gnn_tracking_hpo.slurmcontrol#

Module Contents#

Classes#

SlurmControlAction

Actions for slurm control to take

SlurmControl

Functions#

get_slurm_job_id(→ str)

Get SLURM job ID from environment variables. An empty string is

_parse_slurm_time_str(→ datetime.timedelta)

parse_slurm_time_str(→ datetime.timedelta)

Parse a time string from SLURM into a timedelta

get_slurm_remaining_minutes(→ int)

How many more minutes does the SLURM job have to run?

kill_slurm_job(job_id)

Kill a SLURM job. No error is raised should the killing fail.

class gnn_tracking_hpo.slurmcontrol.SlurmControlAction#

Bases: enum.Enum

Actions for slurm control to take

KILL_NODE#
WAIT#
gnn_tracking_hpo.slurmcontrol.get_slurm_job_id() str#

Get SLURM job ID from environment variables. An empty string is returned if the job is not running on SLURM.

gnn_tracking_hpo.slurmcontrol._parse_slurm_time_str(time_str: str) datetime.timedelta#
gnn_tracking_hpo.slurmcontrol.parse_slurm_time_str(time_str: str) datetime.timedelta#

Parse a time string from SLURM into a timedelta

gnn_tracking_hpo.slurmcontrol.get_slurm_remaining_minutes(job_id: str) int#

How many more minutes does the SLURM job have to run?

gnn_tracking_hpo.slurmcontrol.kill_slurm_job(job_id: str)#

Kill a SLURM job. No error is raised should the killing fail.

class gnn_tracking_hpo.slurmcontrol.SlurmControl#
_refresh()#

Reload the config.

_get_actions(dispatcher_id: str) list[SlurmControlAction]#
_handle_action(action: SlurmControlAction) bool#

Handles action. Returns True if we should refresh the config and check if everything has been resolved afterwards.

__call__(*, dispatcher_id: str) None#