Using tf.get_variable(...) with an unknown batch size

Question

Given that I have some input batch x with shape (batch_size, n_features) how would I use tf.get_variable(...) to create another matrix with shape (batch_size) if it isn't known ahead of time.

For example I am able to do:

b = tf.zeros(shape = (tf.shape(x)[0]))

But I am unable to do:

b = tf.get_variable("b", shape = (tf.shape(x)[0]), initializer = tf.constant_initializer(0.0))

You have a typo right there, after shape= there should be a parenthesis opening, in both code snippets — bluesummers
– bluesummers, Commented May 25, 2017 at 5:56
How does batch size affects one of your Variable ? It is common that they donnot interact. — P-Gn
– P-Gn, Commented May 25, 2017 at 7:14
When you need to create the hidden state for an RNN don't you need to know how many samples are in the batch? — YellowPillow
– YellowPillow, Commented May 25, 2017 at 9:16
if the problem is solved due to a simple typo, just remove the question — Salvador Dali
– Salvador Dali, Commented May 25, 2017 at 19:51

dexhunter · Accepted Answer · 2017-12-21 13:16:22Z

1

You can use tf.Variable(dims=[tf.shape(x)[0]], validate_shape=False) to set a dynamic shape for batch size.

answered Dec 21, 2017 at 13:16

dexhunter

57810 silver badges25 bronze badges

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

Collectives™ on Stack Overflow

Using tf.get_variable(...) with an unknown batch size

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related