AttributeError: 'Tensor' object has no attribute 'shape'

Question

Stack trace

Traceback (most recent call last):
  File "main.py", line 6, in <module>
    connection.start_socket(8089, callback=handler.message_processor)
  File "/mnt/d/workspace/SketchRecognitionWithTensorFlow/src/main/python/connection/python_socket_server.py", line 13, in start_socket
    process_message(connection, callback=callback)
  File "/mnt/d/workspace/SketchRecognitionWithTensorFlow/src/main/python/connection/python_socket_server.py", line 38, in process_message
    result = callback(general_proto)
  File "/mnt/d/workspace/SketchRecognitionWithTensorFlow/src/main/python/recognition/proto_handler.py", line 39, in message_processor
    return train_shape(general_proto.template)
  File "/mnt/d/workspace/SketchRecognitionWithTensorFlow/src/main/python/recognition/proto_handler.py", line 23, in train_shape
    rec.add_training_data(recognition_template.interpretation.label, recognition_template.shape)
  File "/mnt/d/workspace/SketchRecognitionWithTensorFlow/src/main/python/recognition/recognition_manager.py", line 98, in add_training_data
    self.recognizers[label].train(label, points)
  File "/mnt/d/workspace/SketchRecognitionWithTensorFlow/src/main/python/recognition/simple/recognizer.py", line 78, in train
    self.classifier.fit(x=reshaped_tensor, y=target, steps=1)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/estimators/estimator.py", line 173, in fit
    input_fn, feed_fn = _get_input_fn(x, y, batch_size)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/estimators/estimator.py", line 67, in _get_input_fn
    x, y, n_classes=None, batch_size=batch_size)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/io/data_feeder.py", line 117, in setup_train_data_feeder
    X, y, n_classes, batch_size, shuffle=shuffle, epochs=epochs)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/io/data_feeder.py", line 239, in __init__
    self.X.shape, None if self.y is None else self.y.shape, n_classes,
AttributeError: 'Tensor' object has no attribute 'shape'

Here is some code examples:

def create_classifier(self):
        hiddenLayers = [self.num_points, self.num_points * 2, 10]
        self.classifier = tf.contrib.learn.DNNClassifier(hidden_units=hiddenLayers)

label is a string point list is a list of an array of x, y values ex:

[[1,2],[2,3],[3,4],...]


def train(self, label, point_list):
    points = self.resample(point_list, self.num_points)
    utils.strip_ids_from_points(points)
    value_class = 1 if label == self.label else 0
    target = tf.reshape(tf.constant(value_class), [1])
    print 'training classifier to recognize value as: [' + str(value_class) + '] label is ' + label + ' class is ' + self.label
    point_tensor = tf.convert_to_tensor(points, dtype=tf.float32)
    reshaped_tensor = tf.reshape(point_tensor, [1, self.num_points * 2])
    print reshaped_tensor
    print target
    self.classifier.fit(x=reshaped_tensor, y=target, steps=1)

I do not know why it is saying that a tensor does not have a shape attribute or how to fix this problem. Any help appreciated.

dv3 · Accepted Answer · 2017-07-15 14:34:19Z

3

Since there is no accepted answer and I'm myself coming from Google:
Credit goes to mrry with this answer, which reads:
Since TensorFlow 1.0, tf.Tensor now has a tf.Tensor.shape property, which returns the same value as tf.Tensor.get_shape().

answered Jul 15, 2017 at 14:34

dv3

4,5896 gold badges30 silver badges53 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Yan King Yin Over a year ago

So what is the solution??

dv3 Over a year ago

@YanKingYin I don't see the problem - where are you stuck? if this answer doesn't help you, just open another question

jayemar · Accepted Answer · 2016-09-08 23:52:14Z

1

The stack trace doesn't help a lot with code snippets that don't have line numbers. But I can tell you that when I saw that error it was because I used tf.reshape on a numpy array instead of a true tensorflow tensor. The resulting object threw that error when I tried to determine the shape. Hope that helps.

answered Sep 8, 2016 at 23:52

jayemar

3712 silver badges11 bronze badges

Comments

kratenko · Accepted Answer · 2016-09-09 14:33:30Z

0

It's hard to see from the code provided, especially since I cannot figure out, where your error occurs. But for what it's worth, I have run into a related problem before:

It looks to me, like some code is trying to handle a tf-tensor as if it was a np-array. Tensors don't have a shape attribute, since their shape is stored as a more complicated object. If you want to get the information, you are used to from np, you'd have to call my_tensor.get_shape().as_list(). But since non of your code (that you posted) tries to access any shape attribute on a Tensor, I'm not sure how this information could be used to fix your problem.

answered Sep 9, 2016 at 14:33

kratenko

7,6445 gold badges39 silver badges63 bronze badges

Collectives™ on Stack Overflow

AttributeError: 'Tensor' object has no attribute 'shape'

3 Answers 3

2 Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related