这个问题常见且不小心就会遇到,记录下O(∩_∩)O哈！

frr@frr:~/Documents/code/pointnet/models$ python pointnet_SE.py
('input_image', <tf.Tensor 'ExpandDims:0' shape=(32, 1024, 3, 1) dtype=float32>)
Traceback (most recent call last):
  File "pointnet_SE.py", line 138, in <module>
    net, end_points = get_model(inputs, True)
  File "pointnet_SE.py", line 37, in get_model
    transform = input_transform_net(net, is_training, bn_decay, K=3)
  File "/home/frr/Documents/code/pointnet/models/transform_nets.py", line 29, in input_transform_net
    scope='tconv1', bn_decay=bn_decay)
  File "/home/frr/Documents/code/pointnet/models/../utils/tf_util.py", line 167, in conv2d
    bn_decay=bn_decay, scope='bn')
  File "/home/frr/Documents/code/pointnet/models/../utils/tf_util.py", line 536, in batch_norm_for_conv2d
    return batch_norm_template(inputs, is_training, scope, [0, 1, 2], bn_decay)
  File "/home/frr/Documents/code/pointnet/models/../utils/tf_util.py", line 482, in batch_norm_template
    lambda: tf.no_op())
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/util/deprecation.py", line 289, in new_func
    return func(*args, **kwargs)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/control_flow_ops.py", line 1802, in cond
    raise TypeError("pred must not be a Python bool")
TypeError: pred must not be a Python bool

贴出一段先吓一下小伙伴们,有些小伙伴可能刚接触代码没有很久,碰到error就会非常烦躁,甚至直接就give up了,但是作为祖国花朵的我们怎么能说放弃就放弃呢?其实看着这段报错很吓人,有用的只有最后一行!!

TypeError: pred must not be a Python bool

其实就是个类型错误,或者不兼容的问题,
LZ直接定义is_training = True,因为定义的is_training=True不是tf.bool类型,所以导致不兼容,只要定义的时候加上type就行了,举个例子:

is_training = tf.cast(True, tf.bool)

调代码就还是得慢慢来,切勿心浮气躁,小伙伴们加油!哈哈,顺便祝下自己生日快乐,Felaim每天要加油,开心O(∩_∩)O~~