Delete fold_4/best_macro_auprc_model/tokenization_ntv3.py with huggingface_hub
Browse files
fold_4/best_macro_auprc_model/tokenization_ntv3.py
DELETED
|
@@ -1,234 +0,0 @@
|
|
| 1 |
-
from __future__ import annotations
|
| 2 |
-
|
| 3 |
-
import json
|
| 4 |
-
import os
|
| 5 |
-
from typing import Any, Dict, Iterable, List, Optional, Sequence, Tuple
|
| 6 |
-
|
| 7 |
-
from transformers import PreTrainedTokenizer
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
def _as_str(token: Any) -> Optional[str]:
|
| 11 |
-
"""Return the plain string representation for tokens/AddedTokens."""
|
| 12 |
-
if token is None:
|
| 13 |
-
return None
|
| 14 |
-
if isinstance(token, str):
|
| 15 |
-
return token
|
| 16 |
-
return getattr(token, "content", str(token))
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
class _BaseNTv3Tokenizer(PreTrainedTokenizer):
|
| 20 |
-
"""Shared convenience implementation for the NTv3 tokenizers."""
|
| 21 |
-
|
| 22 |
-
vocab_files_names = {"vocab_file": "vocab.json"}
|
| 23 |
-
model_input_names = ["input_ids"]
|
| 24 |
-
|
| 25 |
-
def __init__(
|
| 26 |
-
self,
|
| 27 |
-
*,
|
| 28 |
-
vocab_file: Optional[str],
|
| 29 |
-
unk_token: str,
|
| 30 |
-
pad_token: str,
|
| 31 |
-
mask_token: str,
|
| 32 |
-
cls_token: str,
|
| 33 |
-
eos_token: str,
|
| 34 |
-
bos_token: str,
|
| 35 |
-
default_tokens: Sequence[str],
|
| 36 |
-
standard_tokens: Iterable[str],
|
| 37 |
-
**kwargs: Any,
|
| 38 |
-
) -> None:
|
| 39 |
-
if vocab_file is None:
|
| 40 |
-
token_to_id = {tok: idx for idx, tok in enumerate(default_tokens)}
|
| 41 |
-
else:
|
| 42 |
-
if not os.path.isfile(vocab_file):
|
| 43 |
-
raise ValueError(f"Can't find a vocab file at path '{vocab_file}'.")
|
| 44 |
-
with open(vocab_file, "r", encoding="utf-8") as handle:
|
| 45 |
-
loaded = json.load(handle)
|
| 46 |
-
if isinstance(loaded, dict):
|
| 47 |
-
token_to_id = {str(tok): int(idx) for tok, idx in loaded.items()}
|
| 48 |
-
else:
|
| 49 |
-
token_to_id = {str(tok): idx for idx, tok in enumerate(loaded)}
|
| 50 |
-
|
| 51 |
-
self._token_to_id: Dict[str, int] = dict(token_to_id)
|
| 52 |
-
self._id_to_token: Dict[int, str] = {
|
| 53 |
-
int(idx): tok for tok, idx in self._token_to_id.items()
|
| 54 |
-
}
|
| 55 |
-
|
| 56 |
-
super().__init__(
|
| 57 |
-
unk_token=unk_token,
|
| 58 |
-
pad_token=pad_token,
|
| 59 |
-
mask_token=mask_token,
|
| 60 |
-
cls_token=cls_token,
|
| 61 |
-
eos_token=eos_token,
|
| 62 |
-
bos_token=bos_token,
|
| 63 |
-
**kwargs,
|
| 64 |
-
)
|
| 65 |
-
|
| 66 |
-
self._unk_token_str = _as_str(self.unk_token)
|
| 67 |
-
self._pad_token_str = _as_str(self.pad_token)
|
| 68 |
-
self._mask_token_str = _as_str(self.mask_token)
|
| 69 |
-
self._cls_token_str = _as_str(self.cls_token)
|
| 70 |
-
self._eos_token_str = _as_str(self.eos_token)
|
| 71 |
-
self._bos_token_str = _as_str(self.bos_token)
|
| 72 |
-
|
| 73 |
-
self._special_token_strings = {
|
| 74 |
-
tok
|
| 75 |
-
for tok in (
|
| 76 |
-
self._unk_token_str,
|
| 77 |
-
self._pad_token_str,
|
| 78 |
-
self._mask_token_str,
|
| 79 |
-
self._cls_token_str,
|
| 80 |
-
self._eos_token_str,
|
| 81 |
-
self._bos_token_str,
|
| 82 |
-
)
|
| 83 |
-
if tok is not None
|
| 84 |
-
}
|
| 85 |
-
|
| 86 |
-
self._standard_tokens = {str(tok) for tok in standard_tokens}
|
| 87 |
-
self._unk_literal = (
|
| 88 |
-
self._unk_token_str if self._unk_token_str in self._token_to_id else "<unk>"
|
| 89 |
-
)
|
| 90 |
-
|
| 91 |
-
# ------------------------------------------------------------------
|
| 92 |
-
# Hugging Face required interface
|
| 93 |
-
# ------------------------------------------------------------------
|
| 94 |
-
def get_vocab(self) -> Dict[str, int]:
|
| 95 |
-
return dict(self._token_to_id)
|
| 96 |
-
|
| 97 |
-
@property
|
| 98 |
-
def vocab_size(self) -> int:
|
| 99 |
-
return len(self._token_to_id)
|
| 100 |
-
|
| 101 |
-
# Sub-classes implement `_tokenize`.
|
| 102 |
-
|
| 103 |
-
def _convert_token_to_id(self, token: str) -> int:
|
| 104 |
-
if self._unk_literal in self._token_to_id:
|
| 105 |
-
return self._token_to_id.get(token, self._token_to_id[self._unk_literal])
|
| 106 |
-
return self._token_to_id.get(token, 0)
|
| 107 |
-
|
| 108 |
-
def _convert_id_to_token(self, index: int) -> str:
|
| 109 |
-
idx = int(index)
|
| 110 |
-
return self._id_to_token.get(idx, self._unk_literal)
|
| 111 |
-
|
| 112 |
-
def build_inputs_with_special_tokens(
|
| 113 |
-
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
|
| 114 |
-
) -> List[int]:
|
| 115 |
-
if token_ids_1 is None:
|
| 116 |
-
return token_ids_0
|
| 117 |
-
return token_ids_0 + token_ids_1
|
| 118 |
-
|
| 119 |
-
def get_special_tokens_mask(
|
| 120 |
-
self,
|
| 121 |
-
token_ids_0: List[int],
|
| 122 |
-
token_ids_1: Optional[List[int]] = None,
|
| 123 |
-
already_has_special_tokens: bool = False,
|
| 124 |
-
) -> List[int]:
|
| 125 |
-
length = len(token_ids_0) + (len(token_ids_1) if token_ids_1 else 0)
|
| 126 |
-
if not already_has_special_tokens:
|
| 127 |
-
return [0] * length
|
| 128 |
-
return [
|
| 129 |
-
1 if self._convert_id_to_token(idx) in self._special_token_strings else 0
|
| 130 |
-
for idx in token_ids_0
|
| 131 |
-
] + (
|
| 132 |
-
[
|
| 133 |
-
1
|
| 134 |
-
if self._convert_id_to_token(idx) in self._special_token_strings
|
| 135 |
-
else 0
|
| 136 |
-
for idx in (token_ids_1 or [])
|
| 137 |
-
]
|
| 138 |
-
)
|
| 139 |
-
|
| 140 |
-
def create_token_type_ids_from_sequences(
|
| 141 |
-
self, token_ids_0: List[int], token_ids_1: Optional[List[int]] = None
|
| 142 |
-
) -> List[int]:
|
| 143 |
-
return [0] * (len(token_ids_0) + (len(token_ids_1) if token_ids_1 else 0))
|
| 144 |
-
|
| 145 |
-
# ------------------------------------------------------------------
|
| 146 |
-
# Persistence helpers
|
| 147 |
-
# ------------------------------------------------------------------
|
| 148 |
-
def save_vocabulary(
|
| 149 |
-
self, save_directory: str, filename_prefix: Optional[str] = None
|
| 150 |
-
) -> Tuple[str]:
|
| 151 |
-
os.makedirs(save_directory, exist_ok=True)
|
| 152 |
-
filename = self.vocab_files_names["vocab_file"]
|
| 153 |
-
if filename_prefix:
|
| 154 |
-
filename = f"{filename_prefix}-{filename}"
|
| 155 |
-
path = os.path.join(save_directory, filename)
|
| 156 |
-
with open(path, "w", encoding="utf-8") as handle:
|
| 157 |
-
json.dump(self._token_to_id, handle, ensure_ascii=False, indent=2)
|
| 158 |
-
return (path,)
|
| 159 |
-
|
| 160 |
-
# ------------------------------------------------------------------
|
| 161 |
-
# Optional niceties
|
| 162 |
-
# ------------------------------------------------------------------
|
| 163 |
-
def convert_tokens_to_string(self, tokens: List[str]) -> str:
|
| 164 |
-
return "".join(tokens)
|
| 165 |
-
|
| 166 |
-
def _decode(
|
| 167 |
-
self,
|
| 168 |
-
token_ids: List[int],
|
| 169 |
-
skip_special_tokens: bool = False,
|
| 170 |
-
clean_up_tokenization_spaces: Optional[bool] = None,
|
| 171 |
-
spaces_between_special_tokens: bool = True,
|
| 172 |
-
**kwargs: Any,
|
| 173 |
-
) -> str:
|
| 174 |
-
output_tokens: List[str] = []
|
| 175 |
-
for idx in token_ids:
|
| 176 |
-
token = self._convert_id_to_token(idx)
|
| 177 |
-
if skip_special_tokens and token in self._special_token_strings:
|
| 178 |
-
continue
|
| 179 |
-
output_tokens.append(token)
|
| 180 |
-
return self.convert_tokens_to_string(output_tokens)
|
| 181 |
-
|
| 182 |
-
|
| 183 |
-
class NTv3Tokenizer(_BaseNTv3Tokenizer):
|
| 184 |
-
"""Character-level tokenizer for NTv3 DNA sequences."""
|
| 185 |
-
|
| 186 |
-
def __init__(
|
| 187 |
-
self,
|
| 188 |
-
vocab_file: Optional[str] = None,
|
| 189 |
-
unk_token: str = "<unk>",
|
| 190 |
-
pad_token: str = "<pad>",
|
| 191 |
-
mask_token: str = "<mask>",
|
| 192 |
-
cls_token: str = "<cls>",
|
| 193 |
-
eos_token: str = "<eos>",
|
| 194 |
-
bos_token: str = "<bos>",
|
| 195 |
-
**kwargs: Any,
|
| 196 |
-
) -> None:
|
| 197 |
-
dna_tokens = ("A", "T", "C", "G", "N")
|
| 198 |
-
default_tokens = (
|
| 199 |
-
unk_token,
|
| 200 |
-
pad_token,
|
| 201 |
-
mask_token,
|
| 202 |
-
cls_token,
|
| 203 |
-
eos_token,
|
| 204 |
-
bos_token,
|
| 205 |
-
*dna_tokens,
|
| 206 |
-
)
|
| 207 |
-
super().__init__(
|
| 208 |
-
vocab_file=vocab_file,
|
| 209 |
-
unk_token=unk_token,
|
| 210 |
-
pad_token=pad_token,
|
| 211 |
-
mask_token=mask_token,
|
| 212 |
-
cls_token=cls_token,
|
| 213 |
-
eos_token=eos_token,
|
| 214 |
-
bos_token=bos_token,
|
| 215 |
-
default_tokens=default_tokens,
|
| 216 |
-
standard_tokens=dna_tokens,
|
| 217 |
-
**kwargs,
|
| 218 |
-
)
|
| 219 |
-
|
| 220 |
-
def _tokenize(self, text: str) -> List[str]:
|
| 221 |
-
tokens: List[str] = []
|
| 222 |
-
for char in text:
|
| 223 |
-
candidate = char.upper()
|
| 224 |
-
if candidate in self._standard_tokens or candidate in self._token_to_id:
|
| 225 |
-
tokens.append(candidate)
|
| 226 |
-
else:
|
| 227 |
-
tokens.append(self._unk_literal)
|
| 228 |
-
return tokens
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
__all__ = [
|
| 232 |
-
"NTv3Tokenizer",
|
| 233 |
-
]
|
| 234 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|