Merge pull request #740 from roboflow/feature/depot_ci_gpu

feature/depot ci gpu

Author: alexnorell

.github/workflows/docker.gpu.yml

@@ -4,51 +4,52 @@ on:
   release:
     types: [created]
   push:
-    branches: [ main ]
+    branches: [main]
+  pull_request:
+    branches: [main]
   workflow_dispatch:
     inputs:
       force_push:
         type: boolean
         description: "Do you want to push image after build?"
         default: false
+
 env:
-  VERSION: '0.0.0' # Default version, will be overwritten
+  VERSION: "0.0.0" # Default version, will be overwritten
 
 jobs:
   docker:
     runs-on:
-      group: group8core
+      labels: ubuntu-latest
     timeout-minutes: 120
+    permissions:
+      id-token: write
+      contents: read
     steps:
-      - name: Remove unnecessary files
-        run: |
-          sudo rm -rf /usr/share/dotnet
-          sudo rm -rf "$AGENT_TOOLSDIRECTORY"
-      -
-        name: Set up QEMU
-        uses: docker/setup-qemu-action@v2
-      -
-        name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v2
-      -
-        name: Login to Docker Hub
-        uses: docker/login-action@v2
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
-      - 
-        name: 🛎️ Checkout
-        uses: actions/checkout@v3
-      -
-        name: Read version from file
+      - name: 🛎️ Checkout
+        uses: actions/checkout@v4
+      - name: Read version from file
         run: echo "VERSION=$(DISABLE_VERSION_CHECK=true python ./inference/core/version.py)" >> $GITHUB_ENV
-      -
-        name: Build and Push
-        uses: docker/build-push-action@v4
+      - name: Set up Depot CLI
+        uses: depot/setup-action@v1
+      - name: Build and Push
+        uses: depot/build-push-action@v1
         with:
+          save: true
           push: ${{ github.event_name == 'release' || (github.event.inputs.force_push == 'true')}}
+          project: grl7ffzxd7
           tags: roboflow/roboflow-inference-server-gpu:latest,roboflow/roboflow-inference-server-gpu:${{env.VERSION}}
-          cache-from: type=registry,ref=roboflow/roboflow-inference-server-gpu:cache
-          cache-to: type=registry,ref=roboflow/roboflow-inference-server-gpu:cache,mode=max
           platforms: linux/amd64
-          file: ./docker/dockerfiles/Dockerfile.onnx.gpu
+          file: ./docker/dockerfiles/Dockerfile.onnx.gpu
+      - name: Store build ID
+        run: echo ${{ steps.build.outputs.build-id }} > build_id.txt
+      - name: Upload build ID
+        uses: actions/upload-artifact@v3
+        with:
+          name: build_id
+          path: build_id.txt

.github/workflows/test.nvidia_t4.yml

@@ -1,9 +1,10 @@
 name: Code Quality & Regression Tests - NVIDIA T4
 
 on:
-  push:
-    branches: [main]
-  workflow_dispatch:
+  workflow_run:
+    workflows: ["Build and Push GPU Container"] # The name of the first workflow
+    types:
+      - completed # Trigger this workflow when the first workflow completes
 
 jobs:
   build:
@@ -17,19 +18,39 @@ jobs:
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v2
       - name: 🛎️ Checkout
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           ref: ${{ github.head_ref }}
+      - name: 📦 Cache Python packages
+        uses: actions/cache@v3
+        with:
+          path: ~/.cache/pip
+          key: ${{ runner.os }}-pip-${{ matrix.python-version }}-${{ hashFiles('requirements/**') }}
+          restore-keys: |
+            ${{ runner.os }}-pip-${{ matrix.python-version }}-
+      - name: Read version from file
+        run: echo "VERSION=$(DISABLE_VERSION_CHECK=true python ./inference/core/version.py)" >> $GITHUB_ENV
       - name: 🦾 Install dependencies
         run: |
           python3 -m pip install --upgrade pip
           python3 -m pip install -r requirements/requirements.test.integration.txt
-      - name: �🔨 Build and Push Test Docker - GPU
-        run: |
-          docker build -t roboflow/roboflow-inference-server-gpu:test -f docker/dockerfiles/Dockerfile.onnx.gpu .
+      - name: Set up Depot CLI
+        uses: depot/setup-action@v1
+      - name: Retireve build ID
+        uses: actions/download-artifact@v3
+        with:
+          name: build_id
+          path: .
+      - name: Read build ID
+        run: echo "BUILD_ID=$(cat build_id/build_id.txt)" >> $GITHUB_ENV
+      - name: Set Repo
+        run: echo "INFERENCE_SERVER_REPO=roboflow-inference-server-gpu" >> $GITHUB_ENV
+      - name: Pull GPU Image
+        run: depot pull $BUILD_ID -t roboflow/${INFERENCE_SERVER_REPO}:${VERSION}
       - name: 🔋 Start Test Docker - GPU
         run: |
-          PORT=9101 INFERENCE_SERVER_REPO=roboflow-inference-server-gpu make start_test_docker_gpu
+          docker tag roboflow/${INFERENCE_SERVER_REPO}:${VERSION} roboflow/${INFERENCE_SERVER_REPO}:test
+          PORT=9101 make start_test_docker_gpu
       - name: 🧪 Regression Tests - GPU
         id: regression_tests
         run: |
@@ -85,10 +106,10 @@ jobs:
       - name: 🧪 Regression Tests - SAM2
         id: sam2_tests
         run: |
-           PORT=9101 API_KEY=${{ secrets.API_KEY }} SKIP_SAM2_TESTS=False python3 -m pytest tests/inference/integration_tests/test_sam2.py
+          PORT=9101 API_KEY=${{ secrets.API_KEY }} SKIP_SAM2_TESTS=False python3 -m pytest tests/inference/integration_tests/test_sam2.py
       - name: 🚨 Show server logs on error
         run: docker logs inference-test
         if: ${{ steps.sam2_tests.outcome != 'success' }}
       - name: 🧹 Cleanup Test Docker - SAM2
         run: make stop_test_docker
-        if: success() || failure()
+        if: success() || failure()

docker/dockerfiles/Dockerfile.onnx.gpu

@@ -51,9 +51,6 @@ RUN python3 -m pip install packaging==24.1  && rm -rf ~/.cache/pip
 # Install flash_attn required for Paligemma and Florence2
 RUN python3 -m pip install -r requirements.pali.flash_attn.txt --no-build-isolation && rm -rf ~/.cache/pip
 
-FROM scratch
-COPY --from=base / /
-
 WORKDIR /build
 COPY . .
 RUN ln -s /usr/bin/python3 /usr/bin/python
@@ -82,4 +79,4 @@ ENV ENABLE_STREAM_API=True
 ENV ENABLE_WORKFLOWS_PROFILING=True
 ENV ENABLE_PROMETHEUS=True
 
-ENTRYPOINT uvicorn gpu_http:app --workers $NUM_WORKERS --host $HOST --port $PORT
+ENTRYPOINT uvicorn gpu_http:app --workers $NUM_WORKERS --host $HOST --port $PORT