tf.get_variable returns a variable that doesn't match the given shape

Question

import tensorflow as tf

if __name__ == "__main__":
    weights = tf.get_variable(
        name="weights",
        shape=[9, 3, 1, 64],
        dtype=tf.float32,
        initializer=tf.truncated_normal_initializer()
    ),
    print(tf.shape(weights, name="weights_shape"))

Output:

Tensor("weights_shape:0", shape=(5,), dtype=int32)

Can't figure out why the shape of "weights" does not match the given shape.

Any help would be appreciated!

You can refer https://stackoverflow.com/questions/37085430/tf-shape-get-wrong-shape-in-tensorflow — giser_yugang, Dec 18 '18 at 08:14

score 0 · Accepted Answer · edited Dec 18 '18 at 10:13

0

Oh my god!

found that this mystic error is just occurred by adding an extra "," after tf.get_variable()

But I am still curious why?

edited Dec 18 '18 at 10:13

Bsquare ℬℬ

4,423
11
24
44

answered Dec 18 '18 at 08:12

coco

1
2

adding a comma at the end of an expression is short-hand for creating a tuple, e.g. `a=3,` results in `a` being a tuple, so `a=3,` is equivalent to `a=(3,)` or `a=tuple(3)`. – nemo Dec 18 '18 at 10:13

tf.get_variable returns a variable that doesn't match the given shape

1 Answers1