使用Tensorflow来读取训练自己的数据（三）

本文详解training.py是如何编写的。

import os
import numpy as np
import tensorflow as tf
import input_data
import model

N_CLASSES = 2 # 二分类问题，只有是还是否，即0，1
IMG_W = 208  # resize the image, if the input image is too large, training will be very slow.
IMG_H = 208  # 图像为208*208的尺寸
BATCH_SIZE = 16
CAPACITY = 2000  # 队列最大容量2000
MAX_STEP = 10000 # with current parameters, it is suggested to use MAX_STEP>10k
learning_rate = 0.0001 # with current parameters, it is suggested to use learning rate<0.0001
# 定义开始训练的函数
def run_training():
    # 训练的图片存放的位置
    train_dir = '/Users/arcstone_mems_108/PycharmProjects/catsvsdogs/data/train/'
    # 输出文件的位置
    logs_train_dir = '/Users/arcstone_mems_108/PycharmProjects/catsvsdogs/logs/train/'
    # 调用input_data文件的get_files()函数获得image_list, label_list
    train, train_label = input_data.get_files(train_dir)
    # 获得image_batch, label_batch
    train_batch, train_label_batch = input_data.get_batch(train,
                                                          train_label,
                                                          IMG_W,
                                                          IMG_H,
                                                          BATCH_SIZE,
                                                          CAPACITY)
    # 进行前向训练，获得回归值
    train_logits = model.inference(train_batch, BATCH_SIZE, N_CLASSES)
    # 计算获得损失值loss
    train_loss = model.losses(train_logits, train_label_batch)
    # 对损失值进行优化
    train_op = model.trainning(train_loss, learning_rate)
    # 根据计算得到的损失值，计算出分类准确率
    train__acc = model.evaluation(train_logits, train_label_batch)
    # 将图形、训练过程合并在一起
    summary_op = tf.summary.merge_all()
    # 新建会话
    sess = tf.Session()
    # 将训练日志写入到logs_train_dir的文件夹内
    train_writer = tf.summary.FileWriter(logs_train_dir, sess.graph)
    # 保存变量
    saver = tf.train.Saver()
    # 执行训练过程，初始化变量
    sess.run(tf.global_variables_initializer())
    # 创建一个线程协调器，用来管理之后在Session中启动的所有线程
    coord = tf.train.Coordinator()
    # 启动入队的线程，一般情况下，系统有多少个核，就会启动多少个入队线程（入队具体使用多少个线程在tf.train.batch中定义）;
    threads = tf.train.start_queue_runners(sess=sess, coord=coord)

    try:
        for step in np.arange(MAX_STEP):
            # 使用 coord.should_stop()来查询是否应该终止所有线程，当文件队列（queue）中的所有文件都已经读取出列的时候，
            # 会抛出一个 OutofRangeError 的异常，这时候就应该停止Sesson中的所有线程了;
            if coord.should_stop():
                break
            _, tra_loss, tra_acc = sess.run([train_op, train_loss, train__acc])
            # 每50步打印一次损失值和准确率
            if step % 50 == 0:
                print('Step %d, train loss = %.2f, train accuracy = %.2f%%' % (step, tra_loss, tra_acc * 100.0))
                summary_str = sess.run(summary_op)
                train_writer.add_summary(summary_str, step)
            # 每2000步保存一次训练得到的模型
            if step % 2000 == 0 or (step + 1) == MAX_STEP:
                checkpoint_path = os.path.join(logs_train_dir, 'model.ckpt')
                saver.save(sess, checkpoint_path, global_step=step)
    # 如果读取到文件队列末尾会抛出此异常
    except tf.errors.OutOfRangeError:
        print('Done training -- epoch limit reached')
    finally:
        coord.request_stop()       # 使用coord.request_stop()来发出终止所有线程的命令

    coord.join(threads)            # coord.join(threads)把线程加入主线程，等待threads结束
    sess.close()                   # 关闭会话

def main():
    run_training()


if __name__ == '__main__':
    main()