realizar 0.8.5

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


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

    // ========================================================================
    // KVCache tests
    // ========================================================================

    #[test]
    fn test_kv_cache_new() {
        let cache = KVCache::new(4, 128, 64).expect("cache");
        assert_eq!(cache.num_layers(), 4);
        assert_eq!(cache.max_seq_len(), 128);
        assert_eq!(cache.head_dim(), 64);
        assert_eq!(cache.current_pos(), 0);
    }

    #[test]
    fn test_kv_cache_new_zero_layers_error() {
        let result = KVCache::new(0, 128, 64);
        assert!(result.is_err());
        assert!(result
            .unwrap_err()
            .to_string()
            .contains("num_layers must be > 0"));
    }

    #[test]
    fn test_kv_cache_new_zero_seq_len_error() {
        let result = KVCache::new(4, 0, 64);
        assert!(result.is_err());
        assert!(result
            .unwrap_err()
            .to_string()
            .contains("max_seq_len must be > 0"));
    }

    #[test]
    fn test_kv_cache_new_zero_head_dim_error() {
        let result = KVCache::new(4, 128, 0);
        assert!(result.is_err());
        assert!(result
            .unwrap_err()
            .to_string()
            .contains("head_dim must be > 0"));
    }

    #[test]
    fn test_kv_cache_update_and_advance() {
        let mut cache = KVCache::new(2, 4, 3).expect("cache");
        let key = Tensor::from_vec(vec![3], vec![1.0, 2.0, 3.0]).expect("key");
        let value = Tensor::from_vec(vec![3], vec![4.0, 5.0, 6.0]).expect("value");

        cache.update(0, &key, &value).expect("update");
        cache.advance();

        assert_eq!(cache.current_pos(), 1);
    }

    #[test]
    fn test_kv_cache_update_layer_out_of_bounds() {
        let mut cache = KVCache::new(2, 4, 3).expect("cache");
        let key = Tensor::from_vec(vec![3], vec![1.0, 2.0, 3.0]).expect("key");
        let value = Tensor::from_vec(vec![3], vec![4.0, 5.0, 6.0]).expect("value");

        let result = cache.update(5, &key, &value);
        assert!(result.is_err());
        assert!(result.unwrap_err().to_string().contains("out of bounds"));
    }

    #[test]
    fn test_kv_cache_update_key_size_mismatch() {
        let mut cache = KVCache::new(2, 4, 3).expect("cache");
        let key = Tensor::from_vec(vec![2], vec![1.0, 2.0]).expect("key"); // Wrong size
        let value = Tensor::from_vec(vec![3], vec![4.0, 5.0, 6.0]).expect("value");

        let result = cache.update(0, &key, &value);
        assert!(result.is_err());
        assert!(result.unwrap_err().to_string().contains("Key size"));
    }

    #[test]
    fn test_kv_cache_update_value_size_mismatch() {
        let mut cache = KVCache::new(2, 4, 3).expect("cache");
        let key = Tensor::from_vec(vec![3], vec![1.0, 2.0, 3.0]).expect("key");
        let value = Tensor::from_vec(vec![2], vec![4.0, 5.0]).expect("value"); // Wrong size

        let result = cache.update(0, &key, &value);
        assert!(result.is_err());
        assert!(result.unwrap_err().to_string().contains("Value size"));
    }

    #[test]
    fn test_kv_cache_get_key_empty() {
        let cache = KVCache::new(2, 4, 3).expect("cache");
        let key = cache.get_key(0).expect("key");
        assert_eq!(key.shape(), &[1, 3]); // Returns [1, head_dim] when empty
    }

    #[test]
    fn test_kv_cache_get_value_empty() {
        let cache = KVCache::new(2, 4, 3).expect("cache");
        let value = cache.get_value(0).expect("value");
        assert_eq!(value.shape(), &[1, 3]);
    }

    #[test]
    fn test_kv_cache_get_key_after_update() {
        let mut cache = KVCache::new(2, 4, 3).expect("cache");
        let key = Tensor::from_vec(vec![3], vec![1.0, 2.0, 3.0]).expect("key");
        let value = Tensor::from_vec(vec![3], vec![4.0, 5.0, 6.0]).expect("value");

        cache.update(0, &key, &value).expect("update");
        cache.advance();

        let cached_key = cache.get_key(0).expect("cached_key");
        assert_eq!(cached_key.shape(), &[1, 3]);
        let data = cached_key.data();
        assert!((data[0] - 1.0).abs() < 1e-6);
        assert!((data[1] - 2.0).abs() < 1e-6);
        assert!((data[2] - 3.0).abs() < 1e-6);
    }

    #[test]
    fn test_kv_cache_get_value_after_update() {
        let mut cache = KVCache::new(2, 4, 3).expect("cache");
        let key = Tensor::from_vec(vec![3], vec![1.0, 2.0, 3.0]).expect("key");
        let value = Tensor::from_vec(vec![3], vec![4.0, 5.0, 6.0]).expect("value");

        cache.update(0, &key, &value).expect("update");
        cache.advance();

        let cached_value = cache.get_value(0).expect("cached_value");
        let data = cached_value.data();
        assert!((data[0] - 4.0).abs() < 1e-6);
        assert!((data[1] - 5.0).abs() < 1e-6);
        assert!((data[2] - 6.0).abs() < 1e-6);
    }

    #[test]
    fn test_kv_cache_get_key_layer_out_of_bounds() {
        let cache = KVCache::new(2, 4, 3).expect("cache");
        let result = cache.get_key(5);
        assert!(result.is_err());
    }

    #[test]
    fn test_kv_cache_get_value_layer_out_of_bounds() {
        let cache = KVCache::new(2, 4, 3).expect("cache");
        let result = cache.get_value(5);
        assert!(result.is_err());
    }

    #[test]
    fn test_kv_cache_clear() {
        let mut cache = KVCache::new(2, 4, 3).expect("cache");
        let key = Tensor::from_vec(vec![3], vec![1.0, 2.0, 3.0]).expect("key");
        let value = Tensor::from_vec(vec![3], vec![4.0, 5.0, 6.0]).expect("value");

        cache.update(0, &key, &value).expect("update");
        cache.advance();
        assert_eq!(cache.current_pos(), 1);

        cache.clear();
        assert_eq!(cache.current_pos(), 0);
    }

    #[test]
    fn test_kv_cache_cache_full_error() {
        let mut cache = KVCache::new(1, 2, 2).expect("cache");
        let key = Tensor::from_vec(vec![2], vec![1.0, 2.0]).expect("key");
        let value = Tensor::from_vec(vec![2], vec![3.0, 4.0]).expect("value");

        // Fill cache
        cache.update(0, &key, &value).expect("update");
        cache.advance();
        cache.update(0, &key, &value).expect("update");
        cache.advance();

        // Cache is full now
        let result = cache.update(0, &key, &value);
        assert!(result.is_err());
        assert!(result.unwrap_err().to_string().contains("Cache full"));
    }

    #[test]
    fn test_kv_cache_multiple_layers() {
        let mut cache = KVCache::new(3, 4, 2).expect("cache");
        let key0 = Tensor::from_vec(vec![2], vec![1.0, 2.0]).expect("key0");
        let val0 = Tensor::from_vec(vec![2], vec![3.0, 4.0]).expect("val0");
        let key1 = Tensor::from_vec(vec![2], vec![5.0, 6.0]).expect("key1");
        let val1 = Tensor::from_vec(vec![2], vec![7.0, 8.0]).expect("val1");
        let key2 = Tensor::from_vec(vec![2], vec![9.0, 10.0]).expect("key2");
        let val2 = Tensor::from_vec(vec![2], vec![11.0, 12.0]).expect("val2");

        cache.update(0, &key0, &val0).expect("update");
        cache.update(1, &key1, &val1).expect("update");
        cache.update(2, &key2, &val2).expect("update");
        cache.advance();

        let k0 = cache.get_key(0).expect("k0");
        let k1 = cache.get_key(1).expect("k1");
        let k2 = cache.get_key(2).expect("k2");

        assert!((k0.data()[0] - 1.0).abs() < 1e-6);
        assert!((k1.data()[0] - 5.0).abs() < 1e-6);
        assert!((k2.data()[0] - 9.0).abs() < 1e-6);
    }

    #[test]
    fn test_kv_cache_clone() {
        let cache = KVCache::new(2, 4, 3).expect("cache");
        let cloned = cache.clone();
        assert_eq!(cloned.num_layers(), 2);
        assert_eq!(cloned.max_seq_len(), 4);
    }

    #[test]
    fn test_kv_cache_debug() {
        let cache = KVCache::new(2, 4, 3).expect("cache");
        let debug_str = format!("{:?}", cache);
        assert!(debug_str.contains("KVCache"));
    }

    // ========================================================================
    // ModelConfig tests
    // ========================================================================

    #[test]
    fn test_model_config_new() {
        let config = ModelConfig {
            vocab_size: 32000,
            hidden_dim: 4096,
            num_layers: 32,
            num_heads: 32,
            intermediate_dim: 11008,
            eps: 1e-5,
        };
        assert_eq!(config.vocab_size, 32000);
        assert_eq!(config.hidden_dim, 4096);
        assert_eq!(config.num_layers, 32);
    }

    #[test]
    fn test_model_config_clone() {
        let config = ModelConfig {
            vocab_size: 1000,
            hidden_dim: 256,
            num_layers: 4,
            num_heads: 4,
            intermediate_dim: 1024,
            eps: 1e-5,
        };
        let cloned = config.clone();
        assert_eq!(cloned.vocab_size, 1000);
    }

    #[test]
    fn test_model_config_debug() {
        let config = ModelConfig {
            vocab_size: 1000,
            hidden_dim: 256,
            num_layers: 4,
            num_heads: 4,
            intermediate_dim: 1024,
            eps: 1e-5,
        };
        let debug_str = format!("{:?}", config);
        assert!(debug_str.contains("ModelConfig"));
        assert!(debug_str.contains("vocab_size: 1000"));
    }

    // ========================================================================
    // Embedding tests (basic structure)
    // ========================================================================

    #[test]
    fn test_embedding_new() {
        let embedding = Embedding::new(100, 64).expect("embedding");
        assert_eq!(embedding.vocab_size(), 100);
        assert_eq!(embedding.embed_dim(), 64);
    }

    #[test]
    fn test_embedding_new_zero_vocab_error() {
        let result = Embedding::new(0, 64);
        assert!(result.is_err());
    }

    #[test]
    fn test_embedding_new_zero_dim_error() {
        let result = Embedding::new(100, 0);
        assert!(result.is_err());
    }

    #[test]
    fn test_embedding_forward() {
        let embedding = Embedding::new(100, 4).expect("embedding");
        let indices: &[usize] = &[0, 1];
        let result = embedding.forward(indices).expect("result");
        // Shape should be [2, 4] for 2 tokens with dim 4
        assert_eq!(result.shape(), &[2, 4]);
    }

    #[test]
    fn test_embedding_forward_single_token() {
        let embedding = Embedding::new(50, 8).expect("embedding");
        let indices: &[usize] = &[5];
        let result = embedding.forward(indices).expect("result");
        assert_eq!(result.shape(), &[1, 8]);
    }

    #[test]
    fn test_embedding_forward_out_of_bounds() {
        let embedding = Embedding::new(10, 4).expect("embedding");
        let indices: &[usize] = &[100]; // Out of vocab
        let result = embedding.forward(indices);
        assert!(result.is_err());
    }

    #[test]
    fn test_embedding_clone() {
        let embedding = Embedding::new(100, 64).expect("embedding");
        let cloned = embedding.clone();
        assert_eq!(cloned.vocab_size(), 100);
    }
}