هذا دليل القطارات نموذج الشبكة العصبية لتصنيف الصور من الملابس، مثل أحذية رياضية والقمصان ، ويوفر نموذج المدربين، ومن ثم يقدم مع TensorFlow خدمة . وينصب التركيز على TensorFlow خدمة، بدلا من النمذجة والتدريب في TensorFlow، وذلك لمثال الكامل الذي يركز على النمذجة والتدريب رؤية المثال تصنيف الأساسية .
يستخدم هذا الدليل tf.keras ، وAPI رفيع المستوى لبناء وتدريب النماذج في TensorFlow.
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__))
قم بإنشاء نموذجك
قم باستيراد مجموعة بيانات Fashion MNIST
يستخدم هذا الدليل الأزياء MNIST مجموعة البيانات التي تحتوي على 70،000 الصور الرمادية في 10 فئات. تُظهر الصور مقالات فردية من الملابس بدقة منخفضة (28 × 28 بكسل) ، كما هو موضح هنا:
الشكل 1. عينات الموضة MNIST (عن طريق زالاندو، MIT الترخيص). |
ويهدف الأزياء MNIST بمثابة قطرة في بديل للالكلاسيكية MNIST مجموعة البيانات، غالبا ما تستخدم مثل "مرحبا، العالم" برامج التعلم آلة لرؤية الكمبيوتر. يمكنك الوصول إلى Fashion MNIST مباشرة من TensorFlow ، ما عليك سوى استيراد البيانات وتحميلها.
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
تدريب وتقييم النموذج الخاص بك
لنستخدم أبسط شبكة CNN ممكنة ، لأننا لا نركز على جزء النمذجة.
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
احفظ النموذج الخاص بك
لتحميل نموذج المدربين لدينا في TensorFlow تخدم نحن بحاجة أولا لحفظه في SavedModel الشكل. سيؤدي هذا إلى إنشاء ملف protobuf في تسلسل هرمي محدد جيدًا للدليل ، وسيتضمن رقم الإصدار. TensorFlow التي تخدم تتيح لنا تحديد أي نسخة من نموذج، أو "قابل للعرض" نريد استخدام عندما نجعل طلبات الاستدلال. سيتم تصدير كل إصدار إلى دليل فرعي مختلف ضمن المسار المحدد.
# 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
افحص نموذجك المحفوظ
سنستخدم الأمر أداة سطر saved_model_cli
أن ننظر إلى MetaGraphDefs (النماذج) و SignatureDefs (الأساليب التي يمكن استدعاء) في منطقتنا SavedModel. انظر هذا النقاش من SavedModel CLI في دليل TensorFlow.
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
هذا يخبرنا الكثير عن نموذجنا! في هذه الحالة ، قمنا للتو بتدريب نموذجنا ، لذلك نحن نعرف بالفعل المدخلات والمخرجات ، ولكن إذا لم نقم بذلك ، فستكون هذه معلومات مهمة. إنه لا يخبرنا بكل شيء ، مثل حقيقة أن هذه بيانات صورة بتدرج الرمادي على سبيل المثال ، لكنها بداية رائعة.
قدم نموذجك مع خدمة TensorFlow
أضف عنوان URI لتوزيع خدمة TensorFlow كمصدر حزمة:
نحن التحضير لتثبيت TensorFlow خدمة باستخدام الموهبة منذ يدير هذا Colab في بيئة ديبيان. سنقوم بإضافة tensorflow-model-server
الحزمة إلى قائمة الحزم التي الموهبة يعرف عنه. لاحظ أننا نعمل كجذر.
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.
قم بتثبيت خدمة TensorFlow
هذا كل ما تحتاجه - سطر أوامر واحد!
{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) ...
ابدأ تشغيل خدمة TensorFlow
هذا هو المكان الذي نبدأ فيه تشغيل خدمة TensorFlow وتحميل نموذجنا. بعد أن يتم تحميله يمكننا البدء في تقديم طلبات الاستدلال باستخدام REST. هناك بعض المعايير المهمة:
-
rest_api_port
: المنفذ التي ستستخدمها لطلبات REST. -
model_name
: سوف تستخدم هذا في URL طلبات REST. يمكن أن يكون أي شيء. -
model_base_path
: هذا هو المسار إلى الدليل الذي قمت بحفظ النموذج الخاص بك.
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
قم بتقديم طلب إلى النموذج الخاص بك في خدمة TensorFlow
أولاً ، دعنا نلقي نظرة على مثال عشوائي من بيانات الاختبار الخاصة بنا.
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]]))
حسنًا ، هذا يبدو ممتعًا. ما مدى صعوبة التعرف على ذلك؟ لنقم الآن بإنشاء كائن JSON لمجموعة من ثلاثة طلبات استدلال ، ونرى كيف يتعرف نموذجنا على الأشياء:
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]]]]}
تقديم طلبات REST
أحدث نسخة من
سنرسل طلب توقع باعتباره POST إلى نقطة نهاية REST لخادمنا ، ونمرره بثلاثة أمثلة. سنطلب من الخادم الخاص بنا تزويدنا بأحدث إصدار من الخدمة من خلال عدم تحديد إصدار معين.
# 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]))
نسخة معينة من الخدمة
الآن دعنا نحدد إصدارًا معينًا من ملفنا القابل للخدمة. نظرًا لأن لدينا واحدًا فقط ، فلنختار الإصدار 1. وسنلقي نظرة أيضًا على النتائج الثلاث.
# 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]))