All Tools
S
Dev ToolsFreeOpen Source
SLURM
Open-source cluster workload manager for HPC and GPU jobs
GPL-2.0
ABOUT
Shared GPU and CPU clusters need a scheduler that queues jobs, enforces limits, and accounts for usage. SLURM does that at HPC scale so ML teams run training and inference without a vendor workload manager.
INTEGRATION GUIDE
1. Queue GPU training jobs across a multi-node Linux cluster
2. Allocate CPUs and memory for batch inference or data prep
3. Account cluster usage by user, account, and partition
4. Run interactive debug allocations before a long training job
TAGS
slurmhpcschedulergpuclusterbatchmlopsworkload