TensorFlow中如何确定张量的形状实例

我们可以使用tf.shape()获取某张量的形状张量。

import tensorflow as tf
x = tf.reshape(tf.range(1000), [10, 10, 10])
sess = tf.Session()
sess.run(tf.shape(x))
  
Out[1]: array([10, 10, 10])

我们可以使用tf.shape()在计算图中确定改变张量的形状。

high = tf.shape(x)[0] // 2
width = tf.shape(x)[1] * 2
x_reshape = tf.reshape(x, [high, width, -1])
sess.run(tf.shape(x_reshape))
  
Out: array([ 5, 20, 10])

我们可以使用tf.shape_n()在计算图中得到若干个张量的形状。

y = tf.reshape(tf.range(504), [7,8,9])
sess.run(tf.shape_n([x, y]))
  
Out: [array([10, 10, 10]), array([7, 8, 9])]

我们可以使用tf.size()获取张量的元素个数。

sess.run([tf.size(x), tf.size(y)])

Out: [1000, 504]

tensor.get_shape()或者tensor.shape是无法在计算图中用于确定张量的形状。

In [20]: x.get_shape()
Out[20]: TensorShape([Dimension(10), Dimension(10), Dimension(10)])
  
In [21]: x.get_shape()[0]
Out[21]: Dimension(10)
  
In [22]: type(x.get_shape()[0])
Out[22]: tensorflow.python.framework.tensor_shape.Dimension
  
In [23]: x.get_shape()
Out[23]: TensorShape([Dimension(10), Dimension(10), Dimension(10)])
  
In [24]: sess.run(x.get_shape())
---------------------------------------------------------------------------
TypeError     Traceback (most recent call last)
~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in __init__(self, fetches, contraction_fn)
 299  self._unique_fetches.append(ops.get_default_graph().as_graph_element(
--> 300  fetch, allow_tensor=True, allow_operation=True))
 301 except TypeError as e:
  
~\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py in as_graph_element(self, obj, allow_tensor, allow_operation)
 3477 with self._lock:
-> 3478 return self._as_graph_element_locked(obj, allow_tensor, allow_operation)
 3479
  
~\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py in _as_graph_element_locked(self, obj, allow_tensor, allow_operation)
 3566 raise TypeError("Can not convert a %s into a %s." % (type(obj).__name__,
-> 3567        types_str))
 3568
  
TypeError: Can not convert a TensorShapeV1 into a Tensor or Operation.
  
During handling of the above exception, another exception occurred:
  
TypeError     Traceback (most recent call last)
<ipython-input-24-de007c69e003> in <module>
----> 1 sess.run(x.get_shape())
  
~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in run(self, fetches, feed_dict, options, run_metadata)
 927 try:
 928 result = self._run(None, fetches, feed_dict, options_ptr,
--> 929    run_metadata_ptr)
 930 if run_metadata:
 931  proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)
  
~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in _run(self, handle, fetches, feed_dict, options, run_metadata)
 1135 # Create a fetch handler to take care of the structure of fetches.
 1136 fetch_handler = _FetchHandler(
-> 1137  self._graph, fetches, feed_dict_tensor, feed_handles=feed_handles)
 1138
 1139 # Run request and get response.
  
~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in __init__(self, graph, fetches, feeds, feed_handles)
 469 """
 470 with graph.as_default():
--> 471 self._fetch_mapper = _FetchMapper.for_fetch(fetches)
 472 self._fetches = []
 473 self._targets = []
~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in for_fetch(fetch)
 269  if isinstance(fetch, tensor_type):
 270  fetches, contraction_fn = fetch_fn(fetch)
--> 271  return _ElementFetchMapper(fetches, contraction_fn)
 272 # Did not find anything.
 273 raise TypeError('Fetch argument %r has invalid type %r' % (fetch,
~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in __init__(self, fetches, contraction_fn)
 302  raise TypeError('Fetch argument %r has invalid type %r, '
 303    'must be a string or Tensor. (%s)' %
--> 304    (fetch, type(fetch), str(e)))
 305 except ValueError as e:
 306  raise ValueError('Fetch argument %r cannot be interpreted as a '
TypeError: Fetch argument TensorShape([Dimension(10), Dimension(10), Dimension(10)]) has invalid type <class 'tensorflow.python.framework.tensor_shape.TensorShapeV1'>, must be a string or Tensor. (Can not convert a TensorShapeV1 into a Tensor or Operation.)

我们可以使用tf.rank()来确定张量的秩。tf.rank()会返回一个代表张量秩的张量，可直接在计算图中使用。

In [25]: tf.rank(x)
Out[25]: <tf.Tensor 'Rank:0' shape=() dtype=int32>
  
In [26]: sess.run(tf.rank(x))
Out[26]: 3

补充知识：tensorflow循环改变tensor的值

使用tf.concat()实现4维tensor的循环赋值

alist=[[[[1,1,1],[2,2,2],[3,3,3]],[[4,4,4],[5,5,5],[6,6,6]]],[[[7,7,7],[8,8,8],[9,9,9]],[[10,10,10],[11,11,11],[12,12,12]]]] #2,2,3,3-n,c,h,w
kenel=(np.asarray(alist)*2).tolist()
print(kenel)
inputs=tf.constant(alist,dtype=tf.float32)
kenel=tf.constant(kenel,dtype=tf.float32)
inputs=tf.transpose(inputs,[0,2,3,1]) #n,h,w,c
kenel=tf.transpose(kenel,[0,2,3,1]) #n,h,w,c
uints=inputs.get_shape()
h=int(uints[1])
w=int(uints[2])
encoder_output=[]
for b in range(int(uints[0])):
 encoder_output_c=[]
 for c in range(int(uints[-1])):
  one_channel_in = inputs[b, :, :, c]
  one_channel_in = tf.reshape(one_channel_in, [1, h, w, 1])
  one_channel_kernel = kenel[b, :, :, c]
  one_channel_kernel = tf.reshape(one_channel_kernel, [h, w, 1, 1])
  encoder_output_cc = tf.nn.conv2d(input=one_channel_in, filter=one_channel_kernel, strides=[1, 1, 1, 1], padding="SAME")
  if c==0:
   encoder_output_c=encoder_output_cc
  else:
   encoder_output_c=tf.concat([encoder_output_c,encoder_output_cc],axis=3)
 
 if b==0:
  encoder_output=encoder_output_c
 else:
  encoder_output = tf.concat([encoder_output, encoder_output_c], axis=0)
 
with tf.Session() as sess:
 print(sess.run(tf.transpose(encoder_output,[0,3,1,2])))
 print(encoder_output.get_shape())

输出：

[[[[ 32. 48. 32.]
 [ 56. 84. 56.]
 [ 32. 48. 32.]]
 
 [[ 200. 300. 200.]
 [ 308. 462. 308.]
 [ 200. 300. 200.]]]
 
 
 [[[ 512. 768. 512.]
 [ 776. 1164. 776.]
 [ 512. 768. 512.]]
 
 [[ 968. 1452. 968.]
 [1460. 2190. 1460.]
 [ 968. 1452. 968.]]]]
(2, 3, 3, 2)

以上这篇TensorFlow中如何确定张量的形状实例就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持自学编程网。

编程技巧