From 37d8bed53d88d6fff85b1fd86827f83e5573ca06 Mon Sep 17 00:00:00 2001 From: PeixuanZuo <94887879+PeixuanZuo@users.noreply.github.com> Date: Tue, 14 Nov 2023 11:54:22 +0800 Subject: [PATCH] [ROCm] add migraphx into onnxruntime-training-rocm package (#18339) --- setup.py | 4 ++++ .../github/azure-pipelines/orttraining-pai-ci-pipeline.yml | 1 + tools/ci_build/github/azure-pipelines/templates/rocm.yml | 1 + .../github/linux/docker/Dockerfile.manylinux2_28_rocm | 2 +- tools/ci_build/github/pai/rocm-ci-pipeline-env.Dockerfile | 4 +++- 5 files changed, 10 insertions(+), 2 deletions(-) diff --git a/setup.py b/setup.py index 9eca9845c9..1c04433c9a 100644 --- a/setup.py +++ b/setup.py @@ -219,6 +219,10 @@ try: "librocm_smi64.so.5", "libroctracer64.so.4", "libtinfo.so.6", + "libmigraphx_c.so.3", + "libmigraphx.so.2", + "libmigraphx_onnx.so.2", + "libmigraphx_tf.so.2", ] tensorrt_dependencies = ["libnvinfer.so.8", "libnvinfer_plugin.so.8", "libnvonnxparser.so.8"] diff --git a/tools/ci_build/github/azure-pipelines/orttraining-pai-ci-pipeline.yml b/tools/ci_build/github/azure-pipelines/orttraining-pai-ci-pipeline.yml index ed84b514fb..8d02a5e580 100644 --- a/tools/ci_build/github/azure-pipelines/orttraining-pai-ci-pipeline.yml +++ b/tools/ci_build/github/azure-pipelines/orttraining-pai-ci-pipeline.yml @@ -108,6 +108,7 @@ jobs: FETCHCONTENT_TRY_FIND_PACKAGE_MODE=NEVER \ --use_cache \ --use_rocm \ + --use_migraphx \ --rocm_version=$(RocmVersion) \ --rocm_home ${ROCM_HOME} \ --nccl_home ${ROCM_HOME}\ diff --git a/tools/ci_build/github/azure-pipelines/templates/rocm.yml b/tools/ci_build/github/azure-pipelines/templates/rocm.yml index f41dc1c49f..2e9e6c6b35 100644 --- a/tools/ci_build/github/azure-pipelines/templates/rocm.yml +++ b/tools/ci_build/github/azure-pipelines/templates/rocm.yml @@ -79,6 +79,7 @@ jobs: /onnxruntime_src/tools/ci_build/build.py \ --config ${{ parameters.BuildConfig }} \ --use_rocm \ + --use_migraphx \ --rocm_version=${{ parameters.RocmVersion }} \ --rocm_home=/opt/rocm \ --nccl_home=/opt/rocm \ diff --git a/tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_rocm b/tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_rocm index 19599c9f61..9e12fe8c75 100644 --- a/tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_rocm +++ b/tools/ci_build/github/linux/docker/Dockerfile.manylinux2_28_rocm @@ -31,7 +31,7 @@ RUN yum install -y hipify-clang RUN yum -y install wget # rocm lib -RUN yum install -y miopen-hip-devel rocblas-devel rocrand-devel rccl-devel hipsparse-devel hipfft-devel hipcub-devel hipblas-devel rocthrust-devel +RUN yum install -y miopen-hip-devel rocblas-devel rocrand-devel rccl-devel hipsparse-devel hipfft-devel hipcub-devel hipblas-devel rocthrust-devel migraphx-devel ENV AUDITWHEEL_POLICY=${POLICY} AUDITWHEEL_ARCH=${PLATFORM} AUDITWHEEL_PLAT=${POLICY}_${PLATFORM} ENV LC_ALL=en_US.UTF-8 LANG=en_US.UTF-8 LANGUAGE=en_US.UTF-8 diff --git a/tools/ci_build/github/pai/rocm-ci-pipeline-env.Dockerfile b/tools/ci_build/github/pai/rocm-ci-pipeline-env.Dockerfile index 412bc00d02..2ec826fc8f 100644 --- a/tools/ci_build/github/pai/rocm-ci-pipeline-env.Dockerfile +++ b/tools/ci_build/github/pai/rocm-ci-pipeline-env.Dockerfile @@ -80,7 +80,6 @@ RUN ln -sf /usr/lib/x86_64-linux-gnu/libstdc++.so.6 ${CONDA_ENVIRONMENT_PATH}/bi RUN pip install torch==2.0.1 torchvision==0.15.2 -f https://repo.radeon.com/rocm/manylinux/rocm-rel-${ROCM_VERSION}/ && \ pip install torch-ort --no-dependencies - ##### Install Cupy to decrease CPU utilization # Install non dev openmpi RUN wget https://download.open-mpi.org/release/open-mpi/v4.1/openmpi-4.1.5.tar.bz2 && \ @@ -130,6 +129,9 @@ RUN pip install \ pytest-xdist \ pytest-rerunfailures +# Install migraphx +RUN apt update && apt install -y migraphx + ENV ORTMODULE_ONNX_OPSET_VERSION=15 ARG BUILD_UID=1001