Skip to content

Conversation

@facaiy
Copy link
Member

@facaiy facaiy commented Aug 28, 2019

Fix #428

@facaiy facaiy changed the title Fix: WeightNormalization data init fails WIP: Fix: WeightNormalization data init fails Aug 28, 2019
@facaiy
Copy link
Member Author

facaiy commented Aug 28, 2019

need to check whether it converges.

@seanpmorgan
Copy link
Member

need to check whether it converges.

Looks good to me:
https://colab.research.google.com/drive/17YMqWE3LglqQjUgUTlKY_m8L49t6W6Hd

Will do a review later

@facaiy
Copy link
Member Author

facaiy commented Aug 28, 2019

Good news, thank Sean :-)

trainable=True)
if self.data_init:
self._naked_layer = tf.keras.layers.deserialize(
tf.keras.layers.serialize(self.layer))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just wondering if you could comment on this. Creating a copy of the layer for the data init, but haven't seen this trick before. I guess no other way to copy a layer?

Looks great... thank you very much!

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure, will do, Sean

@seanpmorgan seanpmorgan merged commit cc89403 into tensorflow:master Aug 28, 2019
@facaiy facaiy deleted the BUG/weight_normal_tf_function branch August 28, 2019 22:48
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

WeightNormalization data init fails with Keras experimental_run_tf_function

3 participants