oxionnx 0.1.3

Pure Rust ONNX inference engine — zero C/C++ dependencies
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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309

//! Integration tests for oxionnx using `Session::from_graph()`.

use std::collections::HashMap;

use oxionnx::{Attributes, Graph, Node, OpKind, Session, Tensor};

// ── Helpers ──────────────────────────────────────────────────────────────────

fn make_node(op: OpKind, name: &str, inputs: &[&str], outputs: &[&str]) -> Node {
    Node {
        op,
        name: name.to_string(),
        inputs: inputs.iter().map(|s| s.to_string()).collect(),
        outputs: outputs.iter().map(|s| s.to_string()).collect(),
        attrs: Attributes::default(),
    }
}

fn make_node_with_attrs(
    op: OpKind,
    name: &str,
    inputs: &[&str],
    outputs: &[&str],
    attrs: Attributes,
) -> Node {
    Node {
        op,
        name: name.to_string(),
        inputs: inputs.iter().map(|s| s.to_string()).collect(),
        outputs: outputs.iter().map(|s| s.to_string()).collect(),
        attrs,
    }
}

// ── Test 1: Identity graph ──────────────────────────────────────────────────

#[test]
fn test_identity_graph() {
    let graph = Graph {
        nodes: vec![make_node(OpKind::Identity, "id0", &["x"], &["y"])],
        input_names: vec!["x".to_string()],
        output_names: vec!["y".to_string()],
        ..Default::default()
    };
    let weights: HashMap<String, Tensor> = HashMap::new();

    let session = Session::from_graph(graph, weights).expect("build session");
    let input = Tensor::new(vec![1.0, 2.0, 3.0], vec![1, 3]);
    let outputs = session.run_one("x", input.clone()).expect("run");

    let y = outputs.get("y").expect("output y");
    assert_eq!(y.shape, vec![1, 3]);
    assert_eq!(y.data, vec![1.0, 2.0, 3.0]);
}

// ── Test 2: Add two constants ───────────────────────────────────────────────

#[test]
fn test_add_two_constants() {
    let graph = Graph {
        nodes: vec![make_node(OpKind::Add, "add0", &["a", "b"], &["sum"])],
        input_names: vec![],
        output_names: vec!["sum".to_string()],
        ..Default::default()
    };
    let mut weights: HashMap<String, Tensor> = HashMap::new();
    weights.insert("a".to_string(), Tensor::new(vec![1.0, 2.0, 3.0], vec![3]));
    weights.insert("b".to_string(), Tensor::new(vec![4.0, 5.0, 6.0], vec![3]));

    let session = Session::builder()
        .with_optimization_level(oxionnx::OptLevel::None)
        .build_from_graph(graph, weights)
        .expect("build session");
    let inputs: HashMap<&str, Tensor> = HashMap::new();
    let outputs = session.run(&inputs).expect("run");

    let sum = outputs.get("sum").expect("output sum");
    assert_eq!(sum.shape, vec![3]);
    assert_eq!(sum.data, vec![5.0, 7.0, 9.0]);
}

// ── Test 3: Simple linear layer (MatMul + Add) ─────────────────────────────

#[test]
fn test_linear_layer() {
    // x [1,3] @ W [3,2] => mm [1,2], then mm + b [2] => out [1,2]
    let graph = Graph {
        nodes: vec![
            make_node(OpKind::MatMul, "matmul0", &["x", "W"], &["mm"]),
            make_node(OpKind::Add, "add0", &["mm", "b"], &["out"]),
        ],
        input_names: vec!["x".to_string()],
        output_names: vec!["out".to_string()],
        ..Default::default()
    };

    // W = [[1, 0],
    //      [0, 1],
    //      [1, 1]]  shape [3, 2]
    let w_data = vec![1.0, 0.0, 0.0, 1.0, 1.0, 1.0];
    let b_data = vec![0.5, -0.5];
    let mut weights: HashMap<String, Tensor> = HashMap::new();
    weights.insert("W".to_string(), Tensor::new(w_data, vec![3, 2]));
    weights.insert("b".to_string(), Tensor::new(b_data, vec![2]));

    let session = Session::from_graph(graph, weights).expect("build session");
    // x = [1, 2, 3]  => mm = [1*1+2*0+3*1, 1*0+2*1+3*1] = [4, 5]
    // out = [4+0.5, 5-0.5] = [4.5, 4.5]
    let x = Tensor::new(vec![1.0, 2.0, 3.0], vec![1, 3]);
    let outputs = session.run_one("x", x).expect("run");

    let out = outputs.get("out").expect("output out");
    assert_eq!(out.shape, vec![1, 2]);
    let expected = [4.5, 4.5];
    for (a, b) in out.data.iter().zip(expected.iter()) {
        assert!((a - b).abs() < 1e-6, "expected {b}, got {a}");
    }
}

// ── Test 4: Conv2D + ReLU ───────────────────────────────────────────────────

#[test]
fn test_conv2d_relu() {
    // Conv attributes
    let mut conv_attrs = Attributes::default();
    conv_attrs
        .int_lists
        .insert("strides".to_string(), vec![1, 1]);
    conv_attrs
        .int_lists
        .insert("pads".to_string(), vec![0, 0, 0, 0]);
    conv_attrs
        .int_lists
        .insert("dilations".to_string(), vec![1, 1]);
    conv_attrs.ints.insert("group".to_string(), 1);

    let graph = Graph {
        nodes: vec![
            make_node_with_attrs(
                OpKind::Conv,
                "conv0",
                &["input", "conv_w"],
                &["conv_out"],
                conv_attrs,
            ),
            make_node(OpKind::Relu, "relu0", &["conv_out"], &["relu_out"]),
        ],
        input_names: vec!["input".to_string()],
        output_names: vec!["relu_out".to_string()],
        ..Default::default()
    };

    // Input: 5x5 image, all ones
    let input_data = vec![1.0_f32; 25];
    // Kernel: 3x3, all ones except center is -10 (to produce some negatives)
    let mut kernel_data = vec![1.0_f32; 9];
    kernel_data[4] = -10.0; // center element

    let mut weights: HashMap<String, Tensor> = HashMap::new();
    weights.insert(
        "conv_w".to_string(),
        Tensor::new(kernel_data, vec![1, 1, 3, 3]),
    );

    let session = Session::from_graph(graph, weights).expect("build session");
    let input = Tensor::new(input_data, vec![1, 1, 5, 5]);
    let outputs = session.run_one("input", input).expect("run");

    let relu_out = outputs.get("relu_out").expect("output relu_out");
    assert_eq!(relu_out.shape, vec![1, 1, 3, 3]);

    // Each conv output = sum of 9 kernel values * 1.0 = (8*1 + 1*(-10)) = -2
    // After ReLU: max(0, -2) = 0
    for &v in &relu_out.data {
        assert!(v >= 0.0, "ReLU output should be non-negative, got {v}");
    }
    // With all-ones input and our kernel, every output should be 0
    for &v in &relu_out.data {
        assert!((v - 0.0).abs() < 1e-6, "expected 0.0, got {v}");
    }
}

// ── Test 5: Multi-output Split ──────────────────────────────────────────────

#[test]
fn test_split() {
    let mut split_attrs = Attributes::default();
    split_attrs.ints.insert("axis".to_string(), 1);
    split_attrs
        .int_lists
        .insert("split".to_string(), vec![3, 3]);

    let graph = Graph {
        nodes: vec![make_node_with_attrs(
            OpKind::Split,
            "split0",
            &["x"],
            &["a", "b"],
            split_attrs,
        )],
        input_names: vec!["x".to_string()],
        output_names: vec!["a".to_string(), "b".to_string()],
        ..Default::default()
    };
    let weights: HashMap<String, Tensor> = HashMap::new();

    let session = Session::from_graph(graph, weights).expect("build session");
    let x = Tensor::new(vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0], vec![1, 6]);
    let outputs = session.run_one("x", x).expect("run");

    let a = outputs.get("a").expect("output a");
    let b = outputs.get("b").expect("output b");
    assert_eq!(a.shape, vec![1, 3]);
    assert_eq!(b.shape, vec![1, 3]);
    assert_eq!(a.data, vec![1.0, 2.0, 3.0]);
    assert_eq!(b.data, vec![4.0, 5.0, 6.0]);
}

// ── Test 6: Session builder + from_graph, model_info, export_dot ────────────

#[test]
fn test_session_builder_and_introspection() {
    let graph = Graph {
        nodes: vec![
            make_node(OpKind::MatMul, "matmul_node", &["x", "W"], &["mm"]),
            make_node(OpKind::Relu, "relu_node", &["mm"], &["out"]),
        ],
        input_names: vec!["x".to_string()],
        output_names: vec!["out".to_string()],
        ..Default::default()
    };
    let mut weights: HashMap<String, Tensor> = HashMap::new();
    weights.insert("W".to_string(), Tensor::new(vec![1.0; 6], vec![3, 2]));

    let session = Session::builder()
        .with_optimization_level(oxionnx::OptLevel::None)
        .build_from_graph(graph, weights)
        .expect("build session");

    // model_info
    let info = session.model_info();
    assert!(info.node_count >= 2, "expected at least 2 nodes");
    assert!(
        info.op_histogram.contains_key("MatMul"),
        "histogram should contain MatMul"
    );
    assert!(
        info.op_histogram.contains_key("Relu"),
        "histogram should contain Relu"
    );
    assert_eq!(info.parameter_count, 6); // W has 6 elements

    // export_dot
    let dot = session.export_dot();
    assert!(dot.contains("digraph"), "DOT should start with digraph");
    assert!(
        dot.contains("matmul_node"),
        "DOT should mention matmul_node"
    );
    assert!(dot.contains("relu_node"), "DOT should mention relu_node");

    // input/output names
    assert_eq!(session.input_names(), &["x"]);
    assert_eq!(session.output_names(), &["out"]);
}

// ── Test 7: Profiling ───────────────────────────────────────────────────────

#[test]
fn test_profiling() {
    let graph = Graph {
        nodes: vec![
            make_node(OpKind::Identity, "id_a", &["x"], &["mid"]),
            make_node(OpKind::Identity, "id_b", &["mid"], &["y"]),
        ],
        input_names: vec!["x".to_string()],
        output_names: vec!["y".to_string()],
        ..Default::default()
    };
    let weights: HashMap<String, Tensor> = HashMap::new();

    let session = Session::builder()
        .with_profiling()
        .build_from_graph(graph, weights)
        .expect("build session");

    // Before running, profiling data should be empty
    let initial = session.profiling_results().expect("profiling enabled");
    assert!(initial.is_empty(), "no profiles before first run");

    let x = Tensor::new(vec![1.0, 2.0, 3.0, 4.0], vec![2, 2]);
    let _outputs = session.run_one("x", x).expect("run");

    let profiles = session.profiling_results().expect("profiling enabled");
    assert!(
        !profiles.is_empty(),
        "profiling results should not be empty after run"
    );

    for p in &profiles {
        assert!(!p.node_name.is_empty(), "node_name should not be empty");
        assert!(!p.op_type.is_empty(), "op_type should not be empty");
    }

    // Verify we captured both nodes
    let names: Vec<&str> = profiles.iter().map(|p| p.node_name.as_str()).collect();
    assert!(names.contains(&"id_a"), "profiles should contain id_a");
    assert!(names.contains(&"id_b"), "profiles should contain id_b");
}