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 understand that horovod can dynamically alter the number of worker nodes, which I think I should call it "node-level elasticity". I want to know if horovod can support "GPU-level" elasticity? For example, given a node with 2 GPUs, a elastic job can switch its worker number from 2 to 1 or 1 to 2. Besides, a elastic job can switch its worker number from 4 (across 2 hosts) to 1(1 worker in only 1 host). To sum up, I want to know if it is possible to achieve GPU-level granularity in horovod.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hi,
I understand that horovod can dynamically alter the number of worker nodes, which I think I should call it "node-level elasticity". I want to know if horovod can support "GPU-level" elasticity? For example, given a node with 2 GPUs, a elastic job can switch its worker number from 2 to 1 or 1 to 2. Besides, a elastic job can switch its worker number from 4 (across 2 hosts) to 1(1 worker in only 1 host). To sum up, I want to know if it is possible to achieve GPU-level granularity in horovod.
Thanks.
Beta Was this translation helpful? Give feedback.
All reactions