aprender-core 0.31.2

Next-generation machine learning library in pure Rust
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

//! Integration property tests.

use super::super::*;
pub(crate) use proptest::prelude::*;

pub(super) fn arb_model_type() -> impl Strategy<Value = ModelType> {
    prop_oneof![
        Just(ModelType::LinearRegression),
        Just(ModelType::LogisticRegression),
        Just(ModelType::DecisionTree),
        Just(ModelType::RandomForest),
        Just(ModelType::KMeans),
        Just(ModelType::NaiveBayes),
        Just(ModelType::Custom),
    ]
}

pub(super) fn arb_model_data() -> impl Strategy<Value = Vec<f32>> {
    proptest::collection::vec(
        any::<f32>().prop_filter("finite", |f| f.is_finite()),
        10..500,
    )
}

pub(super) fn arb_large_model_data() -> impl Strategy<Value = Vec<f32>> {
    proptest::collection::vec(
        any::<f32>().prop_filter("finite", |f| f.is_finite()),
        1000..5000,
    )
}

proptest! {
    /// Property: Full metadata stack persists correctly
    #[test]
    fn prop_full_metadata_roundtrip(
        model_name in "[a-zA-Z][a-zA-Z0-9_-]{1,20}",
        description in "[a-zA-Z0-9 ]{1,50}",
        samples in 1usize..100000,
        duration_ms in 1u64..86400000,
        data in arb_model_data()
    ) {
        use tempfile::tempdir;

        #[derive(Debug, serde::Serialize, serde::Deserialize)]
        struct Model { weights: Vec<f32> }

        let model = Model { weights: data.clone() };
        let dir = tempdir().expect("tempdir");
        let path = dir.path().join("full_meta.apr");

        let mut options = SaveOptions::default()
            .with_name(&model_name)
            .with_description(&description);
        options.metadata.training = Some(TrainingInfo {
            samples: Some(samples),
            duration_ms: Some(duration_ms),
            source: Some("test_data".to_string()),
        });

        save(&model, ModelType::Custom, &path, options).expect("save");

        let info = inspect(&path).expect("inspect");
        prop_assert_eq!(info.metadata.model_name.as_deref(), Some(model_name.as_str()));
        prop_assert_eq!(info.metadata.description.as_deref(), Some(description.as_str()));

        let training = info.metadata.training.expect("training");
        prop_assert_eq!(training.samples, Some(samples));
        prop_assert_eq!(training.duration_ms, Some(duration_ms));

        let loaded: Model = load(&path, ModelType::Custom).expect("load");
        prop_assert_eq!(data.len(), loaded.weights.len());
    }

    /// Property: All model types roundtrip correctly
    #[test]
    fn prop_all_model_types_roundtrip(
        model_type in arb_model_type(),
        data in arb_model_data()
    ) {
        use tempfile::tempdir;

        #[derive(Debug, serde::Serialize, serde::Deserialize)]
        struct Model { weights: Vec<f32> }

        let model = Model { weights: data.clone() };
        let dir = tempdir().expect("tempdir");
        let path = dir.path().join("typed.apr");

        save(&model, model_type, &path, SaveOptions::default()).expect("save");

        let info = inspect(&path).expect("inspect");
        prop_assert_eq!(info.model_type, model_type);

        let loaded: Model = load(&path, model_type).expect("load");
        prop_assert_eq!(data.len(), loaded.weights.len());
    }

    /// Property: Large models roundtrip correctly (stress test)
    #[test]
    fn prop_large_model_roundtrip(data in arb_large_model_data()) {
        use tempfile::tempdir;

        #[derive(Debug, serde::Serialize, serde::Deserialize)]
        struct Model { weights: Vec<f32> }

        let model = Model { weights: data.clone() };
        let dir = tempdir().expect("tempdir");
        let path = dir.path().join("large.apr");

        save(&model, ModelType::Custom, &path, SaveOptions::default()).expect("save");

        let loaded: Model = load(&path, ModelType::Custom).expect("load");
        prop_assert_eq!(data.len(), loaded.weights.len());

        // Verify data integrity
        for (i, (orig, loaded_val)) in data.iter().zip(loaded.weights.iter()).enumerate() {
            prop_assert_eq!(
                orig.to_bits(),
                loaded_val.to_bits(),
                "Mismatch at index {}", i
            );
        }
    }

    /// Property: Distillation + License combined
    #[test]
    fn prop_distillation_with_license(data in arb_model_data()) {
        use tempfile::tempdir;

        #[derive(Debug, serde::Serialize, serde::Deserialize)]
        struct Model { weights: Vec<f32> }

        let model = Model { weights: data };
        let dir = tempdir().expect("tempdir");
        let path = dir.path().join("distilled_licensed.apr");

        let distill_info = DistillationInfo {
            method: DistillMethod::Standard,
            teacher: TeacherProvenance {
                hash: "a".repeat(64),
                signature: None,
                model_type: ModelType::Custom,
                param_count: 7_000_000_000,
                ensemble_teachers: None,
            },
            params: DistillationParams {
                temperature: 3.0,
                alpha: 0.7,
                beta: None,
                epochs: 10,
                final_loss: Some(0.42),
            },
            layer_mapping: None,
        };

        let license = LicenseInfo {
            uuid: "12345678-1234-4123-8123-123456789abc".to_string(),
            hash: "b".repeat(64),
            expiry: Some("2025-12-31".to_string()),
            seats: Some(10),
            licensee: Some("Test Corp".to_string()),
            tier: LicenseTier::Enterprise,
        };

        let options = SaveOptions::default()
            .with_distillation_info(distill_info)
            .with_license(license);

        save(&model, ModelType::Custom, &path, options).expect("save");

        let info = inspect(&path).expect("inspect");

        // Both features should be present
        prop_assert!(info.metadata.distillation_info.is_some());
        prop_assert!(info.metadata.license.is_some());
        prop_assert!(info.licensed);

        let restored_distill = info.metadata.distillation_info.expect("distillation");
        prop_assert!(matches!(restored_distill.method, DistillMethod::Standard));

        let restored_license = info.metadata.license.expect("license");
        prop_assert!(matches!(restored_license.tier, LicenseTier::Enterprise));
    }

    /// Property: Multiple saves to same path overwrites correctly
    #[test]
    fn prop_overwrite_preserves_latest(
        data1 in arb_model_data(),
        data2 in arb_model_data()
    ) {
        use tempfile::tempdir;

        #[derive(Debug, serde::Serialize, serde::Deserialize)]
        struct Model { weights: Vec<f32> }

        let dir = tempdir().expect("tempdir");
        let path = dir.path().join("overwrite.apr");

        // Save first model
        let model1 = Model { weights: data1 };
        save(&model1, ModelType::Custom, &path, SaveOptions::default()).expect("save1");

        // Save second model (overwrite)
        let model2 = Model { weights: data2.clone() };
        save(&model2, ModelType::Custom, &path, SaveOptions::default()).expect("save2");

        // Load should return second model
        let loaded: Model = load(&path, ModelType::Custom).expect("load");
        prop_assert_eq!(data2.len(), loaded.weights.len());
    }

    /// Property: File size scales with data size
    #[test]
    fn prop_file_size_scales_with_data(
        small_data in proptest::collection::vec(any::<f32>().prop_filter("finite", |f| f.is_finite()), 10..50),
        large_data in proptest::collection::vec(any::<f32>().prop_filter("finite", |f| f.is_finite()), 500..1000)
    ) {
        use tempfile::tempdir;

        #[derive(Debug, serde::Serialize, serde::Deserialize)]
        struct Model { weights: Vec<f32> }

        let dir = tempdir().expect("tempdir");

        let small_path = dir.path().join("small.apr");
        let model_small = Model { weights: small_data };
        save(&model_small, ModelType::Custom, &small_path, SaveOptions::default()).expect("save small");

        let large_path = dir.path().join("large.apr");
        let model_large = Model { weights: large_data };
        save(&model_large, ModelType::Custom, &large_path, SaveOptions::default()).expect("save large");

        let small_size = std::fs::metadata(&small_path).expect("meta").len();
        let large_size = std::fs::metadata(&large_path).expect("meta").len();

        prop_assert!(large_size > small_size, "Larger data should produce larger file");
    }
}