apeforest commented on a change in pull request #14056: A better
split-2D(SliceChannel) op forward kernel for CPU
URL: https://github.com/apache/incubator-mxnet/pull/14056#discussion_r259210379
##########
File path: src/operator/channel_op_common.h
##########
@@ -101,6 +101,42 @@ void Split(const mshadow::Tensor<xpu, dim, DType> &input,
split_helper<xpu, dim, dim-1>(input, output, dimension, req);
}
}
+
+template<typename cpu, int dim, typename DType>
Review comment:
Current template for `split` and `concatenate` is generic for cpu and gpu.
This Split_2D only works for CPU. Do you plan to add a GPU version as well?
Otherwise, it seems to break the existing pattern.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
