Tensorflow（二） Tensor和Variable

Tensor和Variable都是TensorFlow框架中的基本数据单元，本文详细介绍两者的用法。
参考：
https://tensorflow.google.cn/guide/tensor
https://tensorflow.google.cn/guide/variable

Tensor

基础

下面的例子分别创建一个0维，1维，2维和3维Tensor：

rank_0_tensor = tf.constant(4)
print(rank_0_tensor)

rank_1_tensor = tf.constant([2.0, 3.0, 4.0])
print(rank_1_tensor)

rank_2_tensor = tf.constant([[1, 2],
                             [3, 4],
                             [5, 6]], dtype=tf.float16)
print(rank_2_tensor)

rank_3_tensor = tf.constant([
  [[0, 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]],])

print(rank_3_tensor)

tf.Tensor(4, shape=(), dtype=int32)
tf.Tensor([2. 3. 4.], shape=(3,), dtype=float32)
tf.Tensor(
[[1. 2.]
[3. 4.]
[5. 6.]], shape=(3, 2), dtype=float16)
tf.Tensor(
[[[ 0 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]]], shape=(3, 2, 5), dtype=int32)

到Numpy和Pandas类型的转换

TensorFlow支持转换tensor到numpy的array和pandas的DataFrame

import pandas as pd
import numpy as np

np.array(rank_2_tensor)
rank_2_tensor.numpy()

pd.DataFrame(rank_2_tensor)

注意panda只支持对3维以下的数据转换
用tf.convert_to_tensor,可以将array和DataFrame转换回Tensor

索引和切片

TensorFlow支持对Tensor和Variable的索引和切片，操作与Numpy几乎相同。

rank_1_tensor = tf.constant([0, 1, 1, 2, 3, 5, 8, 13, 21, 34])
print("First:", rank_1_tensor[0].numpy())
print("Second:", rank_1_tensor[1].numpy())
print("Last:", rank_1_tensor[-1].numpy())
print("Everything:", rank_1_tensor[:].numpy())
print("Before 4:", rank_1_tensor[:4].numpy())
print("From 4 to the end:", rank_1_tensor[4:].numpy())
print("From 2, before 7:", rank_1_tensor[2:7].numpy())
print("Every other item:", rank_1_tensor[::2].numpy())
print("Reversed:", rank_1_tensor[::-1].numpy())

First: 0
Second: 1
Last: 34
Everything: [ 0 1 1 2 3 5 8 13 21 34]
Before 4: [0 1 1 2]
From 4 to the end: [ 3 5 8 13 21 34]
From 2, before 7: [1 2 3 5 8]
Every other item: [ 0 1 3 8 21]
Reversed: [34 21 13 8 5 3 2 1 1 0]

print("Second row:", rank_2_tensor[1, :].numpy())
print("Second column:", rank_2_tensor[:, 1].numpy())
print("Last row:", rank_2_tensor[-1, :].numpy())
print("First item in last column:", rank_2_tensor[0, -1].numpy())
print("Skip the first row:")
print(rank_2_tensor[1:, :].numpy(), "\n")

Second row: [3. 4.]
Second column: [2. 4. 6.]
Last row: [5. 6.]
First item in last column: 2.0
Skip the first row:
[[3. 4.]
[5. 6.]]

reshape

和numpy一样，tensor支持reshape

print(tf.reshape(rank_3_tensor, [3*2, 5]), "\n")
print(tf.reshape(rank_3_tensor, [3, -1]))

tf.Tensor(
[[ 0 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]], shape=(6, 5), dtype=int32)
tf.Tensor(
[[ 0 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]], shape=(3, 10), dtype=int32)

数据类型

TensorFlow采用Tensor.dtype数据类型
参考：https://tensorflow.google.cn/api_docs/python/tf/Tensor#dtype
tf.cast可以用来转换数值类型

the_f64_tensor = tf.constant([2.2, 3.3, 4.4], dtype=tf.float64)
the_f16_tensor = tf.cast(the_f64_tensor, dtype=tf.float16)
# Now, cast to an uint8 and lose the decimal precision
the_u8_tensor = tf.cast(the_f16_tensor, dtype=tf.uint8)
print(the_u8_tensor)

tf.Tensor([2 3 4], shape=(3,), dtype=uint8)

TensorFlow也支持string，bool等其它类型

# If you have three string tensors of different lengths, this is OK.
tensor_of_strings = tf.constant(["Gray wolf",
                                 "Quick brown fox",
                                 "Lazy dog"])
# Note that the shape is (3,). The string length is not included.
print(tensor_of_strings)

byte_strings = tf.strings.bytes_split(tf.constant("Duck"))
byte_ints = tf.io.decode_raw(tf.constant("Duck"), tf.uint8)
print("Byte strings:", byte_strings)
print("Bytes:", byte_ints)

# Or split it up as unicode and then decode it
unicode_bytes = tf.constant("アヒル 🦆")
unicode_char_bytes = tf.strings.unicode_split(unicode_bytes, "UTF-8")
unicode_values = tf.strings.unicode_decode(unicode_bytes, "UTF-8")
print("\nUnicode bytes:", unicode_bytes)
print("\nUnicode chars:", unicode_char_bytes)
print("\nUnicode values:", unicode_values)

print(tf.constant([False, True]))

tf.Tensor([b'Gray wolf' b'Quick brown fox' b'Lazy dog'], shape=(3,), dtype=string)
Byte strings: tf.Tensor([b'D' b'u' b'c' b'k'], shape=(4,), dtype=string)
Bytes: tf.Tensor([ 68 117 99 107], shape=(4,), dtype=uint8)
Unicode bytes: tf.Tensor(b'\xe3\x82\xa2\xe3\x83\x92\xe3\x83\xab \xf0\x9f\xa6\x86', shape=(), dtype=string)
Unicode chars: tf.Tensor([b'\xe3\x82\xa2' b'\xe3\x83\x92' b'\xe3\x83\xab' b' ' b'\xf0\x9f\xa6\x86'], shape=(5,), dtype=string)
Unicode values: tf.Tensor([ 12450 12498 12523 32 129414], shape=(5,), dtype=int32)
<tf.Tensor: shape=(2,), dtype=bool, numpy=array([False, True])>

Variable

上一篇介绍过Variable是可以赋值变化的数据单元。

和Tensor的转换

my_tensor = tf.constant([[1.0, 2.0], [3.0, 4.0]])
my_variable = tf.Variable(my_tensor)
tf.convert_to_tensor(my_variable)

赋值

Variable不能用=直接赋值，赋值用TensorFlow里的assign函数，要注意维数和数据长度要相等

a = tf.Variable([2.0, 3.0])
# This will keep the same dtype, float32
a.assign([1, 2]) 
# Not allowed as it resizes the variable: 
try:
  a.assign([1.0, 2.0, 3.0])
except Exception as e:
  print(f"{type(e).__name__}: {e}")

ValueError: Cannot assign value to variable ' Variable:0': Shape mismatch.The variable shape (2,), and the assigned value shape (3,) are incompatible.

另外有assign_add和assign_sub方法进行加减操作

a = tf.Variable([2.0, 3.0])
# Create b based on the value of a
b = tf.Variable(a)
a.assign([5, 6])

# a and b are different
print(a.numpy())
print(b.numpy())

# There are other versions of assign
print(a.assign_add([2,3]).numpy())  # [7. 9.]
print(a.assign_sub([7,9]).numpy())  # [0. 0.]

[5. 6.]
[2. 3.]
[7. 9.]
[0. 0.]

命名

tf.Variable拥有和Python对象一样的生命周期，可以为之命名来进行Debug和Trace：

# Create a and b; they will have the same name but will be backed by
# different tensors.
a = tf.Variable(my_tensor, name="Mark")
# A new variable with the same name, but different value
# Note that the scalar add is broadcast
b = tf.Variable(my_tensor + 1, name="Mark")

# These are elementwise-unequal, despite having the same name
print(a == b)

tf.Tensor(
[[False False]
[False False]], shape=(2, 2), dtype=bool)

变量名在保存和加载模型时被保留。默认情况下，模型中的变量会自动获得唯一的变量名，所以你不需要自己给它们赋值。
通过在创建时将trainable设置为false，可以关闭变量的梯度。用来过滤数据中不需要进行微分运算的变量。不需要渐变的变量的一个例子是训练步长计数器。

step_counter = tf.Variable(1, trainable=False)

性能

TensorFlow会优先使用GPU来进行数据存储运算，用户也可以通过编码，手动分配运用CPU还是GPU。

with tf.device('CPU:0'):

  # Create some tensors
  a = tf.Variable([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])
  b = tf.constant([[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]])
  c = tf.matmul(a, b)

print(c)

tf.Tensor(
[[22. 28.]
[49. 64.]], shape=(2, 2), dtype=float32)

下面的例子用CPU存储数据，用GPU进行计算：

with tf.device('CPU:0'):
  a = tf.Variable([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]])
  b = tf.Variable([[1.0, 2.0, 3.0]])

with tf.device('GPU:0'):
  # Element-wise multiply
  k = a * b

print(k)

tf.Tensor(
[[ 1. 4. 9.]
[ 4. 10. 18.]], shape=(2, 3), dtype=float32)