Rebase dev/sycl-free-func

.github/ISSUE_TEMPLATE/dynamic-skip.yml

@@ -0,0 +1,63 @@
+name: 🐛 Dynamic skip
+description: Create an issue to skip PR unrelated failures dynamically
+title: "[Bug Skip]: "
+labels: ["skipped"]
+
+body:
+- type: markdown
+  attributes:
+    value: >
+      #### Before submitting a bug, please make sure the issue hasn't been already addressed by searching through [the existing and past issues](https://github.com/pytorch/pytorch/issues?q=is%3Aissue+sort%3Acreated-desc+).
+- type: textarea
+  attributes:
+    label: 🐛 Describe the bug with skip template
+    description: |
+      Please provide a clear and concise description of what the bug is.
+      The template for dynamic skip as below:
+
+      ```python
+      # Template(Check in the github action summary)
+      Cases:
+      [Category],[Class name],[Test name]
+      ```
+
+      ```python
+      # example
+      Cases:
+      op_ut,third_party.torch-xpu-ops.test.xpu.test_transformers_xpu.TestTransformersXPU,test_scaled
+      ```
+
+      If the code is too long (hopefully, it isn't), feel free to put it in a public gist and link it in the issue: https://gist.github.com.
+
+      Please also paste or describe the results you observe instead of the expected results. If you observe an error, please paste the error message including the **full** traceback of the exception. It may be relevant to wrap error messages in ```` ```triple quotes blocks``` ````.
+    placeholder: |
+      A clear and concise description of what the bug is and also align the dynamic template.
+      ```
+      # Skippped cases with dynamic template
+      ```
+
+      ```python
+      # Sample code to reproduce the problem
+      ```
+
+      ```
+      The error message you got, with the full traceback.
+      ```
+  validations:
+    required: true
+- type: textarea
+  attributes:
+    label: Versions
+    description: |
+      Please run the following and paste the output below.
+      ```sh
+      wget https://raw.githubusercontent.com/pytorch/pytorch/main/torch/utils/collect_env.py
+      # For security purposes, please check the contents of collect_env.py before running it.
+      python collect_env.py
+      ```
+  validations:
+    required: true
+- type: markdown
+  attributes:
+    value: >
+      Thanks for contributing 🎉!

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

@@ -57,7 +57,7 @@ runs:
       run: |
         # install pytorch
         if [ $(echo "${{ inputs.pytorch }}" |grep -w "release_wheel" -c) -ne 0 ];then
-          pip install torch torchvision torchaudio --pre --index-url https://download.pytorch.org/whl/xpu
+          pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/xpu
         elif [ $(echo "${{ inputs.pytorch }}" |grep -w "test_wheel" -c) -ne 0 ];then
           pip install torch torchvision torchaudio --pre --index-url https://download.pytorch.org/whl/test/xpu
         elif [ $(echo "${{ inputs.pytorch }}" |grep -w "nightly_wheel" -c) -ne 0 ];then
@@ -77,7 +77,7 @@ runs:
         # apply extra PRs for stock pytorch
         pip install requests
         if [ "${{ github.event_name }}" == "pull_request" ];then
-          python ../torch-xpu-ops/.github/scripts/apply_torch_pr.py -e https://github.com/pytorch/pytorch/pull/152940
+          python ../torch-xpu-ops/.github/scripts/apply_torch_pr.py -e https://github.com/mengfei25/pytorch/pull/27
         else
           python ../torch-xpu-ops/.github/scripts/apply_torch_pr.py
         fi
@@ -99,7 +99,7 @@ runs:
             TORCH_XPU_OPS_COMMIT="${{ inputs.torch_xpu_ops }}"
           fi
         fi
-        if [ "${{ github.event_name }}" == "pull_request" ];then
+        if [ "${{ github.event_name }}" == "pull_request" ] && [[ "${{ inputs.pytorch }}" != *"_wheel" ]];then
           cp -r ${{ github.workspace }}/torch-xpu-ops third_party/torch-xpu-ops
           cd third_party/torch-xpu-ops
         else
@@ -115,6 +115,8 @@ runs:
         pip install pandas psutil scipy pyyaml
         cd pytorch
         if [[ "${{ inputs.suite }}" == *"huggingface"* ]];then
+          # for new LLM models
+          pip install accelerate
           pip install -r .ci/docker/ci_commit_pins/huggingface-requirements.txt || pip install transformers==4.54.0 soxr==0.5.0
           TRANSFORMERS_VERSION_ID="$(python -c 'import os; os.chdir("/tmp"); import transformers; print(transformers.__version__)')"
         elif [[ "${{ inputs.suite }}" == *"timm_models"* ]];then
@@ -139,7 +141,7 @@ runs:
           fi
           # for dlrm
           pip install pyre-extensions
-          curl -fsSL https://raw.githubusercontent.com/facebookresearch/dlrm/refs/heads/torchrec-dlrm/requirements.txt |xargs pip install --no-deps
+          curl -fsSL https://raw.githubusercontent.com/facebookresearch/dlrm/refs/heads/torchrec-dlrm/requirements.txt |xargs pip install
           # for soft_actor_critic, temp fix
           pip install git+https://github.com/nocoding03/gym@fix-np
           cd ../pytorch
@@ -152,10 +154,13 @@ runs:
           TORCHBENCH_COMMIT_ID="$(git rev-parse --short  HEAD)"
           sed -i 's/^ *pynvml.*//' requirements.txt
           pip install -r requirements.txt
-          python install.py --continue_on_fail
+          # python install.py --continue_on_fail
+          echo "PYTHONPATH=${PWD}:${PYTHONPATH}" >> ${GITHUB_ENV}
+          pip install dominate
+          python install.py Super_SloMo
           # for dlrm
           pip install pyre-extensions
-          curl -fsSL https://raw.githubusercontent.com/facebookresearch/dlrm/refs/heads/torchrec-dlrm/requirements.txt |xargs pip install --no-deps
+          curl -fsSL https://raw.githubusercontent.com/facebookresearch/dlrm/refs/heads/torchrec-dlrm/requirements.txt |xargs pip install
           cd ../pytorch
         else
           pip install -r ./.ci/docker/requirements-ci.txt
@@ -170,6 +175,11 @@ runs:
         else
           pip install torchao --pre --index-url https://download.pytorch.org/whl/nightly/xpu
         fi
+        if [ "${{ inputs.suite }}" != "None" ];then
+          # To install numpy 1.x for benchmarks as CUDA
+          # yolov requires numpy>=1.23
+          pip install -U numpy==1.26.4
+        fi
     - name: Torch Config
       shell: bash -xe {0}
       run: |

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

@@ -155,7 +155,7 @@ runs:
           tee ${{ github.workspace }}/ut_log/xpu_profiling/test_profiler_tree.log
 
     - name: xpu_distributed
-      shell: timeout 3600 bash -xeu -o pipefail {0}
+      shell: timeout 36000 bash -xeu -o pipefail {0}
       if: ${{ inputs.ut_name == 'xpu_distributed' }}
       run: |
         xpu-smi topology -m
@@ -166,9 +166,13 @@ runs:
           echo -e "[ERROR] XCCL is not enabled"
           exit 1
         fi
+        export CCL_ROOT=$(dirname $(which python))/../
+        export PATH="${CCL_ROOT}/bin/libfabric:${PATH}"
+        export LD_LIBRARY_PATH="${CCL_ROOT}/lib:${LD_LIBRARY_PATH}"
         python run_distributed.py \
           2> ${{ github.workspace }}/ut_log/xpu_distributed/xpu_distributed_test_error.log | \
           tee ${{ github.workspace }}/ut_log/xpu_distributed/xpu_distributed_test.log
+        find ../ -type f -name "*.xml" -exec cp {} ${{ github.workspace }}/ut_log/ \;
 
     # Summary
     - name: UT Test Results Summary

.github/actions/pt2e/action.yml

@@ -29,13 +29,15 @@ runs:
       shell: bash -xe {0}
       run: |
         # dataset
+        dataset_dir="${RUNNER_TEMP}/_datasets/imagenet"
         if [ ! -d ${dataset_dir} ];then
           rm -rf ${dataset_dir} && mkdir -p ${dataset_dir} && cd ${dataset_dir}
           wget -O valprep.sh https://raw.githubusercontent.com/soumith/imagenetloader.torch/master/valprep.sh
           wget -q https://image-net.org/data/ILSVRC/2012/ILSVRC2012_img_val.tar
           tar -xf ILSVRC2012_img_val.tar
           bash valprep.sh
         fi
+        echo "dataset_dir=${dataset_dir}" >> ${GITHUB_ENV}
     - name: PT2E Test (${{ inputs.dt }}  ${{ inputs.scenario }})
       shell: bash -xe {0}
       run: |

.github/ci_expected_accuracy/rolling/inductor_huggingface_inference.csv

@@ -45,3 +45,7 @@ TrOCRForCausalLM,pass,pass,pass,pass,pass
 XGLMForCausalLM,pass,pass,pass,pass,pass
 XLNetLMHeadModel,pass,pass,pass,pass,pass
 YituTechConvBert,pass,pass,pass,pass,pass
+Qwen/Qwen3-0.6B,pass,pass,pass,pass,pass
+google/gemma-2-2b,pass,pass,pass,pass,pass
+meta-llama/Llama-3.2-1B,pass,pass,pass,pass,pass
+openai/whisper-tiny,pass,pass,pass,pass,pass

.github/ci_expected_accuracy/rolling/inductor_huggingface_training.csv

@@ -45,3 +45,8 @@ TrOCRForCausalLM,pass,pass,pass,pass,pass
 XGLMForCausalLM,pass,pass,pass,pass,pass
 XLNetLMHeadModel,pass,pass,pass,pass,pass
 YituTechConvBert,pass,pass,pass,pass,pass
+# https://github.com/intel/torch-xpu-ops/issues/2055
+Qwen/Qwen3-0.6B,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run
+google/gemma-2-2b,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run
+meta-llama/Llama-3.2-1B,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run
+openai/whisper-tiny,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run

.github/scripts/ut_result_check.sh

@@ -198,7 +198,8 @@ if [[ "${ut_suite}" == 'op_regression' || "${ut_suite}" == 'op_regression_dev1'
 fi
 
 if [[ "${ut_suite}" == 'xpu_distributed' ]]; then
-    grep -E "^FAILED" xpu_distributed_test.log | awk '{print $2}' > ./"${ut_suite}"_xpu_distributed_test_failed.log
+    grep -E "^FAILED" xpu_distributed_test.log | awk '{print $3}' > ./"${ut_suite}"_xpu_distributed_test_failed.log
+    sed -i '/^[^.]\+/d' ./"${ut_suite}"_xpu_distributed_test_failed.log
     grep "PASSED" xpu_distributed_test.log | awk '{print $1}' > ./"${ut_suite}"_xpu_distributed_test_passed.log
     echo -e "========================================================================="
     echo -e "Show Failed cases in ${ut_suite} xpu distributed"

.github/workflows/_linux_accelerate.yml

@@ -48,15 +48,16 @@ concurrency:
 defaults:
   run:
     shell: bash {0}
+env:
+  GH_TOKEN: ${{ github.token }}
+  DOCKER_REGISTRY_AUTH_TOKEN: ${{ secrets.DOCKER_HUB_TOKEN }}
 
 jobs:
   conditions-filter:
     name: conditions-filter
     if: ${{ github.event.pull_request.draft == false }}
     runs-on: ubuntu-24.04
     timeout-minutes: 10
-    env:
-      GH_TOKEN: ${{ github.token }}
     outputs:
       disabled_tests: ${{ steps.check-pr-desc.outputs.disabled_tests }}
     steps:

.github/workflows/_linux_build.yml

@@ -31,6 +31,8 @@ permissions: read-all
 defaults:
   run:
     shell: bash -xe {0}
+env:
+  DOCKER_REGISTRY_AUTH_TOKEN: ${{ secrets.DOCKER_HUB_TOKEN }}
 
 jobs:
   runner:
@@ -94,6 +96,7 @@ jobs:
       - name: Build Pytorch on ${{ needs.runner.outputs.hostname }}
         run: |
           export USE_XCCL=1
+          export IS_XPU_CI=1
           # only build pvc for CI
           if [ "${{ github.event_name }}" == "pull_request" ];then
             export TORCH_XPU_ARCH_LIST='pvc'
@@ -200,7 +203,6 @@ jobs:
           python -c "import torchaudio; print(torchaudio.__version__)"
           python pytorch/torch/utils/collect_env.py
           pip list |grep -E 'torch|intel'
-          chmod 777 /__w /github ./ -R
       - name: Upload Torch XPU Wheel
         if: ${{ success() }}
         uses: actions/upload-artifact@v4
@@ -213,3 +215,8 @@ jobs:
         with:
           name: Torch-XPU-Build-Log-${{ github.event.pull_request.number || github.sha }}
           path: ${{ github.workspace }}/build_*.log
+      - name: Cleanup workspace
+        if: ${{ always() }}
+        run: |
+          chmod 777 /__w /github ./ -R
+          find ./ |grep -v "^\./$" |xargs rm -rf

.github/workflows/_linux_e2e.yml

@@ -50,6 +50,7 @@ env:
   GH_TOKEN: ${{ github.token }}
   HF_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
   HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+  DOCKER_REGISTRY_AUTH_TOKEN: ${{ secrets.DOCKER_HUB_TOKEN }}
 
 jobs:
   runner:
@@ -87,7 +88,6 @@ jobs:
         cpus_per_xpu: ${{ needs.runner.outputs.cpus_per_xpu }}
         MODEL_ONLY_NAME: ${{ inputs.model }}
         AGENT_TOOLSDIRECTORY: /tmp/xpu-tool
-        dataset_dir: ${{ runner.temp }}/../_datasets/imagenet
     steps:
       - name: Checkout torch-xpu-ops
         uses: actions/checkout@v4

.github/workflows/_linux_e2e_summary.yml

@@ -30,14 +30,15 @@ jobs:
     steps:
       - name: Checkout torch-xpu-ops
         uses: actions/checkout@v4
-      - name: Install gh-cli
-        run: |
-          sudo apt-get update
-          sudo apt-get install gh rsync ca-certificates -y
       - name: Setup python-${{ inputs.python }}
         uses: actions/setup-python@v5
         with:
           python-version: ${{ inputs.python }}
+      - name: Install gh-cli
+        run: |
+          sudo apt-get update
+          sudo apt-get install gh rsync ca-certificates -y
+          pip install pandas requests
       - name: Download Target Artifact
         run: |
           mkdir target/
@@ -64,30 +65,30 @@ jobs:
       - name: Get summary
         if: ${{ ! cancelled() }}
         run: |
-          pip install pandas requests
+          exit_label=0
           e2e_summary_csv="$(find ./target/ -name "inductor_*.csv" |head -n 1)"
           if [ -f "${e2e_summary_csv}" ];then
             bash ./.github/scripts/e2e_summary.sh ./target ./baseline >> ${GITHUB_STEP_SUMMARY}
             exit_label=$(awk 'BEGIN{sum=0}{if($2>0){sum++}}END{print sum}' /tmp/tmp-result.txt)
             if [ ${exit_label} -ne 0 ];then
               grep -E "(Real failed|to passed|Warning timeout).*: [1-9]|Summary for" /tmp/tmp-*.txt |grep -E "failed|passed|timeout" -B 1
               echo "There are ${exit_label} cases that need look into!!! Please check them"
-              exit ${exit_label}
             fi
           fi
           pt2e_summary_csv="$(find ./target/ -name "summary.csv")"
           if [ -f "${pt2e_summary_csv}" ];then
-            cat ${pt2e_summary_csv}
-            failed_num=$(grep -c ',failed' ${pt2e_summary_csv})
+            failed_num=$(grep -c ',failed' ${pt2e_summary_csv} || true)
             if [ ${failed_num} -ne 0 ];then
               echo "[Warning] PT2E has failures!"
+              grep 'failed' ${pt2e_summary_csv}
             fi
           fi
+          exit ${exit_label}
       - name: Upload Reference Run ID
         if: ${{ endsWith(inputs.test_type, 'ly') }}
         run: |
           gh --repo ${GITHUB_REPOSITORY} issue view ${REFERENCE_ISSUE_ID} --json body -q .body 2>&1 |tee new_body.txt 2>&1
-          has_or_not="$(grep -c 'Inductor-${{ inputs.test_type }}-LTS2' new_body.txt)"
+          has_or_not="$(grep -c 'Inductor-${{ inputs.test_type }}-LTS2' new_body.txt || true)"
           if [ ${has_or_not} -ne 0 ];then
             sed -i "s/Inductor-${{ inputs.test_type }}-LTS2:.*/Inductor-${{ inputs.test_type }}-LTS2: ${GITHUB_RUN_ID}/" new_body.txt
           else

.github/workflows/_linux_op_benchmark.yml

@@ -26,6 +26,7 @@ env:
   GH_TOKEN: ${{ github.token }}
   HF_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
   HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+  DOCKER_REGISTRY_AUTH_TOKEN: ${{ secrets.DOCKER_HUB_TOKEN }}
   reference_issue: 1689
 
 jobs:
@@ -50,8 +51,6 @@ jobs:
   op_benchmark:
     needs: runner
     runs-on: ${{ needs.runner.outputs.runner_id }}
-    permissions: 
-      issues: write
     timeout-minutes: 900
     container:
       image: mengfeili/intel-pvc-driver:1146-1136
@@ -93,6 +92,8 @@ jobs:
   op_benchmark_test_results_check:
     needs: op_benchmark
     runs-on: ubuntu-24.04
+    permissions:
+      issues: write
     steps:
       - name: Install gh-cli
         run: |

.github/workflows/_linux_transformers.yml

@@ -56,6 +56,7 @@ concurrency:
   cancel-in-progress: true
 env:
   HF_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+  DOCKER_REGISTRY_AUTH_TOKEN: ${{ secrets.DOCKER_HUB_TOKEN }}
   HF_HUB_ETAG_TIMEOUT: 120
   HF_HUB_DOWNLOAD_TIMEOUT: 120
   python: ${{ inputs.python != '' && inputs.python || '3.10' }}

.github/workflows/_linux_ut.yml

@@ -33,6 +33,7 @@ env:
   GH_TOKEN: ${{ github.token }}
   HF_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
   HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+  DOCKER_REGISTRY_AUTH_TOKEN: ${{ secrets.DOCKER_HUB_TOKEN }}
   UT_SKIP_ISSUE: 1624
 
 jobs:
@@ -99,11 +100,12 @@ jobs:
 
   test-in-baremetal:
     needs: runner
+    timeout-minutes: 600
     if: ${{ contains(inputs.ut, 'distributed') }}
     runs-on: ${{ needs.runner.outputs.runner_id }}
     env:
       AGENT_TOOLSDIRECTORY: /tmp/xpu-tool
-      PYTEST_ADDOPTS: -v --timeout 600 --timeout_method=thread -n 1
+      PYTEST_ADDOPTS: -v --timeout 3600 --timeout_method=thread -n 1
     steps:
       - name: Checkout torch-xpu-ops
         uses: actions/checkout@v4