Skip to main content
Version: v2.6.0

Allocate device core and memory resource

Allocate device core and memory to container​

To allocate a certain part of device core resource, you need only to assign the iluvatar.ai/vcuda-memory and iluvatar.ai/vcuda-core along with the number of cambricon MLUs you requested in the container using iluvatar.ai/vgpu

apiVersion: v1
kind: Pod
metadata:
  name: poddemo
spec:
  restartPolicy: Never
  containers:
  - name: poddemo
    image: harbor.4pd.io/vgpu/corex_transformers@sha256:36a01ec452e6ee63c7aa08bfa1fa16d469ad19cc1e6000cf120ada83e4ceec1e
    command: 
    - bash
    args:
    - -c
    - |
      set -ex
      echo "export LD_LIBRARY_PATH=/usr/local/corex/lib64:$LD_LIBRARY_PATH">> /root/.bashrc
      cp -f /usr/local/iluvatar/lib64/libcuda.* /usr/local/corex/lib64/
      cp -f /usr/local/iluvatar/lib64/libixml.* /usr/local/corex/lib64/
      source /root/.bashrc
      sleep 360000
    resources:
      requests:
        iluvatar.ai/vgpu: 2
      limits:
        iluvatar.ai/vgpu: 2