AI-Hypercomputer / AI-Hypercomputer/xpk
memory and cpu of Pathways cluster kueue should be configurable
Aperta
enhancement
release-improvements
- Lingua principale
- Python
- Stelle
- 193
- Fork
- 94
- Merge medio
- 24m
- PR unite (30g)
- 1
Descrizione
https://github.com/AI-Hypercomputer/xpk/blob/67b9ac6a424753869d21d22b0bbac4bc6332017e/src/xpk/core/kueue_manager.py#L390-L391
Currently the nominal quota of memory and cpu of Pathways cluster kueue is hardcoded. These values are preventing more Pathways workloads from being scheduled even if the cluster has enough memory and CPUs. We should make these values configurable.
Guida per i contributori
Apri la guida per i contributori
Valutazione
Questa issue non è ancora stata valutata.