntropyWithSoftmax = 3.76826980 * 192; EvalClassificationError = 0.86979167 * 192; time = 0.0205s; samplesPerSecond = 9349.0
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  49-  51, 15.94%]: CrossEntropyWithSoftmax = 3.81256010 * 192; EvalClassificationError = 0.91145833 * 192; time = 0.0207s; samplesPerSecond = 9287.9
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  52-  54, 16.88%]: CrossEntropyWithSoftmax = 3.92133036 * 192; EvalClassificationError = 0.87500000 * 192; time = 0.0205s; samplesPerSecond = 9357.2
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  55-  57, 17.81%]: CrossEntropyWithSoftmax = 3.56128851 * 192; EvalClassificationError = 0.84895833 * 192; time = 0.0205s; samplesPerSecond = 9346.7
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  58-  60, 18.75%]: CrossEntropyWithSoftmax = 3.67830123 * 192; EvalClassificationError = 0.86979167 * 192; time = 0.0205s; samplesPerSecond = 9370.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  61-  63, 19.69%]: CrossEntropyWithSoftmax = 3.36612487 * 192; EvalClassificationError = 0.76041667 * 192; time = 0.0204s; samplesPerSecond = 9392.0
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  64-  66, 20.63%]: CrossEntropyWithSoftmax = 3.41785793 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9351.3
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  67-  69, 21.56%]: CrossEntropyWithSoftmax = 3.37660990 * 192; EvalClassificationError = 0.79166667 * 192; time = 0.0205s; samplesPerSecond = 9358.1
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  70-  72, 22.50%]: CrossEntropyWithSoftmax = 3.68727319 * 192; EvalClassificationError = 0.83333333 * 192; time = 0.0205s; samplesPerSecond = 9371.3
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  73-  75, 23.44%]: CrossEntropyWithSoftmax = 3.64994161 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0204s; samplesPerSecond = 9397.5
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  76-  78, 24.38%]: CrossEntropyWithSoftmax = 3.37700933 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0205s; samplesPerSecond = 9352.2
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  79-  81, 25.31%]: CrossEntropyWithSoftmax = 3.51711918 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0204s; samplesPerSecond = 9393.3
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  82-  84, 26.25%]: CrossEntropyWithSoftmax = 3.47828208 * 192; EvalClassificationError = 0.80208333 * 192; time = 0.0205s; samplesPerSecond = 9368.1
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  85-  87, 27.19%]: CrossEntropyWithSoftmax = 3.54864536 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0206s; samplesPerSecond = 9338.1
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  88-  90, 28.13%]: CrossEntropyWithSoftmax = 3.43454656 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9353.5
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  91-  93, 29.06%]: CrossEntropyWithSoftmax = 3.36875092 * 192; EvalClassificationError = 0.79687500 * 192; time = 0.0205s; samplesPerSecond = 9359.0
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  94-  96, 30.00%]: CrossEntropyWithSoftmax = 3.55401747 * 192; EvalClassificationError = 0.82291667 * 192; time = 0.0205s; samplesPerSecond = 9349.9
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  97-  99, 30.94%]: CrossEntropyWithSoftmax = 3.34809828 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9354.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 100- 102, 31.87%]: CrossEntropyWithSoftmax = 3.49450178 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0206s; samplesPerSecond = 9328.1
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 103- 105, 32.81%]: CrossEntropyWithSoftmax = 3.55445642 * 192; EvalClassificationError = 0.83333333 * 192; time = 0.0205s; samplesPerSecond = 9365.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 106- 108, 33.75%]: CrossEntropyWithSoftmax = 3.49293717 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9361.3
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 109- 111, 34.69%]: CrossEntropyWithSoftmax = 3.32494958 * 192; EvalClassificationError = 0.79687500 * 192; time = 0.0205s; samplesPerSecond = 9372.7
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 112- 114, 35.63%]: CrossEntropyWithSoftmax = 3.28851645 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0206s; samplesPerSecond = 9310.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 115- 117, 36.56%]: CrossEntropyWithSoftmax = 3.19411841 * 192; EvalClassificationError = 0.78125000 * 192; time = 0.0206s; samplesPerSecond = 9310.0
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 118- 120, 37.50%]: CrossEntropyWithSoftmax = 3.25028950 * 192; EvalClassificationError = 0.78125000 * 192; time = 0.0206s; samplesPerSecond = 9329.4
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 121- 123, 38.44%]: CrossEntropyWithSoftmax = 3.53445091 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0204s; samplesPerSecond = 9407.2
MPI Rank 0: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 124- 126, 39.38%]: CrossEntropyWithSoftmax = 3.00326214 * 192; EvalClassificationError = 0.71875000 * 192; time = 0.0206s; samplesPerSecond = 9327.6
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 127- 129, 40.31%]: CrossEntropyWithSoftmax = 3.10787619 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0205s; samplesPerSecond = 9354.0
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 130- 132, 41.25%]: CrossEntropyWithSoftmax = 3.05588285 * 192; EvalClassificationError = 0.76562500 * 192; time = 0.0206s; samplesPerSecond = 9331.3
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 133- 135, 42.19%]: CrossEntropyWithSoftmax = 3.18197888 * 192; EvalClassificationError = 0.74479167 * 192; time = 0.0205s; samplesPerSecond = 9343.5
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 136- 138, 43.13%]: CrossEntropyWithSoftmax = 2.78138232 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0205s; samplesPerSecond = 9349.9
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 139- 141, 44.06%]: CrossEntropyWithSoftmax = 3.17441293 * 192; EvalClassificationError = 0.82291667 * 192; time = 0.0204s; samplesPerSecond = 9388.8
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 142- 144, 45.00%]: CrossEntropyWithSoftmax = 3.03537426 * 192; EvalClassificationError = 0.73437500 * 192; time = 0.0205s; samplesPerSecond = 9347.6
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 145- 147, 45.94%]: CrossEntropyWithSoftmax = 3.00595982 * 192; EvalClassificationError = 0.73958333 * 192; time = 0.0205s; samplesPerSecond = 9346.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 148- 150, 46.88%]: CrossEntropyWithSoftmax = 2.79115323 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0205s; samplesPerSecond = 9358.5
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 151- 153, 47.81%]: CrossEntropyWithSoftmax = 3.04097805 * 192; EvalClassificationError = 0.69791667 * 192; time = 0.0206s; samplesPerSecond = 9339.9
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 154- 156, 48.75%]: CrossEntropyWithSoftmax = 3.15935323 * 192; EvalClassificationError = 0.76562500 * 192; time = 0.0205s; samplesPerSecond = 9351.3
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 157- 159, 49.69%]: CrossEntropyWithSoftmax = 3.11947251 * 192; EvalClassificationError = 0.75000000 * 192; time = 0.0206s; samplesPerSecond = 9338.1
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 160- 162, 50.63%]: CrossEntropyWithSoftmax = 2.86232384 * 192; EvalClassificationError = 0.72395833 * 192; time = 0.0206s; samplesPerSecond = 9341.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 163- 165, 51.56%]: CrossEntropyWithSoftmax = 2.77742114 * 192; EvalClassificationError = 0.70312500 * 192; time = 0.0206s; samplesPerSecond = 9339.4
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 166- 168, 52.50%]: CrossEntropyWithSoftmax = 3.00411374 * 192; EvalClassificationError = 0.73958333 * 192; time = 0.0205s; samplesPerSecond = 9344.4
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 169- 171, 53.44%]: CrossEntropyWithSoftmax = 2.69740857 * 192; EvalClassificationError = 0.68750000 * 192; time = 0.0204s; samplesPerSecond = 9409.0
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 172- 174, 54.37%]: CrossEntropyWithSoftmax = 2.64948203 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0206s; samplesPerSecond = 9338.1
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 175- 177, 55.31%]: CrossEntropyWithSoftmax = 2.71417618 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0205s; samplesPerSecond = 9383.7
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 178- 180, 56.25%]: CrossEntropyWithSoftmax = 2.74281938 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0206s; samplesPerSecond = 9328.1
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 181- 183, 57.19%]: CrossEntropyWithSoftmax = 2.81346277 * 192; EvalClassificationError = 0.72916667 * 192; time = 0.0205s; samplesPerSecond = 9384.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 184- 186, 58.13%]: CrossEntropyWithSoftmax = 2.79862588 * 192; EvalClassificationError = 0.71875000 * 192; time = 0.0206s; samplesPerSecond = 9318.6
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 187- 189, 59.06%]: CrossEntropyWithSoftmax = 2.76655026 * 192; EvalClassificationError = 0.71354167 * 192; time = 0.0205s; samplesPerSecond = 9344.0
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 190- 192, 60.00%]: CrossEntropyWithSoftmax = 2.75908943 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0206s; samplesPerSecond = 9340.8
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 193- 195, 60.94%]: CrossEntropyWithSoftmax = 2.53548199 * 192; EvalClassificationError = 0.63541667 * 192; time = 0.0204s; samplesPerSecond = 9411.8
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 196- 198, 61.88%]: CrossEntropyWithSoftmax = 2.97589950 * 192; EvalClassificationError = 0.73437500 * 192; time = 0.0206s; samplesPerSecond = 9329.9
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 199- 201, 62.81%]: CrossEntropyWithSoftmax = 2.64996722 * 192; EvalClassificationError = 0.63020833 * 192; time = 0.0205s; samplesPerSecond = 9349.4
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 202- 204, 63.75%]: CrossEntropyWithSoftmax = 2.52128374 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0205s; samplesPerSecond = 9345.8
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 205- 207, 64.69%]: CrossEntropyWithSoftmax = 2.64228785 * 192; EvalClassificationError = 0.66145833 * 192; time = 0.0205s; samplesPerSecond = 9350.8
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 208- 210, 65.63%]: CrossEntropyWithSoftmax = 2.57199182 * 192; EvalClassificationError = 0.67708333 * 192; time = 0.0205s; samplesPerSecond = 9344.9
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 211- 213, 66.56%]: CrossEntropyWithSoftmax = 2.58100902 * 192; EvalClassificationError = 0.63020833 * 192; time = 0.0205s; samplesPerSecond = 9347.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 214- 216, 67.50%]: CrossEntropyWithSoftmax = 2.48555431 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0206s; samplesPerSecond = 9334.9
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 217- 219, 68.44%]: CrossEntropyWithSoftmax = 2.75336137 * 192; EvalClassificationError = 0.67187500 * 192; time = 0.0206s; samplesPerSecond = 9340.8
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 220- 222, 69.38%]: CrossEntropyWithSoftmax = 2.49193178 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0205s; samplesPerSecond = 9352.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 223- 225, 70.31%]: CrossEntropyWithSoftmax = 2.46098647 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0206s; samplesPerSecond = 9341.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 226- 228, 71.25%]: CrossEntropyWithSoftmax = 2.74322101 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0205s; samplesPerSecond = 9351.7
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 229- 231, 72.19%]: CrossEntropyWithSoftmax = 2.55837089 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0205s; samplesPerSecond = 9344.0
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 232- 234, 73.13%]: CrossEntropyWithSoftmax = 2.98288510 * 192; EvalClassificationError = 0.76562500 * 192; time = 0.0206s; samplesPerSecond = 9329.4
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 235- 237, 74.06%]: CrossEntropyWithSoftmax = 2.36667287 * 192; EvalClassificationError = 0.58854167 * 192; time = 0.0204s; samplesPerSecond = 9391.5
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 238- 240, 75.00%]: CrossEntropyWithSoftmax = 2.25169614 * 192; EvalClassificationError = 0.61458333 * 192; time = 0.0205s; samplesPerSecond = 9364.0
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 241- 243, 75.94%]: CrossEntropyWithSoftmax = 2.31564120 * 192; EvalClassificationError = 0.57291667 * 192; time = 0.0206s; samplesPerSecond = 9323.6
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 244- 246, 76.88%]: CrossEntropyWithSoftmax = 2.70894816 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0205s; samplesPerSecond = 9351.3
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 247- 249, 77.81%]: CrossEntropyWithSoftmax = 2.44991146 * 192; EvalClassificationError = 0.63020833 * 192; time = 0.0206s; samplesPerSecond = 9322.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 250- 252, 78.75%]: CrossEntropyWithSoftmax = 2.51856232 * 192; EvalClassificationError = 0.66666667 * 192; time = 0.0206s; samplesPerSecond = 9331.3
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 253- 255, 79.69%]: CrossEntropyWithSoftmax = 2.38498228 * 192; EvalClassificationError = 0.61979167 * 192; time = 0.0205s; samplesPerSecond = 9347.2
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 256- 258, 80.63%]: CrossEntropyWithSoftmax = 2.38080698 * 192; EvalClassificationError = 0.58333333 * 192; time = 0.0205s; samplesPerSecond = 9351.7
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 259- 261, 81.56%]: CrossEntropyWithSoftmax = 2.34294967 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0205s; samplesPerSecond = 9352.6
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 262- 264, 82.50%]: CrossEntropyWithSoftmax = 2.30340167 * 192; EvalClassificationError = 0.58854167 * 192; time = 0.0206s; samplesPerSecond = 9324.0
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 265- 267, 83.44%]: CrossEntropyWithSoftmax = 2.08323277 * 192; EvalClassificationError = 0.52604167 * 192; time = 0.0205s; samplesPerSecond = 9359.9
MPI Rank 0: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 268- 270, 84.38%]: CrossEntropyWithSoftmax = 2.43589953 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0206s; samplesPerSecond = 9329.9
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 271- 273, 85.31%]: CrossEntropyWithSoftmax = 2.53399486 * 192; EvalClassificationError = 0.66145833 * 192; time = 0.0206s; samplesPerSecond = 9340.8
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 274- 276, 86.25%]: CrossEntropyWithSoftmax = 2.33995708 * 192; EvalClassificationError = 0.60416667 * 192; time = 0.0205s; samplesPerSecond = 9347.2
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 277- 279, 87.19%]: CrossEntropyWithSoftmax = 2.62970864 * 192; EvalClassificationError = 0.68229167 * 192; time = 0.0206s; samplesPerSecond = 9329.0
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 280- 282, 88.13%]: CrossEntropyWithSoftmax = 2.47993989 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0206s; samplesPerSecond = 9337.6
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 283- 285, 89.06%]: CrossEntropyWithSoftmax = 2.30935254 * 192; EvalClassificationError = 0.63541667 * 192; time = 0.0206s; samplesPerSecond = 9340.3
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 286- 288, 90.00%]: CrossEntropyWithSoftmax = 2.22022265 * 192; EvalClassificationError = 0.59375000 * 192; time = 0.0205s; samplesPerSecond = 9347.6
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 289- 291, 90.94%]: CrossEntropyWithSoftmax = 2.28060247 * 192; EvalClassificationError = 0.56770833 * 192; time = 0.0205s; samplesPerSecond = 9367.7
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 292- 294, 91.88%]: CrossEntropyWithSoftmax = 2.13349744 * 192; EvalClassificationError = 0.57291667 * 192; time = 0.0206s; samplesPerSecond = 9312.7
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 295- 297, 92.81%]: CrossEntropyWithSoftmax = 2.29751600 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0204s; samplesPerSecond = 9405.3
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 298- 300, 93.75%]: CrossEntropyWithSoftmax = 2.29319143 * 192; EvalClassificationError = 0.60416667 * 192; time = 0.0206s; samplesPerSecond = 9342.6
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 301- 303, 94.69%]: CrossEntropyWithSoftmax = 2.14551002 * 192; EvalClassificationError = 0.55729167 * 192; time = 0.0204s; samplesPerSecond = 9401.2
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 304- 306, 95.63%]: CrossEntropyWithSoftmax = 2.26930674 * 192; EvalClassificationError = 0.58333333 * 192; time = 0.0206s; samplesPerSecond = 9329.9
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 307- 309, 96.56%]: CrossEntropyWithSoftmax = 2.17383355 * 192; EvalClassificationError = 0.56770833 * 192; time = 0.0204s; samplesPerSecond = 9399.8
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 310- 312, 97.50%]: CrossEntropyWithSoftmax = 2.43111882 * 192; EvalClassificationError = 0.67187500 * 192; time = 0.0205s; samplesPerSecond = 9356.3
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 313- 315, 98.44%]: CrossEntropyWithSoftmax = 2.18011227 * 192; EvalClassificationError = 0.59895833 * 192; time = 0.0206s; samplesPerSecond = 9338.5
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 316- 318, 99.38%]: CrossEntropyWithSoftmax = 2.21682707 * 192; EvalClassificationError = 0.56250000 * 192; time = 0.0205s; samplesPerSecond = 9361.3
MPI Rank 0: 08/16/2016 03:20:11: Finished Epoch[ 1 of 5]: [Training] CrossEntropyWithSoftmax = 3.03815141 * 20480; EvalClassificationError = 0.73432617 * 20480; totalSamplesSeen = 20480; learningRatePerSample = 0.015625; epochTime=2.24705s
MPI Rank 0: 08/16/2016 03:20:11: SGD: Saving checkpoint model 'C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn.1'
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:11: Starting Epoch 2: learning rate per sample = 0.001953  effective momentum = 0.656119  momentum as time constant = 607.5 samples
MPI Rank 0: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 0: minibatchiterator: epoch 1: frames [20480..40960] (first utterance at frame 20480), data subset 0 of 2, with 1 datapasses
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:11: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[   1-   3, 3.75%]: CrossEntropyWithSoftmax = 2.20431953 * 508; EvalClassificationError = 0.59842520 * 508; time = 0.0273s; samplesPerSecond = 18577.4
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[   4-   6, 7.50%]: CrossEntropyWithSoftmax = 2.19237836 * 492; EvalClassificationError = 0.57723577 * 492; time = 0.0198s; samplesPerSecond = 24866.1
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[   7-   9, 11.25%]: CrossEntropyWithSoftmax = 2.17526222 * 488; EvalClassificationError = 0.59836066 * 488; time = 0.0224s; samplesPerSecond = 21761.4
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  10-  12, 15.00%]: CrossEntropyWithSoftmax = 2.23483521 * 527; EvalClassificationError = 0.59582543 * 527; time = 0.0441s; samplesPerSecond = 11942.0
MPI Rank 0: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.07-seconds latency this time; accumulated time on sync point = 0.07 seconds , average latency = 0.07 seconds
MPI Rank 0: 		(model aggregation stats) 1-th sync:     0.24 seconds since last report (0.01 seconds on comm.); 4289 samples processed by 2 workers (2163 by me);
MPI Rank 0: 		(model aggregation stats) 1-th sync: totalThroughput = 18.06k samplesPerSecond , throughputPerWorker = 9.03k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  13-  15, 18.75%]: CrossEntropyWithSoftmax = 2.00528284 * 473; EvalClassificationError = 0.53911205 * 473; time = 0.1345s; samplesPerSecond = 3516.6
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  16-  18, 22.50%]: CrossEntropyWithSoftmax = 2.08558038 * 511; EvalClassificationError = 0.54990215 * 511; time = 0.0267s; samplesPerSecond = 19109.2
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  19-  21, 26.25%]: CrossEntropyWithSoftmax = 2.08506441 * 506; EvalClassificationError = 0.54940711 * 506; time = 0.0264s; samplesPerSecond = 19198.7
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  22-  24, 30.00%]: CrossEntropyWithSoftmax = 2.12168501 * 513; EvalClassificationError = 0.52241715 * 513; time = 0.0268s; samplesPerSecond = 19165.4
MPI Rank 0: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.05-seconds latency this time; accumulated time on sync point = 0.12 seconds , average latency = 0.06 seconds
MPI Rank 0: 		(model aggregation stats) 2-th sync:     0.18 seconds since last report (0.01 seconds on comm.); 4253 samples processed by 2 workers (2180 by me);
MPI Rank 0: 		(model aggregation stats) 2-th sync: totalThroughput = 23.54k samplesPerSecond , throughputPerWorker = 11.77k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  25-  27, 33.75%]: CrossEntropyWithSoftmax = 2.08058681 * 489; EvalClassificationError = 0.56646217 * 489; time = 0.0896s; samplesPerSecond = 5458.5
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  28-  30, 37.50%]: CrossEntropyWithSoftmax = 2.07411044 * 494; EvalClassificationError = 0.55060729 * 494; time = 0.0275s; samplesPerSecond = 17987.8
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  31-  33, 41.25%]: CrossEntropyWithSoftmax = 2.12310728 * 499; EvalClassificationError = 0.58316633 * 499; time = 0.0270s; samplesPerSecond = 18452.8
MPI Rank 0: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  34-  36, 45.00%]: CrossEntropyWithSoftmax = 2.06918046 * 490; EvalClassificationError = 0.57142857 * 490; time = 0.0263s; samplesPerSecond = 18658.1
MPI Rank 0: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.03-seconds latency this time; accumulated time on sync point = 0.15 seconds , average latency = 0.05 seconds
MPI Rank 0: 		(model aggregation stats) 3-th sync:     0.19 seconds since last report (0.00 seconds on comm.); 4246 samples processed by 2 workers (2144 by me);
MPI Rank 0: 		(model aggregation stats) 3-th sync: totalThroughput = 22.72k samplesPerSecond , throughputPerWorker = 11.36k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  37-  39, 48.75%]: CrossEntropyWithSoftmax = 1.93293190 * 497; EvalClassificationError = 0.50905433 * 497; time = 0.0989s; samplesPerSecond = 5024.7
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  40-  42, 52.50%]: CrossEntropyWithSoftmax = 2.13718669 * 492; EvalClassificationError = 0.59552846 * 492; time = 0.0262s; samplesPerSecond = 18811.7
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  43-  45, 56.25%]: CrossEntropyWithSoftmax = 1.91004653 * 508; EvalClassificationError = 0.54527559 * 508; time = 0.0266s; samplesPerSecond = 19105.6
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  46-  48, 60.00%]: CrossEntropyWithSoftmax = 1.97341931 * 503; EvalClassificationError = 0.52286282 * 503; time = 0.0264s; samplesPerSecond = 19054.5
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  49-  51, 63.75%]: CrossEntropyWithSoftmax = 2.07837958 * 470; EvalClassificationError = 0.57021277 * 470; time = 0.0165s; samplesPerSecond = 28431.4
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  52-  54, 67.50%]: CrossEntropyWithSoftmax = 1.91466321 * 494; EvalClassificationError = 0.54858300 * 494; time = 0.0169s; samplesPerSecond = 29239.4
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  55-  57, 71.25%]: CrossEntropyWithSoftmax = 2.00598279 * 503; EvalClassificationError = 0.52683897 * 503; time = 0.0250s; samplesPerSecond = 20135.3
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  58-  60, 75.00%]: CrossEntropyWithSoftmax = 1.96239369 * 487; EvalClassificationError = 0.54004107 * 487; time = 0.0167s; samplesPerSecond = 29081.6
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  61-  63, 78.75%]: CrossEntropyWithSoftmax = 1.96513086 * 516; EvalClassificationError = 0.51744186 * 516; time = 0.0171s; samplesPerSecond = 30145.5
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  64-  66, 82.50%]: CrossEntropyWithSoftmax = 1.97088895 * 494; EvalClassificationError = 0.55668016 * 494; time = 0.0169s; samplesPerSecond = 29220.4
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  67-  69, 86.25%]: CrossEntropyWithSoftmax = 2.01240053 * 510; EvalClassificationError = 0.56470588 * 510; time = 0.0170s; samplesPerSecond = 29947.2
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  70-  72, 90.00%]: CrossEntropyWithSoftmax = 2.02387287 * 497; EvalClassificationError = 0.54929577 * 497; time = 0.0169s; samplesPerSecond = 29464.1
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  73-  75, 93.75%]: CrossEntropyWithSoftmax = 1.90663199 * 490; EvalClassificationError = 0.55306122 * 490; time = 0.0168s; samplesPerSecond = 29196.2
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  76-  78, 97.50%]: CrossEntropyWithSoftmax = 1.94815649 * 482; EvalClassificationError = 0.52697095 * 482; time = 0.0167s; samplesPerSecond = 28841.6
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  79-  81, 101.25%]: CrossEntropyWithSoftmax = 1.97332721 * 342; EvalClassificationError = 0.50877193 * 342; time = 0.0114s; samplesPerSecond = 29947.5
MPI Rank 0: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.15 seconds , average latency = 0.04 seconds
MPI Rank 0: 		(model aggregation stats) 4-th sync:     0.30 seconds since last report (0.01 seconds on comm.); 7692 samples processed by 2 workers (6788 by me);
MPI Rank 0: 		(model aggregation stats) 4-th sync: totalThroughput = 25.33k samplesPerSecond , throughputPerWorker = 12.67k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:12: Finished Epoch[ 2 of 5]: [Training] CrossEntropyWithSoftmax = 2.05172118 * 20480; EvalClassificationError = 0.55805664 * 20480; totalSamplesSeen = 40960; learningRatePerSample = 0.001953125; epochTime=0.911963s
MPI Rank 0: 08/16/2016 03:20:12: SGD: Saving checkpoint model 'C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn.2'
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:12: Starting Epoch 3: learning rate per sample = 0.000098  effective momentum = 0.656119  momentum as time constant = 2429.9 samples
MPI Rank 0: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 0: minibatchiterator: epoch 2: frames [40960..61440] (first utterance at frame 40960), data subset 0 of 2, with 1 datapasses
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:12: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[   1-   3, 15.00%]: CrossEntropyWithSoftmax = 1.96185095 * 1942; EvalClassificationError = 0.53347065 * 1942; time = 0.0815s; samplesPerSecond = 23818.3
MPI Rank 0: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.07-seconds latency this time; accumulated time on sync point = 0.07 seconds , average latency = 0.07 seconds
MPI Rank 0: 		(model aggregation stats) 1-th sync:     0.21 seconds since last report (0.01 seconds on comm.); 4885 samples processed by 2 workers (2592 by me);
MPI Rank 0: 		(model aggregation stats) 1-th sync: totalThroughput = 23.55k samplesPerSecond , throughputPerWorker = 11.77k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[   4-   6, 30.00%]: CrossEntropyWithSoftmax = 1.94171416 * 1909; EvalClassificationError = 0.55055003 * 1909; time = 0.1664s; samplesPerSecond = 11474.4
MPI Rank 0: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.07 seconds , average latency = 0.03 seconds
MPI Rank 0: 		(model aggregation stats) 2-th sync:     0.15 seconds since last report (0.01 seconds on comm.); 4826 samples processed by 2 workers (2577 by me);
MPI Rank 0: 		(model aggregation stats) 2-th sync: totalThroughput = 33.06k samplesPerSecond , throughputPerWorker = 16.53k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[   7-   9, 45.00%]: CrossEntropyWithSoftmax = 1.98907844 * 1987; EvalClassificationError = 0.55158530 * 1987; time = 0.1247s; samplesPerSecond = 15928.5
MPI Rank 0: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.03-seconds latency this time; accumulated time on sync point = 0.10 seconds , average latency = 0.03 seconds
MPI Rank 0: 		(model aggregation stats) 3-th sync:     0.15 seconds since last report (0.01 seconds on comm.); 4903 samples processed by 2 workers (2577 by me);
MPI Rank 0: 		(model aggregation stats) 3-th sync: totalThroughput = 32.81k samplesPerSecond , throughputPerWorker = 16.40k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 3 of 5]-Minibatch[  10-  12, 60.00%]: CrossEntropyWithSoftmax = 1.94333646 * 1908; EvalClassificationError = 0.54926625 * 1908; time = 0.1201s; samplesPerSecond = 15880.5
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 3 of 5]-Minibatch[  13-  15, 75.00%]: CrossEntropyWithSoftmax = 1.97418902 * 1905; EvalClassificationError = 0.55223097 * 1905; time = 0.0580s; samplesPerSecond = 32831.2
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 3 of 5]-Minibatch[  16-  18, 90.00%]: CrossEntropyWithSoftmax = 1.96248532 * 1913; EvalClassificationError = 0.54887611 * 1913; time = 0.0561s; samplesPerSecond = 34070.1
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 3 of 5]-Minibatch[  19-  21, 105.00%]: CrossEntropyWithSoftmax = 1.97409307 * 1225; EvalClassificationError = 0.54367347 * 1225; time = 0.0289s; samplesPerSecond = 42355.3
MPI Rank 0: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.10 seconds , average latency = 0.02 seconds
MPI Rank 0: 		(model aggregation stats) 4-th sync:     0.17 seconds since last report (0.01 seconds on comm.); 5866 samples processed by 2 workers (5043 by me);
MPI Rank 0: 		(model aggregation stats) 4-th sync: totalThroughput = 34.19k samplesPerSecond , throughputPerWorker = 17.10k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:13: Finished Epoch[ 3 of 5]: [Training] CrossEntropyWithSoftmax = 1.95703393 * 20480; EvalClassificationError = 0.54541016 * 20480; totalSamplesSeen = 61440; learningRatePerSample = 9.7656251e-005; epochTime=0.676426s
MPI Rank 0: 08/16/2016 03:20:13: SGD: Saving checkpoint model 'C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn.3'
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:13: Starting Epoch 4: learning rate per sample = 0.000098  effective momentum = 0.656119  momentum as time constant = 2429.9 samples
MPI Rank 0: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 0: minibatchiterator: epoch 3: frames [61440..81920] (first utterance at frame 61440), data subset 0 of 2, with 1 datapasses
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:13: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[   1-   3, 15.00%]: CrossEntropyWithSoftmax = 1.88963761 * 1923; EvalClassificationError = 0.52366095 * 1923; time = 0.0686s; samplesPerSecond = 28024.7
MPI Rank 0: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.05-seconds latency this time; accumulated time on sync point = 0.05 seconds , average latency = 0.05 seconds
MPI Rank 0: 		(model aggregation stats) 1-th sync:     0.17 seconds since last report (0.01 seconds on comm.); 4901 samples processed by 2 workers (2550 by me);
MPI Rank 0: 		(model aggregation stats) 1-th sync: totalThroughput = 29.57k samplesPerSecond , throughputPerWorker = 14.78k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[   4-   6, 30.00%]: CrossEntropyWithSoftmax = 1.89669303 * 1870; EvalClassificationError = 0.51871658 * 1870; time = 0.1373s; samplesPerSecond = 13617.7
MPI Rank 0: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.03-seconds latency this time; accumulated time on sync point = 0.08 seconds , average latency = 0.04 seconds
MPI Rank 0: 		(model aggregation stats) 2-th sync:     0.13 seconds since last report (0.00 seconds on comm.); 4836 samples processed by 2 workers (2519 by me);
MPI Rank 0: 		(model aggregation stats) 2-th sync: totalThroughput = 38.34k samplesPerSecond , throughputPerWorker = 19.17k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[   7-   9, 45.00%]: CrossEntropyWithSoftmax = 1.91111689 * 1942; EvalClassificationError = 0.54119464 * 1942; time = 0.1103s; samplesPerSecond = 17603.8
MPI Rank 0: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.03-seconds latency this time; accumulated time on sync point = 0.11 seconds , average latency = 0.04 seconds
MPI Rank 0: 		(model aggregation stats) 3-th sync:     0.15 seconds since last report (0.00 seconds on comm.); 4952 samples processed by 2 workers (2551 by me);
MPI Rank 0: 		(model aggregation stats) 3-th sync: totalThroughput = 33.33k samplesPerSecond , throughputPerWorker = 16.67k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[  10-  12, 60.00%]: CrossEntropyWithSoftmax = 1.88475158 * 1885; EvalClassificationError = 0.51458886 * 1885; time = 0.1182s; samplesPerSecond = 15950.5
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[  13-  15, 75.00%]: CrossEntropyWithSoftmax = 1.89480846 * 1870; EvalClassificationError = 0.51497326 * 1870; time = 0.0611s; samplesPerSecond = 30621.6
MPI Rank 0: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[  16-  18, 90.00%]: CrossEntropyWithSoftmax = 1.89539137 * 1873; EvalClassificationError = 0.52108916 * 1873; time = 0.0441s; samplesPerSecond = 42465.9
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 4 of 5]-Minibatch[  19-  21, 105.00%]: CrossEntropyWithSoftmax = 1.93128839 * 1231; EvalClassificationError = 0.52721365 * 1231; time = 0.0292s; samplesPerSecond = 42189.3
MPI Rank 0: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.11 seconds , average latency = 0.03 seconds
MPI Rank 0: 		(model aggregation stats) 4-th sync:     0.16 seconds since last report (0.01 seconds on comm.); 5791 samples processed by 2 workers (4974 by me);
MPI Rank 0: 		(model aggregation stats) 4-th sync: totalThroughput = 35.98k samplesPerSecond , throughputPerWorker = 17.99k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:14: Finished Epoch[ 4 of 5]: [Training] CrossEntropyWithSoftmax = 1.90306770 * 20480; EvalClassificationError = 0.52641602 * 20480; totalSamplesSeen = 81920; learningRatePerSample = 9.7656251e-005; epochTime=0.604274s
MPI Rank 0: 08/16/2016 03:20:14: SGD: Saving checkpoint model 'C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn.4'
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:14: Starting Epoch 5: learning rate per sample = 0.000098  effective momentum = 0.656119  momentum as time constant = 2429.9 samples
MPI Rank 0: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 0: minibatchiterator: epoch 4: frames [81920..102400] (first utterance at frame 81920), data subset 0 of 2, with 1 datapasses
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:14: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[   1-   3, 15.00%]: CrossEntropyWithSoftmax = 1.93021270 * 1863; EvalClassificationError = 0.51851852 * 1863; time = 0.0632s; samplesPerSecond = 29459.7
MPI Rank 0: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.06-seconds latency this time; accumulated time on sync point = 0.06 seconds , average latency = 0.06 seconds
MPI Rank 0: 		(model aggregation stats) 1-th sync:     0.17 seconds since last report (0.01 seconds on comm.); 4919 samples processed by 2 workers (2493 by me);
MPI Rank 0: 		(model aggregation stats) 1-th sync: totalThroughput = 29.52k samplesPerSecond , throughputPerWorker = 14.76k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[   4-   6, 30.00%]: CrossEntropyWithSoftmax = 1.87273976 * 1855; EvalClassificationError = 0.52129380 * 1855; time = 0.1587s; samplesPerSecond = 11685.4
MPI Rank 0: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.03-seconds latency this time; accumulated time on sync point = 0.09 seconds , average latency = 0.04 seconds
MPI Rank 0: 		(model aggregation stats) 2-th sync:     0.14 seconds since last report (0.00 seconds on comm.); 4899 samples processed by 2 workers (2480 by me);
MPI Rank 0: 		(model aggregation stats) 2-th sync: totalThroughput = 35.61k samplesPerSecond , throughputPerWorker = 17.81k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[   7-   9, 45.00%]: CrossEntropyWithSoftmax = 1.87195439 * 1866; EvalClassificationError = 0.52304394 * 1866; time = 0.1007s; samplesPerSecond = 18521.6
MPI Rank 0: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.02-seconds latency this time; accumulated time on sync point = 0.11 seconds , average latency = 0.04 seconds
MPI Rank 0: 		(model aggregation stats) 3-th sync:     0.13 seconds since last report (0.00 seconds on comm.); 4829 samples processed by 2 workers (2470 by me);
MPI Rank 0: 		(model aggregation stats) 3-th sync: totalThroughput = 37.08k samplesPerSecond , throughputPerWorker = 18.54k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  10-  12, 60.00%]: CrossEntropyWithSoftmax = 1.88262131 * 1859; EvalClassificationError = 0.52017214 * 1859; time = 0.1062s; samplesPerSecond = 17508.2
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  13-  15, 75.00%]: CrossEntropyWithSoftmax = 1.82076948 * 1925; EvalClassificationError = 0.49714286 * 1925; time = 0.0664s; samplesPerSecond = 28998.8
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  16-  18, 90.00%]: CrossEntropyWithSoftmax = 1.84625728 * 1860; EvalClassificationError = 0.50967742 * 1860; time = 0.0447s; samplesPerSecond = 41652.7
MPI Rank 0: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  19-  21, 105.00%]: CrossEntropyWithSoftmax = 1.86158884 * 1239; EvalClassificationError = 0.51412429 * 1239; time = 0.0292s; samplesPerSecond = 42419.9
MPI Rank 0: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.11 seconds , average latency = 0.03 seconds
MPI Rank 0: 		(model aggregation stats) 4-th sync:     0.17 seconds since last report (0.01 seconds on comm.); 5833 samples processed by 2 workers (5024 by me);
MPI Rank 0: 		(model aggregation stats) 4-th sync: totalThroughput = 35.03k samplesPerSecond , throughputPerWorker = 17.51k samplesPerSecond
MPI Rank 0: 08/16/2016 03:20:14: Finished Epoch[ 5 of 5]: [Training] CrossEntropyWithSoftmax = 1.88963745 * 20480; EvalClassificationError = 0.51865234 * 20480; totalSamplesSeen = 102400; learningRatePerSample = 9.7656251e-005; epochTime=0.603172s
MPI Rank 0: 08/16/2016 03:20:14: SGD: Saving checkpoint model 'C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn'
MPI Rank 0: 08/16/2016 03:20:14: CNTKCommandTrainEnd: speechTrain
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:14: Action "train" complete.
MPI Rank 0: 
MPI Rank 0: 08/16/2016 03:20:14: __COMPLETED__
MPI Rank 0: ~MPIWrapper
MPI Rank 1: 08/16/2016 03:20:00: Redirecting stderr to file C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr_speechTrain.logrank1
MPI Rank 1: 08/16/2016 03:20:00: -------------------------------------------------------------------
MPI Rank 1: 08/16/2016 03:20:00: Build info: 
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:00: 		Built time: Aug 16 2016 03:09:16
MPI Rank 1: 08/16/2016 03:20:00: 		Last modified date: Fri Aug 12 05:28:23 2016
MPI Rank 1: 08/16/2016 03:20:00: 		Build type: Release
MPI Rank 1: 08/16/2016 03:20:00: 		Build target: GPU
MPI Rank 1: 08/16/2016 03:20:00: 		Math lib: mkl
MPI Rank 1: 08/16/2016 03:20:00: 		CUDA_PATH: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.5
MPI Rank 1: 08/16/2016 03:20:00: 		CUB_PATH: c:\src\cub-1.4.1
MPI Rank 1: 08/16/2016 03:20:00: 		CUDNN_PATH: c:\NVIDIA\cudnn-4.0\cuda
MPI Rank 1: 08/16/2016 03:20:00: 		Build Branch: HEAD
MPI Rank 1: 08/16/2016 03:20:00: 		Build SHA1: 026b1e772b963461e189f8f00aa7ed6951298f84
MPI Rank 1: 08/16/2016 03:20:00: 		Built by svcphil on Philly-Pool1
MPI Rank 1: 08/16/2016 03:20:00: 		Build Path: c:\jenkins\workspace\CNTK-Build-Windows\Source\CNTK\
MPI Rank 1: 08/16/2016 03:20:00: -------------------------------------------------------------------
MPI Rank 1: 08/16/2016 03:20:02: -------------------------------------------------------------------
MPI Rank 1: 08/16/2016 03:20:02: GPU info:
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: 		Device[0]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 1: 08/16/2016 03:20:02: 		Device[1]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 1: 08/16/2016 03:20:02: 		Device[2]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 1: 08/16/2016 03:20:02: 		Device[3]: cores = 2880; computeCapability = 3.5; type = "GeForce GTX 780 Ti"; memory = 3072 MB
MPI Rank 1: 08/16/2016 03:20:02: -------------------------------------------------------------------
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: Running on DPHAIM-24 at 2016/08/16 03:20:02
MPI Rank 1: 08/16/2016 03:20:02: Command line: 
MPI Rank 1: C:\jenkins\workspace\CNTK-Test-Windows-W1\x64\release\cntk.exe  configFile=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN/ParallelBM/cntk.cntk  currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data  RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu  DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data  ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN  OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu  DeviceId=0  timestamping=true  numCPUThreads=12  precision=double  speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]  stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: >>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
MPI Rank 1: 08/16/2016 03:20:02: precision = "float"
MPI Rank 1: command = speechTrain
MPI Rank 1: deviceId = $DeviceId$
MPI Rank 1: parallelTrain = true
MPI Rank 1: speechTrain = [
MPI Rank 1:     action = "train"
MPI Rank 1:     modelPath = "$RunDir$/models/cntkSpeech.dnn"
MPI Rank 1:     deviceId = $DeviceId$
MPI Rank 1:     traceLevel = 1
MPI Rank 1:     SimpleNetworkBuilder = [
MPI Rank 1:         layerSizes = 363:512:512:132
MPI Rank 1:         trainingCriterion = "CrossEntropyWithSoftmax"
MPI Rank 1:         evalCriterion = "ClassificationError"
MPI Rank 1:         layerTypes = "Sigmoid"
MPI Rank 1:         initValueScale = 1.0
MPI Rank 1:         applyMeanVarNorm = true
MPI Rank 1:         uniformInit = true
MPI Rank 1:         needPrior = true
MPI Rank 1:     ]
MPI Rank 1:     ExperimentalNetworkBuilder = [    // the same as above but with BS. Not active; activate by commenting out the SimpleNetworkBuilder entry above
MPI Rank 1:         layerSizes = 363:512:512:132
MPI Rank 1:         trainingCriterion = 'CE'
MPI Rank 1:         evalCriterion = 'Err'
MPI Rank 1:         applyMeanVarNorm = true
MPI Rank 1:         L = Length(layerSizes)-1    // number of model layers
MPI Rank 1:         features = Input(layerSizes[0], 1, tag='feature') ; labels = Input(layerSizes[Length(layerSizes)-1], 1, tag='label')
MPI Rank 1:         featNorm = if applyMeanVarNorm
MPI Rank 1:                    then MeanVarNorm(features)
MPI Rank 1:                    else features
MPI Rank 1:         layers[layer:1..L-1] = if layer > 1
MPI Rank 1:                                then SBFF(layers[layer-1].Eh, layerSizes[layer], layerSizes[layer-1])
MPI Rank 1:                                else SBFF(featNorm, layerSizes[layer], layerSizes[layer-1])
MPI Rank 1:         outLayer = BFF(layers[L-1].Eh, layerSizes[L], layerSizes[L-1])
MPI Rank 1:         outZ = outLayer.z        // + PastValue(layerSizes[L], 1, outLayer.z)
MPI Rank 1:         CE = if trainingCriterion == 'CE'
MPI Rank 1:              then CrossEntropyWithSoftmax(labels, outZ, tag='criterion')
MPI Rank 1:              else Fail('unknown trainingCriterion ' + trainingCriterion)
MPI Rank 1:         Err = if evalCriterion == 'Err' then
MPI Rank 1:               ClassificationError(labels, outZ, tag='evaluation')
MPI Rank 1:               else Fail('unknown evalCriterion ' + evalCriterion)
MPI Rank 1:         logPrior = LogPrior(labels)
MPI Rank 1:         // TODO: how to add a tag to an infix operation?
MPI Rank 1:         ScaledLogLikelihood = Minus (outZ, logPrior, tag='output')
MPI Rank 1:     ]
MPI Rank 1:     SGD = [
MPI Rank 1:         epochSize = 20480
MPI Rank 1:         minibatchSize = 64:256:1024
MPI Rank 1:         learningRatesPerMB = 1.0:0.5:0.1
MPI Rank 1:         numMBsToShowResult = 3
MPI Rank 1:         momentumPerMB = 0.9:0.656119
MPI Rank 1:         dropoutRate = 0.0
MPI Rank 1:         maxEpochs = 5
MPI Rank 1:         keepCheckPointFiles = true
MPI Rank 1:         clippingThresholdPerSample = 1#INF
MPI Rank 1:         ParallelTrain = [
MPI Rank 1:             parallelizationMethod = "BlockMomentumSGD"
MPI Rank 1:             distributedMBReading = true
MPI Rank 1:             syncPerfStats=1
MPI Rank 1:             BlockMomentumSGD = [
MPI Rank 1:                 blockSizePerWorker=2048
MPI Rank 1:                 resetSGDMomentum=true
MPI Rank 1:                 useNesterovMomentum=true
MPI Rank 1:             ]
MPI Rank 1:         ]
MPI Rank 1:         AutoAdjust = [
MPI Rank 1:             reduceLearnRateIfImproveLessThan = 0
MPI Rank 1:             loadBestModel = true
MPI Rank 1:             increaseLearnRateIfImproveMoreThan = 1000000000
MPI Rank 1:             learnRateDecreaseFactor = 0.5
MPI Rank 1:             learnRateIncreaseFactor = 1.382
MPI Rank 1:             autoAdjustLR = "adjustAfterEpoch"
MPI Rank 1:         ]
MPI Rank 1:     ]
MPI Rank 1:     reader = [
MPI Rank 1:         readerType = "HTKMLFReader"
MPI Rank 1:         readMethod = "blockRandomize"
MPI Rank 1:         miniBatchMode = "partial"
MPI Rank 1:         randomize = "auto"
MPI Rank 1:         verbosity = 0
MPI Rank 1:         useMersenneTwisterRand=true
MPI Rank 1:         features = [
MPI Rank 1:             dim = 363
MPI Rank 1:             type = "real"
MPI Rank 1:             scpFile = "glob_0000.scp"
MPI Rank 1:         ]
MPI Rank 1:         labels = [
MPI Rank 1:             mlfFile = "$DataDir$/glob_0000.mlf"
MPI Rank 1:             labelMappingFile = "$DataDir$/state.list"
MPI Rank 1:             labelDim = 132
MPI Rank 1:             labelType = "category"
MPI Rank 1:         ]
MPI Rank 1:     ]
MPI Rank 1: ]
MPI Rank 1: currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 1: RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 1: DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 1: ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN
MPI Rank 1: OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 1: DeviceId=0
MPI Rank 1: timestamping=true
MPI Rank 1: numCPUThreads=12
MPI Rank 1: precision=double
MPI Rank 1: speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]
MPI Rank 1: stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: <<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: >>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
MPI Rank 1: 08/16/2016 03:20:02: precision = "float"
MPI Rank 1: command = speechTrain
MPI Rank 1: deviceId = 0
MPI Rank 1: parallelTrain = true
MPI Rank 1: speechTrain = [
MPI Rank 1:     action = "train"
MPI Rank 1:     modelPath = "C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn"
MPI Rank 1:     deviceId = 0
MPI Rank 1:     traceLevel = 1
MPI Rank 1:     SimpleNetworkBuilder = [
MPI Rank 1:         layerSizes = 363:512:512:132
MPI Rank 1:         trainingCriterion = "CrossEntropyWithSoftmax"
MPI Rank 1:         evalCriterion = "ClassificationError"
MPI Rank 1:         layerTypes = "Sigmoid"
MPI Rank 1:         initValueScale = 1.0
MPI Rank 1:         applyMeanVarNorm = true
MPI Rank 1:         uniformInit = true
MPI Rank 1:         needPrior = true
MPI Rank 1:     ]
MPI Rank 1:     ExperimentalNetworkBuilder = [    // the same as above but with BS. Not active; activate by commenting out the SimpleNetworkBuilder entry above
MPI Rank 1:         layerSizes = 363:512:512:132
MPI Rank 1:         trainingCriterion = 'CE'
MPI Rank 1:         evalCriterion = 'Err'
MPI Rank 1:         applyMeanVarNorm = true
MPI Rank 1:         L = Length(layerSizes)-1    // number of model layers
MPI Rank 1:         features = Input(layerSizes[0], 1, tag='feature') ; labels = Input(layerSizes[Length(layerSizes)-1], 1, tag='label')
MPI Rank 1:         featNorm = if applyMeanVarNorm
MPI Rank 1:                    then MeanVarNorm(features)
MPI Rank 1:                    else features
MPI Rank 1:         layers[layer:1..L-1] = if layer > 1
MPI Rank 1:                                then SBFF(layers[layer-1].Eh, layerSizes[layer], layerSizes[layer-1])
MPI Rank 1:                                else SBFF(featNorm, layerSizes[layer], layerSizes[layer-1])
MPI Rank 1:         outLayer = BFF(layers[L-1].Eh, layerSizes[L], layerSizes[L-1])
MPI Rank 1:         outZ = outLayer.z        // + PastValue(layerSizes[L], 1, outLayer.z)
MPI Rank 1:         CE = if trainingCriterion == 'CE'
MPI Rank 1:              then CrossEntropyWithSoftmax(labels, outZ, tag='criterion')
MPI Rank 1:              else Fail('unknown trainingCriterion ' + trainingCriterion)
MPI Rank 1:         Err = if evalCriterion == 'Err' then
MPI Rank 1:               ClassificationError(labels, outZ, tag='evaluation')
MPI Rank 1:               else Fail('unknown evalCriterion ' + evalCriterion)
MPI Rank 1:         logPrior = LogPrior(labels)
MPI Rank 1:         // TODO: how to add a tag to an infix operation?
MPI Rank 1:         ScaledLogLikelihood = Minus (outZ, logPrior, tag='output')
MPI Rank 1:     ]
MPI Rank 1:     SGD = [
MPI Rank 1:         epochSize = 20480
MPI Rank 1:         minibatchSize = 64:256:1024
MPI Rank 1:         learningRatesPerMB = 1.0:0.5:0.1
MPI Rank 1:         numMBsToShowResult = 3
MPI Rank 1:         momentumPerMB = 0.9:0.656119
MPI Rank 1:         dropoutRate = 0.0
MPI Rank 1:         maxEpochs = 5
MPI Rank 1:         keepCheckPointFiles = true
MPI Rank 1:         clippingThresholdPerSample = 1#INF
MPI Rank 1:         ParallelTrain = [
MPI Rank 1:             parallelizationMethod = "BlockMomentumSGD"
MPI Rank 1:             distributedMBReading = true
MPI Rank 1:             syncPerfStats=1
MPI Rank 1:             BlockMomentumSGD = [
MPI Rank 1:                 blockSizePerWorker=2048
MPI Rank 1:                 resetSGDMomentum=true
MPI Rank 1:                 useNesterovMomentum=true
MPI Rank 1:             ]
MPI Rank 1:         ]
MPI Rank 1:         AutoAdjust = [
MPI Rank 1:             reduceLearnRateIfImproveLessThan = 0
MPI Rank 1:             loadBestModel = true
MPI Rank 1:             increaseLearnRateIfImproveMoreThan = 1000000000
MPI Rank 1:             learnRateDecreaseFactor = 0.5
MPI Rank 1:             learnRateIncreaseFactor = 1.382
MPI Rank 1:             autoAdjustLR = "adjustAfterEpoch"
MPI Rank 1:         ]
MPI Rank 1:     ]
MPI Rank 1:     reader = [
MPI Rank 1:         readerType = "HTKMLFReader"
MPI Rank 1:         readMethod = "blockRandomize"
MPI Rank 1:         miniBatchMode = "partial"
MPI Rank 1:         randomize = "auto"
MPI Rank 1:         verbosity = 0
MPI Rank 1:         useMersenneTwisterRand=true
MPI Rank 1:         features = [
MPI Rank 1:             dim = 363
MPI Rank 1:             type = "real"
MPI Rank 1:             scpFile = "glob_0000.scp"
MPI Rank 1:         ]
MPI Rank 1:         labels = [
MPI Rank 1:             mlfFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/glob_0000.mlf"
MPI Rank 1:             labelMappingFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/state.list"
MPI Rank 1:             labelDim = 132
MPI Rank 1:             labelType = "category"
MPI Rank 1:         ]
MPI Rank 1:     ]
MPI Rank 1: ]
MPI Rank 1: currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 1: RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 1: DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 1: ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN
MPI Rank 1: OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 1: DeviceId=0
MPI Rank 1: timestamping=true
MPI Rank 1: numCPUThreads=12
MPI Rank 1: precision=double
MPI Rank 1: speechTrain=[SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]
MPI Rank 1: stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: <<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: >>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
MPI Rank 1: configparameters: cntk.cntk:command=speechTrain
MPI Rank 1: configparameters: cntk.cntk:ConfigDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\DNN
MPI Rank 1: configparameters: cntk.cntk:currentDirectory=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 1: configparameters: cntk.cntk:DataDir=C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data
MPI Rank 1: configparameters: cntk.cntk:deviceId=0
MPI Rank 1: configparameters: cntk.cntk:numCPUThreads=12
MPI Rank 1: configparameters: cntk.cntk:OutputDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 1: configparameters: cntk.cntk:parallelTrain=true
MPI Rank 1: configparameters: cntk.cntk:precision=double
MPI Rank 1: configparameters: cntk.cntk:RunDir=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu
MPI Rank 1: configparameters: cntk.cntk:speechTrain=[
MPI Rank 1:     action = "train"
MPI Rank 1:     modelPath = "C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn"
MPI Rank 1:     deviceId = 0
MPI Rank 1:     traceLevel = 1
MPI Rank 1:     SimpleNetworkBuilder = [
MPI Rank 1:         layerSizes = 363:512:512:132
MPI Rank 1:         trainingCriterion = "CrossEntropyWithSoftmax"
MPI Rank 1:         evalCriterion = "ClassificationError"
MPI Rank 1:         layerTypes = "Sigmoid"
MPI Rank 1:         initValueScale = 1.0
MPI Rank 1:         applyMeanVarNorm = true
MPI Rank 1:         uniformInit = true
MPI Rank 1:         needPrior = true
MPI Rank 1:     ]
MPI Rank 1:     ExperimentalNetworkBuilder = [    // the same as above but with BS. Not active; activate by commenting out the SimpleNetworkBuilder entry above
MPI Rank 1:         layerSizes = 363:512:512:132
MPI Rank 1:         trainingCriterion = 'CE'
MPI Rank 1:         evalCriterion = 'Err'
MPI Rank 1:         applyMeanVarNorm = true
MPI Rank 1:         L = Length(layerSizes)-1    // number of model layers
MPI Rank 1:         features = Input(layerSizes[0], 1, tag='feature') ; labels = Input(layerSizes[Length(layerSizes)-1], 1, tag='label')
MPI Rank 1:         featNorm = if applyMeanVarNorm
MPI Rank 1:                    then MeanVarNorm(features)
MPI Rank 1:                    else features
MPI Rank 1:         layers[layer:1..L-1] = if layer > 1
MPI Rank 1:                                then SBFF(layers[layer-1].Eh, layerSizes[layer], layerSizes[layer-1])
MPI Rank 1:                                else SBFF(featNorm, layerSizes[layer], layerSizes[layer-1])
MPI Rank 1:         outLayer = BFF(layers[L-1].Eh, layerSizes[L], layerSizes[L-1])
MPI Rank 1:         outZ = outLayer.z        // + PastValue(layerSizes[L], 1, outLayer.z)
MPI Rank 1:         CE = if trainingCriterion == 'CE'
MPI Rank 1:              then CrossEntropyWithSoftmax(labels, outZ, tag='criterion')
MPI Rank 1:              else Fail('unknown trainingCriterion ' + trainingCriterion)
MPI Rank 1:         Err = if evalCriterion == 'Err' then
MPI Rank 1:               ClassificationError(labels, outZ, tag='evaluation')
MPI Rank 1:               else Fail('unknown evalCriterion ' + evalCriterion)
MPI Rank 1:         logPrior = LogPrior(labels)
MPI Rank 1:         // TODO: how to add a tag to an infix operation?
MPI Rank 1:         ScaledLogLikelihood = Minus (outZ, logPrior, tag='output')
MPI Rank 1:     ]
MPI Rank 1:     SGD = [
MPI Rank 1:         epochSize = 20480
MPI Rank 1:         minibatchSize = 64:256:1024
MPI Rank 1:         learningRatesPerMB = 1.0:0.5:0.1
MPI Rank 1:         numMBsToShowResult = 3
MPI Rank 1:         momentumPerMB = 0.9:0.656119
MPI Rank 1:         dropoutRate = 0.0
MPI Rank 1:         maxEpochs = 5
MPI Rank 1:         keepCheckPointFiles = true
MPI Rank 1:         clippingThresholdPerSample = 1#INF
MPI Rank 1:         ParallelTrain = [
MPI Rank 1:             parallelizationMethod = "BlockMomentumSGD"
MPI Rank 1:             distributedMBReading = true
MPI Rank 1:             syncPerfStats=1
MPI Rank 1:             BlockMomentumSGD = [
MPI Rank 1:                 blockSizePerWorker=2048
MPI Rank 1:                 resetSGDMomentum=true
MPI Rank 1:                 useNesterovMomentum=true
MPI Rank 1:             ]
MPI Rank 1:         ]
MPI Rank 1:         AutoAdjust = [
MPI Rank 1:             reduceLearnRateIfImproveLessThan = 0
MPI Rank 1:             loadBestModel = true
MPI Rank 1:             increaseLearnRateIfImproveMoreThan = 1000000000
MPI Rank 1:             learnRateDecreaseFactor = 0.5
MPI Rank 1:             learnRateIncreaseFactor = 1.382
MPI Rank 1:             autoAdjustLR = "adjustAfterEpoch"
MPI Rank 1:         ]
MPI Rank 1:     ]
MPI Rank 1:     reader = [
MPI Rank 1:         readerType = "HTKMLFReader"
MPI Rank 1:         readMethod = "blockRandomize"
MPI Rank 1:         miniBatchMode = "partial"
MPI Rank 1:         randomize = "auto"
MPI Rank 1:         verbosity = 0
MPI Rank 1:         useMersenneTwisterRand=true
MPI Rank 1:         features = [
MPI Rank 1:             dim = 363
MPI Rank 1:             type = "real"
MPI Rank 1:             scpFile = "glob_0000.scp"
MPI Rank 1:         ]
MPI Rank 1:         labels = [
MPI Rank 1:             mlfFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/glob_0000.mlf"
MPI Rank 1:             labelMappingFile = "C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/state.list"
MPI Rank 1:             labelDim = 132
MPI Rank 1:             labelType = "category"
MPI Rank 1:         ]
MPI Rank 1:     ]
MPI Rank 1: ] [SGD=[ParallelTrain=[parallelizationStartEpoch=2]]]
MPI Rank 1: 
MPI Rank 1: configparameters: cntk.cntk:stderr=C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/stderr
MPI Rank 1: configparameters: cntk.cntk:timestamping=true
MPI Rank 1: 08/16/2016 03:20:02: <<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
MPI Rank 1: 08/16/2016 03:20:02: Commands: speechTrain
MPI Rank 1: 08/16/2016 03:20:02: Precision = "double"
MPI Rank 1: 08/16/2016 03:20:02: Using 12 CPU threads.
MPI Rank 1: 08/16/2016 03:20:02: CNTKModelPath: C:\Users\svcphil\AppData\Local\Temp\cntk-test-20160816031852.202534\Speech\DNN_ParallelBM@release_gpu/models/cntkSpeech.dnn
MPI Rank 1: 08/16/2016 03:20:02: CNTKCommandTrainInfo: speechTrain : 5
MPI Rank 1: 08/16/2016 03:20:02: CNTKCommandTrainInfo: CNTKNoMoreCommands_Total : 5
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: ##############################################################################
MPI Rank 1: 08/16/2016 03:20:02: #                                                                            #
MPI Rank 1: 08/16/2016 03:20:02: # Action "train"                                                             #
MPI Rank 1: 08/16/2016 03:20:02: #                                                                            #
MPI Rank 1: 08/16/2016 03:20:02: ##############################################################################
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:02: CNTKCommandTrainBegin: speechTrain
MPI Rank 1: SimpleNetworkBuilder Using GPU 0
MPI Rank 1: reading script file glob_0000.scp ... 948 entries
MPI Rank 1: total 132 state names in state list C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/state.list
MPI Rank 1: htkmlfreader: reading MLF file C:\jenkins\workspace\CNTK-Test-Windows-W1\Tests\EndToEndTests\Speech\Data/glob_0000.mlf ... total 948 entries
MPI Rank 1: ...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
MPI Rank 1: label set 0: 129 classes
MPI Rank 1: minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: Creating virgin network.
MPI Rank 1: Node 'W0' (LearnableParameter operation): Initializing Parameter[512 x 363] <- 0.000000.
MPI Rank 1: Node 'W0' (LearnableParameter operation): Initializing Parameter[512 x 363] <- uniform(seed=1, range=0.050000*1.000000, onCPU=false).
MPI Rank 1: Microsoft::MSR::CNTK::GPUMatrix<ElemType>::SetUniformRandomValue (GPU): creating curand object with seed 1, sizeof(ElemType)==8
MPI Rank 1: Node 'B0' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 1: Node 'B0' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 1: Node 'W1' (LearnableParameter operation): Initializing Parameter[512 x 512] <- 0.000000.
MPI Rank 1: Node 'W1' (LearnableParameter operation): Initializing Parameter[512 x 512] <- uniform(seed=2, range=0.050000*1.000000, onCPU=false).
MPI Rank 1: Node 'B1' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 1: Node 'B1' (LearnableParameter operation): Initializing Parameter[512 x 1] <- 0.000000.
MPI Rank 1: Node 'W2' (LearnableParameter operation): Initializing Parameter[132 x 512] <- 0.000000.
MPI Rank 1: Node 'W2' (LearnableParameter operation): Initializing Parameter[132 x 512] <- uniform(seed=3, range=0.050000*1.000000, onCPU=false).
MPI Rank 1: Node 'B2' (LearnableParameter operation): Initializing Parameter[132 x 1] <- 0.000000.
MPI Rank 1: Node 'B2' (LearnableParameter operation): Initializing Parameter[132 x 1] <- 0.000000.
MPI Rank 1: 
MPI Rank 1: Post-processing network...
MPI Rank 1: 
MPI Rank 1: 7 roots:
MPI Rank 1: 	CrossEntropyWithSoftmax = CrossEntropyWithSoftmax()
MPI Rank 1: 	EvalClassificationError = ClassificationError()
MPI Rank 1: 	InvStdOfFeatures = InvStdDev()
MPI Rank 1: 	MeanOfFeatures = Mean()
MPI Rank 1: 	PosteriorProb = Softmax()
MPI Rank 1: 	Prior = Mean()
MPI Rank 1: 	ScaledLogLikelihood = Minus()
MPI Rank 1: 
MPI Rank 1: Validating network. 25 nodes to process in pass 1.
MPI Rank 1: 
MPI Rank 1: Validating --> labels = InputValue() :  -> [132 x *]
MPI Rank 1: Validating --> W2 = LearnableParameter() :  -> [132 x 512]
MPI Rank 1: Validating --> W1 = LearnableParameter() :  -> [512 x 512]
MPI Rank 1: Validating --> W0 = LearnableParameter() :  -> [512 x 363]
MPI Rank 1: Validating --> features = InputValue() :  -> [363 x *]
MPI Rank 1: Validating --> MeanOfFeatures = Mean (features) : [363 x *] -> [363]
MPI Rank 1: Validating --> InvStdOfFeatures = InvStdDev (features) : [363 x *] -> [363]
MPI Rank 1: Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization (features, MeanOfFeatures, InvStdOfFeatures) : [363 x *], [363], [363] -> [363 x *]
MPI Rank 1: Validating --> W0*features = Times (W0, MVNormalizedFeatures) : [512 x 363], [363 x *] -> [512 x *]
MPI Rank 1: Validating --> B0 = LearnableParameter() :  -> [512 x 1]
MPI Rank 1: Validating --> W0*features+B0 = Plus (W0*features, B0) : [512 x *], [512 x 1] -> [512 x 1 x *]
MPI Rank 1: Validating --> H1 = Sigmoid (W0*features+B0) : [512 x 1 x *] -> [512 x 1 x *]
MPI Rank 1: Validating --> W1*H1 = Times (W1, H1) : [512 x 512], [512 x 1 x *] -> [512 x 1 x *]
MPI Rank 1: Validating --> B1 = LearnableParameter() :  -> [512 x 1]
MPI Rank 1: Validating --> W1*H1+B1 = Plus (W1*H1, B1) : [512 x 1 x *], [512 x 1] -> [512 x 1 x *]
MPI Rank 1: Validating --> H2 = Sigmoid (W1*H1+B1) : [512 x 1 x *] -> [512 x 1 x *]
MPI Rank 1: Validating --> W2*H1 = Times (W2, H2) : [132 x 512], [512 x 1 x *] -> [132 x 1 x *]
MPI Rank 1: Validating --> B2 = LearnableParameter() :  -> [132 x 1]
MPI Rank 1: Validating --> HLast = Plus (W2*H1, B2) : [132 x 1 x *], [132 x 1] -> [132 x 1 x *]
MPI Rank 1: Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax (labels, HLast) : [132 x *], [132 x 1 x *] -> [1]
MPI Rank 1: Validating --> EvalClassificationError = ClassificationError (labels, HLast) : [132 x *], [132 x 1 x *] -> [1]
MPI Rank 1: Validating --> PosteriorProb = Softmax (HLast) : [132 x 1 x *] -> [132 x 1 x *]
MPI Rank 1: Validating --> Prior = Mean (labels) : [132 x *] -> [132]
MPI Rank 1: Validating --> LogOfPrior = Log (Prior) : [132] -> [132]
MPI Rank 1: Validating --> ScaledLogLikelihood = Minus (HLast, LogOfPrior) : [132 x 1 x *], [132] -> [132 x 1 x *]
MPI Rank 1: 
MPI Rank 1: Validating network. 17 nodes to process in pass 2.
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: Validating network, final pass.
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: 12 out of 25 nodes do not share the minibatch layout with the input data.
MPI Rank 1: 
MPI Rank 1: Post-processing network complete.
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: Created model with 25 nodes on GPU 0.
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: Training criterion node(s):
MPI Rank 1: 08/16/2016 03:20:03: 	CrossEntropyWithSoftmax = CrossEntropyWithSoftmax
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: Evaluation criterion node(s):
MPI Rank 1: 08/16/2016 03:20:03: 	EvalClassificationError = ClassificationError
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: Allocating matrices for forward and/or backward propagation.
MPI Rank 1: 
MPI Rank 1: Memory Sharing: Out of 40 matrices, 19 are shared as 8, and 21 are not shared.
MPI Rank 1: 
MPI Rank 1: 	{ W0 : [512 x 363] (gradient)
MPI Rank 1: 	  W0*features+B0 : [512 x 1 x *] }
MPI Rank 1: 	{ W1 : [512 x 512] (gradient)
MPI Rank 1: 	  W1*H1+B1 : [512 x 1 x *] }
MPI Rank 1: 	{ B0 : [512 x 1] (gradient)
MPI Rank 1: 	  H1 : [512 x 1 x *] (gradient)
MPI Rank 1: 	  W1*H1+B1 : [512 x 1 x *] (gradient)
MPI Rank 1: 	  W2*H1 : [132 x 1 x *] }
MPI Rank 1: 	{ H2 : [512 x 1 x *]
MPI Rank 1: 	  W1*H1 : [512 x 1 x *] (gradient) }
MPI Rank 1: 	{ HLast : [132 x 1 x *]
MPI Rank 1: 	  W2 : [132 x 512] (gradient) }
MPI Rank 1: 	{ B1 : [512 x 1] (gradient)
MPI Rank 1: 	  H2 : [512 x 1 x *] (gradient)
MPI Rank 1: 	  HLast : [132 x 1 x *] (gradient) }
MPI Rank 1: 	{ W0*features+B0 : [512 x 1 x *] (gradient)
MPI Rank 1: 	  W1*H1 : [512 x 1 x *] }
MPI Rank 1: 	{ H1 : [512 x 1 x *]
MPI Rank 1: 	  W0*features : [512 x *] (gradient) }
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: Training 516740 parameters in 6 out of 6 parameter tensors and 15 nodes with gradient:
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: 	Node 'B0' (LearnableParameter operation) : [512 x 1]
MPI Rank 1: 08/16/2016 03:20:03: 	Node 'B1' (LearnableParameter operation) : [512 x 1]
MPI Rank 1: 08/16/2016 03:20:03: 	Node 'B2' (LearnableParameter operation) : [132 x 1]
MPI Rank 1: 08/16/2016 03:20:03: 	Node 'W0' (LearnableParameter operation) : [512 x 363]
MPI Rank 1: 08/16/2016 03:20:03: 	Node 'W1' (LearnableParameter operation) : [512 x 512]
MPI Rank 1: 08/16/2016 03:20:03: 	Node 'W2' (LearnableParameter operation) : [132 x 512]
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: Precomputing --> 3 PreCompute nodes found.
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:03: 	MeanOfFeatures = Mean()
MPI Rank 1: 08/16/2016 03:20:03: 	InvStdOfFeatures = InvStdDev()
MPI Rank 1: 08/16/2016 03:20:03: 	Prior = Mean()
MPI Rank 1: minibatchiterator: epoch 0: frames [0..252734] (first utterance at frame 0), data subset 0 of 1, with 1 datapasses
MPI Rank 1: requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:09: Precomputing --> Completed.
MPI Rank 1: 
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:09: Starting Epoch 1: learning rate per sample = 0.015625  effective momentum = 0.900000  momentum as time constant = 607.4 samples
MPI Rank 1: minibatchiterator: epoch 0: frames [0..20480] (first utterance at frame 0), data subset 0 of 1, with 1 datapasses
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:09: Starting minibatch loop.
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[   1-   3, 0.94%]: CrossEntropyWithSoftmax = 4.68332137 * 192; EvalClassificationError = 0.98958333 * 192; time = 0.0394s; samplesPerSecond = 4877.4
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[   4-   6, 1.88%]: CrossEntropyWithSoftmax = 4.42217834 * 192; EvalClassificationError = 0.89062500 * 192; time = 0.0213s; samplesPerSecond = 9003.9
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[   7-   9, 2.81%]: CrossEntropyWithSoftmax = 4.78361173 * 192; EvalClassificationError = 0.93229167 * 192; time = 0.0217s; samplesPerSecond = 8834.9
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  10-  12, 3.75%]: CrossEntropyWithSoftmax = 4.28266373 * 192; EvalClassificationError = 0.90104167 * 192; time = 0.0220s; samplesPerSecond = 8730.8
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  13-  15, 4.69%]: CrossEntropyWithSoftmax = 4.48941255 * 192; EvalClassificationError = 0.89583333 * 192; time = 0.0218s; samplesPerSecond = 8809.4
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  16-  18, 5.63%]: CrossEntropyWithSoftmax = 4.39312485 * 192; EvalClassificationError = 0.94270833 * 192; time = 0.0217s; samplesPerSecond = 8858.5
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  19-  21, 6.56%]: CrossEntropyWithSoftmax = 4.26351131 * 192; EvalClassificationError = 0.98437500 * 192; time = 0.0214s; samplesPerSecond = 8985.4
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  22-  24, 7.50%]: CrossEntropyWithSoftmax = 4.06459300 * 192; EvalClassificationError = 0.90625000 * 192; time = 0.0217s; samplesPerSecond = 8860.6
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  25-  27, 8.44%]: CrossEntropyWithSoftmax = 3.97918086 * 192; EvalClassificationError = 0.90104167 * 192; time = 0.0208s; samplesPerSecond = 9241.0
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  28-  30, 9.38%]: CrossEntropyWithSoftmax = 3.83987406 * 192; EvalClassificationError = 0.85416667 * 192; time = 0.0206s; samplesPerSecond = 9337.2
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  31-  33, 10.31%]: CrossEntropyWithSoftmax = 3.79421365 * 192; EvalClassificationError = 0.84895833 * 192; time = 0.0205s; samplesPerSecond = 9367.7
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  34-  36, 11.25%]: CrossEntropyWithSoftmax = 3.76043875 * 192; EvalClassificationError = 0.86979167 * 192; time = 0.0205s; samplesPerSecond = 9377.7
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  37-  39, 12.19%]: CrossEntropyWithSoftmax = 3.73154819 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0205s; samplesPerSecond = 9370.9
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  40-  42, 13.13%]: CrossEntropyWithSoftmax = 3.68435113 * 192; EvalClassificationError = 0.83333333 * 192; time = 0.0205s; samplesPerSecond = 9360.4
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  43-  45, 14.06%]: CrossEntropyWithSoftmax = 3.89304289 * 192; EvalClassificationError = 0.90104167 * 192; time = 0.0205s; samplesPerSecond = 9371.8
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  46-  48, 15.00%]: CrossEntropyWithSoftmax = 3.76826980 * 192; EvalClassificationError = 0.86979167 * 192; time = 0.0206s; samplesPerSecond = 9301.0
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  49-  51, 15.94%]: CrossEntropyWithSoftmax = 3.81256010 * 192; EvalClassificationError = 0.91145833 * 192; time = 0.0205s; samplesPerSecond = 9371.3
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  52-  54, 16.88%]: CrossEntropyWithSoftmax = 3.92133036 * 192; EvalClassificationError = 0.87500000 * 192; time = 0.0205s; samplesPerSecond = 9381.0
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  55-  57, 17.81%]: CrossEntropyWithSoftmax = 3.56128851 * 192; EvalClassificationError = 0.84895833 * 192; time = 0.0205s; samplesPerSecond = 9369.1
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  58-  60, 18.75%]: CrossEntropyWithSoftmax = 3.67830123 * 192; EvalClassificationError = 0.86979167 * 192; time = 0.0203s; samplesPerSecond = 9436.7
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  61-  63, 19.69%]: CrossEntropyWithSoftmax = 3.36612487 * 192; EvalClassificationError = 0.76041667 * 192; time = 0.0205s; samplesPerSecond = 9370.9
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  64-  66, 20.63%]: CrossEntropyWithSoftmax = 3.41785793 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9363.1
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  67-  69, 21.56%]: CrossEntropyWithSoftmax = 3.37660990 * 192; EvalClassificationError = 0.79166667 * 192; time = 0.0205s; samplesPerSecond = 9384.6
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  70-  72, 22.50%]: CrossEntropyWithSoftmax = 3.68727319 * 192; EvalClassificationError = 0.83333333 * 192; time = 0.0203s; samplesPerSecond = 9439.1
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  73-  75, 23.44%]: CrossEntropyWithSoftmax = 3.64994161 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0205s; samplesPerSecond = 9354.4
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  76-  78, 24.38%]: CrossEntropyWithSoftmax = 3.37700933 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0204s; samplesPerSecond = 9434.9
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  79-  81, 25.31%]: CrossEntropyWithSoftmax = 3.51711918 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0204s; samplesPerSecond = 9397.0
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  82-  84, 26.25%]: CrossEntropyWithSoftmax = 3.47828208 * 192; EvalClassificationError = 0.80208333 * 192; time = 0.0205s; samplesPerSecond = 9369.1
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  85-  87, 27.19%]: CrossEntropyWithSoftmax = 3.54864536 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0205s; samplesPerSecond = 9369.5
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  88-  90, 28.13%]: CrossEntropyWithSoftmax = 3.43454656 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9369.1
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  91-  93, 29.06%]: CrossEntropyWithSoftmax = 3.36875092 * 192; EvalClassificationError = 0.79687500 * 192; time = 0.0204s; samplesPerSecond = 9393.8
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  94-  96, 30.00%]: CrossEntropyWithSoftmax = 3.55401747 * 192; EvalClassificationError = 0.82291667 * 192; time = 0.0205s; samplesPerSecond = 9376.8
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[  97-  99, 30.94%]: CrossEntropyWithSoftmax = 3.34809828 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9345.3
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 100- 102, 31.87%]: CrossEntropyWithSoftmax = 3.49450178 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0205s; samplesPerSecond = 9382.3
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 103- 105, 32.81%]: CrossEntropyWithSoftmax = 3.55445642 * 192; EvalClassificationError = 0.83333333 * 192; time = 0.0205s; samplesPerSecond = 9360.8
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 106- 108, 33.75%]: CrossEntropyWithSoftmax = 3.49293717 * 192; EvalClassificationError = 0.81770833 * 192; time = 0.0204s; samplesPerSecond = 9415.5
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 109- 111, 34.69%]: CrossEntropyWithSoftmax = 3.32494958 * 192; EvalClassificationError = 0.79687500 * 192; time = 0.0208s; samplesPerSecond = 9222.3
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 112- 114, 35.63%]: CrossEntropyWithSoftmax = 3.28851645 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0206s; samplesPerSecond = 9320.4
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 115- 117, 36.56%]: CrossEntropyWithSoftmax = 3.19411841 * 192; EvalClassificationError = 0.78125000 * 192; time = 0.0205s; samplesPerSecond = 9362.7
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 118- 120, 37.50%]: CrossEntropyWithSoftmax = 3.25028950 * 192; EvalClassificationError = 0.78125000 * 192; time = 0.0205s; samplesPerSecond = 9379.6
MPI Rank 1: 08/16/2016 03:20:09:  Epoch[ 1 of 5]-Minibatch[ 121- 123, 38.44%]: CrossEntropyWithSoftmax = 3.53445091 * 192; EvalClassificationError = 0.82812500 * 192; time = 0.0205s; samplesPerSecond = 9370.9
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 124- 126, 39.38%]: CrossEntropyWithSoftmax = 3.00326214 * 192; EvalClassificationError = 0.71875000 * 192; time = 0.0205s; samplesPerSecond = 9374.1
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 127- 129, 40.31%]: CrossEntropyWithSoftmax = 3.10787619 * 192; EvalClassificationError = 0.80729167 * 192; time = 0.0205s; samplesPerSecond = 9359.0
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 130- 132, 41.25%]: CrossEntropyWithSoftmax = 3.05588285 * 192; EvalClassificationError = 0.76562500 * 192; time = 0.0205s; samplesPerSecond = 9361.7
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 133- 135, 42.19%]: CrossEntropyWithSoftmax = 3.18197888 * 192; EvalClassificationError = 0.74479167 * 192; time = 0.0205s; samplesPerSecond = 9359.5
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 136- 138, 43.13%]: CrossEntropyWithSoftmax = 2.78138232 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0204s; samplesPerSecond = 9394.3
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 139- 141, 44.06%]: CrossEntropyWithSoftmax = 3.17441293 * 192; EvalClassificationError = 0.82291667 * 192; time = 0.0205s; samplesPerSecond = 9364.9
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 142- 144, 45.00%]: CrossEntropyWithSoftmax = 3.03537426 * 192; EvalClassificationError = 0.73437500 * 192; time = 0.0205s; samplesPerSecond = 9363.1
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 145- 147, 45.94%]: CrossEntropyWithSoftmax = 3.00595982 * 192; EvalClassificationError = 0.73958333 * 192; time = 0.0205s; samplesPerSecond = 9375.9
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 148- 150, 46.88%]: CrossEntropyWithSoftmax = 2.79115323 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0205s; samplesPerSecond = 9370.9
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 151- 153, 47.81%]: CrossEntropyWithSoftmax = 3.04097805 * 192; EvalClassificationError = 0.69791667 * 192; time = 0.0205s; samplesPerSecond = 9373.2
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 154- 156, 48.75%]: CrossEntropyWithSoftmax = 3.15935323 * 192; EvalClassificationError = 0.76562500 * 192; time = 0.0205s; samplesPerSecond = 9373.6
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 157- 159, 49.69%]: CrossEntropyWithSoftmax = 3.11947251 * 192; EvalClassificationError = 0.75000000 * 192; time = 0.0205s; samplesPerSecond = 9363.6
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 160- 162, 50.63%]: CrossEntropyWithSoftmax = 2.86232384 * 192; EvalClassificationError = 0.72395833 * 192; time = 0.0205s; samplesPerSecond = 9364.0
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 163- 165, 51.56%]: CrossEntropyWithSoftmax = 2.77742114 * 192; EvalClassificationError = 0.70312500 * 192; time = 0.0205s; samplesPerSecond = 9376.8
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 166- 168, 52.50%]: CrossEntropyWithSoftmax = 3.00411374 * 192; EvalClassificationError = 0.73958333 * 192; time = 0.0204s; samplesPerSecond = 9409.5
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 169- 171, 53.44%]: CrossEntropyWithSoftmax = 2.69740857 * 192; EvalClassificationError = 0.68750000 * 192; time = 0.0205s; samplesPerSecond = 9371.3
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 172- 174, 54.37%]: CrossEntropyWithSoftmax = 2.64948203 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0204s; samplesPerSecond = 9399.8
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 175- 177, 55.31%]: CrossEntropyWithSoftmax = 2.71417618 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0205s; samplesPerSecond = 9355.4
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 178- 180, 56.25%]: CrossEntropyWithSoftmax = 2.74281938 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0204s; samplesPerSecond = 9400.2
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 181- 183, 57.19%]: CrossEntropyWithSoftmax = 2.81346277 * 192; EvalClassificationError = 0.72916667 * 192; time = 0.0206s; samplesPerSecond = 9339.0
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 184- 186, 58.13%]: CrossEntropyWithSoftmax = 2.79862588 * 192; EvalClassificationError = 0.71875000 * 192; time = 0.0205s; samplesPerSecond = 9364.9
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 187- 189, 59.06%]: CrossEntropyWithSoftmax = 2.76655026 * 192; EvalClassificationError = 0.71354167 * 192; time = 0.0205s; samplesPerSecond = 9370.4
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 190- 192, 60.00%]: CrossEntropyWithSoftmax = 2.75908943 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0204s; samplesPerSecond = 9420.1
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 193- 195, 60.94%]: CrossEntropyWithSoftmax = 2.53548199 * 192; EvalClassificationError = 0.63541667 * 192; time = 0.0205s; samplesPerSecond = 9365.4
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 196- 198, 61.88%]: CrossEntropyWithSoftmax = 2.97589950 * 192; EvalClassificationError = 0.73437500 * 192; time = 0.0205s; samplesPerSecond = 9355.8
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 199- 201, 62.81%]: CrossEntropyWithSoftmax = 2.64996722 * 192; EvalClassificationError = 0.63020833 * 192; time = 0.0205s; samplesPerSecond = 9375.5
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 202- 204, 63.75%]: CrossEntropyWithSoftmax = 2.52128374 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0205s; samplesPerSecond = 9377.7
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 205- 207, 64.69%]: CrossEntropyWithSoftmax = 2.64228785 * 192; EvalClassificationError = 0.66145833 * 192; time = 0.0205s; samplesPerSecond = 9358.5
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 208- 210, 65.63%]: CrossEntropyWithSoftmax = 2.57199182 * 192; EvalClassificationError = 0.67708333 * 192; time = 0.0205s; samplesPerSecond = 9364.9
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 211- 213, 66.56%]: CrossEntropyWithSoftmax = 2.58100902 * 192; EvalClassificationError = 0.63020833 * 192; time = 0.0205s; samplesPerSecond = 9370.0
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 214- 216, 67.50%]: CrossEntropyWithSoftmax = 2.48555431 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0205s; samplesPerSecond = 9358.1
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 217- 219, 68.44%]: CrossEntropyWithSoftmax = 2.75336137 * 192; EvalClassificationError = 0.67187500 * 192; time = 0.0205s; samplesPerSecond = 9375.0
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 220- 222, 69.38%]: CrossEntropyWithSoftmax = 2.49193178 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0205s; samplesPerSecond = 9373.2
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 223- 225, 70.31%]: CrossEntropyWithSoftmax = 2.46098647 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0205s; samplesPerSecond = 9372.3
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 226- 228, 71.25%]: CrossEntropyWithSoftmax = 2.74322101 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0205s; samplesPerSecond = 9361.3
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 229- 231, 72.19%]: CrossEntropyWithSoftmax = 2.55837089 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0205s; samplesPerSecond = 9360.4
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 232- 234, 73.13%]: CrossEntropyWithSoftmax = 2.98288510 * 192; EvalClassificationError = 0.76562500 * 192; time = 0.0204s; samplesPerSecond = 9406.7
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 235- 237, 74.06%]: CrossEntropyWithSoftmax = 2.36667287 * 192; EvalClassificationError = 0.58854167 * 192; time = 0.0205s; samplesPerSecond = 9376.4
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 238- 240, 75.00%]: CrossEntropyWithSoftmax = 2.25169614 * 192; EvalClassificationError = 0.61458333 * 192; time = 0.0205s; samplesPerSecond = 9352.2
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 241- 243, 75.94%]: CrossEntropyWithSoftmax = 2.31564120 * 192; EvalClassificationError = 0.57291667 * 192; time = 0.0205s; samplesPerSecond = 9366.3
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 244- 246, 76.88%]: CrossEntropyWithSoftmax = 2.70894816 * 192; EvalClassificationError = 0.70833333 * 192; time = 0.0206s; samplesPerSecond = 9341.7
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 247- 249, 77.81%]: CrossEntropyWithSoftmax = 2.44991146 * 192; EvalClassificationError = 0.63020833 * 192; time = 0.0205s; samplesPerSecond = 9369.1
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 250- 252, 78.75%]: CrossEntropyWithSoftmax = 2.51856232 * 192; EvalClassificationError = 0.66666667 * 192; time = 0.0205s; samplesPerSecond = 9375.5
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 253- 255, 79.69%]: CrossEntropyWithSoftmax = 2.38498228 * 192; EvalClassificationError = 0.61979167 * 192; time = 0.0205s; samplesPerSecond = 9373.6
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 256- 258, 80.63%]: CrossEntropyWithSoftmax = 2.38080698 * 192; EvalClassificationError = 0.58333333 * 192; time = 0.0205s; samplesPerSecond = 9364.5
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 259- 261, 81.56%]: CrossEntropyWithSoftmax = 2.34294967 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0205s; samplesPerSecond = 9355.4
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 262- 264, 82.50%]: CrossEntropyWithSoftmax = 2.30340167 * 192; EvalClassificationError = 0.58854167 * 192; time = 0.0204s; samplesPerSecond = 9391.0
MPI Rank 1: 08/16/2016 03:20:10:  Epoch[ 1 of 5]-Minibatch[ 265- 267, 83.44%]: CrossEntropyWithSoftmax = 2.08323277 * 192; EvalClassificationError = 0.52604167 * 192; time = 0.0206s; samplesPerSecond = 9342.2
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 268- 270, 84.38%]: CrossEntropyWithSoftmax = 2.43589953 * 192; EvalClassificationError = 0.64583333 * 192; time = 0.0205s; samplesPerSecond = 9368.1
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 271- 273, 85.31%]: CrossEntropyWithSoftmax = 2.53399486 * 192; EvalClassificationError = 0.66145833 * 192; time = 0.0205s; samplesPerSecond = 9364.5
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 274- 276, 86.25%]: CrossEntropyWithSoftmax = 2.33995708 * 192; EvalClassificationError = 0.60416667 * 192; time = 0.0205s; samplesPerSecond = 9364.0
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 277- 279, 87.19%]: CrossEntropyWithSoftmax = 2.62970864 * 192; EvalClassificationError = 0.68229167 * 192; time = 0.0205s; samplesPerSecond = 9360.4
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 280- 282, 88.13%]: CrossEntropyWithSoftmax = 2.47993989 * 192; EvalClassificationError = 0.64062500 * 192; time = 0.0205s; samplesPerSecond = 9368.6
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 283- 285, 89.06%]: CrossEntropyWithSoftmax = 2.30935254 * 192; EvalClassificationError = 0.63541667 * 192; time = 0.0205s; samplesPerSecond = 9378.7
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 286- 288, 90.00%]: CrossEntropyWithSoftmax = 2.22022265 * 192; EvalClassificationError = 0.59375000 * 192; time = 0.0205s; samplesPerSecond = 9379.1
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 289- 291, 90.94%]: CrossEntropyWithSoftmax = 2.28060247 * 192; EvalClassificationError = 0.56770833 * 192; time = 0.0205s; samplesPerSecond = 9367.7
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 292- 294, 91.88%]: CrossEntropyWithSoftmax = 2.13349744 * 192; EvalClassificationError = 0.57291667 * 192; time = 0.0204s; samplesPerSecond = 9417.3
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 295- 297, 92.81%]: CrossEntropyWithSoftmax = 2.29751600 * 192; EvalClassificationError = 0.65104167 * 192; time = 0.0205s; samplesPerSecond = 9370.9
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 298- 300, 93.75%]: CrossEntropyWithSoftmax = 2.29319143 * 192; EvalClassificationError = 0.60416667 * 192; time = 0.0204s; samplesPerSecond = 9414.5
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 301- 303, 94.69%]: CrossEntropyWithSoftmax = 2.14551002 * 192; EvalClassificationError = 0.55729167 * 192; time = 0.0205s; samplesPerSecond = 9377.7
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 304- 306, 95.63%]: CrossEntropyWithSoftmax = 2.26930674 * 192; EvalClassificationError = 0.58333333 * 192; time = 0.0204s; samplesPerSecond = 9397.9
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 307- 309, 96.56%]: CrossEntropyWithSoftmax = 2.17383355 * 192; EvalClassificationError = 0.56770833 * 192; time = 0.0205s; samplesPerSecond = 9380.0
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 310- 312, 97.50%]: CrossEntropyWithSoftmax = 2.43111882 * 192; EvalClassificationError = 0.67187500 * 192; time = 0.0205s; samplesPerSecond = 9373.6
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 313- 315, 98.44%]: CrossEntropyWithSoftmax = 2.18011227 * 192; EvalClassificationError = 0.59895833 * 192; time = 0.0205s; samplesPerSecond = 9378.2
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 1 of 5]-Minibatch[ 316- 318, 99.38%]: CrossEntropyWithSoftmax = 2.21682707 * 192; EvalClassificationError = 0.56250000 * 192; time = 0.0174s; samplesPerSecond = 11050.4
MPI Rank 1: 08/16/2016 03:20:11: Finished Epoch[ 1 of 5]: [Training] CrossEntropyWithSoftmax = 3.03815141 * 20480; EvalClassificationError = 0.73432617 * 20480; totalSamplesSeen = 20480; learningRatePerSample = 0.015625; epochTime=2.25919s
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:11: Starting Epoch 2: learning rate per sample = 0.001953  effective momentum = 0.656119  momentum as time constant = 607.5 samples
MPI Rank 1: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 1: minibatchiterator: epoch 1: frames [20480..40960] (first utterance at frame 20480), data subset 1 of 2, with 1 datapasses
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:11: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[   1-   3, 3.75%]: CrossEntropyWithSoftmax = 2.19272896 * 260; EvalClassificationError = 0.61538462 * 260; time = 0.0647s; samplesPerSecond = 4017.3
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[   4-   6, 7.50%]: CrossEntropyWithSoftmax = 2.34922865 * 276; EvalClassificationError = 0.70289855 * 276; time = 0.0218s; samplesPerSecond = 12672.8
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[   7-   9, 11.25%]: CrossEntropyWithSoftmax = 2.24157888 * 280; EvalClassificationError = 0.67500000 * 280; time = 0.0264s; samplesPerSecond = 10620.9
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  10-  12, 15.00%]: CrossEntropyWithSoftmax = 2.20817680 * 241; EvalClassificationError = 0.62655602 * 241; time = 0.0164s; samplesPerSecond = 14707.7
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  13-  15, 18.75%]: CrossEntropyWithSoftmax = 2.04015325 * 295; EvalClassificationError = 0.52881356 * 295; time = 0.0175s; samplesPerSecond = 16844.6
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  16-  18, 22.50%]: CrossEntropyWithSoftmax = 2.40322904 * 257; EvalClassificationError = 0.63035019 * 257; time = 0.0127s; samplesPerSecond = 20217.1
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  19-  21, 26.25%]: CrossEntropyWithSoftmax = 2.04484358 * 262; EvalClassificationError = 0.56488550 * 262; time = 0.0128s; samplesPerSecond = 20502.4
MPI Rank 1: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 1-th sync:     0.24 seconds since last report (0.01 seconds on comm.); 4289 samples processed by 2 workers (2126 by me);
MPI Rank 1: 		(model aggregation stats) 1-th sync: totalThroughput = 18.06k samplesPerSecond , throughputPerWorker = 9.03k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  22-  24, 30.00%]: CrossEntropyWithSoftmax = 2.03646309 * 255; EvalClassificationError = 0.53333333 * 255; time = 0.0503s; samplesPerSecond = 5068.3
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  25-  27, 33.75%]: CrossEntropyWithSoftmax = 2.07321525 * 279; EvalClassificationError = 0.58781362 * 279; time = 0.0222s; samplesPerSecond = 12592.0
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  28-  30, 37.50%]: CrossEntropyWithSoftmax = 2.06339373 * 274; EvalClassificationError = 0.52554745 * 274; time = 0.0269s; samplesPerSecond = 10199.9
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  31-  33, 41.25%]: CrossEntropyWithSoftmax = 1.95530128 * 269; EvalClassificationError = 0.53531599 * 269; time = 0.0267s; samplesPerSecond = 10070.8
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  34-  36, 45.00%]: CrossEntropyWithSoftmax = 2.03493597 * 278; EvalClassificationError = 0.61151079 * 278; time = 0.0267s; samplesPerSecond = 10416.3
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  37-  39, 48.75%]: CrossEntropyWithSoftmax = 2.13309941 * 271; EvalClassificationError = 0.54612546 * 271; time = 0.0214s; samplesPerSecond = 12692.0
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  40-  42, 52.50%]: CrossEntropyWithSoftmax = 2.19775096 * 276; EvalClassificationError = 0.58333333 * 276; time = 0.0129s; samplesPerSecond = 21435.2
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  43-  45, 56.25%]: CrossEntropyWithSoftmax = 2.14623378 * 260; EvalClassificationError = 0.58076923 * 260; time = 0.0127s; samplesPerSecond = 20496.6
MPI Rank 1: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 2-th sync:     0.18 seconds since last report (0.01 seconds on comm.); 4253 samples processed by 2 workers (2073 by me);
MPI Rank 1: 		(model aggregation stats) 2-th sync: totalThroughput = 23.55k samplesPerSecond , throughputPerWorker = 11.77k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  46-  48, 60.00%]: CrossEntropyWithSoftmax = 1.98115282 * 265; EvalClassificationError = 0.54716981 * 265; time = 0.0368s; samplesPerSecond = 7192.1
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  49-  51, 63.75%]: CrossEntropyWithSoftmax = 2.13033706 * 298; EvalClassificationError = 0.57718121 * 298; time = 0.0273s; samplesPerSecond = 10911.8
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  52-  54, 67.50%]: CrossEntropyWithSoftmax = 1.96671704 * 274; EvalClassificationError = 0.50364964 * 274; time = 0.0269s; samplesPerSecond = 10169.2
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  55-  57, 71.25%]: CrossEntropyWithSoftmax = 1.98515695 * 265; EvalClassificationError = 0.53962264 * 265; time = 0.0267s; samplesPerSecond = 9929.9
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  58-  60, 75.00%]: CrossEntropyWithSoftmax = 1.85824670 * 281; EvalClassificationError = 0.54448399 * 281; time = 0.0148s; samplesPerSecond = 18949.4
MPI Rank 1: 08/16/2016 03:20:11:  Epoch[ 2 of 5]-Minibatch[  61-  63, 78.75%]: CrossEntropyWithSoftmax = 1.95611759 * 252; EvalClassificationError = 0.52777778 * 252; time = 0.0207s; samplesPerSecond = 12163.9
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  64-  66, 82.50%]: CrossEntropyWithSoftmax = 2.04102834 * 274; EvalClassificationError = 0.51094891 * 274; time = 0.0166s; samplesPerSecond = 16512.0
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  67-  69, 86.25%]: CrossEntropyWithSoftmax = 1.84439304 * 258; EvalClassificationError = 0.49224806 * 258; time = 0.0126s; samplesPerSecond = 20458.3
MPI Rank 1: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 3-th sync:     0.19 seconds since last report (0.01 seconds on comm.); 4246 samples processed by 2 workers (2102 by me);
MPI Rank 1: 		(model aggregation stats) 3-th sync: totalThroughput = 22.76k samplesPerSecond , throughputPerWorker = 11.38k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  70-  72, 90.00%]: CrossEntropyWithSoftmax = 1.97159866 * 271; EvalClassificationError = 0.55719557 * 271; time = 0.0365s; samplesPerSecond = 7430.2
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  73-  75, 93.75%]: CrossEntropyWithSoftmax = 1.97632511 * 278; EvalClassificationError = 0.52877698 * 278; time = 0.0266s; samplesPerSecond = 10459.0
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  76-  78, 97.50%]: CrossEntropyWithSoftmax = 1.95095600 * 286; EvalClassificationError = 0.56293706 * 286; time = 0.0268s; samplesPerSecond = 10683.6
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 2 of 5]-Minibatch[  79-  81, 101.25%]: CrossEntropyWithSoftmax = 1.94334189 * 170; EvalClassificationError = 0.52941176 * 170; time = 0.0221s; samplesPerSecond = 7693.4
MPI Rank 1: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.02-seconds latency this time; accumulated time on sync point = 0.02 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 4-th sync:     0.30 seconds since last report (0.18 seconds on comm.); 7692 samples processed by 2 workers (904 by me);
MPI Rank 1: 		(model aggregation stats) 4-th sync: totalThroughput = 25.25k samplesPerSecond , throughputPerWorker = 12.62k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:12: Finished Epoch[ 2 of 5]: [Training] CrossEntropyWithSoftmax = 2.05172118 * 20480; EvalClassificationError = 0.55805664 * 20480; totalSamplesSeen = 40960; learningRatePerSample = 0.001953125; epochTime=0.91203s
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:12: Starting Epoch 3: learning rate per sample = 0.000098  effective momentum = 0.656119  momentum as time constant = 2429.9 samples
MPI Rank 1: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 1: minibatchiterator: epoch 2: frames [40960..61440] (first utterance at frame 40960), data subset 1 of 2, with 1 datapasses
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:12: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[   1-   3, 15.00%]: CrossEntropyWithSoftmax = 1.94144328 * 1130; EvalClassificationError = 0.53097345 * 1130; time = 0.1167s; samplesPerSecond = 9679.1
MPI Rank 1: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 1-th sync:     0.21 seconds since last report (0.03 seconds on comm.); 4885 samples processed by 2 workers (2293 by me);
MPI Rank 1: 		(model aggregation stats) 1-th sync: totalThroughput = 23.56k samplesPerSecond , throughputPerWorker = 11.78k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[   4-   6, 30.00%]: CrossEntropyWithSoftmax = 1.87694957 * 1163; EvalClassificationError = 0.52794497 * 1163; time = 0.0816s; samplesPerSecond = 14256.6
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[   7-   9, 45.00%]: CrossEntropyWithSoftmax = 1.98228580 * 1085; EvalClassificationError = 0.52718894 * 1085; time = 0.0712s; samplesPerSecond = 15246.5
MPI Rank 1: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.02-seconds latency this time; accumulated time on sync point = 0.02 seconds , average latency = 0.01 seconds
MPI Rank 1: 		(model aggregation stats) 2-th sync:     0.15 seconds since last report (0.01 seconds on comm.); 4826 samples processed by 2 workers (2249 by me);
MPI Rank 1: 		(model aggregation stats) 2-th sync: totalThroughput = 33.08k samplesPerSecond , throughputPerWorker = 16.54k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[  10-  12, 60.00%]: CrossEntropyWithSoftmax = 1.92865528 * 1164; EvalClassificationError = 0.54725086 * 1164; time = 0.0737s; samplesPerSecond = 15788.6
MPI Rank 1: 08/16/2016 03:20:12:  Epoch[ 3 of 5]-Minibatch[  13-  15, 75.00%]: CrossEntropyWithSoftmax = 2.00184241 * 1167; EvalClassificationError = 0.57155099 * 1167; time = 0.0787s; samplesPerSecond = 14830.0
MPI Rank 1: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.02 seconds , average latency = 0.01 seconds
MPI Rank 1: 		(model aggregation stats) 3-th sync:     0.15 seconds since last report (0.02 seconds on comm.); 4903 samples processed by 2 workers (2326 by me);
MPI Rank 1: 		(model aggregation stats) 3-th sync: totalThroughput = 32.79k samplesPerSecond , throughputPerWorker = 16.39k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 3 of 5]-Minibatch[  16-  18, 90.00%]: CrossEntropyWithSoftmax = 1.92549525 * 1159; EvalClassificationError = 0.54270923 * 1159; time = 0.0701s; samplesPerSecond = 16542.0
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 3 of 5]-Minibatch[  19-  21, 105.00%]: CrossEntropyWithSoftmax = 1.98011842 * 823; EvalClassificationError = 0.54921021 * 823; time = 0.0835s; samplesPerSecond = 9857.7
MPI Rank 1: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.02 seconds , average latency = 0.01 seconds
MPI Rank 1: 		(model aggregation stats) 4-th sync:     0.17 seconds since last report (0.07 seconds on comm.); 5866 samples processed by 2 workers (823 by me);
MPI Rank 1: 		(model aggregation stats) 4-th sync: totalThroughput = 34.21k samplesPerSecond , throughputPerWorker = 17.10k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:13: Finished Epoch[ 3 of 5]: [Training] CrossEntropyWithSoftmax = 1.95703393 * 20480; EvalClassificationError = 0.54541016 * 20480; totalSamplesSeen = 61440; learningRatePerSample = 9.7656251e-005; epochTime=0.67636s
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:13: Starting Epoch 4: learning rate per sample = 0.000098  effective momentum = 0.656119  momentum as time constant = 2429.9 samples
MPI Rank 1: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 1: minibatchiterator: epoch 3: frames [61440..81920] (first utterance at frame 61440), data subset 1 of 2, with 1 datapasses
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:13: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[   1-   3, 15.00%]: CrossEntropyWithSoftmax = 1.89485201 * 1149; EvalClassificationError = 0.53176675 * 1149; time = 0.0931s; samplesPerSecond = 12338.1
MPI Rank 1: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 1-th sync:     0.17 seconds since last report (0.01 seconds on comm.); 4901 samples processed by 2 workers (2351 by me);
MPI Rank 1: 		(model aggregation stats) 1-th sync: totalThroughput = 29.68k samplesPerSecond , throughputPerWorker = 14.84k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[   4-   6, 30.00%]: CrossEntropyWithSoftmax = 1.94072250 * 1202; EvalClassificationError = 0.53826955 * 1202; time = 0.0660s; samplesPerSecond = 18218.5
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[   7-   9, 45.00%]: CrossEntropyWithSoftmax = 1.90064937 * 1130; EvalClassificationError = 0.52300885 * 1130; time = 0.0666s; samplesPerSecond = 16970.3
MPI Rank 1: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 2-th sync:     0.13 seconds since last report (0.00 seconds on comm.); 4836 samples processed by 2 workers (2317 by me);
MPI Rank 1: 		(model aggregation stats) 2-th sync: totalThroughput = 38.20k samplesPerSecond , throughputPerWorker = 19.10k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[  10-  12, 60.00%]: CrossEntropyWithSoftmax = 1.85577719 * 1187; EvalClassificationError = 0.50884583 * 1187; time = 0.0591s; samplesPerSecond = 20098.2
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[  13-  15, 75.00%]: CrossEntropyWithSoftmax = 1.94097997 * 1202; EvalClassificationError = 0.54658902 * 1202; time = 0.0818s; samplesPerSecond = 14687.2
MPI Rank 1: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 3-th sync:     0.15 seconds since last report (0.01 seconds on comm.); 4952 samples processed by 2 workers (2401 by me);
MPI Rank 1: 		(model aggregation stats) 3-th sync: totalThroughput = 33.32k samplesPerSecond , throughputPerWorker = 16.66k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[  16-  18, 90.00%]: CrossEntropyWithSoftmax = 1.91747174 * 1199; EvalClassificationError = 0.54211843 * 1199; time = 0.0657s; samplesPerSecond = 18254.9
MPI Rank 1: 08/16/2016 03:20:13:  Epoch[ 4 of 5]-Minibatch[  19-  21, 105.00%]: CrossEntropyWithSoftmax = 1.91766783 * 817; EvalClassificationError = 0.53243574 * 817; time = 0.0473s; samplesPerSecond = 17271.6
MPI Rank 1: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.02-seconds latency this time; accumulated time on sync point = 0.02 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 4-th sync:     0.16 seconds since last report (0.07 seconds on comm.); 5791 samples processed by 2 workers (817 by me);
MPI Rank 1: 		(model aggregation stats) 4-th sync: totalThroughput = 35.86k samplesPerSecond , throughputPerWorker = 17.93k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:14: Finished Epoch[ 4 of 5]: [Training] CrossEntropyWithSoftmax = 1.90306770 * 20480; EvalClassificationError = 0.52641602 * 20480; totalSamplesSeen = 81920; learningRatePerSample = 9.7656251e-005; epochTime=0.604332s
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:14: Starting Epoch 5: learning rate per sample = 0.000098  effective momentum = 0.656119  momentum as time constant = 2429.9 samples
MPI Rank 1: Parallel training (2 workers) using BlockMomentumSGD with block momentum = 0.5000, block momentum time constant (per worker) = 2954.6394, block learning rate = 1.0000, block size per worker = 2048 samples, using Nesterov-style block momentum, resetting SGD momentum after sync.
MPI Rank 1: minibatchiterator: epoch 4: frames [81920..102400] (first utterance at frame 81920), data subset 1 of 2, with 1 datapasses
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:14: Starting minibatch loop, distributed reading is ENABLED.
MPI Rank 1: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[   1-   3, 15.00%]: CrossEntropyWithSoftmax = 1.86643684 * 1209; EvalClassificationError = 0.50372208 * 1209; time = 0.0891s; samplesPerSecond = 13571.8
MPI Rank 1: 		(model aggregation stats): 1-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 1-th sync:     0.17 seconds since last report (0.01 seconds on comm.); 4919 samples processed by 2 workers (2426 by me);
MPI Rank 1: 		(model aggregation stats) 1-th sync: totalThroughput = 29.63k samplesPerSecond , throughputPerWorker = 14.82k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[   4-   6, 30.00%]: CrossEntropyWithSoftmax = 1.94031579 * 1217; EvalClassificationError = 0.53327855 * 1217; time = 0.0696s; samplesPerSecond = 17488.1
MPI Rank 1: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[   7-   9, 45.00%]: CrossEntropyWithSoftmax = 1.89794045 * 1206; EvalClassificationError = 0.51824212 * 1206; time = 0.0600s; samplesPerSecond = 20116.1
MPI Rank 1: 		(model aggregation stats): 2-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 2-th sync:     0.14 seconds since last report (0.00 seconds on comm.); 4899 samples processed by 2 workers (2419 by me);
MPI Rank 1: 		(model aggregation stats) 2-th sync: totalThroughput = 35.49k samplesPerSecond , throughputPerWorker = 17.75k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  10-  12, 60.00%]: CrossEntropyWithSoftmax = 1.96415395 * 1213; EvalClassificationError = 0.53833471 * 1213; time = 0.0770s; samplesPerSecond = 15745.7
MPI Rank 1: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  13-  15, 75.00%]: CrossEntropyWithSoftmax = 1.84516499 * 1147; EvalClassificationError = 0.51264167 * 1147; time = 0.0532s; samplesPerSecond = 21569.1
MPI Rank 1: 		(model aggregation stats): 3-th sync point was hit, introducing a 0.00-seconds latency this time; accumulated time on sync point = 0.00 seconds , average latency = 0.00 seconds
MPI Rank 1: 		(model aggregation stats) 3-th sync:     0.13 seconds since last report (0.00 seconds on comm.); 4829 samples processed by 2 workers (2359 by me);
MPI Rank 1: 		(model aggregation stats) 3-th sync: totalThroughput = 37.07k samplesPerSecond , throughputPerWorker = 18.54k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  16-  18, 90.00%]: CrossEntropyWithSoftmax = 2.02566421 * 1212; EvalClassificationError = 0.55280528 * 1212; time = 0.0762s; samplesPerSecond = 15902.8
MPI Rank 1: 08/16/2016 03:20:14:  Epoch[ 5 of 5]-Minibatch[  19-  21, 105.00%]: CrossEntropyWithSoftmax = 1.89201183 * 809; EvalClassificationError = 0.50679852 * 809; time = 0.0378s; samplesPerSecond = 21404.4
MPI Rank 1: 		(model aggregation stats): 4-th sync point was hit, introducing a 0.03-seconds latency this time; accumulated time on sync point = 0.03 seconds , average latency = 0.01 seconds
MPI Rank 1: 		(model aggregation stats) 4-th sync:     0.17 seconds since last report (0.07 seconds on comm.); 5833 samples processed by 2 workers (809 by me);
MPI Rank 1: 		(model aggregation stats) 4-th sync: totalThroughput = 34.97k samplesPerSecond , throughputPerWorker = 17.49k samplesPerSecond
MPI Rank 1: 08/16/2016 03:20:14: Finished Epoch[ 5 of 5]: [Training] CrossEntropyWithSoftmax = 1.88963745 * 20480; EvalClassificationError = 0.51865234 * 20480; totalSamplesSeen = 102400; learningRatePerSample = 9.7656251e-005; epochTime=0.603086s
MPI Rank 1: 08/16/2016 03:20:14: CNTKCommandTrainEnd: speechTrain
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:14: Action "train" complete.
MPI Rank 1: 
MPI Rank 1: 08/16/2016 03:20:14: __COMPLETED__
MPI Rank 1: ~MPIWrapper