mlf-sample-project.def

# We are going to start with a base image from NVIDIA, which is a Ubuntu system with CUDA and cuDNN installed.
# You can also use a different base image. There are many available on Docker Hub (https://hub.docker.com).
# For instance, there are images for PyTorch, PyTorch Lightning, HuggingFace, MosaicML, etc.
# Most of the aforementioned images start with a NVIDIA base image like the below, so we will use that as our base for simplicity.
# To do so, change the FROM line below to the url of the image tag you want to use.
Bootstrap: docker
From: nvcr.io/nvidia/cuda:12.1.1-cudnn8-runtime-ubuntu22.04

%files
    # Copy the requirements.txt file into the container so it can be accessed during building.
    requirements.txt

%post
    # Update and install Ubuntu packages
    apt-get update && apt-get install -y
    apt-get install -y --no-install-recommends git curl

    apt-get install -y python3 python3-pip python3-venv
    # clean up the apt cache
    apt-get clean
    rm -rf /var/lib/apt/lists/*

    python3 -m venv /venv
    PATH=/venv/bin:$PATH
    pip3 install wheel

    # Install the required Python packages without caching the pip install files
    pip3 install -r requirements.txt --no-cache-dir
    
    # install code-server
    curl -fsSL https://code-server.dev/install.sh | sh

%environment
    PATH=/venv/bin:$PATH