app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load model and tokenizer
model_id = "Rerandaka/Cild_safety_bigbird"
tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=False)
model = AutoModelForSequenceClassification.from_pretrained(model_id)

# Class mapping (optional — edit as needed)
label_map = {
    0: "Safe / Normal",
    1: "Inappropriate / Unsafe"
}

# Inference function
def classify_text(text: str):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=256)
    with torch.no_grad():
        outputs = model(**inputs)
        probs = torch.nn.functional.softmax(outputs.logits, dim=1)
        predicted = torch.argmax(probs, dim=1).item()
        confidence = probs[0][predicted].item()
    return {
        "label": label_map.get(predicted, str(predicted)),
        "confidence": round(confidence, 4)
    }

# Define Gradio Interface
demo = gr.Interface(
    fn=classify_text,
    inputs=gr.Textbox(label="Enter text to classify"),
    outputs=[
        gr.Textbox(label="Predicted Label"),
        gr.Textbox(label="Confidence")
    ],
    title="Child-Safety Text Classifier",
    description="This model detects if text content is unsafe or inappropriate for children.",
    allow_flagging="never"
)

# Expose API endpoint explicitly
demo.launch(api_name="predict")