MNIST (LeNet) Model Plot BatchSize 1 BatchSize 2 BatchSize 4 BatchSize 8 BatchSize 16 BatchSize 32Layer Information Click to See Model Layer Information ↕LAYERNAMELAYERTYPEINPUTNAMESOUTPUTNAMESINPUTSHAPESOUTPUTSHAPESParameter193ConstantInput[[16,4,4,10]][[16,4,4,10]]Parameter193_reshape1_shapeConstantInput[[256,10]][[256,10]]Times212_reshape1ReshapeParameter193;Parameter193_reshape1_shapeParameter193_reshape1[[16,4,4,10],[256,10]][[256,10]]Input3ConstantInput[[1,1,28,28]][[1,1,28,28]]Parameter5ConstantInput[[8,1,5,5]][[8,1,5,5]]Convolution28ConvInput3;Parameter5Convolution28_Output_0[[1,1,28,28],[8,1,5,5]][[1,8,28,28]]Parameter6ConstantInput[[8,1,1]][[8,1,1]]Plus30ElementWiseConvolution28_Output_0;Parameter6Plus30_Output_0[[1,8,28,28],[8,1,1]][[1,8,28,28]]ReLU32ReluPlus30_Output_0ReLU32_Output_0[[1,8,28,28]][[1,8,28,28]]Pooling66PoolingReLU32_Output_0Pooling66_Output_0[[1,8,28,28]][[1,8,14,14]]Parameter87ConstantInput[[16,8,5,5]][[16,8,5,5]]Convolution110ConvPooling66_Output_0;Parameter87Convolution110_Output_0[[1,8,14,14],[16,8,5,5]][[1,16,14,14]]Parameter88ConstantInput[[16,1,1]][[16,1,1]]Plus112ElementWiseConvolution110_Output_0;Parameter88Plus112_Output_0[[1,16,14,14],[16,1,1]][[1,16,14,14]]ReLU114ReluPlus112_Output_0ReLU114_Output_0[[1,16,14,14]][[1,16,14,14]]Pooling160PoolingReLU114_Output_0Pooling160_Output_0[[1,16,14,14]][[1,16,4,4]]Pooling160_Output_0_reshape0_shapeConstantInput[[1,256]][[1,256]]Times212_reshape0ReshapePooling160_Output_0;Pooling160_Output_0_reshape0_shapePooling160_Output_0_reshape0[[1,16,4,4],[1,256]][[1,256]]Times212GemmPooling160_Output_0_reshape0;Parameter193_reshape1Times212_Output_0[[1,256],[256,10]][[1,10]]Parameter194ConstantInput[[1,10]][[1,10]]Plus214ElementWiseTimes212_Output_0;Parameter194Plus214_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 CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSVClick to See Tesla M60 Information ↕Tesla M60 We run the model for different batch sizes.Batch Size 2 PDF CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSVClick to See GRID K520 Information ↕GRID K520 We run the model for different batch sizes.Batch Size 2 PDF CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSVClick to See TITAN Xp Information ↕TITAN Xp We run the model for different batch sizes.Batch Size 2 PDF CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSVClick to See TITAN V Information ↕TITAN V We run the model for different batch sizes.Batch Size 2 PDF CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSVClick to See Tesla V100-SXM2-16GB Information ↕Tesla V100-SXM2-16GB We run the model for different batch sizes.Batch Size 2 PDF CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSVClick to See Quadro RTX 6000 Information ↕Quadro RTX 6000 We run the model for different batch sizes.Batch Size 2 PDF CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSVClick to See Tesla T4 Information ↕Tesla T4 We run the model for different batch sizes.Batch Size 2 PDF CSVBatch Size 4 PDF CSVBatch Size 8 PDF CSVBatch Size 16 PDF CSVBatch Size 32 PDF CSV