File size: 2,514 Bytes
47d601f
 
 
 
 
1ffa386
 
47d601f
 
 
 
 
 
 
 
 
 
efb3b2c
 
 
 
 
 
47d601f
 
 
 
 
efb3b2c
47d601f
 
 
efb3b2c
47d601f
 
1ffa386
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25e037f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
47d601f
 
 
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
"""
Test cases for the tokenizer loading
"""
import unittest

import pytest

from axolotl.utils.dict import DictDefault
from axolotl.utils.models import load_tokenizer


class TestTokenizers(unittest.TestCase):
    """
    test class for the load_tokenizer fn
    """

    def test_default_use_fast(self):
        cfg = DictDefault(
            {
                "tokenizer_config": "huggyllama/llama-7b",
            }
        )
        tokenizer = load_tokenizer(cfg)
        assert "Fast" in tokenizer.__class__.__name__

    def test_dont_use_fast(self):
        cfg = DictDefault(
            {
                "tokenizer_config": "huggyllama/llama-7b",
                "tokenizer_use_fast": False,
            }
        )
        tokenizer = load_tokenizer(cfg)
        assert "Fast" not in tokenizer.__class__.__name__

    def test_special_tokens_modules_to_save(self):
        # setting special_tokens to new token
        cfg = DictDefault(
            {
                "tokenizer_config": "huggyllama/llama-7b",
                "adapter": "lora",
                "special_tokens": {"bos_token": "[INST]"},
            }
        )
        with pytest.raises(
            ValueError,
            match=r".*Please set lora_modules_to_save*",
        ):
            load_tokenizer(cfg)

        # setting special_tokens but not changing from default
        cfg = DictDefault(
            {
                "tokenizer_config": "huggyllama/llama-7b",
                "adapter": "lora",
                "special_tokens": {"bos_token": "<s>"},
            }
        )
        load_tokenizer(cfg)

        # non-adapter setting special_tokens
        cfg = DictDefault(
            {
                "tokenizer_config": "huggyllama/llama-7b",
                "special_tokens": {"bos_token": "[INST]"},
            }
        )
        load_tokenizer(cfg)

    def test_add_additional_special_tokens(self):
        cfg = DictDefault(
            {
                "tokenizer_config": "huggyllama/llama-7b",
                "special_tokens": {"additional_special_tokens": ["<|im_start|>"]},
            }
        )
        tokenizer = load_tokenizer(cfg)
        self.assertEqual(tokenizer("<|im_start|>user")["input_ids"], [1, 32000, 1404])
        self.assertEqual(len(tokenizer), 32001)

        # ensure reloading the tokenizer again from cfg results in same vocab length
        tokenizer = load_tokenizer(cfg)
        self.assertEqual(len(tokenizer), 32001)


if __name__ == "__main__":
    unittest.main()