CODEWITHSUNDEEP
pythontensorflow
hannahrae
hannahrae

Reputation: 61

Tensor returned by tf.tranpose() different when stored?

I am writing an application using TensorFlow and I'm using the tf.transpose() function. The API states that the function returns a transposed tensor, which is what you'd expect. However, I noticed the following phenomenon:

>>> tf.transpose([3, 5])
    <tf.Tensor 'transpose:0' shape=(2,) dtype=int32>
>>> a = tf.transpose([3, 5])
>>> a
   <tf.Tensor 'transpose_1:0' shape=(2,) dtype=int32>
>>> a ==  tf.transpose([3, 5])

Does anyone know why this happens or how it should be used?

Upvotes: 1

Views: 58

Answers (1)

hannahrae
hannahrae

Reputation: 61

Oops, I answered my question as soon as I posted it... I think they are just not equivalent because they are two different tensor objects, even though they have the same value. I was thrown off by the naming convention. We can see this here:

>>> a = tf.transpose([3, 5], name='a')
>>> tf.transpose([3, 5], name='b')
    <tf.Tensor 'b:0' shape=(2,) dtype=int32>
>>> a
    <tf.Tensor 'a:0' shape=(2,) dtype=int32>

Upvotes: 1

Related Questions