cellmap_flow.jobs.site
The numbers that are about the cluster, not about cellmap-flow.
Which GPU queues to offer, how many cores a job asks for, how long it may run, and how long to wait on LSF: all of it is Janelia’s, and it was spread through the launch code as literals. It lives here so that the reasons sit next to the numbers, and so that another site means another profile rather than a search for “gpu_h100”.
Only JANELIA exists. current_site() is the one place that will pick
between profiles once there is more than one.
Attributes
Classes
Functions
|
The profile of the cluster this process submits to. |
Module Contents
- class cellmap_flow.jobs.site.SiteProfile
- gpu_queues: Tuple[Tuple[str, str, str], ...]
- default_charge_group: str = 'cellmap'
- default_walltime: str = '08:00'
- server_cpus: int = 4
- worker_cpus: int = 12
- pending_fallback_seconds: int = 180
- startup_timeout_seconds: int = 300
- bsub_timeout_seconds: int = 30
- property default_queues: List[str]
every GPU queue offered. The dashboard’s GPU pickers start with all of them ticked, and a job goes to whichever has the most free GPUs (queues.candidates); a job pinned to one GPU model says so.
- Type:
The queues a job may run on when nothing says which
- cellmap_flow.jobs.site.JANELIA
- cellmap_flow.jobs.site.current_site() SiteProfile
The profile of the cluster this process submits to.