dfdx 0.9.0

Ergonomic auto differentiation in Rust, with pytorch like apis.
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

use super::*;
use crate::prelude::*;
use num_traits::One;
use rand::prelude::Distribution;
use rand_distr::{Standard, StandardNormal};

/// Something that can be created - currently only implemented for tensors with no tapes.
pub trait TensorCreator: Sized + HasDevice {
    /// Create a new tensor with a `Box<Self::Array>`.
    fn new_boxed(data: Box<Self::Array>) -> Self;

    /// Create a new tensor with `Self::Array` on the stack. This just boxes `Self::Array` and calls [TensorCreator::new_boxed].
    fn new(data: Self::Array) -> Self {
        Self::new_boxed(Box::new(data))
    }

    /// Creates a tensor filled with all 0s.
    fn zeros() -> Self {
        Self::new_boxed(Self::Device::zeros())
    }

    /// Creates a tensor filled with all 1s.
    fn ones() -> Self
    where
        Self::Dtype: One,
    {
        Self::new_boxed(Self::Device::filled(&mut |v| *v = One::one()))
    }

    /// Creates a tensor filled with values sampled from [Standard] distribution.
    fn rand<R: rand::Rng>(rng: &mut R) -> Self
    where
        Standard: Distribution<Self::Dtype>,
    {
        Self::new_boxed(Self::Device::filled(&mut |v| *v = Standard.sample(rng)))
    }

    /// Creates a tensor filled with values sampled from [StandardNormal] distribution.
    fn randn<R: rand::Rng>(rng: &mut R) -> Self
    where
        StandardNormal: Distribution<Self::Dtype>,
    {
        Self::new_boxed(Self::Device::filled(&mut |v| {
            *v = StandardNormal.sample(rng)
        }))
    }
}

macro_rules! tensor_impl {
    ($typename:ident, [$($Vs:tt),*]) => {
impl<$(const $Vs: usize, )*> TensorCreator for $typename<$($Vs, )* NoneTape> {
    /// Returns a new object with `data` and a new [UniqueId].
    fn new_boxed(data: Box<Self::Array>) -> Self {
        Self {
            id: unique_id(),
            data: data.into(),
            tape: Default::default(),
        }
    }
}
    };
}

tensor_impl!(Tensor0D, []);
tensor_impl!(Tensor1D, [M]);
tensor_impl!(Tensor2D, [M, N]);
tensor_impl!(Tensor3D, [M, N, O]);
tensor_impl!(Tensor4D, [M, N, O, P]);

#[cfg(test)]
mod tests {
    use std::collections::HashSet;

    use super::*;
    use crate::unique_id::unique_id;
    use rand::thread_rng;

    #[test]
    fn test_id() {
        let mut ids: HashSet<UniqueId> = Default::default();
        ids.insert(unique_id());

        let x = Tensor0D::new(0.0);
        assert!(!ids.contains(&x.id));
        ids.insert(x.id);

        let x = Tensor0D::new(0.0);
        assert!(!ids.contains(&x.id));
        ids.insert(x.id);

        let x = Tensor1D::<5>::zeros();
        assert!(!ids.contains(&x.id));
        ids.insert(x.id);

        let x = Tensor2D::<3, 2>::ones();
        assert!(!ids.contains(&x.id));
        ids.insert(x.id);

        let x = Tensor3D::<4, 3, 2>::zeros();
        assert!(!ids.contains(&x.id));
        ids.insert(x.id);
    }

    #[test]
    fn test_zeros() {
        assert_eq!(Tensor2D::<3, 2>::zeros().data(), &[[0.0; 2]; 3]);
    }

    #[test]
    fn test_ones() {
        assert_eq!(Tensor2D::<3, 2>::ones().data(), &[[1.0; 2]; 3]);
    }

    #[test]
    fn test_new() {
        let t = [[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]];
        assert_eq!(Tensor2D::new(t).data(), &t);
    }

    #[test]
    fn fuzz_test_rand() {
        let mut rng = thread_rng();
        for &v in Tensor1D::<1000>::rand(&mut rng).data() {
            assert!((0.0..1.0).contains(&v));
        }
    }

    #[test]
    fn test_randn() {
        let mut rng = thread_rng();
        let _t = Tensor1D::<1000>::randn(&mut rng);
    }
}