Nuclio Deploying Functions

Published by onesixx on

Builtin Model 모델 배포

Serverless tutorial by CVAThttps://opencv.github.io/cvat/docs/manual/advanced/serverless-tutorial/

CVAT에서 이미 만들어진 (serverless) Function을 바로 배포만 하면된다.

cvat 프로젝트를 생성

여기에 serverless function과 model을 저장

$ nuctl create project cvat

Deploy function

CVAT에서 제공하는 배포가능한 serverless function 및 모델 ( https://github.com/openvinotoolkit/cvat/tree/develop/serverless)

$ nuctnuctl deploy \\
  --project-name cvat \\
  --path     serverless/openvino/omz/public/mask_rcnn_inception_resnet_v2_atrous_coco/nuclio \\
  --volume   `pwd`/serverless/common:/opt/nuclio/common \\
  --platform local

로컬 환경에서 배포하는 경우에는 —-platform local 옵션을 추가
–http-trigger-service-type NodePort

# 다른 방법
$ serverless/deploy_cpu.sh \\
serverless/openvino/omz/public/mask_rcnn_inception_resnet_v2_atrous_coco/nuclio

$ serverless/deploy_gpu.sh \\
serverless/tensorflow/matterport/mask_rcnn
23.02.09 17:11:16.786                     nuctl (I) Deploying function {"name": ""}
23.02.09 17:11:16.787                     nuctl (I) Building {"builderKind": "docker", "versionInfo": "Label: 1.8.14, Git commit: cbb0774230996a3eb4621c1a2079e2317578005b, OS: linux, Arch: amd64, Go version: go1.17.8", "name": ""}
23.02.09 17:11:17.014                     nuctl (I) Staging files and preparing base images
23.02.09 17:11:17.015                     nuctl (W) Python 3.6 runtime is deprecated and will soon not be supported. Please migrate your code and use Python 3.7 runtime (`python:3.7`) or higher
23.02.09 17:11:17.015                     nuctl (I) Building processor image {"registryURL": "", "imageName": "cvat.openvino.omz.public.mask_rcnn_inception_resnet_v2_atrous_coco:latest"}
23.02.09 17:11:17.015     nuctl.platform.docker (I) Pulling image {"imageName": "quay.io/nuclio/handler-builder-python-onbuild:1.8.14-amd64"}
23.02.09 17:11:20.175     nuctl.platform.docker (I) Pulling image {"imageName": "quay.io/nuclio/uhttpc:0.0.1-amd64"}
23.02.09 17:11:24.389            nuctl.platform (I) Building docker image {"image": "cvat.openvino.omz.public.mask_rcnn_inception_resnet_v2_atrous_coco:latest"}
23.02.09 17:17:55.663            nuctl.platform (I) Pushing docker image into registry {"image": "cvat.openvino.omz.public.mask_rcnn_inception_resnet_v2_atrous_coco:latest", "registry": ""}
23.02.09 17:17:55.663            nuctl.platform (I) Docker image was successfully built and pushed into docker registry {"image": "cvat.openvino.omz.public.mask_rcnn_inception_resnet_v2_atrous_coco:latest"}
23.02.09 17:17:55.663                     nuctl (I) Build complete {"result": {"Image":"cvat.openvino.omz.public.mask_rcnn_inception_resnet_v2_atrous_coco:latest","UpdatedFunctionConfig":{"metadata":{"name":"openvino-mask-rcnn-inception-resnet-v2-atrous-coco","namespace":"nuclio","labels":{"nuclio.io/project-name":"cvat"},"annotations":{"framework":"openvino","name":"Mask RCNN","spec":"[\
  { \\"id\\": 1, \\"name\\": \\"person\\" },\
  { \\"id\\": 2, \\"name\\": \\"bicycle\\" },\
  { \\"id\\": 3, \\"name\\": \\"car\\" },\
  { \\"id\\": 4, \\"name\\": \\"motorcycle\\" },\
  { \\"id\\": 5, \\"name\\": \\"airplane\\" },\
  { \\"id\\": 6, \\"name\\": \\"bus\\" },\
  { \\"id\\": 7, \\"name\\": \\"train\\" },\
  { \\"id\\": 8, \\"name\\": \\"truck\\" },\
  { \\"id\\": 9, \\"name\\": \\"boat\\" },\
  { \\"id\\":10, \\"name\\": \\"traffic_light\\" },\
  { \\"id\\":11, \\"name\\": \\"fire_hydrant\\" },\
  { \\"id\\":13, \\"name\\": \\"stop_sign\\" },\
  { \\"id\\":14, \\"name\\": \\"parking_meter\\" },\
  { \\"id\\":15, \\"name\\": \\"bench\\" },\
  { \\"id\\":16, \\"name\\": \\"bird\\" },\
  { \\"id\\":17, \\"name\\": \\"cat\\" },\
  { \\"id\\":18, \\"name\\": \\"dog\\" },\
  { \\"id\\":19, \\"name\\": \\"horse\\" },\
  { \\"id\\":20, \\"name\\": \\"sheep\\" },\
  { \\"id\\":21, \\"name\\": \\"cow\\" },\
  { \\"id\\":22, \\"name\\": \\"elephant\\" },\
  { \\"id\\":23, \\"name\\": \\"bear\\" },\
  { \\"id\\":24, \\"name\\": \\"zebra\\" },\
  { \\"id\\":25, \\"name\\": \\"giraffe\\" },\
  { \\"id\\":27, \\"name\\": \\"backpack\\" },\
  { \\"id\\":28, \\"name\\": \\"umbrella\\" },\
  { \\"id\\":31, \\"name\\": \\"handbag\\" },\
  { \\"id\\":32, \\"name\\": \\"tie\\" },\
  { \\"id\\":33, \\"name\\": \\"suitcase\\" },\
  { \\"id\\":34, \\"name\\": \\"frisbee\\" },\
  { \\"id\\":35, \\"name\\": \\"skis\\" },\
  { \\"id\\":36, \\"name\\": \\"snowboard\\" },\
  { \\"id\\":37, \\"name\\": \\"sports_ball\\" },\
  { \\"id\\":38, \\"name\\": \\"kite\\" },\
  { \\"id\\":39, \\"name\\": \\"baseball_bat\\" },\
  { \\"id\\":40, \\"name\\": \\"baseball_glove\\" },\
  { \\"id\\":41, \\"name\\": \\"skateboard\\" },\
  { \\"id\\":42, \\"name\\": \\"surfboard\\" },\
  { \\"id\\":43, \\"name\\": \\"tennis_racket\\" },\
  { \\"id\\":44, \\"name\\": \\"bottle\\" },\
  { \\"id\\":46, \\"name\\": \\"wine_glass\\" },\
  { \\"id\\":47, \\"name\\": \\"cup\\" },\
  { \\"id\\":48, \\"name\\": \\"fork\\" },\
  { \\"id\\":49, \\"name\\": \\"knife\\" },\
  { \\"id\\":50, \\"name\\": \\"spoon\\" },\
  { \\"id\\":51, \\"name\\": \\"bowl\\" },\
  { \\"id\\":52, \\"name\\": \\"banana\\" },\
  { \\"id\\":53, \\"name\\": \\"apple\\" },\
  { \\"id\\":54, \\"name\\": \\"sandwich\\" },\
  { \\"id\\":55, \\"name\\": \\"orange\\" },\
  { \\"id\\":56, \\"name\\": \\"broccoli\\" },\
  { \\"id\\":57, \\"name\\": \\"carrot\\" },\
  { \\"id\\":58, \\"name\\": \\"hot_dog\\" },\
  { \\"id\\":59, \\"name\\": \\"pizza\\" },\
  { \\"id\\":60, \\"name\\": \\"donut\\" },\
  { \\"id\\":61, \\"name\\": \\"cake\\" },\
  { \\"id\\":62, \\"name\\": \\"chair\\" },\
  { \\"id\\":63, \\"name\\": \\"couch\\" },\
  { \\"id\\":64, \\"name\\": \\"potted_plant\\" },\
  { \\"id\\":65, \\"name\\": \\"bed\\" },\
  { \\"id\\":67, \\"name\\": \\"dining_table\\" },\
  { \\"id\\":70, \\"name\\": \\"toilet\\" },\
  { \\"id\\":72, \\"name\\": \\"tv\\" },\
  { \\"id\\":73, \\"name\\": \\"laptop\\" },\
  { \\"id\\":74, \\"name\\": \\"mouse\\" },\
  { \\"id\\":75, \\"name\\": \\"remote\\" },\
  { \\"id\\":76, \\"name\\": \\"keyboard\\" },\
  { \\"id\\":77, \\"name\\": \\"cell_phone\\" },\
  { \\"id\\":78, \\"name\\": \\"microwave\\" },\
  { \\"id\\":79, \\"name\\": \\"oven\\" },\
  { \\"id\\":80, \\"name\\": \\"toaster\\" },\
  { \\"id\\":81, \\"name\\": \\"sink\\" },\
  { \\"id\\":83, \\"name\\": \\"refrigerator\\" },\
  { \\"id\\":84, \\"name\\": \\"book\\" },\
  { \\"id\\":85, \\"name\\": \\"clock\\" },\
  { \\"id\\":86, \\"name\\": \\"vase\\" },\
  { \\"id\\":87, \\"name\\": \\"scissors\\" },\
  { \\"id\\":88, \\"name\\": \\"teddy_bear\\" },\
  { \\"id\\":89, \\"name\\": \\"hair_drier\\" },\
  { \\"id\\":90, \\"name\\": \\"toothbrush\\" }\
]\
","type":"detector"}},"spec":{"description":"Mask RCNN inception resnet v2 COCO via Intel OpenVINO","handler":"main:handler","runtime":"python:3.6","env":[{"name":"NUCLIO_PYTHON_EXE_PATH","value":"/opt/nuclio/common/openvino/python3"}],"resources":{"requests":{"cpu":"25m","memory":"1Mi"}},"image":"cvat.openvino.omz.public.mask_rcnn_inception_resnet_v2_atrous_coco:latest","targetCPU":75,"triggers":{"myHttpTrigger":{"class":"","kind":"http","name":"myHttpTrigger","maxWorkers":2,"workerAvailabilityTimeoutMilliseconds":10000,"attributes":{"maxRequestBodySize":33554432}}},"volumes":[{"volume":{"name":"volume-1","hostPath":{"path":"/raid/templates/cvat/serverless/common"}},"volumeMount":{"name":"volume-1","mountPath":"/opt/nuclio/common"}}],"build":{"image":"cvat.openvino.omz.public.mask_rcnn_inception_resnet_v2_atrous_coco","baseImage":"openvino/ubuntu18_dev:2020.2","directives":{"postCopy":[{"kind":"RUN","value":"apt update && DEBIAN_FRONTEND=noninteractive apt install --no-install-recommends -y python3-skimage"},{"kind":"RUN","value":"pip3 install \\"numpy<1.16.0\\""}],"preCopy":[{"kind":"USER","value":"root"},{"kind":"WORKDIR","value":"/opt/nuclio"},{"kind":"RUN","value":"ln -s /usr/bin/pip3 /usr/bin/pip"},{"kind":"RUN","value":"/opt/intel/openvino/deployment_tools/open_model_zoo/tools/downloader/downloader.py --name mask_rcnn_inception_resnet_v2_atrous_coco -o /opt/nuclio/open_model_zoo"},{"kind":"RUN","value":"/opt/intel/openvino/deployment_tools/open_model_zoo/tools/downloader/converter.py --name mask_rcnn_inception_resnet_v2_atrous_coco --precisions FP32 -d /opt/nuclio/open_model_zoo -o /opt/nuclio/open_model_zoo"}]},"codeEntryType":"image"},"platform":{"attributes":{"mountMode":"volume","restartPolicy":{"maximumRetryCount":3,"name":"always"}}},"readinessTimeoutSeconds":120,"securityContext":{},"eventTimeout":"60s"}}}}
23.02.09 17:17:55.671                     nuctl (I) Cleaning up before deployment {"functionName": "openvino-mask-rcnn-inception-resnet-v2-atrous-coco"}
23.02.09 17:17:56.545            nuctl.platform (I) Waiting for function to be ready {"timeout": 120}
23.02.09 17:17:58.560                     nuctl (I) Function deploy complete {"functionName": "openvino-mask-rcnn-inception-resnet-v2-atrous-coco", "httpPort": 49156, "internalInvocationURLs": ["172.17.0.7:8080"], "externalInvocationURLs": []}
모델확인
$ nuctl get functions
  NAMESPACE |                        NAME                        | PROJECT | STATE | REPLICAS | NODE PORT    
  nuclio    | openvino-mask-rcnn-inception-resnet-v2-atrous-coco | cvat    | ready | 1/1      |     49156  

Example using Python

https://nuclio.io/docs/latest/examples/#python-examples
Nuclio==>https://github.com/nuclio/nuclio/tree/master/hack/examples#python-examples
Hello World helloworldA simple function that
showcases unstructured logging and a structured response.
EncryptencryptA function that
uses a 3rd-party Python package to encrypt the event body, and showcases build commands for installing both OS-level and Python packages.
FaceRecognizerfaceA function that
uses Microsoft's face API, configured with function environment variables.
The function uses 3rd-party Python packages, which are installed by using an inline configuration.
Sentiment Analysis sentimentsA function that
uses the vaderSentiment library to classify text strings into a negative or positive sentiment score.
TensorFlowtensorflowA function that
uses the inception model of the TensorFlow open-source machine-learning library to classify images.
The function demonstrates advanced uses of Nuclio with a custom base image, third-party Python packages, pre-loading data into function memory (the AI Model), structured logging, and exception handling.

Hello World

Function-Configuration

function.yaml, main.py, model_handler.py 준비

https://github.com/nuclio/nuclio/blob/development/docs/reference/function-configuration/function-configuration-reference.md

apiVersion: "nuclio.io/v1beta1"
kind: "NuclioFunction"
metadata:
  name: hiworld
spec:
  # image: xxx:latest
  description: For Test Showcases unstructured logging and a structured response.
  runtime: "python"
  handler: main:handler
  minReplicas: 1
  maxReplicas: 1
#from model_handler import ModelHandler

def init_context(context):
    context.logger.info("Init context...  0%")
    # model = ModelHandler()
    # context.user_data.model = model
    context.logger.info("Init context...100%")

def handler(context, event):
    context.logger.info('This is an unstructured log')

    return context.Response(
        body='hihi, from nuclio :]',
        headers={},
        content_type='text/plain',
        status_code=200
    )
# import cv2
# import numpy as np
import os
#from model_loader import ModelLoader

#class ModelHandler:
    # def __init__(self):
    # def handle(self, image, points):

# class AttributesExtractorHandler:
    # def __init__(self):
    # def infer(self, image):

deploy functions

(mmlab38) ~/my/git/mmdetection/cvat$

# $ nuctl deploy hiworld \\
$ nuctl deploy  \\
  --path     serverless/sixx/hiworld/nuclio \\
  --volume   `pwd`/serverless/common:/opt/nuclio/common \\
  --platform local

23.02.16 13:44:32.874                     nuctl (I) Deploying function {"name": "hiworld"}
23.02.16 13:44:32.874                     nuctl (I) Building {"builderKind": "docker", "versionInfo": "Label: 1.8.14, Git commit: cbb0774230996a3eb4621c1a2079e2317578005b, OS: linux, Arch: amd64, Go version: go1.17.8", "name": "hiworld"}
23.02.16 13:44:33.090                     nuctl (I) Staging files and preparing base images
23.02.16 13:44:33.091                     nuctl (I) Building processor image {"registryURL": "", "imageName": "nuclio/processor-hiworld:latest"}
23.02.16 13:44:33.091     nuctl.platform.docker (I) Pulling image {"imageName": "quay.io/nuclio/handler-builder-python-onbuild:1.8.14-amd64"}
23.02.16 13:44:36.511     nuctl.platform.docker (I) Pulling image {"imageName": "quay.io/nuclio/uhttpc:0.0.1-amd64"}
23.02.16 13:44:40.783            nuctl.platform (I) Building docker image {"image": "nuclio/processor-hiworld:latest"}
23.02.16 13:45:33.889            nuctl.platform (I) Pushing docker image into registry {"image": "nuclio/processor-hiworld:latest", "registry": ""}
23.02.16 13:45:33.889            nuctl.platform (I) Docker image was successfully built and pushed into docker registry {"image": "nuclio/processor-hiworld:latest"}
23.02.16 13:45:33.889                     nuctl (I) Build complete {"result": {"Image":"nuclio/processor-hiworld:latest","UpdatedFunctionConfig":{"metadata":{"name":"hiworld","namespace":"nuclio","labels":{"nuclio.io/project-name":"default"}},"spec":{"description":"For Test Showcases unstructured logging and a structured response.","handler":"hiworld:handler","runtime":"python:3.7","resources":{"requests":{"cpu":"25m","memory":"1Mi"}},"image":"nuclio/processor-hiworld:latest","minReplicas":1,"maxReplicas":1,"targetCPU":75,"triggers":{"default-http":{"class":"","kind":"http","name":"default-http","maxWorkers":1}},"volumes":[{"volume":{"name":"volume-1","hostPath":{"path":"/raid/templates/cvat/serverless/common"}},"volumeMount":{"name":"volume-1","mountPath":"/opt/nuclio/common"}}],"build":{"codeEntryType":"image"},"platform":{},"readinessTimeoutSeconds":120,"securityContext":{},"eventTimeout":""}}}}
23.02.16 13:45:33.896                     nuctl (I) Cleaning up before deployment {"functionName": "hiworld"}
23.02.16 13:45:34.728            nuctl.platform (I) Waiting for function to be ready {"timeout": 120}
23.02.16 13:45:36.755                     nuctl (I) Function deploy complete {"functionName": "hiworld", "httpPort": 49159, "internalInvocationURLs": ["172.17.0.9:8080"], "externalInvocationURLs": []}

$ nuctl get functions
  NAMESPACE |                        NAME     | PROJECT | STATE | REPLICAS | NODE PORT  
  nuclio    | hiworld                         | default | ready | 1/1      |     49158  
  nuclio    | openvino-dextr                  | cvat    | ready | 1/1      |     49153  
  nuclio    | openvino-mask-rcnn-inception... | cvat    | ready | 1/1      |     49156  
  nuclio    | openvino-omz-public-yolo-v3-tf  | cvat    | ready | 1/1      |     49155  
  nuclio    | pth-foolwood-siammask           | cvat    | ready | 1/1      |     49157  
  nuclio    | tf-matterport-mask-rcnn         | cvat    | error | 1/1      |            
$ nuctl invoke hiworld \\
 --method POST \\
 --body '{"hihihi":"world"}' \\
 --content-type "application/json"

23.02.16 11:08:15.114    nuctl.platform.invoker (I) Executing function {"method": "POST", "url": "http://:49158", "bodyLength": 17, "headers": {"Content-Type":["application/json"],"X-Nuclio-Log-Level":["info"],"X-Nuclio-Target":["helloworld"]}}
23.02.16 11:08:15.115    nuctl.platform.invoker (I) Got response {"status": "200 OK"}
23.02.16 11:08:15.115                     nuctl (I) >>> Start of function logs
23.02.16 11:08:15.115                helloworld (I) This is an unstrucured log {"time": 1676513295115.0264}
23.02.16 11:08:15.115                     nuctl (I) <<< End of function logs

> Response headers:
Server = nuclio
Date = Thu, 16 Feb 2023 02:08:14 GMT
Content-Type = application/text
Content-Length = 21

> Response body:
Hello, from Nuclio :]
metadata:
  name: hiworld
  labels:
    nuclio.io/project-name: default
spec:
  description: "For Test Showcases unstructured logging and a structured response."
  handler: "hiworld:handler"
  runtime: "python:3.7"
  resources:
    requests:
      cpu: 25m
      memory: 1Mi
  image: "nuclio/processor-hiworld:latest"
  minReplicas: 1
  maxReplicas: 1
  targetCPU: 75
  triggers:
    default-http:
      class: ""
      kind: http
      name: default-http
      maxWorkers: 1
  volumes:
    - volume:
        name: volume-1
        hostPath:
          path: /home/oschung_skcc/my/git/mmdetection/cvat/serverless/common
      volumeMount:
        name: volume-1
        mountPath: /opt/nuclio/common
  build:
    codeEntryType: image
    timestamp: 1676523194
  platform: {}
  readinessTimeoutSeconds: 120
  securityContext: {}
  eventTimeout: ""
  version: 1

모델배포 : custom model - detectron2

https://opencv.github.io/cvat/docs/manual/advanced/serverless-tutorial/#adding-your-own-dl-models

detectron2설치후 Local에서 DL모델 돌려보기

$ python demo/demo.py \\
--config-file configs/COCO-Detection/retinanet_R_101_FPN_3x.yaml \\
--input 1920px-Cat_poster_1.jpg \\
--output out.jpg \\
--opts MODEL.WEIGHTS model_final_971ab9.pkl  MODEL.DEVICE cpu
from detectron2.config import get_cfg
from detectron2.data.detection_utils import read_image

from detectron2.engine.defaults            import DefaultPredictor
from detectron2.data.datasets.builtin_meta import COCO_CATEGORIES

CONFIG_FILE = "configs/COCO-Detection/retinanet_R_101_FPN_3x.yaml"
CONFIG_OPTS = ["MODEL.WEIGHTS", "model_final_971ab9.pkl", "MODEL.DEVICE", "cpu"]
CONFIDENCE_THRESHOLD = 0.5
INPUT_PATH  = "1920px-Cat_poster_1.jpg"

def setup_cfg():
    cfg = get_cfg()
    cfg.merge_from_file(CONFIG_FILE)
    cfg.merge_from_list(CONFIG_OPTS)

    cfg.MODEL.RETINANET.SCORE_THRESH_TEST                      = CONFIDENCE_THRESHOLD
    cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST                      = CONFIDENCE_THRESHOLD
    cfg.MODEL.PANOPTIC_FPN.COMBINE.INSTANCES_CONFIDENCE_THRESH = CONFIDENCE_THRESHOLD
    cfg.freeze()
    return cfg

if __name__ == "__main__":
    cfg = setup_cfg()
\t# INPUT\t
    img = read_image(INPUT_PATH, format="BGR")
    predictor   = DefaultPredictor(cfg)
    predictions = predictor(img)
    
    instances    = predictions['instances']
    pred_boxes   = instances.pred_boxes
    scores       = instances.scores
    pred_classes = instances.pred_classes
\t
    for box, score, label in zip(pred_boxes, scores, pred_classes):
        label = COCO_CATEGORIES[int(label)]["name"]
        print(box.tolist(), float(score), label)
\t# Video

function.yaml

1) Annotation 위해 CVAT이 활용할 serverless function 준비

https://opencv.github.io/cvat/docs/manual/advanced/serverless-tutorial/#dl-model-as-a-serverless-function

함수의 이름 정하기

파라미터 전체설명 : https://nuclio.io/docs/latest/reference/function-configuration/function-configuration-reference/

metadata.name : pth.facebookresearch.detectron2.retinanet_r101

metadata.annotations.name : 보여지는 이름

meradata.annotation.type: serverless function의 종류 . detector

metadata.annotation.framework: 기냥 정보성.. 특별히 정해진건 없으나..OpenVINO, PyTorch, TensorFlow, etc.

function.yaml

metadata:
  name: pth.facebookresearch.detectron2.retinanet_r101
  namespace: cvat
  annotations:
    name: RetinaNet R101
    type: detector
    framework: pytorch
    spec: |
      [
        { "id": 1, "name": "person" },
        { "id": 2, "name": "bicycle" },

        ...

        { "id":89, "name": "hair_drier" },
        { "id":90, "name": "toothbrush" }
      ]      

spec:
  description: RetinaNet R101 from Detectron2
  runtime: 'python:3.8'
  handler: main:handler
  eventTimeout: 30s

  build:
    image: cvat/pth.facebookresearch.detectron2.retinanet_r101
    baseImage: ubuntu:20.04

    directives:
      preCopy:
        - kind: ENV
          value: DEBIAN_FRONTEND=noninteractive
        - kind: RUN
          value: apt-get update && apt-get -y install curl git python3 python3-pip
        - kind: WORKDIR
          value: /opt/nuclio
        - kind: RUN
          value: pip3 install torch==1.8.1+cpu torchvision==0.9.1+cpu torchaudio==0.8.1 -f https://download.pytorch.org/whl/torch_stable.html
        - kind: RUN
          value: pip3 install 'git+https://github.com/facebookresearch/[email protected]'
        - kind: RUN
          value: curl -O https://dl.fbaipublicfiles.com/detectron2/COCO-Detection/retinanet_R_101_FPN_3x/190397697/model_final_971ab9.pkl
        - kind: RUN
          value: ln -s /usr/bin/pip3 /usr/local/bin/pip

  triggers:
    myHttpTrigger:
      maxWorkers: 2
      kind: 'http'
      workerAvailabilityTimeoutMilliseconds: 10000
      attributes:
        maxRequestBodySize: 33554432 # 32MB

  platform:
    attributes:
      restartPolicy:
        name: always
        maximumRetryCount: 3
      mountMode: volume

외부에서 해당 모델을 통해 접근을 위한 설정

nuclio function의 포트번호와 cvat network를 지정

spec.triggers.myHttpTrigger.attributes에 특정 port를 추가하고, 
spec.platform.attributes 에 도커 컴포즈를 실행했을 때 생성된 도커 네트워크(cvat_cvat)를 추가.

  triggers:
    myHttpTrigger:
      maxWorkers: 2
      kind: 'http'
      workerAvailabilityTimeoutMilliseconds: 10000
      attributes:
        maxRequestBodySize: 33554432
        port: 33000 # 추가포트 

  platform:
    attributes:
      restartPolicy:
        name: always
        maximumRetryCount: 3
      mountMode: volume
      network: cvat_cvat # 추가 도커네트웍

2. 로컬에서 DL모델을 실행하기 위한 소스코드를 Nuclio 플랫폼에 적용

2-1 모델을 메모리에 로딩 (init_context(context)함수를 사용하여)

https://nuclio.io/docs/latest/concepts/best-practices-and-common-pitfalls/#use-init_context-instead-of-global-variable-declarations-or-function-calls

def handler(context, event):
    # context.user_data.my_db_connection.query(...)


def init_context(context):
    # Create the DB connection under "context.user_data"
    # setattr(context.user_data, 'my_db_connection', my_db.create_connection())

    cfg = get_config('COCO-Detection/retinanet_R_101_FPN_3x.yaml')
    cfg.merge_from_list(CONFIG_OPTS)
    
    cfg.MODEL.RETINANET.SCORE_THRESH_TEST = CONFIDENCE_THRESHOLD
    cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = CONFIDENCE_THRESHOLD
    cfg.MODEL.PANOPTIC_FPN.COMBINE.INSTANCES_CONFIDENCE_THRESH = CONFIDENCE_THRESHOLD
    
    cfg.freeze()
    predictor = DefaultPredictor(cfg)
    
    context.user_data.model_handler = predictor

2-2 아래 프로세스를 위해 handler에 entry point를 정의하고, main.py에 넣는다.

  • accept incoming HTTP requests
  • run inference
  • reply with detection results

main.py

import json
import base64
import io
from PIL import Image
import yaml

from detectron2.config import get_cfg
from detectron2.data.detection_utils import convert_PIL_to_numpy
from detectron2.engine.defaults import DefaultPredictor
from detectron2.data.datasets.builtin_meta import COCO_CATEGORIES

CONFIG_FILE = "detectron2/configs/COCO-Detection/retinanet_R_101_FPN_3x.yaml"
CONFIG_OPTS = ["MODEL.WEIGHTS", "model_final_971ab9.pkl", "MODEL.DEVICE", "cpu"]
CONFIDENCE_THRESHOLD = 0.5

def handler(context, event):
    context.logger.info("Run retinanet-R101 model")
    
    data      = event.body
    buf       = io.BytesIO(base64.b64decode(data["image"]))
    threshold = float(data.get("threshold", CONFIDENCE_THRESHOLD)) #0.5))
    image     = convert_PIL_to_numpy(Image.open(buf), format="BGR")

    predictions  = context.user_data.model_handler(image)
    
    instances    = predictions['instances']
    pred_boxes   = instances.pred_boxes
    scores       = instances.scores
    pred_classes = instances.pred_classes
    
    results = []
    for box, score, label in zip(pred_boxes, scores, pred_classes):
        label = COCO_CATEGORIES[int(label)]["name"]
        if score >= threshold:
            results.append({
                "confidence": str(float(score)),
                "label": label,
                "points": box.tolist(),
                "type": "rectangle",
            })

    return context.Response(
        body=json.dumps(results), 
        headers={},
        content_type='application/json', 
        status_code=200)


def init_context(context):
    context.logger.info("Init context...  0%")

    cfg = get_config(CONFIG_FILE) # 'COCO-Detection/retinanet_R_101_FPN_3x.yaml')
    cfg.merge_from_list(CONFIG_OPTS)
    cfg.MODEL.RETINANET.SCORE_THRESH_TEST = CONFIDENCE_THRESHOLD
    cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = CONFIDENCE_THRESHOLD
    cfg.MODEL.PANOPTIC_FPN.COMBINE.INSTANCES_CONFIDENCE_THRESH = CONFIDENCE_THRESHOLD
    cfg.freeze()
    
    predictor = DefaultPredictor(cfg)

    #context.user_data.model_handler = predictor
    setattr(context.user_data, 'model_handler', predictor)
    functionconfig = yaml.safe_load(open("/opt/nuclio/function.yaml"))
    labels_spec    = functionconfig['metadata']['annotations']['spec']
    labels         = {item['id']: item['name'] for item in json.loads(labels_spec)}
    setattr(context.user_data, "labels", labels)    
    
    context.logger.info("Init context...100%")

3. deploy

새로운 serverless 함수를 사용하기 위해서는
(위 Builtin model에서 했던것처럼) nuctl 명령어로 deploy를 해야한다.

  • function.yaml
  • main.py
  • model_handler.py

방법1)

$ nuctl deploy --project-name cvat \\
  --path     serverless/pytorch/facebookresearch/detectron2/retinanet/ \\
  --volume   `pwd`/serverless/common:/opt/nuclio/common \\
  --platform local

방법2)

$ serverless/deploy_cpu.sh \\
serverless/pytorch/facebookresearch/detectron2/retinanet/

Issue

https://github.com/opencv/cvat/issues/3457 : Steps for custom model deployment

http://How to upload DL which built by myself? : How to upload DL which built by myself?

https://github.com/opencv/cvat/issues/5551 : Mmdetection MaskRCNN serverless support for semi-automatic annotation

https://github.com/opencv/cvat/issues/4909: Load my own Yolov5 model on cvat by nuclio

mmdetection 모델변환

https://mmdetection.readthedocs.io/en/latest/useful_tools.html#model-conversion

Exporting MMDetection models to ONNX format
https://medium.com/axinc-ai/exporting-mmdetection-models-to-onnx-format-3ec839c38ff

openvino : https://da2so.tistory.com/63

Serving pre-trained ML/DL models
https://docs.mlrun.org/en/stable/tutorial/03-model-serving.html#serving-pre-trained-ml-dl-models

Categories: vision

onesixx

Blog Owner

Subscribe
Notify of
guest

0 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
0
Would love your thoughts, please comment.x
()
x