indirect_inference

Compilation

mkdir -p build
Sinteractive
cd build
../compile.sh
exit

Run

First adjust the params.dat file with the parameters you wish to use, according to variables they are corresponding to in params.var. Adjust the start.sbatch file in order to reflect the required ressources for the run.

The start.sbatch.example is an example that fits well a run with the parameters on params.dat.example. In the example, there are 1000 Monte-Carlo iteration, distributed over 42 tasks (24 iterations per task except for the last one which has 16 iterations). Iterations typically need 45 minutes wall-clock time, however they can sometimes be significantly longer (sometimes up to several hours). In particular, this is the case when having several instances of your matlab code running on the same node, thus sharing memory band and bus. Each job in the array (ie. each task) is hence given 5h59m59s, which should guarantee that the iterations terminate successfully and avoid queuing in >6h job queue.

You can then start the job with:

sbatch start.sbatch

The job array ID is then printed on the screen. The command Squeue shows the status of the running jobs (nothing shown when terminated).

Logs and results

Logs are written for each task in the logs/slurm-%A/slurm-%a.out file, and %A stands for the job ID and %a stands for the array ID. They correspond to what would otherwise be printed on the screen, if jobs were run locally.

Results are written for each Monte-Carlo iteration in the results/run_%A folder.

Scientific-Computing/indirect_inference

About

Resources

Stars

Watchers

Forks

Releases

Packages 0

Contributors 2

Languages

Packages