Singularity images to support a bare-bones minimum image for creating Singularity images for HPC development. This repo specifically has been engineered for use w/ MARCC at JHU.
Supports:
- pytorch with cuda
- tensorflow with cuda
- keras
Additional Package List:
- anaconda
- numpy
- scipy
- scikit-learn
- opencv
- pandas
- pytest
- flake8
- tensorboard
- tensorboardx
- tqdm
- protobuf
- onnx
- spectrum
- nibabel
- mne
Singularity images are built and indexed on https://singularity-hub.org/. To add a new build one should create a branch of this repo, and then activate the branch in your 'Collections' on Singularity-Hub. Each branch has their own 'Singularity' image, which the Hub looks for and builds. Then the uri as
shub://<github_user>/deeplearning_hubs:<branch_name>
# (e.g. shub://adam2392/deeplearning_hubs:pytorch)
can be used to build the corresponding Singularity image on that page. This can then be pulled from shub. For example:
singularity pull --name tensorflow.simg shub://marcc-hpc/tensorflow
Then you can run scripts from this singularity container:
# redefine SINGULARITY_HOME to mount current working directory to base $HOME
export SINGULARITY_HOME=$PWD:/home/$USER
# run signularity image w/ python script
singularity exec --nv ./tensorflow.simg python softmax_regression.py