PyTorch – Tensors and Dynamic neural networks in Python

Re:PyTorch – Tensors and Dynamic neural networks in Python by jjawssd on Wednesday January 18, 2017 @02:05PM (#13429269)

A very large portion of performance problems can be mitigated with the use of cython and the new asyncio stuff.

asyncio success story: https://magic.io/blog/asyncpg-1m-rows-from-postgres-to-pytho...

cython: http://scikit-learn.org/stable/developers/performance.html

[ Parent | on HN ]

Re:PyTorch – Tensors and Dynamic neural networks in Python by vegabook on Wednesday January 18, 2017 @02:16PM (#13429373)

Luajit is at least 10x faster than python and easily obviates the need to mess around with cython. That's an easy win for Lua. Let's be honest: Torch has decided that if you cannot beat them, join them. It is about network effects. Not about Python better than Lua intrinsically.

[ Parent | on HN ]

Re:PyTorch – Tensors and Dynamic neural networks in Python by baq on Wednesday January 18, 2017 @05:02PM (#13430741)

but why do you care if luajit is faster than python if everything that matters is computed on the GPU anyways?

[ Parent | on HN ]

Re:PyTorch – Tensors and Dynamic neural networks in Python by jjawssd on Wednesday January 18, 2017 @02:22PM (#13429432)

Can't argue with that

[ Parent | on HN ]

Re:PyTorch – Tensors and Dynamic neural networks in Python by inlineint on Thursday January 19, 2017 @06:33AM (#13434951)

An alternative to Cython is Numba [1], which speeds up some cycles in pure Python by just adding a single decorator.

[1] http://numba.pydata.org/

[ Parent | on HN ]