Job problem on Dardel

examples/parallel/KTH_PDC/lateral_inhibition/sten_3/Dardel_simulate_lateral-B.job

@@ -7,12 +7,13 @@
 #SBATCH -J Simulate
 #SBATCH -A naiss2023-5-231
 #SBATCH --nodes=60
-#SBATCH --tasks-per-node=128
+#SBATCH --tasks-per-node=50
 #SBATCH --mail-type=ALL
 
 # 2024-02-16: 40 cores per node worked, had 28% free... trying increasing to 45 cores per node
 # 2024-02-18: Increasing back up to 128 cores. Memory free is probably more
 #             dependent on total number of nodes allocated
+# 2024-02-19: 128 cores failade memory.. testar 50 cores nu
 
 # You need to point this as the directory where you created the network in
 #NETWORK_DIR=/cfs/klemming/home/${USER:0:1}/$USER/Snudda/examples/parallel/KTH_PDC/networks/test_10k

examples/parallel/KTH_PDC/lateral_inhibition/sten_4/Dardel_runSnudda_lateral.job

@@ -0,0 +1,83 @@
+#!/bin/bash -l
+#SBATCH --partition=main
+#SBATCH -o log/runSnudda-%j-output.txt
+#SBATCH -e log/runSnudda-%j-error.txt
+#SBATCH -t 5:59:00
+#SBATCH -J Snudda
+#SBATCH -A naiss2023-5-231
+#SBATCH --nodes=2
+#SBATCH -n 256
+#SBATCH --cpus-per-task=2
+#SBATCH --mem-per-cpu=930M
+#SBATCH --mail-type=ALL                                                         
+module load snic-env
+
+
+#..
+#export OMP_STACKSIZE=128G
+ulimit -s unlimited
+
+
+#let NWORKERS="$SLURM_NTASKS-2"
+#let NWORKERS="100"
+let NWORKERS="40"
+
+# REMEMBER TO CREATE THE "log" DIRECTORY
+
+
+export IPNWORKERS=$NWORKERS
+
+
+export IPYTHONDIR="/cfs/klemming/scratch/${USER:0:1}/$USER/.ipython"
+rm -r $IPYTHONDIR
+export IPYTHON_PROFILE=default
+source $HOME/Snudda/snudda_env/bin/activate
+
+
+#.. Start the ipcontroller
+export FI_CXI_DEFAULT_VNI=$(od -vAn -N4 -tu < /dev/urandom)
+srun -n 1 -N 1 -c 2 --exact --overlap --mem=0 ./../../ipcontroller_new.sh &
+
+
+echo ">>> waiting 60s for controller to start"
+sleep 60 
+
+#.. Read in CONTROLLERIP
+CONTROLLERIP=$(<controller_ip.txt)
+
+
+##.. Start the engines
+echo ">>> starting ${IPNWORKERS} engines "
+#srun -n ${IPNWORKERS} -c 2 --exact --overlap ipengine --location=${CONTROLLERIP} --profile=${IPYTHON_PROFILE} --mpi \
+#--ipython-dir=${IPYTHONDIR}  --timeout=30.0 --log-level=DEBUG \
+#--BaseParallelApplication.verbose_crash=True --IPEngine.verbose_crash=True \
+#--Kernel.stop_on_error_timeout=1.0 --IPythonKernel.stop_on_error_timeout=1.0 \
+#Session.buffer_threshold=4096 Session.copy_threshold=250000 \
+#Session.digest_history_size=250000 c.EngineFactory.max_heartbeat_misses=10  c.MPI.use='mpi4py' \
+#1> ipe_${SLURM_JOBID}.out 2> ipe_${SLURM_JOBID}.err &
+
+#srun -n ${IPNWORKERS} -c 2 --exact --overlap valgrind --leak-check=full --show-leak-kinds=all \
+#ipengine --location=${CONTROLLERIP} --profile=${IPYTHON_PROFILE} --mpi \
+#--ipython-dir=${IPYTHONDIR}  --timeout=30.0 c.EngineFactory.max_heartbeat_misses=10  c.MPI.use='mpi4py' \
+#1> ipe_${SLURM_JOBID}.out 2> ipe_${SLURM_JOBID}.err &
+
+export FI_CXI_DEFAULT_VNI=$(od -vAn -N4 -tu < /dev/urandom)
+srun -n ${IPNWORKERS} -c 2 -N ${SLURM_JOB_NUM_NODES} --exact --overlap --mem=0 ipengine \
+--location=${CONTROLLERIP} --profile=${IPYTHON_PROFILE} --mpi \
+--ipython-dir=${IPYTHONDIR}  --timeout=30.0 c.EngineFactory.max_heartbeat_misses=10  c.MPI.use='mpi4py' \
+1> ipe_${SLURM_JOBID}.out 2> ipe_${SLURM_JOBID}.err &
+
+
+echo ">>> waiting 60s for engines to start"
+sleep 30
+
+export FI_CXI_DEFAULT_VNI=$(od -vAn -N4 -tu < /dev/urandom)
+srun -n 1 -N 1 --exact --overlap --mem=0 ./Dardel_runSnudda_lateral.sh
+
+
+echo " "
+
+echo "JOB END "`date` start_time_network_connect.txt
+
+wait
+

examples/parallel/KTH_PDC/lateral_inhibition/sten_4/Dardel_runSnudda_lateral.sh

@@ -0,0 +1,122 @@
+#!/bin/bash
+
+
+
+SNUDDA_DIR=$HOME/Snudda/snudda
+JOBDIR=../networks/sten_4
+
+# SIMSIZE=50000
+
+# If the BasalGangliaData directory exists, then use that for our data
+#/cfs/klemming/scratch/${USER:0:1}/$USER/BasalGangliaData/data
+#BasalGangliaData/Parkinson/PD0
+if [[ -d "$HOME/BasalGangliaData/data" ]]; then
+    export SNUDDA_DATA="$HOME/BasalGangliaData/data"
+    echo "Setting SNUDDA_DATA to $SNUDDA_DATA"
+else
+    echo "SNUDDA_DATA environment variable not changed (may be empty): $SNUDDA_DATA"
+fi
+
+mkdir -p $JOBDIR
+
+echo "Dardel_runSnudda.sh should be started with srun -n 1, to only get one process"
+
+echo "SLURM_PROCID = $SLURM_PROCID"
+
+if [ "$SLURM_PROCID" -gt 0 ]; then
+    mock_string="Not main process"
+else
+
+    # For debug purposes:                                                         
+    echo "PATH: "$PATH
+    echo "IPYTHONDIR: "$IPYTHONDIR
+    echo "PYTHONPATH: "$PYTHONPATH
+    echo "LD_LIBRARY_PATH: "$LD_LIBRARY_PATH
+
+    echo ">>>>>> Main process starting ipcluster"
+    echo
+
+    echo "Start time: " > start_time_network_connect.txt
+    date >> start_time_network_connect.txt
+
+    echo ">>> Init: "`date`
+    # snudda init ${JOBDIR} --size ${SIMSIZE} --overwrite --randomseed 1234
+    python setup_sten_4.py ${JOBDIR}
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during init, aborting!"
+	ipcluster stop	
+	exit -1
+    fi
+
+# WE NOW START IPCLUSTER USING ipcontroller.sh INSTEAD...
+#
+#    echo "SLURM_NODELIST = $SLURM_NODELIST"
+#    let NWORKERS="$SLURM_NTASKS - 1"
+#
+#    echo ">>> NWORKERS " $NWORKERS
+#    echo ">>> Starting ipcluster `date`"
+#    
+#    #.. Start the ipcluster
+#    ipcluster start -n ${NWORKERS} \
+#	      --ip='*' \
+#	      --HeartMonitor.max_heartmonitor_misses=1000 \
+#	      --HubFactory.registration_timeout=600 \
+#	      --HeartMonitor.period=10000 & 
+#
+#    
+#    #.. Sleep to allow engines to start
+#    echo ">>> Wait 120s to allow engines to start"
+#    sleep 120 #60
+
+    echo ">>> Place: "`date`
+    snudda place ${JOBDIR} --verbose
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during placement, aborting!"
+	# ipcluster stop	
+	exit -1
+    fi
+
+    echo ">>> Detect: "`date`
+    snudda detect ${JOBDIR} --hvsize 50 --parallel
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during detection, aborting!"
+	# ipcluster stop	
+	exit -1
+    fi
+
+    echo ">>> Prune: "`date`
+    snudda prune ${JOBDIR} --parallel
+
+    if [ $? != 0 ]; then
+	echo "Something went wrong during pruning, aborting!"
+	# ipcluster stop	
+	exit -1
+    fi
+
+    # Disable input generation at the moment
+
+    # echo ">>> Ablate: "`date`
+    # python ../ablate_network.py ${JOBDIR}
+
+    echo ">>> Input: "`date`
+    # snudda input ${JOBDIR} --parallel --time 18 --input input.json --networkFile ${JOBDIR}/network-synapses-minimal.hdf5
+    snudda input ${JOBDIR} --parallel --time 18 --input input.json
+
+
+    #.. Shut down cluster
+    # ipcluster stop	
+    #.. Shutdown ipcontroller
+    echo "Shutting down ipcontroller"
+
+    python ../../ipcontroller_shutdown.py
+
+
+    date
+    #echo "JOB END "`date` start_time_network_connect.txt
+
+    echo "EXITING Dardel_runjob_lateral.sh"
+
+fi

examples/parallel/KTH_PDC/lateral_inhibition/sten_4/Dardel_simulate_lateral.job

@@ -0,0 +1,95 @@
+#!/bin/bash -l
+#SBATCH --partition=main
+#SBATCH -o log/Simulate-%j-output.txt
+#SBATCH -e log/Simulate-%j-error.txt
+#SBATCH -t 8:59:00
+#SBATCH --time-min=6:59:00
+#SBATCH -J Simulate
+#SBATCH -A naiss2023-5-231
+#SBATCH --nodes=60
+#SBATCH --tasks-per-node=50
+#SBATCH --mail-type=ALL
+
+# 2024-02-16: 40 nodes worked, had 28% free... trying increasing to 45 nodes
+# 2024-03-01: 60 nodes, 50 workers worked...
+
+# You need to point this as the directory where you created the network in
+#NETWORK_DIR=/cfs/klemming/home/${USER:0:1}/$USER/Snudda/examples/parallel/KTH_PDC/networks/test_10k
+NETWORK_DIR=../networks/sten_4
+
+SIMULATION_CONFIG_WITH_SYNAPSES=experiment_config_sten_4-with-synapses-A.json
+SIMULATION_CONFIG_NO_SYNAPSES=experiment_config_sten_4-no-synapses-A.json
+
+
+# NETWORK_WITH_SYNAPSES_OUTPUT=$NETWORK_DIR/simulation/output-with-synapses-sten_1.hdf5
+# NETWORK_NO_SYNAPSES_OUTPUT=$NETWORK_DIR/simulation/output-no-synapses-sten_1.hdf5
+
+
+export N_WORKERS=$SLURM_NTASKS
+
+module load snic-env
+source $HOME/Snudda/snudda_env/bin/activate
+SNUDDA_DIR=/cfs/klemming/home/"${USER:0:1}"/$USER/Snudda
+
+# If the BasalGangliaData directory exists, then use that for our data
+if [[ -d "/cfs/klemming/home/${USER:0:1}/$USER/BasalGangliaData/data" ]]; then
+    export SNUDDA_DATA="/cfs/klemming/home/${USER:0:1}/$USER/BasalGangliaData/data"
+    echo "Setting SNUDDA_DATA to $SNUDDA_DATA"
+    rm mechanisms
+    ln -s $SNUDDA_DATA/neurons/mechanisms/ mechanisms
+else
+    echo "SNUDDA_DATA environment variable not changed (may be empty): $SNUDDA_DATA"
+    rm mechanisms
+    ln -s ../../../../snudda/data/neurons/mechanisms/
+fi
+
+
+NETWORK_INFO_FILE=$NETWORK_DIR/network-synapses.hdf5
+# NETWORK_INFO_FILE=$NETWORK_DIR/network-synapses-minimal.hdf5
+NETWORK_INPUT_FILE=$NETWORK_DIR/input-spikes.hdf5
+# NETWORK_VOLTAGE_FILE=$NETWORK_DIR/simulation/voltage-trace-${SLURM_JOBID}.txt
+
+
+
+echo "Network dir: "$NETWORK_DIR
+
+export PATH=$SNUDDA_DIR/snudda_env/bin/:$PATH
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CRAY_LD_LIBRARY_PATH
+export PYTHONPATH=$SNUDDA_DIR/snudda_env/lib/python3.9/
+
+##############
+
+rm -r x86_64
+
+export CXX=CC
+export CC=cc
+export FC=ftn
+export MPICC=cc
+export MPICXX=CC
+
+CC --version
+
+echo "About to run nrnivmodl"
+which nrnivmodl
+
+# srun -n nrnivmodl mechanisms/
+
+srun -n 1 nrnivmodl -incflags "-lltdl=/usr/lib64/libltdl.so.7 -lreadline=/lib64/libreadline.so.7 -lncurses=/lib64/libncurses.so.6.1" -loadflags "-DLTDL_LIBRARY=/usr/lib64/libltdl.so.7 -DREADLINE_LIBRARY=/lib64/libreadline.so.7 -DNCURSES_LIBRARY=/lib64/libncurses.so.6.1" mechanisms/
+
+# GJ disabled
+# srun -n $N_WORKERS $SNUDDA_DIR/examples/parallel/x86_64/special -mpi -python $SNUDDA_DIR/simulate/simulate.py $NETWORK_INFO_FILE $NETWORK_INPUT_FILE --disableGJ --time 3.5 --voltOut $NETWORK_VOLTAGE_FILE
+
+# GJ active
+# srun -n $N_WORKERS $SNUDDA_DIR/examples/parallel/KTH_PDC/x86_64/special -mpi -python $SNUDDA_DIR/snudda/simulate/simulate.py $NETWORK_INFO_FILE $NETWORK_INPUT_FILE --time 18 --outputFile $NETWORK_WITH_SYNAPSES_OUTPUT
+
+# srun -n $N_WORKERS $SNUDDA_DIR/examples/parallel/KTH_PDC/x86_64/special -mpi -python $SNUDDA_DIR/snudda/simulate/simulate.py $NETWORK_INFO_FILE $NETWORK_INPUT_FILE --time 18 --disableSyn --outputFile $NETWORK_NO_SYNAPSES_OUTPUT
+
+
+# Changed to using the simulation_config
+
+srun -n $N_WORKERS $SNUDDA_DIR/examples/parallel/KTH_PDC/x86_64/special -mpi -python $SNUDDA_DIR/snudda/simulate/simulate.py dummy_file dummy_file --simulation_config $SIMULATION_CONFIG_WITH_SYNAPSES
+
+srun -n $N_WORKERS $SNUDDA_DIR/examples/parallel/KTH_PDC/x86_64/special -mpi -python $SNUDDA_DIR/snudda/simulate/simulate.py dummy_file dummy_file --simulation_config $SIMULATION_CONFIG_NO_SYNAPSES
+
+
+# srun -n $N_WORKERS $SNUDDA_DIR/examples/parallel/KTH_PDC/x86_64/special -mpi -python $SNUDDA_DIR/snudda/simulate/simulate.py $NETWORK_INFO_FILE $NETWORK_INPUT_FILE --time 5 --noVolt

examples/parallel/KTH_PDC/lateral_inhibition/sten_4/experiment_config_sten_4-no-synapses-A.json

@@ -0,0 +1,11 @@
+{
+    "network_file": "../networks/sten_4/network-synapses.hdf5",
+    "input_file": "../networks/sten_4/input-spikes.hdf5",
+    "output_file": "../networks/sten_4/simulation/output-no-synapses-sten_4-A.hdf5",
+    "log_file": "../networks/sten_4/log/network-simulation-no-synapses-log-4A.txt",
+    "sample_dt": 0.01,
+    "time": 18.0,
+    "disable_synapses": true,
+    "record_all_soma": true,
+    "record_all_compartments": [0, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500]
+}

examples/parallel/KTH_PDC/lateral_inhibition/sten_4/experiment_config_sten_4-with-synapses-A.json

@@ -0,0 +1,10 @@
+{
+    "network_file": "../networks/sten_4/network-synapses.hdf5",
+    "input_file": "../networks/sten_4/input-spikes.hdf5",
+    "output_file": "../networks/sten_4/simulation/output-with-synapses-sten_4-A.hdf5",
+    "log_file": "../networks/sten_4/log/network-simulation-with-synapses-log-4A.txt",
+    "sample_dt": 0.01,
+    "time": 18.0,
+    "record_all_soma": true,
+    "record_all_compartments": [0, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 5500, 6000, 6500, 7000, 7500]
+}