jcjohnson/cnn-benchmarks

Speed comparison with different implement frameworks

tumitx opened this issue · 1 comments

tumitx commented

As you stated that the Alexnet is implemented from caffe, while Resnets are implemented from torch.

Is it a fair comparison about the speed?

jcjohnson commented

The AlexNet (and VGG-16 and VGG-19) models were trained in Caffe, but for the benchmark the models were converted to Torch using loadcaffe; after conversion they are composed of normal Torch layers and are absolutely comparable to ResNet models that were trained using Torch.