wcarvalho / jupyter_notebooks Goto Github PK

x = tf.placeholder(tf.float32, shape=(None, ))
q = tf.FIFOQueue(q_capacity, tf.float32, shapes=[x.get_shape()])

x.get_shape() is None, and the code outputs an error

Hi! Thanks for this nice resource.
When giving FIFOQueue the shapes param: shapes=[x.get_shape() for x in sentences_tensors
It raises:
ValueError: Queue shapes must have the same length as dtypes

But when I replace:
q = tf.FIFOQueue(100, tf.int32, shapes=[x.get_shape() for x in sentence_tensors])
With:
q = tf.FIFOQueue(100, dtypes=[tf.int32, tf.int32, tf.int32, tf.int32, tf.int32] , shapes=[x.get_shape() for x in sentences_tensors])
where the len of [tf.int32, ...,..., tf.int32] is just the number of nested lists in sentences_tensors
The error disappears. Did you experience this behaviour?

Thanks!!