Skip to main content

cscope gpus

cscope gpus shows information about available gpus. If its on a Slurm cluster it shows information about every partition, if not, it defaults to the local node.

$ cscope gpus --help
Usage: cscope gpus [OPTIONS]

  Show GPU information.

Options:
  --partition TEXT  Slurm partition name to filter queries (optional)
  --generations     Show only GPU generations
  --counts          Show only GPU counts by type
  --vendor          Show GPU vendor information
  --help            Show this message and exit.

Slurm​

$ cscope gpus
GPU Gen, GPU Count, GPU Vendor, Slurm Partition:
h100, 8, nvidia, h100
h200, 8, nvidia, h200

Slurm gpu generations​

$ cscope gpus --generations
GPU Gen, Slurm Partition:
h100, h100
h200, h200

Slurm gpu counts​

$ cscope gpus --counts
GPU Gen, GPU Count, Slurm Partition:
h100, 8, h100
h200, 8, h200

Slurm gpu vendors​

$ cscope gpus --vendor
GPU Vendors:
nvidia

Slurm Partition Filter​

You can also pass an optional partition arg: ... --partition=<partition-name>, if partition is passed it limits the cscope gpus for only that Slurm partition. It works with any of the other cscope gpus flags.

$ cscope gpus --partition=h100
GPU Gen, GPU Count, GPU Vendor, Slurm Partition:
h100, 8, nvidia, h100