News for hackers. Stuff that matters.
Re:Building a Language and Compiler for Machine Learning by UncleOxidant on Monday December 03, 2018 @07:33PM (#18594689)
That's pretty impressive.

I notice this in the zygote readme:

"The Julia compiler does not yet support all features needed to make Zygote fast, particularly in the presence of control flow."

Any idea of when the Julia compiler will support these features?

Re:Building a Language and Compiler for Machine Learning by viksit on Monday December 03, 2018 @07:54PM (#18594815)
Could you elaborate why this is so impressive? As I understand it, it's because things like TF or PyTorch can only AD specific things as part of a spec vs generic functions like this?
Re:Building a Language and Compiler for Machine Learning by twtw on Monday December 03, 2018 @09:13PM (#18595392)
Yeah, it's impressive because it's differentiating the language - it doesn't require putting 'tf.' in front of everything and hoping tf has the function you want.
Re:Building a Language and Compiler for Machine Learning by FridgeSeal on Tuesday December 04, 2018 @01:52AM (#18596961)
It's impressive and super super cool, because now any function you can write in Julia, the AD can differentiate.

You don't have to hope that your AD/ML package has that function, you can just write it or find it in a package and punch it straight in. That's awesome.

Re:Building a Language and Compiler for Machine Learning by 0-_-0 on Tuesday December 04, 2018 @07:13AM (#18598673)
How is that different from e.g. Tensorflow? You can compose any function you want that has differentiable components, and (re)define the derivatives of elementary or composed functions.
Re:Building a Language and Compiler for Machine Learning by Setepenre on Tuesday December 04, 2018 @07:58PM (#18604910)
No tensorflow does not allow you to do that. Tensorflow differentiate a TF graph. Flux differentiate Julia's IR.

Tensorflow has OP(eration)s that are differentiable that you can compose and that's it. If you want to implement something that is going to be differentiable you need to implement it using Tensorflow OPs or add your own OPs to [tensorflow] with their gradient.

With Flux you can take code written by a random guy on the internet that never thought about using his stuff in ML and Flux will be able to differentiate it anyway.

[1]: https://www.tensorflow.org/guide/extend/op

Re:Building a Language and Compiler for Machine Learning by 0-_-0 on Wednesday December 05, 2018 @04:23AM (#18606805)
Re:Building a Language and Compiler for Machine Learning by FridgeSeal on Wednesday December 05, 2018 @05:28PM (#18613324)
Still a small subset of the language and features

They've also got to manually implement the Python -> autograph translation for a whole variety of language features (so any language features that get added or changed will break autograph until it's updated.

Flux gets this essentially for free, for the entire Julia language, without the need to manually build that language -> tensorflow translation layer. With the added benefit of Julia's non-trivial performance benefits.

Re:Building a Language and Compiler for Machine Learning by byt143 on Wednesday December 05, 2018 @01:24PM (#18610956)
Still a very restricted subset of python, can't autodiff custom types etc