Abjindal/eager mode pipeline (#8870)

* Adding pipeline file for eager mode

* adding the build eager mode flag

* adding torch wheel files for installation

* Changing pytorch version for change in wheel files

* updating requirements file path

* Removing Java and NodeJS from the build

* removing import torch for testing build of eager mode

* changing the build command

* import torch

* building eager mode separately

* removing Java tests

* python path issues

* changing python path location

* changing the build path file loc

* installing torch before build

* setting environment for building eager mode

* Copying the build file and getting rid of flags

* changing python path

* adding missing packages

* moving build eager mode code

* changing python path to python3

* adding amd_hipify

* adding logger file

* install torch before build

* change requirements file location

* install torch before build eager

* modifying eager mode build

* modifying build location

* adding new docker image

* handling gradle move issue

* Typo fix

* changing deps file

* adding java and nodejs

* changing repo name for docker image

* removing pybind

* building only eager mode

* changing the image name

* removing install wheel package

* build complete onnxruntime with eager mode

* building wheel

* enabling pybind

* adding build eager mode flag in unit tests

* removing build java nodejs

* adding build command

* removing java tests

* moving Debug tests before Release

* building Debug only case

* changing debug test code

* running the build eager mode with tests

* adding build dir

* adding build dir path

* changing build dir path

* changing build command for eager mode

* building eager mode and running tests simultaneously

* adding more flags to the pipeline

* chaning flag

* adding Debug and Release

* changing torch to nightly build

* changing torch version for nightly build

* chaning torch version

* move to Ubuntu image

* adding pool

* adding dockerfile for eager mode

* adding python deps file for eager

* modifying python deps file for eager

* changing deps file

* changing deps file statements

* changing python path

* REMOVING ECHO line

* going to original docker file

* changing docker file

* changing to eager requirements file

* changing python deps file

* changing paths

* changing cmake path

* changing build script

* changing python installation

* running debug mode only

* changing pipeline file

* test name

* test name

* test name2

* changing requirements file

* final flags for eager mode

* previous pipeline

* moving to ubuntu image and including some deps

* adding cmake path

* returning to manylinux image

* removing unncecessary files for pipeline

tools/ci_build/github/azure-pipelines/linux-cpu-eager-pipeline.yml

@@ -0,0 +1,45 @@
+jobs:
+- job: Linux_Build
+  timeoutInMinutes: 120
+  workspace:
+    clean: all
+  pool: Linux-CPU-2019
+  steps:
+  - checkout: self
+    clean: true
+    submodules: recursive
+
+  - task: NodeTool@0
+    inputs:
+      versionSpec: '12.16.3'
+
+  - template: templates/get-docker-image-steps.yml
+    parameters:
+      Dockerfile: tools/ci_build/github/linux/docker/Dockerfile.manylinux2014_eager_cpu
+      Context: tools/ci_build/github/linux/docker
+      DockerBuildArgs: "--build-arg BUILD_UID=$( id -u )"
+      Repository: onnxruntimecpubuildeagermode
+
+  - task: CmdLine@2
+    displayName: 'build'
+    inputs:
+      script: |
+        mkdir -p $HOME/.onnx
+        docker run --rm \
+          --volume /data/onnx:/data/onnx:ro \
+          --volume $(Build.SourcesDirectory):/onnxruntime_src \
+          --volume $(Build.BinariesDirectory):/build \
+          --volume $HOME/.onnx:/home/onnxruntimedev/.onnx \
+          -e ALLOW_RELEASED_ONNX_OPSET_ONLY=0 \
+          -e NIGHTLY_BUILD \
+          -e BUILD_BUILDNUMBER \
+          onnxruntimecpubuildeagermode \
+            /opt/python/cp38-cp38/bin/python3 /onnxruntime_src/tools/ci_build/build.py \
+              --build_dir /build --cmake_generator Ninja \
+              --config Debug Release \
+              --skip_submodule_sync \
+              --build_shared_lib \
+              --parallel \
+              --build_eager_mode --enable_training --enable_pybind
+      workingDirectory: $(Build.SourcesDirectory)
+

tools/ci_build/github/linux/docker/Dockerfile.manylinux2014_eager_cpu

@@ -0,0 +1,13 @@
+FROM quay.io/pypa/manylinux2014_x86_64:latest
+
+ENV PATH /usr/local/gradle/bin:/opt/rh/devtoolset-10/root/usr/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+
+ADD scripts /tmp/scripts
+RUN yum remove -y devtoolset\* && yum install -y devtoolset-10-binutils devtoolset-10-gcc devtoolset-10-gcc-c++ devtoolset-10-gcc-gfortran && cd /tmp/scripts && /tmp/scripts/manylinux/install_centos.sh && /tmp/scripts/manylinux/install_deps_eager.sh && rm -rf /tmp/scripts
+
+ARG BUILD_UID=1001
+ARG BUILD_USER=onnxruntimedev
+RUN adduser --uid $BUILD_UID $BUILD_USER
+WORKDIR /home/$BUILD_USER
+USER $BUILD_USER
+

tools/ci_build/github/linux/docker/scripts/manylinux/install_deps_eager.sh

@@ -0,0 +1,109 @@
+#!/bin/bash
+set -e -x
+
+# Development tools and libraries
+yum -y install \
+    graphviz
+
+# Download a file from internet
+function GetFile {
+  local uri=$1
+  local path=$2
+  local force=${3:-false}
+  local download_retries=${4:-5}
+  local retry_wait_time_seconds=${5:-30}
+
+  if [[ -f $path ]]; then
+    if [[ $force = false ]]; then
+      echo "File '$path' already exists. Skipping download"
+      return 0
+    else
+      rm -rf $path
+    fi
+  fi
+
+  if [[ -f $uri ]]; then
+    echo "'$uri' is a file path, copying file to '$path'"
+    cp $uri $path
+    return $?
+  fi
+
+  echo "Downloading $uri"
+  # Use aria2c if available, otherwise use curl
+  if command -v aria2c > /dev/null; then
+    aria2c -q -d $(dirname $path) -o $(basename $path) "$uri"
+  else
+    curl "$uri" -sSL --retry $download_retries --retry-delay $retry_wait_time_seconds --create-dirs -o "$path" --fail
+  fi
+
+  return $?
+}
+
+if [ ! -d "/opt/conda/bin" ]; then
+    PYTHON_EXES=("/opt/python/cp36-cp36m/bin/python3.6" "/opt/python/cp37-cp37m/bin/python3.7" "/opt/python/cp38-cp38/bin/python3.8" "/opt/python/cp39-cp39/bin/python3.9")
+else
+    PYTHON_EXES=("/opt/conda/bin/python")
+fi
+
+os_major_version=$(cat /etc/redhat-release | tr -dc '0-9.'|cut -d \. -f1)
+
+SYS_LONG_BIT=$(getconf LONG_BIT)
+mkdir -p /tmp/src
+GLIBC_VERSION=$(getconf GNU_LIBC_VERSION | cut -f 2 -d \.)
+
+DISTRIBUTOR=$(lsb_release -i -s)
+
+if [[ "$DISTRIBUTOR" = "CentOS" && $SYS_LONG_BIT = "64" ]]; then
+  LIBDIR="lib64"
+else
+  LIBDIR="lib"
+fi
+
+cd /tmp/src
+
+echo "Installing azcopy"
+mkdir -p /tmp/azcopy
+GetFile https://aka.ms/downloadazcopy-v10-linux /tmp/azcopy/azcopy.tar.gz
+tar --strip 1 -xf /tmp/azcopy/azcopy.tar.gz -C /tmp/azcopy
+cp /tmp/azcopy/azcopy /usr/bin
+
+echo "Installing Ninja"
+GetFile https://github.com/ninja-build/ninja/archive/v1.10.0.tar.gz /tmp/src/ninja-linux.tar.gz
+tar -zxf ninja-linux.tar.gz
+cd ninja-1.10.0
+cmake -Bbuild-cmake -H.
+cmake --build build-cmake
+mv ./build-cmake/ninja /usr/bin
+echo "Installing Node.js"
+GetFile https://nodejs.org/dist/v12.16.3/node-v12.16.3-linux-x64.tar.gz /tmp/src/node-v12.16.3-linux-x64.tar.gz
+tar --strip 1 -xf /tmp/src/node-v12.16.3-linux-x64.tar.gz -C /usr
+
+cd /tmp/src
+GetFile https://downloads.gradle-dn.com/distributions/gradle-6.3-bin.zip /tmp/src/gradle-6.3-bin.zip
+unzip /tmp/src/gradle-6.3-bin.zip
+# rm -rf /usr/local/gradle/*
+mv /tmp/src/gradle-6.3 /usr/local/gradle
+
+if ! [ -x "$(command -v protoc)" ]; then
+  source ${0/%install_deps_eager.sh/..\/install_protobuf.sh}
+fi
+
+export ONNX_ML=1
+export CMAKE_ARGS="-DONNX_GEN_PB_TYPE_STUBS=OFF -DONNX_WERROR=OFF"
+
+for PYTHON_EXE in "${PYTHON_EXES[@]}"
+do
+  ${PYTHON_EXE} -m pip install -r ${0/%install_deps_eager\.sh/requirements\.txt}
+  ${PYTHON_EXE} -m pip install -r ${0/%install_deps_eager\.sh/..\/training\/ortmodule\/stage1\/requirements_torch_eager_cpu.txt}
+done
+
+cd /tmp/src
+GetFile 'https://sourceware.org/pub/valgrind/valgrind-3.16.1.tar.bz2' /tmp/src/valgrind-3.16.1.tar.bz2
+tar -jxvf valgrind-3.16.1.tar.bz2
+cd valgrind-3.16.1
+./configure --prefix=/usr --libdir=/usr/lib64 --enable-only64bit --enable-tls
+make -j$(getconf _NPROCESSORS_ONLN)
+make install
+
+cd /
+rm -rf /tmp/src

tools/ci_build/github/linux/docker/scripts/training/ortmodule/stage1/requirements_torch_eager_cpu.txt

@@ -0,0 +1,4 @@
+--pre
+-f https://download.pytorch.org/whl/nightly/cpu/torch_nightly.html
+torch>=1.9.0
+setuptools>=41.4.0