tensorly/tensorly

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1691 stars · 300 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

TensorLy is a Python library for tensor learning, providing tensor decomposition, tensor algebra, and tensor regression with a simple API. Its backend system lets the same code run on NumPy, PyTorch, JAX, TensorFlow, CuPy, or Paddle, on CPU or GPU.

Use cases

• perform tensor decomposition like CP or Tucker in Python
• run tensor algebra operations on GPU with PyTorch or JAX backends
• fit tensor regression models
• switch tensor computation backends without rewriting code
• learn tensor methods for machine learning research
• factorize multi-dimensional arrays

When to choose

• you need tensor decomposition or factorization in Python with a clean high-level API
• you want backend-agnostic tensor code that runs on NumPy, PyTorch, JAX, TensorFlow, or CuPy
• you are doing research or teaching involving tensor methods

When to avoid

• you only need standard 2D matrix operations rather than multi-dimensional tensors
• you need a deep learning framework's full ecosystem rather than a focused tensor library
• you require a permissively documented commercial support arrangement

Facets

library · maturity stable

machine-learning math data-science machine-learning data-science python cross-platform tensor-decomposition tensor-algebra tensor-learning numpy pytorch jax tensorflow cupy backend-agnostic algorithms gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="tensorly/tensorly")