Homec4science
Diffusion solar_potential (master) Browse

solarPOT/sbatch/create_sbatch.sh
72d4dec17468master

Last Change
Raw File

create_sbatch.sh
View Options

#!/bin/bash
if [ "$1" == "-h" ] ; then
echo ""
echo "####################################################"
echo "This script replaces lines in a template file (submit_template.sh) and create a workspace for the jobs submission."
echo "You need to specify/have:"
echo "- a template file 'submit_template.sh' in the directory where you are executing the script"
echo "- a python executable file in your working directory"
echo "- a set of inputs for your jobs, meaning: model, partition (CPU or GPU), memory, input data folder, test dataset, nr. nodes for the ML algorithm."
echo "By invoking this script you will create a workspace with files having the righ formatting for the submission on sbatch system."
echo "Author: roberto.castello@epfl.ch"
echo "Date: 19/03/2018 "
echo "####################################################"
echo ""
exit 0
fi
## STATIC inputs #####################################
TEMPLATE_FILE="submit_template.sh"
GPU_syntax="#SBATCH --partition=gpu --qos=gpu_free --gres=gpu:2"
######################################################
## CUSTOMIZABLE inputs ###############################
######################################################
MAIL="roberto.castello@epfl.ch"
CORES="1"
TIME="00:30:00" #set the maximum time in the format XX:YY:ZZ
VIRTUALENV="hyenergy_py3" ## your virtual env name
MEM="4G" ## memory requirment for the node
MODEL="ELM" ## ML model name
PYTHON_EXECUTABLE="ELM_uncertainties_query" # becomes [..]_query.py or [..]_train.py or [..].py
MYDIR="energy-potential/Solar" # path should start from home
SCRATCHDIR="/scratch/rcastell" ## your scratch
DATADIR="\/scratch\/rcastell" ## --> needed for the different syntax N.B. it is enough to put your scratch location (modulo a dataset/ folder is existing)
## Arrays
PART_VEC=( "GPU" ) ## Partition: "CPU" or "GPU" or both
DATA_VEC=( "2012-2012_grid100_SIS" ) ## Dataset name
QUERY_VEC=( "grid1600" ) ## Query name
NODE_VEC=( "200" ) ## nr. nodes of ELM
EST_VEC=( "2" "5" "10" ) ## nr. estimators for the ensamble
######################################################
######################################################
for DATA in ${DATA_VEC[@]}
do
echo "[MSG] Creating workspace for the dataset $DATA"
echo ""
## == Creating dir if not already existing and copying the python executable into it
if [ ! -d $SCRATCHDIR\/workspace_$MODEL\_$DATA ]
then
mkdir $SCRATCHDIR\/workspace_$MODEL\_$DATA
mkdir $SCRATCHDIR\/workspace_$MODEL\_$DATA\/output
fi
## == Copying the python executable in the folder
cp $HOME\/$MYDIR\/$PYTHON_EXECUTABLE.py $SCRATCHDIR\/workspace_$MODEL\_$DATA\/.
## == Printing lines of the template (can be skipped)
#cat $TEMPLATE_FILE | while read line_input
#do
# [[ $line_input = \#\#* ]] && continue
# echo $line_input
#done
for QUERY in ${QUERY_VEC[@]}
do
for PART in ${PART_VEC[@]}
do
for NODE in ${NODE_VEC[@]}
do
for EST in ${EST_VEC[@]}
do
echo "..."
cp $TEMPLATE_FILE $SCRATCHDIR\/workspace_$MODEL\_$DATA\/.
## == Replacing the custom choice in a temporary file
sed -i "s/\_mem/$MEM/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_time/$TIME/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_cores/$CORES/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_mail/$MAIL/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_virtenv/$VIRTUALENV/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_datadir/$DATADIR/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_model/$MODEL/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_python/$PYTHON_EXECUTABLE/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_dataset/$DATA/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_accel/$PART/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_qry/$QUERY/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_estimators/$EST/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
sed -i "s/\_node/$NODE/g" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
echo ".."
## == Adding the GPU syntax (if needed)
if [ $PART == GPU ]
then
sed -i "s/\_part/$GPU_syntax/" $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
else
sed -i 's/\_part/\#\#/' $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
fi
## == customizing for train or validation ==
if [[ $PYTHON_EXECUTABLE == ELM_*_train ]]
then
cp $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$MODEL\_$PART\_N$NODE\_E$EST\_TRN.sh
elif [[ $PYTHON_EXECUTABLE == ELM_*_query ]]
then
cp $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$MODEL\_$PART\_N$NODE\_E$EST\_$QUERY\_QRY\.sh
elif [[ $PYTHON_EXECUTABLE == RF_* ]]
then
cp $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$MODEL\_$PART\_N$NODE\_E$EST\_$QUERY\.sh
else
echo "[WARNING] The python file you are trying to execute is not recognized"
fi
## == Cleaning
rm $SCRATCHDIR\/workspace_$MODEL\_$DATA\/$TEMPLATE_FILE
echo "."
echo ""
done
done
done
done
## == Creating a script (run.sh) in the workspace for the sbatch submission (N.B. not executing, just creating)
cd $SCRATCHDIR\/workspace_$MODEL\_$DATA\/
suffix=".sh"
if [ ! -f runAll.sh ]; then
echo "#!/bin/bash" >> runAll.sh
chmod +x runAll.sh
for f in ./*_*.sh; do
name=$(basename "$f")
echo "sbatch -J" ${name%$suffix} $name >> runAll.sh
done
else
for f in ./*_*.sh; do
exist=false
name=$(basename "$f")
while IFS= read -r line; do
if [[ $line == *"$name" ]] ; then
exist=true
fi
done < runAll.sh
if [ $exist == false ]; then
echo "sbatch -J" ${name%$suffix} $name >> runAll.sh
fi
done
fi
cd ../
## == Terminating
echo "[DONE] The files and relative scripts for the submission on batch queues have been created inside "$SCRATCHDIR"/workspace_"$MODEL"_"$DATA"/"
done
c4science · Help