Repository navigation

Install each device stack only where its driver is claimed, adding CPU pools and AMD serving - #556

Merged

haarchri merged 19 commits into

modelplaneai:mainfrom

haarchri:feature/multi-accelerator-cpu-dra

Oct 30, 2026

Merged

Install each device stack only where its driver is claimed, adding CPU pools and AMD serving#556
haarchri merged 19 commits into
modelplaneai:mainfrom
haarchri:feature/multi-accelerator-cpu-dra

Commits on Oct 30, 2026
- Install each device stack only where its driver is claimed
  
  haarchri
  authored and
  negz
  committed
  845adc8
- Regenerate schemas and aicr stacks
  
  haarchri
  authored and
  negz
  committed
  f8f061a
- Fake a pool per device stack in the cloud e2e legs
  
  haarchri
  authored and
  negz
  committed
  2ddbafe
- Add Qwen2.5-0.5B recipes for an AMD GPU and for CPUs
  
  haarchri
  authored and
  negz
  committed
  41cd36f
- Observe only NVIDIA GPU pools with the cluster agent
  negz
  committed
  0848772
- Replace MetricMapping's cumulative boolean with a fromType enum
  negz
  committed
  65e4a25
- Note that CPU pools need Kubernetes 1.36 on GKE, AKS and Nebius
  negz
  committed
  1a0f54e
- Regenerate schemas
  negz
  committed
  d36e26c
- Rename the AMD Qwen2.5-0.5B recipe to qwen2.5-0.5b-amd
  negz
  committed
  8de9535
- Check compose-inference-cluster's supported drivers against the stack's
  negz
  committed
  42700fe
- Name every unsupported class, and gate only Existing out
  negz
  committed
  934fb2c
- Declare external DRA drivers in every cloud's module
  negz
  committed
  1664320
- Pin the AMD GPU operator's metrics exporter image
  negz
  committed
  be957b9
- Reject provisioned classes that claim dra.cpu beside another driver
  negz
  committed
  f93d6ee
- Regenerate schemas
  negz
  committed
  12bee5d
- Reject a dra.cpu claim that spans more than one device
  negz
  committed
  d02d198
- Declare CPU pool taints like GPU pool taints
  negz
  committed
  28aa28c
- Correct serving stack comments the device stacks made stale
  negz
  committed
  4dc4f9c
- Update the docs for per-driver device stacks, CPU pools and AMD
  negz
  committed
  4693b4b

Commits

Commits on Oct 30, 2026