Thank you and the team so much for all the new toys!
New Keras release: 2.0.9. Lots of changes since the previous version. https://github.com/fchollet/keras/releases/tag/2.0.9 … Thanks to all who contributed to this effort!
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Thx
@fchollet and all who helped!Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Can I use Bidirectional for CuDNNLSTM in 2.0.9? CuDNNLSTM don't have argument for go_backwards. Is it intended limitation for gpu?
-
It should be supported in the future. We'll take a look at it
End of conversation
New conversation -
-
-
Thank you so much for the update and all the manpower behind the scenes. Will make a good use of it.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Thank you all for consistently improving such a user-friendly library!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Questions: Do CuDNNLSTMs defs automatically transformed into LSTMs in case TF-GPU version is not supported when we load back trained models?
-
You can now load a CuDNNLSTM's weights into a LSTM. But you will have to manually change CuDNNLSTM to LSTM in your models.
End of conversation
New conversation -
-
-
Love Keras, curious how best to think about Keras stand alone revisions vs. Embedded TensorFlow versions? (spelled out anyplace?)
-
tf.keras will generally be more compatible with TF internals and should be preferred as part of a pure-TF workflow
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.