tensorly/tensorly

TensorLy: Tensor Learning in Python.

What it solves

TensorLy simplifies tensor learning by providing a unified Python library for tensor decomposition, tensor algebra, and tensor learning. It removes the need to rewrite complex tensor operations for different deep learning frameworks.

How it works

It operates as a high-level wrapper with a flexible backend system. Users can perform computations using NumPy by default, or switch to PyTorch, JAX, TensorFlow, CuPy, or Paddle to leverage GPU acceleration and scale their methods.

Who it’s for

It is designed for researchers and developers working with multi-dimensional arrays (tensors) who need to perform advanced decompositions or algebraic operations across various computational backends.

Highlights

  • Multi-backend support: Seamlessly switch between NumPy, PyTorch, TensorFlow, JAX, CuPy, and Paddle.
  • Tensor Decomposition: Built-in support for methods like Tucker decomposition and CP tensors.
  • Hardware Acceleration: Ability to run computations on both CPU and GPU depending on the chosen backend.
  • Tensor Algebra: Tools for creating, unfolding, and folding tensors.

Related

  • Project
    elixir-nx/axon

    A neural network framework for Elixir that provides high-level APIs for model creation and training, powered by Nx for numerical acceleration.

  • Project
    Minglink/dsh-infinite-gen-4

    A cybersecurity red-teaming and robustness benchmarking suite for DeepSeek-V4 series models, using dual-layer prompt injection to test model defenses.

  • Project
    EnzymeAD/Enzyme

    A high-performance automatic differentiation plugin for LLVM and MLIR that synthesizes fast gradients for optimized code, including GPU kernels and parallel programs.

  • Project
    mlrun/mlrun

    An open-source AI orchestration platform that automates the delivery of ML and generative AI pipelines, from data management and training to serverless deployment and monitoring.