Re:NGraph: A New Open Source Compiler for Deep Learning Systems
by amelius on Tuesday March 20, 2018 @03:48PM
(#16631894)
That's nice, until someone decides to prune certain functionality from future versions of the library because they are not needed for Deep Learning. Or until someone decides: hey, we can replace this floating point computation by a fixed point computation, because it doesn't matter for any kind of Deep Learning anyway.
[ Parent
| on HN ]
Re:NGraph: A New Open Source Compiler for Deep Learning Systems
by claytonjy on Tuesday March 20, 2018 @04:50PM
(#16632492)
The maintainers of major DL frameworks like TF and PyTorch intentionally target an ever-wider set of use-cases beyond deep learning, and would not chop out non-DL functionality.
There's many cool non-DL projects using these libraries for fast computation like Pyro (PyTorch) & Greta (TensorFlow), and the community is starting to shift away from DL-specific vocabulary towards terms like "differentiable programming". Deep learning is just one use of this broader paradigm.
Re:NGraph: A New Open Source Compiler for Deep Learning Systems
by nl on Tuesday March 20, 2018 @06:06PM
(#16633174)
That's really not how it works at all.
Frameworks are layered, and the acceleratored tensor libraries are useful independently of the neural network architectures built on top of them.
It's basically a numpy/scikit like relationship, except both come from the same project.