Merge pull request #9218 from eclipse/ag_build_fixes

Ag build fixes

.github/actions/install-cmake-linux/action.yml

@@ -0,0 +1,11 @@
+name: Install protobuf linux
+runs:
+  using: composite
+  steps:
+    - name: Install protobuf linux
+      shell: bash
+      run: |
+        apt-get -yq  update && apt-get install -y build-essential unzip libssl-dev
+        curl -fsSL http://cmake.org/files/v3.19/cmake-3.19.0.tar.gz | tar xz && cd cmake-3.19.0
+                          ./configure --prefix=/opt/cmake && make -j2 && make install && cd .. && rm -r cmake-3.19.0
+        echo "/opt/cmake/bin" >> $GITHUB_PATH

.github/actions/update-deps-linux/action.yml

@@ -0,0 +1,63 @@
+name: Install protobuf linux
+runs:
+  using: composite
+  steps:
+    - name: Install protobuf linux
+      shell: bash
+      run: |
+        sudo apt-get install build-essential make zlib1g-dev wget
+        wget https://www.gnupg.org/ftp/gcrypt/libgpg-error/libgpg-error-1.36.tar.bz2
+        tar jxf libgpg-error-1.36.tar.bz2
+        cd libgpg-error-1.36
+        ./configure
+        make
+        sudo make install
+        cd ..
+        wget https://www.gnupg.org/ftp/gcrypt/libgcrypt/libgcrypt-1.8.4.tar.bz2
+        tar jxf libgcrypt-1.8.4.tar.bz2
+        cd libgcrypt-1.8.4
+        ./configure
+        make
+        sudo make install
+        cd ..
+        wget https://www.gnupg.org/ftp/gcrypt/libksba/libksba-1.3.5.tar.bz2
+        tar jxf libksba-1.3.5.tar.bz2
+        cd libksba-1.3.5
+        ./configure
+        make
+        sudo make install
+        cd ..
+        wget https://www.gnupg.org/ftp/gcrypt/libassuan/libassuan-2.5.3.tar.bz2
+        tar jxf libassuan-2.5.3.tar.bz2
+        cd libassuan-2.5.3
+        ./configure
+        make
+        sudo make install
+        cd ..
+        wget https://www.gnupg.org/ftp/gcrypt/ntbtls/ntbtls-0.1.2.tar.bz2
+        tar jxf ntbtls-0.1.2.tar.bz2
+        cd ntbtls-0.1.2
+        ./configure
+        make
+        sudo make install
+        cd ..
+        wget https://www.gnupg.org/ftp/gcrypt/npth/npth-1.6.tar.bz2
+        tar jxf npth-1.6.tar.bz2
+        cd npth-1.6
+        ./configure
+        make
+        sudo make install
+        cd ..
+        echo 'include /usr/local/lib/' | sudo tee -a /etc/ld.so.conf
+        sudo ldconfig -v
+        wget https://www.gnupg.org/ftp/gcrypt/gnupg/gnupg-2.2.17.tar.bz2
+        tar jxf gnupg-2.2.17.tar.bz2
+        cd gnupg-2.2.17/
+        ./configure
+        make
+        sudo make install
+        cd ..
+        sudo apt-get install pinentry-curses
+        #echo 'pinentry-program /usr/bin/pinentry-curses' | tee -a ~/.gnupg/gpg-agent.conf
+        #gpg-connect-agent reloadagent /bye
+        sudo apt-get install ca-certificates

.github/workflows/build-android-x86_64.yml

@@ -3,8 +3,20 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   android-x86_64:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - uses: AutoModality/action-clean@v1
       - name: Cancel Previous Runs
@@ -16,7 +28,9 @@ jobs:
         with:
           ndk-version: r18b
       - uses: actions/checkout@v2
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:
@@ -29,11 +43,13 @@ jobs:
       - name: Build on linux-x86_64
         env:
           ANDROID_NDK: ${{ steps.setup-ndk.outputs.ndk-path }}
-          LIBND4J_HOME: "${GITHUB_WORKSPACE}/libnd4j"
+          LIBND4J_HOME: ${GITHUB_WORKSPACE}/libnd4j
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           MAVEN_USERNAME: ${{ secrets.SONATYPE_USER_1 }}
           MAVEN_PASSWORD: ${{ secrets.SONATYPE_USER1_PASS }}
           MAVEN_GPG_PASSPHRASE: ${{ secrets.PACKAGES_GPG_PASS }}
+          OPENBLAS_PATH: ${HOME}/.javacpp/cache/openblas-0.3.13-1.5.5-android-x86_64.jar/lib/
+          DEPLOY_TO: ossrh
         run: |
           echo "Verifying programs on path. Path is $PATH"
           echo "Path post update is $PATH. Maven is at `which mvn` cmake is at `which cmake` protoc is at `which protoc`"
@@ -41,6 +57,8 @@ jobs:
           cmake --version
           protoc --version
           clang --version
+          export LIBND4J_HOME=${GITHUB_WORKSPACE}/libnd4j
+          export OPENBLAS_PATH=${HOME}/.javacpp/cache/openblas-0.3.13-1.5.5-android-x86_64.jar/lib/
           mvn  -X -Dorg.bytedeco.javacpp.logger.debug=true -Possrh -pl ":nd4j-native,:libnd4j" --also-make  \
           -Djavacpp.platform=android-x86_64  \
           -Dlibnd4j.platform=android-x86_64 -Dlibnd4j.chip=cpu \

.github/workflows/build-deploy-android-arm32.yml

@@ -3,9 +3,21 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   #Note: no -pl here because we publish everything from this branch and use this as the basis for all uploads.
   android-arm32:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - uses: AutoModality/action-clean@v1
       - name: Cancel Previous Runs
@@ -13,7 +25,9 @@ jobs:
         with:
           access_token: ${{ github.token }}
       - uses: actions/checkout@v2
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:

.github/workflows/build-deploy-android-arm64.yml

@@ -3,9 +3,21 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   #Note: no -pl here because we publish everything from this branch and use this as the basis for all uploads.
   android-arm64:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - uses: AutoModality/action-clean@v1
       - name: Cancel Previous Runs
@@ -13,7 +25,9 @@ jobs:
         with:
           access_token: ${{ github.token }}
       - uses: actions/checkout@v2
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:
@@ -36,6 +50,7 @@ jobs:
           MAVEN_USERNAME: ${{ secrets.SONATYPE_USER_1 }}
           MAVEN_PASSWORD: ${{ secrets.SONATYPE_USER1_PASS }}
           MAVEN_GPG_PASSPHRASE: ${{ secrets.PACKAGES_GPG_PASS }}
+          DEPLOY_TO: ossrh
         run: |
           mvn --version
           cmake --version

.github/workflows/build-deploy-linux-arm32.yml

@@ -3,9 +3,21 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   #Note: no -pl here because we publish everything from this branch and use this as the basis for all uploads.
   linux-arm32:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - uses: AutoModality/action-clean@v1
       - name: Cancel Previous Runs
@@ -13,7 +25,9 @@ jobs:
         with:
           access_token: ${{ github.token }}
       - uses: actions/checkout@v2
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:

.github/workflows/build-deploy-linux-arm64.yml

@@ -3,9 +3,21 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   #Note: no -pl here because we publish everything from this branch and use this as the basis for all uploads.
   linux-arm64:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - uses: AutoModality/action-clean@v1
       - name: Cancel Previous Runs
@@ -13,7 +25,9 @@ jobs:
         with:
           access_token: ${{ github.token }}
       - uses: actions/checkout@v2
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:

.github/workflows/build-deploy-linux-cuda-11.0.yml

@@ -3,9 +3,20 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
-
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-16.04
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   linux-x86_64-cuda_11-0:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - name: Cancel Previous Runs
         uses: styfle/cancel-workflow-action@0.8.0
@@ -23,7 +34,9 @@ jobs:
         env:
           cuda: 11.0.167
           GCC: 9
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:

.github/workflows/build-deploy-linux-cuda-11.2.yml

@@ -3,8 +3,20 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-16.04
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   linux-x86_64-cuda-11-2:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - name: Cancel Previous Runs
         uses: styfle/cancel-workflow-action@0.8.0
@@ -22,7 +34,9 @@ jobs:
         env:
           cuda: 11.2.1_461
           GCC: 9
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:

.github/workflows/build-deploy-linux-x86_64.yml

@@ -4,16 +4,30 @@ on:
   workflow_dispatch:
   push:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   #Note: no -pl here because we publish everything from this branch and use this as the basis for all uploads.
   linux-x86_64:
-    runs-on: ubuntu-18.04
+    needs: pre-ci
+    runs-on: ubuntu-16.04
     steps:
       - name: Cancel Previous Runs
         uses: styfle/cancel-workflow-action@0.8.0
         with:
           access_token: ${{ github.token }}
       - uses: actions/checkout@v2
+      - uses: ./.github/actions/update-deps-linux
       - uses: ./.github/actions/install-protobuf-linux
+      - uses: ./.github/actions/install-cmake-linux
       - name: Set up Java for publishing to GitHub Packages
         uses: actions/setup-java@v1
         with:
@@ -40,4 +54,3 @@ jobs:
           sudo apt-get clean
           mvn  -X -Possrh  -Djavacpp.platform=linux-x86_64 -Dlibnd4j.chip=cpu -Pcpu --batch-mode deploy -DskipTests
 
-

.github/workflows/build-deploy-mac.yml

@@ -3,7 +3,19 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   mac-x86_64:
+    needs: pre-ci
     runs-on: macos-10.15
     steps:
       - name: Cancel Previous Runs

.github/workflows/build-deploy-windows-cuda-11.0.yml

@@ -3,7 +3,19 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   windows-x86_64-cuda-11-0:
+    needs: pre-ci
     runs-on: windows-2019
     steps:
       - name: Cancel Previous Runs

.github/workflows/build-deploy-windows-cuda-11.2.yml

@@ -3,7 +3,19 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   windows-x86_64-cuda-11-2:
+    needs: pre-ci
     runs-on: windows-2019
     steps:
       - name: Cancel Previous Runs

.github/workflows/build-deploy-windows.yml

@@ -3,8 +3,19 @@ on:
     - cron: "0 */12 * * *"
   workflow_dispatch:
 jobs:
-
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   windows-x86_64:
+    needs: pre-ci
     runs-on: windows-2019
     steps:
       - name: Cancel Previous Runs

.github/workflows/cpu-integration-tests.yaml

@@ -1,7 +1,19 @@
 on:
   push:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   linux-x86_64:
+    needs: pre-ci
     runs-on: ubuntu-18.04
     steps:
       - name: Cancel Previous Runs

.github/workflows/cpu-sanity-check-tests.yaml

@@ -1,7 +1,19 @@
 on:
   push:
 jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
   linux-x86_64:
+    needs: pre-ci
     runs-on: ubuntu-18.04
     steps:
       - name: Cancel Previous Runs

.github/workflows/run-gpu-tests-sanity-checks.yml

@@ -0,0 +1,40 @@
+on:
+  push:
+
+jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+  linux-x86_64-cuda-11-2:
+    needs: pre-ci
+    runs-on: [self-hosted]
+    steps:
+      - uses: AutoModality/action-clean@v1
+      - name: Cancel Previous Runs
+        uses: styfle/cancel-workflow-action@0.8.0
+        with:
+          access_token: ${{ github.token }}
+      - uses: ./.github/actions/download-dl4j-test-resources-linux
+      - uses: actions/checkout@v2
+      - name: Run cuda tests
+        shell: bash
+        env:
+          DEBIAN_FRONTEND: noninteractive
+        run: |
+          export PATH="/opt/protobuf/bin:/usr/local/cuda-11.2/bin:$PATH"
+          nvcc --version
+          mvn --version
+          cmake --version
+          protoc --version
+          bash ./change-cuda-versions.sh 11.2
+          export OMP_NUM_THREADS=1
+          mvn  -Ptestresources -pl ":deeplearning4j-modelimport,:deeplearning4j-core,:nd4j-cuda-11.2,:samediff-import,:libnd4j" -Dlibnd4j.compute="5.0 5.2 5.3 6.0 6.2 8.0"  -Ptest-nd4j-cuda  --also-make -Dlibnd4j.chip=cuda clean test
+

.github/workflows/test_multiple_arch.yaml

@@ -0,0 +1,92 @@
+on:
+  push:
+jobs:
+  # Wait for up to a minute for previous run to complete, abort if not done by then
+  pre-ci:
+    runs-on: ubuntu-latest
+    timeout-minutes: 1
+    steps:
+      - name: 'Block Concurrent Executions'
+        uses: softprops/turnstyle@v1
+        with:
+          poll-interval-seconds: 10
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+  build_and_test_cross_platform:
+    needs: pre-ci
+    # The host should always be linux
+    runs-on: ubuntu-16.04
+    name: Build on ${{ matrix.distro }} ${{ matrix.arch }}
+
+    # Run steps on a matrix of 2 arch/distro combinations
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - arch: aarch64
+            distro: ubuntu16.04
+    steps:
+      - name: Cancel Previous Runs
+        uses: styfle/cancel-workflow-action@0.8.0
+        with:
+          access_token: ${{ github.token }}
+      - uses: actions/checkout@v2
+      - uses: ./.github/actions/download-dl4j-test-resources-linux
+      - uses: ./.github/actions/install-protobuf-linux
+      - uses: uraimo/run-on-arch-action@v2.0.8
+        name: Build and test with ${{ matrix.arch }} ${{ matrix.distro }}
+        id: build
+        with:
+          arch: ${{ matrix.arch }}
+          distro: ${{ matrix.distro }}
+          # Pass some environment variables to the container
+          env: | # YAML, but pipe character is necessary
+            DEBIAN_FRONTEND: noninteractive
+            M2_HOME: /opt/maven
+
+          # The shell to run commands with in the container
+          shell: /bin/sh
+          run: |
+               echo "Running install on architecture ${{ matrix.arch }}"
+               apt-get -yq  update && apt-get install -y build-essential unzip libssl-dev curl openjdk-8-jdk-headless
+               echo "After install on architecture ${{ matrix.arch }}"
+               mkdir -p "/opt/maven"
+               curl -fsSL http://apache.osuosl.org/maven/maven-3/3.6.3/binaries/apache-maven-3.6.3-bin.tar.gz | tar -xzC "/opt/maven" --strip-components=1
+               apt-get -yq  update && apt-get install -y build-essential unzip libssl-dev
+               curl -fsSL http://cmake.org/files/v3.19/cmake-3.19.0.tar.gz | tar xz && cd cmake-3.19.0
+                                 ./configure --prefix=/opt/cmake && make -j2 && make install && cd .. && rm -r cmake-3.19.0
+               export PATH="/opt/maven/bin:$PATH"
+                   echo "RUNNING ARCH ${{ matrix.arch }}"
+                            case "${{ matrix.arch }}" in
+                              armv7)
+                                  export JAVA_HOME="/usr/lib/jvm/java-8-openjdk-armhf/jre
+                                  ;;
+                                aarch64)
+                                  export JAVA_HOME="/usr/lib/jvm/java-8-openjdk-arm64/jre"
+                                  ;;
+                              ppc64le)
+                                    export JAVA_HOME="/usr/lib/jvm/java-8-openjdk-ppc64el/jre
+                                  ;;
+                  esac
+                  mvn --version
+                  export PATH="/opt/cmake/bin:/opt/maven/bin:/opt/protobuf/bin:$PATH"
+                  cmake --version
+                  wget https://github.com/KonduitAI/dl4j-test-resources/archive/master.zip && unzip master.zip
+                  cd dl4j-test-resources-master
+                  mvn clean install -DskipTests
+                  echo "Extracted test resources"
+                  echo "Running build in ${pwd}"
+                  curl -fsSL https://github.com/google/protobuf/releases/download/v3.5.1/protobuf-cpp-3.5.1.tar.gz \
+                                       | tar xz && \
+                                       cd protobuf-3.5.1 && \
+                                       ./configure --prefix=/opt/protobuf && \
+                                       make -j2 && \
+                                       make install && \
+                                       cd .. && \
+                                       rm -rf protobuf-3.5.1
+                  echo "/opt/protobuf/bin" >> $GITHUB_PATH
+                  mvn  -X -Possrh -pl :nd4j-tests  -Djavacpp.platform=linux-${{matrix.arch}} -Dlibnd4j.chip=cpu -Pcpu --batch-mode clean test -Dtest=org.nd4j.smoketests.SmokeTest
+
+
+
+

libnd4j/buildnativeoperations.sh

@@ -430,9 +430,9 @@ fi
 
 if [ -z "$COMPUTE" ]; then
   if [ "$ARCH" == "x86-64" ]; then
-   COMPUTE="5.0 5.2 5.3 6.0 6.2 8.0"
+   COMPUTE="5.0 5.2 5.3 6.0 8.0"
   else
-      COMPUTE="5.0 5.2 5.3 6.0 6.2"
+      COMPUTE="5.0 5.2 5.3 6.0"
   fi
 fi
 

nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/ir/TensorNamespace.java

@@ -5411,14 +5411,12 @@ public final class TensorNamespace {
      * Serializations can either use one of the fields above, or use this
      * raw bytes field. The only exception is the string case, where one is
      * required to store the content in the repeated bytes string_data field.
-     *
      * When this raw_data field is used to store tensor value, elements MUST
      * be stored in as fixed-width, little-endian order.
      * Floating-point data types MUST be stored in IEEE 754 format.
      * Complex64 elements must be written as two consecutive FLOAT values, real component first.
      * Complex128 elements must be written as two consecutive DOUBLE values, real component first.
      * Boolean type MUST be written one byte per tensor element (00000001 for true, 00000000 for false).
-     *
      * Note: the advantage of specific field rather than the raw_data field is
      * that in some cases (e.g. int data), protobuf does a better packing via
      * variable length storage, and may lead to smaller binary footprint.
@@ -5657,7 +5655,6 @@ public final class TensorNamespace {
   /**
    * <pre>
    * Tensors
-   *
    * A serialized tensor value.
    * </pre>
    *
@@ -7013,14 +7010,12 @@ public final class TensorNamespace {
      * Serializations can either use one of the fields above, or use this
      * raw bytes field. The only exception is the string case, where one is
      * required to store the content in the repeated bytes string_data field.
-     *
      * When this raw_data field is used to store tensor value, elements MUST
      * be stored in as fixed-width, little-endian order.
      * Floating-point data types MUST be stored in IEEE 754 format.
      * Complex64 elements must be written as two consecutive FLOAT values, real component first.
      * Complex128 elements must be written as two consecutive DOUBLE values, real component first.
      * Boolean type MUST be written one byte per tensor element (00000001 for true, 00000000 for false).
-     *
      * Note: the advantage of specific field rather than the raw_data field is
      * that in some cases (e.g. int data), protobuf does a better packing via
      * variable length storage, and may lead to smaller binary footprint.
@@ -7771,7 +7766,6 @@ public final class TensorNamespace {
     /**
      * <pre>
      * Tensors
-     *
      * A serialized tensor value.
      * </pre>
      *
@@ -9086,14 +9080,12 @@ public final class TensorNamespace {
        * Serializations can either use one of the fields above, or use this
        * raw bytes field. The only exception is the string case, where one is
        * required to store the content in the repeated bytes string_data field.
-       *
        * When this raw_data field is used to store tensor value, elements MUST
        * be stored in as fixed-width, little-endian order.
        * Floating-point data types MUST be stored in IEEE 754 format.
        * Complex64 elements must be written as two consecutive FLOAT values, real component first.
        * Complex128 elements must be written as two consecutive DOUBLE values, real component first.
        * Boolean type MUST be written one byte per tensor element (00000001 for true, 00000000 for false).
-       *
        * Note: the advantage of specific field rather than the raw_data field is
        * that in some cases (e.g. int data), protobuf does a better packing via
        * variable length storage, and may lead to smaller binary footprint.
@@ -9110,14 +9102,12 @@ public final class TensorNamespace {
        * Serializations can either use one of the fields above, or use this
        * raw bytes field. The only exception is the string case, where one is
        * required to store the content in the repeated bytes string_data field.
-       *
        * When this raw_data field is used to store tensor value, elements MUST
        * be stored in as fixed-width, little-endian order.
        * Floating-point data types MUST be stored in IEEE 754 format.
        * Complex64 elements must be written as two consecutive FLOAT values, real component first.
        * Complex128 elements must be written as two consecutive DOUBLE values, real component first.
        * Boolean type MUST be written one byte per tensor element (00000001 for true, 00000000 for false).
-       *
        * Note: the advantage of specific field rather than the raw_data field is
        * that in some cases (e.g. int data), protobuf does a better packing via
        * variable length storage, and may lead to smaller binary footprint.
@@ -9140,14 +9130,12 @@ public final class TensorNamespace {
        * Serializations can either use one of the fields above, or use this
        * raw bytes field. The only exception is the string case, where one is
        * required to store the content in the repeated bytes string_data field.
-       *
        * When this raw_data field is used to store tensor value, elements MUST
        * be stored in as fixed-width, little-endian order.
        * Floating-point data types MUST be stored in IEEE 754 format.
        * Complex64 elements must be written as two consecutive FLOAT values, real component first.
        * Complex128 elements must be written as two consecutive DOUBLE values, real component first.
        * Boolean type MUST be written one byte per tensor element (00000001 for true, 00000000 for false).
-       *
        * Note: the advantage of specific field rather than the raw_data field is
        * that in some cases (e.g. int data), protobuf does a better packing via
        * variable length storage, and may lead to smaller binary footprint.

nd4j/nd4j-backends/nd4j-backend-impls/nd4j-cuda/pom.xml

@@ -287,10 +287,7 @@
             <plugin>
                 <artifactId>maven-jar-plugin</artifactId>
             </plugin>
-            <plugin>
+
-                <groupId>org.apache.maven.plugins</groupId>
-                <artifactId>maven-surefire-plugin</artifactId>
-            </plugin>
             <plugin>
                 <groupId>org.apache.maven.plugins</groupId>
                 <artifactId>maven-enforcer-plugin</artifactId>

nd4j/nd4j-backends/nd4j-tests/src/test/java/org/nd4j/smoketests/SmokeTest.java

@@ -0,0 +1,59 @@
+/*
+ *
+ *  *  ******************************************************************************
+ *  *  *
+ *  *  *
+ *  *  * This program and the accompanying materials are made available under the
+ *  *  * terms of the Apache License, Version 2.0 which is available at
+ *  *  * https://www.apache.org/licenses/LICENSE-2.0.
+ *  *  *
+ *  *  *  See the NOTICE file distributed with this work for additional
+ *  *  *  information regarding copyright ownership.
+ *  *  * Unless required by applicable law or agreed to in writing, software
+ *  *  * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+ *  *  * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
+ *  *  * License for the specific language governing permissions and limitations
+ *  *  * under the License.
+ *  *  *
+ *  *  * SPDX-License-Identifier: Apache-2.0
+ *  *  *****************************************************************************
+ *
+ *
+ */
+
+package org.nd4j.smoketests;
+
+import lombok.extern.slf4j.Slf4j;
+import org.junit.Test;
+import org.nd4j.linalg.api.buffer.DataType;
+import org.nd4j.linalg.api.ndarray.INDArray;
+import org.nd4j.linalg.factory.Nd4j;
+import org.nd4j.linalg.profiler.ProfilerConfig;
+
+@Slf4j
+public class SmokeTest {
+
+
+    @Test
+    public void testBasic() {
+        Nd4j.getEnvironment().setDebug(true);
+        Nd4j.getExecutioner().setProfilingConfig(ProfilerConfig.builder()
+                .checkForNAN(true)
+                .checkForINF(true)
+                .checkLocality(true)
+                .checkElapsedTime(true)
+                .checkWorkspaces(true)
+                .build());
+        INDArray arr = Nd4j.randn(2,2);
+        INDArray arr2 = Nd4j.randn(2,2);
+        for(DataType dataType : DataType.values()) {
+            log.info("Testing matrix multiply on data type {}",dataType);
+            INDArray casted = arr.castTo(dataType);
+            INDArray casted2 = arr2.castTo(dataType);
+            INDArray result = casted.mmul(casted2);
+            log.info("Result for data type {} was {}",dataType,result);
+
+        }
+    }
+
+}