Spaces:
Sleeping
Sleeping
File size: 10,644 Bytes
a20a7ca |
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 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import numpy as np
from sklearn.feature_extraction.text import TfidfVectorizer
import joblib
import pandas as pd
from datetime import datetime
import logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class TextClassificationPipeline:
def __init__(self, model_path='./models', method='bertbased'):
"""
Initialize the classification pipeline
Args:
model_path: Path to saved models
method: 'bertbased' or 'baseline'
"""
try:
self.method = method
if method == 'bertbased':
logger.info("Loading BERT model...")
self.tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased')
self.model = AutoModelForSequenceClassification.from_pretrained(f"{model_path}/bert-model")
self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
self.model.to(self.device)
self.model.eval()
logger.info(f"BERT model loaded successfully. Using device: {self.device}")
else:
logger.info("Loading baseline model...")
self.tfidf = joblib.load(f"{model_path}/baseline-model/tfidf_vectorizer.pkl")
self.baseline_model = joblib.load(f"{model_path}/baseline-model/baseline_model.pkl")
logger.info("Baseline model loaded successfully")
# Load label encoder for both methods
self.label_encoder = joblib.load(f"{model_path}/label_encoder.pkl")
except Exception as e:
logger.error(f"Error initializing model: {str(e)}")
raise
# def preprocess_text(self, text):
# """Clean and preprocess text"""
# if isinstance(text, str):
# # Basic cleaning
# text = text.strip()
# text = ' '.join(text.split()) # Remove extra whitespace
# return text
# return text
def preprocess_text(self, text):
"""Clean and preprocess text"""
if isinstance(text, str):
# Basic cleaning
text = text.strip()
text = ' '.join(text.split()) # Remove extra whitespace
# Capitalize first letter to match training data format
text = text.title() # This will capitalize first letter of each word
return text
return text
def preprocess(self, text):
"""
Preprocess the input text based on method
"""
try:
# Clean text first
text = self.preprocess_text(text)
if self.method == 'bertbased':
# BERT preprocessing
encodings = self.tokenizer(
text,
truncation=True,
padding=True,
max_length=512,
return_tensors='pt'
)
encodings = {k: v.to(self.device) for k, v in encodings.items()}
return encodings
else:
# Baseline preprocessing
return self.tfidf.transform([text] if isinstance(text, str) else text)
except Exception as e:
logger.error(f"Error in preprocessing: {str(e)}")
raise
def predict(self, text, return_probability=False):
"""
Predict using either BERT or baseline model
Args:
text: Input text or list of texts
return_probability: Whether to return probability scores
Returns:
Predictions with metadata
"""
try:
# Handle both single string and list of strings
if isinstance(text, str):
text = [text]
# Preprocess
inputs = self.preprocess(text)
if self.method == 'bertbased':
# BERT predictions
with torch.no_grad():
outputs = self.model(**inputs)
probabilities = torch.softmax(outputs.logits, dim=-1)
predictions = torch.argmax(probabilities, dim=-1)
predictions = predictions.cpu().numpy()
probabilities = probabilities.cpu().numpy()
else:
# Baseline predictions
predictions = self.baseline_model.predict(inputs)
probabilities = self.baseline_model.predict_proba(inputs)
# Convert numeric predictions to original labels
predicted_labels = self.label_encoder.inverse_transform(predictions)
# Ensure consistent casing with training data
predicted_labels = [label.title() for label in predicted_labels]
if return_probability:
results = []
for t, label, prob, probs in zip(text, predicted_labels,
probabilities.max(axis=1),
probabilities):
result = {
'text': t[:200] + '...' if len(t) > 200 else t,
'predicted_label': label.title(), # Ensure consistent casing
'confidence': float(prob),
'model_type': self.method,
'probabilities': {
self.label_encoder.inverse_transform([i])[0].title(): float(p) # Consistent casing
for i, p in enumerate(probs)
},
# ... rest of the result dictionary ...
}
results.append(result)
return results[0] if len(text) == 1 else results
return predicted_labels[0] if len(text) == 1 else predicted_labels
except Exception as e:
logger.error(f"Error in prediction: {str(e)}")
raise
def predict_old(self, text, return_probability=False):
"""
Predict using either BERT or baseline model
Args:
text: Input text or list of texts
return_probability: Whether to return probability scores
Returns:
Predictions with metadata
"""
try:
# Handle both single string and list of strings
if isinstance(text, str):
text = [text]
# Preprocess
inputs = self.preprocess(text)
if self.method == 'bertbased':
# BERT predictions
with torch.no_grad():
outputs = self.model(**inputs)
probabilities = torch.softmax(outputs.logits, dim=-1)
predictions = torch.argmax(probabilities, dim=-1)
predictions = predictions.cpu().numpy()
probabilities = probabilities.cpu().numpy()
else:
# Baseline predictions
predictions = self.baseline_model.predict(inputs)
probabilities = self.baseline_model.predict_proba(inputs)
# Convert numeric predictions to original labels
predicted_labels = self.label_encoder.inverse_transform(predictions)
if return_probability:
results = []
for t, label, prob, probs in zip(text, predicted_labels,
probabilities.max(axis=1),
probabilities):
# Create detailed result dictionary
result = {
'text': t[:200] + '...' if len(t) > 200 else t, # Truncate long text
'predicted_label': label,
'confidence': float(prob),
'model_type': self.method,
'probabilities': {
self.label_encoder.inverse_transform([i])[0]: float(p)
for i, p in enumerate(probs)
},
'timestamp': datetime.now().isoformat(),
'metadata': {
'model_name': 'BERT' if self.method == 'bertbased' else 'Baseline',
'text_length': len(t),
'preprocessing_steps': ['cleaning', 'tokenization']
}
}
results.append(result)
return results[0] if len(text) == 1 else results
return predicted_labels[0] if len(text) == 1 else predicted_labels
except Exception as e:
logger.error(f"Error in prediction: {str(e)}")
raise
def get_model_info(self):
"""Return model information"""
return {
'model_type': self.method,
'model_name': 'BERT' if self.method == 'bertbased' else 'Baseline',
'device': str(self.device) if self.method == 'bertbased' else 'CPU',
'max_sequence_length': 512 if self.method == 'bertbased' else None,
'number_of_classes': len(self.label_encoder.classes_),
'classes': list(self.label_encoder.classes_)
}
def load_and_process_pdf(url_or_file):
"""
Load and process PDF from URL or file
Returns extracted text
"""
try:
# Your PDF processing code here
# Return extracted text
pass
except Exception as e:
logger.error(f"Error processing PDF: {str(e)}")
raise
# Example usage
if __name__ == "__main__":
# Test the pipeline
classifier = TextClassificationPipeline()
# Test single prediction
text = "Example construction document text"
result = classifier.predict(text, return_probability=True)
print("\nSingle Prediction Result:")
print(result)
# Test batch prediction
texts = ["First document", "Second document"]
results = classifier.predict(texts, return_probability=True)
print("\nBatch Prediction Results:")
for result in results:
print(f"\nText: {result['text']}")
print(f"Prediction: {result['predicted_label']}")
print(f"Confidence: {result['confidence']:.4f}") |