tenferro

tenferro is a dense tensor computation library for Rust users who want PyTorch- and JAX-style tensor workflows with explicit Rust ownership and backend control.

It supports:

• eager tensor operations with Tensor, TypedTensor, and a backend,
• eager scalar-loss reverse-mode AD with EagerTensor,
• lazy traced execution with TracedTensor and Engine,
• transform AD with grad, vjp, jvp, and HVP composition,
• einsum and linear algebra,
• CUDA execution through the feature-gated CubeCL backend.

Start Here

• Getting Started
• Choosing an API
• Custom Tensor Operations
• Devices and GPU
• API Reference

First CPU Example

use tenferro::{CpuBackend, Tensor};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut backend = CpuBackend::new();

    let a = Tensor::from_vec(vec![2, 2], vec![1.0_f64, 3.0, 2.0, 4.0]);
    let b = Tensor::from_vec(vec![2, 2], vec![5.0_f64, 7.0, 6.0, 8.0]);

    let c = a.matmul(&b, &mut backend)?;

    assert_eq!(c.shape(), &[2, 2]);
    assert_eq!(c.as_slice::<f64>().unwrap(), &[19.0, 43.0, 22.0, 50.0]);

    Ok(())
}

Mental Model

Workflow	Use
Direct CPU computation	Tensor or TypedTensor with CpuBackend
Scalar-loss eager AD	EagerTensor with EagerContext
Transform AD and graph optimization	TracedTensor with Engine
CUDA execution	tenferro::cuda::CudaBackend with the cuda feature and explicit upload/download