Python实战 | 基于 Flask 部署 Keras 深度学习模型开发者社区

link之家

链接快照平台

输入网页链接，自动生成快照
标签化管理网页链接

Python实战 | 基于 Flask 部署 Keras 深度学习模型

from keras.models import Model
from keras.applications.vgg16 import VGG16, preprocess_input
import keras.backend.tensorflow_backend as KTF
import tensorflow as tf
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1" #使用GPU
# 按需占用GPU显存
gpu_options = tf.GPUOptions(allow_growth=True)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
KTF.set_session(sess)
# 构建model
base_model = VGG16(weights=‘imagenet’, include_top=True)
model = Model(inputs=base_model.input,
outputs=base_model.get_layer(layer).output) # 获取指定层的输出值，layer为层名
# 进行预测
img = load_image(img_name, target_size=(224, 224))  # 加载图片并resize成224x224
# 图像预处理
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
feature = model.predict(x) # 提取特征

# 使用python中的time模块
import time
t0 = time.time()
图像处理和特征提取
print(time.time()-t0) #耗时，以秒为单位

pip install redis
import redis

# 1. 导入redis
from redis import StrictRedis
# 2. 连接数据库，指定host,端口号，数据库
r = StrictRedis(host=‘localhost’, port=6379, db=2)
# 3. 存储到redis中
r.set('test1', 'value1')  # 单个数据存储
r.set('test2', 'value2')
# 4. 从redis中获取值
r.get('test1')
# 5. 批量操作
r.mset(k1='v1', k2='v2')
r.mset({'k1':'v1', 'k2':'v2'})
r.mget('k1', 'k2')
r.mget(['k1', 'k2'])

import numpy as np
from redis import StrictRedis
r = StrictRedis(host=‘localhost’, port=6379, db=2)
x1 = np.array(([0.2,0.1,0.6],[10.2,4.2,0.9]))
r.set('test1', x1)
r.get('test1')
>>> b'[[ 0.2  0.1  0.6]\n [10.2  4.2  0.9]]'
type(r.get('test1')) #获取后的数据类型
>>> <class 'bytes'>

import pickle
r.set('test2', pickle.dumps(x1))
pickle.loads(r.get('test2'))
>>> array([[ 0.2,  0.1,  0.6],
         [10.2,  4.2,  0.9]])

pip install flask
import flask # 导入
flask.__version__ # 版本
>>> '1.1.1' #当前版本

# hello.py
from flask import Flask, request
app = Flask(__name__) #创建Flask类的实例，第一个参数是模块或者包的名称
app.config['JSON_AS_ASCII']=False # 支持中文显示
@app.route('/', methods=['GET', 'POST']) # 使用methods参数处理不同HTTP方法
def home():
    return 'Hello, Flask'
if __name__ == '__main__':
    app.run()

app.run(host="0.0.0.0", port="5000", debug=True, processes=2, threaded=False)

@app.route("/me")
def me_api():
    user = get_current_user()
    return {
        "username": user.username,
        "theme": user.theme,
        "image": url_for("user_image", filename=user.image),

@app.route("/users")
def users_api():
    users = get_all_users()
    return jsonify([user.to_json() for user in users])

$ export FLASK_APP=my_application
$ export FLASK_ENV=development
$ flask run

if __name__ == '__main__':
    app.run()

pip install gunicorn

pip install gevent
pip install eventlet
pip install greenlet

# gunicorn_conf.py
bind: '0.0.0.0:5000' # 监听地址和端口号
workers = 2 # 进程数
worker_class = 'sync' #工作模式，可选sync, gevent, eventlet, gthread, tornado等
threads = 1 # 指定每个进程的线程数，默认为1
worker_connections = 2000 # 最大客户并发量
timeout = 30 # 超时时间，默认30s
reload = True # 开发模式，代码更新时自动重启
daemon = False # 守护Gunicorn进程，默认False
accesslog = './logs/access.log' # 访问日志文件
errorlog = './logs/error.log'
loglevel = 'debug' # 日志输出等级，debug, info, warning, error, critical

#flask_feature.app
import numpy as np
from flask import Flask, jsonify
from keras.models import Model
from keras.applications.vgg16 import VGG16
from keras.backend.tensorflow_backend import set_session
app = Flask(__name__)
app.config['JSON_AS_ASCII']=False
@app.route("/", methods=["GET", "POST"])
def feature():
    img_feature = extract()
    return jsonify({'result':'true', 'msg':'成功'})
def extract(img_name):
    # 图像预处理
    img = load_image(img_name, target_size=(feature_params["size"], feature_params["size"]))
    x = image.img_to_array(img)
    x = np.expand_dims(x, axis=0)
    x = preprocess_input(x)
    with graph.as_default():
        set_session(sess)
        res = model.predict(x)
    return res
if __name__ == '__main__':
    tf_config = some_custom_config
    sess = tf.Session(config=tf_config)
    set_session(sess)
    base_model = VGG16(weights=model_weights, include_top=True)
    model = Model(inputs=base_model.input,
                    outputs=base_model.get_layer(layer).output)
    graph = tf.get_default_graph()
    app.run()

"Tensor Tensor(\"pooling/Mean:0\", shape=(?, 1280), dtype=float32) is not an element of this graph.",

import tensorflow as tf
global graph, model
graph = tf.get_default_graph()
#当需要进行预测的时候
with graph.as_default():
    y = model.predict(x)

2 root error(s) found.\n
(0) Failed precondition: Error while reading resource variable block5_conv2/kernel from Container: localhost. This could mean that the variable was uninitialized. Not found: Container localhost does not exist. (Could not find resource: localhost/block5_conv2/kernel)\n\t [[{{node block5_conv2/convolution/ReadVariableOp}}]]\n\t [[fc2/Relu/_7]]\n
(1) Failed precondition: Error while reading resource variable block5_conv2/kernel from Container: localhost. This could mean that the variable was uninitialized. Not found: Container localhost does not exist. (Could not find resource: localhost/block5_conv2/kernel)\n\t [[{{node block5_conv2/convolution/ReadVariableOp}}]]\n0 successful operations.\n0 derived errors ignored."

from tensorflow.python.keras.backend import set_session
from tensorflow.python.keras.models import load_model
tf_config = some_custom_config
sess = tf.Session(config=tf_config)
graph = tf.get_default_graph()
# IMPORTANT: models have to be loaded AFTER SETTING THE SESSION for keras!
# Otherwise, their weights will be unavailable in the threads after the session there has been set
set_session(sess)
model = load_model(...)
# 在每一个request中：
global sess
global graph
with graph.as_default():
    set_session(sess)
    model.predict(...)

# on thread 1
session = tf.Session(graph=tf.Graph())
with session.graph.as_default():
    k.backend.set_session(session)
    model = k.models.load_model(filepath)
# on thread 2
with session.graph.as_default():
    k.backend.set_session(session)

Python实战 | 基于 Flask 部署 Keras 深度学习模型

Python实战 | 基于 Flask 部署 Keras 深度学习模型

1、项目简介

2、项目流程

2.1 图像分类模型

1. 模型的选择

2. 框架选择

3. 代码示例

2.2 模型性能测试

2.3 Redis 使用

1. 基本使用

2. Redis 存储数组

2.4 web 开发框架——Flask

1. 安装使用

2. Flask 响应

3. 运行开发服务器

2.5 使用 Gunicorn

1. 安装及使用

3、代码示例

4、遇到的问题

4.1 Flask 多线程与多进程问题

4.1 Flask 与 Keras 问题

Q1：Tensor is not an element of this graph

Q2：使用 Flask 启动服务，加载两次模型，占用两份显存

4.2 gunicorn 启动服务相关问题

Q1: Failed precondition