onnx-export-rs 0.1.1

Export canonical Rust machine-learning models to ONNX
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

use crate::canonical::{SvmClassifier, SvmKernel, SvmRegressor};
use crate::graph_builder::{
    assemble_model, floats_attribute, int_attribute, ints_attribute, make_node,
    make_typed_value_info, make_value_info, string_attribute, Dimension, INT64,
};
use crate::proto::{GraphProto, ModelProto, OperatorSetIdProto};
use crate::{Result, IR_VERSION, ML_OPSET_VERSION, OPSET_VERSION};

fn kernel_attributes(kernel: SvmKernel) -> Vec<crate::proto::AttributeProto> {
    vec![
        string_attribute("kernel_type", kernel.onnx_name()),
        floats_attribute("kernel_params", kernel.onnx_parameters()),
    ]
}

fn add_ml_opset(mut model: ModelProto) -> ModelProto {
    model.opset_import.push(OperatorSetIdProto {
        domain: "ai.onnx.ml".into(),
        version: ML_OPSET_VERSION,
    });
    model
}

/// Exports an ONNX-ML SVM regressor or one-class SVM.
#[must_use]
pub fn export_svm_regressor(model: &SvmRegressor) -> ModelProto {
    let mut attributes = kernel_attributes(model.kernel);
    attributes.extend([
        floats_attribute(
            "support_vectors",
            model
                .support_vectors
                .iter()
                .map(|&value| value as f32)
                .collect(),
        ),
        floats_attribute(
            "coefficients",
            model
                .coefficients
                .iter()
                .map(|&value| value as f32)
                .collect(),
        ),
        floats_attribute("rho", vec![model.rho as f32]),
        int_attribute("n_supports", model.support_vectors.nrows() as i64),
        int_attribute("one_class", i64::from(model.one_class)),
        string_attribute("post_transform", b"NONE".as_slice()),
    ]);
    let mut node = make_node("SVMRegressor", ["X"], ["Y"], attributes);
    node.domain = "ai.onnx.ml".into();
    let graph = GraphProto {
        node: vec![node],
        name: "svm_regressor".into(),
        initializer: Vec::new(),
        doc_string: String::new(),
        input: vec![make_value_info(
            "X",
            &[
                Dimension::Symbolic("batch".into()),
                Dimension::Fixed(model.support_vectors.ncols()),
            ],
        )],
        output: vec![make_value_info(
            "Y",
            &[Dimension::Symbolic("batch".into()), Dimension::Fixed(1)],
        )],
        value_info: Vec::new(),
    };
    add_ml_opset(assemble_model(graph, OPSET_VERSION, IR_VERSION))
}

/// Exports an integer-labelled ONNX-ML SVM classifier.
///
/// # Errors
///
/// Returns an error when the canonical raw ONNX coefficient layout is invalid.
pub fn export_svm_classifier(model: &SvmClassifier) -> Result<ModelProto> {
    model.validate()?;
    let mut attributes = kernel_attributes(model.kernel);
    attributes.extend([
        ints_attribute("classlabels_ints", model.class_labels.clone()),
        ints_attribute(
            "vectors_per_class",
            model
                .vectors_per_class
                .iter()
                .map(|&value| value as i64)
                .collect(),
        ),
        floats_attribute(
            "support_vectors",
            model
                .support_vectors
                .iter()
                .map(|&value| value as f32)
                .collect(),
        ),
        floats_attribute(
            "coefficients",
            model
                .coefficients
                .iter()
                .map(|&value| value as f32)
                .collect(),
        ),
        floats_attribute("rho", model.rho.iter().map(|&value| value as f32).collect()),
        floats_attribute(
            "prob_a",
            model.prob_a.iter().map(|&value| value as f32).collect(),
        ),
        floats_attribute(
            "prob_b",
            model.prob_b.iter().map(|&value| value as f32).collect(),
        ),
        string_attribute("post_transform", b"NONE".as_slice()),
    ]);
    let mut node = make_node("SVMClassifier", ["X"], ["label", "scores"], attributes);
    node.domain = "ai.onnx.ml".into();
    let graph = GraphProto {
        node: vec![node],
        name: "svm_classifier".into(),
        initializer: Vec::new(),
        doc_string: String::new(),
        input: vec![make_value_info(
            "X",
            &[
                Dimension::Symbolic("batch".into()),
                Dimension::Fixed(model.support_vectors.ncols()),
            ],
        )],
        output: vec![
            make_typed_value_info("label", &[Dimension::Symbolic("batch".into())], INT64),
            make_value_info(
                "scores",
                &[
                    Dimension::Symbolic("batch".into()),
                    Dimension::Fixed(model.class_labels.len()),
                ],
            ),
        ],
        value_info: Vec::new(),
    };
    Ok(add_ml_opset(assemble_model(
        graph,
        OPSET_VERSION,
        IR_VERSION,
    )))
}