_images/BeyondML_horizontal-color.png

Welcome to BeyondML’s documentation!

BeyondML is a Python package which enables creating sparse multitask artificial neural networks (MANNs) compatible with TensorFlow and PyTorch. This package contains custom layers and utilities to facilitate the training and optimization of models using the Reduction of Sub-Network Neuroplasticity (RSN2) training procedure developed by AI Squared, Inc.

View this Documentation in PDF Format

Installation

This package is available through Pypi and can be installed by running the following command:

pip install beyondml

Alternatively, the latest version of the software can be installed directly from GitHub using the following command:

pip install git+https://github.com/beyond-ml-labs/beyondml

Documentation:

Changelog

  • Version 0.1.0
    • Refactored existing MANN repository to rename to BeyondML

  • Version 0.1.1
    • Added the SparseDense, SparseConv, SparseMultiDense, and SparseMultiConv layers to

      beyondml.tflow.layers, giving users the functionality to utilize sparse tensors during inference

  • Version 0.1.2
    • Added the MaskedMultiHeadAttention, MaskedTransformerEncoderLayer, and MaskedTransformerDecoderLayer layers to beyondml.pt.layers to add pruning to the transformer architecture

    • Added MaskedConv3D, MultiMaskedConv3D, MultiConv3D, MultiMaxPool3D, SparseConv3D, and SparseMultiConv3D layers to beyondml.tflow.layers

    • Added MaskedConv3D, MultiMaskedConv3D, MultiConv3D, MultiMaxPool3D, SparseConv3D, SparseMultiConv3D, and MultiMaxPool2D layers to beyondml.pt.layers

  • Version 0.1.3
    • Added beyondml.pt compatibility with more native PyTorch functionality for using models on different devices and datatypes

    • Added train_model function to beyondml.tflow.utils

    • Added MultitaskNormalization layer to beyondml.tflow.layers and beyondml.pt.layers

  • Version 0.1.4
    • Updated documentation to use Sphinx