Enable MetaX GPU sharing

HAMi now supports metax.com/gpu by implementing most device-sharing features as NVIDIA GPUs, device-sharing features include the following:

• GPU Sharing: Tasks can request a fraction of a GPU rather than the entire GPU card, allowing multiple tasks to share the same GPU.

• Device Memory Control: Tasks can be allocated a specific amount of GPU memory, with strict enforcement to ensure usage does not exceed the assigned limit.

• Compute Core Limiting: Tasks can be allocated a specific percentage of GPU compute cores (e.g., 60 means the container can use 60% of the GPU’s compute cores).

Prerequisites

• MetaX Driver >= 2.31.0
• MetaX GPU Operator >= 0.10.1
• Kubernetes >= 1.23

Enabling GPU-sharing Support

• Deploy MetaX GPU Operator on metax nodes (Please consult your device provider to acquire its package and document)

• Deploy HAMi using the online installation guide

Running MetaX jobs

MetaX GPUs can now be requested by a container using the metax-tech.com/sgpu resource type:

apiVersion: v1
kind: Pod
metadata:
  name: gpu-pod1
spec:
  containers:
    - name: ubuntu-container
      image: cr.metax-tech.com/public-ai-release/c500/colossalai:2.24.0.5-py38-ubuntu20.04-amd64
      imagePullPolicy: IfNotPresent
      command: ["sleep","infinity"]
      resources:
        limits:
          metax-tech.com/sgpu: 1 # requesting 1 GPU
          metax-tech.com/vcore: 60 # each GPU use 60% of total compute cores
          metax-tech.com/vmemory: 4 # each GPU require 4 GiB device memory

备注

You can find more examples in examples/sgpu folder.