PiNNAcLe (PiNN Activated Learning) is a collection of workflows built for activated learning and sampling of interatomic potentials. The workflows are implemented in the nextflow language to enable their scalable execution.
By default, PiNNAcLe workflows are executed with containerized environments, so the only dependencies required are singularity and nextflow:
nextflow run teoroo-cmc/pinnacle -entry h2o-demoPiNNAcLe also collects workflow configurations for known supercomputer centres that the developers have access. For those resources, profiles are provided that can be easily used:
nextflow run teoroo-cmc/pinnacle -entry h2o-demo -profile dardelSee a complete list in the documentation, along with guides to build your own profile.
The default workflow in PiNNAcLe is called acle (activated learning). Each implemented workflow has its set of parameters that can be set at runtime:
nextflow run yqshao/pinnacle --proj=testrun --initDs=myDs.{yml,tfr} --initModel=myModel.yml
The job history is automatically logged by nextflow, which one can recover by
nextflow log. For a better record of setup, you may also chose to use a
parameter file. Available parameters, along with parameter templates are given
for each workflow entry in the documentation.
The workflows in PiNNAcLe are modularized such that extension of the workflow is possible. If you wish to use PiNNAcLe as a starting point for your own project, use the copier template:
copier gh:teoroo-cmc/pinnaclePiNNAcLe is developed by Yunqi Shao at the TeC group in Uppsala University, Sweden.