tenferro-linalg 0.2.0

Linear algebra traced APIs, eager helpers, extension runtime, and optional AD rules for tenferro.
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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207

use super::*;

#[test]
fn test_solve_zero_dim_returns_zeros() {
    let mut backend = CpuBackend::new();
    let a = Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 0], vec![]).unwrap());
    let b = Tensor::F64(TypedTensor::from_vec_col_major(vec![0, 1], vec![]).unwrap());
    let x = backend.solve(&a, &b).unwrap();
    assert_eq!(x.shape(), &[0, 1]);
}

#[test]
fn test_solve_with_1d_vector_rhs() {
    let a =
        Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 2], vec![2.0, 1.0, 0.0, 3.0]).unwrap());
    let b = Tensor::F64(TypedTensor::from_vec_col_major(vec![2], vec![5.0, 7.0]).unwrap());
    let mut backend = CpuBackend::new();
    let x = backend.solve(&a, &b).unwrap();
    assert_eq!(x.shape(), &[2]);
    let expected = matmul_f64(
        &[2.0, 1.0, 0.0, 3.0],
        &[get_f64(&x, &[0]), get_f64(&x, &[1])],
        2,
        2,
        1,
    );
    assert_f64_close_tol(expected[0], 5.0, 1e-10);
    assert_f64_close_tol(expected[1], 7.0, 1e-10);
}

#[test]
fn test_solve_with_batched_vector_rhs() {
    let a = Tensor::F64(
        TypedTensor::from_vec_col_major(
            vec![2, 2, 2],
            vec![2.0, 1.0, 0.0, 3.0, 1.0, 0.0, 1.0, 2.0],
        )
        .unwrap(),
    );
    let b =
        Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 2], vec![5.0, 7.0, 3.0, 4.0]).unwrap());
    let mut backend = CpuBackend::new();
    let x = backend.solve(&a, &b).unwrap();
    assert_eq!(x.shape(), &[2, 2]);
}

#[test]
fn test_triangular_solve_dtype_mismatch_and_unsupported() {
    let mut backend = CpuBackend::new();
    let a_f32 = Tensor::F32(
        TypedTensor::from_vec_col_major(vec![2, 2], vec![1.0f32, 0.0, 0.0, 1.0]).unwrap(),
    );
    let b_f64 = Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 1], vec![1.0, 2.0]).unwrap());
    let err = backend
        .triangular_solve(&a_f32, &b_f64, true, true, false, false)
        .unwrap_err();
    assert!(matches!(
        err,
        tenferro_tensor::Error::DTypeMismatch {
            op: "triangular_solve",
            ..
        }
    ));

    let a_i64 =
        Tensor::I64(TypedTensor::from_vec_col_major(vec![2, 2], vec![1_i64, 0, 0, 1]).unwrap());
    let b_i64 = Tensor::I64(TypedTensor::from_vec_col_major(vec![2, 1], vec![1_i64, 2]).unwrap());
    let err = backend
        .triangular_solve(&a_i64, &b_i64, true, true, false, false)
        .unwrap_err();
    assert!(matches!(
        err,
        tenferro_tensor::Error::BackendFailure {
            op: "triangular_solve",
            ..
        }
    ));
}

#[test]
fn test_linalg_returns_errors_for_unsupported_dtypes() {
    let mut backend = CpuBackend::new();
    let i64_matrix =
        Tensor::I64(TypedTensor::from_vec_col_major(vec![2, 2], vec![1_i64, 0, 0, 1]).unwrap());
    let i64_rhs = Tensor::I64(TypedTensor::from_vec_col_major(vec![2, 1], vec![1_i64, 2]).unwrap());

    assert!(backend.cholesky(&i64_matrix).is_err());
    assert!(backend.svd(&i64_matrix).is_err());
    assert!(backend.qr(&i64_matrix).is_err());
    assert!(backend.eigh(&i64_matrix).is_err());
    assert!(backend.eig(&i64_matrix).is_err());
    assert!(backend.solve(&i64_matrix, &i64_rhs).is_err());
    assert!(backend
        .triangular_solve(&i64_matrix, &i64_rhs, true, true, false, false)
        .is_err());
}

#[test]
fn test_solve_zero_dim_rhs_returns_zeros() {
    let mut backend = CpuBackend::new();
    let a =
        Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 2], vec![1.0, 0.0, 0.0, 1.0]).unwrap());
    let b = Tensor::F64(TypedTensor::from_vec_col_major(vec![0], vec![]).unwrap());
    let x = backend.solve(&a, &b).unwrap();
    assert_eq!(x.shape(), &[0]);
}

#[test]
fn test_solve_with_regular_matrix_rhs() {
    let a =
        Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 2], vec![2.0, 1.0, 0.0, 3.0]).unwrap());
    let b =
        Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 2], vec![5.0, 7.0, 3.0, 4.0]).unwrap());
    let mut backend = CpuBackend::new();
    let x = backend.solve(&a, &b).unwrap();
    assert_eq!(x.shape(), &[2, 2]);
    let x_data = match &x {
        Tensor::F64(inner) => inner.host_data().unwrap().to_vec(),
        _ => panic!("expected f64 tensor"),
    };
    let recon = matmul_f64(&[2.0, 1.0, 0.0, 3.0], &x_data, 2, 2, 2);
    assert_f64_close_tol(recon[0], 5.0, 1e-10);
    assert_f64_close_tol(recon[1], 7.0, 1e-10);
    assert_f64_close_tol(recon[2], 3.0, 1e-10);
    assert_f64_close_tol(recon[3], 4.0, 1e-10);
}

#[test]
fn test_lu_unsupported_dtype_returns_error() {
    let input =
        Tensor::I64(TypedTensor::from_vec_col_major(vec![2, 2], vec![1_i64, 0, 0, 1]).unwrap());
    let mut backend = CpuBackend::new();
    assert!(backend.lu(&input).is_err());
}

#[test]
fn test_lu_zero_sized_batch_outputs_empty_parity() {
    let input = Tensor::F64(TypedTensor::from_vec_col_major(vec![2, 2, 0], Vec::new()).unwrap());
    let mut backend = CpuBackend::new();
    let outputs = backend.lu(&input).unwrap();

    assert_eq!(outputs.len(), 4);
    assert_eq!(outputs[0].shape(), &[2, 2, 0]);
    assert_eq!(outputs[1].shape(), &[2, 2, 0]);
    assert_eq!(outputs[2].shape(), &[2, 2, 0]);
    assert_eq!(outputs[3].shape(), &[0]);
    for output in outputs {
        match output {
            Tensor::F64(inner) => assert!(inner.host_data().unwrap().is_empty()),
            other => panic!("expected f64 tensor, got {:?}", other.dtype()),
        }
    }
}

#[test]
fn test_svd_unsupported_dtype_returns_error() {
    let input =
        Tensor::I64(TypedTensor::from_vec_col_major(vec![2, 2], vec![1_i64, 0, 0, 1]).unwrap());
    let mut backend = CpuBackend::new();
    assert!(backend.svd(&input).is_err());
}

#[cfg(feature = "cpu-faer")]
#[test]
fn test_faer_svd_decomposition_failure_returns_error() {
    let ctx = CpuContext::with_threads(1).unwrap();
    let mut buffers = BufferPool::new();
    let input = TypedTensor::from_vec_col_major(vec![2, 2], vec![f64::NAN, 0.0, 0.0, 1.0]).unwrap();

    let err = faer_linalg::svd(&ctx, &mut buffers, &input).unwrap_err();

    assert!(err.to_string().contains("svd"), "unexpected error: {err}");
}

#[cfg(feature = "cpu-faer")]
#[test]
fn test_faer_eig_decomposition_failure_returns_error() {
    let ctx = CpuContext::with_threads(1).unwrap();
    let mut buffers = BufferPool::new();
    let input = Tensor::F64(
        TypedTensor::from_vec_col_major(vec![2, 2], vec![f64::NAN, 0.0, 0.0, 1.0]).unwrap(),
    );

    let err = faer_linalg::eig(&ctx, &mut buffers, &input).unwrap_err();

    assert!(err.to_string().contains("eig"), "unexpected error: {err}");
}

#[test]
fn test_qr_unsupported_dtype_returns_error() {
    let input =
        Tensor::I64(TypedTensor::from_vec_col_major(vec![2, 2], vec![1_i64, 0, 0, 1]).unwrap());
    let mut backend = CpuBackend::new();
    assert!(backend.qr(&input).is_err());
}

#[test]
fn test_eig_returns_complex_outputs_for_real_input() {
    let input = Tensor::F64(
        TypedTensor::from_vec_col_major(vec![2, 2], vec![0.0, -1.0, 1.0, 0.0]).unwrap(),
    );
    let mut backend = CpuBackend::new();
    let outputs = backend.eig(&input).unwrap();
    assert_eq!(outputs.len(), 2);
    assert_eq!(outputs[0].shape(), &[2]);
    assert_eq!(outputs[1].shape(), &[2, 2]);
}