Considerable speedup(VGG model:1.5x, AlexNet:1.1x)

akrizhevsky / cuda-convnet2

Automatically exported from code.google.com/p/cuda-convnet2

Apache License 2.0

774 stars 288 forks source link

Considerable speedup(VGG model:1.5x, AlexNet:1.1x) #16

Closed bestimage-tencent closed 9 years ago

bestimage-tencent commented 9 years ago

Considerable speedup(1.5x under VGG model with miniBatch of 32, 1.1x under AlexNet with miniBatch of 128), and the optimizations focus on fully employing gpu-releated functions.