05_run_afe.sh

#!/bin/bash
#SBATCH -o PATH_TO_LOGS/%x_%a.slurm.out
#SBATCH -e PATH_TO_LOGS/%x_%a.slurm.err
#SBATCH -n N_TASKS
#SBATCH --gres=gpu:N_GPUS
#SBATCH --cpus-per-gpu N_CPUS

eval "$(conda shell.bash hook)"
conda activate obss

#try without this: #SBATCH --mem MEMORY

start=`date +%s`

# get ligand pair and lambda string from input variables
transformation=$1
lambdastring="$2"

# set variables in meze.py
engine=ENGINE
repeats=N_REPEATS 
outputs_dir=OUTPUTS_DIR

echo "$lambdastring"
# read in lambda string
INPUT_FILE_STREAM=" " read -ra lambdas <<< "$lambdastring"

# use slurm array task id as lambda window index
id=$SLURM_ARRAY_TASK_ID
lambda=${lambdas[$id]}

for stage in "unbound" "bound"
do
	for (( i=1; i<=$repeats; i++)) 
	do
		lambda_directory=$outputs_dir/${engine}_${i}/$transformation/$stage/lambda_$lambda
		echo "Lambda directory is: $lambda_directory"
		
		echo "Using ${engine}_${i} for $transformation at $stage lambda $lambda"
		
		if [[ $engine == *"SOMD"* ]]; then	

			echo "Running AFE transformation..."
			cd $lambda_directory
		
			$BSS_HOME/somd-freenrg -C ./somd.cfg -l $lambda -c ./somd.rst7 -t ./somd.prm7 -m ./somd.pert -p CUDA 1> ./somd.out 2> ./somd.err
		else
			echo "Engine $engine is not supported yet."
		fi
	done
done

end=`date +%s`
runtime=$((end - start))

echo "Finished in $runtime seconds, or $((runtime/60)) minutes"