apiVersion: apps/v1
kind: Deployment
metadata:
name: resnetinfer1-1-deploy
labels:
app: infers
spec:
replicas: 1
selector:
matchLabels:
app: infers
template:
metadata:
labels:
app: infers
host-arch: huawei-arm # Select the os arch. If the os arch is arm, change it to huawei-x86.
ring-controller.atlas: ascend-npu # Add this Label, used to check job kind.
annotations:
sp-block: "8"
ra-block: "8"
huawei.com/schedule_policy: "chip8-node8-ra64-sp"
spec:
schedulerName: volcano
automountServiceAccountToken: false
containers:
- image: ubuntu-infer:v1 # Inference image name
imagePullPolicy: IfNotPresent
name: resnet50infer
env:
# ASCEND_VISIBLE_DEVICES env variable is used by ascend-docker-runtime when in the whole card scheduling scene with volcano scheduler. please delete it when in the static or dynamic vNPU scheduling scene or without volcano.
- name: ASCEND_VISIBLE_DEVICES
valueFrom:
fieldRef:
fieldPath: metadata.annotations['huawei.com/npu']
resources:
requests:
huawei.com/npu: 8
limits:
huawei.com/npu: 8 # The value should be the same as that of requests .
volumeMounts:
- name: slog
mountPath: /var/log/npu/conf/slog/ #Log path
- name: localtime #The container time must be the same as the host time.
mountPath: /etc/localtime
volumes:
- name: slog
hostPath:
path: /var/log/npu/conf/slog/
- name: localtime
hostPath:
path: /etc/localtime