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
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137

use crate::optimizers::*;
use crate::cpu_params::*;
use crate::util::*;

use std::collections::HashMap;

pub struct OptimizerSGD {
    pub learn_rate: f32,
    pub momentum: f32,
    pub delta: HashMap<u64, HashMap<i32, VariantParam>>,
}

impl OptimizerSGD {
    pub fn new(learn_rate: f32, momentum: f32) -> Self {
        Self {
            learn_rate,
            momentum,
            delta: HashMap::new(),
        }
    }
}

impl Default for OptimizerSGD {
    fn default() -> Self {
        Self {
            learn_rate: 1e-2,
            momentum: 0.8,
            delta: HashMap::new(),
        }
    }
}

impl OptimizerSGD {
    fn optimize_layer(
        buf: &mut [f32],
        buf_grad: &[f32],
        delta: &mut [f32],
        learn_rate: &f32,
        momentum: &f32,
    ) {
        for ((buf_v, buf_grad_v), delta_v) in buf.iter_mut().zip(buf_grad.iter()).zip(delta.iter_mut()) {
            if *buf_grad_v == 0.0 {
                continue;
            }

            *buf_v += *momentum * *delta_v;
            *delta_v = *learn_rate * buf_grad_v;
            *buf_v += *delta_v;
        }
    }
}

impl Optimizer for OptimizerSGD {
    fn optimize_params(&mut self, lp: &mut CpuParams, opt_prms: TrainableBufsIds) {
        if !self.delta.contains_key(&lp.id) {
            self.delta.insert(lp.id, HashMap::new());
        }

        let delta_val = self.delta.get_mut(&lp.id).unwrap();

        for (buf_id, buf_grad_id) in opt_prms.0.iter().zip(opt_prms.1.iter()) {
            let buf_grad = lp.get_param(*buf_grad_id);

            if !delta_val.contains_key(buf_grad_id) {
                let zeroed_param = VariantParam::copy_zeroed_shape_from(&buf_grad);
                delta_val.insert(*buf_grad_id, zeroed_param);
            }

            let delta_m = delta_val.get_mut(buf_grad_id).unwrap();

            match delta_m {
                VariantParam::Array1(arr1) => {
                    let buf_grad_slice = buf_grad.get_arr_1d();
                    let buf_grad_slice = buf_grad_slice.borrow();
                    let buf_grad_slice = buf_grad_slice.as_slice().unwrap();

                    let buf_slice = lp.get_1d_buf(*buf_id);
                    let mut buf_slice = buf_slice.borrow_mut();
                    let buf_slice = buf_slice.as_slice_mut().unwrap();

                    let delta_slice = arr1.as_slice_mut().unwrap();
                    OptimizerSGD::optimize_layer(
                        buf_slice,
                        buf_grad_slice,
                        delta_slice,
                        &self.learn_rate,
                        &self.momentum
                    );
                }
                VariantParam::Array2(arr2) => {
                    let buf_grad_slice = buf_grad.get_arr_2d();
                    let buf_grad_slice = buf_grad_slice.borrow();
                    let buf_grad_slice = buf_grad_slice.as_slice().unwrap();

                    let buf_slice = lp.get_2d_buf(*buf_id);
                    let mut buf_slice = buf_slice.borrow_mut();
                    let buf_slice = buf_slice.as_slice_mut().unwrap();

                    let delta_slice = arr2.as_slice_mut().unwrap();
                    OptimizerSGD::optimize_layer(
                        buf_slice,
                        buf_grad_slice,
                        delta_slice,
                        &self.learn_rate,
                        &self.momentum
                    );
                }
            }
        }
    }
}

impl WithParams for OptimizerSGD {
    fn cfg(&self) -> HashMap<String, Variant> {
        let mut cfg_params = HashMap::new();

        cfg_params.insert("type".to_string(), Variant::String("sgd".to_string()));
        cfg_params.insert("learning_rate".to_string(), Variant::Float(self.learn_rate));
        cfg_params.insert("momentum".to_string(), Variant::Float(self.momentum));

        cfg_params
    }

    fn set_cfg(&mut self, args: &HashMap<String, Variant>) {
        if args.contains_key("learning_rate") {
            if let Variant::Float(v) = args.get("learning_rate").unwrap() {
                self.learn_rate = *v;
            }
        }

        if args.contains_key("momentum") {
            if let Variant::Float(v) = args.get("momentum").unwrap() {
                self.momentum = *v;
            }
        }
    }
}