Docker_GPU_Deployment.md

2. 🐳 Docker GPU Deployment Section

Requires NVIDIA Container Toolkit to run GPU containers: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html

Run Official GPU Container:

docker run -d --gpus all \
  -p 7860:7860 \
  linzeyi/hivision_idphotos

For a custom GPU-specialized image, please refer to docker/gpu/Dockerfile.

---

3. 📁 Docker GPU Structure and Files

3.1. Directory Structure

HivisionIDPhotos/
├── docker/
│   ├── gpu/
│   │   ├── Dockerfile
│   │   ├── README_GPU_DOCKER.md
│   │   └── entrypoint.sh (Optional)

3.2. File: docker/gpu/Dockerfile

FROM python:3.10-slim

# Enable NVIDIA GPU inside Docker
ENV NVIDIA_VISIBLE_DEVICES=all
ENV NVIDIA_DRIVER_CAPABILITIES=all

WORKDIR /app

# Install system libs
RUN apt-get update && apt-get install -y git wget && rm -rf /var/lib/apt/lists/*

# Copy project
COPY . /app

# Install base requirements
RUN pip install -r requirements.txt
RUN pip install -r requirements-app.txt

# Install GPU execution provider
RUN pip install onnxruntime-gpu==1.18.0

# Optional: Install CUDA PyTorch
# RUN pip install torch --index-url [https://download.pytorch.org/whl/cu121](https://download.pytorch.org/whl/cu121)

EXPOSE 7860

CMD ["python3", "app.py"]

3.3. File: docker/gpu/README_GPU_DOCKER.md

GPU-Enabled Docker Image for HivisionIDPhotos

This Dockerfile enables CUDA GPU acceleration through onnxruntime-gpu.

Requirements

• NVIDIA GPU
• NVIDIA Drivers
• NVIDIA Container Toolkit: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html

Build

docker build -t hivision_idphotos_gpu -f docker/gpu/Dockerfile .

Run

docker run -d --gpus all \
  -p 7860:7860 \
  hivision_idphotos_gpu

Visit: http://127.0.0.1:7860