File size: 7,108 Bytes
96c0ca2 a03d44f 96c0ca2 a03d44f 96c0ca2 a03d44f 96c0ca2 a03d44f |
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 |
#!/usr/bin/env python
# encoding: utf-8
import os
import json
import itertools
from typing import Sequence, List
from transformers import PreTrainedTokenizer
gene_standard_toks = ['1', '2', '3', '4', '5', '.', '-', '*']
prot_standard_toks = ['L', 'A', 'G', 'V', 'S', 'E', 'R', 'T', 'I', 'D', 'P', 'K', 'Q', 'N', 'F', 'Y', 'M', 'H', 'W', 'C', 'X', 'B', 'U', 'Z', 'O', 'J', '.', '-', '*']
gene_prot_standard_toks = ['1', '2', '3', '4', '5', 'L', 'A', 'G', 'V', 'S', 'E', 'R', 'T', 'I', 'D', 'P', 'K', 'Q', 'N', 'F', 'Y', 'M', 'H', 'W', 'C', 'X', 'B', 'U', 'Z', 'O', 'J', '.', '-', '*']
gene_prot_prepend_toks = ['[PAD]', '[UNK]']
gene_prot_append_toks = ['[CLS]', '[SEP]', '[MASK]']
class Alphabet(object):
def __init__(
self,
standard_toks: Sequence[str] = gene_prot_standard_toks,
prepend_toks: Sequence[str] = gene_prot_prepend_toks,
append_toks: Sequence[str] = gene_prot_append_toks,
prepend_bos: bool = True,
append_eos: bool = True
):
self.standard_toks = list(standard_toks)
self.prepend_toks = list(prepend_toks)
self.append_toks = list(append_toks)
self.prepend_bos = prepend_bos
self.append_eos = append_eos
self.all_toks = list(self.prepend_toks)
self.all_toks.extend(self.append_toks)
self.all_toks.extend(self.standard_toks)
self.tok_to_idx = {tok: i for i, tok in enumerate(self.all_toks)}
self.unk_idx = self.tok_to_idx["[UNK]"]
self.padding_idx = self.get_idx("[PAD]")
self.pad_token_id = self.padding_idx
self.cls_idx = self.get_idx("[CLS]")
self.mask_idx = self.get_idx("[MASK]")
self.eos_idx = self.get_idx("[SEP]")
self.all_special_tokens = prepend_toks + append_toks
self.all_special_token_idx_list = [self.tok_to_idx[v] for v in self.all_special_tokens]
self.unique_no_split_tokens = self.all_toks
self.vocab_size = self.__len__()
def __len__(self):
return len(self.all_toks)
def get_idx(self, tok):
return self.tok_to_idx.get(tok, self.unk_idx)
def get_tok(self, ind):
return self.all_toks[ind]
def to_dict(self):
return self.tok_to_idx.copy()
@classmethod
def from_predefined(cls, name: str):
if name.lower() == "prot":
standard_toks = prot_standard_toks
elif name.lower() == "gene":
standard_toks = gene_standard_toks
elif name.lower() in ["gene_prot", "prot_gene"]:
standard_toks = gene_prot_standard_toks
else:
raise Exception("Not support tokenizer name: %s" % name)
prepend_toks = gene_prot_prepend_toks
append_toks = gene_prot_append_toks
prepend_bos = True
append_eos = True
return cls(standard_toks, prepend_toks, append_toks, prepend_bos, append_eos)
@classmethod
def from_pretrained(cls, dir_path):
import os, pickle
return pickle.load(open(os.path.join(dir_path, "alphabet.pkl"), "rb"))
def save_pretrained(self, save_dir):
import os, pickle
with open(os.path.join(save_dir, "alphabet.pkl"), 'wb') as outp:
pickle.dump(self, outp, pickle.HIGHEST_PROTOCOL)
def _tokenize(self, text) -> str:
return text.split()
def tokenize(self, text, **kwargs) -> List[str]:
def split_on_token(tok, text):
result = []
split_text = text.split(tok)
for i, sub_text in enumerate(split_text):
if i < len(split_text) - 1:
sub_text = sub_text.rstrip()
if i > 0:
sub_text = sub_text.lstrip()
if i == 0 and not sub_text:
result.append(tok)
elif i == len(split_text) - 1:
if sub_text:
result.append(sub_text)
else:
pass
else:
if sub_text:
result.append(sub_text)
result.append(tok)
return result
def split_on_tokens(tok_list, text):
if not text.strip():
return []
tokenized_text = []
text_list = [text]
for tok in tok_list:
tokenized_text = []
for sub_text in text_list:
if sub_text not in self.unique_no_split_tokens:
tokenized_text.extend(split_on_token(tok, sub_text))
else:
tokenized_text.append(sub_text)
text_list = tokenized_text
return list(
itertools.chain.from_iterable(
(
self._tokenize(token)
if token not in self.unique_no_split_tokens
else [token]
for token in tokenized_text
)
)
)
no_split_token = self.unique_no_split_tokens
tokenized_text = split_on_tokens(no_split_token, text)
return tokenized_text
def encode(self, text):
return [self.tok_to_idx[tok] for tok in self.tokenize(text)]
class AlphabetTokenizer(PreTrainedTokenizer):
def __init__(
self,
alphabet: Alphabet = Alphabet(),
**kwargs
):
super().__init__(**kwargs)
self.alphabet = alphabet
self.pad_token = '[PAD]'
self.cls_token = '[CLS]'
self.sep_token = '[SEP]'
self.mask_token = '[MASK]'
self.unk_token = '[UNK]'
def _tokenize(self, text: str):
# Use your Alphabet class's tokenize method
return self.alphabet.tokenize(text)
def convert_tokens_to_ids(self, tokens):
# Use the Alphabet class's get_idx method
return [self.alphabet.get_idx(token) for token in tokens]
def convert_ids_to_tokens(self, ids):
# Use the Alphabet class's get_tok method
return [self.alphabet.get_tok(index) for index in ids]
def save_vocabulary(self, save_directory, filename_prefix=None):
# Save the tokenizer vocabulary, required by Hugging Face
vocab_file = os.path.join(save_directory, (filename_prefix or "") + "vocab.json")
with open(vocab_file, "w", encoding="utf-8") as vocab_writer:
json.dump(self.alphabet.to_dict(), vocab_writer, ensure_ascii=False)
return (vocab_file,)
def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
# Add special tokens to input ids, if required
cls_token = [self.alphabet.cls_idx]
sep_token = [self.alphabet.eos_idx]
if token_ids_1:
return cls_token + token_ids_0 + sep_token + token_ids_1 + sep_token
return cls_token + token_ids_0 + sep_token
|