TF数据读取队列机制详解
-
TFR文件多线程队列读写操作:
-
-
TFRecod文件写入操作:
-
import tensorflow as tf def _int64_feature(value): # value必须是可迭代对象 # 非int的数据使用bytes取代int64即可 return tf.train.Feature(int64_list=tf.train.Int64List(value=[value])) num_shards = 2 instance_perPshard = 2 for i in range(num_shards): filename = (‘FTR/data.tfrecords-%.5d-of-%.5d‘ % (i, num_shards)) writer = tf.python_io.TFRecordWriter(filename) #<---------书写器打开 for j in range(instance_perPshard): example = tf.train.Example(features=tf.train.Features(feature={ #<---------书写入缓冲区 ‘i‘:_int64_feature(i), ‘j‘:_int64_feature(j) })) writer.write(example.SerializeToString()) #<---------书写入实际文件 writer.close() #<---------书写器关闭
-
-
TFRecod文件读取操作:
-
默认多线程,这个默认的多线程过程用于维护文件名队列
‘‘‘读取TFR‘‘‘ files = ["FTR/data.tfrecords-00000-of-00002","FTR/data.tfrecords-00001-of-00002"] # files = tf.train.match_filenames_once("FTR/data.tfrecords-*") # 输入文件名列表 # 返回QueueRunner & FIFOQueue # 打乱顺序&加入队列 和 输出队列获取文件 属于单独的线程 filename_queue = tf.train.string_input_producer(files, shuffle=False) #<---------输入文件队列 reader = tf.TFRecordReader() #<---------读取器打开 _,serialized_example = reader.read(filename_queue) #<---------读取原始文件 features = tf.parse_single_example( #<---------读取解析后文件 serialized_example, features={ ‘i‘:tf.FixedLenFeature([],tf.int64), ‘j‘:tf.FixedLenFeature([],tf.int64) }) with tf.Session() as sess: tf.global_variables_initializer().run() coord = tf.train.Coordinator() #<---------多线程 threads = tf.train.start_queue_runners(sess=sess,coord=coord) #<---------文件名队列填充线程启动 for i in range(6): print(sess.run([features[‘i‘],features[‘j‘]])) #<---------实际会话中启动读取过程 coord.request_stop() #<---------多线程 coord.join(threads) #<---------多线程
-
-
TFRecod文件打包操作:
-
打包机制:
——————多线程调用前面的节点计算入队
——————批量出队并打包
所以不需要修改解析读取数据过程为循环之类的,可以说很是方便
example_batch, label_batch = tf.train.batch([example, label], #<---------多线程batch生成 batch_size=batch_size, num_threads=3, capacity=capacity)
example_batch, label_batch = tf.train.shuffle_batch([example, label], #<---------多线程随机batch生成 batch_size=batch_size, num_threads=3, capacity=capacity, min_after_dequeue=30) 由于元素太少随机意义就不大了,所以多了个参数
files = ["FTR/data.tfrecords-00000-of-00002","FTR/data.tfrecords-00001-of-00002"] # files = tf.train.match_filenames_once("FTR/data.tfrecords-*") # 输入文件名列表 # 返回QueueRunner & FIFOQueue # 打乱顺序&加入队列 和 输出队列获取文件 属于单独的线程 filename_queue = tf.train.string_input_producer(files, shuffle=False) #<---------输入文件队列 reader = tf.TFRecordReader() #<---------读取 _,serialized_example = reader.read(filename_queue) #<---------读取 features = tf.parse_single_example( #<---------读取 serialized_example, features={ ‘i‘:tf.FixedLenFeature([],tf.int64), ‘j‘:tf.FixedLenFeature([],tf.int64) }) example, label = features[‘i‘], features[‘j‘] batch_size = 2 capacity = 1000 + 3 * batch_size # 入队单个样例,出队batch # 可以指定多个线程同时执行入队操作 example_batch, label_batch = tf.train.batch([example, label], #<---------多线程batch生成 batch_size=batch_size, num_threads=3, capacity=capacity) with tf.Session() as sess: tf.global_variables_initializer().run() coord = tf.train.Coordinator() #<---------多线程管理器 threads = tf.train.start_queue_runners(sess=sess,coord=coord) #<---------文件名队列填充线程启动 for i in range(3): cur_example_batch, cur_label_batch = sess.run([example_batch, label_batch]) print(cur_example_batch, cur_label_batch) coord.request_stop() #<---------多线程关闭 coord.join(threads)
这个输出每一行前为image(代指),后为label,第一行的数据对实际为0-0,0-1:
[0 0] [0 1] [1 1] [0 1] [0 0] [0 1]
-
图片文件使用TFR读写测试:
read的二进制数据直接进行_bytes_feature化就可以写入文件,使用tf.string类型读出图片数据后可以直接decode解码之(推测tf中string对应二进制数据类型)。
把一张图片写入TFR中:
import tensorflow as tf import matplotlib.pyplot as plt def _bytes_feature(value): return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value])) def _int64_feature(value): return tf.train.Feature(int64_list=tf.train.Int64List(value=[value])) img_raw = tf.gfile.FastGFile(‘123123.jpeg‘,‘rb‘).read() filename = (‘FTR/image.tfrecords‘) writer = tf.python_io.TFRecordWriter(filename) #<---------书写 example = tf.train.Example(features=tf.train.Features(feature={ #<---------书写 ‘image‘:_bytes_feature(img_raw), ‘label‘:_int64_feature(1) })) writer.write(example.SerializeToString()) #<---------书写 writer.close()
从TFR中读取图片数据并解码绘制出来:
filename_queue = tf.train.string_input_producer([‘FTR/image.tfrecords‘], shuffle=False) #<---------输入文件队列 reader = tf.TFRecordReader() #<---------读取 _,serialized_example = reader.read(filename_queue) #<---------读取 features = tf.parse_single_example( #<---------读取 serialized_example, features={ ‘image‘:tf.FixedLenFeature([],tf.string), ‘label‘:tf.FixedLenFeature([],tf.int64) }) img = tf.image.decode_jpeg(features[‘image‘]) with tf.Session() as sess: tf.global_variables_initializer().run() coord = tf.train.Coordinator() # <---------多线程 threads = tf.train.start_queue_runners(sess=sess, coord=coord) # <---------文件名队列填充线程启动 # img_raw, label = sess.run([features[‘image‘], features[‘label‘]]) image = sess.run(img) plt.imshow(image) plt.show() coord.request_stop() # <---------多线程 coord.join(threads) # <---------多线程
-
图片文件直接使用队列读写操作:
仅仅示范了维护图片文件名队列的读写,没有过多的其他操作
reader = tf.WholeFileReader():新的读取器,应该是范用性二进制文件读取器
# 导入tensorflow import tensorflow as tf # 新建一个Session with tf.Session() as sess: # 我们要读三幅图片A.jpg, B.jpg, C.jpg filename = [‘123.png‘, ‘123123.jpeg‘] # string_input_producer会产生一个文件名队列 filename_queue = tf.train.string_input_producer(filename, shuffle=False, num_epochs=5) # reader从文件名队列中读数据。对应的方法是reader.read reader = tf.WholeFileReader() #<---------注意读取器不一样了 key, value = reader.read(filename_queue) # tf.train.string_input_producer定义了一个epoch变量,要对它进行初始化 tf.local_variables_initializer().run() # 使用start_queue_runners之后,才会开始填充队列 threads = tf.train.start_queue_runners(sess=sess) i = 0 while True: i += 1 # 获取图片数据并保存 image_data = sess.run(value) with open(‘test_%d.jpg‘ % i, ‘wb‘) as f: f.write(image_data)
-
书上的队列文件使用样例:
文件名队列创建->读取解析文件->打包解析好的文件->多线程启动图训练(多线程指被使用的部分其实还是文件读取)
import tensorflow as tf ‘‘‘创建文件列表‘‘‘ files = tf.train.match_filenames_once("Records/output.tfrecords") filename_queue = tf.train.string_input_producer(files, shuffle=False) ‘‘‘解析TFRecord文件里的数据‘‘‘ # 读取文件。 reader = tf.TFRecordReader() _,serialized_example = reader.read(filename_queue) # 解析读取的样例。 features = tf.parse_single_example( serialized_example, features={ ‘image_raw‘:tf.FixedLenFeature([],tf.string), ‘pixels‘:tf.FixedLenFeature([],tf.int64), ‘label‘:tf.FixedLenFeature([],tf.int64) }) decoded_images = tf.decode_raw(features[‘image_raw‘],tf.uint8) retyped_images = tf.cast(decoded_images, tf.float32) labels = tf.cast(features[‘label‘],tf.int32) #pixels = tf.cast(features[‘pixels‘],tf.int32) images = tf.reshape(retyped_images, [784]) ‘‘‘将文件以100个为一组打包‘‘‘ min_after_dequeue = 10000 batch_size = 100 capacity = min_after_dequeue + 3 * batch_size image_batch, label_batch = tf.train.shuffle_batch([images, labels], batch_size=batch_size, capacity=capacity, min_after_dequeue=min_after_dequeue) ‘‘‘训练模型‘‘‘ def inference(input_tensor, weights1, biases1, weights2, biases2): layer1 = tf.nn.relu(tf.matmul(input_tensor, weights1) + biases1) return tf.matmul(layer1, weights2) + biases2 # 模型相关的参数 INPUT_NODE = 784 OUTPUT_NODE = 10 LAYER1_NODE = 500 REGULARAZTION_RATE = 0.0001 TRAINING_STEPS = 5000 weights1 = tf.Variable(tf.truncated_normal([INPUT_NODE, LAYER1_NODE], stddev=0.1)) biases1 = tf.Variable(tf.constant(0.1, shape=[LAYER1_NODE])) weights2 = tf.Variable(tf.truncated_normal([LAYER1_NODE, OUTPUT_NODE], stddev=0.1)) biases2 = tf.Variable(tf.constant(0.1, shape=[OUTPUT_NODE])) y = inference(image_batch, weights1, biases1, weights2, biases2) # 计算交叉熵及其平均值 cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=label_batch) cross_entropy_mean = tf.reduce_mean(cross_entropy) # 损失函数的计算 regularizer = tf.contrib.layers.l2_regularizer(REGULARAZTION_RATE) regularaztion = regularizer(weights1) + regularizer(weights2) loss = cross_entropy_mean + regularaztion # 优化损失函数 train_step = tf.train.GradientDescentOptimizer(0.01).minimize(loss) # 初始化回话并开始训练过程。 with tf.Session() as sess: tf.global_variables_initializer().run() coord = tf.train.Coordinator() threads = tf.train.start_queue_runners(sess=sess, coord=coord) # 循环的训练神经网络。 for i in range(TRAINING_STEPS): if i % 1000 == 0: print("After %d training step(s), loss is %g " % (i, sess.run(loss))) sess.run(train_step) coord.request_stop() coord.join(threads)
时间: 2024-10-01 05:56:10