Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

[mkldnn-v1.0]rm int8 sum workaround #16623

Merged
merged 2 commits into from
Oct 28, 2019
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -161,8 +161,11 @@ static void MKLDNNQuantizedElemwiseAddForward(const nnvm::NodeAttrs& attrs, cons
std::vector<mkldnn::memory::desc> in_desc;
in_desc.push_back(dataA_mem->get_desc());
in_desc.push_back(dataB_mem->get_desc());
auto output_desc = dataA_mem->get_desc();
output_desc.data.data_type = static_cast<mkldnn_data_type_t>(output_data_type);
const auto in_shape = in_data[quantized_elemwise_add_enum::kDataA].shape();
mkldnn::memory::dims i_dims(in_shape.begin(), in_shape.end());
auto output_desc = mkldnn::memory::desc(i_dims,
output_data_type,
mkldnn::memory::format_tag::any);
mkldnn::sum::primitive_desc pdesc(output_desc, scales, in_desc, engine);
auto mem = CreateMKLDNNMem(out_data[quantized_elemwise_add_enum::kOut],
pdesc.dst_desc(),
Expand Down