File size: 5,898 Bytes
d5a62f1 | 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 | import itertools
import os
import json
import re
from typing import List, Optional, Tuple
from transformers import PreTrainedTokenizer
class DNAKmerTokenizer(PreTrainedTokenizer):
def __init__(self, k, add_bos_token=True, add_eos_token=False, **kwargs):
self.k = k
self.add_bos_token = add_bos_token
self.add_eos_token = add_eos_token
self.special_tokens = [
"<oov>",
"<s>",
"</s>",
"<pad>",
"<mask>",
"<bog>",
"<eog>",
"<bok>",
"<eok>",
"<+>",
"<->",
"<cds>",
"<pseudo>",
"<tRNA>",
"<rRNA>",
"<ncRNA>",
"<miscRNA>",
"<mam>",
"<vrt>",
"<inv>",
"<pln>",
"<fng>",
"<prt>",
"<arc>",
"<bct>",
"<mit>",
"<plt>",
"<plm>",
"<vir>",
"<sp0>",
"<sp1>",
"<sp2>",
]
self.kmers = [
"".join(kmer) for kmer in itertools.product("ATCG", repeat=self.k)
]
self.vocab = {
token: i for i, token in enumerate(self.special_tokens + self.kmers)
}
self.ids_to_tokens = {v: k for k, v in self.vocab.items()}
self.special_token_pattern = re.compile(
"|".join(re.escape(token) for token in self.special_tokens)
)
self.dna_pattern = re.compile(f"[A-Z]{{{self.k}}}|[A-Z]+")
kwargs.setdefault("unk_token", "<oov>")
kwargs.setdefault("bos_token", "<s>")
kwargs.setdefault("eos_token", "</s>")
kwargs.setdefault("pad_token", "<pad>")
kwargs.setdefault("mask_token", "<mask>")
super().__init__(**kwargs)
@property
def vocab_size(self):
return len(self.vocab)
def get_vocab(self):
return dict(self.vocab)
def _tokenize(self, text, **kwargs) -> List[str]:
tokens = []
pos = 0
while pos < len(text):
special_match = self.special_token_pattern.match(text, pos)
if special_match:
tokens.append(special_match.group())
pos = special_match.end()
else:
dna_match = self.dna_pattern.match(text, pos)
if dna_match:
dna_seq = dna_match.group()
tokens.append(dna_seq)
pos = dna_match.end()
else:
tokens.append(text[pos])
pos += 1
return tokens
def _convert_token_to_id(self, token: str) -> int:
return self.vocab.get(token, self.vocab["<oov>"])
def _convert_id_to_token(self, index: int) -> str:
return self.ids_to_tokens.get(index, "<oov>")
def convert_tokens_to_string(self, tokens: List[str]) -> str:
return "".join(tokens)
def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
bos = [self.bos_token_id] if self.add_bos_token else []
eos = [self.eos_token_id] if self.add_eos_token else []
if token_ids_1 is None:
return bos + token_ids_0 + eos
# Dual sequence case: bos + token_ids_0 + bos + token_ids_1 + eos
return bos + token_ids_0 + bos + token_ids_1 + eos
def get_special_tokens_mask(
self, token_ids_0, token_ids_1=None, already_has_special_tokens=False
):
if already_has_special_tokens:
return super().get_special_tokens_mask(
token_ids_0, token_ids_1, already_has_special_tokens=True
)
bos_mask = [1] if self.add_bos_token else []
eos_mask = [1] if self.add_eos_token else []
if token_ids_1 is None:
return bos_mask + ([0] * len(token_ids_0)) + eos_mask
# Dual sequence case: bos + token_ids_0 + bos + token_ids_1 + eos
return bos_mask + ([0] * len(token_ids_0)) + bos_mask + ([0] * len(token_ids_1)) + eos_mask
def prepare_for_model(self, *args, **kwargs):
encoding = super().prepare_for_model(*args, **kwargs)
if "token_type_ids" in encoding:
del encoding["token_type_ids"]
return encoding
def save_vocabulary(
self, save_directory: str, filename_prefix: Optional[str] = None
) -> Tuple[str]:
if not os.path.exists(save_directory):
os.makedirs(save_directory)
vocab_file = os.path.join(
save_directory,
(filename_prefix + "-" if filename_prefix else "") + "vocab.txt",
)
with open(vocab_file, "w", encoding="utf-8") as f:
for token, idx in sorted(self.vocab.items(), key=lambda x: x[1]):
f.write(f"{token} {idx}\n")
return (vocab_file,)
def save_pretrained(self, save_directory: str, **kwargs):
vocab_files = super().save_pretrained(save_directory, **kwargs)
tokenizer_config_path = os.path.join(save_directory, "tokenizer_config.json")
# Read existing config or create new one
if os.path.exists(tokenizer_config_path):
with open(tokenizer_config_path, "r", encoding="utf-8") as f:
config = json.load(f)
else:
config = {}
# Add auto_map configuration
config.update({
"auto_map": {
"AutoTokenizer": [
"tokenizer.DNAKmerTokenizer",
None
]
},
"k": self.k,
"add_bos_token": self.add_bos_token,
"add_eos_token": self.add_eos_token,
})
# Save config
with open(tokenizer_config_path, "w", encoding="utf-8") as f:
json.dump(config, f, ensure_ascii=False, indent=2)
return vocab_files
|