3.1.6. PBS: options for PBS job control¶
Option aliases:
site-needed → site_needed
Recognized options:
- A: {
None} |object account name(s) or number(s)
- S: {
'/bin/bash'} |str shell to execute PBS job
- W: {
''} |str PBS W setting, usually for setting group
- aoe: {
None} |str architecture operating environment
- e: {
None} |object explicit STDERR file name
- j: {
'oe'} |str PBS ‘join’ setting
- k: {
None} |str PBS ‘keep’ setting; live updates to STD(OUT|ERR)
- mem: {
None} |int Requested cpu memory (on gpu node)
- model: {
None} |str model type/architecture
- mpiprocs: {
None} |int number of MPI processes per node
- ncpus: {
None} |int number of cores (roughly CPUs) per node
- ngpus: {
None} |int number of gpus per node
- o: {
None} |object explicit STDOUT file name
- ompthreads: {
None} |int number of OMP threads
- p: {
None} |object PBS priority
- place: {
'excl'} |'shared' Share or exclusive access to requested gpu nodes
- q: {
'normal'} |str PBS queue name
- r: {
'n'} |'y' rerun-able setting
- select: {
1} |int number of nodes
- site_needed: {
None} |list[str] list of manually requested hard drives to mount
- walltime: {
'8:00:00'} |str maximum job wall time
