This is needed to get performance on backends other than GPU in Imperative Gluon (think MKLDNN layouts or non CPU/non GPU memory). We're not including these other backends in this first PR, and we haven't fully optimized Gluon for CPU through nGraph, but this is setting the stage for upcoming PRs.
[ Full content available at: https://github.com/apache/incubator-mxnet/pull/12502 ] This message was relayed via gitbox.apache.org for [email protected]
