test oneapi 2025.3

Author: mengfei25

.github/actions/linux-uttest/action.yml

@@ -181,9 +181,6 @@ runs:
           echo -e "[ERROR] XCCL is not enabled"
           exit 1
         fi
-        export CCL_ROOT=$(dirname $(which python))/../
-        export PATH="${CCL_ROOT}/bin/libfabric:${PATH}"
-        export LD_LIBRARY_PATH="${CCL_ROOT}/lib:${LD_LIBRARY_PATH}"
         python run_distributed_local.py \
           2> ${{ github.workspace }}/ut_log/xpu_distributed/xpu_distributed_test_error.log | \
           tee ${{ github.workspace }}/ut_log/xpu_distributed/xpu_distributed_test.log

.github/scripts/build.sh

@@ -46,30 +46,30 @@ cd ${WORKSPACE}/pytorch
 python -m pip install requests
 git submodule sync && git submodule update --init --recursive
 python -m pip install -r requirements.txt
-python -m pip install mkl-static==2025.2.0 mkl-include==2025.2.0
+python -m pip install mkl-static mkl-include
 export USE_STATIC_MKL=1
 if [ "${XPU_ONEAPI_PATH}" == "" ];then
     export PYTORCH_EXTRA_INSTALL_REQUIREMENTS=" \
-        intel-cmplr-lib-rt==2025.2.1 | \
-        intel-cmplr-lib-ur==2025.2.1 | \
-        intel-cmplr-lic-rt==2025.2.1 | \
-        intel-sycl-rt==2025.2.1 | \
-        oneccl-devel==2021.16.1 | \
-        oneccl==2021.16.1 | \
-        impi-rt==2021.16.1 | \
-        onemkl-sycl-blas==2025.2.0 | \
-        onemkl-sycl-dft==2025.2.0 | \
-        onemkl-sycl-lapack==2025.2.0 | \
-        onemkl-sycl-rng==2025.2.0 | \
-        onemkl-sycl-sparse==2025.2.0 | \
-        dpcpp-cpp-rt==2025.2.1 | \
-        intel-opencl-rt==2025.2.1 | \
-        mkl==2025.2.0 | \
-        intel-openmp==2025.2.1 | \
-        tbb==2022.2.0 | \
-        tcmlib==1.4.0 | \
-        umf==0.11.0 | \
-        intel-pti==0.13.1
+        intel-cmplr-lib-rt | \
+        intel-cmplr-lib-ur | \
+        intel-cmplr-lic-rt | \
+        intel-sycl-rt | \
+        oneccl-devel | \
+        oneccl | \
+        impi-rt | \
+        onemkl-sycl-blas | \
+        onemkl-sycl-dft | \
+        onemkl-sycl-lapack | \
+        onemkl-sycl-rng | \
+        onemkl-sycl-sparse | \
+        dpcpp-cpp-rt | \
+        intel-opencl-rt | \
+        mkl | \
+        intel-openmp | \
+        tbb | \
+        tcmlib | \
+        umf | \
+        intel-pti
     "
 fi
 

.github/workflows/_linux_build.yml

@@ -59,7 +59,7 @@ jobs:
     if: ${{ ! endsWith(inputs.pytorch, '_wheel') }}
     runs-on: ${{ needs.runner.outputs.runner_id }}
     container:
-      image: 'pytorch/manylinux2_28-builder:xpu-2.9'
+      image: 'intelgpu/ubuntu-22.04-lts2:2523.31'
       volumes:
         - ${{ github.workspace }}:${{ github.workspace }}
       env:
@@ -72,21 +72,30 @@ jobs:
     steps:
       - name: Install gh-cli
         run: |
+          rm -rf ./*.whl ./*.log
           cat /etc/os-release
           hostname && id
           # install gh
-          dnf install -y 'dnf-command(config-manager)'
-          dnf config-manager --add-repo https://cli.github.com/packages/rpm/gh-cli.repo
-          dnf install -y gh --repo gh-cli
-          gh --version
+          sudo apt update
+          sudo apt install -y gpg-agent wget curl cmake git unzip zip libgl1 zlib1g-dev numactl \
+            libglib2.0-dev rsync jq gcc-11 g++-11 python3-dev python3-venv gh
       - name: Setup python-${{ inputs.python }}
         run: |
           rm -rf /tmp/xpu-tool/myvenv
-          local_python=$(echo ${{ inputs.python }} |awk -F. '{printf("cp%s%s-cp%s%s", $1, $2, $1, $2)}')
-          /opt/python/${local_python}/bin/python -m venv /tmp/xpu-tool/myvenv
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          source $HOME/.local/bin/env
+          uv venv /tmp/xpu-tool/myvenv --python 3.10 --clear
+          source /tmp/xpu-tool/myvenv/bin/activate
           which python && python -V
           which pip && pip list
-          pip install -U pip wheel setuptools
+          uv pip install -U pip wheel setuptools
+      - name: Install oneapi
+        run: |
+          rm -rf /opt/intel/oneapi
+          wget https://registrationcenter-download.intel.com/akdlm/IRC_NAS/aa5447b5-3644-43c8-8ec4-72d53f6ecc19/intel-deep-learning-essentials-2025.3.0.338_offline.sh
+          sudo bash intel-deep-learning-essentials-2025.3.0.338_offline.sh -a -s --eula accept
+          source /opt/intel/oneapi/setvars.sh
+          icpx --version
       - name: Checkout torch-xpu-ops
         uses: actions/checkout@v4
         with:
@@ -113,8 +122,6 @@ jobs:
             TORCH_XPU_OPS_REPO="https://github.com/intel/torch-xpu-ops.git"
             TORCH_XPU_OPS_COMMIT="${{ inputs.torch_xpu_ops }}"
           fi
-          # gcc 11
-          source /opt/rh/gcc-toolset-11/enable
           source ${{ github.workspace }}/torch-xpu-ops/.github/scripts/env.sh
           ${{ github.workspace }}/torch-xpu-ops/.github/scripts/build.sh \
             --WORKSPACE="${{ github.workspace }}" \
@@ -129,8 +136,6 @@ jobs:
           fi
       - name: Build Torchvision and Torchaudio
         run: |
-          # gcc 11
-          source /opt/rh/gcc-toolset-11/enable
           cd ./pytorch
           TORCHVISION_COMMIT_ID="$(cat .github/ci_commit_pins/vision.txt)"
           TORCHAUDIO_COMMIT_ID="$(cat .github/ci_commit_pins/audio.txt)"
@@ -177,9 +182,6 @@ jobs:
               curl -sSL https://raw.githubusercontent.com/intel/intel-xpu-backend-for-triton/${TRITON_COMMIT_ID}/python/triton/__init__.py 2>&1 |\
                       grep '__version__' |head -n 1 |awk -F "'" '{print $2}'
             )"
-            # gcc 13
-            dnf install -y gcc-toolset-13-gcc-c++ zlib-devel
-            source /opt/rh/gcc-toolset-13/enable
             pip install cmake ninja pybind11
             python .github/scripts/build_triton_wheel.py --device xpu --commit-hash ${TRITON_COMMIT_ID} --triton-version ${TRITON_VERSION_NAME} \
               2>&1 |tee ${{ github.workspace }}/build_triton_${TRITON_COMMIT_ID}.log

.github/workflows/_linux_ut.yml

@@ -166,6 +166,7 @@ jobs:
           else
             ut_list="${{ inputs.ut }}"
           fi
+          cp ${{ github.workspace }}/.github/scripts/ut_result_check.sh ./
           for ut_name in ${ut_list}
           do
             cp Known_issue.log.tmp Known_issue.log