Decouple djl and serving in dockerfiles and gradle

and add serving publish to repo
deepjavalibrary · Oct 15, 2024 · d87235b · d87235b
1 parent 94644d1
commit d87235b
Show file tree

Hide file tree

Showing 11 changed files with 173 additions and 49 deletions.
diff --git a/.github/workflows/docker-nightly-publish.yml b/.github/workflows/docker-nightly-publish.yml
@@ -58,6 +58,13 @@ jobs:
         with:
           path: ~/.gradle/caches
           key: ${{ runner.os }}-gradle-${{ hashFiles('**/*.gradle*') }}
+      - name: Extract DJL and DJL Serving versions from TOML
+        id: get-versions
+        run: |
+          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          SERVING_VERSION=$(awk -F '=' '/serving / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          echo "DJL_VERSION=$DJL_VERSION" >> $GITHUB_ENV
+          echo "SERVING_VERSION=$SERVING_VERSION" >> $GITHUB_ENV
       - name: Build serving package for nightly
         if: ${{ inputs.mode == '' || inputs.mode == 'nightly' }}
         run: |
@@ -66,17 +73,21 @@ jobs:
         if: ${{ inputs.mode == '' || inputs.mode == 'nightly' }}
         working-directory: serving/docker
         run: |
-          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' ../../gradle/libs.versions.toml)
           export NIGHTLY="-nightly"
-          docker compose build --no-cache --build-arg djl_version=${DJL_VERSION}-SNAPSHOT ${{ matrix.arch }}
+          docker compose build --no-cache \
+          --build-arg djl_version=${{ env.DJL_VERSION }} \
+          --build-arg djl_serving_version=${{ env.SERVING_VERSION }} \
+          ${{ matrix.arch }}
           docker compose push ${{ matrix.arch }}
       - name: Build and push temp image
         if: ${{ inputs.mode == 'temp' }}
         working-directory: serving/docker
         run: |
-          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' ../../gradle/libs.versions.toml)
           export NIGHTLY="-nightly"
-          docker compose build --no-cache --build-arg djl_version=${DJL_VERSION}-SNAPSHOT ${{ matrix.arch }}
+          docker compose build --no-cache \
+          --build-arg djl_version=${{ env.DJL_VERSION }} \
+          --build-arg djl_serving_version=${{ env.SERVING_VERSION }} \
+          ${{ matrix.arch }}
           repo="185921645874.dkr.ecr.us-east-1.amazonaws.com/djl-ci-temp"
           aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin $repo
           tempTag="$repo:${{ matrix.arch }}-${GITHUB_SHA}"
@@ -86,17 +97,18 @@ jobs:
         if: ${{ inputs.mode == 'release' }}
         working-directory: serving/docker
         run: |
-          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' ../../gradle/libs.versions.toml)
-          export BASE_RELEASE_VERSION="${DJL_VERSION}"
-          export RELEASE_VERSION="${DJL_VERSION}-"
-          docker compose build --no-cache --build-arg djl_version=${DJL_VERSION} ${{ matrix.arch }}
+          export BASE_RELEASE_VERSION="${{ env.SERVING_VERSION }}"
+          export RELEASE_VERSION="${{ env.SERVING_VERSION }}-"
+          docker compose build --no-cache \
+          --build-arg djl_version=${{ env.DJL_VERSION }} \
+          --build-arg djl_serving_version=${{ env.SERVING_VERSION }} \
+          ${{ matrix.arch }}
           docker compose push ${{ matrix.arch }}
       - name: Retag image for release
         if: ${{ matrix.arch == 'cpu' && inputs.mode == 'release' }}
         working-directory: serving/docker
         run: |
-          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' ../../gradle/libs.versions.toml)
-          docker tag deepjavalibrary/djl-serving:${DJL_VERSION} deepjavalibrary/djl-serving:latest
+          docker tag deepjavalibrary/djl-serving:${{ env.SERVING_VERSION }} deepjavalibrary/djl-serving:latest
           docker push deepjavalibrary/djl-serving:latest
 
   create-runner:
@@ -147,6 +159,13 @@ jobs:
         with:
           path: ~/.gradle/caches
           key: ${{ runner.os }}-gradle-${{ hashFiles('**/*.gradle*') }}
+      - name: Extract DJL and DJL Serving versions from TOML
+        id: get-versions
+        run: |
+          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          SERVING_VERSION=$(awk -F '=' '/serving / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          echo "DJL_VERSION=$DJL_VERSION" >> $GITHUB_ENV
+          echo "SERVING_VERSION=$SERVING_VERSION" >> $GITHUB_ENV
       - name: Build serving package for nightly
         if: ${{ inputs.mode == '' || inputs.mode == 'nightly' }}
         run: |
@@ -155,17 +174,21 @@ jobs:
         if: ${{ inputs.mode == '' || inputs.mode == 'nightly' }}
         working-directory: serving/docker
         run: |
-          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' ../../gradle/libs.versions.toml)
           export NIGHTLY="-nightly"
-          docker compose build --no-cache --build-arg djl_version=${DJL_VERSION}-SNAPSHOT aarch64
+          docker compose build --no-cache \
+          --build-arg djl_version=${{ env.DJL_VERSION }} \
+          --build-arg djl_serving_version=${{ env.SERVING_VERSION }} \
+          aarch64
           docker compose push aarch64
       - name: Build and push temp image
         if: ${{ inputs.mode == 'temp' }}
         working-directory: serving/docker
         run: |
-          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' ../../gradle/libs.versions.toml)
           export NIGHTLY="-nightly"
-          docker compose build --no-cache --build-arg djl_version=${DJL_VERSION}-SNAPSHOT aarch64
+          docker compose build --no-cache \
+          --build-arg djl_version=${{ env.DJL_VERSION }} \
+          --build-arg djl_serving_version=${{ env.SERVING_VERSION }} \
+          aarch64
           repo="185921645874.dkr.ecr.us-east-1.amazonaws.com/djl-ci-temp"
           aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin $repo
           tempTag="$repo:aarch64-${GITHUB_SHA}"
@@ -175,10 +198,12 @@ jobs:
         if: ${{ inputs.mode == 'release' }}
         working-directory: serving/docker
         run: |
-          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' ../../gradle/libs.versions.toml)
-          export BASE_RELEASE_VERSION="${DJL_VERSION}"
-          export RELEASE_VERSION="${DJL_VERSION}-"
-          docker compose build --no-cache --build-arg djl_version=${DJL_VERSION} aarch64
+          export BASE_RELEASE_VERSION="${{ env.SERVING_VERSION }}"
+          export RELEASE_VERSION="${{ env.SERVING_VERSION }}-"
+          docker compose build --no-cache \
+          --build-arg djl_version=${{ env.DJL_VERSION }} \
+          --build-arg djl_serving_version=${{ env.SERVING_VERSION }} \
+          aarch64
           docker compose push aarch64
 
 

diff --git a/.github/workflows/serving-publish.yml b/.github/workflows/serving-publish.yml
@@ -0,0 +1,84 @@
+name: Serving publish
+on:
+  schedule:
+    - cron: '0 12 * * *'
+  push:
+    branches:
+      - decouple-djl  # Trigger on push events to the decouple-djl branch
+  workflow_dispatch:
+    inputs:
+      mode:
+        description: 'staging/snapshot, default is snapshot'
+        required: true
+        default: 'snapshot'
+      repo-id:
+        description: 'staging repository id for djl api packages'
+        required: false
+        default: ''
+permissions:
+  id-token: write
+  contents: read
+jobs:
+  publish:
+    if: github.repository == 'deepjavalibrary/djl-serving'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set up JDK 17
+        uses: actions/setup-java@v4
+        with:
+          distribution: 'corretto'
+          java-version: 17
+      # Enable gradle cache: https://github.com/actions/cache/blob/master/examples.md#java---gradle
+      - uses: actions/cache@v4
+        with:
+          path: ~/.gradle/caches
+          key: ${{ runner.os }}-gradle-${{ hashFiles('*/build.gradle.kts', 'engines/**/build.gradle.kts', 'extensions/**/build.gradle.kts') }}
+          restore-keys: |
+            ${{ runner.os }}-gradle-
+      - name: Configure AWS Credentials
+        uses: aws-actions/configure-aws-credentials@v4
+        with:
+          role-to-assume: arn:aws:iam::425969335547:role/djl-ci-publish-role
+          aws-region: us-east-2
+      - name: Copy serving snapshot artifacts to S3
+        if: ${{ github.event.inputs.mode == '' || github.event.inputs.mode == 'snapshot' }}
+        run: |
+          ./gradlew :serving:createDeb -Psnapshot
+          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          aws s3 cp serving/build/distributions/*.deb s3://djl-ai/publish/djl-serving/
+          aws cloudfront create-invalidation --distribution-id E371VB8JQ6NRVY --paths "/djl-serving/djl-serving_${DJL_VERSION}*"
+      - name: Copy serving release artifacts to S3
+        if: ${{ github.event.inputs.mode == 'staging' }}
+        run: |
+          ./gradlew :serving:dZ :serving:createDeb -Pstaging
+          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          aws s3 cp serving/build/distributions/*.deb s3://djl-ai/publish/djl-serving/
+          aws cloudfront create-invalidation --distribution-id E371VB8JQ6NRVY --paths "/djl-serving/djl-serving_${DJL_VERSION}*"
+          if [[ $(aws s3 ls s3://djl-ai/publish/djl-serving/serving-$DJL_VERSION.tar | wc -l) -eq 0 ]]; \
+            then aws s3 cp serving/build/distributions/*.tar s3://djl-ai/publish/djl-serving/; \
+            else echo serving tarball published already!; fi
+          aws s3 cp serving/build/distributions/*.zip s3://djl-ai/publish/djl-serving/
+          aws cloudfront create-invalidation --distribution-id E371VB8JQ6NRVY --paths "/djl-serving/serving-${DJL_VERSION}*"
+      - name: Copy benchmark release artifacts to S3
+        if: ${{ github.event.inputs.mode == 'staging' }}
+        run: |
+          ./gradlew :benchmark:dZ :benchmark:createDeb -Pstaging
+          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          aws s3 cp benchmark/build/distributions/*.tar s3://djl-ai/publish/djl-bench/${DJL_VERSION}/
+          aws s3 cp benchmark/build/distributions/*.deb s3://djl-ai/publish/djl-bench/${DJL_VERSION}/
+          aws s3 cp benchmark/build/distributions/*.zip s3://djl-ai/publish/djl-bench/${DJL_VERSION}/
+          aws cloudfront create-invalidation --distribution-id E371VB8JQ6NRVY --paths "/djl-bench/${DJL_VERSION}/*"
+      - name: Copy awscurl snapshot artifacts to S3
+        if: ${{ github.event.inputs.mode == '' || github.event.inputs.mode == 'snapshot' }}
+        run: |
+          ./gradlew :awscurl:jar
+          aws s3 cp awscurl/build/awscurl s3://djl-ai/publish/awscurl/
+          aws cloudfront create-invalidation --distribution-id E371VB8JQ6NRVY --paths "/awscurl/awscurl"
+      - name: Copy awscurl staging artifacts to S3
+        if: ${{ github.event.inputs.mode == 'staging' }}
+        run: |
+          ./gradlew :awscurl:jar -Pstaging
+          DJL_VERSION=$(awk -F '=' '/djl / {gsub(/ ?"/, "", $2); print $2}' gradle/libs.versions.toml)
+          aws s3 cp awscurl/build/awscurl s3://djl-ai/publish/awscurl/${DJL_VERSION}/
+          aws cloudfront create-invalidation --distribution-id E371VB8JQ6NRVY --paths "/awscurl/${DJL_VERSION}/*"
diff --git a/buildSrc/src/main/kotlin/ai/djl/javaBase.gradle.kts b/buildSrc/src/main/kotlin/ai/djl/javaBase.gradle.kts
@@ -13,7 +13,7 @@ plugins {
 val libs = the<LibrariesForLibs>()
 var servingVersion: String? = System.getenv("DJL_VERSION")
 val stagingRepo: String? = System.getenv("DJL_STAGING")
-servingVersion = if (servingVersion == null) libs.versions.djl.get() else servingVersion
+servingVersion = if (servingVersion == null) libs.versions.serving.get() else servingVersion
 if (!project.hasProperty("staging")) {
     servingVersion += "-SNAPSHOT"
 }

diff --git a/gradle/libs.versions.toml b/gradle/libs.versions.toml
@@ -3,6 +3,7 @@ format.version = "1.1"
 
 [versions]
 djl = "0.30.0"
+serving = "0.30.0"
 onnxruntime = "1.19.0"
 commonsCli = "1.9.0"
 commonsCodec = "1.17.1"

diff --git a/serving/docker/Dockerfile b/serving/docker/Dockerfile
@@ -10,7 +10,8 @@
 # BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, express or implied. See the License for
 # the specific language governing permissions and limitations under the License.
 FROM ubuntu:22.04 AS base
-ARG djl_version=0.30.0~SNAPSHOT
+ARG djl_version
+ARG djl_serving_version
 
 COPY scripts scripts/
 RUN mkdir -p /opt/djl/conf && \
@@ -20,10 +21,10 @@ COPY config.properties /opt/djl/conf/
 COPY distribution[s]/ ./
 RUN mv *.deb djl-serving_all.deb || true
 
-RUN scripts/install_djl_serving.sh $djl_version && \
+RUN scripts/install_djl_serving.sh $djl_version $djl_serving_version && \
     scripts/install_s5cmd.sh x64 && \
     mkdir -p /opt/djl/bin && cp scripts/telemetry.sh /opt/djl/bin && \
-    echo "${djl_version} cpu" > /opt/djl/bin/telemetry && \
+    echo "${djl_serving_version} cpu" > /opt/djl/bin/telemetry && \
     scripts/security_patch.sh cpu && \
     apt-get clean -y && rm -rf /var/lib/apt/lists/* && \
     rm -rf scripts
@@ -55,7 +56,7 @@ LABEL dlc_major_version="1"
 LABEL com.amazonaws.ml.engines.sagemaker.dlc.framework.djl.cpu="true"
 LABEL com.amazonaws.sagemaker.capabilities.multi-models="true"
 LABEL com.amazonaws.sagemaker.capabilities.accept-bind-to-port="true"
-LABEL djl-version=$djl_version
+LABEL djl-version=$djl_serving_version
 
 
 FROM base AS cpu-full
@@ -64,7 +65,7 @@ ARG torch_version=2.4.0
 
 COPY scripts scripts/
 RUN scripts/install_python.sh && \
-    scripts/install_djl_serving.sh $djl_version $torch_version && \
+    scripts/install_djl_serving.sh $djl_version $djl_serving_version $torch_version && \
     djl-serving -i ai.djl.pytorch:pytorch-native-cpu:$torch_version:linux-x86_64 && \
     djl-serving -i ai.djl.onnxruntime:onnxruntime-engine:$djl_version && \
     djl-serving -i com.microsoft.onnxruntime:onnxruntime:1.18.0 && \

diff --git a/serving/docker/aarch64.Dockerfile b/serving/docker/aarch64.Dockerfile
@@ -10,7 +10,8 @@
 # BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, express or implied. See the License for
 # the specific language governing permissions and limitations under the License.
 FROM arm64v8/ubuntu:22.04
-ARG djl_version=0.30.0~SNAPSHOT
+ARG djl_version
+ARG djl_serving_version
 ARG torch_version=2.3.1
 
 EXPOSE 8080
@@ -38,13 +39,13 @@ COPY config.properties /opt/djl/conf/
 COPY distribution[s]/ ./
 RUN mv *.deb djl-serving_all.deb || true
 
-RUN scripts/install_djl_serving.sh $djl_version && \
-    scripts/install_djl_serving.sh $djl_version $torch_version && \
-    scripts/install_djl_serving.sh $djl_version $torch_version && \
+RUN scripts/install_djl_serving.sh $djl_version $djl_serving_version && \
+    scripts/install_djl_serving.sh $djl_version $djl_serving_version $torch_version && \
+    scripts/install_djl_serving.sh $djl_version $djl_serving_version $torch_version && \
     scripts/install_s5cmd.sh aarch64 && \
     djl-serving -i ai.djl.pytorch:pytorch-native-cpu-precxx11:$torch_version:linux-aarch64 && \
     mkdir -p /opt/djl/bin && cp scripts/telemetry.sh /opt/djl/bin && \
-    echo "${djl_version} aarch" > /opt/djl/bin/telemetry && \
+    echo "${djl_serving_version} aarch" > /opt/djl/bin/telemetry && \
     rm -f /usr/local/djl-serving-*/lib/mxnet-* && \
     rm -f /usr/local/djl-serving-*/lib/tensorflow-* && \
     rm -f /usr/local/djl-serving-*/lib/tensorrt-* && \
@@ -61,4 +62,5 @@ LABEL com.amazonaws.ml.engines.sagemaker.dlc.framework.djl.v0-30-0.aarch64="true
 LABEL com.amazonaws.sagemaker.capabilities.multi-models="true"
 LABEL com.amazonaws.sagemaker.capabilities.accept-bind-to-port="true"
 LABEL djl-version=$djl_version
+LABEL djl-serving-version=$djl_serving_version
 LABEL torch-version=$torch_version
diff --git a/serving/docker/lmi.Dockerfile b/serving/docker/lmi.Dockerfile
@@ -12,7 +12,8 @@
 ARG version=12.4.1-devel-ubuntu22.04
 FROM nvidia/cuda:$version
 ARG cuda_version=cu124
-ARG djl_version=0.30.0~SNAPSHOT
+ARG djl_version
+ARG djl_serving_version
 # Base Deps
 ARG python_version=3.10
 ARG torch_version=2.4.0
@@ -88,14 +89,14 @@ RUN mv *.deb djl-serving_all.deb || true
 
 RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -yq libaio-dev libopenmpi-dev g++ \
     && scripts/install_openssh.sh \
-    && scripts/install_djl_serving.sh $djl_version \
-    && scripts/install_djl_serving.sh $djl_version ${torch_version} \
+    && scripts/install_djl_serving.sh $djl_version $djl_serving_version \
+    && scripts/install_djl_serving.sh $djl_version $djl_serving_version ${torch_version} \
     && djl-serving -i ai.djl.onnxruntime:onnxruntime-engine:$djl_version \
     && djl-serving -i com.microsoft.onnxruntime:onnxruntime_gpu:$onnx_version \
     && scripts/install_python.sh ${python_version} \
     && scripts/install_s5cmd.sh x64 \
     && mkdir -p /opt/djl/bin && cp scripts/telemetry.sh /opt/djl/bin \
-    && echo "${djl_version} lmi" > /opt/djl/bin/telemetry \
+    && echo "${djl_serving_version} lmi" > /opt/djl/bin/telemetry \
     && pip3 cache purge \
     && apt-get clean -y && rm -rf /var/lib/apt/lists/*
 
@@ -132,6 +133,7 @@ LABEL com.amazonaws.ml.engines.sagemaker.dlc.framework.djl.v0-30-0.lmi="true"
 LABEL com.amazonaws.sagemaker.capabilities.multi-models="true"
 LABEL com.amazonaws.sagemaker.capabilities.accept-bind-to-port="true"
 LABEL djl-version=$djl_version
+LABEL djl-serving-version=$djl_serving_version
 LABEL cuda-version=$cuda_version
 # To use the 535 CUDA driver, CUDA 12.4 can work on this one too
 LABEL com.amazonaws.sagemaker.inference.cuda.verified_versions=12.4
diff --git a/serving/docker/pytorch-gpu.Dockerfile b/serving/docker/pytorch-gpu.Dockerfile
@@ -13,7 +13,8 @@ ARG version=12.4.1-devel-ubuntu22.04
 
 FROM nvidia/cuda:$version as base
 
-ARG djl_version=0.30.0~SNAPSHOT
+ARG djl_version
+ARG djl_serving_version
 ARG cuda_version=cu124
 ARG torch_version=2.4.0
 ARG torch_vision_version=0.19.0
@@ -52,8 +53,8 @@ RUN mv *.deb djl-serving_all.deb || true
 COPY scripts scripts/
 SHELL ["/bin/bash", "-c"]
 RUN chmod +x /usr/local/bin/dockerd-entrypoint.sh && \
-    scripts/install_djl_serving.sh $djl_version && \
-    scripts/install_djl_serving.sh $djl_version ${torch_version} && \
+    scripts/install_djl_serving.sh $djl_version $djl_serving_version && \
+    scripts/install_djl_serving.sh $djl_version $djl_serving_version ${torch_version} && \
     djl-serving -i ai.djl.onnxruntime:onnxruntime-engine:$djl_version && \
     djl-serving -i com.microsoft.onnxruntime:onnxruntime_gpu:$onnx_version && \
     scripts/install_python.sh ${python_version} && \
@@ -62,7 +63,7 @@ RUN chmod +x /usr/local/bin/dockerd-entrypoint.sh && \
     scripts/patch_oss_dlc.sh python && \
     scripts/security_patch.sh pytorch-gpu && \
     mkdir -p /opt/djl/bin && cp scripts/telemetry.sh /opt/djl/bin && \
-    echo "${djl_version} pytorchgpu" > /opt/djl/bin/telemetry && \
+    echo "${djl_serving_version} pytorchgpu" > /opt/djl/bin/telemetry && \
     useradd -m -d /home/djl djl && \
     chown -R djl:djl /opt/djl && \
     rm -rf scripts && pip3 cache purge && \
@@ -81,6 +82,7 @@ LABEL com.amazonaws.ml.engines.sagemaker.dlc.framework.djl.v0-30-0.pytorch-cu124
 LABEL com.amazonaws.sagemaker.capabilities.multi-models="true"
 LABEL com.amazonaws.sagemaker.capabilities.accept-bind-to-port="true"
 LABEL djl-version=$djl_version
+LABEL djl-serving-version=$djl_serving_version
 LABEL cuda-version=$cuda_version
 LABEL torch-version=$torch_version
 # To use the 535 CUDA driver