v2.11.0
Release highlights
- Whole-GPU allocations export metrics directly through NVML without mounting HAMi-core.
- Cluster Autoscaler support and real-usage monitoring for Flexible MIG workloads.
- Running vLLM with Remote CUDA.
- PodGroup scheduling, mutex policy combinations, and GPU/CPU NUMA alignment.
- AMD GPU and Ascend 950 support, plus libvnpu performance optimizations.
What's changed
See the v2.11.0 release notes for the complete list of features, fixes, and other changes.
Committers: Contributors
- moezdil (@moezdil)
- devGPP23 (@devGPP23)
- maishivamhoo123 (@maishivamhoo123)
- KunwarSidhu47 (@KunwarSidhu47)
- saiyam1814 (@saiyam1814)
- Kevindua26 (@Kevindua26)
- asadjan4611 (@asadjan4611)
- im-Toqeer-506 (@im-Toqeer-506)
- iemAnshuman (@iemAnshuman)
- haitwang-cloud (@haitwang-cloud)
- Gurkaran18 (@Gurkaran18)
- FouoF (@FouoF)
- Utkarsh56016 (@Utkarsh56016)
- archlitchi (@archlitchi)
- spencercjh (@spencercjh)
- togettoyou (@togettoyou)
- peachest (@peachest)
- curry30000 (@curry30000)
- wjluo (@wjluo)
- Eshiv-Pandey (@Eshiv-Pandey)
- Shouren (@Shouren)
- shkatara (@shkatara)
- sanket-jadhav-cse (@sanket-jadhav-cse)
- maverick-woo (@maverick-woo)
- usr-bin-ksh (@usr-bin-ksh)
- vesta-mlp (@vesta-mlp)
- adity1raut (@adity1raut)
- B-White84 (@B-White84)
- tsingst (@tsingst)
- falconlee236 (@falconlee236)
- heuristicwave (@heuristicwave)
- panpan0000 (@panpan0000)
- AyushSrivastava1818 (@AyushSrivastava1818)
- blackdragoon26 (@blackdragoon26)
- SK8-infi (@SK8-infi)
- IIITManjeet (@IIITManjeet)
- iron-prog (@iron-prog)
- Woolgathererer (@Woolgathererer)
- 0z5a (@0z5a)
- magic-peach (@magic-peach)
- rapsealk (@rapsealk)
- ManthanNimodiya (@ManthanNimodiya)
- diegodevxd (@diegodevxd)
- phantom5125 (@phantom5125)
- Samurai007AK (@Samurai007AK)
- loiht2 (@loiht2)
- 0718lol (@0718lol)
- Shenhan11 (@Shenhan11)
- dongjiang1989 (@dongjiang1989)
- kinglong08 (@kinglong08)
- karthikeyan-lgk (@karthikeyan-lgk)
Full Changelog: https://github.com/Project-HAMi/HAMi/compare/v2.10.0...v2.11.0
