Modify build workflow (#1359)

intel · Feb 24, 2025 · d4b8901 · d4b8901
1 parent c9dc598
commit d4b8901
Show file tree

Hide file tree

Showing 8 changed files with 24 additions and 14 deletions.
diff --git a/.github/actions/inductor-xpu-e2e-test/action.yml b/.github/actions/inductor-xpu-e2e-test/action.yml
@@ -132,7 +132,10 @@ runs:
                     bash inductor_xpu_test.sh ${suite} ${dt} ${mode} ${scenario} xpu ${xpu_id/:*} static ${#xpu_list[*]} ${xpu_id/*:} &
                   done
                 else
-                  bash inductor_xpu_test.sh ${suite} ${dt} ${mode} ${scenario} xpu 0 static 1 0 ${MODEL_ONLY_NAME} &
+                  for test_model in $(echo ${MODEL_ONLY_NAME} |sed 's/,/ /g')
+                  do
+                    bash inductor_xpu_test.sh ${suite} ${dt} ${mode} ${scenario} xpu 0 static 1 0 ${test_model}
+                  done
                 fi
                 wait
                 # summarize pass rate

diff --git a/.github/scripts/e2e_summary.sh b/.github/scripts/e2e_summary.sh
@@ -30,7 +30,7 @@ function get_model_result() {
             do
                 for mode in training inference
                 do
-                    colorful=$(grep "${model}" "/tmp/tmp-${suite}-${mode}-${dtype}.txt" 2>&1 |awk 'BEGIN{
+                    colorful=$(grep -w "${model}" "/tmp/tmp-${suite}-${mode}-${dtype}.txt" 2>&1 |awk 'BEGIN{
                         color = "black";
                         exit_label = 0;
                     }{

diff --git a/.github/workflows/_linux_build.yml b/.github/workflows/_linux_build.yml
@@ -113,7 +113,7 @@ jobs:
             WERROR=1 python setup.py bdist_wheel 2>&1 | tee pytorch_${current_commit}_build.log
 
             if [[ ${is_fork_pr} == "false" ]]; then
-              if [ -f dist/torch*.whl -a "${last_commit}" != "${current_commit}" ]; then
+              if [ -f dist/torch*.whl && "${last_commit}" != "${current_commit}"] && [[ "${{ inputs.pytorch }}" == "main" || "${{ inputs.pytorch }}" == "release/"* ]]; then
                 echo "Wheel build successful, update last commit in the issue https://github.com/intel/torch-xpu-ops/issues/1280"
                 gh --repo $repo issue view $commit_issue --json body -q .body | sed "s;${last_commit};${current_commit};g" | sed '/^$/d' > new_body.txt
                 gh --repo $repo issue edit $commit_issue --body-file new_body.txt
@@ -170,21 +170,21 @@ jobs:
           cd ../pytorch
           echo "TORCH_BRANCH_ID=$(git rev-parse --abbrev-ref HEAD)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "TORCH_COMMIT_ID=$(git rev-parse HEAD)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
-          echo "DRIVER_VERSION=$(dkms status 2>&1 |grep 'intel-i915-dkms' |sed 's/.*\///;s/,.*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
+          echo "DRIVER_VERSION=$(sycl-ls |grep 'opencl:gpu' |awk '{print $NF}' |sort |uniq -c |sed 's/ //g;s/\[/*[/')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "KERNEL_VERSION=$(uname -rv 2>&1)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
-          echo "BUNDLE_VERSION=$(dpcpp --version 2>&1 |grep 'DPC++/C++' |sed 's/.*(//;s/).*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
+          echo "BUNDLE_VERSION=$(icpx --version 2>&1 |grep 'DPC++/C++' |sed 's/.*(//;s/).*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           . /etc/os-release
           echo "OS_PRETTY_NAME=${PRETTY_NAME}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "GCC_VERSION=$(gcc -dumpversion)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo ${GITHUB_ENV}
       - name: Upload Torch XPU Wheel
-        if: always()
+        if: ${{ ! cancelled() }}
         uses: actions/upload-artifact@v4
         with:
           name: Torch-XPU-Wheel-${{ github.event.pull_request.number || github.sha }}-${{ inputs.abi }} 
           path: ${{ github.workspace }}/torch*.whl
       - name: Upload Build Log
-        if: always()
+        if: ${{ ! cancelled() }}
         uses: actions/upload-artifact@v4
         with:
           name: Torch-XPU-Build-Log-${{ github.event.pull_request.number || github.sha }}-${{ inputs.abi }}

diff --git a/.github/workflows/_linux_ut.yml b/.github/workflows/_linux_ut.yml
@@ -280,7 +280,7 @@ jobs:
             bash ut_result_check.sh ${ut_suite}
           done
       - name: Upload Inductor XPU UT Log
-        if: always()
+        if: ${{ ! cancelled() }}
         uses: actions/upload-artifact@v4
         with:
           name: Inductor-XPU-UT-Data-${{ github.event.pull_request.number || github.sha }}-${{ inputs.abi }}-${{ env.UT_NAME }}

diff --git a/.github/workflows/nightly_ondemand.yml b/.github/workflows/nightly_ondemand.yml
@@ -68,7 +68,7 @@ concurrency:
 jobs:
   Linux-Nightly-Ondemand-Build:
     secrets: inherit
-    if: always()
+    if: ${{ ! cancelled() }}
     name: linux-nightly-ondemand
     permissions:
       issues: write
@@ -168,9 +168,9 @@ jobs:
           echo "TRANSFORMERS_VERSION=$(<.ci/docker/ci_commit_pins/huggingface.txt)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "TIMM_COMMIT_ID=$(<.ci/docker/ci_commit_pins/timm.txt)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "MODEL_ONLY_NAME=${{ inputs.model }}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
-          echo "DRIVER_VERSION=$(dkms status 2>&1 |grep 'intel-i915-dkms' |sed 's/.*\///;s/,.*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
+          echo "DRIVER_VERSION=$(sycl-ls |grep 'opencl:gpu' |awk '{print $NF}' |sort |uniq -c |sed 's/ //g;s/\[/*[/')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "KERNEL_VERSION=$(uname -rv 2>&1)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
-          echo "BUNDLE_VERSION=$(dpcpp --version 2>&1 |grep 'DPC++/C++' |sed 's/.*(//;s/).*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
+          echo "BUNDLE_VERSION=$(icpx --version 2>&1 |grep 'DPC++/C++' |sed 's/.*(//;s/).*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           . /etc/os-release
           echo "OS_PRETTY_NAME=${PRETTY_NAME}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "GCC_VERSION=$(gcc -dumpversion)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
@@ -287,6 +287,7 @@ jobs:
           tar zcf xpu-inductor-${GITHUB_RUN_ID}.tgz -C ${{ github.workspace }}/upload_files/ . # backup logs
           # Print summary
           rm -rf /tmp/tmp-*.txt
+          source activate e2e_ci
           bash ${{ github.workspace }}/.github/scripts/e2e_summary.sh ${{ github.workspace }}/upload_files >> ${GITHUB_STEP_SUMMARY}
           exit_label=$(awk 'BEGIN{sum=0}{if($2>0){sum++}}END{print sum}' /tmp/tmp-result.txt)
           if [ ${exit_label} -ne 0 ];then

diff --git a/.github/workflows/nightly_ondemand_rolling.yml b/.github/workflows/nightly_ondemand_rolling.yml
@@ -173,9 +173,9 @@ jobs:
           echo "TRANSFORMERS_VERSION=$(<.ci/docker/ci_commit_pins/huggingface.txt)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "TIMM_COMMIT_ID=$(<.ci/docker/ci_commit_pins/timm.txt)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "MODEL_ONLY_NAME=${{ inputs.model }}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
-          echo "DRIVER_VERSION=$(dkms status 2>&1 |grep 'intel-i915-dkms' |sed 's/.*\///;s/,.*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
+          echo "DRIVER_VERSION=$(sycl-ls |grep 'opencl:gpu' |awk '{print $NF}' |sort |uniq -c |sed 's/ //g;s/\[/*[/')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "KERNEL_VERSION=$(uname -rv 2>&1)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
-          echo "BUNDLE_VERSION=$(dpcpp --version 2>&1 |grep 'DPC++/C++' |sed 's/.*(//;s/).*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
+          echo "BUNDLE_VERSION=$(icpx --version 2>&1 |grep 'DPC++/C++' |sed 's/.*(//;s/).*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           . /etc/os-release
           echo "OS_PRETTY_NAME=${PRETTY_NAME}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "GCC_VERSION=$(gcc -dumpversion)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
@@ -205,6 +205,7 @@ jobs:
           echo "$GITHUB_ENV"
           rm -rf ../pytorch/inductor_log
           rm -rf /tmp/torchinductor_*
+          rm -rf ~/.triton/cache
 
       # Nihglty launch
       - name: Nightly Huggingface FP32/BF16/FP16 Inference & Training Accuracy Test
@@ -298,6 +299,7 @@ jobs:
           tar zcf xpu-inductor-${GITHUB_RUN_ID}.tgz -C ${{ github.workspace }}/upload_files/ . # backup logs
           # Print summary
           rm -rf /tmp/tmp-*.txt
+          source activate e2e_ci
           bash ${{ github.workspace }}/.github/scripts/e2e_summary.sh ${{ github.workspace }}/upload_files >> ${GITHUB_STEP_SUMMARY}
           exit_label=$(awk 'BEGIN{sum=0}{if($2>0){sum++}}END{print sum}' /tmp/tmp-result.txt)
           if [ ${exit_label} -ne 0 ];then

diff --git a/.github/workflows/nightly_ondemand_whl.yml b/.github/workflows/nightly_ondemand_whl.yml
@@ -134,11 +134,12 @@ jobs:
           echo "TORCHAUDIO_COMMIT_ID=$(python -c 'import torchaudio; print(torchaudio.version.git_version)')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "TRITON_COMMIT_ID=$(python -c 'import triton; print(triton.__version__)')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           cd ../pytorch
+          source .github/scripts/env.sh
           echo "TORCHBENCH_COMMIT_ID=$(<.github/ci_commit_pins/torchbench.txt)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "TRANSFORMERS_VERSION=$(<.ci/docker/ci_commit_pins/huggingface.txt)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "TIMM_COMMIT_ID=$(<.ci/docker/ci_commit_pins/timm.txt)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "MODEL_ONLY_NAME=${{ inputs.model }}" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
-          echo "DRIVER_VERSION=$(dkms status 2>&1 |grep 'intel-i915-dkms' |sed 's/.*\///;s/,.*//')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
+          echo "DRIVER_VERSION=$(sycl-ls |grep 'opencl:gpu' |awk '{print $NF}' |sort |uniq -c |sed 's/ //g;s/\[/*[/')" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "KERNEL_VERSION=$(uname -rv 2>&1)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           echo "BUNDLE_VERSION=$(pip list |grep cmplr |head -n 1)" |tee -a "${GITHUB_OUTPUT}" >> "${GITHUB_ENV}"
           . /etc/os-release
@@ -150,6 +151,7 @@ jobs:
           echo "$GITHUB_ENV"
           rm -rf ../pytorch/inductor_log
           rm -rf /tmp/torchinductor_*
+          rm -rf ~/.triton/cache
 
       # Nihglty launch
       - name: Nightly Huggingface FP32/BF16/FP16 Inference & Training Accuracy Test
@@ -243,6 +245,7 @@ jobs:
           tar zcf xpu-inductor-${GITHUB_RUN_ID}.tgz -C ${{ github.workspace }}/upload_files/ . # backup logs
           # Print summary
           rm -rf /tmp/tmp-*.txt
+          source activate e2e_ci
           bash ${{ github.workspace }}/.github/scripts/e2e_summary.sh ${{ github.workspace }}/upload_files >> ${GITHUB_STEP_SUMMARY}
           exit_label=$(awk 'BEGIN{sum=0}{if($2>0){sum++}}END{print sum}' /tmp/tmp-result.txt)
           if [ ${exit_label} -ne 0 ];then

diff --git a/.github/workflows/pull.yml b/.github/workflows/pull.yml
@@ -195,6 +195,7 @@ jobs:
           cp -r ${{ github.workspace }}/../pytorch/inductor_log ${{ github.workspace }}/upload_files
           # Print summary
           rm -rf /tmp/tmp-*.txt
+          source activate e2e_ci
           bash ${{ github.workspace }}/.github/scripts/e2e_summary.sh ${{ github.workspace }}/upload_files >> ${GITHUB_STEP_SUMMARY}
           exit_label=$(awk 'BEGIN{sum=0}{if($2>0){sum++}}END{print sum}' /tmp/tmp-result.txt)
           if [ ${exit_label} -ne 0 ];then