版本:v2.4.1
Allocate device memory by percentage
Allocate a part of device memory by percentage to container
To allocate a certain size of GPU device memory by percentage, you need only to assign nvidia.com/gpumem-percentage besides nvidia.com/gpu.
apiVersion: v1
kind: Pod
metadata:
name: gpu-pod
spec:
containers:
- name: ubuntu-container
image: ubuntu:18.04
command: ["bash", "-c", "sleep 86400"]
resources:
limits:
nvidia.com/gpu: 2 # requesting 2 vGPUs
nvidia.com/gpumem-percentage: 50 # each vGPU requests 50% of device memory
备注
建议选择一种方式指定显存。若同时设置这两个参数,且 nvidia.com/gpumem 为大于 0 的整数,则显存申请量以 nvidia.com/gpumem 为准。详见显存分配说明。