ai-dataloader 0.6.2

Rust implementation to the PyTorch DataLoader
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

/// Implementation for Sequence.
///
/// Currently `BinaryHeap`, `BTreeSet`, `HashSet` and `LinkedList` are not supported because the current implementation
/// require indexing for doing the transpose.
///
use super::super::Collate;
use super::TorchCollate;
use std::collections::VecDeque;

impl<T> Collate<Vec<T>> for TorchCollate
where
    Self: Collate<T>,
    T: Clone,
{
    type Output = Vec<<Self as Collate<T>>::Output>;
    fn collate(&self, batch: Vec<Vec<T>>) -> Self::Output {
        let elem_size = batch
            .first()
            .expect("Batch should contain at least one element")
            .len();

        assert!(
            batch.iter().all(|vec| vec.len() == elem_size),
            "Each Vec in the batch should have equal size"
        );

        let mut collated = Vec::with_capacity(batch.len());

        for i in 0..batch[0].len() {
            let vec: Vec<_> = batch.iter().map(|sample| sample[i].clone()).collect();
            collated.push(self.collate(vec));
        }
        collated
    }
}

impl<T> Collate<VecDeque<T>> for TorchCollate
where
    Self: Collate<T>,
    T: Clone,
{
    type Output = Vec<<Self as Collate<T>>::Output>;
    fn collate(&self, batch: Vec<VecDeque<T>>) -> Self::Output {
        let elem_size = batch
            .first()
            .expect("Batch should contain at least one element")
            .len();

        assert!(
            batch.iter().all(|vec| vec.len() == elem_size),
            "Each Vec in the batch should have equal size"
        );

        let mut collated = Vec::with_capacity(batch.len());

        for i in 0..batch[0].len() {
            let vec: Vec<_> = batch.iter().map(|sample| sample[i].clone()).collect();
            collated.push(self.collate(vec));
        }
        collated
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    use tch::Tensor;

    #[test]
    fn vec_of_vec() {
        assert_eq!(
            TorchCollate.collate(vec![vec![1]]),
            vec![Tensor::from_slice(&[1])]
        );
        assert_eq!(
            TorchCollate.collate(vec![vec![1, 2], vec![3, 4]]),
            vec![Tensor::from_slice(&[1, 3]), Tensor::from_slice(&[2, 4])]
        );
        // different type
        assert_eq!(
            TorchCollate.collate(vec![vec![true, false], vec![true, false]]),
            vec![
                Tensor::from_slice(&[true, true]),
                Tensor::from_slice(&[false, false])
            ]
        );

        assert_eq!(
            TorchCollate.collate(vec![vec![1, 2, 3], vec![4, 5, 6]]),
            vec![
                Tensor::from_slice(&[1, 4]),
                Tensor::from_slice(&[2, 5]),
                Tensor::from_slice(&[3, 6])
            ]
        );
        // batch_size 3
        assert_eq!(
            TorchCollate.collate(vec![vec![1, 2], vec![3, 4], vec![5, 6]]),
            vec![
                Tensor::from_slice(&[1, 3, 5]),
                Tensor::from_slice(&[2, 4, 6])
            ]
        );
        // batch_size 10
        assert_eq!(
            TorchCollate.collate(vec![
                vec![1, 2],
                vec![3, 4],
                vec![5, 6],
                vec![7, 8],
                vec![9, 10],
                vec![11, 12],
                vec![13, 14],
                vec![15, 16],
                vec![17, 18],
                vec![19, 20]
            ]),
            vec![
                Tensor::from_slice(&[1, 3, 5, 7, 9, 11, 13, 15, 17, 19]),
                Tensor::from_slice(&[2, 4, 6, 8, 10, 12, 14, 16, 18, 20])
            ]
        );
    }

    #[test]
    fn specialized() {
        assert_eq!(
            TorchCollate.collate(vec![
                vec![String::from("a"), String::from("b")],
                vec![String::from("c"), String::from("d")]
            ]),
            vec![
                vec![String::from('a'), String::from('c')],
                vec![String::from('b'), String::from('d')],
            ]
        );
    }
}