Tensor0

Tensor0 is an experimental symmetric tensor library with Rust-backed structural metadata and JAX-backed tensor storage; it is not yet a production-ready tensor-network framework, and the API may change while the core representation is being developed.

Features

• Built-in sector families including U1Irrep, SU2Irrep, cyclic ZNIrrep aliases, fermion parity, and selected product-sector aliases.
• space(...) and hom(codomain, domain) constructors for symmetric linear map spaces.
• TensorMap storage backed by JAX arrays.
• Block access, blockwise composition, compact SVD, and truncation helpers.
• permute, braid, transpose, and repartition transform helpers.
• to_dense(...) and from_dense(...) for small correctness checks.
• JAX jit and grad support through pytree registration.

Installation

Tensor0 is currently installed from source. It requires Python 3.11 or newer and a Rust toolchain.

The recommended setup is to install Tensor0 into a virtual environment:

git clone https://github.com/yitan1/Tensor0.git
cd Tensor0

python -m venv .venv
source .venv/bin/activate
python -m pip install .

If you manage your own Python environment, python -m pip install . can also be run without creating a virtual environment.

Local Development

From a checkout:

uv sync --group dev
uv run maturin develop

Run the Python and Rust tests:

uv run pytest tests/python -q
cargo test

Run the public example:

uv run python examples/basic_usage.py

Minimal Example

import jax.numpy as jnp

from tensor0 import TensorMap, U1Irrep, get_degeneracystructure, hom, space

v = space(U1Irrep, {0: 2, 1: 3})
h = hom((v,), (v,))

total_dim = get_degeneracystructure(h).total_dim
tensor = TensorMap(h, jnp.arange(total_dim, dtype=jnp.float32))

for sector, block in tensor.blocks():
    print(sector, block.shape)

See docs/usage.md and examples/basic_usage.py for more examples.

Acknowledgments

Tensor0's main design and architecture are based on TensorKit.jl. Tensor0 focuses on bringing these ideas to Python with JAX-backed tensor storage and automatic differentiation support.