In order to learn more about the tech underlying DNN, I've been tinkering with a toy, very statically-typed, TensorFlow-like system in Scala. It is nowhere near ready for useful work and likely never will be, but lemme know if this piques your interest:https://github.com/avibryant/nda
-
-
Let me tell you about that time one of my co-workers wasted two weeks of RNN training because of an incredibly hard-to-find bug that was obscured by python's permissive typing...
-
I'm not even going to claim the python ML ecosystem isn't good (it is), but dynamic typing bugs are a real thing.
-
Win, lose or draw, I
their mascot.
End of conversation
New conversation -
-
-
In my mind python is also just a wrapper for C/C++ in this world. No reason we can’t have a strongly typed library on top of the same c code. Theoretically makes even more sense with the TF graphs/sessions.
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.

... If I could find it 