Merge branch 'master' into as/npuw_parallel_for_improve

.github/workflows/job_pytorch_layer_tests.yml

@@ -7,10 +7,6 @@ on:
         description: 'Machine on which the tests would run'
         type: string
         required: true
-      shell:
-        description: "shell to override the default shell settings in the runner's operating system."
-        type: string
-        required: true
       container:
         description: 'JSON to be converted to the value of the "container" configuration for the job'
         type: string
@@ -20,12 +16,15 @@ on:
         description: 'Components that are affected by changes in the commit defined by the Smart CI Action'
         type: string
         required: true
+      python-version:
+        description: 'Python version to setup. E.g., "3.11"'
+        type: string
+        required: true
 
 permissions: read-all
 
 env:
   PIP_CACHE_PATH: /mount/caches/pip/linux
-  PYTHON_VERSION: '3.11'
 
 jobs:
   PyTorch_Layer_Tests:
@@ -35,7 +34,7 @@ jobs:
     container: ${{ fromJSON(inputs.container) }}
     defaults:
       run:
-        shell: ${{ inputs.shell }}
+        shell: ${{ contains(inputs.runner, 'win') && 'pwsh' || 'bash' }}
     env:
       DEBIAN_FRONTEND: noninteractive # to prevent apt-get from waiting user input
       OPENVINO_REPO: ${{ github.workspace }}/openvino
@@ -55,12 +54,6 @@ jobs:
           name: openvino_tests
           path: ${{ env.INSTALL_TEST_DIR }}
 
-      - name: Download OpenVINO tokenizers extension
-        uses: actions/download-artifact@fa0a91b85d4f404e444e00e005971372dc801d16 # v4.1.8
-        with:
-          name: openvino_tokenizers_wheel
-          path: ${{ env.INSTALL_DIR }}
-
       # Needed as ${{ github.workspace }} is not working correctly when using Docker
       - name: Setup Variables
         if: runner.os != 'Windows'
@@ -98,10 +91,10 @@ jobs:
           sparse-checkout-cone-mode: false
           path: 'openvino'
 
-      - name: Setup Python ${{ env.PYTHON_VERSION }}
+      - name: Setup Python ${{ inputs.python-version }}
         uses: ./openvino/.github/actions/setup_python
         with:
-          version: ${{ env.PYTHON_VERSION }}
+          version: ${{ inputs.python-version }}
           pip-cache-path: ${{ runner.os == 'Linux' && env.PIP_CACHE_PATH || '' }}
           should-setup-pip-paths: ${{ runner.os == 'Linux' }}
           self-hosted-runner: ${{ runner.os == 'Linux' }}
@@ -112,43 +105,39 @@ jobs:
           # Install the core OV wheel
           python3 -m pip install ${INSTALL_DIR}/tools/openvino-*.whl
 
-          # Install the core OV Tokenizers wheel
-          python3 -m pip install ${INSTALL_DIR}/openvino_tokenizers-*.whl
-
       - name: Install OpenVINO Python wheels (Windows)
         if: runner.os == 'Windows'
         run: |
           # Find and install the core OV wheel
           $ovCoreWheelPath=Get-ChildItem -Path ${{ env.INSTALL_DIR }}\tools -Filter openvino-*.whl | % { $_.FullName }
           python3 -m pip install "$ovCoreWheelPath"
 
-          # Find and install the core OV Tokenizers wheel
-          $ovCoreWheelPath=Get-ChildItem -Path ${{ env.INSTALL_DIR }} -Filter openvino_tokenizers-*.whl | % { $_.FullName }
-          python3 -m pip install "$ovCoreWheelPath"
-
       - name: Install Pytorch Layer tests dependencies
         run: |
           # pytorch test requirements
           python3 -m pip install -r ${{ env.INSTALL_TEST_DIR }}/requirements_pytorch
 
       - name: PyTorch Layer Tests
         if: ${{ fromJSON(inputs.affected-components).PyTorch_FE.test && runner.arch != 'ARM64' }} # Ticket: 126287, 142196
-        run: python3 -m pytest ${{ env.LAYER_TESTS_INSTALL_DIR }}/pytorch_tests -n logical -m precommit --junitxml=${{ env.INSTALL_TEST_DIR }}/TEST-pytorch.xml
+        # due to CVS-152795, parallel run is not possible on Windows
+        run: python3 -m pytest ${{ env.LAYER_TESTS_INSTALL_DIR }}/pytorch_tests ${PARALLEL} -m precommit --junitxml=${{ env.INSTALL_TEST_DIR }}/TEST-pytorch.xml
         env:
           TEST_DEVICE: CPU
           TEST_PRECISION: FP32
+          PARALLEL: ${{ runner.os == 'Windows' && ' ' || '-n logical'}}
 
       - name: PyTorch torch.export Layer Tests
-        if: ${{ fromJSON(inputs.affected-components).PyTorch_FE.test && runner.arch != 'ARM64' }} # Ticket: 126287
+        if: ${{ fromJSON(inputs.affected-components).PyTorch_FE.test && runner.arch != 'ARM64' && runner.os != 'Windows' }} # Ticket: 126287
         run: |
-          python3 -m pytest ${{ env.LAYER_TESTS_INSTALL_DIR }}/pytorch_tests -n logical -m precommit_torch_export --junitxml=${{ env.INSTALL_TEST_DIR }}/TEST-pytorch.xml
+          python3 -m pytest ${{ env.LAYER_TESTS_INSTALL_DIR }}/pytorch_tests ${PARALLEL} -m precommit_torch_export --junitxml=${{ env.INSTALL_TEST_DIR }}/TEST-pytorch.xml
         env:
           TEST_DEVICE: CPU
           TEST_PRECISION: FP32
           PYTORCH_TRACING_MODE: EXPORT
+          PARALLEL: ${{ runner.os == 'Windows' && ' ' || '-n logical'}}
 
       - name: PyTorch torch.compile TORCHFX Layer Tests
-        if: ${{ fromJSON(inputs.affected-components).PyTorch_FE.test && runner.os != 'macOS' && runner.arch != 'ARM64' }} # Ticket: 126287
+        if: ${{ fromJSON(inputs.affected-components).PyTorch_FE.test && runner.os != 'macOS' && runner.arch != 'ARM64' && runner.os != 'Windows' }} # Ticket: 126287
         run: |
           python3 -m pytest ${{ env.LAYER_TESTS_INSTALL_DIR }}/pytorch_tests -m precommit_fx_backend --junitxml=${{ env.INSTALL_TEST_DIR }}/TEST-pytorch.xml
         env:

.github/workflows/job_tensorflow_layer_tests.yml

@@ -7,10 +7,6 @@ on:
         description: 'Machine on which the tests would run'
         type: string
         required: true
-      shell:
-        description: "shell to override the default shell settings in the runner's operating system."
-        type: string
-        required: true
       container:
         description: 'JSON to be converted to the value of the "container" configuration for the job'
         type: string
@@ -20,12 +16,15 @@ on:
         description: 'Components that are affected by changes in the commit defined by the Smart CI Action'
         type: string
         required: true
+      python-version:
+        description: 'Python version to setup. E.g., "3.11"'
+        type: string
+        required: true
 
 permissions: read-all
 
 env:
   PIP_CACHE_PATH: /mount/caches/pip/linux
-  PYTHON_VERSION: '3.11'
 
 jobs:
   TensorFlow_Layer_Tests:
@@ -35,7 +34,7 @@ jobs:
     container: ${{ fromJSON(inputs.container) }}
     defaults:
       run:
-        shell: ${{ inputs.shell }}
+        shell: ${{ contains(inputs.runner, 'win') && 'pwsh' || 'bash' }}
     env:
       DEBIAN_FRONTEND: noninteractive # to prevent apt-get from waiting user input
       OPENVINO_REPO: ${{ github.workspace }}/openvino
@@ -98,10 +97,10 @@ jobs:
           sparse-checkout-cone-mode: false
           path: 'openvino'
 
-      - name: Setup Python ${{ env.PYTHON_VERSION }}
+      - name: Setup Python ${{ inputs.python-version }}
         uses: ./openvino/.github/actions/setup_python
         with:
-          version: ${{ env.PYTHON_VERSION }}
+          version: ${{ inputs.python-version }}
           pip-cache-path: ${{ runner.os == 'Linux' && env.PIP_CACHE_PATH || '' }}
           should-setup-pip-paths: ${{ runner.os == 'Linux' }}
           self-hosted-runner: ${{ runner.os == 'Linux' }}

.github/workflows/linux_arm64.yml

@@ -173,19 +173,19 @@ jobs:
     uses: ./.github/workflows/job_tensorflow_layer_tests.yml
     with:
       runner: 'aks-linux-16-cores-arm'
-      shell: bash
       container: '{"image": "${{ fromJSON(needs.docker.outputs.images).ov_test.ubuntu_20_04_arm64 }}", "volumes": ["/mount:/mount"]}'
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   Pytorch_Layer_Tests:
     name: Pytorch Layer Tests
-    needs: [ Build, Docker, Smart_CI, Openvino_tokenizers ]
+    needs: [ Build, Docker, Smart_CI ]
     uses: ./.github/workflows/job_pytorch_layer_tests.yml
     with:
       runner: 'aks-linux-16-cores-arm'
-      shell: bash
       container: '{"image": "${{ fromJSON(needs.docker.outputs.images).ov_test.ubuntu_20_04_arm64 }}", "volumes": ["/mount:/mount"]}'
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   CPU_Functional_Tests:
     name: CPU functional tests

.github/workflows/mac.yml

@@ -276,17 +276,17 @@ jobs:
     uses: ./.github/workflows/job_tensorflow_layer_tests.yml
     with:
       runner: 'macos-13'
-      shell: bash
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   Pytorch_Layer_Tests:
     name: Pytorch Layer Tests
-    needs: [ Build, Smart_CI, Openvino_tokenizers ]
+    needs: [ Build, Smart_CI ]
     uses: ./.github/workflows/job_pytorch_layer_tests.yml
     with:
       runner: 'macos-13'
-      shell: bash
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   CPU_Functional_Tests:
     name: CPU functional tests

.github/workflows/mac_arm64.yml

@@ -275,17 +275,17 @@ jobs:
     uses: ./.github/workflows/job_tensorflow_layer_tests.yml
     with:
       runner: 'macos-13-xlarge'
-      shell: bash
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   Pytorch_Layer_Tests:
     name: Pytorch Layer Tests
-    needs: [ Build, Smart_CI, Openvino_tokenizers ]
+    needs: [ Build, Smart_CI ]
     uses: ./.github/workflows/job_pytorch_layer_tests.yml
     with:
       runner: 'macos-13-xlarge'
-      shell: bash
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   CPU_Functional_Tests:
     name: CPU functional tests

.github/workflows/ubuntu_22.yml

@@ -305,19 +305,19 @@ jobs:
     uses: ./.github/workflows/job_tensorflow_layer_tests.yml
     with:
       runner: 'aks-linux-4-cores-16gb'
-      shell: bash
       container: '{"image": "${{ fromJSON(needs.docker.outputs.images).ov_test.ubuntu_22_04_x64 }}", "volumes": ["/mount:/mount"]}'
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   Pytorch_Layer_Tests:
     name: Pytorch Layer Tests
-    needs: [ Docker, Build, Smart_CI, Openvino_tokenizers ]
+    needs: [ Docker, Build, Smart_CI ]
     uses: ./.github/workflows/job_pytorch_layer_tests.yml
     with:
       runner: 'aks-linux-4-cores-16gb'
-      shell: bash
       container: '{"image": "${{ fromJSON(needs.docker.outputs.images).ov_test.ubuntu_22_04_x64 }}", "volumes": ["/mount:/mount"]}'
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   CPU_Functional_Tests:
     name: CPU functional tests

.github/workflows/ubuntu_24.yml

@@ -133,6 +133,16 @@ jobs:
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
       python-version: '3.12'
 
+  Pytorch_Layer_Tests:
+    name: Pytorch Layer Tests
+    needs: [ Docker, Build, Smart_CI ]
+    uses: ./.github/workflows/job_pytorch_layer_tests.yml
+    with:
+      runner: 'aks-linux-4-cores-16gb'
+      container: '{"image": "${{ fromJSON(needs.docker.outputs.images).ov_test.ubuntu_24_04_x64 }}", "volumes": ["/mount:/mount"]}'
+      affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.12'
+
   Overall_Status:
     name: ci/gha_overall_status_ubuntu_24
     needs: [Smart_CI, Build, Debian_Packages, Samples, Python_Unit_Tests]

.github/workflows/windows_vs2019_release.yml

@@ -404,17 +404,17 @@ jobs:
     uses: ./.github/workflows/job_tensorflow_layer_tests.yml
     with:
       runner: 'aks-win-8-cores-16gb'
-      shell: pwsh
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   Pytorch_Layer_Tests:
     name: Pytorch Layer Tests
-    needs: [ Build, Smart_CI, Openvino_tokenizers ]
+    needs: [ Build, Smart_CI ]
     uses: ./.github/workflows/job_pytorch_layer_tests.yml
     with:
       runner: 'aks-win-8-cores-16gb'
-      shell: pwsh
       affected-components: ${{ needs.smart_ci.outputs.affected_components }}
+      python-version: '3.11'
 
   CXX_Unit_Tests:
     name: C++ unit tests

src/plugins/intel_gpu/src/kernel_selector/kernels/fully_connected/fully_connected_kernel_bf_tiled.cpp

@@ -534,6 +534,7 @@ JitConstants FullyConnected_bf_tiled::GetJitConstants(const fully_connected_para
     size_t tile_k_ofm_packed = tile_k_ofm;
     size_t quantize_grp_size = get_dynamic_quantize_group_size(params);
 
+    bool add_decompress_scale_post_op = false;
     WeightsType weights_dt = params.weights.GetDType();
     if (weights_dt == WeightsType::UINT4 || weights_dt == WeightsType::INT4) {
         tile_k_ofm_packed /= 2;
@@ -542,7 +543,7 @@ JitConstants FullyConnected_bf_tiled::GetJitConstants(const fully_connected_para
         const size_t scale_group_size = params.weights.IFM().v / params.decompression_scale.Feature().v;
         // Do not use SCALE_POST_OP for SLM kernel, since it demonstrates worse performance
         if (scale_group_size % simd == 0 && !dispatchData.use_slm)
-            jit.AddConstant(MakeJitConstant("DECOMPRESSION_SCALE_POST_OP", 1));
+            add_decompress_scale_post_op = true;
     }
     if (params.weights.GetLayout() == WeightsLayout::os_is_yx_osv32_isv2) {
         jit.AddConstant(MakeJitConstant("W_IDX", "fi * TILE_K + kii"));
@@ -619,6 +620,8 @@ JitConstants FullyConnected_bf_tiled::GetJitConstants(const fully_connected_para
         jit.AddConstant(MakeJitConstant("DQ_TYPE", "char"));
         jit.AddConstant(MakeJitConstant("QUANTIZE_GROUP_SIZE", quantize_grp_size));
     } else {
+        if (add_decompress_scale_post_op)
+            jit.AddConstant(MakeJitConstant("DECOMPRESSION_SCALE_POST_OP", 1));
         jit.AddConstant(MakeJitConstant("DYNAMIC_QUANTIZE", 0));
         jit.AddConstant(MakeJitConstant("QUANTIZE_GROUP_SIZE", min_quantize_grp_size));
     }

tests/layer_tests/pytorch_tests/test_bitwise_ops.py

@@ -4,6 +4,8 @@
 import numpy as np
 import pytest
 import torch
+from packaging import version
+
 from pytorch_layer_test_class import PytorchLayerTest, skip_if_export
 
 
@@ -69,10 +71,12 @@ def forward_not_out(self, tensor_a, out):
     )
     @pytest.mark.parametrize("out", [False, skip_if_export(True)])
     def test_bitwise_mixed_dtypes(
-        self, op_type, out, lhs_dtype, rhs_dtype, lhs_shape, rhs_shape, ie_device, precision, ir_version
+            self, op_type, out, lhs_dtype, rhs_dtype, lhs_shape, rhs_shape, ie_device, precision, ir_version
     ):
         if ie_device == "GPU" and (lhs_dtype != "bool" or rhs_dtype != "bool"):
             pytest.xfail(reason="bitwise ops are not supported on GPU")
+        if out and version.parse(np.__version__) >= version.parse("2.0.0"):
+            pytest.xfail(reason="CVS-154082: incorrect handling out type")
         self._test(
             *self.create_model(op_type, out),
             ie_device,

tests/requirements_pytorch

@@ -1,10 +1,14 @@
+# test ovc with NumPy 2.x on Ubuntu 24 with default Python 3.12
+# test against NumPy 1.x with older Python versions
 # optimum still requires numpy<2.0.0
-numpy==1.26.4
+numpy==1.26.4; python_version < "3.12"
+numpy==2.1.1; python_version >= "3.12"
 torch==2.4.1; platform_system != "Darwin" or platform_machine != "x86_64"
-torch==2.2.0; platform_system == "Darwin" and platform_machine == "x86_64"
+torch==2.2.2; platform_system == "Darwin" and platform_machine == "x86_64"
 --extra-index-url https://download.pytorch.org/whl/cpu
 
-torchvision==0.19.1
+torchvision==0.19.1; platform_system != "Darwin" or platform_machine != "x86_64"
+torchvision==0.17.2; platform_system == "Darwin" and platform_machine == "x86_64"
 # transformers 4.45.1 is available
 # but optimum still requires <4.45.0
 transformers==4.44.2
@@ -13,30 +17,30 @@ pytest-html==4.1.1
 pytest-xdist[psutil]==3.6.1
 defusedxml==0.7.1
 
-auto-gptq==0.7.1; platform_system == "Linux" and platform_machine == "x86_64"
+auto-gptq==0.7.1; platform_system == "Linux" and platform_machine == "x86_64" and python_version < "3.12"
 av==13.0.0
-basicsr==1.4.2
+basicsr==1.4.2; python_version < "3.12"
 datasets==3.0.1
 easyocr==1.7.2
-facexlib==0.3.0
-librosa==0.10.2
-optimum==1.22.0
+facexlib==0.3.0; python_version < "3.12"
+librosa==0.10.2; python_version < "3.12"
+optimum==1.22.0; python_version < "3.12"
 packaging==24.1
 pandas==2.2.3
 protobuf==5.28.2
-pyctcdecode==0.5.0
+pyctcdecode==0.5.0; python_version < "3.12"
 sacremoses==0.1.1
 sentencepiece==0.2.0
 soundfile==0.12.1
-super-image==0.1.7
+super-image==0.1.7; python_version < "3.12"
 timm==1.0.8
 torchaudio==2.4.1
 wheel==0.44.0
 PyYAML==6.0.2
 kornia==0.7.3
 
 # use latest released version once it's available
-git+https://github.com/huggingface/optimum-intel.git@main
+git+https://github.com/huggingface/optimum-intel.git@main; python_version < "3.12"
 # set 'export HF_HUB_ENABLE_HF_TRANSFER=1' to benefits from hf_transfer
 hf_transfer==0.1.8