Dl4J? Chainer? Dynet? Give me a break. There's tons out there now. https://twitter.com/xamat/status/913654270280339456 …
This Tweet is unavailable.
-
Caffe itself is legacy, and Caffe2 has no traction. "Dead" is too strong a qualifier, this is a gradually declining ecosystem.
The biggest mystery in deep learning these days is why PyTorch is getting more traction than the original Chainer. Branding & marketing?...
-
-
I don't think that's it. Many switched from chainer to pytorch
-
What's the difference though?
- Show replies
New conversation -
-
-
Backing. It's hard to get TF users to look at a no-name thing, but if it's supported by Facebook, diff story.
-
(and TF is so terrible to work with, of course you would switch to a sane alternative once you saw it)
End of conversation
New conversation -
-
-
I can give one data point in pytorch mistery. I am looking at it because of Jeremy Howard's fast.ai. Trensetters in the field matter. A lot!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@fchollet what do you think about http://openmined.org (encrypted, decentralized ai) -
To start with its much easier to implement a new CUDA method In PyTorch compared with TF.
End of conversation
New conversation -
-
-
Haven’t used Chainer. My guess (only to some ext): people just don’t wanna buy sth from Japan or whatever they’re not familiar with.
-
A lot of it is just marketing. Of note here though: Most people "using" DL frameworks s don't build from scratch. They clone pre existing.
- Show replies
New conversation -
-
-
No one should be surprised that branding and marketing influences people. I use mxnet/gluon
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.