[ROCm] add manylinux build test for ROCm CI

tools/ci_build/github/azure-pipelines/orttraining-pai-ci-pipeline.yml

@@ -11,6 +11,14 @@ pr:
     - 'onnxruntime/core/providers/js'
 name: 'orttraining_ci_$(Date:yyyyMMdd)_$(Rev:r)'
 
+resources:
+  repositories:
+  - repository: manylinux
+    type: Github
+    endpoint: Microsoft
+    name: pypa/manylinux
+    ref: 5eda9aded5462201e6310105728d33016e637ea7
+
 variables:
   - name: video
     value: 44
@@ -22,7 +30,101 @@ variables:
     value: Release
 
 jobs:
-- job: Linux_Build
+- job: Linux_Build_manylinux
+  variables:
+    skipComponentGovernanceDetection: true
+    CCACHE_DIR: $(Pipeline.Workspace)/ccache
+    TODAY: $[format('{0:dd}{0:MM}{0:yyyy}', pipeline.startTime)]
+  workspace:
+    clean: all
+  pool: onnxruntime-Ubuntu2004-AMD-CPU
+  timeoutInMinutes: 120
+
+  steps:
+  - task: mspremier.PostBuildCleanup.PostBuildCleanup-task.PostBuildCleanup@3
+    displayName: 'Clean Agent Directories'
+    condition: always()
+
+  - checkout: self
+    clean: true
+    submodules: recursive
+
+  - template: templates/get-docker-image-steps.yml
+    parameters:
+      Dockerfile: tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_rocm
+      Context: tools/ci_build/github/linux/docker
+      DockerBuildArgs: >-
+        --build-arg INSTALL_DEPS_EXTRA_ARGS=-tmur
+        --network=host --build-arg POLICY=manylinux_2_28 --build-arg PLATFORM=x86_64
+        --build-arg BUILD_UID=$(id -u)
+        --build-arg ROCM_VERSION=$(RocmVersion)
+        --build-arg DEVTOOLSET_ROOTPATH=/opt/rh/gcc-toolset-12/root
+        --build-arg PREPEND_PATH=/opt/rh/gcc-toolset-12/root/usr/bin:
+        --build-arg LD_LIBRARY_PATH_ARG=/opt/rh/gcc-toolset-12/root/usr/lib64:/opt/rh/gcc-toolset-12/root/usr/lib:/opt/rh/gcc-toolset-12/root/usr/lib64/dyninst:/opt/rh/gcc-toolset-12/root/usr/lib/dyninst:/usr/local/lib64:/usr/local/lib
+      Repository: onnxruntimetrainingrocm-cibuild-rocm$(RocmVersion)-manylinux-build
+
+  - task: Cache@2
+    inputs:
+      key: '"manylinux" | "$(TODAY)" | "$(Build.SourceBranch)" | "$(Build.SourceVersion)"'
+      path: $(CCACHE_DIR)
+      cacheHitVar: CACHE_RESTORED
+      restoreKeys: |
+        "manylinux" | "$(TODAY)" | "$(Build.SourceBranch)"
+        "manylinux" | "$(TODAY)" |
+    displayName: Cache Task
+
+  - script: mkdir -p $(CCACHE_DIR)
+    condition: ne(variables.CACHE_RESTORED, 'true')
+    displayName: Create Cache Dir
+
+  - task: CmdLine@2
+    inputs:
+      script: |-
+        export ROCM_HOME=/opt/rocm
+        docker run --rm \
+          --ipc=host \
+          --network=host \
+          --cap-add=SYS_PTRACE \
+          --security-opt seccomp=unconfined \
+          --shm-size=1024m \
+          --user $UID:$(id -g $USER) \
+          -e CC=/opt/rh/gcc-toolset-12/root/usr/bin/cc -e CXX=/opt/rh/gcc-toolset-12/root/usr/bin/c++ -e CFLAGS="-Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fstack-protector-strong -fstack-clash-protection -fcf-protection -O3 -Wl,--strip-all" -e CXXFLAGS="-Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fstack-protector-strong -fstack-clash-protection -fcf-protection -O3 -Wl,--strip-all" \
+          -e CCACHE_DIR=/cache \
+          --volume $(Build.SourcesDirectory):/onnxruntime_src \
+          --volume $(Build.BinariesDirectory):/build \
+          --volume $(CCACHE_DIR):/cache \
+          --workdir /onnxruntime_src \
+          onnxruntimetrainingrocm-cibuild-rocm$(RocmVersion)-manylinux-build \
+          /bin/bash -c "
+            set -ex; \
+            ccache -s; \
+            /opt/python/cp38-cp38/bin/python3 tools/ci_build/build.py \
+              --config $(BuildConfig) \
+              --enable_training \
+              --mpi_home /opt/ompi \
+              --cmake_extra_defines \
+                CMAKE_HIP_COMPILER=${ROCM_HOME}/llvm/bin/clang++ \
+                onnxruntime_BUILD_UNIT_TESTS=OFF \
+                FETCHCONTENT_TRY_FIND_PACKAGE_MODE=NEVER \
+              --use_cache \
+              --use_rocm \
+              --rocm_version=$(RocmVersion) \
+              --rocm_home ${ROCM_HOME} \
+              --nccl_home ${ROCM_HOME}\
+              --update \
+              --build_dir /build \
+              --build \
+              --parallel \
+              --build_wheel \
+              --skip_submodule_sync \
+              --skip_tests; \
+            ccache -sv; \
+            ccache -z"
+    displayName: 'Build onnxruntime'
+
+  - template: templates/explicitly-defined-final-tasks.yml
+
+- job: Linux_Build_ubuntu
   variables:
     skipComponentGovernanceDetection: true
     CCACHE_DIR: $(Pipeline.Workspace)/ccache
@@ -115,8 +217,7 @@ jobs:
 
   - template: templates/explicitly-defined-final-tasks.yml
 
-
-- job: Linux_Test
+- job: Linux_Test_ubuntu
   workspace:
     clean: all
   pool: AMD-GPU

tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_rocm

@@ -185,6 +185,13 @@ RUN cd /tmp/scripts && \
      rm -rf /tmp/scripts
 
 
+# Install ccache to reuse this dockerfile for CI
+RUN mkdir -p /tmp/ccache && \
+    cd /tmp/ccache && \
+    wget -q -O - https://github.com/ccache/ccache/releases/download/v4.7.4/ccache-4.7.4-linux-x86_64.tar.xz | tar --strip 1 -J -xf - && \
+    cp /tmp/ccache/ccache /usr/bin && \
+    rm -rf /tmp/ccache
+
 ARG BUILD_UID=1001
 ARG BUILD_USER=onnxruntimedev
 RUN adduser --uid $BUILD_UID $BUILD_USER