How do you use them to batch text of different lengths?
-
This Tweet is unavailable.
-
This Tweet is unavailable.
-
This Tweet is unavailable.
-
This Tweet is unavailable.
-
It's possible that newer versions of Keras have simplified this so that what we did wouldn't be necessary today. But still, in my experience, explicit masks are way better than implicit masks. You only occasionally need the mask.
Any reason why you wouldn't just explicitly pass the mask to the layers that need it? Every Keras layer that supports masking takes a mask argument, like this: y = LSTM(32)(inputs, mask=mask)
-
-
Yes, this is what we do now. I don't remember why we didn't do this originally; we were probably biased by trying to follow how `supports_masking` worked pre Keras 2.0, where it sure seemed like implicit masks was the encouraged design pattern.
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.