Removing all-zero rows and columns of the feature map matrix by a GPU routine

wenwei202 / caffe

Caffe for Sparse and Low-rank Deep Neural Networks

Other

379 stars 134 forks source link

Removing all-zero rows and columns of the feature map matrix by a GPU routine #3

Open wenwei202 opened 7 years ago

wenwei202 commented 7 years ago

In GPU mode with conv_mode: LOWERED_CCNMM , we need to first remove all-zero columns and rows in the feature map matrix col_buffer_. This concatenation process is temporally using the corresponding CPU routine. We plan to substitute it with a GPU routine. Please pull request if anyone implements this.

Code branch: https://github.com/wenwei202/caffe/tree/scnn