MCPcopy Create free account
hub / github.com/SpecterOps/DeepPass2 / preprocess

Function preprocess

utils/finetune-BERT-chunked.py:73–96  ·  view source on GitHub ↗

Extracts tokens from text and applies tokenization.

(examples)

Source from the content-addressed store, hash-verified

71
72 # Tokenize dataset for token classification
73 def preprocess(examples):
74 """Extracts tokens from text and applies tokenization."""
75
76 doc = examples['document']
77
78 tokenized = tokenizer(doc, padding="max_length", truncation=True, return_offsets_mapping=True)
79
80 # Generate labels (1 for credentials, 0 otherwise)
81 labels = [0] * len(tokenized["input_ids"])
82
83 credentials = examples["cred"]
84 if credentials != []:
85 for credential in credentials:
86 start_idx = doc.find(credential)
87
88 end_idx = start_idx + len(credential)
89
90 # Label tokens within credential span
91 for i, (tok_start, tok_end) in enumerate(tokenized["offset_mapping"]):
92 if tok_start >= start_idx and tok_end <= end_idx:
93 labels[i] = 1 # Mark as credential token
94
95 tokenized["labels"] = labels
96 return tokenized
97
98
99 # Apply preprocessing to dataset

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected