realizar 0.8.6

Pure Rust ML inference engine built from scratch - model serving for GGUF and safetensors
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
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362


// ============================================================================
// Tests
// ============================================================================

#[cfg(test)]
mod tests {
    use super::*;

    /// Simple linear model for testing
    struct LinearModel {
        weights: Vec<f32>,
        bias: f32,
    }

    impl LinearModel {
        fn new(weights: Vec<f32>, bias: f32) -> Self {
            Self { weights, bias }
        }
    }

    impl Explainable for LinearModel {
        fn predict(&self, instance: &[f32]) -> Result<f32, ExplainError> {
            if instance.len() != self.weights.len() {
                return Err(ExplainError::InvalidInput {
                    expected: self.weights.len(),
                    actual: instance.len(),
                });
            }
            let dot: f32 = instance.iter().zip(&self.weights).map(|(x, w)| x * w).sum();
            Ok(dot + self.bias)
        }

        fn n_features(&self) -> usize {
            self.weights.len()
        }
    }

    /// Simple tree model for testing
    struct SimpleTreeModel {
        structure: TreeStructure,
    }

    impl SimpleTreeModel {
        fn new(tree: DecisionTree) -> Self {
            let n_features = tree
                .feature
                .iter()
                .filter(|&&f| f >= 0)
                .map(|&f| f as usize + 1)
                .max()
                .unwrap_or(1);
            Self {
                structure: TreeStructure {
                    n_trees: 1,
                    n_features,
                    trees: vec![tree],
                },
            }
        }
    }

    impl Explainable for SimpleTreeModel {
        fn predict(&self, instance: &[f32]) -> Result<f32, ExplainError> {
            let sum: f32 = self
                .structure
                .trees
                .iter()
                .map(|t| t.predict(instance))
                .sum();
            Ok(sum / self.structure.n_trees as f32)
        }

        fn n_features(&self) -> usize {
            self.structure.n_features
        }

        fn is_tree_model(&self) -> bool {
            true
        }

        fn get_tree_structure(&self) -> Option<&TreeStructure> {
            Some(&self.structure)
        }
    }

    // === ShapExplanation Tests ===

    #[test]
    fn test_shap_explanation_new() {
        let exp = ShapExplanation::new(0.5, vec![0.1, -0.2, 0.3], 0.7);
        assert_eq!(exp.base_value, 0.5);
        assert_eq!(exp.shap_values.len(), 3);
        assert_eq!(exp.prediction, 0.7);
        assert_eq!(exp.feature_names.len(), 3);
    }

    #[test]
    fn test_shap_explanation_with_feature_names() {
        let exp = ShapExplanation::new(0.5, vec![0.1, -0.2], 0.4)
            .with_feature_names(vec!["age".to_string(), "income".to_string()]);
        assert_eq!(exp.feature_names, vec!["age", "income"]);
    }

    #[test]
    fn test_shap_explanation_top_features() {
        let exp = ShapExplanation::new(0.0, vec![0.1, -0.3, 0.2], 0.0).with_feature_names(vec![
            "a".to_string(),
            "b".to_string(),
            "c".to_string(),
        ]);

        let top = exp.top_features(2);
        assert_eq!(top.len(), 2);
        // Should be sorted by absolute value
        assert_eq!(top[0].0, "b"); // |-0.3| = 0.3
        assert_eq!(top[1].0, "c"); // |0.2| = 0.2
    }

    #[test]
    fn test_shap_explanation_verify_consistency() {
        // Perfect consistency
        let exp = ShapExplanation::new(0.5, vec![0.2, 0.3], 1.0);
        assert!(exp.verify_consistency(0.01));

        // Not consistent
        let exp_bad = ShapExplanation::new(0.5, vec![0.2, 0.3], 2.0);
        assert!(!exp_bad.verify_consistency(0.01));
    }

    #[test]
    fn test_shap_explanation_display() {
        let exp = ShapExplanation::new(0.5, vec![0.1, -0.2], 0.4);
        let display = format!("{exp}");
        assert!(display.contains("SHAP Explanation"));
        assert!(display.contains("Base value"));
        assert!(display.contains("Prediction"));
    }

    // === Decision Tree Tests ===

    #[test]
    fn test_decision_tree_predict_simple() {
        // Simple tree: if feature_0 <= 0.5 then 1.0 else 2.0
        let tree = DecisionTree::new(
            vec![0, -1, -1],     // feature indices (-1 = leaf)
            vec![0.5, 0.0, 0.0], // thresholds
            vec![1, 0, 0],       // left children
            vec![2, 0, 0],       // right children
            vec![0.0, 1.0, 2.0], // values
        );

        assert_eq!(tree.predict(&[0.3]), 1.0); // <= 0.5, go left
        assert_eq!(tree.predict(&[0.7]), 2.0); // > 0.5, go right
    }

    #[test]
    fn test_decision_tree_n_nodes() {
        let tree = DecisionTree::new(
            vec![0, -1, -1],
            vec![0.5, 0.0, 0.0],
            vec![1, 0, 0],
            vec![2, 0, 0],
            vec![0.0, 1.0, 2.0],
        );
        assert_eq!(tree.n_nodes(), 3);
    }

    #[test]
    fn test_decision_tree_is_leaf() {
        let tree = DecisionTree::new(
            vec![0, -1, -1],
            vec![0.5, 0.0, 0.0],
            vec![1, 0, 0],
            vec![2, 0, 0],
            vec![0.0, 1.0, 2.0],
        );
        assert!(!tree.is_leaf(0)); // root is not leaf
        assert!(tree.is_leaf(1)); // leaf
        assert!(tree.is_leaf(2)); // leaf
    }

    // === ShapExplainer Tests ===

    #[test]
    fn test_shap_explainer_new() {
        let background = vec![vec![1.0, 2.0], vec![3.0, 4.0]];
        let explainer = ShapExplainer::new(background);
        assert_eq!(explainer.nsamples, 100);
        assert_eq!(explainer.feature_names.len(), 2);
    }

    #[test]
    fn test_shap_explainer_with_nsamples() {
        let explainer = ShapExplainer::new(vec![vec![1.0]]).with_nsamples(50);
        assert_eq!(explainer.nsamples, 50);
    }

    #[test]
    fn test_shap_explainer_with_feature_names() {
        let explainer = ShapExplainer::new(vec![vec![1.0, 2.0]])
            .with_feature_names(vec!["age".to_string(), "income".to_string()]);
        assert_eq!(explainer.feature_names, vec!["age", "income"]);
    }

    #[test]
    fn test_shap_explainer_linear_model() {
        let model = LinearModel::new(vec![1.0, 2.0], 0.0);
        let background = vec![vec![0.0, 0.0], vec![1.0, 1.0]];
        let explainer = ShapExplainer::new(background);

        let instance = vec![2.0, 3.0]; // prediction = 2*1 + 3*2 = 8
        let explanation = explainer.explain(&model, &instance).expect("test");

        assert_eq!(explanation.prediction, 8.0);
        assert_eq!(explanation.shap_values.len(), 2);
    }

    #[test]
    fn test_shap_explainer_tree_model() {
        // Simple tree: if feature_0 <= 0.5 then 1.0 else 2.0
        let tree = DecisionTree::new(
            vec![0, -1, -1],
            vec![0.5, 0.0, 0.0],
            vec![1, 0, 0],
            vec![2, 0, 0],
            vec![0.0, 1.0, 2.0],
        );
        let model = SimpleTreeModel::new(tree);

        let background = vec![vec![0.3], vec![0.7]];
        let explainer = ShapExplainer::new(background);

        let explanation = explainer.explain(&model, &[0.3]).expect("test");
        assert_eq!(explanation.prediction, 1.0);
    }

    #[test]
    fn test_shap_explainer_invalid_input() {
        let model = LinearModel::new(vec![1.0, 2.0], 0.0);
        let background = vec![vec![0.0, 0.0]];
        let explainer = ShapExplainer::new(background);

        let result = explainer.explain(&model, &[1.0, 2.0, 3.0]); // 3 features, model expects 2
        assert!(matches!(result, Err(ExplainError::InvalidInput { .. })));
    }

    #[test]
    fn test_shap_explainer_empty_background() {
        let model = LinearModel::new(vec![1.0], 0.0);
        let explainer = ShapExplainer::new(vec![]); // Empty background

        let result = explainer.explain(&model, &[1.0]);
        assert!(matches!(result, Err(ExplainError::NoBackground)));
    }

    // === ExplainError Tests ===

    #[test]
    fn test_explain_error_display() {
        let err = ExplainError::UnsupportedModel {
            reason: "not a tree".to_string(),
        };
        assert!(err.to_string().contains("not a tree"));

        let err = ExplainError::InvalidInput {
            expected: 3,
            actual: 2,
        };
        assert!(err.to_string().contains("expected 3"));
        assert!(err.to_string().contains("got 2"));

        let err = ExplainError::NoBackground;
        assert!(err.to_string().contains("Background"));

        let err = ExplainError::ComputationError("overflow".to_string());
        assert!(err.to_string().contains("overflow"));
    }

    // === Binomial Tests ===

    #[test]
    fn test_binomial_coefficient() {
        assert_eq!(binomial(5, 0), 1);
        assert_eq!(binomial(5, 1), 5);
        assert_eq!(binomial(5, 2), 10);
        assert_eq!(binomial(5, 3), 10);
        assert_eq!(binomial(5, 4), 5);
        assert_eq!(binomial(5, 5), 1);
        assert_eq!(binomial(5, 6), 0); // k > n
    }

    #[test]
    fn test_binomial_edge_cases() {
        assert_eq!(binomial(0, 0), 1);
        assert_eq!(binomial(1, 0), 1);
        assert_eq!(binomial(1, 1), 1);
        assert_eq!(binomial(10, 5), 252);
    }

    // === TreeStructure Tests ===

    #[test]
    fn test_tree_structure_serialization() {
        let tree = DecisionTree::new(
            vec![0, -1, -1],
            vec![0.5, 0.0, 0.0],
            vec![1, 0, 0],
            vec![2, 0, 0],
            vec![0.0, 1.0, 2.0],
        );
        let structure = TreeStructure {
            n_trees: 1,
            n_features: 1,
            trees: vec![tree],
        };

        let json = serde_json::to_string(&structure).expect("test");
        assert!(json.contains("n_trees"));
        assert!(json.contains("n_features"));
    }

    #[test]
    fn test_shap_explanation_serialization() {
        let exp = ShapExplanation::new(0.5, vec![0.1, -0.2], 0.4);
        let json = serde_json::to_string(&exp).expect("test");
        let parsed: ShapExplanation = serde_json::from_str(&json).expect("test");

        assert_eq!(parsed.base_value, exp.base_value);
        assert_eq!(parsed.shap_values, exp.shap_values);
        assert_eq!(parsed.prediction, exp.prediction);
    }

    // === Additional Edge Case Tests ===

    #[test]
    fn test_shap_explanation_empty_values() {
        let exp = ShapExplanation::new(0.5, vec![], 0.5);
        assert!(exp.verify_consistency(0.01));
        assert!(exp.top_features(3).is_empty());
    }

    #[test]
    fn test_decision_tree_empty_instance() {
        let tree = DecisionTree::new(
            vec![-1], // Just a leaf
            vec![0.0],
            vec![0],
            vec![0],
            vec![5.0],
        );
        assert_eq!(tree.predict(&[]), 5.0);
    }

    #[test]
    fn test_linear_model_batch_predict() {
        let model = LinearModel::new(vec![1.0, 2.0], 0.0);
        let instances = vec![vec![1.0, 1.0], vec![2.0, 2.0]];
        let predictions = model.predict_batch(&instances).expect("test");
        assert_eq!(predictions, vec![3.0, 6.0]);
    }
}