With most Python questions the answer is the accepted StackOverflow answer, with TensorFlow you really have to read every question that is even remotely similar to yours and then the answer is the unaccepted answer right at the bottom of the 9th question you read. 
-
-
I think they could start by writing the documentation to actually reflect what the code does. Especially those who release their weights pretending like it's open science. It's not open if nobody knows what the hell you did. Often they didn't even do what's in the preprint!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
In terms of how to design their library (it reminds me a lot of web dev stuff where nothing is ever permanent) and in terms of how to actually address individual science problems.
-
Yes. I'm with you.

-
In fairness the "instability", as with web dev, is in response to rapidly changing needs of extremely diverse userbases. I'm really quite partial to stability though.
-
I think other libraries might be more stable, or maybe wrapper libraries at least?

-
I haven't had a chance to check out caffe, theano, keras, etc., yet!

-
It's all Greek to me xD
-
Ah, my 2 years of ancient Greek in school...fails me =_=
- 2 more replies
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.