Containers

Containers#

Get an image#

# Apptainer
apptainer pull zoomy_jax.sif oras://ghcr.io/zoomylab/zoomy_jax_sif:latest

# Docker
docker pull ghcr.io/zoomylab/zoomy_jax:latest

Every image below exists for both. Replace zoomy_jax by the image you need. Which backend for what: Backends.

Image

Backend

zoomy_numpy

NumPy

zoomy_jax

JAX, CPU and GPU

zoomy_amrex

AMReX, CPU and GPU

zoomy_openfoam

OpenFOAM with preCICE

zoomy_dmplex

DMPlex

zoomy_firedrake

Firedrake

zoomy_postprocess

plotting and HDF5

Run#

The first argument chooses the mode.

apptainer run zoomy_jax.sif 8080      # solver for the GUI on localhost:8080
apptainer run zoomy_jax.sif jupyter   # JupyterLab on localhost:8888
apptainer run zoomy_jax.sif shell     # shell

Docker needs the ports and your files:

docker run --rm -p 8080:8080 ghcr.io/zoomylab/zoomy_jax:latest
docker run --rm -p 8888:8888 -v $PWD:/workspace ghcr.io/zoomylab/zoomy_jax:latest jupyter

GUI#

Start the solver on port 8080, open the GUI and connect it (GUI, “Connect backend”, http://localhost:8080). Details: Connecting your own backend.

Jupyter#

apptainer run --bind $PWD:/workspace zoomy_jax.sif jupyter

Open the printed address. The kernel has the backend installed, your files are in /workspace.

CPU and GPU#

zoomy_jax and zoomy_amrex run on the CPU, and on the GPU when the host driver is brought in.

apptainer exec zoomy_jax.sif python run.py          # CPU
apptainer exec --nv zoomy_jax.sif python run.py     # GPU
docker run --rm --gpus all ghcr.io/zoomylab/zoomy_jax:latest   # GPU with Docker
Contents