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

Transformations: API 2.0 transition part 3 for LPT transformations (#19610)

Browse Source 
* lpt transformations: transition to api 2.0, ngraph -> openvino

* use ov namespace for lpt transformations

* fix low_precision usings

* includes refactoring

* delete RecurrentGraphRewrite and RecurrentMatcher as unused classes

* use ov header for itt; delete the disabled test

* delete the unused function

* suppress doxygen warning

* fix link in the documentation
This commit is contained in:
Ivan Tikhonov authored and GitHub committed 2023-09-13 12:30:31 +00:00
1 parent 3454139931
commit d0213301a5
547 files changed
+2797 -3176
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

docs/snippets/lpt_intel_cpu_plugin.cpp
+15 -15
View File
@@ -33,9 +33,9 @@ ov::pass::Manager manager;
auto pass_config = manager.get_pass_config();
//! [lpt_common]
// check if the function is quantized to ignore LPT transformations for not quantized function to speed up model loading
const bool useLpt = ngraph::pass::low_precision::LowPrecision::isFunctionQuantized(nGraphFunc);
const bool useLpt = ov::pass::low_precision::LowPrecision::isFunctionQuantized(nGraphFunc);
auto defaultPrecisions =
useLpt ? ngraph::pass::low_precision::precision_set::int8_support : std::vector<ov::element::Type>{};
useLpt ? ov::pass::low_precision::precision_set::int8_support : std::vector<ov::element::Type>{};
if (useLpt) {
// disable constant folding on dequantization subgraphs so they can be processed by LPT
manager.register_pass<ov::pass::MarkDequantizationSubgraph>(defaultPrecisions);
@@ -45,7 +45,7 @@ if (useLpt) {
if (useLpt) {
// convert subtract constant to INT8 to prevent unnecessary FP16 to FP32 conversion
manager.register_pass<ngraph::pass::low_precision::ConvertSubtractConstant>(defaultPrecisions);
manager.register_pass<ov::pass::low_precision::ConvertSubtractConstant>(defaultPrecisions);
}
// nGraph common transformations happen here
@@ -53,12 +53,12 @@ if (useLpt) {
if (useLpt) {
// convert not supported cases FakeQuantize -> Convert -> Convert -> Subtract -> Multiply to a single FakeQuantize
pass_config->set_callback<ov::pass::ConvertQuantizeDequantize>([&defaultPrecisions](const std::shared_ptr<const ngraph::Node> &node) -> bool {
return ngraph::pass::low_precision::NetworkHelper::areQuantizeAndDequantizeSupportedForMultiply(node, defaultPrecisions);
return ov::pass::low_precision::NetworkHelper::areQuantizeAndDequantizeSupportedForMultiply(node, defaultPrecisions);
});
// convert not supported cases FakeQuantize -> Convert -> Convert -> Subtract -> Multiply to a single FakeQuantize
pass_config->set_callback<ov::pass::ConvertSubtract>([&defaultPrecisions](const std::shared_ptr<const ngraph::Node> &node) -> bool {
return ngraph::pass::low_precision::NetworkHelper::areQuantizeAndDequantizeSupportedForSubtract(node, defaultPrecisions);
return ov::pass::low_precision::NetworkHelper::areQuantizeAndDequantizeSupportedForSubtract(node, defaultPrecisions);
});
}
@@ -66,7 +66,7 @@ manager.run_passes(nGraphFunc);
//! [lpt_common]
//! [lpt_execution]
using namespace ngraph::pass::low_precision;
using namespace ov::pass::low_precision;
if (useLpt) {
// Low precision transformations plugin specific configuration: restrictions definition
auto supportedPrecisions = std::vector<PrecisionsRestriction>({
@@ -96,7 +96,7 @@ if (useLpt) {
// Low precision transformations instantiation and registration in pass manager
ov::pass::Manager lptManager;
lptManager.register_pass<ngraph::pass::low_precision::LowPrecision>(supportedPrecisions, perTensorQuantization);
lptManager.register_pass<ov::pass::low_precision::LowPrecision>(supportedPrecisions, perTensorQuantization);
// Low precision transformations plugin specific configuration: transformation callbacks definition
lptManager.get_pass_config()->set_callback<MarkupPrecisions>([](const std::shared_ptr<const ngraph::Node>& node) -> bool {
@@ -130,7 +130,7 @@ int lpt_supported_precisions() {
std::shared_ptr<ov::Model> nGraphFunc;
ov::pass::Manager manager;
using namespace ngraph::pass::low_precision;
using namespace ov::pass::low_precision;
//! [lpt_supported_precisions]
auto supportedPrecisions = std::vector<PrecisionsRestriction>({
PrecisionsRestriction::create<ov::opset1::Convolution>({
@@ -140,7 +140,7 @@ auto supportedPrecisions = std::vector<PrecisionsRestriction>({
});
ov::pass::Manager lptManager;
lptManager.register_pass<ngraph::pass::low_precision::LowPrecision>(supportedPrecisions);
lptManager.register_pass<ov::pass::low_precision::LowPrecision>(supportedPrecisions);
lptManager.run_passes(nGraphFunc);
//! [lpt_supported_precisions]
@@ -154,7 +154,7 @@ return 0;
int per_tensor_quantization() {
std::shared_ptr<ov::Model> nGraphFunc;
//! [per_tensor_quantization]
using namespace ngraph::pass::low_precision;
using namespace ov::pass::low_precision;
const std::vector<PrecisionsRestriction> emptyRestrictions;
@@ -163,7 +163,7 @@ auto perTensorQuantization = std::vector<QuantizationGranularityRestriction>({
});
ov::pass::Manager lptManager;
lptManager.register_pass<ngraph::pass::low_precision::LowPrecision>(emptyRestrictions, perTensorQuantization);
lptManager.register_pass<ov::pass::low_precision::LowPrecision>(emptyRestrictions, perTensorQuantization);
lptManager.run_passes(nGraphFunc);
//! [per_tensor_quantization]
@@ -177,10 +177,10 @@ auto pass_config = manager.get_pass_config();
//! [asymmetric_quantization]
using namespace ngraph::pass::low_precision;
using namespace ov::pass::low_precision;
ov::pass::Manager lptManager;
lptManager.register_pass<ngraph::pass::low_precision::LowPrecision>();
lptManager.register_pass<ov::pass::low_precision::LowPrecision>();
lptManager.get_pass_config()->set_callback<ConvolutionBackpropDataTransformation>([&defaultPrecisions](const std::shared_ptr<const ngraph::Node>& node) -> bool {
return LayerTransformation::isAsymmetricQuantization(node, defaultPrecisions) || WeightableLayerTransformation::isAsymmetricOnWeights(node);
});
@@ -194,7 +194,7 @@ int lpt_markup_pipeline() {
std::shared_ptr<ov::Model> nGraphFunc;
ov::pass::Manager manager;
using namespace ngraph::pass::low_precision;
using namespace ov::pass::low_precision;
//! [lpt_markup_pipeline]
auto supportedPrecisions = std::vector<PrecisionsRestriction>({
PrecisionsRestriction::create<ov::opset1::Convolution>({
@@ -208,7 +208,7 @@ auto perTensorQuantization = std::vector<QuantizationGranularityRestriction>({
});
ov::pass::Manager lptManager;
lptManager.register_pass<ngraph::pass::low_precision::LowPrecision>(supportedPrecisions, perTensorQuantization);
lptManager.register_pass<ov::pass::low_precision::LowPrecision>(supportedPrecisions, perTensorQuantization);
lptManager.run_passes(nGraphFunc);
//! [lpt_markup_pipeline]