metatensor 0.4.2

Self-describing sparse tensor data format for atomistic machine learning and beyond
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

use metatensor::{Labels, TensorMap};

mod utils;
use utils::{example_tensor, example_block, make_fill_value};

#[test]
fn sorted_samples() {
    let keys_to_move = Labels::empty(vec!["key_2"]);
    let tensor = example_tensor().keys_to_samples(&keys_to_move, make_fill_value(0.0), true).unwrap();

    let keys = tensor.keys();
    assert_eq!(keys.count(), 3);
    assert_eq!(keys.names(), ["key_1"]);
    assert_eq!(keys[0], [0]);
    assert_eq!(keys[1], [1]);
    assert_eq!(keys[2], [2]);

    // The first two blocks are not modified
    let block_1 = tensor.block_by_id(0);
    let values = block_1.values().to_ndarray_lock::<f64>().read().unwrap();
    assert_eq!(*values, ndarray::Array::from_elem(vec![3, 1, 1], 1.0));

    let samples = block_1.samples();
    assert_eq!(samples.count(), 3);
    assert_eq!(samples[0], [0, 0]);
    assert_eq!(samples[1], [2, 0]);
    assert_eq!(samples[2], [4, 0]);

    let block_2 = tensor.block_by_id(1);
    let values = block_2.values().to_ndarray_lock::<f64>().read().unwrap();
    assert_eq!(*values, ndarray::Array::from_elem(vec![3, 1, 3], 2.0));

    let samples = block_2.samples();
    assert_eq!(samples.count(), 3);
    assert_eq!(samples[0], [0, 0]);
    assert_eq!(samples[1], [1, 0]);
    assert_eq!(samples[2], [3, 0]);

    // The new third block contains the old third and fourth blocks merged
    let block_3 = tensor.block_by_id(2);
    let samples = block_3.samples();
    assert_eq!(samples.names(), ["samples", "key_2"]);
    assert_eq!(samples.count(), 8);
    assert_eq!(samples[0], [0, 2]);
    assert_eq!(samples[1], [0, 3]);
    assert_eq!(samples[2], [1, 3]);
    assert_eq!(samples[3], [2, 3]);
    assert_eq!(samples[4], [3, 2]);
    assert_eq!(samples[5], [5, 3]);
    assert_eq!(samples[6], [6, 2]);
    assert_eq!(samples[7], [8, 2]);

    assert_eq!(block_3.components().len(), 1);
    let component = &block_3.components()[0];

    assert_eq!(component.names(), ["components"]);
    assert_eq!(component.count(), 3);
    assert_eq!(component[0], [0]);
    assert_eq!(component[1], [1]);
    assert_eq!(component[2], [2]);

    let properties = block_3.properties();
    assert_eq!(properties.names(), ["properties"]);
    assert_eq!(properties.count(), 1);
    assert_eq!(properties[0], [0]);

    let expected = ndarray::Array::from_shape_vec(vec![8, 3, 1], vec![
        3.0, 3.0, 3.0,
        4.0, 4.0, 4.0,
        4.0, 4.0, 4.0,
        4.0, 4.0, 4.0,
        3.0, 3.0, 3.0,
        4.0, 4.0, 4.0,
        3.0, 3.0, 3.0,
        3.0, 3.0, 3.0,
    ]).unwrap();
    let values = block_3.values().to_ndarray_lock::<f64>().read().unwrap();
    assert_eq!(*values, expected);

    let gradient_3 = block_3.gradient("parameter").unwrap();
    let samples = gradient_3.samples();
    assert_eq!(samples.names(), ["sample", "parameter"]);
    assert_eq!(samples.count(), 3);
    assert_eq!(samples[0], [1, 1]);
    assert_eq!(samples[1], [4, -2]);
    assert_eq!(samples[2], [5, 3]);

    let expected = ndarray::Array::from_shape_vec(vec![3, 3, 1], vec![
        14.0, 14.0, 14.0,
        13.0, 13.0, 13.0,
        14.0, 14.0, 14.0,
    ]).unwrap();
    let values = gradient_3.values().to_ndarray_lock::<f64>().read().unwrap();
    assert_eq!(*values, expected);
}

#[test]
fn unsorted_samples() {
    let keys_to_move = Labels::empty(vec!["key_2"]);
    let tensor = example_tensor().keys_to_samples(&keys_to_move, make_fill_value(0.0), false).unwrap();

    let block_3 = tensor.block_by_id(2);
    let samples = block_3.samples();
    assert_eq!(samples.names(), ["samples", "key_2"]);
    assert_eq!(samples.count(), 8);
    assert_eq!(samples[0], [0, 2]);
    assert_eq!(samples[1], [3, 2]);
    assert_eq!(samples[2], [6, 2]);
    assert_eq!(samples[3], [8, 2]);
    assert_eq!(samples[4], [0, 3]);
    assert_eq!(samples[5], [1, 3]);
    assert_eq!(samples[6], [2, 3]);
    assert_eq!(samples[7], [5, 3]);
}

#[test]
fn user_provided_entries() {
    let keys_to_move = Labels::new(["key_2"], [[3]]);
    let result = example_tensor().keys_to_samples(&keys_to_move, make_fill_value(0.0), false);

    assert_eq!(
        result.unwrap_err().message,
        "invalid parameter: user provided values for the keys to move is \
        not yet implemented, `keys_to_move` should not contain any entry \
        when calling keys_to_samples"
    );
}

#[test]
#[allow(clippy::vec_init_then_push)]
fn empty_samples() {
    let mut blocks = Vec::new();
    blocks.push(example_block(
        /* samples          */ vec![[0], [2], [4]],
        /* components       */ vec![[0]],
        /* properties       */ vec![[0], [1], [2], [3]],
        /* gradient_samples */ vec![],
        /* values           */ 1.0,
        /* gradient_values  */ 11.0,
    ));

    blocks.push(example_block(
        /* samples          */ vec![],
        /* components       */ vec![[0]],
        /* properties       */ vec![[0], [1], [2], [3]],
        /* gradient_samples */ vec![],
        /* values           */ 0.0,
        /* gradient_values  */ 0.0,
    ));
    let keys = Labels::new(["key_1", "key_2"], [[0, 0], [1, 0]]);

    let tensor = TensorMap::new(keys, blocks).unwrap();

    let keys_to_move = Labels::empty(vec!["key_1"]);
    let tensor = tensor.keys_to_samples(&keys_to_move, make_fill_value(0.0), true).unwrap();

    assert_eq!(
        tensor.block_by_id(0).samples(),
        Labels::new(["samples", "key_1"], [[0, 0], [2, 0], [4, 0]])
    );
}