文章目录
张量的阶和数据类型：
1：TensorFlow 的基本数据格式
2：一个类型化的N维数组（tf.Tensor）
3:三部分，名字，形状，数据类型
张量属性：
.graph 张量所属的默认图
.op 张量的操作名
.name 张量的字符串描述
.shape 张量形状
TensorFlow：打印出来的形状表示
0维：()
1维：(n)
2维：(n,m)
3维: (n,m,l)
1.张量的动态形状与静态形状
.TensorFlow中，张量具有静态形状和动态形状
.静态形状：创建一个张量，初始状态的形状
.tf.Tensor.get_shape:获取静态形状
.tf.Tensor.set_shape():更新Tensor对象的静态形状
.动态形状：一种描述原始张量在执行过程中的一种形状（动态变化）
.tf.reshape:创建一个具有不同动态形状的新张量
设置静态形状
import tensorflow as tf
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\dtypes.py:517: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
_np_qint8 = np.dtype([("qint8", np.int8, 1)])
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\dtypes.py:518: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
_np_quint8 = np.dtype([("quint8", np.uint8, 1)])
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\dtypes.py:519: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
_np_qint16 = np.dtype([("qint16", np.int16, 1)])
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\dtypes.py:520: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
_np_quint16 = np.dtype([("quint16", np.uint16, 1)])
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\dtypes.py:521: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
_np_qint32 = np.dtype([("qint32", np.int32, 1)])
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\dtypes.py:526: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
np_resource = np.dtype([("resource", np.ubyte, 1)])
plt = tf.placeholder(tf.float32,[None,2])
plt
<tf.Tensor 'Placeholder:0' shape=(?, 2) dtype=float32>
plt.set_shape([3,2])
plt
<tf.Tensor 'Placeholder:0' shape=(3, 2) dtype=float32>
错误的修改方式1：静态形状不能跨维度修改
plt.set_shape([3,2,2])
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\tensor_shape.py in merge_with(self, other)
578 try:
--> 579 self.assert_same_rank(other)
580 new_dims = []
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\tensor_shape.py in assert_same_rank(self, other)
623 raise ValueError("Shapes %s and %s must have the same rank" % (self,
--> 624 other))
625
ValueError: Shapes (3, 2) and (3, 2, 2) must have the same rank
During handling of the above exception, another exception occurred:
ValueError Traceback (most recent call last)
<ipython-input-4-c9e5d4ac7673> in <module>
----> 1 plt.set_shape([3,2,2])
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\ops.py in set_shape(self, shape)
468 """
469 if not _USE_C_API:
--> 470 self._shape_val = self._shape_val.merge_with(shape)
471 return
472 if not isinstance(shape, tensor_shape.TensorShape):
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\tensor_shape.py in merge_with(self, other)
583 return TensorShape(new_dims)
584 except ValueError:
--> 585 raise ValueError("Shapes %s and %s are not compatible" % (self, other))
586
587 def concatenate(self, other):
ValueError: Shapes (3, 2) and (3, 2, 2) are not compatible
错误的修改方式2：再次设置静态形状（会报错，只能设置一次）
plt.set_shape([4,2])
# ValueError: Shapes (3, 2) and (4, 2) are not compatible
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\tensor_shape.py in merge_with(self, other)
581 for i, dim in enumerate(self._dims):
--> 582 new_dims.append(dim.merge_with(other[i]))
583 return TensorShape(new_dims)
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\tensor_shape.py in merge_with(self, other)
139 other = as_dimension(other)
--> 140 self.assert_is_compatible_with(other)
141 if self._value is None:
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\tensor_shape.py in assert_is_compatible_with(self, other)
112 raise ValueError("Dimensions %s and %s are not compatible" % (self,
--> 113 other))
114
ValueError: Dimensions 3 and 4 are not compatible
During handling of the above exception, another exception occurred:
ValueError Traceback (most recent call last)
<ipython-input-5-3a8be0759c51> in <module>
----> 1 plt.set_shape([4,2])
2
3 # ValueError: Shapes (3, 2) and (4, 2) are not compatible
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\ops.py in set_shape(self, shape)
468 """
469 if not _USE_C_API:
--> 470 self._shape_val = self._shape_val.merge_with(shape)
471 return
472 if not isinstance(shape, tensor_shape.TensorShape):
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\tensor_shape.py in merge_with(self, other)
583 return TensorShape(new_dims)
584 except ValueError:
--> 585 raise ValueError("Shapes %s and %s are not compatible" % (self, other))
586
587 def concatenate(self, other):
ValueError: Shapes (3, 2) and (4, 2) are not compatible
通过动态张量再次修改（创建一个新的张量）
plt_reshape = tf.reshape(plt,[2,3])
plt_reshape
<tf.Tensor 'Reshape:0' shape=(2, 3) dtype=float32>
动态形状修改一定注意元素数量匹配
plt_reshape2 = tf.reshape(plt,[3,3])
plt_reshape2
# ValueError: Cannot reshape a tensor with 6 elements to shape [3,3] (9 elements) for 'Reshape_2' (op: 'Reshape') with input shapes: [3,2], [2] and with input tensors computed as partial shapes: input[1] = [3,3].
---------------------------------------------------------------------------
InvalidArgumentError Traceback (most recent call last)
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\common_shapes.py in _call_cpp_shape_fn_impl(op, input_tensors_needed, input_tensors_as_shapes_needed, require_shape_fn)
685 graph_def_version, node_def_str, input_shapes, input_tensors,
--> 686 input_tensors_as_shapes, status)
687 except errors.InvalidArgumentError as err:
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\errors_impl.py in __exit__(self, type_arg, value_arg, traceback_arg)
515 compat.as_text(c_api.TF_Message(self.status.status)),
--> 516 c_api.TF_GetCode(self.status.status))
517 # Delete the underlying status object from memory otherwise it stays alive
InvalidArgumentError: Cannot reshape a tensor with 6 elements to shape [3,3] (9 elements) for 'Reshape_1' (op: 'Reshape') with input shapes: [3,2], [2] and with input tensors computed as partial shapes: input[1] = [3,3].
During handling of the above exception, another exception occurred:
ValueError Traceback (most recent call last)
<ipython-input-7-55ae297cd228> in <module>
----> 1 plt_reshape2 = tf.reshape(plt,[3,3])
2 plt_reshape2
3
4 # ValueError: Cannot reshape a tensor with 6 elements to shape [3,3] (9 elements) for 'Reshape_2' (op: 'Reshape') with input shapes: [3,2], [2] and with input tensors computed as partial shapes: input[1] = [3,3].
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\ops\gen_array_ops.py in reshape(tensor, shape, name)
5099 if _ctx.in_graph_mode():
5100 _, _, _op = _op_def_lib._apply_op_helper(
-> 5101 "Reshape", tensor=tensor, shape=shape, name=name)
5102 _result = _op.outputs[:]
5103 _inputs_flat = _op.inputs
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\op_def_library.py in _apply_op_helper(self, op_type_name, name, **keywords)
785 op = g.create_op(op_type_name, inputs, output_types, name=scope,
786 input_types=input_types, attrs=attr_protos,
--> 787 op_def=op_def)
788 return output_structure, op_def.is_stateful, op
789
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\ops.py in create_op(self, op_type, inputs, dtypes, input_types, name, attrs, op_def, compute_shapes, compute_device)
3271 op_def=op_def)
3272 self._create_op_helper(ret, compute_shapes=compute_shapes,
-> 3273 compute_device=compute_device)
3274 return ret
3275
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\ops.py in _create_op_helper(self, op, compute_shapes, compute_device)
3311 # compute_shapes argument.
3312 if op._c_op or compute_shapes: # pylint: disable=protected-access
-> 3313 set_shapes_for_outputs(op)
3314 # TODO(b/XXXX): move to Operation.__init__ once _USE_C_API flag is removed.
3315 self._add_op(op)
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\ops.py in set_shapes_for_outputs(op)
2499 return _set_shapes_for_outputs_c_api(op)
2500 else:
-> 2501 return _set_shapes_for_outputs(op)
2502
2503
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\ops.py in _set_shapes_for_outputs(op)
2472 shape_func = _call_cpp_shape_fn_and_require_op
2473
-> 2474 shapes = shape_func(op)
2475 if shapes is None:
2476 raise RuntimeError(
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\ops.py in call_with_requiring(op)
2402
2403 def call_with_requiring(op):
-> 2404 return call_cpp_shape_fn(op, require_shape_fn=True)
2405
2406 _call_cpp_shape_fn_and_require_op = call_with_requiring
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\common_shapes.py in call_cpp_shape_fn(op, require_shape_fn)
625 res = _call_cpp_shape_fn_impl(op, input_tensors_needed,
626 input_tensors_as_shapes_needed,
--> 627 require_shape_fn)
628 if not isinstance(res, dict):
629 # Handles the case where _call_cpp_shape_fn_impl calls unknown_shape(op).
C:\Anaconda\envs\tensorflow16\lib\site-packages\tensorflow\python\framework\common_shapes.py in _call_cpp_shape_fn_impl(op, input_tensors_needed, input_tensors_as_shapes_needed, require_shape_fn)
689 missing_shape_fn = True
690 else:
--> 691 raise ValueError(err.message)
692
693 if missing_shape_fn:
ValueError: Cannot reshape a tensor with 6 elements to shape [3,3] (9 elements) for 'Reshape_1' (op: 'Reshape') with input shapes: [3,2], [2] and with input tensors computed as partial shapes: input[1] = [3,3].
总结
对于静态形状来说，一旦张量形状固定了，不能再次设置静态形状，不能跨维度修改
动态形状可以去创建一个新的张量,改变的时候一定要注意元素数量要匹配
2.张量操作
生成张量
固定值张量：
tf.zeros(shepe,dtype = tf.float32,name = None) # 创建所有元素值为0的张量
tf.ones(shape,dtype = tf.float32,name = None) # 创建一个所有元素值为1的张量
生成全部元素为0和元素为1的张量
zero = tf.zeros([3,4],tf.float32)
zero
<tf.Tensor 'zeros:0' shape=(3, 4) dtype=float32>
one = tf.ones([3,4],tf.float32)
one
<tf.Tensor 'ones:0' shape=(3, 4) dtype=float32>
结果输出
with tf.Session() as sess:
print(sess.run([zero,one]))
[array([[0., 0., 0., 0.],
[0., 0., 0., 0.],
[0., 0., 0., 0.]], dtype=float32), array([[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]], dtype=float32)]
随机值张量：
tf.random_normal(shape,mean = 0.0,stddev = 1.0,dtype = tf.float32,seed = None,name = None) # 从正态分布中输出随机值，由随机正态分布的数字组成的矩阵
生成随机张量
random_1 = tf.random_normal([3,4],mean = 0.0,stddev = 1,dtype = tf.float32)
random_1
<tf.Tensor 'random_normal:0' shape=(3, 4) dtype=float32>
结果输出
with tf.Session() as sess:
print(random_1,'\n')
print(sess.run(random_1),'\n')
print(random_1)
Tensor("random_normal:0", shape=(3, 4), dtype=float32)
[[ 1.1654192 0.7380267 0.44780177 0.9135542 ]
[ 0.22852856 -0.0483947 0.3628924 0.13235609]
[ 0.6036132 -0.47465008 0.10996567 -1.3579988 ]]
Tensor("random_normal:0", shape=(3, 4), dtype=float32)
张量变换
张量变换（改变张量中数据类型）
tf.cast(x,dtype,name = None)
tf.sring_to_number(string_tensor,out_type = None,name = None)
tf.to_double(x,name = ‘ToDouble’);tf.to_float(x,name = ‘ToFloat’);…
类型转换(其中cast最常用)
constant_2 = tf.constant([1,2,3,4,5,6],shape = [2,3])
zeros_2 = tf.zeros([3.0,3.0],dtype = tf.float32)
ones_2 = tf.ones([3,3],dtype = tf.int32)
convert_1 = tf.cast(constant_2,tf.float32)
convert_2 = tf.cast(zeros_2,tf.int32)
convert_3 = tf.to_int32(ones_2,name = 'ToInt32')
转换结果输出
with tf.Session() as sess:
print(sess.run(convert_1))
print(sess.run(convert_2))
print(sess.run(convert_3))
[[1. 2. 3.]
[4. 5. 6.]]
[[0 0 0]
[0 0 0]
[0 0 0]]
[[1 1 1]
[1 1 1]
[1 1 1]]
张量变换（改变形状）
tf.reshape(tensor,shape,name = None)
a = tf.ones([2,4])
with tf.Session() as sess:
print(sess.run(tf.reshape(a,[4,2])))
[[1. 1.]
[1. 1.]
[1. 1.]
[1. 1.]]
张量合并
tf.concat()
a = tf.ones([2,2])
b = tf.zeros([2,2])
with tf.Session() as sess:
print(sess.run(tf.concat([a,b],axis = 0)))
print(sess.run(tf.concat([a,b],axis = 1)))
[[1. 1.]
[1. 1.]
[0. 0.]
[0. 0.]]
[[1. 1. 0. 0.]
[1. 1. 0. 0.]]
版权声明：本文为qq_33489955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。