Skip to content

S
s5_cet_baseline_release
Switch branch/tag
Find file
Normal viewHistoryPermalink
final.config 19.7 KB
EditWeb IDE
Newer Older
Initial commit
1dd607be
 
ntut committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 
# This file was created by the command:
# steps/nnet3/xconfig_to_configs.py --xconfig-file exp/chain/tdnn_1d_sp/configs/network.xconfig --config-dir exp/chain/tdnn_1d_sp/configs/
# It contains the entire neural network.

input-node name=ivector dim=100
input-node name=input dim=43
component name=lda type=FixedAffineComponent matrix=exp/chain/tdnn_1d_sp/configs/lda.mat
component-node name=lda component=lda input=Append(Offset(input, -1), input, Offset(input, 1), ReplaceIndex(ivector, t, 0))
component name=tdnn1.affine type=NaturalGradientAffineComponent input-dim=229 output-dim=1536  max-change=0.75 l2-regularize=0.01
component-node name=tdnn1.affine component=tdnn1.affine input=lda
component name=tdnn1.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnn1.relu component=tdnn1.relu input=tdnn1.affine
component name=tdnn1.batchnorm type=BatchNormComponent dim=1536 target-rms=1.0
component-node name=tdnn1.batchnorm component=tdnn1.batchnorm input=tdnn1.relu
component name=tdnn1.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnn1.dropout component=tdnn1.dropout input=tdnn1.batchnorm
component name=tdnnf2.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-1,0 orthonormal-constraint=-1.0
component-node name=tdnnf2.linear component=tdnnf2.linear input=tdnn1.dropout
component name=tdnnf2.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,1
component-node name=tdnnf2.affine component=tdnnf2.affine input=tdnnf2.linear
component name=tdnnf2.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf2.relu component=tdnnf2.relu input=tdnnf2.affine
component name=tdnnf2.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf2.batchnorm component=tdnnf2.batchnorm input=tdnnf2.relu
component name=tdnnf2.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf2.dropout component=tdnnf2.dropout input=tdnnf2.batchnorm
component name=tdnnf2.noop type=NoOpComponent dim=1536
component-node name=tdnnf2.noop component=tdnnf2.noop input=Sum(Scale(0.66, tdnn1.dropout), tdnnf2.dropout)
component name=tdnnf3.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-1,0 orthonormal-constraint=-1.0
component-node name=tdnnf3.linear component=tdnnf3.linear input=tdnnf2.noop
component name=tdnnf3.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,1
component-node name=tdnnf3.affine component=tdnnf3.affine input=tdnnf3.linear
component name=tdnnf3.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf3.relu component=tdnnf3.relu input=tdnnf3.affine
component name=tdnnf3.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf3.batchnorm component=tdnnf3.batchnorm input=tdnnf3.relu
component name=tdnnf3.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf3.dropout component=tdnnf3.dropout input=tdnnf3.batchnorm
component name=tdnnf3.noop type=NoOpComponent dim=1536
component-node name=tdnnf3.noop component=tdnnf3.noop input=Sum(Scale(0.66, tdnnf2.noop), tdnnf3.dropout)
component name=tdnnf4.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-1,0 orthonormal-constraint=-1.0
component-node name=tdnnf4.linear component=tdnnf4.linear input=tdnnf3.noop
component name=tdnnf4.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,1
component-node name=tdnnf4.affine component=tdnnf4.affine input=tdnnf4.linear
component name=tdnnf4.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf4.relu component=tdnnf4.relu input=tdnnf4.affine
component name=tdnnf4.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf4.batchnorm component=tdnnf4.batchnorm input=tdnnf4.relu
component name=tdnnf4.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf4.dropout component=tdnnf4.dropout input=tdnnf4.batchnorm
component name=tdnnf4.noop type=NoOpComponent dim=1536
component-node name=tdnnf4.noop component=tdnnf4.noop input=Sum(Scale(0.66, tdnnf3.noop), tdnnf4.dropout)
component name=tdnnf5.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=0 orthonormal-constraint=-1.0
component-node name=tdnnf5.linear component=tdnnf5.linear input=tdnnf4.noop
component name=tdnnf5.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0
component-node name=tdnnf5.affine component=tdnnf5.affine input=tdnnf5.linear
component name=tdnnf5.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf5.relu component=tdnnf5.relu input=tdnnf5.affine
component name=tdnnf5.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf5.batchnorm component=tdnnf5.batchnorm input=tdnnf5.relu
component name=tdnnf5.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf5.dropout component=tdnnf5.dropout input=tdnnf5.batchnorm
component name=tdnnf5.noop type=NoOpComponent dim=1536
component-node name=tdnnf5.noop component=tdnnf5.noop input=Sum(Scale(0.66, tdnnf4.noop), tdnnf5.dropout)
component name=tdnnf6.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf6.linear component=tdnnf6.linear input=tdnnf5.noop
component name=tdnnf6.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf6.affine component=tdnnf6.affine input=tdnnf6.linear
component name=tdnnf6.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf6.relu component=tdnnf6.relu input=tdnnf6.affine
component name=tdnnf6.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf6.batchnorm component=tdnnf6.batchnorm input=tdnnf6.relu
component name=tdnnf6.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf6.dropout component=tdnnf6.dropout input=tdnnf6.batchnorm
component name=tdnnf6.noop type=NoOpComponent dim=1536
component-node name=tdnnf6.noop component=tdnnf6.noop input=Sum(Scale(0.66, tdnnf5.noop), tdnnf6.dropout)
component name=tdnnf7.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf7.linear component=tdnnf7.linear input=tdnnf6.noop
component name=tdnnf7.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf7.affine component=tdnnf7.affine input=tdnnf7.linear
component name=tdnnf7.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf7.relu component=tdnnf7.relu input=tdnnf7.affine
component name=tdnnf7.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf7.batchnorm component=tdnnf7.batchnorm input=tdnnf7.relu
component name=tdnnf7.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf7.dropout component=tdnnf7.dropout input=tdnnf7.batchnorm
component name=tdnnf7.noop type=NoOpComponent dim=1536
component-node name=tdnnf7.noop component=tdnnf7.noop input=Sum(Scale(0.66, tdnnf6.noop), tdnnf7.dropout)
component name=tdnnf8.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf8.linear component=tdnnf8.linear input=tdnnf7.noop
component name=tdnnf8.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf8.affine component=tdnnf8.affine input=tdnnf8.linear
component name=tdnnf8.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf8.relu component=tdnnf8.relu input=tdnnf8.affine
component name=tdnnf8.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf8.batchnorm component=tdnnf8.batchnorm input=tdnnf8.relu
component name=tdnnf8.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf8.dropout component=tdnnf8.dropout input=tdnnf8.batchnorm
component name=tdnnf8.noop type=NoOpComponent dim=1536
component-node name=tdnnf8.noop component=tdnnf8.noop input=Sum(Scale(0.66, tdnnf7.noop), tdnnf8.dropout)
component name=tdnnf9.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf9.linear component=tdnnf9.linear input=tdnnf8.noop
component name=tdnnf9.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf9.affine component=tdnnf9.affine input=tdnnf9.linear
component name=tdnnf9.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf9.relu component=tdnnf9.relu input=tdnnf9.affine
component name=tdnnf9.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf9.batchnorm component=tdnnf9.batchnorm input=tdnnf9.relu
component name=tdnnf9.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf9.dropout component=tdnnf9.dropout input=tdnnf9.batchnorm
component name=tdnnf9.noop type=NoOpComponent dim=1536
component-node name=tdnnf9.noop component=tdnnf9.noop input=Sum(Scale(0.66, tdnnf8.noop), tdnnf9.dropout)
component name=tdnnf10.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf10.linear component=tdnnf10.linear input=tdnnf9.noop
component name=tdnnf10.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf10.affine component=tdnnf10.affine input=tdnnf10.linear
component name=tdnnf10.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf10.relu component=tdnnf10.relu input=tdnnf10.affine
component name=tdnnf10.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf10.batchnorm component=tdnnf10.batchnorm input=tdnnf10.relu
component name=tdnnf10.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf10.dropout component=tdnnf10.dropout input=tdnnf10.batchnorm
component name=tdnnf10.noop type=NoOpComponent dim=1536
component-node name=tdnnf10.noop component=tdnnf10.noop input=Sum(Scale(0.66, tdnnf9.noop), tdnnf10.dropout)
component name=tdnnf11.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf11.linear component=tdnnf11.linear input=tdnnf10.noop
component name=tdnnf11.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf11.affine component=tdnnf11.affine input=tdnnf11.linear
component name=tdnnf11.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf11.relu component=tdnnf11.relu input=tdnnf11.affine
component name=tdnnf11.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf11.batchnorm component=tdnnf11.batchnorm input=tdnnf11.relu
component name=tdnnf11.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf11.dropout component=tdnnf11.dropout input=tdnnf11.batchnorm
component name=tdnnf11.noop type=NoOpComponent dim=1536
component-node name=tdnnf11.noop component=tdnnf11.noop input=Sum(Scale(0.66, tdnnf10.noop), tdnnf11.dropout)
component name=tdnnf12.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf12.linear component=tdnnf12.linear input=tdnnf11.noop
component name=tdnnf12.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf12.affine component=tdnnf12.affine input=tdnnf12.linear
component name=tdnnf12.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf12.relu component=tdnnf12.relu input=tdnnf12.affine
component name=tdnnf12.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf12.batchnorm component=tdnnf12.batchnorm input=tdnnf12.relu
component name=tdnnf12.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf12.dropout component=tdnnf12.dropout input=tdnnf12.batchnorm
component name=tdnnf12.noop type=NoOpComponent dim=1536
component-node name=tdnnf12.noop component=tdnnf12.noop input=Sum(Scale(0.66, tdnnf11.noop), tdnnf12.dropout)
component name=tdnnf13.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf13.linear component=tdnnf13.linear input=tdnnf12.noop
component name=tdnnf13.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf13.affine component=tdnnf13.affine input=tdnnf13.linear
component name=tdnnf13.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf13.relu component=tdnnf13.relu input=tdnnf13.affine
component name=tdnnf13.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf13.batchnorm component=tdnnf13.batchnorm input=tdnnf13.relu
component name=tdnnf13.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf13.dropout component=tdnnf13.dropout input=tdnnf13.batchnorm
component name=tdnnf13.noop type=NoOpComponent dim=1536
component-node name=tdnnf13.noop component=tdnnf13.noop input=Sum(Scale(0.66, tdnnf12.noop), tdnnf13.dropout)
component name=tdnnf14.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf14.linear component=tdnnf14.linear input=tdnnf13.noop
component name=tdnnf14.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf14.affine component=tdnnf14.affine input=tdnnf14.linear
component name=tdnnf14.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf14.relu component=tdnnf14.relu input=tdnnf14.affine
component name=tdnnf14.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf14.batchnorm component=tdnnf14.batchnorm input=tdnnf14.relu
component name=tdnnf14.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf14.dropout component=tdnnf14.dropout input=tdnnf14.batchnorm
component name=tdnnf14.noop type=NoOpComponent dim=1536
component-node name=tdnnf14.noop component=tdnnf14.noop input=Sum(Scale(0.66, tdnnf13.noop), tdnnf14.dropout)
component name=tdnnf15.linear type=TdnnComponent input-dim=1536 output-dim=160 l2-regularize=0.01 max-change=0.75 use-bias=false time-offsets=-3,0 orthonormal-constraint=-1.0
component-node name=tdnnf15.linear component=tdnnf15.linear input=tdnnf14.noop
component name=tdnnf15.affine type=TdnnComponent input-dim=160 output-dim=1536 l2-regularize=0.01 max-change=0.75 time-offsets=0,3
component-node name=tdnnf15.affine component=tdnnf15.affine input=tdnnf15.linear
component name=tdnnf15.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=tdnnf15.relu component=tdnnf15.relu input=tdnnf15.affine
component name=tdnnf15.batchnorm type=BatchNormComponent dim=1536
component-node name=tdnnf15.batchnorm component=tdnnf15.batchnorm input=tdnnf15.relu
component name=tdnnf15.dropout type=GeneralDropoutComponent dim=1536 dropout-proportion=0.0 continuous=true
component-node name=tdnnf15.dropout component=tdnnf15.dropout input=tdnnf15.batchnorm
component name=tdnnf15.noop type=NoOpComponent dim=1536
component-node name=tdnnf15.noop component=tdnnf15.noop input=Sum(Scale(0.66, tdnnf14.noop), tdnnf15.dropout)
component name=prefinal-l type=LinearComponent input-dim=1536 output-dim=256  orthonormal-constraint=-1.0 max-change=0.75 l2-regularize=0.01
component-node name=prefinal-l component=prefinal-l input=tdnnf15.noop
component name=prefinal-chain.affine type=NaturalGradientAffineComponent input-dim=256 output-dim=1536 l2-regularize=0.01 max-change=0.75
component-node name=prefinal-chain.affine component=prefinal-chain.affine input=prefinal-l
component name=prefinal-chain.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=prefinal-chain.relu component=prefinal-chain.relu input=prefinal-chain.affine
component name=prefinal-chain.batchnorm1 type=BatchNormComponent dim=1536
component-node name=prefinal-chain.batchnorm1 component=prefinal-chain.batchnorm1 input=prefinal-chain.relu
component name=prefinal-chain.linear type=LinearComponent input-dim=1536 output-dim=256 l2-regularize=0.01 max-change=0.75 orthonormal-constraint=-1 
component-node name=prefinal-chain.linear component=prefinal-chain.linear input=prefinal-chain.batchnorm1
component name=prefinal-chain.batchnorm2 type=BatchNormComponent dim=256
component-node name=prefinal-chain.batchnorm2 component=prefinal-chain.batchnorm2 input=prefinal-chain.linear
component name=output.affine type=NaturalGradientAffineComponent input-dim=256 output-dim=6160  l2-regularize=0.002 max-change=1.5 param-stddev=0.0 bias-stddev=0.0
component-node name=output.affine component=output.affine input=prefinal-chain.batchnorm2
output-node name=output input=output.affine objective=linear
component name=prefinal-xent.affine type=NaturalGradientAffineComponent input-dim=256 output-dim=1536 l2-regularize=0.01 max-change=0.75
component-node name=prefinal-xent.affine component=prefinal-xent.affine input=prefinal-l
component name=prefinal-xent.relu type=RectifiedLinearComponent dim=1536 self-repair-scale=1e-05
component-node name=prefinal-xent.relu component=prefinal-xent.relu input=prefinal-xent.affine
component name=prefinal-xent.batchnorm1 type=BatchNormComponent dim=1536
component-node name=prefinal-xent.batchnorm1 component=prefinal-xent.batchnorm1 input=prefinal-xent.relu
component name=prefinal-xent.linear type=LinearComponent input-dim=1536 output-dim=256 l2-regularize=0.01 max-change=0.75 orthonormal-constraint=-1 
component-node name=prefinal-xent.linear component=prefinal-xent.linear input=prefinal-xent.batchnorm1
component name=prefinal-xent.batchnorm2 type=BatchNormComponent dim=256
component-node name=prefinal-xent.batchnorm2 component=prefinal-xent.batchnorm2 input=prefinal-xent.linear
component name=output-xent.affine type=NaturalGradientAffineComponent input-dim=256 output-dim=6160  learning-rate-factor=5.0 l2-regularize=0.002 max-change=1.5 param-stddev=0.0 bias-stddev=0.0
component-node name=output-xent.affine component=output-xent.affine input=prefinal-xent.batchnorm2
component name=output-xent.log-softmax type=LogSoftmaxComponent dim=6160
component-node name=output-xent.log-softmax component=output-xent.log-softmax input=output-xent.affine
output-node name=output-xent input=output-xent.log-softmax objective=linear