esp-idf-unda 0.1.2

Embedded neural network forward feeding for precompiled models created by the unda crate
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43

use crate::network::{layer::{layers::Layer, dense::Dense}, input::Input, activations::Activations};

pub struct SerializedLayer {
    pub name: char,
    pub rows: usize,
    pub cols: usize,
    pub weights: String,
    pub bias: String
}

impl SerializedLayer {
    pub fn from(&self, activation: Activations) -> Box<dyn Layer> {
        match self.name {
            'D' => {
                let weights_f32: Vec<f32> = self.weights.split(" ").into_iter().map(|val| val.parse().unwrap()).collect();
                let bias_f32: Vec<f32> = self.bias.split(" ").into_iter().map(|val| val.parse().unwrap()).collect();
                let dense_layer: Dense = Dense::new_ser(self.rows, self.cols, weights_f32, bias_f32, activation);
                return Box::new(dense_layer)
            },
            _ => panic!("Not a supported type"),
        };
    }
    fn flatten_string(data: &Vec<Vec<f32>>) -> String {
        data.to_param()
            .into_iter()
            .map(|d| d.to_string() + " ")
            .collect::<String>().trim_end().to_string()
    }
    pub fn to_string(&self) -> String {
        format!("{}|{}|{}|{}|{}", self.name, self.rows, self.cols, self.weights, self.bias)
    }
    pub fn from_string(data: String) -> Self {
        let mut parse_res = data.split("|");
        let name: char = parse_res.next().unwrap().chars().next().unwrap();
        let rows: usize = str::parse(parse_res.next().unwrap()).unwrap();
        let cols: usize = str::parse(parse_res.next().unwrap()).unwrap();
        let weights = parse_res.next().unwrap().to_string();
        let bias = parse_res.next().unwrap().to_string();

        Self { name, rows, cols, weights, bias }
    }
}