To ease the use of common python module, a apptainer container is defined with a conda environment including standard packages.

The container launches a jupyter notebook with a lig microscopy image analysis environment.

Launch a container with a jupyter lab on a remote machine with a conda environment.

• environment.yml : image formats, scikit, scipy, pytorch, cellpose, tensorflow, etc
• bioimaging.def: container based on mambaorg/micromamba

Build the container from the def and yml file:

apptainer build --force bioimaging.sif bioimaging.def

Run as on a login node (note the --bind option to get access to data drives)

ssh <username@loginnode>
apptainer run --writable-tmpfs --bind /path/to/data:/path/ bioimaging.sif

Run on a compute node:

ssh <username@loginnode>
 srun --partition=gpu -c 112 --pty bash -i
 apptainer run  --nv --writable-tmpfs --bind /path/to/data:/path/to/data bioimaging.sif 

or on a gpu node:

ssh <username@loginnode>
srun -p ml --gres=gpu:4 --pty bash -i
apptainer run  --nv --writable-tmpfs --bind /path/to/data:/path/to/data bioimaging.sif 

Connect to the jupyter kernel.

https://arcdocs.leeds.ac.uk/usage/conda-containers.html