tensorflow2报错AttributeError: module ‘tensorflow‘ has no attribute ‘truncated_normal‘

淡淡的烟草味﹌ 2023-02-28 04:28 113阅读 0赞

tensorflow2报错AttributeError: module ‘tensorflow’ has no attribute ‘truncated_normal’

代码

  1. tn = tf.truncated_normal([prev_num_nodes, num_nodes], stddev=w_stddev)

报错信息

  1. ---------------------------------------------------------------------------
  2. AttributeError                            Traceback (most recent call last)
  3. <ipython-input-14-05dbe3f468af> in <module>()
  4.       5 
  5.       6 # new_alex = TransferedAlexNet(old_alex_graph, num_tgt_classes=2)
  6. ----> 7 new_alex = TransferedAlexNet(old_alex_graph)
  7.       8 print(new_alex)
  8.       9 inputs   = old_alex_graph.get_tensor_by_name('inputs/images:0')
  10. <ipython-input-12-d49ea2971fb0> in __init__(s, init_graph)
  11.      21             with tf.name_scope('new_top_layer'):
  12.      22                 # the old fc8 (which we are replacing) had 1000 nodes for 1000 classes
  13. ---> 23                 s.logits = helpers_07.fully_connected_xavier_relu_layer(orig_fc7)
  14.      24                 print( s.logits)
  15.      25 
  17.  in fully_connected_xavier_relu_layer(incoming_layer, num_nodes, b_val, keep_prob, name)
  18.      84                                  activation_fn = tf.nn.relu,
  19.      85                                  keep_prob=keep_prob,
  20. ---> 86                                  name=name)
  21.      87 
  22.      88 if __name__ == "__main__":
  24. in fully_connected_layer(incoming_layer, num_nodes, w_stddev, b_val, activation_fn, keep_prob, name)
  25.      56         # FIXME:  consider stddev=np.sqrt(2.0 / prev_num_nodes)
  26.      57         #         [fancy initialization]
  27. ---> 58         tn = tf.truncated_normal([prev_num_nodes, num_nodes], stddev=w_stddev)
  28.      59         W = tf.Variable(tn, name='W')
  29.      60         const = tf.constant(b_val, shape=[num_nodes])
  31. AttributeError: module 'tensorflow' has no attribute 'truncated_normal'

解决方法

  1. tf.random.truncated_normal([prev_num_nodes, num_nodes], stddev=w_stddev)

发表评论

表情：
评论列表 （有 0 条评论，113人围观）

还没有评论，来说两句吧...

相关阅读