vkml 0.0.2

High-level Vulkan-based machine learning library
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

mod common;

use std::path::Path;
use vkml::{ComputeManager, Tensor};

use crate::common::get_test_data_path;

/// Helper: load a PNG and return bytes matching the model input tensor size.
fn load_image_as_input(path: &Path) -> Box<[u8]> {
    let raw = image::open(path)
        .expect("Failed to open image")
        .into_bytes();
    let mut out: Vec<u8> = Vec::with_capacity(raw.len() * 4);
    for b in raw.into_iter() {
        let f: f32 = (b as f32) / 255.0;
        out.extend_from_slice(&f.to_le_bytes());
    }
    out.into()
}

fn argmax(data: &[f32]) -> usize {
    data.iter()
        .enumerate()
        .max_by(|(_, a), (_, b)| a.partial_cmp(b).unwrap())
        .map(|(i, _)| i)
        .unwrap()
}

#[test]
fn test_mnist_mlp_fp32() -> Result<(), Box<dyn std::error::Error>> {
    let file_path = get_test_data_path("mnist_mlp_png_probs.onnx");
    let file = file_path.to_str().unwrap();

    let mut cm = ComputeManager::new_from_onnx_path(file)?;

    let img0_path = get_test_data_path("0.png");
    let img1_path = get_test_data_path("1.png");

    let input_ids = cm.tensor_graph.get_input_tensor_ids().to_vec();
    let input_desc = cm.tensor_read(input_ids[0]).desc().clone();

    let bytes0 = load_image_as_input(&img0_path);
    let bytes1 = load_image_as_input(&img1_path);

    let batch0 = Tensor::new_cpu(input_desc.clone(), bytes0);
    let batch1 = Tensor::new_cpu(input_desc.clone(), bytes1);

    let ref0 = [
        1.0,
        3.050_845_1e-9,
        3.666_505e-10,
        2.464_466_1e-12,
        1.580_271_2e-10,
        1.084_859_15e-10,
        1.675_991_4e-8,
        3.684_312_6e-9,
        1.579_724e-12,
        4.535_866_4e-8,
    ];
    let ref1 = [
        1.058_358_9e-11,
        0.999_994_75,
        5.817_719_3e-8,
        1.388_941_5e-10,
        3.135_667_4e-7,
        2.058_546_8e-8,
        6.364_984_4e-7,
        6.640_671e-7,
        3.540_966_9e-6,
        1.474_830_8e-10,
    ];

    println!("Running inference for mnist_mlp_fp32 (0.png)...");
    let out0 = cm.forward(vec![batch0])?;
    let out0_t = cm.tensor_read(out0[0]).read();
    let res0 = bytemuck::cast_slice::<u8, f32>(&out0_t);
    println!("0.png Prediction: {} (Results: {:?})", argmax(res0), res0);
    common::assert_tensors_close(res0, &ref0, 1e-4, 1e-5, "mnist_mlp_fp32 Image 0");

    println!("Running inference for mnist_mlp_fp32 (1.png)...");
    let out1 = cm.forward(vec![batch1])?;
    let out1_t = cm.tensor_read(out1[0]).read();
    let res1 = bytemuck::cast_slice::<u8, f32>(&out1_t);
    println!("1.png Prediction: {} (Results: {:?})", argmax(res1), res1);
    common::assert_tensors_close(res1, &ref1, 1e-4, 1e-5, "mnist_mlp_fp32 Image 1");

    Ok(())
}

#[test]
fn test_mnist_12_fp32() -> Result<(), Box<dyn std::error::Error>> {
    let file_path = get_test_data_path("mnist-12.onnx");
    let file = file_path.to_str().unwrap();

    let mut cm = ComputeManager::new_from_onnx_path(file)?;

    let img0_path = get_test_data_path("0.png");
    let img1_path = get_test_data_path("1.png");

    let input_ids = cm.tensor_graph.get_input_tensor_ids().to_vec();
    let input_desc = cm.tensor_read(input_ids[0]).desc().clone();

    let ref0 = [
        17.405_083,
        -15.782_351,
        3.530_790_8,
        -6.299_926_3,
        -6.745_813,
        -2.670_010_8,
        6.638_413,
        -5.180_913,
        -2.657_142_2,
        0.046_438_437,
    ];
    let ref1 = [
        -1.514_844_2,
        11.094_378,
        -1.328_422_5,
        -6.250_932,
        -0.783_710_6,
        -2.500_538_8,
        -5.469_571,
        1.325_823_7,
        -3.772_811_7,
        -5.065_256,
    ];

    println!("Running inference for mnist-12 (0.png)...");
    let bytes0 = load_image_as_input(&img0_path);
    let batch0 = Tensor::new_cpu(input_desc.clone(), bytes0);
    let out0 = cm.forward(vec![batch0])?;
    let out0_t = cm.tensor_read(out0[0]).read();
    let res0 = bytemuck::cast_slice::<u8, f32>(&out0_t);
    println!("0.png Prediction: {} (Results: {:?})", argmax(res0), res0);
    common::assert_tensors_close(res0, &ref0, 1e-4, 1e-5, "mnist-12 Image 0");

    println!("Running inference for mnist-12 (1.png)...");
    let bytes1 = load_image_as_input(&img1_path);
    let batch1 = Tensor::new_cpu(input_desc.clone(), bytes1);
    let out1 = cm.forward(vec![batch1])?;
    let out1_t = cm.tensor_read(out1[0]).read();
    let res1 = bytemuck::cast_slice::<u8, f32>(&out1_t);
    println!("1.png Prediction: {} (Results: {:?})", argmax(res1), res1);
    common::assert_tensors_close(res1, &ref1, 1e-4, 1e-5, "mnist-12 Image 1");

    Ok(())
}

#[test]
fn test_multi_chain_add_fp32() -> Result<(), Box<dyn std::error::Error>> {
    let file_path = get_test_data_path("multi_chain_add-10.onnx");
    let file = file_path.to_str().unwrap();

    let mut cm = ComputeManager::new_from_onnx_path(file)?;

    let input_ids = cm.tensor_graph.get_input_tensor_ids().to_vec();
    let input_desc = cm.tensor_read(input_ids[0]).desc().clone();

    let input_bytes = vec![0u8; input_desc.size_in_bytes()];
    let batch = Tensor::new_cpu(input_desc.clone(), input_bytes.into_boxed_slice());

    println!("Running inference for multi_chain_add...");
    let out = cm.forward(vec![batch])?;
    let out_t = cm.tensor_read(out[0]).read();

    println!("multi_chain_add result bytes len: {}", out_t.len());
    Ok(())
}