This pull request adds the first part of submitting a large job to the API through the script job_submission.py. Essentially the user gives it a path to a folder with .obj and the script will run num_simulations_per_object experiments.
Experiment management
While there is no automatic experiment management provided natively through the api the user can use the variableworking_dir. This sets the inductiva.working_dir variable ensuring all the API logs are sent to that directory. This logging is already done automatically through the API.
For example, setting WORKING_DIR=../../wind_tunnel_experiments/experiment_1 will create under the given working dir:
Where task_metadata.json contains information about all simulations submited.
~### Extras~
~* Changed the __init__ and added a setup.py file to make the make lib instalable with pip install -e .;~
~### Simplifications with respect to previous iterations~
~Here the submission code is much cleaner as the API is already storing the logs and the inputs to each task. This removes the extra necessary code for experiment management. To get a better understanding of what is no longer necessary look here, here and here~
This pull request is using a very outdated version of the code.
With the current codebase large scale job submission is very easy to do, see batch_run.py
Contents
This pull request adds the first part of submitting a large job to the API through the script
job_submission.py
. Essentially the user gives it a path to a folder with.obj
and the script will runnum_simulations_per_object
experiments.Experiment management
While there is no automatic experiment management provided natively through the api the user can use the variable
working_dir
. This sets theinductiva.working_dir
variable ensuring all the API logs are sent to that directory. This logging is already done automatically through the API.For example, setting
WORKING_DIR=../../wind_tunnel_experiments/experiment_1
will create under the given working dir:Where
task_metadata.json
contains information about all simulations submited.~### Extras~
~* Changed the
__init__
and added asetup.py
file to make the makelib
instalable withpip install -e .
;~~### Simplifications with respect to previous iterations~
~Here the submission code is much cleaner as the API is already storing the logs and the inputs to each task. This removes the extra necessary code for experiment management. To get a better understanding of what is no longer necessary look here, here and here~