File size: 14,705 Bytes
060ac52 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 |
import random
import torch
import logging
from transformers import BertTokenizer, BertForMaskedLM
from nltk.corpus import stopwords
import nltk
from transformers import RobertaTokenizer, RobertaForMaskedLM
from tqdm import tqdm
# Set logging to WARNING for a cleaner terminal.
logging.basicConfig(level=logging.WARNING, format="%(asctime)s - %(levelname)s - %(message)s")
logger = logging.getLogger(__name__)
# Ensure stopwords are downloaded
try:
nltk.data.find('corpora/stopwords')
except LookupError:
nltk.download('stopwords')
class MaskingProcessor:
def __init__(self, tokenizer, model):
self.tokenizer = tokenizer
self.model = model.to(torch.device("cuda" if torch.cuda.is_available() else "cpu"))
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
self.stop_words = set(stopwords.words('english'))
tqdm.write(f"[MaskingProcessor] Initialized on device: {self.device}")
def remove_stopwords(self, words):
return [word for word in words if word.lower() not in self.stop_words]
def adjust_ngram_indices(self, original_words, common_ngrams):
logger.info("Adjusting n-gram indices.")
non_stop_words = self.remove_stopwords(original_words)
original_to_non_stop = []
non_stop_idx = 0
for original_idx, word in enumerate(original_words):
if word.lower() not in self.stop_words:
original_to_non_stop.append((original_idx, non_stop_idx))
non_stop_idx += 1
adjusted_ngrams = {}
for ngram, positions in common_ngrams.items():
adjusted_positions = []
for start, end in positions:
try:
new_start = next(non_stop for orig, non_stop in original_to_non_stop if orig == start)
new_end = next(non_stop for orig, non_stop in original_to_non_stop if orig == end)
adjusted_positions.append((new_start, new_end))
except StopIteration:
continue
adjusted_ngrams[ngram] = adjusted_positions
return adjusted_ngrams
def mask_sentence_random(self, sentence, common_ngrams):
tqdm.write(f"[MaskingProcessor] Masking (random) sentence: {sentence}")
original_words = sentence.split()
has_punctuation = False
punctuation = ''
if original_words and any(original_words[-1].endswith(p) for p in ['.', ',', '!', '?', ';', ':']):
has_punctuation = True
punctuation = original_words[-1][-1]
original_words = original_words[:-1]
non_stop_words = self.remove_stopwords(original_words)
adjusted_ngrams = self.adjust_ngram_indices(original_words, common_ngrams)
mask_indices = []
ngram_positions = [pos for positions in adjusted_ngrams.values() for pos in positions]
if ngram_positions:
first_ngram_start = ngram_positions[0][0]
if first_ngram_start > 0:
mask_index_before_ngram = random.randint(0, first_ngram_start-1)
mask_indices.append(mask_index_before_ngram)
for i in range(len(ngram_positions) - 1):
end_prev = ngram_positions[i][1]
start_next = ngram_positions[i + 1][0]
if start_next > end_prev + 1:
mask_index_between_ngrams = random.randint(end_prev + 1, start_next - 1)
mask_indices.append(mask_index_between_ngrams)
last_ngram_end = ngram_positions[-1][1]
if last_ngram_end < len(non_stop_words) - 1:
mask_index_after_ngram = random.randint(last_ngram_end + 1, len(non_stop_words) - 1)
mask_indices.append(mask_index_after_ngram)
non_stop_to_original = {}
non_stop_idx = 0
for orig_idx, word in enumerate(original_words):
if word.lower() not in self.stop_words:
non_stop_to_original[non_stop_idx] = orig_idx
non_stop_idx += 1
original_mask_indices = [non_stop_to_original[idx] for idx in mask_indices]
masked_words = original_words.copy()
for idx in original_mask_indices:
masked_words[idx] = self.tokenizer.mask_token
if has_punctuation:
masked_words.append(punctuation)
logger.info(f"Masked sentence (random): {' '.join(masked_words)}")
return " ".join(masked_words), original_mask_indices
def mask_sentence_pseudorandom(self, sentence, common_ngrams):
logger.info(f"Masking sentence using pseudorandom strategy: {sentence}")
random.seed(3)
original_words = sentence.split()
has_punctuation = False
punctuation = ''
if original_words and any(original_words[-1].endswith(p) for p in ['.', ',', '!', '?', ';', ':']):
has_punctuation = True
punctuation = original_words[-1][-1]
original_words = original_words[:-1]
non_stop_words = self.remove_stopwords(original_words)
adjusted_ngrams = self.adjust_ngram_indices(original_words, common_ngrams)
mask_indices = []
ngram_positions = [pos for positions in adjusted_ngrams.values() for pos in positions]
if ngram_positions:
first_ngram_start = ngram_positions[0][0]
if first_ngram_start > 0:
mask_index_before_ngram = random.randint(0, first_ngram_start-1)
mask_indices.append(mask_index_before_ngram)
for i in range(len(ngram_positions) - 1):
end_prev = ngram_positions[i][1]
start_next = ngram_positions[i + 1][0]
if start_next > end_prev + 1:
mask_index_between_ngrams = random.randint(end_prev + 1, start_next - 1)
mask_indices.append(mask_index_between_ngrams)
last_ngram_end = ngram_positions[-1][1]
if last_ngram_end < len(non_stop_words) - 1:
mask_index_after_ngram = random.randint(last_ngram_end + 1, len(non_stop_words) - 1)
mask_indices.append(mask_index_after_ngram)
non_stop_to_original = {}
non_stop_idx = 0
for orig_idx, word in enumerate(original_words):
if word.lower() not in self.stop_words:
non_stop_to_original[non_stop_idx] = orig_idx
non_stop_idx += 1
original_mask_indices = [non_stop_to_original[idx] for idx in mask_indices]
masked_words = original_words.copy()
for idx in original_mask_indices:
masked_words[idx] = self.tokenizer.mask_token
if has_punctuation:
masked_words.append(punctuation)
logger.info(f"Masked sentence (pseudorandom): {' '.join(masked_words)}")
return " ".join(masked_words), original_mask_indices
def mask_sentence_entropy(self, sentence, common_ngrams):
logger.info(f"Masking sentence using entropy strategy: {sentence}")
original_words = sentence.split()
has_punctuation = False
punctuation = ''
if original_words and any(original_words[-1].endswith(p) for p in ['.', ',', '!', '?', ';', ':']):
has_punctuation = True
punctuation = original_words[-1][-1]
original_words = original_words[:-1]
non_stop_words = self.remove_stopwords(original_words)
adjusted_ngrams = self.adjust_ngram_indices(original_words, common_ngrams)
mask_indices = []
ngram_positions = [pos for positions in adjusted_ngrams.values() for pos in positions]
non_stop_to_original = {}
non_stop_idx = 0
for orig_idx, word in enumerate(original_words):
if word.lower() not in self.stop_words:
non_stop_to_original[non_stop_idx] = orig_idx
non_stop_idx += 1
if ngram_positions:
first_ngram_start = ngram_positions[0][0]
if first_ngram_start > 0:
candidate_positions = range(0, first_ngram_start)
entropies = [(pos, self.calculate_word_entropy(sentence, non_stop_to_original[pos])) for pos in candidate_positions]
mask_indices.append(max(entropies, key=lambda x: x[1])[0])
for i in range(len(ngram_positions) - 1):
end_prev = ngram_positions[i][1]
start_next = ngram_positions[i + 1][0]
if start_next > end_prev + 1:
candidate_positions = range(end_prev + 1, start_next)
entropies = [(pos, self.calculate_word_entropy(sentence, non_stop_to_original[pos])) for pos in candidate_positions]
mask_indices.append(max(entropies, key=lambda x: x[1])[0])
last_ngram_end = ngram_positions[-1][1]
if last_ngram_end < len(non_stop_words) - 1:
candidate_positions = range(last_ngram_end + 1, len(non_stop_words))
entropies = [(pos, self.calculate_word_entropy(sentence, non_stop_to_original[pos])) for pos in candidate_positions]
mask_indices.append(max(entropies, key=lambda x: x[1])[0])
original_mask_indices = [non_stop_to_original[idx] for idx in mask_indices]
masked_words = original_words.copy()
for idx in original_mask_indices:
masked_words[idx] = self.tokenizer.mask_token
if has_punctuation:
masked_words.append(punctuation)
logger.info(f"Masked sentence (entropy): {' '.join(masked_words)}")
return " ".join(masked_words), original_mask_indices
def calculate_mask_logits(self, original_sentence, original_mask_indices):
logger.info(f"Calculating mask logits for sentence: {original_sentence}")
words = original_sentence.split()
mask_logits = {}
for idx in original_mask_indices:
masked_words = words.copy()
masked_words[idx] = self.tokenizer.mask_token
masked_sentence = " ".join(masked_words)
input_ids = self.tokenizer(masked_sentence, return_tensors="pt")["input_ids"].to(self.device)
mask_token_index = torch.where(input_ids == self.tokenizer.mask_token_id)[1]
with torch.no_grad():
outputs = self.model(input_ids)
logits = outputs.logits
mask_logits_tensor = logits[0, mask_token_index, :]
top_mask_logits, top_mask_indices = torch.topk(mask_logits_tensor, 100, dim=-1)
top_tokens = []
top_logits = []
seen_words = set()
for token_id, logit in zip(top_mask_indices[0], top_mask_logits[0]):
token = self.tokenizer.convert_ids_to_tokens(token_id.item())
if token.startswith('##'):
continue
word = self.tokenizer.convert_tokens_to_string([token]).strip()
if word and word not in seen_words:
seen_words.add(word)
top_tokens.append(word)
top_logits.append(logit.item())
if len(top_tokens) == 50:
break
mask_logits[idx] = {
"tokens": top_tokens,
"logits": top_logits
}
logger.info("Completed calculating mask logits.")
return mask_logits
def calculate_word_entropy(self, sentence, word_position):
logger.info(f"Calculating word entropy for position {word_position} in sentence: {sentence}")
words = sentence.split()
masked_words = words.copy()
masked_words[word_position] = self.tokenizer.mask_token
masked_sentence = " ".join(masked_words)
input_ids = self.tokenizer(masked_sentence, return_tensors="pt")["input_ids"].to(self.device)
mask_token_index = torch.where(input_ids == self.tokenizer.mask_token_id)[1]
with torch.no_grad():
outputs = self.model(input_ids)
logits = outputs.logits
probs = torch.nn.functional.softmax(logits[0, mask_token_index], dim=-1)
entropy = -torch.sum(probs * torch.log(probs + 1e-9))
logger.info(f"Computed entropy: {entropy.item()}")
return entropy.item()
def process_sentences(self, sentences_list, common_grams, method="random"):
tqdm.write(f"[MaskingProcessor] Processing sentences using method: {method}")
results = {}
for sentence, ngrams in tqdm(common_grams.items(), desc="Masking Sentences"):
words = sentence.split()
last_word = words[-1]
if any(last_word.endswith(p) for p in ['.', ',', '!', '?', ';', ':']):
words[-1] = last_word[:-1]
punctuation = last_word[-1]
processed_sentence = " ".join(words) + " " + punctuation
else:
processed_sentence = sentence
if method == "random":
masked_sentence, original_mask_indices = self.mask_sentence_random(processed_sentence, ngrams)
elif method == "pseudorandom":
masked_sentence, original_mask_indices = self.mask_sentence_pseudorandom(processed_sentence, ngrams)
else: # entropy
masked_sentence, original_mask_indices = self.mask_sentence_entropy(processed_sentence, ngrams)
logits = self.calculate_mask_logits(processed_sentence, original_mask_indices)
results[sentence] = {
"masked_sentence": masked_sentence,
"mask_logits": logits
}
logger.info(f"Processed sentence: {sentence}")
tqdm.write("[MaskingProcessor] Completed processing sentences.")
return results
if __name__ == "__main__":
sentences = [
"The quick brown fox jumps over small cat the lazy dog everyday again and again .",
]
result_dict = {
'The quick brown fox jumps over small cat the lazy dog everyday again and again .': {
'brown fox': [(2, 3)],
'cat': [(7, 7)],
'dog': [(10, 10)]
}
}
processor = MaskingProcessor(
BertTokenizer.from_pretrained("bert-large-cased-whole-word-masking"),
BertForMaskedLM.from_pretrained("bert-large-cased-whole-word-masking")
)
results_entropy = processor.process_sentences(sentences_list, common_grams, method="random")
for sentence, output in results_entropy.items():
logger.info(f"Original Sentence (Random): {sentence}")
logger.info(f"Masked Sentence (Random): {output['masked_sentence']}")
|