深度学习：TensorFlow2构建、保存、加载神经网络模型【经典流程】

2024-09-02 03:48
文章标签 加载 流程 学习 构建 保存 深度 模型 经典 神经网络 tensorflow2

本文主要是介绍深度学习：TensorFlow2构建、保存、加载神经网络模型【经典流程】，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

一、network.save_weights、network.load_weights

保存模型的参数，加载已保存的参数的network的结构必须和之前的network的所有结构一模一样

import osos.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'  # 放在 import tensorflow as tf 之前才有效import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers, optimizers, datasets# 一、获取数据集
(X_train, Y_train), (X_test, Y_test) = datasets.mnist.load_data()
print('X_train.shpae = {0}，Y_train.shpae = {1}------------type(X_train) = {2}，type(Y_train) = {3}'.format(X_train.shape, Y_train.shape, type(X_train), type(Y_train)))# 二、数据处理
# 预处理函数：将numpy数据转为tensor
def preprocess(x, y):x = tf.cast(x, dtype=tf.float32) / 255.x = tf.reshape(x, [28 * 28])y = tf.cast(y, dtype=tf.int32)y = tf.one_hot(y, depth=10)return x, y# 2.1 处理训练集
# print('X_train.shpae = {0}，Y_train.shpae = {1}------------type(X_train) = {2}，type(Y_train) = {3}'.format(X_train.shape, Y_train.shape, type(X_train), type(Y_train)))
dataset_train = tf.data.Dataset.from_tensor_slices((X_train, Y_train))  # 此步骤自动将numpy类型的数据转为tensor
dataset_train = dataset_train.map(preprocess)  # 调用map()函数批量修改每一个元素数据的数据类型
dataset_train = dataset_train.shuffle(len(X_train))  # 打散dataset_train中的样本顺序，防止图片的原始顺序对神经网络性能的干扰
print('dataset_train = {0}，type(dataset_train) = {1}'.format(dataset_train, type(dataset_train)))
batch_size_train = 20000  # 每个batch里的样本数量设置100-200之间合适。
dataset_batch_train = dataset_train.batch(batch_size_train)  # 将dataset_batch_train中每sample_num_of_each_batch_train张图片分为一个batch，读取一个batch相当于一次性并行读取sample_num_of_each_batch_train张图片
print('dataset_batch_train = {0}，type(dataset_batch_train) = {1}'.format(dataset_batch_train, type(dataset_batch_train)))
# 2.2 处理测试集
dataset_test = tf.data.Dataset.from_tensor_slices((X_test, Y_test))  # 此步骤自动将numpy类型的数据转为tensor
dataset_test = dataset_test.map(preprocess)  # 调用map()函数批量修改每一个元素数据的数据类型
dataset_test = dataset_test.shuffle(len(X_test))  # 打散样本顺序，防止图片的原始顺序对神经网络性能的干扰
batch_size_test = 5000  # 每个batch里的样本数量设置100-200之间合适。
dataset_batch_test = dataset_test.batch(batch_size_test)  # 将dataset_test中每sample_num_of_each_batch_test张图片分为一个batch，读取一个batch相当于一次性并行读取sample_num_of_each_batch_test张图片# 三、构建神经网络结构：Dense 表示全连接神经网络，激活函数用 relu
network = keras.Sequential([layers.Dense(500, activation=tf.nn.relu),  # 降维：784-->500layers.Dense(300, activation=tf.nn.relu),  # 降维：500-->300layers.Dense(100, activation=tf.nn.relu),  # 降维：300-->100layers.Dense(10)])  # 降维：100-->10，最后一层一般不需要在此处指定激活函数，在计算Loss的时候会自动运用激活函数
network.build(input_shape=[None, 784])  # 28*28=784，None表示样本数量，是不确定的值。
network.summary()  # 打印神经网络model的简要信息# 四、设置神经网络各个参数
network.compile(optimizer=optimizers.Adam(lr=0.01),loss=tf.losses.CategoricalCrossentropy(from_logits=True),metrics=['accuracy'])# 五、给神经网络喂数据，训练神经网络模型参数
print('\n++++++++++++++++++++++++++++++++++++++++++++Training 阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
network.fit(dataset_batch_train, epochs=5, validation_data=dataset_batch_test, validation_freq=2)  # validation_freq参数表示每多少个epoch做一次验证/validation
print('++++++++++++++++++++++++++++++++++++++++++++Training 阶段：结束++++++++++++++++++++++++++++++++++++++++++++')# 六、模型评估 test/evluation
print('\n++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
network.evaluate(dataset_batch_test)
print('++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++')network.save_weights('weights.ckpt')
print('\n================saved weights================')
del network
print('================del network================')# 七、创建一个和所加载参数的原始network一模一样的network
print('================创建一个和所加载参数的原始network一模一样的network================')
network = keras.Sequential([layers.Dense(500, activation=tf.nn.relu),  # 降维：784-->500layers.Dense(300, activation=tf.nn.relu),  # 降维：500-->300layers.Dense(100, activation=tf.nn.relu),  # 降维：300-->100layers.Dense(10)])  # 降维：100-->10，最后一层一般不需要在此处指定激活函数，在计算Loss的时候会自动运用激活函数
network.build(input_shape=[None, 784])  # 28*28=784，None表示样本数量，是不确定的值。
network.compile(optimizer=optimizers.Adam(lr=0.01),loss=tf.losses.CategoricalCrossentropy(from_logits=True),metrics=['accuracy'])
network.load_weights('weights.ckpt')
print('================loaded weights================')# 八、模型评估 test/evluation
print('\n++++++++++++++++++++++++++++++++++++++++++++加载weights后--->Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
network.evaluate(dataset_batch_test)
print('++++++++++++++++++++++++++++++++++++++++++++加载weights后--->Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++')# 九、模型上线应用
sample = next(iter(dataset_batch_test))  # 从 dataset_batch_test 中取一个batch数据做模拟
x = sample[0]
y = sample[1]  # one-hot
pred = network.predict(x)  # [b, 10]
y = tf.argmax(y, axis=1)  # convert back to number
pred = tf.argmax(pred, axis=1)
print('\n++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
print(pred)
print(y)
print('++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：结束++++++++++++++++++++++++++++++++++++++++++++')

打印结果：

X_train.shpae = (60000, 28, 28)，Y_train.shpae = (60000,)------------type(X_train) = <class 'numpy.ndarray'>，type(Y_train) = <class 'numpy.ndarray'>
dataset_train = <ShuffleDataset shapes: ((784,), (10,)), types: (tf.float32, tf.float32)>，type(dataset_train) = <class 'tensorflow.python.data.ops.dataset_ops.ShuffleDataset'>
dataset_batch_train = <BatchDataset shapes: ((None, 784), (None, 10)), types: (tf.float32, tf.float32)>，type(dataset_batch_train) = <class 'tensorflow.python.data.ops.dataset_ops.BatchDataset'>
Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 500)               392500    
_________________________________________________________________
dense_1 (Dense)              (None, 300)               150300    
_________________________________________________________________
dense_2 (Dense)              (None, 100)               30100     
_________________________________________________________________
dense_3 (Dense)              (None, 10)                1010      
=================================================================
Total params: 573,910
Trainable params: 573,910
Non-trainable params: 0
_________________________________________________________________++++++++++++++++++++++++++++++++++++++++++++Training 阶段：开始++++++++++++++++++++++++++++++++++++++++++++
Epoch 1/5
3/3 [==============================] - 2s 113ms/step - loss: 2.7174 - accuracy: 0.1086
Epoch 2/5
3/3 [==============================] - 3s 492ms/step - loss: 2.6596 - accuracy: 0.1666 - val_loss: 1.6333 - val_accuracy: 0.4709
Epoch 3/5
3/3 [==============================] - 2s 115ms/step - loss: 1.5516 - accuracy: 0.4968
Epoch 4/5
3/3 [==============================] - 2s 255ms/step - loss: 1.0690 - accuracy: 0.6475 - val_loss: 0.7587 - val_accuracy: 0.7859
Epoch 5/5
3/3 [==============================] - 2s 115ms/step - loss: 0.7137 - accuracy: 0.7955
++++++++++++++++++++++++++++++++++++++++++++Training 阶段：结束++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++
2/2 [==============================] - 0s 22ms/step - loss: 0.5240 - accuracy: 0.8493
++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++================saved weights================
================del network================
================创建一个和所加载参数的原始network一模一样的network================
================loaded weights================++++++++++++++++++++++++++++++++++++++++++++加载weights后--->Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++
2/2 [==============================] - 0s 22ms/step - loss: 0.5223 - accuracy: 0.8486
++++++++++++++++++++++++++++++++++++++++++++加载weights后--->Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：开始++++++++++++++++++++++++++++++++++++++++++++
tf.Tensor([6 3 7 ... 5 1 0], shape=(5000,), dtype=int64)
tf.Tensor([6 3 7 ... 3 1 0], shape=(5000,), dtype=int64)
++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：结束++++++++++++++++++++++++++++++++++++++++++++Process finished with exit code 0

二、network.save()、network.load()

保存整个模型，加载后再根据network的通常做法进行操作。

import osos.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'  # 放在 import tensorflow as tf 之前才有效import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers, optimizers, datasets# 一、获取数据集
(X_train, Y_train), (X_test, Y_test) = datasets.mnist.load_data()
print('X_train.shpae = {0}，Y_train.shpae = {1}------------type(X_train) = {2}，type(Y_train) = {3}'.format(X_train.shape, Y_train.shape, type(X_train), type(Y_train)))# 二、数据处理
# 预处理函数：将numpy数据转为tensor
def preprocess(x, y):x = tf.cast(x, dtype=tf.float32) / 255.x = tf.reshape(x, [28 * 28])y = tf.cast(y, dtype=tf.int32)y = tf.one_hot(y, depth=10)return x, y# 2.1 处理训练集
# print('X_train.shpae = {0}，Y_train.shpae = {1}------------type(X_train) = {2}，type(Y_train) = {3}'.format(X_train.shape, Y_train.shape, type(X_train), type(Y_train)))
dataset_train = tf.data.Dataset.from_tensor_slices((X_train, Y_train))  # 此步骤自动将numpy类型的数据转为tensor
dataset_train = dataset_train.map(preprocess)  # 调用map()函数批量修改每一个元素数据的数据类型
dataset_train = dataset_train.shuffle(len(X_train))  # 打散dataset_train中的样本顺序，防止图片的原始顺序对神经网络性能的干扰
print('dataset_train = {0}，type(dataset_train) = {1}'.format(dataset_train, type(dataset_train)))
batch_size_train = 20000  # 每个batch里的样本数量设置100-200之间合适。
dataset_batch_train = dataset_train.batch(batch_size_train)  # 将dataset_batch_train中每sample_num_of_each_batch_train张图片分为一个batch，读取一个batch相当于一次性并行读取sample_num_of_each_batch_train张图片
print('dataset_batch_train = {0}，type(dataset_batch_train) = {1}'.format(dataset_batch_train, type(dataset_batch_train)))
# 2.2 处理测试集
dataset_test = tf.data.Dataset.from_tensor_slices((X_test, Y_test))  # 此步骤自动将numpy类型的数据转为tensor
dataset_test = dataset_test.map(preprocess)  # 调用map()函数批量修改每一个元素数据的数据类型
dataset_test = dataset_test.shuffle(len(X_test))  # 打散样本顺序，防止图片的原始顺序对神经网络性能的干扰
batch_size_test = 5000  # 每个batch里的样本数量设置100-200之间合适。
dataset_batch_test = dataset_test.batch(batch_size_test)  # 将dataset_test中每sample_num_of_each_batch_test张图片分为一个batch，读取一个batch相当于一次性并行读取sample_num_of_each_batch_test张图片# 三、构建神经网络结构：Dense 表示全连接神经网络，激活函数用 relu
network = keras.Sequential([layers.Dense(500, activation=tf.nn.relu),  # 降维：784-->500layers.Dense(300, activation=tf.nn.relu),  # 降维：500-->300layers.Dense(100, activation=tf.nn.relu),  # 降维：300-->100layers.Dense(10)])  # 降维：100-->10，最后一层一般不需要在此处指定激活函数，在计算Loss的时候会自动运用激活函数
network.build(input_shape=[None, 784])  # 28*28=784，None表示样本数量，是不确定的值。
network.summary()  # 打印神经网络model的简要信息# 四、设置神经网络各个参数
network.compile(optimizer=optimizers.Adam(lr=0.01),loss=tf.losses.CategoricalCrossentropy(from_logits=True),metrics=['accuracy'])# 五、给神经网络喂数据，训练神经网络模型参数
print('\n++++++++++++++++++++++++++++++++++++++++++++Training 阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
network.fit(dataset_batch_train, epochs=5, validation_data=dataset_batch_test, validation_freq=2)  # validation_freq参数表示每多少个epoch做一次验证/validation
print('++++++++++++++++++++++++++++++++++++++++++++Training 阶段：结束++++++++++++++++++++++++++++++++++++++++++++')# 六、模型评估 test/evluation
print('\n++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
network.evaluate(dataset_batch_test)
print('++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++')network.save('model.h5')
print('\n================saved total model================')
del network
print('================del network================')# 七、从磁盘加载保存的整体模型(包括所有参数、结构...)
print('================loaded model from file================')
network = tf.keras.models.load_model('model.h5', compile=False)
network.compile(optimizer=optimizers.Adam(lr=0.01),loss=tf.losses.CategoricalCrossentropy(from_logits=True),metrics=['accuracy'])# 八、模型评估 test/evluation
print('\n++++++++++++++++++++++++++++++++++++++++++++从磁盘加载整个model后--->Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
network.evaluate(dataset_batch_test)
print('++++++++++++++++++++++++++++++++++++++++++++从磁盘加载整个model后--->Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++')# 九、模型上线应用
sample = next(iter(dataset_batch_test))  # 从 dataset_batch_test 中取一个batch数据做模拟
x = sample[0]
y = sample[1]  # one-hot
pred = network.predict(x)  # [b, 10]
y = tf.argmax(y, axis=1)  # convert back to number
pred = tf.argmax(pred, axis=1)
print('\n++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：开始++++++++++++++++++++++++++++++++++++++++++++')
print(pred)
print(y)
print('++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：结束++++++++++++++++++++++++++++++++++++++++++++')

打印结果：

X_train.shpae = (60000, 28, 28)，Y_train.shpae = (60000,)------------type(X_train) = <class 'numpy.ndarray'>，type(Y_train) = <class 'numpy.ndarray'>
dataset_train = <ShuffleDataset shapes: ((784,), (10,)), types: (tf.float32, tf.float32)>，type(dataset_train) = <class 'tensorflow.python.data.ops.dataset_ops.ShuffleDataset'>
dataset_batch_train = <BatchDataset shapes: ((None, 784), (None, 10)), types: (tf.float32, tf.float32)>，type(dataset_batch_train) = <class 'tensorflow.python.data.ops.dataset_ops.BatchDataset'>
Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 500)               392500    
_________________________________________________________________
dense_1 (Dense)              (None, 300)               150300    
_________________________________________________________________
dense_2 (Dense)              (None, 100)               30100     
_________________________________________________________________
dense_3 (Dense)              (None, 10)                1010      
=================================================================
Total params: 573,910
Trainable params: 573,910
Non-trainable params: 0
_________________________________________________________________++++++++++++++++++++++++++++++++++++++++++++Training 阶段：开始++++++++++++++++++++++++++++++++++++++++++++
Epoch 1/5
3/3 [==============================] - 2s 119ms/step - loss: 2.4869 - accuracy: 0.2464
Epoch 2/5
3/3 [==============================] - 2s 514ms/step - loss: 3.5169 - accuracy: 0.3786 - val_loss: 1.5471 - val_accuracy: 0.5026
Epoch 3/5
3/3 [==============================] - 2s 116ms/step - loss: 1.4532 - accuracy: 0.5238
Epoch 4/5
3/3 [==============================] - 2s 273ms/step - loss: 0.9930 - accuracy: 0.6789 - val_loss: 0.6357 - val_accuracy: 0.8010
Epoch 5/5
3/3 [==============================] - 2s 112ms/step - loss: 0.6005 - accuracy: 0.8118
++++++++++++++++++++++++++++++++++++++++++++Training 阶段：结束++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++
2/2 [==============================] - 0s 24ms/step - loss: 0.4489 - accuracy: 0.8735
++++++++++++++++++++++++++++++++++++++++++++Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++================saved total model================
================del network================
================loaded model from file================++++++++++++++++++++++++++++++++++++++++++++从磁盘加载整个model后--->Evluation 阶段：开始++++++++++++++++++++++++++++++++++++++++++++
2/2 [==============================] - 0s 21ms/step - loss: 0.4505 - accuracy: 0.8729
++++++++++++++++++++++++++++++++++++++++++++从磁盘加载整个model后--->Evluation 阶段：结束++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：开始++++++++++++++++++++++++++++++++++++++++++++
tf.Tensor([9 0 9 ... 5 1 9], shape=(5000,), dtype=int64)
tf.Tensor([9 0 9 ... 5 1 9], shape=(5000,), dtype=int64)
++++++++++++++++++++++++++++++++++++++++++++加载weights后--->应用阶段：结束++++++++++++++++++++++++++++++++++++++++++++Process finished with exit code 0

三、tf.saved_model.save()、tf.saved_model.load()、

保存为可以被其他语言(比如：C++)调用的格式
在这里插入图片描述

这篇关于深度学习：TensorFlow2构建、保存、加载神经网络模型【经典流程】的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！


http://www.chinasem.cn/article/1128904。

相关文章

Java中Redisson 的原理深度解析

Java中Redisson 的原理深度解析

《Java中Redisson的原理深度解析》Redisson是一个高性能的Redis客户端,它通过将Redis数据结构映射为Java对象和分布式对象,实现了在Java应用中方便地使用Redis,本文... 目录前言一、核心设计理念二、核心架构与通信层1. 基于 Netty 的异步非阻塞通信2. 编解码器三、
阅读更多...
Java HashMap的底层实现原理深度解析

Java HashMap的底层实现原理深度解析

《JavaHashMap的底层实现原理深度解析》HashMap基于数组+链表+红黑树结构,通过哈希算法和扩容机制优化性能,负载因子与树化阈值平衡效率,是Java开发必备的高效数据结构,本文给大家介绍... 目录一、概述：HashMap的宏观结构二、核心数据结构解析1. 数组（桶数组）2. 链表节点（Node
阅读更多...
Java 虚拟线程的创建与使用深度解析

Java 虚拟线程的创建与使用深度解析

《Java虚拟线程的创建与使用深度解析》虚拟线程是Java19中以预览特性形式引入,Java21起正式发布的轻量级线程,本文给大家介绍Java虚拟线程的创建与使用,感兴趣的朋友一起看看吧... 目录一、虚拟线程简介1.1 什么是虚拟线程？1.2 为什么需要虚拟线程？二、虚拟线程与平台线程对比代码对比示例：三
阅读更多...
Nginx分布式部署流程分析

Nginx分布式部署流程分析

《Nginx分布式部署流程分析》文章介绍Nginx在分布式部署中的反向代理和负载均衡作用,用于分发请求、减轻服务器压力及解决session共享问题,涵盖配置方法、策略及Java项目应用,并提及分布式事... 目录分布式部署NginxJava中的代理代理分为正向代理和反向代理正向代理反向代理Nginx应用场景
阅读更多...
Python函数作用域与闭包举例深度解析

Python函数作用域与闭包举例深度解析

《Python函数作用域与闭包举例深度解析》Python函数的作用域规则和闭包是编程中的关键概念,它们决定了变量的访问和生命周期,：本文主要介绍Python函数作用域与闭包的相关资料,文中通过代码... 目录1. 基础作用域访问示例1：访问全局变量示例2：访问外层函数变量2. 闭包基础示例3：简单闭包示例4
阅读更多...
MyBatis延迟加载与多级缓存全解析

MyBatis延迟加载与多级缓存全解析

《MyBatis延迟加载与多级缓存全解析》文章介绍MyBatis的延迟加载与多级缓存机制,延迟加载按需加载关联数据提升性能,一级缓存会话级默认开启,二级缓存工厂级支持跨会话共享,增删改操作会清空对应缓... 目录MyBATis延迟加载策略一对多示例一对多示例MyBatis框架的缓存一级缓存二级缓存MyBat
阅读更多...
Spring Boot分层架构详解之从Controller到Service再到Mapper的完整流程(用户管理系统为例)

Spring Boot分层架构详解之从Controller到Service再到Mapper的完整流程(用户管理系统为例)

《SpringBoot分层架构详解之从Controller到Service再到Mapper的完整流程(用户管理系统为例)》本文将以一个实际案例（用户管理系统）为例,详细解析SpringBoot中Co... 目录引言：为什么学习Spring Boot分层架构？第一部分：Spring Boot的整体架构1.1
阅读更多...
nodejs打包作为公共包使用的完整流程

nodejs打包作为公共包使用的完整流程

《nodejs打包作为公共包使用的完整流程》在Node.js项目中,打包和部署是发布应用的关键步骤,：本文主要介绍nodejs打包作为公共包使用的相关资料,文中通过代码介绍的非常详细,需要的朋友可... 目录前言一、前置准备二、创建与编码三、一键构建四、本地“白嫖”测试（可选）五、发布公共包六、常见踩坑提醒
阅读更多...
Ubuntu向多台主机批量传输文件的流程步骤

Ubuntu向多台主机批量传输文件的流程步骤

《Ubuntu向多台主机批量传输文件的流程步骤》：本文主要介绍在Ubuntu中批量传输文件到多台主机的方法,需确保主机互通、用户名密码统一及端口开放,通过安装sshpass工具,准备包含目标主机信... 目录Ubuntu 向多台主机批量传输文件1.安装 sshpass2.准备主机列表文件3.创建一个批处理脚
阅读更多...
一个Java的main方法在JVM中的执行流程示例详解

一个Java的main方法在JVM中的执行流程示例详解

《一个Java的main方法在JVM中的执行流程示例详解》main方法是Java程序的入口点,程序从这里开始执行,：本文主要介绍一个Java的main方法在JVM中执行流程的相关资料,文中通过代码... 目录第一阶段：加载 (Loading)第二阶段：链接 (Linking)第三阶段：初始化 (Initia
阅读更多...

Copyright China编程 23002807@qq.com

沪ICP备2023033072号-2