test5

Author: sandeepgupta12

.ci/docker/manywheel/Dockerfile_ppc64le

@@ -1,4 +1,4 @@
-# Use UBI 9.3 as base image
+# Use UBI 9 as base image
 FROM registry.access.redhat.com/ubi9/ubi:9.5 AS base
 
 # Install necessary dependencies
@@ -27,6 +27,6 @@ RUN pip install -r requirements.txt
 RUN mkdir -p /workspace/pytorch
 
 ENTRYPOINT []
-CMD ["sleep", "infinity"]
+CMD ["/bin/bash"]
 
 

.github/scripts/ppc64le-build.sh

@@ -8,18 +8,12 @@ cd /workspace/$PACKAGE_NAME
 # Clean up old artifacts
 rm -rf build/ dist/ torch.egg-info/
 
-# Set CPU-specific compilation flags
-export CMAKE_CXX_FLAGS="-mcpu=power9 -mtune=power9"
-export CMAKE_C_FLAGS="-mcpu=power9 -mtune=power9"
-export TORCH_CMAKE_ARGS="-DCMAKE_CXX_FLAGS='-mcpu=power9 -mtune=power9' -DCMAKE_C_FLAGS='-mcpu=power9 -mtune=power9'"
-
-
 # Build and install PyTorch wheel
 MAX_JOBS=$(nproc)
 python setup.py bdist_wheel 2>&1 | tee build_log.txt
 
 if [ $? -ne 0 ]; then
-    echo "❌ Build failed! Check build_log.txt for details."
+    echo "Build failed! Check build_log.txt for details."
     exit 1
 fi
 
@@ -42,6 +36,6 @@ if ! pytest "$PACKAGE_NAME/test/test_utils.py"; then
 
 else
     echo "------------------$PACKAGE_NAME:install_and_test_both_success-------------------------"
-    echo "✅ All tests passed successfully!"
+    echo "All tests passed successfully!"
     exit 0
 fi

.github/scripts/ppc64le-ci/README.md

@@ -3,16 +3,31 @@
 ## Install prerequisites.
 
 ```
-Install Docker 
+$ sudo dnf install podman podman-docker jq
 ```
-## Clone pytorch repository
-
 ## Add services.
 
 ```
 $ sudo cp self-hosted-builder/*.service /etc/systemd/system/
 $ sudo systemctl daemon-reload
 ```
+
+## Rebuild the image
+
+First build s390x builder image `docker.io/pytorch/ubippc64le-builder`,
+using following commands:
+
+```
+$ cd ~
+$ git clone https://github.com/pytorch/pytorch
+$ cd pytorch
+$ git submodule update --init --recursive
+$ GPU_ARCH_TYPE=cpu-ppc64le "$(pwd)/.ci/docker/manywheel/build.sh" ubippc64le-builder
+$ docker image tag localhost/pytorch/ubippc64le-builder docker.io/pytorch/ubippc64le-builder:cpu-ppc64le
+$ docker image save -o ~/ubi-ppc64le.tar docker.io/pytorch/ubippc64le-builder:cpu-ppc64le
+```
+
+
 Next step is to build `actions-runner` image using:
 
 ```

.github/scripts/ppc64le-ci/self-hosted-builder/actions-runner.Dockerfile

@@ -96,7 +96,7 @@ USER runner
 # Set working directory
 WORKDIR /opt/runner
 
-COPY --chown=runner:runner ubi-ppc64le.tar /opt/runner/ubi-ppc64le.tar
+COPY --chown=runner:runner pytorch-ubi-ppc64le.tar /opt/runner/pytorch-ubi-ppc64le.tar
 
 # Define entry point and command
 ENTRYPOINT ["/usr/bin/entrypoint"]

.github/scripts/ppc64le-ci/self-hosted-builder/fs/usr/bin/actions-runner

@@ -2,52 +2,21 @@
 
 set -e -u
 
-trap cleanup EXIT
-
 # first import docker image
-if [ -f ./ubi-ppc64le.tar ] ; then
-        docker image load --input ubi-ppc64le.tar
+if [ -f ./pytorch-ubi-ppc64le.tar ] ; then
+        docker image load --input pytorch-ubi-ppc64le.tar
         docker image tag docker.io/pytorch/ubippc64le-builder:cpu-ppc64le docker.io/pytorch/ubippc64le-builder:cpu-ppc64le-main
         rm -f ubi-ppc64le.tar
 fi
 
 token_file=registration-token.json
 
-# Function to clean up and unregister the runner
-cleanup() {
-    echo "Cleaning up temporary files..."
-    [ -f "$token_file" ] && rm -f "$token_file"
-    [ -f "runner-id.json" ] && rm -f "runner-id.json"
-
-    echo "Unregistering the runner from GitHub..."
-    ACCESS_TOKEN="$(cat /run/runner_secret)"
-    runner_id=$(curl -s \
-        -H "Accept: application/vnd.github.v3+json" \
-        -H "Authorization: Bearer ${ACCESS_TOKEN}" \
-        "https://api.github.com/repos/${OWNER}/${REPO}/actions/runners" | \
-        jq --raw-output '.runners[] | select(.name=="'"${NAME}"'") | .id')
-
-    if [ -n "$runner_id" ]; then
-        curl -s \
-            -X DELETE \
-            -H "Accept: application/vnd.github.v3+json" \
-            -H "Authorization: Bearer ${ACCESS_TOKEN}" \
-            "https://api.github.com/repos/${OWNER}/${REPO}/actions/runners/$runner_id"
-        echo "Runner unregistered successfully."
-    else
-        echo "Warning: Runner ID for ${NAME} not found. It may already be removed."
-    fi
-
-    unset ACCESS_TOKEN runner_id
-}
-
 # Fetch GitHub access token
 if [ ! -f /run/runner_secret ]; then
     echo "Error: Access token file not found at /run/runner_secret."
     exit 1
 fi
 
-
 ACCESS_TOKEN="$(cat /run/runner_secret)"
 
 # Generate registration token
@@ -60,6 +29,8 @@ curl \
 
 unset ACCESS_TOKEN
 
+rm -f /run/runner_secret
+
 # register runner as ephemeral runner
 # it does one job, stops and unregisters
 registration_token=$(jq --raw-output .token "$token_file")

.github/workflows/_linux-build.yml

@@ -363,5 +363,5 @@ jobs:
         shell: bash
         run: |
          
-          docker ps -q | xargs -r docker stop
-          docker ps -aq | xargs -r docker rm
+          docker stop -a || true
+          docker kill -a || true

.github/workflows/ppc64le.yml

@@ -12,19 +12,21 @@ concurrency:
 
 jobs:
 
-  ppc64le-UBI-9-3-Build-and-Test:
-    name: ppc64le-UBI-9-3-Build-and-Test
+  ppc64le-UBI-9-Build-and-Test:
+    name: ppc64le-UBI-9-Build-and-Test
     runs-on: linux.ppc64le
+    env:
+      IMAGE_NAME: docker.io/pytorch/ubippc64le-builder:cpu-ppc64le-main  # Define the image variable
+
     steps:
       - name: Checkout PyTorch
         uses: pytorch/pytorch/.github/actions/checkout-pytorch@main
-        
         with:
           no-sudo: true
 
       - name: Execute Build and Tests inside ppc64le Docker Container
         run: |
-          docker run --rm -v ${GITHUB_WORKSPACE}:/workspace/pytorch docker.io/pytorch/ubippc64le-builder:cpu-ppc64le-main /workspace/pytorch/.github/scripts/ppc64le-build.sh
+          docker run --rm -v ${GITHUB_WORKSPACE}:/workspace/pytorch $IMAGE_NAME /workspace/pytorch/.github/scripts/ppc64le-build.sh
       
       - name: Archive ppc64le artifacts into zip
         run: |
@@ -43,9 +45,9 @@ jobs:
         if: always()
         shell: bash
         run: |
-          # on s390x stop the container for clean worker stop
-          docker stop -a || true
-          docker kill -a || true
+          # Stop and remove all containers
+          docker ps -q | xargs -r docker stop
+          docker ps -aq | xargs -r docker rm