einops 0.2.1

Simplistic API for deep learning tensor operations
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

use tch::nn::Module;
use tch::Tensor;

use crate::backend::{Backend, Operation, RearrangeFn, ReduceFn, RepeatFn};
use crate::{Rearrange, Reduce, Repeat};

impl Backend for Tensor {
    fn shape(&self) -> Vec<usize> {
        self.size().iter().map(|&x| x as usize).collect::<Vec<_>>()
    }

    fn reshape(&self, shape: &[usize]) -> Self {
        self.reshape(&shape.iter().map(|&x| x as i64).collect::<Vec<_>>())
    }

    fn transpose(&self, axes: &[usize]) -> Self {
        self.permute(&axes.iter().map(|&x| x as i64).collect::<Vec<_>>())
    }

    fn reduce_axes(&self, operation: Operation, axes: &[usize]) -> Self {
        let mut output = self.shallow_clone();

        let mut axes = axes.to_vec();
        axes.sort_unstable();

        for &axis in axes.iter().rev() {
            output = match operation {
                Operation::Min => output.min_dim(axis as i64, false).0,
                Operation::Max => output.max_dim(axis as i64, false).0,
                Operation::Sum => output.sum_dim_intlist(&[axis as i64], false, output.kind()),
                Operation::Mean => output.mean_dim(&[axis as i64], false, output.kind()),
                Operation::Prod => output.prod_dim_int(axis as i64, false, output.kind()),
            };
        }

        output
    }

    fn add_axes(&self, naxes: usize, pos2len: &[(usize, usize)]) -> Self {
        let mut output = self.shallow_clone();

        let mut repeats = vec![1; naxes];

        for &(axis_pos, axis_len) in pos2len {
            output = output.unsqueeze(axis_pos as i64);
            repeats[axis_pos] = axis_len as i64;
        }

        output.repeat(&repeats)
    }
}

impl Module for Rearrange {
    fn forward(&self, xs: &Tensor) -> Tensor {
        self.recipe.apply(xs).unwrap()
    }
}

impl Module for Reduce {
    fn forward(&self, xs: &Tensor) -> Tensor {
        self.recipe.apply(xs).unwrap()
    }
}

impl Module for Repeat {
    fn forward(&self, xs: &Tensor) -> Tensor {
        self.recipe.apply(xs).unwrap()
    }
}

impl RearrangeFn for Tensor {}

impl ReduceFn for Tensor {}

impl RepeatFn for Tensor {}

#[cfg(test)]
mod tests {
    // TODO Add more tests
    use super::*;
    use crate::EinopsError;
    use tch::{Device, Kind, Tensor};

    #[test]
    fn tch_layers() -> Result<(), EinopsError> {
        let input = Tensor::randn(&[10, 3, 32, 32], (Kind::Float, Device::Cpu));

        let output1 = input.reduce("b c (h 2) (w 2) -> b c h w", Operation::Max)?;
        let output2 = input.max_pool2d_default(2);

        assert_eq!(output1, output2);

        Ok(())
    }

    #[test]
    fn tch_reduce() {
        let tests = vec![(
            Tensor::of_slice(&[
                0.66984287, 0.52894678, 0.85415958, 0.17721198, 0.81804799, 0.80991797, 0.64868822,
                0.96697902, 0.08047191, 0.46024353, 0.21955009, 0.31731976, 0.05446258, 0.39454557,
                0.40949016, 0.21366165, 0.2357463, 0.93699481, 0.64522596, 0.4383618, 0.54871827,
                0.87823442, 0.01261184, 0.90636503,
            ])
            .reshape(&[4, 2, 3]),
            Operation::Min,
            &[0],
            Tensor::of_slice(&[
                0.05446258, 0.39454557, 0.08047191, 0.17721198, 0.01261184, 0.31731976,
            ])
            .reshape(&[2, 3]),
        )];

        for (tensor, operation, axes, expected) in tests {
            assert_eq!(tensor.reduce_axes(operation, axes), expected);
        }
    }

    #[test]
    fn tch_transpose() {
        let tests = vec![(
            Tensor::arange(2 * 3 * 4 * 5, (Kind::Float, Device::Cpu)).reshape(&[2, 3, 4, 5]),
            &[3, 0, 2, 1],
            Tensor::of_slice(&[
                0, 20, 40, 5, 25, 45, 10, 30, 50, 15, 35, 55, 60, 80, 100, 65, 85, 105, 70, 90,
                110, 75, 95, 115, 1, 21, 41, 6, 26, 46, 11, 31, 51, 16, 36, 56, 61, 81, 101, 66,
                86, 106, 71, 91, 111, 76, 96, 116, 2, 22, 42, 7, 27, 47, 12, 32, 52, 17, 37, 57,
                62, 82, 102, 67, 87, 107, 72, 92, 112, 77, 97, 117, 3, 23, 43, 8, 28, 48, 13, 33,
                53, 18, 38, 58, 63, 83, 103, 68, 88, 108, 73, 93, 113, 78, 98, 118, 4, 24, 44, 9,
                29, 49, 14, 34, 54, 19, 39, 59, 64, 84, 104, 69, 89, 109, 74, 94, 114, 79, 99, 119,
            ])
            .reshape(&[5, 2, 4, 3]),
        )];

        for (tensor, axes, expected) in tests {
            assert_eq!(Backend::transpose(&tensor, axes), expected);
        }
    }

    #[test]
    fn tch_add_axes() {
        let tests = vec![(
            Tensor::arange(1 * 2 * 3, (Kind::Float, Device::Cpu)).reshape(&[1, 2, 3]),
            5,
            &[(0, 5), (3, 3)],
            Tensor::of_slice(&[
                0, 1, 2, 0, 1, 2, 0, 1, 2, 3, 4, 5, 3, 4, 5, 3, 4, 5, 0, 1, 2, 0, 1, 2, 0, 1, 2, 3,
                4, 5, 3, 4, 5, 3, 4, 5, 0, 1, 2, 0, 1, 2, 0, 1, 2, 3, 4, 5, 3, 4, 5, 3, 4, 5, 0, 1,
                2, 0, 1, 2, 0, 1, 2, 3, 4, 5, 3, 4, 5, 3, 4, 5, 0, 1, 2, 0, 1, 2, 0, 1, 2, 3, 4, 5,
                3, 4, 5, 3, 4, 5,
            ])
            .reshape(&[5, 1, 2, 3, 3]),
        )];

        for (tensor, naxes, pos2len, expected) in tests {
            assert_eq!(tensor.add_axes(naxes, pos2len), expected);
        }
    }
}