本文詳解training.py是如何編寫的。

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

N_CLASSES = 2 # 二分類問題，只有是還是否，即0，1
IMG_W = 208  # resize the image, if the input image is too large, training will be very slow.
IMG_H = 208  # 影象為208*208的尺寸
BATCH_SIZE = 16
CAPACITY = 2000  # 佇列最大容量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
# 定義開始訓練的函式
def run_training():
    # 訓練的圖片存放的位置
    train_dir = '/Users/arcstone_mems_108/PycharmProjects/catsvsdogs/data/train/'
    # 輸出檔案的位置
 
    logs_train_dir = '/Users/arcstone_mems_108/PycharmProjects/catsvsdogs/logs/train/'
    # 呼叫input_data檔案的get_files()函式獲得image_list, label_list
    train, train_label = input_data.get_files(train_dir)
    # 獲得image_batch, label_batch
    train_batch, train_label_batch = input_data.get_batch(train,
                                                          train_label,
                                                          IMG_W,
                                                          IMG_H,
                                                          BATCH_SIZE,
                                                          CAPACITY)
    # 進行前向訓練，獲得迴歸值
 
    train_logits = model.inference(train_batch, BATCH_SIZE, N_CLASSES)
    # 計算獲得損失值loss
    train_loss = model.losses(train_logits, train_label_batch)
    # 對損失值進行優化
    train_op = model.trainning(train_loss, learning_rate)
    # 根據計算得到的損失值，計算出分類準確率
    train__acc = model.evaluation(train_logits, train_label_batch)
    # 將圖形、訓練過程合併在一起
    summary_op = tf.summary.merge_all()
    # 新建會話
    sess = tf.Session()
    # 將訓練日誌寫入到logs_train_dir的資料夾內
    train_writer = tf.summary.FileWriter(logs_train_dir, sess.graph)
    # 儲存變數
    saver = tf.train.Saver()
    # 執行訓練過程，初始化變數
    sess.run(tf.global_variables_initializer())
    # 建立一個執行緒協調器，用來管理之後在Session中啟動的所有執行緒
    coord = tf.train.Coordinator()
    # 啟動入隊的執行緒，一般情況下，系統有多少個核，就會啟動多少個入隊執行緒（入隊具體使用多少個執行緒在tf.train.batch中定義）;
    threads = tf.train.start_queue_runners(sess=sess, coord=coord)

    try:
        for step in np.arange(MAX_STEP):
            # 使用 coord.should_stop()來查詢是否應該終止所有執行緒，當檔案佇列（queue）中的所有檔案都已經讀取出列的時候，
            # 會丟擲一個 OutofRangeError 的異常，這時候就應該停止Sesson中的所有執行緒了;
            if coord.should_stop():
                break
            _, tra_loss, tra_acc = sess.run([train_op, train_loss, train__acc])
            # 每50步列印一次損失值和準確率
            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步儲存一次訓練得到的模型
            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)
    # 如果讀取到檔案佇列末尾會丟擲此異常
    except tf.errors.OutOfRangeError:
        print('Done training -- epoch limit reached')
    finally:
        coord.request_stop()       # 使用coord.request_stop()來發出終止所有執行緒的命令

    coord.join(threads)            # coord.join(threads)把執行緒加入主執行緒，等待threads結束
    sess.close()                   # 關閉會話

def main():
    run_training()


if __name__ == '__main__':
    main()