WIP: Create GitHub actions for ROCm

.github/workflows/ci_gpu.yml

@@ -16,7 +16,13 @@ concurrency:
 
 jobs:
   ci-gpu:
-    runs-on: [self-hosted, ci-gpu]
+    runs-on:
+      - self-hosted
+      - ci-gpu
+      - ${{ matrix.gpu-type }}]
+    strategy:
+      matrix:
+        gpu-type: [cuda, rocm]
     steps:
       - name: Clean up previous run
         run: |
@@ -41,9 +47,14 @@ jobs:
         uses: actions/checkout@v3
         with:
           submodules: recursive
-      - name: Install dependencies
+      - name: Install dependencies for CUDA
+        if: matrix.gpu-type == 'cuda'
         run: |
           python ts_scripts/install_dependencies.py --environment=dev --cuda=cu121
+      - name: Install dependencies for ROCm
+        if: matrix.gpu-type == 'rocm'
+        run: |
+          python ts_scripts/install_dependencies.py --environment=dev --rocm=rocm62
       - name: Torchserve Sanity
         uses: nick-fields/retry@v3
         with:

.github/workflows/kserve_gpu_tests.yml

@@ -8,7 +8,13 @@ on:
 
 jobs:
   kserve-gpu-tests:
-    runs-on: [self-hosted, regression-test-gpu]
+    runs-on:
+      - self-hosted
+      - regression-test-gpu
+      - ${{ matrix.gpu-type }}
+    strategy:
+      matrix:
+        gpu-type: [cuda, rocm]
     steps:
       - name: Clean up previous run
         run: |

.github/workflows/regression_tests_gpu.yml

@@ -17,7 +17,13 @@ concurrency:
 jobs:
   regression-gpu:
     # creates workflows on self hosted runner
-    runs-on: [self-hosted, regression-test-gpu]
+    runs-on:
+      - self-hosted
+      - regression-test-gpu
+      - ${{ matrix.gpu-type }}
+    strategy:
+      matrix:
+        gpu-type: [cuda, rocm]
     steps:
       - name: Clean up previous run
         run: |
@@ -44,9 +50,14 @@ jobs:
         uses: actions/checkout@v3
         with:
           submodules: recursive
-      - name: Install dependencies
+      - name: Install dependencies for CUDA
+        if: matrix.gpu-type == 'cuda'
         run: |
           python ts_scripts/install_dependencies.py --environment=dev --cuda=cu121
+      - name: Install dependencies for ROCm
+        if: matrix.gpu-type == 'rocm'
+        run: |
+          python ts_scripts/install_dependencies.py --environment=dev --rocm=rocm62
       - name: Torchserve Regression Tests
         run: |
           export TS_RUN_IN_DOCKER=False

docker/Dockerfile

@@ -37,12 +37,12 @@ ARG BRANCH_NAME
 ARG REPO_URL=https://github.com/pytorch/serve.git
 ENV PYTHONUNBUFFERED TRUE
 
-RUN --mount=type=cache,id=apt-dev,target=/var/cache/apt \
+RUN --mount=type=cache,sharing=locked,id=apt-dev,target=/var/cache/apt \
     apt-get update && \
     apt-get upgrade -y && \
     apt-get install software-properties-common -y && \
     add-apt-repository -y ppa:deadsnakes/ppa && \
-    apt remove python-pip  python3-pip && \
+    apt remove -y python-pip python3-pip && \
     DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y \
         ca-certificates \
         g++ \
@@ -55,6 +55,13 @@ RUN --mount=type=cache,id=apt-dev,target=/var/cache/apt \
         git \
     && rm -rf /var/lib/apt/lists/*
 
+RUN --mount=type=cache,sharing=locked,id=apt-dev,target=/var/cache/apt \
+    if [ "$USE_ROCM_VERSION" ]; then \
+        apt-get update && \
+        DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y rocm-dev amd-smi-lib \
+        && rm -rf /var/lib/apt/lists/* ; \
+    fi
+
 # Make the virtual environment and "activating" it by adding it first to the path.
 # From here on the python$PYTHON_VERSION interpreter is used and the packages
 # are installed in /home/venv which is what we need for the "runtime-image"
@@ -67,6 +74,7 @@ RUN python -m pip install -U pip setuptools
 RUN export USE_CUDA=1
 
 ARG USE_CUDA_VERSION=""
+ARG USE_ROCM_VERSION=""
 
 COPY ./ serve
 
@@ -76,7 +84,6 @@ RUN \
         git clone --recursive $REPO_URL -b $BRANCH_NAME serve; \
     fi
 
-
 WORKDIR "serve"
 
 RUN cp docker/dockerd-entrypoint.sh /usr/local/bin/dockerd-entrypoint.sh
@@ -90,6 +97,14 @@ RUN \
         else \
             python ./ts_scripts/install_dependencies.py;\
         fi; \
+    elif echo "${BASE_IMAGE}" | grep -q "rocm/"; then \
+        # Install ROCm version specific binary when ROCm version is specified as a build arg
+        if [ "$USE_ROCM_VERSION" ]; then \
+            python ./ts_scripts/install_dependencies.py --rocm $USE_ROCM_VERSION;\
+        # Install the binary with the latest CPU image on a ROCm base image
+        else \
+            python ./ts_scripts/install_dependencies.py; \
+        fi; \
     # Install the CPU binary
     else \
         python ./ts_scripts/install_dependencies.py; \
@@ -111,13 +126,14 @@ FROM ${BASE_IMAGE} AS production-image
 # Re-state ARG PYTHON_VERSION to make it active in this build-stage (uses default define at the top)
 ARG PYTHON_VERSION
 ENV PYTHONUNBUFFERED TRUE
+ARG USE_ROCM_VERSION
 
-RUN --mount=type=cache,target=/var/cache/apt \
+RUN --mount=type=cache,sharing=locked,target=/var/cache/apt \
     apt-get update && \
     apt-get upgrade -y && \
     apt-get install software-properties-common -y && \
     add-apt-repository ppa:deadsnakes/ppa -y && \
-    apt remove python-pip  python3-pip && \
+    apt remove -y python-pip python3-pip && \
     DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y \
     python$PYTHON_VERSION \
     python3-distutils \
@@ -130,13 +146,25 @@ RUN --mount=type=cache,target=/var/cache/apt \
     && rm -rf /var/lib/apt/lists/* \
     && cd /tmp
 
+RUN --mount=type=cache,sharing=locked,id=apt-dev,target=/var/cache/apt \
+    if [ "$USE_ROCM_VERSION" ]; then \
+        apt-get update && \
+        DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y rocm-dev amd-smi-lib \
+        && rm -rf /var/lib/apt/lists/* ; \
+    fi
+
 RUN useradd -m model-server \
     && mkdir -p /home/model-server/tmp
 
 COPY --chown=model-server --from=compile-image /home/venv /home/venv
 COPY --from=compile-image /usr/local/bin/dockerd-entrypoint.sh /usr/local/bin/dockerd-entrypoint.sh
 ENV PATH="/home/venv/bin:$PATH"
 
+RUN \
+    if [ "$USE_ROCM_VERSION" ]; then \
+        python -m pip install /opt/rocm/share/amd_smi; \
+    fi
+
 RUN chmod +x /usr/local/bin/dockerd-entrypoint.sh \
     && chown -R model-server /home/model-server
 
@@ -157,13 +185,14 @@ FROM ${BASE_IMAGE} AS ci-image
 ARG PYTHON_VERSION
 ARG BRANCH_NAME
 ENV PYTHONUNBUFFERED TRUE
+ARG USE_ROCM_VERSION
 
-RUN --mount=type=cache,target=/var/cache/apt \
+RUN --mount=type=cache,sharing=locked,target=/var/cache/apt \
     apt-get update && \
     apt-get upgrade -y && \
     apt-get install software-properties-common -y && \
     add-apt-repository -y ppa:deadsnakes/ppa && \
-    apt remove python-pip  python3-pip && \
+    apt remove -y python-pip python3-pip && \
     DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y \
     python$PYTHON_VERSION \
     python3-distutils \
@@ -183,13 +212,24 @@ RUN --mount=type=cache,target=/var/cache/apt \
     && rm -rf /var/lib/apt/lists/* \
     && cd /tmp
 
+RUN --mount=type=cache,sharing=locked,id=apt-dev,target=/var/cache/apt \
+    if [ "$USE_ROCM_VERSION" ]; then \
+        apt-get update && \
+        DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y rocm-dev amd-smi-lib \
+        && rm -rf /var/lib/apt/lists/* ; \
+    fi
 
 COPY --from=compile-image /home/venv /home/venv
 
 ENV PATH="/home/venv/bin:$PATH"
 
 RUN python -m pip install --no-cache-dir -r https://raw.githubusercontent.com/pytorch/serve/$BRANCH_NAME/requirements/developer.txt
 
+RUN \
+    if [ "$USE_ROCM_VERSION" ]; then \
+        python -m pip install /opt/rocm/share/amd_smi; \
+    fi
+
 RUN mkdir /home/serve
 ENV TS_RUN_IN_DOCKER True
 
@@ -203,11 +243,13 @@ ARG PYTHON_VERSION
 ARG BRANCH_NAME
 ARG BUILD_FROM_SRC
 ARG LOCAL_CHANGES
+ARG USE_ROCM_VERSION
 ARG BUILD_WITH_IPEX
 ARG IPEX_VERSION=1.11.0
 ARG IPEX_URL=https://software.intel.com/ipex-whl-stable
 ENV PYTHONUNBUFFERED TRUE
-RUN --mount=type=cache,target=/var/cache/apt \
+
+RUN --mount=type=cache,sharing=locked,target=/var/cache/apt \
     apt-get update && \
     apt-get upgrade -y && \
     apt-get install software-properties-common -y && \
@@ -227,9 +269,15 @@ RUN --mount=type=cache,target=/var/cache/apt \
     # https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1009905
     openjdk-17-jdk \
     build-essential \
+    wget \
     curl \
     vim \
     numactl \
+    nodejs \
+    npm \
+    zip \
+    unzip \
+    && npm install -g [email protected] newman-reporter-htmlextra markdown-link-check \
     && if [ "$BUILD_WITH_IPEX" = "true" ]; then apt-get update && apt-get install -y libjemalloc-dev libgoogle-perftools-dev libomp-dev && ln -s /usr/lib/x86_64-linux-gnu/libjemalloc.so /usr/lib/libjemalloc.so && ln -s /usr/lib/x86_64-linux-gnu/libtcmalloc.so /usr/lib/libtcmalloc.so && ln -s /usr/lib/x86_64-linux-gnu/libiomp5.so /usr/lib/libiomp5.so; fi \
     && rm -rf /var/lib/apt/lists/*
 
@@ -243,10 +291,17 @@ RUN \
 
 COPY --from=compile-image /home/venv /home/venv
 ENV PATH="/home/venv/bin:$PATH"
+
+RUN \
+    if [ "$USE_ROCM_VERSION" ]; then \
+        python -m pip install /opt/rocm/share/amd_smi; \
+    fi
+
 WORKDIR "serve"
+
 RUN python -m pip install -U pip setuptools \
     && python -m pip install --no-cache-dir -r requirements/developer.txt \
-    && python ts_scripts/install_from_src.py \
+    && python ts_scripts/install_from_src.py --environment=dev\
     && useradd -m model-server \
     && mkdir -p /home/model-server/tmp \
     && cp docker/dockerd-entrypoint.sh /usr/local/bin/dockerd-entrypoint.sh \