mod
This commit is contained in:
l.gabrysiak
parent
d073a1733b
commit
c55fbe8632
339
hft.py
339
hft.py
|
|
@ -1,21 +1,30 @@
|
||||||
import os
|
import os
|
||||||
import torch
|
import torch
|
||||||
import torch.nn as nn
|
import random
|
||||||
from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, Trainer, DataCollatorForLanguageModeling
|
|
||||||
from datasets import Dataset
|
|
||||||
import re
|
import re
|
||||||
import json
|
import json
|
||||||
import PyPDF2
|
import PyPDF2
|
||||||
import docx2txt
|
import docx2txt
|
||||||
import pytesseract
|
import pytesseract
|
||||||
|
import numpy as np
|
||||||
from PIL import Image
|
from PIL import Image
|
||||||
from collections import defaultdict
|
from collections import defaultdict
|
||||||
|
from multiprocessing import cpu_count
|
||||||
|
from concurrent.futures import ThreadPoolExecutor
|
||||||
|
from transformers import (
|
||||||
|
AutoTokenizer,
|
||||||
|
AutoModelForCausalLM,
|
||||||
|
TrainingArguments,
|
||||||
|
Trainer,
|
||||||
|
DataCollatorForLanguageModeling
|
||||||
|
)
|
||||||
|
from datasets import Dataset
|
||||||
|
from nlpaug import Augmenter, CharAugmenter, WordAugmenter
|
||||||
from huggingface_hub import login
|
from huggingface_hub import login
|
||||||
|
|
||||||
# Konfiguracja
|
# Konfiguracja
|
||||||
os.environ['TORCH_USE_CUDA_DSA'] = '1'
|
|
||||||
os.environ["TOKENIZERS_PARALLELISM"] = "false"
|
os.environ["TOKENIZERS_PARALLELISM"] = "false"
|
||||||
login(token="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")
|
login(token="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") # Zastąp prawdziwym tokenem
|
||||||
|
|
||||||
class SourceMapper:
|
class SourceMapper:
|
||||||
def __init__(self):
|
def __init__(self):
|
||||||
|
|
@ -33,227 +42,215 @@ class SourceMapper:
|
||||||
def get_source(self, idx):
|
def get_source(self, idx):
|
||||||
return self.idx_to_source.get(idx, "Unknown")
|
return self.idx_to_source.get(idx, "Unknown")
|
||||||
|
|
||||||
def load_file_catalog(catalog_path):
|
class LegalDataProcessor:
|
||||||
try:
|
def __init__(self, catalog_path):
|
||||||
with open(catalog_path, 'r', encoding='utf-8') as file:
|
self.catalog = self.load_catalog(catalog_path)
|
||||||
return json.load(file)
|
self.augmenter = WordAugmenter.AntonymAug()
|
||||||
except Exception as e:
|
|
||||||
print(f"Błąd wczytywania katalogu plików: {str(e)}")
|
|
||||||
return {}
|
|
||||||
|
|
||||||
def identify_legal_document(filename, file_catalog):
|
def load_catalog(self, path):
|
||||||
base_name = os.path.splitext(filename)[0].lower()
|
|
||||||
return file_catalog.get(base_name, "Opracowanie własne")
|
|
||||||
|
|
||||||
def extract_text_from_file(file_path):
|
|
||||||
try:
|
try:
|
||||||
_, ext = os.path.splitext(file_path)
|
with open(path, 'r', encoding='utf-8') as f:
|
||||||
ext = ext.lower()
|
return json.load(f)
|
||||||
|
except:
|
||||||
|
return defaultdict(str)
|
||||||
|
|
||||||
if ext in ['.txt', '.md']:
|
def identify_document(self, filename):
|
||||||
with open(file_path, 'r', encoding='utf-8') as file:
|
base = os.path.splitext(filename)[0].lower()
|
||||||
return file.read()
|
return self.catalog.get(base, "Opracowanie własne")
|
||||||
elif ext == '.pdf':
|
|
||||||
text = ""
|
def extract_text(self, file_path):
|
||||||
|
ext = os.path.splitext(file_path)[1].lower()
|
||||||
try:
|
try:
|
||||||
with open(file_path, 'rb') as file:
|
if ext == '.pdf':
|
||||||
reader = PyPDF2.PdfReader(file)
|
return self._extract_pdf(file_path)
|
||||||
for page in reader.pages:
|
|
||||||
text += page.extract_text() or ""
|
|
||||||
except Exception as e:
|
|
||||||
print(f"Błąd PDF: {str(e)}")
|
|
||||||
return text
|
|
||||||
elif ext in ['.doc', '.docx']:
|
elif ext in ['.doc', '.docx']:
|
||||||
return docx2txt.process(file_path)
|
return docx2txt.process(file_path)
|
||||||
elif ext in ['.jpg', '.jpeg', '.png', '.bmp', '.tiff']:
|
elif ext in ['.jpg', '.jpeg', '.png']:
|
||||||
return pytesseract.image_to_string(Image.open(file_path))
|
return self._extract_ocr(file_path)
|
||||||
else:
|
else:
|
||||||
print(f"Nieobsługiwany format pliku: {ext}")
|
with open(file_path, 'r', encoding='utf-8') as f:
|
||||||
return ""
|
return f.read()
|
||||||
except Exception as e:
|
except Exception as e:
|
||||||
print(f"Błąd ekstrakcji tekstu: {str(e)}")
|
print(f"Błąd przetwarzania {file_path}: {str(e)}")
|
||||||
return ""
|
return ""
|
||||||
|
|
||||||
def prepare_dataset(directory, catalog_path, source_mapper):
|
def _extract_pdf(self, path):
|
||||||
file_catalog = load_file_catalog(catalog_path)
|
text = ""
|
||||||
data = []
|
with open(path, 'rb') as f:
|
||||||
|
reader = PyPDF2.PdfReader(f)
|
||||||
|
for page in reader.pages:
|
||||||
|
text += page.extract_text() + "\n"
|
||||||
|
return re.sub(r'\s+', ' ', text)
|
||||||
|
|
||||||
print(f"\n{'='*50}\nDIAGNOSTYKA DANYCH\n{'='*50}")
|
def _extract_ocr(self, path):
|
||||||
|
return pytesseract.image_to_string(
|
||||||
|
Image.open(path),
|
||||||
|
config='--psm 4 --oem 3 -c preserve_interword_spaces=1'
|
||||||
|
)
|
||||||
|
|
||||||
for root, _, files in os.walk(directory):
|
def process_legal(self, text, doc_type):
|
||||||
for file in files:
|
articles = re.split(
|
||||||
file_path = os.path.join(root, file)
|
r'(?ix)(Art\.?\s*\d+[a-z]*|§\s*\d+|Rozdział\s+[IVXLCDM]+)\b',
|
||||||
print(f"\nPrzetwarzanie pliku: {file_path}")
|
text
|
||||||
|
)
|
||||||
|
processed = []
|
||||||
|
current_header = ""
|
||||||
|
|
||||||
try:
|
for item in articles:
|
||||||
text = extract_text_from_file(file_path)
|
if item and re.match(r'(?i)(Art|§|Rozdział)', item):
|
||||||
if not text.strip():
|
if current_header:
|
||||||
print("Pominięto - brak tekstu")
|
processed.append(current_header)
|
||||||
continue
|
current_header = item.strip()
|
||||||
|
elif current_header:
|
||||||
print(f"Długość tekstu: {len(text)} znaków")
|
processed.append(current_header + " " + item.strip())
|
||||||
|
current_header = ""
|
||||||
doc_type = identify_legal_document(file, file_catalog)
|
|
||||||
print(f"Rozpoznany typ dokumentu: {doc_type}")
|
|
||||||
|
|
||||||
if doc_type != "Opracowanie własne":
|
|
||||||
articles = re.split(r'(?i)(Art[\.\s]+\d+[\.\s]?)', text)
|
|
||||||
articles = [a.strip() for a in articles if a.strip()]
|
|
||||||
|
|
||||||
print(f"Znaleziono {len(articles)} fragmentów")
|
|
||||||
|
|
||||||
for i in range(0, len(articles)-1, 2):
|
|
||||||
article_number = articles[i]
|
|
||||||
article_content = articles[i+1]
|
|
||||||
|
|
||||||
if len(article_content) < 50:
|
|
||||||
continue
|
|
||||||
|
|
||||||
source = f"{doc_type}, {article_number}"
|
|
||||||
source_mapper.add_source(source)
|
|
||||||
data.append({
|
|
||||||
"text": f"{article_number} {article_content}",
|
|
||||||
"source_idx": source_mapper.get_idx(source)
|
|
||||||
})
|
|
||||||
else:
|
else:
|
||||||
|
processed.append(item.strip())
|
||||||
|
|
||||||
|
return [
|
||||||
|
(f"[{doc_type}] {p}", doc_type)
|
||||||
|
for p in processed if len(p) > 30
|
||||||
|
]
|
||||||
|
|
||||||
|
def process_custom(self, text):
|
||||||
clean_text = re.sub(r'\s+', ' ', text).strip()
|
clean_text = re.sub(r'\s+', ' ', text).strip()
|
||||||
chunks = [clean_text[i:i+512] for i in range(0, len(clean_text), 512)]
|
chunk_size = 384
|
||||||
chunks = [c for c in chunks if c.strip()]
|
overlap = 128
|
||||||
|
|
||||||
for chunk in chunks:
|
chunks = [
|
||||||
data.append({
|
clean_text[i:i+chunk_size]
|
||||||
"text": chunk,
|
for i in range(0, len(clean_text), chunk_size - overlap)
|
||||||
"source_idx": -1
|
]
|
||||||
})
|
return [("[Custom] " + c, "Custom") for c in chunks if c.strip()]
|
||||||
print(f"Dodano {len(chunks)} chunków")
|
|
||||||
|
|
||||||
except Exception as e:
|
class EnhancedDataCollator(DataCollatorForLanguageModeling):
|
||||||
print(f"Błąd podczas przetwarzania pliku: {str(e)}")
|
|
||||||
continue
|
|
||||||
|
|
||||||
print(f"\nPodsumowanie przygotowania danych:")
|
|
||||||
print(f"Łączna liczba przykładów: {len(data)}")
|
|
||||||
if data:
|
|
||||||
print("Przykładowy wpis:")
|
|
||||||
print(json.dumps(data[0], indent=2, ensure_ascii=False))
|
|
||||||
else:
|
|
||||||
print("BRAK DANYCH - sprawdź diagnostykę powyżej")
|
|
||||||
|
|
||||||
return data
|
|
||||||
|
|
||||||
class CustomModel(nn.Module):
|
|
||||||
def __init__(self, model_name, config):
|
|
||||||
super().__init__()
|
|
||||||
self.base_model = AutoModelForCausalLM.from_pretrained(model_name, config=config)
|
|
||||||
self.source_embedding = nn.Embedding(10000, config.hidden_size, padding_idx=-1)
|
|
||||||
|
|
||||||
for param in self.base_model.parameters():
|
|
||||||
param.requires_grad = False
|
|
||||||
for param in self.base_model.get_output_embeddings().parameters():
|
|
||||||
param.requires_grad = True
|
|
||||||
|
|
||||||
def forward(self, input_ids=None, attention_mask=None, labels=None, source_idx=None, **kwargs):
|
|
||||||
if source_idx is not None:
|
|
||||||
valid_indices = torch.clamp(source_idx, 0, self.source_embedding.num_embeddings-1)
|
|
||||||
source_embeds = self.source_embedding(valid_indices).unsqueeze(1)
|
|
||||||
inputs_embeds = self.base_model.get_input_embeddings()(input_ids) + source_embeds
|
|
||||||
return self.base_model(
|
|
||||||
inputs_embeds=inputs_embeds,
|
|
||||||
attention_mask=attention_mask,
|
|
||||||
labels=labels,
|
|
||||||
**kwargs
|
|
||||||
)
|
|
||||||
return self.base_model(
|
|
||||||
input_ids=input_ids,
|
|
||||||
attention_mask=attention_mask,
|
|
||||||
labels=labels,
|
|
||||||
**kwargs
|
|
||||||
)
|
|
||||||
|
|
||||||
def generate(self, *args, **kwargs):
|
|
||||||
return self.base_model.generate(*args, **kwargs)
|
|
||||||
|
|
||||||
class CustomDataCollator(DataCollatorForLanguageModeling):
|
|
||||||
def torch_call(self, examples):
|
def torch_call(self, examples):
|
||||||
# Przetwórz podstawowe pola
|
batch = super().torch_call(examples)
|
||||||
input_ids = torch.stack([torch.tensor(ex["input_ids"]) for ex in examples])
|
|
||||||
attention_mask = torch.stack([torch.tensor(ex["attention_mask"]) for ex in examples])
|
|
||||||
labels = torch.stack([torch.tensor(ex["labels"]) for ex in examples])
|
|
||||||
|
|
||||||
batch = {
|
|
||||||
"input_ids": input_ids,
|
|
||||||
"attention_mask": attention_mask,
|
|
||||||
"labels": labels
|
|
||||||
}
|
|
||||||
|
|
||||||
# Dodaj source_idx jeśli istnieje
|
|
||||||
if "source_idx" in examples[0]:
|
if "source_idx" in examples[0]:
|
||||||
source_idx = torch.stack([torch.tensor(ex["source_idx"]) for ex in examples])
|
batch["source_idx"] = torch.tensor(
|
||||||
batch["source_idx"] = source_idx
|
[ex["source_idx"] for ex in examples],
|
||||||
|
dtype=torch.long
|
||||||
|
)
|
||||||
return batch
|
return batch
|
||||||
|
|
||||||
def main():
|
def main():
|
||||||
|
# Konfiguracja
|
||||||
source_mapper = SourceMapper()
|
source_mapper = SourceMapper()
|
||||||
model_name = "crumb/nano-mistral"
|
processor = LegalDataProcessor("file_catalog.json")
|
||||||
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
tokenizer = AutoTokenizer.from_pretrained("crumb/nano-mistral")
|
||||||
tokenizer.pad_token = tokenizer.eos_token
|
tokenizer.pad_token = tokenizer.eos_token
|
||||||
|
|
||||||
# Przygotowanie danych
|
# Przygotowanie danych
|
||||||
catalog_path = "file_catalog.json"
|
data = []
|
||||||
data = prepare_dataset("files", catalog_path, source_mapper)
|
|
||||||
|
|
||||||
if not data:
|
def process_file(file_path):
|
||||||
print("\nBrak danych do treningu!")
|
nonlocal data
|
||||||
|
text = processor.extract_text(file_path)
|
||||||
|
if not text:
|
||||||
return
|
return
|
||||||
|
|
||||||
|
doc_type = processor.identify_document(os.path.basename(file_path))
|
||||||
|
if doc_type != "Opracowanie własne":
|
||||||
|
processed = processor.process_legal(text, doc_type)
|
||||||
|
else:
|
||||||
|
processed = processor.process_custom(text)
|
||||||
|
|
||||||
|
for text, source in processed:
|
||||||
|
source_mapper.add_source(source)
|
||||||
|
data.append({
|
||||||
|
"text": text,
|
||||||
|
"source_idx": source_mapper.get_idx(source)
|
||||||
|
})
|
||||||
|
|
||||||
|
# Przetwarzanie wielowątkowe
|
||||||
|
with ThreadPoolExecutor(max_workers=cpu_count()) as executor:
|
||||||
|
futures = []
|
||||||
|
for root, _, files in os.walk("files"):
|
||||||
|
for file in files:
|
||||||
|
futures.append(executor.submit(
|
||||||
|
process_file,
|
||||||
|
os.path.join(root, file)
|
||||||
|
))
|
||||||
|
|
||||||
|
for future in futures:
|
||||||
|
try:
|
||||||
|
future.result()
|
||||||
|
except Exception as e:
|
||||||
|
print(f"Błąd: {str(e)}")
|
||||||
|
|
||||||
|
# Augmentacja
|
||||||
|
print(f"Przed augmentacją: {len(data)} przykładów")
|
||||||
|
augmented = []
|
||||||
|
for item in data:
|
||||||
|
for _ in range(2): # 2 dodatkowe warianty
|
||||||
|
sentences = item['text'].split('. ')
|
||||||
|
random.shuffle(sentences)
|
||||||
|
augmented.append({
|
||||||
|
"text": '. '.join(sentences),
|
||||||
|
"source_idx": item["source_idx"]
|
||||||
|
})
|
||||||
|
data += augmented
|
||||||
|
print(f"Po augmentacji: {len(data)} przykładów")
|
||||||
|
|
||||||
|
# Przygotowanie datasetu
|
||||||
dataset = Dataset.from_list(data)
|
dataset = Dataset.from_list(data)
|
||||||
|
|
||||||
def tokenize_function(examples):
|
def tokenize_fn(examples):
|
||||||
tokenized = tokenizer(
|
tokenized = tokenizer(
|
||||||
examples["text"],
|
examples["text"],
|
||||||
truncation=True,
|
|
||||||
padding="max_length",
|
|
||||||
max_length=512,
|
max_length=512,
|
||||||
|
padding="max_length",
|
||||||
|
truncation=True,
|
||||||
return_tensors="pt"
|
return_tensors="pt"
|
||||||
)
|
)
|
||||||
return {
|
return {
|
||||||
"input_ids": tokenized["input_ids"].squeeze(),
|
"input_ids": tokenized["input_ids"].squeeze(),
|
||||||
"attention_mask": tokenized["attention_mask"].squeeze(),
|
"attention_mask": tokenized["attention_mask"].squeeze(),
|
||||||
"labels": tokenized["input_ids"].squeeze().clone(),
|
"labels": tokenized["input_ids"].squeeze(),
|
||||||
"source_idx": examples["source_idx"] # Dodano bez konwersji do tensora
|
"source_idx": examples["source_idx"]
|
||||||
}
|
}
|
||||||
|
|
||||||
tokenized_dataset = dataset.map(tokenize_function, batched=True, batch_size=16)
|
tokenized_ds = dataset.map(
|
||||||
|
tokenize_fn,
|
||||||
|
batched=True,
|
||||||
|
batch_size=32,
|
||||||
|
num_proc=4
|
||||||
|
)
|
||||||
|
|
||||||
model = CustomModel(model_name, AutoModelForCausalLM.from_pretrained(model_name).config)
|
# Model
|
||||||
model.source_mapper = source_mapper
|
model = AutoModelForCausalLM.from_pretrained(
|
||||||
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
"crumb/nano-mistral",
|
||||||
model.to(device)
|
trust_remote_code=True
|
||||||
|
)
|
||||||
|
model.resize_token_embeddings(len(tokenizer))
|
||||||
|
|
||||||
|
# Trening
|
||||||
training_args = TrainingArguments(
|
training_args = TrainingArguments(
|
||||||
output_dir="./results",
|
output_dir="./results",
|
||||||
num_train_epochs=3,
|
num_train_epochs=5,
|
||||||
per_device_train_batch_size=2,
|
per_device_train_batch_size=2,
|
||||||
gradient_accumulation_steps=4,
|
gradient_accumulation_steps=8,
|
||||||
learning_rate=2e-5,
|
learning_rate=1e-4,
|
||||||
fp16=torch.cuda.is_available(),
|
fp16=torch.cuda.is_available(),
|
||||||
logging_steps=10,
|
logging_steps=20,
|
||||||
save_strategy="steps",
|
save_strategy="epoch",
|
||||||
save_steps=1000,
|
report_to="none"
|
||||||
report_to="none",
|
|
||||||
remove_unused_columns=False
|
|
||||||
)
|
)
|
||||||
|
|
||||||
trainer = Trainer(
|
trainer = Trainer(
|
||||||
model=model,
|
model=model,
|
||||||
args=training_args,
|
args=training_args,
|
||||||
train_dataset=tokenized_dataset,
|
train_dataset=tokenized_ds,
|
||||||
data_collator=CustomDataCollator(tokenizer=tokenizer, mlm=False)
|
data_collator=EnhancedDataCollator(tokenizer=tokenizer, mlm=False)
|
||||||
)
|
)
|
||||||
|
|
||||||
print("\nRozpoczęcie treningu...")
|
print("Rozpoczęcie treningu...")
|
||||||
trainer.train()
|
trainer.train()
|
||||||
|
print("Trening zakończony!")
|
||||||
|
|
||||||
|
# Zapisz model
|
||||||
|
model.save_pretrained("./trained_model")
|
||||||
|
tokenizer.save_pretrained("./trained_model")
|
||||||
|
|
||||||
if __name__ == "__main__":
|
if __name__ == "__main__":
|
||||||
main()
|
main()
|
||||||
Loading…
Reference in New Issue