Running in Containers

AMPLFI also provides a container build available via the github container repository

You can pull the container locally with either docker or apptainer

apptainer

$ apptainer pull ${AMPLFI_CONTAINER_ROOT}/amplfi.sif docker://ghcr.io/ml4gw/amplfi/amplfi:main

Supported python versions: 3.10-3.12.

docker

$ docker pull ghcr.io/ml4gw/amplfi/amplfi:main

Supported python versions: 3.10-3.12.

AMPLFI train and tune commands can now be run inside the container. For example, to train a flow inside the container, you can run

$ APPTAINERENV_AMPLFI_OUTDIR=/path/to/outdir APPTAINERENV_AMPLFI_OUTDIR=/path/to/datadir/ \
    apptainer run $AMPLFI_CONTAINER_ROOT/amplfi.sif --nv \ 
    amplfi-flow-cli fit --config /path/to/config.yaml

Note that we must map in the APPTAINERENV_AMPLFI_OUTDIR and APPTAINERENV_AMPLFI_DATADIR environment variables so they can be accessed inside the container.

Mounting Local Code

During development, often times you will make code changes and want to test them inside the container. To do so, you could rebuild the entire container locally so that your changes are reflected in the container,

apptainer build $AMPLFI_CONTAINER_ROOT/amplfi.sif apptainer.def

but it can be quite cumbersome to do this each time you make a minor tweak to the code base.

Fortunately, AMPLFI is installed editably inside the container at the path /opt/amplfi. So, it is possible to map local changes into the container at runtime by bind mounting your local repository. As an example, the above training command can be modified to do so using the -B flag (replace /path/to/local/amplfi with the location where your local amplfi lives).

APPTAINERENV_AMPLFI_OUTDIR=/path/to/outdir APPTAINERENV_AMPLFI_OUTDIR=/path/to/datadir/ \
    apptainer run -B /path/to/local/amplfi:/opt/amplfi --nv \
    $AMPLFI_CONTAINER_ROOT/amplfi.sif amplfi-flow-cli fit \
    --config /path/to/config.yaml

Note

If you make modifications the the environment like adding python dependencies you will have to rebuild the containers!