Implementing cuSPARSE within cuDNN

Hi,

I am looking for cuSPARSE support within cuDNN. By making filter weight matrices sparse in convolutions, I want to accelerate the forward pass. What would be best way to enable Sparse matrix operations in cuDNN?

Thanks!

Hi,

We currently don’t have cuSPARSE+cuDNN sample code.

For cuDNN,
There are some samples located at /home/ubuntu/cudnn_samples_v5/ after installing with JetPack.

For cuSPARSE,
Please find document for more information:
http://docs.nvidia.com/cuda/cusparse/index.html#ixzz4aWEDdaET

Hello,

Is there any sample code implementation of the convolution layer using cuSPARSE ?(.cu file)

thank you in advance

Hi,

We don’t have an official sample for cuSPARSE on DL.
Here is the resource from the community for your reference:
https://github.com/wenwei202/caffe/tree/scnn

Thanks.