MNIST (LeNet) Model Plot BatchSize 1 BatchSize 2 BatchSize 4 BatchSize 8 BatchSize 16 BatchSize 32 Layer Information Click to See Model Layer Information ↕ LAYERNAME LAYERTYPE INPUTNAMES OUTPUTNAMES INPUTSHAPES OUTPUTSHAPES Parameter193 ConstantInput [[16,4,4,10]] [[16,4,4,10]] Parameter193_reshape1_shape ConstantInput [[256,10]] [[256,10]] Times212_reshape1 Reshape Parameter193;Parameter193_reshape1_shape Parameter193_reshape1 [[16,4,4,10],[256,10]] [[256,10]] Input3 ConstantInput [[1,1,28,28]] [[1,1,28,28]] Parameter5 ConstantInput [[8,1,5,5]] [[8,1,5,5]] Convolution28 Conv Input3;Parameter5 Convolution28_Output_0 [[1,1,28,28],[8,1,5,5]] [[1,8,28,28]] Parameter6 ConstantInput [[8,1,1]] [[8,1,1]] Plus30 ElementWise Convolution28_Output_0;Parameter6 Plus30_Output_0 [[1,8,28,28],[8,1,1]] [[1,8,28,28]] ReLU32 Relu Plus30_Output_0 ReLU32_Output_0 [[1,8,28,28]] [[1,8,28,28]] Pooling66 Pooling ReLU32_Output_0 Pooling66_Output_0 [[1,8,28,28]] [[1,8,14,14]] Parameter87 ConstantInput [[16,8,5,5]] [[16,8,5,5]] Convolution110 Conv Pooling66_Output_0;Parameter87 Convolution110_Output_0 [[1,8,14,14],[16,8,5,5]] [[1,16,14,14]] Parameter88 ConstantInput [[16,1,1]] [[16,1,1]] Plus112 ElementWise Convolution110_Output_0;Parameter88 Plus112_Output_0 [[1,16,14,14],[16,1,1]] [[1,16,14,14]] ReLU114 Relu Plus112_Output_0 ReLU114_Output_0 [[1,16,14,14]] [[1,16,14,14]] Pooling160 Pooling ReLU114_Output_0 Pooling160_Output_0 [[1,16,14,14]] [[1,16,4,4]] Pooling160_Output_0_reshape0_shape ConstantInput [[1,256]] [[1,256]] Times212_reshape0 Reshape Pooling160_Output_0;Pooling160_Output_0_reshape0_shape Pooling160_Output_0_reshape0 [[1,16,4,4],[1,256]] [[1,256]] Times212 Gemm Pooling160_Output_0_reshape0;Parameter193_reshape1 Times212_Output_0 [[1,256],[256,10]] [[1,10]] Parameter194 ConstantInput [[1,10]] [[1,10]] Plus214 ElementWise Times212_Output_0;Parameter194 Plus214_Output_0 [[1,10],[1,10]] [[1,10]] Per Layer Timing We evaluate the MNIST across all systems and we show the results below. Click to See Tesla K80 Information ↕ Tesla K80 We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV Click to See Tesla M60 Information ↕ Tesla M60 We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV Click to See GRID K520 Information ↕ GRID K520 We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV Click to See TITAN Xp Information ↕ TITAN Xp We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV Click to See TITAN V Information ↕ TITAN V We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV Click to See Tesla V100-SXM2-16GB Information ↕ Tesla V100-SXM2-16GB We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV Click to See Quadro RTX 6000 Information ↕ Quadro RTX 6000 We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV Click to See Tesla T4 Information ↕ Tesla T4 We run the model for different batch sizes. Batch Size 2 PDF CSV Batch Size 4 PDF CSV Batch Size 8 PDF CSV Batch Size 16 PDF CSV Batch Size 32 PDF CSV