You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I would like to see an extra section right before https://docs.mila.quebec/Userguide.html#information-on-partitions-nodes where there is some kind of narrative to give an explanation about what might happen if someone was to submit 100 jobs on main in the middle of the night when the cluster is empty. And to long, for that matter. If those jobs run for 12 hours each, I'd like a more precise description of why that would make other researchers bring out the pitchforks in the morning.
I think the "Max Resource Usage" is also ambiguous in the sense that it's not clear that we're talking about resources for one job, or resources for all the jobs submitted. It could be read both ways at a glance.
I messed up in the cheat sheet when it came to saying that we have preemption on the Mila cluster. I thought it was just as simple as that, but it isn't. I still don't have a specific section of the documentation to reference if I want to direct people to more detailed explanations (hence this ticket).
The text was updated successfully, but these errors were encountered:
I would like to see an extra section right before
https://docs.mila.quebec/Userguide.html#information-on-partitions-nodes
where there is some kind of narrative to give an explanation about what might happen if someone was to submit 100 jobs onmain
in the middle of the night when the cluster is empty. And tolong
, for that matter. If those jobs run for 12 hours each, I'd like a more precise description of why that would make other researchers bring out the pitchforks in the morning.I think the "Max Resource Usage" is also ambiguous in the sense that it's not clear that we're talking about resources for one job, or resources for all the jobs submitted. It could be read both ways at a glance.
I messed up in the cheat sheet when it came to saying that we have preemption on the Mila cluster. I thought it was just as simple as that, but it isn't. I still don't have a specific section of the documentation to reference if I want to direct people to more detailed explanations (hence this ticket).
The text was updated successfully, but these errors were encountered: