Device sharing
HAMi offers robust device-sharing capabilities, enabling multiple tasks to share the same GPU, MLU, or NPU device, maximizing the utilization of heterogeneous AI computing resources.
HAMi's device sharing enables:
- Multi-task sharing: The same device can be shared by multiple tasks, each utilizing only partial device.
- Device memory control: Memory can be allocated by MB or percentage.
- Use specific device: Allows selecting specific types of heterogeneous AI devices or targeting a device using its UUID.
- In-container hard limits: Imposes a hard limit on streaming multiprocessors.
- Non-intrusive control: Requires zero changes to existing programs while managing resource allocation.
- Dynamic MIG support: Supports on-the-fly MIG adjustments using mig-parted for dynamic-mig.
By leveraging these features, HAMi enhances resource efficiency and security in shared-device environments. Organizations can optimize their AI infrastructure for greater flexibility and performance while meeting diverse computational demands.