Enable opensource images multi-platform building

.github/workflows/docker-image.yml

@@ -8,39 +8,45 @@ on:
 jobs:
   build:
     runs-on: ubuntu-latest
+    env:
+      BUILDX_NO_DEFAULT_ATTESTATIONS: 1
 
     steps:
     - name: Checkout repository
       uses: actions/checkout@v4
 
+    - name: Set up QEMU
+      uses: docker/setup-qemu-action@v3
+
     - name: Set up Docker Buildx
-      uses: docker/setup-buildx-action@v2
+      uses: docker/setup-buildx-action@v3
 
     - name: Log in to Docker Hub
-      uses: docker/login-action@v2
+      uses: docker/login-action@v3
       with:
         username: ${{ secrets.DOCKER_USERNAME }}
         password: ${{ secrets.DOCKER_PASSWORD }}
 
     - name: Log in to ACR
-      uses: docker/login-action@v2
+      uses: docker/login-action@v3
       with:
         username: ${{ secrets.ACR_USERNAME }}
         password: ${{ secrets.ACR_PASSWORD }}
         registry: ${{ secrets.ACR_REGISTRY }}
 
     - name: Extract Docker metadata (tags, labels)
       id: meta
-      uses: docker/metadata-action@v4
+      uses: docker/metadata-action@v5
       with:
         images: |
           opencsg/csghub-server
           ${{ secrets.ACR_REGISTRY }}/opencsg_public/csghub_server
 
     - name: Build and push Docker image
-      uses: docker/build-push-action@v4
+      uses: docker/build-push-action@v6
       with:
         context: .
         file: ./docker/Dockerfile
         push: true
         tags: ${{ steps.meta.outputs.tags }}
+        platforms: linux/amd64,linux/arm64

docker/Dockerfile

@@ -1,12 +1,12 @@
-FROM opencsg-registry.cn-beijing.cr.aliyuncs.com/public/starhub-server:base-1.0 AS build-env
+FROM opencsg-registry.cn-beijing.cr.aliyuncs.com/opencsg_public/csghub_server:base-1.0 AS build-env
 ENV GOPROXY=https://goproxy.cn,direct
 WORKDIR /starhub
 COPY . .
 RUN  CGO_ENABLED=1 GOOS=linux go build  -v -o /go/bin/starhub ./cmd/csghub-server && \
      rm -rf /go/pkg && \
      rm -rf /starhub
 
-FROM opencsg-registry.cn-beijing.cr.aliyuncs.com/public/starhub-server:base-1.0    
+FROM opencsg-registry.cn-beijing.cr.aliyuncs.com/opencsg_public/csghub_server:base-1.0
 WORKDIR /starhub-bin
 ENV GIN_MODE=release
 COPY --from=build-env /go/bin/starhub .

docker/Dockerfile-base

@@ -7,7 +7,7 @@ RUN mkdir -p ~/.duckdb/extensions/v0.9.2/linux_amd64 && \
     wget https://opencsg-public-resource.oss-cn-beijing.aliyuncs.com/tools/duckdb/httpfs.duckdb_extension -O ~/.duckdb/extensions/v0.9.2/linux_amd64/httpfs.duckdb_extension && \
     chmod 755 ~/.duckdb/extensions/v0.9.2/linux_amd64/httpfs.duckdb_extension
 
-RUN sed -i "s@/deb.debian.org@/mirrors.cloud.aliyuncs.com@g" /etc/apt/sources.list && \
+RUN sed -i "s@/deb.debian.org@/mirrors.tuna.tsinghua.edu.cn@g" /etc/apt/sources.list && \
     apt-get update && apt-get install -y cron && rm -rf /var/lib/apt/lists/*
 
 # docker build -t opencsg-registry.cn-beijing.cr.aliyuncs.com/public/starhub-server:base-1.0 -f Dockerfile-base .

docker/README.md

@@ -0,0 +1,22 @@
+# CSGHUB Server Base Images Building 
+
+## Login Container Registry
+```bash
+OPENCSG_ACR="opencsg-registry.cn-beijing.cr.aliyuncs.com"
+OPENCSG_ACR_USERNAME=""
+OPENCSG_ACR_PASSWORD=""
+echo "$OPENCSG_ACR_PASSWORD" | docker login $OPENCSG_ACR -u $OPENCSG_ACR_USERNAME --password-stdin
+```
+
+## Build Multi-Platform Images
+```bash
+export BUILDX_NO_DEFAULT_ATTESTATIONS=1
+export IMAGE_TAG=1.0
+docker buildx build --platform linux/amd64,linux/arm64 \
+  -t ${OPENCSG_ACR}/opencsg_public/csghub_server:base-${IMAGE_TAG} \
+  -t ${OPENCSG_ACR}/opencsg_public/csghub_server:base-latest \
+  -f Dockerfile.nginx \
+  --push .
+```
+*The above command will create `linux/amd64` and `linux/arm64` images with the tags `base-${IMAGE_TAG}` and `base-latest` at the same time.*
+

docker/finetune/Dockerfile.llamafactory

@@ -1,55 +1,57 @@
-# pull from devel image instead of base
+# Pull from devel image instead of base
 FROM nvidia/cuda:12.1.1-devel-ubuntu22.04
+
 # Set bash as the default shell
-ENV SHELL=/bin/bash
-ENV JUPYTERHUB_SERVICE_PREFIX=/proxy/
-ENV GRADIO_ROOT_PATH=/proxy/7860/
-ENV TZ=Asia/Shanghai
-ENV NCCL_IB_DISABLE=1 NCCL_P2P_DISABLE=1
-ENV HF_HOME=/workspace/.cache
+ENV SHELL=/bin/bash \
+    JUPYTERHUB_SERVICE_PREFIX=/proxy/ \
+    GRADIO_ROOT_PATH=/proxy/7860/ \
+    TZ=Asia/Shanghai \
+    NCCL_IB_DISABLE=1 NCCL_P2P_DISABLE=1 \
+    HF_HOME=/workspace/.cache \
+    DEBIAN_FRONTEND=noninteractive
 
 # Build with some basic utilities
-RUN apt-get update && apt-get install -y \
-    python3-pip apt-utils \
-    wget curl vim \
-    git git-lfs \
-    supervisor \
-    unzip
-# set timezone
-ARG DEBIAN_FRONTEND=noninteractive
-RUN apt-get install -y tzdata \
-    && ln -snf /usr/share/zoneinfo/$TZ /etc/localtime \
-    && echo $TZ > /etc/timezone \
-    && dpkg-reconfigure -f noninteractive tzdata
-
-# alias python='python3'
-RUN ln -s /usr/bin/python3 /usr/bin/python
-RUN pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    python3-pip apt-utils wget curl vim \
+    git git-lfs supervisor unzip tzdata && \
+    apt-get clean && rm -rf /var/lib/apt/lists/*
+
+# Set timezone
+RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && \
+    echo $TZ > /etc/timezone && \
+    dpkg-reconfigure -f noninteractive tzdata
 
 # Install the appropriate torch version 
-#RUN pip install torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 --index-url https://download.pytorch.org/whl/cu121
-RUN pip install --no-cache-dir jupyterlab numpy==1.26.4 \
+RUN ln -sf /usr/bin/python3 /usr/bin/python && \
+    pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple && \
+    pip install --no-cache-dir jupyterlab numpy==1.26.4 \
     torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 \
-    jupyter-server-proxy==4.2.0 
+    jupyter-server-proxy==4.2.0
+
 # Create a working directory
 WORKDIR /etc/csghub
-RUN git clone https://github.com/hiyouga/LLaMA-Factory.git --branch v0.8.3 --single-branch
-RUN cd LLaMA-Factory && pip install --no-cache-dir -e ".[metrics,deepspeed]"
-# setup supervisord
-RUN mkdir -p /var/log/supervisord
+
+RUN git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git --branch v0.8.3 --single-branch && cd LLaMA-Factory && \
+    pip install --no-cache-dir -e ".[metrics,deepspeed]"
+
+# Setup supervisord
 COPY script/supervisord.conf /etc/supervisor/conf.d/supervisord.conf
 COPY script/jupyter_notebook_config.py /root/.jupyter/jupyter_notebook_config.py
 COPY script/ /etc/csghub/
 COPY script/handlers.py /usr/local/lib/python3.10/dist-packages/jupyter_server_proxy/handlers.py
-RUN chmod +x /etc/csghub/*.sh
-#use dark mode
-RUN mkdir -p /root/.jupyter/lab/user-settings/@jupyterlab/apputils-extension && \
-	    echo '{"theme":"JupyterLab Dark"}' > /root/.jupyter/lab/user-settings/@jupyterlab/apputils-extension/themes.jupyterlab-settings && \
-	    mkdir -p /root/.jupyter/lab/user-settings/@jupyterlab/notebook-extension && \
-	    echo '{"codeCellConfig":{"lineNumbers":true }}' >   /root/.jupyter/lab/user-settings/@jupyterlab/notebook-extension/tracker.jupyterlab-settings
-#fix gradio proxy issue
-RUN pip uninstall gradio && pip install https://git-devops.opencsg.com/opensource/gradio/-/raw/3a207a08755b4820541915e9ea63e6abc1b4b424/gradio-4.41.0-py3-none-any.whl
+
+RUN mkdir -p /var/log/supervisord && \
+    chmod +x /etc/csghub/*.sh && \
+    mkdir -p /root/.jupyter/lab/user-settings/@jupyterlab/apputils-extension && \
+	echo '{"theme":"JupyterLab Dark"}' > /root/.jupyter/lab/user-settings/@jupyterlab/apputils-extension/themes.jupyterlab-settings && \
+	mkdir -p /root/.jupyter/lab/user-settings/@jupyterlab/notebook-extension && \
+	echo '{"codeCellConfig":{"lineNumbers":true }}' > /root/.jupyter/lab/user-settings/@jupyterlab/notebook-extension/tracker.jupyterlab-settings
+
+# Fix gradio proxy issue
+RUN pip uninstall -y gradio && \
+    pip install --no-cache-dir https://git-devops.opencsg.com/opensource/gradio/-/raw/3a207a08755b4820541915e9ea63e6abc1b4b424/gradio-4.41.0-py3-none-any.whl
+
 # Create a working directory
 WORKDIR /workspace/
-ENTRYPOINT ["/usr/bin/supervisord", "-c", "/etc/supervisor/conf.d/supervisord.conf"]
-EXPOSE 8000
+EXPOSE 8000
+ENTRYPOINT ["/usr/bin/supervisord", "-c", "/etc/supervisor/conf.d/supervisord.conf"]

docker/finetune/README.md

@@ -1,30 +1,38 @@
-# CSGHUB finetune images
+# CSGHUB Finetune Images Building
 
-## base image
-https://docs.nvidia.com/deeplearning/frameworks/support-matrix/index.html
-https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch/tags
+## Base Images
+- https://docs.nvidia.com/deeplearning/frameworks/support-matrix/index.html
+- https://catalog.ngc.nvidia.com/orgs/nvidia/containers/pytorch/tags
 
-## build images
+## Login Container Registry
 ```bash
-docker build -f Dockerfile.llamafactory .
+OPENCSG_ACR="opencsg-registry.cn-beijing.cr.aliyuncs.com"
+OPENCSG_ACR_USERNAME=""
+OPENCSG_ACR_PASSWORD=""
+echo "$OPENCSG_ACR_PASSWORD" | docker login $OPENCSG_ACR -u $OPENCSG_ACR_USERNAME --password-stdin
 ```
 
-## push images
-```
-docker login opencsg-registry.cn-beijing.cr.aliyuncs.com
-docker push xxx
-```
-## latest images
-```
-#for llama-factory image
-opencsg-registry.cn-beijing.cr.aliyuncs.com/public/llama-factory:1.20-cuda12.1-devel-ubuntu22.04-py310-torch2.1.2
-```
-## Run image locally
+## Build Multi-Platform Images
+```bash
+export BUILDX_NO_DEFAULT_ATTESTATIONS=1
+export IMAGE_TAG=1.21-cuda12.1-devel-ubuntu22.04-py310-torch2.1.2
+docker buildx build --platform linux/amd64,linux/arm64 \
+  -t ${OPENCSG_ACR}/public/llama-factory:${IMAGE_TAG} \
+  -t ${OPENCSG_ACR}/public/llama-factory:latest \
+  -f Dockerfile.llamafactory \
+  --push .
 ```
+*Note: The above command will create `linux/amd64` and `linux/arm64` images with the tags `${IMAGE_TAG}` and `latest` at the same time.*
 
-docker run -d -e ACCESS_TOKEN=xxx -e REPO_ID="OpenCSG/csg-wukong-1B"  -e HF_ENDPOINT=https://hub.opencsg.com/hf  opencsg-registry.cn-beijing.cr.aliyuncs.com/public/llama-factory:1.20-cuda12.1-devel-ubuntu22.04-py310-torch2.1.2
-
+## Run Finetune Image Locally
+```bash
+docker run -d \
+  -e ACCESS_TOKEN=xxx \
+  -e REPO_ID="OpenCSG/csg-wukong-1B" \
+  -e HF_ENDPOINT=https://opencsg.com/hf \
+  -p 8000:8000 \
+  ${OPENCSG_ACR}/public/llama-factory:${IMAGE_TAG}
 ```
-Note: HF_ENDPOINT should be use the real csghub address
+*Note: HF_ENDPOINT should be use the real csghub address.*
 
 

docker/finetune/build.sh

@@ -0,0 +1,47 @@
+#!/usr/bin/env bash
+
+if [ -z "$1" ] || [ -z "$2" ] || [ -z "$3" ]; then
+    echo "Usage: $0 <OPENCSG_ACR_USERNAME> <OPENCSG_ACR_PASSWORD> <IMAGE_TAG>"
+    echo "Tag example: 1.22-cuda12.1-devel-ubuntu22.04-py310-torch2.1.2"
+    exit 1
+fi
+
+OS=$(uname -s)
+echo "Enable docker buildx with QEMU for ${OS}"
+if [ "$OS" = "Darwin" ]; then
+    echo "QEMU enabled default..."
+elif [ "$OS" = "Linux" ]; then
+    echo "Install QEMU support..."
+    docker run --privileged --rm tonistiigi/binfmt --install all
+else
+    echo "Unknown OS: $OS"
+fi
+
+export DOCKER_BUILDKIT=1
+export BUILDX_NO_DEFAULT_ATTESTATIONS=1
+DOCKER_CONTAINERS=$(docker buildx ls | grep docker-container)
+if [[ ! -z "$DOCKER_CONTAINERS" ]]; then
+    BUILDER=$(echo "$DOCKER_CONTAINERS" | awk 'NR==1{gsub(/\*$/, "", $1); print $1}')
+    docker buildx use ${BUILDER}
+else
+    docker buildx create --name container-builder --driver docker-container --use --bootstrap
+fi
+
+OPENCSG_ACR_USERNAME=$1
+OPENCSG_ACR_PASSWORD=$2
+OPENCSG_ACR=${OPENCSG_ACR:-"opencsg-registry.cn-beijing.cr.aliyuncs.com"}
+OPENCSG_ACR_NAMESPACE=${OPENCSG_ACR_NAMESPACE:-"public"}
+DOCKER_IMAGE_PREFIX="$OPENCSG_ACR/$OPENCSG_ACR_NAMESPACE"
+
+echo "Logging in to OpenCSG ACR..."
+echo "$OPENCSG_ACR_PASSWORD" | docker login "$OPENCSG_ACR" -u "$OPENCSG_ACR_USERNAME" --password-stdin
+
+echo "Building images..."
+export IMAGE_TAG=$3
+docker buildx build --platform linux/amd64,linux/arm64 \
+    -t ${DOCKER_IMAGE_PREFIX}/llama-factory:${IMAGE_TAG} \
+    -t ${DOCKER_IMAGE_PREFIX}/llama-factory:latest \
+    -f Dockerfile.llamafactory \
+    --push .
+
+echo "Done! New image pushed with tag: $NEW_TAG"