昇腾 npu 的容器内算力和显存使用率，没有执行查询，而是直接返回0.

[tianzhiqiang3]

最新版本的 hami-webui，依然存在昇腾 npu 环境的容器内算力和显存使用率，没有执行查询，而是直接返回0的问题，请问什么时候解决？

// 任务算力利用率
func (s *MetricsGenerator) taskCoreUsed(ctx context.Context, provider, namespace, pod, container, podUUID, deviceUUID, hostname string, deviceIndex int) (float32, error) {
	query := ""
	switch provider {
	case biz.NvidiaGPUDevice:
		//query = fmt.Sprintf("avg(Device_utilization_desc_of_container{deviceuuid=\"%s\", podnamespace=\"%s\", podname=\"%s\", ctrname=\"%s\"})", deviceUUID, namespace, pod, container)
		//		queryTemplate := `last_over_time((Device_utilization_desc_of_container{deviceuuid="%s", podnamespace="%s", podname="%s", ctrname="%s"} != 0)[1m:])
		//or
		//last_over_time(Device_utilization_desc_of_container{deviceuuid="%s", podnamespace="%s", podname="%s", ctrname="%s"}[1m:])`
		//		query = fmt.Sprintf(queryTemplate, deviceUUID, namespace, pod, container, deviceUUID, namespace, pod, container)
		queryTemplate := fmt.Sprintf("Device_utilization_desc_of_container{deviceuuid=\"%s\", podnamespace=\"%s\", podname=\"%s\", ctrname=\"%s\"}", deviceUUID, namespace, pod, container)
		query = fmt.Sprintf("sum_over_time(%s[1m]) == 0 or (sum_over_time(%s[10m:]) / count_over_time(( %s !=0)[10m:])) ", queryTemplate, queryTemplate, queryTemplate)
		//query = queryTemplate
	case biz.CambriconGPUDevice:
		query = fmt.Sprintf("avg(mlu_utilization * on(uuid) group_right mlu_container{namespace=\"%s\",pod=\"%s\",container=\"%s\",type=\"mlu370.smlu.vcore\"})", namespace, pod, container)
	case biz.AscendGPUDevice:
		return 0, nil
	case biz.HygonGPUDevice:
		query = fmt.Sprintf("avg(vdcu_percent{pod_uuid=\"%s\", container_name=\"%s\"})", podUUID, container)
	case biz.MetaxGPUDevice, metax.MetaxGPUDevice:
		query = fmt.Sprintf("avg(mx_gpu_usage{uuid=\"%s\", exported_namespace=\"%s\", exported_pod=\"%s\", exported_container=\"%s\"})", deviceUUID, namespace, pod, container)
	case metax.MetaxSGPUDevice:
		query = fmt.Sprintf("avg(mx_sgpu_usage{Hostname=\"%s\", deviceId=\"%d\", exported_namespace=\"%s\", exported_pod=\"%s\", exported_container=\"%s\"})",
			hostname, deviceIndex, namespace, pod, container)
	default:
		return 0, errors.New("provider not exists")
	}
	return s.queryInstantVal(ctx, query)
}
// 任务显存使用量
func (s *MetricsGenerator) taskMemoryUsed(ctx context.Context, provider, namespace, pod, container, podUUID, deviceUUID, hostname string, deviceIndex int) (float32, error) {
	query := ""
	switch provider {
	case biz.NvidiaGPUDevice:
		query = fmt.Sprintf("avg(vGPU_device_memory_usage_in_bytes{deviceuuid=\"%s\", podnamespace=\"%s\", podname=\"%s\", ctrname=\"%s\"})", deviceUUID, namespace, pod, container)
	case biz.CambriconGPUDevice:
		query = fmt.Sprintf("avg(mlu_memory_utilization * on(uuid) group_right mlu_container{namespace=\"%s\",pod=\"%s\",container=\"%s\",type=\"mlu370.smlu.vmemory\"})", namespace, pod, container)
	case biz.AscendGPUDevice:
		return 0, nil
	case biz.HygonGPUDevice:
		query = fmt.Sprintf("avg(vdcu_usage_memory_size{pod_uuid=\"%s\", container_name=\"%s\"})", podUUID, container)
	case metax.MetaxGPUDevice:
		query = fmt.Sprintf("avg(mx_memory_used{uuid=\"%s\", exported_namespace=\"%s\", exported_pod=\"%s\", exported_container=\"%s\", type=\"vram\"})", deviceUUID, namespace, pod, container)
	case metax.MetaxSGPUDevice:
		query = fmt.Sprintf("avg(mx_sgpu_used_memory{Hostname=\"%s\", deviceId=\"%d\", exported_namespace=\"%s\", exported_pod=\"%s\", exported_container=\"%s\"})",
			hostname, deviceIndex, namespace, pod, container)
	default:
		return 0, errors.New("provider not exists")
	}
	return s.queryInstantVal(ctx, query)
}