Skip to content

2024‐02‐21

Jump to bottom
Teagan King edited this page Feb 21, 2024 · 7 revisions

February 21, 2024

Agenda:

  1. General updates (would be great to have everyone there, ~15 minutes)
    • Meeting structure going forward
    • Work in progress:
      • Parallelization
      • Getting Started documentation
      • Sea ice notebook / GeoCAT collaboration
      • Per-group status updates / future plans / short-term goals (if anything to report)
  2. Collaborative work time (Join if interested)
    • Setting up notebook using Lev's workflow
    • Small group on config file setup / breaking out into multiple component-specific config files
    • Polar stereograph plots (see issue from geoCAT & Dave B)

Slides

Notes

  • Parallelization:
    • Use qinteractive to request multiple cores on casper
    • Mike has been working with this to speed up notebooks
    • We can look into playing with less coarse resolution land notebooks with this
    • Atmosphere notebook fairly quick in serial
  • Documentation
    • Lev has put together getting started notebook to add notebooks
  • Dave Bailey mentioned PCMDI diagnostics package uses python scripts that are run from within notebook -- is this a performance benefit?
    • Probably the only way to do things in parallel that we know if is through dask, so this likely wouldn't be faster within python script
    • We do want to support notebooks and/or scripts
    • Good to be aware of overlap here
  • Group updates:
    • Ocean group: will continue working on diagnostics notebooks and going forward make them compatible (not necessarily dealing with CUPiD framework). Gustavo will start running in CUPiD-Analysis environment so that we know if we need to add to it.
    • Documentation: descriptions for each function, etc would be useful to add at a future point.
      • NCAR-specific documentation can be moved out of the README and into an NCAR-specific tools page
    • Lev mentioned that we could be more specific about what you can actually do without using CUPiD directly
    • Brian Medeiros was looking at what we can pull out of ADF
    • ADF: want more 'real' run as opposed to quick run
    • We should add an 'adding scripts' page to the documentation as well.
  • It is useful to make a fork and branch for contributing to code (see workflow here)
  • ADF has notebook version, too
  • LMWG can work with ADF to see if can use CLM history output in ADF
  • Dani mentioned there are some diagnostic standards created by MDTF that could be useful to consider during the design phase
Clone this wiki locally