IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[GPU] Added WA for MarkDequantizationSubgraph to work correctly in the plugin (#20767)

Browse Source
This commit is contained in:
Roman Lyamin 2023-10-30 21:58:34 +04:00 committed by GitHub
parent df3e51f818
commit 7eb53586f5
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 7 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
src/plugins/intel_gpu/src/plugin/transformations_pipeline.cpp
View File

@ -109,6 +109,7 @@
#include "transformations/convert_precision.hpp"
#include "transformations/init_node_info.hpp"
#include "transformations/rt_info/fused_names_attribute.hpp"
#include "transformations/rt_info/keep_const_precision.hpp"
#include "transformations/smart_reshape/matmul_sr.hpp"
#include "plugin/transformations/convert_matmul_to_fc.hpp"
@ -159,6 +160,12 @@ void TransformationsPipeline::apply(std::shared_ptr<ov::Model> func) {
auto pass_config = manager.get_pass_config();
manager.set_per_pass_validation(false);
// Temporary solution, global rt info cleanup is needed
for (auto& node : func->get_ops()) {
ov::enable_constant_folding(node);
ov::disable_keep_const_precision(node);
}
enableInt8 = config.get_property(ov::intel_gpu::enable_lp_transformations) && ov::pass::low_precision::LowPrecision::isFunctionQuantized(func);
if (enableInt8) {
manager.register_pass<ov::pass::MarkDequantizationSubgraph>(