Implement the join_inputs_design into the current SDA workflows.
#3634
Conversation
DongchenZ
requested review from
infotroph and
mdietze
and removed request for
mdietze
There was a problem hiding this comment.
FYI, this provides an immediate fix, but really you should be passing in the new input_design object too, not just the new ensemble.size variable. That change is needed to get the desired result of sampling jointly once at the start of the SDA and then keeping that design as the SDA iterates.
There was a problem hiding this comment.
Just checked are was reminded that input_design is now the first argument to write.ensemble.configs and it's not optional, so adding that fix is actually required in this PR, not just a good next step for a future PR.
pecan/modules/uncertainty/R/ensemble.R
Line 228 in ccc5135
I'd recommend that you add a call to modules/uncertainty/R/generate_joint_ensemble_design once in your code early on, that you save the design to restart, and that you read that design during restart rather than generating a new one (i.e., that if you have an "if restart" option, then the input_design is generated in the
"else")
DongchenZ
changed the title
join_inputs_design into the current SDA workflows.
| TimeseriesPlot = FALSE, | ||
| debug = FALSE, | ||
| pause = FALSE, | ||
| Profiling = FALSE, |
There was a problem hiding this comment.
Removed unnecessary flags.
| control=list(TimeseriesPlot = FALSE, | ||
| OutlierDetection=FALSE, | ||
| parallel_qsub = TRUE, | ||
| parallel_qsub = FALSE, |
There was a problem hiding this comment.
Set the parallel_qsub as FALSE because we can't use both local and remote qsub for the model execution.
| run_parallel = TRUE, | ||
| MCMC.args = NULL), | ||
| MCMC.args = NULL, | ||
| local.execution = TRUE), |
There was a problem hiding this comment.
Add the local.execution when we just want to run the jobs locally.
base/logger/R/logger.R
Outdated
| "latin-ascii" | ||
| ) | ||
| if (nchar(long.msg) > 20 && wrap) { | ||
| if (wrap) { |
There was a problem hiding this comment.
The hard-coded length in the if control should be removed. The wrap argument should fully define it.
|
|
||
| export(met2model.STICS) | ||
| export(model2netcdf.STICS) | ||
| export(pecan2stics) |
There was a problem hiding this comment.
This is generated by building the existing pecan files.
There was a problem hiding this comment.
The newest commits provide the new debias workflow (which has already been implemented in the current SDA workflows), and the old script has been moved to the inst folder. This is cleaner and easier to understand, which could also benefit any future development of the debiasing algorithm.
| } | ||
| cores <- cores - 1 | ||
| # if we only have one CPU. | ||
| if (cores < 1) cores <- 1 |
There was a problem hiding this comment.
Update the parallel registration.
| #' `qsub_parallel`, where we first combine jobs and submit them into the SCC. | ||
| #' @param debias List: R list containing the covariance directory and the start year. | ||
| #' covariance directory should include GeoTIFF files named by year. | ||
| #' start year is numeric input which decide when to start the debiasing feature. |
There was a problem hiding this comment.
Removed the previous debias workflow and replaced it with the new workflow.
This reverts commit 77bda48.
| } | ||
| PEcAn.logger::logger.info("SMAP SMP Prep Completed!") | ||
| list(SMP_Output = SMAP_Output, time_points = time_points, var = "SoilMoist") | ||
| list(SMP_Output = SMAP_Output, time_points = time_points, var = "SoilMoistFrac") |
There was a problem hiding this comment.
Correct the variable name from SoilMoist to SoilMoistFrac.
2 participants
Description
Motivation and Context
Review Time Estimate
Types of changes
Checklist: