Crate tensorrs 

Tensors

Tensors is a lightweight machine learning library in Rust. It provides a simple and efficient way to create and train machine learning models with minimal dependencies.

Dependencies

The library uses the following dependencies:

• rayon - for parallel computations on CPU.
• rand - for random number generation.
• serde - for saving models.
• serde_json - for loading models.

Example

use tensorrs::activation::Function;
use tensorrs::DataType;
use tensorrs::linalg::{Matrix, Vector};
use tensorrs::nn::{Linear, Sequential};
use tensorrs::optim::Adam;
use tensorrs::loss::MSE;
use tensorrs::loss::Loss;

let x = Matrix::from(Vector::range(-1.0, 1.0, 0.125).unwrap());
let y:Matrix<f32> = 8.0 * &x - 10.0;

let layers: Vec<Box< dyn Function<f32>>> = vec![Box::new(Linear::new(1, 1, true))];
let mut optim = Adam::new(0.001, &layers);
let mut model = Sequential::new(layers);
let loss = MSE::new(DataType::f32());

for _ in 0..1000 {
    model.train(x.transpose(), y.transpose(), &mut optim, &loss);
}

Thanks for using Tensors!!!

Modules

activation

Activation Functions

linalg

Linear Algebra

loss

Loss functions

nn

Building Blocks for Neural Networks

optim

Optimization algorithms

utils

Macros

matrix

Matrix definition

tensor

vector

Vector definition

Structs

DataType

Structure to improve readability

Traits

Float

Float type

Num

Numeric type