TensorFlow2-数据类型(4)

小灰灰 2022-11-29 03:04 144阅读 0赞

### 文章目录 ###

*  数据类型
 *   *  常见数据仓
     *  常见数据类型
     *  常见数据类型使用
     *  常见数据属性使用
     *  常见数据类型检测
     *  Convert 数据类型转换
     *  数据转换--Int与Bool型
     *  tf.Variable
     *  数据类型转换 TO numpy

# 数据类型 #

## 常见数据仓 ##

*  list
 *  np.array
 *  tf.Tensor
    
     *  scalar:标量、0维：1.1
     *  vector：\[1.1\], \[1.1,1.2,…\]
     *  matrix: \[\[1,2,3\],\[2,4,5\],\[23,45\]\]
     *  tensor: rank(维度) > 2

## 常见数据类型 ##

*  int,float,double
 *  bool
 *  string

## 常见数据类型使用 ##

import tensorflow as tf
    import numpy as np

# 老版本：常量,数值不能改变
    # 新版本是普通的变量
    # int32
    tf.constant(1)

<tf.Tensor: shape=(), dtype=int32, numpy=1>

# float32
    tf.constant(1.)

<tf.Tensor: shape=(), dtype=float32, numpy=1.0>

# 手动指定类型
    # double -> float64
    tf.constant(2.2, dtype=tf.double)

<tf.Tensor: shape=(), dtype=float64, numpy=2.2>

# bool
    tf.constant([True, False])

<tf.Tensor: shape=(2,), dtype=bool, numpy=array([ True, False])>

# strins
    tf.constant("hello,world.")

<tf.Tensor: shape=(), dtype=string, numpy=b'hello,word.'>

## 常见数据属性使用 ##

# cpu与gpu使用环境
    with tf.device("cpu"):
        a = tf.constant([1])
        
    with tf.device("gpu"):
        b = tf.range(4)

# 显示所在运行环境
    # 默认在gpu上
    a.device
    b.device

'/job:localhost/replica:0/task:0/device:CPU:0'

# cpu转gpu
    aa = a.gpu()
    aa.device

WARNING:tensorflow:From <ipython-input-15-7d7915ff2fc1>:1: _EagerTensorBase.gpu (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
    Instructions for updating:
    Use tf.identity instead.
    
    
    
    ---------------------------------------------------------------------------
    
    RuntimeError                              Traceback (most recent call last)
    
    <ipython-input-15-7d7915ff2fc1> in <module>
    ----> 1 aa = a.gpu()
          2 aa.device
    
    
    ~/.virtualenvs/py38/lib/python3.8/site-packages/tensorflow/python/util/deprecation.py in new_func(*args, **kwargs)
        322               'in a future version' if date is None else ('after %s' % date),
        323               instructions)
    --> 324       return func(*args, **kwargs)
        325     return tf_decorator.make_decorator(
        326         func, new_func, 'deprecated',
    
    
    ~/.virtualenvs/py38/lib/python3.8/site-packages/tensorflow/python/framework/ops.py in gpu(self, gpu_index)
       1099       as this Tensor.
       1100     """
    -> 1101     return self._copy(context.context(), "GPU:" + str(gpu_index))
       1102 
       1103   def set_shape(self, shape):
    
    
    ~/.virtualenvs/py38/lib/python3.8/site-packages/tensorflow/python/framework/ops.py in _copy(self, ctx, device_name)
       1041   def _copy(self, ctx=None, device_name=None):
       1042     """Copies tensor to dest device."""
    -> 1043     new_tensor = self._copy_nograd(ctx, device_name)
       1044     # Record the copy on tape and define backprop copy as well.
       1045     if context.executing_eagerly():
    
    
    ~/.virtualenvs/py38/lib/python3.8/site-packages/tensorflow/python/framework/ops.py in _copy_nograd(self, ctx, device_name)
       1034     try:
       1035       ctx.ensure_initialized()
    -> 1036       new_tensor = self._copy_to_device(device_name)
       1037     except core._NotOkStatusException as e:
       1038       six.raise_from(core._status_to_exception(e.code, e.message), None)
    
    
    RuntimeError: GPU:0 unknown device.

# gpu转cpu
    bb = b.cpu()
    bb.device

WARNING:tensorflow:From <ipython-input-14-57e39c66408d>:1: _EagerTensorBase.cpu (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
    Instructions for updating:
    Use tf.identity instead.
    
    
    
    
    
    '/job:localhost/replica:0/task:0/device:CPU:0'

# 或者numpy类型数据
    b.numpy()

array([0, 1, 2, 3], dtype=int32)

# 数据形状
    b.shape

TensorShape([4])

# 数据维度
    b.ndim

# 返回一个Tensor的数据维度
    tf.rank(b)

<tf.Tensor: shape=(), dtype=int32, numpy=1>

tf.ones([1,4,2,5]).shape

TensorShape([1, 4, 2, 5])

tf.rank(tf.ones([1,4,2,5]))

<tf.Tensor: shape=(), dtype=int32, numpy=4>

# 1.x版本的遗留
    # b.name

---------------------------------------------------------------------------
    
    AttributeError                            Traceback (most recent call last)
    
    <ipython-input-25-45bc9755420c> in <module>
    ----> 1 b.name
    
    
    ~/.virtualenvs/py38/lib/python3.8/site-packages/tensorflow/python/framework/ops.py in name(self)
       1120   @property
       1121   def name(self):
    -> 1122     raise AttributeError(
       1123         "Tensor.name is meaningless when eager execution is enabled.")
       1124 
    
    
    AttributeError: Tensor.name is meaningless when eager execution is enabled.

## 常见数据类型检测 ##

a = tf.constant([1.])
    b = tf.constant([True,False])
    c = tf.constant("hello.world.")
    d = np.arange(4)

# 判断变量是否是Tensor类型
    # 某种情况不适用
    isinstance(a, tf.Tensor)

True

# 判断数据是否是Tensor类型
    # 推荐这种方式
    tf.is_tensor(b)

True

tf.is_tensor(d)

False

# 数据类型查看
    a.dtype, b.dtype, c.dtype, c.dtype

(tf.float32, tf.bool, tf.string, tf.string)

# 数据类型比较
    a.dtype == tf.float32

True

c.dtype == tf.string

True

## Convert 数据类型转换 ##

a = np.arange(5)

a.dtype

dtype('int64')

# 数据类型转为Tensor
    aa = tf.convert_to_tensor(a)
    aa

<tf.Tensor: shape=(5,), dtype=int64, numpy=array([0, 1, 2, 3, 4])>

# 数据类型转为Tensor，指定数据类型
    aa = tf.convert_to_tensor(a,dtype=tf.int32)
    aa

<tf.Tensor: shape=(5,), dtype=int32, numpy=array([0, 1, 2, 3, 4], dtype=int32)>

# cast 一个专用的数据转换api
    tf.cast(aa, dtype=tf.float32)

<tf.Tensor: shape=(5,), dtype=float32, numpy=array([0., 1., 2., 3., 4.], dtype=float32)>

aaa = tf.cast(aa, dtype=tf.double)
    aaa

<tf.Tensor: shape=(5,), dtype=float64, numpy=array([0., 1., 2., 3., 4.])>

tf.cast(aaa, dtype=tf.int32)

<tf.Tensor: shape=(5,), dtype=int32, numpy=array([0, 1, 2, 3, 4], dtype=int32)>

## 数据转换–Int与Bool型 ##

b=tf.constant([0,1])
    b

<tf.Tensor: shape=(2,), dtype=float32, numpy=array([0., 1.], dtype=float32)>

bb=tf.cast(b,dtype=tf.bool)
    bb

<tf.Tensor: shape=(2,), dtype=bool, numpy=array([False,  True])>

tf.cast(bb,dtype=tf.int32)

<tf.Tensor: shape=(2,), dtype=int32, numpy=array([0, 1], dtype=int32)>

## tf.Variable ##

*  优化数据类型
 *  使变量具有一些特性，使之可求导特性

# 生成数据
    a=tf.range(5)
    a

<tf.Tensor: shape=(5,), dtype=int32, numpy=array([0, 1, 2, 3, 4], dtype=int32)>

# 包装数据
    b=tf.Variable(a)
    print(b.dtype)
    print(b.name)
    print(b.trainable)

<dtype: 'int32'>
    Variable:0
    True

# 指定名称，不推荐
    b=tf.Variable(a, name="input_data")
    # 数据类型
    print(b.dtype)
    # 变量名称
    print(b.name)
    # 设定trainable=False 可以防止该变量被数据流图的 GraphKeys.TRAINABLE_VARIABLES 收集, 这样我们就不会在训练的时候尝试更新它的值。
    print(b.trainable)

<dtype: 'int32'>
    input_data:0
    True

# 查看数据类型
    isinstance(b,tf.Tensor)

isinstance(b,tf.Variable)

True

tf.is_tensor(b)

True

b.numpy()

array([0, 1, 2, 3, 4], dtype=int32)

## 数据类型转换 TO numpy ##

# 生成数据
    a=tf.range(5)
    a

<tf.Tensor: shape=(5,), dtype=int32, numpy=array([0, 1, 2, 3, 4], dtype=int32)>

# 包装数据
    b=tf.Variable(a)