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


// ============================================================================
// Tests for APR Tokenizer (PMAT-802: T-COV-95)
// ============================================================================

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

    // -------------------------------------------------------------------------
    // SimpleTokenizer Tests
    // -------------------------------------------------------------------------

    #[test]
    fn test_simple_tokenizer_new() {
        let vocab = vec!["hello".to_string(), "world".to_string()];
        let tokenizer = SimpleTokenizer::new(vocab, Some(0), Some(1));
        assert_eq!(tokenizer.vocab_size(), 2);
        assert_eq!(tokenizer.bos_token_id, Some(0));
        assert_eq!(tokenizer.eos_token_id, Some(1));
    }

    #[test]
    fn test_simple_tokenizer_vocab_size() {
        let vocab = vec!["a".to_string(), "b".to_string(), "c".to_string()];
        let tokenizer = SimpleTokenizer::new(vocab, None, None);
        assert_eq!(tokenizer.vocab_size(), 3);
    }

    #[test]
    fn test_simple_tokenizer_is_eos() {
        let vocab = vec!["<s>".to_string(), "</s>".to_string()];
        let tokenizer = SimpleTokenizer::new(vocab, Some(0), Some(1));
        assert!(tokenizer.is_eos(1));
        assert!(!tokenizer.is_eos(0));
        assert!(!tokenizer.is_eos(2)); // Out of range
    }

    #[test]
    fn test_simple_tokenizer_is_eos_none() {
        let vocab = vec!["hello".to_string()];
        let tokenizer = SimpleTokenizer::new(vocab, None, None);
        assert!(!tokenizer.is_eos(0));
        assert!(!tokenizer.is_eos(1));
    }

    #[test]
    fn test_simple_tokenizer_is_bos() {
        let vocab = vec!["<s>".to_string(), "</s>".to_string()];
        let tokenizer = SimpleTokenizer::new(vocab, Some(0), Some(1));
        assert!(tokenizer.is_bos(0));
        assert!(!tokenizer.is_bos(1));
    }

    #[test]
    fn test_simple_tokenizer_is_bos_none() {
        let vocab = vec!["hello".to_string()];
        let tokenizer = SimpleTokenizer::new(vocab, None, None);
        assert!(!tokenizer.is_bos(0));
    }

    // -------------------------------------------------------------------------
    // BpeTokenizer Tests
    // -------------------------------------------------------------------------

    #[test]
    fn test_bpe_tokenizer_encode_empty() {
        let tokenizer = BpeTokenizer {
            token_to_id: HashMap::new(),
            id_to_token: vec![],
            merge_rules: vec![],
            bos_id: None,
            eos_id: None,
            special_tokens: HashMap::new(),
        };
        let result = tokenizer.encode("");
        assert!(result.is_empty());
    }

    #[test]
    fn test_bpe_tokenizer_encode_simple() {
        let mut token_to_id = HashMap::new();
        token_to_id.insert("h".to_string(), 0);
        token_to_id.insert("i".to_string(), 1);

        let tokenizer = BpeTokenizer {
            token_to_id,
            id_to_token: vec!["h".to_string(), "i".to_string()],
            merge_rules: vec![],
            bos_id: None,
            eos_id: None,
            special_tokens: HashMap::new(),
        };
        let result = tokenizer.encode("hi");
        assert_eq!(result, vec![0, 1]);
    }

    #[test]
    fn test_bpe_tokenizer_encode_with_space() {
        let mut token_to_id = HashMap::new();
        token_to_id.insert("Ġ".to_string(), 0); // Space
        token_to_id.insert("a".to_string(), 1);

        let tokenizer = BpeTokenizer {
            token_to_id,
            id_to_token: vec!["Ġ".to_string(), "a".to_string()],
            merge_rules: vec![],
            bos_id: None,
            eos_id: None,
            special_tokens: HashMap::new(),
        };
        let result = tokenizer.encode(" a");
        assert_eq!(result, vec![0, 1]);
    }

    // -------------------------------------------------------------------------
    // bpe_encode Tests
    // -------------------------------------------------------------------------

    #[test]
    fn test_bpe_encode_empty() {
        let vocab: HashMap<String, u32> = HashMap::new();
        let merges: Vec<(String, String)> = vec![];
        let special: HashMap<String, u32> = HashMap::new();
        let result = bpe_encode("", &vocab, &merges, &special);
        assert!(result.is_empty());
    }

    #[test]
    fn test_bpe_encode_simple_chars() {
        let mut vocab = HashMap::new();
        vocab.insert("a".to_string(), 0);
        vocab.insert("b".to_string(), 1);
        vocab.insert("c".to_string(), 2);
        let special: HashMap<String, u32> = HashMap::new();

        let result = bpe_encode("abc", &vocab, &[], &special);
        assert_eq!(result, vec![0, 1, 2]);
    }

    #[test]
    fn test_bpe_encode_with_merge() {
        let mut vocab = HashMap::new();
        vocab.insert("a".to_string(), 0);
        vocab.insert("b".to_string(), 1);
        vocab.insert("ab".to_string(), 2);
        let special: HashMap<String, u32> = HashMap::new();

        let merges = vec![("a".to_string(), "b".to_string())];
        let result = bpe_encode("ab", &vocab, &merges, &special);
        assert_eq!(result, vec![2]); // "ab" merged
    }

    #[test]
    fn test_bpe_encode_space_handling() {
        let mut vocab = HashMap::new();
        vocab.insert("Ġ".to_string(), 0); // Space becomes Ġ
        let special: HashMap<String, u32> = HashMap::new();

        let result = bpe_encode(" ", &vocab, &[], &special);
        assert_eq!(result, vec![0]);
    }

    #[test]
    fn test_bpe_encode_newline_handling() {
        let mut vocab = HashMap::new();
        vocab.insert("Ċ".to_string(), 0); // Newline becomes Ċ
        let special: HashMap<String, u32> = HashMap::new();

        let result = bpe_encode("\n", &vocab, &[], &special);
        assert_eq!(result, vec![0]);
    }

    #[test]
    fn test_bpe_encode_unknown_tokens() {
        let vocab: HashMap<String, u32> = HashMap::new();
        let special: HashMap<String, u32> = HashMap::new();
        let result = bpe_encode("xyz", &vocab, &[], &special);
        assert!(result.is_empty()); // Unknown tokens filtered out
    }

    // -------------------------------------------------------------------------
    // byte_to_bpe_char Tests
    // -------------------------------------------------------------------------

    #[test]
    fn test_byte_to_bpe_char_space() {
        assert_eq!(byte_to_bpe_char(b' '), "Ġ");
    }

    #[test]
    fn test_byte_to_bpe_char_newline() {
        assert_eq!(byte_to_bpe_char(b'\n'), "Ċ");
    }

    #[test]
    fn test_byte_to_bpe_char_tab() {
        assert_eq!(byte_to_bpe_char(b'\t'), "ĉ");
    }

    #[test]
    fn test_byte_to_bpe_char_ascii() {
        assert_eq!(byte_to_bpe_char(b'a'), "a");
        assert_eq!(byte_to_bpe_char(b'Z'), "Z");
        assert_eq!(byte_to_bpe_char(b'0'), "0");
        assert_eq!(byte_to_bpe_char(b'!'), "!");
    }

    #[test]
    fn test_byte_to_bpe_char_non_printable() {
        // Non-printable bytes get hex encoding
        assert_eq!(byte_to_bpe_char(0x00), "<0x00>");
        assert_eq!(byte_to_bpe_char(0x7F), "<0x7F>");
        assert_eq!(byte_to_bpe_char(0xFF), "<0xFF>");
    }

    // -------------------------------------------------------------------------
    // GH-189: Special Token Tests
    // -------------------------------------------------------------------------

    #[test]
    fn test_bpe_tokenizer_with_special_tokens() {
        // GH-189: Verify special tokens are tokenized atomically
        let mut token_to_id = HashMap::new();
        token_to_id.insert("<|im_start|>".to_string(), 100);
        token_to_id.insert("<|im_end|>".to_string(), 101);
        token_to_id.insert("h".to_string(), 0);
        token_to_id.insert("i".to_string(), 1);

        let mut special_tokens = HashMap::new();
        special_tokens.insert("<|im_start|>".to_string(), 100);
        special_tokens.insert("<|im_end|>".to_string(), 101);

        let tokenizer = BpeTokenizer {
            token_to_id,
            id_to_token: vec![
                "h".to_string(),
                "i".to_string(),
                // ... padding ...
            ],
            merge_rules: vec![],
            bos_id: None,
            eos_id: None,
            special_tokens,
        };

        // Test that special tokens are kept atomic
        let result = tokenizer.encode("<|im_start|>hi<|im_end|>");
        assert_eq!(result.len(), 4); // <|im_start|>, h, i, <|im_end|>
        assert_eq!(result[0], 100); // <|im_start|>
        assert_eq!(result[3], 101); // <|im_end|>
    }

    #[test]
    fn test_bpe_encode_with_special_tokens() {
        // GH-189: Test bpe_encode directly with special tokens
        let mut vocab = HashMap::new();
        vocab.insert("<|im_start|>".to_string(), 100);
        vocab.insert("<|im_end|>".to_string(), 101);
        vocab.insert("H".to_string(), 0);
        vocab.insert("i".to_string(), 1);

        let mut special = HashMap::new();
        special.insert("<|im_start|>".to_string(), 100);
        special.insert("<|im_end|>".to_string(), 101);

        let result = bpe_encode("<|im_start|>Hi<|im_end|>", &vocab, &[], &special);
        // Should be: [100, 0, 1, 101] = <|im_start|>, H, i, <|im_end|>
        assert!(result.contains(&100)); // <|im_start|>
        assert!(result.contains(&101)); // <|im_end|>
        // Verify order: special token should be first
        assert_eq!(result[0], 100);
    }

    #[test]
    fn test_split_by_special_tokens() {
        let mut special_tokens = HashMap::new();
        special_tokens.insert("<|im_start|>".to_string(), 100);
        special_tokens.insert("<|im_end|>".to_string(), 101);

        let text = "<|im_start|>user\nHello<|im_end|>";
        let segments = split_by_special_tokens(text, &special_tokens);

        // Should have 3 segments: Special(100), Regular("user\nHello"), Special(101)
        assert_eq!(segments.len(), 3);
        match &segments[0] {
            TextSegment::Special(id) => assert_eq!(*id, 100),
            TextSegment::Regular(_) => panic!("Expected Special segment"),
        }
        match &segments[1] {
            TextSegment::Regular(s) => assert_eq!(s, "user\nHello"),
            TextSegment::Special(_) => panic!("Expected Regular segment"),
        }
        match &segments[2] {
            TextSegment::Special(id) => assert_eq!(*id, 101),
            TextSegment::Regular(_) => panic!("Expected Special segment"),
        }
    }
}