服务器之家

服务器之家 > 正文

浅谈tensorflow中dataset.shuffle和dataset.batch dataset.repeat注意点

时间:2020-06-09 10:19     来源/作者:青盏

batch很好理解,就是batch size。注意在一个epoch中最后一个batch大小可能小于等于batch size

dataset.repeat就是俗称epoch,但在tf中与dataset.shuffle的使用顺序可能会导致个epoch的混合

dataset.shuffle就是说维持一个buffer size 大小的 shuffle buffer,图中所需的每个样本从shuffle buffer中获取,取得一个样本后,就从源数据集中加入一个样本到shuffle buffer中。

?
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
import os
os.environ['CUDA_VISIBLE_DEVICES'] = ""
import numpy as np
import tensorflow as tf
np.random.seed(0)
x = np.random.sample((11,2))
# make a dataset from a numpy array
print(x)
print()
dataset = tf.data.Dataset.from_tensor_slices(x)
dataset = dataset.shuffle(3)
dataset = dataset.batch(4)
dataset = dataset.repeat(2)
 
# create the iterator
iter = dataset.make_one_shot_iterator()
el = iter.get_next()
 
with tf.Session() as sess:
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
?
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
#源数据集
[[ 0.5488135  0.71518937]
 [ 0.60276338 0.54488318]
 [ 0.4236548  0.64589411]
 [ 0.43758721 0.891773 ]
 [ 0.96366276 0.38344152]
 [ 0.79172504 0.52889492]
 [ 0.56804456 0.92559664]
 [ 0.07103606 0.0871293 ]
 [ 0.0202184  0.83261985]
 [ 0.77815675 0.87001215]
 [ 0.97861834 0.79915856]]
 
# 通过shuffle batch后取得的样本
[[ 0.4236548  0.64589411]
 [ 0.60276338 0.54488318]
 [ 0.43758721 0.891773 ]
 [ 0.5488135  0.71518937]]
[[ 0.96366276 0.38344152]
 [ 0.56804456 0.92559664]
 [ 0.0202184  0.83261985]
 [ 0.79172504 0.52889492]]
[[ 0.07103606 0.0871293 ]
 [ 0.97861834 0.79915856]
 [ 0.77815675 0.87001215]] #最后一个batch样本个数为3
[[ 0.60276338 0.54488318]
 [ 0.5488135  0.71518937]
 [ 0.43758721 0.891773 ]
 [ 0.79172504 0.52889492]]
[[ 0.4236548  0.64589411]
 [ 0.56804456 0.92559664]
 [ 0.0202184  0.83261985]
 [ 0.07103606 0.0871293 ]]
[[ 0.77815675 0.87001215]
 [ 0.96366276 0.38344152]
 [ 0.97861834 0.79915856]] #最后一个batch样本个数为3

1、按照shuffle中设置的buffer size,首先从源数据集取得三个样本:
shuffle buffer:
[ 0.5488135 0.71518937]
[ 0.60276338 0.54488318]
[ 0.4236548 0.64589411]
2、从buffer中取一个样本到batch中得:
shuffle buffer:
[ 0.5488135 0.71518937]
[ 0.60276338 0.54488318]
batch:
[ 0.4236548 0.64589411]
3、shuffle buffer不足三个样本,从源数据集提取一个样本:
shuffle buffer:
[ 0.5488135 0.71518937]
[ 0.60276338 0.54488318]
[ 0.43758721 0.891773 ]
4、从buffer中取一个样本到batch中得:
shuffle buffer:
[ 0.5488135 0.71518937]
[ 0.43758721 0.891773 ]
batch:
[ 0.4236548 0.64589411]
[ 0.60276338 0.54488318]
5、如此反复。这就意味中如果shuffle 的buffer size=1,数据集不打乱。如果shuffle 的buffer size=数据集样本数量,随机打乱整个数据集

?
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
import os
os.environ['CUDA_VISIBLE_DEVICES'] = ""
import numpy as np
import tensorflow as tf
np.random.seed(0)
x = np.random.sample((11,2))
# make a dataset from a numpy array
print(x)
print()
dataset = tf.data.Dataset.from_tensor_slices(x)
dataset = dataset.shuffle(1)
dataset = dataset.batch(4)
dataset = dataset.repeat(2)
 
# create the iterator
iter = dataset.make_one_shot_iterator()
el = iter.get_next()
 
with tf.Session() as sess:
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
 
[[ 0.5488135  0.71518937]
 [ 0.60276338 0.54488318]
 [ 0.4236548  0.64589411]
 [ 0.43758721 0.891773 ]
 [ 0.96366276 0.38344152]
 [ 0.79172504 0.52889492]
 [ 0.56804456 0.92559664]
 [ 0.07103606 0.0871293 ]
 [ 0.0202184  0.83261985]
 [ 0.77815675 0.87001215]
 [ 0.97861834 0.79915856]]
 
[[ 0.5488135  0.71518937]
 [ 0.60276338 0.54488318]
 [ 0.4236548  0.64589411]
 [ 0.43758721 0.891773 ]]
[[ 0.96366276 0.38344152]
 [ 0.79172504 0.52889492]
 [ 0.56804456 0.92559664]
 [ 0.07103606 0.0871293 ]]
[[ 0.0202184  0.83261985]
 [ 0.77815675 0.87001215]
 [ 0.97861834 0.79915856]]
[[ 0.5488135  0.71518937]
 [ 0.60276338 0.54488318]
 [ 0.4236548  0.64589411]
 [ 0.43758721 0.891773 ]]
[[ 0.96366276 0.38344152]
 [ 0.79172504 0.52889492]
 [ 0.56804456 0.92559664]
 [ 0.07103606 0.0871293 ]]
[[ 0.0202184  0.83261985]
 [ 0.77815675 0.87001215]
 [ 0.97861834 0.79915856]]

注意如果repeat在shuffle之前使用:

官方说repeat在shuffle之前使用能提高性能,但模糊了数据样本的epoch关系

?
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
import os
os.environ['CUDA_VISIBLE_DEVICES'] = ""
import numpy as np
import tensorflow as tf
np.random.seed(0)
x = np.random.sample((11,2))
# make a dataset from a numpy array
print(x)
print()
dataset = tf.data.Dataset.from_tensor_slices(x)
dataset = dataset.repeat(2)
dataset = dataset.shuffle(11)
dataset = dataset.batch(4)
 
# create the iterator
iter = dataset.make_one_shot_iterator()
el = iter.get_next()
 
with tf.Session() as sess:
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
  print(sess.run(el))
 
[[ 0.5488135  0.71518937]
 [ 0.60276338 0.54488318]
 [ 0.4236548  0.64589411]
 [ 0.43758721 0.891773 ]
 [ 0.96366276 0.38344152]
 [ 0.79172504 0.52889492]
 [ 0.56804456 0.92559664]
 [ 0.07103606 0.0871293 ]
 [ 0.0202184  0.83261985]
 [ 0.77815675 0.87001215]
 [ 0.97861834 0.79915856]]
 
[[ 0.56804456 0.92559664]
 [ 0.5488135  0.71518937]
 [ 0.60276338 0.54488318]
 [ 0.07103606 0.0871293 ]]
[[ 0.96366276 0.38344152]
 [ 0.43758721 0.891773 ]
 [ 0.43758721 0.891773 ]
 [ 0.77815675 0.87001215]]
[[ 0.79172504 0.52889492#出现相同样本出现在同一个batch中
 [ 0.79172504 0.52889492]
 [ 0.60276338 0.54488318]
 [ 0.4236548  0.64589411]]
[[ 0.07103606 0.0871293 ]
 [ 0.4236548  0.64589411]
 [ 0.96366276 0.38344152]
 [ 0.5488135  0.71518937]]
[[ 0.97861834 0.79915856]
 [ 0.0202184  0.83261985]
 [ 0.77815675 0.87001215]
 [ 0.56804456 0.92559664]]
[[ 0.0202184  0.83261985]
 [ 0.97861834 0.79915856]]     #可以看到最后个batch为2,而前面都是4

使用案例:

?
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
def input_fn(filenames, batch_size=32, num_epochs=1, perform_shuffle=False):
  print('Parsing', filenames)
  def decode_libsvm(line):
    #columns = tf.decode_csv(value, record_defaults=CSV_COLUMN_DEFAULTS)
    #features = dict(zip(CSV_COLUMNS, columns))
    #labels = features.pop(LABEL_COLUMN)
    columns = tf.string_split([line], ' ')
    labels = tf.string_to_number(columns.values[0], out_type=tf.float32)
    splits = tf.string_split(columns.values[1:], ':')
    id_vals = tf.reshape(splits.values,splits.dense_shape)
    feat_ids, feat_vals = tf.split(id_vals,num_or_size_splits=2,axis=1)
    feat_ids = tf.string_to_number(feat_ids, out_type=tf.int32)
    feat_vals = tf.string_to_number(feat_vals, out_type=tf.float32)
    #feat_ids = tf.reshape(feat_ids,shape=[-1,FLAGS.field_size])
    #for i in range(splits.dense_shape.eval()[0]):
    #  feat_ids.append(tf.string_to_number(splits.values[2*i], out_type=tf.int32))
    #  feat_vals.append(tf.string_to_number(splits.values[2*i+1]))
    #return tf.reshape(feat_ids,shape=[-1,field_size]), tf.reshape(feat_vals,shape=[-1,field_size]), labels
    return {"feat_ids": feat_ids, "feat_vals": feat_vals}, labels
 
  # Extract lines from input files using the Dataset API, can pass one filename or filename list
  dataset = tf.data.TextLineDataset(filenames).map(decode_libsvm, num_parallel_calls=10).prefetch(500000# multi-thread pre-process then prefetch
 
  # Randomizes input using a window of 256 elements (read into memory)
  if perform_shuffle:
    dataset = dataset.shuffle(buffer_size=256)
 
  # epochs from blending together.
  dataset = dataset.repeat(num_epochs)
  dataset = dataset.batch(batch_size) # Batch size to use
 
  #return dataset.make_one_shot_iterator()
  iterator = dataset.make_one_shot_iterator()
  batch_features, batch_labels = iterator.get_next()
  #return tf.reshape(batch_ids,shape=[-1,field_size]), tf.reshape(batch_vals,shape=[-1,field_size]), batch_labels
  return batch_features, batch_labels

到此这篇关于浅谈tensorflow中dataset.shuffle和dataset.batch dataset.repeat注意点的文章就介绍到这了,更多相关tensorflow中dataset.shuffle和dataset.batch dataset.repeat内容请搜索服务器之家以前的文章或继续浏览下面的相关文章希望大家以后多多支持服务器之家! 

原文链接:https://blog.csdn.net/qq_16234613/article/details/81703228

相关文章

热门资讯

2022年最旺的微信头像大全 微信头像2022年最新版图片
2022年最旺的微信头像大全 微信头像2022年最新版图片 2022-01-10
蜘蛛侠3英雄无归3正片免费播放 蜘蛛侠3在线观看免费高清完整
蜘蛛侠3英雄无归3正片免费播放 蜘蛛侠3在线观看免费高清完整 2021-08-24
背刺什么意思 网络词语背刺是什么梗
背刺什么意思 网络词语背刺是什么梗 2020-05-22
yue是什么意思 网络流行语yue了是什么梗
yue是什么意思 网络流行语yue了是什么梗 2020-10-11
暖暖日本高清免费中文 暖暖在线观看免费完整版韩国
暖暖日本高清免费中文 暖暖在线观看免费完整版韩国 2021-05-08
返回顶部