Build Tensorflow Serving Job

This section introduces how to customly build a Tensorflow serving job.

Path

pkg/apis/serving.TFServingJobBuilder

Function

func NewTFServingJobBuilder() *TFServingJobBuilder

Parameters

TFServingJobBuilder has following functions to custom your Tensorflow serving job.

function	description	matches cli option
Name(name string) *TFServingJobBuilder	specify the job name	--name
Namespace(namespace string) *TFServingJobBuilder	specify the namespace	-n/--namespace
Command(args []string) *TFServingJobBuilder	specify the command	-
GPUCount(count int) *TFServingJobBuilder	specify the gpu count	--gpus
GPUMemory(memory int) *TFServingJobBuilder	specify the gpu memory(gpushare)	--gpumemory
Image(image string) *TFServingJobBuilder	specify the image	--image
ImagePullPolicy(policy string) *TFServingJobBuilder	specify the image pull policy	--image-pull-policy
CPU(cpu string) *TFServingJobBuilder	specify the cpu limitation of job	--cpu
Memory(memory string) *TFServingJobBuilder	specify the memory limitation	--memory
Envs(envs map[string]string) *TFServingJobBuilder	specify the envs of container	--env
Replicas(count int) *TFServingJobBuilder	specify the replicas	--replicas
Port(port int) *TFServingJobBuilder	specify the grpc service port	--port
RestfulPort(port int) *TFServingJobBuilder	specify the restful service port	--restfulPort
EnableIstio() *TFServingJobBuilder	enable istio	--enable-istio
ExposeService() *TFServingJobBuilder	expose service	--expose-service
Version(version string) *TFServingJobBuilder	specify the job version	--version
Tolerations(tolerations []string) *TFServingJobBuilder	specify the tolerations of node taints	--toleration
NodeSelectors(selectors map[string]string) *TFServingJobBuilder	specify the node selectors	--selector
Annotations(annotations map[string]string) *TFServingJobBuilder	specify the annotations	--annotation
Datas(volumes map[string]string) *TFServingJobBuilder	specify the pvc name which stores dataset	--data
DataDirs(volumes map[string]string) *TFServingJobBuilder	specify the host path which stores dataset	--data-dir
VersionPolicy(policy string) *TFServingJobBuilder	specify the version policy	--version-policy
ModelConfigFile(filePath string) *TFServingJobBuilder	specify the model configuration file	--modeConfigFile
ModelName(name string) *TFServingJobBuilder	specify the model name	--model-name
ModelPath(path string) *TFServingJobBuilder	specify the model path	--model-path
Build() (*Job, error)	build the job	-

Example

package main

import (
    "fmt"
    "time"

    "github.com/kubeflow/arena/pkg/apis/arenaclient"
    "github.com/kubeflow/arena/pkg/apis/serving"
    "github.com/kubeflow/arena/pkg/apis/types"
)

func main() {
    jobName := "test"
    jobType := types.TFServingJob
    // create arena client
    client, err := arenaclient.NewArenaClient(types.ArenaClientArgs{
        Kubeconfig: "",
        LogLevel:   "info",
        Namespace:  "default",
    })
    // create tfjob
    /* command:
    arena serve tensorflow \
    --name=mymnist1 \
    --model-name=mnist1  \
    --gpus=1   \
    --image=tensorflow/serving:latest-gpu \
    --data=tfmodel:/tfmodel \
    --model-path=/tfmodel/mnist \
    --versionPolicy=specific:1
    */
    job, err := serving.NewTFServingJobBuilder().
        Name(jobName).
        GPUCount(1).
        ModelName("mnist1").
        Image("tensorflow/serving:latest-gpu").
        Data(map[string]string{"tfmodel": "/tfmodel"}).
        ModelPath("/tfmodel/mnist").
        VersionPolicy("specific:1").Build()
    if err != nil {
        fmt.Printf("failed to build tensorflow serving job,reason: %v\n", err)
        return
    }
    // submit tfjob
    if err := client.Serving().Submit(job); err != nil {
        fmt.Printf("failed to submit job,reason: %v\n", err)
        return
    }
}