Rust implementation for lib like keras

I would like tensorflow team to contribute on my project, update it and add it to its library in rust

It’s super speed than python, suitable for all devices including embedded and low level devices

The example you see is just using Linear activation function, you can also try relu, sigmoid or tanh, the softmax is not implemented yet

Take a look

use rkl::prelude::*;

fn main() {
    let x = array![[1., 2.], [3., 4.], [5., 6.]];
    let y = array![[3.], [7.], [11.]];

    let mut model = Sequential::new(&[
        Dense::new(4, 2, Activation::Linear),
        Dense::new(2, 4, Activation::Linear),
        Dense::new(1, 2, Activation::Linear),
    model.compile(Optimizer::SGD(0.01), Loss::MSE);, y, 1000, true);
    let x_test = array![[2., 3.]];
    let y_test = array![[5.]];
    let eval = model.evaluate(x_test.clone(), y_test);
    println!("\ncost: {}\n", eval);
    let prediction = model.predict(x_test);
    println!("prediction: {}", prediction);"./test.model");
1 Like

There’s a Rust Sig group GitHub - tensorflow/rust: Rust language bindings for TensorFlow
Have you tried it?
maybe you could contribute to it too.

1 Like

Thanks for your reply, i will contribute to this repository

Ahmed, I also have a similar library based on my experience working with Keras, its still getting going and lacks documentation but check it out sometime!