tch 0.0.4

Rust wrappers for the PyTorch C++ api (libtorch).
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
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179

/// Optimizers to be used for gradient-descent based training.
use super::c_optimizer::COptimizer;
use super::var_store::VarStore;
use crate::{Scalar, Tensor};

pub struct Optimizer {
    opt: COptimizer,
    trainable_variables: Vec<Tensor>,
}

/// Parameters for the SGD optimizer.
pub struct Sgd {
    pub momentum: f64,
    pub dampening: f64,
    pub wd: f64,
    pub nesterov: bool,
}

impl Default for Sgd {
    fn default() -> Self {
        Sgd {
            momentum: 0.,
            dampening: 0.,
            wd: 0.,
            nesterov: false,
        }
    }
}

/// Parameters for the Adam optimizer.
pub struct Adam {
    pub beta1: f64,
    pub beta2: f64,
    pub wd: f64,
}

impl Default for Adam {
    fn default() -> Self {
        Adam {
            beta1: 0.9,
            beta2: 0.999,
            wd: 0.,
        }
    }
}

/// Parameters for the RmsProp optimizer.
pub struct RmsProp {
    pub alpha: f64,
    pub eps: f64,
    pub wd: f64,
    pub momentum: f64,
    pub centered: bool,
}

impl Default for RmsProp {
    fn default() -> Self {
        RmsProp {
            alpha: 0.99,
            eps: 1e-8,
            wd: 0.,
            momentum: 0.,
            centered: false,
        }
    }
}

impl Optimizer {
    pub fn sgd(vs: &VarStore, lr: f64, s: Sgd) -> Optimizer {
        let mut opt = COptimizer::sgd(lr, s.momentum, s.dampening, s.wd, s.nesterov);
        let trainable_variables = vs.trainable_variables();
        opt.add_parameters(&trainable_variables);
        Optimizer {
            opt,
            trainable_variables,
        }
    }

    pub fn adam(vs: &VarStore, lr: f64, a: Adam) -> Optimizer {
        let mut opt = COptimizer::adam(lr, a.beta1, a.beta2, a.wd);
        let trainable_variables = vs.trainable_variables();
        opt.add_parameters(&trainable_variables);
        Optimizer {
            opt,
            trainable_variables,
        }
    }

    pub fn rms_prop(vs: &VarStore, lr: f64, r: RmsProp) -> Optimizer {
        let mut opt = COptimizer::rms_prop(lr, r.alpha, r.eps, r.wd, r.momentum, r.centered);
        let trainable_variables = vs.trainable_variables();
        opt.add_parameters(&trainable_variables);
        Optimizer {
            opt,
            trainable_variables,
        }
    }

    pub fn zero_grad(&self) {
        self.opt.zero_grad()
    }

    /// Clips gradient value at some specified maximum value.
    pub fn clip_grad_value(&self, max: f64) {
        for tensor in self.trainable_variables.iter() {
            let _t = tensor
                .grad()
                .clamp_(&Scalar::float(-max), &Scalar::float(max));
        }
    }

    pub fn step(&self) {
        self.opt.step()
    }

    pub fn backward_step(&self, loss: &Tensor) {
        self.zero_grad();
        loss.backward();
        self.step();
    }

    pub fn backward_step_clip(&self, loss: &Tensor, max: f64) {
        self.zero_grad();
        loss.backward();
        self.clip_grad_value(max);
        self.step();
    }

    /// Sets the optimizer learning rate.
    pub fn set_lr(&mut self, lr: f64) {
        self.opt.set_learning_rate(lr)
    }

    /// Sets the optimizer momentum.
    pub fn set_momentum(&mut self, m: f64) {
        self.opt.set_momentum(m)
    }
}

pub fn sgd(
    vs: &VarStore,
    lr: f64,
    momentum: f64,
    dampening: f64,
    wd: f64,
    nesterov: bool,
) -> Optimizer {
    let sgd = Sgd {
        momentum,
        dampening,
        wd,
        nesterov,
    };
    Optimizer::sgd(vs, lr, sgd)
}

pub fn adam(vs: &VarStore, lr: f64, beta1: f64, beta2: f64, wd: f64) -> Optimizer {
    let adam = Adam { beta1, beta2, wd };
    Optimizer::adam(vs, lr, adam)
}

pub fn rms_prop(
    vs: &VarStore,
    lr: f64,
    alpha: f64,
    eps: f64,
    wd: f64,
    momentum: f64,
    centered: bool,
) -> Optimizer {
    let rmsprop = RmsProp {
        alpha,
        eps,
        wd,
        momentum,
        centered,
    };
    Optimizer::rms_prop(vs, lr, rmsprop)
}