Crate tenferro 

tenferro: traced tensor computation with StableHLO-style IR.

This crate provides a tracing-based tensor computation framework where operations are recorded into a StableHLO-compatible intermediate representation, then compiled and executed on a backend (e.g., CPU).

Examples

use tenferro::{CpuBackend, Engine, TracedTensor};

let mut engine = Engine::new(CpuBackend::default());
// ... build and execute traced computations

Re-exports

pub use engine::Engine;

pub use sym_dim::SymDim;

pub use traced::TracedTensor;

Modules

compiler

eager_einsum

Eager einsum helpers exposed through the tenferro facade.

eager_exec

einsum

N-ary einsum with configurable contraction strategy.

engine

error

Error types for the tenferro crate.

exec

segment

ExecProgram segmentation and segment-based dispatch.

shape_infer

Output-shape and output-dtype inference for StdTensorOp.

sym_dim

traced

Structs

CpuBackend

CPU execution backend.

DotGeneralConfig

DotGeneral dimension configuration.

EagerContext

Shared eager execution context for tensors on a backend.

EagerTensor

Eager tensor with reverse-mode autodiff over concrete tensor values.

GatherConfig

StableHLO gather dimension configuration.

PadConfig

StableHLO pad configuration.

ScatterConfig

StableHLO scatter dimension configuration.

SliceConfig

Slice configuration.

TypedTensor

Contiguous column-major typed tensor storage.

Enums

DType

Runtime scalar dtype tag.

Tensor

Dynamic tensor enum over the supported scalar types.

Traits

TensorBackend

Standard runtime backend over dynamic Tensor values.

TensorScalar

Sealed trait for scalar types that can be stored in a Tensor.

Functions

cholesky

Cholesky factorization.

convert

Convert a traced tensor to a different dtype.

det

Determinant from slogdet.

eig

Non-symmetric eigendecomposition.

eigh

Hermitian eigenvalue decomposition with a default numerical epsilon.

eigh_with_eps

Hermitian eigenvalue decomposition with an explicit numerical epsilon.

eigvals

General eigenvalues only.

eigvalsh

Hermitian eigenvalues only.

inv

Matrix inverse via solve(a, eye).

lu

LU decomposition with partial pivoting.

matmul

Matrix multiplication helper for rank-2 traced tensors.

norm

Vector or matrix norm.

pinv

Moore-Penrose pseudoinverse via the SVD.

pinv_with_rtol

Moore-Penrose pseudoinverse via the SVD with an explicit relative cutoff.

pow

Elementwise power helper with NumPy-style broadcasting.

qr

QR decomposition.

slogdet

Sign and log-absolute-determinant from the LU factorization.

solve

Solve a linear system using LU decomposition and triangular solves.

svd

Singular value decomposition with a default numerical epsilon.

svd_with_eps

Singular value decomposition with an explicit numerical epsilon.

triangular_solve

Solve a triangular linear system.