echidna 0.9.0

A high-performance automatic differentiation library for Rust
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

//! CUDA kernel parity tests mirroring `gpu_kernel_parity.rs`.
//!
//! Exercises the 3A fixes (atan2 overflow, asinh/acosh large-|a|, powf
//! a≤0 safety, max/min NaN routing, fract truncation) against the CUDA
//! reverse_sweep / tangent_forward / tangent_reverse kernels.

#![cfg(feature = "gpu-cuda")]

use echidna::gpu::{CudaContext, GpuBackend, GpuTapeData};
use echidna::{record, BReverse};
use num_traits::Float;

fn cuda_context() -> Option<CudaContext> {
    CudaContext::new()
}

// ── atan2 ───────────────────────────────────────────────────────────

#[test]
fn cuda_atan2_large_magnitudes_gradient_finite() {
    let ctx = match cuda_context() {
        Some(c) => c,
        None => return,
    };

    let x0 = [1.0_f32, 1.0_f32];
    let (tape, _) = record(|v: &[BReverse<f32>]| v[0].atan2(v[1]), &x0);

    let large = 1e20_f32; // a*a = 1e40 overflows f32::MAX ≈ 3.4e38
    let gpu_data = GpuTapeData::from_tape(&tape).unwrap();
    let gpu_tape = ctx.upload_tape(&gpu_data);
    let (_, g) = ctx.gradient_batch(&gpu_tape, &[large, large], 1).unwrap();

    assert!(
        g[0].is_finite(),
        "d/dy atan2 should be finite; got {}",
        g[0]
    );
    assert!(
        g[1].is_finite(),
        "d/dx atan2 should be finite; got {}",
        g[1]
    );
    assert!(g[0] != 0.0, "d/dy atan2 underflowed to zero");
    assert!(g[1] != 0.0, "d/dx atan2 underflowed to zero");
}

// ── asinh / acosh ───────────────────────────────────────────────────

#[test]
fn cuda_asinh_large_derivative_finite() {
    let ctx = match cuda_context() {
        Some(c) => c,
        None => return,
    };

    let x0 = [1.0_f32];
    let (tape, _) = record(|v: &[BReverse<f32>]| v[0].asinh(), &x0);

    let large = 1e20_f32;
    let gpu_data = GpuTapeData::from_tape(&tape).unwrap();
    let gpu_tape = ctx.upload_tape(&gpu_data);
    let (_, g) = ctx.gradient_batch(&gpu_tape, &[large], 1).unwrap();

    assert!(g[0].is_finite(), "asinh derivative should be finite");
    let rel_err = (g[0] as f64 - 1e-20).abs() / 1e-20;
    assert!(rel_err < 1e-5, "g[0] = {}, expected ≈ 1e-20", g[0]);
}

#[test]
fn cuda_acosh_large_derivative_finite() {
    let ctx = match cuda_context() {
        Some(c) => c,
        None => return,
    };

    let x0 = [2.0_f32];
    let (tape, _) = record(|v: &[BReverse<f32>]| v[0].acosh(), &x0);

    let large = 1e20_f32;
    let gpu_data = GpuTapeData::from_tape(&tape).unwrap();
    let gpu_tape = ctx.upload_tape(&gpu_data);
    let (_, g) = ctx.gradient_batch(&gpu_tape, &[large], 1).unwrap();

    assert!(g[0].is_finite(), "acosh derivative should be finite");
    let rel_err = (g[0] as f64 - 1e-20).abs() / 1e-20;
    assert!(rel_err < 1e-5, "g[0] = {}, expected ≈ 1e-20", g[0]);
}

// ── powf ────────────────────────────────────────────────────────────

#[test]
fn cuda_powf_negative_base_integer_exponent() {
    let ctx = match cuda_context() {
        Some(c) => c,
        None => return,
    };

    let x0 = [-2.0_f32];
    let (tape, _) = record(
        |v: &[BReverse<f32>]| v[0].powf(BReverse::constant(3.0)),
        &x0,
    );

    let gpu_data = GpuTapeData::from_tape(&tape).unwrap();
    let gpu_tape = ctx.upload_tape(&gpu_data);
    let (_, g) = ctx.gradient_batch(&gpu_tape, &[-2.0_f32], 1).unwrap();

    assert!(g[0].is_finite(), "gradient must be finite (not NaN)");
    let rel_err = (g[0] as f64 - 12.0).abs() / 12.0;
    assert!(rel_err < 1e-5, "g[0] = {}, expected 12", g[0]);
}

// ── max / min NaN routing ───────────────────────────────────────────

#[test]
fn cuda_max_with_nan_operand() {
    let ctx = match cuda_context() {
        Some(c) => c,
        None => return,
    };

    let x0 = [1.5_f32, f32::NAN];
    let (tape, _) = record(|v: &[BReverse<f32>]| v[0].max(v[1]), &x0);

    let gpu_data = GpuTapeData::from_tape(&tape).unwrap();
    let gpu_tape = ctx.upload_tape(&gpu_data);
    let (_, g) = ctx
        .gradient_batch(&gpu_tape, &[1.5_f32, f32::NAN], 1)
        .unwrap();

    assert_eq!(g[0], 1.0, "adjoint should route to non-NaN operand");
    assert_eq!(g[1], 0.0, "adjoint to NaN operand should be 0");
}

// ── fract ───────────────────────────────────────────────────────────

#[test]
fn cuda_fract_negative_input_matches_cpu() {
    let ctx = match cuda_context() {
        Some(c) => c,
        None => return,
    };

    let x0 = [-1.3_f32];
    let (tape, _) = record(|v: &[BReverse<f32>]| v[0].fract(), &x0);

    let gpu_data = GpuTapeData::from_tape(&tape).unwrap();
    let gpu_tape = ctx.upload_tape(&gpu_data);
    let out = ctx.forward_batch(&gpu_tape, &[-1.3_f32], 1).unwrap();

    let expected = -1.3_f32.fract(); // truncation convention: -0.3
    assert!(
        (out[0] - expected).abs() < 1e-6,
        "fract(-1.3) on CUDA = {}, expected ≈ {}",
        out[0],
        expected,
    );
    assert!(
        out[0] < 0.0,
        "GPU fract should be negative for negative input"
    );
}