6월 14-15일 우크라이나 DevFest에 참여하세요. 지금 온라인 등록

TensorFlow Serving으로 TensorFlow 모델 학습 및 제공

이 가이드는 분류하는 신경망 모델 훈련 운동화와 티셔츠 같은 옷의 이미지를 , 훈련 모델을 저장하고, 다음으로 서비스를 제공 TensorFlow 서빙 . TensorFlow 그렇게 모델링과 훈련이 볼에 초점을 맞춘 완전한 예를 들어, 오히려 TensorFlow의 모델링 및 교육보다는 봉사에 초점은 기본 분류 예 .

이 가이드는 사용 tf.keras , TensorFlow에서 빌드 기차 모델에 대한 높은 수준의 API를.

import sys

# Confirm that we're using Python 3
assert sys.version_info.major == 3, 'Oops, not running Python 3. Use Runtime > Change runtime type'
# TensorFlow and tf.keras
print("Installing dependencies for Colab environment")
!pip install -Uq grpcio==1.26.0

import tensorflow as tf
from tensorflow import keras

# Helper libraries
import numpy as np
import matplotlib.pyplot as plt
import os
import subprocess

print('TensorFlow version: {}'.format(tf.__version__))

모델 만들기

Fashion MNIST 데이터세트 가져오기

이 가이드는 사용 패션 MNIST의 10 개 개의 카테고리에서 70,000 그레이 스케일 이미지가 포함 된 데이터 집합을. 이미지는 다음과 같이 저해상도(28 x 28픽셀)의 개별 의류 품목을 보여줍니다.

패션 MNIST 스프라이트
그림 1. 패션 - MNIST 샘플 (인 Zalando, MIT 라이센스에 의해).

패션 MNIST는 드롭 인 교체 고전에 대한위한 것입니다 MNIST 데이터 세트 종종 컴퓨터 비전을위한 기계 학습 프로그램의 "안녕, 세계"로 사용. TensorFlow에서 직접 Fashion MNIST에 액세스할 수 있으며 데이터를 가져오고 로드하기만 하면 됩니다.

fashion_mnist = keras.datasets.fashion_mnist
(train_images, train_labels), (test_images, test_labels) = fashion_mnist.load_data()

# scale the values to 0.0 to 1.0
train_images = train_images / 255.0
test_images = test_images / 255.0

# reshape for feeding into the model
train_images = train_images.reshape(train_images.shape[0], 28, 28, 1)
test_images = test_images.reshape(test_images.shape[0], 28, 28, 1)

class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']

print('\ntrain_images.shape: {}, of {}'.format(train_images.shape, train_images.dtype))
print('test_images.shape: {}, of {}'.format(test_images.shape, test_images.dtype))
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz
32768/29515 [=================================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz
26427392/26421880 [==============================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz
8192/5148 [===============================================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz
4423680/4422102 [==============================] - 0s 0us/step

train_images.shape: (60000, 28, 28, 1), of float64
test_images.shape: (10000, 28, 28, 1), of float64

모델 학습 및 평가

모델링 부분에 초점을 맞추지 않았기 때문에 가능한 가장 간단한 CNN을 사용합시다.

model = keras.Sequential([
  keras.layers.Conv2D(input_shape=(28,28,1), filters=8, kernel_size=3, 
                      strides=2, activation='relu', name='Conv1'),
  keras.layers.Flatten(),
  keras.layers.Dense(10, name='Dense')
])
model.summary()

testing = False
epochs = 5

model.compile(optimizer='adam', 
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=[keras.metrics.SparseCategoricalAccuracy()])
model.fit(train_images, train_labels, epochs=epochs)

test_loss, test_acc = model.evaluate(test_images, test_labels)
print('\nTest accuracy: {}'.format(test_acc))
2021-12-04 10:29:34.128871: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'libcusolver.so.10'; dlerror: libcusolver.so.10: cannot open shared object file: No such file or directory
2021-12-04 10:29:34.129907: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1757] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.
Skipping registering GPU devices...
Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
Conv1 (Conv2D)               (None, 13, 13, 8)         80        
_________________________________________________________________
flatten (Flatten)            (None, 1352)              0         
_________________________________________________________________
Dense (Dense)                (None, 10)                13530     
=================================================================
Total params: 13,610
Trainable params: 13,610
Non-trainable params: 0
_________________________________________________________________
Epoch 1/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.7204 - sparse_categorical_accuracy: 0.7549
Epoch 2/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.3997 - sparse_categorical_accuracy: 0.8611
Epoch 3/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.3580 - sparse_categorical_accuracy: 0.8754
Epoch 4/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.3399 - sparse_categorical_accuracy: 0.8780
Epoch 5/5
1875/1875 [==============================] - 4s 2ms/step - loss: 0.3232 - sparse_categorical_accuracy: 0.8849
313/313 [==============================] - 0s 1ms/step - loss: 0.3586 - sparse_categorical_accuracy: 0.8738

Test accuracy: 0.8737999796867371

모델 저장

우리에게에 저장하는 최초의 필요성 봉사 TensorFlow에 우리의 훈련 모델을로드하려면 SavedModel의 형식을. 이렇게 하면 잘 정의된 디렉토리 계층에 protobuf 파일이 생성되고 버전 번호가 포함됩니다. TensorFlow 서빙은 우리가 우리가 추론 요청을 할 때 사용하고자하는 모델의 버전, 또는 "게재 가능한"를 선택할 수 있습니다. 각 버전은 지정된 경로 아래의 다른 하위 디렉토리로 내보내집니다.

# Fetch the Keras session and save the model
# The signature definition is defined by the input and output tensors,
# and stored with the default serving key
import tempfile

MODEL_DIR = tempfile.gettempdir()
version = 1
export_path = os.path.join(MODEL_DIR, str(version))
print('export_path = {}\n'.format(export_path))

tf.keras.models.save_model(
    model,
    export_path,
    overwrite=True,
    include_optimizer=True,
    save_format=None,
    signatures=None,
    options=None
)

print('\nSaved model:')
!ls -l {export_path}
export_path = /tmp/1
2021-12-04 10:29:53.392905: W tensorflow/python/util/util.cc:348] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.
INFO:tensorflow:Assets written to: /tmp/1/assets

Saved model:
total 88
drwxr-xr-x 2 kbuilder kbuilder  4096 Dec  4 10:29 assets
-rw-rw-r-- 1 kbuilder kbuilder 78055 Dec  4 10:29 saved_model.pb
drwxr-xr-x 2 kbuilder kbuilder  4096 Dec  4 10:29 variables

저장된 모델 검토

우리는 명령 행 유틸리티 사용합니다 saved_model_cli 상기 볼 MetaGraphDefs (모델)과 SignatureDefs 우리 SavedModel에 (방법은 당신이 호출 할 수 있습니다). 참조 SavedModel CLI의이 토론 TensorFlow 설명서를.

saved_model_cli show --dir {export_path} --all
MetaGraphDef with tag-set: 'serve' contains the following SignatureDefs:

signature_def['__saved_model_init_op']:
  The given SavedModel SignatureDef contains the following input(s):
  The given SavedModel SignatureDef contains the following output(s):
    outputs['__saved_model_init_op'] tensor_info:
        dtype: DT_INVALID
        shape: unknown_rank
        name: NoOp
  Method name is: 

signature_def['serving_default']:
  The given SavedModel SignatureDef contains the following input(s):
    inputs['Conv1_input'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 28, 28, 1)
        name: serving_default_Conv1_input:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['Dense'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 10)
        name: StatefulPartitionedCall:0
  Method name is: tensorflow/serving/predict

Defined Functions:
  Function Name: '__call__'
    Option #1
      Callable with:
        Argument #1
          Conv1_input: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='Conv1_input')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #2
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #3
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None
    Option #4
      Callable with:
        Argument #1
          Conv1_input: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='Conv1_input')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None

  Function Name: '_default_save_signature'
    Option #1
      Callable with:
        Argument #1
          Conv1_input: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='Conv1_input')

  Function Name: 'call_and_return_all_conditional_losses'
    Option #1
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #2
      Callable with:
        Argument #1
          Conv1_input: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='Conv1_input')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None
    Option #3
      Callable with:
        Argument #1
          Conv1_input: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='Conv1_input')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #4
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28, 1), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None

그것은 우리 모델에 대해 많은 것을 말해줍니다! 이 경우 우리는 우리의 모델을 훈련시켰기 때문에 이미 입력과 출력을 알고 있지만, 그렇지 않다면 이것은 중요한 정보가 될 것입니다. 예를 들어 이것이 회색조 이미지 데이터라는 사실과 같이 모든 것을 알려주지는 않지만 훌륭한 시작입니다.

TensorFlow Serving으로 모델 제공

TensorFlow Serving 배포 URI를 패키지 소스로 추가합니다.

우리는 TensorFlow 사용 서빙 설치 준비하고 적성을 이 Colab 데비안 환경에서 실행하기 때문이다. 우리는 추가 할 것입니다 tensorflow-model-server 적성가 알고있는 패키지 목록에 패키지를. 우리는 루트로 실행 중입니다.

import sys
# We need sudo prefix if not on a Google Colab.
if 'google.colab' not in sys.modules:
  SUDO_IF_NEEDED = 'sudo'
else:
  SUDO_IF_NEEDED = ''
# This is the same as you would do from your command line, but without the [arch=amd64], and no sudo
# You would instead do:
# echo "deb [arch=amd64] http://storage.googleapis.com/tensorflow-serving-apt stable tensorflow-model-server tensorflow-model-server-universal" | sudo tee /etc/apt/sources.list.d/tensorflow-serving.list && \
# curl https://storage.googleapis.com/tensorflow-serving-apt/tensorflow-serving.release.pub.gpg | sudo apt-key add -

!echo "deb http://storage.googleapis.com/tensorflow-serving-apt stable tensorflow-model-server tensorflow-model-server-universal" | {SUDO_IF_NEEDED} tee /etc/apt/sources.list.d/tensorflow-serving.list && \
curl https://storage.googleapis.com/tensorflow-serving-apt/tensorflow-serving.release.pub.gpg | {SUDO_IF_NEEDED} apt-key add -
!{SUDO_IF_NEEDED} apt update
deb http://storage.googleapis.com/tensorflow-serving-apt stable tensorflow-model-server tensorflow-model-server-universal
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  2943  100  2943    0     0  15571      0 --:--:-- --:--:-- --:--:-- 15571
OK
Hit:1 http://asia-east1.gce.archive.ubuntu.com/ubuntu bionic InRelease
Hit:2 http://asia-east1.gce.archive.ubuntu.com/ubuntu bionic-updates InRelease
Hit:3 http://asia-east1.gce.archive.ubuntu.com/ubuntu bionic-backports InRelease
Hit:4 https://nvidia.github.io/libnvidia-container/stable/ubuntu18.04/amd64  InRelease
Get:5 https://nvidia.github.io/nvidia-container-runtime/ubuntu18.04/amd64  InRelease [1481 B]
Get:6 https://nvidia.github.io/nvidia-docker/ubuntu18.04/amd64  InRelease [1474 B]
Ign:7 http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64  InRelease
Get:8 http://storage.googleapis.com/tensorflow-serving-apt stable InRelease [3012 B]
Hit:9 http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64  Release
Get:10 http://security.ubuntu.com/ubuntu bionic-security InRelease [88.7 kB]
Get:11 https://packages.cloud.google.com/apt eip-cloud-bionic InRelease [5419 B]
Get:12 http://packages.cloud.google.com/apt google-cloud-logging-wheezy InRelease [5483 B]
Hit:13 http://archive.canonical.com/ubuntu bionic InRelease
Err:11 https://packages.cloud.google.com/apt eip-cloud-bionic InRelease
  The following signatures couldn't be verified because the public key is not available: NO_PUBKEY FEEA9169307EA071 NO_PUBKEY 8B57C5C2836F4BEB
Get:15 http://storage.googleapis.com/tensorflow-serving-apt stable/tensorflow-model-server amd64 Packages [339 B]
Err:12 http://packages.cloud.google.com/apt google-cloud-logging-wheezy InRelease
  The following signatures couldn't be verified because the public key is not available: NO_PUBKEY FEEA9169307EA071 NO_PUBKEY 8B57C5C2836F4BEB
Get:16 http://storage.googleapis.com/tensorflow-serving-apt stable/tensorflow-model-server-universal amd64 Packages [348 B]
Fetched 106 kB in 1s (103 kB/s)



119 packages can be upgraded. Run 'apt list --upgradable' to see them.
W: An error occurred during the signature verification. The repository is not updated and the previous index files will be used. GPG error: https://packages.cloud.google.com/apt eip-cloud-bionic InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY FEEA9169307EA071 NO_PUBKEY 8B57C5C2836F4BEB
W: An error occurred during the signature verification. The repository is not updated and the previous index files will be used. GPG error: http://packages.cloud.google.com/apt google-cloud-logging-wheezy InRelease: The following signatures couldn't be verified because the public key is not available: NO_PUBKEY FEEA9169307EA071 NO_PUBKEY 8B57C5C2836F4BEB
W: Failed to fetch https://packages.cloud.google.com/apt/dists/eip-cloud-bionic/InRelease  The following signatures couldn't be verified because the public key is not available: NO_PUBKEY FEEA9169307EA071 NO_PUBKEY 8B57C5C2836F4BEB
W: Failed to fetch http://packages.cloud.google.com/apt/dists/google-cloud-logging-wheezy/InRelease  The following signatures couldn't be verified because the public key is not available: NO_PUBKEY FEEA9169307EA071 NO_PUBKEY 8B57C5C2836F4BEB
W: Some index files failed to download. They have been ignored, or old ones used instead.

TensorFlow Serving 설치

이것은 당신이 필요한 전부입니다 - 하나의 명령줄!

{SUDO_IF_NEEDED} apt-get install tensorflow-model-server
The following packages were automatically installed and are no longer required:
  linux-gcp-5.4-headers-5.4.0-1040 linux-gcp-5.4-headers-5.4.0-1043
  linux-gcp-5.4-headers-5.4.0-1044 linux-gcp-5.4-headers-5.4.0-1049
Use 'sudo apt autoremove' to remove them.
The following NEW packages will be installed:
  tensorflow-model-server
0 upgraded, 1 newly installed, 0 to remove and 119 not upgraded.
Need to get 335 MB of archives.
After this operation, 0 B of additional disk space will be used.
Get:1 http://storage.googleapis.com/tensorflow-serving-apt stable/tensorflow-model-server amd64 tensorflow-model-server all 2.7.0 [335 MB]
Fetched 335 MB in 7s (45.2 MB/s)
Selecting previously unselected package tensorflow-model-server.
(Reading database ... 264341 files and directories currently installed.)
Preparing to unpack .../tensorflow-model-server_2.7.0_all.deb ...
Unpacking tensorflow-model-server (2.7.0) ...
Setting up tensorflow-model-server (2.7.0) ...

TensorFlow Serving 실행 시작

여기에서 TensorFlow Serving 실행을 시작하고 모델을 로드합니다. 로드된 후 REST를 사용하여 추론 요청을 시작할 수 있습니다. 몇 가지 중요한 매개변수가 있습니다.

  • rest_api_port : 당신은 REST 요청을 사용합니다하는 포트입니다.
  • model_name : 당신은 REST 요청의 URL이 사용됩니다. 무엇이든 될 수 있습니다.
  • model_base_path : 이것은 당신이 당신의 모델을 저장 한 디렉토리 경로입니다.
os.environ["MODEL_DIR"] = MODEL_DIR
nohup tensorflow_model_server \
  --rest_api_port=8501 \
  --model_name=fashion_model \
  --model_base_path="${MODEL_DIR}" >server.log 2>&1
tail server.log

TensorFlow Serving에서 모델에 요청하기

먼저 테스트 데이터에서 임의의 예를 살펴보겠습니다.

def show(idx, title):
  plt.figure()
  plt.imshow(test_images[idx].reshape(28,28))
  plt.axis('off')
  plt.title('\n\n{}'.format(title), fontdict={'size': 16})

import random
rando = random.randint(0,len(test_images)-1)
show(rando, 'An Example Image: {}'.format(class_names[test_labels[rando]]))

png

알겠습니다. 흥미롭군요. 당신은 그것을 인식하기가 얼마나 어렵습니까? 이제 3개의 추론 요청 배치에 대한 JSON 객체를 생성하고 모델이 사물을 얼마나 잘 인식하는지 살펴보겠습니다.

import json
data = json.dumps({"signature_name": "serving_default", "instances": test_images[0:3].tolist()})
print('Data: {} ... {}'.format(data[:50], data[len(data)-52:]))
Data: {"signature_name": "serving_default", "instances": ...  [0.0], [0.0], [0.0], [0.0], [0.0], [0.0], [0.0]]]]}

REST 요청 만들기

최신 버전의 서빙 가능

우리는 서버의 REST 끝점에 POST로 예측 요청을 보내고 세 가지 예를 전달합니다. 특정 버전을 지정하지 않음으로써 서버에 최신 버전의 서빙 가능을 제공하도록 요청할 것입니다.

# docs_infra: no_execute
!pip install -q requests

import requests
headers = {"content-type": "application/json"}
json_response = requests.post('http://localhost:8501/v1/models/fashion_model:predict', data=data, headers=headers)
predictions = json.loads(json_response.text)['predictions']

show(0, 'The model thought this was a {} (class {}), and it was actually a {} (class {})'.format(
  class_names[np.argmax(predictions[0])], np.argmax(predictions[0]), class_names[test_labels[0]], test_labels[0]))

특정 버전의 servable

이제 servable의 특정 버전을 지정해 보겠습니다. 하나만 있으므로 버전 1을 선택하겠습니다. 세 가지 결과도 모두 살펴보겠습니다.

# docs_infra: no_execute
headers = {"content-type": "application/json"}
json_response = requests.post('http://localhost:8501/v1/models/fashion_model/versions/1:predict', data=data, headers=headers)
predictions = json.loads(json_response.text)['predictions']

for i in range(0,3):
  show(i, 'The model thought this was a {} (class {}), and it was actually a {} (class {})'.format(
    class_names[np.argmax(predictions[i])], np.argmax(predictions[i]), class_names[test_labels[i]], test_labels[i]))