luma-optim 0.1.0

Optimizer for update grads
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
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

use luma_io::lpk::LumaPack;
use luma_tensor::{Device, DynTensor, GradStore, Scalar, Tensor, no_grad};

use super::Optimizer;

#[derive(Clone, Debug)]
pub struct RMSPropConfig {
    pub lr: f64,
    pub alpha: f64, // 平滑常数（通常设为 0.99）
    pub eps: f64,   // 极小值（通常设为 1e-8）
    pub weight_decay: f64,
}

impl Default for RMSPropConfig {
    fn default() -> Self {
        Self { lr: 1e-2, alpha: 0.99, eps: 1e-8, weight_decay: 0.0 }
    }
}

struct RMSPropParam<D: Device> {
    param: Tensor<D>,
    square_avg: Tensor<D>,
}

pub struct RMSProp<D: Device> {
    params: Vec<RMSPropParam<D>>,
    config: RMSPropConfig,
}

impl<D: Device> RMSProp<D> {
    pub fn new(params: impl Into<Vec<Tensor<D>>>, config: RMSPropConfig) -> luma_tensor::Result<Self> {
        let params = params
            .into()
            .into_iter()
            .map(|param| {
                let square_avg = param.zeros_like()?;
                Ok(RMSPropParam { param, square_avg })
            })
            .collect::<luma_tensor::Result<Vec<_>>>()?;
        Ok(Self { params, config })
    }
}

impl<D: Device> Optimizer for RMSProp<D> {
    type Device = D;

    fn set_lr(&mut self, lr: f64) {
        self.config.lr = lr;
    }

    fn get_lr(&self) -> f64 {
        self.config.lr
    }

    fn step(&mut self, grads: &GradStore<Self::Device>) -> luma_tensor::Result<()> {
        no_grad!();
        let lr = self.config.lr;
        let alpha = self.config.alpha;
        let eps = self.config.eps;
        let weight_decay = self.config.weight_decay;

        for RMSPropParam { param, square_avg } in self.params.iter_mut() {
            if let Some(g) = grads.get(&param) {
                let g = g.clone();

                /*
                    1. weight decay

                    w_t = w_{t-1} - lr*g
                    w_t = (1-lr*eta)w_{t-1} - lr*g
                    w_t = w_{t-1} - lr*(g+eta*w_{t+1})

                    g = g+eta*w_{t+1}
                */
                if weight_decay != 0.0 {
                    g.add_(&param.mul_scalar(weight_decay)?)?;
                }

                // 2. update square_avg
                // square_avg = alpha*square_avg + (1-alpha)*g^2
                square_avg.mul_scalar_(alpha)?;
                square_avg.add_(&g.pow(2.0)?.mul_scalar(1. - alpha)?)?;

                // 3. update param
                let denom = square_avg.sqrt()?.add_scalar(eps)?;
                param.sub_(&g.div(&denom)?.mul_scalar(lr)?)?
            }
        }

        Ok(())
    }

    fn state_dict(&self) -> luma_tensor::Result<LumaPack<Self::Device>> {
        let mut pack = LumaPack::new();
        for (i, p) in self.params.iter().enumerate() {
            pack.tensors.insert(format!("{i}.square_avg"), DynTensor::Float(p.square_avg.clone()));
        }
        pack.scalars.insert("lr".into(), Scalar::F64(self.config.lr));
        pack.scalars.insert("alpha".into(), Scalar::F64(self.config.alpha));
        pack.scalars.insert("eps".into(), Scalar::F64(self.config.eps));
        pack.scalars.insert("weight_decay".into(), Scalar::F64(self.config.weight_decay));
        Ok(pack)
    }

    fn load_state_dict(&mut self, pack: &LumaPack<Self::Device>) -> luma_tensor::Result<()> {
        if let Some(v) = pack.scalars.get("lr").and_then(|s| s.to_f64()) {
            self.config.lr = v;
        }
        if let Some(v) = pack.scalars.get("alpha").and_then(|s| s.to_f64()) {
            self.config.alpha = v;
        }
        if let Some(v) = pack.scalars.get("eps").and_then(|s| s.to_f64()) {
            self.config.eps = v;
        }
        if let Some(v) = pack.scalars.get("weight_decay").and_then(|s| s.to_f64()) {
            self.config.weight_decay = v;
        }
        for (i, p) in self.params.iter_mut().enumerate() {
            let key = format!("{i}.square_avg");
            if let Some(dt) = pack.tensors.get(&key) {
                if let Some(src) = dt.as_float() {
                    p.square_avg.copy_(src)?;
                }
            }
        }
        Ok(())
    }
}