New script to launch slurm array

UD_any/launchBatches.py

 #! /usr/bin/env python3
 
-import sys
 import os
 import subprocess
-import time
+import sys
+sys.path.insert(1, '../scripts')
+from launchSlurmArray import launchSlurmArray
 
 ###############################################################################
 def printUsageAndExit() :
@@ -19,121 +20,121 @@ def prepareExperiment(lang, template, expName) :
 
 ###############################################################################
 def addNamesAndCommandsTrain(names, commands, mode, expName, arguments, seed) :
-  names.append("\""+expName+"\"")
+  names.append(expName)
   
-  commands.append("'"+"./train.sh {} bin/{} {} --silent --seed {}".format(mode, expName,arguments,seed)+"'")
+  commands.append("./train.sh {} bin/{} {} --silent --seed {}".format(mode, expName,arguments,seed))
 ###############################################################################
 
 ###############################################################################
 def addNamesAndCommandsDecode(names, commands, mode, expName) :
-  names.append("\""+expName+"\"")
+  names.append(expName)
   
-  commands.append("\""+"./evaluate.sh {} bin/{} --silent".format(mode, expName)+"\"")
+  commands.append("./evaluate.sh {} bin/{} --silent".format(mode, expName))
 ###############################################################################
 
 ###############################################################################
-def launchArray(names, commands, mode, jobName, device, nbHours, limit) :
-  filename = "{}.{}.slurm".format(mode,jobName)
-  sFile = open(filename, "w")
-
-  hostname = os.getenv("HOSTNAME")
-
-  commandList = " ".join(commands)
-
-  if hostname == "jean-zay1" :
-    print("""#! /usr/bin/env bash
-
-#SBATCH --array=0-{}%{}
-#SBATCH --job-name={}:{}
-#SBATCH --output=%A_%a.out
-#SBATCH --error=%A_%a.err
-#SBATCH --open-mode=append
-#SBATCH --ntasks=1
-#SBATCH --cpus-per-task=10
-#SBATCH --gres=gpu:1
-#SBATCH --hint=nomultithread
-#SBATCH --partition=gpu_p1
-#SBATCH --qos={}
-#SBATCH --time={}:00:00
-
-module purge
-module load gcc/9.1.0
-module load python/3.7.5
-
-names=({})
-commands=({})
-
-newOut=${{names[$SLURM_ARRAY_TASK_ID]}}".stdout"
-newErr=${{names[$SLURM_ARRAY_TASK_ID]}}".stderr"
-oldOut=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".out"
-oldErr=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".err"
-tmpFile=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".tmp"
-
-touch $newOut
-
-cp $newOut $tmpFile
-mv $oldOut $newOut
-cat $tmpFile >> $newOut
-
-touch $newErr
-
-cp $newErr $tmpFile
-mv $oldErr $newErr
-cat $tmpFile >> $newErr
-
-rm $tmpFile
-
-eval "${{commands[$SLURM_ARRAY_TASK_ID]}}"
-""".format(len(names)-1, limit, mode, jobName, "qos_gpu-t4" if nbHours > 20 else "qos_gpu-t3", nbHours, " ".join(names), " ".join(commands)), file=sFile)
-    sFile.close()
-  elif hostname == "sms.liscluster" :
-    print('''#! /usr/bin/env bash
-
-#SBATCH --array=0-{}%{}
-#SBATCH --job-name={}:{}
-#SBATCH --output=%A_%a.out
-#SBATCH --error=%A_%a.err
-#SBATCH --open-mode=append
-#SBATCH --ntasks=1
-#SBATCH --cpus-per-task=1
-#SBATCH --hint=nomultithread
-#SBATCH --partition={}
-#SBATCH --exclude=sensei1,lifnode1,asfalda1
-#SBATCH --time={}:00:00
-
-module purge
-
-names=({})
-commands=({})
-
-newOut=${{names[$SLURM_ARRAY_TASK_ID]}}".stdout"
-newErr=${{names[$SLURM_ARRAY_TASK_ID]}}".stderr"
-oldOut=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".out"
-oldErr=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".err"
-tmpFile=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".tmp"
-
-touch $newOut
-
-cp $newOut $tmpFile
-mv $oldOut $newOut
-cat $tmpFile >> $newOut
-
-touch $newErr
-
-cp $newErr $tmpFile
-mv $oldErr $newErr
-cat $tmpFile >> $newErr
-
-rm $tmpFile
-
-eval "${{commands[$SLURM_ARRAY_TASK_ID]}}"
-'''.format(len(names)-1, limit, mode, jobName, "cpu" if device == "cpu" else "gpu\n#SBATCH --gres=gpu", nbHours, " ".join(names), commandList), file=sFile)
-    sFile.close()
-  else :
-    print("ERROR : Unknown hostname \'%s\'"%hostname)
-    exit(1)
-
-  subprocess.Popen("sbatch {}".format(filename), shell=True).wait()
+#def launchArray(names, commands, mode, jobName, device, nbHours, limit) :
+#  filename = "{}.{}.slurm".format(mode,jobName)
+#  sFile = open(filename, "w")
+#
+#  hostname = os.getenv("HOSTNAME")
+#
+#  commandList = " ".join(commands)
+#
+#  if hostname == "jean-zay1" :
+#    print("""#! /usr/bin/env bash
+#
+##SBATCH --array=0-{}%{}
+##SBATCH --job-name={}:{}
+##SBATCH --output=%A_%a.out
+##SBATCH --error=%A_%a.err
+##SBATCH --open-mode=append
+##SBATCH --ntasks=1
+##SBATCH --cpus-per-task=10
+##SBATCH --gres=gpu:1
+##SBATCH --hint=nomultithread
+##SBATCH --partition=gpu_p1
+##SBATCH --qos={}
+##SBATCH --time={}:00:00
+#
+#module purge
+#module load gcc/9.1.0
+#module load python/3.7.5
+#
+#names=({})
+#commands=({})
+#
+#newOut=${{names[$SLURM_ARRAY_TASK_ID]}}".stdout"
+#newErr=${{names[$SLURM_ARRAY_TASK_ID]}}".stderr"
+#oldOut=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".out"
+#oldErr=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".err"
+#tmpFile=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".tmp"
+#
+#touch $newOut
+#
+#cp $newOut $tmpFile
+#mv $oldOut $newOut
+#cat $tmpFile >> $newOut
+#
+#touch $newErr
+#
+#cp $newErr $tmpFile
+#mv $oldErr $newErr
+#cat $tmpFile >> $newErr
+#
+#rm $tmpFile
+#
+#eval "${{commands[$SLURM_ARRAY_TASK_ID]}}"
+#""".format(len(names)-1, limit, mode, jobName, "qos_gpu-t4" if nbHours > 20 else "qos_gpu-t3", nbHours, " ".join(names), " ".join(commands)), file=sFile)
+#    sFile.close()
+#  elif hostname == "sms.liscluster" :
+#    print('''#! /usr/bin/env bash
+#
+##SBATCH --array=0-{}%{}
+##SBATCH --job-name={}:{}
+##SBATCH --output=%A_%a.out
+##SBATCH --error=%A_%a.err
+##SBATCH --open-mode=append
+##SBATCH --ntasks=1
+##SBATCH --cpus-per-task=1
+##SBATCH --hint=nomultithread
+##SBATCH --partition={}
+##SBATCH --exclude=sensei1,lifnode1,asfalda1
+##SBATCH --time={}:00:00
+#
+#module purge
+#
+#names=({})
+#commands=({})
+#
+#newOut=${{names[$SLURM_ARRAY_TASK_ID]}}".stdout"
+#newErr=${{names[$SLURM_ARRAY_TASK_ID]}}".stderr"
+#oldOut=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".out"
+#oldErr=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".err"
+#tmpFile=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".tmp"
+#
+#touch $newOut
+#
+#cp $newOut $tmpFile
+#mv $oldOut $newOut
+#cat $tmpFile >> $newOut
+#
+#touch $newErr
+#
+#cp $newErr $tmpFile
+#mv $oldErr $newErr
+#cat $tmpFile >> $newErr
+#
+#rm $tmpFile
+#
+#eval "${{commands[$SLURM_ARRAY_TASK_ID]}}"
+#'''.format(len(names)-1, limit, mode, jobName, "cpu" if device == "cpu" else "gpu\n#SBATCH --gres=gpu", nbHours, " ".join(names), commandList), file=sFile)
+#    sFile.close()
+#  else :
+#    print("ERROR : Unknown hostname \'%s\'"%hostname)
+#    exit(1)
+#
+#  subprocess.Popen("sbatch {}".format(filename), shell=True).wait()
 ###############################################################################
 
 ###############################################################################
@@ -168,6 +169,6 @@ if __name__ == "__main__" :
         else :
           addNamesAndCommandsDecode(names, commands, xp['mode'],xp['expName'])
 
-  launchArray(names, commands, mode, name, device, nbHours, limit)
+  launchSlurmArray(names, commands, name, device, nbHours, limit)
 ###############################################################################
 

scripts/launchSlurmArray.py

+import os
+import subprocess
+
+def launchSlurmArray(names, commands, jobName, device, nbHours, limit) :
+  commands = ["'%s'"%s for s in commands]
+  names = ["'%s'"%s for s in names]
+
+  filename = "{}.slurm".format(jobName)
+  sFile = open(filename, "w")
+
+  hostname = os.getenv("HOSTNAME")
+
+  commandList = " ".join(commands)
+
+  if hostname == "jean-zay1" :
+    print("""#! /usr/bin/env bash
+
+#SBATCH --array=0-{}%{}
+#SBATCH --job-name={}
+#SBATCH --output=%A_%a.out
+#SBATCH --error=%A_%a.err
+#SBATCH --open-mode=append
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=10
+#SBATCH --gres=gpu:1
+#SBATCH --hint=nomultithread
+#SBATCH --partition=gpu_p1
+#SBATCH --qos={}
+#SBATCH --time={}:00:00
+
+module purge
+module load gcc/9.1.0
+module load python/3.7.5
+
+names=({})
+commands=({})
+
+newOut=${{names[$SLURM_ARRAY_TASK_ID]}}".stdout"
+newErr=${{names[$SLURM_ARRAY_TASK_ID]}}".stderr"
+oldOut=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".out"
+oldErr=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".err"
+tmpFile=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".tmp"
+
+touch $newOut
+
+cp $newOut $tmpFile
+mv $oldOut $newOut
+cat $tmpFile >> $newOut
+
+touch $newErr
+
+cp $newErr $tmpFile
+mv $oldErr $newErr
+cat $tmpFile >> $newErr
+
+rm $tmpFile
+
+eval "${{commands[$SLURM_ARRAY_TASK_ID]}}"
+""".format(len(names)-1, limit, jobName, "qos_gpu-t4" if nbHours > 20 else "qos_gpu-t3", nbHours, " ".join(names), " ".join(commands)), file=sFile)
+    sFile.close()
+  elif hostname == "sms.liscluster" :
+    print('''#! /usr/bin/env bash
+
+#SBATCH --array=0-{}%{}
+#SBATCH --job-name={}
+#SBATCH --output=%A_%a.out
+#SBATCH --error=%A_%a.err
+#SBATCH --open-mode=append
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=1
+#SBATCH --hint=nomultithread
+#SBATCH --partition={}
+#SBATCH --exclude=sensei1,lifnode1,asfalda1
+#SBATCH --time={}:00:00
+
+module purge
+
+names=({})
+commands=({})
+
+newOut=${{names[$SLURM_ARRAY_TASK_ID]}}".stdout"
+newErr=${{names[$SLURM_ARRAY_TASK_ID]}}".stderr"
+oldOut=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".out"
+oldErr=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".err"
+tmpFile=$SLURM_ARRAY_JOB_ID"_"$SLURM_ARRAY_TASK_ID".tmp"
+
+touch $newOut
+
+cp $newOut $tmpFile
+mv $oldOut $newOut
+cat $tmpFile >> $newOut
+
+touch $newErr
+
+cp $newErr $tmpFile
+mv $oldErr $newErr
+cat $tmpFile >> $newErr
+
+rm $tmpFile
+
+eval "${{commands[$SLURM_ARRAY_TASK_ID]}}"
+'''.format(len(names)-1, limit, jobName, "cpu" if device == "cpu" else "gpu\n#SBATCH --gres=gpu", nbHours, " ".join(names), commandList), file=sFile)
+    sFile.close()
+  else :
+    print("ERROR : Unknown hostname \'%s\'"%hostname)
+    exit(1)
+
+  subprocess.Popen("sbatch {}".format(filename), shell=True).wait()
+