numr 0.5.2

High-performance numerical computing with multi-backend GPU acceleration (CPU/CUDA/WebGPU)
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

//! WebGPU runtime operation helpers and TensorOps implementation.
//!
//! Trait implementations (CompareOps, ScalarOps, LogicalOps, etc.) live in `ops/wgpu/`.
//! This module provides shared helpers, native GPU launchers, and TensorOps.

pub(crate) mod helpers;
pub(crate) mod native;
mod tensor;

// ============================================================================
// Tests
// ============================================================================

#[cfg(test)]
mod tests {
    use crate::ops::{ActivationOps, BinaryOps, CompareOps, MatmulOps, ReduceOps, ScalarOps};
    use crate::runtime::Runtime;
    use crate::runtime::wgpu::{WgpuDevice, WgpuRuntime, is_wgpu_available};
    use crate::tensor::Tensor;

    fn create_test_tensor(data: &[f32], shape: &[usize]) -> Tensor<WgpuRuntime> {
        let device = WgpuDevice::new(0);
        Tensor::<WgpuRuntime>::from_slice(data, shape, &device)
    }

    #[test]
    fn test_wgpu_add() {
        if !is_wgpu_available() {
            println!("No GPU available, skipping test");
            return;
        }

        let device = WgpuDevice::new(0);
        let client = WgpuRuntime::default_client(&device);

        let a = create_test_tensor(&[1.0, 2.0, 3.0, 4.0], &[2, 2]);
        let b = create_test_tensor(&[5.0, 6.0, 7.0, 8.0], &[2, 2]);

        let result = client.add(&a, &b).unwrap();
        let data: Vec<f32> = result.to_vec();

        assert_eq!(data, vec![6.0, 8.0, 10.0, 12.0]);
    }

    #[test]
    fn test_wgpu_matmul() {
        if !is_wgpu_available() {
            println!("No GPU available, skipping test");
            return;
        }

        let device = WgpuDevice::new(0);
        let client = WgpuRuntime::default_client(&device);

        // 2x3 @ 3x2 = 2x2
        let a = create_test_tensor(&[1.0, 2.0, 3.0, 4.0, 5.0, 6.0], &[2, 3]);
        let b = create_test_tensor(&[1.0, 2.0, 3.0, 4.0, 5.0, 6.0], &[3, 2]);

        let result = client.matmul(&a, &b).unwrap();
        assert_eq!(result.shape(), &[2, 2]);

        let data: Vec<f32> = result.to_vec();
        assert_eq!(data, vec![22.0, 28.0, 49.0, 64.0]);
    }

    #[test]
    fn test_wgpu_relu() {
        if !is_wgpu_available() {
            println!("No GPU available, skipping test");
            return;
        }

        let device = WgpuDevice::new(0);
        let client = WgpuRuntime::default_client(&device);

        let a = create_test_tensor(&[-1.0, 0.0, 1.0, 2.0], &[4]);
        let result = client.relu(&a).unwrap();
        let data: Vec<f32> = result.to_vec();

        assert_eq!(data, vec![0.0, 0.0, 1.0, 2.0]);
    }

    #[test]
    fn test_wgpu_sum() {
        if !is_wgpu_available() {
            println!("No GPU available, skipping test");
            return;
        }

        let device = WgpuDevice::new(0);
        let client = WgpuRuntime::default_client(&device);

        let a = create_test_tensor(&[1.0, 2.0, 3.0, 4.0], &[2, 2]);
        let result = client.sum(&a, &[0], false).unwrap();
        let data: Vec<f32> = result.to_vec();

        assert_eq!(data, vec![4.0, 6.0]);
    }

    #[test]
    fn test_wgpu_mul_scalar() {
        if !is_wgpu_available() {
            println!("No GPU available, skipping test");
            return;
        }

        let device = WgpuDevice::new(0);
        let client = WgpuRuntime::default_client(&device);

        let a = create_test_tensor(&[1.0, 2.0, 3.0, 4.0], &[4]);
        let result = client.mul_scalar(&a, 2.0).unwrap();
        let data: Vec<f32> = result.to_vec();

        assert_eq!(data, vec![2.0, 4.0, 6.0, 8.0]);
    }

    #[test]
    fn test_wgpu_eq() {
        if !is_wgpu_available() {
            println!("No GPU available, skipping test");
            return;
        }

        let device = WgpuDevice::new(0);
        let client = WgpuRuntime::default_client(&device);

        let a = create_test_tensor(&[1.0, 2.0, 3.0, 4.0], &[4]);
        let b = create_test_tensor(&[1.0, 0.0, 3.0, 0.0], &[4]);

        let result = client.eq(&a, &b).unwrap();
        let data: Vec<f32> = result.to_vec();

        assert_eq!(data, vec![1.0, 0.0, 1.0, 0.0]);
    }
}