app.py

# Importing necessary libraries
from datasets import load_dataset
from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments, TextClassificationPipeline
from sklearn.metrics import accuracy_score, precision_recall_fscore_support
import gradio as gr

# Load the dataset
ds = load_dataset("GonzaloA/fake_news")

# Load pre-trained tokenizer
tokenizer = AutoTokenizer.from_pretrained('bert-base-uncased')

# Define tokenization function
def tokenize_function(examples):
    return tokenizer(examples['text'], padding='max_length', truncation=True, max_length=128)

# Apply tokenization
tokenized_datasets = ds.map(tokenize_function, batched=True)

# Load pre-trained BERT model for sequence classification
model = AutoModelForSequenceClassification.from_pretrained('bert-base-uncased', num_labels=2)

# Define training arguments
training_args = TrainingArguments(
    output_dir='./results',
    num_train_epochs=3,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=8,
    evaluation_strategy='epoch',
    logging_dir='./logs',
)

# Create trainer instance
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_datasets['train'].shuffle().select(range(1000)),
    eval_dataset=tokenized_datasets['test'].shuffle().select(range(1000)),
)

# Start training
trainer.train()

# Define function to compute metrics
def compute_metrics(pred):
    labels = pred.label_ids
    preds = pred.predictions.argmax(-1)
    precision, recall, f1, _ = precision_recall_fscore_support(labels, preds, average='binary')
    acc = accuracy_score(labels, preds)
    return {'accuracy': acc, 'f1': f1, 'precision': precision, 'recall': recall}

# Update trainer to include custom metrics
trainer.compute_metrics = compute_metrics

# Evaluate the model
eval_result = trainer.evaluate()
print(eval_result)

# Save the fine-tuned model and tokenizer
trainer.save_model('TeamQuad-fine-tuned-bert')
tokenizer.save_pretrained('TeamQuad-fine-tuned-bert')

# Load the fine-tuned model and tokenizer
new_model = AutoModelForSequenceClassification.from_pretrained('TeamQuad-fine-tuned-bert')
new_tokenizer = AutoTokenizer.from_pretrained('TeamQuad-fine-tuned-bert')

# Create a classification pipeline
classifier = TextClassificationPipeline(model=new_model, tokenizer=new_tokenizer)

# Add label mapping for fake news detection (assuming LABEL_0 = 'fake' and LABEL_1 = 'true')
label_mapping = {0: 'fake', 1: 'true'}

# Function to classify input text
def classify_news(text):
    result = classifier(text)
    # Extract the label and score
    label = result[0]['label']  # 'LABEL_0' or 'LABEL_1'
    score = result[0]['score']  # Confidence score
    mapped_result = {'label': label_mapping[int(label.split('_')[1])], 'score': score}
    return f"Label: {mapped_result['label']}, Score: {mapped_result['score']:.4f}"

# Create a Gradio interface
iface = gr.Interface(
    fn=classify_news,  # The function to process the input
    inputs=gr.Textbox(lines=10, placeholder="Enter a news headline or article to classify..."),
    outputs="text",  # Output will be displayed as text
    title="Fake News Detection",
    description="Enter a news headline or article and see whether the model classifies it as 'Fake News' or 'True News'.",
)

# Launch the interface
iface.launch(share=True)