tensorflow入门:tfrecord 和tf.data.TFRecordDataset的使用

时间:2022-04-21 03:33:24

1.创建tfrecord

tfrecord支持写入三种格式的数据:string,int64,float32,以列表的形式分别通过tf.train.BytesList、tf.train.Int64List、tf.train.FloatList写入tf.train.Feature,如下所示:

?
1
2
3
tf.train.Feature(bytes_list=tf.train.BytesList(value=[feature.tostring()])) #feature一般是多维数组,要先转为list
tf.train.Feature(int64_list=tf.train.Int64List(value=list(feature.shape))) #tostring函数后feature的形状信息会丢失,把shape也写入
tf.train.Feature(float_list=tf.train.FloatList(value=[label]))

通过上述操作,以dict的形式把要写入的数据汇总,并构建tf.train.Features,然后构建tf.train.Example,如下:

?
1
2
3
4
5
6
7
def get_tfrecords_example(feature, label):
 tfrecords_features = {}
 feat_shape = feature.shape
 tfrecords_features['feature'] = tf.train.Feature(bytes_list=tf.train.BytesList(value=[feature.tostring()]))
 tfrecords_features['shape'] = tf.train.Feature(int64_list=tf.train.Int64List(value=list(feat_shape)))
 tfrecords_features['label'] = tf.train.Feature(float_list=tf.train.FloatList(value=label))
 return tf.train.Example(features=tf.train.Features(feature=tfrecords_features))

把创建的tf.train.Example序列化下,便可通过tf.python_io.TFRecordWriter写入tfrecord文件,如下:

?
1
2
3
4
5
tfrecord_wrt = tf.python_io.TFRecordWriter('xxx.tfrecord') #创建tfrecord的writer,文件名为xxx
exmp = get_tfrecords_example(feats[inx], labels[inx]) #把数据写入Example
exmp_serial = exmp.SerializeToString()  #Example序列化
tfrecord_wrt.write(exmp_serial)  #写入tfrecord文件
tfrecord_wrt.close()  #写完后关闭tfrecord的writer

代码汇总:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
import tensorflow as tf
from tensorflow.contrib.learn.python.learn.datasets.mnist import read_data_sets
 
mnist = read_data_sets("MNIST_data/", one_hot=True)
#把数据写入Example
def get_tfrecords_example(feature, label):
 tfrecords_features = {}
 feat_shape = feature.shape
 tfrecords_features['feature'] = tf.train.Feature(bytes_list=tf.train.BytesList(value=[feature.tostring()]))
 tfrecords_features['shape'] = tf.train.Feature(int64_list=tf.train.Int64List(value=list(feat_shape)))
 tfrecords_features['label'] = tf.train.Feature(float_list=tf.train.FloatList(value=label))
 return tf.train.Example(features=tf.train.Features(feature=tfrecords_features))
#把所有数据写入tfrecord文件
def make_tfrecord(data, outf_nm='mnist-train'):
 feats, labels = data
 outf_nm += '.tfrecord'
 tfrecord_wrt = tf.python_io.TFRecordWriter(outf_nm)
 ndatas = len(labels)
 for inx in range(ndatas):
 exmp = get_tfrecords_example(feats[inx], labels[inx])
 exmp_serial = exmp.SerializeToString()
 tfrecord_wrt.write(exmp_serial)
 tfrecord_wrt.close()
 
import random
nDatas = len(mnist.train.labels)
inx_lst = range(nDatas)
random.shuffle(inx_lst)
random.shuffle(inx_lst)
ntrains = int(0.85*nDatas)
 
# make training set
data = ([mnist.train.images[i] for i in inx_lst[:ntrains]], \
 [mnist.train.labels[i] for i in inx_lst[:ntrains]])
make_tfrecord(data, outf_nm='mnist-train')
 
# make validation set
data = ([mnist.train.images[i] for i in inx_lst[ntrains:]], \
 [mnist.train.labels[i] for i in inx_lst[ntrains:]])
make_tfrecord(data, outf_nm='mnist-val')
 
# make test set
data = (mnist.test.images, mnist.test.labels)
make_tfrecord(data, outf_nm='mnist-test')

2.tfrecord文件的使用:tf.data.TFRecordDataset

从tfrecord文件创建TFRecordDataset:

?
1
dataset = tf.data.TFRecordDataset('xxx.tfrecord')

解析tfrecord文件的每条记录,即序列化后的tf.train.Example;使用tf.parse_single_example来解析:

?
1
feats = tf.parse_single_example(serial_exmp, features=data_dict)

其中,data_dict是一个dict,包含的key是写入tfrecord文件时用的key,相应的value则是tf.FixedLenFeature([], tf.string)、tf.FixedLenFeature([], tf.int64)、tf.FixedLenFeature([], tf.float32),分别对应不同的数据类型,汇总即有:

?
1
2
3
4
5
6
7
def parse_exmp(serial_exmp):  #label中[10]是因为一个label是一个有10个元素的列表,shape中的[x]为shape的长度
feats = tf.parse_single_example(serial_exmp, features={'feature':tf.FixedLenFeature([], tf.string),\
 'label':tf.FixedLenFeature([10],tf.float32), 'shape':tf.FixedLenFeature([x], tf.int64)})
image = tf.decode_raw(feats['feature'], tf.float32)
label = feats['label']
shape = tf.cast(feats['shape'], tf.int32)
return image, label, shape

解析tfrecord文件中的所有记录,使用dataset的map方法,如下:

?
1
dataset = dataset.map(parse_exmp)

map方法可以接受任意函数以对dataset中的数据进行处理;另外,可使用repeat、shuffle、batch方法对dataset进行重复、混洗、分批;用repeat复制dataset以进行多个epoch;如下:

?
1
dataset = dataset.repeat(epochs).shuffle(buffer_size).batch(batch_size)

解析完数据后,便可以取出数据进行使用,通过创建iterator来进行,如下:

?
1
2
iterator = dataset.make_one_shot_iterator()
batch_image, batch_label, batch_shape = iterator.get_next()

要把不同dataset的数据feed进行模型,则需要先创建iterator handle,即iterator placeholder,如下:

?
1
2
3
4
handle = tf.placeholder(tf.string, shape=[])
iterator = tf.data.Iterator.from_string_handle(handle, \
 dataset_train.output_types, dataset_train.output_shapes)
image, label, shape = iterator.get_next()

然后为各个dataset创建handle,以feed_dict传入placeholder,如下:

?
1
2
3
4
with tf.Session() as sess:
 handle_train, handle_val, handle_test = sess.run(\
 [x.string_handle() for x in [iter_train, iter_val, iter_test]])
    sess.run([loss, train_op], feed_dict={handle: handle_train}

汇总:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
import tensorflow as tf
 
train_f, val_f, test_f = ['mnist-%s.tfrecord'%i for i in ['train', 'val', 'test']]
 
def parse_exmp(serial_exmp):
 feats = tf.parse_single_example(serial_exmp, features={'feature':tf.FixedLenFeature([], tf.string),\
 'label':tf.FixedLenFeature([10],tf.float32), 'shape':tf.FixedLenFeature([], tf.int64)})
 image = tf.decode_raw(feats['feature'], tf.float32)
 label = feats['label']
 shape = tf.cast(feats['shape'], tf.int32)
 return image, label, shape
 
 
def get_dataset(fname):
 dataset = tf.data.TFRecordDataset(fname)
 return dataset.map(parse_exmp) # use padded_batch method if padding needed
 
epochs = 16
batch_size = 50 # when batch_size can't be divided by nDatas, like 56,
 # there will be a batch data with nums less than batch_size
 
# training dataset
nDatasTrain = 46750
dataset_train = get_dataset(train_f)
dataset_train = dataset_train.repeat(epochs).shuffle(1000).batch(batch_size) # make sure repeat is ahead batch
  # this is different from dataset.shuffle(1000).batch(batch_size).repeat(epochs)
  # the latter means that there will be a batch data with nums less than batch_size for each epoch
  # if when batch_size can't be divided by nDatas.
nBatchs = nDatasTrain*epochs//batch_size
 
# evalation dataset
nDatasVal = 8250
dataset_val = get_dataset(val_f)
dataset_val = dataset_val.batch(nDatasVal).repeat(nBatchs//100*2)
 
# test dataset
nDatasTest = 10000
dataset_test = get_dataset(test_f)
dataset_test = dataset_test.batch(nDatasTest)
 
# make dataset iterator
iter_train = dataset_train.make_one_shot_iterator()
iter_val  = dataset_val.make_one_shot_iterator()
iter_test  = dataset_test.make_one_shot_iterator()
 
# make feedable iterator
handle = tf.placeholder(tf.string, shape=[])
iterator = tf.data.Iterator.from_string_handle(handle, \
 dataset_train.output_types, dataset_train.output_shapes)
x, y_, _ = iterator.get_next()
train_op, loss, eval_op = model(x, y_)
init = tf.initialize_all_variables()
 
# summary
logdir = './logs/m4d2a'
def summary_op(datapart='train'):
 tf.summary.scalar(datapart + '-loss', loss)
 tf.summary.scalar(datapart + '-eval', eval_op)
 return tf.summary.merge_all()
summary_op_train = summary_op()
summary_op_test = summary_op('val')
 
with tf.Session() as sess:
 sess.run(init)
 handle_train, handle_val, handle_test = sess.run(\
 [x.string_handle() for x in [iter_train, iter_val, iter_test]])
    _, cur_loss, cur_train_eval, summary = sess.run([train_op, loss, eval_op, summary_op_train], \
  feed_dict={handle: handle_train, keep_prob: 0.5} )
    cur_val_loss, cur_val_eval, summary = sess.run([loss, eval_op, summary_op_test], \
  feed_dict={handle: handle_val, keep_prob: 1.0})

3.mnist实验

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
import tensorflow as tf
 
train_f, val_f, test_f = ['mnist-%s.tfrecord'%i for i in ['train', 'val', 'test']]
 
def parse_exmp(serial_exmp):
 feats = tf.parse_single_example(serial_exmp, features={'feature':tf.FixedLenFeature([], tf.string),\
 'label':tf.FixedLenFeature([10],tf.float32), 'shape':tf.FixedLenFeature([], tf.int64)})
 image = tf.decode_raw(feats['feature'], tf.float32)
 label = feats['label']
 shape = tf.cast(feats['shape'], tf.int32)
 return image, label, shape
 
 
def get_dataset(fname):
 dataset = tf.data.TFRecordDataset(fname)
 return dataset.map(parse_exmp) # use padded_batch method if padding needed
 
epochs = 16
batch_size = 50 # when batch_size can't be divided by nDatas, like 56,
 # there will be a batch data with nums less than batch_size
 
# training dataset
nDatasTrain = 46750
dataset_train = get_dataset(train_f)
dataset_train = dataset_train.repeat(epochs).shuffle(1000).batch(batch_size) # make sure repeat is ahead batch
  # this is different from dataset.shuffle(1000).batch(batch_size).repeat(epochs)
  # the latter means that there will be a batch data with nums less than batch_size for each epoch
  # if when batch_size can't be divided by nDatas.
nBatchs = nDatasTrain*epochs//batch_size
 
# evalation dataset
nDatasVal = 8250
dataset_val = get_dataset(val_f)
dataset_val = dataset_val.batch(nDatasVal).repeat(nBatchs//100*2)
 
# test dataset
nDatasTest = 10000
dataset_test = get_dataset(test_f)
dataset_test = dataset_test.batch(nDatasTest)
 
# make dataset iterator
iter_train = dataset_train.make_one_shot_iterator()
iter_val  = dataset_val.make_one_shot_iterator()
iter_test  = dataset_test.make_one_shot_iterator()
 
# make feedable iterator, i.e. iterator placeholder
handle = tf.placeholder(tf.string, shape=[])
iterator = tf.data.Iterator.from_string_handle(handle, \
 dataset_train.output_types, dataset_train.output_shapes)
x, y_, _ = iterator.get_next()
 
# cnn
x_image = tf.reshape(x, [-1,28,28,1])
w_init = tf.truncated_normal_initializer(stddev=0.1, seed=9)
b_init = tf.constant_initializer(0.1)
cnn1 = tf.layers.conv2d(x_image, 32, (5,5), padding='same', activation=tf.nn.relu, \
 kernel_initializer=w_init, bias_initializer=b_init)
mxpl1 = tf.layers.max_pooling2d(cnn1, 2, strides=2, padding='same')
cnn2 = tf.layers.conv2d(mxpl1, 64, (5,5), padding='same', activation=tf.nn.relu, \
 kernel_initializer=w_init, bias_initializer=b_init)
mxpl2 = tf.layers.max_pooling2d(cnn2, 2, strides=2, padding='same')
mxpl2_flat = tf.reshape(mxpl2, [-1,7*7*64])
fc1 = tf.layers.dense(mxpl2_flat, 1024, activation=tf.nn.relu, \
 kernel_initializer=w_init, bias_initializer=b_init)
keep_prob = tf.placeholder('float')
fc1_drop = tf.nn.dropout(fc1, keep_prob)
logits = tf.layers.dense(fc1_drop, 10, kernel_initializer=w_init, bias_initializer=b_init)
 
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y_))
optmz = tf.train.AdamOptimizer(1e-4)
train_op = optmz.minimize(loss)
 
def get_eval_op(logits, labels):
 corr_prd = tf.equal(tf.argmax(logits,1), tf.argmax(labels,1))
 return tf.reduce_mean(tf.cast(corr_prd, 'float'))
eval_op = get_eval_op(logits, y_)
 
init = tf.initialize_all_variables()
 
# summary
logdir = './logs/m4d2a'
def summary_op(datapart='train'):
 tf.summary.scalar(datapart + '-loss', loss)
 tf.summary.scalar(datapart + '-eval', eval_op)
 return tf.summary.merge_all()
summary_op_train = summary_op()
summary_op_val = summary_op('val')
 
# whether to restore or not
ckpts_dir = 'ckpts/'
ckpt_nm = 'cnn-ckpt'
saver = tf.train.Saver(max_to_keep=50) # defaults to save all variables, using dict {'x':x,...} to save specified ones.
restore_step = ''
start_step = 0
train_steps = nBatchs
best_loss = 1e6
best_step = 0
 
# import os
# os.environ["CUDA_VISIBLE_DEVICES"] = "0"
# config = tf.ConfigProto()
# config.gpu_options.per_process_gpu_memory_fraction = 0.9
# config.gpu_options.allow_growth=True # allocate when needed
# with tf.Session(config=config) as sess:
with tf.Session() as sess:
 sess.run(init)
 handle_train, handle_val, handle_test = sess.run(\
 [x.string_handle() for x in [iter_train, iter_val, iter_test]])
 if restore_step:
 ckpt = tf.train.get_checkpoint_state(ckpts_dir)
 if ckpt and ckpt.model_checkpoint_path: # ckpt.model_checkpoint_path means the latest ckpt
  if restore_step == 'latest':
  ckpt_f = tf.train.latest_checkpoint(ckpts_dir)
  start_step = int(ckpt_f.split('-')[-1]) + 1
  else:
  ckpt_f = ckpts_dir+ckpt_nm+'-'+restore_step
  print('loading wgt file: '+ ckpt_f)
  saver.restore(sess, ckpt_f)
 summary_wrt = tf.summary.FileWriter(logdir,sess.graph)
 if restore_step in ['', 'latest']:
 for i in range(start_step, train_steps):
  _, cur_loss, cur_train_eval, summary = sess.run([train_op, loss, eval_op, summary_op_train], \
   feed_dict={handle: handle_train, keep_prob: 0.5} )
  # log to stdout and eval validation set
  if i % 100 == 0 or i == train_steps-1:
  saver.save(sess, ckpts_dir+ckpt_nm, global_step=i) # save variables
  summary_wrt.add_summary(summary, global_step=i)
  cur_val_loss, cur_val_eval, summary = sess.run([loss, eval_op, summary_op_val], \
   feed_dict={handle: handle_val, keep_prob: 1.0})
  if cur_val_loss < best_loss:
   best_loss = cur_val_loss
   best_step = i
  summary_wrt.add_summary(summary, global_step=i)
  print 'step %5d: loss %.5f, acc %.5f --- loss val %0.5f, acc val %.5f'%(i, \
   cur_loss, cur_train_eval, cur_val_loss, cur_val_eval)
  # sess.run(init_train)
 with open(ckpts_dir+'best.step','w') as f:
  f.write('best step is %d\n'%best_step)
 print 'best step is %d'%best_step
 # eval test set
 test_loss, test_eval = sess.run([loss, eval_op], feed_dict={handle: handle_test, keep_prob: 1.0})
 print 'eval test: loss %.5f, acc %.5f'%(test_loss, test_eval)

实验结果:

tensorflow入门:tfrecord 和tf.data.TFRecordDataset的使用

以上这篇tensorflow入门:tfrecord 和tf.data.TFRecordDataset的使用就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持服务器之家。

原文链接:https://blog.csdn.net/yeqiustu/article/details/79793454