Merge branch 'master' into feature/snippets/aarch64/prelu_sqrt_round

.github/workflows/linux_riscv.yml

@@ -139,7 +139,6 @@ jobs:
             git submodule update --init -- ${OPENVINO_REPO}/thirdparty/gflags
             git submodule update --init -- ${OPENVINO_REPO}/thirdparty/telemetry
             git submodule update --init -- ${OPENVINO_REPO}/src/plugins/intel_cpu
-            git submodule update --init -- ${OPENVINO_REPO}/thirdparty/open_model_zoo
             git submodule update --init -- ${OPENVINO_REPO}/thirdparty/flatbuffers/flatbuffers
           popd
 

.gitmodules

@@ -52,9 +52,6 @@
 [submodule "thirdparty/onednn_gpu"]
 	path = src/plugins/intel_gpu/thirdparty/onednn_gpu
 	url = https://github.com/oneapi-src/oneDNN.git
-[submodule "tools/pot/thirdparty/open_model_zoo"]
-	path = thirdparty/open_model_zoo
-	url = https://github.com/openvinotoolkit/open_model_zoo.git
 [submodule "thirdparty/json/nlohmann_json"]
 	path = thirdparty/json/nlohmann_json
 	url = https://github.com/nlohmann/json.git

docs/articles_en/openvino-workflow/running-inference/inference-devices-and-modes/cpu-device/performance-hint-and-thread-scheduling.rst

@@ -63,19 +63,19 @@ the model precision and the ratio of P-cores and E-cores.
 
 Then the default settings for low-level performance properties on Windows and Linux are as follows:
 
-+--------------------------------------+-----------------------------------------------------------------------+-----------------------------------------------------------------------+
-| Property                             | Windows                                                               | Linux                                                                 |
-+======================================+=======================================================================+=======================================================================+
-| ``ov::num_streams``                  | 1                                                                     | 1                                                                     |
-+--------------------------------------+-----------------------------------------------------------------------+-----------------------------------------------------------------------+
-| ``ov::inference_num_threads``        | is equal to the number of P-cores or P-cores+E-cores on one numa node | is equal to the number of P-cores or P-cores+E-cores on one numa node |
-+--------------------------------------+-----------------------------------------------------------------------+-----------------------------------------------------------------------+
-| ``ov::hint::scheduling_core_type``   | :ref:`Core Type Table of Latency Hint <core_type_latency>`            | :ref:`Core Type Table of Latency Hint <core_type_latency>`            |
-+--------------------------------------+-----------------------------------------------------------------------+-----------------------------------------------------------------------+
-| ``ov::hint::enable_hyper_threading`` | No                                                                    | No                                                                    |
-+--------------------------------------+-----------------------------------------------------------------------+-----------------------------------------------------------------------+
-| ``ov::hint::enable_cpu_pinning``     | No / Not Supported                                                    | Yes except using P-cores and E-cores together                         |
-+--------------------------------------+-----------------------------------------------------------------------+-----------------------------------------------------------------------+
++--------------------------------------+--------------------------------------------------------------------+--------------------------------------------------------------------+
+| Property                             | Windows                                                            | Linux                                                              |
++======================================+====================================================================+====================================================================+
+| ``ov::num_streams``                  | 1                                                                  | 1                                                                  |
++--------------------------------------+--------------------------------------------------------------------+--------------------------------------------------------------------+
+| ``ov::inference_num_threads``        | is equal to the number of P-cores or P-cores+E-cores on one socket | is equal to the number of P-cores or P-cores+E-cores on one socket |
++--------------------------------------+--------------------------------------------------------------------+--------------------------------------------------------------------+
+| ``ov::hint::scheduling_core_type``   | :ref:`Core Type Table of Latency Hint <core_type_latency>`         | :ref:`Core Type Table of Latency Hint <core_type_latency>`         |
++--------------------------------------+--------------------------------------------------------------------+--------------------------------------------------------------------+
+| ``ov::hint::enable_hyper_threading`` | No                                                                 | No                                                                 |
++--------------------------------------+--------------------------------------------------------------------+--------------------------------------------------------------------+
+| ``ov::hint::enable_cpu_pinning``     | No / Not Supported                                                 | Yes except using P-cores and E-cores together                      |
++--------------------------------------+--------------------------------------------------------------------+--------------------------------------------------------------------+
 
 .. note::
 
@@ -96,7 +96,7 @@ Then the default settings for low-level performance properties on Windows and Li
    Starting from 5th Gen Intel Xeon Processors, new microarchitecture enabled new sub-NUMA clusters
    feature. A sub-NUMA cluster (SNC) can create two or more localization domains (numa nodes)
    within a socket by BIOS configuration. 
-   By default OpenVINO with latency hint uses single NUMA node for inference. Although such
+   By default OpenVINO with latency hint uses single socket for inference. Although such
    behavior allows to achive best performance for most of the models, there might be corner
    cases which require manual tuning of ``ov::num_streams`` and ``ov::hint::enable_hyper_threading parameters``.
    Please find more detail about `Sub-NUMA Clustering <https://www.intel.com/content/www/us/en/developer/articles/technical/xeon-processor-scalable-family-technical-overview.html>`__ 

docs/openvino_custom_sphinx_sitemap/openvino_custom_sphinx_sitemap/__init__.py

@@ -126,33 +126,36 @@ def process_coveo_meta(meta, url, link):
             if tag_name == 'ovdoctype':
                 ET.SubElement(namespace_element, tag_name).text = process_link(link)
             elif tag_name == 'ovcategory' and loc_element is not None:
-                ET.SubElement(namespace_element, tag_name).text = extract_hierarchy(loc_element.text)
+                ET.SubElement(namespace_element, tag_name).text = extract_categories(loc_element.text)
             elif tag_name == 'ovversion':
                 ET.SubElement(namespace_element, tag_name).text = tag_value
 
 def process_link(link):
     if '/' in link:
-        return link.split('/')[0].replace("-", " ")
-    return link.split('.html')[0].replace("-", " ")
+        return format_segment(link.split('/')[0].replace("-", " "))
+    return format_segment(link.split('.html')[0].replace("-", " "))
 
-def extract_hierarchy(link):
+def extract_categories(link):
     path = link.split("://")[-1]
     segments = path.split('/')[1:]
     if segments and segments[-1].endswith('.html'):
         segments = segments[:-1]
 
+    if segments:
+        segments = segments[1:]
+
     if segments and '.' in segments[0]:
         year, *rest = segments[0].split('.')
         if year.isdigit() and len(year) == 4:
             segments[0] = year
 
     segments = [format_segment(segment) for segment in segments]
 
-    hierarchy = []
-    for i in range(1, len(segments) + 1):
-        hierarchy.append('|'.join(segments[:i]))
-
-    return ';'.join(hierarchy)
+    if segments:
+        hierarchy = ['|'.join(segments[:i]) for i in range(1, len(segments) + 1)]
+        return ';'.join(hierarchy)
+    
+    return "No category"
 
 def format_segment(segment):
     if segment == 'c_cpp_api': segment = 'C/C++_api'

src/frontends/paddle/src/default_opset.hpp

@@ -2,13 +2,13 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
-#include "openvino/opsets/opset9.hpp"
+#include "openvino/opsets/opset14.hpp"
 
 namespace ov {
 namespace frontend {
 namespace paddle {
 namespace op {
-namespace default_opset = ov::opset9;
+namespace default_opset = ov::opset14;
 
 }  // namespace op
 }  // namespace paddle

src/frontends/paddle/src/op/elu.cpp

@@ -0,0 +1,23 @@
+// Copyright (C) 2018-2024 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "default_opset.hpp"
+#include "openvino/frontend/paddle/node_context.hpp"
+#include "openvino/frontend/paddle/visibility.hpp"
+
+namespace ov {
+namespace frontend {
+namespace paddle {
+namespace op {
+NamedOutputs elu(const NodeContext& node) {
+    auto data = node.get_input("X");
+    auto alpha = node.get_attribute<float>("alpha", 1.0);
+    const auto& elu_node = std::make_shared<default_opset::Elu>(data, alpha);
+    return node.default_single_output_mapping({elu_node}, {"Out"});
+}
+
+}  // namespace op
+}  // namespace paddle
+}  // namespace frontend
+}  // namespace ov

src/frontends/paddle/src/op/expand_v2.cpp

@@ -19,8 +19,16 @@ NamedOutputs expand_v2(const NodeContext& node) {
         auto inputs = node.get_ng_inputs("expand_shapes_tensor");
         ov::NodeVector node_vec;
         for (auto& input : inputs) {
+            if (input.get_partial_shape().rank().get_length() == 0) {
+                // should unsqueeze the input with non-shape.
+                auto unsqueeze_scalar = default_opset::Constant::create(ov::element::i32, {}, {0});
+                input = std::make_shared<default_opset::Unsqueeze>(input, unsqueeze_scalar);
+            }
+            PADDLE_OP_CHECK(node,
+                            input.get_partial_shape().rank().get_length() == 1,
+                            "the rank of conv input must == 1");
             auto cast = std::make_shared<Convert>(input, element::i32);
-            node_vec.push_back(cast);
+            node_vec.emplace_back(cast);
         }
         shape_expected_node = std::make_shared<Concat>(node_vec, 0);
     } else {

src/frontends/paddle/src/op/eye.cpp

@@ -0,0 +1,36 @@
+// Copyright (C) 2018-2024 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "default_opset.hpp"
+#include "openvino/frontend/paddle/node_context.hpp"
+
+namespace ov {
+namespace frontend {
+namespace paddle {
+namespace op {
+NamedOutputs eye(const NodeContext& node) {
+    auto row = node.get_attribute<int64_t>("num_rows");
+    auto col = node.get_attribute<int64_t>("num_columns", row);
+    auto dtype = node.get_attribute<ov::element::Type>("dtype", ov::element::f32);
+
+    const auto& row_node = std::make_shared<default_opset::Constant>(ov::element::i64, Shape{}, (row));
+    const auto& col_node = std::make_shared<default_opset::Constant>(ov::element::i64, Shape{}, (col));
+    const auto& diagonal_index_node = std::make_shared<default_opset::Constant>(ov::element::i32, Shape{}, (0));
+
+    std::shared_ptr<Node> out_node;
+    if (dtype == ov::element::i32 || dtype == ov::element::i64) {
+        out_node = std::make_shared<default_opset::Eye>(row_node, col_node, diagonal_index_node, dtype);
+    } else {
+        const auto& eye_node =
+            std::make_shared<default_opset::Eye>(row_node, col_node, diagonal_index_node, ov::element::i32);
+        out_node = std::make_shared<default_opset::Convert>(eye_node, dtype);
+    }
+
+    return node.default_single_output_mapping({out_node}, {"Out"});
+}
+
+}  // namespace op
+}  // namespace paddle
+}  // namespace frontend
+}  // namespace ov

src/frontends/paddle/src/op/fill_constant.cpp

@@ -29,6 +29,10 @@ NamedOutputs fill_constant(const NodeContext& node) {
         PADDLE_OP_CHECK(node, false, "fill_constant only supports i32, f32, i64");
     }
 
+    if (shape.empty()) {
+        shape.emplace_back(1);
+    }
+
     PADDLE_OP_CHECK(node,
                     shape.size() > 0 || node.has_input("ShapeTensor") || node.has_input("ShapeTensorList"),
                     "fill_constant shape not set");

src/frontends/paddle/src/op/interp.cpp

@@ -4,6 +4,7 @@
 
 #include "default_opset.hpp"
 #include "openvino/frontend/paddle/node_context.hpp"
+#include "openvino/opsets/opset4.hpp"
 
 namespace ov {
 namespace frontend {
@@ -147,8 +148,9 @@ static NamedOutputs interpolate(const NodeContext& node,
     attrs.pads_begin = {0, 0, 0, 0};
     attrs.pads_end = {0, 0, 0, 0};
 
-    return node.default_single_output_mapping({std::make_shared<Interpolate>(x, target_spatial_shape, scales, attrs)},
-                                              {"Out"});
+    return node.default_single_output_mapping(
+        {std::make_shared<ov::opset4::Interpolate>(x, target_spatial_shape, scales, attrs)},
+        {"Out"});
 }
 
 NamedOutputs linear_interp_v2(const NodeContext& node) {

src/frontends/paddle/src/op/reduce_ops.hpp

@@ -31,6 +31,10 @@ NamedOutputs reduce_ops(const NodeContext& node) {
         dims = node.get_attribute<std::vector<int64_t>>("dim");
     }
 
+    std::transform(dims.begin(), dims.end(), dims.begin(), [&input_rank](int64_t value) {
+        return value >= 0 ? value : value + input_rank;
+    });
+
     int64_t axis_size = static_cast<int64_t>(dims.size());
     reduce_all = reduce_all || (axis_size == input_rank || axis_size == 0);
 

src/frontends/paddle/src/op_table.cpp

@@ -39,9 +39,11 @@ OP_CONVERTER(elementwise_sub);
 OP_CONVERTER(equal);
 OP_CONVERTER(greater_equal);
 OP_CONVERTER(not_equal);
+OP_CONVERTER(elu);
 OP_CONVERTER(embedding);
 OP_CONVERTER(exp);
 OP_CONVERTER(expand_v2);
+OP_CONVERTER(eye);
 OP_CONVERTER(flip);
 OP_CONVERTER(flatten_contiguous_range);
 OP_CONVERTER(floor);
@@ -173,9 +175,11 @@ std::map<std::string, CreatorFunction> get_supported_ops() {
             {"elementwise_sub", op::elementwise_sub},
             {"dropout", op::dropout},
             {"elementwise_pow", op::elementwise_pow},
+            {"elu", op::elu},
             {"equal", op::equal},
             {"exp", op::exp},
             {"expand_v2", op::expand_v2},
+            {"eye", op::eye},
             {"fill_any_like", op::fill_any_like},
             {"fill_constant", op::fill_constant},
             {"fill_constant_batch_size_like", op::fill_constant_batch_size_like},

src/frontends/paddle/tests/op_fuzzy.cpp

@@ -188,6 +188,7 @@ static const std::vector<std::string> models{
     std::string("elementwise_floordiv_int64_2/elementwise_floordiv_int64_2.pdmodel"),
     std::string("elementwise_floordiv_int64_3/elementwise_floordiv_int64_3.pdmodel"),
     std::string("elementwise_mul_bool1/elementwise_mul_bool1.pdmodel"),
+    std::string("elu/elu.pdmodel"),
     std::string("embedding_0/embedding_0.pdmodel"),
     std::string("embedding_sparse/embedding_sparse.pdmodel"),
     std::string("embedding_none_weight/embedding_none_weight.pdmodel"),
@@ -201,6 +202,9 @@ static const std::vector<std::string> models{
     std::string("expand_v2_tensor_list/expand_v2_tensor_list.pdmodel"),
     std::string("expand_v2_tensor_list2/expand_v2_tensor_list2.pdmodel"),
     std::string("exp_test_float32/exp_test_float32.pdmodel"),
+    std::string("eye/eye.pdmodel"),
+    std::string("eye_int32/eye_int32.pdmodel"),
+    std::string("eye_int64/eye_int64.pdmodel"),
     std::string("flip_1/flip_1.pdmodel"),
     std::string("flip_2/flip_2.pdmodel"),
     std::string("flip_3/flip_3.pdmodel"),

src/frontends/paddle/tests/test_models/gen_scripts/generate_elu.py

@@ -0,0 +1,44 @@
+# Copyright (C) 2018-2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+#
+# relu6 paddle model generator
+#
+import numpy as np
+from save_model import saveModel
+import paddle
+import sys
+
+
+def elu(name: str, x, alpha=None, data_type='float32'):
+    paddle.enable_static()
+
+    with paddle.static.program_guard(paddle.static.Program(), paddle.static.Program()):
+        node_x = paddle.static.data(name='x', shape=x.shape, dtype=data_type)
+
+        if paddle.__version__ >= '2.0.0':
+            out = paddle.nn.functional.elu(node_x, alpha, name='elu')
+        else:
+            out = paddle.fluid.layers.elu(node_x, alpha, name='elu')
+        cpu = paddle.static.cpu_places(1)
+        exe = paddle.static.Executor(cpu[0])
+        # startup program will call initializer to initialize the parameters.
+        exe.run(paddle.static.default_startup_program())
+
+        outs = exe.run(
+            feed={'x': x},
+            fetch_list=[out])             
+
+        saveModel(name, exe, feed_vars=[node_x], fetchlist=[out],
+                  inputs=[x], outputs=[outs[0]], target_dir=sys.argv[1])
+
+    return outs[0]
+
+
+def main():
+    data_type = 'float32'
+    data = np.random.randn(2, 3, 4).astype('float32')
+    elu("elu", data)
+
+if __name__ == "__main__":
+    main()

src/frontends/paddle/tests/test_models/gen_scripts/generate_eye.py

@@ -0,0 +1,41 @@
+# Copyright (C) 2018-2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+#
+# fill_const paddle model generator
+#
+import numpy as np
+from save_model import saveModel
+import paddle
+import sys
+
+
+def eye(name : str, rows, cols = None, dtype = None):
+    paddle.enable_static()
+    with paddle.static.program_guard(paddle.static.Program(), paddle.static.Program()):
+        if paddle.__version__ >= '2.0.0':
+            x1 = paddle.eye(num_rows=rows, num_columns=cols, dtype=dtype, name='fill')
+            x2 = paddle.eye(num_rows=rows, num_columns=cols, dtype=dtype, name='fill')
+        else:
+            x1 = paddle.fluid.layers.eye(num_rows=rows, num_columns=cols, dtype=dtype, name='fill_constant')
+            x2 = paddle.fluid.layers.eye(num_rows=rows, num_columns=cols, dtype=dtype, name='fill_constant')
+        out = paddle.add(x1, x2)
+        cpu = paddle.static.cpu_places(1)
+        exe = paddle.static.Executor(cpu[0])
+        # startup program will call initializer to initialize the parameters.
+        exe.run(paddle.static.default_startup_program())
+
+        outs = exe.run(
+            fetch_list=[out])             
+
+        saveModel(name, exe, feed_vars=[], fetchlist=[out], inputs=[], outputs=[outs[0]], target_dir=sys.argv[1])
+
+    return outs[0]
+
+def main():
+    eye("eye", 3)
+    eye("eye_int32", 2, 3, "int32")
+    eye("eye_int64", 2, 3, "int64")
+
+if __name__ == "__main__":
+    main()

tools/CMakeLists.txt

@@ -5,18 +5,10 @@
 # Python tools
 #
 
-# MO
-
-add_subdirectory(mo)
-
 if(ENABLE_PYTHON)
     # Benchmark Tool
     add_subdirectory(benchmark_tool)
 
     # OpenVino Conversion Tool
     add_subdirectory(ovc)
 endif()
-
-# wheel openvino-dev
-
-add_subdirectory(openvino_dev)