heldsuarez.jl driver, this demo shows how to run multiple experiments and plot their differences using one bash script.
Setup the pipeline bash script in the
ClimateMachine.jldirectory as shown here, and change the highlighted lines.
- Setup the
exp_parametersfile. This is a list of
par1 val1 par2 val2 ..., where each line represents one model run. Note that parameter names must include all characters before the
=sign, and values all of the characters after the
=sign, as is in the experiment run file. For example, this script will modify the
/experiments/AtmosGCM/heldsuarez.jlfile and initiate three experiments of ΔT_y=(30K,60K,90K), each running for 0.5 days:
ΔT_y FT(60) n_days::FT 0.5 ΔT_y FT(30) n_days::FT 0.5 ΔT_y FT(90) n_days::FT 0.5
helper_mod.shwhich contains functions for the automated parameter swapping.
Run script using
- Your specified output folder should contain
.../netcdf/containing the diagnostics output
.../restart/containing all restart
.jld2files from individual nodes, for each experiment
- View your notebook on a local machine
- On both local and remote machines:
- ensure JupyterLab is installed on both local and remote machines (check version of Julia and necessary packages for your notebook) - Remote host:
cdinto your output directory
jupyter notebook --no-browser --port=XXXX- Local host:
ssh -N -f -L YYYY:localhost:XXXX <remoteuser>@<remote-cluster-node>
- In your local browser type
- You may get a prompt to authenticate using a token (printed when launched the notebook on the remote host), then you’re good to go!