pengzhao-intel commented on a change in pull request #14819: [MKLDNN] Enable
more convolution + activation fusion
URL: https://github.com/apache/incubator-mxnet/pull/14819#discussion_r279192978
##########
File path: src/operator/subgraph/mkldnn/mkldnn_conv.cc
##########
@@ -373,6 +374,19 @@ void SgMKLDNNConvOperator::Forward(const OpContext &ctx,
if (mkldnn_param.with_sum) {
full_conv_param.sum_scale = output_scale / sum_in_scale;
}
+ if (mkldnn_param.with_act &&
+ full_conv_param.act_param.alg ==
mkldnn::algorithm::eltwise_bounded_relu) {
+ if (mkldnn_param.with_sum) {
+ LOG(ERROR) << "mkldnn doesn't support conv + relu + sum fusion yet.";
Review comment:
Is this a `warning` and make the unfused flow can work too?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to 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
