• V0.9.5 (latest)

    2017.11.08 - macOS support

    The distinctive features of Neural Network Libraries V0.9.5 are as following:

     * It can be installed with pip on macOS (CUDA extension is not available yet).
     * C++ stand-alone mode is compatible with various OS (macOS, Windows).
     * CUDA is available with C++ standalone.
     * Linear quantization, quadratic quantization, training functionalities have been added.
     * New functions have been implemented including BatchMatmul, MatrixDiag, SELU, Swish, etc.

    Install the latest nnabla by:

    pip install -U nnabla
    pip install -U nnabla_ext_cuda  # For CUDA users

    Additional setup may be required depending on your OS or environment. Please check Python Package Installation Guide for details.

    To use C++ inference feature, follow the demonstration on MNIST inference in C++.

    For distributed training, you need to build a binary from source. See the guide for building multi-GPU training package.

  • V0.9.4

    2017.08.22 - C++ inference

  • V0.9.3

    2017.08.04 - Distributed training and faster dilated-Convolution with cuDNN

  • V0.9.2

    2017.07.21 - Python3 Support

  • V0.9.1

    2017.06.27 - The First Release of Neural Network Libraries