What's wrong with `tf.zeros_initializer` in `get_variable()`? - tensorflow

Question

I want to train some adversarial examples using CW algorithm, and I used an example from here and a CW implementation from here. But I encountered an error about tf.zeros_initializer:
ValueError: The initializer passed is not valid. It should be a callable with no arguments and the shape should not be provided or an instance of
'tf.keras.initializers.*' and `shape` should be fully defined.
Edit: It seems that non-fully defined shape conflicts with using initializers. How can I fix it?
Here's a piece of code:
# ... omitted
with tf.variable_scope('model', reuse=tf.AUTO_REUSE):
# CW
_, env.adv_cw, _ = cw.cw(model, env.x)
Here's env.x:
env.x = tf.placeholder(tf.float32, (None, width, height, channels), name='x')
When I run the code, I get the error message:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-39-712c8b007d37> in <module>()
8 with tf.variable_scope('model', reuse=tf.AUTO_REUSE):
9 # CW
---> 10 _, env.adv_cw, _ = cw.cw(model, env.x)
5 frames
/content/cw.py in cw(model, x, y, eps, ord_, T, optimizer, alpha, min_prob, clip)
50 """
51 xshape = x.get_shape().as_list()
---> 52 noise = tf.get_variable('noise', shape=xshape, dtype=tf.float32,
53 initializer=tf.zeros_initializer)
54
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/variable_scope.py in get_variable(name, shape, dtype, initializer, regularizer, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter, constraint, synchronization, aggregation)
1494 constraint=constraint,
1495 synchronization=synchronization,
-> 1496 aggregation=aggregation)
1497
1498
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/variable_scope.py in get_variable(self, var_store, name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter, constraint, synchronization, aggregation)
1237 constraint=constraint,
1238 synchronization=synchronization,
-> 1239 aggregation=aggregation)
1240
1241 def _get_partitioned_variable(self,
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/variable_scope.py in get_variable(self, name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter, constraint, synchronization, aggregation)
560 constraint=constraint,
561 synchronization=synchronization,
--> 562 aggregation=aggregation)
563
564 def _get_partitioned_variable(self,
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/variable_scope.py in _true_getter(name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource, constraint, synchronization, aggregation)
512 constraint=constraint,
513 synchronization=synchronization,
--> 514 aggregation=aggregation)
515
516 synchronization, aggregation, trainable = (
/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/variable_scope.py in _get_single_variable(self, name, shape, dtype, initializer, regularizer, partition_info, reuse, trainable, collections, caching_device, validate_shape, use_resource, constraint, synchronization, aggregation)
906 variable_dtype = None
907 else:
--> 908 raise ValueError("The initializer passed is not valid. It should "
909 "be a callable with no arguments and the "
910 "shape should not be provided or an instance of "
ValueError: The initializer passed is not valid. It should be a callable with no arguments and the shape should not be provided or an instance of `tf.keras.initializers.*' and `shape` should be fully defined.
But Google's TensorFlow Guide gives an example of the usage of get_variable:
my_int_variable = tf.get_variable("my_int_variable", [1, 2, 3], dtype=tf.int32,
initializer=tf.zeros_initializer)
Environment: Google Colab, TensorFlow 1.14.0-rc1, Python 3.6

Answer 1

Just make changes according to your placeholder dimension, let me take example of your placeholder variable.
** x = placeholder(t f. float 32, (None, width, height, channels), name='x')**.
It has 4 dimension : [None, width, height, channels], but width, height, channels is not defined, means, for an image width =6, height=6, channels=3 is defined so the tensor dimension is [6 x 6 x 3].
What you can do is, the image you read, get the all three dimension value in different variable and pass it to your placeholder variable.
Ex.
Image A = 32 x 32 x 3
width = A.shape[0]
height = A.shape[1]
channels =A.shape[2]
Or you can directly provide the value to width, height, channels (If you know your input data shape) by this way the placeholder will be defined.