Create the new aio-local-ai-cuda12 image based on nvidia-cuda12 locai-ai docker image

.circleci/config.yml

@@ -0,0 +1,62 @@
+version: 2.1
+
+jobs:
+  build-and-push:
+    parameters:
+      tag-name:
+        type: string
+        default: "v2"
+    machine:
+      image: ubuntu-2404:2024.08.1
+      docker_layer_caching: true
+    resource_class: medium
+    environment:
+      DOCKER_BUILDKIT: 1
+    steps:
+      - checkout
+
+      - run:
+          name: Show disk usage (1st time)
+          command: |
+            df -h
+
+      - run:
+          name: Pull base GPU image (from Dockerfile.cuda12)
+          command: |
+            BASE_GPU_IMAGE=$(awk '/^FROM / {print $2}' Dockerfile.cuda12)
+            echo "Base GPU image found: $BASE_GPU_IMAGE"
+            echo "Pulling Docker image: $BASE_GPU_IMAGE"
+            docker pull "$BASE_GPU_IMAGE"
+
+      - run:
+          name: Show disk usage (2nd time)
+          command: |
+            df -h
+
+      - run:
+          name: Build Docker image (no-cache)
+          command: |
+            echo "Building Docker image with tag: << parameters.tag-name >>"
+            docker build \
+              --platform linux/amd64 \
+              -f Dockerfile.cuda12 \
+              -t "luzfcb/aio-local-ai-cuda12:<< parameters.tag-name >>" \
+              .
+      - run:
+          name: Show disk usage (3nd time)
+          command: |
+            df -h
+      - run:
+          name: Push Docker image to Docker Hub
+          command: |
+            # Assumes you've set DOCKERHUB_USERNAME and DOCKERHUB_PASSWORD as environment variables in CircleCI
+            echo "$DOCKERHUB_PASSWORD" | docker login -u "$DOCKERHUB_USERNAME" --password-stdin
+            echo "Pushing image: luzfcb/aio-local-ai-cuda12:<< parameters.tag-name >>"
+            docker push "luzfcb/aio-local-ai-cuda12:<< parameters.tag-name >>"
+
+
+workflows:
+  build-and-push-workflow:
+    jobs:
+      - build-and-push:
+          tag-name: "v2"  # OPTIONAL: override default if desired

.github/workflows/docker-build.yml

@@ -7,12 +7,23 @@ on:
         description: "Tag name"
         required: true
         default: 'v2'
+      buildCPU:
+        description: "Build and push CPU image"
+        type: boolean
+        required: true
+        default: true
+      buildGPU:
+        description: "Build and push GPU image"
+        type: boolean
+        required: true
+        default: true
 
 jobs:
-  push_to_registry:
+  push_cpu_image_to_registry:
     runs-on: ubuntu-latest
+    if: ${{ github.event.inputs.buildCPU == 'true' }}
 
-    name: Build docker image and push to dockerhub
+    name: Build docker CPU image and push to dockerhub
 
     steps:
       - name: Release space from worker
@@ -64,12 +75,165 @@ jobs:
           password: ${{ secrets.DOCKERHUB_TOKEN }}
 
       - name: Build container image
-        uses: docker/build-push-action@v5
+        uses: docker/build-push-action@v6
         with:
           push: true
           platforms: linux/amd64,linux/arm64
           context: './'
           no-cache: true
           file: 'Dockerfile'
-          tags: szaimen/aio-local-ai:${{ github.event.inputs.tagName }}
+          tags: luzfcb/aio-local-ai:${{ github.event.inputs.tagName }}
+          provenance: false
+
+  push_gpu_image_to_registry:
+    runs-on: ubuntu-22.04
+    environment: "DockerHubAccess"
+    if: ${{ github.event.inputs.buildGPU == 'true' }}
+
+    name: Build docker GPU image and push to dockerhub
+
+    steps:
+      - name: Available storage before
+        run: df -h
+
+
+      - name: Maximize build space
+        uses: AdityaGarg8/[email protected]
+        with:
+          remove-dotnet: 'true'
+          remove-android: 'true'
+          remove-haskell: 'true'
+          remove-codeql: 'true'
+          remove-docker-images: 'true'
+          remove-large-packages: 'true'
+          remove-cached-tools: 'true'
+          remove-swapfile: 'true'
+      - name: Available storage after
+        run: df -h
+
+
+      - name: Release space from worker
+        run: |
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 50 <<< "${pkgs}"
+          echo "Available storage before:"
+          sudo df -h
+          echo
+          sudo apt-get remove -y '^llvm-.*|^libllvm.*' || true
+          sudo apt-get remove --auto-remove android-sdk-platform-tools || true
+          sudo apt-get purge --auto-remove android-sdk-platform-tools || true
+          sudo rm -rf /usr/local/lib/android
+          sudo apt-get purge -y '^dotnet-.*|^aspnetcore-.*' || true
+          sudo apt-get purge -y '^temurin-.*|^ca-certificates-java.*' || true
+          sudo apt-get purge -y '^libclang-.*' || true
+          sudo rm -rf /usr/share/dotnet
+          sudo apt-get purge -y '^mono-.*' || true
+          sudo apt-get purge -y '^ghc-.*' || true
+          sudo apt-get purge -y '.*jdk.*|.*jre.*' || true
+          sudo apt-get purge -y 'php.*' || true
+          sudo apt-get purge -y hhvm powershell firefox monodoc-manual msbuild || true
+          sudo apt-get purge -y '^google-.*' || true
+          sudo apt-get purge -y azure-cli || true
+          sudo apt-get purge -y '^mongo.*-.*|^postgresql-.*|^mysql-.*|^mssql-.*' || true
+          sudo apt-get purge -y '^gfortran-.*' || true
+          sudo apt-get purge -y '^microsoft-edge-stable.*|^firefox.*|powershell.*' || true
+          sudo apt-get autoremove -y
+          sudo apt-get clean
+          sudo rm -rf /usr/share/dotnet || true
+          sudo rm -rf /opt/hostedtoolcache/CodeQL || true
+          echo
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 50 <<< "${pkgs}"
+          echo
+          sudo rm -rfv build || true
+          sudo docker system prune --all --volumes --force || true
+          echo "Available storage after:"
+          df -h
+
+      - name: Available storage after
+        run: df -h
+
+      - name: Check out the repo
+        uses: actions/checkout@v4
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
+      - name: Available storage before build
+        run: df -h
+
+      - name: Extract Base GPU Docker Image name
+        id: extract-image
+        run: |
+          BASE_GPU_IMAGE=$(awk '/^FROM / {print $2}' Dockerfile.cuda12)
+          echo "Base GPU image found: $BASE_GPU_IMAGE"
+          echo "BASE_GPU_IMAGE=$BASE_GPU_IMAGE" >> $GITHUB_ENV
+
+      - name: Pull Base Docker Image
+        run: |
+          echo "Pulling Docker image: $BASE_GPU_IMAGE"
+          docker pull "$BASE_GPU_IMAGE"
+        env:
+          BASE_GPU_IMAGE: ${{ env.BASE_GPU_IMAGE }}
+
+      - name: Create a TMPDIR for docker build on a different disk
+        run: |
+          mkdir -p /tmp/docker-build
+          export TMPDIR=/tmp/docker-build
+          echo "TMPDIR=$TMPDIR" >> $GITHUB_ENV
+
+      - name: Available storage before build
+        run: df -h
+
+      - name: Verify Docker image size
+        uses: luzfcb/docker-image-size-limit@master
+        with:
+          image: "${{ env.BASE_GPU_IMAGE }}"
+          size: "47GB"
+          show_current_size: true
+          dry_run: true
+
+
+      - name: Build container image
+        uses: docker/build-push-action@v6
+        with:
+          push: false
+          platforms: linux/amd64
+          context: './'
+          file: 'Dockerfile.cuda12'
+          tags: luzfcb/aio-local-ai-cuda12:${{ github.event.inputs.tagName }}
+          provenance: false
+        env:
+          TMPDIR: ${{ env.TMPDIR }}
+
+      - name: Verify Docker image size
+        uses: luzfcb/docker-image-size-limit@master
+        with:
+          image: "luzfcb/aio-local-ai-cuda12:${{ github.event.inputs.tagName }}"
+          size: "47GB"
+          show_current_size: true
+          dry_run: true
+
+      - name: Available storage after build
+        run: df -h
+
+      - name: Push container image
+        uses: docker/build-push-action@v6
+        with:
+          push: true
+          platforms: linux/amd64
+          context: './'
+          file: 'Dockerfile.cuda12'
+          tags: luzfcb/aio-local-ai-cuda12:${{ github.event.inputs.tagName }}
           provenance: false

.github/workflows/docker-lint.yml

@@ -26,7 +26,7 @@ jobs:
 
       - name: run lint
         run: |
-          DOCKERFILES="$(find ./ -name Dockerfile)"
+          DOCKERFILES="$(find ./ -name 'Dockerfile*')"
           mapfile -t DOCKERFILES <<< "$DOCKERFILES"
           for file in "${DOCKERFILES[@]}"; do
             # DL4006 warning: Set the SHELL option -o pipefail before RUN with a pipe in it. If you are using /bin/sh in an alpine image or if your shell is symlinked to busybox then consider explicitly setting your SHELL to /bin/ash, or disable this check

.gitignore

@@ -0,0 +1,2 @@
+.idea
+

Dockerfile

@@ -1,4 +1,4 @@
-# From https://github.com/go-skynet/LocalAI/blob/master/Dockerfile
+# From https://github.com/mudler/LocalAI/blob/master/Dockerfile
 FROM quay.io/go-skynet/local-ai:v2.24.2-aio-cpu
 
 COPY --chmod=775 start.sh /start.sh

Dockerfile.cuda12

@@ -0,0 +1,18 @@
+# From https://github.com/mudler/LocalAI/blob/master/Dockerfile
+FROM quay.io/go-skynet/local-ai:v2.24.2-aio-gpu-nvidia-cuda-12
+
+COPY --chmod=775 start.sh /start.sh
+
+# hadolint ignore=DL3008
+RUN set -ex; \
+    \
+    export DEBIAN_FRONTEND=noninteractive; \
+    apt-get update; \
+    apt-get install -y --no-install-recommends \
+        netcat-openbsd \
+    ; \
+    rm -rf /var/lib/apt/lists/*
+
+# hadolint ignore=DL3002
+USER root
+ENTRYPOINT [ "/start.sh" ]

docker-compose.yml

@@ -0,0 +1,22 @@
+# a quick way to build and run the docker image
+services:
+  cpu:
+    build:
+        context: .
+        dockerfile: Dockerfile
+    image: szaimen/aio-local-ai:v2
+
+  cuda12:
+    build:
+      platforms:
+        - linux/amd64
+      context: .
+      dockerfile: Dockerfile.cuda12
+    image: luzfcb/aio-local-ai-cuda12:v2
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: 1
+              capabilities: [gpu]

start.sh

@@ -19,6 +19,33 @@ else
     echo "CPU: no AVX512 found"
 fi
 
+echo -e "\nGPU info:"
+echo    "    Note: Detecting a GPU does not guarantee compatibility with Local-AI. "
+echo -e "    For more details on supported GPUs and requirements, please refer to the documentation.\n"
+
+# Filter only GPU-related devices using device class codes (0300, 0302, 0380)
+if command -v lspci &>/dev/null; then
+    echo "Detected GPUs:"
+    lspci -nn | awk '/\[03(00|02|80)\]/'
+else
+    echo "    'lspci' not available to detect GPUs."
+fi
+
+
+if command -v nvidia-smi &>/dev/null; then
+    echo -e "\nDetected NVIDIA GPUs:"
+    nvidia-smi
+else
+    echo -e "\n    NVIDIA GPU not detected or 'nvidia-smi' is not installed."
+    echo      "    To enable NVIDIA GPU support, please ensure the following:"
+    echo      "      1. Install the Nvidia Container Toolkit on the Docker host machine."
+    echo      "         Refer to the official Nvidia Container Toolkit documentation for installation instructions."
+    echo      "      2. Use the 'aio-local-ai-cuda12' Docker image, specifically designed for Nvidia GPU support."
+    echo      "      3. Run the container with the '--gpus all' flag to allocate GPU resources."
+    echo -e "    For more details, please consult the Nextcloud Local-AI community container documentation.\n"
+fi
+
+
 while ! nc -z nextcloud-aio-nextcloud 9001; do
     echo "Waiting for nextcloud to start"
     sleep 5
@@ -33,6 +60,14 @@ done
 set -x
 THREADS="$(nproc)"
 export THREADS
+
+# enable GPU
+LOCALAI_F16=true
+export LOCALAI_F16
+
+
 set +x
 
+
+
 ./local-ai --preload-models-config "/nextcloud/admin/files/nextcloud-aio-local-ai/models.yaml"