Unanswered
On A Related Line But More Complicated: How Can We Ask The Autoscaler To Queue, Say, N Jobs On An N-Gpu Machine, Please? For Example, On Aws, Nvidia A100 Gpus Are Only Available On Instances With 8X A100, Which Is Overkill For A Single-Gpu Job, So Might A
@<1541954607595393024:profile|BattyCrocodile47>
Is that instance only able to handle one task at a time?
You could have multiple agents on the same machine, each one with its own dedicated GPU, but you will not be able to change the allocation (i.e. now I want 2 GPUs on one agent) without restarting the agents on the instance. In either case, this is for a "bare-metal" machine, and in the AWS autoscaler case, this goes under "dynamic" GPUs (see above)
165 Views
0
Answers
one year ago
one year ago