Over the past few days I've seen lots of green accounts on HN pushing the narrative that PyTorch is "taking over" DL research...
-
-
Show this thread
-
...this is utterly contradicted by every metric I monitor with regard to usage in the research community
Show this thread -
Keras has added more researchers (and general users) in the past 6 months than PyTorch has
Show this thread
End of conversation
New conversation -
-
-
What's the ratio of tf:theano in keras?
-
Not sure when it comes to researchers specifically, but in general TF is ~80-85% of Keras users.
- Show replies
New conversation -
-
-
How is theano still that big? Historical inertia? What are they up to? Did they get lots of partnerships/contracts ?
-
Historical inertia. Lots of big DL lab still have all their code in Theano.
- Show replies
New conversation -
-
-
Not just search metrics either.. TensorFlow leads in almost every KPI: GitHub stars, forks, mailing list traffic, etc.
-
I wrote a paper / put together internal presentation about this while at Watson. Lots of data to support TF's lead and that it is expanding
End of conversation
New conversation -
-
-
is there a reason to use tf alone without keras?
-
implement a completely new architecture (new type of layer for example) is easier with TF because of lower level.
- Show 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.