feat: add dockerfiles and cd workflow (#712)

* feat: add dockerfiles and GH runer * fix: turn on docker build action * fix: docker push runner * fix: docker push runner * fix: docker push runner * fix: test cuda dockerfile * fix: test cuda dockerfile * fix: cuda * fix: cuda * fix: cuda * fix: cuda * fix: cuda * fix: cuda * fix: cuda * fix: cuda * fix: update runner
jina-ai · May 11, 2022 · edf0d86 · edf0d86
1 parent 72d69c7
commit edf0d86
Show file tree

Hide file tree

Showing 9 changed files with 253 additions and 3 deletions.
diff --git a/.dockerignore b/.dockerignore
@@ -0,0 +1,4 @@
+.git
+.github
+scripts
+docs
diff --git a/.github/workflows/force-docker-build.yml b/.github/workflows/force-docker-build.yml
@@ -0,0 +1,128 @@
+name: Manual Docker Build
+
+on:
+  workflow_dispatch:
+    inputs:
+      release_token:
+        description: 'Your release token'
+        required: true
+      triggered_by:
+        description: 'CD | TAG | MANUAL'
+        required: false
+        default: MANUAL
+
+jobs:
+  token-check:
+    runs-on: ubuntu-latest
+    steps:
+      - run: echo "success!"
+        if: "${{ github.event.inputs.release_token }} == ${{ env.release_token }}"
+        env:
+          release_token: ${{ secrets.CAS_RELEASE_TOKEN }}
+
+  regular-release:
+    needs: token-check
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        pip_tag: [ "", "onnx"]  # default: "" = torch
+        engine_tag: ["", "cuda"]  # default: "" = cpu
+    steps:
+      - uses: actions/checkout@v2
+      - name: Set envs and versions
+        run: |
+          VCS_REF=${{ github.ref }}
+          echo "VCS_REF=$VCS_REF" >> $GITHUB_ENV
+          echo "Will build $VCS_REF"
+          echo "BUILD_DATE=$(date -u +'%Y-%m-%dT%H:%M:%SZ')" >> $GITHUB_ENV
+          echo "BUILD_TARGET=clip_executor" >> $GITHUB_ENV
+
+          CAS_VERSION=$(sed -n '/^__version__/p' ./server/clip_server/__init__.py | cut -d \' -f2)
+          V_CAS_VERSION=v${CAS_VERSION}
+          CAS_MINOR_VERSION=${CAS_VERSION%.*}
+          CAS_MAJOR_VERSION=${CAS_MINOR_VERSION%.*}
+          
+          ENGINE_TAG=${{matrix.engine_tag}}
+          if [ -n "${ENGINE_TAG}" ]; then
+            ENGINE_TAG=-${ENGINE_TAG//./}
+          fi
+
+          PIP_TAG=${{ matrix.pip_tag }}
+          if [ -n "${PIP_TAG}" ]; then
+              PIP_TAG=-${PIP_TAG}
+          fi
+
+          if [[ "${{ github.event.inputs.triggered_by }}" == "CD" ]]; then
+            # on every CD release
+            echo "TAG_ALIAS=\
+                            jinaai/clip_executor:master${PIP_TAG}${ENGINE_TAG}" \
+                            >> $GITHUB_ENV
+
+          elif [[ "${{ github.event.inputs.triggered_by }}" == "TAG" ]]; then
+            # on every tag release
+            echo "TAG_ALIAS=\
+                            jinaai/clip_executor:latest${PIP_TAG}${ENGINE_TAG}, \
+                            jinaai/clip_executor:${CAS_VERSION}${PIP_TAG}${ENGINE_TAG}, \
+                            jinaai/clip_executor:${CAS_MINOR_VERSION}${PIP_TAG}${ENGINE_TAG}, \
+                            jinaai/clip_executor:${CAS_MAJOR_VERSION}${PIP_TAG}${ENGINE_TAG} \
+                            " >> $GITHUB_ENV
+            
+          elif [[ "${{ github.event.inputs.triggered_by }}" == "MANUAL" ]]; then
+            # on every manual release
+            echo "TAG_ALIAS=\
+                            jinaai/clip_executor:${CAS_VERSION}${PIP_TAG}${ENGINE_TAG} \
+                            " >> $GITHUB_ENV
+          else
+            # test
+            echo "Debugging Workflow!"
+            echo "TAG_ALIAS=\
+                            jinaai/clip_executor:${CAS_VERSION}${PIP_TAG}${ENGINE_TAG} \
+                            " >> $GITHUB_ENV
+          fi
+
+          echo "CAS_VERSION=${CAS_VERSION}" >> $GITHUB_ENV
+
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@v2
+        with:
+          install: true
+      - name: Login to DockerHub
+        uses: docker/login-action@v2
+        with:
+          username: ${{ secrets.DOCKERHUB_DEVBOT_USER }}
+          password: ${{ secrets.DOCKERHUB_DEVBOT_TOKEN }}
+      - run: |
+          # https://github.com/docker/buildx/issues/464#issuecomment-741507760
+          # https://github.com/kubernetes-sigs/azuredisk-csi-driver/pull/808/files
+          docker run --privileged --rm tonistiigi/binfmt --uninstall qemu-aarch64
+          docker run --rm --privileged tonistiigi/binfmt --install all
+      - name: CPU Build and push
+        uses: docker/build-push-action@v2
+        if: ${{ matrix.engine_tag }} == ''
+        with:
+          context: .
+          file: Dockerfiles/base.Dockerfile
+          platforms: linux/amd64
+          push: true
+          tags: ${{env.TAG_ALIAS}}
+          build-args: |
+            BUILD_DATE=${{env.BUILD_DATE}}
+            CAS_VERSION=${{env.CAS_VERSION}}
+            VCS_REF=${{env.VCS_REF}}
+            PIP_TAG=${{matrix.pip_tag}}
+      - name: CUDA Build and push
+        uses: docker/build-push-action@v2
+        if: ${{ matrix.engine_tag }} == 'cuda'
+        with:
+          context: .
+          file: Dockerfiles/cuda.Dockerfile
+          platforms: linux/amd64
+          push: true
+          tags: ${{env.TAG_ALIAS}}
+          build-args: |
+            BUILD_DATE=${{env.BUILD_DATE}}
+            CAS_VERSION=${{env.CAS_VERSION}}
+            VCS_REF=${{env.VCS_REF}}
+            PIP_TAG=${{matrix.pip_tag}}
diff --git a/Dockerfiles/base.Dockerfile b/Dockerfiles/base.Dockerfile
@@ -0,0 +1,40 @@
+# !!! An ARG declared before a FROM is outside of a build stage, so it can’t be used in any instruction after a FROM
+ARG JINA_VERSION=3.3.25
+
+FROM jinaai/jina:${JINA_VERSION}-py38-standard
+
+ARG PIP_TAG
+
+# constant, wont invalidate cache
+LABEL org.opencontainers.image.vendor="Jina AI Limited" \
+      org.opencontainers.image.licenses="Apache 2.0" \
+      org.opencontainers.image.title="Clip-As-Service" \
+      org.opencontainers.image.description="Embed images and sentences into fixed-length vectors with CLIP" \
+      org.opencontainers.image.authors="hello@jina.ai" \
+      org.opencontainers.image.url="clip-as-service" \
+      org.opencontainers.image.documentation="https://clip-as-service.jina.ai/"
+
+RUN pip3 install --no-cache-dir torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu
+
+# copy will almost always invalid the cache
+COPY . /clip-as-service/
+
+RUN echo '\
+jtype: CLIPEncoder\n\
+with:\n\
+  name: ${{ env.MODEL_NAME }}\n\
+  device: ${{ env.DEVICE }}\n\
+  minibatch_size: ${{ env.MINIBATCH_SIZE }}\n\
+metas:\n\
+  py_modules:\n\
+    - server/clip_server/executors/clip_${{ env.ENGINE }}.py\n\
+' > /tmp/config.yml
+
+RUN cd /clip-as-service && \
+    if [ -n "$PIP_TAG" ]; then pip3 install --no-cache-dir server/"[$PIP_TAG]" ; fi && \
+    pip3 install --no-cache-dir "server/"
+
+WORKDIR /clip-as-service
+
+
+ENTRYPOINT ["jina", "executor", "--uses", "/tmp/config.yml"]
diff --git a/Dockerfiles/cuda.Dockerfile b/Dockerfiles/cuda.Dockerfile
@@ -0,0 +1,44 @@
+ARG CUDA_VERSION=11.4.2
+
+FROM nvcr.io/nvidia/cuda:${CUDA_VERSION}-cudnn8-runtime-ubuntu20.04
+ENV DEBIAN_FRONTEND=noninteractive
+
+ARG JINA_VERSION=3.3.25
+ARG PIP_TAG
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    python3-setuptools python3-wheel python3-pip \
+    && apt-get clean && rm -rf /var/lib/apt/lists/*;
+
+RUN python3 -m pip install --default-timeout=1000 --no-cache-dir torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113
+RUN python3 -m pip install --default-timeout=1000 --no-cache-dir "jina[standard]==${JINA_VERSION}"
+
+RUN python3 -m pip install nvidia-pyindex
+
+# copy will almost always invalid the cache
+COPY . /clip-as-service/
+
+
+RUN echo '\
+jtype: CLIPEncoder\n\
+with:\n\
+  name: ${{ env.MODEL_NAME }}\n\
+  device: ${{ env.DEVICE }}\n\
+  minibatch_size: ${{ env.MINIBATCH_SIZE }}\n\
+metas:\n\
+  py_modules:\n\
+    - server/clip_server/executors/clip_${{ env.ENGINE }}.py\n\
+' > /tmp/config.yml
+
+RUN cd /clip-as-service && \
+    if [ -n "${PIP_TAG}" ]; then python3 -m pip install --no-cache-dir server/"[${PIP_TAG}]" ; fi && \
+    python3 -m pip install --no-cache-dir "server/"
+
+WORKDIR /clip-as-service
+
+ENTRYPOINT ["jina", "executor", "--uses", "/tmp/config.yml"]
+
+
+
+
+
diff --git a/Dockerfiles/tensorrt.Dockerfile b/Dockerfiles/tensorrt.Dockerfile
@@ -0,0 +1,34 @@
+# Dockerfile to run Clip-as-Service with TensorRT, CUDA integration
+
+ARG TENSORRT_VERSION=22.04
+
+FROM nvcr.io/nvidia/tensorrt:${TENSORRT_VERSION}-py3
+
+ARG JINA_VERSION=3.3.25
+ARG PIP_VERSION
+
+RUN pip3 install --default-timeout=1000 --no-cache-dir torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113
+RUN pip3 -m pip install --default-timeout=1000 --no-cache-dir "jina[standard]==${JINA_VERSION}"
+
+
+# copy will almost always invalid the cache
+COPY . /clip-as-service/
+
+RUN pip3 install --no-cache-dir "server/[tensorrt]"
+
+RUN echo '\
+jtype: CLIPEncoder\n\
+with:\n\
+  name: ${{ env.MODEL_NAME }}\n\
+  device: cuda\n\
+  minibatch_size: ${{ env.MINIBATCH_SIZE }}\n\
+metas:\n\
+  py_modules:\n\
+    - server/clip_server/executors/clip_${{ env.ENGINE }}.py\n\
+' > /tmp/config.yml
+
+
+WORKDIR /clip-as-service
+
+ENTRYPOINT ["jina", "executor", "--uses", "/tmp/config.yml"]
+
diff --git a/docs/user-guides/server.md b/docs/user-guides/server.md
@@ -132,7 +132,7 @@ executors:
       jtype: CLIPEncoder
       metas:
         py_modules:
-          - executors/clip_trt.py
+          - executors/clip_tensorrt.py
 ```
 ````
 

diff --git a/server/clip_server/executors/clip_trt.py → ...er/clip_server/executors/clip_tensorrt.py b/server/clip_server/executors/clip_trt.py → ...er/clip_server/executors/clip_tensorrt.py
diff --git a/server/clip_server/tensorrt-flow.yml b/server/clip_server/tensorrt-flow.yml
@@ -8,5 +8,5 @@ executors:
       jtype: CLIPEncoder
       metas:
         py_modules:
-          - executors/clip_trt.py
+          - executors/clip_tensorrt.py
 
diff --git a/tests/conftest.py b/tests/conftest.py
@@ -39,7 +39,7 @@ def make_torch_flow(port_generator, request):
 
 @pytest.fixture(scope='session', params=['torch'])
 def make_trt_flow(port_generator, request):
-    from clip_server.executors.clip_trt import CLIPEncoder
+    from clip_server.executors.clip_tensorrt import CLIPEncoder
 
     f = Flow(port=port_generator()).add(name=request.param, uses=CLIPEncoder)
     with f:
-Original file line number
+Diff line change
@@ -0,0 +1,4 @@
+    .git
+    .github
+    scripts
+    docs