/Neural-Layers

Chainer-based framework for neural networks

Primary LanguagePython

Neural-Layers: nula - Chainer-based framework for neural networks

nula is a small package with a set of chainer/nula-functions which can be used to build neural networks in python. These functions are available in nula/functions. nula also provides some prebuild models available in nula/models.

Dependencies:

  • Python 3.4+
  • Chainer 1.3+
  • cython 0.22+
  • numpy 1.9+

nula reimplements some functions which are already available in chainer, but some of these reimplementations can be more efficient and use somewhat more efficient cuda-code (using raw cuda-code instead of elementwise-kernels).