tensor-trait

A small Rust library providing unified traits for tensor operations in machine learning models.

Features

• Standardized tensor trait interfaces
• Adapters for popular tensor crates
• Custom error handling for tensor ops

Usage

Add to your Cargo.toml:

tensor-trait = "0.1"

Implementing Tensor Traits

use tensor_trait::tensor_trait::Tensor;

struct MyTensor {
    data: Vec<f32>,
    shape: Vec<usize>,
}

impl Tensor for MyTensor {
    type Elem = f32;
    fn ndim(&self) -> usize {
        self.shape.len()
    }
    fn shape(&self) -> &[usize] {
        &self.shape
    }
    // Other trait methods...
}

Using Adapters

use tensor_trait::adapters::ndarray_adapter::NdArrayTensor;
use ndarray::Array2;

let arr = Array2::<f32>::zeros((2, 3));
let tensor = NdArrayTensor::from(arr);
assert_eq!(tensor.ndim(), 2);

Error Handling Example

use tensor_trait::error::TensorOpError;

fn do_tensor_op() -> Result<(), TensorOpError> {
    // ...something that can fail
    Err(TensorOpError::ShapeMismatch)
}

Documentation

Detailed trait docs and adapters can be found in the crate docs.