plano/model_server/Dockerfile.gpu

# Use NVIDIA CUDA base image to enable GPU support
FROM nvidia/cuda:12.1.0-cudnn8-runtime-ubuntu22.04 as base
ENV DEBIAN_FRONTEND=noninteractive

# Install Python 3.10
RUN apt-get update && \
    apt-get install -y python3.10 python3-pip python3-dev python-is-python3 && \
    rm -rf /var/lib/apt/lists/*



#
# builder
#
FROM base AS builder

WORKDIR /src

# Upgrade pip
RUN pip install --upgrade pip

# Install git for cloning repositories
RUN apt-get update && apt-get install -y git && apt-get clean

# Copy requirements.txt
COPY requirements.txt /src/

# Install Python dependencies
RUN pip install --force-reinstall -r requirements.txt

RUN apt-get update && \
    apt-get install -y cuda-toolkit-12-2

# Check for NVIDIA GPU and CUDA support and install EETQ if detected
RUN if command -v nvcc >/dev/null 2>&1; then \
        echo "CUDA and NVIDIA GPU detected, installing EETQ..." && \
        git clone https://github.com/NetEase-FuXi/EETQ.git && \
        cd EETQ && \
        git submodule update --init --recursive && \
        pip install .; \
    else \
        echo "CUDA or NVIDIA GPU not detected, skipping EETQ installation."; \
    fi

COPY . /src

# Specify list of models that will go into the image as a comma separated list
ENV MODELS="katanemo/bge-large-en-v1.5-onnx"
ENV DEBIAN_FRONTEND=noninteractive

COPY /app /app
WORKDIR /app

# Install required tools
RUN apt-get update && apt-get install -y \
  curl \
  && rm -rf /var/lib/apt/lists/*

# Uncomment if you want to install the model during the image build
# RUN python install.py && \
#   find /root/.cache/torch/sentence_transformers/ -name onnx -exec rm -rf {} +

# Set the default command to run the application
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "80"]