trueno 0.17.4

High-performance SIMD compute library with GPU support for matrix 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

use super::super::super::super::*;
use proptest::prelude::*;

// ========================================================================
// Property tests for leaky_relu() - Leaky Rectified Linear Unit
// ========================================================================

proptest! {
    /// Property test: leaky_relu() preserves positive values exactly
    #[test]
    fn test_leaky_relu_preserves_positive_property(
        a in prop::collection::vec(-100.0f32..100.0, 1..100),
        alpha in 0.0f32..1.0
    ) {
        let va = Vector::from_slice(&a);
        let result = va.leaky_relu(alpha).unwrap();

        for (i, &val) in a.iter().enumerate() {
            if val > 0.0 {
                prop_assert!(
                    (result.data[i] - val).abs() < 1e-6,
                    "Positive value {} should be preserved, got {}",
                    val, result.data[i]
                );
            }
        }
    }
}

proptest! {
    /// Property test: leaky_relu() scales negative values by alpha
    #[test]
    fn test_leaky_relu_scales_negative_property(
        a in prop::collection::vec(-100.0f32..100.0, 1..100),
        alpha in 0.01f32..0.5 // Use smaller range to avoid precision issues
    ) {
        let va = Vector::from_slice(&a);
        let result = va.leaky_relu(alpha).unwrap();

        for (i, &val) in a.iter().enumerate() {
            if val < 0.0 {
                let expected = alpha * val;
                prop_assert!(
                    (result.data[i] - expected).abs() < 1e-4,
                    "Negative value {} should be scaled by {}: expected {}, got {}",
                    val, alpha, expected, result.data[i]
                );
            }
        }
    }
}

proptest! {
    /// Property test: leaky_relu() is monotonically increasing
    /// If x < y, then leaky_relu(x) < leaky_relu(y)
    #[test]
    fn test_leaky_relu_monotonic_property(
        a in prop::collection::vec(-50.0f32..50.0, 2..100),
        alpha_raw in 1u32..50
    ) {
        let alpha = alpha_raw as f32 / 100.0; // 0.01..0.50, avoids proptest float sampler panic
        let va = Vector::from_slice(&a);
        let result = va.leaky_relu(alpha).unwrap();

        // Check all pairs for monotonicity
        for i in 0..a.len() {
            for j in 0..a.len() {
                if a[i] < a[j] {
                    prop_assert!(
                        result.data[i] < result.data[j] + 1e-5,
                        "Monotonicity violated: {} < {} but leaky_relu({}) = {} >= leaky_relu({}) = {}",
                        a[i], a[j], a[i], result.data[i], a[j], result.data[j]
                    );
                }
            }
        }
    }
}

// ========================================================================
// Property tests for elu() - Exponential Linear Unit
// ========================================================================

proptest! {
    /// Property test: elu() preserves positive values exactly
    #[test]
    fn test_elu_preserves_positive_property(
        a in prop::collection::vec(-100.0f32..100.0, 1..100),
        alpha in 0.1f32..5.0
    ) {
        let va = Vector::from_slice(&a);
        let result = va.elu(alpha).unwrap();

        for (i, &val) in a.iter().enumerate() {
            if val > 0.0 {
                prop_assert!(
                    (result.data[i] - val).abs() < 1e-6,
                    "Positive value {} should be preserved, got {}",
                    val, result.data[i]
                );
            }
        }
    }
}

proptest! {
    /// Property test: elu() produces values >= -alpha for negative inputs
    /// ELU saturates to -alpha as x -> -inf
    #[test]
    fn test_elu_bounded_below_property(
        a in prop::collection::vec(-100.0f32..100.0, 1..100),
        alpha in 0.1f32..5.0
    ) {
        let va = Vector::from_slice(&a);
        let result = va.elu(alpha).unwrap();

        for &val in result.as_slice() {
            prop_assert!(
                val >= -alpha - 0.01,
                "ELU output {} should be >= -alpha = {}",
                val, -alpha
            );
        }
    }
}

proptest! {
    /// Property test: elu() is monotonically increasing
    /// If x < y, then elu(x) < elu(y)
    #[test]
    fn test_elu_monotonic_property(
        a in prop::collection::vec(-20.0f32..20.0, 2..50),
        alpha in 0.5f32..2.0
    ) {
        let va = Vector::from_slice(&a);
        let result = va.elu(alpha).unwrap();

        // Check all pairs for monotonicity
        for i in 0..a.len() {
            for j in 0..a.len() {
                if a[i] < a[j] {
                    prop_assert!(
                        result.data[i] < result.data[j] + 1e-5,
                        "Monotonicity violated: {} < {} but elu({}) = {} >= elu({}) = {}",
                        a[i], a[j], a[i], result.data[i], a[j], result.data[j]
                    );
                }
            }
        }
    }
}