File size: 1,317 Bytes
d458774
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from datasets import load_dataset
from tokenizers import trainers, Tokenizer, normalizers, ByteLevelBPETokenizer
from tokenizers import AddedToken, Regex, Tokenizer, decoders, normalizers, pre_tokenizers, trainers

model_dir = "./"  # ${MODEL_DIR}

# load dataset
dataset = load_dataset("mc4", "bn", split="train", streaming=True)

Instantiate tokenizer
tokenizer = ByteLevelBPETokenizer()

# Instantiate normalizer
tokenizer.normalizer = normalizers.Sequence(
    [
        normalizers.Nmt(),
        normalizers.NFKC(),
        normalizers.Replace(Regex(" {2,}"), " "),
        normalizers.Replace("\u09e4", "\u0964"),
        normalizers.Replace("\u09e5", "\u0965"),
        normalizers.Replace("\u007c", "\u0964"),
        normalizers.Replace("\u09f7", "\u0964"),
        normalizers.Replace(Regex(r"(?<=[\u0980-\u09ff]):"), "\u0983"),
        normalizers.Lowercase(),
    ]
)

def batch_iterator(batch_size=1000):
    for i in range(0, len(dataset), batch_size):
        yield dataset[i: i + batch_size]["text"]

# Customized training
tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
    "<|endoftext|>",
])

# Save files to disk
tokenizer.save(f"{model_dir}/tokenizer.json")

# f = open("demofile3.txt", "w")
# f.write(next(iter(dataset))['text'])
# f.close()