Tensorflow text_generation

Question

I'm working through the code
https://www.tensorflow.org/tutorials/sequences/text_generation

When I arrive at the line the following error is produced.

 sampled_indices = tf.random.categorical(example_batch_predictions[0], num_samples=1)
 sampled_indices = tf.squeeze(sampled_indices,axis=-1).numpy()

Error

AttributeError
Traceback (most recent call last)

in

----> 1 sampled_indices = tf.random.categorical(example_batch_predictions[0], num_samples=1)
2 sampled_indices = tf.squeeze(sampled_indices,axis=-1).numpy()

AttributeError: module 'tensorflow._api.v1.random' has no attribute 'categorical'

System information - TensorFlow version: 1.12 Jupyter NoteBooks on Ubuntu

Any pointers for what could be missing? I'm wondering if there is an import missing?

You can now update to tf 2.0. This issue is fixed in 2.0.

Prabhat Doongarwal
– Prabhat Doongarwal

2019-10-01 11:24:02 +00:00
Commented Oct 1, 2019 at 11:24 — Prabhat Doongarwal
– Prabhat Doongarwal, Commented Oct 1, 2019 at 11:24

Leo Lee · Accepted Answer · 2019-01-16 03:53:43Z

4

tf.random.categorical probably has been changed to tf.random.multinomial, assuming example_batch_predictions are logits: https://www.tensorflow.org/api_docs/python/tf/random/multinomial

answered Jan 16, 2019 at 3:53

Leo Lee

664 bronze badges

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

1 Comment

satnhak Over a year ago

Looks like tf.keras.losses.sparse_categorical_crossentropy can be replaced by tf.keras.losses.sparse_categorical_crossentropy. So return tf.nn.sparse_softmax_cross_entropy_with_logits(labels=labels, logits=logits).

Collectives™ on Stack Overflow

Tensorflow text_generation

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related