These are examples of useful scripts for visualising output from
- Run LES experiment and plot some output
- Set up multi-experiment GCM runs with performance info tracking on GPU
- Basic GCM: run and plot
- Output and plot the last timestep before crashing
- Run multiple experiments and plot their differences
NB: these demos are based on the ClimateMachine.jl v0.2 release.
VizCLIMA contains Julia-based scripts for analysis of output from the
ClimateMachine.jl. It can be used as part of the end-to-end pipeline, as shown above, or for stand-alone analysis of
.nc output. Here are some examples:
To apply a Julia script on ClimateMachine.jl output, and convert it into a Jupyter Notebook using Literate, run:
VIZCLIMA_HOME=<location-of-your-VizCLIMA.jl> VIZCLIMA_SCRIPT=<your-VizCLIMA.jl-script> CLIMA_ANALYSIS=<location-of-your-NetCDF-file(s)> julia --project=$VIZCLIMA_HOME -e 'using Pkg; Pkg.instantiate(); Pkg.API.precompile()' VIZCLIMA_LITERATE=$VIZCLIMA_HOME'/src/utils/make_literate.jl' julia --project=$VIZCLIMA_HOME $VIZCLIMA_LITERATE --input-file $CLIMA_ANALYSIS/$VIZCLIMA_SCRIPT --output-dir $CLIMA_ANALYSIS
- when analysing large data files, Jupyter Notebooks can be very slow, so the computation is recommended to be done using Julia scripts and on a compute node (e.g.
hpc-89-24). For optimising your code in Julia see, for example, this tutorial.
ncviewfor fast exploration of large GCM output when on Caltech Cluster
Gallery (use the scripts above)
- CliMA Wiki:
- end-to-end bash scripts
- Setup JupyterLab on Caltech Cluster
- setup SLURM environment on Caltech’s Cluster and run ClimateMachine.jl (updated with each release): choose to run on CPU or GPU
- initial GCM physics testing