Here is the same dynamic RNN implemented in 4 different frameworks (TensorFlow/Keras, MXNet/Gluon, Chainer, PyTorch). Can you tell which is which?pic.twitter.com/nsfuTULlKS
Show this thread
This API style was initially introduced by Chainer in 2015, and in 2017 multiple other frameworks adopted it (including those listed above). In the context of Keras, I call it "the Model subclassing API". It's a good fit for NLP research in particular!
-
-
You'll hear some say, "you can't implement X in framework Y, you need framework Z". It was already nonsense before all frameworks supported the same API style, but it's even more ridiculous now that you can port code from one library to any other with only minor syntax changes
Show this thread -
It's in the same category of nonsense as people claiming "you get better accuracy if you implement this model in framework X!" -- the output of an algorithm does not depend on the language you implement it in.
Show this thread
End of conversation
New conversation -
-
-
Interesting that you have called it "Model subclassing API". Have you identified any other interesting styles and patterns?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.