rotta_rs 0.0.5

a Deep Learning library with rust language
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
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58

use crate::rotta_rs_module::{
    arrayy::broadcasting,
    arrayy::to_shape_arr,
    arrayy::sum_axis_arr,
    arrayy::Arrayy,
    BackwardLabel,
    NodeType,
    Tensor,
};

// broadcasting_tensor
pub fn broadcasting_tensor_non_panic(tensor_arr: &Tensor, broadcast_shape: Vec<usize>) -> Tensor {
    let arr = broadcasting(&tensor_arr.value(), broadcast_shape);

    let tensor = Tensor::from_arrayy(arr);
    tensor.update_parent(vec![tensor_arr.node.clone()]);
    tensor.node.lock().unwrap().label = Some(
        BackwardLabel::Broadcasting(tensor_arr.node.clone(), tensor.value())
    );

    tensor
}

pub fn d_broadcasting_tensor(tensor_arr: &NodeType, broad_arr: Arrayy, grad: Arrayy) {
    let mut tensor_arr = tensor_arr.lock().unwrap();

    if tensor_arr.requires_grad {
        let broadcasted_shape = &broad_arr.shape;

        let pre_shape = tensor_arr.value.shape.clone();
        let mut sum_list = vec![];

        let mut broad_rev = broadcasted_shape.clone();
        broad_rev.reverse();

        let mut pre_shape_rev = pre_shape.clone();
        pre_shape_rev.reverse();

        for d in 0..broad_rev.len() {
            if let Some(pre) = pre_shape_rev.get(d) {
                if &broad_rev[d] != pre {
                    sum_list.push(broad_rev.len() - 1 - d);
                }
            } else {
                sum_list.push(broad_rev.len() - 1 - d);
            }
        }

        let mut sum = grad;

        for sum_d in sum_list {
            sum = sum_axis_arr(&sum, &[sum_d as i32]);
        }

        let d_arr = to_shape_arr(&sum, pre_shape);
        tensor_arr.add_grad(d_arr);
    }
}