lol, especially for NLP researchers that don't want to bring about apocalyptic panopticism by using @PyTorch 
As of TensorFlow 1.7, with eager execution, you can use a Chainer-like API in tf.keras to create models with dynamic control flow. This should be especially useful for NLP researchers working with recursive models!pic.twitter.com/ZfAs3u0HJ0
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Finally! Thanks. :)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Gotta love Keras' syntax and clean style.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Life imitates art!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
This Tweet is unavailable.
-
Depends on your use case. If you don't need it you can just keep it turned off, but if you do it adds some flexibility to the use of tensorflow api calls etc...
End of conversation
-
-
-
Really nice! I look forward to adding wrappers for use in spaCy. We already have this underway for PyTorch models :) Is there an easy way to pass in a gradient wrt output to update the graph from a given node? (You don't need to tell me what it is; I can get that from the docs)
-
this would be brilliant Matthew! looking forward to getting my hands on them
End of conversation
New conversation -
-
-
Good news, perhaps some improvements in speed as well, since tf.keras has good execution time for big text data.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
How benchmarks of eager in 1.7 compared to
@PyTorch or even static graph in tf ? -
On a single GPU, it's generally about as fast as graph TF.
End of conversation
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.