New Features
Omnia now executes exclusively within a virtual environment created by the
prereq.shscriptPython version upgraded to 3.11 (Previously 3.9)
Ansible version upgraded to 9.5.1 (Previously 7.7.0)
Kubernetes version upgraded to 1.29.5 (Previously 1.26.12)
Pre-enablement for Intel Gaudi 3 accelerators:
Software stack installation (See the support matrix for the supported Intel firmware version)
Inventory tagging for the Gaudi accelerators (
compute_gpu_intel)Monitoring for the Gaudi accelerators via:
Omnia telemetry
iDRAC telemetry
Kubernetes telemetry via Prometheus exporter
Visualization of the Kubernetes telemetry and Intel Gaudi accelerator metrics using Grafana
AI tools enablement:
DeepSpeed
Kubeflow
vLLM
Sample playbook for a pre-trained Generative AI model - Llama 3.1
CSI drivers for Kubernetes to access PowerScale storage with an option to enable the SmartConnect feature (without SSL certificates)
Added support for NVIDIA container toolkit for NVIDIA accelerators in a Kubernetes cluster
Added support for corporate proxy on RHEL, Rocky Linux, and Ubuntu clusters
Set OS Kernel command-line parameters and/or configure additional NICs on the nodes using a single playbook
The internal OpenLDAP server can now be configured as a proxy server
If you have any feedback about Omnia documentation, please reach out at omnia.readme@dell.com.