Piyapawashe/RESUME_CLASSIFICATION_BERT

1import gradio as gr2import torch3import pickle4from transformers import BertTokenizer, BertForSequenceClassification5 6# Load model and tokenizer from local folder7def load_model():8    model = BertForSequenceClassification.from_pretrained("./")9    tokenizer = BertTokenizer.from_pretrained("./")10    return model, tokenizer11 12# Load label encoder13def load_encoder():14    with open("label_encoder.pkl", "rb") as f:15        return pickle.load(f)16 17model, tokenizer = load_model()18label_encoder = load_encoder()19 20# Prediction function21def predict_resume(text):22    if text.strip() == "":23        return "⚠️ Please enter resume text"24    25    inputs = tokenizer(26        text,27        return_tensors="pt",28        truncation=True,29        padding=True,30        max_length=51231    )32    33    with torch.no_grad():34        outputs = model(**inputs)35    36    pred_id = torch.argmax(outputs.logits, dim=1).item()37    category = label_encoder.inverse_transform([pred_id])[0]38    39    confidence = torch.softmax(outputs.logits, dim=1)[0][pred_id].item()40    41    return f"🔮 Predicted Category: **{category}**\n📊 Confidence: {confidence:.2%}"42 43# Gradio Interface44interface = gr.Interface(45    fn=predict_resume,   # <-- FIXED: added function here46    inputs=gr.Textbox(lines=10, placeholder="Paste resume text here..."),47    outputs="text",48    title="Resume Classification using BERT",49    description="This app classifies resumes into job categories using a fine-tuned BERT model.",50    css="style.css"   # optional external CSS file51)52 53interface.launch()54