Job#

class nnodes.job.DTN(job: dict, state: list)#: Oak Ridge National Lab Data Transfer Node.

class nnodes.job.Job(job: dict, state: list)#

Base class for clusters.

property aborted#: Any task failed twice during execution.

create(dst: Optional[str] = None)#: Creates a directory as job workspace.

property failed#: Any task failed during execution.

property inqueue: bool#: Job is allocated from scheduler (enables automatic requeue and mpiexec timeout).

mpiexec(cmd: str, nprocs: int, cpus_per_proc: int = 1, gpus_per_proc: int = 0, mps: Optional[int] = None) → str#: Run a MPI task.

property paused#: Job paused due to insuffcient time.

property remaining: float#: Remaining walltime in minutes.

requeue()#: Resubmit current job.

write(cmd: str, dst: str)#: Write job submission script to target directory.

class nnodes.job.LSF(job: dict, state: list)#

LSF-based cluster.

property inqueue#: Job is allocated from scheduler (enables automatic requeue and mpiexec timeout).

mpiexec(cmd: str, nprocs: int, cpus_per_proc: int = 1, gpus_per_proc: int = 0, mps: Optional[int] = None)#: Get the command to call MPI.

requeue()#: Run current job again.

write(cmd, dst)#: Write job submission script to target directory.

class nnodes.job.Local(job: dict, state: list)#: Local computer using multiprocessing instead of MPI.

class nnodes.job.LocalMPI(job: dict, state: list)#

Local computer with MPI installed.

mpiexec(cmd: str, nprocs: int, *_)#: Get the command to call MPI.

class nnodes.job.Slurm(job: dict, state: list)#

Slurm-based cluster.

property inqueue#: Job is allocated from scheduler (enables automatic requeue and mpiexec timeout).

mpiexec(cmd: str, nprocs: int, cpus_per_proc: int = 1, gpus_per_proc: int = 0, mps: Optional[int] = None)#: Get the command to call MPI.

requeue()#: Run current job again.

write(cmd, dst)#: Write job submission script to target directory.

class nnodes.job.Summit(job: dict, state: list)#

class nnodes.job.Tiger(job: dict, state: list)#: Princeton TigerGPU

class nnodes.job.Traverse(job: dict, state: list)#: Princeton Traverse