-
Notifications
You must be signed in to change notification settings - Fork 2
/
how-to-launch.txt
77 lines (51 loc) · 4.06 KB
/
how-to-launch.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
=========================================================================
To launch EMEWS workflows in MN4 we need to run some commands beforehand:
# assign more memory for swift VM
salloc -p interactive --cpus-per-task=2
# Required modules
module load python java R/3.4.0 swiftt/1.4.3
The first command will request 2 login nodes, and we are interested in that because we can't launch swift-t with 2GB of ram. Using 2 nodes allow us to gather 4GB, and thus to launch swift-t without issues.
The second command will load swift-t and dependencies.
=========================================================================
This project forked from Gerard's repo, implements a basic model of the TNF fate model pulse experiments.
The code for the multi-scale model is placed in data/PhysiBoSSv2.
Before running any exploration, the TNF model must be compiled running the following commands:
cd data/PhysiBoSSv2
make
cd ../..
That's it! If it works, a binary file named spheroid_TNF_v2 should be generated.
To perform the model exploration there are currently two implemented exploration strategies:
=============
- GA search
=============
GA search uses a cost function f (defined in python/get_metrics.py) to find parameters that minimize f.
Currently f is the minimization of the number of alive cells.
The command to run the GA exploration
bash swift/swift_run_eqpy.sh <EXPERIMENT_ID> <GA_PARAMS_FILE> (e.g. ${script_name} experiment_1 data/ga_params.json)"
* <EXPERIMENT_ID>: and string that will be used as a name for the results folder ./experiments/<EXPERIMENT_ID>
* <GA_PARAMS_FILE>: a json file with the input a parameters for the genetic algorithm
If you wish to run the GA for calibrating PhysiBoSSv2 parameters, you can issue the following command:
bash swift/swift_run_eqpy_compare.sh <EXPERIMENT_ID> <GA_PARAMS_FILE> <DISTANCE_TYPE>(e.g. ${script_name} experiment_1 data/ga_params.json euclidean)"
* <DISTANCE_TYPE>: One out of three implemented distance types {'euclidean', 'dtw', 'l1'}
In this case, the output of PhysiBoSSv2 simulations is compared against two known and validated golden standard cases.
===================
- Parameter sweep
===================
The sweep is a regular/random grid search that runs one simulation for each set of parameters present in the input file.
In addition to the parallel execution of the simulations this workflow also runs a script to summarize simulation outputs (scripts/summarize_simulation.py). To input for the sweep is a .txt with one json dictionary per line. The dictionary contains the parameters to be tested. There is a python script (scripts/generate_json_sweep.py to generate sweep input file using a template
The command to run the parameter sweep:
bash swift/swift_run_sweep.sh <EXPERIMENT_ID> <SWEEP_INPUT> (e.g. ${script_name} experiment_1 data/input.txt)
In both cases, the .sh script will start the slurm job, so there is no need to call the sbatch command.
To modify the job options, we need to modify directly the run_ga.sh file.
The key parameters to tweak in the swift/swift_run_sweep.sh are the following four:
PROCS=48 # total number of processes
- QUEUE=main # defaul queue
- WALLTIME=47:59:00 # total time assigned to the run
- PPN=3 # Processes per computing node
- TURBINE_SBATCH_ARGS=--qos=bsc_ls # This is for BSC MN4 queue system
- EXECUTABLE_SOURCE=$EMEWS_PROJECT_ROOT/data/<MODEL BINARY> # Path to the executable model file
- DEFAULT_XML_SOURCE=$EMEWS_PROJECT_ROOT/data/<MODEL PARAMETER TEMPLATE> # Path to the input params template
- the first is an integer indicating the number of experiment (to store several experiments or runs)
- the second is the file with genetic algorithm global parameters
- the third is the file with an initial population of ga_params values (population initialization, i.e. PhysiCell parameters to explore)
- the fourth is the number of threads each PhysiCell simulation will use.