add: all vlm & all llm

.github/workflows/deploy.yml

@@ -0,0 +1,80 @@
+name: Build and Push Deploy Images
+
+on:
+  push:
+    branches: [ main, master ]
+    paths:
+      - 'src/**'
+      - 'docker/RK3576/VLM/deploy_llm.dockerfile'
+      - 'docker/RK3576/LLM/deploy_llm.dockerfile'
+      - 'docker/RK3588/VLM/deploy_vlm.dockerfile'
+      - 'docker/RK3588/LLM/deploy_llm.dockerfile'
+      - '.github/workflows/deploy.yml'
+      - 'lib/**'
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  packages: write
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        include:
+          - platform: rk3576
+            type: llm
+            dockerfile: docker/RK3576/LLM/deploy_llm.dockerfile
+          - platform: rk3576
+            type: vlm
+            dockerfile: docker/RK3576/VLM/deploy_vlm.dockerfile
+          - platform: rk3588
+            type: llm
+            dockerfile: docker/RK3588/LLM/deploy_llm.dockerfile
+          - platform: rk3588
+            type: vlm
+            dockerfile: docker/RK3588/VLM/deploy_vlm.dockerfile
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Set up QEMU (for ARM64 emulation)
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Log in to GHCR
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      # Generate image tags:
+      #   Image name: ghcr.io/<owner>/deploy_<type>
+      #   Tags: <platform>-latest, <platform>-<short-sha>
+      - name: Generate image tags
+        id: meta
+        run: |
+          OWNER_LOWER=$(echo "${{ github.repository_owner }}" | tr '[:upper:]' '[:lower:]')
+          IMAGE_NAME="ghcr.io/${OWNER_LOWER}/deploy_${{ matrix.type }}"
+          SHORT_SHA=$(echo "${{ github.sha }}" | cut -c1-7)
+          TAGS="${IMAGE_NAME}:${{ matrix.platform }}-latest"
+          TAGS="${TAGS},${IMAGE_NAME}:${{ matrix.platform }}-${SHORT_SHA}"
+          echo "tags=${TAGS}" >> $GITHUB_OUTPUT
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          file: ${{ matrix.dockerfile }}
+          platforms: linux/arm64
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: |
+            org.opencontainers.image.source=${{ github.server_url }}/${{ github.repository }}
+            org.opencontainers.image.revision=${{ github.sha }}
+            org.opencontainers.image.version=${{ github.ref_name }}

.gitignore

@@ -28,7 +28,7 @@ Thumbs.db
 *.db
 
 # Environment variables
-.env
+.env/
 .env.local
 
 # Jupyter

VLM.md

@@ -1,4 +1,4 @@
-# LLM Fast start
+# VLM Fast start
 
 ## Step 1: Install Docker
 

docker/RK3576/LLM/DeepSeek-R1-Distill-Qwen.dockerfile

@@ -34,4 +34,4 @@ ENV RKLLM_MODEL_PATH=/app/models/${MODEL_FILE}
 
 EXPOSE 8001
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3576"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3576.sh && python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3576"]

docker/RK3576/LLM/Qwen3.dockerfile

@@ -34,4 +34,4 @@ ENV RKLLM_MODEL_PATH=/app/models/${MODEL_FILE}
 
 EXPOSE 8001
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3576"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3576.sh && python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3576"]

docker/RK3576/LLM/deploy_llm.dockerfile

@@ -0,0 +1,25 @@
+# 第一阶段：环境准备
+FROM python:3.10-slim AS base
+
+RUN apt-get update && \
+    apt-get install -y wget curl git sudo libgomp1 && \
+    rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# 安装依赖
+COPY ./src/llm/requirements.txt /app/requirements.txt
+RUN pip install --no-cache-dir -r /app/requirements.txt
+
+# 拷贝运行时库和工具
+COPY ./lib/librkllmrt.so /usr/lib/librkllmrt.so
+COPY ./src/fix_freq_rk3576.sh /app/fix_freq_rk3576.sh
+RUN chmod +x /app/fix_freq_rk3576.sh
+
+# --- 最终阶段 ---
+FROM base AS final
+
+COPY ./src/llm/fastapi_server_llm.py /app/
+
+EXPOSE 8001
+

docker/RK3576/VLM/DeepseekOCR.dockerfile

@@ -48,4 +48,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3576.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3576/VLM/InternVL3.dockerfile

@@ -22,6 +22,7 @@ COPY ./lib/librknnrt.so /usr/lib/librknnrt.so
 COPY ./src/fix_freq_rk3576.sh /app/fix_freq_rk3576.sh
 RUN chmod +x /app/fix_freq_rk3576.sh
 
+
 # --- 最终阶段 ---
 FROM base AS final
 ARG LLM_URL
@@ -48,4 +49,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3576.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3576/VLM/Qwen2.5-VL.dockerfile

@@ -22,6 +22,7 @@ COPY ./lib/librknnrt.so /usr/lib/librknnrt.so
 COPY ./src/fix_freq_rk3576.sh /app/fix_freq_rk3576.sh
 RUN chmod +x /app/fix_freq_rk3576.sh
 
+
 # --- 最终阶段 ---
 FROM base AS final
 ARG LLM_URL
@@ -48,4 +49,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3576.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3576/VLM/Qwen3-VL.dockerfile

@@ -48,4 +48,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3576.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3576/VLM/deploy_vlm.dockerfile

@@ -0,0 +1,28 @@
+# 第一阶段：环境准备
+FROM python:3.10-slim AS base
+
+RUN apt-get update && \
+    apt-get install -y wget curl git sudo libgomp1 && \
+    rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# 安装依赖
+COPY ./src/vlm/requirements.txt /app/requirements.txt
+COPY ./src/vlm/fastapi_server_vlm.py /app/fastapi_server_vlm.py
+RUN pip install --no-cache-dir -r /app/requirements.txt
+
+# 拷贝运行时库和工具
+COPY ./lib/librkllmrt.so /usr/lib/librkllmrt.so 
+COPY ./lib/librkllm_service.so /usr/lib/librkllm_service.so 
+COPY ./lib/librknnrt.so /usr/lib/librknnrt.so
+
+
+COPY ./src/fix_freq_rk3576.sh /app/fix_freq_rk3576.sh
+RUN chmod +x /app/fix_freq_rk3576.sh
+
+# --- 最终阶段 ---
+FROM base AS final
+
+EXPOSE 8002
+

docker/RK3588/LLM/DeepSeek-R1-Distill-Qwen.dockerfile

@@ -34,4 +34,4 @@ ENV RKLLM_MODEL_PATH=/app/models/${MODEL_FILE}
 
 EXPOSE 8001
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3588"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3588"]

docker/RK3588/LLM/Gemma3.dockerfile

@@ -34,4 +34,4 @@ ENV RKLLM_MODEL_PATH=/app/models/${MODEL_FILE}
 
 EXPOSE 8001
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3588"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3588"]

docker/RK3588/LLM/Qwen3.dockerfile

@@ -34,4 +34,4 @@ ENV RKLLM_MODEL_PATH=/app/models/${MODEL_FILE}
 
 EXPOSE 8001
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3588"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_llm.py --rkllm_model_path ${RKLLM_MODEL_PATH} --target_platform rk3588"]

docker/RK3588/LLM/deploy_llm.dockerfile

@@ -0,0 +1,25 @@
+# 第一阶段：环境准备
+FROM python:3.10-slim AS base
+
+RUN apt-get update && \
+    apt-get install -y wget curl git sudo libgomp1 && \
+    rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# 安装依赖
+COPY ./src/llm/requirements.txt /app/requirements.txt
+RUN pip install --no-cache-dir -r /app/requirements.txt
+
+# 拷贝运行时库和工具
+COPY ./lib/librkllmrt.so /usr/lib/librkllmrt.so
+COPY ./src/fix_freq_rk3588.sh /app/fix_freq_rk3588.sh
+RUN chmod +x /app/fix_freq_rk3588.sh
+
+# --- 最终阶段 ---
+FROM base AS final
+
+COPY ./src/llm/fastapi_server_llm.py /app/
+
+EXPOSE 8001
+

docker/RK3588/VLM/DeepseekOCR.dockerfile

@@ -48,4 +48,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3588/VLM/InternVL3.dockerfile

@@ -19,8 +19,8 @@ COPY ./lib/librkllm_service.so /usr/lib/librkllm_service.so
 COPY ./lib/librknnrt.so /usr/lib/librknnrt.so
 
 
-COPY ./src/fix_freq_rk3576.sh /app/fix_freq_rk3576.sh
-RUN chmod +x /app/fix_freq_rk3576.sh
+COPY ./src/fix_freq_rk3588.sh /app/fix_freq_rk3588.sh
+RUN chmod +x /app/fix_freq_rk3588.sh
 
 # --- 最终阶段 ---
 FROM base AS final
@@ -48,4 +48,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3588/VLM/Qwen2-VL.dockerfile

@@ -50,4 +50,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3588/VLM/Qwen2.5-VL.dockerfile

@@ -50,4 +50,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3588/VLM/Qwen3-VL.dockerfile

@@ -22,6 +22,7 @@ COPY ./lib/librknnrt.so /usr/lib/librknnrt.so
 COPY ./src/fix_freq_rk3588.sh /app/fix_freq_rk3588.sh
 RUN chmod +x /app/fix_freq_rk3588.sh
 
+
 # --- 最终阶段 ---
 FROM base AS final
 ARG LLM_URL
@@ -50,4 +51,4 @@ ENV VISION_MODEL_PATH=/app/models/${VISION_FILE}
 
 EXPOSE 8002
 
-CMD ["sh", "-c", "python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]
+CMD ["sh", "-c", "bash /app/fix_freq_rk3588.sh && python3 /app/fastapi_server_vlm.py --llm_model ${LLM_MODEL_PATH} --encoder_model ${VISION_MODEL_PATH}"]

docker/RK3588/VLM/deploy_vlm.dockerfile

@@ -0,0 +1,27 @@
+# 第一阶段：环境准备
+FROM python:3.10-slim AS base
+
+RUN apt-get update && \
+    apt-get install -y wget curl git sudo libgomp1 && \
+    rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# 安装依赖
+COPY ./src/vlm/requirements.txt /app/requirements.txt
+COPY ./src/vlm/fastapi_server_vlm.py /app/fastapi_server_vlm.py
+RUN pip install --no-cache-dir -r /app/requirements.txt
+
+# 拷贝运行时库和工具
+COPY ./lib/librkllmrt.so /usr/lib/librkllmrt.so 
+COPY ./lib/librkllm_service.so /usr/lib/librkllm_service.so 
+COPY ./lib/librknnrt.so /usr/lib/librknnrt.so
+
+
+COPY ./src/fix_freq_rk3588.sh /app/fix_freq_rk3588.sh
+RUN chmod +x /app/fix_freq_rk3588.sh
+
+# --- 最终阶段 ---
+FROM base AS final
+
+EXPOSE 8002