Merge branch 'master' into an/mvn_refactor_2

.github/ISSUE_TEMPLATE/bug.yml

@@ -58,7 +58,6 @@ body:
         - Caffe
         - ONNX
         - PyTorch
-        - mxnet
         - PaddlePaddle
     validations:
       required: false

.github/workflows/coverage.yml

@@ -51,7 +51,6 @@ jobs:
           # For running TensorFlow frontend unit tests
           python3 -m pip install -r ${{ github.workspace }}/src/frontends/tensorflow/tests/requirements.txt
           # For MO unit tests
-          python3 -m pip install -r ${{ github.workspace }}/tools/mo/requirements_mxnet.txt
           python3 -m pip install -r ${{ github.workspace }}/tools/mo/requirements_caffe.txt
           python3 -m pip install -r ${{ github.workspace }}/tools/mo/requirements_kaldi.txt
           python3 -m pip install -r ${{ github.workspace }}/tools/mo/requirements_onnx.txt

.github/workflows/job_python_unit_tests.yml

@@ -99,10 +99,6 @@ jobs:
 
           extras_to_install="caffe,kaldi,onnx,tensorflow2,pytorch"
 
-          if [[ "${{ runner.arch }}" != "ARM64" ]]; then
-            extras_to_install="mxnet,$extras_to_install"
-          fi
-
           # Find and install OV dev wheel
           pushd ${INSTALL_DIR}/tools
             ov_dev_wheel_name=$(find . -name 'openvino_dev*.whl')
@@ -142,8 +138,7 @@ jobs:
 
           # Skips under tickets: 133405, 122666
           python3 -m pytest -s ${INSTALL_TEST_DIR}/mo/unit_tests \
-              --junitxml=${INSTALL_TEST_DIR}/TEST-ModelOptimizer.xml \
-              --ignore-glob="**/mo/unit_tests/mo/front/mxnet/**"
+              --junitxml=${INSTALL_TEST_DIR}/TEST-ModelOptimizer.xml
 
       - name: Python ONNX operators tests
         if: (fromJSON(inputs.affected-components).Python_API.test ||

.github/workflows/mo.yml

@@ -46,7 +46,6 @@ jobs:
           # For UT
           pip install unittest-xml-reporting==3.0.2
           # MO requirements
-          pip install -r requirements_mxnet.txt
           pip install -r requirements_caffe.txt
           pip install -r requirements_kaldi.txt
           pip install -r requirements_onnx.txt

.github/workflows/windows_conditional_compilation.yml

@@ -265,7 +265,7 @@ jobs:
 
   CC_Build:
     name: Conditional Compilation
-    timeout-minutes: 20
+    timeout-minutes: 25
     needs: [Build, Smart_CI]
     defaults:
       run:

.github/workflows/windows_vs2019_release.yml

@@ -286,7 +286,7 @@ jobs:
 
           # Find and install the dev OV wheel
           $ovDevWheelPath=Get-ChildItem -Path "${{ env.INSTALL_DIR }}\tools" -Filter openvino_dev*.whl | % { $_.FullName }
-          python3 -m pip install "$ovDevWheelPath[mxnet,caffe,kaldi,onnx,tensorflow2,pytorch]"
+          python3 -m pip install "$ovDevWheelPath[caffe,kaldi,onnx,tensorflow2,pytorch]"
 
       - name: Install Python API tests dependencies
         run: |
@@ -316,7 +316,7 @@ jobs:
         if: fromJSON(needs.smart_ci.outputs.affected_components).MO.test
         shell: cmd
         run: |
-          python3 -m pytest -s ${{ env.INSTALL_TEST_DIR }}/mo/unit_tests --ignore=${{ env.INSTALL_TEST_DIR }}/mo/unit_tests/mo/front/mxnet --junitxml=${{ env.INSTALL_TEST_DIR }}/TEST-ModelOptimizer.xml
+          python3 -m pytest -s ${{ env.INSTALL_TEST_DIR }}/mo/unit_tests --junitxml=${{ env.INSTALL_TEST_DIR }}/TEST-ModelOptimizer.xml
 
       - name: Install Python Layer tests dependencies
         run: |
@@ -491,7 +491,7 @@ jobs:
       - name: Intel CPU plugin func tests (parallel)
         run: |
           . "${{ env.INSTALL_DIR }}/setupvars.ps1"
-          python3 ${{ env.PARALLEL_TEST_SCRIPT }} -e ${{ env.INSTALL_TEST_DIR }}/ov_cpu_func_tests.exe -c ${{ env.PARALLEL_TEST_CACHE }} -w ${{ env.INSTALL_TEST_DIR }} -s suite -- --gtest_filter=*smoke*
+          python3 ${{ env.PARALLEL_TEST_SCRIPT }} -e ${{ env.INSTALL_TEST_DIR }}/ov_cpu_func_tests.exe -c ${{ env.PARALLEL_TEST_CACHE }} -w ${{ env.INSTALL_TEST_DIR }} -s suite -rf 0 -- --gtest_filter=*smoke*
         timeout-minutes: 60
 
       - name: Save tests execution time

docs/articles_en/about-openvino/release-notes-openvino.rst

@@ -305,21 +305,21 @@ Known Issues
 | ID: n/a
 | Description:
 |   For ARM binaries, the `1.16 ONNX library <https://vcpkg.link/ports/onnx/versions>`__
-    is not yet available, while the current latest has shown two significant vulnerabilities:
+    is not yet available.
+|   The ONNX library for ARM, version 1.15, does not include the latest
+    functional and security updates. OpenVINO, version 2024.5 is targeted to be released in
+    Q4 2024 and will include additional functional and security updates. Users should update to
+    the latest version as it becomes available.
+|   The current vulnerabilities are:
     `CVE-2024-27318 <https://nvd.nist.gov/vuln/detail/CVE-2024-27318>`__ and
     `CVE-2024-27319 <https://nvd.nist.gov/vuln/detail/CVE-2024-27319>`__.
-    The vulnerabilities are less severe in the context of OpenVINO and will be fixed as soon as
-    the most recent version of the library is available for ARM, expected at the 2024.5 release.
-
-
-
-
-
-
-
-
-
 
+| **Component: Kaldi**
+| ID: n/a
+| Description:
+|   There is a known issue with the Kaldi DL framework support on the Python version 3.12 due
+    to the numpy version incompatibilities. As Kaldi support in OpenVINO is currently deprecated
+    and will be discontinued with version 2025.0, the issue will not be addressed.
 
 
 Previous 2024 releases

docs/articles_en/assets/snippets/ov_extensions.py

@@ -47,7 +47,7 @@ def conversion(node):
 #! [py_frontend_extension_aten_hardtanh]
 import torch
 from openvino.frontend import ConversionExtension, NodeContext
-from openvino.tools.mo import convert_model
+from openvino import convert_model
 
 
 class HardTanh(torch.nn.Module):
@@ -69,5 +69,5 @@ def convert_hardtanh(node: NodeContext):
 
 model = HardTanh(min_val=0.1, max_val=2.0)
 hardtanh_ext = ConversionExtension("aten::hardtanh", convert_hardtanh)
-ov_model = convert_model(input_model=model, extensions=[hardtanh_ext])
+ov_model = convert_model(input_model=model, extension=[hardtanh_ext])
 #! [py_frontend_extension_aten_hardtanh]

docs/articles_en/documentation/legacy-features.rst

@@ -78,7 +78,7 @@ offering.
 Discontinued:
 #############
 
-.. dropdown:: Apache MXNet, Caffe, and Kaldi model formats
+.. dropdown:: Caffe, and Kaldi model formats
 
    |   *New solution:* conversion to ONNX via external tools
    |   *Old solution:* model support discontinued with OpenVINO 2024.0

docs/articles_en/documentation/legacy-features/transition-legacy-conversion-api/legacy-conversion-api/[legacy]-model-optimizer-faq.rst

@@ -234,7 +234,7 @@ Q14. What does the message "Cannot infer shape for node {} because there is no C
 Q15. What does the message "Framework name can not be deduced from the given options. Use --framework to choose one of Caffe, TensorFlow, MXNet" mean?
 ######################################################################################################################################################
 
-**A:** You have run Model Optimizer without a flag ``--framework caffe|tf|mxnet``. Model Optimizer tries to deduce the framework by the extension of input model file (``.pb`` for TensorFlow, ``.caffemodel`` for Caffe, ``.params`` for Apache MXNet). Your input model might have a different extension and you need to explicitly set the source framework. For example, use ``--framework caffe``.
+**A:** You have run Model Optimizer without a flag ``--framework caffe|tf``. Model Optimizer tries to deduce the framework by the extension of input model file (``.pb`` for TensorFlow, ``.caffemodel`` for Caffe, ``.params`` for Apache MXNet). Your input model might have a different extension and you need to explicitly set the source framework. For example, use ``--framework caffe``.
 
 .. _question-16:
 

docs/dev/ov_dependencies.txt

@@ -147,7 +147,6 @@ mccabe
 mistune
 mongo-python-driver
 more-itertools
-mxnet
 mypy
 mypy-extensions
 networkx

docs/dev/pypi_publish/pypi-openvino-dev.md

@@ -83,7 +83,6 @@ pip install openvino-dev[extras]
 | :-------------------------------| :------------------------------------------------------------------------------- |
 | caffe                           |   [Caffe*](https://caffe.berkeleyvision.org/)                                    |
 | kaldi                           |   [Kaldi*](https://github.com/kaldi-asr/kaldi)                                   |
-| mxnet                           |   [Apache MXNet*](https://mxnet.apache.org/)                                     |
 | onnx                            |   [ONNX*](https://github.com/microsoft/onnxruntime/)                             |
 | pytorch                         |   [PyTorch*](https://pytorch.org/)                                               |
 | tensorflow                      |   [TensorFlow* 1.x](https://www.tensorflow.org/versions#tensorflow_1)            |
@@ -149,14 +148,14 @@ Users in China might encounter errors while downloading sources via PIP during O
 If you use zsh (Z shell) interpreter, that is the default shell for macOS starting with version 10.15 (Catalina), you may encounter the following error while installing `openvino-dev` package with extras:
 
 ```sh
-pip install openvino-dev[tensorflow2,mxnet,caffe]
-zsh: no matches found: openvino-dev[tensorflow2,mxnet,caffe]
+pip install openvino-dev[tensorflow2,caffe]
+zsh: no matches found: openvino-dev[tensorflow2,caffe]
 ```
 
 By default zsh interprets square brackets as an expression for pattern matching. To resolve this issue, you need to escape the command with quotes:
 
 ```sh
-pip install 'openvino-dev[tensorflow2,mxnet,caffe]'
+pip install 'openvino-dev[tensorflow2,caffe]'
 ```
 
 To avoid such issues you can also disable globbing for PIP commands by defining an alias in `~/.zshrc` file:

docs/sphinx_setup/_static/download/supported_models.csv

@@ -211,11 +211,6 @@ facenet,Object Detection,onnx,FP16,,,+
 facenet,Object Detection,onnx,FP16-INT8,,,+
 facenet-20180408-102900,Object Detection,tf,FP16,+,+,+
 facenet-20180408-102900,Object Detection,tf,FP32,+,+,
-face-recognition-mobilefacenet-arcface,Object Detection,mxnet,FP16,,,+
-face-recognition-mobilefacenet-arcface,Object Detection,mxnet,FP16-INT8,,,+
-face-recognition-resnet50-arcface,Object Detection,mxnet,FP16,+,+,+
-face-recognition-resnet50-aws,Object Detection,mxnet,FP16,,,+
-face-recognition-resnet50-aws,Object Detection,mxnet,FP32,,,+
 face-reidentification-retail-0095,Object Detection,onnx,FP16,,,+
 face-reidentification-retail-0095,Object Detection,onnx,FP16-INT8,,,+
 facial-landmarks-35-adas-0002,Object Detection,caffe,FP16,,,+
@@ -428,8 +423,6 @@ ocrnet-hrnet-w18,Text Detection,paddle,FP16-INT8,+,+,
 ocrnet-hrnet-w18,Text Detection,paddle,FP32,+,,
 ocrnet-hrnet-w48,Text Detection,paddle,FP16,+,,
 ocrnet-hrnet-w48,Text Detection,paddle,FP32,+,,
-octave-resnext-101-0.25,Image Classification,mxnet,FP16,+,+,+
-octave-resnext-101-0.25,Image Classification,mxnet,FP32,+,,
 openchat-3.6-8b-20240522,Large Language Model,pytorch,intel-optimum default,,+,
 open-closed-eye-0001,Image Classification,onnx,FP16,,,+
 open-closed-eye-0001,Image Classification,onnx,FP16-INT8,,,+
@@ -629,8 +622,6 @@ squeezenet1.0-12,Image Classification,onnx,FP16,,,+
 squeezenet1.0-12,Image Classification,onnx,FP32,+,+,+
 squeezenet1.1-caffe,Image Classification,caffe,FP16,+,+,+
 squeezenet1.1-caffe,Image Classification,caffe,FP32,+,+,+
-squeezenet1.1-mxnet,Image Classification,mxnet,FP16,+,+,+
-squeezenet1.1-mxnet,Image Classification,mxnet,FP32,+,+,+
 squeezenet1.1-onnx,Image Classification,onnx,FP32,+,,
 srgan-onnx,"Image Processing, Enhancement",onnx,FP16,+,,+
 srgan-tf,"Image Processing, Enhancement",tf,FP16,+,+,

docs/sphinx_setup/_static/html/banner_data.json

@@ -1,6 +1,6 @@
 {
-    "p": "",
+    "p": "This page is a nightly version. It may be incomplete or faulty in both content and functionality. <a href='https://docs.openvino.ai/'>Go to the most recent official documentation version, 2024.</a>",
     "version": "1",
     "style": "warning",
-    "styles": "warning info critical"
-}
+    "_styles": "warning info critical"
+}

src/common/transformations/src/transformations/transpose_sinking/ts_unsqueeze.cpp

@@ -190,9 +190,19 @@ TSUnsqueezeBackward::TSUnsqueezeBackward() {
                 return false;
             }
         } else {
-            auto rank = main_node->get_output_partial_shape(0).rank();
-            non_negative_axes =
-                util::try_get_normalized_axis_vector(unsqueeze_axes->get_tensor_view(), rank, *main_node);
+            const auto& axes = unsqueeze_axes->cast_vector<int64_t>();
+            if (std::all_of(axes.begin(), axes.end(), [](int64_t axis) {
+                    return axis >= 0;
+                })) {
+                non_negative_axes = std::vector<size_t>(axes.begin(), axes.end());
+            } else {
+                auto rank = main_node->get_output_partial_shape(0).rank();
+                if (rank.is_dynamic()) {
+                    return false;
+                }
+                non_negative_axes =
+                    util::try_get_normalized_axis_vector(unsqueeze_axes->get_tensor_view(), rank, *main_node);
+            }
         }
 
         auto transpose_order_values = transpose_order->cast_vector<size_t>();

src/common/transformations/tests/transpose_sinking/ts_common_test.cpp

@@ -1636,6 +1636,47 @@ auto test_backward_reshape_unsqueeze = []() {
 INSTANTIATE_TEST_SUITE_P(TransposeSinkingCommonReshapeUnsqueezeBackward,
                          TSTestFixture,
                          test_backward_reshape_unsqueeze());
+
+auto test_backward_unsqueeze_dyn_rank = []() {
+    TestCase test_case;
+
+    // Initialize common attributes
+    test_case.transformation = CREATE_PASS_FACTORY(TSUnsqueezeBackward);
+    test_case.num_main_ops = {1};
+    test_case.inputs_to_main = {
+        parameter(element::f32, PartialShape::dynamic()),
+        constant<int64_t>(element::i32, {2}, {-1}),
+    };
+
+    auto dyn_transpose = [](const vector<size_t>& idxs, const OutputVector& out_vec) -> OutputVector {
+        OutputVector result = out_vec;
+        for (const auto& idx : idxs) {
+            const auto& out = out_vec[idx];
+
+            // fill the order const with the stub values {-1, -2}
+            auto order = make_shared<Constant>(element::i32, Shape{2}, vector<int64_t>{-1, -2});
+            auto transpose = make_shared<Transpose>(out, order);
+            result[idx] = transpose;
+        }
+        return result;
+    };
+
+    // Test model description:
+    test_case.model.main_op = {CREATE_BINARY_FACTORY(Unsqueeze)};
+    test_case.model.preprocess_outputs_of_main = {{dyn_transpose}, {{0}}};
+    test_case.model.model_template = create_model;
+
+    // Ref model description, the same as the original model, the transformation is not applied
+    // it's expected.
+    test_case.model_ref.main_op = {CREATE_BINARY_FACTORY(Unsqueeze)};
+    test_case.model_ref.preprocess_outputs_of_main = {{dyn_transpose}, {{0}}};
+    test_case.model_ref.model_template = create_model;
+    return wrapper(test_case);
+};
+
+INSTANTIATE_TEST_SUITE_P(TransposeSinkingCommonUnsqueezeBackwardDynRank,
+                         TSTestFixture,
+                         test_backward_unsqueeze_dyn_rank());
 }  // namespace common
 }  // namespace testing
 }  // namespace transpose_sinking

src/plugins/intel_cpu/src/emitters/plugin/aarch64/jit_eltwise_emitters.cpp

@@ -1198,6 +1198,58 @@ std::set<std::vector<element::Type>> jit_logical_not_emitter::get_supported_prec
     return {{element::f32}};
 }
 
+/// LOGICAL_XOR ///
+jit_logical_xor_emitter::jit_logical_xor_emitter(dnnl::impl::cpu::aarch64::jit_generator* host,
+                                                 dnnl::impl::cpu::aarch64::cpu_isa_t host_isa,
+                                                 const std::shared_ptr<ov::Node>& node)
+        : jit_emitter(host, host_isa, node, get_arithmetic_binary_exec_precision(node)) {
+    prepare_table();
+}
+
+jit_logical_xor_emitter::jit_logical_xor_emitter(dnnl::impl::cpu::aarch64::jit_generator* host,
+                                                 dnnl::impl::cpu::aarch64::cpu_isa_t host_isa,
+                                                 const ov::element::Type exec_prc)
+    : jit_emitter(host, host_isa, exec_prc) {
+    prepare_table();
+}
+
+size_t jit_logical_xor_emitter::get_inputs_count() const { return 2; }
+
+size_t jit_logical_xor_emitter::get_aux_vecs_count() const { return 1; }
+
+size_t jit_logical_xor_emitter::get_aux_gprs_count() const { return 1; }
+
+void jit_logical_xor_emitter::emit_impl(const std::vector<size_t> &in_vec_idxs, const std::vector<size_t> &out_vec_idxs) const {
+    if (host_isa_ == dnnl::impl::cpu::aarch64::asimd) {
+        emit_isa<dnnl::impl::cpu::aarch64::asimd>(in_vec_idxs, out_vec_idxs);
+    } else {
+        OV_CPU_JIT_EMITTER_THROW("Can't create jit eltwise kernel");
+    }
+}
+
+template <dnnl::impl::cpu::aarch64::cpu_isa_t isa>
+void jit_logical_xor_emitter::emit_isa(const std::vector<size_t> &in_vec_idxs, const std::vector<size_t> &out_vec_idxs) const {
+    OV_CPU_JIT_EMITTER_ASSERT(exec_prc_ == ov::element::f32, "unsupported precision: " + exec_prc_.to_string());
+
+    using TReg = typename dnnl::impl::cpu::aarch64::cpu_isa_traits<isa>::TReg;
+    const TReg src1 = TReg(in_vec_idxs[0]);
+    const TReg src2 = TReg(in_vec_idxs[1]);
+    const TReg dst = TReg(out_vec_idxs[0]);
+    const TReg aux = TReg(aux_vec_idxs[0]);
+
+    h->eor(dst.b16, src1.b16, src2.b16);
+    h->ld1r(aux.s, table_val2("one"));
+    h->and_(dst.b16, dst.b16, aux.b16);
+}
+
+void jit_logical_xor_emitter::register_table_entries() {
+    push_arg_entry_of("one", 0x3f800000, true);
+}
+
+std::set<std::vector<element::Type>> jit_logical_xor_emitter::get_supported_precisions(const std::shared_ptr<ov::Node>& node) {
+    return {{element::f32, element::f32}};
+}
+
 /// MAX ///
 jit_maximum_emitter::jit_maximum_emitter(dnnl::impl::cpu::aarch64::jit_generator* host,
                                          dnnl::impl::cpu::aarch64::cpu_isa_t host_isa,

src/plugins/intel_cpu/src/emitters/plugin/aarch64/jit_eltwise_emitters.hpp

@@ -549,6 +549,34 @@ class jit_logical_not_emitter : public jit_emitter {
     void register_table_entries() override;
 };
 
+class jit_logical_xor_emitter : public jit_emitter {
+public:
+    jit_logical_xor_emitter(dnnl::impl::cpu::aarch64::jit_generator *host,
+                            dnnl::impl::cpu::aarch64::cpu_isa_t host_isa,
+                            const ov::element::Type exec_prc = ov::element::f32);
+
+    jit_logical_xor_emitter(dnnl::impl::cpu::aarch64::jit_generator *host,
+                            dnnl::impl::cpu::aarch64::cpu_isa_t host_isa,
+                            const std::shared_ptr<ov::Node>& n);
+
+    size_t get_inputs_count() const override;
+
+    size_t get_aux_vecs_count() const override;
+
+    size_t get_aux_gprs_count() const override;
+
+    static std::set<std::vector<element::Type>> get_supported_precisions(
+            const std::shared_ptr<ov::Node>& node = nullptr);
+
+private:
+    void emit_impl(const std::vector<size_t> &in_vec_idxs, const std::vector<size_t> &out_vec_idxs) const override;
+
+    template <dnnl::impl::cpu::aarch64::cpu_isa_t isa>
+    void emit_isa(const std::vector<size_t> &in_vec_idxs, const std::vector<size_t> &out_vec_idxs) const;
+
+    void register_table_entries() override;
+};
+
 class jit_mod_emitter : public jit_emitter {
 public:
     jit_mod_emitter(dnnl::impl::cpu::aarch64::jit_generator *host,

src/plugins/intel_cpu/src/nodes/executors/aarch64/jit_eltwise.cpp

@@ -35,6 +35,7 @@ bool JitEltwiseExecutor::isSupported(
                                      Algorithm::EltwiseIsNaN,
                                      Algorithm::EltwiseLessEqual,
                                      Algorithm::EltwiseLogicalNot,
+                                     Algorithm::EltwiseLogicalXor,
                                      Algorithm::EltwiseMaximum,
                                      Algorithm::EltwiseMinimum,
                                      Algorithm::EltwiseMish,