Dockerfile for faster-whisper in GCP Cloud Run GPU

Dockerfile

#
# Dockerfile for a simple PoC faster-whisper translation server in GCP GPU Cloud Run
# 20240925 rickt
#
#  build: $ docker build -t gcr.io/<gcp_project_name>/<service_name> .
#   push: $ docker push gcr.io/<gcp_project_name>/<service_name>
# deploy: $ gcloud beta run deploy <service_name> --region us-central1 --image gcr.io/<gcp_project_name>/<service_name> \
#           --port 8080 --cpu 8 --memory 32Gi --gpu 1 --gpu-type nvidia-l4 --max-instances 1 --allow-unauthenticated
#

# use latest ubuntu 24 NVIDIA CUDA base image
FROM nvidia/cuda:12.6.1-runtime-ubuntu24.04

# expose GPU(s)
ENV NVIDIA_VISIBLE_DEVICES=all
ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility

# update the system pkg lists & install only what we need
RUN apt-get update && apt-get install -y \
    ffmpeg \
    curl \
    wget \
    git \
    python3-pip \
    python3-venv \
    && rm -rf /var/lib/apt/lists/*

# nvidia stuff
RUN curl "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.0-1_all.deb" -o cuda.deb && \
    dpkg -i cuda.deb && rm cuda.deb
RUN apt-get update -y && \
    apt-get install -y --no-install-recommends \
    cuda-cudart-11-8 cuda-nvcc-11-8 cuda-nvrtc-11-8 \
    'libcudnn8=8.9.1.*+cuda11.8' libcublas-11-8 && \
    rm -rf /var/*/apt

# create & activate a venv
RUN python3 -m venv /opt/venv 
RUN /opt/venv/bin/python -m pip install --upgrade pip 

# install python modules
RUN /opt/venv/bin/pip3 install faster-whisper[server] \
    torch --extra-index-url https://download.pytorch.org/whl/cu118 \
    nvidia-cublas-cu11 nvidia-cublas-cu12 \
    google-cloud-logging \
    flask \
    uuid

# download/preload small whisper model to the Docker image
RUN /opt/venv/bin/python3 -c "from faster_whisper import WhisperModel; model = WhisperModel('small', device='cpu')"

# app port
EXPOSE 8080

# directory
WORKDIR /app
COPY . /app

# entry point
CMD ["/opt/venv/bin/python3", "server.py"]

# EOF