Skip to main content
Clusters are the compute backends that run your FlexAI workloads. Each cluster is provisioned with a specific accelerator type (for example, NVIDIA H100 or A100) and is made available to one or more organizations. By default, FlexAI selects a cluster automatically based on the --device-arch flag of the command you run. If you need to pin a workload to a specific cluster — for instance, to target a particular accelerator model — you can pass --affinity cluster=<cluster_name> to flexai training run, flexai training debug-ssh, or flexai inference serve. Use the flexai cluster set of subcommands to discover which clusters your organization can use.

Available subcommands