wenwei202 / caffe

Caffe for Sparse and Low-rank Deep Neural Networks
Other
377 stars 134 forks source link

how to get speedup on GPU using conv_mode: LOWERED_CCNMM #20

Open MingSun-Tse opened 6 years ago

MingSun-Tse commented 6 years ago

I have a alexnet caffemodel with zero-column and zero-row weights. Using conv_mode: LOWERED_CCNMM, I got speedup on CPU (like structured sparsity=75%, speedup=3.1x), but on GPU, there is no speedup at all, what should I do to get speedup on GPU? I use the build/tools/caffe time tool to evaluate inference time. Anyone know sth. about this? thx a lot !!

wenwei202 commented 6 years ago

@MingSun-Tse you may refer to the third comment here

MingSun-Tse commented 6 years ago

@wenwei202 Do you mean the slow-down in GPU mode is because the LOWERED_CCNMM is implemented by CPU for now? If so, could you tell me how to reimplement the GPU speedup in your paper "Learning Structured Sparsity in Deep Neural Networks". There are GPU speedup results there in Table4 with AlexNet on ILSVRC 2012. Any hints will be okay, thanks~