@fchollet I'm hoping they'll at least wrap the LSTM implementation in cuDNN v5.
I feel like RNNs in TensorFlow are getting increasingly slower with each new TF release.
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@fchollet do you have some benchmarks?Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@fchollet Any opinion on theano vs tensorflow backends at this point? When would be a good time to switch from theano (if ever?) -
@_DaveSullivan the TF backend is very good for anything but RNNs. For RNNs, prefer the Theano backend.
End of conversation
New conversation -
-
-
@fchollet somehow for me it is still almost 2x faster than in Keras (regardless of a backend)
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@coffeephoenix@fchollet from me it was always TensorOOMThanks. 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.