一. from_tensor_slices
import tensorflow as tf
if __name__ == '__main__':
#把给定的元组、列表和张量等数据进行特征切片,张量
dataset = tf.data.Dataset.from_tensor_slices([1,2,3,4,5])
for ele in dataset:
print(ele)
'''
tf.Tensor(1, shape=(), dtype=int32)
tf.Tensor(2, shape=(), dtype=int32)
tf.Tensor(3, shape=(), dtype=int32)
tf.Tensor(4, shape=(), dtype=int32)
tf.Tensor(5, shape=(), dtype=int32)
'''
#第一维度进行切分
dataset = tf.data.Dataset.from_tensor_slices([[1,2], [3,4], [5,6], [7,8], [9,10]])
for ele in dataset:
print(ele)
'''
tf.Tensor([1 2], shape=(2,), dtype=int32)
tf.Tensor([3 4], shape=(2,), dtype=int32)
tf.Tensor([5 6], shape=(2,), dtype=int32)
tf.Tensor([7 8], shape=(2,), dtype=int32)
tf.Tensor([ 9 10], shape=(2,), dtype=int32)
'''
for ele in dataset:
print(ele.numpy())
'''
[1 2]
[3 4]
[5 6]
[7 8]
[ 9 10]
'''
#字典
dataset_dic = tf.data.Dataset.from_tensor_slices(
{
'a': [1, 2, 3, 4],
'b': [5, 6, 7, 8],
'c': [9,10,11,12]
}
)
print(dataset_dic)
#<TensorSliceDataset shapes: {a: (), b: (), c: ()}, types: {a: tf.int32, b: tf.int32, c: tf.int32}>
for ele in dataset_dic:
print(ele)
'''
'a': <tf.Tensor: shape=(), dtype=int32, numpy=1>, 'b': <tf.Tensor: shape=(), dtype=int32, numpy=5>, 'c': <tf.Tensor: shape=(), dtype=int32, numpy=9>}
{'a': <tf.Tensor: shape=(), dtype=int32, numpy=2>, 'b': <tf.Tensor: shape=(), dtype=int32, numpy=6>, 'c': <tf.Tensor: shape=(), dtype=int32, numpy=10>}
{'a': <tf.Tensor: shape=(), dtype=int32, numpy=3>, 'b': <tf.Tensor: shape=(), dtype=int32, numpy=7>, 'c': <tf.Tensor: shape=(), dtype=int32, numpy=11>}
{'a': <tf.Tensor: shape=(), dtype=int32, numpy=4>, 'b': <tf.Tensor: shape=(), dtype=int32, numpy=8>, 'c': <tf.Tensor: shape=(), dtype=int32, numpy=12>}
'''
二. shuffle，repeat，square
dataset = tf.data.Dataset.from_tensor_slices([1,2,3,4,5,6,7])
dataset = dataset.shuffle(7)#乱序,7个元素乱序一次
for ele in dataset:
print(ele.numpy())
#5 3 7 2 4 1 6
#dataset = dataset.shuffle(7).repeat(3)
dataset =dataset.repeat(3)#乱序三次
'''
2 4 6 7 5 1 3
2 4 3 7 1 5 6
3 5 4 2 6 7 1
'''
#dataset = dataset.batch(3)#三个一组
'''
[1 5 2]
[6 7 4]
[3 3 7]
[6 4 5]
[2 1 6]
[7 4 2]
[3 1 5]
'''
dataset = tf.data.Dataset.from_tensor_slices([1, 2, 3, 4, 5, 6, 7])
dataset = dataset.map(tf.square)#平放
'''
1
4
9
16
25
36
49
'''
三. 实例
import tensorflow as tf
if __name__ == '__main__':
#mnist数据集
(train_images,train_labels),(test_images,test_labels) = tf.keras.datasets.mnist.load_data
train_images = train_images/255#(60000, 28, 28)
test_images = test_images/255#(10000, 28, 28)
ds_train_img = tf.data.Dataset.from_tensor_slices(train_images)
ds_train_labels = tf.data.Dataset.from_tensor_slices(train_labels)
ds_test_img = tf.data.Dataset.from_tensor_slices(test_images)
ds_test_labels = tf.data.Dataset.from_tensor_slices(test_labels)
ds_train = tf.data.Dataset.zip((ds_train_img,ds_train_labels))#()元组形式合并
#<ZipDataset shapes: ((28, 28), ()), types: (tf.float64, tf.uint8)> ()代表一个数字
ds_test = tf.data.Dataset.zip((ds_test_img,ds_test_labels))
ds_train = ds_train.shuffle(60000).repeat().batch(64)#10000张图片，64张图片一组，无限乱序
ds_test = ds_test.batch(64)
model = tf.keras.Sequential([
tf.keras.layers.Flatten(input_shape=(28,28)),
tf.keras.layers.Dense(128,activation='relu'),
tf.keras.layers.Dense(10,activation='softmax')
])
model.compile(
optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy']
)
steps_per_epochs = train_images.shape[0]//64#数据集总数/batch_size等于有多少个batch.//整除
history = model.fit(
ds_train,
epochs=5,
steps_per_epoch = steps_per_epochs,#循环多少个batch算一个epoch
validation_data=ds_test,#打印测试集
validation_steps=10000//64
)
print(history)
Epoch 1/5
937/937 [==============================] - 1s 1ms/step - loss: 0.2947 - accuracy: 0.9177 - val_loss: 0.1588 - val_accuracy: 0.9527
Epoch 2/5
937/937 [==============================] - 1s 1ms/step - loss: 0.1355 - accuracy: 0.9607 - val_loss: 0.1145 - val_accuracy: 0.9649
Epoch 3/5
937/937 [==============================] - 1s 1ms/step - loss: 0.0951 - accuracy: 0.9723 - val_loss: 0.0944 - val_accuracy: 0.9719
Epoch 4/5
937/937 [==============================] - 1s 1ms/step - loss: 0.0733 - accuracy: 0.9782 - val_loss: 0.0857 - val_accuracy: 0.9738
Epoch 5/5
937/937 [==============================] - 1s 1ms/step - loss: 0.0584 - accuracy: 0.9825 - val_loss: 0.0804 - val_accuracy: 0.9743
版权声明：本文为weixin_40519901原创文章，遵循CC 4.0 BY-SA版权协议，转载请附上原文出处链接和本声明。